Clinical Chemistry (ISSN 0009-9147) Is Published Monthly by the American Association for Clinical Chemistry, 1850 K Street, Suite 625, Washington, DC 20006

Contents VolumeClinical 53, NumberChemistry 12

EDITORIAL

Untargeted Metabolomic Analysis Hits the Target. M.J. Bennett 2037

OPINION

Glucose: A Simple Molecule That Is Not Simple to Quantify. R. Gambino 2040

ARTICLES

MOLECULAR DIAGNOSTICS AND GENETICS

Tentacle ProbesTM: Differentiation of Difficult Single-Nucleotide Polymorphisms and Deletions by Presence or Absence of a Signal in Real-Time PCR. B.C. Satterfield, D.A. Kulesh, D.A. Norwood, L.P. Wasieloski, Jr., M.R. Caplan, and J.A.A. West 2042

Development of a Focused Oligonucleotide-Array Comparative Genomic Hybridization Chip for Clinical Diagnosis of Genomic Imbalance. Y. Shen, M. Irons, D.T. Miller, S.W. Cheung, V. Lip, X. Sheng, K. Tomaszewicz, H. Shao, H. Fang, H.S. Tang, C.A. Walsh, O. Platt, J.F. Gusella, and B.-L. Wu 2051

Global Sequencing Approach for Characterizing the Molecular Background of Hereditary Iron Disorders. S. Cunat, M. Giansily-Blaizot, M. Bismuth, F. Blanc, O. Dereure, D. Larrey, A. Le Quellec, P. Pouderoux, C. Rose, I. Raingeard, E. Renard, J.-F. Schved, P. Aguilar-Martinez, and the CHU Montpellier AOI 2004 Working Group 2060

Detailed Technical Analysis of Urine RNA-Based Tumor Diagnostics Reveals ETS2/Urokinase Plasminogen Activator to Be a Novel Marker for Bladder Cancer. M. Hanke, I. Kausch, G. Dahmen, D. Jocham, and J.M. Warnecke 2070

IL18 Haplotypes Are Associated with Serum IL-18 Concentrations in a Population-Based Study and a Cohort of Individuals with Premature Coronary Heart Disease. S.R. Thompson, P.A. McCaskie, J.P. Beilby, J. Hung, M. Jennens, C. Chapman, P. Thompson, and S.E. Humphries 2078

EVIDENCE-BASED LABORATORY MEDICINE AND TEST UTILIZATION

National Academy of Clinical Biochemistry Laboratory Medicine Practice Guidelines: Use of Cardiac Troponin and B-Type Natriuretic Peptide or N-Terminal proB-Type Natriuretic Peptide for Etiologies Other than Acute Coronary Syndromes and Heart Failure. NACB Writing Group Members: A.H.B. Wu, A.S. Jaffe, F.S. Apple, R.L. Jesse, G.L. Francis, D.A. Morrow, L. Kristin Newby, J. Ravkilde, W.H. Wilson Tang, and R.H. Christenson, NACB Committee Members: R.H. Christenson, F.S. Apple, C.P. Cannon, G.L. Francis, R.L. Jesse, D.A. Morrow, L. Kristin Newby, J. Ravkilde, A.B. Storrow, W.H. Wilson Tang, and A.H.B. Wu 2086

Renal Dysfunction Is a Confounder for Plasma Natriuretic Peptides in Detecting Heart Dysfunction in Uremic and Idiopathic Dilated Cardiomyopathies. M. Codognotto, A. Piccoli, M. Zaninotto, M. Mion, M. Plebani, U. Vertolli, F. Tona, L. Ruzza, A. Barchita, and G.M. Boffa 2097

Clinical Chemistry (ISSN 0009-9147) is published monthly by the American Association for Clinical Chemistry, 1850 K Street, Suite 625, Washington, DC 20006. New subscriptions, renewals, changes of address, back issues, and all customer service questions should be addressed to: AACC, Subscription Department, 1850 K Street, Suite 625, Washington, DC 20006. Telephone (202) 857-0717 or 1 (800) 892-1400; fax (202) 887-5093; e-mail [email protected]. Subscription rates: Institutional subscription USA $927, elsewhere $1069. Individual subscription USA $279, elsewhere $434. Airmail outside USA is an additional $260. Individual subscriptions are for personal use and not to be used in a library. Periodicals postage paid at Washington, DC and at additional mailing offices. Postmaster: Send address changes to Clinical Chemistry, 1850 K Street, Suite 625, Washington, DC 20006. Copyright © 2007 The American Association for Clinical Chemistry.

2A Clinical Chemistry December 2007

PROTEOMICS AND PROTEIN MARKERS

Modified Form of the Fibrinogen B␤ Chain (des-Gln B␤), a Potential Long-Lived Marker of Pancreatitis. D. Schmidt and S.O. Brennan 2105

Risk Stratification for Heart Failure and Death in an Acute Coronary Syndrome Population Using Inflammatory Cytokines and N-Terminal Pro-Brain Natriuretic Peptide. P.A. Kavsak, D.T. Ko, A.M. Newman, G.E. Palomaki, V. Lustig, A.R. MacRae, and A.S. Jaffe 2112

CANCER DIAGNOSTICS

MS-FLAG, a Novel Real-Time Signal Generation Method for Methylation-Specific PCR. C. Bonanno, E. Shehi, D. Adlerstein, and G.M. Makrigiorgos 2119

LIPIDS,LIPOPROTEINS, AND CARDIOVASCULAR RISK FACTORS

Development of a Homogeneous Assay to Measure Remnant Lipoprotein Cholesterol. K. Miyauchi, N. Kayahara, M. Ishigami, H. Kuwata, H. Mori, H. Sugiuchi, T. Irie, A. Tanaka, S. Yamashita, and T. Yamamura 2128

DRUG MONITORING AND TOXICOLOGY

Buprenorphine and Norbuprenorphine in Hair of Pregnant Women and Their Infants after Controlled Buprenorphine Administration. R.S. Goodwin, D.G. Wilkins, O. Averin, R.E. Choo, J.R. Schroeder, D.R. Jasinski, R.E. Johnson, H.E. Jones, and M.A. Huestis 2136

ENDOCRINOLOGY AND METABOLISM

Analytical Validation and Biological Evaluation of a High–Molecular-Weight Adiponectin ELISA. M.K. Sinha, T. Songer, Q. Xiao, J.H. Sloan, J. Wang, S. Ji, W.E. Alborn, R.A. Davis, M.M. Swarbrick, K.L. Stanhope, B.M. Wolfe, P.J. Havel, T. Schraw, R.J. Konrad, P.E. Scherer, and J.S. Mistry 2144

Association of C-Reactive Protein with Surrogate Measures of Insulin Resistance among Nondiabetic US Adults: Findings from National Health and Nutrition Examination Survey 1999–2002. Y.-X. Meng, E.S. Ford, C. Li, A. Quarshie, A.M. Al-Mahmoud, W. Giles, G.H. Gibbons, and G. Strayhorn 2152

Determination of Bioavailable Testosterone [Non–Sex Hormone–Binding Globulin (SHBG)-Bound Testosterone] in a Population of Healthy French Men: Influence of Androstenediol on Testosterone Binding to SHBG. F. Giton, S. Urien, C. Born, J. Tichet, J. Gue´chot, J. Callebert, F. Bronsard, J.P. Raynaud, and J. Fiet 2160

AUTOMATION AND ANALYTICAL TECHNIQUES

Metabolomics Identifies Perturbations in Human Disorders of Propionate Metabolism. W.R. Wikoff, J.A. Gangoiti, B.A. Barshop, and G. Siuzdak 2169

Visual Recognition and Efficient Isolation of Apoptotic Cells with Fluorescent-Magnetic-Biotargeting Multifunctional Nanospheres. E.-Q. Song, G.-P. Wang, H.-Y. Xie, Z.-L. Zhang, J. Hu, J. Peng, D.-C. Wu, Y.-B. Shi, and D.-W. Pang 2177

CLINICAL IMMUNOLOGY

Antibodies against Synthetic Deamidated Gliadin Peptides as Predictors of Celiac Disease: Prospective Assessment in an Adult Population with a High Pretest Probability of Disease. S. Niveloni, E. Sugai, A. Cabanne, H. Vazquez, J. Argonz, E. Smecuol, M.L. Moreno, F. Nachman, R. Mazure, Z. Kogan, J.C. Gomez, E. Maurin˜o, and J.C. Bai 2186

(Continued) 3A Contents Volume 53, Number 12

INFECTIOUS DISEASE

Midregional Pro-A-Type Natriuretic Peptide and Carboxy-Terminal Provasopressin May Predict Prognosis in Community-Acquired Pneumonia. M. Masia´, J. Papassotiriou, N.G. Morgenthaler, I. Herna´ndez, C. Shum, and F. Gutie´rrez 2193

TECHNICAL BRIEFS

Effects of 7 Hemoglobin Variants on the Measurement of Glycohemoglobin by 14 Analytical Methods. S.-T. Lee, C.W. Weykamp, Y.-W. Lee, J.-W. Kim, and C.-S. Ki 2202

Mass Spectrometry–Based Detection of Hemoglobin E Mutation by Allele-Specific Base Extension Reaction. J.C.H. Tsang, P. Charoenkwan, K.C.K. Chow, Y. Jin, C. Wanapirak, T. Sanguansermsri, Y.M.D. Lo, and R.W.K. Chiu 2205

Development and Validation of an Automated Thawing and Mixing Workcell. C.D. Hawker, W.L. Roberts, A. DaSilva, G.D. Stam, W.E. Owen, D. Curtis, B.-S. Choi, and T.A. Ring 2209

Detection of Factor VIII Gene Mutations by High-Resolution Melting Analysis. A.D. Laurie, M.P. Smith, and P.M. George 2211

LETTERS TO THE EDITOR

The Origin of Circulating Free DNA. M. van der Vaart and P.J. Pretorius 2215

A Receptor-Mediated Mechanism to Support Clinical Observation of Altered Albumin Variants. J.T. Andersen and I. Sandlie 2216

Stability of Plasma Homocysteine, S-Adenosylmethionine, and S-Adenosylhomocysteine in EDTA, Acidic Citrate, and Primavette™ Collection Tubes. U. Hu¨ bner, H. Schorr, R. Eckert, J. Geisel, and W. Herrmann 2217

From Syndrome to Spectrum: What Evolution Suggests about the Status of the Metabolic Syndrome. A. Matheson 2218

Effect of Corticosteroid Therapy on Low-Molecular–Weight Protein Markers of Kidney Function. A. Bo¨kenkamp, C.A.R.C. Laarman, K.I. Braam, J.A.E. van Wijk, W.A. Kors, M. Kool, J. de Valk, A.A. Bouman, M.D. Spreeuwenberg, and B. Stoffel-Wagner 2219

Ammonium 5-Bromo-7-fluorobenzo-2-oxa-1,3-diazole-4-sulphonate: A New Fluorogenic Reagent for the Determination of Aminothiols by HPLC. G. Cevasco, A.M. Mumot, C. Scapolla, and S. Thea 2221

44 Single-Nucleotide Polymorphisms Expressed by Placental RNA: Assessment for Use in Noninvasive Prenatal Diagnosis of Trisomy 21. A.T.J.I. Go, A. Visser, M.A.M. Mulders, M.A. Blankenstein, J.M.G. van Vugt, and C.B.M. Oudejans 2223

BOOK,SOFTWARE, AND WEB SITE REVIEWS

Tietz’s Applied Laboratory Medicine, 2nd ed. Mitchell G. Scott, Ann M. Gronowski, and Charles S. Eby, eds. K. Jung 2225

CORRECTION

Can Whole-Blood Samples Be Stored over 24 Hours without Compromising Stability of C-Reactive Protein, Retinol, Ferritin, Folic Acid, and Fatty Acids in Epidemiologic Research. M. van Eijsden, M.F. van der Wal, G. Hornstra, and G.J. Bonsel 2226

4A (Continued) Clinical Chemistry December 2007

THE CLINICAL CHEMIST

D.E. Bruns 2227

INVITED REVIEWERS—2007 2229

ACCENTா—CONTINUING EDUCATION CREDIT FOR READERS OF CLINICAL CHEMISTRY

For more information go to www.aacc.org/ccj/accent

5A Editorial

Untargeted Metabolomic Analysis Hits the Target

Propionic (PA) and methylmalonic (MMA) acidemias ways have been used mainly in targeted studies of are inborn errors of the metabolism of several important candidate alternate pathways such as studies of the in- amino acids, the side chain of the cholesterol molecule, volvement of the Krebs cycle in PA and MMA. Targeted and dietary odd chain-length fatty acids. PA results analysis is likely to provide incomplete knowledge of all from a deficiency of mitochondrial propionyl-CoA car- metabolic interactions that may result from a single gene boxylase, an enzyme that requires biotin as a cofactor defect. and converts propionyl-CoA to D-methylmalonyl-CoA. In this issue of Clinical Chemistry, Wikoff et al. (6) Classical MMA results from deficiency of methylmalonyl- provide proof of principle that untargeted metabolomic CoA mutase, a cobalamin- (vitamin B12) dependent en- analysis may generate additional information regarding zyme, which converts L-methylmalonyl-CoA to succinyl- the complexity of the metabolic ramifications of a meta- CoA. The 2 enzymes are almost adjacent in this important bolic disease. In this study, methanol-extracted plasma metabolic pathway, and the 2 defects are very similar samples from patients with PA or mutase-deficient MMA from the perspective of clinical presentation and meta- were subjected to an approximately 1.25-h long capillary bolic biomarkers (1). For example, in expanded screening C-18 reversed-phase gradient liquid chromatography sep- programs for metabolic diseases in newborns, observa- aration process. Peaks eluting from the column were tion of increased propionyl-carnitine (C3-carnitine) by detected by electrospray ionization time-of-flight mass tandem mass spectrometric analysis is predictive of the spectrometry, and spectral data accumulation was across occurrence of both defects (2). Currently, the 2 defects a mass range of m/z 75–1000. A total of more than 4000 are differentiated biochemically with targeted analysis features (peaks) were detected with the equivalent of just by gas chromatography/mass spectrometry of character- 8 ␮L of plasma. Some of these features had characteristic istic urinary organic acid. Specifically, large amounts of and known spectra for which positive identification was MMA acid and several propionate metabolites are possible and some were not known but were putatively found in the urine from patients with MMA, whereas identified. Other unidentified compounds were not increased MMA acid has not been found in tests of urine, subjected to further studies to help identify them. The and more recently dried blood spots, from patients with authors focused their major efforts on identification of PA (3). features that can be used to distinguish samples from Although both of these disorders have been studied for patients with PA or MMA from control samples and from more than 4 decades, several features and metabolic each other. issues remain unresolved. Before the introduction of As proof of the principle that untargeted analysis can diagnostic methods to measure urinary organic acids, hit the right note, the most characteristic and statistically these disorders were defined as a syndrome called ketotic significant feature that distinguished both types of disease hyperglycinemia (4). The mechanism that leads to increased blood glycine has never been determined, and samples from healthy samples was a compound subse- most patients remain hyperglycinemic throughout life. It quently confirmed to be propionyl-carnitine. This com- is not known if hyperglycinemia contributes at all to the pound is used for identification of both conditions in phenotype. It has been established, however, that several newborn screening programs and is a routine targeted alternate pathways of propionyl-CoA metabolism that metabolic marker for monitoring disease status (2). All exist as minor pathways in healthy individuals are used patients with these 2 diseases have consistently increased extensively in patients with these diseases. The extent to concentrations of propionyl-carnitine, which may fluctu- which the use of these alternative pathways leads to the ate according to metabolic status. clinical phenotype has not been determined. One of these A 2nd probable acyl-carnitine species, which may be alternate pathways involves the condensation of propio- the 6:1 or a branched-chain methyl carnitine, was also nyl-CoA with oxaloacetate to form methylcitrate. This identified as being significantly increased in both groups ␥ pathway uses enzymes of the Krebs cycle (5). Whether of patients. Less consistently, -butyrobetain, an interme- this additional input into a vital metabolic pathway is diate on the carnitine biosynthetic pathway, was also beneficial or harmful is not known. increased. The contribution of exogenous carnitine ther- Our understanding of the pathogenesis of metabolic apy, which is routinely provided for PA and MMA diseases is maturing. Initially, the concept was held that a patients, may partly explain these findings of carnitine single gene defect results only in a 1-step involvement in metabolic changes. PA and MMA patients are on chronic a single well-defined metabolic pathway. Now we have long-term high-dose therapy, whereas the control patients come to the realization that abnormalities in 1 step in a that took carnitine in this study did so for a short time only. single pathway are very likely to impact multiple other The relative effects of carnitine medication, long-term or pathways. Furthermore, we are recognizing that involve- short-term, on metabolic processes are unknown. It also ment of alternate pathways may additionally impact the remains to be seen if the basic metabolic defect leads to disease process. The evolving metabolomic tools that alteration of carnitine metabolism itself, as was indicated by allow us to simultaneous study multiple metabolic path- the increased concentrations of ␥-butyrobetaine.

Clinical Chemistry 53, No. 12, 2007 2037 2038 Bennett: Untargeted Metabolomic Analysis Hits the Target

A total of 71 features differentiated PA from MMA, a identification is important for clinical application of finding that is quite remarkable given the very close these predominantly proteomic biomarkers (11–14). proximity of the 2 metabolic defects. The most notable Further advances in untargeted metabolomic analysis feature that differed between the 2 was the presence of of metabolic diseases requires absolute identification of isovaleryl carnitine (IV-carnitine), which was increased important analytes so that we can unravel the metabolic by a factor of 5.1 in MMA but did not differ from complexity of these diseases and perhaps design better normal in PA. No other features that differed be- therapeutic regimens. We already have targeted bio- tween PA and MMA were positively identified in this markers for most of the metabolic diseases, and identi- study, but identification remains possible on the basis fication of markers for diagnosis is less of an issue. This of chromatographic retention time and the mass spec- field requires identification of the novel features that tral information that was obtained. We should antici- can help us explain the complex clinical phenotypes. pate further experimentation to reveal features that The present study, although providing few new in- uncover important metabolomic interactions in PA and sights, is the first to use untargeted analysis. That MMA. predictable features were found is good. It is now time Most intramitochondrial diseases with primary accu- to use this novel and potentially useful technique to mulation of coenzyme A intermediates, such as PA and dig deeper in the search for unidentified but clinically MMA, eventually lead to the accumulation of acylcarnitine species, presumably owing to the interaction of different features. otherwise minor carnitine esterification pathways. This study clearly identifies involvement of carnitine metabolism in the complex metabolic perturbations seen in PA References 1. Fenton WA, Gravel RA, Rosenblatt DS. Disorders of propionate and methyl- and MMA and appears to have identified a considerable malonate metabolism. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. The number of different features between PA and MMA. The Metabolic and Molecular Bases of Inherited Disease, 8th ed. New York: observation of increased IV-carnitine in MMA but not McGraw-Hill, 2001:2165–93. PA was clearly unanticipated and at the moment not 2. Chace DH, Kalas TA, Naylor EW. Use of tandem mass spectrometry for multianalyte screening of dried blood specimens from newborns. Clin Chem easily explained. Increased accumulation of IV-carnitine 2003;49:1797–1817. presumably results from increased accumulation of 3. La Marca G, Malvagia S, Pasquini E, Innocenti M, Donati MA, Zammarchi E. isovaleryl-CoA, which is the major targeted marker for Rapid 2nd tier test for measurement of 3-OH propionic and methylmalonic isovaleric acidemia, a disorder of leucine metabolism, and acids on dried blood spots: reducing the false-positive rate for propionylcarnitine during expanded newborn screening by liquid chromatography-tandem also of multiple acyl-CoA dehydrogenation defects result- mass spectrometry. Clin Chem 2007;53:1364–9. ing from impaired electron transport (7, 8). Leucine, how- 4. Rosenberg LE, Lilljeqvist A-C, Hsia YE. Methylmalonic aciduria: an inborn ever, is not a precursor of propionyl-CoA nor is it known error leading to metabolic acidosis, long-chain ketonuria and intermittent to be a precursor or downstream metabolic product of hyperglycinemia. N Engl J Med 1968;278:1319–22. L-methylmalonyl-CoA. Patients with MMA do not have 5. Ando T, Rasmussen K, Wright JM, Nyhan WL. Isolation and identification of methylcitrate, a major metabolic product of propionate in patients with all of the features of isovaleric acidemia, including the propionic academia. J Biol Chem 1972;247:2200–4. characteristic odor of sweaty feet, which is related to 6. Wikoff WR, Gangoiti JA, Barshop BA, Siuzdak G. Metabolomics identifies isovaleryl-CoA accumulation. The finding of increased perturbations in human disorders of propionate metabolism. Clin Chem IV-carnitine requires additional investigation to discern 2007;53:2169–76. 7. Sweetman L, Williams JC. Branched chain organic acidurias. In: Scriver CR, whether or not it is an important observation that will Beaudet AL, Sly WS, Valle D, eds. The Metabolic and Molecular Bases of lead to better understanding of the complexity of MMA Inherited Disease, 8th ed. New York: McGraw-Hill 2001:2125–63. and perhaps provide an insight into a different disease 8. Frerman FE, Goodman SI. Defects of electron transfer flavoprotein and mechanism for PA. electron transfer flavoprotein oxidoreductase: glutaric academia type II. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. The Metabolic and Molecular At the end of this study, there remained many Bases of Inherited Disease, 8th ed. New York: McGraw-Hill, 2001:2357–65. unidentified features. The identification of useful com- 9. Poon TCW, Yip TT, Chan ATC, Yip C, Yip V, Mok TSH, et al. Comprehensive ponents among this large list of potentially useful proteomic profiling identifies serum proteomic signatures for detection of biomarkers presents a major technological challenge, hepatocellular carcinoma and its subtypes. Clin Chem 2003;49:752–60. and whether technology can answer those important 10. Kam RKT, Poon TCW, Chan HLY, Wong N, Hui AY, Sung JJY. High throughput quantitative profiling of serum N-glycome by MALDI-TOF mass spectrometry clinical and metabolic questions remains unknown un- and N-glycomic fingerprint of liver fibrosis. Clin Chem 2007;53:1254–63. til these additional features are identified. In other 11. Koomen JM, Donghui L, Xiao L-C, Liu TC, Coombes KR, Abbruzzese J, et al. emerging areas in which untargeted study of proteomic Direct tandem mass spectrometry reveals limitations in protein profiling biomarkers seeks primarily to identify new disease experiments for plasma biomarker discovery. J Proteome Res 2005;4: 972–81. biomarkers to aid diagnosis, the need for absolute 12. Song J, Patel M, Rosenzweig CN, Chan-Li Y, Sokoll LJ, Fung ET, et al. analyte identification is controversial. To some investi- Quantification of fragments of human serum inter-␣-trypsin inhibitor heavy gators, it matters less exactly what component X might chain 4 by a surface-enhanced laser desorption/ionization-based immuno- be as long as it is measurable and shown to be the best assay. Clin Chem 2006;52:1045–53. biomarker for that particular disease state [Refs. (9, 10) 13. Villanueva J, Shaffer DR, Philip J, Chaparro CA, Erdjument-Bromage H, Olshen AB, et al. Differential exoprotease activities confer tumor-specific show some of the many recent examples in which serum peptidome patterns. J Clin Invest 2006;116:271–84. positive identification was not deemed necessary]. 14. Hortin GL. Can mass spectrometric protein profiling meet desired standards Other investigators express the view that molecular of clinical laboratory practice. Clin Chem 2005;51:3–5. Clinical Chemistry 53, No. 12, 2007 2039

Michael J. Bennett 3516 Civic Center Blvd. Philadelphia, PA 19104 Department of Pathology and Fax 215 590 19 Laboratory Medicine E-mail [email protected] The Children’s Hospital of Philadelphia and University of Pennsylvania DOI: 10.1373/clinchem.2007.096438 Abramson Pediatric Research Center Opinion

Glucose: A Simple Molecule That Is Not Simple to Quantify

Small increments in blood glucose substantially increase the NIST SRM for glucose in frozen human serum was not the risk of developing diabetes mellitus; but preanalytical available (3). This situation forced manufacturers to use and analytical variables, such as the absence of harmoni- older retained calibrators to establish set points for new zation for glucose assays, make it difficult to correctly calibrators. Some of these secondary calibrator set points apply these epidemiological insights to individual pa- were too low by as much as 4.5%. Other calibrator tients. Harmonization can be improved if 3 variables are problems include the variable instability of glucose in addressed: changing proficiency test grading from con- lyophilized serum samples and frozen serum samples sensus based to accuracy based, effectively controlling (4, 5). In 1996, SRM 909a, which was a lyophilized prod- glycolysis, and taking into account the time of day blood uct, manifested variable decreases in glucose content, a was collected. problem that was traced to a too-high moisture content in The continuous and graded quantitative relationship of approximately 10% of the vials (6). I recommend the fasting glucose measurements to the risk of developing parallel use of stabilized aqueous glucose standards to- diabetes was well documented recently by Tirosh et al. gether with protein-based calibrators—both traceable to (1). They found an increased risk of type 2 diabetes across NIST. The aqueous standards are more stable, and their quintiles of fasting plasma glucose (FPG) concentrations parallel use will ensure the accuracy and stability of the within the newly defined reference range, Ͻ5.55 mmol/L protein-containing calibrators. (Ͻ100 mg/dL). For example, a person with an FPG In addition, accuracy and precision must be separately between 4.83 and 5.00 mmol/L (87 and 90 mg/dL) has an graded. Neither analytical bias nor imprecision should be age-adjusted risk of developing diabetes that is 1.81 times permitted to consume the entire allowable total error of a that of a person with an FPG Ͻ4.55 mmol/L (82 mg/dL; proficiency test sample. In the absence of separately 95% CI 1.16–2.83). Thus, a difference as small as 0.28 defined limits for bias and imprecision it is possible for a mmol/L (5 mg/dL) nearly doubles the risk. Higher laboratory operating with very high precision to pass a concentrations of FPG were correlated with higher risk proficiency test challenge in spite of clinically unaccept- ratios, going as high as 3.05 times higher when FPG was able analytical bias (7). between 5.27 and 5.49 mmol/L (95 and 99 mg/dL; CI The most practical way to control glycolysis is to 1.78–5.18). Unfortunately, until glucose measurements measure glucose immediately in whole blood or to sepa- are harmonized these epidemiologically correct cut points rate serum or plasma from cells within 30 min of collec- cannot be applied with confidence to individual patients. tion (8), true even if the specimen is collected in a tube Three major variables must be addressed to achieve that contains sodium fluoride. It is well documented, but harmonization. First, proficiency test programs for glu- not widely recognized, that the rates of decrease of cose in the US should be accuracy based rather than glucose in the 1st hour after sample collection in tubes consensus based. Second, glycolysis in the specimen must with and without fluoride are virtually identical (9). The be effectively limited. Third, the time of day blood is decrease in glucose after2hinafluoride tube can collected must be taken into account. sometimes exceed 0.50 mmol/L (9 mg/dL). A change of The absence of an accuracy-based proficiency test pro- this magnitude straddles 2 of the 4 top quintiles described gram leads to unacceptably large differences among by Tirosh et al. (1, 10). methods. For example, in the most recent College of Glucose concentrations change in relation to the time of day fasting blood is collected. Researchers at the CDC American Pathologists survey C-B 2007, the mean glucose measured FPG in individuals who were randomly as- result for sample CHM-08—for different methods and signed to have blood collected either in the morning or instruments—ranged from a low of 7.92 mmol/L (144 afternoon (11). The mean fasting glucose was 5.41 mg/dL) to a high of 8.68 mmol/L (157.8 mg/dL) (2). The mmol/L (97.4 mg/dL) in the morning and 5.13 mmol/L potential difference is even greater when you include in (92.4 mg/dL) in the afternoon—a 5% difference. Most of the analysis the mean imprecision (expressed as CV) of the decrease took place by 10:00 AM. 1.9% for the method with the lowest result, and 2.5% for Even if we control all of these variables, fasting and 2-h the method with the highest result. Thus, one-third of the postprandial glucose concentrations inherently have more time the result for an individual measurement could biological day-to-day variability than does hemoglobin range between 7.75 mmol/L (141 mg/dL) and 8.91 A1c (HbA1c) (12). In addition, HbA1c has been harmo- mmol/L (162 mg/dL). This variation of 6.9% above or nized (13). Therefore, consideration should be given to below the mean indicates that one-third of the time the including HbA1c in any screening program for diabetes. difference between 2 different glucose results for an Inclusion of HbA1c, however, would not eliminate the individual patient—assayed in 2 different laboratories— need for rigorous harmonization of glucose testing. could differ by more than 14% even if the true difference was zero, an unacceptable situation. Much of the disharmony is attributable to differences in calibrator accuracy. One cannot assume that a calibrator’s Grant/funding support: None declared. set point is accurate. For example, in 2003 and part of 2004 Financial disclosures: None declared.

2040 Clinical Chemistry 53, No. 12, 2007 Clinical Chemistry 53, No. 12, 2007 2041

References 10. Gambino R, Reichberg S, Schwartz JG. Normal fasting plasma glucose level 1. Tirosh A, Shai I, Tekes-Manova D, Israeli E, Pereg D, Shochat T, et al. Normal and type 2 diabetes. N Engl J Med 2006;354:87–8. fasting plasma glucose levels and type 2 diabetes in young men. N Engl 11. Troisi RJ, Cowie CC, Haris MI. Diurnal variation in fasting plasma glucose. J Med 2005;353:1454–62. Implications for diagnosis of diabetes in patients examined in the afternoon. 2. College of American Pathologists. C-B Chemistry/Therapeutic Drug Monitor- JAMA 2000;284:3157. ing. CHM-06 to CHM-10 for glucose. Surveys2007:28–9. 12. Selvin E, Crainiceanu C, Brancati FL, Coresh J. Short-term variability in measures of glycemia and implications for the classification of diabetes. 3. Schwartz JG, Reichberg SB, Gambino RS. Glucose testing variability and the Arch Intern Med 2007;167:1545–51. need for an expert oversight committee. CAP Today 2005;19:12–6. 13. Sacks DB. Global harmonization of hemoglobin A1c. Clin Chem 2004;51: 4. Giampietro O, Navalesi R, Buzzigoli G, Boni C, Benzl L. Decrease in plasma 681–3. glucose concentration during storage at Ϫ20 oC. Clin Chem 1980;28: 1710–2. 5. Miller WG. How useful are reference materials? Clin Chem 1996;42: Raymond Gambino 1733–4. 6. Gills TE, May WE. Comments concerning report of defective NIST SRM 909a Quest Diagnostics, Inc. vials. Clin Chem 1996;42:1878. 7. Klee GG. Tolerance limits for short-term analytical bias and analytical 1300 East Newport Center Dr. imprecision derived from clinical assay specificity. Clin Chem 1993;39: Deerfield Beach, FL 33442-7727 1514–8. Fax 954-281-3873 8. Stahl M, Jørgensen LGM, Petersen PH, Brandslunc I, Olivarius DF, Borch- Hohnsen K. Optimization of preanalytical conditions and analysis of plasma E-mail [email protected]. glucose. 1. Impact of the new WHO and ADA recommendations on diagnosis of diabetes mellitus. Scand J Clin Lab Invest 2001;61:169–80. 9. Chan AYW, Swaminathan R, Cockram CS. Effectiveness of sodium fluoride DOI: 10.1373/clinchem.2007.094466 as a preservative of glucose in blood. Clin Chem 1989;35:315–7. Clinical Chemistry 53:12 2042–2050 (2007) Molecular Diagnostics and Genetics

Tentacle ProbesTM: Differentiation of Difficult Single-Nucleotide Polymorphisms and Deletions by Presence or Absence of a Signal in Real-Time PCR

Brent C. Satterfield,1,2 David A. Kulesh,3 David A. Norwood,3 Leonard P. Wasieloski, Jr.,3 Michael R. Caplan,1* and Jay A.A. West2

Background: False-positive results are a common prob- samples of near neighbors. With Tentacle Probes no lem in real-time PCR identification of DNA sequences false-positive results occurred. that differ from near neighbors by a single-nucleotide Conclusions: The high specificity exhibited by Tentacle polymorphism (SNP) or deletion. Because of a lack Probes may eliminate melting curve analysis for SNP of sufficient probe specificity, post-PCR analysis, such and deletion mutation detection, allowing the diagnos- as a melting curve, is often required for mutation tic use of previously difficult targets. differentiation. © 2007 American Association for Clinical Chemistry Methods: Tentacle ProbesTM, cooperative reagents with both a capture and a detection probe based on specific Use of real-time PCR for infectious disease identification cell-targeting principles, were developed as a replace- in a point-of-care clinical setting requires rapid analysis ment for 2 chromosomal TaqMan–minor groove binder with a very low rate of false positives. Additionally, (MGB) assays previously developed for Yersinia pestis interpretation of the results must be straightforward, with and Bacillus anthracis detection. We compared Taq- little to no post-PCR analysis. Finally, because initial Man-MGB probes to Tentacle Probes for SNP and concentrations and identities of clinical samples are un- deletion detection based on the presence or absence of a known, assay independence of initial concentration and growth curve. purity is essential. Identical criteria are also important for Results: With the TaqMan-MGB Y. pestis yp48 assays, use of DNA detectors in the field for detection of patho- false-positive results for Yersinia pseudotuberculosis oc- gens released for bioterror or biowarfare purposes. curred at every concentration tested, and with the Taq- The Roche LightCycler 2.0 and its field-adapted coun- Man-MGB B. anthracis gyrA assays, false-positive re- terpart, the Ruggedized Advanced Pathogen Identifica- sults occurred in 21 of 29 boil preps of environmental tion Device, are 2 devices with potential for point-of-care and field applications because of their automated software calling of growth curves. These instruments are based on a series of technologies designed largely by Wittwer et al. (1–3). The primary features of these devices include fast-cycling PCR with hot-air thermal cycling of 1 Harrington Department of Bioengineering, Arizona State University, reagents in glass capillaries. The detector software on both Tempe, AZ. devices has the advantage of being intuitive to users 2 Arcxis Biotechnologies, Pleasanton, CA. without technical backgrounds in real-time PCR analysis 3 United States Army Medical Research Institute of Infectious Diseases, Frederick, MD. because it can identify the presence or absence of a signal * Address correspondence to this author at: Arizona State University, PO for automated target identification. Box 879709, Tempe, AZ 85287-9709. Fax 480-727-7624; e-mail Michael. [email protected]. All opinions expressed in this paper are the authors’ and do not necessarily reflect the policies and views of DHS, DOE, DOD, or ORISE. Opinions, interpretations, conclusions, and recommendations are those of the author and 4 Nonstandard abbreviations: SNP, single-nucleotide polymorphism; are not necessarily endorsed by the U.S. Army. MGB, minor groove binder; RAPID, Ruggedized Advanced Pathogen Identi- Received May 4, 2007; accepted September 17, 2007. fication Device; USAMRIID, United States Army Medical Research Institute of Previously published online at DOI: 10.1373/clinchem.2007.091488 Infectious Disease.

2042 Clinical Chemistry 53, No. 12, 2007 2043

Automatic calling can be problematic, however, when anthracis under laboratory conditions; however, in this genetic sequences are very similar [i.e., differ by a single- study, we show that Hurtle et al.’s assay does not perform nucleotide polymorphisms (SNP)4 or deletion]. Assays nearly as well under more challenging experimental con- designed to detect these genetic differences often generate ditions, such as boil preps, that are commonly used to low-level signal in the near-neighbor genotype, and this screen environmental samples. signal is pronounced when high concentrations of near- Tentacle Probes act through cooperative binding with neighbor template are present. Although the difference in both a detection and a capture probe designed around the wild-type and low-level near-neighbor signals is discern- principles of specific cell targeting (12–16). These probes able when concentration and sample integrity are known, have been shown to have increased binding kinetics signals from samples of dubious origin in which PCR compared to standard stem-loop fluorescent DNA probes, efficiency may be inhibited could be more difficult to to have concentration-independent melting curves, and to resolve, and growth curves from samples with unknown increase specificity without sacrificing sensitivity (7). integrity and concentrations cannot be called with any These attributes allow the development of a binary style degree of confidence. In these cases, signal generated assay that generates a yes or no signal in the presence of from a high concentration of the near neighbor could be wild-type or variant forms. In this study we compare the interpreted as a low level of wild-type organism. Many rate of false positives generated by TaqMan-MGB and select agents have near neighbors whose genomes differ Tentacle Probes in real-time PCR applications. We com- only by SNPs or deletions (4–6). Any signal from such a pared the previously developed and optimized TaqMan- near neighbor would contribute to false positives with the MGB assay to single-iteration designs of Tentacle Probes detector software of the LightCycler or Ruggedized Ad- for detecting the gyrA gene of B. anthracis and the yp48 vanced Pathogen Identification Device. gene of Y. pestis (GenBank accession nos. AL031866, Recently our laboratories have designed and tested NC_004088, NC_003143, NC_005810, and NC_006155). assays for chromosomal genes (Bacillus anthracis gyrA These targets are chromosomal genes from CDC Category GenBank accession no. AY281534 and AY291535 and A pathogens that differ from their near neighbors by an Yersinia pestis yp48 GenBank accession nos. AL031866, SNP and deletion, respectively. This class of probes NC_004088, NC_003143, NC_005810, and NC_006155) should be generally applicable to clinical genetic testing in from 2 CDC Category A pathogens, Y. pestis (Bubonic which detection of SNPs and deletions is required. plague) and B. anthracis (Anthrax) using Tentacle Probes™, which we recently showed increased specificity experimental procedures in SNP differentiation using a nonamplified fluorescence Probe synthesis. Tentacle Probes are synthesized using detection assay (7). In this study we applied Tentacle standard controlled-pore glass chemistry (Biosearch Tech- Probes to real-time PCR for the 1st time. These assays nologies) and contain both a capture and detection probe. were selected because of ongoing difficulties with the The capture probe is a strand of linear DNA, and the selective detection of these genes in quantitative PCR detection probe contains a stem-loop structure with a using TaqMan–minor groove binder (MGB) probes (8). Y. fluorophore and a quencher. Each probe was dual HPLC pestis differs from its near neighbor Yersinia pseudotuber- purified and was not modified before use. The design of culosis by a 25-base deletion in the yp48 gene. Prior efforts the Tentacle Probes was based on generalizations of the to differentiate Y. pestis from Y. pseudotuberculosis with principles described by Satterfield et al. (7). Briefly, TaqMan-MGB probes demonstrated that post-PCR anal- Mfold (http://frontend.bioinfo.rpi.edu/applications/ ysis was required (8). Although TaqMan-MGB probes mfold/cgi-bin/dna-form1.cgi) was used to calculate have increased utility in differentiating genetic sequences melting temperatures of the detection probe stem, the differing by SNPs (9, 10), their ability to successfully capture probe-target duplex, and the detection probe- differentiate such sequences under more routine condi- target duplex independent of each other. The targeted tions is not clear. In the course of the studies by Chase et SNP/deletion was located in the center of the detection al. (8), 10 probes were synthesized to attempt the differ- probe. We designed the detection and capture probes entiation of Y. pestis from Y. pseudotuberculosis under with predicted melting temperatures 5 °C below the assay real-time PCR conditions, without success. temperature of 60 °C (55 °C) because the principles of B. anthracis differs from its near neighbor Bacillus cereus cooperativity indicate that the individual binding sites by an SNP in the gyrA gene (GenBank accession nos. should have affinities slightly weaker than the affinity AY291534 and AY291535). Hurtle et al. (11) synthesized 6 required to achieve binding under the conditions present probes in an attempt to find a B. anthracis probe that in the assay (7). The melting temperature for the stem- demonstrated good specificity from its near neighbor B. loop structure was chosen using the standard for Molec- cereus for this SNP. In the process, the annealing/fluores- ular Beacons, 7–10 °C above the assay conditions (70 °C) cence monitoring temperature was increased to 67 °C to (17). TaqMan-MGB probes and primers are designed as improve specificity, a procedure that departed from the reported in Chase et al. (8) and Hurtle et al. (11) using standard protocol and prevented multiplexing of the Applied Biosystems Primer Express software, version 2.0. assay. Hurtle et al. (11) did attain specificity for B. Genomic DNA samples for B. anthracis, B. cereus, Y. pestis, 2044 Satterfield et al.: Tentacle Probes Differentiate SNPs and Deletions

and Y. pseudotuberculosis were obtained from the United of PCR. Amplification products were run on a gel to States Army Medical Research Institute of Infectious verify successful PCR results. Disease (USAMRIID). PCR reagents were purchased from Idaho Technology. Probes and primers are summarized in gyrA assay. The real-time PCR probes developed by Hur- Table 1. tle et al. (11) for gyrA differentiation were used and directly compared with Tentacle Probes targeting the yp48 assay. The real-time PCR conditions and probes same amplicon under identical PCR conditions. Those developed by Chase et al. (8) for yp48 differentiation were conditions were 15 ␮L of master mix (10.2 ␮L deionized ␮ ϫ used and directly compared with Tentacle Probes target- water, 2 Lof10 reaction buffer in 50 mmol/L MgCl2, ing the same amplicon. The reaction conditions were 15 2 ␮Lof10ϫ dNTPs, 0.2 ␮L each of 50 ␮mol/L forward ␮L of master mix (10.2 ␮L deionized water, 2 ␮L10ϫ and reverse primers, 0.2 ␮Lof10␮mol/L probe, and 0.2 ␮ ϫ ␮ ␮ reaction buffer in 50 mmol/L MgCl2,2 Lof10 dNTPs, L of Platinum Taq polymerase) with 5 L of template. 0.2 ␮L each of 50 ␮mol/L forward and reverse primers, The temperature cycles included a 2-min denaturation at 0.2 ␮Lof10␮mol/L probe, 0.2 ␮LofTaq polymerase) 95 °C, followed by 45 cycles of 95 °C for 0 seconds and with 5 ␮L of template. TaqMan-MGB assay used Taq 60 °C for 20 seconds. Standard dilutions of template were platinum polymerase with a 2-min denaturation at 95 °C, used from 20 copies to 20 000 copies with 3 replicates followed by 45 cycles of 95 °C for 0 seconds and 60 °C for each. 20 seconds. The Tentacle Probe assay used Taq TSP Both assays were assessed for specificity with boil polymerase (exonuclease deficient) with a 2-min denatur- preps of 29 environmental liquid collected air samples ation at 95 °C, followed by 45 cycles of 95 °C for 0 seconds, known to contain near neighbors to B. anthracis. Boil preps 60 °C for 10 seconds, and 70 °C for 10 seconds (mecha- were prepared from 2 to 3 isolated colonies taken from nism in Fig. 1A). Taq TSP polymerase, which is exonucle- overnight cultures on streaked plates. Colonies were ase deficient, was used for Tentacle Probes to allow removed from the plate, suspended in 500 ␮L Dulbecco’s fluorescence monitoring at temperatures other than the PBS (8 g/L sodium chloride, 1.15 g/L sodium phosphate annealing temperature, such as during the extension step. diabasic, 0.2 g/L potassium chloride, 0.2 g/L potassium TaqMan-MGB probes required degradation, so Taq plati- phosphate monobasic), and centrifuged/washed twice. num polymerase was used in reactions with TaqMan- They were then suspended in 100 ␮L deionized water at MGB probes. Because TaqMan-MGB probes were de- 95 °C for 15 min. The samples were centrifuged at 15 000g graded, fluorescence monitoring at temperatures other for 10 min, and the supernatant was collected for PCR than the annealing temperature was not beneficial. Stan- analysis. The average nucleic acid concentration of each dard dilutions of template were used from 20 copies to boil prep was 500 mg/L, and 5 ␮L was used in each 20 000 copies with 3 replicates each for subsequent rounds reaction. The reaction conditions for both Tentacle Probes

Table 1. Probe and target sequences with near neighbors are provided.a Target sequence and near neighbor B. anthracis gyrA 5Ј-CCActtctacgcatgaccatattcTATTCTTCACTAataaagggaaagtataccgTACG 1710 B. cereus gyrA 5Ј-CCActtctacgcatgatcatattcTATTCTTCACTAataaggggaaagtataccgTACG 1710 Y. pestis yp48 5Ј-gagtattcgtctgggggGCGTGCGGGAAAtcgaggtcaggtg*************************gcacgTTAAAGTGG 840 Y. pseudoTB yp48 5Ј-gagtattcgtctgggggGCGTGCGGGAAAtcgaggtcaggtgGATACCGCCGCCGCTCGTTCAGGTGagcacgTTAAAGTGG 865

Primers and probes B. anthracis gyrA Tentacle Probe FAM-CTTCTA(CGCATGACCATATTC) gcgtagaag-BHQ-PEG9 -ATAAAGGGAAAGTATACCG-Carb3 TaqMan-MGB FAM-CGCATGACCATATTC-MGBNFQ Forward primer GGGAACAAATGATGATGATTTCGT Reverse primer ACTCTGGGATTTCATATCCTTTCGT Y. pestis yp48 Tentacle Probe GAGTATTCGTCTGGGGG-PEG9-T(FAM)-ccc CG(AGGTTCAGGTGAGCACG) ctcgggga-BHQ TaqMan-MGB FAM-AGGTTCAGGTGAGCACG-MGBNFQ Forward primer GCAGGAAATGCGCAATGC Reverse primer GGGCGGATCCCCACTTTA a Carb and PEG represent the linker types of carbon chains or poly(ethylene glycol), respectively. The numbers following the PEG and Carb abbreviations represent the polymer length. FAM and BHQ are the fluorophore and quencher pair used. MGBNFQ is MGB and nonfluorescent quencher. Lowercase bases in Tentacle Probes represent bases added to help form the stem. Parenthesis around nucleotides in the Tentacle Probe is the region analogous to the corresponding TaqMan-MGB Probe. Lowercase bases in the templates represent probe binding regions. Lowercases in italics are regions where the capture probe binds, and nonitalicized text is where the detection probe binds. Bold bases represent polymorphisms and asterisks represent deleted bases. Clinical Chemistry 53, No. 12, 2007 2045

Fig. 1. Mechanism of exonuclease- deficient (A) and exonuclease-active (B) PCR with Tentacle Probes. Tentacle Probes consist of a linked capture and a detection probe. Under denaturing conditions, secondary structure of Tentacle Probes is removed and template is denatured. At the annealing temperature, the capture probe binds to the template providing affinity while the strong secondary structure of the detection probe provides specificity. The detection probe opens in the presence of the wild type, but remains closed in the presence of near neighbors. (A), because the probe is not digested, fluorescence can be read at any temperature or stage, not just at the primer annealing temperature, facilitating optimization of reaction specificity. The probe melts off at the extension temperature allowing primer extension. (B), the specificity of exonuclease-active PCR is limited to the primer annealing temperature because the probe hybridizes and is digested at this temperature, separating the fluorophore from the quencher, causing an increase in fluorescence. The detection region of a Tentacle Probe remains closed in the presence of a mismatch, causing the detection probe to be cleaved off, increasing the specificity of the assay. (In general, Tentacle Probes require binding to the proximal capture probe to open the detection probe; if the capture probe and detection probe are separated, then the detection probe cannot open.) and TaqMan-MGB were 15 ␮L of master mix (10.2 ␮L and variant produced no false negatives or false positives deionized water, 2 ␮L10ϫ reaction buffer in 50 mmol/L (Fig. 2, C and D). When boil preps of 29 environmental ␮ ϫ ␮ ␮ MgCl2,2 Lof10 dNTPs, 0.2 L each of 50 mol/L samples were used, however, TaqMan-MGB results in- forward and reverse primers, 0.2 ␮Lof10␮mol/L probe, cluded 21 false positives, whereas Tentacle Probes had no and 0.2 ␮LofTaq Platinum polymerase) with 5 ␮Lof false positives for any of the samples (Fig. 3). A gel was template undergoing a 2-min denaturation at 95 °C, fol- run for all the PCR products from the boil preps for both lowed by 45 cycles of 95 °C for 0 seconds and 60 °C for 20 TaqMan-MGB and Tentacle Probes (Fig. 4). The presence seconds. A 2nd set of experiments was performed with an of amplification products in each indicated that the DNA annealing temperature of 67 °C. One positive control was equally amplified in both experiments. containing 20 000 copies of the wild-type B. anthracis When Hurtle et al. (11) failed to achieve specificity of chromosome was run simultaneously with 29 boil preps reaction with a 60 °C annealing temperature even after 6 from environmental samples known to contain near designs of TaqMan-MGB, they used the best probe design neighbors to B. anthracis. Amplification products were run at an increased annealing temperature, 67 °C. Accord- on a gel to verify successful PCR results. Tentacle Probe ingly, we repeated the TaqMan-MGB gyrA experiment reaction mechanism with exonuclease active polymerase with 29 boil preps of environmental samples at this high is shown in Fig. 1B. annealing temperature and still obtained 7 false positives for the 29 samples (data not shown) compared to zero Results false positive results for Tentacle Probes at the more The Y. pestis TaqMan-MGB exhibited results similar to conventional 60 °C. those described previously (8). At all concentrations of near-neighbor Y. pseudotuberculosis, false positives oc- Discussion curred approximately 3 cycles later than detection of an The ability to perform PCR without false-positive results equivalent concentration of Y. pestis. In contrast, Tentacle is imperative for clinical and field diagnostics. Further- Probes yielded no false positives at any concentration more, the need to perform highly selective assays is tested (Fig. 2, A and B). Clean bands approximately 100 complicated by the presence of near neighbors that often bases in size appeared for each PCR product when run in differ from the target organism by a single SNP or an agarose gel, indicating that lack of amplification was deletion. In this study we have shown improvements in 2 not the cause of the increased specificity (data not shown). assays for the detection of priority infectious disease With both TaqMan-MGB and Tentacle Probes, the gyrA pathogens, with which false positives from near neigh- assay using 20–20 000 purified copies of both wild type bors have been greatly reduced if not eliminated. 2046 Satterfield et al.: Tentacle Probes Differentiate SNPs and Deletions

Fig. 2. Standard curve for Y. pestis (ࡗ) and Y. pseudotuberculosis (Ⅺ) for TaqMan-MGB assay (A) and Tentacle Probe assay (B). Only 1 standard curve is shown for Tentacle Probes because no false positives were recorded at any concentration. PCR curves are shown for 20 000 copies of Y. pestis (top curve) and an equal number of Y. pseudotuberculosis (bottom curve)inthebottom right hand corner of each graph. Standard curve for B. anthracis for TaqMan-MGB (C) and Tentacle Probes (D) for purified samples. Representative PCR curves are shown for each in the bottom right hand corner of the graphs, where 4 concentrations of B. anthrax are shown in contrast to no signal for either probe type of 20 000 copies of B. cereus.

Y. pestis is an example of an organism for which few allowing the 2 flanking sequences matching the probe to differences in the genome exist from its near neighbors form a perfect match to the probe. (b) Additionally, 7 (18). Although detection can be performed based on bases flanking the 5Ј end of the insertion are repeated in virulence plasmids (19–21), concerns regarding the devel- the 3Ј end of the insertion, causing exact matches to opment of genetically modified organisms and the impli- shorter probes, an especially problematic feature for cation that the plasmids alone are not responsible for the probes that depend on short lengths for increased speci- virulence of Y. pestis (5) require the addition of a chromo- ficity. TaqMan-MGB is an example of a probe that re- somal assay. The yp48 gene contains one of the greatest quires short lengths for improved differentiation, thus differences between Y. pestis and Y. pseudotuberculosis a explaining why false positives were present even after 10 25-bp deletion in Y. pestis (22). design iterations of TaqMan-MGB probes (8). Because It might be thought that detecting a 25-bp deletion Tentacle Probes use a strong hairpin for enhanced differ- would be easy; however, detecting the 25 deleted bases in entiation and do not require short probes, they are ideal the near neighbor does not rule out the presence of Y. candidates for specifically detecting Y. pestis. pestis. The differentiation of the 2 species in this region is In spite of the progress Tentacle Probes achieve in difficult for 2 reasons (Fig. 5). (a) The insertion is flexible, differentiation without post-PCR analysis, there is still Clinical Chemistry 53, No. 12, 2007 2047

threshold was supported by repeats of the experiment with exonuclease-active polymerase, which resulted in similar cycle thresholds for both Tentacle Probes and TaqMan-MGB (data not shown). Further optimization of Tentacle Probes to avoid this limitation is possible, and it should be noted that the Tentacle Probes studied in this report were not optimized (1st-iteration design). Anthrax detection can be performed by amplifying the virulence plasmids (23–25), but as with Y. pestis, there is concern over genetic modification of near neighbors and false negatives from B. anthracis without one or more of the plasmids. Moreover, B. cereus and Bacillus thuringiensis isolates have been found that harbor 1 or both B. anthracis plasmids with and without anthrax toxin genes (26–31). Accordingly, a chromosomal assay may augment the ability to distinguish B. anthracis from other environmental Bacillus isolates, but unfortunately there are few specific chromosomal targets to differentiate B. anthracis from other Bacillus species. Until now, the gyrA sequence in B. anthracis has appeared to be unique; however, the single- base mutation present in near neighbors has proved difficult to reliably reject even after 6 TaqMan-MGB probe designs (11). Although TaqMan-MGB has been used with greater success in the gyrA assay at an annealing temperature of 67 °C, this temperature departs from our standard protocol, precluding the option of multiplexing this chromosomal assay with virulence plasmid genes, and still reported 7 false positives out of 29 samples. For this reason, we chose to contrast the TaqMan-MGB assay and the Tentacle Probe assay at 60 °C. TaqMan-MGB has been used with success to differentiate SNPs in other organisms (9, 32) and to detect B. anthracis via gyrA in laboratory samples; however, when boil preps of these samples were used, TaqMan-MGB reliability diminished sharply. One possibility for this failure is that variants other than the 49 strains of B. cereus and B. thuringiensis (testing reported in the original publication) were encountered (11). Another explanation is that the large amount of starting material present in boil preps may cause nonspecific amplification to occur, an explanation we consider unlikely, because gel images of the PCR product show that the samples that caused the most false positives (circled in Fig. 4) were predominately Fig. 3. Boil preps of 29 environmental samples of various strains of B. comprised of a single PCR product and were at approxi- cereus and B. thuringensis and 1 positive control (B. anthracis) were mately the same concentration as those that did not cause run for TaqMan-MGB (top) and Tentacle Probes (bottom). false positives. TaqMan-MGB experienced 21 false positives out of 29 samples. Tentacle Probes Alternatively, TaqMan-MGB loss in specificity may be had no false positives. due to the presence of cell lysate, which contains salts, surfactants, and other molecules that may influence the room for improvement. For example, the Y. pestis assay binding behavior of the probe. Because the cells were for Tentacle Probes required up to 7 additional cycles for lysed in deionized water, however, surfactant or chemical detection (Fig. 2). Because this experiment was performed contamination is not possible, and salt carryover from with exonuclease-deficient polymerase and growth previous steps should be minimal. Additionally, the DI- curves were monitored during the extension step, we NAMelt server does not register any change in the pre- believe the longer cycle detection times were attributable dicted melting temperature for such a small amount of to decreased amplification efficiency due to high probe salt contamination (http://frontend.bioinfo.rpi.edu/ap- melting temperatures. This explanation for loss in cycle plications/hybrid/twostate.php). A final hypothesis is 2048 Satterfield et al.: Tentacle Probes Differentiate SNPs and Deletions

Fig. 4. DNA gel of PCR product from boil preps of the gyrA assay for both TaqMan-MGB (top) and Tentacle Probes (bottom)at60°C. Samples that caused false positives at both 60 °C and 67 °C are circled. Ladder size is from 100 to 1500 bp in 100-bp increments. The presence of amplicon indicates that the lack of false positives in Tentacle Probes was not from failure to amplify but due to probe specificity.

the presence of cellular proteins and debris; however, the Whatever the cause of the failure mode of the TaqMan- amount of these should be minimal after a rigorous MGB assay, Tentacle Probes maintained specificity under centrifugation step. Any soluble proteins should have identical conditions. We previously reported Tentacle been denatured by heat during the lysis step. Further- Probes with concentration-independent specificity, which more, if an interaction with cellular proteins causes a PCR prevents a loss of specificity, even in the presence of probe to fail, then its utility in clinical diagnostics is unexpected levels of nonspecific amplification (7). Sec- questionable. ond, the large stems in the hairpin of the detection probe provide a stabilized form that may be less susceptible to nonspecific binding in general. Although we cannot state the exact cause of the loss in specificity of TaqMan-MGB, we are certain that Tentacle Probes experienced no false positives under identical test conditions as are used with the TaqMan-MGB probes. There are a variety of other probes and PCR techniques that have been successfully used in the differentiation of SNPs and deletions (33); however, Tentacle Probes is the only probe, to our knowledge, that exhibits cooperative binding between probe and target DNA. Simpler meth- Fig. 5. Design of probes specific for deletions is difficult. ods, such as allele-specific amplification, fail to confer the The 25 deleted bases in Y. pestis are present in Y. pseodotuberculosis (dashed line) and allow the 2 regions flanking the insertion matching the probe (solid required specificity, because they do not completely in- lines) to come together, forming a perfect match, and consequently false hibit amplification of the variant (34, 35). In clinical sam- positives, to a typical Y. pestis detection probe. Additionally, there is a repeat of the 5Ј flanking region (gray) in the insert, causing a perfect match to shorter ples, for which starting concentrations and purity are probes and consequently false positives. unknown, delayed amplification cannot differentiate be- Clinical Chemistry 53, No. 12, 2007 2049

tween the presence of a small amount of wild type and a Portions of this research were funded by the Department of large amount of variant. Of the hybridization probes, Homeland Security under a Small Business Innovative Re- Scorpion primers and molecular beacons have shown search Grant awarded to Arcxis Biotechnologies under con- excellent SNP differentiation in controlled laboratory en- tract no. NBCHC060031. Portions of the research described vironments (36–39). However, although TaqMan-MGB herein were sponsored by the Defense Threat Reduction showed excellent SNP differentiation of the gyrA gene in Agency project 8.10006 07 RD B. a controlled environment across a panel of 49 strains of Financial disclosures: Arcxis Biotechnologies is a for-profit near neighbors without any false positives (11), it does corporation and owns the rights to Tentacle Probes, a com- not perform equally well with field samples, as demon- mercial product. strated in this study. Similarly, although there are no data Acknowledgments: We thank Dr. Seth Stern for helpful to indicate that Scorpion primers would yield false posi- review of this manuscript. We also thank Leslie Wachter for tives with field samples, there also are no data to indicate facilitating research at USAMRIID facilities. they will be successful. The basic difference between Scorpion primers and References Tentacle Probes is in the interaction of the stem-loop 1. Wittwer CT, Fillmore GC, Hillyard DR. Automated polymerase chain region with its target. Scorpion primers consist of a reaction in capillary tubes with hot air. Nucleic Acids Res 1989; stem-loop detection region connected to a primer; the 17:4353–7. stem-loop region interacts with the extended primer re- 2. Wittwer CT, Fillmore GC, Garling DJ. Minimizing the time required for DNA amplification by efficient heat transfer to small samples. sulting in a unimolecular detection event. Although sim- Anal Biochem 1990;186:328–31. ilar in appearance, Tentacle Probes exhibit a cooperative 3. Wittwer CT, Garling DJ. Rapid cycle DNA amplification: time and interaction with the target DNA through their capture temperature optimization. Biotechniques 1991;10:76–83. segment (compare to the primer region of the Scorpion 4. Helgason E, Okstad OA, Caugant DA, Johansen HA, Fouet A, Mock primer) and also through their stem-loop region (instead M, et al. Bacillus anthracis, Bacillus cereus, and Bacillus thurin- of binding to the extended primer as in the Scorpion giensis: one species on the basis of genetic evidence. Appl primer). Thus, Tentacle Probes have 2 interactions be- Environ Microbiol 2000;66:2627–30. tween the probe and the target DNA compared to the 1 5. Chain PS, Carniel E, Larimer FW, Lamerdin J, Stoutland PO, Regala WM, et al. Insights into the evolution of Yersinia pestis through interaction for Scorpion primers. As described above, we whole-genome comparison with Yersinia pseudotuberculosis. Proc cannot be certain as to the cause of the improved perfor- Natl Acad Sci U S A 2004;101:13826–31. mance of Tentacle Probes over TaqMan-MGB, but a 6. Cowley R, Greenaway PJ. Nucleotide sequence comparison of strong candidate is the cooperativity generated by this homologous genomic regions from variola, monkeypox, and vac- 2nd interaction between probe and target DNA. If this cinia viruses. J Med Virol 1990;31:267–71. hypothesis is correct, Tentacle Probes would likely exhibit 7. Satterfield BC, West JAA, Caplan MR. Tentacle Probes: eliminating advantages over Scorpion primers similar to the advan- false positives without sacrificing sensitivity. Nucleic Acids Res tages over TaqMan-MGB. 2007;35:e76. 8. Chase CJ, Ulrich MP, Wasieloski LP Jr, Kondig JP, Garrison J, Lindler LE, et al. Real-time PCR assays targeting a unique chro- In summary, we developed a 1st iteration Tentacle Probe mosomal sequence of Yersinia pestis. Clin Chem 2005;51: design that detects 2 priority select agents without any 1778–85. false positives, in contrast to multiple iteration designs of 9. Kutyavin IV, Afonina IA, Mills A, Gorn VV, Lukhtanov EA, Belousov TaqMan-MGB, which continue to generate false-positive ES, et al. 3Ј-minor groove binder-DNA probes increase sequence results from near neighbors. The high degree of specificity specificity at PCR extension temperatures. Nucleic Acids Res attainable with Tentacle Probes will allow more precise 2000;28:655–61. identification of genetic mutations, including SNPs and 10. Yao Y, Nellaker C, Karlsson H. Evaluation of minor groove binding deletions, without requiring melting-curve analysis. probe and Taqman probe PCR assays: influence of mismatches and template complexity on quantification. Mol Cell Probes 2006; 20:311–6. 11. Hurtle W, Bode E, Kulesh DA, Kaplan RS, Garrison J, Bridge D, et Grant/funding support: M.R.C. is funded by the National al. Detection of the Bacillus anthracis gyrA gene by using a minor Institute of Dental and Craniofacial Research Career Devel- groove binder probe. J Clin Microbiol 2004;42:179–85. opment and Faculty Transition Award (K22 DE014846) and a 12. Mammen M, Choi S, Whitesides GM. Polyvalent Interactions in National Institutes of Health Exploratory Research Grant Biological Systems: Implications for Design and Use of Multivalent (R21 NS051310). B.C.S. performed this research while on Ligands and Inhibitors. Angewandte Chemie International Edition appointment as a US Department of Homeland Security 1998;37:2754–94. (DHS) Fellow under the DHS Scholarship and Fellowship 13. Kiessling LL, Gestwicki JE, Strong LE. Synthetic multivalent ligands in the exploration of cell-surface interactions. Curr Opin Program, a program administered by the Oak Ridge Institute Chem Biol 2000;4:696–703. for Science and Education (ORISE) for DHS through an 14. Holland PM, Abramson RD, Watson R, Gelfand DH. Detection of interagency agreement with the US Department of Energy specific polymerase chain reaction product by utilizing the 5Ј-3Ј (DOE). ORISE is managed by Oak Ridge Associated Univer- exonuclease activity of Thermus aquaticus DNA polymerase. Proc sities under DOE contract number DE-AC05-00OR22750. Natl Acad Sci U S A 1991;88:7276–80. 2050 Satterfield et al.: Tentacle Probes Differentiate SNPs and Deletions

15. Chandler DP, Newton GJ, Small JA, Daly DS. Sequence versus pXO2 plasmid and genomic sequence from closely related bacte- structure for the direct detection of 16S rRNA on planar oligonu- ria. BMC Genomics 2002;3:34. cleotide microarrays. Appl Environ Microbiol 2003;69:2950–8. 28. Hoffmaster AR, Hill KK, Gee JE, Marston CK, De BK, Popovic T, et 16. Caplan MR, Rosca EV. Targeting drugs to combinations of recep- al. Characterization of Bacillus cereus isolates associated with tors: a modeling analysis of potential specificity. Ann Biomed Eng fatal pneumonias: strains are closely related to Bacillus anthracis 2005;33:1113–24. and harbor B. anthracis virulence genes. J Clin Microbiol 2006; 17. Vet JA, Marras SA. Design and optimization of molecular beacon 44:3352–60. real-time polymerase chain reaction assays. Methods Mol Biol 29. Hoffmaster AR, Ravel J, Rasko DA, Chapman GD, Chute MD, 2005;288:273–90. Marston CK, et al. Identification of anthrax toxin genes in a 18. Achtman M, Zurth K, Morelli G, Torrea G, Guiyoule A, Carniel E. Bacillus cereus associated with an illness resembling inhalation Yersinia pestis, the cause of plague, is a recently emerged clone anthrax. Proc Natl Acad Sci U S A 2004;101:8449–54. of Yersinia pseudotuberculosis. Proc Natl Acad Sci U S A 1999; 30. Van der Auwera GA, Andrup L, Mahillon J. Conjugative plasmid 96:14043–8. pAW63 brings new insights into the genesis of the Bacillus 19. Leal NC, Almeida AM. Diagnosis of plague and identification of anthracis virulence plasmid pXO2 and of the Bacillus thuringiensis virulence markers in Yersinia pestis by multiplex-PCR. Rev Inst plasmid pBT9727. BMC Genomics 2005;6:103. Med Trop Sao Paulo 1999;41:339–42. 31. Klee SR, Ozel M, Appel B, Boesch C, Ellerbrok H, Jacob D, et al. Characterization of Bacillus anthracis-like bacteria isolated from 20. Melo AC, Almeida AM, Leal NC. Retrospective study of a plague wild great apes from Cote d’Ivoire and Cameroon. J Bacteriol outbreak by multiplex-PCR. Lett Appl Microbiol 2003;37:361–4. 2006;188:5333–44. 21. Tomaso H, Reisinger EC, Al Dahouk S, Frangoulidis D, Rakin A, 32. de Kok JB, Wiegerinck ET, Giesendorf BA, Swinkels DW. Rapid Landt O, et al. Rapid detection of Yersinia pestis with multiplex genotyping of single nucleotide polymorphisms using novel minor real-time PCR assays using fluorescent hybridisation probes. groove binding DNA oligonucleotides (MGB probes). Hum Mutat FEMS Immunol Med Microbiol 2003;38:117–26. 2002;19:554–9. 22. Buchrieser C, Rusniok C, Frangeul L, Couve E, Billault A, Kunst F, 33. Marras SA, Tyagi S, Kramer FR. Real-time assays with molecular et al. The 102-kilobase pgm locus of Yersinia pestis: sequence beacons and other fluorescent nucleic acid hybridization probes. analysis and comparison of selected regions among different Clin Chim Acta 2006;363:48–60. Yersinia pestis and Yersinia pseudotuberculosis strains. Infect 34. Newton CR, Graham A, Heptinstall LE, Powell SJ, Summers C, Immun 1999;67:4851–61. Kalsheker N, et al. Analysis of any point mutation in DNA. The 23. Bell CA, Uhl JR, Hadfield TL, David JC, Meyer RF, Smith TF, et al. amplification refractory mutation system (ARMS). Nucleic Acids Detection of Bacillus anthracis DNA by LightCycler PCR. J Clin Res 1989;17:2503–16. Microbiol 2002;40:2897–902. 35. Germer S, Holland MJ, Higuchi R. High-throughput SNP allele- 24. Beyer W, Glockner P, Otto J, Bohm R. A nested PCR method for the frequency determination in pooled DNA samples by kinetic PCR. detection of Bacillus anthracis in environmental samples col- Genome Res 2000;10:258–66. lected from former tannery sites. Microbiol Res 1995;150:179– 36. Whitcombe D, Theaker J, Guy SP, Brown T, Little S. Detection of 86. PCR products using self-probing amplicons and fluorescence. Nat 25. Reif TC, Johns M, Pillai SD, Carl M. Identification of capsule- Biotechnol 1999;17:804–7. forming Bacillus anthracis spores with the PCR and a novel 37. Tyagi S, Kramer FR. Molecular beacons: probes that fluoresce dual-probe hybridization format. Appl Environ Microbiol 1994;60: upon hybridization. Nat Biotechnol 1996;14:303–8. 1622–5. 38. Thelwell N, Millington S, Solinas A, Booth J, Brown T. Mode of 26. Pannucci J, Okinaka RT, Sabin R, Kuske CR. Bacillus anthracis action and application of Scorpion primers to mutation detection. pXO1 plasmid sequence conservation among closely related Nucleic Acids Res 2000;28:3752–61. bacterial species. J Bacteriol 2002;184:134–41. 39. Marras SA, Kramer FR, Tyagi S. Multiplex detection of single- 27. Pannucci J, Okinaka RT, Williams E, Sabin R, Ticknor LO, Kuske nucleotide variations using molecular beacons. Genet Anal 1999; CR. DNA sequence conservation between the Bacillus anthracis 14:151–6. Clinical Chemistry 53:12 2051–2059 (2007) Molecular Diagnostics and Genetics

Development of a Focused Oligonucleotide-Array Comparative Genomic Hybridization Chip for Clinical Diagnosis of Genomic Imbalance

Yiping Shen,1,2† David T. Miller,1,3,4† Sau Wai Cheung,5 Va Lip,1 Xiaoming Sheng,1 Keith Tomaszewicz,1 Hong Shao,1 Hong Fang,1 Hung Siv Tang,1 Mira Irons,3,4 Christopher A. Walsh,3,4,6 Orah Platt,1,4 James F. Gusella,2,4 and Bai-Lin Wu1,4*

Background: Submicroscopic genomic imbalance un- imbalance events, as detected by bacterial artificial derlies well-defined microdeletion and microduplica- chromosome–based array CGH, FISH, or multiplex tion syndromes and contributes to general developmen- ligation-dependent probe amplification. tal disorders such as mental retardation and autism. Results: Focused array CGH detected all known regions Array comparative genomic hybridization (CGH) com- of genomic imbalance in 51 validation samples with plements routine cytogenetic methods such as karyotyp- 100% concordance and an excellent signal-to-noise ratio. ing and fluorescence in situ hybridization (FISH) for the The mean SD among log2 ratios of all noncontrol fea- detection of genomic imbalance. Oligonucleotide arrays tures without copy number alteration was 0.062 (me- in particular offer advantages in ease of manufacturing, dian, 0.055). Clinical testing of another 211 samples from but standard arrays for single-nucleotide polymorphism individuals with developmental delay, unexplained genotyping or linkage analysis offer variable coverage mental retardation, dysmorphic features, or multiple in clinically relevant regions. We report the design congenital anomalies revealed genomic imbalance in 25 and validation of a focused oligonucleotide-array samples (11.9%). CGH assay for clinical laboratory diagnosis of genomic Conclusions: This focused oligonucleotide-array CGH imbalance. assay, a flexible, robust method for clinically diagnos- Methods: We selected >10 000 60-mer oligonucleotide ing genetic disorders associated with genomic imbal- features from Agilent’s eArray probe library to interro- ance, offers appreciable advantages over currently avail- gate all subtelomeric and pericentromeric regions and able platforms. 95 additional clinically relevant regions for a total of 179 © 2007 American Association for Clinical Chemistry loci. Sensitivity and specificity were measured for 105 patient samples, including 51 with known genomic- Genomic imbalance causes a variety of human genetic disorders, ranging from imbalance of entire chromosomes, as in Down syndrome, to submicroscopic rearrangements, as in the 22q11 deletion that causes 1 Department of Laboratory Medicine, Children’s Hospital Boston, Boston, MA. DiGeorge/velocardiofacial syndrome. Genomic imbal- 2 Center for Human Genetic Research, Massachusetts General Hospital, ance also causes idiopathic mental retardation (1, 2) and Boston, MA. is detectable in approximately 3%–4% of cases (3) by 3 Division of Genetics, Children’s Hospital Boston, Boston, MA. 4 traditional cytogenetic methods, such as karyotype and Departments of Medicine, Neurology, Pathology, Pediatrics, Harvard 7 Medical School, Boston, MA. fluorescence in situ hybridization (FISH) analyses. These 5 Department of Molecular and Human Genetics, Baylor College of Med- traditional cytogenetic methods are labor intensive, espe- icine, Houston, TX. cially when multiple genomic regions are interrogated. 6 Howard Hughes Medical Institute, Chevy Chase, MD. † These authors contributed equally to this work. * Address correspondence to this author at: Departments of Laboratory Medicine and Pathology, Children’s Hospital and Harvard Medical School, 300 Longwood Ave., Boston, MA 02115. Fax 617-730-0338; e-mail bai- 7 Nonstandard abbreviations: FISH, fluorescence in situ hybridization; [email protected]. MLPA, multiplex ligation-dependent probe amplification; CGH, comparative Received April 20, 2007; accepted September 11, 2007. genomic hybridization; BAC, bacterial artificial chromosome; CNV, copy- Previously published online at DOI: 10.1373/clinchem.2007.090290 number variant; SNR, signal-to-noise ratio.

2051 2052 Shen et al.: Oligonucleotide-Array CGH Chip for Clinical Diagnosis

Molecular techniques such as multiplex ligation- clinical samples dependent probe amplification (MLPA) and real-time After assay validation, we performed clinical array CGH PCR are alternatives to multiple FISH assays for evaluat- testing of 211 consecutively submitted samples from ing the genomic copy number of multiple targets (4). presumably unrelated children. Samples were submitted Microarray-based comparative genomic hybridization after referral to specialists in the Divisions of Clinical (CGH) offers the ability to interrogate many more Genetics and Developmental Medicine, and the Depart- genomic regions in a single assay. Early CGH arrays were ment of Neurology for clinical molecular-diagnostic test- composed of large-insert bacterial artificial chromosome ing. The referring diagnoses for these patients included (BAC) clones (5). BAC-based arrays have revolutionized developmental delay, mental retardation, dysmorphic the detection of genomic imbalance in clinical cytogenetic features, or multiple congenital anomalies. All samples laboratories (6, 7)but are challenging to develop, validate, were compared with a reference sample for standard and manufacture. The fact that BAC clones in standard 2-color array CGH, either a 46,XY male or a 46,XX female libraries may be inaccurately mapped could lead to diag- sample. Reference DNA was purchased from Promega. nostic errors without careful validation (8). Additionally, Genomic DNA was extracted from whole blood for all BAC clone inserts average approximately 150 kb, limiting samples with a D50K PureGene DNA-isolation reagent the resolution of detectable copy-number variants (CNVs) set (Qiagen/Gentra) according to the manufacturer’s in- Ϫ to the size of a BAC insert. Deletion breakpoints that structions. All DNA was stored at 20 °C. extend beyond the BAC clone cannot be accurately determined. Once validated, BAC arrays are much more effi- chip design cient than multiplex FISH analysis, but genetic informa- This focused oligonucleotide chip covers 179 clinically tion is constantly changing. Consequently, updates to a relevant regions of genomic imbalance, including all BAC-based array require successive rounds of extensive subtelomeric and pericentromeric regions, and 95 regions responsible for well-defined microdeletion/microdupli- probe validation. cation syndromes, mental retardation, and autism (for a Oligonucleotide-based arrays offer advantages over summary of array coverage, see Supplemental Data 1 in BAC-based arrays, and many platforms are available. the Data Supplement that accompanies the online version Oligonucleotide arrays designed for genotyping single- of this article at http://www.clinchem.org/content/ nucleotide polymorphisms may not provide uniform cov- vol53/issue12). A total of 10 207 region-specific features erage at all sites of genomic imbalance (9, 10). Custom and 603 quality-control and negative-control features se- oligonucleotide arrays that are based on libraries of vali- lected from Agilent’s eArray library (11) are randomly dated synthetic probes can interrogate clinically relevant located on the array with an average spatial resolution of genomic regions without the need for large-insert clone Ͻ35 kb within the targeted regions. Each subtelomeric libraries. We describe an array based on Agilent’s eArray region has a minimum coverage of 5 Mb. A subset of 660 library, a large collection of 60-mer oligonucleotides spe- features is duplicated on each block as a quality-control cifically selected for robust copy-number analysis (11). measure. Arrays were manufactured with Agilent’s Sure- This targeted oligonucleotide-based array provides a flex- Print Inkjet technology. In designing the targeted oligo- ible and adaptable method for CGH to detect genomic nucleotide-based array, we consulted the Database of copy-number imbalance in the clinical diagnostic Genomic Variants (http://projects.tcag.ca/variation/) to laboratory. avoid CNVs with no apparent clinical relevance.

Materials and Methods cgh validation samples Oligonucleotide-array CGH was performed according to DNA was obtained from the material remaining from 105 the manufacturer’s Oligonucleotide Array–Based CGH samples after previous clinical assays had been completed for Genomic DNA Analysis protocol (version 3; Agilent for patients who originally had been referred for genetic Technologies; see Supplemental Data 2 in the online Data testing with BAC-based array CGH, FISH, karyotyping, Supplement for a summary of the protocol). or MLPA in the DNA Diagnostic Laboratory at Children’s Dye-swap verification was performed on all samples Hospital Boston and the Medical Genetics Laboratories at with positive findings. For other confirmation assays, we Baylor College of Medicine. Genomic imbalance was carried out BAC-array CGH and FISH confirmation as previously identified in 51 (49%) of the 105 samples. described previously (12). MLPA confirmation was per- Samples with positive results from prior testing were formed as described previously (13). The MLPA oligonu- assigned to a “validation set” and subjected to oligonucleotides for the CTNS8 (cystinosis, nephropathic) gene cleotide-array CGH analysis in these 2 laboratories with are as follows: exon 2, GTTTTCACACTGGGCGAAGG the array platform described below. Laboratory personnel were blinded to prior testing results. The Children’s Hospital Boston Institutional Review Board approved this 8 Human genes: CTNS, cystinosis, nephropathic; NPHP1, nephronophthi- project. sis 1 (juvenile); CARKL, carbohydrate kinase-like. Clinical Chemistry 53, No. 12, 2007 2053

Ͼ GAGGACT and CCTGAGCTCTGCCTCTTCCAGTAA of the target features had a mean SNR 4. Mean log2 CATTG; exon 6, CCGAGGATACGCTTTCTTGTGATCC ratios were distributed symmetrically around the zero and GCAGCAGCGCCATTAGCATCATAAACC; exon value. Only a small fraction of features (52 of 10 025, 12, CAACCAAGTTTGGACTCGGGGT and CTTCTC 0.52%) exhibited mean values Ͼ0.1 or ϽϪ0.1. These CATCGTCTT CGACGTCGTC. features were excluded from the dataset before further

analysis. The mean SD of the log2 ratio of all noncontrol data analysis features was 0.062 (median SD, 0.055). Ͼ Scanned images were quantified with Feature Extraction We demonstrated a log2 ratio SD of 0.1 for 714 software (version 9.0; Agilent Technologies). We used the features (7.12%); we categorized these features as non- signal-to-noise ratio (SNR) and the normalized log2 ratio ideal targets and excluded them from further analysis. (test:reference) with 40 nonpathologic individual DNA More than 90% of the features passed the feature level samples to evaluate the quality and variability of each Ͻ filter criteria: an absolute mean log2 ratio 0.1, a log2 ratio feature/target. The SNR was calculated by dividing the SD Ͻ0.1, and a mean SNR Ͼ4. After excluding nonideal mean signal intensity of each feature by the mean back- features, the dataset quality improved dramatically. For Ͻ ground signal intensity. Features with a mean SNR 4or example, the SD of the log ratio dropped from Ͼ0.06 to Ͼ 2 an SD of the log2 ratio 0.1 were considered to have poor Ͻ 0.03. Features with a large log2 SD largely overlapped signal quality and high variability and were filtered out those with a low SNR, further validating this filtering before further analysis. These thresholds were chosen approach. Because the nonideal targets are approximately empirically and are similar to those used in similar evenly distributed across the target regions, the overall studies (1, 14). resolution of the chip is not appreciably affected. We visualized the filtered data further with CGH Several key variables were used to evaluate chip qual- Analytics software (version 3.4; Agilent Technologies) ity and to describe the quality of the dataset as a whole. and evaluated the quality of each test with the quality- These variables included probe-to-probe log ratio noise control metrics generated with CGH Analytics software. 2 (DLRSpread), the median signal intensity of both chan- Copy-number aberration was indicated with the Aberra- nels, background noise for both channels, and SNR. The tion Detection Method 2 algorithm for the data that following cutoffs were used to pass our quality-control passed quality-control testing. The Aberration Detection testing: DLRSpread Յ0.25, median signal intensity Ն50, Method 2 algorithm finds intervals of varying size with a background noise Յ10, and SNR Ն15. None of the sam- consistent, appreciably low, or high log ratio. An aberra- 2 ples failed quality-control testing because of poor chip tion filter was set to indicate regions with at least 3 targets quality or problems with hybridization. Two samples showing the same direction in copy-number change. The failed testing because of DNA impurities; both samples mean log ratio of each region of potential imbalance was 2 passed quality-control tests after we repurified the DNA. calculated and compared with the SD for the whole dataset. A copy-number gain was called if the mean log2 ratio was greater than twice the SD of the whole dataset, chip validation with blinded samples and a loss was called if the mean was less than Ϫ2 SDs. We next blindly tested 65 samples for further chip vali- These thresholds were chosen empirically and are similar dation. Genomic imbalance had previously been detected to those used in other such studies (15, 16). Cutoff values in 51 of the 65 samples by BAC-array CGH, FISH, for genomic imbalance can be adjusted and set accord- karyotyping, or MLPA, or by some combination of these ingly with the threshold function of CGH Analytics analyses, and these 51 samples served as positive controls software, especially when a potential mosaic scenario is for validation. The remaining 14 samples had previously encountered. Variants not known to be pathogenic were been tested by targeted BAC-array CGH with nonpatho- compared with the Database of Genomic Variants (http:// logic results and thus served as negative controls for projects.tcag.ca/variation/) to facilitate interpretation. validation. All samples were traceable to the technologist who performed the hybridizations. Results Of the 51 validation samples with a previously de- evaluation of target loci and overall chip tected genomic imbalance (see Supplemental Data 3 in the performance online Data Supplement), the samples from 2 cases dem- Forty sex-matched samples from healthy individuals onstrated aneuploidy for an entire chromosome, 3 cases were analyzed on the array, including 2 self-self hybrid- involved unbalanced chromosomal rearrangements, 13 izations, to evaluate each feature on the array. Signal cases had subtelomeric deletions/duplications, 17 cases quality, log2 ratio variability, mean SNR, and SD of the had interstitial deletions/duplications, and 15 cases were log2 ratio were calculated for each noncontrol feature. The associated with known segmental aneuploidy regions, mean and median signal intensities of all the noncontrol including Angelman/Prader–Willi syndrome (4 cases), features were 251 and 178, respectively. The mean and atypical Angelman syndrome on 22q13.3 (1 case), an median values of the mean SNRs from all noncontrol autism phenotype associated with duplication of 15q11- features were 9.05 and 6.39, respectively. More than 91% q13 (1 case), a velocardiofacial/DiGeorge syndrome re- 2054 Shen et al.: Oligonucleotide-Array CGH Chip for Clinical Diagnosis

gion (5 cases with deletions and 2 with duplication), and Oligonucleotide-array CGH was able to detect cryptic Williams–Beuren syndrome (2 cases). rearrangements, submicroscopic alterations, and even Across all samples, results from oligonucleotide-array single-gene deletions. The smallest imbalance event de- CGH were consistent with the results obtained with the 4 tected in this study was a heterozygous genomic deletion prior methods, but oligonucleotide-based array CGH proof 3 consecutive probes covering a minimal 23-kb interval. vided the most precise breakpoint boundaries. Fig. 1A Fig. 2A shows the targeted array CGH data, and deletion shows a genomic-imbalance event (2q37.3 deletion) iden- of the 3 targets was confirmed by dye-swap hybridization tified by oligonucleotide-based array CGH, and Fig. 1B (green in the forward hybridization and red in the reverse shows the FISH confirmation of the 2q37.3 deletion. hybridization). In this case, we repeated the CGH analysis Oligonucleotide-array CGH detected no appreciable with Agilent’s 244K whole-genome oligonucleotide array imbalance events in any of the 14 negative controls, with and confirmed the deletion, which includes the entire the exception of several reported CNVs. The dye-swap CARKL (carbohydrate kinase-like) gene and part of CTNS scheme essentially eliminated false-positive results. (Fig. 2B). The partial deletion of CTNS was independently To further evaluate the confidence of each imbalance confirmed by MLPA analysis. Fig. 2C shows a 1-copy call by oligonucleotide-array CGH analysis, we calculated deletion for CTNS exons 2 and 6 and a typical dosage for CTNS exon 12. To further characterize the deletion, we the mean log2 ratios for each detected imbalance region and compared them with the SD for the whole dataset. amplified the deletion junction by the PCR and confirmed by sequencing (Fig. 2D) that the deletion detected by The value of the mean log2 ratio/SD indicates the sepa- rability of each imbalance event from the background array CGH is the common “European” deletion associ- noise of the whole dataset. For the majority of deletion ated with cystinosis (17). This case further demonstrated events detected, the value was less than Ϫ3.2, whereas the the excellent resolution and sensitivity of the custom value was Ͼ2.6 for the majority of gain events detected. oligonucleotide-array CGH method. genomic imbalance detected in clinical genomic imbalance probably causing a disorder samples Table 1B lists 5 samples with interstitial deletions/dupli- We used focused oligonucleotide-array CGH to test 211 cations that may be clinically relevant. The list includes 3 clinical samples that had been ascertained to have come interesting cases: a 3.3 Mb duplication at 17p11.2, which is from individuals with developmental delay, unexplained similar to that of a recently reported 17p11.2-duplication mental retardation, dysmorphic features, or multiple con- syndrome (18), and 2 cases of a 546 kb de novo deletion at genital anomalies. In this cohort, the detection rate for 16p11.2, which is within a region of frequently observed genomic imbalance was approximately 11.9% (25 of 211 cytogenetic polymorphism but is not observed in the samples). All abnormal findings were first verified with a CNV database. Neither individual with the de novo dye-swap array CGH and then independently confirmed 546-kb 16p11.2 deletion had specific dysmorphic features, by either FISH or MLPA. All the genomic-imbalance but the absence of this deletion in the parents suggests events, including CNVs with unknown significance, were that the deletion is responsible for the phenotype of divided into 3 categories, as is described below. The developmental delay. imbalance events associated with known disorders or likely to cause disease are listed in Table 1. genomic CNVs with unknown significance Nine patients had unreported CNVs with relatively small genomic imbalance associated with genomic deletions of between 50 and 200 kb (data not well-defined disorders shown). The clinical significance of these imbalance Table 1A lists 12 samples with 10 genomic imbalance events is unclear. Although this custom oligonucleotide events associated with known genetic disorders. In this array was designed to avoid CNVs, many new CNV loci group, 2 of the samples revealed a complex pattern have been reported since the design of the array involving both gain and loss on 2 different chromosomes (http://projects.tcag.ca/variation/). (case 1) or on the same chromosome (case 2), 2 samples had a well-defined microdeletion syndrome (cases 3 and Discussion 4), 1 sample had a subtelomeric deletion (case 5), and 2 Array CGH is a valuable clinical diagnostic assay for samples had whole-chromosome aneuploidy (cases 6 and patients with mental retardation and other genetic condi- 7). One dominantly inherited disorder could be diagnosed tions. Although high-resolution whole-genome oligonu- by the detection of haploinsufficiency for the relevant cleotide microarrays are commercially available for re- gene (cases 8 and 9), and 3 cases involved carriers of a search, targeted array CGH offers several advantages in a recessive allele: deletion at the NPHP1 [nephronophthisis clinical diagnostic laboratory (8). We chose genomic re- 1 (juvenile); cases 10 and 11] and CTNS loci (case 12). The gions with well-documented clinical relevance, analogous last 3 cases featured deletions of Ͻ100 kb, each of which to those of the currently accepted BAC-based arrays covered defined disease genes. designed by the leading array CGH laboratories (19, 20). Clinical Chemistry 53, No. 12, 2007 2055

Fig. 1. Genomic-imbalance event (2q37.3 deletion) identified by oligonucleotide-based array CGH and FISH confirmation of the deletion.

(A), chromosome view of array CGH results showing a terminal deletion on 2q37.3. Arrowhead points to the targets with a downward-shifted log2 ratio. (B), 2q37.3 deletion confirmed by metaphase FISH. The small arrowhead indicates the signal of the region-specific BAC clone RP11-367H1, and the larger arrowheads indicate the signals of the chromosome 2 centromere–specific BAC clone. 2056 Shen et al.: Oligonucleotide-Array CGH Chip for Clinical Diagnosis

Table 1. Genomic imbalance identified in clinical samples by oligonucleotide-array CGH and confirmed with alternative methods. Genomic imbalance detected by focused oligonucleotide- Cytogenetics (FISH/karyotype), MLPA, array CGH whole-genome CGH/targeted PCR Confirmation (A) Associated with disorders 13qter gain, 20.7 Mb; 18qter loss, 5.6 Mb 46,XX, add(18)(q2?1.3) ish Consistent (partial trisomy 13q and partial monosomy 18q) der(18)t(13;18)(D13S327ϩ, 18qtel11–) 18pter-p11.21 loss, 13.4 Mb; 18p11.21 gain, 46,XX ish 18pter(D18S552 ϫ 1), 18p11.21 (RP11- Consistent 1.3 Mb (partial monosomy 18p and partial 720L3 ϫ 3) trisomy 18p) 17p11.2 loss, 3.6 Mb (Smith–Magenis syndrome) 46,XX, ish del(17)(p11.2 p11.2) Consistent 1p36.21 loss, 1.8 Mb (1p36 deletion syndrome) 46,XY, ish del(1)(p36.2) Consistent 4q35.2 loss, 1.1 Mb (autism spectrum disorder) 46,XY, ish del(4)(qter–) Consistent X gain (aneusomy X) 47,XXY Consistent Y gain (aneusomy Y) 47,XYY Consistent Yp11.2 loss, 2.7 Mb: 2 cases ish Yp11.2 (RP11-115H13 ϫ 0) Consistent 2q13 loss, 100 kb (NPHP1 deletion): 2 cases Confirmed by whole-genome array CGH Consistent 17p13 loss, 23 kb (CARKL and CTNS deletion; Confirmed by whole-genome array CGH, MLPA, Consistent familial) parental array CGH, and PCR flanking the deletion (B) Likely clinically relevant 17p11.2 gain, 3.3 Mb 46,XY, nuc ish 17p11.2 (RP11-363P3 ϫ 3) Consistent 16p11.2 loss, 546 kb (de novo): 2 cases 46,XY, ish del(16)(p11.2 p11.2) Consistent 15q13.3 gain, 1.5 Mb nuc ish 15q13.3 (RP11-303I13 ϫ 3) Consistent 5q22.1–q23.1 loss, 8.5 Mb Confirmed by whole-genome array CGH Consistent

Genomic imbalance identified on the targeted oligonucle- tive samples, we were able to identify smaller imbalance otide array can be verified with existing FISH or MLPA events (unreported CNVs; data not shown) that were not probes, whereas secondary methods are not readily avail- detected with other methods. able for a whole-genome array. Oligonucleotide platforms can quickly and easily ac- Uniformly manufactured arrays facilitate reproducible commodate changes in genomic coverage. Manufacturing results in the clinical laboratory. On BAC-based arrays, costs are not prohibitive and the list of available probe each BAC clone may include both unique and repetitive sequences extends the length of the genome. Thus, up- sequences. Each BAC must be propagated in culture, dates to oligonucleotide-based arrays can be accom- introducing possible variability in BAC inserts or DNA plished more quickly and with less postproduction vali- contamination between batches and potentially affecting dation than BAC-based arrays, which require new BAC the reproducibility and consistency of the manufactured clones to be individually validated and DNA to be pre- arrays. In contrast, 60-mer oligonucleotide probes are pared from each clone before chips can be manufactured. synthesized robotically in situ and have a fixed GC This focused oligonucleotide-based array CGH plat- content and melting temperature that facilitate uniform form detected all genomic imbalance events in the 65 hybridization. Although all targets in Agilent’s eArray validation samples, with 100% concordance with BAC- library were designed in silico with narrow melting based array CGH, FISH/karyotyping, or MLPA. Cover- temperature ranges, the actual performance of each target age of clinically relevant loci (see Supplemental Data 1 in needed to be evaluated in practice. Some features exhib- the online Data Supplement) is equivalent to other BAC- ited large variability, possibly due to properties of these based targeted array CGH platforms (19, 20). The en- unique sequences. Genomic CNVs unquestionably con- hanced sensitivity and specificity of oligonucleotide-array tribute to the variability in the performance of some CGH compared with other methods are attributable to probes. better resolution and the custom design, respectively. The sensitivity and specificity of the oligonucleotide- Despite the superior technical performance of oligonu- array platform were excellent because of the high SNRs cleotide-based array CGH, clinical interpretation can be and low SDs. We observed the SDs of all targets to be challenging. We encountered several scenarios: (a) known consistently Ͻ0.1—with the majority being Ͻ0.08—with genomic-imbalance events with well-documented clinical this oligonucleotide-based array. Given the nice separa- relevance (Table 1A) that were verified with dye-swap- bility of imbalance events above the baseline, we can ping and either FISH or MLPA (on smaller regions); (b) identify genomic imbalance events with a high level of known genomic-imbalance events with possible clinical confidence. In addition to the 100% concordance between relevance (Table 1B) that were verified by dye-swapping the oligonucleotide-array CGH results and the results and the 244K whole-genome array for further character- generated by other methods for both positive and nega- ization of the genomic-imbalance events; and (c) novel Clinical Chemistry 53, No. 12, 2007 2057

Fig. 2. Oligonucleotide-array CGH detection of a heterozygous genomic deletion of 3 consecutive probes covering a minimal 23-kb interval. (A), the deletion of 3 consecutive targets (underlined with rectangular bar) was confirmed by dye-swap array CGH. Note the symmetrical opposite ratios between forward-labeled (downward-shift) and reverse-labeled (upward-shift) array CGH. (B), 244K whole-genome oligonucleotide-array CGH confirmation of the deletion. The rectangular bar underlines the deleted region. Note the entire CARKL gene deletion and the partial CTNS gene deletion (forward labeling only). (C), MLPA confirmation of the deletion. The trace in lighter gray is a sample from a healthy control individual; the trace in darker gray is the patient sample. The dosage of CTNS exons 2 and 6 appears reduced by half (underlined with the rectangular bar), whereas the dosage of CTNS exon 12 is unchanged. (D), a segment of the sequencing trace around the deletion junction of the common European deletion. One base (C) overlaps. 2058 Shen et al.: Oligonucleotide-Array CGH Chip for Clinical Diagnosis

genomic imbalance with uncertain clinical relevance, In conclusion, the custom oligonucleotide array that we cases of which were verified by dye-swapping, 244K have described offers a sensitive and specific clinical assay whole-genome array analysis, and examination of paren- for detecting genomic imbalance. The platform can easily tal samples to determine whether the imbalance was a de accommodate new information to keep pace with re- novo event. search advances. This methodology offers a proof of Adding to the complexity of interpretation is that an principle that well-designed oligonucleotide-based arrays apparent familial variant can actually be pathogenic. For offer substantial technical advantages over currently example, a deletion variant that appears in a healthy available platforms and at a comparable cost. We antici- parent and an affected child could be acting as a recessive pate widespread adoption of this methodology in the allele in the parent. Likewise, a de novo region of genomic clinical diagnostic setting. imbalance could be a previously undescribed nonpatho- genic CNV or a case of nonpaternity. Dozens of CNVs are present in every human genome (21, 22). Grant/funding support: This work was supported in part by Although this custom oligonucleotide array has an National Institutes of Health (NIH) Grant NS24279 (to J.F.G.) average resolution of approximately 35 kb within the and National Institute of Neurological Disorders and Stroke targeted regions, the fact that array formats continue to Grant R37 NS35129 (to C.A.W.). Y.S. is the recipient of a evolve provides ample opportunities to increase resolu- Young Investigator Award from the Children’s Tumor Foun- tion through the selection of additional probes from dation. B.-L.W. received support from the NIH-CETT Pro- available oligonucleotide databases or the design of cus- gram and the MDA Foundation for the development of tom oligonucleotide arrays. The flexibility of the oligonu- cost-effective assays for genetic diagnostics. cleotide-array design allows individual laboratories to Financial disclosures: None declared. Acknowledgments: We thank the referring physicians and interrogate different genomic regions at the most biolog- genetic counselors at the Children’s Hospital Boston for their ically relevant level of resolution. For example, arrays can cooperation, Dr. Arthur L. Beaudet of Baylor College of be designed to detect single-exon deletions and duplica- Medicine for thoughtful discussion of the study, and Dr. tions in a gene of interest. Arthur R. Brothman and Emily Aston of the University of Resolution beyond that provided by this custom oligo- Utah for FISH confirmation of clinical cases. nucleotide array may not be necessary or desirable in all cases, however. A majority of the CNVs detected on Ͻ References high-resolution oligonucleotide platforms are 150 kb 1. Shaw-Smith C, Redon R, Rickman L, Rio M, Willatt L, Fiegler H, et (23, 24), and the detection of these CNVs complicates the al. Microarray based comparative genomic hybridisation (array- clinical interpretation. CNV databases are highly valuable CGH) detects submicroscopic chromosomal deletions and dupli- research tools, but they have limitations in the clinical cations in patients with learning disability/mental retardation and setting. CNVs in databases are often determined by a dysmorphic features. J Med Genet 2004;41:241–8. variety of methods that may not provide accurate break- 2. Schoumans J, Ruivenkamp C, Holmberg E, Kyllerman M, Anderlid BM, Nordenskjold M. Detection of chromosomal imbalances in points. These databases are often not annotated with children with idiopathic mental retardation by array based compar- clinical information and thus do not indicate genotype- ative genomic hybridisation (array-CGH). J Med Genet 2005;42: phenotype associations. 699–705. As the knowledge of genotype-phenotype associations 3. Shevell M, Ashwal S, Donley D, Flint J, Gingold M, Hirtz D, et al. in disorders of genomic imbalance accumulates, however, Practice parameter: evaluation of the child with global develop- precise and accurate determination of the genomic se- mental delay: report of the Quality Standards Subcommittee of the American Academy of Neurology and The Practice Committee of quences involved is critically important in the clinical the Child Neurology Society. Neurology 2003;60:367–80. diagnostic setting. FISH and BAC clones can detect im- 4. de Vries BB, Pfundt R, Leisink M, Koolen DA, Vissers LE, Janssen balance equal to or larger than the size of clone insert, IM, et al. Diagnostic genome profiling in mental retardation. Am J typically approximately 150 kb; thus, smaller deletions or Hum Genet 2005;77:606–16. duplications can be missed on a BAC-based array. Even 5. Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D, et al. genome-wide tiling BAC arrays do not have a resolution High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet of less than approximately 46 kb (25, 26). 1998;20:207–11. Oligonucleotide-based arrays provide an unprece- 6. Bejjani BA, Shaffer LG. Application of array-based comparative dented degree of resolution for determining both the size genomic hybridization to clinical diagnostics. J Mol Diagn 2006; and breakpoints of regions of genomic imbalance. Our 8:528–33. current design was able to precisely define a heterozy- 7. Bejjani BA, Theisen AP, Ballif BC, Shaffer LG. Array-based compar- gous deletion of 23 kb, a level of resolution not possible on ative genomic hybridization in clinical diagnosis. Expert Rev Mol Diagn 2005;5:421–9. a BAC-based array. This improvement in resolution could 8. Bejjani BA, Saleki R, Ballif BC, Rorem EA, Sundin K, Theisen A, et be important in the clinical interpretation of cases in al. Use of targeted array-based CGH for the clinical diagnosis of which the inclusion or exclusion of certain genes may chromosomal imbalance: is less more? Am J Med Genet A predict clinical features of the condition. 2005;134:259–67. Clinical Chemistry 53, No. 12, 2007 2059

9. Friedman JM, Baross A, Delaney AD, Ally A, Arbour L, Armstrong L, cystinosis: development of a PCR-based detection assay. Am J et al. Oligonucleotide microarray analysis of genomic imbalance in Hum Genet 1999;65:353–9. children with mental retardation. Am J Hum Genet 2006;79:500– 18. Potocki L, Bi W, Treadwell-Deering D, Carvalho CM, Eifert A, 13. Friedman EM, et al. Characterization of Potocki-Lupski syndrome 10. Ylstra B, van den Ijssel P, Carvalho B, Brakenhoff RH, Meijer GA. (dup(17)(p11.2p11.2)) and delineation of a dosage-sensitive crit- BAC to the future! or oligonucleotides: a perspective for micro ical interval that can convey an autism phenotype. Am J Hum array comparative genomic hybridization (array CGH). Nucleic Genet 2007;80:633–49. Acids Res 2006;34:445–50. 19. Baylor College of Medicine. Chromosomal Microarray Analysis. 11. Agilent Technologies. eArray 4.5. http://earray.chem.agilent. http://www.bcm.edu/cma (accessed August 2006). com/earray/login.do (accessed August 2006). 20. Signature Genomic Laboratories. Home page. http://www.signa 12. Cheung SW, Shaw CA, Yu W, Li J, Ou Z, Patel A, et al. Development turegenomics.com (accessed August 2006). and validation of a CGH microarray for clinical cytogenetic diagno- 21. Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, Qi Y, et sis. Genet Med 2005;7:422–32. al. Detection of large-scale variation in the human genome. Nat 13. Shao H, Lip V, Wu BL. Effectiveness of multiplex ligation-depen- Genet 2004;36:949–51. dent probe amplification assay used for detecting deletion of 22. Eichler EE. Widening the spectrum of human genetic variation. Nat Prader-Willi syndrome. Beijing Da Xue Xue Bao 2005;37:64–7. Genet 2006;38:9–11. 14. Wong KK, deLeeuw RJ, Dosanjh NS, Kimm LR, Cheng Z, Horsman DE, et al. A comprehensive analysis of common copy-number 23. Redon R, Ishikawa S, Fitch KR, Feuk L, Perry GH, Andrews TD, et variations in the human genome. Am J Hum Genet 2007;80:91– al. Global variation in copy number in the human genome. Nature 104. 2006;444:444–54. 15. Locke DP, Sharp AJ, McCarroll SA, McGrath SD, Newman TL, 24. Komura D, Shen F, Ishikawa S, Fitch KR, Chen W, Zhang J, et al. Cheng Z, et al. Linkage disequilibrium and heritability of copy- Genome-wide detection of human copy number variations using number polymorphisms within duplicated regions of the human high-density DNA oligonucleotide arrays. Genome Res 2006;16: genome. Am J Hum Genet 2006;79:275–90. 1575–84. 16. Sharp AJ, Hansen S, Selzer RR, Cheng Z, Regan R, Hurst JA, et al. 25. Osoegawa K, Mammoser AG, Wu C, Frengen E, Zeng C, Catanese Discovery of previously unidentified genomic disorders from the JJ, et al. A bacterial artificial chromosome library for sequencing duplication architecture of the human genome. Nat Genet 2006; the complete human genome. Genome Res 2001;11:483–96. 38:1038–42. 26. Krzywinski M, Bosdet I, Smailus D, Chiu R, Mathewson C, Wye N, 17. Forestier L, Jean G, Attard M, Cherqui S, Lewis C, van’t Hoff W, et et al. A set of BAC clones spanning the human genome. Nucleic al. Molecular characterization of CTNS deletions in nephropathic Acids Res 2004;32:3651–60. Clinical Chemistry 53:12 2060–2069 (2007) Molecular Diagnostics and Genetics

Global Sequencing Approach for Characterizing the Molecular Background of Hereditary Iron Disorders

Se´verine Cunat,1 Muriel Giansily-Blaizot,1 Michael Bismuth,2 Franc¸ois Blanc,3 Olivier Dereure,4 Dominique Larrey,2 Alain Le Quellec,5 Philippe Pouderoux,6 Christian Rose,7 Isabelle Raingeard,8 Eric Renard,8 Jean-Franc¸ois Schved,1 Patricia Aguilar-Martinez,1* and the CHU Montpellier AOI 2004 Working Group

Background: New genetic forms of hereditary hemo- slightly increased TS, with or without iron overload chromatosis (HH) or hereditary hyperferritinemia (HF) (groups 2A and 2B, respectively). have been identified over the last few years, and abnor- Results: With this strategy we identified single-gene malities of various genes may interact in a single pa- and multigene abnormalities, including 6 previously tient. This study aimed to develop a rapid automated undescribed abnormalities in HFE (c.794dupA), HFE2 method for sequencing the main genes involved. (c.–89–4dupT), and SLC40A1 (c.262A>G, c.533G>A, Methods: We used a standard 96-well microplate with a c.1468G>A, and c.–59_–45del). single PCR condition in an adaptation of the SCAIP Conclusion: This method is a simple approach for (single-condition amplification with internal primer) investigating hereditary iron overload or HF and allows method to sequence the HFE (hemochromatosis), HAMP rapid evaluation of patients. (hepcidin antimicrobial peptide), HFE2/HJV [hemochro- © 2007 American Association for Clinical Chemistry matosis type 2 (juvenile)], SLC40A1 (ferroportin), and TFR2 (transferrin receptor 2) genes, and the 5؅ untrans- Hereditary hemochromatosis (HH)9 is the main cause of lated region of the FTL (ferritin, light polypeptide) gene. inherited iron overload. When present in a homozygous To further simplify the method, we adjusted PCR con- state, the Cys282Tyr mutation in the hemochromatosis ditions to avoid the use of an internal primer and (HFE)10 gene product is the most common genetic back- applied this single-condition amplification method to ground of HH (1). Ten years after the description of this 38 selected, unrelated patients. We tailored the genetic genotype, further investigations of the disease have re- investigation according to the clinical picture, with the vealed considerable heterogeneity. patients falling into 2 groups. Group 1 consisted of First, other HFE genotypes, including the compound patients with hyperferritinemia and high transferrin heterozygote encoding the Cys282Tyr and His63Asp mu- saturation (TS) (classic adult and juvenile HH forms, tations, have been associated with an HH phenotype, groups 1A and 1B, respectively), and group 2 consisted although individuals with such genotypes accumulate of patients with hyperferritinemia and low, typical, or less iron than Cys282Tyr homozygotes (1–3). Other rare private HFE mutations responsible for HH have also been reported. Second, Cys282Tyr homozygotes have been demonstrated to lack full penetrance (4), and some indi- 1 Laboratory of Haematology, 2 Department of Hepatology, 3 Department of Internal Medicine E, 4 Department of Dermatology, 5 Department of Internal Medicine A, CHU of Montpellier, Montpellier, France. 6 Department of Hepato-Gastroenterology, CHU Nîmes, Nîmes, France. 7 Department of Oncology and Haematology, Lille University Hospital, 9 Nonstandard abbreviations: HH, hereditary hemochromatosis; JH, Lille, France. juvenile hemochromatosis; HF, hereditary hyperferritinemia; HHCS, he- 8 Department of Endocrinology, CHU of Montpellier, Montpellier, France. reditary hyperferritinemia cataract syndrome; UTR, untranslated region; * Address correspondence to this author at: Laboratoire d’He´matologie, SCAIP, single-condition amplification with internal primer; TS, transferrin CHU de Montpellier, Hoˆpital Saint Eloi, 34295 Montpellier CEDEX 5, France. saturation; SCA, single-condition amplification. Fax 33-467-33-70-36; e-mail [email protected]. 10 Human genes: HFE, hemochromatosis; HFE2/HJV, hemochromatosis Received April 30, 2007; accepted September 14, 2007. type 2 (juvenile); HAMP, hepcidin antimicrobial peptide; TFR2, transferrin Previously published online at DOI: 10.1373/clinchem.2007.090605 receptor 2; FTL, ferritin, light polypeptide; SLC40A1, ferroportin.

2060 Clinical Chemistry 53, No. 12, 2007 2061

viduals homozygous for the Cys282Tyr mutant remain ration of Helsinki, 2002. Written informed consent was asymptomatic throughout life. Investigators have also obtained from all patients or their parents in accordance described several other genes that present a phenotype with French bioethics regulations. similar to that of HFE-related HH, indicating that hemochromatosis is a multigene disorder. Nowadays, HH is phenotypic assessment considered an autosomal-recessive inherited disorder Patients enrolled in this study had a phenotypic diagnosis with 2 clinical forms, adult and juvenile HH (5, 6), de- of HH or HF according to previously proposed classifica- fined according to the age of onset. Juvenile hemochro- tion criteria (5, 6). We screened for the usual HFE geno- matosis (JH) is due to a mutation in 1 of 2 genes: HFE2 type (encoding the Cys282Tyr and His63Asp mutations) [hemochromatosis type 2 (juvenile); also known as HJV before sequencing the other genes (except in cases of and encoding hemojuvelin] (7) and, more rarely, HAMP HHCS). To target the genes of interest, we divided (hepcidin antimicrobial peptide) (8). Adult HH is mainly patients into 2 groups according to their clinical profile due to mutation in HFE, and TFR2 (transferrin receptor 2) (5). Group 1 included patients with a classic HH pheno- mutations are rarely found (9). Associations between type, i.e., displaying at least a high transferrin saturation mutations of different genes leading to modified clinical (TS) associated with hyperferritinemia and increased iron presentations have been described for some patients. Such in the liver, as assessed by MRI or liver biopsy. Acquired findings demonstrate that hemochromatosis can be a causes of iron overload were excluded for each patient. multigenic or at least a digenic disease. Indeed, HAMP This group was divided into 2 subgroups. Group 1A (10), HFE2 (11),orTFR2 (12) mutations that have been consisted of patients with adult HH forms without a associated with HFE genotypes have been described to classic HFE genotype (i.e., neither Cys282Tyr homozy- aggravate the clinical presentation. gotes nor Cys282Tyr/His63Asp compound heterozy- On the other hand, clinical pictures characterized by gotes). Group 1B included patients with either a JH high ferritinemia have recently emerged [for review, see phenotype or patients with an HFE Cys282Tyr/ (5)]. Hereditary hyperferritinemia (HF) mostly has a Cys282Tyr or Cys282Tyr/His63Asp genotype and early dominant inheritance pattern and may or may not feature (age Ͻ25 years) or severe clinical expression of hemochro- iron overload. No iron overload is associated with hered- matosis. Group 2 consisted of individuals with hyperfer- itary hyperferritinemia cataract syndrome (HHCS) (13, 14), ritinemia and low, typical, or slightly increased TS for which is due to mutations in the 5Ј untranslated region whom other potential causes of hyperferritinemia, such as (UTR) of FTL (ferritin, light polypeptide). On the other inflammation, chronic anemia, cancers, or liver patholo- hand, alterations in SLC40A1 (ferroportin) cause iron gies, could be excluded. Group 2 patients were divided overload (15, 16); in its classic form, however, this disor- into group 2A (hyperferritinemia with iron overload der, termed ferroportin disease, has clinical and biological evidenced by liver biopsy or MRI), which essentially presentations that are considerably different from HH targeted the SLC40A1 gene, and group 2B (hyperferritine- (17). A 2nd clinical class of ferroportin-related disorders mia without iron overload and possibly with cataracts), that produces a phenotype very similar to classic HH has which targeted suspected HHCS cases. been described for a group of different mutations (18). These new findings increase the diagnostic and prog- molecular studies nostic interest in the precise identification of the molecu- DNA-sequencing strategy. We studied 3 sets of genes selected lar disorder in patients with HH not related to a classic according the patient’s initial phenotypic presentation. We HFE genotype or with suspected HF, a technically chal- first targeted HFE and TFR2 genes in group 1A and prefer- lenging task given that the total number of genes to test entially screened HAMP and HFE2 genes in group 1B. In has now increased and corresponds to a total of approx- group 2, we first tested patients with unexplained hyperfer- imately 90 amplicons (including both the sense and anti- ritinemia without iron overload (group 2B) for mutations in sense strands). To address this obstacle, we set up in our the 5Ј UTR of the FTL gene. When iron overload was present laboratory a sequencing procedure based on the SCAIP (group 2A), we sequenced the SLC40A1 gene. In the studied method (single-condition amplification with internal patients, we had no suspicion of ceruloplasmin deficiency or primer) described by Flanigan et al. (19) in 2003. We used other clinical forms of hereditary iron overload with neuro- a simplified adaptation of this technique to facilitate the logic symptoms; thus, genes such as those encoding cerulo- straightforward sequencing of most of the genes impli- plasmin or frataxin were excluded from our experiments. cated in human diseases related to iron metabolism. Patients with an ambiguous clinical presentation were screened for all 3 sets of genes. Materials and Methods patients electronic database information Consent and ethical issues. The study was approved by the GenBank sequence accession numbers for the SLC40A1, ethics committee of the French Data Protection Authority FTL, HFE, HFE2, HAMP, and TFR2 genes are (Commission Nationale de l’Informatique et des Liberte´s) NC_000002.10, NC_000019.8, NC_000006.10, NC_000001.9, in accordance with the World Medical Association Decla- NC_000019.8, and NC_000007.12, respectively. Mutations 2062 Cunat et al.: Rapid DNA Sequencing of Iron Genes

were named in accordance with the standard interna- our group or others had previously identified to possess tional nomenclature guidelines recommended by the Hu- mutations. man Genome Variation Society (HGVS at http://www. hgvs.org/mutnomen/recs.html). analyses of individual mutations and of parents or siblings single-condition amplification method We confirmed an identified mutation using a novel PCR We extracted DNA from patient’s peripheral blood leu- product either by restriction enzyme analysis when avail- kocytes with standard procedures. Direct sequence anal- able or by another sequencing experiment. We carried out ysis of all the targeted iron metabolism-related genes was a segregation analysis of family members of patients with based on the SCAIP method (19), which allows simulta- an identified mutation. neous amplification, purification, and sequence analysis of multiple DNA regions on a standardized microplate Results with a single PCR condition. We used a simplified version the sca sequencing method adapted to iron of this method, the single-condition amplification (SCA) metabolism-related genes method, by adjusting PCR conditions according to Roux Successive adaptations, including primer choice, step- et al. (20), thereby avoiding the need for an internal down PCR conditions, and purification conditions, were primer. We chose amplicons to cover the 39 exons, exon/ necessary to permit easy screening of the 6 targeted genes intron boundaries, and most of the 5Ј and 3Ј UTRs of HFE, related to iron metabolism. We optimized the conditions TFR2, HFE2, HAMP, and SLC40A1, as well as the 5Ј UTR to apply the method to the 2 main phenotypic profiles: of FTL. This set constituted 44 amplicons spanning 16.4 kb HH/group 1 (HFE, TFR2, HFE2, and HAMP)orHF/ over a total of 64.8 kb of genomic DNA sequence. Primer group 2 (SLC40A1 and FTL). The TFR2 gene had to be sets were adapted from those described in previously processed separately because the annealing temperatures for TFR2 were substantially higher than for the other HH published articles or were designed for this study (see genes (see Table 2 in the online Data Supplement). The Table 1 in the Data Supplement that accompanies the sample requirements for a maximum of 44 amplicons per online version of this article at http://www.clinchem. patient are relatively low, approximately 44 ␮g DNA. One org/content/vol53/issue12). Each amplicon was de- cycle-sequencing plate can be used for a complete study signed for an optimal size range of 200–500 bp to main- of the different genes for a single patient or to screen a tain uniform PCR conditions and tested for compatibility given gene for different patients. The implementation of under temperature “stepdown” conditions (21), an adap- the SCA technique in our laboratory has allowed us to tation of the “touchdown” PCR method (22). The appro- greatly shorten the time required to complete a diagnosis priate amplification gradient used for each gene is given of iron-related disorders. Generally, only 1 week is nec- in Table 2 in the online Data Supplement. essary to complete the sequencing steps for a 96-well plate, including analysis of the sequencing results. If a dna sequencing ␮ gene defect is identified, further investigation is needed to Sequencing reactions were carried out in a 10- L volume confirm the mutation. containing 10–50 fmol PCR template, 1.5 ␮L5ϫ BigDye® ␮ Terminator v1.1 Sequencing Buffer (Applera), 5 pmol/ L application to the identification of novel ␮ ® sequencing primer, 1.5 L ABI PRISM BigDye Termina- mutations in genes related to iron metabolism ␮ tor v1.1 reaction mixture (Applera), and 4 L sterile We studied 38 patients with this method, including 16 water. Sequencing products were then purified on Mon- ௢ patients with a suspected diagnosis of HH and 22 patients tage -SEQ96 microplates (Millipore) according to the with suspected HF. The patients with potentially caus- ␮ ௢ manufacturer’s instructions. We added 15 L of Hi-Di ative mutations are described in Table 1. In addition, we formamide (Applera) to each well as recommended (22) identified several previously known or undescribed poly- and electrophoresed the samples on an ABI PRISM 310 morphisms during the course of the study (see Table 3 in DNA analyzer (Applera). We then analyzed sequence- the online Data Supplement) and found no mutation in trace files with Sequencing Analysis Software (version the targeted genes for 29 of the patients. ® 5.1.1; Applera) and used SeqScape software (version In group 1, 1 patient had a new HFE mutation, 1 2.1.1; Applera) to compare the data with the GenBank patient had a novel HFE2 mutation, and 2 patients had a reference sequence. new SLC40A1 mutation. The new HFE mutant was a frameshift mutation, c.794dupA, in HFE exon 4 of patient control of dna sequencing HG3572 and was in trans of the Cys282Tyr mutant (see DNA from an anonymous control individual with typical Fig. 1 in the online Data Supplement). This patient was erythrocytes and nonpathologic values for iron variables classified into group 1A (classic HH). The HFE2 mutation was sequenced with the SCA method in parallel with a was found in patient HG2280, a 23-year-old woman classic sequencing method for all of the studied genes referred for discrepancies between her HFE genotype and (23). For internal positive controls, we used samples that the clinical phenotype. Although she was homozygous for Clinical Chemistry 53, No. 12, 2007 2063

the Cys282Tyr mutation and had a classic adult form of ␮g/L) associated with high TS (83%). She experienced HH, as did her mother, the patient had abnormally high chronic fatigue and a cataract and was heterozygous for values of iron-related variables for her age (group 1B). A the His63Asp substitution. An MRI evaluation demon- heterozygous T insertion 5Ј to exon 2 (c.–89–4dupT) was strated a heavy hepatic iron overload (350 ␮mol/g). The identified in the HFE2 gene inherited from the father, who FTL gene was wild type. This patient’s phenotype placed was a Cys282Tyr/His63Asp compound heterozygote and her in group 1A. The apparent phenotypic discrepancy affected with an adult form of HH. This insertion was not between these 2 patients with the same mutation is present in the mother. The HFE, HAMP, and TFR2 genes discussed below. were wild type. A final assignment of causative status for Finally, patient HG4421 had a personal and family this mutation will require analysis of HFE2 transcripts. history of hyperferritinemia. She had a typical TS and no Patient HG3943, age 73 years, was referred for suspected common HFE mutation and was classified in group 2A. A classic HH because of a high ferritin concentration (1500 new mutation in the SLC40A1 gene (c.533GϾA) was in ␮ g/L) and high TS (80%) (group 1A). The HFE genotype heterozygous condition (Table 1). was simply a His63Asp heterozygote. An MRI examination of the liver revealed iron overload with a hepatic iron Discussion ␮ concentration of 240 mol/g. The patient also had liver Clinical genetic testing in iron-overload disorders rou- enzyme anomalies. The genes involved in HH, namely tinely focuses on the detection of the Cys282Tyr and HFE, TFR2, HFE2, and HAMP, were wild type. An unde- His63Asp mutations encoded by the HFE gene. Recently, scribed short deletion, c.–59_–45del, was identified in the investigators have recognized the need to expand muta- Ј 5 UTR of the SLC40A1 gene. The 2nd patient, HG3135, is tional investigation to other genes associated with iron described below. overload. Sequencing analysis via the classic methods of Group 2 included 2 unrelated 58-year-old female pa- sequencing individual samples is time-consuming be- tients who were referred for isolated hyperferritinemia cause of the number and, in some cases, the size of the without iron overload and with cataracts (Table 1, group iron metabolism-related genes (TFR2 has 18 exons) and 2B). Both patients had a previously described mutation in because such genes have to be studied with a variety of the 5Ј UTR of the FTL gene. Three unrelated patients had annealing and PCR conditions. Consequently, such strat- novel heterozygous mutations in the SLC40A1 gene (Ta- egies are considered labor intensive and expensive. A ble 1, group 2A). limited number of other global approaches, mainly in- Patient HG3181, a 38-year-old man, was referred for volving denaturing HPLC, for the screening of mutations high hyperferritinemia with slightly increased TS and a in genes involved in iron metabolism have been proposed family history of iron overload. He had a wild-type HFE in the literature (10, 25). Other DNA-screening methods, gene, and the genes involved in HH were also wild type. such as microelectronic DNA chips (26), have been ap- Analysis of the SLC40A1 gene revealed a missense muta- plied to the study of single iron-related genes. These tion (c.262AϾG) in exon 3 leading to an Arg88Gly substitution. This modification was reported simultaneously by methods, although useful, require expertise to set up the us and another group (Aguilar-Martinez et al., unpub- analysis for each individual gene. They also may have lished data; Jouanolle et al., personal communication, limitations in sensitivity, and some molecular defects may January 30, 2004). It is noteworthy that the patient had be missed. Another drawback is that these approaches are become anemic under a phlebotomy program, which screening methods, and a subsequent sequencing step is therefore had to be stopped. A liver biopsy performed at needed to precisely identify the sequence variation and to another center, which was initially unavailable to us, distinguish known and unknown polymorphisms from recently confirmed the iron overload in mixed parenchy- deleterious mutations. We have described an efficient and mal and Kupffer cells typically seen in “ferroportin dis- simple sequencing method (SCA) for the molecular diag- ease”. The patient’s father, who had the same mutation, nosis of HH or HF that we have adapted from the SCAIP recently died of liver cancer. technique of Flanigan et al. (19). This method, which is Two apparently unrelated patients (HG3142 and based on a SCA followed by gene sequencing, permits the HG3134) who were referred independently shared the amplification of large numbers of amplicons at a deter- same previously undescribed c.1468GϾA nucleotide sub- mined set of PCR temperatures. We used stepdown PCR stitution in SLC40A1. This mutation produces a Gly490Ser conditions, an approach derived from touchdown PCR substitution at the same position of another substitution, (22) that conveniently bypasses spurious amplification Gly490Asp, which has previously been described to lead (21, 27). This strategy makes primer choice easier, because to ferroportin disease (24). Patient HG3142, who at 24 the melting temperature of each primer can be situated in years at the time of diagnosis was the youngest of both, an area determined by the stepdown conditions. In our had isolated hyperferritinemia with typical TS and no case, the melting temperature was chosen to permit the additional biochemical abnormality, including hepatic simultaneous amplification of iron metabolism–related enzymes (group 2A). The 2nd patient (HG3134) was a genes that we clustered according to 3 specified clinical 71-year-old woman who had high hyperferritinemia (2710 profiles. 2064

Table 1. Main clinical, biochemical, histological, and genotypic data of the patients. Group 1 (hyperferritinemia with high TS and parenchymal iron overload)

Group 1A (classic adult form of HH) Group 1B (juvenile form of HH)

Patient no. HG3572 HG3134 HG3943 HG2280 HG3081 (positive control) Sex Male Female Male Female Female Age at molecular 47 71 73 28 8 Genes Iron of Sequencing DNA Rapid al.: et Cunat diagnosis, years Hepatopathy No NA NA NA No Hypogonadotropic No NA NA NA No hypogonadism Cardiomyopathy No NA NA NA No Other clinical signs No Fatigue, cataract No No No Serum iron, ␮mol/L NA 6.7 34 NA 46.4 Serum ferritin, ␮g/L 1007 2710 1400 780 199 TS, % 69 83 80 88 90 HIC,a ␮mol/g 140 350 240 135 260 HIC/age, ␮mol/g/years 3.0 4.9 3.3 NA 32.5 HFE p.Cys282Tyr CY CC CC YY CC p.His63Asp HH HD HD HH DD Other HFE mutation c.794dupA No No No No (p.Trp267LeufsX80) HFE polymorphism No ND c.͓340ϩ4TϾC͔ϩ͓ϭ͔b No No TFR2 c.͓1851CϾT͔ϩ͓ϭ͔ (p.Ala617) ND c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ c.͓2085GϾC͔ϩ͓ϭ͔; (p.Ser695)c HAMP c.͓ϭ͔ϩ͓ϭ͔ ND c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ HFE2 c.͓ϭ͔ϩ͓ϭ͔ ND c.͓ϭ͔ϩ͓ϭ͔ c.͓–89–4dupT͔ϩ͓ϭ͔ c.͓526CϾT͔ϩ͓959GϾT͔ p.͓Arg176Cys͔ϩ͓Gly320Val͔ SLC40A1 c.͓ϭ͔ϩ͓ϭ͔ c.͓1468GϾA͔ϩ͓ϭ͔ c.͓–59_–45del͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ p.͓Gly490Ser͔ϩ͓ϭ͔ (5Ј UTR) FTL,5Ј UTR ND c.͓ϭ͔ϩ͓ϭ͔ ND ND ND Table 1. Continued

Group 2 (hyperferritinemia with typical, low, or slightly increased TS)

Group 2A (evidence of tissue iron overload) Group 2B (cataract, no tissue iron overload)

Patient no. HG3142 HG3181 HG4421 D235 D167 Sex Male Male Female Female Female Age at molecular 24 41 70 58 58 diagnosis, years Hepatopathy No Liver siderosis and fibrosis (liver biopsy) NA NA NA Hypogonadotropic No No NA NA NA hypogonadism Cardiomyopathy No No NA NA NA Chemistry Clinical Other clinical signs Chronic fatigue, joint pain No Fatigue, joint pain Congenital cataract Cataract Serum iron, ␮mol/L 17.5 6.2 12.3 NA NA Serum ferritin, ␮g/L 1262 2200 1031 1120 1011 TS, % 32 7.1, after phlebotomies 23 28 6 HIC, ␮mol/g NA 65 NA ND ND HIC/age, ␮mol/g/years NA 1.6 NA ND ND 3 o 2 2007 12, No. 53, HFE p.Cys282Tyr CC CC CC ND ND p.His63Asp HH HH HH ND ND Other HFE mutation No No ND ND ND HFE polymorphism ND c.͓892ϩ48GϾA͔ϩ͓ϭ͔d ND ND ND TFR2 c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ ND ND ND HAMP c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ ND ND ND HFE2 c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ ND ND ND SLC40A1 c.͓1468GϾA͔ϩ͓ϭ͔ c.͓262AϾG͔ϩ͓ϭ͔ c.͓533GϾA͔ϩ͓ϭ͔ ND ND p.͓Gly490Ser͔ϩ ͓ϭ͔ p.͓Arg88Gly͔ϩ͓ϭ͔ p.͓Arg178Gln͔ϩ͓ϭ͔ FTL,5Ј UTR c.͓ϭ͔ϩ͓ϭ͔ c.͓ϭ͔ϩ͓ϭ͔ ND c.͓–168GϾT͔ϩ͓ϭ͔ c.͓–168GϾA͔ϩ͓ϭ͔ (known as ϩ32GϾT) (37) (known as ϩ32GϾA) (38) a HIC, hepatic iron concentration; CY, Cys282/Tyr282 heterozygote; CC, Cys282/Cys282 homozygote; YY, Tyr282/Tyr282 homozygote; HH, His63/His63 homozygote; HD, His63/Asp63 heterozygote; DD, Asp63/Asp63 homozygote; ND, not done; NA, not available. b Known polymorphism in HFE intervening sequence (intron) 2 (IVS2) (39). c A rare polymorphism of the TFR2 gene in exon 17 (2085GϾC) does not produce an amino acid change at residue 695 (Ser695). This sequence modification occurred in a young girl with JH caused by compound heterozygosity for HFE2 mutations (36). This individual was included as a control in this study. It is unlikely this nucleotide change played a role in the increased values for iron-related variables in this girl, because this change was also present in her father, who had typical values for serum iron indices. This nucleotide substitution was not found in any of the other tested patients. d Known HFE polymorphism (40). For details on the polymorphisms identified in this study, see Table 3 in the online Data Supplement. 2065 2066

Table 2. Arguments for a causative role for the newly identified mutants. Associated mutation in iron- Study of 100 related gene (HFE, HAMP, control Patient no./positive segregation for Genotype-phenotype Gene Mutation type Location (gene) Protein modification HFE2, SLC40A1) chromosomes family correlation HFE c.794dupA Exon 4 p.Trp267LeufsX80 HFE (p.Cys282Tyr Absent HG3572/Daughter (age 20 years) Yes (heterozygous (frameshift mutation compound heterozygous for c.794dupA;

frameshift) creating a premature heterozygote) asymptomatic Genes Iron of Sequencing DNA Rapid al.: et Cunat stop codon) HFE2 c.–89–4dupT 5Ј exon 2 (IVS1a Nob HFE (p.Cys282Tyr Absent HG2280/Mother: HFE, Yes (heterozygous splicing site?) homozygote) p.Cys282Tyr homozygote intronic point (classic HH adult form); HFE2, mutation) wild type; Father: HFE, p.Cys282Tyr and p.His63Asp compound heterozygote; HFE2, c.–89–4dupT heterozygote (classic HH adult form) SLC40A1 c.1468GϾA Exon 8 p.Gly490Ser (another No Absent Observed in 2 unrelated Yes (heterozygous substitution, individuals (HG3134 and missense) p.Gly490Asp, is known HG3142)/ NA at the same position) (24) SLC40A1 c.533GϾA Exon 6 p.Arg178Gln ND Absent HG4421/NA Yes (heterozygous missense) SLC40A1 c.262AϾG Exon 3 p.Arg88Gly No Absent HG3181/Father and paternal Yes (heterozygous uncle heterozygous and clinically missense) affected SLC40A1 c.–59_–45del 5Ј UTR (del 15 Nob No Absent HG3943/NA Yes (heterozygous bp)c short deletion) a IVS1, intervening sequence (intron) 1; NA, not available; ND, not done. b RNA analysis required to ascertain the causative role. c In silico modification of transcription factor binding sites (loss of E4 and myogenin site, new Adf1 (alcohol dehydrogenase distal factor 1) site with Alibaba 2.1 software (http://darwin.nmsu.edu/ϳmolb470/ fall2003/Projects/solorz/aliBaba_2_1.htm) (41). Clinical Chemistry 53, No. 12, 2007 2067

Although the technical steps have been greatly simpli- investigated in patients homozygous for the HFE fied, reading this large number of sequences still remains Cys282Tyr substitution who have an abnormally severe a time-consuming step, even with the help of automated or early clinical presentation, as was previously de- computer programs. The SCA is an interesting method for scribed. HFE2 seems to be the more frequently affected mutation analysis, but, like other sequencing-based tech- gene (9), and although additional studies are needed to nologies, its usefulness is limited for detecting gross ascertain the causative role of this mutation, this par- rearrangements (deletions or duplications). Although this ticular situation seems to be the case for patient HG2280 kind of lesion does not seem frequent in genes related to (Tables 1 and 2). iron metabolism, specific testing of these defects could be Finally, some patients in our series with ferroportin performed in combination with the SCA method, as, for mutations initially had diagnoses of classic HH (group example, with multiplex amplifiable probe hybridization 1A patients HG3134 and HG3943; Table 1). The absence analysis of duplications (28), multiplex ligation-depen- of mutations in HH genes led to the correct diagnosis. dent probe amplification (29), or quantitative multiplex This finding is in keeping with the existence of 2 kinds PCR of short fragments (30). We have planned to set up of clinical profiles associated with different types of such methods to screen gross rearrangements of iron- ferroportin mutations [for review, see (35)]. The 1st related genes, especially in patients for whom SCA se- class of mutations leads to the originally described quencing has identified no causative mutation. Indeed, by ferroportin disease, and the 2nd class displays a phe- applying the SCA method to genes involved in iron notype indistinguishable from HH. These 2 classes can metabolism, we were able to identify 9 potentially caus- be difficult to separate by only phenotype, however, ative mutations among 38 patients with suspected HH or especially in older patients (e.g., patients HG3134 and HF. The absence of detectable mutations in the remaining HG3943, ages 71 and 73 years, respectively), and func- patients has not yet been explained. One possibility is that tional assays are needed to ascertain the effect of the these patients have an unrecognized acquired condition. mutation. The correct phenotypic assessment of patients with The application of an easy-to-use, flexible sequenc- increased ferritin concentrations or even with increased ing method for the diagnosis of hereditary iron over- ferritin concentration and TS is often not complete load and HF has become a subject of great interest. The before the patient is referred to the genetics laboratory. precise diagnosis of these disorders is clinically useful, The establishment of rules could avoid unnecessary especially from a therapeutic point of view. Phleboto- sequencing. mies, the main treatment for iron overload in nonane- Unlike the use of such screening strategies as single- mic patients, are a well-tolerated and effective way to strand conformation polymorphism analysis, denatur- treat HH, but they can lead to anemia in class 1 ing gradient gel electrophoresis, or denaturing HPLC, ferroportin disease (17) and are contraindicated in the SCA method permits immediate identification of a HHCS. The availability of the causative mutation is also sequence variant. Known polymorphisms can be useful for genetic counseling of family members. This readily recognized so that rare mutations can be con- information offers the possibility of early diagnosis in sidered, depending on the position in the gene se- relatives and may help prevent severe complications, as quence and the mutation’s possible role in causing the in JH (36). phenotype. We used an extensive literature review and the criteria proposed by Cotton and Scriver (31) to ascertain the deleterious role of newly identified mu- Grant/funding support: This study was supported in part by tants (Table 2). Apart from 2 patients with previously a grant from La Recherche Clinique, CHU de Montpellier, identified FTL mutations, the other patients had new AOI 2004. undescribed variations in iron metabolism-related Financial disclosures: None declared. genes. Acknowledgments: We thank He´le`ne Igual, Sabrina Julien, Our attempt to make phenotype-to-genotype rela- and Anouk Bernard for technical assistance and Dr. Yves tionships among the identified defects stimulated inter- Lecorre and Hubert Andreani for patient referral. We are also esting lines of discussion. We can outline 3 primary grateful to Drs./Profs. Marc Ferriere, Christian Jorgensen, situations. First, digenism for mutations in other iron- Amadou Konate, Pascal Perney, Marie Christine Picot, related genes has been described for patients heterozy- Jeanne Ramos, and Jean-Franc¸ois Rossi from the AOI 2004 gous for HFE Cys282Tyr mutation and an iron overload Working Group, CHU Montpellier, for their constant compatible with a diagnosis of adult HH (9–11, 32); support. however, the HFE gene must initially be screened because additional private mutations can be found in References these patients (32–34). This is illustrated by patient 1. Feder JN, Gnirke A, Thomas W, Tsuchihashi Z, Ruddy DA, HG3572, who had a 2nd frameshift mutation in the Basava A, et al. A novel MHC class I-like gene is mutated in other HFE allele (Tables 1 and 2). Second, digenism patients with hereditary haemochromatosis. Nat Genet 1996; involving HFE2, HAMP,orTFR2 mutations must be 13:399–408. 2068 Cunat et al.: Rapid DNA Sequencing of Iron Genes

2. Aguilar-Martinez P, Biron C, Blanc F, Masmejean C, Jeanjean P, 19. Flanigan KM, von Niederhausern A, Dunn DM, Alder J, Mendell JR, Michel H, et al. Compound heterozygotes for hemochromatosis Weiss RB. Rapid direct sequence analysis of the dystrophin gene. gene mutations: may they help to understand the pathophysiology Am J Hum Genet 2003;72:931–9. of the disease? Blood Cells Mol Dis 1997;23:269–76. 20. Roux AF, Faugere V, Le Guedard S, Pallares-Ruiz N, Vielle A, 3. Walsh A, Dixon JL, Ramm GA, Hewett DG, Lincoln DJ, Anderson GJ, Chambert S, et al. Survey of the frequency of USH1 gene et al. The clinical relevance of compound heterozygosity for the mutations in a cohort of Usher patients shows the importance of C282Y and H63D substitutions in hemochromatosis. Clin Gastro- cadherin 23 and protocadherin 15 genes and establishes a enterol Hepatol 2006;4:1403–10. detection rate of above 90%. J Med Genet 2006;43:763–8. 4. Beutler E, Felitti VJ, Koziol JA, Ho NJ, Gelbart T. Penetrance of 21. Hecker KH, Roux KH. High and low annealing temperatures 845G–Ͼ A (C282Y) HFE hereditary haemochromatosis mutation increase both specificity and yield in touchdown and stepdown in the U S A. Lancet 2002;359:211–8. PCR. Biotechniques 1996;20:478–85. 5. Aguilar-Martinez P, Schved JF, Brissot P. The evaluation of 22. Don RH, Cox PT, Wainwright BJ, Baker K, Mattick JS. ‘Touchdown’ hyperferritinemia: an updated strategy based on advances in PCR to circumvent spurious priming during gene amplification. detecting genetic abnormalities. Am J Gastroenterol 2005;100: Nucleic Acids Res 1991;19:4008. 1185–94. 23. Giansily-Blaizot M, Aguilar-Martinez P, Biron-Andreani C, Jeanjean P, Igual H, Schved JF. Analysis of the genotypes and phenotypes 6. Pietrangelo A. Molecular insights into the pathogenesis of hered- of 37 unrelated patients with inherited factor VII deficiency. Eur J itary haemochromatosis. Gut 2006;55:564–8. Hum Genet 2001;9:105–12. 7. Papanikolaou G, Samuels ME, Ludwig EH, MacDonald ML, 24. Jouanolle AM, Douabin-Gicquel V, Halimi C, Loreal O, Fergelot P, Franchini PL, Dube MP, et al. Mutations in HFE2 cause iron Delacour T, et al. Novel mutation in ferroportin 1 gene is associ- overload in chromosome 1q-linked juvenile hemochromatosis. Nat ated with autosomal dominant iron overload. J Hepatol 2003;39: Genet 2004;36:77–82. 286–9. 8. Biasiotto G, Roetto A, Daraio F, Polotti A, Gerardi GM, Girelli D, et 25. Biasiotto G, Belloli S, Ruggeri G, Zanella I, Gerardi G, Corrado M, al. Identification of new mutations of hepcidin and hemojuvelin in et al. Identification of new mutations of the HFE, hepcidin, and patients with HFE C282Y allele. Blood Cells Mol Dis 2004;33: transferrin receptor 2 genes by denaturing HPLC analysis of 338–43. individuals with biochemical indications of iron overload. Clin 9. Camaschella C, Roetto A, Cali A, De Gobbi M, Garozzo G, Carella Chem 2003;49:1981–8. M, et al. The gene TFR2 is mutated in a new type of haemochro- 26. Ferrari F, Foglieni B, Arosio P, Camaschella C, Daraio F, Levi S, et matosis mapping to 7q22. Nat Genet 2000;25:14–5. al. Microelectronic DNA chip for hereditary hyperferritinemia cata- 10. Merryweather-Clarke AT, Cadet E, Bomford A, Capron D, Viprakasit ract syndrome, a model for large-scale analysis of disorders of V, Miller A, et al. Digenic inheritance of mutations in HAMP and iron metabolism. Hum Mutat 2006;27:201–8. HFE results in different types of haemochromatosis. Hum Mol 27. Chiang FT, Hsu KL, Chen WM, Tseng CD, Tseng YZ. Determination Genet 2003;12:2241–7. of angiotensin-converting enzyme gene polymorphisms: stepdown 11. Le Gac G, Scotet V, Ka C, Gourlaouen I, Bryckaert L, Jacolot S, et PCR increases detection of heterozygotes. Clin Chem 1998;44: al. The recently identified type 2A juvenile haemochromatosis 1353–6. gene (HJV), a second candidate modifier of the C282Y homozy- 28. Dent KM, Dunn DM, von Niederhausern AC, Aoyagi AT, Kerr L, gous phenotype. Hum Mol Genet 2004;13:1913–8. Bromberg MB, et al. Improved molecular diagnosis of dystrophi- 12. Pietrangelo A, Caleffi A, Henrion J, Ferrara F, Corradini E, Kulaksiz nopathies in an unselected clinical cohort. Am J Med Genet A H, et al. Juvenile hemochromatosis associated with pathogenic 2005;134:295–8. mutations of adult hemochromatosis genes. Gastroenterology 29. Schouten JP, McElgunn CJ, Waaijer R, Zwijnenburg D, Diepvens F, 2005;128:470–9. Pals G. Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification. Nucleic Acids 13. Beaumont C, Leneuve P, Devaux I, Scoazec JY, Berthier M, Res 2002;30:e57. Loiseau MN, et al. Mutation in the iron responsive element of the L ferritin mRNA in a family with dominant hyperferritinaemia and 30. Casilli F, Di Rocco ZC, Gad S, Tournier I, Stoppa-Lyonnet D, cataract. Nat Genet 1995;11:444–6. Frebourg T, et al. Rapid detection of novel BRCA1 rearrangements in high-risk breast-ovarian cancer families using multi- 14. Girelli D, Corrocher R, Bisceglia L, Olivieri O, De Franceschi L, plex PCR of short fluorescent fragments. Hum Mutat 2002;20: Zelante L, et al. Molecular basis for the recently described 218–26. hereditary hyperferritinemia-cataract syndrome: a mutation in the 31. Cotton RG, Scriver CR. Proof of “disease causing” mutation. Hum iron-responsive element of ferritin L-subunit gene (the “Verona Mutat 1998;12:1–3. mutation”). Blood 1995;86:4050–3. 32. Ka C, Le Gac G, Dupradeau FY, Rochette J, Ferec C. The Q283P 15. Montosi G, Donovan A, Totaro A, Garuti C, Pignatti E, Cassanelli S, amino-acid change in HFE leads to structural and functional et al. Autosomal-dominant hemochromatosis is associated with a consequences similar to those described for the mutated 282Y mutation in the ferroportin (SLC11A3) gene. J Clin Invest 2001; HFE protein. Hum Genet 2005;117:467–75. 108:619–23. 33. Beutler E, Griffin MJ, Gelbart T, West C. A previously unde- 16. Njajou OT, Vaessen N, Joosse M, Berghuis B, van Dongen JW, scribed nonsense mutation of the HFE gene. Clin Genet 2002; Breuning MH, et al. A mutation in SLC11A3 is associated with 61:40–2. autosomal dominant hemochromatosis. Nat Genet 2001;28: 34. Barton JC, West C, Lee PL, Beutler E. A previously undescribed 213–4. frameshift deletion mutation of HFE (c.del277; G93fs) associated 17. Pietrangelo A. The ferroportin disease. Blood Cells Mol Dis with hemochromatosis and iron overload in a C282Y heterozy- 2004;32:131–8. gote. Clin Genet 2004;66:214–6. 18. Drakesmith H, Schimanski LM, Ormerod E, Merryweather-Clarke 35. De Domenico I, Ward DM, Musci G, Kaplan J. Iron overload due to AT, Viprakasit V, Edwards JP, et al. Resistance to hepcidin is mutations in ferroportin. Haematologica 2006;91:92–5. conferred by hemochromatosis-associated mutations of ferropor- 36. Aguilar-Martinez P, Lok CY, Cunat S, Cadet E, Robson K, Rochette tin. Blood 2005;106:1092–7. J. Juvenile hemochromatosis caused by a novel combination of Clinical Chemistry 53, No. 12, 2007 2069

hemojuvelin G320V/R176C mutations in a 5-year old girl. Haema- ferritin gene in two families with the hereditary hyperferritinemia- tologica 2007;92:421–2. cataract syndrome. Blood 1998;91:319–23. 37. Cazzola M, Bergamaschi G, Tonon L, Arbustini E, Grasso M, 39. Beutler E, West C. New diallelic markers in the HLA region of Vercesi E, et al. Hereditary hyperferritinemia-cataract syndrome: chromosome 6. Blood Cells Mol Dis 1997;23:219–29. relationship between phenotypes and specific mutations in the 40. Totaro A, Grifa A, Carella M, D’Ambrosio L, Valentino M, Roth MP, iron-responsive element of ferritin light-chain mRNA. Blood 1997; et al. Hereditary hemochromatosis: a HpaI polymorphism within 90:814–21. the HLA-H gene. Mol Cell Probes 1997;11:229–30. 38. Martin ME, Fargion S, Brissot P, Pellat B, Beaumont C. A point 41. Grabe N. AliBaba2: context specific identification of transcription mutation in the bulge of the iron-responsive element of the L factor binding sites. In Silico Biol 2002;2:S1–15. Clinical Chemistry 53:12 2070–2077 (2007) Molecular Diagnostics and Genetics

Detailed Technical Analysis of Urine RNA-Based Tumor Diagnostics Reveals ETS2/Urokinase Plasminogen Activator to Be a Novel Marker for Bladder Cancer

Merle Hanke,1,2 Ingo Kausch,3 Gerlinde Dahmen,4 Dieter Jocham,3 and Jens M. Warnecke1,2*

Background: The noninvasive detection of RNA tumor Conclusions: The described methodology for RNA markers in body fluids represents an attractive diagnos- marker detection in urine appears to be clinically appli- tic option, but diagnostic performance of tissue-derived cable. The ratio of ETS2 mRNA to uPA mRNA in urine markers is often poorer when measured in body fluids is a potential marker for bladder cancer. rather than in tumors. We aimed to develop a procedure © 2007 American Association for Clinical Chemistry for measurement of tumor RNA in urine that would minimize donor-dependent influences on the results. Gene expression signatures of tumor cells have been Methods: RNA isolated from urinary cell pellet, cell- investigated for a variety of human cancers (1) to gain depleted fraction, and whole urine was quantified by insight into tumor development and progression. Mi- reverse transcription quantitative–PCR. The donor- croarray gene expression analysis by Dyrskjøt et al. (2), dependent influence of urine background on individual for example, revealed stage-specific gene clusters whose steps of the standardized procedure was analyzed using expression pattern allowed a classification of bladder an external RNA standard. Using a test set of samples cancer into 3 distinct histopathological groups: noninva- from 61 patients with bladder cancer and 37 healthy sive Ta tumors, invasive T1 tumors, and T2–T4 tumors. donors, we compared 4 putative RNA tumor markers However, validation of these consolidated findings via identified in whole urine with 5 established, tissue- reverse transcription quantitative–PCR (RT-qPCR)5 failed derived RNA tumor markers for the detection of blad- although the same sample material was used (3). The der cancer. statistical basis, the RNA isolation procedure, and the Results: Of the markers analyzed by this system, the stability of PCR amplicons, as well as the amount and RNA ratio of v-ets erythroblastosis virus E26 oncogene particularly the quality of RNA, are some factors that homolog 2 (avian; ETS2) to urokinase plasminogen influence the outcome of gene expression studies. The activator (uPA) enabled the most specific (100%) and development of one general standard operating protocol sensitive (75.4%) detection of bladder cancer from whole that should be used without restrictions in any laboratory urine, with an area under the curve of 0.929 (95% CI institution—including the sampling, isolation, quantifica- 0.882–0.976). tion (use of identical PCR amplicons), and normalization of RNA—is therefore mandatory to allow a comparison of gene expression data. 1 Kompetenzzentrum fuer Drug Design und Target Monitoring, Luebeck, To avoid differences in processing, the use of body Germany. fluids without any pretreatment (e.g., centrifugation) is 2 Institut fuer Molekulare Medizin, 3 Klinik und Poliklinik fu¨ r Urologie, most desirable for a noninvasive diagnostic test. This and 4 Institut fuer Medizinische Biometrie und Statistik, UK-S-H, Campus Luebeck, Luebeck, Germany. application is challenging, since body fluids from differ- * Address correspondence to this author at: Institut fuer Molekulare ent donors differ with respect to the amount, origin, and Medizin, Universitaetsklinikum Schleswig-Holstein, Campus Luebeck, Ratze- burger Allee 160, 23538 Luebeck, Germany. Fax 49-451-500-2729; e-mail [email protected]. Received July 9, 2007; accepted August 31, 2007. 5 Nonstandard abbreviations: RT-qPCR, reverse transcription quantita- DOI: 10.1373/clinchem.2007.091363. tive–PCR; GTC, guanidinium thiocyanate; AUC, area under the curve.

2070 Clinical Chemistry 53, No. 12, 2007 2071

integrity of cells and nucleic acids. The cells may contain To investigate selected tumor markers, we used whole RNA that is fragmented by necrotic and apoptotic pro- urine from 98 donors. For a control group, we chose cesses that also give rise to cell-free RNA in urine (4). healthy donors who best matched the age and sex distri- Cell-free RNA is resistant to the activity of urinary bution of the bladder cancer group. Detailed clinical RNases because of packaging into apoptotic bodies. In information is shown in Table 1 in the Data Supplement addition, cell death–independent processes have been that accompanies the online version of this article at described for the generation of extracellular RNA (5).To http://www.clinchem.org/content/vol53/issue12. An avoid exclusion of patients, however, a diagnostic routine overview of the experimental procedures is given in application based on body fluids ideally is not dependent Fig. 1. on the integrity of the RNA. Despite these obstacles, the detection of RNA tumor preparation of the external rna standard markers in urine has been reported as an emerging tool (rnaluc) for noninvasive tumor diagnosis (6). One of the most We synthesized truncated luciferase RNA (RNALUC)byin frequently studied markers for bladder cancer is the vitro runoff transcription using T7 RNA-polymerase. The mRNA of human telomerase reverse transcriptase T7-transcript was prepared from a purified EcoRV-re- ϩ (hTERT),6 because its concentration correlates with telom- stricted pet24a( ) vector (Promega) fragment containing the Photinus pyralis luciferase gene. The reaction mixture erase activity, which is absent in most human somatic ␮ ϫ cells but detectable in 85% of human tumors (7). RNA (50 L) for T7 transcription contained 1 reaction buffer extracts from tumor material (8–10), bladder washings (MBI Fermentas), 2 mmol/L each nucleotide triphos- phate, 40 units RNase inhibitor (Ambion), 30 units T7 (11, 12), and exfoliated cells (13) have been used for RT-qPCR–based hTERT mRNA quantification for the detection of bladder cancer, resulting in reported diagnostic sensitivities of 50% to 100%. When urine sediment was used as sample material in a larger prospective study, however, hTERT mRNA expression was often found to be near the detection limit, resulting in a sensitivity of 55% (14). In this study, we aimed to develop a RT-qPCR–based test of whole urine for robust and sensitive diagnosis of bladder cancer in a clinical setting. Toward this goal, we 1st analyzed factors influencing the RNA isolation and quantification procedure. We then applied a standardized process to compare RNA tumor markers originating from our screening experiments with established markers for bladder cancer.

Materials and Methods clinical samples The study was approved by the local research ethics committee. All 139 samples were obtained with written informed consent of the participants. For the determination of assay characteristics, urine of 21 donors (13 men, 8 women; median age 66 years) was used for preparation of urine fractions. For the investigation of daytime profiles, spontaneously voided urine of 10 participants (6 men, 4 women; median age 31 years) was collected in the morning (1st void of the day), at midday, and in the evening. We selected 10 further urine samples for enrichment experiments (5 men, 5 women; median age 30 years). Fig. 1. Experimental setup. Urine was collected from 139 participants. Based on availability and sample volume, 41 attendants were selected for the assay characterization (left box). To determine the influence of urine composition on individual steps of the process, 6 Human genes: hTERT, human telomerase reverse transcriptase; GAPDH, a defined amount of RNALUC was added as “spike” at individual steps and glyceraldehyde-3-phosphate dehydrogenase; RPLP0, ribosomal protein, large, subsequently quantified by RT-qPCR (black arrows). We collected 98 urine P0; UBC, ubiquitin C; ETS2, v-ets erythroblastosis virus E26 oncogene homolog samples for the investigation of tumor markers (gray box, right side). Hexagonal

2 (avian); uPA, urokinase plasminogen activator; HTATIP2, HIV-1 Tat interac- boxes denote standard operating procedure-based steps. RNALUC was added to tive protein 2, 30kDa; UPK1A, uroplakin 1A. each sample as external standard before RNA isolation (white arrow). 2072 Hanke et al.: RNA Quantification in Urine

RNA-Polymerase (MBI Fermentas), and 0.8 ␮g DNA quantitative pcr and data analysis

template. After transcription of RNALUC (1570 nucleo- TaqMan-based quantification of cDNA was always run in tides), the DNA template was digested using DNase I triplicate, and the non–reverse transcription reaction was (Ambion). The reaction mixture was gel-filtrated through run in duplicate. The protocol is described in detail in the a NAP-5 column (Amersham), extracted by chloroform- online Data Supplement, together with primer sequences phenol, and precipitated with ethanol. Integrity and size and amplicon characteristics (see Table 2 in the online

of RNALUC were controlled by denaturing agarose gel Data Supplement). We performed data analysis with SDS electrophoresis. RNA concentration was determined by 2.1 software (Applied Biosystems). The threshold cycle ϭ absorbance at 260 nm (1 A260 40 mg RNA/L) of a values of amplified targets were transformed into abso- dilution series of the purified RNALUC in vitro transcript lute RNA copy numbers using standard curves, allowing using NanoDrop ND-1000 UV-Vis Spectrophotometer an absolute quantification of target RNA copy numbers. (NanoDrop Technologies). We used mean values to cal-

culate RNALUC copy numbers. assay characterization 7 RT-qPCR. RNALUC (10 copies) were directly reverse- processing of urine samples transcribed, and we quantified RNALUC cDNA by use of Spontaneously voided urine was adjusted to a final con- qPCR in triplicate. We performed independent qPCR centration of 3 mol/L guanidinium thiocyanate (GTC), assay runs using a standard curve for every run to

0.025 mol/L sodium acetate, and 0.25% N-lauroylsar- determine the interrun variation of RNALUC quantifica- cosine. For the monovette-based sample collection, we tion. To determine the donor-dependent influence of the transferred 3.54 g GTC powder into a 10-mL urine urine background on the qPCRs, 5 ϫ 105 copies of

monovette (Sarstedt) and adjusted the stamp to the 7-mL RNALUC were added to RNA extracts from different mark of the monovette. Further handling of the donors. Copy numbers of RNALUC were determined by monovette was performed according to manufacturer’s the standard RT-qPCR procedure. instructions. After the GTC powder was dissolved in the collected urine, the sample was transported to the analytic Interrun imprecision of the complete RNA isolation and quan- laboratory. Adjustment to 0.025 mol/L sodium acetate tification procedure. We added 107 copies of in vitro–

and 0.25% N-lauroylsarcosine and addition of 1 mol/L transcribed RNALUC to 8 aliquots of spontaneously HEPES (pH 7) to a final volume of 10 mL were carried out voided urine of a single donor after adjustment to 3 before storage at Ϫ80 °C. mol/L GTC. For each sample, we determined copy num-

bers of RNALUC, glyceraldehyde-3-phosphate dehydroge- preparation of urinary cell pellet nase (GAPDH), ribosomal protein large P0 (RPLP0), and The cellular fraction was obtained by centrifugation of 3 ubiquitin C (UBC) by independent isolation and quanti- mL voided urine at 400g for 5 min. We discarded the fication procedures. supernatant and dissolved the pellet in a special lysis buffer (15). The lysate was frozen in liquid nitrogen and Intradonor variability. For daytime profiles, spontaneously stored at Ϫ80 °C until RNA isolation. voided urine was collected in the morning, at midday, 7 and in the evening. We added 10 copies of RNALUC to preparation of cell-depleted urine whole urine after adjustment to 3 mol/L GTC. RNA was Cell-depleted urine was obtained by passing voided urine isolated in duplicate and quantified in triplicate via ␮ through a 5- m filter (Sartorius). The urine filtrate was RT-qPCR. Copy numbers of RNALUC, GAPDH, RPLP0, treated as described for whole urine. and UBC were determined by standardized RT-qPCR procedure. standard procedure for the isolation and reverse transcription of total rna Interdonor variability. For the investigation of donor- 7 Before RNA isolation, 10 copies of RNALUC were added. dependent influence of urine background on RNA recov- 7 Total RNA was isolated using the RNeasy Midi Kit ery, we added 10 copies of RNALUC to whole urine, (Qiagen) according to the manufacturer’s instructions, cell-depleted urine, or the cellular fraction before RNA except that special lysis buffer was used instead of RLT isolation and after adjustment to 3 mol/L GTC. RNA was buffer. The procedure included an on-column digestion of isolated in duplicate, and we determined copy numbers of

genomic DNA with DNase I. RNA was eluted twice with RNALUC by the standard procedure for RNA isolation ␮ 160 L nuclease-free H2O and lyophilized. The RNA and quantification. In addition, we determined copy ␮ pellets were resolved in 20 L nuclease-free H2O. We numbers of endogenous GAPDH. used 10 ␮L RNA extract for cDNA synthesis and non- reverse transcription reaction. Additionally, 107 copies of statistical analysis and software

RNALUC were directly reverse-transcribed to verify ap- The statistical analysis of RNA marker ratios is described plied copies. A detailed description of the cDNA synthe- in detail in the online Data Supplement. The diagnostic sis protocol is provided in the online Data Supplement. power of selected markers was analyzed by ROC curves. Clinical Chemistry 53, No. 12, 2007 2073

Table 1. Interrun imprecision of the RNA isolation and quantification process. Steps Number of replicates CVa Detected RNA, copies Experimental setup

RT-qPCR 10 0.12 RNALUC Detection of in vitro RNA transcript ϩ Isolation RT-qPCR 8 0.16 RNALUC RNALUC added to 1 urine sample Isolation ϩ RT-qPCR 8 0.15 Endogenous GAPDH Aliquots of 1 urine sample Isolation ϩ RT-qPCR 8 0.18 Endogenous RPLP0 Aliquots of 1 urine sample Isolation ϩ RT-qPCR 8 0.16 Endogenous UBC Aliquots of 1 urine sample a Coefficient of intraassay variation.

For the comparison of 2 datasets (comparison of RNALUC interrun imprecision yields and housekeeping gene ratios), we used the non- Interrun imprecision (CV) for the isolation and quantifi- parametric paired Wilcoxon test. For all statistical tests, cation process of RNALUC from 8 aliquots of whole urine 2-sided P values Յ0.05 were considered statistically was 0.16 (Table 1). Endogenous RNAs for GAPDH, significant. RPLP0, and UBC were detected with a comparable impre-

Statistical analyses were performed using SAS statisti- cision (CVGAPDH 0.15; CVRPLP0 0.18; CVUBC 0.16). cal software (version 9.1, SAS Institute), SPSS statistical software (version 13.0, SPSS GmbH Software), R language interdonor variability and environment for statistical computing (version 2.3.1, We next determined the donor-dependent influence of R Foundation for Statistical Computing), and STATA/SE urine background on assay imprecision (Table 2). The statistical software (version 9.1, Stata). overall RNALUC yield of the RNA isolation and quantification process was quantified for different fractions (cell- Results depleted, cellular fraction, and whole urine) prepared RNA tumor marker quantification in a clinical setting from urine of different donors (n ϭ 21). The yield of requires a characterized process to allow precise and RNALUC was best for the cellular fraction and lowest for reproducible processing. We developed a monovette- cell-depleted urine (see Fig. 1 in the online Data Supple- based system containing GTC powder to ensure fast, ment). RNALUC yields for the cellular fraction (mean 27%) standardized processing of urine samples. For the system- were significantly higher (Wilcoxon test P ϭ 0.002) than atic characterization of the complete RNA isolation and for whole urine (mean 16%). The highest variability is quantification process, an external RNA standard was introduced by the RNA isolation step, since downstream added before RNA isolation and quantified by TaqMan- steps turned out to be robust (CVRT-qPCR 0.14; Table 2). based RT-qPCR. Reproducible processing was achieved Accordingly, for all subsequent experiments, RNALUC- using standard operating procedures for each step, in- normalized copy numbers were determined to allow a cluding sample collection. An overview of the experi- comparison of urine samples from different donors. ments that were performed to determine the assay characteristics is shown in Fig. 1 (left panel). intradonor variability The amount of endogenous housekeeping genes such as assay characterization GAPDH RNA in urine varied by 4 orders of magnitude The assay characteristics are summarized in Tables 1 (Table 2). To allow a comparison between individual and 2. samples, we calculated the ratios of housekeeping gene

Table 2. Donor-dependent influence on the RNA isolation and quantification process. Interdonor variability Number of donors CVa RNA detectability, range RT-qPCRb n ϭ 10 0.14 1.068–1.687 c External RNA calibrator, RNALUC recovery Whole urine n ϭ 21 1.01 0.007–0.479 Cell-depleted urine n ϭ 21 1.00 0.001–0.226 Cellular fraction n ϭ 21 0.82 0.016–0.901 Endogenous RNA calibrator, GAPDHd Whole urine n ϭ 21 3.87 3.7 ϫ 103 to 10.2 ϫ 107 Cell-depleted urine n ϭ 21 3.77 1.7 ϫ 103 to 0.4 ϫ 107 Cellular fraction n ϭ 21 3.46 0.4 ϫ 103 to 8.7 ϫ 107 a Coefficient of variation for RNA detectability. b ϭ RNALUC was added to RNA extracts from total urine; readout RNALUC detected divided by RNALUC input. c ϭ RNALUC was added to different urine samples; readout RNALUC detected divided by RNALUC input. d Readout ϭ copies GAPDH per mL urine. 2074 Hanke et al.: RNA Quantification in Urine

ϭ ϫ 6 RNAs. Systematic differences in RNA composition were numbers/mL; medianwhole urine 1.89 10 copy num- ϭ ϫ 5 analyzed by investigation of housekeeping gene ratios at bers/mL) and UBC RNA (mediancell-depleted 5.12 10 ϭ ϭ ϫ 6 3 times of the day (n 10). copy numbers/mL; medianwhole urine 5.58 10 copy Ratios were calculated from RNA copy numbers of numbers/mL). In some samples, the copy numbers of GAPDH, RPLP0, and UBC. The intradonor variability for GAPDH RNA detected in whole urine consisted almost GAPDH:RPLP0 (Fig. 2A) was smaller (mean CV 0.25; exclusively of cell-free RNA (see Fig. 2B, samples 2, 11, 15, range 0.09–0.61) than for GAPDH:UBC (mean CV 0.41; and 21). The cell-free RNA portion of the whole urine range 0.07–0.71). Pairwise comparison using nonparamet- fraction was higher for GAPDH RNA (mean 46.9%) com- ric Wilcoxon test provides no evidence for a significant pared with UBC RNA (mean 16.9%). influence of time of day on the GAPDH:UBC ratio (P ϭ These results provided a rationale for the reinvestiga- 0.13). Only the comparison of the GAPDH:RPLP0 ratio tion of different cell-based tumor markers for their appli- from morning and midday showed a significant differ- cability using whole urine. Their diagnostic performance ence (P ϭ 0.05). To obtain conclusive data independently was analyzed together with markers derived from RT- of patients’ compliance with instructions, we used only qPCR–based screening of RNA isolated from whole urine spontaneously voided urine (midday) instead of morning (data not shown). urine for further experiments. investigation of rna tumor markers in whole rna composition in urine fractions urine The GAPDH:UBC ratios in total differed significantly RNA was isolated from urine of 37 healthy donors and 61 from the cellular fraction (Wilcoxon test, P ϭ 0.01; n ϭ 21) patients with bladder cancer. A set of 8 different RNA tumor in contrast to the GAPDH:RPLP0 RNA ratios (Wilcoxon markers (Table 3) containing putative as well as established test, P ϭ 0.07; n ϭ 21). The cell-depleted fraction differed bladder cancer markers and the housekeeping gene GAPDH significantly from whole urine and the cellular fraction for were quantified as specified in Fig. 1, right panel. the 2 RNA ratios (Wilcoxon test, P Ͻ0.001 for GAPDH: Multivariate analysis of individual marker ratios was UBC and GAPDH:RPLP0,nϭ 21). performed to test the ability to separate the 98 donors into We next determined the amount of cell-free RNA that those with cancer and those without. We selected 3 contributes to the detected RNA in whole urine (Fig. 2B). important marker ratios using classification trees. In a Quantification of housekeeping gene RNA in the cell- logistic regression model with backward selection includ- depleted urine fractions (n ϭ 21) revealed high amounts ing the selected variables and all interactions, only the ϭ ϫ 6 of GAPDH RNA (mediancell-depleted 1.06 10 copy ratio of v-ets erythroblastosis virus E26 oncogene ho-

Fig. 2. Detection of housekeeping gene RNA in whole urine. (A), intraday fluctuations of GAPDH:RPLP0 RNA ratios in whole urine. Morning, midday, and evening urine was collected from 10 donors, and total RNA was isolated in duplicate for each donor. RNA was reverse-transcribed, and copy numbers were determined in triplicate by qPCR. Ratios were calculated for each experiment separately. Bars represent mean values of 2 independent RNA isolations and quantification procedures. SDs are depicted by error bars.(B), GAPDH and UBC RNA copy numbers/mL urine for whole urine and cell-depleted fractions of 21 donors as evaluated by qPCR. Total RNA isolation and the reverse transcription reactions were performed in duplicate; the qPCR was run in triplicate. Detected copy numbers were normalized with RNALUC to compensate for variations in total RNA isolation. White bars indicate GAPDH copy numbers/mL whole urine; white hatched bars indicate UBC copy numbers/mL whole urine. GAPDH copy numbers in the cell-depleted fraction are represented by black bars and UBC copy numbers by gray hatched bars. Bars represent mean values of 2 independent RNA isolations and quantification procedures. SDs are depicted by error bars. The term RNA instead of mRNA is used to indicate that RNA integrity was not determined. Clinical Chemistry 53, No. 12, 2007 2075

Table 3. Selected targets for RNA tumor marker analysis in whole urine. Symbol (name) Function/description Reason for validation Reference BAX (BCL2-associated X protein) Proapoptotic protein Screeninga (30) BCL2L1 (BCL2-like 1) Antiapoptotic protein Screeninga (31) HTATIP2 (30-kDa HIV-1 Tat interactive protein Proapoptotic protein, suppressor of metastasis Tumor marker (32) 2) ETS2 Transcription factor, overexpressed in prostate Screeninga (33) and breast cancer GAPDH Oxidoreductase in glycolysis and glyconeogenesis Housekeeping geneb (34) Ki-67 (antigen identified by monoclonal Only expressed in proliferating cells Tumor marker (15) antibody Ki-67) STMN1 (stathmin 1/oncoprotein 18) Ubiquitous cytosolic phosphoprotein, regulates Tumor marker (35) the microtubule filament system uPA/PLAU (plasminogen activator, urokinase) Serine protease, degradation of extracellular Tumor marker and screeninga (28) matrix; highly expressed in bladder cancer UPK1A (uroplakin 1A) Transmembrane protein, mediates signal Tumor marker (36) transduction; highly expressed in bladder cancer a RT-qPCR–based array (RT2Profiler™ PCR Array; SuperArray), detecting 84 genes involved in transformation and tumorigenesis. b GAPDH mRNA expression is known to be regulated. molog 2 (avian; ETS2) to urokinase plasminogen activator urine of patients with bladder cancer and healthy controls (uPA) allowed a statistically significant separation. To are presented in Fig. 3A. ROC curves were calculated that confirm the result of the logistic regression model, we represent the diagnostic power of the RNA marker com- performed 2 nonparametric diagonal linear discriminant binations. ROC analysis (Fig. 3B) of the study population analyses, one with the 3 important marker ratios and by use of the ETS2:uPA ratio revealed an area under the another with ETS2:uPA. The classification results of the curve (AUC) of 0.929 (95% CI 0.882–0.976), indicating the diagonal linear discriminant analyses were not different strong diagnostic power of the test. Setting the specificity (P Ͼ0.05). Therefore, the RNA ratio of ETS2:uPA was at 100% (cutoff value 0.96), a sensitivity of 75.4% was shown to be the only independent marker for the detec- achieved. For the group of low-grade tumors, we deter- tion of bladder cancer. The ETS2:uPA ratios in whole mined a sensitivity of 53.9%. Sensitivity can be enhanced

Fig. 3. Diagnostic performance of the ETS2:uPA RNA ratio. (A), box plot of ETS2:uPA RNA ratios in whole urine from healthy donors compared with patients with bladder cancer stratified according to tumor grade. The line inside each box denotes median, whereas the boxes mark 25th and 75th percentiles. Error bars mark 5th and 95th percentiles. Symbols indicate outlying data points. (B), ROC curves for selected tumor markers on the basis of 37 healthy donors and 61 patients with bladder cancer. The thick black line represents the ROC of the ETS2:uPA RNA ratio, the gray line indicates the ROC for GAPDH-normalized ETS2 RNA, and the thin black line indicates the ROC of GAPDH-normalized uPA RNA. 2076 Hanke et al.: RNA Quantification in Urine

to 79.9% for low-grade tumors and 89.1% for the group of entities (16, 17) accentuates the need to investigate the all tumors, if the specificity is reduced to 89.2%. As variability and source of RNA species in whole urine. Our demonstrated by the ROC analysis in Fig. 3B, the use of experiments indicate that RNA tumor markers derived GAPDH-normalized ETS2 RNA and GAPDH-normalized from gene expression analysis of cells or tumor material uPA RNA as individual markers did not turn out to be (e.g., hTERT, UPK1A, HTATIP2) cannot be transferred informative. The AUC values for the other marker com- unconditionally to RT-qPCR–based analysis of whole binations tested are listed in Table 3 in the online Data urine. Based on our data, one major reason is the presence Supplement. of cell-free RNA in significant amounts. When GAPDH is used for normalization, the high concentrations of cell- Discussion free GAPDH RNA, in addition to the well-known overex- RT-qPCR–based marker detection in urine sediment al- pression of GAPDH RNA in tumor tissue, mask the lows a noninvasive, sensitive, and specific detection of diagnostic power of cellular markers, as demonstrated bladder cancer. The feasibility of the diagnostic applica- here for uPA/GAPDH (AUC 0.359; see Fig. 3B). In contrast tion has been demonstrated by prospective studies (14), to extracellular DNA in urine (18, 19), the presence of but different authors report substantially different diag- specific extracellular RNA has not been reported. In nostic accuracies, even if the same markers were used. addition to apoptosis and necrosis, cell death–indepen- Using the monovette-based sampling of whole urine dent generation of extracellular RNA has been demon- presented here provides several technical advantages. strated (5). The initial step of sampling, i.e., the transfer of urine into RT-qPCR array–based screening experiments (data not the monovette containing GTC powder, does not require shown) using whole urine and the subsequent evaluation any laboratory equipment. It is therefore simple to imple- of putative tumor markers revealed the ratio of ETS2:uPA ment into hospital routine, thereby avoiding increased pro- to be a suitable marker for the detection of bladder cancer. cessing time. Inactivation of RNA-degrading enzymes oc- When the specificity was set at 100%, a sensitivity of curs rapidly by the dissolved GTC. The urine samples were 75.4% was achieved. This result outperforms the sensitiv- thus stabilized immediately, transported, and frozen within ity (28%–76%) of urinary cytology (20) as the standard 6 h after voiding. This is an important step toward the noninvasive method and most of the molecular bladder development of a standard method for the sampling of cancer markers when the specificity is set at 100% (21, 22). urine, comparable to EDTA monovettes for the collection of Stratification by tumor grade (Fig. 3A) indicated a higher blood samples. To obtain a high informative rate, the RNA sensitivity for high-grade tumors (81.3%, grade 2–3), an isolates have to be concentrated. We have recently demon- observation that has been described for other urine-based strated the feasibility of ethanol precipitation for this pur- RNA markers (14, 23, 24). pose (15). To allow maximum standardization, this step was ETS2 is a member of the ETS family of transcription substituted by lyophilization. factors that regulate a variety of biological processes (25). The use of the “1st void of the day” (morning) urine is ETS2 RNA and protein concentrations were found to be quite common in diagnostic studies because of the higher increased in breast cancer samples compared with normal content of nucleic acids. Investigation of intraday variance tissue (26). Buggy et al. (26) showed a significant corre- indicated that the RNA ratio of GAPDH:RPLP0 showed a lation of protein expression between ETS2 and uPA, significant difference between morning and midday which contains ETS2 binding sites in its promoter region. urine. This RNA ratio–dependent intraday variation in- The role of the uPA system in tumorigenesis has been dicates an advantage of spontaneously voided urine (mid- intensively studied (27). In bladder cancer, urinary pro- day), since the results are independent of the patient’s tein concentrations of uPA [e.g., (28)] as well as the uPA compliance with instructions. mRNA content in tissue [e.g., (29)], have been suitable for Using whole urine, differences in yield of the complete tumor diagnostics. The experimental data of our study RNA isolation and quantification process ranged from 0.7% population indicate a higher ETS2 RNA concentration to 47.9%. Accordingly, RNA tumor markers with low abun- compared with uPA in the case of bladder cancer, result- dance have to be excluded from this application. This holds ing in an increased ETS2:uPA RNA ratio. particularly for the analysis of hTERT RNA, in which our analysis indicated insufficient abundance for reliable quan- In conclusion, we describe a fully standardized process tification (data not shown). This is in line with results from for the isolation and quantification of RNA tumor mark- the prospective study of Weikert et al. (14), in which hTERT ers from total urine. The new tumor marker ratio of mRNA expression in exfoliated cells was often found to be ETS2:uPA is promising for the detection of bladder cancer near the detection limit. In contrast to hTERT mRNA, ETS2 with high specificity and sensitivity in a clinical setting. and uPA RNA copy numbers/mL were sufficiently high in ϭ ϫ 5 whole urine (medianETS2 4.64 10 copy numbers/mL; ϭ ϫ 5 medianuPA 4.84 10 copy numbers/mL). Grant/funding support: M.H. is supported by European The emerging interest in using RNA-tumor marker Union Grant 3ASH2000/32/19535. detection in whole urine for the diagnosis of other tumor Financial disclosures: None declared. Clinical Chemistry 53, No. 12, 2007 2077

Acknowledgments: We thank Drs. C. Ho¨ppner, M. Horn, T. 17. Bai VU, Kaseb A, Tejwani S, Divine GW, Barrack ER, Menon M, et Menke, S. Thomas, and J. Tra¨der for collecting samples, A. al. Identification of prostate cancer mRNA markers by averaged Kru¨ ger and B. Thode for technical assistance, and G. Sczakiel differential expression and their detection in biopsies, blood, and urine. Proc Natl Acad Sci U S A 2007;104:2343–8. for critically reading the manuscript. 18. Tong YK, Lo YM. Diagnostic developments involving cell-free (circulating) nucleic acids. Clin Chim Acta 2006;363:187–96. References 19. Mao L, Schoenberg MP, Scicchitano M, Erozan YS, Merlo A, 1. Wadlow R, Ramaswamy S. DNA microarrays in clinical cancer Schwab D, et al. Molecular detection of primary bladder cancer by research. Curr Mol Med 2005;5:111–20. microsatellite analysis. Science 1996;271:659–62. 2. Dyrskjøt L, Thykjaer T, Kruhoffer M, Jensen JL, Marcussen N, 20. Konety BR. Molecular markers in bladder cancer: a critical ap- Hamilton-Dutoit S, et al. Identifying distinct classes of bladder praisal. Urol Oncol 2006;24:326–37. carcinoma using microarrays. Nat Genet 2003;33:90–6. 21. Wang XS, Zhang Z, Wang HC, Cai JL, Xu QW, Li MQ, et al. Rapid 3. Schultz IJ, Wester K, Straatman H, Kiemeney LA, Babjuk M, Mares identification of UCA1 as a very sensitive and specific unique J, et al. Prediction of recurrence in Ta urothelial cell carcinoma by marker for human bladder carcinoma. Clin Cancer Res 2006;12: real-time quantitative PCR analysis: a microarray validation study. 4851–8. Int J Cancer 2006;119:1915–9. 22. Weikert S, Christoph F, Schrader M, Krause H, Miller K, Muller M. 4. Bryzgunova OE, Skvortsova TE, Kolesnikova EV, Starikov AV, Quantitative analysis of survivin mRNA expression in urine and Rykova EY, Vlassov VV, et al. Isolation and comparative study of tumor tissue of bladder cancer patients and its potential rele- cell-free nucleic acids from human urine. Ann N Y Acad Sci vance for disease detection and prognosis. Int J Cancer 2005; 2006;1075:334–40. 116:100–4. 5. Stroun M, Anker P, Beljanski M, Henri J, Lederrey C, Ojha M, et al. 23. Christoph F, Muller M, Schostak M, Soong R, Tabiti K, Miller K. Presence of RNA in the nucleoprotein complex spontaneously Quantitative detection of cytokeratin 20 mRNA expression in released by human lymphocytes and frog auricles in culture. bladder carcinoma by real-time reverse transcriptase-polymerase Cancer Res 1978;38:3546–54. chain reaction. Urology 2004;64:157–61. 6. Goessl C. Noninvasive molecular detection of cancer: the bench 24. Schultz IJ, Kiemeney LA, Willems JL, Swinkels DW, Witjes JA, de and the bedside. Curr Med Chem 2003;10:691–706. Kok JB. Survivin and MKI67 mRNA expression in bladder washings 7. Nakamura TM, Morin GB, Chapman KB, Weinrich SL, Andrews WH, of patients with superficial urothelial cell carcinoma correlate with Lingner J, et al. Telomerase catalytic subunit homologs from tumor stage and grade but do not predict tumor recurrence. Clin fission yeast and human. Science 1997;277:955–9. Chem 2006;52:1440–2. 8. Ito H, Kyo S, Kanaya T, Takakura M, Inoue M, Namiki M. 25. Foulds CE, Nelson ML, Blaszczak AG, Graves BJ. Ras/mitogen- Expression of human telomerase subunits and correlation with activated protein kinase signaling activates Ets-1 and Ets-2 by telomerase activity in urothelial cancer. Clin Cancer Res 1998;4: CBP/p300 recruitment. Mol Cell Biol 2004;24:10954–64. 1603–8. 26. Buggy Y, Maguire TM, McDermott E, Hill AD, O’Higgins N, Duffy 9. De Kok JB, Schalken JA, Aalders TW, Ruers TJ, Willems HL, MJ. Ets2 transcription factor in normal and neoplastic human Swinkels DW. Quantitative measurement of telomerase reverse breast tissue. Eur J Cancer 2006;42:485–91. transcriptase (hTERT) mRNA in urothelial cell carcinomas. Int J 27. Duffy MJ, Duggan C. The urokinase plasminogen activator system: Cancer 2000;87:217–20. a rich source of tumour markers for the individualised manage- 10. de Kok JB, Ruers TJ, van Muijen GN, van Bokhoven A, Willems HL, ment of patients with cancer. Clin Biochem 2004;37:541–8. Swinkels DW. Real-time quantification of human telomerase reverse transcriptase mRNA in tumors and healthy tissues. Clin 28. Casella R, Shariat SF, Monoski MA, Lerner SP. Urinary levels of Chem 2000;46:313–8. urokinase-type plasminogen activator and its receptor in the 11. Isurugi K, Suzuki Y, Tanji S, Fujioka T. Detection of the presence detection of bladder carcinoma. Cancer 2002;95:2494–9. of catalytic subunit mRNA associated with telomerase gene in 29. Bhuvarahamurthy V, Schroeder J, Denkert C, Kristiansen G, exfoliated urothelial cells from patients with bladder cancer. J Urol Schnorr D, Loening SA, et al. In situ gene expression of urokinase- 2002;168:1574–7. type plasminogen activator and its receptor in transitional cell 12. de Kok JB, van Balken MR, Roelofs RW, van Aarssen YA, Swinkels carcinoma of the human bladder. Oncol Rep 2004;12:909–13. DW, Klein Gunnewiek JM. Quantification of hTERT mRNA and 30. Reed JC. Proapoptotic multidomain Bcl-2/Bax-family proteins: telomerase activity in bladder washings of patients with recurrent mechanisms, physiological roles, and therapeutic opportunities. urothelial cell carcinomas. Clin Chem 2000;46:2003–7. Cell Death Differ 2006;13:1378–86. 13. Bowles L, Bialkowska-Hobrzanska H, Bukala B, Nott L, Razvi H. A 31. Le Bras M, Rouy I, Brenner C. The modulation of inter-organelle prospective evaluation of the diagnostic and potential prognostic cross-talk to control apoptosis. Med Chem 2006;2:1–12. utility of urinary human telomerase reverse transcriptase mRNA in 32. Ito M, Jiang C, Krumm K, Zhang X, Pecha J, Zhao J, et al. TIP30 patients with bladder cancer. Can J Urol 2004;11:2438–44. deficiency increases susceptibility to tumorigenesis. Cancer Res 14. Weikert S, Krause H, Wolff I, Christoph F, Schrader M, Emrich T, 2003;63:8763–7. et al. Quantitative evaluation of telomerase subunits in urine as 33. Hsu T, Trojanowska M, Watson DK. Ets proteins in biological biomarkers for noninvasive detection of bladder cancer. Int J control and cancer. J Cell Biochem 2004;91:896–903. Cancer 2005;117:274–80. 34. de Kok JB, Roelofs RW, Giesendorf BA, Pennings JL, Waas ET, 15. Menke TB, Boettcher K, Kruger S, Kausch I, Boehle A, Sczakiel G, Feuth T, et al. Normalization of gene expression measurements in et al. Ki-67 protein concentrations in urothelial bladder carcino- tumor tissues: comparison of 13 endogenous control genes. Lab mas are related to Ki-67-specific RNA concentrations in urine. Clin Invest 2005;85:154–9. Chem 2004;50:1461–3. 35. Mistry SJ, Atweh GF. Stathmin expression in immortalized and 16. Groskopf J, Aubin SM, Deras IL, Blase A, Bodrug S, Clark C, et al. oncogene transformed cells. Anticancer Res 1999;19:573–7. APTIMA PCA3 molecular urine test: development of a method to 36. Kageyama S, Yoshiki T, Isono T, Tanaka T, Kim CJ, Yuasa T, et al. aid in the diagnosis of prostate cancer. Clin Chem 2006;52: High expression of human uroplakin Ia in urinary bladder transi- 1089–95. tional cell carcinoma. Jpn J Cancer Res 2002;93:523–31. Clinical Chemistry 53:12 2078–2085 (2007) Molecular Diagnostics and Genetics

IL18 Haplotypes Are Associated with Serum IL-18 Concentrations in a Population-Based Study and a Cohort of Individuals with Premature Coronary Heart Disease

Simon R. Thompson,1 Pamela A. McCaskie,2 John P. Beilby,3 Joseph Hung,4,5 Michelle Jennens,3 Caroline Chapman,3 Peter Thompson,4 and Steve E. Humphries1*

Background: Interleukin (IL)-18 is a proinflammatory with significantly lower IL-18 than others. In CUDAS, cytokine that has been implicated in several diseases, those carrying 2 copies had IL-18 concentrations 15% in ;(0.002 ؍ including atherosclerosis, and increased circulating lower than those carrying no copies (P These .(0.004 ؍ IL-18 concentrations increase risk of future coronary CUPID, the difference was 22% (P heart disease (CHD). We evaluated the effect of common associations remained significant after adjustment for variation within the IL18 gene on concentrations of age, sex, hypertension, HDL cholesterol, waist-to-hip circulating IL-18. ratio, and alcohol consumption. Despite being associ- Methods: We measured IL-18, by ELISA, in the popula- ated with differences in IL-18 concentrations, the hap- tion-based study group [Carotid Ultrasound Disease lotypes did not occur at different frequencies in those Assessment Study (CUDAS)] and a predominantly male with or without carotid atherosclerotic plaques. cohort with premature cardiovascular disease [Carotid Conclusions: Variation within IL18 affects IL-18 con- Ultrasound in Patients with Ischaemic Heart Disease centrations in healthy and diseased individuals and (CUPID)]. Using a tagging single-nucleotide polymor- thus may influence the pathophysiology of plaques at phism (SNP) approach that captured >90% of genetic all stages of CHD progression. variation, we identified 4 common (>10%) haplotypes. © 2007 American Association for Clinical Chemistry Results: A common SNP was associated with differences in IL-18 concentrations; in CUDAS individuals Interleukin (IL)6-18, a pleiotropic cytokine involved in carrying 2 copies of the rare allele, concentrations were both the innate and the adaptive immune response, is -widely expressed in monocytes/macrophages, adipo .(0.002 ؍ higher than in those with no copies (P 13% Haplotypes were also associated with significant differ- cytes, keratinocytes, Kupffer cells, and osteoblasts (1). ences in IL-18 concentrations in CUDAS and CUPID. Originally identified as an interferon-␥–inducing factor Haplotype GTATA (frequency 23%) was associated (2), it stimulates interferon-␥ production in T lymphocytes and natural killer cells, both of which are key components of atherosclerotic plaque progression and 1 Department of Cardiovascular Genetics, University College London, stability (3). Indeed, increased IL-18 expression is seen in London, United Kingdom. atherosclerotic plaques and is associated with plaque 2 Laboratory for Genetic Epidemiology, Western Australian Institute for instability (4). The use of animal models has further Medical Research, University of Western Australia, Perth, Australia. 3 Clinical Biochemistry, Western Australia Centre for Pathology and Med- ical Research, University of Western Australia, Perth, Australia. 4 Sir Charles Gairdner Hospital Campus, Heart Research Institute of Western Australia, Perth, Australia. 6 Nonstandard abbreviations: IL, interleukin; IL-18BP, IL-18 binding pro- 5 School of Medicine and Pharmacology, University of Western Australia, tein; CHD, coronary heart disease; CUDAS, Carotid Ultrasound Disease Perth, Australia. Assessment Study; CUPID, Carotid Ultrasound in Patients with Ischaemic * Address correspondence to this author at: Department of Cardiovascular Heart Disease; IMT, intima-media thickness; IIPGA, Innate Immunity PGA; Genetics, University College London, 5 University St., London, WC1E 6JF, IMT, intima-medial wall thickness; tSNP, tagging single-nucleotide polymor- United Kingdom. Fax 0207-679-6212; e-mail [email protected]. phism; SNP, single-nucleotide polymorphism; HWE, Hardy–Weinberg equi- Received May 29, 2007; accepted October 2, 2007. librium; LD, linkage disequilibrium; hGTATA, GTATA haplotype; hTTATC, Previously published online at DOI: 10.1373/clinchem.2007.092692 TTATC haplotype.

2078 Clinical Chemistry 53, No. 12, 2007 2079

demonstrated the influence of IL-18 and the benefit of its age decile between 20 and 70 years. Participants took part inhibition [through IL-18 binding protein (IL-18BP), its in the original 1989 Australian National Heart Foundation intrinsic inhibitor] on plaque progression and composi- Perth Risk Factor Prevalence Survey and agreed to attend tion (5–8). the study clinic between June 1995 and December 1996, Although IL-18 experimental data appear to be consis- having not previously had carotid artery surgery. The tent, clinical studies have proven inconsistent and contro- CUPID cohort was collected in 1995 and consisted of 556 versial. A large study in healthy, middle-aged European individuals (487 men) between 26 and 60 years of age men has shown that total plasma IL-18 concentrations are with premature CHD. All CUPID participants were med- an independent predictor of coronary events (9), and ically stable at the time of data collection but had a history importantly, variation within the IL-18 gene (IL18)7 has of angina, unstable angina, or myocardial infarction and been shown to influence circulating concentrations of angiographically demonstrated CHD with at least 1 cor- IL-18 and clinical outcome in patients with coronary onary vessel with Ͼ50% stenosis. Written informed con- heart disease (CHD) (10). However, recent findings from sent was obtained from all study participants, and the the Monitoring of Cardiovascular Disease (MONICA)/ Institutional Ethics Committee of the University of West- Cooperative Health Research in the Augsburg Region ern Australia approved the study protocol. (KORA) Ausburg Case-Cohort Study show no association between IL-18 and risk of future CHD in apparently assessments healthy individuals (11). Following on from the experi- A self-administered questionnaire was used to report the mental data outlined above, several groups have sought prevalence of smoking, physician-diagnosed hyperten- to clarify the role of IL-18 and IL-18BP in the development sion, diabetes, angina pectoris, myocardial infarction, of plaque instability in cross-sectional studies (12–14), but stroke, and medication use among study participants. We again with inconsistent results. calculated smoking lifetime exposure by pack-years, mea- The exact nature of IL-18’s influence on CHD risk sured resting systolic and diastolic blood pressures with a remains unclear. IL-18 shares similar downstream signal- mercury column manometer, and recorded anthropomet- ing pathways with several other cytokines [interaction ric measurements (waist and hip circumferences, height, with MyD88 (15), nuclear translocation of NF-␬B (16)], and weight). A fasting venous blood sample was collected but this may not be the case for all cell types (17). The fact from each participant, and sera were stored at Ϫ70 °C that administration of IL-18 does not induce fever in until analysis. We performed bilateral carotid B-mode animals (18), yet both IL-6 and IL-1 do (19), suggests ultrasound using a 7.5-MHz annular phased-array trans- substantial differences in site and method of action be- ducer on an Interspec (Apogee) CX 200 ultrasound ma- tween the cytokines. Also, IL-18 was found to be predic- chine (20, 21). The ultrasound study was used to deter- tive of future CHD events independent of C-reactive mine the presence of focal carotid plaque and measure the protein (9), which further supports this hypothesis. mean common carotid artery intima-medial wall thick- Therefore, if IL-18 is part of an alternative inflammatory ness (IMT) as described (20, 21). Presence of an athero- pathway in CHD, knowledge of important genetic effects sclerotic plaque was defined as a clearly identified area of may improve risk prediction beyond that of other inflam- focal increased thickness (Ͼ1 mm) of the intima-media matory mediators. layer. The purpose of this study was to investigate the effect of genetic variation within IL18 on circulating IL-18 con- IL18 single-nucleotide polymorphism centrations in a large cross-sectional community-based identification and genotyping study and in a medically stable cohort with angiographi- Tagging single-nucleotide polymorphism identification. We cally demonstrated CHD, thereby examining the impact used IL-18 resequencing data from Innate Immunity of genetic variation within IL18 and its potential impor- PGA (IIPGA, http://innateimmunity.net) in conjunction tance in determining CHD risk. with a haplotype-tagging single-nucleotide polymorphism (tSNP) program, tagSNPs.exe (22). A previous report Materials and Methods gives greater detail of how the single-nucleotide poly- study groups morphisms (SNPs) were selected (23). Individuals were selected from 2 Western Australian studies: the Carotid Ultrasound Disease Assessment snp genotyping Study (CUDAS) (20) and the Carotid Ultrasound in We carried out SNP genotyping using TaqMan protocols Patients with Ischaemic Heart Disease (CUPID) study (for primer and probe pairs, see the Data Supplement that (21). Both groups are predominantly European-Austra- accompanies the online version of this article at http:// lian. The CUDAS group consists of 1109 participants with www.clinchem.org/content/vol53/issue12) using a 5-␮L an equal male-to-female ratio and equal numbers in each reaction. Samples were cycled for 40 cycles on MJ Re- search cyclers and read using a Victor2 fluorometer (Perkin-Elmer). Undetermined samples were regeno- 7 Human gene: IL18, interleukin 18 (interferon-␥–inducing factor). typed using tetra-primer amplification refraction muta- 2080 Thompson et al.: IL18 Haplotypes Are Associated with IL-18 Concentrations

tion system-PCR or restriction fragment length poly- polynomial terms. Haplotypes were inferred for individ- morphism protocols (for primer and probe pairs, see uals with ambiguous phase, and haplotype frequencies online Data Supplement). Genotypes for all SNPs were were estimated using an expectation-maximization algo- recorded for Ͼ99% of individuals in CUDAS and Ͼ98% in rithm as described by Excoffier and Slatkin (30). Using CUPID. this approach, it was not possible to statistically test for differences in haplotype frequencies between the study biochemical analysis groups, as the frequencies are estimated and not directly We measured serum IL-18 by use of a commercially observed. We used SimHap (www.genepi.com.au/ available ELISA method (MBL Co. Ltd.) as described (9). projects/simhap), R (31), and JLIN (32) to manage and The ELISA uses monoclonal antibodies against 2 different analyze data. P values were derived via empirical simu- epitopes of human IL-18, one of which is peroxidase- lation where possible. Statistical significance was defined conjugated. The within-run CV was 5.4% at a mean value at a nominal 5% level. No adjustments were made for of 400 ␮g/L (28 samples); between-run CV was 8.2% at multiple testing, because this has been suggested to lead 298 ␮g/L (9 samples) and 7.8% at 496 ␮g/L (9 samples). to errors in interpretation (33). Total cholesterol, HDL, and triglyceride concentrations were measured enzymatically (24). Results statistical analysis baseline characteristics The primary quantitative outcome variable of the associ- The baseline characteristics of both study groups are ation analyses was IL-18 concentration. The principal shown in Table 1. CUPID is a diseased cohort, while explanatory variables were the 5 genotyped IL18 poly- CUDAS is a population-based sample; as such, the pro- morphisms and inferred haplotypes. The SNP genotypes portion of smokers and those with metabolic syndrome is were coded into 3 classes and analyzed categorically, with far greater in CUPID than CUDAS. Furthermore, the the most common homozygous genotype as a reference number of individuals using cholesterol-lowering and category to which the effect of the other 2 genotypes was antihypertensive medication was far greater in CUPID compared. Haplotypes were recoded as independent vari- than CUDAS, and therefore mean lipid concentrations ables into 3 classes (0, 1, or 2), representing the number of and blood pressure are similar in the 2 groups. That copies of each haplotype that made up an individual’s concentrations of IL-18 were significantly higher in diplotype. They were then analyzed categorically with CUPID than CUDAS is likely to be accounted for by an zero copies as the baseline. The genetic effect types of the increased prevalence of obesity and metabolic syndrome mutations for SNPs and haplotypes were determined to in CUPID; IL-18 concentrations strongly associate with be additive, dominant, or recessive by analyzing the trend obesity and metabolic syndrome risk traits (24, 34). of the ␤-coefficients for each category. We tested each SNP for departure from Hardy–Wein- tSNP selection and validation berg equilibrium (HWE) using a modified Markov-chain A tSNP set comprising SNPs Ϫ9731 GϾT, Ϫ5848 TϾC, random walk algorithm (25). We analyzed pairwise link- ϩ4860 AϾC, ϩ8855 TϾA, and ϩ11015 AϾC was selected age disequilibrium (LD) by use of a likelihood-ratio test, (rs1946519, rs2043055, rs549908, rs360729, and rs3882891), whose empirical distribution was obtained via a permu- based on haplotypes derived from resequencing data of tation procedure (26). Lewontin disequilibrium coeffi- white individuals (23). Genotypes for all 5 SNPs were cient DЈ was calculated for each pairwise comparison (27). determined in all study groups. LD was significant and Sex was analyzed as a dichotomous variable. All other high across the whole region, as shown in Fig. 1 (all DЈ measurements were analyzed as continuous, gaussian- Ͼ0.72 in CUDAS). ϩ4860 AϾC was the only exonic SNP distributed variables. IL-18 concentration was not gauss- found within IL18 by IIPGA; however, it is nonsynony- ian distributed and was therefore logarithmically (base mous—no other potentially functional SNPs were found. 10) transformed before analysis. Multivariate analyses used generalized linear models (linear and logistic regression) (28) to model the effects of multiple covariates on allele and haplotype frequency ϩ Ͼ continuous and dichotomous outcomes. We used both SNP frequency. The genotype distribution of 8855 T A forward and backward stepwise variable selection proce- differs significantly from that expected by HWE, with dures to determine a useful subset of independent pre- fewer heterozygotes observed than expected in both dictors on the outcomes of interest. Checks of goodness of CUDAS and CUPID (Table 2). This may be due to the model fit included examination of Akaiki information clustering methods used in TaqMan genotyping. The criteria to determine the models that best fit the data. In intermediate cluster (relating to heterozygotes) tends to the multivariate analysis presented here those predictors be less focused, therefore making undetermined hetero- were age, sex, hypertension, HDL, waist-to-hip ratio, and zygote samples more likely. The distribution of ϩ8855 alcohol consumption. Checks of goodness of fit (29) genotypes in previous population-based studies we have included an investigation of the need for interaction or genotyped (23) have also shown that it has a tendency to Clinical Chemistry 53, No. 12, 2007 2081

Table 1. Baseline characteristics of the 2 study groups.a CUDAS CUPID n 1109 556 Male, n (%) 558 (50) 487 (87) Age, years 53 (13) 50 (5)* Body mass index, kg/m2 26.1 (4.1) 28.2 (4.1)* Waist-to-hip ratio, cm 0.84 (0.09) 0.92 (0.07)* Waist circumference, cm 84.65 (12.46) NA Systolic blood pressure, mmHg 128 (19) 126 (10)* Diastolic blood pressure, mmHg 80 (10) 81 (10)* Total cholesterol, mmol/L 5.6 (1.0) 5.2 (1.0)* LDL cholesterol, mmol/L 3.6 (0.9) 3.2 (1.0)* Triglycerides, mmol/L 1.3 (0.7) 2.0 (1.4)* HDL cholesterol, mmol/L 1.3 (0.4) 1.1 (0.3)* IL-18 protein, ␮g/L 327.8 (146.6) 366.2 (156.0)* Physician-diagnosed diabetes, n (%) 23 (2.1) 88 (15.8)* History of myocardial infarction or stroke, n (%) 5.7 (63) 61.3 (341)* Cholesterol-lowering medication, n (%) 75 (6.8) 371 (66.7)* Ever smoked, n (%) 549 (49.5) 406 (72.8)* Plaque, n (%) 284 (25.6) 329 (59.5)* IMT, mm 0.71 (0.14) 0.68 (0.12) Metabolic syndrome, n (%) 209 (18.8) 213 (38.3)* a Data are mean (SD) unless noted otherwise. NA, not applicable. * P Ͻ0.05. deviate from HWE; therefore, we elected to analyze the As shown in Table 2, there were no significant data cautiously and have included and excluded ϩ8855 differences in allele frequency between CUDAS and in the analyses to follow. CUPID.

Fig. 1. Location, linkage disequilibrium (in CUDAS), and common haplotypes of the selected tSNP set within IL18. 2082 Thompson et al.: IL18 Haplotypes Are Associated with IL-18 Concentrations

Table 2. SNP minor allele frequency (MAF) and HWE and rare-allele homozygotes had IL-18 concentrations P values for both study groups. higher than common-allele homozygotes, though not significantly (P ϭ 0.05 and 0.74, respectively). Furthermore, CUDAS CUPID Between-group both Ϫ9731 and ϩ11015 were significantly associated SNP MAF HWE p MAF HWE P P with differences in IL-18 concentrations (P ϭ 0.02 and Ϫ9731 0.401 0.40 0.421 0.30 0.31 0.01, respectively); in both cases the rare allele was asso- Ϫ5848 0.362 0.81 0.367 0.35 0.79 ciated with higher IL-18 concentrations. These associa- ϩ 4860 0.352 0.06 0.319 0.06 0.76 tions remained significant in multivariate analysis. ϩ8855 0.313 0.04 0.321 0.05 0.63 ϩ 11015 0.422 0.06 0.427 0.08 0.83 il-18 concentrations: haplotype analysis To ensure sufficient statistical power in the following haplotype frequency analyses, analysis was restricted to common haplotypes Ͼ The 5 IL18 polymorphisms generated 4 common haplo- observed at a frequency 10%. types (frequency Ͼ10%) accounting for 93% of the inferred haplotypes for both cohorts. There was no substan- cudas tial difference in haplotype frequencies or organization Haplotypes were significantly associated with differences between study groups (Fig. 1). As expected from the in IL-18 concentrations in CUDAS (Table 4). More specif- observation of no difference in allele frequency, it appears ically, the GTATA haplotype (hGTATA) was associated Ͻ there is no difference in genetic architecture between the with significantly lower IL-18 concentrations (P 0.001) study groups. and explained 1.7% of variance in IL-18 concentrations. In those carrying 2 copies of the haplotype, IL-18 concentra- il-18 concentrations: single snp analysis tions were 15% lower than in those carrying no copies There were no consistent single SNP associations with (P ϭ 0.002). The effect appeared to be additive, with those IL-18 concentrations across both studies (Table 3), and individuals carrying 1 copy of the haplotype having none of the SNPs explained substantial proportions of the concentrations midway between those carrying 2 copies variance in IL-18 concentrations in either study (CUDAS: and no copies. hGCATA also appeared to have an effect all SNPs Ͻ0.9%; CUPID: all SNPs Ͻ1.2%). In CUDAS, on IL-18 concentrations, with those carrying 2 copies of IL-18 concentrations were only significantly different by the haplotype having significantly higher IL-18 concen- Ϫ5848 genotype (P ϭ 0.003), with the rare allele associ- trations than those carrying no copies (P ϭ 0.02). Globally, ated with higher concentrations than the common allele. hGCATA explained 0.4% of the variance in IL-18 concen- In CUPID, however, Ϫ5848 was not associated with IL-18 trations and was marginally associated with differences in concentrations globally (P ϭ 0.14), but both heterozygotes IL-18 concentration (P ϭ 0.04). The effect appeared to be

Table 3. Univariate and multivariate SNP analysis of IL18 genotype and IL-18 concentration in both CUDAS and CUPID.a SNP CUDAS P1 P2 CUPID P1 P2 Ϫ9731 GG 298.7 (286.6–311.4) 322.5 (304.1–342.0) GT 299.0 (288.0–310.4) 0.98 0.77 335.5 (320.1–351.6) 0.31 0.58 TT 308.4 (290.2–327.8) 0.40 0.54 372.4 (342.9–404.5) 0.006 0.005 Ϫ5848 TT 286.5 (275.5–298.0) 324.0 (307.0–342.1) TC 306.5 (295.4–318.1) 0.01 0.02 348.4 (331.8–365.8) 0.05 0.07 CC 324.3 (302.8–347.3) 0.002 0.002 330.1 (300.2–363.1) 0.74 0.47 ϩ4860 AA 296.6 (286.0–307.5) 334.3 (317.9–351.7) AC 303.5 (291.7–315.8) 0.40 0.62 338.2 (321.8–355.4) 0.75 0.89 CC 307.4 (285.1–331.5) 0.40 0.44 343.9 (305.9–386.5) 0.67 0.76 ϩ8855 TT 297.1 (286.6–308.0) 332.4 (316.0–349.6) TA 304.1 (292.1–316.6) 0.40 0.67 339.4 (323.1–356.5) 0.56 0.63 AA 303.7 (281.4–327.7) 0.61 0.57 343.9 (305.9–386.5) 0.60 0.68 ϩ11 015 AA 295.1 (282.7–308.0) 314.4 (296.1–333.9) AC 301.7 (290.8–313.1) 0.44 0.37 343.2 (327.6–359.5) 0.02 0.09 CC 307.8 (290.5–326.1) 0.25 0.23 362.7 (333.8–394.1) 0.007 0.01 a Data are estimated geometric mean IL-18 pg/mL (95% CI). P1, univariate P value compared to reference; P2, multivariate P value compared to reference. Clinical Chemistry 53, No. 12, 2007 2083

Table 4. Univariate and multivariate haplotype analysis of IL18 haplotypes and IL-18 concentrations in both CUDAS and CUPID.a Haplotype and copy number CUDAS P1 P2 CUPID P1 P2 GCATA 0 291.4 (280.9–302.3) 334.2 (319.4–349.6) 1 305.7 (294.2–317.7) 0.08 0.10 340.4 (322.5–359.4) 0.65 0.55 2 322.8 (298.1–349.5) 0.02 0.02 335.7 (301.3–374.1) 0.95 0.68 TTCAC 0 301.1 (290.9–311.6) 338.1 (321.7–355.3) 1 298.2 (286.1–310.8) 0.72 0.58 333.4 (317.3–350.4) 0.70 0.67 2 307.1 (281.5–335.1) 0.68 0.85 353.2 (308.5–404.4) 0.54 0.55 GTATA 0 314.8 (304.7–325.2) 358.1 (342.3–374.6) 1 283.6 (271.7–296.0) Ͻ0.001 0.002 314.1 (297.5–331.5) Ͻ0.001 Ͻ0.001 2 269.1 (245.2–295.3) 0.002 Ͻ0.001 279.3 (237.4–328.5) 0.004 0.002 TTATC 0 299.2 (290.9–307.6) 327.3 (315.7–339.4) 1 309.1 (291.6–327.6) 0.33 0.11 373.3 (345.7–403.0) 0.003 0.01 2 268.8 (218.0–331.4) 0.32 0.23 507.0 (356.5–720.9) 0.02 0.004 a Data are estimated geometric mean IL-18 pg/mL (95% CI). P1, univariate P value compared to reference; P2, multivariate P value compared to reference. additive, although for individuals carrying only 1 copy of tion and progression. When this hypothesis was tested in hGCATA the association did not reach statistical signifi- CUDAS and CUPID, none of the IL18 haplotypes were cance (P ϭ 0.08). Neither of these associations was atten- associated with risk of carotid plaque or mean IMT (data uated in multivariate analysis or in analysis that excluded not shown). However, hTTATC, shown to be associated ϩ8855 (data not shown). with higher IL-18 concentrations, was borderline significantly associated with a higher risk of carotid plaque in cupid CUPID (P ϭ 0.05). Both hGTATA and the TTATC haplotype (hTTATC) were associated with statistically significant differences in IL-18 Discussion concentrations (P Ͻ0.001 and P ϭ 0.001, respectively; The data presented here demonstrate that genetic varia- Table 4), and explained 3.0% and 2.1% of the variance in tion within the IL18 gene affects IL-18 protein concentra- IL-18 concentrations, respectively. Furthermore, both as- tions and may therefore affect IL-18 concentrations at the sociations were additive. Individuals with 2 copies of site of disease. hGTATA had IL-18 concentrations, on average, 22% In single SNP analysis, only 1 SNP, Ϫ5848, was as- lower than those with no copies of the haplotype (P ϭ sociated with differences in serum IL-18 concentra- 0.004), whereas those carrying 1 copy of the haplotype tions. Ϫ5848C was seen only on 1 common haplotype, had concentrations 12% lower than those with no copies hGCATA, and as might be expected, there was evidence (P Ͻ0.001). hTTATC operated similarly but in the oppo- that this haplotype was associated with significantly site direction, with those carrying 2 copies having esti- higher IL-18 concentrations. By contrast, Ϫ5848T was seen mated geometric mean IL-18 concentrations 55% higher on 3 common haplotypes. One of these, hGTATA, was than those carrying no copies (P ϭ 0.02), whereas those associated with significant differences in IL-18 concentra- with 1 copy had concentrations 15% higher than those tions in both study groups. This haplotypic effect was of with no copies (P ϭ 0.003). Both associations were not greater impact, in terms of difference in IL-18 concentra- attenuated in multivariate analysis, or in analysis that tions, than that seen with hGCATA and was of greater excluded ϩ8855 (data not shown). statistical significance. Of interest, hGCATA was not significantly associated with differences in IL-18 concen- carotid plaque and mean imt haplotype trations in CUPID (P ϭ 0.88), suggesting that the effect analysis observed in the healthy CUDAS group may be attenuated All participants in both studies were assessed for carotid or masked in diseased individuals. IMT and evidence of plaques by ultrasound. Because Further evidence for hGTATA’s effect on IL-18 concen- IL-18 is produced in atherosclerotic plaques and related to trations comes from the AtheroGene cohort. Tiret et al. plaque instability (4, 8), it is possible that genetic variants (10) highlighted the role of the IL18 gene in cardiovascu- that alter IL-18 concentrations could affect plaque initia- lar disease, demonstrating that IL18 haplotypes caused 2084 Thompson et al.: IL18 Haplotypes Are Associated with IL-18 Concentrations

variation in IL-18 serum concentrations and were associ- IL-18 in CHD has not yet been established, so it is entirely ated with cardiovascular mortality. The 5 IL18 tSNPs possible that IL-18 is simply reflecting burden of disease genotyped here were not the same as those used by Tiret and is not directly causing it (reverse causality). et al. (10). Because both are studies of whites, however, it Several studies have shown IL-18BP to be a major can be deduced, with IIPGA IL18 resequencing (http:// determinant of free IL-18 concentrations (38–40), and the innateimmunity.net/), that hGTATA marks the same absence of IL-18BP measures in these analyses is a limi- haplotype previously found to be associated with lower tation. Without such measures, it is not possible to infer IL-18 concentrations and a protective effect on risk. Fur- that the differences observed here are strong enough to ther data have demonstrated that this haplotype has a affect free IL-18 concentrations, and whether they may significant effect on IL18 mRNA concentrations in trans- affect disease progression. With regard to assessing func- formed lymphoblastoid cell lines (35), although the exact tionality, it is our belief that the functional SNP lies on the functional mechanism was not elucidated. Furthermore, haplotype but has not been genotyped directly. Because of Frayling et al. (36) studied the relationship of the the tSNP methodology, we have no means of identifying rs5744256 SNP on IL-18 production and found the rare functional SNPs on hGTATA, and the task has added allele to be associated with lower IL-18 concentrations. difficulty given the high degree of LD between the tSNPs Again, using the IIPGA gene-wide haplotypes, we find and other SNPs within the gene. that the rare allele is seen only on 1 common haplotype, which was represented by hGTATA in this study. There- In conclusion, we show that common genetic variation fore the same haplotype has been associated with lower within IL18 is associated with interindividual differences IL-18 concentrations in 4 separate studies by 3 separate in concentrations of IL-18 protein. Such an association laboratories, as well as differences in body mass index in may prove important in individual prediction of CHD a further 2 disease cohorts (23). The SNP effect of Ϫ5848 risk (and other diseases in which IL-18 has been implicated). may itself be functional, or it is most probably derived from the haplotype effect, with the true functional SNP being elsewhere on the haplotype, as none of the other Grant/funding support: S.R.T. was supported by a British haplotypes carrying Ϫ5848T showed a consistent effect. Heart Foundation studentship FS/04/039, and S.E.H. re- Furthermore, hGCATA may also be mirroring hGTATA’s ceived support from Programme Grant 2000/015. P.A.M. association, in a reverse direction. was supported by an Australian Postgraduate Award. The hTTATC was also associated with significant differ- study was supported by a grant-in-aid from the National ences in plasma IL-18 concentrations in CUPID. However, Health and Medical Research Council (211980). C.C. and concentrations associated with this haplotype in CUDAS J.P.B. were supported by HeartSearch (Perth, Western did not replicate the pattern seen in CUPID. This may Australia). suggest that this genetic association, which seems distinct Financial disclosures: None declared. from that associated with hGTATA, appears only when the IL-18 system is “stressed”, by either advanced dis- References ease or other parameters that differ in CUPID compared 1. Nakanishi K, Yoshimoto T, Tsutsui H, Okamura H. Interleukin-18 with CUDAS (e.g., smoking, medication, or metabolic regulates both Th1 and Th2 responses. Annu Rev Immunol syndrome). 2001;19:423–74. The lack of any strong association between IL18 genetic 2. Okamura H, Tsutsi H, Komatsu T, Yutsudo M, Hakura A, Tanimoto T, et al. Cloning of a new cytokine that induces IFN-gamma variation and presence of carotid plaque or carotid IMT, production by T cells. Nature 1995;378:88–91. despite IL-18 concentrations being associated with carotid 3. Gupta S, Pablo AM, Jiang X, Wang N, Tall AR, Schindler C. IMT in previous studies (37) but not those studied here IFN-gamma potentiates atherosclerosis in ApoE knock-out mice. (34), is suggestive of several explanations. First, assuming J Clin Invest 1997;99:2752–61. IL-18 to be causal in CHD, it may be that environmental/ 4. Mallat Z, Corbaz A, Scoazec A, Besnard S, Leseche G, Chvatchko disease-associated variables are playing a far greater role Y, et al. Expression of interleukin-18 in human atherosclerotic in determining the influence of IL-18 on disease than the plaques and relation to plaque instability. Circulation 2001;104: genetic variants studied here (this seems especially likely 1598–603. because the proportion of variance in IL-18 concentrations 5. de Nooijer R, der Thusen JH, Verkleij CJ, Kuiper J, Jukema JW, van explained by genetic variation in IL18 is low), or that IL-18 der Wall EE, et al. Overexpression of IL-18 decreases intimal production at the site of disease is not regulated in the collagen content and promotes a vulnerable plaque phenotype in apolipoprotein-E-deficient mice. Arterioscler Thromb Vasc Biol same manner as systemic IL-18 production. Presence of 2004;24:2313–9. carotid plaque was seen to be associated only with 6. Mallat Z, Corbaz A, Scoazec A, Graber P, Alouani S, Esposito B, et hTTATC. Because hTTATC was associated with some of al. Interleukin-18/interleukin-18 binding protein signaling modu- the highest IL-18 concentrations observed, this association lates atherosclerotic lesion development and stability. Circ Res might reflect the need for higher IL-18 concentrations to 2001;89:E41–5. affect plaque progression than that produced by the 7. Tenger C, Sundborger A, Jawien J, Zhou X. IL-18 accelerates genetic variants described here. Second, causality for atherosclerosis accompanied by elevation of IFN-gamma and Clinical Chemistry 53, No. 12, 2007 2085

CXCL16 expression independently of T cells. Arterioscler Thromb subjects with an example from the Multiethnic Cohort Study. Hum Vasc Biol 2005;25:791–6. Hered 2003;55:27–36. 8. Whitman SC, Ravisankar P, Daugherty A. Interleukin-18 enhances 23. Thompson SR, Sanders J, Stephens JW, Miller GJ, Humphries SE. atherosclerosis in apolipoprotein E(Ϫ/Ϫ) mice through release of A common interleukin 18 haplotype is associated with higher body interferon-gamma. Circ Res 2002;90:E34–8. mass index in subjects with diabetes and coronary heart disease. 9. Blankenberg S, Luc G, Ducimetiere P, Arveiler D, Ferrieres J, Metabolism 2007;56:662–9. Amouyel P, et al. Interleukin-18 and the risk of coronary heart 24. Hung J, McQuillan BM, Chapman CM, Thompson PL, Beilby JP. disease in European men: the Prospective Epidemiological Study Elevated interleukin-18 levels are associated with the metabolic of Myocardial Infarction (PRIME). Circulation 2003;108:2453–9. syndrome independent of obesity and insulin resistance. Arterio- 10. Tiret L, Godefroy T, Lubos E, Nicaud V, Tregouet DA, Barbaux S, et scler Thromb Vasc Biol 2005;25:1268–73. al. Genetic analysis of the interleukin-18 system highlights the 25. Guo SW, Thompson EA. Performing the exact test of Hardy- role of the interleukin-18 gene in cardiovascular disease. Circula- Weinberg proportion for multiple alleles. Biometrics 1992;48: tion 2005;112:643–50. 361–72. 11. Koenig W, Khuseyinova N, Baumert J, Thorand B, Loewel H, 26. Slatkin M, Excoffier L. Testing for linkage disequilibrium in geno- Chambless L, et al. Increased concentrations of C-reactive protein typic data using the Expectation-Maximization algorithm. Heredity and IL-6 but not IL-18 are independently associated with incident 76:377–83, 1996. coronary events in middle-aged men and women. Results from the 27. Lewontin RC. The interaction of selection and linkage. I. General MONICA/KORA Augsburg Case-Cohort Study, 1984–2002. Arte- considerations; heterotic models. Genetics 1964;49:49–67. rioscler Thromb Vasc Biol 2006;26:2745–51. 28. Armitage P, Barry G. Statistical Methods in Medical Research. 12. Kawasaki D, Tsujino T, Morimoto S, Fujioka Y, Naito Y, Okumura Oxford, UK: Blackwell Scientific Publications, 1994. T, et al. Usefulness of circulating interleukin-18 concentration in 29. McCullagh P, Nelder J. Generalised Linear Models. London, UK: acute myocardial infarction as a risk factor for late restenosis Chapman and Hall, 1989. after emergency coronary angioplasty. Am J Cardiol 2003;91: 30. Excoffier L, Slatkin M. Maximum-likelihood estimation of molecu- 1258–61. lar haplotype frequencies in a diploid population. Mol Biol Evol 13. Narins CR, Lin DA, Burton PB, Jin ZG, Berk BC. Interleukin-18 and 1995;12:921–7. interleukin-18 binding protein levels before and after percutane- 31. Ihaka R, Gentleman R. R: a language for data analysis and ous coronary intervention in patients with and without recent graphics. J Comput Graph Stat 1996;5:299–314. myocardial infarction. Am J Cardiol 2004;94:1285–7. 32. Carter KW, McCaskie PA, Palmer LJ. JLIN: a java based linkage disequilibrium plotter. BMC Bioinformatics 2006;7:60. 14. Rosso R, Roth A, Herz I, Miller H, Keren G, George J. Serum levels 33. Rothman KJ. No adjustments are needed for multiple compari- of interleukin-18 in patients with stable and unstable angina sons. Epidemiology 1990;1:43–6. pectoris. Int J Cardiol 2005;98:45–8. 34. Chapman CM, McQuillan BM, Beilby JP, Thompson PL, Hung 15. Adachi O, Kawai T, Takeda K, Matsumoto M, Tsutsui H, Sakagami J. Interleukin-18 levels are not associated with subclinical carotid M, et al. Targeted disruption of the MyD88 gene results in loss of atherosclerosis in a community population. The Perth Carotid IL-1- and IL-18-mediated function. Immunity 1998;9:143–50. Ultrasound Disease Assessment Study (CUDAS). Atherosclerosis 16. Matsumoto S, Tsuji-Takayama K, Aizawa Y, Koide K, Takeuchi M, 2006;189:414–9. Ohta T, et al. Interleukin-18 activates NF-kappaB in murine T 35. Barbaux S, Poirier O, Godefroy T, Kleinert H, Blankenberg S, helper type 1 cells. Biochem Biophys Res Commun 1997;234: Cambien F, et al. Differential haplotypic expression of the inter- 454–7. leukin-18 gene. Eur J Hum Genet 2007;15:856–63. 17. Lee JK, Kim SH, Lewis EC, Azam T, Reznikov LL, Dinarello CA. 36. Frayling TM, Rafiq S, Murray A, Hurst AJ, Weedon MN, Henley W, Differences in signaling pathways by IL-1beta and IL-18. Proc Natl et al. An interleukin-18 polymorphism is associated with reduced Acad Sci U S A 2004;101:8815–20. serum concentrations and better physical functioning in older 18. Li S, Goorha S, Ballou LR, Blatteis CM. Intracerebroventricular people. J Gerontol A Biol Sci Med Sci 2007;62:73–8. interleukin-6, macrophage inflammatory protein-1 beta and IL-18: 37. Yamagami H, Kitagawa K, Hoshi T, Furukado S, Hougaku H, Nagai pyrogenic and PGE(2)-mediated? Brain Res 2003;992:76–84. Y, et al. Associations of serum IL-18 levels with carotid intima- 19. Netea MG, Kullberg BJ, van der Meer JW. Circulating cytokines as media thickness. Arterioscler Thromb Vasc Biol 2005;25:1458– mediators of fever. Clin Infect Dis 2000;31(Suppl 5):S178–84. 62. 20. McQuillan BM, Beilby JP, Nidorf M, Thompson PL, Hung J. Hyper- 38. Lonnemann G, Novick D, Rubinstein M, Dinarello CA. Interleukin- homocysteinemia but not the C677T mutation of methylenetetra- 18, interleukin-18 binding protein and impaired production of hydrofolate reductase is an independent risk determinant of interferon-gamma in chronic renal failure. Clin Nephrol 2003;60: carotid wall thickening. The Perth Carotid Ultrasound Disease 327–34. Assessment Study (CUDAS). Circulation 1999;99:2383–8. 39. Neumayr G, Ludwiczek O, Hoertnagl H, Pfister R, Mitterbauer G, 21. McCaskie PA, Cadby G, Hung J, McQuillan BM, Chapman CM, Moschen A, et al. The impact of prolonged strenuous endurance Carter KW, et al. The C-480T hepatic lipase polymorphism is exercise on interleukin 18 and interleukin 18 binding protein in associated with HDL-C but not with risk of coronary heart disease. recreational cyclists. Int J Sports Med 2005;26:836–40. Clin Genet 2006;70:114–21. 40. Novick D, Schwartsburd B, Pinkus R, Suissa D, Belzer I, Sthoeger 22. Stram DO, Haiman CA, Hirschhorn JN, Altshuler D, Kolonel LN, Z, et al. A novel IL-18BP ELISA shows elevated serum IL-18BP in Henderson BE, et al. Choosing haplotype-tagging SNPS based on sepsis and extensive decrease of free IL-18. Cytokine 2001;14: unphased genotype data using a preliminary sample of unrelated 334–42. Clinical Chemistry 53:12 Evidence-Based 2086–2096 (2007) Laboratory Medicine and Test Utilization

NACB WRITING GROUP MEMBERS Alan H.B. Wu,1 Allan S. Jaffe,2 Fred S. Apple,3 Robert L. Jesse,4 Gary L. Francis,5 David A. Morrow,6 L. Kristin Newby,7 Jan Ravkilde,8 W.H. Wilson Tang,5 and Robert H. Christenson9*

NACB COMMITTEE MEMBERS Robert H. Christenson, Chair; Fred S. Apple; Christopher P. Cannon6; Gary L. Francis; Robert L. Jesse; David A. Morrow; L. Kristin Newby; Jan Ravkilde; Alan B. Storrow; W.H. Wilson Tang; and Alan H.B. Wu, Section Leader

I. OVERVIEW OF OTHER ETIOLOGIES...... 5 1. Biomarkers in chronic renal failure...... 8 A. Background and Scope...... 5 III. USE OF BIOMARKERS IN THE EVALUATION OF II. USE OF CARDIAC BIOMARKERS IN THE EVAL- OTHER NONISCHEMIC ETIOLOGIES ...... 11 UATION OF PATIENTS WITH CHRONIC RENAL A. Utilization of Cardiac Biochemical Marker in the FAILURE ...... 7 Setting of Nonischemic Etiologies...... 11 A. Utilization of cardiac biochemical marker in the 1. Cardiac troponin and BNP/NTproBNP in other setting of chronic renal failure ...... 7 nonischemic etiologies ...... 12 2. Cardiac troponin and BNP/NTproBNP in pulmonary embolism ...... 12 1 University of California at San Francisco, San Francisco, CA. 3. Cardiac troponin in critical care patients ...... 14 2 Mayo Clinic, Rochester, MN. IV. USE OF BIOMARKERS AFTER NONCARDIAC 3 Hennepin County Medical Center, Minneapolis, MN. 4 Medical College of Virginia, Richmond, VA. SURGERY ...... 14 5 Cleveland Clinic, Cleveland, OH. A. Use of Cardiac Biochemical Markers after Non- 6 Harvard University, Boston, MA. cardiac Surgery ...... 14 7 Duke University Medical Center, Durham, NC. 1. Biomarkers after noncardiac surgery...... 15 8 Aarhus University Hospital (Aalborg), Aalborg, Denmark. 9 University of Maryland School of Medicine, Baltimore, MD. V. BIOMARKER USE AFTER PERCUTANEOUS COR- * Address correspondence to this author at: University of Maryland School ONARY INTERVENTION (PCI)...... 16 of Medicine, Laboratories of Pathology, University of Maryland Medical A. Use of Cardiac Biochemical Markers after Center, 22 South Greene St., Baltimore, MD 21201. Fax 410-328-5880; e-mail [email protected]. PCI ...... 16 The materials in this publication represent the opinions of the authors and 1. Biomarkers after PCI...... 17 committee members, and do not represent the official position of the NACB. VI. USE OF BIOMARKERS AFTER CARDIAC SUR- The NACB is the academy of the American Association for Clinical Chemistry. GERY ...... 18 Received July 29, 2007; accepted September 14, 2007. Previously published online at DOI: 10.1373/clinchem.2007.095679 A. Utilization of Cardiac Biochemical Markers after © 2007 American Association for Clinical Chemistry Cardiac Surgery ...... 18

2086 Clinical Chemistry 53, No. 12, 2007 2087

1. Biomarkers after cardiac surgery...... 19 aacc.org/AACC/members/nacb/LMPG/OnlineGuide/ VII. REFERENCES ...... 21 PublishedGuidelines/ACSHeart/heartpdf.htm. The writing group of this document represents a mul- Preamble tidisciplinary subcommittee consisting of laboratorians and specialist clinicians from the NACB guidelines com- Over the past decade, cardiac troponin (cTn)10 has become mittee with the addition of 1 ad hoc member (A.S.J.). The the cornerstone laboratory medicine measurement for process for developing the recommendations was collab- assessment of myocardial infarction (MI) in suspected orative and iterative with input from all members of the acute coronary syndrome (ACS) patients. In the past 5–7 writing group; the effort and process was chaired by years, methods for measuring the natriuretic peptide R.H.C.; write-up of the document was spearheaded by the B-type natriuretic peptide (BNP) and its inert cometabo- section leader A.H.B.W. lite N-terminal proBNP (NT-proBNP) have become avail- The strength of scientific data supporting each recom- able, and much knowledge has accumulated regarding mendation is characterized using the scoring criteria their clinical use in the context of heart failure and adopted from the American Heart Association/American hemodynamic stress. In addition to ACS and heart failure, College of Cardiology, as summarized in Table 1. For each there are common and clinically important patient cohorts recommendation, the designations I, IIa, IIb, and III in whom these measurements can aid in diagnosis and describe the indications, and the uppercase letters A management. For this reason, the National Academy of through C describe the weight of evidence. Clinical Biochemistry (NACB) formed a Laboratory Med- The target groups for this guideline are both laboratory icine Practice Guidelines (LMPG) committee to extend professionals and clinicians. These guidelines were devel- cardiac biochemical marker recommendations and estab- oped with best available evidence and incorporated sub- lish modern guidelines for utilization in etiologies other stantial input from acknowledged experts and profes- than ACS and heart failure. During development, up- sional organizations. As such, they represent the current dated draft revisions of the guidelines were prepared and best practice for utilization of biochemical cardiac mark- placed for comment on the NACB World Wide Web site ers in etiologies other than ACS and heart failure. (http://www.aacc.org/AACC/members/nacb/LMPG/ OnlineGuide/DraftGuidelines/BioHearFailure/) begin- I. Overview of Other Etiologies ning in August 2004. The draft LMPG and suggested a. background and scope revisions were presented for public and stakeholder com- Currently available cardiac biomarkers can define cell ment at the October 2004 Arnold O. Beckman Conference death, damage, or dysfunction to particular organ sys- titled Cardiac Markers: Establishing Guidelines and Improving Results. The resulting draft guidelines can be viewed in their entirety at on the NACB World Wide Web site Table 1. American College of Cardiology/American Heart (http://www.aacc.org/AACC/members/nacb/LMPG/ Association classifications. OnlineGuide/PublishedGuidelines/ACSHeart/heartpdf. Summary of htm). indications The scope of the guidelines presented here represent I Conditions for which there is evidence and/or general agreement that a given procedure or the writing group’s recommendations for cTn, BNP, and treatment is useful and effective. NT-proBNP in the rapidly evolving area of etiologies II Conditions for which there is conflicting evidence other than ACS and heart failure in adult patients. Other and/or a divergence of opinion about the sections of the biochemical cardiac marker guidelines usefulness/efficacy of a procedure or treatment. involving clinical and analytical issues of ACS and heart IIa Weight of evidence/opinion is in favor of usefulness/efficacy. failure and issues of logistics and point-of-care testing are IIb Usefulness/efficacy is less well established by available at http://www.aacc.org/AACC/members/nacb/ evidence/opinion. LMPG/OnlineGuide/PublishedGuidelines. More explicit III Conditions for which there is evidence and/or detail on guideline development and the groups provid- general agreement that the procedure/treatment ing stakeholder input is available in the Preamble to the is not useful/effective and in some cases may be harmful. overall guideline, which is available at http://www. Weight of evidence A Data derived from multiple randomized clinical trials or properly conducted cohort studies that 10 Nonstandard abbreviations: cTn, cardiac troponin; MI, myocardial in- involved large numbers of patients. farction; ACS, acute coronary syndrome; BNP, B-type natriuretic peptide; B Data derived from a limited number of randomized NT-proBNP, N-terminal proBNP; NACB, National Academy of Clinical Bio- trials that involved small numbers of patients or chemistry; LMPG, Laboratory Medicine Practice Guidelines; cTnT, cardiac from careful analyses of nonrandomized studies troponin T; cTnI, cardiac troponin I; AMI, acute MI; ESRD, end-stage renal or observational registries. disease; ED, emergency department; PE, pulmonary embolism; NRMI, Na- C Expert consensus as the primary basis for the tional Registry of Myocardial Infarction; CK, creatine kinase; OR, odds ratio; recommendation. PCI, percutaneous coronary intervention. 2088 Wu et al.: Use of Troponin and Natriuretic Peptides in Other Etiologies

tems but cannot define the mechanisms of the effects seen. Table 3. Increased concentrations of BNP/NT-proBNP Cardiologists, emergency medicine physicians, and clini- without overt heart failure (references). cal laboratorians have often acted as if cardiac biomarkers ● Inflammatory cardiac diseases (3–5) such as cardiac troponins T (cTnT) and I (cTnI), BNP, and ● Systemic arterial hypertension with left ventricular hypertrophy NT-proBNP are specific only to the presence of ACS and (6–8) heart failure, and imply diagnosis of ACS and heart ● Pulmonary hypertension (9–11) failure, and their etiology. Although it is true that release ● Acute or chronic renal failure (12, 13) of cTn into blood is the result of myocardial damage, it is ● Ascitic liver cirrhosis (14–16) not necessarily damage related to a coronary artery ab- ● Endocrine disorders normality, or even the result of acute myocardial isch- Hyperaldosteronism (17, 18) emia. Therefore the diagnosis of acute MI (AMI) must Adrenal tumors (19) always be framed in the correct clinical context; this Hyperthyroidism (20, 22) includes the caveat that the combination of ischemia plus necrosis does not in and of itself imply a coronary etiology. This is particularly more important when lower In a similar manner, increases in BNP and NT-proBNP cutoff values are used, such as the 99th percentile limit of concentrations are not specific for heart failure alone. a healthy population, which detect a variety of more Table 3 lists conditions other than heart failure than can subtle abnormalities (1). Table 2 shows a list of conditions cause an increase in BNP and NT-proBNP (3–22). Renal that can cause an increase in cTn in the absence of overt dysfunction is a confounder for cTnT and cTnI as well as ischemic heart disease (2). BNP and NT-proBNP. Cardiac surgery will release cardiac biomarkers into blood because of the damage to the Table 2. Elevations of cardiac troponins without overt heart during the procedure itself. Although increased ischemic heart disease.a concentrations of cTn and natriuretic peptides are not ● Trauma (including contusion, ablation, pacing, implantable specific to ischemic damage or heart failure, a detectable cardioverter defibrillator firings including atrial defibrillators, increase in these cardiac biomarkers has been associated cardioversion, endomyocardial biopsy, cardiac surgery, after with worse prognosis regardless of the etiology of the interventional closure of atrial septal defects) increase. ● Congestive heart failure—acute and chronic Finally, it must also be clinically recognized that false- ● Aortic valve disease and hypertrophic obstructive cardiomyopathy with significant left ventricular hypertrophy positive increases in cTn, BNP, and NT-proBNP can ● Hypertension occur, although very infrequently, as a result of analysis ● Hypotension, often with arrhythmias errors. Although the incidence of assay interferences ● Postoperative noncardiac surgery patients who seem to do well caused by atypical antibodies has been reduced, all im- ● Renal failure munoassays have the potential for both false-positive and ● Critically ill patients, especially with diabetes, respiratory failure, false-negative interference (23, 24). gastrointestinal bleeding, sepsis The recommendations in this section will focus on ● Drug toxicity, e.g., adriamycin, 5-fluorouracil, herceptin, snake venoms, carbon monoxide poisoning other etiologies that can affect the interpretation of cTn ● Hypothyroidism and the natriuretic peptides. Recommendations will be ● Abnormalities in coronary vasomotion, including coronary provided where there is sufficient scientific and medical vasospasm evidence. Note that an overview of ACS and heart failure ● Apical ballooning syndrome containing definitions, pathogenesis, and management ● Inflammatory diseases, e.g., myocarditis, parvovirus B19, Kawasaki disease, sarcoid, smallpox vaccination, or myocardial from the perspective of biochemical markers is presented extension of bacterial endocarditis in Chapters 1 and 3. Chapters 1 and 3 also include specific ● Post-PCI patients who appear not to have complications recommendations for the use of biochemical markers for ● Pulmonary embolismPE, severe pulmonary hypertension the diagnosis and risk stratification of patients and clinical ● Sepsis decisions in the context of ACS and heart failure. Recom- ● Burns, especially if total surface burn area is Ͼ30% mendations addressing analytical issues for cardiac bi- ● Infiltrative diseases, including amyloidosis, hemochromatosis, omarkers of ACS and heart failure are presented in sarcoidosis, scleroderma ● Acute neurological disease, including cerebrovascular accident, Chapters 2 and 4, respectively. subarachnoid bleeds ● Rhabdomyolysis with cardiac injury II. Use of Cardiac Biomarkers in the Evaluation of Patients ● Transplant vasculopathy with Chronic Renal Failure ● Vital exhaustion a. utilization of cardiac biochemical marker a Table from “Troponin: the biomarker of choice for the detection of cardiac in the setting of chronic renal failure injury.” Reprinted from CMAJ 08-Nov-05;173 (10):1191–202 by permission of Recommendations for use of biochemical markers in the setting the publisher. © 2005 Canadian Medical Association. of chronic renal failure Clinical Chemistry 53, No. 12, 2007 2089

class i observed (28–30). Given the prognostic importance of the acute increases, which is similar to that seen in the 1. In renal failure patients with symptoms (e.g., acute absence of ESRD, some have suggested that the benefits of chest pain) or electrocardiograph or other clinical acute pharmacologic or invasive interventions in patients evidence suggesting myocardial ischemia, measure- with evidence of ischemia and increased cTn outweigh ment of cTn is warranted for evaluation of MI (Level the risks of bleeding and increased renal dysfunction (31). of evidence A). Acute changes in cTn are an important consideration in 2. For end-stage renal disease (ESRD) patients, as for those ESRD patients with chronic increases that are not in all patients who may have baseline elevations of and of themselves benign. These abnormal concentrations cTn, who present with possible ACS, relying on have been significant predictors of an adverse short- dynamic changes in the cTn values of 20% or more and/or long-term prognosis in nearly every available should be used to define those with AMI (Level of study. In a study of 224 individuals, increased concentra- evidence B). tions of cTnT were strongly associated with diffuse coronary artery disease and an independent predictor of death class iib (32). Virtually identical results were obtained for the 1. cTnT and cTnI can be used as aids for defining the outcome of death by cTnT concentrations at 1, 2, and 3 risk of mortality in ESRD patients and provide years of follow-up for cTnT among 733 patients; odds baseline values for comparison when measured in ratios (ORs) ranged from 2.2 to 2.5 and there was a the setting of an acute clinical change (Level of gradation of risk related to the magnitude of the increases evidence B). (33). The incidence of abnormal values in this cohort was 2. In renal failure patients, BNP or NT-proBNP testing considerably higher for cTnT than for cTnI; 82% of ESRD can be used in the acute setting to rule out or to patients had an increase in cTnT compared with only 6% confirm the diagnosis of heart failure among pa- for cTnI when the 99th percentile cutoff limit was used tients presenting with ambiguous signs and symp- (33). Furthermore, the prevalence of increased cTn con- toms. However, different decision point (cutoff) centrations varied by which assay was used for measure- values must be used than in patients with estimated ment. A substantially greater number of patients had an Ϫ glomerular filtration rate Ͼ60 mL min 1 (1.73 increased cTnT (85%) compared with cTnI by either cTnI Ϫ m2) 1 (Level of evidence B). method (19% Beckman, 5% Dade). Percentage of agreement between the Beckman and Dade assays for increased cTnI was 85% (␬ 0.32). Two-year mortality rates based on class iii an increased cTnI regardless of cTnT status was 61% for the Dade assay and 47% for the Beckman assay (34). 1. Routine BNP/NT-proBNP measurement is not war- Therefore, both markers are predictive of risk in ESRD. ranted in asymptomatic ESRD patients (Level of cTnI appears to be less useful on a routine basis, evidence B). however, because the frequency of increased values associated with increased risk of adverse events is markedly 1. biomarkers in chronic renal failure lower. Although the exact reason for this difference is Increases in the concentration of cTnT and cTnI in the unknown, it is likely related to the mechanism by which setting of ESRD indicate cardiac damage. In patients with cTns are differentially released into the circulation, de- suspected ACS, a dynamic change in the cTnI indicates a graded, and/or cleared from the circulation. There is diagnosis of AMI (25, 26) and warrants further investiga- some suggestion that the dialysis process itself may affect tion/treatment. A recommended cutoff cTn value of cTn values, although changes in cTn concentrations due Ն20% in the 6–9 h after presentation represents a signif- to the procedure are not large (35). Recently, the Food and icant (3 SD) change in cTn on the basis of a 5%–7% Drug Administration has cleared cTnT as a biomarker for analytical CV typical for most assays in the concentration risk stratification in ESRD patients for all-cause death, and range indicating AMI. Patients with ESRD who have the use of cTnT for this indication is suggested by the increased cTn concentrations have a higher (long-term) Kidney Disease Outcomes Quality Initiative (29).Inthe risk of death than corresponding patients without in- absence of myocardial ischemia, there are no specific creases in cTn in this setting. Using a cut point of 0.03 therapeutic interventions known to reduce cardiovascular ␮g/L for cTnT (10% CV), Aviles et al. (27) reported a risk that can be recommended based solely on the results 2.7-fold higher (95% CI 1.9–3.8) risk of MI or death at 30 of cTn testing in patients with ESRD. However, the days among patients with suspected ACS in the lowest availability of such baseline values would simplify the quartile for creatinine clearance. Recent data suggest that care of patients with ESRD who present with a variety of even if there is an increased baseline concentration, fur- problems for emergency department (ED) and/or hospi- ther increase above that baseline occurs in acute ischemic tal evaluation and care. Increases in the concentration of damage. Thus acute increases can be differentiated from BNP and NT-proBNP have also been observed to have more chronic elevations when a rising pattern of results is prognostic significance in patients with ESRD; however, 2090 Wu et al.: Use of Troponin and Natriuretic Peptides in Other Etiologies

recommended reference intervals for BNP and NT- cTnT or cTnI measurements are not warranted proBNP have not been validated for patients with chronic among cancer patients undergoing chemotherapies renal failure. that are toxic to the heart (except those receiving Zoccali et al. (36) reported a relative OR of 6.7 (95% CI adriamycin) (Level of evidence C). 2.4–18.5) for cardiovascular death among patients on dialysis with an increased concentration of BNP. Several 1. ctn and bnp/ntprobnp in other nonischemic investigators have combined the results of biomarkers to etiologies determine whether they provide additive risk stratifica- Years after the release of the first commercial cTn assays, tion information. In an analysis of cTnT, cTnI, and atrial there have been no basic or clinical studies that have and BNPs in chronic dialysis patients, only cTnT was an shown that cTn can be released from any tissues except independent predictor of death (37). However, Apple et the heart. Therefore, cTn found in concentrations exceed- al. (34) found that cTnT, cTnI, and high-sensitivity C-re- ing the 99th percentile of a reference population reflects active protein were each independent predictors of death recent myocardial damage. However, increases in cTnT or in ESRD patients. In his study, NT-proBNP had prognos- cTnI neither imply an ischemic etiology for the damage tic value, but it was not independent of cTn. Comparing nor are necessarily associated with an acute coronary BNP results directly against NT-proBNP in patients with event. cTn can be observed in nonischemic injuries to the chronic kidney disease, Vickery et al. (38) suggested that heart, among patients with critical illnesses (40–47), che- NT-proBNP concentrations were more affected by declin- motherapy (48–53), myocarditis (54), blunt chest trauma ing kidney function. NT-proBNP does not have receptors (55, 56), stroke (57), pulmonary embolism (PE) (58–62), and is not degraded by neutral endopeptidases but is sepsis (63–65), and other conditions (66, 67). These find- excreted in the urine (39). ings are believed to represent ongoing myocardial damage. Regardless, there are several situations in which detection of increased cTn values may be clinically help- III. Use of Biomarkers in the Evaluation of Other ful. In a metaanalysis conducted on 6 studies of blunt Nonischemic Etiologies chest trauma in which myocardial contusion was sus- a. use of cardiac biomarkers in the setting of pected (56), it was concluded that cTn was a sensitive nonischemic etiologies indicator of myocardial damage. Because myocardial con- Recommendations for use of biochemical markers in other tusion is known to cause QTc prolongation, which can be nonischemic etiologies associated with malignant life-threatening arrhythmias, monitoring such individuals for arrhythmias is a rational class iib but as yet unproven adjunct to care. Troponin can also be released in patients undergoing chemotherapy, such as 1. Increased cardiac telemetry may be warranted for with the anthracyclines, who can develop heart failure. patients who have increased cTn values after blunt Recent studies suggest that for patients with increased chest trauma (Level of evidence B). cTn, treatment with angiotensin-converting enzyme in- 2. The measurement of cTn can be used to define risk hibitors dramatically reduces the frequency of heart fail- among patients who are critically ill (Level of ure (49). evidence A). 3. Increased cTn values identify individuals at in- 2. ctn and bnp/ntprobnp in pe creased risk for the development of congestive heart Among patients with PE, data substantiate the association failure when treated with adriamycin therapy for of cTn increases with worse prognosis. La Vecchia et al. (59) demonstrated that when cTnI was Յ0.6 ␮g/L, the cancer (Level of evidence B). Ͼ ␮ 4. Increased cTn values identify individuals at in- mortality was 4.8% vs 36% for cTnI 0.6 g/L. In a study creased risk of acute pulmonary embolism (Level of of 56 consecutive patients with confirmed PE, in-hospital evidence B). mortality was 44% among patients who were cTnT positive (Ͼ0.1 ␮g/L) vs 3% among those in whom cTnT was 5. Routine BNP/NT-proBNP measurements may be Յ ␮ warranted among patients with nonischemic etiolo- 0.1 g/L (59). In addition to mortality, the use of gies such as sepsis, myocarditis, or pulmonary em- inotropic drugs, need for resuscitation, and mechanical ventilation were all significantly higher among the cTnT- bolism (Level of evidence C). positive patients. This result is likely because increases in cTn correlate with the degree of right ventricular dysfunc- class iii tion, a factor known to be associated with prognosis among patients with PE (60). Although the therapeutic 1. Release of cTn from patients with cancer undergo- implications of such increases are not clear, it is important ing cardiotoxic chemotherapies represents myocar- to assign the correct diagnosis in these patients, because dial damage, which may be associated with a worse mortality among cTn-positive PE patients is substantially prognosis (Level of evidence B). However, routine higher than that among AMI patients, with the exception Clinical Chemistry 53, No. 12, 2007 2091

of those with cardiogenic shock. In both of the studies 3. Increases of cTn postoperatively are associated with cited above, the rate of cTn-positive PE patients was adverse prognosis and should prompt clinical fol- approximately 30%, whereas only approximately 5% of low-up (Level of evidence B). AMI patients developed cardiogenic shock in the Na- tional Registry of Myocardial Infarction (NRMI)-2 and class iii NRMI-3 registries (1994–2000). It is unlikely that AMI risk adjustment models would reflect the high PE mortality 1. Routine BNP/NT-proBNP measurements are not rates; although mortality for cardiogenic shock is high, the warranted among patients undergoing noncardiac incidence of shock is low. Thus assigning the diagnosis of surgery (Level of evidence C). AMI to patients with PE would most probably produce a skewed observed:expected AMI mortality ratio that could contribute to a mortality outlier status for the facility. 1. biomarkers after noncardiac surgery It is unclear at present what to do therapeutically in Ischemic myocardial damage can occur in patients under- response to such increases and whether cTn and/or BNP going surgery that does not involve the myocardium. or NT-proBNP increases provide more information. Some Creatine kinase (CK) and CK-MB are less reliable biomar- experts have recommended consideration of fibrinolytic kers than cTn for assessing ischemic myocardial compli- therapy or invasive thrombectomy for those identified as cations because these enzymes are released from skeletal high risk on the basis of increased cTn and/or BNP or muscle damage as the result of the surgery (68). cTn is NT-proBNP among patients with submassive PE (62). specific to heart damage and is not normally released in noncardiac surgery (69). Therefore, increased cTn concen- 3. ctn in critical care patients trations after noncardiac surgery are a marker of myocar- Increases of cTn are common among critically ill patients, dial damage and are predictive of an adverse outcome at e.g., those with sepsis. Although closely related to the 6 and 12 months (70–72). Increases in cTnT above 0.03 extent of left ventricular dysfunction and the need for ␮g/L (10% CV cut point) were indicative of occult myo- inotropic support and prognosis (63, 64), the therapeutic cardial necrosis and an independent marker of mortality implications of such increases have not yet been fully (OR 14.9, 95% CI 3.7–60.3) (73). Similar results have been defined. These principles may hold for many other situ- shown for cTnI (OR 9.8, 95% CI 3.0–32) (72). Although it ations in which there are increases of cTn. However, until appears that increases of cTn provide prognostic informa- systematic studies are conducted addressing the utility of tion in many surgical settings, the etiology of increases, the cTn for diagnosis, prognosis, and treatment in these the absolute value of the increases, and whether there is nonischemic etiologies, the relationship will remain un- short- or long-term prognostic significance may vary. For clear. BNP and NT-proBNP have also been used in many example, in vascular surgery patients, increases of cTn are of these same clinical conditions (65–67), but these studies closely associated with severity and duration of ST- are less extensive at present than those with cTn. segment changes in a group of patients known to have a high incidence of underlying coronary artery disease, and IV. Use of Biomarkers after Noncardiac Surgery these increases are highly prognostic (74). Furthermore, a. use of cardiac biochemical markers after increases are associated with both early and late clinical noncardiac surgery consequences, suggesting the need for intervention Recommendations for use of cardiac markers after noncardiac acutely when increases occur in the hospital (68). It is only surgery for vascular surgery patients that present evidence suggests a role for the routine monitoring of cardiac markers. class iib Though less well studied, increases in orthopedic patients (depending on age and other characteristics) might not be 1. cTnT and cTnI are recommended for patients unrelated to ischemic heart disease but might more likely be dergoing noncardiac surgery if there is a question of associated with pulmonary emboli, which are frequent cardiac ischemia. Cutoff concentrations that are postoperative complications (75). Thus, each surgical used for diagnosis of MI are appropriate (Level of group should be evaluated individually. It is only for evidence C). vascular surgery patients that the present evidence sug- 2. cTnT and cTnI may be considered for postsurgical gests a role for monitoring of cTn. Currently, there is no assessment of patients undergoing vascular surgery evidence for the potential role of BNP/NT-proBNP in given the high frequency of underlying coronary noncardiac surgery. artery disease and associated perioperative events. Such increases appear to be due to ischemia and are V. Biomarker Use after Percutaneous Coronary Intervention highly prognostic for both short- and long-term a. use of cardiac biochemical markers after mortality. Cutoff concentrations that are used for Percutaneous Coronary Intervention diagnosis of MI are appropriate (Level of Recommendations for use of biomarkers after percutaneous evidence B). coronary intervention 2092 Wu et al.: Use of Troponin and Natriuretic Peptides in Other Etiologies

class iib perfusion grade and intracoronary myocardial contrast echocardiography (80). Although the mechanism is un- 1. It is appropriate to measure cTnT or cTnI before and known, decreased tissue perfusion in patients with high after percutaneous coronary intervention (PCI) to cTn may be responsible for the increased incidence of determine the presence of ischemic cardiac damage adverse cardiac events. if the baseline preprocedural value is less than the Recent data have challenged the concept that postpro- 99th percentile for the reference control population. cedure biomarker elevations carry prognostic signifi- Any increase is indicative of cardiac damage. There cance. When one uses baseline cTn in the analysis, the is currently insufficient evidence to recommend the prognostic significance of postprocedure values is totally specific cTn cutoff concentration (Level of obviated, suggesting that it is the preprocedure value that evidence C). defines risk. If the baseline cTn value is increased, at- tempting to distinguish PCI-induced injury from the class iii injury leading to the admission is probably not something 1. Routine BNP/NT-proBNP measurements are not that can be accomplished unless there is substantial time warranted among patients undergoing PCI (Level of (2 samples at least 6 h apart) before the PCI. Similar evidence C). principles should be applied for CK-MB, whose rise may 2. If the preprocedural baseline cTn is increased above not be detected due to its substantial lack of sensitivity the 99th percentile of a reference control population, compared with troponin. In this situation, the baseline then biochemical markers should not be used to value predicts subsequent risk (83). If the cTn values are estimate whether increases are related to the proce- stable over time, using criteria for reinfarction is sug- dure or to progression of the underlying disease gested. When the baseline cTn value is normal, increases state that caused the need for the procedure (Level in both cTn and CK-MB are modest and unassociated of evidence C). If serial preprocedural cTn values with long-term events (83) but at least in this situation, are available, a falling trend followed by a postpro- the mechanism can be more clearly ascertained. The cedural increase of 20% or more may be indicative European Society of Cardiology Task Force on Invasive of new myocardial injury, even if any or all of the Cardiology recommended a CK-MB cutoff concentration preprocedural results are above the 99th percentile. of 5 times the upper limit of normal (84). The previous convention has been the use of a 3-fold increase (85). 1. biomarkers after pci Periprocedural myocardial damage has been the subject VI. Use of Biomarkers after Cardiac Surgery of debate since the inception of the technique 30 years ago a. use of cardiac biochemical markers after (76). cTn release after PCI ranges in incidence from 14% to cardiac surgery 48% (77–82). This wide variation is caused by the assay Recommendations for use of biomarkers after cardiac surgery and corresponding cutoff concentrations used, as well as the underlying indication for the revascularization proce- class iia dure (e.g., acute vs elective) and the type of procedure 1. In addition to Ͼ5-fold increase in cTn after the performed. In the majority of these studies, cTnT or cTnI procedure, clinical and other (nonlaboratory medi- cutoff concentrations were higher than either the 99th cine) diagnostic testing criteria should be used to percentile or the 10% CV value. These and other studies distinguish components related to the operative have consistently shown that postprocedural increases in procedure and cardioprotection from vascular cTn and/or CK-MB are associated with major adverse events (Level of evidence C). clinical events. Indeed, minor increases in CK-MB are 2. Preprocedural baseline cTn increases help to define associated with an increase in 6-month mortality, a risk risk among patients undergoing cardiac surgery similar to that observed with spontaneous AMI at any (Level of evidence C). given CK-MB concentration (82). In a study of elective PCI, increased cTnI (13.6% of patients) was associated class iii with the presence of procedural side branch occlusions and thrombus formation (79). In the 481 patients with 1. At this time, there is insufficient evidence to recom- ACS and PCI enrolled in the SYMPHONY trial, 48% had mend routine measurement of BNP/NT-proBNP increased cTnI, which was associated with 90-day events before or after cardiac surgery (Level of evidence C). of MI, severe recurrent ischemia, and the combination of death or MI (78). Similar results have been reported for 1. biomarkers after cardiac surgery cTnT, in which abnormal concentrations resulted in an It has been recognized for many years that patients OR for death or MI of 2.6 (95% CI 1.4–5.1) (77). Postpro- undergoing cardiac surgery release some amount of car- cedural increases in cTnI were also associated with de- diac proteins such as CK and cTn. There is a relationship creased tissue-level perfusion as measured by angiogra- between increases of biomarkers and the details of the phy using TIMI (thrombolysis in MI) myocardial procedure itself, such as the duration of the cross clamp Clinical Chemistry 53, No. 12, 2007 2093

and cardiopulmonary bypass times, the nature of the mittee for the redefinition of myocardial infarction. J Am Coll cardioplegic solution, cold vs warm solutions, and so on Cardiol 2000;36:959–69. (86–94). Recent MRI data suggest that most of the damage 2. Babuin L, Jaffe AS. Troponin: the biomarker of choice for the detection of cardiac injury. CMAJ 2005;173:1191–202. observed after bypass surgery is subendocardial and 3. Mocelin AO, Issa VS, Bacal F, Guimaraes GV, Cunha E, Bocchi apical and likely is a result of issues related to cardiac EA. The influence of aetiology on inflammatory and neurohumoral preservation (90). Transmural damage is observed only activation in patients with severe heart failure: a prospective with very marked increases of cTn that and are potentially study comparing Chagas’ heart disease and idiopathic dilated related to a primary vascular event (88, 90). Therefore, if cardiomyopathy. Eur J Heart Fail 2005;7:869–73. the diagnosis of MI must reflect a primary vascular event, 4. Finkelstein A, Michowitz Y, Abashidze A, Miller H, Keren G, an appropriate cutoff value is difficult to define. Several George J. Temporal association between circulating proteolytic, inflammatory and neurohormonal markers in patients with coro- clinical studies have demonstrated that cTnT and cTnI nary ectasia. Atherosclerosis 2005;179:353–9. concentrations that are much greater than the AMI cutoffs 5. Karadag O, Calguneri M, Yavuz B, Atalar E, Akdogan A, Kalyoncu are associated with in-hospital and long-term morbidity U, et al. B-type natriuretic peptide (BNP) levels in female systemic and mortality (95–100); in general, the higher the value, lupus erythematosus patients: what is the clinical significance? the worse the prognosis (89). In addition, the higher the Clin Rheumatol 2007;26:1701–4. value, the greater the likelihood of transmural involve- 6. Park MH, Uber PA, Scott RL, Mehra MR. B-type natriuretic peptide ment, which some might equate to a vascular event as in heart transplantation: an important marker of allograft performance. Heart Fail Rev 2003;8:359–63. opposed to cardiac damage related to the procedure itself 7. Pruszczyk P. N-terminal pro-brain natriuretic peptide as an indi- (88). However, studies using angiography to define graft cator of right ventricular dysfunction. J Card Fail 2005;11(Suppl and/or native vessel occlusion have found that there 5):S65–9. remains substantial overlap between the values in those 8. Persu A, De Plaen JF. Recent insights in the development of with graft occlusion and those without (101). In a recent organ damage caused by hypertension. Acta Cardiol 2004;59: series that used a marked increase of cTn to define 369–81. possible graft occlusion (102), only 67 of 118 patients had 9. Burke MA, Cotts WG. Interpretation of B-type natriuretic peptide in cardiac disease and other comorbid conditions. Heart Fail Rev a primary vascular event. Thus, additional criteria over 2007;12:23–36. and above markers are needed to define a vascular event 10. Yap LB. B-type natriuretic peptide and the right heart. Heart Fail after bypass surgery. Given that fact and the need to Rev 2004;9:99–105. evaluate patients with a variety of surgical types (off 11. Yap LB, Ashrafian H, Mukerjee D, Coghlan JG, Timms PM. The pump, for example), using a low cutoff value (for example natriuretic peptides and their role in disorders of right heart a 5-fold increase) together with other criteria might be dysfunction and pulmonary hypertension. Clin Biochem 2004; advised. Nonetheless, regardless of the mechanism, the 37:847–56. 12. Safley DM, Awad A, Sullivan RA, Sandberg KR, Mourad I, higher the value, the greater the likelihood of subsequent Boulware M, et al. Changes in B-type natriuretic peptide levels in adverse clinical events. hemodialysis and the effect of depressed left ventricular func- At present, there are few data on the use of BNP and tion. Adv Chronic Kidney Dis 2005;12:117–24. NT-proBNP for risk stratification for adverse events after 13. Horl WH. Natriuretic peptides in acute and chronic kidney dis- cardiac surgery. In contrast to cTn, most studies on BNP ease and during renal replacement therapy. J Invest Med 2005; have focused on the predictive value of preoperative 53:366–70. concentrations. One study suggested that preoperative 14. Yap LB, Mukerjee D, Timms PM, Ashrafian H, Coghlan JG. Natriuretic peptides, respiratory disease, and the right heart. values of BNP predicted the need for intraaortic balloon Chest 2004;126:1330–6. pumps, prolonged hospital length of stay, and 1-year 15. Yildiz R, Yildirim B, Karincaoglu M, Harputluoglu M, Hilmioglu F. mortality (103). In another study, increase preoperative Brain natriuretic peptide and severity of disease in non-alcoholic BNP predicted postoperative atrial fibrillation (104). Such cirrhotic patients. J Gastroenterol Hepatol 2005;20:1115–20. studies address the suitability of patients undergoing 16. Henriksen JH, Gotze JP, Fuglsang S, Christensen E, Bendtsen F, cardiac surgery and not perioperative complications or Moller S. Increased circulating pro-brain natriuretic peptide events. (proBNP) and brain natriuretic peptide (BNP) in patients with cirrhosis: relation to cardiovascular dysfunction and severity of disease. Gut 2003;52:1511–7. 17. Kato J, Etoh T, Kitamura K, Eto T. Atrial and brain natriuretic Grant/funding support: None declared. peptides as markers of cardiac load and volume retention in Financial disclosures: All financial disclosures and relation- primary aldosteronism. Am J Hypertens 2005;18:354–7. ships with industry for the writing group members are 18. Lee YJ, Lin SR, Shin SJ, Lai YH, Lin YT, Tsai JH. Brain natriuretic reported online at: http://www.aacc.org/AACC/members/ peptide is synthesized in the human adrenal medulla and its nacb/LMPG/OnlineGuide/PublishedGuidelines/ACSHeart/. messenger ribonucleic acid expression along with that of atrial natriuretic peptide are enhanced in patients with primary aldosteronism. J Clin Endocrinol Metab 1994;79:1476–82. References 19. Totsune K, Takahashi K, Murakami O, Satoh F, Sone M, Ohneda 1. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial M, et al. Immunoreactive brain natriuretic peptide in human infarction redefined: a consensus document of the joint Euro- adrenal glands and adrenal tumors. Eur J Endocrinol 1996;135: pean Society of Cardiology/American College of Cardiology Com- 352–6. 2094 Wu et al.: Use of Troponin and Natriuretic Peptides in Other Etiologies

20. Wei T, Zeng C, Tian Y, Chen Q, Wang L. B-type natriuretic peptide patients with CKD: relationship to renal function and left ventric- in patients with clinical hyperthyroidism. J Endocrinol Invest ular hypertrophy. Am J Kidney Dis 2005;46:610–20. 2005;28:8–11. 39. Ng LL, Geeranavar S, Jennings SC, Loke I, O’Brien RJ. Diagnosis 21. Schultz M, Faber J, Kistorp C, Jarlov A, Pedersen F, Wiinberg N, of heart failure using urinary natriuretic peptides. Clin Sci 2004; et al. N-terminal-pro-B-type natriuretic peptide (NT-pro-BNP) in 106:129–33. different thyroid function states. Clin Endocrinol 2004;60:54–9. 40. Guest TM, Ramanathan AV, Tuteur PG, Schechtman KB, Laden- 22. Kohno M, Horio T, Yasunari K, Yokokawa K, Ikeda M, Kurihara N, son JH, Jaffe AS. Myocardial injury in critically ill patients: a et al. Stimulation of brain natriuretic peptide release from the frequently unrecognized complication. JAMA 1995;273:1945–9. heart by thyroid hormone. Metabolism 1993;42:1059–64. 41. Wright RS, Williams BA, Lewis L, Ladenson JH, Cramner H, 23. Kricka LJ. Interferences in immunoassay: still a threat. Clin Gallahue F, et al. Elevations of cardiac troponin I are associated Chem 2000;46:1037–8. with increased short-term mortality in non-cardiac critically ill 24. Eriksson S, Halenius H, Pilkki K, Hellman J, Pettersson K. emergency department patients. Am J Cardiol 2002;90:634–6. Negative interference in cardiac troponin I immunoassays by 42. Wu TT, Yuan A, Chen CY, Chen WJ, Luh KT, Kuo SH, et al. Cardiac circulating cardiac troponin autoantibodies. Clin Chem 2005;51: troponin I levels are a risk factor for mortality and multiple organ 839–47. failure in noncardiac critically ill patients and have an additive 25. McLaurin MD, Apple FS, Falahati A, Murakami MA, Miller EA, effect to the APACHE II score in outcome prediction. Shock Sharkey SW. Cardiac troponin I and creatine kinase MB mass to 2004;22:95–101. rule out myocardial injury in hospitalized patients with renal 43. Landesberg G, Vesselov Y, Einav S, Goodman S, Sprung CL, insufficiency. Am J Cardiol 1998;82:973–5. Weissman C. Myocardial ischemia, cardiac troponin, and long- 26. McCullough PA, Nowack RM, Foreback C, Tokarski G, Tomianov- term survival of high-cardiac risk critically ill intensive care unit ich MC, Khoury KR, et al. Performance of multiple cardiac patients. Crit Care Med 2005;33:1281–7. biomarkers measured in the emergency department in patients 44. King DA, Codish S, Novack V, Barski L, Almog Y. The role of with chronic kidney disease and chest pain. Acad Emerg Med cardiac troponin I as a prognosticator in critically ill medical 2002;9:1389–96. patients: a prospective observational cohort study. Crit Care 2005;9:R390–5. 27. Aviles RJ, Askari AT, Lindahl B, Wallentin L, Jia G, Ohman EM, et al. Troponin T levels in patients with acute coronary syndromes, 45. Relos RP, Hasinoff IK, Beilman GJ. Moderately increased serum with and without renal dysfunction. N Engl J Med 2002;346: troponin concentrations are associated with increased morbidity 2047–52. and mortality rates in surgical intensive care unit patients. Crit Care Med 2003;31:2598–603. 28. Le EHY, Klootwijk PJ, Weimar W, Zietse R. Significance of acute 46. Quenot JP, Le Teuff G, Quantin C, Doise JM, Abrahamowicz M, versus chronic troponin T elevation in dialysis patients. Nephron Masson D, et al. Myocardial injury in critically ill patients: relation Clin Prac 2004;98:c87–92. to increased cardiac troponin I and hospital mortality. Chest 29. Kidney Disease Outcomes Quality Initiative. Clinical practice 2005;128:2758–64. guidelines for cardiovascular disease in dialysis patients. Am J 47. Minkin R, Cotiga D, Noack S, Dobrescu A, Homel P, Shapiro JM. Kidney Dis 2005;45(Suppl 3):S1–154. Use of admission troponin in critically ill medical patients. 30. Jaffe AS. Chasing troponin: how low can you go if you can see the J Intensive Care Med 2005;20:334–8. rise? J Am Coll Cardiol 2006;48:1763–4. 48. Cardinale D, Sandri MGT, Martinoni A, Tricca A, Civelli M, 31. Scirica BM, Morrow DA. Troponins in acute coronary syndromes. Lamantgia G, et al. Left ventricular dysfunction predicted by early Prog Cardiovasc Dis 2004;47:177–88. troponin I release after high-dose chemotherapy. J Am Coll 32. deFilippi C, Wasserman S, Rosanio S, Tiblier E, Sperger H, Tocchi Cardiol 2000;36:517–22. M, et al. Cardiac troponin T and C-reactive protein for predicting 49. Cardinale D, Colombo A, Sandri MT, Lamantia G, Colombo N, prognosis, coronary atherosclerosis, and cardiomyopathy in pa- Civilli M, et al. Prevention of high-dose chemotherapy-induced tients undergoing long-term hemodialysis. JAMA 2003;290: cardiotoxicity in high-risk patients by angiotensin-converting en- 353–9. zyme inhibition. Circulation 2006;114:2474–81. 33. Apple FS, Murakami MM, Pearce LA, Herzog CA. Predictive value 50. Kilickap S, Barista I, Akgul E, Aytemir K, Aksoyek S, Aksoy S, et of cardiac troponin I and T for subsequent death in end-stage al. cTnT can be a useful marker for early detection of anthracy- renal disease. Circulation 2002;106:2941–5. cline cardiotoxicity. Ann Oncology 2005;16:798–804. 34. Apple FS, Murakami MM, Pearce LA, Herzog CA. Multi-biomarker 51. Specchia G, Buquicchio C, Pansini N, Di Serio F, Liso V, Pastore risk stratification of N-terminal pro-B-type natriuretic peptide, D, et al. Monitoring of cardiac function on the basis of serum high-sensitivity C-reactive protein, and cardiac troponin T and I in troponin I levels in patients with acute leukemia treated with end-stage renal disease for all cause death. Clin Chem 2004; anthracyclines. J Lab Clin Med 2005;145:212–20. 50:2279–85. 52. Lipshultz SE, Rifai N, Dalton VM, Levy DE, Silverman LB, Lipsitz 35. Wayand D, Baum H, Schatzle G, Scharf J, Neumeier D. Cardiac SR, et al. The effect of dexrazoxane on myocardial injury in troponin T and I in end-stage renal failure. Clin Chem 2000;46: doxorubicin-treated children with acute lymphoblastic leukemia. 1345–50. N Engl J Med 2004;351:145–53. 36. Zoccali C, Mallamaci F, Benedetto FA, Tripepi G, Parlongo S, 53. Lipshultz SE, Rifai N, Sallan SE, Lipsitz SR, Dalton V, Sacks DB, Cataliotti A, et al. Cardiac natriuretic peptides are related to left et al. Predictive value of cardiac troponin T in pediatric patients ventricular mass and function and predict mortality in dialysis at risk for myocardial injury. Circulation 1997;96:2641–8. patients. J Am Soc Nephrol 2001;12:1508–15. 54. Greaves K, Oxford JS, Price CP, Clarke GH, Crake T. The 37. Ishii J, Nomura M, Okuma T, Minagawa T, Naruse H, Mori Y, et al. prevalence of myocarditis and skeletal muscle injury during acute Risk stratification using serum concentrations of cardiac tropo- viral infection in adults: measurement of cardiac troponins I and nin T in patients with end-stage renal disease on chronic T in 152 patients with acute influenza infection. Arch Intern Med maintenance dialysis. Clin Chim Acta 2001;312:69–79. 2003;163:165–8. 38. Vickery S, Price CP, John RI, Abbas NA, Webb MC, Kempson ME, 55. Bertinchant JP, Polge A, Mohty D, Nguyen-Ngoc-Lam R, Estorc J, et al. B-type natriuretic peptide and amino-terminal proBNP in Cohendy R, et al. Evaluation of incidence, clinical significance, Clinical Chemistry 53, No. 12, 2007 2095

and prognostic value of circulating cardiac troponin I and T major noncardiac surgery in patients at risk of coronary artery elevation in hemodynamically stable patients with suspected disease. J Am Coll Cardiol 2003;42:1767–76. myocardial contusion after blunt chest trauma. J Traum Injury 73. Oscarsson A, Eintrei C, Anskar S, Engdahl O, Fagerstrom L, Infect Crit Care 2000;48:924–31. Blomqvist P, et al. Troponin T-values provide long-term prognosis 56. Jackson L, Stewart A. Use of cardiac troponin for the diagnosis of in elderly patients undergoing non-cardiac surgery. Acta Anaes- myocardial contusion after blunt chest trauma. Emerg Med J thesiol Scand 2004;48:1071–9. 2005;22:193–5. 74. Landesberg G, Jaffe AS. A small step for man, a leap for 57. di Angelantonio E, Fiorelli M, Toni D, Sacchetti ML, Lorenzano S, postoperative management. J Am Coll Cardiol 2003;42:1555–7. Falcou A, et al. Prognostic significance of admission levels of 75. Ol ED, Caprini JA, Colwell CW, Frostick SP, Haas S, Hull RD, et al. cardiac troponin I in patients with acute ischaemic stroke. Fatal vascular outcomes following major orthopedic surgery. J Neurol Neurosurg Psych 2005;76:76–81. Thromb Haemost 2005;93:860–6. 58. Giannitis E, Muller-Bardorff M, Kurowski V, Weidtmann B, Wie- 76. Herrmann J. Peri-procedural myocardial injury: 2005 update. Eur gand U, Kampmann M, et al. Independent prognostic value of Heart J 2005;26:2493–519. cardiac troponin T in patients with confirmed pulmonary emboli. 77. Nageh T, Sherwood RA, Harris BM, Thomas MR. Prognostic role Circulation 2000;102:211–7. of cardiac troponin I after percutaneous coronary intervention in 59. La Vecchia L, Ottani F, Favero L, Spadaro GL, Rubboli A, Boanno stable coronary disease. Heart 2005;91:1181–5. C, et al. Increased cardiac troponin I on admission predicts 78. Cantor WJ, Newby LK, Christenson RH, Tuttle RH, Hasselblad V, in-hospital mortality in acute pulmonary embolism. Heart 2004; Armstrong PW, et al. SYMPHONY and 2nd SYMPHONY Cardiac 90:633–7. Markers Substudy Investigators. Prognostic significance of in- 60. Scridon T, Scridon C, Skali H, Alverez A, Goldhaber SZ, Solomon creased cardiac troponin I after percutaneous coronary interven- SD. Prognostic significance of troponin elevation and right ven- tion. J Am Coll Cardiol 2002;39:1738–44. tricular enlargement in acute pulmonary embolism. Am J Cardiol 79. Ricciardi MJ, Davidson CJ, Gubernikoff G, Beohar N, Eckman LJ, 2005;96:303–5. Parker MA, et al. Troponin I elevation and cardiac events after 61. Kucher N, Printzen G, Doernhoefer T, Windecker S, Meier B, Hess percutaneous coronary intervention. Am Heart J 2003;145: OM, et al. Low pro-brain natriuretic peptide levels predict benign 522–8. clinical outcome in acute pulmonary embolism. Circulation 80. Bolognese L, Ducci K, Angioli P, Falsini G, Liistro F, Baldassarre 2003;107:1576–8. S, et al. Elevations in cardiac troponin I after percutaneous 62. Piazza G, Goldhaber SZ. Acute pulmonary embolism: part II: coronary interventions are associated with abnormal tissue-level treatment and prophylaxis. Circulation 2006;114:e42–7. perfusion in high-risk patients with non-ST segment elevation 63. Maeder M, Fehr T, Rickli H, Ammann P. Sepsis-associated acute coronary syndromes. Circulation 2004;110:1592–7. myocardial dysfunction: diagnostic and prognostic impact of 81. Kizer JR, Muttrej MR, Matthai WH, McConnell J, Nardone H, cardiac troponins and natriuretic peptides. Chest 2006;129: Sonel AF, et al. Role of cardiac troponin T in the long-term risk 1349–66. stratification of patients undergoing percutaneous coronary in- 64. Spies C, Haude V, Fitzner R, Schroder K, Overbeck M, Runkel N, tervention. Eur Heart J 2003;24:1314–22. et al. Serum cardiac troponin T as a prognostic marker in early 82. Akkerhuis KM, Alexander JH, Tardiff BE, Boersma E, Harrington sepsis. Chest 1998;113:1055–63. RA, Lincoll AM, et al. Minor myocardial damage and prognosis: 65. Ammann P, Fehr T, Minder EI, Gunter C, Bertel O. Elevation of Are spontaneous and percutaneous coronary intervention-related troponin I in sepsis and septic shock. Int Care Med 2001;27: events different? Circulation 2002;105:554–6. 965–9. 83. Miller WL, Garratt KN, Burrit MF, Lennon RJ, Reeder GS, Jaffe AS. 66. Castillo, JR, Zagler A, Carrillo-Jimenez R, Hennekens CH. Brain Baseline troponin level: key to understanding the importance of natriuretic peptide: a potential marker for mortality in septic post-PCI cardiac troponin elevations. Eur Heart J 2006;27: shock. Int J Infect Dis 2004;8:271–4. 1061–9. 67. Maron BJ, Link MS, Wang PJ, Estes NA III. Clinical profile of 84. Silber S, Albertsson P, Aviles FF, Camici P, Colombo A, Hamm C, commotion cordis: an under appreciated cause of sudden death et al., for the Task Force for Percutaneous Coronary Interventions in the young during sports and other activities. J Cardiovasc of the European Society of Cardiology. Guidelines for percutane- Electrophysiol 1999;10:114–20. ous coronary interventions. Eur Heart J 2005;226:804–47. 68. Landesberg G, Shatz V, Akopnik I, Wolf YG, Mayer M, Berlatzky Y, 85. Califf RM, Abdelmeguid AE, Kuntz RE, Popma JJ, Davidson CJ, et al. Association of cardiac troponin, CK-MB, and postoperative Cohen EA, et al. Myonecrosis after revascularization procedures. myocardial ischemia with long-term survival after major vascular J Am Coll Cardiol 1998;31:241–51. surgery. J Am Coll Cardiol 2003;42:1547–54. 86. Thielmann M, Massoudy P, Schmermund A, Neuhauser M, 69. Akhtar S, Silverman DG. Assessment and management of pa- Marggraf G, Kamler M, et al. Diagnostic discrimination between tients with ischemic heart disease. Crit Care Med 2004;32(Sup- graft-related and non-graft-related perioperative myocardial in- pl):S126–36. farction with cardiac troponin I after coronary artery bypass 70. Bursi F, Babuin L, Barbieri A, Politi L, Zennaro M, Grimaldi T, et surgery. Eur Heart J 2005;26:2440–7. al. Vascular surgery patients: perioperative and long-term risk 87. Onorati F, De Feo M, Mastroroberto P, Cristodoro L, Pezzo F, according to the ACC/AHA guidelines, the additive role of post- Renzulli A, et al. Determinants and prognosis of myocardial operative cardiac troponin elevation 2005;26:2448–56. damage after coronary artery bypass grafting. Ann Thorac Surg 71. Lopez-Jimenez F, Goldman L, Sacks DB, Thomas EJ, Johnson PA, 2005;79:837–45. Cook EF, et al. Prognostic value of cardiac troponin T after 88. Selvanayagam JB, Pigott D, Balacumaraswami L, Petersen SE, noncardiac surgery: 6-month follow-up data. J Am Coll Cardiol Neubauer S, Taggart DP. Relationship of irreversible myocardial 1997;29:1241–5. injury to troponin I and creatine kinase-MB elevation after 72. Filipovic M, Jeger R, Probst C, Girard T, Pfisterer M, Gurke L, et coronary artery bypass surgery: insights from cardiovascular al. Heart rate variability and cardiac troponin I are incremental magnetic resonance imaging. J Am Coll Cardiol 2005;45:629– and independent predictors of one-year all-cause mortality after 31. 2096 Wu et al.: Use of Troponin and Natriuretic Peptides in Other Etiologies

89. Croal BL, Hillis GS, Gibson PH, Fazal MT, El-Shafei H, Gibson G, morbidity and mortality after elective open heart surgery. Clin et al. Relationship between postoperative cardiac troponin I Chem 2004;50:1560–7. levels and outcome of cardiac surgery. Circulation 2006;114: 98. Baggish AL, MacGillivray TE, Hoffman W, Newell JB, Lewand- 1468–75. rowski KB, Lee-Lewandrowski E, et al. Postoperative troponin-T 90. Steuer J, Bjerner T, Duvernoy O, Jideus L, Johansson L, Ahlstrom predicts prolonged intensive care unit length of stay following H, et al. Visualization and quantification of peri-operative myo- cardiac surgery. Crit Care Med 2004;32:1866–71. cardial infarction after coronary artery bypass surgery with con- 99. Fellahi JL, Gue X, Richomme X, Monier E, Guillou L, Riou B. Short- trast-enhanced magnetic resonance imaging. Eur Heart J 2004; and long-term prognostic value of postoperative cardiac troponin 25:1293–9. I concentrations in patients undergoing coronary artery bypass 91. Thielmann M, Massoudy P, Neuhauser M, Tsagakis K, Marggraf grafting. Anesthesiol 2003;99:270.4. G, Kamler M, et al. Prognostic value of preoperative cardiac troponin I in patients undergoing emergency coronary artery 100. Lasocki S, Provenchere S, Benessiano J, Vicaut E, Lecharny JB, bypass surgery with non-ST-elevation or ST-elevation acute coro- Desmonts JM, et al. Cardiac troponin I is an independent nary syndromes. Circulation 2006(Suppl 1);114:I448–53. predictor of in-hospital death after adult cardiac surgery. Anes- 92. Greaves SC, Rutherford JD, Aranki SF, Cohn LH, Couper GS, thesiol 2002;97:405–11. Adams DH, et al. Current incidence and determinants of periop- 101. Holmvang L, Jurlander B, Rasmussen C, Thiis JJ, Grande P, erative myocardial infarction in coronary artery surgery. Am Clemmensen P. Use of biochemical markers of infarction for Heart J 1996;132:572–8. diagnosing perioperative myocardial infarction and early graft 93. Harff GA, Jeurissen RWM, Dijkstra JB, Reitjens TJM, Schonberger occlusion after coronary artery bypass surgery. Chest 2002;121: JPAM. Differentiation between transmural perioperative myocar- 103–11. dial infarction and subendocardial injury after coronary artery 102. Thielmann M, Massoudy P, Jaeger BR, Neuhauser M, Marggraf bypass grafting using biochemical tests, elaborated by cluster G, Sack S, et al. Emergency re-revascularization with percutane- and discriminant analysis. Clin Chim Acta 1998;274:29–41. ous coronary intervention, reoperation, or conservative treat- 94. Murphy CO, Pan-Chih, Gott JP, Guyton RA. Microvascular reactiv- ment in patients with acute perioperative graft failure following ity after crystalloid, cold blood, and warm blood cardioplegic coronary artery bypass surgery. Eur J Cardiothorac Surg 2006; arrest. Ann Thoracic Surg 1005;60:1021–7. 30:117–24. 95. Januzzi JL, Lewandrowski K, MacGillivray TE, Newell JB, Kathire- san S, Servoss SJ, et al. A comparison of cardiac troponin T and 103. Hutfless R, Kazanegra R, Madani M, Bhalla MA, Tulua-Tata A, creatine kinase-MB for patient evaluation after cardiac surgery. Chen A, et al. Utility of B-type natriuretic peptide in predicting J Am Coll Cardiol 2002;39:1518–23. postoperative complications and outcomes in patients undergo- 96. Greenson N, Macoviak J, Krishnaswamy P, Morrisey R, James C, ing heart surgery. J Am Coll Cardiol 2004;43:1873–9. Clopton P, et al. Usefulness of cardiac troponin I in patients 104. Wazni OM, Martin DO, Marrouche NF, Latif AA, Ziada K, undergoing open heart surgery. Am Heart J 2001;141:447–55. Shaaraoui M, et al. Plasma B-type natriuretic peptide levels 97. Lehrke S, Steen H, Sievers HH, Peters H, Opitz A, Muller-Bardorff predict postoperative atrial fibrillation in patients undergoing M, et al. Cardiac troponin T for prediction of short- and long-term cardiac surgery. Circulation 2004;110:124–7. Clinical Chemistry 53:12 Evidence-Based 2097–2104 (2007) Laboratory Medicine and Test Utilization

Renal Dysfunction Is a Confounder for Plasma Natriuretic Peptides in Detecting Heart Dysfunction in Uremic and Idiopathic Dilated Cardiomyopathies

Marta Codognotto,1 Antonio Piccoli,1* Martina Zaninotto,2 Monica Mion,2 Mario Plebani,2 Ugo Vertolli,1 Francesco Tona,3 Luisa Ruzza,3 Agata Barchita,3 and Giovanni M. Boffa3

Background: The diagnostic value of natriuretic pep- HD), diastolic pattern, left atrial volume, and body mass tides in uremic cardiomyopathy has not been defined, index. nor has the effect of a hemodialysis (HD) session on Conclusions: Renal dysfunction was a confounder for peptides. natriuretic peptides, which were present in higher con- Methods: We performed an observational study of 100 centrations in the uremic patients with milder cardiac white adult outpatients in New York Heart Association dysfunction than in those with idiopathic DCM without class I–II, with neither diabetes nor ischemic heart renal dysfunction. Left diastolic function pattern and disease, 50 of whom had idiopathic dilated cardiomyop- atrial volume were cardiac determinants of peptide athy (DCM) and 50 of whom had uremic cardiomyopa- concentrations in DCM and HD. thy and were undergoing HD. We measured plasma © 2007 American Association for Clinical Chemistry N-terminal proB-type natriuretic peptide (NT-proBNP), BNP, and atrial natriuretic peptide (ANP) both before The cardiac natriuretic peptides of clinical relevance in- and after a dialysis session. Doppler echocardiograms clude atrial natriuretic peptide (ANP),4 B-type natriuretic were evaluated. We performed multiple regression anal- peptide (BNP), and the N-terminal fragment of proBNP ysis on the logarithm of peptide concentrations using (NT-proBNP). Circulating natriuretic peptide concentra- clinical, laboratory, and echocardio-Doppler data as ex- tions are greatly increased in patients with either acute or planatory variables. chronic heart failure, acute myocardial infarction, and Results: Mean peptide concentrations were higher in renal failure (1, 2). Several articles have reported in- the HD group, with an HD:DCM ratio of 25 for NT- creased peptide concentrations in hemodialysis (HD) pa- proBNP and 5 for BNP and ANP. Peptides were corre- tients, possibly due to cardiac dysfunction, fluid overload, > lated with each other (r 0.85). After HD, NT-proBNP and decreased renal clearance of peptides (2, 3). The significantly increased by 14%, BNP decreased by 17%, function and structure of atrial and ventricular chambers and ANP decreased by 56%. Predialysis concentrations in uremic cardiomyopathy are sensitive to volume and correlated with postdialysis values (r > 0.85). A multiple pressure overload through a remodeling process charac- regression equation significantly fitted the observed terized by an increase in myocardial fibrosis (4–7). Both peptide concentrations, both pre- and postdialysis, us- myocyte disarray and fibrosis in uremic patients are ing the same set of 4 variables: disease group (DCM or

Divisions of 1 Nephrology, 2 Laboratory Medicine, and 3 Cardiology, Uni- 4 Nonstandard abbreviations: ANP, A-type natriuretic peptide; BNP, B- versity Hospital, University of Padova, Italy. type natriuretic peptide; NT-proBNP, N-terminal fragment of proBNP; HD, * Address correspondence to this author at: Division of Nephrology, hemodialysis; DCM, dilated cardiomyopathy; GFR, glomerular filtration rate; University Hospital, University of Padova, Policlinico IV piano, Via Giustini- BMI, body mass index; NYHA, New York Heart Association; CRP, C-reactive ani, 2, I-35128 Padova, Italy. Fax 39-049 618157; e-mail [email protected]. protein; Ln, natural logarithmic; N-AR, normal or abnormal relaxation dia- Received March 30, 2007; accepted September 18, 2007. stolic pattern; Pn-R, pseudonormal or restrictive diastolic pattern; NS, not Previously published online at DOI: 10.1373/clinchem.2007.089656 significant.

2097 2098 Codognotto et al.: Natriuretic Peptides in Uremic and Dilated Cardiomyopathies

similar to those observed in patients with idiopathic frequency, BIA-101 analyzer, RJL Systems) (11), and rou- dilated cardiomyopathy (DCM) (8, 9). tine clinical laboratory tests (plasma Na, K, Ca, urea, In this study, we compared uremic HD patients with hemoglobin). An alkaline picrate method was used for idiopathic DCM patients who showed more severe heart creatinine measurement (calibration traceable to isotope involvement and normal or slightly decreased renal func- dilution mass spectrometry, Roche Diagnostics). C-reaction. We hypothesized that if natriuretic peptide concen- tive protein (CRP) was measured in lithium-heparin trations depended on the severity of heart failure regard- plasma using a high-sensitivity assay (Cardiophase less of renal function, higher peptide concentrations hsCRP), with total CVs ranging from 2.1% to 5.7% at 0.5 to should be found in the DCM group; moreover, we hy- 56.0 mg/L. NT-proBNP was assayed on lithium-heparin pothesized that DCM patients with lower vs higher plasma (PBNP, Dade Behring), with a total CV of 3.1% to glomerular filtration rates (GFRs) have similar peptide 5.7% at 159 to 3734 ng/L and decision threshold of 125 or concentrations. 450 ng/L for participants younger than or older than 75 years, respectively. BNP was assayed on plasma K3- Materials and Methods EDTA (BNP, Bayer Diagnostics) with total CVs of 2.8% to This cross-sectional observational study enrolled 100 5.4% at 42 to 1615 ng/L, and a decision threshold of 100 white, adult cardiomyopathy patients, made up of 2 ng/L. ANP was assayed on plasma K3-EDTA (Shionogi) groups of 50 individuals each: a dialysis group with with total CVs of 5.7% to 6.0% at 20 to 563 ng/L without uremic cardiomyopathy (36 men, 14 women) and a non- an established decision threshold. dialysis group with idiopathic DCM (31 men, 19 women). Doppler echocardiography was performed following In accordance with the Declaration of Helsinki, the pro- recommendations of the American Society of Echocardi- tocol was approved by the ethics committee of our ography (13). Complete M-mode, 2-dimensional, and institution, and written informed consent was obtained Doppler echocardiograms were performed with a from all of the participants. Uremic outpatients undergo- Hewlett-Packard 5500 Sonos system by use of a 2.5-MHz ing thrice-weekly, standard bicarbonate dialysis with combined imaging and Doppler transducer. The left ven- low-flux dialyzers for 6 months or longer, with uremic tricular mass was calculated with the Devereux formula cardiomyopathy (defined as an abnormality of echocar- (14) and was indexed by body surface area (g/m2). An diographic parameters with systolic or diastolic dysfunc- end-diastolic volume index Ͼ70 mL/m2 was considered a tion in the absence of other cardiomyopathies) were ventricle dilation. A left ventricular ejection fraction considered. All patients were anuric. Compensated out- Ͻ50% was defined as systolic dysfunction. Left ventricu- patients with idiopathic DCM with a current left ventric- lar diastolic function was classified into 4 patterns ular ejection fraction Ͻ50%, an end-diastolic volume (15, 16): normal, E/A Ͼ1, DT Ͻ220 ms, IRT Ͻ100 ms; index Ͼ70 mL/m2, and normal or slightly decreased renal abnormal relaxation, E/A Ͻ1 in individuals Ͻ55 years or function (chronic kidney disease class I–II) made up the E/A Ͻ0.8 in individuals Ͼ55 years, DT Ͼ220 ms, IRT control group. The DCM group was divided into 2 Ͼ100 ms; pseudonormal: 1 Ͻ E/A Ͻ 2, 150 ms Ͻ DT Ͻ subgroups with GFR values below and above the median 200 ms, IRT Ͻ100 ms; restrictive filling, E/A Ͼ2, DT Ͻ150 Ϫ Ϫ rates [85 mL ⅐ min 1 ⅐ (1.73 m2) 1]. Diagnosis of idiopathic ms, IRT Ͻ60 ms, where E is the peak E-wave velocity and DCM had been previously made according to the current A is the peak A-wave velocity of the mitral inflow, DT is WHO definition (10). Inclusion criteria for both groups the deceleration time of the E wave, and IRT is the left were age 25–80 years, body mass index (BMI) 17–35 ventricular isovolumic relaxation time (13). The index of kg/m2, New York Heart Association (NYHA) class I–II, myocardial performance (Tei index) was calculated as the normal body hydration as assessed by the bioelectrical ratio isovolumic contraction time plus isovolumic relax- impedance vector analysis (11), and normal cardiac tro- ation time)/ejection time, which reflects the global myo- ponin I (Ͻ0.15 ␮g/L, decision level for myocardial dam- cardial performance (reference value Ͻ0.4) and combines age, with a CV Ͻ10%, on RxL Dimension, Dade Behring) systolic and diastolic elements (17). (12). Exclusion criteria for both groups were atrial fibril- In DCM patients, GFR was estimated according to lation, pacemakers, previous surgical heart procedures, kidney/Dialysis Outcomes Quality Initiative recommen- valvular and congenital heart disorders, myocardial isch- dations using the simplified Modification of Diet in Renal emic events, pericardial effusion, edema, diabetes, pulmo- Disease equation (18) not applicable to anuric HD pa- nary disorders, liver disorders, alcohol abuse (Ͼ90 g/day tients. In the dialysis patients, body weight, blood pres- longer than 5 years), cancer, and treatment with cytotoxic sure, bioelectrical impedance, routine laboratory tests, drugs. and natriuretic peptides were recorded before and after a Protocol measurements were carried out on the same midweek dialysis session. day and included height, body weight, BMI (weight in Statistical calculations were performed with SPSS sta- kg/squared height in m2), systolic and diastolic blood tistical software (SPSS, version 14). After the natural pressure, mean blood pressure, heart rate, electrocardio- logarithmic (Ln) transformation of the natriuretic pep- gram, echocardio-Doppler, whole body bioelectrical im- tides and CRP values (distributions skewed to the right), pedance (resistance and reactance components at 50 kHz we obtained log-normal distributions (Kolmogorov–Smir- Clinical Chemistry 53, No. 12, 2007 2099

nov test) that were suitable for parametric statistical tests (1- and 2-way ANOVA and Student t-test). We used the Wilcoxon test to analyze the skewed distributions of differences with negative values (log transformation not allowed). We analyzed frequency distributions using Fisher exact test. The diagnostic performance of natriuretic peptides was evaluated by use of ROC curves. The relationships between natriuretic peptides and protocol variables were evaluated with the simple linear correlation coefficient r (determination coefficient r2) and multiple linear regression analysis (multiple correlation coefficient R, and multiple determination coefficient R2). Manual backward elimination was performed on blocks of variables until regression models with only significant variables were obtained. P Ͻ0.05 was considered significant in all the statistical tests.

Results Demographic and metabolic characteristics of patients (predialysis in HD patients) are shown in Table 1. The DCM group was divided into 2 subgroups with median Ϫ Ϫ GFRs of 85–126 mL ⅐ min 1 ⅐ (1.73 m2) 1 and 52–84 Ϫ Ϫ mL ⅐ min 1 ⅐ (1.73 m2) 1. Patients in both disease groups and in both GFR subgroups belonged to the same NYHA Fig. 1. Scattergram of the logarithmic distributions of NT-proBNP and class (I and II), age, and BMI range, and were free from BNP by disease group. diabetes and heart ischemic events. BNP, ANP, and NT-proBNP were highly correlated to HD patients and 24% of the 50 DCM patients [17% with Ϫ Ϫ one another within groups (0.81 Ͻ r Ͻ 0.97, P Ͻ0.001) and GFR Ͼ85 mL ⅐ min 1 ⅐ (1.73 m2) 1] had normal NT- also in the pooled groups (0.86 Ͻ r Ͻ 0.90, P Ͼ0.001). The proBNP concentrations. All the HD patients and 76% of Ϫ correlation between Ln NT-proBNP and Ln BNP is de- the DCM patients [61% with GFR Ͼ85 mL ⅐ min 1 ⅐ (1.73 Ϫ picted in the scattergram in Fig. 1. Mean peptide concen- m2) 1] had abnormal NT-proBNP concentrations. Normal trations were higher in the HD group, with an HD:DCM BNP concentrations were found in 14% of the HD patients ratio of 24.7 for NT-proBNP (1584–274 995 ng/L in HD vs and in 60% of the DCM patients [30% with GFR Ͼ85 Ϫ Ϫ 10–9162 ng/L in DCM), 4.7 for BNP (28–7625 ng/L in HD mL ⅐ min 1 ⅐ (1.73 m2) 1]. Abnormal BNP concentrations vs 2–1387 ng/L in DCM), and 4.8 for ANP (24–904 ng/L were found in 86% of the HD patients and 40% of the Ϫ in HD vs 3–413 ng/L in DCM) (Table 1). None of the 50 DCM patients [80% with GFR Ͼ85 mL ⅐ min 1 ⅐ (1.73

Table 1. Demographic and metabolic characteristics of patients by GFR in the DCM group and by disease group.a DCM DCM GFR Ͻ85 GFR Ͼ85 P DCM HD P mL ⅐ min؊1 ⅐ (1.73 m2)؊1 mL ⅐ min؊1 ⅐ (1.73 m2)؊1 Age, years 57.3 (2.7) 49.9 (2.6) NS 53.7 (1.9) 62.8 (2.2) 0.002 BMI, kg/m2 24.2 (0.7) 25.6 (0.8) NS 24.9 (0.5) 24.8 (0.4) NS Systolic blood pressure, mmHg 123.2 (2.7) 133.0 (3.3) 0.03 128.1 (2.2) 142.4 (3.1) Ͻ0.001 Diastolic blood pressure, mmHg 79.4 (1.8) 78.4 (1.9) NS 78.9 (1.3) 78.3 (1.8) NS Mean blood pressure, mmHg 93.9 (1.9) 96.7 (2.2) NS 95.3 (1.4) 99.7 (2.1) NS Urea, mmol/L 6.1 (0.2) 8.6 (0.5) Ͻ0.001 7.3 (0.3) 22.3 (0.8) Ͻ0.001 Creatinine, ␮mol/L 72.0 (2.2) 99.8 (3.2) Ͻ0.001 88.9 (2.8) 796.4 (22.3) Ͻ0.001 Hemoglobin, g/L 137.0 (2.3) 138.7 (3.1) NS 138.9 (1.9) 112.7 (1.6) Ͻ0.001 C-reactive protein, mg/L 1.2 (1.2) 1.7 (1.2) NS 1.4 (1.2) 4.8 (1.2) Ͻ0.001 Resistance/height, Ohm/m 297.9 (12.8) 295.2 (10.9) NS 296.5 (8.3) 307.0 (6.9) NS Reactance/height, Ohm/m 30.3 (1.1) 31.4 (1.1) NS 30.8 (0.8) 26.8 (0.9) 0.001 NT-proBNP, ng/L 785 (1.3) 191 (1.3) 0.001 387 (1.3) 9509 (1.2) Ͻ0.001 BNP, ng/L 135 (1.3) 40 (1.3) 0.004 74 (1.2) 348 (1.2) Ͻ0.001 ANP, ng/L 56 (1.3) 23 (1.3) 0.006 36 (1.2) 174 (1.1) Ͻ0.001 a Data are mean (SE); geometric mean of CRP, NT-proBNP, BNP, and ANP. The 95% CI of the geometric mean is the antilog of Ln(M) Ϯ 2 Ln(SE). SD ϭ 2.24 SE for all variables. Resistance and Reactance are the bioimpedance vector components. 2100 Codognotto et al.: Natriuretic Peptides in Uremic and Dilated Cardiomyopathies

Ϫ m2) 1]. Mean individual peptide concentrations by sex and NYHA class were comparable within the HD and DCM groups (2-way ANOVA: sex P ϭ 0.55–0.97, NYHA P ϭ 0.29–0.94). Most echocardiographic findings were significantly different in the disease groups (Table 2). In the HD group, the relative wall thickness and the left atrial volume index were greater than those in the DCM group. The heart was more severely affected in the DCM patients in whom the ejection fraction was decreased (ejection fraction Ͻ50% in 100% of DCM vs 10% of HD group), the diastolic volume was increased (dilation in 100% of DCM vs 36% of HD), the index of myocardial performance was worse (0.7 in DCM vs 0.4 in HD), and the left ventricular mass index was increased (172 g/m2 in DCM vs 151 g/m2 in HD, marginally significant, P ϭ 0.05). The frequency distribution of diastolic patterns was comparable in the disease Fig. 2. Means with SE of Ln(NT-proBNP), Ln(BNP), and Ln(ANP) by groups (Fisher exact test) [borderline normal with abnor- disease group and by Doppler diastolic pattern (normal or abnormal mal relaxation patterns in one category (N-AR) vs relaxation, N-AR, vs pseudonormal or restrictive, Pn-R). The influence of dialysis is indicated by a line connecting the peptide concen- pseudonormal with restrictive patterns in another cate- trations in the HD and DCM groups with the same diastolic pattern. gory (Pn-R)]. The 3 natriuretic peptides were significantly increased in patients with a Pn-R vs N-AR pattern in both disease groups, but this difference was obscured by the from 0.10 to 0.26 for the 3 peptides. The areas under the greater increase in peptide concentrations found in the ROC curves modestly increased to the values of 0.23 to HD patients compared to the DCM patients with the same 0.40 for the 3 peptides in detecting an ejection fraction diastolic pattern (Fig. 2). For instance, a BNP value of 150 Ͻ40%. In the DCM group, all patients had an ejection ng/L (5 in the Ln scale in Fig. 2) can be found in an HD fraction Ͻ50% by inclusion criteria, whereas 54% had an patient with a N-AR diastolic pattern as well as in a DCM ejection fraction Ͻ40%. The area under the ROC curve patient with a Pn-R pattern. increased to 0.64 for NT-proBNP [not significant (NS)], The influence of extreme GFR values on natriuretic 0.66 for BNP (NS), and 0.70 for ANP (P ϭ 0.02) in peptides was stronger than the degree of heart involve- detecting an ejection fraction Ͻ40%. ment. As a consequence, we obtained misleading ROC In DCM patients, the concentration of the 3 natriuretic curve patterns of natriuretic peptides in detecting systolic peptides linearly increased with decreasing GFR irrespec- dysfunction (ejection fraction Ͻ50%). The areas under the tive of the severity of heart involvement [significant ROC curve for all 100 participants were small, ranging inverse correlation, P Ͻ0.001, between GFR and Ln NT-

Table 2. Mean values of Doppler echocardiography variables by GFR in the DCM group and by disease group.a DCM DCM GFR Ͻ85 GFR Ͼ85 P DCM HD P mL ⅐ min؊1 ⅐ (1.73 m2)؊1 mL ⅐ min؊1 ⅐ (1.73 m2)؊1 Posterior wall thickness, mm 11.0 (0.3) 10.1 (0.3) NS 10.6 (0.2) 13.6 (0.5) Ͻ0.001 Interventricular wall thickness, mm 10.2 (0.3) 9.7 (0.3) NS 9.9 (0.2) 12.5 (0.4) Ͻ0.001 Ventricular diameter, mm 68.1 (2.4) 65.2 (2.2) NS 66.6 (1.6) 50.1 (1.4) Ͻ0.001 Relative wall thickness 0.3 (0.01) 0.3 (0.01) NS 0.3 (0.01) 0.5 (0.03) Ͻ0.001 End-diastolic volume, mL/m2 123.9 (9.4) 98.3 (5.3) 0.02 111.1 (5.6) 62.4 (2.9) Ͻ0.001 End-systolic volume, mL/m2 82.9 (8.3) 60.4 (4.2) 0.02 71.6 (4.9) 24.7 (1.6) Ͻ0.001 Ejection fraction, % 35.4 (2.2) 39.7 (1.3) NS 37.6 (1.3) 60.7 (1.2) Ͻ0.001 Ventricular mass, g/m2 188.3 (11.9) 154.8 (9.3) 0.03 171.6 (7.9) 150.5 (7.3) 0.05 Left atrial volume, mL/m2 41.1 (3.3) 39.1 (2.7) NS 40.1 (2.1) 54.3 (2.8) Ͻ0.001 Peak E wave/peak A wave 1.4 (0.2) 1.2 (0.1) NS 1.3 (0.1) 1.0 (0.1) 0.03 Deceleration time E wave, ms 215.4 (16.2) 227.9 (14.6) NS 221.6 (10.8) 173.3 (8.6) 0.01 Myocardial performance index 0.7 (0.1) 0.7 (0.1) NS 0.7 (0.04) 0.4 (0.02) Ͻ0.001 Pulmonary systolic wave, ms 51.1 (3.6) 55.3 (4.4) NS 52.9 (2.8) 51.7 (2.1) Ͻ0.001 Pulmonary diastolic wave, ms 47.4 (3.0) 48.5 (3.6) NS 47.9 (2.3) 50.7 (1.9) Ͻ0.001 Pulmonary diastolic/systolic 0.9 (0.1) 0.9 (0.1) NS 0.9 (0.04) 1.1 (0.1) NS a Data are mean (SE). SD can be calculated as 2.24 SE. Clinical Chemistry 53, No. 12, 2007 2101

proBNP, r ϭ –0.54 (Fig. 3A); Ln BNP, r ϭ –0.50 (Fig. 3B); influence on natriuretic peptide concentrations both be- and Ln ANP, r ϭ –0.49]. As a consequence, mean natri- fore and after a dialysis session. The prediction of NT- uretic peptide concentrations were higher in patients with proBNP concentration achieved the highest R value both Ϫ Ϫ GFR below the median [52 to 84 mL ⅐ min 1 ⅐ (1.73 m2) 1 before (R ϭ 0.85, P Ͻ0.001) and after (R ϭ 0.83, P Ͻ0.001) Ϫ compared to those above it (85 to mL ⅐ min 1 ⅐ (1.73 the dialysis session, followed by Ln BNP (R ϭ 0.65 and Ϫ m2) 1], with a ratio of 4.1 for NT-proBNP, 3.4 for BNP, 0.62, P Ͻ0.001) and Ln ANP (R ϭ 0.53 and 0.59, P Ͻ0.001). and 2.4 for ANP, although most echo-Doppler variables Standardized partial regression coefficients (␤) indicated were comparable in the 2 GFR groups (Tables 1 and 2). that uremic status was the most important variable, both Natriuretic peptides were linearly correlated with sev- before and after dialysis, followed in decreasing order by eral continuous protocol variables in both groups (0.20 Ͻ diastolic pattern, atrial volume index, and BMI. The r Ͻ 0.54, P ϭ 0.04 to P Ͻ0.001). The highest correlations following regression equations significantly fitted natri- were found between natriuretic peptides and the left uretic peptide concentration before dialysis (disease en- atrial volume index in both disease groups (r ϭ 0.54, P coded 0 if DCM, 1 if HD; D pattern encoded 0 if N-AR, 1 Ͻ0.001). Most correlation coefficients were of the same if Pn-R): order due to the high mutual correlation between peptides. Multiple regression analysis was used to identify Ln(NT-proBNP) ϭ 7.48 ϩ 3.05 Disease among the correlated variables those with an independent ϩ 1.29 D Pattern ϩ 0.02 Atrial volume

Ϫ 0.11BMI

Ln(BNP) ϭ 5.94 ϩ 1.42 Disease ϩ 1.37 D Pattern

ϩ 0.02 Atrial volume Ϫ 0.12 BMI

Ln(ANP) ϭ 4.49 ϩ 1.44 Disease ϩ 0.80 D Pattern

ϩ 0.02 Atrial volume Ϫ 0.08 BMI.

The diastolic pattern, atrial volume, and BMI significantly fitted the peptide distribution also within each disease group [Ln NT-proBNP, R ϭ 0.60 and 0.62 (P Ͻ0.001); Ln BNP, R ϭ 0.65 and 0.62 (P Ͻ0.001); Ln ANP, R ϭ 0.53 and 0.59 (P Ͻ0.001), in the DCM and HD groups, respectively]. In the DCM group, GFR did not significantly correlate with left ventricular mass and volumes (–0.31 Ͻ r Ͻ –0.20). Although mean values of left ventricular mass and diastolic and systolic volumes were significantly increased in the lower GFR group (P ϭ 0.03, Table 2), GFR was the only significant explanatory variable for natriuretic peptides in a multiple regression equation that included GFR and left ventricular mass and volumes (0.52 Ͻ R Ͻ0.57, P Ͻ0.001). After HD sessions (low-flux dialyzer), an average of 2900 mL fluid was removed by ultrafiltration, systolic and diastolic blood pressure decreased by 13 mmHg (P Ͻ0.01) and 5 mmHg (P Ͻ0.01), respectively, and heart rate increased by 2 bpm (P ϭ 0.03). The peptides reacted differently to a dialysis session. NT-proBNP increased by 14% (from 9509 to 9711 ng/L, P ϭ 0.03), BNP decreased by 17% (from 348 to 271 ng/L, P Ͻ0.01), and ANP decreased by 56% (from 174 to 70 ng/L, P Ͻ0.01). The correlation between pre- and postdialysis measurements of any peptide was very high (P Ͻ0.001, r ϭ 0.94 for NT-proBNP, 0.94 for BNP, and 0.85 for ANP), and the same set of Fig. 3. Scattergram of Ln(NT-proBNP) (A) and Ln(BNP) (B) with GFR in significant variables entered both pre- and post-HD mul- the DCM group. tiple regression equations. 2102 Codognotto et al.: Natriuretic Peptides in Uremic and Dilated Cardiomyopathies

Discussion be independent of other clinical conditions. We provide We compared the concentrations of natriuretic peptides in evidence that natriuretic peptide concentrations depend uremic patients with abolished renal function with those more on GFR than on cardiac involvement, both in the in cardiac patients with normal or slightly decreased renal extreme reduction of GFR and in the initial renal dysfunc- function to establish whether uremic status or cardiomy- tion. The confounding effect of renal dysfunction explains opathy affects plasma concentrations of the peptides in why natriuretic peptides are unable to detect systolic HD patients. Although patients in the HD and DCM dysfunction. Therefore, although the validity of decision groups belonged to the same NYHA class and both thresholds should probably be restricted to GFR Ͼ100 Ϫ Ϫ cardiomyopathies were likely characterized by a similar mL ⅐ min 1 ⅐ (1.73 m2) 1 (Fig. 3), changes in peptide con- myocyte disarray and intermyocyte fibrosis (17), the centrations over time may be meaningful in the same groups differed in their extreme values of renal function patient with stable renal function. For instance, our equa- and Doppler-echocardiographic parameters. Heart in- tion predicts that a 3.5-fold increase in the NT-proBNP volvement was more severe in the DCM patients. If heart value from 2640 to 9500 ng/L in an HD patient would function had been the determinant for plasma peptide correspond to the increase from 125 to 450 ng/L in a DCM concentrations, increased concentrations of the peptides patient with comparable diastolic pattern, atrial volume, should have been found in the DCM group owing to and BMI. Specific outcome studies are needed to test our increased wall stress (9). Peptide concentrations instead preliminary hypothesis and to establish whether our were increased 5- to 25-fold in the HD patients, indicating equations can be used for a correction procedure in that uremic status and not heart involvement was the uremic patients in a larger study group. major determinant of the peptide increase. Peptides were After an HD session, a variable decrease in circulating also increased 2- to 4-fold in the DCM patients with early peptides has been reported, especially with high-flux renal dysfunction compared to those with preserved GFR, dialyzers (2, 3). The effects of our dialysis session with reflecting the inverse linear relation with GFR. If heart low-flux dialyzers differed among the peptides. We have function had been the determinant for plasma peptide no clear explanation for the different behaviors of the concentrations in DCM patients, no difference should peptides, which can reflect different half-lives combined have been found in the patients with lower vs higher GFR with the different molecular sizes. However, the cross- values because measures of heart involvement were sim- correlation between peptides, the high correlation be- ilar in the 2 groups. tween pre- and postdialysis measurements of any pep- Increased concentrations of NT-proBNP, BNP, and tide, and the same set of significant variables in the ANP have been reported in dialysis patients compared to multiple regression equations indicate that the predialysis patients with mild and moderate renal dysfunction ranking of individual values is preserved after dialysis, (2, 17, 19). We extended the investigation to cardiac pa- meaning that long- rather than short-term regulation is tients with preserved renal function. However, the hydration status of both groups, as assessed by bioelectrical stronger. impedance vector analysis (20), was normal and Several reports support a regulatory role of left ven- comparable. tricular diastolic pattern and left atrial volume on circu- We suggest 4 explanatory variables that significantly lating peptide concentrations (5, 8, 25–29). The meaning fitted the 3 different natriuretic peptide concentrations in of the inverse relationship between natriuretic peptide the DCM and HD groups: uremic status, diastolic pattern, concentration and BMI is not clear in the literature (30– left atrial volume, and BMI. Abolished GFR had the 32). In our study, ANP, BNP, and NT-proBNP decreased 2 greatest influence on the increased peptide concentra- with the increase of BMI in the 17 to 35 kg/m range. In tions. Although higher peptide concentrations are associ- the absence of other data, we consider BMI as a scale ated with a worse prognosis, there is agreement in the factor for peptide metabolism in different body literature that current decision thresholds for BNP with- compartments. out adjustment for kidney function should not be applied Other potential explanatory variables reported in the (18, 21–24). Empirical thresholds have been suggested in literature, such as blood pressure (33, 34), C-reactive mild-to-moderate renal dysfunction, e.g., BNP 300 ng/L protein (35), ventricular wall thickness and mass (33, 36), instead of 100 ng/L for GFR of 18–60 mL/min (22) and and ejection fraction (36), although correlated with pep- 390 ng/L in HD patients (21). We found that even in tides, were not able to enter the regression equation with cardiac patients with normal GFR, the relationship be- a significant contribution in our study. The strict selection tween peptides and GFR was continuous and linear (Fig. criteria of patients in our protocol rules out confounding 3), therefore preventing the identification of any fixed factors such as ischemia, pulmonary disorders, diabetes, decision threshold. Peptide concentration expression cor- obesity, abnormal body hydration, or differences in the rected for GFR is impractical and cannot be used at all in NYHA class (2, 5–7). uremic patients with abolished GFR. A limitation of the present study is the lack of a Diagnostic testing is useful for identifying a decision follow-up period evaluating the association between pep- threshold with high sensitivity and specificity that should tide concentrations and outcomes in the 2 cardiomyopa- Clinical Chemistry 53, No. 12, 2007 2103

thies. In addition, we studied only patients in NYHA class 15. Ommen SR, Nishimura RA, Appleton CP, Miller FA, Oh JK, Redfield I and II without diabetes and ischemic heart disease. MM, et al. Clinical utility of Doppler echocardiography and tissue Doppler imaging in the estimation of left ventricular filling pressures: a comparative simultaneous Doppler-catheterization study. In conclusion, abolished renal function rather than cardio- Circulation 2000;102:1788–94. myopathy was the cause of increased peptides in HD, the 16. Nishimura RA, Tajik AJ. Evaluation of diastolic filling of left long-term regulation of which was stronger than 1 HD ventricle in health and disease: Doppler echocardiography is the session. Natriuretic peptide concentrations were depen- clinician’s Rosetta Stone. J Am Coll Cardiol 1997;30:8–18. dent on GFR even at the onset of renal dysfunction. 17. National Kidney Foundation: K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis 2002;39(Suppl 1):S1–266. Grant/funding support: Italian government, Ministry of Uni- 18. Safley DM, Awad A, Sullivan RA, Sandberg KR, Mourad I, Boulware versity and Research, “Ex 60%” fund. M, et al. Changes in B-type natriuretic peptide levels in hemodialysis and the effect of depressed left ventricular function. Adv Financial disclosures: None declared. Chronic Kidney Dis 2005;12:117–24. 19. Apple FS, Murakami MM, Pearce LA, Herzog CH. Multi-biomarker References risk stratification of N-terminal pro-B-type natriuretic peptide, 1. Clerico A, Emdin M. Diagnostic accuracy and prognostic relevance high-sensitivity C-reactive protein, and cardiac troponin T and I in of the measurement of cardiac natriuretic peptides: a review. Clin end-stage renal disease for all-cause death. Clin Chem 2004;50: Chem 2004;50:33–50. 2279–85. 2. Wahl HG, Graf S, Renz H, Fassbinder W. Elimination of the car- 20. Piccoli A, for the Italian HD-BIA study group. Identification of diac natriuretic peptides B-type natriuretic peptide (BNP) and operational clues to dry weight prescription in hemodialysis using N-terminal proBNP by hemodialysis. Clin Chem 2004;50:1071–4. bioimpedance vector analysis. Kidney Int 1998;53:1036–43. 3. Suresh M, Farrington K. Natriuretic peptides and the dialysis 21. Goto T, Takase H, Toriyama T, Sugiura T, Kurita Y, Tsuru N, et al. patient. Semin Dial 2005;18:409–19. Increased circulating levels of natriuretic peptides predict future 4. Tyralla K, Amann K. Cardiovascular changes in renal failure. Blood cardiac events in patients with chronic hemodialysis. Nephron Purif 2002;20:462–5. 2002;92:610–5. 5. Casserly LF, Givertz MM. Congestive heart failure and its manage- 22. Mueller C, Laule-Kilian K, Scholer A, Nusbaumer C, Zeller T, Staub ment in end-stage renal failure. In: Loscalzo J, London GM, eds. D, et al. B-type natriuretic peptide for acute dyspnea in patients Cardiovascular Disease in End-Stage Renal Failure. New York: with kidney disease: insights from a randomized comparison. Oxford University Press Inc., 2000:289–315. Kidney Int 2005;67:278–84. 6. Mall G, Huther W, Schneider J, Lundin P, Ritz E. Diffuse intermyo- 23. McCullough PA, Duc P, Omland T, McCord J, Nowak RM, Hollander cardiocytic fibrosis in uremic patients. Nephrol Dial Transplant JE, et al. B-type natriuretic peptide and renal function in the 1990;5:39–44. diagnosis of heart failure: an analysis from the Breathing Not 7. McKie PM, Burnett JC Jr. B-type natriuretic peptide as a biomarker Properly Multinational Study. Am J Kidney Dis 2003;41:571–9. beyond heart failure: speculations and opportunities. Mayo Clin 24. Mark PB, Stewart GA, Gansevoort RT, Petrie CJ, McDonagh TA, Proc 2005;80:1029–36. Dargie HJ, et al. Diagnostic potential of circulating natriuretic 8. Aoki J, Ikari Y, Nakajima H, Mori M, Sugimoto T, Hatori M, et al. peptides in chronic kidney disease. Nephrol Dial Transplant Clinical and pathologic characteristics of dilated cardiomyopathy 2006;21:402–10. in hemodialysis patients. Kidney Int 2005;67:333–40. 25. Fruhwald FM, Fahrleitner A, Watzinger N, Fruhwald S, Dobnig H, 9. Mizuno Y, Yoshimura M, Harada E, Nakayama M, Sakamoto T, Schumacher M, et al. Natriuretic peptides in patients with dia- Shimasaki Y, et al. Plasma levels of A- and B-type natriuretic stolic dysfunction due to idiopathic dilated cardiomyopathy. Eur peptides in patients with hypertrophic cardiomyopathy or idio- Heart J 1999;20:1415–23. pathic dilated cardiomyopathy. Am J Cardiol 2000;86:1036–40. 10. Richardson P, McKenna W, Bristow M, Maisch B, Mautner B, 26. Mottram PM, Leano R, Marwick TH. Usefulness of B-type natri- O’Connel J, et al. Report of the 1995 World Health Organization/ uretic peptide in hypertensive patients with exertional dyspnea International Society and Federation of Cardiology Task Force on and normal left ventricular ejection fraction and correlation with the definition and classification of cardiomyopathies. Circulation new echocardiographic indexes of systolic and diastolic function. 1996;93:841–2. Am J Cardiol 2003;92:1434–8. 11. Piccoli A, Nigrelli S, Caberlotto A, Bottazzo S, Rossi B, Pillon L, et 27. Luchner A, Stevens TL, Borgeson DD, Redfield M, Wei CM, Porter al. Bivariate normal values of the bioelectrical impedance vector in JG, et al. Differential atrial and ventricular expression of myocar- adult and elderly populations. Am J Clin Nutr 1995;61:269–70. dial BNP during evolution of heart failure. Am J Physiol Heart Circ 12. The Joint European Society of Cardiology/American College of Physiol 1998;274:H1684–9. Cardiology Committee for the redefinition of myocardial infarction. 28. Atherton JJ, Moore TD, Thomson HL, Frenneaux MP. Restrictive Myocardial infarction redefined: a consensus document of the left ventricular filling patterns are predictive of diastolic ventricular Joint European Society of Cardiology/American College of Cardi- interaction in chronic heart failure. J Am Coll Cardiol 1998;31: ology Committee for the Redefinition of Myocardial Infarction. 413–8. J Am Coll Cardiol 2000;36:959–69. 29. Wang TJ, Larson MG, Levy D, Benjamin EJ, Leip EP, Wilson PW, et 13. Anderson B. Echocardiography: The Normal Examination and al. Impact of obesity on plasma natriuretic peptide levels. Circu- Echocardiographic Measurements. Manly, Queensland: MGA lation 2004;109:594–600. Graphics, 2000. 30. Krauser DG, Lloyd-Jones DM, Chae CU, Cameron R, Anwaruddin S, 14. Devereux RB, Alonso DR, Lutas EM, Gottlieb GJ, Campo E, Sachs Baggish AL, et al. Effect of body mass index on natriuretic peptide I, et al. Echocardiographic assessment of left ventricular hyper- levels in patients with acute congestive heart failure: a ProBNP trophy: comparison to necropsy findings. Am J Cardiol 1986;57: Investigation of Dyspnea in the Emergency Department (PRIDE) 450–8. substudy. Am Heart J 2005;149:744–50. 2104 Codognotto et al.: Natriuretic Peptides in Uremic and Dilated Cardiomyopathies

31. Das SR, Drazner MH, Dries DL, Vega GL, Stanek HG, Abdullah 34. Ortega O, Gallar P, Munoz M, Rodriguez I, Carreno A, Ortiz M, et al. SM, et al. Impact of body mass and body composition on Association between C-reactive protein levels and N-terminal circulating levels of natriuretic peptides. Circulation 2005;112: pro-B-type natriuretic peptide in pre-dialysis patients. Nephrol Clin 2163–8. Pract 2004;97:125–30. 32. Nishikimi T, Yoshihara F, Morimoto A, Ishikawa K, Ishimitsu T, 35. Zoccali C, Mallamaci F, Benedetto FA, Tripepi G, Parlongo S, Saito Y, et al. Relationship between left ventricular geometry and Cataliotti A, et al. Cardiac natriuretic peptides are related to left natriuretic peptide levels in essential hypertension. Hypertension ventricular mass and function and predict mortality in dialysis 1996;28:22–30. patients. J Am Soc Nephrol 2001;12:1508–15. 33. Yasumoto K, Takata M, Ueno H, Tomita S, Tomoda F, Inoue H. 36. Vickery S, Price CP, John RI, Abbas NA, Webb MC, Kempson ME, Relation of plasma brain and atrial natriuretic peptides to left et al. B-type natriuretic peptide (BNP) and amino-terminal proBNP ventricular geometric patterns in essential hypertension. Am J in patients with CKD: relationship to renal function and left Hypertens 1999;12:921–4. ventricular hypertrophy. Am J Kidney Dis 2005;46:610–20. Clinical Chemistry 53:12 2105–2111 (2007) Proteomics and Protein Markers

Modified Form of the Fibrinogen B␤ Chain (des-Gln B␤), a Potential Long-Lived Marker of Pancreatitis

David Schmidt and Stephen O. Brennan*

Background: During an investigation of genetic vari- provide clinicians with retrospective evidence of disease. ants of fibrinogen, we observed a novel form of the B␤ © 2007 American Association for Clinical Chemistry chain, with a mass decrease of approximately 128 Da, in one of the controls. The plasma sample originated from Fibrin is the primary protein of blood clots, and its correct an individual who had experienced acute pancreatitis a deposition is essential for preserving the integrity of the week earlier but whose serum amylase activity had hemovascular system. Its precursor, fibrinogen, is a 340- returned to normal. We investigated the structure of the kDa glycoprotein composed of 6 polypeptide chains [i.e., ␣ ␤ ␥ modified fibrinogen and explored its relationship to (A ,B , )2] linked by 29 disulfide bonds. The molecule pancreatic disease. has a symmetrical trinodal structure with a central E Method: Fibrinogen was isolated from the plasma of 9 domain linked to 2 peripheral D domains in a linear individuals with increased pancreatic amylase activity D-E-D configuration. The E domain contains the N ter- (114–1826 U/L) and presumed pancreatitis and from 6 mini of all 6 chains, and the outer D domains are formed control individuals with amylase activities <56 U/L. from the C-terminal regions of the B␤ and ␥ chains (1, 2). Fibrinogen (or fibrin) B␤ chains were isolated by re- On activation by thrombin, cleavage of the A and B versed-phase HPLC and analyzed directly by electro- peptides from the respective N termini of the A␣ and B␤ spray ionization mass spectrometry. Tryptic and CNBr chains initiates the polymerization process. The newly peptide mapping and thrombin treatment pinpointed exposed Gly-Pro-Arg and Gly-His-Arg sequences dock the location of the 128-Da loss in mass. with preformed binding sites located in homologous Results: The acquired fibrinogen B␤ chain modification regions of the D domains of neighboring molecules. This was attributable to the loss of its C-terminal glutamine binding leads to the formation of a half-staggered bimo- residue. Incubating purified fibrinogen with pancreatic lecular array of fibrin molecules, and these protofibrils carboxypeptidase A (CpA) produced an identical modi- subsequently condense both longitudinally and laterally ␤ fication. The des-Gln B fibrinogen accounted for to form the clot matrix (3). ␤ >80% of the B chains in 3 of the individuals with The circulating fibrinogen molecule displays a vast ␤ increased amylase but only approximately 5% of the B amount of genetic and acquired variation, and these pre- chains in control samples. and posttranslational modifications have important ef- Conclusion: Pancreatic CpA activity is used as an index fects on function (4, 5). There are alternative transcripts of acute pancreatic disease, but given that the circulatory for both the A␣ and ␥ chains, with the A␣ being phos- half-lives of fibrinogen and CpA are approximately 4 phorylated nonstoichiometrically (1) and both the B␤ and ␤ days and only 2.5 h, respectively, measuring des-Gln B ␥ chains containing biantennary oligosaccharide side fibrinogen, the in vivo product of CpA activity, could chains that terminate either with 2 sialic acids or with 1 sialic acid and 1 galactose residue (5). The existence of common genetic polymorphisms for the A␣ (Thr/Ala312) and B␤ chains (Arg/Lys448 and Pro/Leu235) further Molecular Pathology Laboratory, Canterbury Health Laboratories, Christchurch, New Zealand. confound comparisons of mass measurements, particu- * Address correspondence to this author at: Molecular Pathology Labora- larly when comparing B␤-chain masses between individ- tory, Canterbury Health Laboratories, PO Box 151, Christchurch 8140, New uals (6). This inherent B␤-chain variation is further com- Zealand. Fax 64-3-3640545; e-mail [email protected]. Received September 12, 2007; accepted September 13, 2007. plicated by variable phosphorylation of the proline Previously published online at DOI: 10.1373/clinchem.2007.093179 residue at position 31 (1).

2105 2106 Schmidt and Brennan: Modified Fibrinogen and Pancreatitis

Notwithstanding these limitations, the mean mass of formed onto a true molecular-mass scale with the maxi- the major B␤-chain component reported for 6 individuals mum entropy algorithm. with typical coagulation profiles was 54 200 Da, with an additional peak at ϩ291 Da corresponding to the disialy- preparation of fibrin ␤ chains and lated isomer (5). During our investigations of the molec- fibrinopeptides ular basis of dysfibrinogenemia and hypofibrinogenemia, We redissolved the purified fibrinogen (0.7 mg) in 300 ␮L we noted an additional B␤ peak at Ϫ130 Da in a control of 25 mmol/L Tris-HCl and 25 mmol/L NaCl, pH 7.4, and sample from a patient with pancreatitis. We describe the added 3 U bovine thrombin. After 25 min, the resulting structure of this isoform and our measurements of its clot was recovered by winding around a glass rod; the proportion in relation to the activity of pancreatic amy- supernatant was retained for fibrinopeptide isolation. The lase, a marker of pancreatic disease. clot was solubilized in 8 mol/L urea, and the individual chains were separated as described above for fibrinogen. Materials and Methods The fibrinopeptide-containing supernatant was heated to measurement of amylase activity 96 °C for 5 min to inactivate thrombin and to precipitate We measured the activity of pancreatic amylase in the any residual soluble fibrin. After centrifugation, the fi- plasma with an Architect c8000 chemistry analyzer (Ab- brinopeptides were chromatographed on a Nova-Pak C18 bott Laboratories) with Abbott reagents and with proto- column (150 mm ϫ 3.9 mm; Waters Corporation). A linear cols that use an antibody inhibitor of salivary amylase gradient of 33%–56% solvent B was applied over 12 min. in a colorimetric assay with a protected p-nitrophenyl– Solvent B was 500 mL/L acetonitrile in 49 mmol/L maltoheptaoside substrate. KH2PO4, pH 2.9; solvent A was 49 mmol/L KH2PO4,pH 2.9 (7). The A and B peptides were dried at 55 °C under fibrinogen isolation and chain separation nitrogen. The peptides were redissolved in water, and the Fibrinogen was isolated from heparin-anticoagulated phosphate salts were removed by extracting the peptides plasma by precipitation with 23% saturated ammonium onto C18 resin (8). After elution with 600 mL/L acetoni- sulfate and washing 3 times with 25% saturated ammo- trile, we analyzed the peptides by electrospray ionization 1 nium sulfate (7). The starting plasma solution and the (ESI) mass spectrometry (MS). washing solutions all contained 1 mmol/L phenylmethylsulfonyl fluoride, 10 mmol/L ␧-aminocaproic trypsin digestion ␤ ␮ acid, 10 mmol/L EDTA, and 5 mmol/L cysteamine Isolated B chains (approximately 200 g) were dried ␮ hydrochloride. under nitrogen and redissolved in 50 L 50 mmol/L ␮ Purified fibrinogen (or fibrin) was dissolved to a con- NH4HCO3 containing 10 g trypsin. After overnight centration of 5–10 g/L in 8 mol/L urea containing 100 incubation at 37 °C, the digest was dried, redissolved in mmol/L Tris-HCl, pH 8.0, and 15 mmol/L dithiothreitol. 500 mL/L acetonitrile and 20 mL/L formic acid in water, The solution was incubated at 37 °C for 3 h, and then and analyzed by ESI MS (6). either 2 ␮L (analytical) or 50 ␮L (preparative) was loaded ϫ CNBr digestion onto a Jupiter C4 column (250 mm 4.6 mm; Phenome- ␤ ␮ nex), and the protein solution was fractionated with a Isolated B chains (approximately 200 g) were redissolved in a 25-␮L volume of 700 mL/L formic acid that linear gradient of 61%–81% solvent B over 20 min (5). ␮ Solvent B was 600 mL/L acetonitrile and 0.5 mL/L contained 200 g CNBr. After overnight incubation at room temperature, the mixture was dried under reduced trifluoroacetic acid in water; solvent A was 0.5 mL/L pressure over NaOH. We added 25 ␮L water, vortex- trifluoroacetic acid. The column was monitored at 215 nm; mixed the suspension, transferred the soluble peptides to peak crests and the bulk of the material were collected a new tube, and added 25 ␮L acetonitrile and 1 ␮L formic separately. acid. We injected 20 ␮L of this solution directly into the mass spectrometer. electrospray ionization mass spectrometry Twenty microliters of the fibrin or fibrinogen chain peak carboxypeptidase a treatment crests were injected into the ion source of a VG Platform We redissolved fibrinogen to 5 g/L in 50 mmol/L II quadrupole instrument (Micromass) at a flow rate of 5 NH HCO containing 0.18 g/L bovine carboxypeptidase ␮L/min (5). The system was operated in positive-ion 4 3 A (CpA) and incubated the fibrinogen solution overnight mode, the probe was charged at ϩ3000 V, and the source at room temperature (9). We then reprecipitated the temperature was maintained at 60 °C. The m/z range of fibrinogen with 25% saturated ammonium sulfate, sepa- 850-1600 was scanned for 2.5 s with an interscan time of 100 ms and a cone voltage ramp of 40–65 V. At least 80 scans were collected and averaged for each run. The raw uncalibrated data were processed with MassLynx Mass 1 Nonstandard abbreviations: ESI, electrospray ionization; MS, mass spec- Spectrometry software (Waters Corporation) and trans- trometry; CpA, carboxypeptidase A. Clinical Chemistry 53, No. 12, 2007 2107

rated the chains by reversed-phase HPLC, and measured their masses by ESI MS. fibrin polymerization Purified fibrinogen (500 ␮L of an approximately 2.0 g/L solution) was dialyzed over 16 h against 5 250-mL changes of buffer (20 mmol/L HEPES and 150 mmol/L NaCl, pH 7.4). After dialysis, we measured the fibrinogen concentration from the difference in absorbance at 280 nm and 320 nm and with an absorptivity of 15.1; we then adjusted the concentration to 0.444 g/L with dialysis buffer. We added 180 ␮L of the fibrinogen solution (containing 80 ␮g fibrinogen) in triplicate to wells of a black isoplate and then added 20 ␮L 20 U/L human ␣-thrombin. The absorbance at 350 nm was monitored at 12-s intervals during a 1-h period.

Results Fibrinogen was purified from the plasma of 9 patients with indications of pancreatitis (pancreatic amylase activities from 114-1826 U/L) and from 6 individuals with typical amylase activities (Ͻ56 U/L). We detected no differences between the 2 groups in gel patterns after polyacrylamide gel electrophoresis with SDS (data not shown). On nonreducing gels, both groups showed the expected 340- and 305-kDa bands associated with fully ␣ ␤ ␥ intact (A ,B, )2 molecules and molecules with 1 cleaved A␣ chain, respectively; on reducing gels, both groups showed the same pattern of A␣ (64 kDa), B␤ (52 kDa), and ␥ chains (48 kDa). When the individual fibrinogen chains were separated by reversed-phase HPLC, the 2 groups again showed similar elution profiles; a typical pattern is shown in Fig. 1. When we analyzed the individual peak crests directly by ESI MS, however, the B␤-chain peak showed marked alterations in the proportions of the different isoforms (Fig. 2). The major signal at 54 194 (13) Da [mean (SD), n ϭ 15] represents the previously characterized monosia- lylated form of the glycoprotein chain, and the species at 54 487 (13) Da is its disialylated derivative (5). The masses of both of these peaks were quantitatively decreased by 129 Da in samples with high amylase activities, and Fig. 1. Reversed-phase HPLC profile of reduced fibrinogen chains paired typical and altered (Ϫ129 Da) signals were clearly showing separation of A␣,B␤, and ␥ chains. visible in plasma samples with intermediate amylase B␤ peak crests were collected for direct mass analysis, and the remainder of the crest and bulk material was dried for mass mapping. activities. The presence of the same modification in the 2 B␤ chains with different biantennary oligosaccharide termini suggested an alteration in the polypeptide itself chains by reversed-phase HPLC, and analyzed the ␤ peak rather than in the carbohydrate structure. by MS (Fig. 3). As expected, both sets of ␤ chains showed To help locate the modification site, we incubated the characteristic decrease of 1535 Da corresponding to purified fibrinogen with thrombin and separated the the loss of fibrinopeptide B; however, the fact that the ␤ soluble fibrinopeptides from the resulting fibrin polymer. chain from a high-amylase individual had a mass 130 Da Our reversed-phase HPLC analysis of the fibrinopeptides lower than its typical counterpart suggested a protein derived from the N termini showed a typical pattern of A modification of the C terminus rather than the N and B peptides, and mass analysis confirmed the pre- terminus. dicted peptide masses and sequences (data not shown). We used tryptic peptide mapping to confirm this We redissolved the fibrin clot in 8 mol/L urea containing supposition and to obtain a more accurate measurement 15 mmol/L dithiothreitol, separated the individual fibrin of the decrease in mass. Maps for chains with high and 2108 Schmidt and Brennan: Modified Fibrinogen and Pancreatitis

Fig. 2. Transformed ESI MS profiles of isolated fibrinogen B␤ chains. (A), typical profile of a sample from a control individual with an amylase activity of 23 U/L; the paired peaks represent the mono- and disialylated isoforms of the glycoprotein chain. (B), sample from an individual with an amylase activity of 114 U/L. (C), sample from an individual with an amylase activity of 1671 U/L. (D), isolated B␤ chains from fibrinogen treated with CpA. The ordinate shows the response relative to the most intense peak in the spectrum. The amylase activity and the percentage of modified B␤ chains are indicated in each panel. ND, amylase activity not determined. low degrees of modification were very similar, but those high-amylase individual was the truncated form for an individual with high amylase activity lacked the (KIRPFFPQ) at 517 m/z (Fig. 4). The presence of both peak at 1032 m/z that was seen in the control individual peaks in each spectrum was because the fibrinogen sam- (Fig. 4). This signal at 1032 m/z represents the Mϩ1H ion ple from the control individual was 9% modified, whereas of T-53 (IRPFFPQQ), the C-terminal peptide of the B␤ the patient’s fibrinogen was 77% modified. chain, and loss of its C-terminal glutamine (128 Da) would The data to this point provide compelling evidence cause the peptide to appear at 904 m/z. The presence of that the degree of C-terminal modification of the fibrino- some 904 m/z ions in the control reflects partial modifica- gen B␤ chain was loosely related to pancreatic disease, tion of the B␤ chains, because previous measurements of and because such disease is associated with increased the intact chains in this control had shown them to be 30% plasma activities of amylase, lipase, and various pro- modified. teases, CpA in particular, we investigated the effect of this Because the B␤ chain has a methionine residue conve- exoprotease on fibrinogen. After incubating native fibrin- niently located 9 residues in from the C terminus, we used ogen with CpA, we isolated the B␤ chains and analyzed CNBr digestion to confirm the putative pruning of the them by ESI MS. CpA incubation produced a pattern terminal glutamine residue. The predicted C-terminal identical to that identified in the patients with acute fragment (KIRPFFPQQ) would have an Mϩ2H ion at pancreatitis: there was complete conversion of both the 581.2 m/z, and its truncated counterpart would have a mono- and disialylated isoforms to their des-Gln forms corresponding ion at 517.1 m/z. A direct analysis of the (Fig. 2). water-soluble CNBr peptides showed that although the We then examined the kinetics of thrombin-catalyzed control had a dominant signal for the full-length peptide fibrin polymerization to determine whether the modifica- at 581 m/z, the predominant form in the digest from the tion affected fibrinogen function. In triplicate experi- Clinical Chemistry 53, No. 12, 2007 2109

decrease in mass. This decrease can be entirely accounted for by the loss of the C-terminal glutamine residue. Incubating native fibrinogen with CpA reproduced the in vivo findings exactly, with only the terminal glutamine residue in the KIRPFFPQQ sequence being cleaved. A similar proteolytic modification has been reported for serum albumin in association with pancreatic pseudocysts (11). The principal modified form, des-Leu albumin, lacked the C-terminal leucine residue, and incubation with CpA reproduced the cleavage that had occurred in vivo (9). Although typical baseline CpA activities are very low (12), increased activity has been demonstrated in plasma samples from patients with pancreatitis (13–16). CpA, the major carboxypeptidase produced by the pancreas, preferentially cleaves uncharged C-terminal residues (17); however, a number of CpAs with slightly different spec- ificities have recently been identified in several mammalian species (18). Whereas CpA 1 prefers branched aliphatic amino acids at the C terminus, CpA 2 cleaves bulky aromatic residues (19, 20). The bovine CpA we used has a combined specificity for both aliphatic and aromatic amino acids (17), but we presume that CpA 1 is the modifying enzyme in humans because of the more aliphatic character of the glutamine residue. Cleavage of the penultimate glutamine is probably prevented by the neighboring proline. Previous searches for novel biomarkers of pancreatic inflammation have focused on the enzymes released from necrotic cells (e.g., amylase, lipase, and CpA), and little attention has been paid to the products generated by these enzymes. The results of the present study and the previous study on albumin (11) provide valuable information on modifications that affect 2 of the major plasma proteins during pancreatitis; however, further research on the kinetics of the process and on the correlation between the Fig. 3. Transformed mass spectrum of fibrin ␤ chains isolated from proportion of the variant polypeptide and the course of thrombin-treated fibrinogen. the disease is necessary to evaluate the diagnostic benefit (A), sample from a control individual with 3% of ␤ chains modified. (B), sample of this type of marker. This consideration is particularly from an individual with Ͼ90% of ␤ chains modified. After removal of the important because of the differences in the half-lives of N-terminal fibrinopeptide (1553 Da), the mass decrease of approximately 131 Da is still evident in the C-terminal portion of the protein. The amylase activity the causative enzymes and their circulating protein sub- and the percentage of modified ␤ chains are indicated in each panel. ND, strates. Fig. 5 shows a plot of the percentage of des-Gln B␤ amylase activity not determined. as a function of pancreatic amylase activity. Of note is that one of the controls with 30% modification had a typical ments, purified fibrinogen with high (85%) and low (5%) activity of 25 U/L. Given that the half-life of CpA is degrees of modification showed similar lag times, maxi- approximately 2.5 h (21) and that of fibrinogen is approx- mum velocity values, and final turbidities (data not imately 4 days, this result may indicate an individual shown), and these results were well within the range of recovering from acute pancreatic injury, because the con- nonpathologic variation (10). trol samples were all from individuals for whom analyses of pancreatic amylases had been requested. Removing Discussion this outlier suggests a typical mean percentage of approx- An approximate molecular-mass decrease of 130 Da was imately 5%, a proportion consistent with previous (un- observed for both major isoforms of the fibrinogen B␤ published data) observations made during an investiga- chain derived from individuals with high pancreatic tion of functionally abnormal fibrinogens associated with amylase activities. Tryptic and CNBr peptide mapping clotting abnormalities. located the modification site to the C terminus and Quantification of des-Gln B␤ by fibrinogen purifica- yielded a more precise value of 128 Da for the actual tion, HPLC, and protein ESI MS may seem impracticable 2110 Schmidt and Brennan: Modified Fibrinogen and Pancreatitis

Fig. 4. Peptide maps. Tryptic digests of isolated B␤ chains from individuals with low (A) and high (B) amylase activities. The ion at 1032 m/z (2) arising from the C-terminal peptide is missing and has been replaced by the ion at 904 m/z (1). CNBr digests of B␤ chains from individuals with low (C) and high (D) levels of modification. The Mϩ2H ion of the C-terminal peptide (KIRPFFPQQ) is largely missing at 581 m/z (2) in the high-amylase sample and has been replaced by the signal at 517 m/z (1). The amylase activity and the percentage of modified chains are indicated in each panel. in a routine diagnostic setting, but such an approach may assay. In addition, with the increasing availability of MS be useful in a research context. The development of a technology in clinical laboratories, a “bottom up” ap- monoclonal antibody capable of recognizing the trun- proach that uses tryptic or CNBr digests with signature- cated molecule should be possible, however, and such an ion quantification by direct-injection MS could be feasible. antibody could easily be incorporated into a routine Either fibrinogen pelleted from an ammonium sulfate solution or a washed fibrin clot could be a convenient fibrin(ogen) source. To assess the possible consequences of the truncation, we examined fibrin polymerization, because the functionally important D domain contains C termini of both the B␤ and ␥ chains. A comparison of fibrin polymerization kinetics revealed no significant differences, however. Given that the very end of the B␤ chain has no defined function and that the important C terminus of the ␥ chain seems to be intact in patient-derived fibrinogen (data not shown), this result is not surprising.

Grant/funding support: None declared. Financial disclosures: None declared.

References 1. Henschen A, McDonagh J. Fibrinogen, fibrin and factor XIII. In: Fig. 5. Graph showing variation in the percentage of des-Gln B␤ chains Zwaal FFA, Hemker HC, eds. Blood Coagulation. Amsterdam: as a function of pancreatic amylase activity. Elsevier Science Publishers BV, 1986;171–241. The regression line illustrates the correlation between the established marker 2. Doolittle RF. The molecular biology of fibrin. In:. Stamatoyanno- and the novel marker. poulos G, Nienhuis AW, Majerus PW, Varmus H, eds. The Molec- Clinical Chemistry 53, No. 12, 2007 2111

ular Basis of Blood Diseases. Philadelphia: WB Saunders Com- 12. Stewart JD, Gilvarg C. Determination of the activity of carboxypep- pany, 1994;701–23. tidase A in the blood of healthy human adults. Clin Chim Acta 3. Mosesson MW, Sibenlist KR, Mah DA. The structure and biological 1999;281:19–28. features of fibrinogen and fibrin. Ann NY Acad Sci 2001;936:11– 13. Roth M, Rohner A. Determination of pancreatic carboxypeptidase 30. A in human blood serum. Clin Chim Acta 1983;135:65–71. 4. Cote HC, Lord ST, Pratt KP. ␥-Chain dysfibrinogenemias: molecular 14. Kazmierczak SC, Van Lente F. Measuring carboxypeptidase A structure-function relationships of naturally occurring mutations in activity with a centrifugal analyzer: analytical and clinical consid- the gamma chain of human fibrinogen. Blood 1998;92:2195– erations. Clin Chem 1989;35:251–5. 212. 15. Peterson LM, Holmquist B, Bethune JL. A unique activity assay for 5. Brennan SO. Electrospray ionisation analysis of human fibrinogen. carboxypeptidase A in human serum. Anal Biochem 1982;125: Thromb Haemost 1997;78:1055–8. 420–6. 16. Brown KS, Kingsbury WD, Hall NM, Dunn GL, Gilvarg C. Determi- 6. Brennan SO, Fellowes AP, Faed JL, George PM. Hypofibrinogen- nation of carboxypeptidase A using N-acetyl-phenylalanyl-3-thia- emia in an individual with two coding (␥82 A3G and B␤235 P3L) phenylalanine as substrate: application to a direct serum assay. and two non-coding mutations. Blood 2000;95:1709–13. Anal Biochem 1987;161:219–25. 7. Brennan SO, Hammonds B, George PM. Aberrant hepatic process- 17. Petra PH, Hermodson MA, Walsh KA, Neurath H. Characterization ing causes removal of activation peptide and primary polymeriza- of bovine carboxypeptidase A. Biochemistry 1971;10:4023–5. tion site from fibrinogen Canterbury (A␣20 Val3Asp). J Clin Invest 18. Pascual R, Burgos FJ, Salva M, Soriano F, Mendez E, Avile´s FX. 1995;96:2854–8. Purification and properties of five different forms of human 8. Liew OW, Chong JPC, Yandle TG, Brennan SO. Preparation of procarboxypeptidases. Eur J Biochem 1989;179:609–16. recombinant thioredoxin fused N-terminal proCNP: analysis of 19. Gardell SJ, Craik CS, Clauser E, Goldsmith EJ, Stewart CB, Graf M, enterokinase cleavage products reveals new enterokinase cleav- et al. A novel rat carboxypeptidase, CPA2: characterization, mo- age sites. Protein Expr Purif 2005;41:332–40. lecular cloning, and evolutionary implications on substrate speci- 9. Brennan SO, Fellows AP, George PM. Albumin Banks Peninsula: a ficity in the carboxypeptidase gene family. J Biol Chem 1988;263: new termination variant characterised by electrospray mass spec- 17828–36. trometry. Biochim Biophys Acta 1999;1433:321–6. 20. Reznik SE, Fricker LD. Carboxypeptidases from A to Z: implica- 10. Maghzal GJ, Brennan SO, George PM. Fibrinogen B␤ polymor- tions in embryonic development and Wnt binding. Cell Mol Life Sci phisms do not directly contribute to an altered in vitro clot 2001;58:1790–804. structure in humans. Thromb and Haemost 2003;90:1021–8. 21. Ton GN, Weichert JP, Longino MA, Fine JP, Kwon GS. Methoxypoly- 11. Brennan SO, George PM. Three truncated forms of serum albumin (ethylene glycol)-conjugated carboxypeptidase A for solid tumor associated with pancreatic pseudocyst. Biochim Biophys Acta targeting: part II: pharmacokinetics and biodistribution in normal 2000;1481:337–43. and tumor-bearing rodents. J Control Release 2005;104:155–66. Clinical Chemistry 53:12 2112–2118 (2007) Proteomics and Protein Markers

Risk Stratification for Heart Failure and Death in an Acute Coronary Syndrome Population Using Inflammatory Cytokines and N-Terminal Pro-Brain Natriuretic Peptide

Peter A. Kavsak,1* Dennis T. Ko,2 Alice M. Newman,2 Glenn E. Palomaki,3 Viliam Lustig,4 Andrew R. MacRae,4,5 and Allan S. Jaffe6

Background: Inflammation in acute coronary syndrome onset. Kaplan–Meier analysis demonstrated that indi- (ACS) can identify those at greater long-term risks for viduals with increased NT-proBNP (>183 ng/L) or cyto- heart failure (HF) and death. The present study assessed kines (IL-6 > 6.4 ng/L; above upper limit of normal for the performance of interleukin (IL)-6, IL-8, and mono- IL-8 or MCP-1) had a greater probability of death or HF cyte chemoattractant protein-1 (MCP-1) (cytokines in- in the following 8 years (P <0.05). In a Cox proportional volved in the activation and recruitment of leukocytes) hazard model adjusted for both CRP and troponin I, in addition to known biomarkers [e.g., N-terminal pro- increased IL-6, MCP-1, and NT-proBNP remained sig- brain natriuretic peptide (NT-proBNP)] for predicting nificant risk factors. Combining all 3 biomarkers re- HF and death in an ACS population. sulted in a higher likelihood ratio for death or HF than Methods: In a cohort of 216 ACS patients, NT-proBNP models restricted to any 2 of these biomarkers. (Elecsys®; Roche) and IL-6, IL-8, and MCP-1 (evidence Conclusion: IL-6, MCP-1, and NT-proBNP are indepen- investigator™; Randox) were measured in serial spec- dent predictors of long-term risk of death or HF, high- -lighting the importance of identifying leukocyte activa .(723 ؍ imens collected early after symptom onset (n We collected at least 2 specimens from each partici- tion and recruitment in ACS patients. pant: an early specimen (median 2 h; interquartile © 2007 American Association for Clinical Chemistry range 2–4 h) and a later specimen (9 h; 9–9 h), and used the later specimens’ biomarker concentrations Even modestly increased concentrations of C-reactive for risk stratification. protein (CRP),7 a marker of inflammation, have been Results: An increase in both IL-6 and NT-proBNP was shown to be predictive for both short- and long-term risk observed but not for IL-8 or MCP-1 early after pain of heart failure (HF) and death, but not acute myocardial infarction, in patients with acute coronary syndrome (ACS) (1–4). A recent report also indicated that increased CRP concentrations are associated with new HF in non- 1 Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada. ACS patients with stable coronary artery disease (5). CRP 2 Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, is an acute-phase reactant, but its concentration takes time Ontario, Canada. to increase during an acute event (6). Furthermore, in- 3 Department of Pathology, Women and Infants Hospital, Providence, RI. creases are not specific for vascular inflammation (7). 4 Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada. Interleukin (IL)-6 is a proinflammatory cytokine thought 5 Department of Biochemistry and Medical Genetics, University of Mani- to be the most important proximate stimulator for CRP; it toba, Winnipeg, Manitoba, Canada. also stimulates activation of leukocytes (7, 8). IL-8 and 6 Cardiovascular Division and Division of Laboratory Medicine, Mayo Clinic, Rochester, MN. * Address correspondence to this author at: Hamilton Regional Laboratory Medicine Program, Henderson General Hospital (Core Laboratory Section), 711 Concession St. Hamilton, Ontario, Canada L8V 1C3. Fax 905-575-2581; 7 Nonstandard abbreviations: CRP, C-reactive protein; HF, heart failure; e-mail [email protected]. ACS, acute coronary syndrome; IL, interleukin; MCP-1, monocyte chemoat- Received April 19, 2007; accepted September 17, 2007. tractant protein-1; NT-proBNP, N-terminal pro-brain natriuretic peptide; IQR, Previously published online at DOI: 10.1373/clinchem.2007.090613 interquartile range; cTnI, cardiac troponin I; HR, hazard ratio.

2112 Clinical Chemistry 53, No. 12, 2007 2113

monocyte chemoattractant protein-1 (MCP-1) are both tients) were collected at specified intervals (from time of chemokines that recruit neutrophils and monocytes, re- symptom onset hourly until 6 h, and then at 9, 12, 24, and spectively, to the inflammatory process (9). We have 48 h) until the patient was discharged, declined further reported that increased baseline concentrations of CRP participation, or was removed from the study by those are predictive of long-term risk for HF and death in an responsible for his or her care. All specimens from 1996 ACS population (4). In the present study, we sought to were frozen, predominantly at –70 °C, until 2003 when the determine if these more specific and earlier mediators of heparin specimens were thawed and cardiac troponin I leukocyte activation and recruitment, coupled with as- (cTnI) and CRP measurements were performed on the sessment of heart dysfunction based on N-terminal pro- Access® (AccuTnI assay) and Immage® (high-sensitivity brain natriuretic peptide (NT-proBNP) concentrations, CRP assay) instruments, respectively, from Beckman could provide prognostic information beyond that of CRP Coulter (4, 10–12). We have confirmed the stability of and troponin in a cohort of chest pain patients presenting troponin I and CRP over time in our cohort (4). For the to the emergency department. Moreover, because the present study, we selected only those individuals (n ϭ optimal timing for measuring cytokines is unknown in 216) who had at least 2 EDTA specimens available in this setting, we wanted to document the cytokine profile storage. The median number of specimens per partici- early after chest pain onset. pant was 3 [interquartile range (IQR) 2–5], and all specimens (n ϭ 723) were measured with the cytokine Materials and Methods array and NT-proBNP to provide a serial and temporal study population profile of the biomarkers. For the outcome analysis, we The study population and its characteristics have been selected only 2 specimens per participant based on the reported (4, 10–12). At the time of study enrollment in following criteria: the earliest available (1st specimen) 1996, 448 consecutive unique patients presenting with and the closest to 9 h after onset (2nd specimen). In the symptoms suggestive of cardiac ischemia to the emer- event that the 1st specimen obtained was Ͼ6 h after gency department in a community hospital were re- onset, then the next specimen at least 3 h later was cruited for a retrospective cardiac marker study. Time of selected as the 2nd specimen. Thus the minimum symptom onset was solicited, and blood samples (both interval between specimen pairs was 3 h [median 6.5 h EDTA and heparin anticoagulated blood from all pa- (IQR 5–8)].

Table 1. Study cohort characteristics. Female Male Total P value n 84 132 216 Median age (IQR), years 71 (57–76) 64 (53–74) 66 (53–76) 0.009 Previous myocardial infarction No 66 (78.6) 91 (68.9) 157 (72.7) 0.121 Yes 18 (21.4) 41 (31.1) 59 (27.3) History of HF 9 (10.7) 12 (9.1) 21 (9.7) 0.695 In hospital course Thrombolytic No 80 (95.2) 117 (88.6) 197 (91.2) 0.095 Yes Ͻ5 15 (11.4) Acetylsalicylic acid No 68 (81.0) 90 (68.2) 158 (73.1) 0.039 Yes 16 (19.0) 42 (31.8) 58 (26.9) Nitroglycerin No 41 (48.8) 55 (41.7) 96 (44.4) 0.303 Yes 43 (51.2) 77 (58.3) 120 (55.6) Median hospital stay (IQR), days 4 (2–7) 4 (2–7) 4 (2–7) 0.704 Acute myocardial infarction diagnosis 1996 diagnosis No 70 (83.3) 104 (78.8) 174 (80.6) 0.411 Yes 14 (16.7) 28 (21.2) 42 (19.4) STEMI (Q wave) 5 (6.0) 17 (12.9) 22 (10.2) 0.101 Outcomes (HF/death) 6 months 12 (14.3) 21 (15.9) 33 (15.3) 0.746 2 years 27 (32.1) 31 (23.5) 58 (26.9) 0.162 8 years 45 (53.6) 61 (46.2) 106 (49.1) 0.292 2114 Kavsak et al.: Cytokines and NT-proBNP in ACS for HF/Death

1000 * mortality outcomes and the Canadian Institute for Health Information Discharge Abstract Database for hospital discharges associated with HF (11). Both the Registered 100 Persons Data Base and Canadian Institute for Health * Information Discharge Abstract Database (i.e., adminis- 10 Log Conc (ng/L) trative databases) have been reported to be highly accurate in obtaining these endpoints (18–21). Based on the death date and earliest subsequent readmission for HF, 1

P 8 P 1 indicators were created to reflect whether an event (death -6 -8 -1 -6 -8 -1 P -6 - -1 -6 -8 - P L P N L L P N L IL P N L L P N I IL C B I I C I C I I C B o oB roB M o or HF readmission) occurred within 8 years postpresen- M pr M r M r - -p -p -p T T T T N N N N tation (in patients who died without previous HF read- Time from Onset Groups: 3 h 4-6 h 7-9 h 10 h mission, follow-ups were censored at the date of death). Median Time (IQR): 2 (2-3) 5 (4-6) 9 (9-9) 12 (12-12) # of specimens: n=196 n=205 n=175 n=147 We used biomarker concentrations (e.g., NT-proBNP, IL-6, IL-8, MCP-1) in the later (2nd) specimen to deter- P * <0.05 for IL-6 and NT-proBNP by Kruskal-Wallis test across the time groups mine risk based on the hypothesis that this specimen Fig. 1. Time-concentration profile for IL-6, IL-8, MCP-1, and NT-proBNP would reflect the severity of cardiac dysfunction and the after symptom onset. inflammatory response of the patient better than the Each point with error bars represents the median and IQR for each of the biomarkers, respectively. earlier specimen. Of note, exploratory analyses using logistic regression models with the cytokines for the combined endpoint death/HF suggested no difference in biomarker measurements long-term outcomes (e.g., at 8 years) between the 1st and In 2006, the EDTA specimens were thawed for the 1st 2nd specimen; however, the 2nd specimen tended to be time, and a cytokine array was measured using the more predictive for early outcomes (e.g., 30 days and 1 evidence investigator™ (Randox) biochip platform year) than the 1st specimen. The analyses were based on (9, 13). The biochip can assay 12 cytokines; however, a the following classifications. For both IL-6 and NT- priori the decision was made to evaluate only IL-6, IL-8, proBNP, we used the median concentrations at the 2nd and MCP-1 for the present study assessing HF and death Ͼ Ͼ in an ACS population. The interassay (n ϭ 20 assays) time point (IL-6 6.4 ng/L and NT-proBNP 183 ng/L). imprecision (CV), determined by measuring 3 levels of We used the upper limit of normal (97.5th percentile) for Ͼ Ͼ quality control material, ranged from 10.9% to 16.2% for IL-8 ( 7.5 ng/L) and MCP-1 ( 156 ng/L). For our study ϭ IL-6, 8.3% to 16.1% for IL-8, and 7.6% to 13.0% for MCP-1. population (n 216), the combined endpoint (death/HF) We measured NT-proBNP by use of the Elecsys® 1010 was used for all analyses to maximize the number of events. (Roche), with interassay impression Ͻ7%. There is evi- We constructed Kaplan–Meier curves to display time dence to support the stability of the cytokines measured to an event (death/HF) and assessed differences between in the present study after 10 years’ storage, in that the groups by use of the log-rank test. We used the Cox reference ranges for IL-8 and MCP-1 published in 2006 for proportional hazard model to compare time to an event this method were derived from samples collected as early for the increased biomarkers. In keeping with our previ- as 1994 (9, 14). IL-6 and NT-proBNP also appear to be ous analyses in this cohort (4, 11), we used different stable during long-term storage (15–17). models to assess the risk associated with increased cytokines: model 1 adjusts for age (continuous variable), sex, health outcomes and statistical analysis history of HF, and STEMI (Q wave) at presentation; model Research ethics board approval was obtained to measure 2 adjusts for age (continuous variable), sex, history of HF, biomarkers in the stored samples and to make health STEMI (Q wave), presentation CRP concentration Ͼ7.44 outcome linkages to the Registered Persons Data Base for mg/L (the concentration that optimized performance of

Table 2. Biochemical characteristics for specimen set.a Variable 1st specimen 2nd specimen P value n 216 216 Time from onset, h 2 (2–4) 9 (9–9) cTnI, ␮g/L 0.01 (0.00–0.04) 0.03 (0.01–0.26) Ͻ0.001 IL-6, ng/L 4.0 (1.8–11.5) 6.4 (2.8–17.0) Ͻ0.001 IL-8, ng/L 6.1 (4.5–9.5) 6.5 (4.8–10.2) 0.116 MCP-1, ng/L 150 (116–202) 155 (114–203) 0.102 NT-proBNP, ng/L 147 (47–867) 183 (56–854) Ͻ0.001 a Data are median (IQR). Selected pairs of specimens in the cohort were at least 3 h apart, maximum 12 h apart. Time between pairs of specimens was 6.5 h (5.0–8.0). Clinical Chemistry 53, No. 12, 2007 2115

Fig. 2. Kaplan–Meier survival curves for IL-6, NT-proBNP, IL-8, and MCP-1. IL-6 (A) and NT-proBNP (B) group assignments based on median values; IL-8 (C) and MCP-1 (D) group assignments based on published reference ranges (9).

the model), and cTnI peak categories. The cTnI peak Results concentrations were categorized into 4 groups with values For the 216 participants (61% male), the median age defined as Յ0.01 ␮g/L, 0.02–0.03 ␮g/L, 0.04–0.10 ␮g/L, (25th–75th percentile) was 66 years (53–76) (Table 1). At and Ͼ0.10 ␮g/L. We designated values of 0.00–0.01 ␮g/L the time of study enrollment in 1996, 19.4% of the partic- as the reference group based on previous analyses show- ipants were diagnosed with acute myocardial infarction ing that increases in risk occur at values Ն0.02 ␮g/L based on WHO Monitoring Cardiovascular Disease (11, 22). We used the highest cTnI concentration in the (MONICA) criteria (10, 23). Applying the European Soci- model to assess the possibility that the inflammatory ety of Cardiology/American College of Cardiology crite- markers measured increased in proportion to the extent of ria retrospectively based on the peak cTnI concentrations necrosis and thus reflected infarct size, a known determi- resulted in 44.4% of participants having a cTnI concentra- nant of prognosis. Significance of the association was tion Ͼ99th percentile (Ͼ0.04 ␮g/L) (10, 24, 25). There based on the Wald ␹2 statistic, with significance set at P were increases in both NT-proBNP and IL-6 concentra- Ͻ0.05. We used Cox proportional hazard models to assess tions, but not IL-8 or MCP-1, early after the onset of chest the ability of combinations of IL-6, NT-proBNP, and pain (Fig. 1). This result was also evident by the increased MCP-1 and their interactions to predict death/HF. Mod- concentrations in the 2nd specimen (median 9 h after els were constructed for IL-6 alone, IL-6 and MCP-1 (with onset) vs the 1st specimen (median 2 h after onset) and without interaction), IL-6 and NT-proBNP (with and (Table 2). without interaction), and IL-6, MCP-1, and NT-proBNP Kaplan–Meier analysis using the concentrations from (with and without interactions), with the likelihood ratio the later (2nd) specimen demonstrated that increased as well as the significance of the association based on the concentrations of each of the cytokines and NT-proBNP Wald ␹2 statistic (P Ͻ0.05). We based between-group resulted in a greater probability for death/HF over the 8 comparisons of central tendency on the Wilcoxon and years after emergency department presentation (Fig. 2). Kruskal–Wallis tests. Analyses were performed using SAS Cox proportional hazard models adjusting for age, sex, version 9.1.3 and GraphPad Prism version 5.00. history of HF, and STEMI at presentation yielded signif- 2116 Kavsak et al.: Cytokines and NT-proBNP in ACS for HF/Death

Table 3. HRs for death/HF as determined by IL-6, MCP-1, and NT-proBNP.a Time since presentation HR relative to IL-6 <6.4 ng/L 95% CI P value IL-6 Model 1 6 months 4.59 1.68–12.50 0.003 2 years 4.35 2.08–9.09 Ͻ0.001 8 years 3.11 1.98–4.88 Ͻ0.001 Model 2 6 months 3.79 1.28–11.19 0.016 2 years 3.27 1.48–7.20 0.003 8 years 2.57 1.58–4.19 Ͻ0.001 HR relative to MCP-1 <156 ng/L MCP-1 Model 1 6 months 2.65 1.22–5.77 0.014 2 years 1.92 1.11–3.31 0.019 8 years 1.58 1.07–2.33 0.023 Model 2 6 months 2.70 1.23–5.95 0.014 2 years 2.27 1.28–4.03 0.005 8 years 1.74 1.17–2.59 0.006 HR relative to NT-proBNP <183 ng/L NT-proBNP Model 1 6 months 3.45 1.32–9.04 0.012 2 years 3.87 1.80–8.30 Ͻ0.001 8 years 3.19 1.97–5.17 Ͻ0.001 Model 2 6 months 2.15 0.80–5.73 0.128 2 years 2.66 1.23–5.78 0.013 8 years 2.64 1.61–4.33 Ͻ0.001 a Model 1 adjusts for age, sex, history of HF, and STEMI (Q wave); model 2 adjusts for age, sex, history of HF, STEMI (Q wave), presentation CRP Ͼ7.44 mg/L, and cTnI peak categories (Ͻ0.02 ␮g/L as reference). icant hazard ratios (HRs) for IL-6, MCP-1, and NT- HF in patients with ACS (1–4, 26–28). This study is proBNP, but not for IL-8, at all 3 time points. After different from others in that the analyzed cytokines were adjusting for high CRP and peak cTnI concentrations, measured early after the onset of pain, a time when they only the increased IL-6 and MCP-1 groups were signifi- have roles in the activation and recruitment of leukocytes cantly associated with an increased risk for death/HF at 6 during the inflammatory process. In addition, they were months; however, at 2 and 8 years, increased IL-6, MCP-1, assessed as predictors of both short-term (6 months) and and NT-proBNP all had significant HRs (Table 3). To long-term (8 years) outcomes after adjusting for cTnI, assess whether there was a synergistic relationship be- using a cutoff value below the 99th percentile previously tween IL-6, NT-proBNP, and MCP-1 for predicting long- shown to be of prognostic importance (11), and after term death/HF in our population, a time-to-event analy- adjusting for CRP values previously found to be predic- sis (Cox proportional model) was performed with IL-6 tive (4). alone and with MCP-1 and NT-proBNP alone and to- A rising pattern was observed for both IL-6 and gether to assess their combinations and interactions (Ta- NT-proBNP but not for the chemokines (IL-8 and MCP-1), ble 4). Including IL-6, MCP-1, and NT-proBNP in the suggesting that IL-6 and/or NT-proBNP production may model resulted in a significantly higher likelihood ratio be an acute response to myocardial injury. However, it for death/HF compared with IL-6 alone and the combi- did not appear to be related to the extent of cardiac injury nation of IL-6 with either MCP-1 or NT-proBNP. as reflected by the Cox proportional analysis using peak cTnI concentrations to correct for this possibility. Thus, it Discussion seems that the exuberance of the acute inflammatory The present study confirms earlier work indicating that response is an important predictor of both intermediate inflammation is a strong predictor of future death and/or and long-term prognosis independent of the extent of

Table 4. Cox proportional hazard models for assessing combinations of IL-6, MCP-1, and NT-proBNP for death/HF at 8 years. Model Markers in model Likelihood ratio Degrees of freedom Models compared P value 1 IL-6 alone 16.5 1 Reference 2 IL-6, MCP-1 19.3 2 2 vs 1 0.094 3 IL-6 ϩ MCP-1 ϩ interactions 22.7 3 3 vs 2 0.066 4 IL-6, NT-proBNP 36.4 2 4 vs 1 Ͻ0.001 5 IL-6 ϩ NT-proBNP ϩ interaction 40.1 3 5 vs 4 0.056 6 IL-6, MCP-1, NT-proBNP 41.4 3 6 vs 2; 6 vs 4 Ͻ0.001; 0.025 7 IL-6, MCP-1, NT-proBNP ϩ 2-way interactions 43.6 6 7 vs 6 0.542 Clinical Chemistry 53, No. 12, 2007 2117

myocardial injury. IL-6, as a potent proinflammatory Acknowledgments: Special thanks to the staff at the Clinical cytokine, by itself may exacerbate the damage that results Research and Clinical Trials Laboratory at the Hamilton from minor myocardial necrosis/injury or may itself Regional Laboratory Medicine Program and Randox Labora- stimulate muscle atrophy and myocardial failure during tories Ltd. for technical support. ACS (29). For immune-mediated damage, recruitment/ redirection of leukocytes is required. This process may be References the reason MCP-1 has been implicated in ACS (30, 31) and 1. Toss H, Lindahl B, Siegbahn A, Wallentin L. Prognostic influence of HF (32, 33), but to our knowledge this result is the 1st increased fibrinogen and C-reactive protein levels in unstable indication that increases are independent predictors for coronary artery disease. Circulation 1997;96:4204–10. 2. James SK, Armstrong P, Barnathan E, Califf R, Lindahl B, Sieg- long-term death/HF in an ACS population. bahn A, et al. Troponin and C-reactive protein have different Because of the small sample size, we cannot separate relations to subsequent mortality and myocardial infarction after the risks associated for either HF or death alone or include acute coronary syndrome. J Am Coll Cardiol 2003;41:916–24. a large number of covariates in our modeling. Moreover, 3. Suleiman M, Khatib R, Agmon Y, Mahamid R, Boulos M, Kapelio- our population does not represent a contemporary cohort. vich M, et al. Early inflammation and risk of long-term development This approach is a strength in the sense that it provides a of heart failure and mortality in survivors of acute myocardial better natural history; however, prospective studies are infarction. J Am Coll Cardiology 2006;47:962–8. necessary that take into account present medical manage- 4. Kavsak PA, MacRae AR, Newman AM, Lustig V, Palomaki GE, Ko DT, et al. Elevated C-reactive protein in acute coronary syndrome ment in patients with ACS (more medical and mechanical presentation is an independent predictor of long-term mortality interventions). Also, because our aim was to evaluate and and heart failure. Clin Biochem 2007;40:326–9. document changes in biomarkers over time, we opted to 5. Sabatine MS, Morrow DA, Jablonski KA, Rice MM, Warnica W, include only individuals with at least 2 specimens at least Domanski MJ, et al. Prognostic significance of the Centers for 3 h apart. The 216 individuals selected for the study were, Disease Control/American Heart Association high-sensitivity C-re- on average, older than the 232 individuals who were active protein cut points for cardiovascular and other outcomes in excluded (64.7 vs 60.5 years; P ϭ 0.002), they stayed in the patients with stable coronary artery disease. Circulation 2007; 115:1528–36. hospital longer for their initial event, and they had higher ␮ 6. Yip HK, Hang CL, Fang CY, Hsieh YK, Yang CH, Hung WC, et al. cTnI peak concentrations (median 0.04 vs 0.01 g/L). Level of high-sensitivity C-reactive protein is predictive of 30-day However, important for this analysis, there was no differ- outcomes in patients with acute myocardial infarction undergoing ence in presentation CRP concentrations, and the end- primary coronary intervention. Chest 2005;127:803–8. points at 30 days and 1 year between the included and 7. Yu H, Rifai N. High-sensitivity C-reactive protein and atherosclero- excluded patient groups were not different. sis: from theory to therapy. Clin Biochem 2000;33:601–10. This study builds on previous work by using a vali- 8. Delves PJ, Roitt IM. The immune system. Second of two parts. dated analytical platform for cytokine measurements and N Engl J Med 2000;343:108–17. reference ranges for IL-8 and MCP-1 for classifying ele- 9. Berrahmoune H, Lamont JV, Herbeth B, Fitzgerald PS, Visvikis- vations (9, 13). Moreover, the cutoffs used in this study Siest S. Biological determinants of reference values for plasma interleukin-8, monocyte chemoattractant protein-1, epidermal closely resemble other reported values in the literature for growth factor, and vascular endothelial growth factor: Results from risk stratification for IL-6 (26) and the median value for the STANISLAS cohort. Clin Chem 2006;52:504–10. NT-proBNP in an ACS population (34). 10. Kavsak PA, MacRae AR, Lustig V, Bhargava R, Vandersluis R, Palomaki GE, et al. The impact of the ESC/ACC redefinition of In conclusion, these data support the findings that inflam- myocardial infarction and new sensitive troponin assays on the mation and cardiac dysfunction in ACS indicate a poor frequency of acute myocardial infarction. Am Heart J 2006;152: prognosis, independent of the extent of myocardial necro- 118–25. 11. Kavsak PA, Newman AM, Lustig V, MacRae AR, Palomaki GE, Ko sis, in that ACS patients with increased IL-6, MCP-1, and DT, et al. Long-term health outcomes associated with detectable NT-proBNP are at greater risk for subsequent HF and troponin I concentrations. Clin Chem 2007;53:220–7. death. 12. Kavsak PA, MacRae AR, Palomaki GE, Newman AM, Ko DT, Lustig V, et al. Health outcomes categorized by current and previous definitions of acute myocardial infarction in an unselected cohort of troponin naı¨ve emergency department patients. Clin Chem Grant/funding support: This work was supported by a grant 2006;52:2028–35. from the Canadian Institutes of Health Research. 13. Fitzgerald SP, Lamont JV, McConnell RI, Benchikh EO. Develop- Financial disclosures: None declared (P.A.K., D.T.K., A.M.N., ment of a high-throughput automated analyzer using biochip array G.E.P., A.R.M.). A.S.J. receives research support from and is technology. Clin Chem 2005;51:1165–76. a consultant for Beckman-Coulter, Dade-Behring, Ortho Di- 14. Siest G, Visvikis S, Herbeth B, Gueguen R, Vincent-Viry M, Sass C, et al. Objectives, design and recruitment of a familial and longi- agnostics, Critical Diagnostics, Intermune, Pfizer, Bayer, and tudinal cohort for studying gene-environment interactions in the GlaxoSmithKline. He is or has been at one time or another, a field of cardiovascular risk: the Stanislas cohort. Clin Chem Lab consultant to most of the major diagnostic companies. V.L. Med 1998;36:35–42. has received financial support for lecturing on cardiac mark- 15. Kenis G, Teunissen C, De Jongh R, Bosmans E, Steinbusch H, ers from Roche Diagnostics. Maes M. Stability of interleukin 6, soluble interleukin 6 receptor, 2118 Kavsak et al.: Cytokines and NT-proBNP in ACS for HF/Death

interleukin 10 and CC16 in human serum. Cytokine 2002;19: kinase MB mass for use with European Society of Cardiology/ 228–35. American College of Cardiology consensus recommendations. 16. Blankenberg S, McQueen M, Smieja M, Pogue J, Balion C, Lonn E, Clin Chem 2003;49:1331–6. et al. Comparative impact of multiple biomarkers and N-terminal 26. Lindmark E, Diderholm E, Wallentin L, Siegbahn A. Relationship pro-brain natriuretic peptide in the context of conventional risk between interleukin 6 and mortality in patients with unstable factors for the prediction of recurrent cardiovascular events in the coronary artery disease. JAMA 2001;286:2107–13. heart outcomes prevention evaluation (HOPE) study. Circulation 27. Jernberg T, Lindahl B, Siegbahn A, Andren B, Frostfeldt G, 2006;114:201–8. Lagerqvist B, et al. N-terminal pro-brain natriuretic peptide in 17. Kragelund C, Gronning B, Kober L, Hildebrandt P, Steffensen R. relation to inflammation, myocardial necrosis, and the effect of an N-terminal pro-B-type natriuretic peptide and long-term mortality in invasive strategy in unstable coronary artery disease. J Am Coll stable coronary heart disease. N Engl J Med 2005;352:666–75. Cardiol 2003;42:1909–16. 18. Austin PC, Daly PA, Tu JV. A multicentre study of the coding 28. James SK, Lindahl B, Timmer JR, Ottervanger JP, Siegbahn A, accuracy of hospital discharge administrative data for patients Stridsberg M, et al. Usefulness of biomarkers for predicting admitted to cardiac care units in Ontario. Am Heart J 2002;144: long-term mortality in patients with diabetes mellitus and non-ST- 290–6. elevation acute coronary syndromes. Am J Cardiol 2006;97:167– 19. Tu JV, Pashos C, Naylor CD, Chen E, Normand S, Newhouse JP, et 72. al. Use of cardiac procedures and outcomes in elderly patients 29. Janssen SP, Gayan-Ramirez G, Bergh AV, Herijgers P, Maes K, with myocardial infarction in the United States and Canada. N Engl Verbeken E, et al. Interleukin-6 causes myocardial failure and J Med 1997;336:1500–5. skeletal muscle atrophy in rats. Circulation 2005;111: 20. Lee DS, Donovan L, Austin PC, Gong Y, Liu PP, Rouleau JL, et al. 996–1005. Comparison of coding of heart failure and comorbidities in admin- 30. Aukrust P, Berg RK, Ueland T, Aaser E, Damas JK, Wikeby L, et al. istrative and clinical data for use in outcomes research. Med Care Interaction between chemokines and oxidative stress: possible 2005;43:182–8. pathogenic role in acute coronary syndromes. J Am Coll Cardiol 21. Roger VL, Weston SA, Redfield MM, Hellermann-Homan JP, Killian 2001;37:485–91. J, Yawn BP, et al. Trends in heart failure incidence and survival in 31. De Lemos JA, Morrow DA, Sabatine MS, Murphy SA, Gibson CM, a community-based population. JAMA 2004;292:344–50. Antman EM, et al. Association between plasma levels of mono- 22. Zethelius B, Johnston N, Venge P. Troponin I as a predictor of cyte chemoattractant protein-1 and long-term clinical outcomes in coronary heart disease and mortality in 70-year-old men: a com- patients with acute coronary syndromes. Circulation 2003;107: munity-based cohort study. Circulation 2006;113:1071–8. 690–5. 23. Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakan- 32. Aukrust P, Ueland T, Muller F, Andreassen AK, Nordov I, Aas H, et gas A, Pajak A. Myocardial infarction and coronary deaths in the al. Elevated circulating levels of C-C chemokines in patients with World Health Organization MONICA project: registration proce- congestive heart failure. Circulation 1998;97:1136–43. dures, event rates, and case fatality rates in 38 populations from 33. Hayashidani S, Tsutsui H, Shiomi T, Ikeuchi M, Matsusaka H, 21 countries in four continents. Circulation 1994;90:583–612. Suematsu N, et al. Anti-monocyte chemoattractant protein-1 gene 24. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial infarc- therapy attenuates left ventricular remodeling and failure after tion redefined: a consensus document of the Joint European experimental myocardial infarction. Circulation 2003;108:2134– Society of Cardiology/American College of Cardiology Committee 40. for the redefinition of myocardial infarction. J Am Coll Cardiol 34. Cameron SJ, Green GB, White CN, Laterza OF, Clarke W, Kim H, et 2000;36:959–69. al. Assessment of BNP and NT-proBNP in emergency department 25. Apple FS, Quist HE, Doyle PJ, Otto AP, Murakami MM. Plasma patients presenting with suspected acute coronary syndromes. 99th percentile reference limits for cardiac troponin and creatine Clin Biochem 2006;39:11–8. Clinical Chemistry 53:12 2119–2127 (2007) Cancer Diagnostics

MS-FLAG, a Novel Real-Time Signal Generation Method for Methylation-Specific PCR

Cinzia Bonanno,1 Erlet Shehi,2 Daniel Adlerstein,2* and G. Mike Makrigiorgos3

Background: Aberrant promoter methylation is a major multiplex MS-FLAG assay for GATA5 and RASSF1 mechanism for silencing tumor suppressor genes in promoters. cancer. Detection of hypermethylation is used as a Conclusion: MS-FLAG provides a new, quantitative, molecular marker for early cancer diagnosis, as a prog- high-throughput method for detecting gene promoter nostic index, or to define therapeutic targets for rever- methylation and is a convenient alternative to agarose sion of aberrant methylation. We report on a novel gel-based MSP for screening methylation. In addition to signal generation technology for real-time PCR to detect methylation, FLAG-based real-time signal generation gene promoter methylation. may have broad applications in DNA diagnostics. Methods: FLAG (fluorescent amplicon generation) is a © 2007 American Association for Clinical Chemistry homogeneous signal generation technology based on the exceptionally thermostable endonuclease PspGI. Cytosine-5 methylation of CpG dinucleotides is associ- FLAG provides real-time signal generation during PCR ated with repression of gene expression and is recognized by PspGI-mediated cleavage of quenched fluorophores as an important event in carcinogenesis (1, 2). In verte- ,at the 5؅ end of double-stranded PCR products. Methy- brates, certain CpG-rich regions within gene promoters known as CpG islands, are typically unmethylated (3, 4) lation-specific PCR (MSP) applied on bisulfite-treated to facilitate gene transcription. Exceptions are CpG is- DNA was adapted to a real-time format (methylation- lands associated with tissue-specific genes, imprinted specific FLAG; MS-FLAG) for quantifying methylation genes (5, 6), or genes located on the X chromosome (7). in the promoter of CDKN2A (p16), GATA5, and RASSF1. The epigenetic pattern (epigenome) of an individual is We validated MS-FLAG on plasmids and genomic DNA established early during embryogenesis and is main- with known methylation status and applied it to detec- tained by DNA-methyltransferases during mitotic cell tion of methylation in a limited number of clinical divisions (8). DNA methylation represses transcription by samples. We also conducted bisulfite sequencing on recruiting histone deacetylases, thereby inducing a com- these samples. pact chromatin structure (4, 9) and/or altering DNA Results: Real-time PCR results obtained via MS-FLAG binding sites of transcription factors. Aberrant promoter agreed with results obtained via conventional, gel-based hypermethylation is a major mechanism for silencing MSP. The new technology showed high specificity, tumor suppressor genes in human cancer (10, 11). Genes sensitivity (2–3 plasmid copies), and selectivity (0.01% involved in cell-cycle regulation [p16, also called of methylated DNA) on control samples. It enabled CDKN2A4 (cyclin-dependent kinase inhibitor 2A); correct prediction of the methylation status of all 3 gene RASSF1 (Ras association (RalGDS/AF-6) domain family promoters in 21 lung adenocarcinoma samples, as con- 1)], apoptosis (DAPK, death-associated protein kinase), firmed by bisulfite sequencing. We also developed a and DNA repair (MGMT, O-6-methylguanine-DNA methyltransferase) are frequently silenced in cancer (12, 13). For these genes detection of promoter methylation status from surgical tumor specimens or bodily 1 University of Milano Bicocca, Milano, Italy. fluids provides a sensitive, specific, and relatively nonin- 2 DiaSorin SpA, Saluggia (VC), Italy. 3 Dana Farber-Brigham and Women’s Cancer Center, Harvard Medical School, Boston, MA. * Address correspondence to this author at: Diasorin SpA, Viale Pasteur 10, 20014 Nerviano (MI), Italy. Fax 0039-03311547; e-mail daniel.adlerstein@ 4 Human genes: CDKN2A, cyclin-dependent kinase inhibitor 2A (mela- diasorin.it. noma, p16, inhibits CDK4); RASSF1, Ras association (RalGDS/AF-6) domain Received June 27, 2007; accepted September 26, 2007. family 1; GATA5, GATA binding protein 5; DAPK, death-associated protein Previously published online at DOI: 10.1373/clinchem.2007.094011 kinase; MGMT, O-6-methylguanine-DNA methyl-transferase.

2119 2120 Bonanno et al.: MS-FLAG for Methylation Detection

vasive way to screen for cancer. CpG hypermethylation methylation (methylation-specific FLAG; MS-FLAG). We can be used as a molecular marker for early cancer first validated this new assay on DNA with known diagnosis (14–16), as a prognostic index (17, 18), and to methylation status and then demonstrated the detection define therapeutic targets for reversion of aberrant meth- of methylation for 3 tumor suppressor genes that are often ylation (11). found to be methylated in lung adenocarcinoma: Several techniques have been developed for analyzing CDKN2A, RASSF1, and GATA5 (GATA binding protein 5). the DNA methylation status of CpG islands, including methylation-sensitive restriction enzymes (19), MALDI- Materials and Methods TOF/mass spectrometry (20), and microarrays (21).A generation of plasmid dna and sources of major contribution to the detection of DNA methylation genomic dna in epigenetic studies, described by Frommer and col- To test the reliability of FLAG signal-generation technol- leagues (22, 23) in the early 1990s, was the laboratory use ogy on controls that eliminate the variability introduced of sodium bisulfite. Sodium bisulfite converts unmethyl- by the bisulfite treatment (22), we generated plasmid ated cytosines into uracils while leaving methylated cy- methylated and unmethylated control sequences for tosines relatively intact, thus creating sequence differ- RASSF1 promoter. We synthesized 2 177-bp DNA frag- ences between genomes that originally differ only in their ments, each simulating the sequence for fully methylated CpG methylation pattern. The nucleotide differences are or completely unmethylated RASSF1 promoter that then detected by sequencing (24), restriction enzyme would result after complete bisulfite conversion. We then analysis (25), PCR (26), and other methods. One of the cloned these fragments into PCR-Blunt 2.1 plasmids (In- most widely used methods is methylation-specific PCR vitrogen). The resulting methylated (M) and unmethyl- (MSP),5 described by Herman et al. (26). MSP uses ated (U) recombinant control plasmids were cultured in primers that bind to and amplify bisulfite-converted se- Escherichia coli strain TOP10 cells and extracted with quences only if CpG dinucleotides on these sequences NucleoSpin Plasmid Kit (Macherey-Nagel), and their remain unaffected by the chemical treatment (i.e., the identity was confirmed by automated sequencing (MWG cytosines are methylated). Alternatively, primers that Biotech). We used these plasmids as template in experi- bind specifically only to unmethylated cytosines within ments evaluating the specificity, sensitivity, and selectiv- the primer sequence can be used, thus revealing the ity of MS-FLAG. absence of methylation. MSP can detect 1 methylated The genomic DNA samples used as positive (M) and allele in the presence of 1000 unmethylated (normal) negative (U) controls were CpGenome Universal Methyl- alleles (26). Limitations in the originally described MSP ated DNA and CpGenome Universal Unmethylated DNA are the requirement for gel electrophoresis and insuffi- (Chemicon), respectively. Surgical lung adenocarcinoma cient quantification due to the endpoint–based PCR de- tumor samples were obtained from the Massachusetts tection format. Real-time PCR technologies using the General Hospital Tumor Bank, Boston, after we obtained TaqMan probe approach for signal generation (27) have internal review board approval. DNA was extracted from ௢ been adapted for detecting methylation (28, 29). Real-time these samples by use of the Dneasy Tissue Kit (Qiagen). technology eliminates the need for gel separation, provides quantitative information on the degree of DNA bisulfite treatment and sequencing methylation in a given sample with a sensitivity ap- To convert unmethylated cytosines to uracils, we treated proaching 1/10 000, and has the throughput and conve- 300 ng human genomic DNA from lung adenocarcinoma nience lacking in MSP (30). However, TaqMan-based samples and genomic DNA controls with sodium bisulfite approaches such as MethyLight can occasionally miss by use of the CpGenome DNA Modification Kit (Chemi- methylated samples that are detectable via MSP, possibly con) according to the manufacturer’s protocol. We per- because both primers and probe must hybridize correctly formed MS-FLAG assays on 1 ␮L bisulfite-treated DNA for signal generation. In partially methylated clinical (corresponding to approximately 5 ng starting material) to samples, this requirement is not always fulfilled. Further- investigate hypermethylation of promoter regions of more, multiplexing of more than 1 gene is relatively RASSF1, p16, and GATA5. We also examined the methyl- difficult when TaqMan approaches are used because of ation status of clinical samples via bisulfite sequencing. the multiple oligonucleotides that must be used in the After bisulfite treatment, the MS-FLAG target region was reactions. PCR-amplified using primers external to the MS-FLAG In this report we describe fluorescent amplicon gener- region, and the amplified products were processed via ation (FLAG), a new method for real-time signal genera- dideoxy sequencing. tion during PCR that is adapted to the detection of CpG methylight assay for p16 Sequence of primers and probe used in p16 MethyLight assay were as follows: forward primer p16_Fw (TGG AGT 5 Nonstandard abbreviations: MSP, methylation-specific PCR; FLAG, fluorescent amplicon generation; MS-FLAG, methylation-specific FLAG; NTC, TTT CGG TTG ATT GGT T), reverse primer P16_Rv no-target control. (AGG AGG TGC GGG CGT TGT T), and TaqMan probe Clinical Chemistry 53, No. 12, 2007 2121

Fig. 1. Principle of signal generation via MS-FLAG. MS-FLAG primers are designed to anneal to regions harboring methylated CpG sites and contain an oligonucleotide 5Ј tail carrying a quencher and a fluorophore separated by the recognition sequence of the PspGI endonuclease. The highly thermostable PspGI enzyme is present during the PCR reaction. (A), MS-FLAG forward primer an- neals to methylated target DNA (Step 1) and is extended by the DNA polymerase (Step 2). The reverse primer synthe- sizes the opposite strand (Step 3) and generates a double-stranded recognition sequence for PspGI. Cleavage by PspGI (Step 4) leads to separation of the quencher and fluorophore, generating fluorescence. (B), if the interrogated target DNA is not methylated, the binding of the primers is inefficient and no PCR product or fluorescence is generated.

p16_Probe (FAM-ACC CGA CCC CGA ACC GCG- p16. Forward primer CB78QF (TTT CCA GGT TTC GAT BHQ1). Reaction mixtures contained 300 nmol/L of each TCG TGT ACG ACG TTG), reverse primer CB79QF (TTT primer, 200 nmol/L probe, and 1ϫ TaqMan Universal CCA GGT TTG CAA CCG CGC GCA AA), which gener- PCR Master Mix (Applied Biosystems) in 30 ␮L total ate a 192-bp product. volume. After an initial denaturation of 10 min at 95 °C, the amplification protocol consisted of 40–45 cycles (95 °C RASSF1. Forward primer CB107QF (TTT CCA GGT TTA 15 s, 57 °C 15 s, and 60 °C 1 min). CGA GAG CGC GTT TAG TTT CGT TTT C), reverse primer CB108QF (TTT CCA GGT TTA GCT AAC AAA ms-flag assay for p16, gata5, and rassf1 CGC GAA CCG AAC G), which generate a 188-bp The restriction endonuclease used for FLAG signal gen- product. eration was PspGI (New England Biolabs). An 11-nucleotide oligonucleotide (TTT CCA GGT TT) containing the GATA5. Forward primer CB131QR (TTT CCA GGT TTC PspGI recognition sequence (underlined) was added to GTT GGG GTT TCG GTC GTA), reverse primer CB132QR (TTT CCA GGT TTA CTA ATC CGA ACT CCG CGC the 5Ј end of the gene-specific portion of the primers TA), which generate a 129-bp product. (primer tail). The primer tail was doubly labeled with Reaction mixtures contained 300 nmol/L of each Iowa-Black FQ quencher (Integrated DNA Technologies) primer, 300 ␮mol/L dNTPs, and 1ϫ JumpStart PCR at the 5Ј end and a fluorophore at the 3Ј end. We used buffer with 10 units PspGI enzyme (New England Biolab) fluorescein for p16 and RASSF1 assays and MAX (Inte- and 1 unit JumpStart Taq Polymerase (Sigma) in 20 ␮L grated DNA Technologies) dye for GATA5 assays. Fluo- total volume. After an initial denaturation of 3 min at rescence was detected and quantified on a Chromo4 94 °C, the amplification protocol consisted of 45–50 cycles real-time PCR machine (MJ Research) using the FAM of (94 °C 30 s, 68 °C 30 s, and 72 °C 1 min). To facilitate detection channel for fluorescein and VIC detection chan- multiplex reactions, the MS-FLAG primers were designed nel for MAX. We visualized the gene-specific amplifica- to operate under a single annealing temperature. tion products via ethidium bromide-stained 2% agarose gel electrophoresis. All MS-FLAG primers were designed Multiplex MS-FLAG. Assay conditions were slightly dif- with the support of VisualOmp software (DNA Software) ferent for the duplex GATA5/RASSF1 MS-FLAG, to ac- and synthesized by Integrated DNA Technologies. Se- count for the simultaneous amplification of 2 gene frag- quences of primers for bisulfite-converted DNA used in ments with different efficiencies. The reaction mixture in each assay were as follows. this case contained 300 nmol/L GATA5 primers 2122 Bonanno et al.: MS-FLAG for Methylation Detection

(CB131QR and CB132QR) labeled with MAX dye, 200 nmol/L RASSF1 primers (CB107QF, CB108QF) labeled with fluorescein, 500 ␮mol/L dNTPs, and 1ϫ JumpStart PCR buffer with 10 units PspGI enzyme (New England Biolab) and 1 unit JumpStart Taq Polymerase (Sigma) in a final volume of 20 ␮L.

Results ms-flag principle FLAG is a novel signal generation technology for real- time PCR (Fig. 1) that includes the endonuclease PspGI in the reaction. Amplification is performed using primers that have a target-specific 3Ј region and a 5Ј-oligonucleotide tail containing a fluorophore-quencher pair separated by nucleotides carrying the recognition sequence of an exceptionally thermostable restriction endonuclease, PspGI (28). PspGI has a half-life of2hat95°C(31) and remains active throughout the PCR reaction. At each amplification cycle, the complete double-stranded recognition site for this enzyme is generated at both ends (primer tails) of the amplified product. Cleavage of the primer tails by the endonuclease results in an increase in fluorescence due to loss of fluorescence resonance energy transfer quenching (Fig. 1). Because only the introduced 5Ј tail of the PCR amplicons is digested, amplicons retain their primer-binding regions, which serve as templates for primer binding and amplification. MS-FLAG primers are specific for methylated sequences (Fig. 1) such that fluorescence signals are generated only if the interrogated sample contains methylated CpG DNA. validation of ms-flag principle on plasmid controls: specificity, sensitivity, and selectivity To test MS-FLAG in regard to the reliability and specificity of real-time signal generation, 2 recombinant plasmids were synthesized to act as positive and negative controls. The plasmids contained the anticipated sequence that the RASSF1 promoter region will have after complete bisulfite conversion of fully methylated (positive) or completely unmethylated (negative) samples. In the 1st case, each CpG spot of the original target sequence is supposed to be methylated, whereas in the 2nd case, no methylated CpGs are present. These plasmid controls eliminate the variability introduced by bisulfite treatment (22) and allow an independent evaluation of the novel signal Fig. 2. Validation of MS-FLAG real-time signal generation on engineered plasmids. generation method without the complications introduced (A), specificity: real-time growth curves of duplicate independent experiments on by chemical treatment. fully methylated (M) or completely unmethylated (U) control plasmid DNA Real-time signal generation via MS-FLAG performed samples. The specific amplification PCR product (inset) and the corresponding fluorescence are selectively generated only in controls mimicking methylated on control plasmids is depicted in Fig. 2. Only plasmids samples. Duplicate independent experiments are depicted (overlapping growth representing fully methylated sequences generated real- curves). (B), sensitivity: the assay was performed on samples containing time signals (Fig. 2A), indicating the specificity of the decreasing amounts of RASSF1 plasmid DNA. Fluorescence was generated in samples down to 10 ag (approximately 2–3 copies of M plasmid DNA), with an designed primers for methylated CpG. Electrophoretic excellent degree of linearity, r2 ϭ 0.998 (inset). (C), selectivity: decreasing separation on ethidium bromide-stained agarose gels amounts of methylated (M) RASSF1 DNA were added to RASSF1 unmethylated (U) DNA. Methylation was detected with excellent linearity (r2 ϭ 0.999), down to (Fig. 2A, inset) depicts no amplification products for 0.01% methylated-to-unmethylated plasmid DNA (inset). no-target control (NTC) or unmethylated plasmid controls, whereas an approximately 188-bp band is present in Clinical Chemistry 53, No. 12, 2007 2123

the methylated plasmid controls. The sensitivity of MS- detection of CpG methylation in bisulfite- FLAG signal generation was then tested on 10-fold serial treated genomic dna controls dilutions of the methylated control plasmid, ranging from After validation of the real-time signal-generation 100 fg to 10 ag of target DNA. Signal generation was properties of MS-FLAG, the assay was applied to detected down to 10 ag (corresponding to 2–3 plasmid bisulfite-treated human genomic DNA that was either copies) of methylated DNA. A semilogarithmic plot of the completely unmethylated or fully methylated. We used thresholds obtained demonstrated an almost perfect lin- 1 ␮L bisulfite-treated DNA, corresponding to approxi- earity (r2 ϭ 0.998) (Fig. 2B, inset). The selectivity of mately 5 ng starting material, as a target for MS-FLAG MS-FLAG signal generation was tested by enriching the assays to identify the methylation status of the pro- RASSF1 methylated control plasmid into its unmethyl- moter region of the genes GATA5, RASSF1, and p16. ated counterpart, to form a series of dilutions with meth- Results from triplicate independent experiments are ylated-to-unmethylated ratios ranging from 1:10 to 1:106 shown in Fig. 3, A–C. The MS-FLAG thresholds [thresh- (0.0001%). MS-FLAG detected the methylated sequence in old cycle (SD)], as derived by the 3 independent re- 104-fold higher amounts of unmethylated sequences (Fig. peats, were 33.0 (0.5) (GATA5), 31.3 (0.2) (RASSF1), and 2C); the detection was quantitative (r2 ϭ 0.999; Fig. 2C 32.7 (0.5) (p16). Methylated samples generated fluores- inset). Overall, real-time signal generation using MS- cent signals corresponding to the formation of specific FLAG is highly sensitive and quantitative, and the prim- amplification products for each assay, as also verified ers designed for the RASSF1 promoter are selective for by agarose gel electrophoresis-based sizing of the final sequences expected to form after bisulfite treatment of PCR products (insets of Fig. 3, A–C). In terms of methylated CpG-containing DNA. fluorescence intensity and threshold cycle, MS-FLAG

Fig. 3. MS-FLAG on genomic DNA controls and comparison to MethyLight. MS-FLAG assay was performed on 1 ␮L commercially available fully methylated (M) or completely unmethylated (U) genomic DNA controls (Chemicon) after bisulfite treatment. For all 3 investigated genes, GATA5 (A), RASSF1 (B), and p16 (C), triplicate independent MS-FLAG experiments are depicted. The data show fluorescent signals corresponding to the formation of specific amplification products from methylated DNA, as also confirmed by ethidium bromide–stained gel electrophoresis (insets, conventional MSP). Unmethylated samples demonstrate no fluorescence and no gel electrophoresis (conventional MSP) bands. MethyLight assay for p16 performed in triplicate independent experiments on the same genomic samples (D) showed threshold cycle values and fluorescence intensities similar to those of MS-FLAG. 2124 Bonanno et al.: MS-FLAG for Methylation Detection

data were very similar to those of the well-established MethyLight technique, as shown for p16 in Fig. 3D. multiplex ms-flag for gata5 and rassf1 on genomic dna controls After validation of MS-FLAG in detecting the methylation status of DNA in simplex reactions, we also designed and optimized duplex assays that allow for simultaneous detection of the methylation status of 2 different genes in a single reaction. To this end, MS-FLAG primers for RASSF1 and GATA5 were synthesized with primer tails containing distinct fluorophores, FAM and MAX, respectively (the fluorescence of MAX can be monitored on the VIC channel of most real-time PCR machines). When the interrogated genomic DNA control sample was methylated for both genes, fluorescence was produced by PspGI cleavage of the primer tails in both sequences, generating real-time signals simultaneously in the FAM and VIC channels (Fig. 4A). Electrophoretic analysis of the PCR end products on agarose gels showed 2 bands (Fig. 4B), each representing a single methylated gene-specific amplification product (129 bp for GATA5 and 188 bp for RASSF1). methylation analysis of clinical samples To field test the new technology on clinical samples, MS-FLAG was used to investigate the methylation status

Fig. 5. Investigation of the methylation status of GATA5, p16, and RASSF1 in lung adenocarcinoma clinical samples. For each gene, MS-FLAG real-time growth curves are depicted for representative methylated and unmethylated surgical specimens (curves with error bars representing independent triplicate experiments), together with methylated or unmethylated genomic DNA controls.

of GATA5, p16, and RASSF1 on a limited number of Fig. 4. Multiplex MS-FLAG assay. samples from lung adenocarcinoma surgical specimens. Multiplex detection of the methylation status of 2 different genes (RASSF1 and GATA5) in a single reaction tube. Two different sets of primers were used, GATA5 was methylated in 10 of 21 samples (47%), labeled with 2 different fluorophores: fluorescein for RASSF1 (green curves) and whereas p16 and RASSF1 were methylated in 85% and MAX for GATA5 (red curves, detected on VIC channel). (A), methylated samples generated fluorescence in both FAM and VIC channels. (B), electrophoretic 57% of samples, respectively. MS-FLAG growth curves analysis of the PCR end products on ethidium bromide–stained agarose gel. indicating the methylation status of the interrogated Clinical Chemistry 53, No. 12, 2007 2125

genes were obtained in repeated independent experi- low fraction of methylated alleles within unmethylated ments. Representative real-time growth curves of methyl- alleles. Therefore clinical samples that contain a low ated and unmethylated samples are depicted in Fig. 5, fraction of methylated alleles may conceivably appear together with DNA controls. For an additional confirma- unmethylated when bisulfite sequencing is used but can tion of the results, we performed bisulfite sequencing on still demonstrate substantial methylation with MSP or samples analyzed via MS-FLAG to examine whether, MS-FLAG. In the samples examined in this investigation, indeed, the MS-FLAG primers amplify DNA containing a discrepancy between the 2 methods was not observed, methylated cytosines in CpG sites. Fig. 6 depicts repre- i.e., the samples were either fully methylated or fully sentative results from GATA5 for clinical samples TL58 unmethylated at the MS-FLAG primer-binding regions. and TL25 presented in Fig. 5. The presence of cytosines for Accordingly, the bisulfite sequencing data were consis- MS-FLAG–positive sample TL58 [as revealed by observ- tent with the conclusions obtained from MS-FLAG ap- ing guanines (see arrows) on the opposite strand of the plied to clinical samples. electropherogram] indicates methylated CpG sites on the MS-FLAG primer-binding sites (boxed area; Fig. 6A). In Discussion contrast, MS-FLAG–negative sample TL25 indicated thy- Quantitative and precise assays for methylation bear an midines at the same positions, suggesting full conversion increasingly important role in elucidating the influence of of unmethylated C to T by the bisulfite treatment (Fig. epigenetic modifications in cancer and differentiation 6C). Bisulfite sequencing of the methylated control is also (30). MS-FLAG is a new real-time PCR method that uses depicted (Fig. 6B). It is noteworthy that MSP is much the endonuclease PspGI for fluorescence signal generation more sensitive than bisulfite sequencing in identifying a and is adapted to the detection of DNA methylation.

Fig. 6. Bisulfite sequencing of clinical lung adenocarcinoma samples. Representative electropherograms of the samples TL58 and TL25 examined via MS-FLAG in Fig. 5 for GATA5 are shown. (A), the presence of cytosines for bisulfite-treated sample TL58 [as revealed by observing guanines (arrows) on the opposite strand of the electropherogram] indicates methylated CpG sites on the MS-FLAG primer-binding sites (boxed area). (B), the same GATA5 region was sequenced for the fully methylated genomic DNA control. A similar methylation pattern at the primer-binding sites is depicted. (C), the presence of thymidines for bisulfite-treated sample TL25 [as revealed by observing adenines (arrows) on the opposite strand of the electropherogram] indicates unmethylated CpG sites on the MS-FLAG primer-binding sites (boxed area). The data confirm the methylation status of samples TL58 and TL25 as predicted by MS-FLAG. 2126 Bonanno et al.: MS-FLAG for Methylation Detection

Because of the unusually high thermostability of PspGI References (31), which has an activity half-life of approximately2hat 1. Feinberg AP, Tycko B. The history of cancer epigenetics. Nat Rev 95 °C, the relief of fluorescence quenching from the 5Ј Cancer 2004;4:143–53. ends of double-stranded PCR products proceeds through- 2. Esteller M. Aberrant DNA methylation as a cancer-inducing mech- out the entire PCR reaction, resulting in robust fluores- anism. Annu Rev Pharmacol Toxicol 2005;45:629–56. 3. Takai D, Jones PA. Comprehensive analysis of CpG islands in cence signals. The reproducibility of MS-FLAG corre- human chromosomes 21 and 22. Proc Natl Acad Sci U S A sponds to SDs of 0.2–0.5 for the real-time PCR threshold, 2002;99:3740–5. which is very satisfactory considering the variability that 4. Jones PA, Baylin SB. The fundamental role of epigenetic events in can be introduced by the bisulfite treatment itself (30).A cancer. Nat Rev Genet 2002;3:415–28. potential limitation of MS-FLAG is that the amplified 5. Li E, Beard C, Jaenisch R. Role for DNA methylation in genomic region must not contain PspGI sites other than those imprinting. Nature 1993;366:362–5. incorporated at the 5Ј ends of the primers. This require- 6. Barlow DP. Gametic imprinting in mammals. Science 1995;270: ment restricts the application to amplicons that are devoid 1610–3. of 5Ј-CCTGG-3Ј sequence strings. For application to the 7. Costello JF, Plass C. Methylation matters. J Med Genet 2001;38: field of methylation, however, this restriction should not 285–303. be an issue, because bisulfite treatment removes the 8. Robertson KD. DNA methylation and chromatin-unraveling the tangled web. Oncogene 2002;21:5361–79. cytosines from PspGI recognition sequences and prevents 9. Wade PA. Methyl CpG binding proteins: coupling chromatin archi- illegitimate digestion of target sequences. tecture to gene regulation. Oncogene 2001;20:3166–73. Real-time PCR methods can be distinguished from those 10. Herman JG, Baylin SB. Promoter-region hypermethylation and using a hybridization probe (27, 32–34) or those using la- gene silencing in human cancer. Curr Top Microbiol Immunol beled PCR primers (35–40). The former provide high spec- 2000;249:35–54. ificity and sensitivity in target quantification; however, the 11. Esteller M. CpG island hypermethylation and tumor suppressor requirement for a 3rd oligonucleotide in addition to the genes: a booming present, a brighter future. Oncogene 2002;21: primers in the PCR reaction limits their multiplexing capa- 5427–40. bility, owing to primer–probe interactions. Approaches us- 12. Esteller M, Corn PG, Baylin SB, Herman JG. A gene hypermethylation profile of human cancer. Cancer Res 2001;61:3225–9. ing universally labeled PCR primers can be more cost- 13. Attwood JT, Yung RL, Richardson BC. DNA methylation and the effective as long as the formation of primer-dimers or regulation of gene transcription. Cell Mol Life Sci 2002;59:241– nonspecific amplification products does not adversely affect 57. specificity. Optimal primer design is therefore required in 14. Belinsky SA. Gene-promoter hypermethylation as a biomarker in the latter case. This requirement is particularly relevant for lung cancer. Nat Rev Cancer 2004;4:707–17. bisulfite-treated DNA, because the conversion of C to T 15. Palmisano WA, Divine KK, Saccomanno G, Gilliland FD, Baylin SB, degrades the DNA, lowers the DNA annealing temperature, Herman JG, et al. Predicting lung cancer by detecting aberrant and reduces the specificity of primers. In part because of promoter methylation in sputum. Cancer Res 2000;60:5954–8. these technical difficulties, multiplexed real-time MSP di- 16. Teodoridis JM, Strathdee G, Brown R. Epigenetic silencing mediated by CpG island methylation: potential as a therapeutic target rectly from genomic DNA has not, to our knowledge, been and as a biomarker. Drug Resist Updat 2004;7:267–78. previously reported. Our data demonstrate the successful 17. Hiranuma C, Kawakami K, Oyama K, Ota N, Omura K, Watanabe identification of methylation-specific primers for MS-FLAG G. Hypermethylation of the MYOD1 gene is a novel prognostic and PCR cycling conditions that avoid primer-dimer forma- factor in patients with colorectal cancer. Int J Mol Med 2004;13: tion while allowing effective PspGI-mediated signal genera- 413–7. tion. The demonstration of MS-FLAG multiplexing capabil- 18. Friedrich MG, Chandrasoma S, Siegmund KD, Weisenberger DJ, ities for the genes GATA5 and RASSF1 indicates the broad Cheng JC, Toma MI, et al. Prognostic relevance of methylation potential of this approach for increasing the throughput of markers in patients with non-muscle invasive bladder carcinoma. methylation detection. Overall, MS-FLAG combines the Eur J Cancer 2005;41:2769–78. 19. Singer-Sam J, Grant M, LeBon JM, Okuyama K, Chapman V, Monk broad sensitivity of MSP with the convenience and through- M, et al. Use of a HpaII-polymerase chain reaction assay to study put of real-time PCR, while enabling a more straightforward DNA methylation in the Pgk-1 CpG island of mouse embryos at the multiplexing than TaqMan-based approaches. time of X-chromosome inactivation. Mol Cell Biol 1990;10: The FLAG approach is expected to find broad addi- 4987–9. tional applications in real-time PCR, such as in virology, 20. Gut IG. DNA analysis by MALDI-TOF mass spectrometry. Hum genotyping, and infectious diseases. Furthermore, as pre- Mutat 2004;23:437–41. liminary results in our laboratory indicate, FLAG can be 21. Gitan RS, Shi H, Chen CM, Yan PS, Huang TH. Methylation-specific used for real-time RT-PCR for the detection of viral RNA. oligonucleotide microarray: a new potential for high-throughput methylation analysis. Genome Res 2002;12:158–64. 22. Frommer M, McDonald LE, Millar DS, Collis CM, Watt F, Grigg GW, et al. A genomic sequencing protocol that yields a positive display Grant/funding support: This work was supported by Dia- of 5-methylcytosine residues in individual DNA strands. Proc Natl Sorin SpA and in part by National Institutes of Health Grants Acad Sci U S A 1992;89:1827–31. CA111994-01 and CA115439-01 (to G.M.M.). 23. Clark SJ, Harrison J, Paul CL, Frommer M. High sensitivity mapping Financial disclosures: None declared. of methylated cytosines. Nucleic Acids Res 1994;22:2990–7. Clinical Chemistry 53, No. 12, 2007 2127

24. Grunau C, Clark SJ, Rosenthal A. Bisulfite genomic sequencing: exonuclease activity of Thermus aquaticus DNA polymerase. Proc systematic investigation of critical experimental parameters. Nu- Natl Acad Sci U S A 1991;88:7276–80. cleic Acids Res 2001;29:E65–5. 33. Tyagi S, Kramer FR. Molecular beacons: probes that fluoresce 25. Xiong Z, Laird PW. COBRA: a sensitive and quantitative DNA upon hybridization. Nat Biotechnol 1996;14:303–8. methylation assay. Nucleic Acids Res 1997;25:2532–4. 34. Wittwer CT, Herrmann MG, Moss AA, Rasmussen RP. Continuous 26. Herman JG, Graff JR, Myohanen S, Nelkin BD, Baylin SB. Methyl- fluorescence monitoring of rapid cycle DNA amplification. Biotech- ation-specific PCR: a novel PCR assay for methylation status of niques 1997;22:130–1, 134–8. CpG islands. Proc Natl Acad Sci U S A 1996;93:9821–6. 35. Nazarenko IA, Bhatnagar SK, Hohman RJ. A closed tube format for 27. Heid C, Stevens J, Livak K, Williams P. Real time quantitative PCR. amplification and detection of DNA based on energy transfer. Genome Res 1996;6:986–94. Nucleic Acids Res 1997;25:2516–21. 36. Nazarenko I, Lowe B, Darfler M, Ikonomi P, Schuster D, Rashtch- 28. Eads CA, Danenberg KD, Kawakami K, Saltz LB, Blake C, Shibata ian A. Multiplex quantitative PCR using self-quenched primers D, et al. MethyLight: a high-throughput assay to measure DNA labeled with a single fluorophore. Nucleic Acids Res 2002;30: methylation. Nucleic Acids Res 2000;28:E32. e37. 29. Lo YM, Wong IH, Zhang J, Tein MS, Ng MH, Hjelm NM. Quantitative 37. Whitcombe D, Theaker J, Guy SP, Brown T, Little S. Detection of analysis of aberrant p16 methylation using real-time quantitative PCR products using self-probing amplicons and fluorescence. Nat methylation-specific polymerase chain reaction. Cancer Res Biotechnol 1999;17:804–7. 1999;59:3899–903. 38. Todd AV, Fuery CJ, Impey HL, Applegate TL, Haughton MA. 30. Ogino S, Kawasaki T, Brahmandam M, Cantor M, Kirkner GJ, DzyNA-PCR: use of DNAzymes to detect and quantify nucleic acid Spiegelman D, et al. Precision and performance characteristics of sequences in a real-time fluorescent format. Clin Chem 2000;46: bisulfite conversion and real-time PCR (MethyLight) for quantita- 625–30. tive DNA methylation analysis. J Mol Diagn 2006;8:209–17. 39. Zhang Y, Zhang D, Li W, Chen J, Peng Y, Cao W. A novel real-time 31. Morgan R, Xiao J, Xu S. Characterization of an extremely thermo- quantitative PCR method using attached universal template stable restriction enzyme, PspGI, from a Pyrococcus strain and probe. Nucleic Acids Res 2003;31:e123. cloning of the PspGI restriction-modification system in Escherichia 40. Li J, Wang F, Mamon H, Kulke MH, Harris L, Maher E, et al. coli. Appl Environ Microbiol 1998;64:3669–73. Antiprimer quenching-based real-time PCR and its application to 32. Holland PM, Abramson RD, Watson R, Gelfand DH. Detection of the analysis of clinical cancer samples. Clin Chem 2006;52:624– specific polymerase chain reaction product by utilizing the 5Ј—3Ј 33. Clinical Chemistry 53:12 Lipids, Lipoproteins, 2128–2135 (2007) and Cardiovascular Risk Factors

Development of a Homogeneous Assay to Measure Remnant Lipoprotein Cholesterol

Kazuhito Miyauchi,1 Norihiko Kayahara,2* Masato Ishigami,3 Hideyuki Kuwata,4 Hideharu Mori,4 Hiroyuki Sugiuchi,5 Tetsumi Irie,6 Akira Tanaka,7 Shizuya Yamashita,8 and Taku Yamamura3

Background: Quantification of triglyceride-rich lipoprotein (apo) E and apo C-III from IDL particles in the poprotein (TRL) remnants is useful for risk assessment presence of cholesterol esterase (CHER), and the particle of coronary artery disease and the diagnosis of type III size distribution of IDLs became smaller after the reac- hyperlipoproteinemia. Although an immunoseparation tion. These results revealed that IDL particles are spe- procedure for remnant-like particle cholesterol has been cifically modified in the presence of CHER and POE- evaluated extensively in recent years, available methods POB, making their component cholesterol available for for measuring TRL remnants have not achieved wide enzymatic assay. Addition of phospholipase D im- use in routine laboratory practice, suggesting a need for proved the reactivity toward chylomicron remnants 1.018x– 0.01 ؍ a homogeneous assay that can measure TRL remnant (CMRs). We found a high correlation [y between the proposed assay [(160 ؍ n) 0.962 ؍ cholesterol in serum or plasma without pretreatment. mmol/L, r Methods: We screened for suitable surfactants that and the immunoseparation assay in serum from healthy exhibited favorable selectivity toward the VLDL rem- individuals. nant (VLDLR) fraction, including intermediate-density Conclusion: The homogeneous assay described in this lipoproteins (IDLs). We investigated the principal char- report can measure TRL remnant cholesterol, including acteristics of this assay by gel filtration of lipoproteins CMRs, VLDLRs, and IDLs, with high sensitivity and and their particle size distribution. We developed a specificity. simple assay and evaluated its performance with the © 2007 American Association for Clinical Chemistry Hitachi-7170 analyzer. Results: Polyoxyethylene-polyoxybutylene block copol- Previous clinical and experimental studies have sug- ymer (POE-POB) exhibited favorable selectivity toward gested that triglyceride-rich lipoprotein (TRL)9 remnants VLDLR and IDL fractions. POE-POB removed apoli- play an important role in atherogenesis (1). It has also been reported that postprandial hyperlipidemia induces fat accumulation in arterial walls (2) and is strongly 1 Scientific and Technical Affairs Department, Kyowa Medex Co., Ltd., associated with the progression of coronary artery disease Tokyo, Japan. (CAD) (3). Many studies have shown a high correlation 2 Research and Development Department, Kyowa Medex Co., Ltd., Tokyo, between postprandial hyperlipidemia and atherosclerosis Japan. 3 Division of Health Sciences, Osaka University Graduate School of Med- (4, 5). In addition, the Montreal Heart Study showed an icine, Osaka, Japan. independent contribution of increased VLDL remnants 4 Research Laboratory Department, Kyowa Medex Co., Ltd., Shizuoka, (VLDLRs) plus intermediate-density lipoproteins (IDLs) Japan. to the progression of CAD and related clinical events (6). 5 Kumamoto Health Science University, Kumamoto, Japan. 6 Faculty of Medical and Pharmaceutical Sciences, Kumamoto University, Furthermore, a report from the Framingham Heart Study Kumamoto, Japan. found that remnant-like particle cholesterol (RLP-C), 7 Laboratory of Clinical Nutrition and Medicine, Kagawa Nutrition Uni- versity, Saitama, Japan. 8 Department of Cardiovascular Medicine, Osaka University Graduate School of Medicine, Osaka, Japan. 9 Nonstandard abbreviations: TRL, triglyceride-rich lipoprotein; CAD, * Address correspondence to this author at: Research and Development coronary artery disease; VLDLR, VLDL remnant; IDL, intermediate-density Department, Kyowa Medex Co., Ltd., 1-8-10, Harumi, Chuo-ku, Tokyo 104- lipoprotein; RLP-C, remnant-like particle cholesterol; apo, apolipoprotein; CM, 6004, Japan. Fax 81-3-6219-7614; e-mail [email protected]. chylomicron; LPL, lipoprotein lipase; CMR, CM remnant; TG, triglyceride; Received May 21, 2007; accepted September 5, 2007. PL-D, phospholipase D; CHER, cholesterol esterase; CHOD, cholesterol oxi- Previously published online at DOI: 10.1373/clinchem.2007.092296 dase; POE-POB, polyoxyethylene-polyoxybutylene block copolymer.

2128 Clinical Chemistry 53, No. 12, 2007 2129

which provides an estimate of TRL remnants, is an University School of Medicine and Tokyo Medical and independent risk factor for CAD in women (7). Dental University. We used hyperlipidemic sera for lipopro- TRL remnants are metabolic intermediates formed tein fractions isolated by ultracentrifugation to obtain VLDL when apolipoprotein (apo) B-48–containing chylomi- (0.96–1.006 kg/L), IDL (1.006–1.019 kg/L), LDL (1.019– crons (CMs) of intestinal origin and apo B-100–containing 1.063 kg/L), and HDL (1.063–1.021 kg/L) fractions VLDL of hepatic origin are hydrolyzed by lipoprotein (15, 21, 22). Blood was collected in plastic serum tubes lipase (LPL) on the surface of vascular endothelium; they (Venoject II; inner diameter 12 mm; length 10 cm; with are referred to as CM remnants (CMRs) and VLDLRs, procoagulant agent; Terumo). respectively (8). VLDLRs are found not only in TRL (d Ͻ 1.006 kg/L), but also in IDLs (1.006 to 1.019 kg/L). separation of lipoproteins by Therefore IDLs are also referred to as remnants (1). ultracentrifugation and gel filtration Compared with nascent VLDLs and CMs, TRL remnants The density of fasting serum obtained from a patient with and IDLs contain fewer triglycerides (TGs) and phospho- type III hyperlipoproteinemia (apo E phenotype E2/2) lipids, but relatively more esterified cholesterol and apo E was adjusted to 1.006 and 1.019 kg/L, and the treated (9–12). TRL remnants have been identified, isolated, and serum was ultracentrifuged at 131 918g for 20 h (Beckman: quantified in plasma on the basis of their density, charge, 50.4 Ti rotor). We collected the supernatant fraction after size, specific lipid components, apo composition, or apo tube-slicing with Pasteur pipette to obtain lipoprotein immunospecificity (1, 13). Major methods include ultra- fractions with densities of Ͻ1.006 kg/L and Ͻ1.019 kg/L. centrifugation for IDLs (14, 15); agarose gel electrophore- For separation of lipoprotein fractions, we used Superose sis for ␤-VLDLs and slow pre-␤ VLDLs and PAGE for 6 HR 10/30 columns (bed volume 24 mL; inner diameter IDLs (presence or absence of a midband) (11, 14, 16); and 10 mm; length 30 cm; count 2 columns; Pharmacia) immunoseparation for RLP-C as an unbound fraction not (23, 24). Each lipoprotein fraction was eluted with 0.15 adsorbed by anti–apo A-I or anti–apo B-100 antibodies mol/L NaCl and 1 mmol/L EDTA (pH 7.4) at a flow rate (17, 18). These methods require special equipment and of 0.3 mL/min. We collected the column effluent in are time-consuming; hence, they are not well suited as 0.5-mL fractions, in which total cholesterol concentration routine tests in hospitals and clinics. There is a need for a was measured by enzymatic assay (Kyowa Medex) and simple, rapid, automated method to quantify TRL rem- apo E concentration was measured by immunoturbidi- nants in serum or plasma. metric assay (Daiichi Pure Chemicals). Specific interaction of surfactants with lipoproteins has been successfully used for developing direct lipoprotein characterization by gel filtration analysis cholesterol assays, in which the surfactants are able to We added a mixture of POE-POB 12 g/L, CHER 1.5 kU/L, recognize differences in hydrated density, net charge, or and MOPS buffer 20 mmol/L, pH 6.5, as the reaction size of the various lipoprotein fractions (19, 20). We have reagent to the IDL, LDL, and HDL fractions obtained by therefore attempted to establish a new homogeneous ultracentrifugation at a 1:1 ratio, and mixtures were then assay for remnant lipoprotein cholesterol that can be incubated at 37 °C for 5 min. For gel filtration analysis carried out by using a universal autoanalyzer without any of the reaction mixture, we used Superose 6 HR 10/30 pretreatment. Herein we describe such an assay for rem- columns (23, 24). With each sample, 0.2 mL was ap- nant lipoprotein cholesterol by use of a synthetic nonionic plied with 0.15 mol/L NaCl and 1 mmol/L EDTA (pH surfactant and phospholipase D (PL-D). 7.4) at a flow rate of 0.5 mL/min, and 0.2-mL fractions were collected. Concentrations of total cholesterol and Materials and Methods TGs in the fractions were measured by enzymatic assay materials (Kyowa Medex), and the concentrations of apo E, apo We used cholesterol esterase (CHER) (EC3.1.1.34: 134 kDa, C-III, apo B, and apo A-I were measured by immunotur- Pseudomonas species) from Toyobo, cholesterol oxidase bidimetric assay (Daiichi Pure Chemicals). For gel filtra- (CHOD) (EC1.1.3.6: 58 kDa, recombinant Escherichia coli) tion analysis of serum samples obtained after feeding a from Kyowa Hakko, PL-D (EC3.1.4.4: 46 kDa, Streptomyces high-fat meal, we used TSK gel Lipopropak XL columns species) from Asahi Kasei, Good’s buffer MOPS from (inner diameter 7.8 mm; length 30 cm; count 2; Tosoh) Dojindo Laboratories, 4-aminoantipyrine from Kanto Chem- (24, 25). A postprandial serum sample was obtained from ical, and N-ethyl-N-(3-methylphenyl)-NЈ-succinylethylene- a single individual 10 h after a high-fat meal containing diamine manufactured by Kyowa Medex. Polyoxyethylene- 100 g of fat and 75 g of alcohol. Of the serum sample, polyoxybutylene block copolymer (POE-POB) was obtained 0.2 mL was subjected to gel filtration with 0.15 mol/mL from NOF Corporation, and the immunoseparation reagents NaCl and 1 mmol/mL EDTA (pH 7.4) at a flow rate of for RLP-C measurements were obtained from Otsuka Phar- 0.5 mL/min, and 0.2-mL fractions were collected. Concen- maceutical. Human serum samples were obtained from 160 trations of total cholesterol, TGs, apo E, apo C-III, and healthy volunteers (100 men and 60 women, age 20 to 59 apo B were measured as described above, and apo B-48 years) who had fasted for 12 h and from hyperlipidemic was measured by chemiluminescence enzyme immuno- patients with the approval of ethics committees at Osaka assay (Fujirebio). 2130 Miyauchi et al.: Assay for Remnant Lipoprotein Cholesterol

characterization by particle size distribution analysis For analysis of lipoprotein size distribution, we used a fiber optic dynamic light-scattering photometer (particle size range from 3 nm to 3 ␮m; FDLS-3000, Otsuka Electronics). The VLDL, IDL, LDL, and HDL fractions obtained by ultracentrifugation were mixed with the reaction reagent at a 1:1 ratio, incubated at 37 °C for 0 to 9 min, and diluted 100-fold for testing. analytical procedure For the measurement of remnant lipoprotein cholesterol, the final formulation of the 1st reagent was N-ethyl-N-(3- methylphenyl)-NЈ-succinylethylenediamine (1.1 mmol/ L), POE-POB (8 g/L), and MOPS buffer (20 mmol/L, pH 6.5); that of the 2nd reagent was PL-D (8 kU/L), CHER (1.5 kU/L), CHOD (3 kU/L), peroxidase (20 kU/L), 4-aminoantipyrine (2.5 mmol/L), and MOPS buffer (20 mmol/L, pH 6.8). Serum samples were stored at 4 °C after collection and Fig. 1. Gel filtration profile of the d Ͻ 1.019 (VLDLϩIDL) and d Ͻ 1.006 were subjected to testing (Hitachi-7170 autoanalyzer) (VLDL) fractions from a patient serum with type III hyperlipoproteinemia within 2 days, within which time the values obtained by fractionated by Superose 6 HR 10/30 columns. the assay were stable. To 3.8 ␮L of each serum sample, E, total cholesterol in d Ͻ1.019; F, total cholesterol in d Ͻ1.006; dotted line, IDL cholesterol calculated by difference. we added 180 ␮L of the 1st reagent and incubated the resulting solution at 37 °C for 5 min. Next, we added 60 ␮L of the 2nd reagent and incubated the reaction filtration profile of the lipoproteins from the patient mixture at 37 °C for 5 min. The chromophore formed in a serum. Whereas the main peak of the VLDL fraction is coupled reaction with peroxidase was measured spectro- ordinarily detected at void volume (24), that of the d Ͻ photometrically at dual wavelengths of 600 nm (main) 1.006 kg/L fraction from the patient was detected at void and 700 nm (subsidiary), and the end-point method was volume and at a smaller particle size. Compared with the used for the calculation. A serum-base calibrator prepared void volume fraction, the smaller peak contained more from TG-rich human sera (Kyowa Medex, 0.592 mmol/L) apo E. Because the density of IDL is 1.006 to 1.019 kg/L, was used to estimate remnant lipoprotein cholesterol. We IDL elution can be identified by subtracting the concen- performed the RLP-C immunoseparation method accord- tration of the d Ͻ 1.006 kg/L fraction from that of the d Ͻ ing to the package insert provided by Otsuka Pharma- 1.019 kg/L fraction (Fig. 1). The smaller VLDL particle ceutical, and serum samples were stored at 4 °C after size area of the gel filtration (d Ͻ 1.006 kg/L) mostly collection and tested within 2 days. For PAGE, Lipophor overlapped with the elution profile of IDL. In the present (Jokoh) was used. The total imprecision (CV) of the study, the fractions 14 through 19 from the d Ͻ 1.019 kg/L methods was as follows: total cholesterol Ͻ1.2%, TG fraction, as shown in Fig. 1, which were considered to Ͻ1.1%, apo A1 Ͻ5%, apo CIII Ͻ5%, apo E Ͻ5%, apo B contain VLDLR including IDL, were mixed to become the Ͻ5%, and apo B-48 Ͻ6.4%. The intraassay imprecision VLDLR fraction. (CV) of the immunoseparation method was Ͻ15%. We examined 180 surfactants with diverse structures, with which selectivity toward cholesterol in VLDLRs statistics was investigated. Of the surfactants tested, POE-POB of We used Pearson correlation coefficient analysis and 10 042 Da (see Supplemental Data Fig. 1), which has simple regression to assess the relation between the relatively high solubility as well as high selectivity to- proposed assay and the immunoseparation assay. Statis- ward cholesterol in VLDLRs, was selected for further tical analysis was performed with Excel 2003 (Microsoft) consideration. with the add-in software Statcel 2 (26). effects of poe-pob on gel filtration of Results lipoproteins and their size distribution separation of lipoproteins by Using gel filtration and size distribution studies, we ultracentrifugation and gel filtration and gained insight into the mechanism by which POE-POB selection of surfactant exhibits its selectivity toward VLDLRs. Fig. 2A shows the To obtain the VLDLR fraction from the serum of a patient profiles of cholesterol, TGs, apo B, apo E, and apo C-III with type III hyperlipoproteinemia, we conducted ultra- after applying the IDL fraction to gel filtration obtained centrifugation and gel filtration. Fig. 1 shows the gel by ultracentrifugation. Total cholesterol, TGs, apo B, Clinical Chemistry 53, No. 12, 2007 2131

apo E, and apo C-III were eluted at mostly the same retention time. Next, the IDL fraction was mixed with the reaction reagents at a 1:1 ratio, incubated at 37 °C for 5 min, and applied to gel filtration (Fig. 2B). When comparing the retention times of apos, apo B was eluted at approximately the same time as total cholesterol and TGs, but apo E and apo C-III were eluted with significant delay. When LDLs and HDLs were analyzed by the same methods, in contrast to IDLs, different apo peaks were not seen before and after the reaction reagent treatment. We examined the effect of POE-POB on the size distribution of lipoprotein fractions by use of a fiber optic dynamic light-scattering photometer (FDLS-3000), which quantifies particle size based on light scattering. As shown in Fig. 2C, the particle size of the IDL fraction decreased with time after mixing with the reaction reagent. By contrast, changes in particle sizes of the VLDL, LDL, and HDL fractions were relatively small, particularly with LDL and HDL.

improved reactivity to cmrs With the reagent containing POE-POB (8 g/L), which exhibited a high selectivity toward VLDLR in the presence of CHER (1.5 kU/L) and CHOD (3 kU/L), the correlation with the immunoseparation method for RLP-C was investigated using serum samples, including those with high TG concentrations. The results showed poor correlations with nonfast- ing samples having high TG concentrations. To clarify the cause of the differences, detailed reactivity was compared using lipoproteins isolated by gel filtration (24, 25). A serum sample collected from an individual after a high-fat meal was subjected to gel filtration. Fig. 3A shows gel filtration profiles of lipids and apos. Large CMs were eluted in the void volume (25), but apo B-48 and apo E were observed mostly in smaller particles, thus suggesting the presence of CMRs. Next, in the presence of CHER (1.5 kU/L) and CHOD (3 kU/L), the reactivity of reagent containing POE- POB (8 g/L) toward CMRs was examined, but the reactivity was very low in the region corresponding to CMRs. This low reactivity toward CMRs was thought to be the cause of the differences with the immunoseparation method, which detects all particles containing apo B-48. Therefore, PL-D (5 kU/L) was added to the reagent containing POE-POB (8 g/L) in the presence of CHER (1.5 kU/L) and CHOD (3 kU/L), and the reactivity of the fractions containing CMR significantly improved (Fig. 3A). As shown in Fig. 3B, the reagent containing POE-POB and PL-D in the presence of CHER and CHOD exhibited the required selectivity toward cholesterol in VLDLRs, including IDL, when a serum from a patient with type III hyperlipoproteinemia was used. The above results confirmed that the proposed assay made it possible to detect cholesterol in CMRs, VLDLRs, and IDLs.

Fig. 2. Effect of POE-POB on gel filtration profile of IDL fraction linearity study and lower limit of detection fractionated on Superose 6 HR 10/30 columns and its particle size Five serum samples were diluted with 155 mmol/L NaCl, distribution. and linearity of the method was examined. The linearity (A), IDL fraction without POE-POB. (B), IDL fraction with POE-POB. E, total cholesterol; F, TGs; ‚, apo B; Œ, apo E; Ⅺ, apo CIII. (C), particle size distribution was obtained up to 2.1 mmol/L cholesterol. The lower of IDL fractions before (0) and 3, 6, and 9 min after adding POE-POB. detection limit was 0.005 mmol/L. 2132 Miyauchi et al.: Assay for Remnant Lipoprotein Cholesterol

of the method was 2.75% at 0.111 mmol/L, 1.54% at 0.199 mmol/L, and 1.59% at 0.499 mmol/L.

effects of interfering substances To pooled serum samples, we added conjugated bilirubin (up to 0.68 mmol/L), free bilirubin (up to 0.68 mmol/L), hemoglobin (up to 4.9 g/L), and Intrafat (up to 10%) (Nihon Seiyaku Kogyo). None of these compounds produced more than a 5% error in the assay result.

comparison with the rlp-c immunoseparation method The proposed assay was compared with the immunoseparation method in fasting serum samples from 160 healthy volunteers. As shown in Fig. 4A, the regression equation for the comparison was y ϭ 1.018x– 0.01 (r ϭ 0.96). The mean (SD) of 160 sera was 0.18 (0.15) mmol/L for the immunoseparation method and 0.17 (0.16) mmol/L for the proposed method, respectively. The relation with the immunoseparation method was also examined by use of sera from diabetic patients, which showed that the proposed method occasionally exhibited higher concentrations of remnant lipoprotein cholesterol than the immunoseparation method (Fig. 4B). When these samples were analyzed by PAGE, a large amount of midband (14) was detected, indicating increased concentrations of IDLs.

Discussion Various methods have been developed to detect remnant lipoproteins in recent years. These methods are not well suited for use in regular laboratory practice because they are time-consuming and require special equipment. We used a special surfactant, POE-POB, and an enzyme, PL-D, to develop a homogenous assay that conveniently measures TRL remnant cholesterol in serum or plasma. The proposed assay does not require any sample pretreatment and uses 3.8 ␮L of serum, with the assay taking only 10 min on a regular autoanalyzer. The proposed assay measures cholesterol in CMRs, VLDLRs, and IDLs by selectively modifying these 3 remnant lipoproteins with POE-POB and PL-D in the presence of CHER and CHOD. Polyoxamers including POE-POP have been successfully used as surface modifi- ers for improving the stability of latex particles (27) and as Fig. 3. Gel filtration profiles of serum obtained after a high-fat meal in vehicles for transdermal drug delivery (28). Furthermore, the presence and absence of PL-D in reaction reagents (A) and those POE-POP, with a molecular weight of 3850 Da and a high of a patient serum with type III hyperlipoproteinemia (B). polyoxypropylene content, was found to be LDL selective, TSK gel Lipopropak XL columns and Superose 6 HR 10/30 columns were used in A and B, respectively. E, total cholesterol; ‚, cholesterol detected by reagent and it has been used in assays for LDL cholesterol (29). containing [CHER (1.5 kU/L) ϩ CHOD (3 kU/L) ϩ POE-POB (8 g/L)]; F, POE-POB, with a molecular weight of 10 042 Da, was cholesterol detected by reagent containing [CHER (1.5 kU/L) ϩ CHOD (3 kU/L) ϩ POE-POB (8 g/L) ϩ PL-D (8 kU/L)]; downward arrows, location of apos; upward selected for this proposed assay. When the IDL fraction arrow, improvement of reactivity of cholesterol in CMRs. (VLDLRs) incubated with POE-POB in the presence of CHER was applied to gel filtration, apo E and apo C-III imprecision were eluted from gel filtration columns with significant The within-run imprecision (CV) of the method was delay compared with the retention time of total choles- 1.25% at 0.100 mmol/L, 0.96% at 0.317 mmol/L, and terol, TGs, and apo B (Fig. 2, A and B). This result 1.8% at 0.412 mmol/L. The run-to-run imprecision (CV) indicates that apo E and apo C-III were removed from the Clinical Chemistry 53, No. 12, 2007 2133

size. In addition, with the LDL and HDL fractions, small changes of elution time after reaction reagent addition also suggested changes in particle shape, but apo B in LDLs and apo A-I and apo E in HDLs were not removed from the respective lipoprotein particles. Reaction reagent treatment also made IDLs smaller with time, thus clarifying that IDL modification was enhanced in the presence of CHER and POE-POB. The above findings suggested that, in the presence of CHER with POE-POB, apo E and apo C-III—which are relatively hydrophilic, of low molecular weight, and rich in VLDLR—were removed, and this phenomenon triggers the modification of VLDLR. Using postprandial sera with high TGs, there were substantial differences between the immunoseparation method and the reagent with POE-POB in the presence of CHER and CHOD. To determine the cause of discrepancies, postprandial serum was collected from an individual after a high-fat meal. Whereas large CMs were eluted in the void volume during gel filtration by use of a serum sample collected from this individual, lipoprotein fractions containing apo B-48 together with apo E and apo C-III were detected in smaller particles. Although CMs of varying size are secreted postprandially (1), the smaller particles include some CMRs, in which TGs and some phospholipids have been hydrolyzed by LPLs. In the presence of CHER and CHOD, the reactivity of POE-POB toward the fractions containing such CMRs was found to be very weak, and this reaction was evidently the cause of the discrepancies. Therefore, in an attempt to improve the reactivity toward CMRs, PL-D was added to the reagent with CHER, CHOD, and POE-POB. The results showed no changes in the reactivity to total cholesterol in lipoproteins other than the fractions containing CMRs, but the reactivity toward cholesterol in these fractions improved significantly. PL-D acts on the phosphodiester bond of the polar area of phospholipids to catalyze hydrolysis (30). In the absence of POE-POB, PL-D increased the reactivity of the assay for all lipoprotein fractions, probably through the hydrolysis of phospholipids on the surfaces of all the lipoprotein particles, thereby triggering favorable modification of the particles, allowing detection of its cholesterol Fig. 4. Correlation between the immunoseparation method and the in the presence of CHER and CHOD. On the other hand, proposed method. POE-POB interfered with the ability of lipoprotein cho-

(A), Bland-Altman difference plot between the immunoseparation method (Cis) lesterol, with the exception of TRL remnant cholesterol, to and the proposed method (Cp) in fasting serum samples from 160 healthy participate in the subsequent enzymatic reactions. There- volunteers. (B), Bland-Altman difference plot between Cis and Cp in sera from F diabetic patients. , samples in which the difference between Cis and Cp fore, the combination of PL-D and POE-POB seems to exceeded 0.17 mmol/L. increase the reactivity of the assay selectively for CMRs (Fig. 3A). lipoprotein surfaces in the presence of CHER with POE- Lipoproteins are essentially microemulsions in which a POB and formed isolated small particles. When treated hydrophobic core containing nonpolar lipids, such as TG with the reaction reagent, the elution time for IDL parti- and cholesterol esters, is surrounded by a monolayer of cles themselves was reduced. However, the size of IDLs apos and polar lipids, such as phospholipids and free treated with the reaction reagent became progressively cholesterol (31). Remnant lipoproteins are produced smaller with time (Fig. 2C), suggesting that the interaction when CM and VLDL lipids are hydrolyzed by LPLs. It has between the reaction reagent and IDL particles may affect been reported that hydrolysis of TGs by LPLs occurs at the shape of IDL particles, thus increasing their apparent the same time as removal of polar components on TRL 2134 Miyauchi et al.: Assay for Remnant Lipoprotein Cholesterol

surfaces (10). In other words, the surface structures of 4. Havel RJ. Postprandial hyperlipidemia and remnant lipoproteins. remnant lipoproteins differ from those of other lipopro- Curr Opin Lipidol 1994;5:102–9. teins and are thought to be modified by LPLs. Because 5. Tanaka A. Postprandial hyperlipidemia and atherosclerosis. there are no fundamental differences in the overall struc- J Atheroscler Thromb 2004;11:322–9. ture of CMs, VLDLs, CMRs, and VLDLRs, however, it has 6. Phillips NR, Waters D, Havel RJ. Plasma lipoproteins and progression of coronary artery disease evaluated by angiography and been difficult to differentiate these lipoproteins. By releas- clinical events. Circulation 1993;88:2762–70. ing apo E, apo C-III, and phospholipids from TRL rem- 7. McNamara JR, Shah PK, Nakajima K, Cupples LA, Wilson PW, nants, the reagent containing CHER, CHOD, POE-POB, Ordovas JM, et al. Remnant-like particle (RLP) cholesterol is an and PL-D may promote cholesterol reactivity specifically independent cardiovascular disease risk factor in women: results in these modified particles. from the Framingham Heart Study. Atherosclerosis 2001;154: Although a high correlation was observed between the 229–36. immunoseparation and proposed assay methods in 8. Eisenberg S. Remnant lipoprotein metabolism. In: Crepaldi G, healthy individuals, there were cases with discrepancies Tiengo A, Manzato E, eds. Diabetes, Obesity and Hyperlipidemia: among diabetic patients, who displayed increased con- V. The Plurimetabolic Syndrome. New York: Elsevier Science Publishers, 1993:7–14. centrations of IDLs. These sera were analyzed by Li- 9. Sata T, Havel RJ, Jones AL. Characterization of subfractions of pophor, in which IDLs were detected as a “midband” triglyceride-rich lipoproteins separated by gel chromatography (14). We found that such sera contained more IDLs rather from blood plasma of normolipemic and hyperlipemic humans. J than large VLDLs. This result suggested the possibility Lipid Res 1972;13:757–68. that the proposed assay can quantify IDLs with high 10. Mjøs OD, Faergeman O, Hamilton RL, Havel RJ. Characterization sensitivity. of remnants produced during the metabolism of triglyceride-rich There is a spectrum of sizes for CMRs and VLDLRs lipoproteins of blood plasma and intestinal lymph in the rat. J Clin depending on their degree of lipolysis. Therefore, further Invest 1975;56:603–15. studies need to be performed to determine the specificity 11. Pagnan A, Havel RJ, Kane JP, Kotite L. Characterization of human very low density lipoproteins containing two electrophoretic popu- of their assays to the various-sized remnant particles lations: double pre-beta lipoproteinemia and primary dysbetali- produced in the postprandial state. poproteinemia. J Lipid Res 1977;18:613–22. The above findings indicate that this method detects 12. Marcoux C, Tremblay M, Nakajima K, Davignon J, Cohn JS. not only TG-rich large particle size lipoproteins, such as Characterization of remnant-like particles isolated by immuno- CMRs, but also smaller remnants, such as VLDLRs and affinity gel from the plasma of type III and type IV hyperlipo- IDLs, with high sensitivity. Therefore, the proposed assay proteinemic patients. J Lipid Res 1999;40:636–47. can quantify the VLDLR and IDL fractions, which reflect 13. Cohn JS, Marcoux C, Davignon J. Detection, quantification, and CAD progression, as reported in the Montreal Heart characterization of potentially atherogenic triglyceride-rich rem- Study (6). Furthermore, the proposed assay does not nant lipoproteins. Arterioscler Thromb Vasc Biol 1999;19:2474– 86. require any sample pretreatment and can be performed in 14. Kameda K, Matsuzawa Y, Kubo M, Ishikawa K, Maejima I, a short period of time with the use of an autoanalyzer. It Yamamura T, et al. Increased frequency of lipoprotein disorders may therefore be useful for risk assessment of CAD and similar to type III hyperlipoproteinemia in survivors of myocardial the diagnosis of type III hyperlipoproteinemia and other infarction in Japan. Atherosclerosis 1984;51:241–9. dyslipidemias characterized by accumulation of TRL 15. Tatami R, Mabuchi H, Ueda K, Ueda R, Haba T, Kametani T, et al. remnants. Intermediate-density lipoprotein and cholesterol-rich very low density lipoprotein in angiographically determined coronary artery disease. Circulation 1981;64:1174–84. Grant/funding support: This study was supported by 16. Carlson LA, Ericsson M. Quantitative and qualitative serum li- Kyowa Medex Co., Ltd. poprotein analysis. Part 1. Studies in healthy men and women. Atherosclerosis 1975;21:417–33. Financial disclosures: None declared. 17. Nakajima K, Saito T, Tamura A, Suzuki M, Nakano T, Adachi M, Acknowledgments: We thank Dr. Richard Havel (University et al. Cholesterol in remnant-like lipoproteins in human serum of California, San Francisco, Medical Center) and Dr. G. using monoclonal anti apo B-100 and anti apo A-I immunoaffinity Russell Warnick (Berkeley HeartLab, Inc.) for a critical re- mixed gels. Clin Chim Acta 1993;223:53–71. view of an earlier draft of this manuscript. 18. Nakajima K, Okazaki M, Tanaka A, Pullinger C, Wang T, Nakano T, et al. Separation and determination of remnant-like particles in References human serum using monoclonal antibodies to apoB-100 and 1. Havel RJ. Determination and clinical significance of triglyceride- apoA-I. J Clin Ligand Assay 1996;19:177–83. rich lipoprotein remnants. In: Rifai N, Warnick GR, Dominiczak MH, 19. Nauck M, Wiebe D, Warnick GR. Measurement of high-density eds. Handbook of Lipoprotein Testing, 2nd ed. Washington, DC: lipoprotein cholesterol. In: Rifai N, Warnick GR, Dominiczak MH, American Association of Clinical Chemistry Press, 2000:565–80. eds. Handbook of Lipoprotein Testing, 2nd ed. Washington, DC: 2. Zilversmit DB. Atherosclerosis: a postprandial phenomenon. Cir- American Association of Clinical Chemistry Press, 2000:221–44. culation 1979;60:473–85. 20. Bachorik PS. Measurement of low-density lipoprotein cholesterol. 3. Karpe F, Steiner G, Uffelman K, Olivecrona T, Hamsten A. Post- In: Rifai N, Warnick GR, Dominiczak MH, eds. Handbook of prandial lipoproteins and progression of coronary atherosclerosis. Lipoprotein Testing, 2nd ed. Washington, DC: American Associa- Atherosclerosis 1994;106:83–97. tion of Clinical Chemistry Press, 2000:245–64. Clinical Chemistry 53, No. 12, 2007 2135

21. Havel RJ, Eder HA, Bragdon JH. The distribution and chemical 26. Yanai H. Statcel: The useful add-in software forms on Excel, composition of ultracentrifugally separated lipoproteins in human 2nd ed. Tokyo: OMS, 2004. serum. J Clin Invest 1955;34:1345–53. 27. Juha´sz J, Lenaerts V, Bellemare M, Pimienta C, Ong H. Stability 22. Hatch FT, Lees RS. Practical methods for plasma lipoprotein and absorption to polymeric surfaces of rat ANF in poloxamer 407 analysis. Adv Lipid Res 1968;6:1–68. solutions. Int J Pharm 1991;77:309–13. 28. Cappel MJ, Kreuter J. Effect of nonionic surfactants on transder- 23. Kieft KA, Bocan TM, Krause BR. Rapid online determination of mal drug delivery: II. Poloxamer and poloxamine surfactants. Int cholesterol distribution among plasma lipoproteins after high- J Pharm 1991;69:155–67. performance gel filtration chromatography. J Lipid Res 1991;32: 29. Sugiuchi H, Irie T, Uji Y, Ueno T, Chaen T, Uekama K, et al. 859–66. Homogeneous assay for measuring low-density lipoprotein choles- 24. Okazaki M, Usui S, Hosaki S. Analysis of plasma lipoproteins by terol in serum with triblock copolymer and ␣-cyclodextrin sulfate. gel permeation chromatography. In: Rifai N, Warnick GR, Dominic- Clin Chem 1998;44:522–31. zak MH, eds. Handbook of Lipoprotein Testing, 2nd ed. Washing- 30. Takayama M, Itoh S, Nagasaki T, Tanimizu I. A new enzymatic ton, DC: American Association of Clinical Chemistry Press, 2000: method for determination of serum choline-containing phospho- 647–69. lipids. Clin Chim Acta 1977;79:93–8. 25. Kitamura T, Ito S, Moriyama H, Kato Y, Sasamoto K, Okazaki M. 31. Havel RJ. Origin, metabolic fate, and metabolic function of plasma Quantitative analysis of serum lipoproteins (CM, VLDL, LDL and lipoproteins. In: Steinberg D, JM Olefsky, eds. Contemporary HDL) by high-performance gel filtration chromatography. Chroma- Issues in Endocrinology and Metabolism. Vol. 3. New York: tography 1996;17:33–7. Churchill Livingstone, 1987:117–41. Clinical Chemistry 53:12 2136–2143 (2007) Drug Monitoring and Toxicology

Buprenorphine and Norbuprenorphine in Hair of Pregnant Women and Their Infants after Controlled Buprenorphine Administration

Robert S. Goodwin,1 Diana G. Wilkins,2 Olga Averin,2 Robin E. Choo,1,3 Jennifer R. Schroeder,4 Donald R. Jasinski,5 Rolley E. Johnson,6 Hendre´e E. Jones,7 and Marilyn A. Huestis1*

Background: Buprenorphine is under investigation as a Results: Buprenorphine concentrations were signifi- ؍ pharmacotherapeutic agent for treating opioid depen- cantly greater in unwashed hair than washed hair (P dence in pregnant women. We hypothesized that there 0.031). Norbuprenorphine concentrations were signifi- would be a relationship between the cumulative mater- cantly greater than buprenorphine concentrations in .(0.0033 ؍ and infant hair (P (0.0097 ؍ nal dose of buprenorphine during pregnancy and the both maternal (P concentration of buprenorphine and norbuprenorphine There were statistically significant associations between in maternal and infant hair. the cumulative maternal dose of buprenorphine admin- Methods: This study examined buprenorphine and nor- istered and the concentrations of buprenorphine ؍ buprenorphine concentrations in hair obtained from 9 (washed, P <0.0001; unwashed, P 0.0004), norbu- ؍ buprenorphine-maintained pregnant women and 4 of prenorphine (washed, P <0.0001; unwashed, P their infants. Specimens were analyzed by liquid chro- 0.0005), and buprenorphine plus norbuprenorphine ؍ matography-tandem mass spectrometry with limits of (washed, P <0.0001; unwashed, P 0.0005) for both quantification of 3.0 pg/mg. All maternal hair specimens washed and unwashed maternal hair specimens. There was a significant positive association between concen- were washed with methylene chloride before analysis, trations of buprenorphine and norbuprenorphine in and when sufficient amounts of maternal hair were ؍ maternal hair (washed, P <0.0001; unwashed, P available, specimens also were analyzed without wash- ؍ a trend for this association in infant hair (P ,(0.0003 ing. Infant hair specimens were not washed. 0.08), and an association between buprenorphine concentrations in maternal unwashed hair and infant hair -The buprenorphine:norbuprenorphine ra .(0.0002 ؍ P)

1 Chemistry and Drug Metabolism Section, National Institute on Drug tio increased in distal segments. Abuse–Intramural Research Program, National Institutes of Health, Baltimore, Conclusion: Buprenorphine treatment during gestation MD. provides an opportunity for monitoring drug disposi- 2 Center for Human Toxicology, Department of Pharmacology and Toxi- tion in maternal and fetal tissues under controlled cology, University of Utah, Salt Lake City, UT. 3 Department of Biology, University of Pittsburgh at Titusville, Titusville, conditions. PA. © 2007 American Association for Clinical Chemistry 4 Office of the Clinical Director, National Institute on Drug Abuse– Intramural Research Program, National Institutes of Health, Baltimore, MD. Illicit drugs or alcohol were used by 13% of pregnant 5 Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD. women in the US in 1999 and 2000, with opiates account- 6 Reckitt Benckiser Pharmaceuticals, Inc., Richmond, VA. ing for 19% of the abused substances (1). Buprenorphine, 7 Department of Psychiatry and Behavioral Sciences, Johns Hopkins Uni- a partial mu opioid agonist and kappa antagonist ap- versity School of Medicine, Johns Hopkins Bayview Medical Center, Baltimore, MD. proved in the US for treatment of nonpregnant opioid- * Address correspondence to this author at: Chief, Chemistry and Drug dependent adults, may decrease the incidence and/or Metabolism, National Institute on Drug Abuse–Intramural Research Program, severity of neonatal abstinence syndrome often observed National Institutes of Health, 5500 Nathan Shock Dr., Baltimore, MD 21224. after prenatal exposure to full mu agonists (2). Although Fax 410-550-2971; e-mail [email protected]. Received May 7, 2007; accepted September 18, 2007. buprenorphine is not yet approved in the US for use Previously published online at DOI: 10.1373/clinchem.2007.091413 during pregnancy, physicians are prescribing it to preg-

2136 Clinical Chemistry 53, No. 12, 2007 2137

nant patients because the medication can be taken at contrast to previous reports, norbuprenorphine was de- home and also has a favorable safety profile (1). tected in higher concentrations than buprenorphine. For ethical and safety reasons, it is important to limit Vincent et al. (15) also reported higher norbuprenor- exposure to medications during pregnancy because of phine concentrations in hair of 4 of 5 participants receiv- potential harm to the developing fetus. Untreated chronic ing buprenorphine. Hair specimens were prepared ac- illnesses, however, such as opioid dependence, are asso- cording to the Kintz et al. (12) solid-phase extraction and ciated with increased morbidity and mortality. Buprenor- GC-MS methods. Cirimele et al. (16) attempted to resolve phine treatment during gestation is similar to methadone the discrepancy by analyzing the wash used to decontam- in benefits for mother and child (3). Although benefits inate specimens. In 40 of 66 washes, buprenorphine was during pregnancy appear to outweigh risks, accumulating present in higher quantities than norbuprenorphine, sug- data regarding dose-concentration relationships are criti- gesting that the former was preferentially lost during cal to understanding how this medication impacts mater- washing. nal and neonatal outcomes. Several additional aspects of buprenorphine therapy Scheidweiler et al. (4) demonstrated the dose-related and pharmacokinetics remain unclear. There are no data distribution of codeine, cocaine, and metabolites into relating the dose of buprenorphine administered in preg- human hair. However, the disposition of illicit drugs into nancy to concentration of drug in maternal and fetal maternal and fetal hair cannot be rigorously studied in tissues and fluids. No studies have examined the associ- pregnant women because of ethical and safety issues. ation between cumulative buprenorphine dose and drug Buprenorphine treatment during pregnancy presents an concentrations in hair. To evaluate the ability of hair opportunity for studying the excretion of buprenorphine analysis to predict magnitude of gestational drug expo- and its primary metabolite, norbuprenorphine, into ma- sure, we designed a study to investigate the relationship ternal and fetal hair under controlled conditions, and for between dose of buprenorphine administered to 9 preg- evaluating dose-concentration relationships in pregnant nant women and neonatal and maternal buprenorphine and norbuprenorphine concentrations in washed and women and their infants. unwashed hair. Neonatal hair analysis has been evaluated for monitoring gestational exposure to drugs (5, 6). Hair analysis has Materials and Methods advantages over blood and urine testing because expo- study participants sure can be monitored over a period of several months We recruited study participants from heroin-dependent (7, 8). Neonatal hair reflects drug exposure during the 3rd pregnant women participating in the Center for Addiction trimester, when hair begins to be formed, and remains and Pregnancy’s multidisciplinary treatment program be- positive for the drug under investigation for up to 3 tween May 2000 and March 2003. Hair specimens from 4 months after birth (6). participants’ infants also were included. The remaining Buprenorphine has an elimination half-life of about infants were not included because they did not have 37 h (9) and is metabolized by N-dealkylation to norbu- adequate hair or because their mothers did not agree to prenorphine, which has an even longer elimination rate inclusion. The study protocol was approved by the Johns (10). Kintz (11) quantified buprenorphine and norbu- Hopkins Bayview Medical Center and National Institute prenorphine concentrations in hair from 3 individuals on Drug Abuse Intramural Research Program Institu- who used buprenorphine daily. Higher concentrations of tional Review Boards. Adult participants provided writ- parent compound than metabolite were observed in all 3 ten informed consent for themselves and for infants. individuals. In a subsequent study of 14 buprenorphine Inclusion criteria for pregnant women were age 21–40 users, Kintz et al. (12) found parent compound in the hair years; estimated gestational age by sonogram of 16–30 of all participants, but norbuprenorphine in only 11. weeks; DSM-IV (Diagnostic and Statistical Manual of Mental Additionally, buprenorphine was present at concentra- Disorders, 4th Edition) diagnosis of current heroin depen- tions approximately 2 to 18 times higher than metabolite. dence; request for maintenance pharmacotherapy; recent Maximum buprenorphine and norbuprenorphine concen- self-reported heroin use (use on more than 4 of the past 7 trations were 0.59 and 0.15 ng/mg, respectively. days); and an opiate-positive urine specimen at intake. Tracqui et al. (13) analyzed hair from 6 participants Exclusion criteria were urine positive for undocumented treated with oral buprenorphine in a detoxification pro- methadone at enrollment, current DSM-IV diagnosis of gram. Concentrations ranged from 4 to 140 pg/mg and 0 alcohol abuse or dependence, self-reported regular use of to 67 pg/mg for buprenorphine and norbuprenorphine, benzodiazepines, current medication use for another Axis respectively. In all cases, buprenorphine exceeded norbu- I disorder, presence of a serious concurrent medical prenorphine concentrations. illness, diagnosis of preterm labor, evidence of fetal mal- Wilkins et al. (14) reported buprenorphine hair conformation, positive HIV test, or positive sickle cell trait centrations for 12 participants receiving 8 mg of bu- (2). Nearly 1500 women were screened, and only 57 prenorphine daily for up to 180 days. Buprenorphine or qualified for this phase II study. Because of the experi- norbuprenorphine was detected in 11 of 12 individuals. In mental nature of the drug intervention, treatment was not 2138 Goodwin et al.: Buprenorphine and Norbuprenorphine in Hair

permissible until later gestational ages. Strict inclusion weeks). At the time of delivery, mean (SD) buprenorphine and exclusion criteria were necessary for safety reasons, dose was 18.7 (3.5) mg/d (range 14–24 mg/d) (2). especially given the limited knowledge about buprenorphine during pregnancy. Of 57 consenting women, 30 hair specimens were subsequently randomized to study treatment; 15 Drug disposition in hair occurs from multiple sources received methadone, and 15 received buprenorphine. including blood, sweat, and sebum. In an attempt to Nine of 15 women completed the buprenorphine arm of account for total disposition of drug in hair from all the study. The final sample size enrolled in treatment at sources, the entire hair specimen was analyzed. Maternal delivery was 9 women (8 African-American and 1 white) hair specimens were obtained from the posterior vertex (Table 1). The mean (SD) age of the participants was 30 approximately every 4 weeks. The 1st specimen analyzed (3.4) years (range 22–32 years). This study was part of a from participants A–F (see Tables 2 and 4, and Supple- previously published larger trial (2). mental Data Table 1 that accompanies the online version of this article at http://www.clinchem.org/content/ urine drug screening at intake and during vol53/issue12) was obtained before dosing. The next buprenorphine maintenance specimen for these individuals and the 1st specimen for The Dade Behring enzyme-multiplied immunoassay tech- participants G–I were obtained, on average, 21 days after nique (Emit) was used at enrollment to screen participants dosing was initiated. The specimens (n ϭ 51) weighed for opiates, cocaine, benzodiazepines, cannabinoids, and 3.0–34.9 mg and represented up to 15 months hair methadone. During buprenorphine maintenance, urine growth. Hair specimens were cut into 3-cm segments and specimens were collected 3 times per week and assayed minced with scissors. Individual hair segments (n ϭ 148) for the above drug classes. One specimen, selected weekly weighed 2.7–10.3 mg. Infant hair specimens (n ϭ 4) were at random, was sent for assay for amphetamines, canna- collected below the crown of the head within 48 h of birth. binoids by immunoassay and for antidepressants, barbi- Aliquots of each hair segment were placed into si- turates, opioids, phencyclidine, and other drug testing by lanized glass vials. When sufficient specimen was avail- thin-layer chromatography. Breath samples (Alco-Sensor able (n ϭ 18), 2 aliquots of each segment were prepared to III, Alcopro) were randomly obtained once weekly for compare quantitative data for washed vs unwashed hair. monitoring alcohol ingestion. Washing was performed with methylene chloride as per Cirimele et al. (16). Briefly, hair segments were washed dosing twice with 2 mL of methylene chloride for 2 min at room The mean (SD) initial buprenorphine dose was 8.2 (2.1) temperature and allowed to evaporate to dryness. The mg/d (range 4–12 mg/d). Buprenorphine was adminis- 2nd aliquot, if available, was not washed before analysis. tered sublingually with observation by research staff. Infant hair specimens were analyzed unwashed because Double-blind changes in medication dose were made there was insufficient hair to test washed and unwashed, through protocol-driven clinical decision criteria. Dose and it was not known if washing would remove excess changes were limited to once every 2 weeks unless amounts of drug, owing to the fine texture of hair. clinically indicated. To maintain blinding for research staff, dose changes were known only to pharmacy staff. digestion, extraction, and analysis by mass Participants received 4–24 mg daily sublingual buprenor- spectrometry phine during pregnancy and up to 10 weeks after delivery Buprenorphine-d4 and norbuprenorphine-d3 (1 ng/mg for a mean (SD) period of 21.1 (4.8) weeks (range 14–30 each) were added as internal standards to 3.5–10.1 mg of adult or 1.6–5.3 mg of infant hair. Hair was digested (solubilized) in 2 mol/L NaOH overnight at room tem- Table 1. Patient demographic information.ª perature. Specimen pH was adjusted to 10.5 with 6 mol/L Patient identification Ethnicity Hair color HCl, and extracted with n-butyl chloride/acetonitrile/ A African-American Black ethylacetate (4:1:1 vol/vol). Chromatographic separation ϫ B African-American Dark brown was achieved on a YMC ODS-AQ 2.0 150 mm S-3 120 Å C African-American Dark brown/black reversed-phased HPLC column. The liquid chromatogra- D African-American Dark brown/black phy–tandem mass spectrometer was operated in selected E African-American Black reaction monitoring mode with electrospray ionization. F African-American Dark brown/black Four selected reactions were monitored (m/z): 468.23 G White Blond 396.1 for buprenorphine, 472.23400.1 for buprenorphine- 3 3 H African-American Black d4, 414.2 100.9 for norbuprenorphine, and 417.2 100.9 I African-American Black for norbuprenorphine-d3. Analyte concentrations were ª Pregnant women were maintained on sublingual buprenorphine during preg- determined from peak area of native drug to peak area of nancy and up to 10 weeks after delivery for an average of 21.1 (4.8) weeks deuterated internal standard, and ratio comparison to the (range 14–30 weeks). Mean age of participants was 30 (3.4) years (range calibration curve generated from analysis of human hair 22–32 years). fortified with known buprenorphine and norbuprenor- Clinical Chemistry 53, No. 12, 2007 2139

Table 2. Cumulative dose of buprenorphine administered (mg), analyte concentrations (pg/mg), and analyte ratios in washed and unwashed hair specimens of 9 pregnant women maintained on buprenorphine.a BUPb in NBUPc in BUP in NBUP in Ratio of BUP: Ratio of BUP: Cumulative washed hair, washed hair, unwashed hair, unwashed hair, NBUP in NBUP in Patient dose, mg pg/mg pg/mg pg/mg pg/mg washed hair unwashed hair A0NDd ND 246 ND ND 682 3.6 56.4 0.064 1234 16.4 262.2 24.1 240.0 0.063 0.101 1774 18.9 301.7 24.3 421.6 0.063 0.058 B 0 ND ND 118 ND ND ND ND 478 2.8 39.0 0.073 1006 7.6 102.2 13.7 158.1 0.075 0.087 1418 10.3 134.5 0.076 2174 15.0 180.5 0.083 C 0 ND ND 324 3.5 12.8 0.276 744 26.0 424.5 0.061 1302 27.6 381.5 0.072 1590 33.1 458.5 0.072 2454 83.6 704.9 111.0 846.5 0.119 0.131 D 0 ND ND 270 10.4 186.4 11.2 185.9 0.056 0.060 918 51.7 1230.8 0.042 1292 24.7 571.4 0.043 1724 30.6 714.6 47.0 1218.9 0.043 0.039 1886 32.0 760.8 70.8 2018.6 0.042 0.035 2210 88.9 1220.5 0.073 2808 143.8 1583.4 0.091 E 0 ND ND 414 7.0 75.2 0.094 970 29.9 312.0 0.096 1242 11.1 124.1 18.6 199.0 0.090 0.094 F 0 0.9 ND 256 20.3 31.7 22.9 46.5 0.640 0.493 794 33.9 245.5 17.9 177.7 0.138 0.101 1416 60.3 392.8 60.9 499.8 0.153 0.122 2040 63.4 263.2 0.241 2280 44.4 374.6 0.118 G18NDND 386 13.0 ND 870 24.1 12.5 1.933 1544 31.9 21.3 1.503 2292 40.5 26.3 1.541 H 176 33.4 33.7 0.990 568 4.0 66.1 0.061 934 20.2 181.1 0.111 1590 31.5 330.9 35.6 336.7 0.095 0.106 2044 28.2 317.2 36.4 334.2 0.089 0.109 2458 44.1 511.9 59.7 765.7 0.086 0.078 I 532 ND 6.2 0.9 10.8 0.000 0.088 728 2.6 36.8 3.3 48.4 0.070 0.068 1302 50.0 589.7 0.085 1400 14.1 94.2 0.150 1636 26.8 422.8 73.8 1174.7 0.065 a Hair was collected at approximately 4-week intervals and analyzed after dividing into 3-cm segments and washing with methylene chloride. Concentrations listed are of the entire hair specimen, i.e., all segments combined. If sufficient specimen was available, hair was analyzed unwashed as well. The 1st specimen analyzed from individuals A–F was obtained prior to the beginning of dosing. The next specimen for each of these study participants and the 1st specimen for individuals G–I were obtained, on average, 21 days after dosing was initiated. b Buprenorphine (BUP) concentration (pg/mg) of entire hair specimen. c Norbuprenorphine (NBUP) concentration (pg/mg) of entire hair specimen. d ND, None detected. 2140 Goodwin et al.: Buprenorphine and Norbuprenorphine in Hair

phine and internal standard concentrations. Hair calibra- measures linear regression with norbuprenorphine con- tors extracted in each batch were prepared by adding centration as the dependent variable and buprenorphine drug-free human hair at concentrations from 3 to 10 000 concentration as the independent variable. pg/mg. QC hair samples to assess accuracy and precision To determine whether the ratio of buprenorphine: of each batch also were included at 25, 100, and 1000 norbuprenorphine in maternal hair changed according to pg/mg. Intra- and interassay imprecision were within segment, a repeated-measures linear regression model 11% for buprenorphine and norbuprenorphine at these was fit with ratio as the dependent variable and hair concentrations. QC samples fortified only with buprenor- segment as the independent variable. Segment was coded phine demonstrated no conversion to norbuprenorphine as a class variable to do pairwise comparisons, i.e., during digestion and extraction. Limits of quantification comparing each segment to the 1st hair segment with P were 3.0 pg/mg of hair for buprenorphine and values adjusted using the Dunnett-Hsu adjustment to norbuprenorphine. keep overall alpha at 0.05. Analysis of the relationship between concentrations of statistical analysis maternal buprenorphine, norbuprenorphine, or norbu- Analyses were done using Statistical Analysis Systems prenorphine plus buprenorphine in hair and concentra- (SAS) version 9 (SAS Institute). A Spearman correlation tions in infant hair was performed with repeated mea- matrix was calculated to determine whether a correlation sures linear regression with maternal hair concentration existed between cumulative maternal dose and buprenor- as dependent variable and infant concentration as the phine or norbuprenorphine concentration in infant hair. independent variable. The cumulative buprenorphine dose administered to the mother immediately before birth was used in each case. Results Most analyses were performed using repeated mea- Of 45 washed maternal hair specimens obtained after sures linear regression (SAS Proc Mixed), making best use initiation of dosing, 41 were positive for buprenorphine of available data. To test hypotheses of equality (compar- [30.1 (27.3) pg/mg, range 0.8–143.8 pg/mg] and norbu- ing analyte concentrations in washed vs unwashed hair, prenorphine [336.5 (359.5) pg/mg, range 6.2–1583.4 pg/ comparing buprenorphine vs norbuprenorphine concen- mg] in at least 1 of the hair segments tested (Table 2). Of trations, comparing analyte ratios in maternal vs infant the 18 unwashed specimens collected after dosing began, hair), intercept-only models were used, with difference 17 were positive for buprenorphine [40.4 (28.7) pg/mg, score (washed minus unwashed, maternal minus infant) range 0.9–111.0 pg/mg] and norbuprenorphine [563.1 as the dependent variable. If the intercept was signifi- (500.3) pg/mg, range 10.8–2018.6 pg/mg]. Concentra- cantly different from zero, then a difference was consid- tions of buprenorphine and norbuprenorphine were ered to exist. greater in unwashed hair: mean difference was 10.8 (15.6) To determine whether there were associations between pg/mg for buprenorphine (n ϭ 16 pairs, t ϭϪ2.70, df ϭ the cumulative maternal dose of buprenorphine and 7, P ϭ 0.031) and 190.0 (345.7) pg/mg for norbuprenor- analyte concentrations, the cumulative dose was the in- phine (n ϭ 17 pairs, t ϭϪ2.16, df ϭ 7, P ϭ 0.067). Ratios dependent variable in repeated measures linear regres- of buprenorphine and norbuprenorphine in washed [0.24 sion having as dependent variables total buprenorphine, (0.44) pg/mg, range 0.04–1.93 pg/mg] and unwashed total norbuprenorphine, and total buprenorphine plus specimens [0.10 (0.10) pg/mg, range 0.04–0.49] were not norbuprenorphine, for washed and unwashed hair. This significantly different (t ϭ 0.27, df ϭ 6, P ϭ 0.80). All examination was limited to maternal data only (n ϭ 9). infant hair specimens were positive for buprenorphine Associations between concentrations of buprenorphine and norbuprenorphine (Table 3). and norbuprenorphine in maternal (washed and un- Norbuprenorphine concentrations were significantly washed) and infant hair were assessed with repeated greater than buprenorphine concentrations in maternal

Table 3. Cumulative dose of buprenorphine administered, buprenorphine and norbuprenorphine concentrations, and buprenorphine:norbuprenorphine ratios in hair of 4 infants whose mothers were maintained on buprenorphine.a Cumulativeb maternal dose of BUP,c NBUP,d BUP؉NBUP, Ratio of Infant buprenorphine, mg pg/mg pg/mg pg/mg BUP:NBUP E 1242 36.8 732.2 769.0 0.050 F 2280 62.0 793.3 855.3 0.078 H 1590 37.7 579.9 617.6 0.065 I 1400 82.1 1037.1 1119.2 0.079 a Pregnant women were maintained on buprenorphine for up to 23 weeks during gestation. Infant hair was collected within 48 h of birth. b The cumulative maternal dose was the total dose of buprenorphine administered to the mother immediately prior to the birth of the infant. c Buprenorphine (BUP) concentration (pg/mg) of entire infant hair specimen. d Norbuprenorphine (NBUP) concentration (pg/mg) of entire infant hair specimen. Clinical Chemistry 53, No. 12, 2007 2141

hair (washed, P ϭ 0.0131; unwashed, P ϭ 0.0097) and 4 infants born to mothers maintained on buprenorphine infant hair (P ϭ 0.0033) (Tables 2 and 3). Only patient G, during the 3rd trimester, all had hair specimens positive a participant with blond hair, had buprenorphine concen- for buprenorphine and norbuprenorphine. Accordingly, trations exceeding those of norbuprenorphine. This indi- the presence of buprenorphine or norbuprenorphine in vidual had buprenorphine concentrations similar to dark- hair can be used to document buprenorphine exposure in haired participants but much lower norbuprenorphine pregnant women and their infants. Also, concentrations of concentrations. Results from infant hair analyses should buprenorphine and norbuprenorphine in maternal hair be considered tentative because these data were based on estimate the degree of exposure during gestation on an only 4 observations. individual basis. There were significant positive associations between A small percentage (5.6%–8.9%) of washed and un- cumulative maternal buprenorphine dose and total con- washed hair specimens were negative for either bu- centrations of buprenorphine (washed, P Ͻ0.0001; un- prenorphine or norbuprenorphine. Generally, these sam- washed, P ϭ 0.0004), norbuprenorphine (washed, P ples were the 1st or 2nd specimens collected after dosing Ͻ0.0001; unwashed, P ϭ 0.0005), and buprenorphine plus began. Some contributing factors include the fact that the norbuprenorphine (washed, P Ͻ0.0001; unwashed, P ϭ hair root and follicle beneath the scalp may contain 0.0005) in whole hair specimens (all segments included), additional drug, but these concentrations were not mea- both washed and unwashed (Table 2). These examina- sured, and if low concentrations of drugs were present, tions were limited to maternal data only (n ϭ 9). Signifi- they may have been below the method limits of quantifi- cant relationships between cumulative maternal bu- cation of 3 pg/mg. Drug incorporated into the hair root prenorphine dose and analyte concentrations also were from blood generally does not reach the level of the scalp observed for several hair segments in washed hair (see for 7 to 10 days; however, drug present in sweat may be Supplemental Data Table 1) and for the 1st segments from deposited on hair within minutes to hours. Additionally, unwashed hair specimens. maternal plasma concentrations of buprenorphine and There was a statistically significant association be- norbuprenorphine are unknown and may have varied tween concentrations of buprenorphine and norbuprenor- considerably within and between individuals, and the phine in maternal hair (washed, P Ͻ0.0001; unwashed, minimum detectable dose in hair has not been P ϭ 0.0003) and a trend for the association in infant hair established. (P ϭ 0.08). A quantitative relationship between cumulative mater- The buprenorphine:norbuprenorphine ratio increased nal dose of buprenorphine and concentration of bu- as hair segments became more distal (P Ͻ0.0001) (see prenorphine or norbuprenorphine in infant hair was not Table 4 and Supplemental Data Table 1). Ratios in seg- established, possibly because of small sample size (n ϭ 4). ments 4 and 5 were significantly greater than in segment Other factors were that infant hair forms during the 3rd 1 (adjusted P Ͻ0.01), although segments 2 and 3 were not trimester and as a result does not reflect the entire (adjusted P Ͼ0.05). Stratifying the data into homogeneous buprenorphine administration period, and fetal hair is groups with respect to same number of hair segments exposed to drug in amniotic fluid. (e.g., 2 segments, 3 segments) showed that the greatest Norbuprenorphine concentrations were significantly increase was from segment 1 to segment 2 (the ratio more greater than buprenorphine concentrations in both mater- than doubled), then leveled off for subsequent segments nal and infant hair. This finding is consistent with the (2 to 3, 3 to 4, etc.). work of Wilkins et al. (14) and Vincent et al. (15), but There was no correlation between cumulative maternal contrasts with the results of Kintz (11) and Tracqui et al. dose and buprenorphine or norbuprenorphine infant hair (13). Cirimele et al. (16) attributed this discrepancy to loss concentrations. There was an association between bu- of buprenorphine to a greater extent during methylene prenorphine concentrations in unwashed maternal hair chloride washing. A limitation of the present study was and infant hair (P ϭ 0.0002); however, this association was that the discarded methylene chloride wash was not not present for norbuprenorphine. analyzed for the presence of buprenorphine or norbuprenorphine. Although concentrations of buprenorphine Discussion and norbuprenorphine were higher in unwashed hair in This study demonstrated a significant relationship be- this study, the explanation of Cirimele et al. (16) cannot tween maternal buprenorphine dose and buprenorphine account for greater concentrations of the metabolite, be- and norbuprenorphine concentrations in pregnant wom- cause statistically significant greater concentrations of en’s hair. Cumulative maternal buprenorphine dose was norbuprenorphine than buprenorphine were found in positively associated with concentrations of buprenor- both washed and unwashed maternal hair. Additionally, phine, norbuprenorphine, and buprenorphine plus nor- in the present study, there was no statistically significant buprenorphine in maternal hair. This relationship was difference between ratios of buprenorphine:norbuprenor- found for whole hair specimens (washed and unwashed) phine in washed compared to unwashed hair. Further- as well as with several individual hair segments for more, Kintz (11) and Tracqui et al. (13) used a methylene washed hair and the 1st segment of unwashed hair. Of the chloride wash before analysis, as was done in this inves- 2142 Goodwin et al.: Buprenorphine and Norbuprenorphine in Hair

Table 4. Ratio of buprenorphine:norbuprenorphine concentration in 3 cm washed hair segments from 9 buprenorphine maintained pregnant women.a Cumulative Ratio 1st Ratio 2nd Ratio 3rd Ratio 4th Ratio 5th Patient dose, mg segment segment segment segment segment A0 246 682 0.064 1234 0.057 0.186 1774 0.065 0.098 B0 118 478 0.073 1006 0.077 1418 0.081 2174 0.084 0.126 C0 324 0.436 0.496 744 0.057 0.075 0.085 1302 0.056 0.316 0.223 0.267 1590 0.065 0.105 0.138 0.122 2454 0.093 0.253 0.146 0.210 D0 270 0.056 918 0.042 1292 0.043 1724 0.042 0.048 1886 0.040 0.076 2210 0.073 2808 0.091 E0 414 0.094 970 0.096 1242 0.086 0.107 F0 256 0.551 0.744 794 0.092 0.517 1416 0.071 0.442 0.445 2040 1.117 0.096 1.011 2280 0.070 0.397 0.537 G18 386 870 0.699 1.773 2.302 1544 0.471 1.333 2.368 2.945 3.468 2292 0.430 1.275 1.405 6.274 3.577 H 176 0.584 1.753 1.356 568 0.061 934 0.108 0.199 1590 0.092 0.166 2044 0.081 0.127 0.348 2458 0.082 0.121 I 532 728 0.070 1302 0.084 0.086 1400 0.153 0.147 1636 0.057 0.091 0.142 a Pregnant women were maintained on buprenorphine for up to 23 weeks during gestation. Hair was collected at approximately 4-week intervals and analyzed after dividing into 3-cm segments. No ratio was calculated if either the numerator or denominator was zero. The 1st specimen analyzed from individuals A–F was obtained prior to the beginning of dosing. The next specimen for each of these individuals and the 1st specimen for individuals G–I were obtained, on average, 21 days after dosing was initiated. Clinical Chemistry 53, No. 12, 2007 2143

tigation. It also is unlikely that the acidic extraction used Resources; a grant from the National Institutes of Health; and by Vincent et al. (15), Kintz (11), and Tracqui et al. (13) funds from the National Institute on Drug Abuse Intramural could account for these discrepant results, because Vin- Research Program. cent et al. also reported higher norbuprenorphine hair Financial disclosures: None declared. concentrations. Specimen preparation procedures in Wilkins et al. (14) and the current protocol differed from References other studies by including dissolution of hair with strong 1. Nocon JJ. Buprenorphine in pregnancy: the advantages. Addiction base, and extraction of buprenorphine and norbuprenor- 2006;101:608. phine at pH 10.5. This pH was selected to enhance 2. Jones HE, Johnson RE, Jasinski DR, O’Grady KE, Chisholm CA, Choo RE, et al. Buprenorphine versus methadone in the treatment recovery of norbuprenorphine. We found that acid diges- of pregnant opioid-dependent patients: effects on the neonatal tion diminished the amount of norbuprenorphine recov- abstinence syndrome. Drug Alcohol Depend 2005;79:1–10. ered. Because the only participant in this study with 3. Lejeune C, Simmat-Durand L, Gourarier L, Aubisson S. Prospective buprenorphine concentrations exceeding norbuprenor- multicenter observational study of 260 infants born to 259 phine concentrations was one with blond hair, the possi- opiate-dependent mothers on methadone or high-dose bupreno- bility that norbuprenorphine is preferentially incorpo- phine substitution. Drug Alcohol Depend 2006;82:250–7. rated into pigmented hair should be considered. All 4. Scheidweiler KB, Cone EJ, Moolchan ET, Huestis MA. Dose-related distribution of codeine, cocaine, and metabolites into human hair participants in the study by Wilkins et al. (14) had following controlled oral codeine and subcutaneous cocaine ad- dark-colored hair; hair color was not described for the ministration. J Pharmacol Exp Ther 2005;313:909–15. other investigations. Analysis of melanin content in the 5. Vinner E, Vignau J, Thibault D, Codaccioni X, Brassart C, Humbert present study was not performed. L, et al. Hair analysis of opiates in mothers and newborns for The ratio of buprenorphine:norbuprenorphine in- evaluating opiate exposure during pregnancy. Forensic Sci Int creased distally along the hair shaft. This distribution 2003;133:57–62. could indicate that norbuprenorphine is less tightly 6. Bar-Oz B, Klein J, Karaskov T, Koren G. Comparison of meconium and neonatal hair analysis for detection of gestational exposure to bound to hair than buprenorphine, and as a result there is drugs of abuse. Arch Dis Child Fetal Neonatal Ed 2003;88:F98– increased loss of norbuprenorphine in distal segments 100. from routine hygienic care. Other explanations are that 7. Cirimele V, Kintz P, Mangin P. Testing human hair for cannabis. there is enhanced transport of buprenorphine along the Forensic Sci Int 1995;70:175–82. hair follicle or greater deposition and adsorption of bu- 8. Cone EJ, Gorodetzky CW, Yousefnejad D, Buchwald WF, Johnson prenorphine from sweat. RE. The metabolism and excretion of buprenorphine in humans. This study demonstrated that buprenorphine and nor- Drug Metab Dispos 1984;12:577–81. 9. Thompson. Suboxone. 2005 Physicians Desk Reference. 2005: buprenorphine are incorporated into fetal hair in utero 2830. and that analysis of hair can be used to study dose- 10. Kuhlman JJ, Lalani S, Magluilo J, Levine B, Darwin WD, Johnson concentration relationships in pregnant women on an RE, et al. Human pharmacokinetics of intravenous, sublingual, individual basis. An association between concentrations and buccal buprenorphine. J Anal Toxicol 1996;20:369–78. of these analytes in hair has been described for the 1st 11. Kintz P. Determination of buprenorphine and its dealkylated time. The fact that an association was established between metabolite in human hair. J Anal Toxicol 1993;17:443–4. the cumulative dose of buprenorphine and concentrations 12. Kintz P, Cirimele V, Edel Y, Jamey C, Mangin P. Hair analysis for buprenorphine and its dealkylated metabolite by RIA and confir- of drug and metabolite in hair suggests that it may be mation by LC/ECD. J Forensic Sci 1994;39:1497–503. possible to estimate degree of exposure of individual 13. Tracqui A, Kintz P, Mangin P. HPLC-MS determination of buprenor- women during pregnancy using analyte concentrations. phine and norbuprenorphine in biological fluids and hair samples. Additionally, data indicate that neonatal hair concentra- J Forensic Sci 1997;42:111–4. tions could be used to qualitatively identify drug expo- 14. Wilkins DG, Rollins DE, Valdez AS, Mizuno A, Krueger GC, Cone EJ. sure during gestation, but not the degree of exposure. A retrospective study of buprenorphine and norbuprenorphine in human hair after multiple doses. J Anal Toxicol 1999;23:409–15. 15. Vincent F, Bessard J, Vacheron J, Mallaret M, Bessard G. Deter- mination of buprenorphine and norbuprenorphine in urine and hair Grant/funding support: This research was supported by by gas chromatography-mass spectrometry. J Anal Toxicol 1999; Grant DA R01 12220 from the National Institute on Drug 23:270–9. Abuse; Grant M01RR-02719 from the General Clinical Re- 16. Cirimele V, Kintz P, Lohner S, Ludes B. Buprenorphine to norbu- search Centers Program of the National Center of Research prenorphine ratio in human hair. J Anal Toxicol 2000;24:448–9. Clinical Chemistry 53:12 2144–2151 (2007) Endocrinology and Metabolism

Analytical Validation and Biological Evaluation of a High–Molecular-Weight Adiponectin ELISA

Madhur K. Sinha,1* Traci Songer,1 Qiang Xiao,1 John H. Sloan,1 Jin Wang,1 Shaoquen Ji,1 William E. Alborn,2 Randy A. Davis,2 Michael M. Swarbrick,3 Kimber L. Stanhope,3 Bruce M. Wolfe,4 Peter J. Havel,3 Todd Schraw,5 Robert J. Konrad,2 Philipp E. Scherer,5 and Jehangir S. Mistry1

Background: Of the 3 circulating multimeric forms of centage of HMW adiponectin than for total adiponectin. adiponectin, the high–molecular-weight (HMW) form, as HMW and total adiponectin increased after bypass measured by size-exclusion and/or immunoblotting tech- surgery, but changes in HMW adiponectin were more niques, is a better index of insulin sensitivity for monitor- pronounced and preceded changes in total adiponectin. ing health and disease than is total adiponectin. We aimed Conclusion: This simple, rapid ELISA for HMW adi- to develop a simple ELISA to measure HMW adiponectin. ponectin recognizes the HMW isoform, produces results Methods: We pretreated serum or plasma samples with closely correlated with those obtained with Western digestion solution containing proteinase K (Millipore, blotting, and appears to better distinguish BMI-, sex-, ESDS). HMW (Millipore, EZHMWA-64K) and total adi- and weight loss–associated differences than assays for ponectin (Millipore, EZHADP-61K) concentrations were total adiponectin. measured in treated and untreated samples, respec- © 2007 American Association for Clinical Chemistry tively, from 108 individuals and from 20 morbidly obese patients before and at 1, 3, 6, and 12 months after Adiponectin, the only secretory hormone of fat cells that gastric-bypass surgery. is paradoxically decreased in obesity (1), also has insulin- Results: The ELISA has a dynamic range of 3–200 ␮g/L sensitizing effects (2–4); all other known adipokines are and a detection limit of 0.8 ␮g/L. Intraassay and inter- increased in obesity and induce insulin resistance (5). assay CVs were <4% and <10%, respectively. Sample- Given that adipose tissue is the primary site of adiponec- dilution curves paralleled the calibration curves. Fast tin secretion, the decrease in the production and secretion protein liquid chromatography profiles of the protein- of adiponectin with increasing adiposity is an intriguing ase K-treated samples revealed predominantly HMW phenomenon. Substantial and convincing evidence dem- adiponectin. Values for HMW adiponectin produced onstrates that hypoadiponectinemia is associated with with this method are comparable with those obtained insulin resistance and obesity, diabetes, and metabolic .(syndrome including cardiovascular abnormalities (6, 7 ؍ n ;0.96 ؍ 0.77x – 0.15; r ؍ with Western blot analysis (y 56). Body mass index (BMI)- and sex-related changes Dietary, lifestyle, and/or pharmacological interventions were more pronounced for HMW adiponectin and per- (thiazolidinediones) that improve insulin sensitivity also increase circulating adiponectin concentrations (8, 9).In addition, decreased adiponectin could be a risk factor for the progression of insulin resistance and type 2 diabetes in 1 Millipore Bioscience Division, Millipore Corporation, St. Charles, MO. healthy individuals, with or without impaired glucose 2 Division of Laboratory and Experimental Medicine, Eli-Lilly Company, Indianapolis, IN. tolerance (10, 11). 3 Department of Nutrition, University of California, Davis, CA. Adiponectin is a 28-kDa protein with a collagen-like 4 Division of General Surgery, Oregon Health and Science University, structure; it circulates predominantly in 3 multimeric Portland, OR. 5 Touchstone Diabetes Center, The University of Texas Southwestern forms, i.e., trimer, hexamer, and high–molecular-weight 6 Medical Center, Dallas, TX. (HMW) multimers (12- and 18-mers), which are evident * Address correspondence to this author at: Millipore Bioscience Division, 6 Research Park Dr., St. Charles, MO 63304. Fax 636-442-6087; e-mail Madhur_ [email protected]. Received April 20, 2007; accepted September 28, 2007. 6 Nonstandard abbreviations: HMW, high–molecular-weight; BMI, body Previously published online at DOI: 10.1373/clinchem.2007.090670 mass index; FPLC, fast protein liquid chromatography.

2144 Clinical Chemistry 53, No. 12, 2007 2145

from size-fractionation (gel filtration, velocity gradient, evaluate the effect of weight loss on HMW adiponectin as and gel electrophoresis) and immunoblotting experiments measured with this new ELISA method. The changes in (5). These 3 multimeric adiponectin isoforms have differ- body composition, measures of insulin resistance, and the ent biological activities, with HMW adiponectin probably concentrations of adiponectin and its isoforms that oc- being the active form (5, 12, 13). Tonelli et al. (14) and curred after gastric-bypass surgery have previously been Pajvani et al. (15) initially showed that the distribution of described (22). The Institutional Review Board of the these multimers, particularly as measured by the ratio University of California, Davis, approved the experimen- of HMW adiponectin to total adiponectin (i.e., percentage tal protocol, and all study individuals provided written of HMW adiponectin), correlates better with thiazo- informed consent to participate in this study. lidinedione-mediated improvement in insulin sensitivity than does the total adiponectin concentration. Percentage fast protein liquid chromatography analysis of HMW adiponectin appears to be a better index of Samples of human serum and plasma (100 ␮Lofa1/5 insulin sensitivity (14, 15) and of metabolic syndrome dilution in sample-digestion buffer, EDGB) were fraction- trait cluster(s) (16). ated before and after proteinase K (sample-digestion To measure HMW adiponectin and/or percentage of solution, ESDS) treatment by size-exclusion fast protein HMW adiponectin, most studies have combined size liquid chromatography (FPLC) on a Superdex 200 10/300 fractionation and immunoblotting (14–21), a cumbersome Tricorn column (GE Healthcare). Sample treatment with approach that cannot easily be adopted for high-through- the enzyme selectively degrades hexamer and trimer put analyses of several samples. Therefore, we have isoforms without affecting HMW adiponectin. The mobile developed an ELISA to measure HMW adiponectin in phase was 25 mmol/L sodium phosphate, pH 7.5, con- human serum and plasma samples. taining 100 mg/L phenylmethylsulfonyl fluoride, with a flow rate of 0.5 mL/min and a 0.25-mL fraction size. FPLC Materials and Methods fractions of the enzyme-treated and untreated plasma study participants sample were then assayed for adiponectin immunoreac- We obtained human serum and/or plasma samples from tivity with the Human Adiponectin ELISA (EZHADP- 4 different sources. In the initial development and valida- 61K; Millipore) and Human HMW Adiponectin ELISA tion of the HMW adiponectin ELISA, plasma (containing (EZHMWA-64K, Millipore) reagent sets. EDTA dibasic sodium salt, 1.5 g/L) and serum samples from 30 individuals (4 men and 26 women) were obtained western blot analysis from a commercial source (Bioreclamation). After assay We used Western blotting to measure the trimer, hexamer, development, we obtained another 56 plasma samples and HMW adiponectin isoforms in 56 human plasma sam- from Bioreclamation to compare HMW adiponectin val- ples. We first fractionated plasma samples by polyacryl- ues obtained with the new ELISA and with Western amide gel electrophoresis under nondenaturing and nonre- blotting; we obtained no biochemical or anthropometric ducing conditions with a 1.5-mm 4%–20% polyacrylamide data on these individuals. In a retrospective study, we Tris-glycine gel cassette (Invitrogen, EC6028BOX), 10ϫ obtained another 108 plasma samples from Bioreclama- Novex TrisGlycine SDS Running Buffer (Invitrogen, LC2675; tion and measured HMW adiponectin and total adiponec- diluted 1/10), and 2ϫ TrisGlycine SDS Sample Buffer (In- tin to evaluate the effect of body mass index (BMI) and vitrogen, LC2675) in a XCell SureLock MiniCell (Invitrogen, sex. Table 1 summarizes the anthropometric data for these EI0001) for 1.5 h at 150 V. We ran SeeBlue Plus2 Pre-Stained individuals. Plasma samples obtained from 20 morbidly Standard (Invitrogen, LC5925) with each electrophoresis obese patients before and at 1, 3, 6, and 12 months after separation. Proteins were transferred from the gel to a Roux-en-Y gastric-bypass surgery were also measured to nitrocellulose membrane at 100 V for2hinaMiniTrans-Blot

Table 1. Anthropometric variables for the study individuals.a Group Sex, n Age, years Race, n Systolic BPb, mmHg Diastolic BP, mmHg BMI, kg/m2 Lean (n ϭ 25) M, 17 39.0 (12.3) C, 5 128 (12) 77 (8) 21.56 (1.85) F, 8 AA, 16 H, 4 Overweight (n ϭ 28) M, 12 37.8 (12.0) C, 8 121 (12) 76 (8) 26.79 (1.34) F, 16 AA, 18 H, 2 Obese (n ϭ 55) M, 25 36.9 (10.7) C, 16 126 (14) 79 (7) 35.76 (5.13) F, 30 AA, 30 H, 9 a Data are presented as the mean (SD) where appropriate. b BP, blood pressure; C, Caucasian; AA, African American; H, Hispanic; M, Male; F, Female. 2146 Sinha et al.: HMW Adiponectin ELISA Validation

Cell with 200 mL/L methanol, 2 g/L sodium dodecyl sulfate HMW, and total adiponectin values are expressed as the and 40 mL 25ϫ Novex Tris-Glycine Transfer Buffer (Invitro- mean (SE), whereas all other data including those of gen, LC3675) in 760 mL distilled water. After blocking anthropometric variables and assay characteristics are ௢ overnight at 4 °C with Blocker Casein in Tris-buffered expressed as the mean (SD). saline (1% by weight Casein Hammersten Grade in Tris- buffered saline, Kathon as preservative, pH 7.4, Pierce, Results 37532), we incubated the nitrocellulose membrane in the characteristics of the hmw adiponectin assay same buffer for 1.5 h with mouse antihuman adiponectin The calibration curve for this HMW Adiponectin ELISA monoclonal antibody labeled with horseradish peroxidase (n ϭ 8, Fig. 1A) has a concentration range of 3–200 ␮g/L. (1/1000 dilution) (R&D Systems, MAB10651). After washing Assay imprecision (CV) ranged between 4% and 8% for away the excess antibody, we added detection reagents 1 absorbance and was Ͻ2% for the back-calculated values, and 2 in equal proportions (Amersham/GE Healthcare, as evaluated by fixed weighted 5-parameter (asymmetri- RPN2106V1 and RPN2106V2), incubated the membrane for cal) logistic curve fitting (StatLIA software; Brendan Tech- 1 min, and exposed the membrane for 1–4 min to BioMax nologies). This program yielded an assay sensitivity of 0.8 Light Film (Kodak, 869358). We digitally photographed the ␮g/L, which is the mean ϩ 2 SDs of the minimum immunoblot and quantified the trimer, hexamer, and HMW detection limit (n ϭ 8). Interassay imprecision (4 assays, adiponectin bands with AlphaImager and AlphaEase FC 10 samples) and intraassay imprecision (5 results within a software V.4.1.0 (Alpha Innotech). single assay, 8 samples; Table 2) were measured after the samples had each been pretreated and diluted (1/200 final measurement of total human adiponectin and dilution). HMW adiponectin values reported in Table 2 hmw adiponectin are after sample pretreatment and dilution from HMW We measured total adiponectin in serum and plasma adiponectin assays. Intraassay CVs were between 1.0% samples with Millipore’s Human Adiponectin ELISA. We and 3.4% for sample HMW adiponectin values between used the newly developed HMW Adiponectin ELISA to 6.0 ␮g/L and 65.3 ␮g/L. Interassay CVs were between measure HMW adiponectin. We diluted 20-␮L serum 3.0% and 8.1% for HMW adiponectin values between 13.3 samples 1/10 with sample-digestion buffer (Millipore, ␮g/L and 61.5 ␮g/L. To evaluate the robustness of the EDGB) and pretreated the samples with sample-digestion assay, we analyzed 18 different serum samples on 2 solution (Millipore, ESDS) for2hat37°C.Enzyme- different occasions. We observed no significant differ- treated samples were diluted further to 1/20 with 1ϫ sample dilution buffer (Millipore, ESDB) and then assayed for HMW adiponectin with the Human HMW Adiponectin ELISA. The pairs of antibodies for capture and detection were different for the total (Millipore, EZHADP-61K) and HMW (Millipore, EZHMWA-64K) adiponectin ELISA reagent sets. The calibrator for the total adiponectin ELISA was pooled human serum calibrated against recombinant human adiponectin produced in mammalian cells (Millipore, 1061-K). We used the same calibrator for the HMW Adiponectin ELISA after we corrected for the proportion of HMW adiponectin, as measured by FPLC analysis.

statistical analysis We carried out a regression analysis to compare HMW adiponectin values obtained with the HMW ELISA reagent set and with Western blotting. Differences in the concentrations of total adiponectin and HMW adiponectin between lean, overweight, and obese individuals, between male and female individuals, and before and after gastric-bypass surgery were compared by means of 2-tailed Student t-tests under the assumption of unequal variances. Multivariate analysis was performed by stepwise regression (Minitab 14; Minitab). BMI was entered as a response variable, and sex, age, race, diastolic and Fig. 1. Evaluation of the HMW Adiponectin ELISA. systolic blood pressures, total adiponectin, HMW adi- (A), calibration curve with mean absorbance values (n ϭ 8; f) and an imprecision profile (Ⅺ) of CVs based on the same 8 calibration curves. (B), linearity of the ponectin, and percentage of HMW adiponectin were dilution curves for 10 serum samples after pretreating and diluting the sample; entered as predictor variables. HMW, percentage of serial dilutions were with ELISA buffer. Clinical Chemistry 53, No. 12, 2007 2147

Table 2. Interassay and intraassay CVs. (5 ؍ Intraassay (n (4 ؍ Interassay (n

Sample no. Mean concentration, ␮g/L CV, % Sample no. Mean concentration, ␮g/L CV, % 1 13.3 4.1 11 6.0 1.0 2 21.2 8.1 12 11.1 1.3 3 23.2 4.1 13 13.7 3.4 4 23.7 3.7 14 21.3 2.1 5 28.7 3.1 15 25.4 3.3 6 26.6 5.0 16 27.3 2.6 7 34.8 7.8 17 39.5 3.1 8 38.4 9.1 18 65.3 9 60.8 3.0 10 61.5 3.8 ences (i.e., P Ͼ0.05) in HMW adiponectin values [mean extent before the simple enzyme pretreatment. After (SD), 8.7 (4.5) mg/L vs 9.0 (4.8) mg/L] with a mean pretreatment, the HMW adiponectin peak remains unal- difference of Ϫ1.6% (7.1%) in the mean values obtained tered, whereas both hexamer and trimer peaks virtually with the 2 replicate assays. Recovery was evaluated by disappear because of their preferential degradation by the adding different concentrations of HMW adiponectin enzyme (Fig. 2A). The efficacy of the HMW Adiponectin (obtained from human serum after enzyme digestion and ELISA for measuring HMW adiponectin can be ascribed calibration with the HMW Adiponectin ELISA) to 8 to both this ELISA’s lower sensitivity for recognizing pretreated and diluted samples. The data in the recovery hexamers and trimers and preferential enzymatic degra- experiments were expressed as a percentage of the ex- dation of these 2 isoforms. pected value (basal plus added HMW adiponectin). The mean recovery was 101.2% (1.9%) for 3.125 ␮g/L, 106.6% (2.3%) for 25 ␮g/L, and 112.4% (3.1%) for 100 ␮g/L. An evaluation of the linearity of endogenous HMW adiponectin concentrations in serial dilutions (1/8, 1/4, and 1/2) of 10 initially diluted and pretreated serum samples yielded values between 95% and 110% of the expected values (value of a 1/1 dilution times the dilution factor). The dilution curves were linear for all 8 serum samples (Fig. 1B) and paralleled the calibration curve. Insulin, glucagons, leptin, glucagon-like peptide 1, amylin, resis- tin, acylation-stimulating protein, different interleukins, interferon ␥, tumor necrosis factor ␣, transforming growth factor ␤, and plasminogen activator inhibitor 1 diluted in assay buffer (20–1000 ␮g/L) showed no cross-reactivity (Ͻ0.01%) with the HMW Adiponectin ELISA in direct tests. Values for paired serum (x) and plasma (y) samples (n ϭ 21) showed excellent correlation in the same assay (y ϭ 0.95x Ϫ 0.23; r ϭ 0.99); however, plasma values [7.0 (5.1) mg/L] were slightly but significantly lower [mean difference, 8.0% (5.7%); P Ͻ0.01] than serum values [7.5 (5.3) mg/L]. fplc profiles Six serum samples with or without proteinase K pretreatment were fractionated by FPLC on a Superdex G-200 column. The total adiponectin ELISA for FPLC fractions of untreated samples detected 3 distinct peaks for the HMW, hexamer, and trimer isoforms. These results suggest that the total adiponectin ELISA recognizes all 3 Fig. 2. Comparison of ELISA and Western blot analysis. adiponectin multimeric isoforms (Fig. 2A). The HMW (A), a representative FPLC elution profile of human serum before (ࡗ) and after ({) sample pretreatment and subsequent ELISA of total and HMW adiponectin. Adiponectin ELISA recognizes all 3 isoforms, but hexam- (B), comparison with Western blotting results (ratio of HMW adiponectin to total ers and trimers are recognized to an appreciably lesser adiponectin x total adiponectin). 2148 Sinha et al.: HMW Adiponectin ELISA Validation

western blot analysis [10.75 (0.98) mg/L; P Ͻ0.05] and obese individuals [9.45 We measured the relative distributions of adiponectin (0.51) mg/L; P Ͻ0.001], compared with lean individuals isoforms in 56 human plasma samples by Western blot- [13.80 (1.08) mg/L]; however, the differences for HMW ting with nondenaturing and nonreducing gels. HMW adiponectin concentrations were even more pronounced adiponectin concentrations in these samples were then in overweight individuals [4.44 (0.55) mg/L; P Ͻ0.005] extrapolated by multiplying the percentage HMW adi- and obese individuals [3.60 (0.32) mg/L; P Ͻ0.0001], ponectin values obtained by Western blotting by the total compared with lean individuals [7.06 (0.68) mg/L]. Sim- adiponectin concentrations obtained with the Millipore ilarly, the differences in percentage of HMW adiponectin human adiponectin ELISA. The HMW adiponectin values were also more pronounced than for the total adiponectin indirectly derived via Western blotting were then com- differences in comparisons of overweight individuals pared with those obtained for the HMW Adiponectin [38.71% (2.55%); P Ͻ0.0002] and obese individuals ELISA after sample pretreatment. HMW adiponectin con- [35.44% (1.61%); P Ͻ0.000005] with lean individu- centrations obtained with these 2 different methods show als [50.48% (2.21%)]. The differences in HMW adiponectin high correlation (y ϭ 0.77 x Ϫ 0.15; r ϭ 0.96) (Fig. 2B). Of (P Ͻ0.0005) and percentage of HMW adiponectin (P note is that the apparent difference (approximately 23%) Ͻ0.0005) were also more pronounced in the obese group in absolute values could be due to these 2 very different [mean BMI, 35.75 (5.13) kg/m2] than in the combined lean methods and the possibilities of differential recognition and overweight groups [mean BMI, 24.32 (3.07) kg/m2], and/or different affinities of the antibodies used. compared with the differences for total adiponectin (P Ͻ0.005). The ratio of HMW adiponectin to total adiponec- effect of body weight and sex tin has previously been demonstrated to be higher in Mean (SE) values for total adiponectin, HMW adiponec- female individuals than in male individuals (3, 15). Be- tin, and percentage of HMW adiponectin (relative to total cause of the disproportionately higher ratio of males to adiponectin) in lean [mean BMI, 21.56 (1.85) kg/m2], females in the lean group than in the overweight and overweight [mean BMI, 26.79 (1.34) kg/m2], and obese obese groups, the weight-related decrease in HMW adi- [mean BMI, 35.67 (5.13) kg/m2] individuals are shown in ponectin concentration may have been compromised due Fig. 3A. As expected, concentrations of total adiponectin to a suboptimal selection of the study population. were significantly decreased in overweight individuals As shown in Fig. 3B with mean (SE) values, the subpopulation of African American women [mean BMI, 29.74 (5.81) kg/m2;nϭ 26] and men [mean BMI, 28.97 (7.78) kg/m2;nϭ 35], featured more pronounced alterations for HMW adiponectin [5.68 (0.64) mg/L and 3.63 (0.39) mg/L, respectively; P Ͻ0.02] and percentage of HMW adiponectin [44.42% (2.38%) and 35.04% (2.26%), respectively; P Ͻ0.01] than for total adiponectin [12.27 (1.08) mg/L and 9.54 (0.64) mg/L, respectively; P Ͻ0.05]. Analysis of the entire dataset of 54 female and 54 male individuals revealed significantly higher values in females for HMW adiponectin (P Ͻ0.05), but not for percentage of HMW adiponectin (P Ͼ0.05) and total adiponectin (P Ͼ0.05). In a multivariate analysis with stepwise regression, the best-fitting model included HMW adiponectin (␤ ϭ Ϫ1.01; P Ͻ0.0001), sex (␤ ϭ 3.5; P Ͻ0.01), and diastolic pressure (␤ ϭ 0.16; P Ͼ0.05). This model accounted for 20.5% of the variance in BMI (adjusted R2). Total adiponectin was not a significant predictor and was thus left out of the model.

gastric-bypass surgery Fig. 4 summarizes the effects of weight loss on total, HMW, and percentage of HMW adiponectin at 1, 3, 6, and Fig. 3. Total adiponectin, HMW adiponectin, and percentage of HMW 12 months after gastric-bypass surgery in 20 morbidly adiponectin according to body weight and sex. obese adults (19 women and 1 man). Gastric-bypass (A), values for lean (n ϭ 25), overweight (n ϭ 28), and obese (n ϭ 55) surgery produced appreciable weight loss and improved individuals. *, P Ͻ0.05; **, P Ͻ0.001; ϩ, P Ͻ0.005; ϩϩ, P Ͻ0.0001; †, P insulin sensitivity in all of the patients (22). Before sur- Ͻ0.0002; ††, P Ͻ0.000005. (B), values for African American women (n ϭ 26) and men (n ϭ 35). *, P Ͻ0.05; ϩϩ, P Ͻ0.02; †††, P Ͻ0.01. Columns represent gery, the mean (SE) concentrations of total adiponectin the mean (SE). and HMW adiponectin in the morbidly obese patients Clinical Chemistry 53, No. 12, 2007 2149

electrophoresis, gel filtration, and velocity gradient centrifugation and then used immunoblotting with anti- adiponectin antibody to recognize the multimers. These time-consuming methods require experience in protein- separation techniques. In addition, these methods are not practical for analyzing several samples. Immunoassays, particularly ELISA, can be an alternative. Currently, 2 different ELISAs are commercially available (23, 24). Blu- her et al. (25) evaluated this HMW Adiponectin ELISA (ALPCO Diagnostics), an RIA for total adiponectin, and Fig. 4. Total, HMW, and percentage of HMW adiponectin in morbidly another ELISA (Mediagnost) to determine the best adi- obese individuals after gastric-bypass surgery. ponectin assay for predicting improvement in insulin *, P Ͻ0.02; **, P Ͻ0.001; ***, P Ͻ0.00002; ϩ, P Ͻ0.002; ϩϩ, P Ͻ0.0005; ϩϩϩ, P Ͻ0.0002; ϩϩϩϩ, P Ͻ0.00002; †, P Ͻ0.0005; ††, P Ͻ0.00001; sensitivity after exercise in healthy individuals, those with †††, P Ͻ0.000005; ††††, P Ͻ0.0000005. Columns (percentage of baseline impaired glucose tolerance, and diabetic individuals. value) represent the mean (SE). These investigators found that the total adiponectin RIA, not the HMW Adiponectin ELISA, was the better predic- were 7.2 (0.5) mg/L and 2.2 (0.3) mg/L, respectively, and tor of insulin sensitivity. Another HMW Adiponectin the mean (SE) percentage of HMW adiponectin value was ELISA developed by Nakano et al. (24) uses an antibody 28.1% (2.3%). Total adiponectin concentrations were un- raised against HMW adiponectin but involves no enzyme altered at 1 month after surgery [102.47% (7.66%) of pretreatment of the sample. Although this adiponectin presurgery concentrations, P Ͼ0.05] but were significantly ELISA is simple, it also measures the hexameric isoform to increased at 3 months [124.52% (9.02%), P Ͻ0.02], 6 some degree (24). months [141.3% (10.50%), P Ͻ0.001], and 12 months Our HMW Adiponectin ELISA method is simple, [166.12% (11.44%), P Ͻ0.00002]; however, the changes in rapid, and specific with respect to the multimeric (HMW) HMW adiponectin were significantly more pronounced isoform. After a simple enzyme pretreatment of the sam- than for total adiponectin at 1 month [147.00% (14.48%), P ple, the new ELISA recognizes predominantly the HMW Ͻ0.0005], 3 months [173.65% (20.58%), P Ͻ0.002], 6 adiponectin isoform. This HMW Adiponectin ELISA months [210.5% (23.44%), P Ͻ0.0002], and 12 months meets most of the analytical-robustness criteria of a good [271.72% (30.67%), P Ͻ0.00002] after gastric-bypass sur- immunoassay. In addition, the HMW adiponectin values gery. In addition, the changes in percentage of HMW obtained with this method and with a more conventional adiponectin were also more significant than for total size-exclusion and Western blotting method are well adiponectin at 1 month [134.61% (9.17%), P Ͻ0.00005], 3 correlated (r ϭ 0.96) over a wide range of HMW adiponec- months [136.08% (8.26%), P Ͻ0.0005], 6 months [146.81% tin concentrations; however, there is a 23% difference in (7.64%), P Ͻ0.00001], and 12 months [160.74% (7.74%), P the absolute values, which is not surprising considering Ͻ0.0000005] after surgery. the 2 very different methodologies. The relevance of this new HMW Adiponectin ELISA is Discussion demonstrated by comparing the values for HMW adi- That adiponectin has an insulin-sensitizing effect (2–4) ponectin and total adiponectin in the context of BMI, sex, and is an important biomarker for insulin sensitivity in and weight loss in morbidly obese patients. Obesity- various pathophysiological conditions associated with related changes were more apparent for HMW adiponec- insulin resistance is well documented (1, 5–7). In addition, tin concentration and percentage of HMW adiponectin the circulating adiponectin concentration seems to be a than for total concentration when these variables were good predictor of the progression of insulin resistance in evaluated for lean, overweight, and obese individuals healthy and/or glucose-intolerant individuals (10, 11). (26–28). Similarly, we noted larger differences between Therapeutic or dietary interventions intended to improve men and women for HMW adiponectin concentration and insulin sensitivity are usually associated with increased percentage of HMW adiponectin than for the concentra- adiponectin concentrations (8, 9). Of the 3 distinct multi- tion of total adiponectin (3, 15, 29); however, our studies meric isoforms in the circulation, HMW adiponectin (12- may have been compromised by confounding factors to 18-mers) is believed to be the most biologically active introduced by a suboptimal control population. Gastric- (5, 12, 13). The HMW adiponectin concentration and/or bypass surgery is an effective means of weight loss and percentage of HMW adiponectin are reported to be a leads to improved insulin sensitivity (30). Similar to the better index of insulin sensitivity than total adiponectin findings of Swarbrick et al. (22) with a different ELISA (14, 21). (23), the present study also demonstrated a significant Testing the importance of HMW adiponectin in vari- increase in HMW adiponectin concentration and percent- ous metabolic states requires a simple, accurate, and rapid age of HMW adiponectin (but not for the concentration of methodology. A majority of the studies have fractionated total adiponectin) at 1 month after gastric-bypass surgery, the different adiponectin multimers by size with gel when appreciable weight loss and an improvement in 2150 Sinha et al.: HMW Adiponectin ELISA Validation

insulin resistance are typically observed. The concentra- 10. Lindsay RS, Funahashi T, Hanson RL, Matsuzawa Y, Tanaka S, tion of HMW adiponectin increased progressively at 3, 6, Tataranni PA, et al. Adiponectin and development of type 2 and 12 months after surgery, along with further decreases diabetes in the Pima Indian population. Lancet 2002;360:57–8. in body weight and insulin resistance (homeostasis model 11. Schulze MB, Shai I, Rimm EB, Li T, Rifai N, Hu FB. Adiponectin and future coronary heart disease events among men with type 2 assessment), and the increases were proportionately diabetes. Diabetes 2005;54:534–9. greater than for total adiponectin concentrations. 12. Waki H, Yamauchi T, Kamon J, Ito Y, Uchida S, Kita S, et al. Impaired multimerization of human adiponectin mutants associ- In summary, we have developed a simple, robust, rapid, ated with diabetes. Molecular structure and multimer formation of and size-specific ELISA for measuring HMW adiponectin. adiponectin. J Biol Chem 2003;278:40352–63. The results are comparable with those obtained with a 13. Kobayashi H, Ouchi N, Kihara S, Walsh K, Kumada M, Abe Y, et al. Western blotting method. More pronounced differences Selective suppression of endothelial cell apoptosis by the high molecular weight form of adiponectin. Circ Res 2004;94:e27–31. were observed for HMW adiponectin than for total adi- 14. Tonelli J, Li W, Kishore P, Pajvani UB, Kwon E, Weaver C, et al. ponectin with respect to obesity, weight loss, and sex. This Mechanisms of early insulin-sensitizing effects of thiazolidinedio- assay may become a useful tool for high-throughput nes in type 2 diabetes. Diabetes 2004;53:1621–9. analyses of HMW adiponectin and may facilitate further 15. Pajvani UB, Hawkins M, Combs TP, Rajala MW, Doebber T, Berger elucidation of the importance of HMW adiponectin as a JP, et al. Complex distribution, not absolute amount of adiponec- biomarker for monitoring the progression and treatment tin, correlates with thiazolidinedione-mediated improvement in of diabetes and cardiovascular diseases. insulin sensitivity. J Biol Chem 2004;279:12152–62. 16. Lara-Castro C, Luo N, Wallace P, Klein RL, Garvey WT. Adiponectin multimeric complexes and the metabolic syndrome trait cluster. Diabetes 2006;55:249–5. Grant/funding support: None declared. 17. Fisher FF, Trujillo ME, Hanif W, Barnett AH, McTernan PG, Scherer Financial disclosures: None declared. PE, et al. Serum high molecular weight complex of adiponectin Acknowledgments: We thank Dr. Rick Ryan, Millipore Bio- correlates better with glucose tolerance than total serum adi- science Division, for his keen interest and encouragement ponectin in Indo-Asian males. Diabetologia 2005;48:1084–7. and Deborah Droll for statistical analysis. 18. Bodles AM, Banga A, Rasouli N, Ono F, Kern PA, Owens RJ. Pioglitazone increases secretion of high-molecular-weight adi- References ponectin from adipocytes. Am J Physiol 2006;291:E1100–5. 1. Arita Y, Kihara S, Ouchi N, Takahashi M, Maeda K, Miyagawa J, et 19. Bobbert T, Rochlitz H, Wegewitz U, Akpulat S, Mai K, Weickert MO, al. Paradoxical decrease of an adipose-specific protein adiponec- et al. Changes of adiponectin oligomer composition by moderate tin in obesity. Biochem Biophys Res Commun 1999;257:79–83. weight reduction. Diabetes 2005;54:2712–9. 2. Berg AH, Combs TP, Du X, Brownlee M, Scherer PE. The adipocyte- 20. Hammarstedt A, Sopasakis VR, Gogg S, Jansson PA, Smith U. secreted protein Acrp30 enhances hepatic insulin action. Nat Med Improved insulin sensitivity and adipose tissue dysregulation after 2001;7:947–53. short-term treatment with pioglitazone in non-diabetic, insulin- resistant subjects. Diabetologia 2005;48:96–104. 3. Pajvani UB, Du X, Combs TP, Berg AH, Rajala MW, Schulthess T, 21. Halperin F, Beckman JA, Patti ME, Trujillo ME, Garvin M, Creger et. al Structure-function studies of the adipocyte-secreted hor- MA, et al. The role of total and high-molecular-weight complex of mone Acrp30/adiponectin: implications for metabolic regulation adiponectin in vascular function in offspring whose parents both and bioactivity. J Biol Chem 2003;278:9073–85. had type 2 diabetes. Diabetologia 2005;48:2147–54. 4. Satoh H, Nguyen MT, Trujillo M, Imamura T, Usui I, Scherer PE, et 22. Swarbrick MM, Austrheim-Smith IT, Stanhope KL, Van Loan MD, al. Adenovirus-mediated adiponectin expression augments skele- Ali MR, Wolfe BM, et al. Circulating concentrations of high- tal muscle insulin sensitivity in male Wistar rats. Diabetes 2005; molecular-weight adiponectin are increased following Roux-eb-Y 54:1304–13. gastric bypass surgery. Diabetologia 2006;49:2552–8. 5. Trujillo ME, Scherer PE. Adiponectin-journey from an adipocyte 23. Ebinuma H, Miyazaki O, Yago H, Hara K, Yamauchi T, Kadowaki T. secretory protein to biomarker of the metabolic syndrome. J Intern A novel ELISA system for selective measurement of human Med 2005;257:167–75. adiponectin multimers by using proteases. Clin Chim Acta 2006; 6. Weyer C, Funahashi T, Tanaka S, Hotta K, Matsuzawa Y, Pratley 372:47–53. RE, et al. Hypoadiponectinemia in obesity and type 2 diabetes: 24. Nakano Y, Tajima S, Yoshimi A, Akiyama H, Tsushima M, Tanioka Close association with insulin resistance and hyperinsulinemia. T, et al. A novel enzyme-linked immunosorbent assay specific for J Clin Endocrinol Metab 2001;86:1930–5. high-molecular-weight adiponectin. J Lip Res 2006;47:1572–82. 7. Cnop M, Havel PH, Utzschneider KM, Carr DB, Sinha MK, Boyko 25. Bluher M, Brennan AM, Kelesidis T, Kratzsch J, Fasshauer M, EJ, et al. Relationship of adiponectin to body fat distribution, Kralisch S, et al. Total and high-molecular weight adiponectin in insulin sensitivity and plasma lipoproteins: evidence for indepen- relation to metabolic variables at baseline and in response to an dent roles of age and sex. Diabetologia 2003;46:459–69. exercise treatment program: comparative evaluation of three 8. Yang WS, Lee WJ, Funahashi T, Tanaka S, Matsuzawa Y, Chao CL. assays. Diabetes Care 2007;30:280–5. et al. Weight reduction increases plasma levels of an adipose- 26. Hara K, Horikoshi M, Yamauchi T, Yago H, Miyazaki O, Ebinuma H, derived anti-inflammatory protein, adiponectin. J Clin Endocrinol et al. Measurement of the high-molecular-weight form of adiponec- Metab 2001;86:3815–9. tin in plasma is useful for the prediction of insulin resistance and 9. Yu JG, Javorschi S, Hevener AL, Kruszynska YT, Norman RA, Sinha metabolic syndrome. Diabetes Care 2006;29:1357–62. M, et al. The effect of thiazolldinedlones on plasma adiponectin 27. Aso Y, Yamamoto R, Wakabayashi S, Uchida T, Takayanagi K, levels in normal, obese, and type 2 diabetic subjects. Diabetes Takebayashi K, et al. Comparison of serum high-molecular-weight 2002;51:2968–74. (HMW) adiponectin with total adiponectin concentrations in type 2 Clinical Chemistry 53, No. 12, 2007 2151

diabetic patients with coronary artery disease using a novel 29. Andersen KK, Frystayk J, Wolthers OD, Heuck C, Flyvbjerg A. enzyme-linked immunosorbent assay to detect HMW adiponectin. Gender differences of oligomers and total adiponectin during Diabetes 2006;55:1954–60. puberty: a cross-sectional study of 859 Danish school children. 28. Nakashima R, Kamei N, Yamane K, Nakanishi S, Nakashima A, J Clin Endocrinol Metab 2007;92:3674–85. Kohno N. Decreased total and high-molecular-weight adiponectin 30. Pories WJ, MacDonald KG Jr, Morgan EJ, Sinha MK, Dohm GL, are independent risk factors for the development of type 2 Swanson MS, et al. Surgical treatment of obesity and its effect on diabetes in Japanese-Americans. J Clin Endocrinol Metab 2006; diabetes: 10-y follow-up. Am J Clin Nutr. 1992;55(Suppl 2):582S– 91:3873–7. 5S. Clinical Chemistry 53:12 2152–2159 (2007) Endocrinology and Metabolism

Association of C-Reactive Protein with Surrogate Measures of Insulin Resistance among Nondiabetic US Adults: Findings from National Health and Nutrition Examination Survey 1999–2002

Yuan-Xiang Meng,1* Earl S. Ford,2 Chaoyang Li,2 Alexander Quarshie,3 Ahmad M. Al-Mahmoud,3 Wayne Giles,2 Gary H. Gibbons,4 and Gregory Strayhorn1

Background: Increased C-reactive protein (CRP) con- regression coefficient ranges from 0.06 to 0.09). The centration and insulin resistance (IR) are associated association of CRP (>3vs<1 mg/L) with the homeosta- with increased rates of adverse cardiovascular events. sis model for assessment of IR (>75th vs <75th percen- We sought to examine the relationship of CRP with tile) was statistically significant among people with a surrogate measures of IR among nondiabetic adults in body mass index >30 kg/m2 (odds ratio, 2.6; 95% CI, the US. 1.3–5.1) or with a body mass index <25 kg/m2 (odds Methods: We conducted analyses using data from the ratio, 2.5; 95% CI, 1.5–4.2). National Health and Nutrition Examination Survey Conclusions: CRP was significantly associated with the 1999–2002. We analyzed a nationally representative surrogate measures of IR among nondiabetic adults. sample of 2514 men and nonpregnant women age >20 Obesity may play an important role in the association of years who were non-Hispanic white, non-Hispanic CRP with IR in this nationally representative sample. black, or Mexican American. © 2007 American Association for Clinical Chemistry Results: After adjustment for age, sex, race/ethnicity, smoking status, systolic blood pressure, and serum Insulin resistance (IR)5 is a pathophysiological state char- concentrations of HDL cholesterol, LDL cholesterol, and acterized by a subnormal physiological response to insu- triglyceride, CRP was significantly associated with 10 IR lin concentrations within reference intervals. Experimen- measures (all P values <0.01). The strength of the tal and epidemiological studies have demonstrated a association attenuated after further adjustment for waist strong association of IR with many diseases or metabolic circumference (change in adjusted regression coeffi- abnormalities, including coronary heart disease, stroke, cients ranging from 60.0% to 75.1%). The association of type 2 diabetes, hypertension, dyslipidemia, systemic CRP with each IR surrogate was similar (standardized inflammation, and atherogenesis (1, 2). It is estimated that approximately 20.6 million adults aged 20 years or older in the US have diabetes, with the 1 Department of Family Medicine, Morehouse School of Medicine, Atlanta, majority of these having type 2 diabetes in 2005 (3). The GA. prevalence of hyperinsulinemia has increased by 35% 2 Division of Adult and Community Health, National Center for Chronic among nondiabetic adults in the US in the past decade (4). Disease Prevention and Health Promotion, Centers for Disease Control and It is believed that IR and subsequently compensatory Prevention, Atlanta, GA. ␤ 3 Clinical Research Center and 4 Cardiovascular Research Institute, More- hyperinsulinemia develop much earlier than -cell dys- house School of Medicine, Atlanta, GA. function and may exist and progress years before even * Address correspondence to this author at: Yuan-Xiang Meng, Depart- prediabetes would be diagnosed by the detection of ment of Family Medicine, Morehouse School of Medicine, 1513 East Cleveland Ave., Bldg. 100, Suite 300A, East Point, GA 30344. Fax 404-756-1229; e-mail [email protected]. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the Centers for Disease Control and 5 Nonstandard abbreviations: IR, insulin resistance; CRP, C-reactive pro- Prevention. tein; TG, triglyceride; HDL-C, HDL cholesterol; BMI, body mass index; Received March 14, 2007; accepted September 28, 2007. HOMA, homeostasis model assessment; SBP, systolic blood pressure; LDL-C, Previously published online at DOI: 10.1373/clinchem.2007.088930 LDL cholesterol.

2152 Clinical Chemistry 53, No. 12, 2007 2153

impaired fasting glucose or impaired glucose tolerance. ing blood drawn with fasting time Ն8 h, and had CRP Several studies have demonstrated that early interven- concentration Յ10 mg/L (n ϭ 4040). tions, including lifestyle modification and pharmacological treatment, can effectively delay the onset of diabetes procedures in prediabetic individuals and therefore decrease the Detailed descriptions of blood collection and processing incidence of cardiovascular disease and other diabetes- have been previously provided (24). In brief, serum related chronic illnesses (5, 6). Thus, early identification specimens were frozen at ϽϪ70 °C, shipped on dry ice, of individuals who have developed IR is particularly and stored at ϽϪ70 °C until analysis. CRP concentrations important in clinical practice, but diagnosis is challenging. were quantified at the University of Washington Medical Although IR can be determined by a variety of meth- Center, Seattle, WA, by latex-enhanced nephelometry, a ods, they are difficult to apply in daily clinical practice, high-sensitivity assay, on a BN II nephelometer (Dade particularly in outpatient care settings. Evidence indi- Behring). Control materials of 2 concentrations from Bio- cates that compensatory hyperinsulinemia is highly cor- Rad Laboratories were used for QC purposes, and CVs related with IR (7) and may offer a more useful way to ranged from 4.93% to 7.84%. identify insulin-resistant patients than measurements of Serum insulin concentration was measured using an glucose intolerance. However, analytic methods for insu- RIA reagent set from Pharmacia Diagnostics. The cross- lin measurements are not standardized, and it is difficult reactivity of Pharmacia insulin antibody with proinsulin to relate absolute values of plasma insulin concentrations is approximately 40%. All insulin assays for NHANES from one laboratory to another (8). 1999–2002 were performed by the same laboratory at the C-reactive protein (CRP) is a marker for systemic subclin- University of Missouri-Columbia. Identical laboratory ical inflammation and may have prognostic value in identi- procedures for insulin assays and their QC were per- fying persons who are at an increased risk of developing formed. The overall CVs were 3.3%–5.4% in NHANES 1999–2002. Plasma glucose concentration was measured type 2 diabetes and subsequent cardiovascular complica- using an enzymatic reaction. Serum TG concentration tions (9). Given its ease of measurement, biological stability, was measured enzymatically after hydrolyzation to glyc- and improved high-sensitivity method, CRP may be useful erol, and HDL cholesterol (HDL-C) was measured after as a clinical measure for identifying individuals at risk for IR the precipitation of other lipoproteins with a heparin- (10). CRP has been associated with increased adiposity (11), manganese chloride mixture. but only a few studies have demonstrated that CRP is Up to 4 blood pressure readings were obtained in associated with IR independent of obesity (12). the mobile examination center. The mean of the last 2 Because it is not feasible to directly measure IR in measurements for participants who had 3 or 4 mea- large epidemiologic studies, surrogate measures using surements, the last measurement for participants with fasting insulin or the combination of fasting insulin 2 measurements, and the single measurement for par- with fasting glucose or triglyceride (TG) have been ticipants who had only 1 measurement were used to proposed (13–21). Little is known, however, about the establish blood pressure status. Body mass index association of CRP with different surrogate measures of [BMI ϭ weight (kg)/height (m)2] was calculated using IR. In this study, we used data from the National Health measured weight and height, and BMIs were catego- and Nutrition Examination Survey (NHANES) 1999–2002 rized into 3 groups (1, Ͻ25; 2, 25–29.9; and 3, Ն30) to examine the association of CRP with a series of surrogate according to WHO criteria (25). Waist circumference measures of IR. was measured with a steel measuring tape to the nearest 0.1 cm at the high point of the iliac crest at Materials and Methods minimal respiration. study design and participants NHANES 1999–2002 is a stratified, multistage probability surrogate measures of ir sample of the civilian noninstitutionalized US population. A total of 10 surrogate measures of IR were used in this Trained interviewers, using a computer-assisted personal study, including fasting insulin, homeostasis model as- survey system, spoke to participants at home. Respon- sessment (HOMA) of IR, log (1/HOMA), empirical fast- dents were then asked to attend a mobile examination ing IR index, fasting insulin:fasting glucose ratio, center, where they completed additional questionnaires, Raynaud index, Bennetts index, quantitative insulin sen- underwent various examinations, and provided blood sitivity check index, Avignon index, and McAuley index samples. Details of the NHANES 1999–2002 surveys may (see Appendix). Nine indices were derived from the be found elsewhere (22, 23). This analysis was limited to measures of fasting insulin and fasting glucose or TG. men and nonpregnant women who were Ն20 years old, The formulas for calculating these indices are provided were non-Hispanic white, non-Hispanic black, or Mexi- in the Appendix. In addition, we included the C-peptide can-American Hispanics, did not use hormone therapy, concentrations as a marker of insulin secretion because it had no diagnosed diabetes or hypoglycemia medications is cosecreted with insulin in equimolar amounts but is not or fasting glucose Ն126 mg/dL (7.0 mmol/L), had morn- subject to hepatic clearance (26). 2154 Meng et al.: CRP and IR

statistical analysis significant differences in geometric means of CRP (P ϭ We assessed continuous variables for gaussian distribu- 0.38) and waist circumferences (P ϭ 0.45) among race/ tion and performed logarithm transformation for fasting ethnic groups. Black participants had higher mean fasting insulin, CRP, and C-peptide to approximate a gaussian insulin:fasting glucose ratios and McAuley indices than distribution. Means and Pearson correlation coefficients whites. were calculated for all surrogate measures of IR and The unadjusted Pearson correlation coefficients of CRP CRP. Multiple linear regression analyses were performed with the 10 measures of IR ranged from 0.32 to 0.39 to assess the association of CRP with IR surrogate mea- among whites (P Ͻ0.001), from 0.28 to 0.36 among blacks sures, adjusting for potential confounders including age, (P Ͻ0.01), and from 0.29 to 0.36 among Mexican Ameri- sex, race/ethnicity, smoking status, systolic blood pres- cans (P Ͻ0.001) (Table 2). The correlation coefficients were sure (SBP), and concentrations of HDL-C, LDL cholesterol not statistically significant between CRP and log fasting (LDL-C), and TG. Additional adjustment for waist cir- insulin, HOMA, log (1/HOMA), empirical fasting IR cumference was made to examine the role of central index, and fasting insulin:fasting glucose ratio across the obesity in the association. Using the standardized score of 3 racial or ethnic groups. In contrast, the correlation dependent and independent variables, a standardized coefficients were statistically significant between CRP and regression coefficient of CRP for each IR measure was Raynaud index, Bennetts index, quantitative insulin sen- calculated to facilitate the comparisons of association sitivity check index, Avignon index, and McAuley index between CRP and the IR surrogate measures with differ- across the 3 racial or ethnic groups. ent units. Multiple linear regression analyses showed that CRP Because no standardization of insulin assay is avail- was significantly associated with all 10 measures of IR able, a universal cutoff value of fasting insulin or HOMA after adjustment for age, sex, race or ethnicity, smoking to define IR is not available. Therefore, we defined the status, SBP, HDL-C, LDL-C, TG (except for the McAuley HOMA-IR using the 75th percentile of HOMA generated index) (model 1; Table 3). The associations were attenu- among nondiabetic adults in NHANES III as a cutoff ated after further adjustment for waist circumference value according to the suggestion of the European Group (model 2; Table 3). The change in adjusted regression for the Study of Insulin Resistance (27). In addition, we coefficient of CRP on IR surrogate measures between categorized CRP into 3 groups (Ͻ1,1to3,andϾ3 mg/L) model 1 and model 2 ranged from 60.0% to 75.1%. The (28). To examine the role of obesity for the association of partial R2 of CRP decreased after additional adjustment CRP with IR surrogate measures, odds ratios and 95% CI for waist circumference. of CRP for HOMA-IR were estimated in multiple logistic The associations of CRP with the surrogate measures of regression models stratified by the 3 BMI categories. An ␣ IR were statistically significant among participants with of 0.05 was used to define statistical significance for BMI Ͻ25 kg/m2 (P ranged from Ͻ0.001 to 0.005) and 2-sided tests. All analyses were conducted using SAS among participants with BMI Ն30 kg/m2 (P Ͻ0.001 for all (version 8.2) and SUDAAN software (Release 9.0, Re- measures), but not statistically significant among partici- search Triangle Institute), to account for the complex pants whose BMI was between 25 and Ͻ30 kg/m2 (P sampling design. ranged from 0.18 to 0.67) (Table 4). Participants with CRP Ͼ3 mg/L and BMI Ն30 kg/m2 Results had the highest prevalence of IR (76.2%), whereas those Among participants who met inclusion criteria for our with CRP Ͻ1 mg/L and BMI Ͻ25 kg/m2 had the lowest analyses (n ϭ 4040), we further excluded those with prevalence of IR (6.6%) (Fig. 1). After adjustment for age, missing data for fasting insulin, glucose, C-peptide, TG, sex, race/ethnicity, smoking status, SBP, serum concen- HDL-C, or LDL-C (n ϭ 1392), and for smoking, SBP, trations of HDL-C, LDL-C, and TG, the association of CRP weight, height, or waist circumference (n ϭ 134). The final (Ͼ3vsϽ1 mg/L) with HOMA-IR (Ն75th vs Ͻ75th analytic sample (n ϭ 2514; 62.3%) comprised 56.7% males, percentile) was statistically significant among people with 79.7% whites, 12.0% blacks, and 8.3% Mexican Americans. a BMI Ն30 kg/m2 (odds ratio, 2.6; 95% CI, 1.3–5.1) or with The percentage of current smokers was 25.9%, 27.3%, and a BMI Ͻ25 kg/m2 (odds ratio, 2.5; 95% CI, 1.5–4.2). The 23.7% for white, black, and Mexican-American respon- linear trend in the odds ratios of CRP (Ͻ1,1to3,Ͼ3 dents, respectively. As shown in Table 1, black and mg/L) for HOMA IR was statistically significant among Mexican-American participants were younger than people with a BMI Ͻ25 kg/m2 (P ϭ 0.0003) and with a whites (P Ͻ0.0001). Compared to white participants, BMI Ն30 kg/m2 (P ϭ 0.01), but not statistically significant blacks had higher SBP (P Ͻ0.01), fasting insulin (P Ͻ0.01), among people with a BMI between 25 and Ͻ30 kg/m2 HDL-C (P Ͻ0.01), and body weight (P Ͻ0.01), but lower (P ϭ 0.77). There was a marginally significant interaction fasting glucose (P Ͻ0.01), C-peptide (P Ͻ0.01), TG (P between CRP and BMI on HOMA-IR (P ϭ 0.07). Ͻ0.0001), and LDL-C (P Ͻ0.01). In contrast, Mexican Americans had lower SBP (P Ͻ0.01), HDL-C (P Ͻ0.01), Discussion LDL-C (P Ͻ0.01), and body weight (P Ͻ0.01), but higher Using the recent nationally representative sample, we fasting insulin (P Ͻ0.0001) than whites. There were no found that CRP was significantly associated with all 10 Clinical Chemistry 53, No. 12, 2007 2155

Table 1. Means or percentagesa of demographic characteristics, variables, and surrogate measures of IR by race or ethnicity among US adults (age >20 years), NHANES 1999–2002. Variable Total Non-Hispanic White Non-Hispanic Black Mexican American P valueb Male, n (%) 1432 (56.71) 798 (80.54) 264 (10.75) 370 (8.71) 0.02 Female, n (%) 1082 (43.29) 549 (78.57) 232 (13.74) 301 (7.69) Age, year 44.76 Ϯ 0.55 46.16 Ϯ 0.67 41.08 Ϯ 0.60c 36.63 Ϯ 0.53c Ͻ0.001 SBP, mmHg 121.51 Ϯ 0.48 121.50 Ϯ 0.59 124.19 Ϯ 0.51d 117.68 Ϯ 0.76f Ͻ0.001 CRP, mg/Le 1.61 Ϯ 0.05 1.58 Ϯ 0.06 1.75 Ϯ 0.12 1.67 Ϯ 0.13 0.38 Fasting insulin, 56.27 Ϯ 0.96 54.68 Ϯ 1.15 60.56 Ϯ 1.54d 66.64 Ϯ 1.69c Ͻ0.001 pmol/Le Fasting glucose, 5.38 Ϯ 0.54 5.41 Ϯ 0.53 5.21 Ϯ 0.54d 5.45 Ϯ 0.53 Ͻ0.001 mmol/L C-peptide, nmol/Le 0.68 Ϯ 0.01 0.69 Ϯ 0.01 0.63 Ϯ 0.01d 0.73 Ϯ 0.02 Ͻ0.001 TG, mmol/Le 1.41 Ϯ 0.03 1.46 Ϯ 0.03 1.10 Ϯ 0.04c 1.46 Ϯ 0.03 Ͻ0.001 HDL-C, mmol/L 1.30 Ϯ 0.01 1.29 Ϯ 0.02 1.39 Ϯ 0.02d 1.23 Ϯ 0.01d Ͻ0.001 LDL-C, mmol/L 3.21 Ϯ 0.02 3.25 Ϯ 0.03 3.05 Ϯ 0.05d 3.10 Ϯ 0.05d Ͻ0.001 Waist circumference, 94.96 Ϯ 0.40 95.15 Ϯ 0.53 94.44 Ϯ 0.62 93.94 Ϯ 0.74 0.45 cm Body weight, kg 80.35 Ϯ 0.47 80.25 Ϯ 0.63 83.95 Ϯ 0.92d 76.06 Ϯ 0.96d Ͻ0.001 Smoking status Nonsmoker, n (%) 1264.00 (48.48) 604.00 (45.70) 282.00 (60.18) 378.00 (58.21) Ͻ0.001 Current smoker, 593.00 (25.88) 313.00 (25.88) 136.00 (27.31) 144.00 (23.73) n (%) Former smoker, 657.00 (25.64) 430.00 (28.41) 78.00 (12.52) 149.00 (18.06) n (%) Indicators of IR HOMA 15.87 Ϯ 0.32 15.38 Ϯ 0.40 17.05 Ϯ 0.62 18.88 Ϯ 0.58c 0.0001 Log (1/HOMA) Ϫ2.58 Ϯ 0.02 Ϫ2.56 Ϯ 0.02 Ϫ2.63 Ϯ 0.03 Ϫ2.76 Ϯ 0.03c Ͻ0.0001 FIRI 14.28 Ϯ 0.29 13.84 Ϯ 0.36 15.34 Ϯ 0.56 16.99 Ϯ 0.52c 0.0001 Insulin:glucose 12.24 Ϯ 0.21 11.79 Ϯ 0.26 13.84 Ϯ 0.41f 14.26 Ϯ 0.40c Ͻ0.0001 ratio Raynaud index 0.82 Ϯ 0.02 0.84 Ϯ 0.02 0.77 Ϯ 0.02 0.69 Ϯ 0.02c Ͻ0.0001 Bennetts index 0.15 Ϯ 0.01 0.15 Ϯ 0.001 0.15 Ϯ 0.001 0.15 Ϯ 0.001c Ͻ0.0001 QUICKI 0.18 Ϯ 0.01 0.18 Ϯ 0.001 0.18 Ϯ 0.001 0.17 Ϯ 0.001 Ͻ0.0001 Avignon index 33001.15 Ϯ 842.23 33790.26 Ϯ 1033.61 30934.33 Ϯ 1011.91 28406.66 Ϯ 1056.19࿣ 0.0004 McAuley index 4.33 Ϯ 0.04 4.32 Ϯ 0.04 4.58 Ϯ 0.05d 4.09 Ϯ 0.05࿣ Ͻ0.0001 a Data are presented as weighted mean (SE) or sample size (weighted percentage). FIRI, empirical fasting insulin resistance index; QUICKI, quantitative insulin sensitivity check index. b Test for the overall difference in the means across 3 racial or ethnic groups. Significance level for multiple comparison with Bonferroni adjustment is 0.05/3 Ϸ 0.017. c P Ͻ0.0001. d P Ͻ0.01. e Geometric mean. f P Ͻ0.001; non-Hispanic whites served as a reference group. surrogate measures of IR among nondiabetic US adults. the 10 surrogate measures of IR, CRP appeared to be more This association appeared to be consistent across the 3 strongly associated with fasting insulin, Raynaud index, racial or ethnic groups. Approximately 60%–75% of the quantitative insulin sensitivity check index, and McAuley variance for the association of CRP with IR was explained index. The common feature of these 4 IR indices is that by central obesity status as measured by waist circumfer- they involve only fasting insulin or an addition of fasting ence. Among obese participants with a high concentration glucose or TG, suggesting that the fasting glucose concen- of CRP (Ͼ3 mg/L), approximately 76% had IR as defined tration does not always correctly reflect the status of IR or using the 75th percentile of HOMA. More importantly, insulin action. Previous studies have shown that fasting our data showed that CRP was significantly associated insulin alone may be a simple and effective surrogate with surrogate measures of IR among people with a BMI measure of IR (29). Our results are consistent with these Ͻ25 kg/m2. studies in that fasting insulin was at least as good as other The unique results of our study are the consistent complex indices of IR in relation to CRP. significant associations of CRP with all 10 surrogate IR Our results were consistent with previous studies that measures and across the 3 racial or ethnic groups. Among have examined associations of some measures of IR, such 2156 Meng et al.: CRP and IR

Table 2. Pearson correlation coefficientsa between CRP and surrogate measures of IR among US adults (aged >20 years), NHANES 1999–2002. Surrogate measure of IR Total Non-Hispanic White Non-Hispanic Black Mexican American P valueb Log fasting insulin, pmol/L 0.37 0.37 0.33 0.36 0.28 HOMA 0.32 0.33 0.32 0.29 0.92 Log (1/HOMA) Ϫ0.37 Ϫ0.38 Ϫ0.34 Ϫ0.35 0.23 C-peptide, nmol/L 0.38 0.39 0.36 0.33 0.06 FIRI 0.32 0.33 0.32 0.29 0.92 Insulin:glucose ratio 0.31 0.32 0.28 0.32 0.40 Raynaud index Ϫ0.34 Ϫ0.35 Ϫ0.28 Ϫ0.36 0.004 Bennetts index Ϫ0.36 Ϫ0.36 Ϫ0.32 Ϫ0.33 0.013 QUICKI Ϫ0.37 Ϫ0.37 Ϫ0.33 Ϫ0.36 0.04 Avignon index Ϫ0.34 Ϫ0.34 Ϫ0.32 Ϫ0.35 0.017 McAuley index Ϫ0.37 Ϫ0.39 Ϫ0.34 Ϫ0.33 0.009 a Weighted zero-order correlation coefficients; P values Ͻ0.001 for all correlation coefficients. FIRI, empirical fasting insulin resistance index; QUICKI, quantitative insulin sensitivity check index. b Test for the overall difference in Pearson correlation coefficients across 3 racial or ethnic groups. as fasting insulin and HOMA, with CRP in diabetic with IR. Numerous studies have shown that both CRP patients (30) and nondiabetic Asians (31). It is interesting and IR are related to obesity (28, 32–34). However, that although measurements of fasting glucose concentra- whether obesity plays a role as a moderator or mediator tions are necessary for the diagnosis of prediabetes and for the association of CRP with IR is still elusive. Recent diabetes according to the current clinical guidelines, in studies have demonstrated that obesity was a major nondiabetic individuals these measurements may not be determinant for the association of CRP with metabolic necessary for identifying IR, whereas measurements of syndrome among patients with type 2 diabetes (35) and in fasting insulin can be quite useful. Commercial laborato- the general adult population (36). Our results add further ries should consider reporting not only the absolute support for the notion that central obesity as measured by insulin concentration but also information as to where the waist circumference or overall obesity as measured by value falls within the laboratory’s frequency distribution BMI could be a mediator for the association of CRP with of nondiabetic individuals. IR among nondiabetic adults. As shown in our study, approximately 60%–75% of the In addition, consistent with recent studies (34),we variance for the association between CRP and IR can be found that this relationship was also significant among explained by waist circumference, suggesting that central people who had normal weight (BMI Ͻ25 kg/m2). Al- obesity plays an important role in the association of CRP though the mechanism for the association between CRP

Table 3. Association of CRP with surrogate measures of IR among US adults (age >20 years), NHANES 1999–2002. Model 1, without waist circumference Model 2, with waist circumference Surrogate measure ␤ a ␤ b 2c 2d ␤ ␤ 2 2 ⌬␤ of IR (SE) s P value R Partial R (SE) s P value R Partial R ,% Log fasting insulin, 0.12 (0.01) 0.29 Ͻ0.001 0.33 0.08 0.04 (0.01) 0.09 Ͻ0.001 0.49 0.01 66.7 pmol/L HOMA 2.17 (0.19) 0.16 Ͻ0.001 0.25 0.06 0.54 (0.20) 0.06 0.01 0.41 0.004 75.1 Log (1/HOMA) Ϫ0.12 (0.01) Ϫ0.27 Ͻ0.001 0.34 0.08 Ϫ0.04 (0.01) Ϫ0.08 Ͻ0.001 0.50 0.01 66.7 C-peptide, nmol/L 0.07 (0.01) 0.24 Ͻ0.001 0.32 0.06 0.02 (0.01) 0.07 Ͻ0.001 0.47 0.005 71.4 FIRI 1.95 (0.17) 0.16 Ͻ0.001 0.25 0.06 0.49 (0.18) 0.06 0.01 0.41 0.004 74.9 Insulin:glucose ratio 1.52 (0.14) 0.21 Ͻ0.001 0.25 0.06 0.41 (0.14) 0.07 0.01 0.41 0.006 73.0 Raynaud index Ϫ0.09 (0.01) Ϫ0.29 Ͻ0.001 0.28 0.06 Ϫ0.03 (0.01) Ϫ0.09 Ͻ0.001 0.40 0.01 66.7 Bennetts index Ϫ0.005 (0.001) Ϫ0.26 Ͻ0.001 0.33 0.07 Ϫ0.002 (0.001) Ϫ0.08 Ͻ0.001 0.45 0.01 60.0 QUICKI Ϫ0.004 (0.001) Ϫ0.29 Ͻ0.001 0.34 0.08 Ϫ0.001 (0.001) Ϫ0.09 Ͻ0.001 0.48 0.01 75.0 Avignon index Ϫ5754.07 (580.30) Ϫ0.3 Ͻ0.001 0.29 0.08 Ϫ1602.25 (582.54) Ϫ0.07 0.01 0.46 0.01 72.2 McAuley indexe Ϫ0.21 (0.02) Ϫ0.26 Ͻ0.001 0.39 0.07 Ϫ0.08 (0.02) Ϫ0.09 Ͻ0.001 0.49 0.01 61.9 a ␤, regress regression coefficients. Model 1 was adjusted for age, sex, race or ethnicity, smoking status, SBP, HDL-C, LDL-C, and TGs; model 2 was further adjusted for waist circumference based on Model 1. FIRI, empirical fasting insulin resistance index; QUICKI, quantitative insulin sensitivity check index; ⌬␤, relative difference ␤ ␤ Ϫ ␤ ␤ ϫ in between Model 2 and Model 1, calculated as ( model 1 model 2)/ model 1 100%. b ␤ s, standardized regression coefficients. c R2, multiple r-squared of the model. d Partial R2, partial r-squared of CRP. e TG was not included in the regression model as a covariate. Clinical Chemistry 53, No. 12, 2007 2157

Table 4. Association of CRP with surrogate measures of IR by BMI categories among US adults (aged >20 years), NHANES 1999–2002. (702 ؍ BMI >30 kg/m2 (n (937 ؍ BMI <30 kg/m2 (n > 25 (875 ؍ BMI <25 kg/m2 (n

␤ a 2b 2c ␤ 2 2 ␤ 2 2 Surrogate measure of IR s P value R Partial R s P value R Partial R s P value R Partial R Log fasting insulin, pmol/L 0.12 0.001 0.18 0.021 0.05 0.35 0.16 0.003 0.31 Ͻ0.0001 0.27 0.076 HOMA 0.04 0.002 0.18 0.015 0.02 0.32 0.16 0.002 0.26 Ͻ0.0001 0.19 0.059 Log (1/HOMA) Ϫ0.11 0.002 0.19 0.021 Ϫ0.04 0.44 0.18 0.002 Ϫ0.29 Ͻ0.0001 0.27 0.077 C-peptide, nmol/L 0.09 0.001 0.23 0.017 0.06 0.18 0.23 0.005 0.31 Ͻ0.0001 0.24 0.056 FIRI 0.04 0.002 0.18 0.015 0.02 0.32 0.16 0.002 0.26 Ͻ0.0001 0.19 0.059 Insulin:glucose ratio 0.05 0.002 0.14 0.015 0.04 0.23 0.14 0.004 0.32 Ͻ0.0001 0.22 0.057 Raynaud index Ϫ0.19 0.002 0.15 0.020 Ϫ0.04 0.42 0.16 0.002 Ϫ0.19 Ͻ0.0001 0.26 0.064 Bennetts index Ϫ0.15 0.005 0.20 0.020 Ϫ0.02 0.67 0.19 0.001 Ϫ0.22 Ͻ0.0001 0.27 0.066 QUICKI Ϫ0.15 0.002 0.19 0.021 Ϫ0.04 0.48 0.18 0.001 Ϫ0.26 Ͻ0.0001 0.27 0.074 Avignon index Ϫ0.2 0.002 0.17 0.021 Ϫ0.01 0.62 0.16 0.001 Ϫ0.1 Ͻ0.0001 0.28 0.056 McAuley indexd Ϫ0.2 Ͻ0.001 0.20 0.020 Ϫ0.04 0.46 0.26 0.001 Ϫ0.16 0.0003 0.30 0.056 a ␤ s, standardized regression coefficients; adjusted for age, sex, race or ethnicity, smoking status, SBP, HDL-C, LDL-C, and TGs. FIRI, empirical fasting insulin resistance index; QUICKI, quantitative insulin sensitivity check index. b R2, multiple r-squared of the model. c Partial R2, partial r-squared of CRP. d TG was not included in the regression model as a covariate. and IR is not fully understood, study evidence suggests of IR among people who were overweight (i.e., BMI that low-grade, chronic inflammation state may lead to IR between 25 and 29.9 kg/m2). The exact physiological and because of the role of inflammatory cytokines released biochemical mechanisms for this observation are un- from adipocytes (34). This and other previous studies (34) known, but this finding might be attributable to the poor reporting the association of CRP with IR independent of discriminatory power of BMI for body fat and lean mass. obesity suggest that another pathway linking inflamma- A recent study demonstrated that a BMI Ն30 kg/m2 tion and IR among nonobese individuals is also plausible. had good specificity but poor sensitivity, whereas a BMI In the high IR tertile, about 1 in 6 people was of normal Ն25 kg/m2 had good sensitivity but poor specificity to weight (37); thus the search for clinically useful biomar- detect obesity as defined by body fat Ͼ25% in men and kers of IR among this subpopulation is necessary. Al- 35% in women (38). Small increases of BMI as seen in though BMI, a less expensive and convenient measure, overweight people could be due to increases in body fat may serve as an indicator for IR among people with or increments in lean mass or both. Preserved and in- excessive weight, CRP could be 1 of the novel biomarkers creased lean mass have been associated with better fitness for IR among people with normal weight. and exercise capacity, whereas excessive body fatness has One of the interesting findings in our study was the been associated with adverse metabolic profiles (39). lack of association between CRP and surrogate measures Further studies are warranted to examine the mechanisms for the interrelations of CRP, IR, and body composition. Alternative methods might be needed to accurately characterize people who truly have excessive body fat vs those who have increased muscle mass, especially when their BMI is mildly increased. Our study has several strengths. First, we used 10 surrogate measures of IR and C-peptide as a measurement of ␤-cell function (26) to examine the association of CRP with IR and insulin secretion. Our results showed that the association appeared to be consistent using any of these measures. Second, we used a large representative sample of US adults; therefore, we were able to conduct analyses stratified by race or ethnicity or body weight status. Our results indicated that the association of CRP with IR is potentially generalizable across different racial Fig. 1. Age-adjusted prevalence of insulin resistance by the categories of CRP and body mass index. or ethnic groups because the association between CRP Insulin resistance was defined using the 75th percentile of HOMA among and some surrogate measures of IR (e.g., fasting insulin nondiabetic adults in NHANES III. and HOMA) appeared to be similar by race/ethnicity. 2158 Meng et al.: CRP and IR

There are several limitations in the present study. First, tions in patients who are at greatest risk of developing we used a cross-sectional design; therefore the observed diabetes, the metabolic syndrome, or cardiovascular association between CRP and surrogate measures of IR complications. may not be assumed causal. It is likely that IR could contribute to the increase of CRP (33). Second, we did not have a direct measure of IR in our data; thus, we were Grant/funding support: Y.-X.M. was a Clinical Research Edu- unable to validate the surrogate measures. Previous eval- cation and Career Development fellow partially supported by uation studies have shown that simple indices, particu- National Institutes of Health Grant 1R25RR017694-01. larly fasting insulin, are valid and reliable surrogate Financial disclosures: None declared. measures of IR in large epidemiologic studies. The 3rd Acknowledgments: We thank the Morehouse School of Med- limitation was related to the use of a single insulin assay icine Master of Science in Clinical Research Program and measured with the Pharmacia Insulin RIA reagent set. Department of Family Medicine. Because no standardization of insulin assays is available thus far, caution may be needed when comparing our References results with other assays. The cross-reactivity of Pharma- 1. Kernan WN, Inzucchi SE, Viscoli CM, Brass LM, Bravata DM, cia insulin antibody with proinsulin (approximately 40%) Horwitz RI. Insulin resistance and risk for stroke. Neurology may overestimate the true insulin concentrations in the 2002;59:809–15. 2. Reaven GM. Banting Lecture 1988. Role of insulin resistance in population. However, because proinsulin concentration is human disease. Diabetes 1988;37:1595–607. relatively low among people without diabetes (40), the 3. Centers for Disease Control and Prevention. National diabetes impact of cross-reactivity between insulin and proinsulin fact sheet: general information and national estimates on diabe- on our results could be minimal. tes in the United States, 2005. Atlanta, GA: U.S. Department of Health and Human Services, Center for Disease Control and In conclusion, because we are unable to routinely measure Prevention, 2005. IR in clinical practice, efforts to find simple measures for 4. Li C, Ford ES, McGuire LC, Mokdad AH, Little RR, Reaven GM. Trends in hyperinsulinemia among nondiabetic adults in the U.S. IR are ongoing. One of the clinical challenges of identify- Diabetes Care 2006;29:2396–402. ing individuals with IR is the cumbersome nature of the 5. Esposito K, Pontillo A, Di Palo C, Giugliano G, Masella M, Marfella assays. Obesity, particularly central obesity, plays an R, et al. Effect of weight loss and lifestyle changes on vascular important role for the association between CRP and IR; inflammatory markers in obese women: a randomized trial. JAMA however, it is conceivable that the use of CRP in clinical 2003;289:1799–804. settings could facilitate an earlier identification of IR 6. Lindstrom J, Ilanne-Parikka P, Peltonen M, Aunola S, Eriksson JG, among people who may not achieve certain clinical Hemio K, et al. Sustained reduction in the incidence of type 2 diabetes by lifestyle intervention: follow-up of the Finnish Diabetes thresholds for measures of adiposity or fasting glucose. In Prevention Study. Lancet 2006;368:1673–9. addition, by providing a simple and cost-effective way to 7. Yeni-Komshian H, Carantoni M, Abbasi F, Reaven GM. Relation- identify high-risk individuals with IR, the measurement ship between several surrogate estimates of insulin resistance of CRP concentration might facilitate preventive interven- and quantification of insulin-mediated glucose disposal in 490 healthy nondiabetic volunteers. Diabetes Care 2000;23:171–5. 8. Robbins DC, Andersen L, Bowsher R, Chance R, Dinesen B, Frank Appendix B, et al. Report of the American Diabetes Association’s Task Force

a on standardization of the insulin assay. Diabetes 1996;45:242– Formulas for calculating selected surrogate measures of IR 56. Surrogate 9. Ridker PM, Koenig W, Fuster V. C-reactive protein and coronary measure of IR Formula References heart disease. N Engl J Med 2004;351:295–8. Fasting insulin, NA (7) 10. Ridker PM, Wilson PW, Grundy SM. Should C-reactive protein be pmol/L added to metabolic syndrome and to assessment of global HOMA [FI (pmmol/L) * FG (mmol/L)]/22.5 (17) cardiovascular risk? Circulation 2004;109:2818–25. Log 1/HOMA ln (1/HOMA) (19) 11. Visser M, Bouter LM, McQuillan GM, Wener MH, Harris TB. C-peptide, NA (26) Elevated C-reactive protein levels in overweight and obese adults. nmol/L JAMA 1999;282:2131–5. FIRI [FI (pmmol/L) * FG (mmol/L)]/25 (14) 12. Taniguchi A, Nagasaka S, Fukushima M, Sakai M, Okumura T, I:G ratio FI/FG (pmol/mmol) (15) Yoshii S, et al. C-reactive protein and insulin resistance in Raynaud index 40/[FI (pmmol/L)] (20) non-obese Japanese type 2 diabetic patients. Metabolism 2002; Bennetts index 1/[ln FG (mmol)/L * ln FI (pmmol)/L] (21) 51:1578–81. QUICKI 1/[ln FG(mmol/L) ϩ ln FI (pmmol/L)] (16) 13. Avignon A, Boegner C, Mariano-Goulart D, Colette C, Monnier L. 8 Assessment of insulin sensitivity from plasma insulin and glucose Avignon index 10 /[FI (pmmol/L) * FG (mmol/L) * Vd], (13) Vd ϭ 0.65 ϫ 26% of body weight in the fasting or post oral glucose-load state. Int J Obes Relat Ϫ Metab Disord 1999;23:512–7. McAuley index e[2.63–0.28*ln(FI pmmol/L) 0.31*ln(FTG mmol/L)] (18) 14. Duncan MH, Singh BM, Wise PH, Carter G, Alaghband-Zadeh J. A a NA, not applicable; FI, fasting insulin; FG, fasting glucose; FTG, fasting simple measure of insulin resistance. Lancet 1995;346:120–1. triglyceride; I, insulin; G, glucose; FIRI, empirical fasting insulin resistance index; 15. Hanson RL, Pratley RE, Bogardus C, Narayan KM, Roumain JM, QUICKI, quantitative insulin sensitivity check index. Imperatore G, et al. Evaluation of simple indices of insulin Clinical Chemistry 53, No. 12, 2007 2159

sensitivity and insulin secretion for use in epidemiologic studies. 28. Ridker PM. Clinical application of C-reactive protein for cardiovas- Am J Epidemiol 2000;151:190–8. cular disease detection and prevention. Circulation 2003;107: 16. Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, Sullivan G, 363–9. et al. Quantitative insulin sensitivity check index: a simple, 29. Laakso M. How good a marker is insulin level for insulin resis- accurate method for assessing insulin sensitivity in humans. tance? Am J Epidemiol 1993;137:959–65. J Clin Endocrinol Metab 2000;85:2402–10. 30. Iwasaki T, Nakajima A, Yoneda M, Terauchi Y. Relationship 17. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, between the serum concentrations of C-reactive protein and Turner RC. Homeostasis model assessment: insulin resistance parameters of adiposity and insulin resistance in patients with and beta-cell function from fasting plasma glucose and insulin type 2 diabetes mellitus. Endocr J 2006;53:345–56. concentrations in man. Diabetologia 1985;28:412–9. 18. McAuley KA, Williams SM, Mann JI, Walker RJ, Lewis-Barned NJ, 31. Nakanishi N, Shiraishi T, Wada M. Association between C-reactive Temple LA, et al. Diagnosing insulin resistance in the general protein and insulin resistance in a Japanese population: the population. Diabetes Care 2001;24:460–4. Minoh Study. Intern Med 2005;44:542–7. 19. Garcia-Estevez DA, Araujo-Vilar D, Fiestras-Janeiro G, Saavedra- 32. Ford ES. Body mass index, diabetes, and C-reactive protein Gonzalez A, Cabezas-Cerrato J. Comparison of several insulin among U.S. adults. Diabetes Care 1999;22:1971–7. sensitivity indices derived from basal plasma insulin and glucose 33. McLaughlin T, Abbasi F, Lamendola C, Liang L, Reaven G, Schaaf levels with minimal model indices. Horm Metab Res 2003;35: P, et al. Differentiation between obesity and insulin resistance in 13–7. the association with C-reactive protein. Circulation 2002;106: 20. Raynaud E, Perez-Martin A, Brun JF, Benhaddad AA, Mercier J. 2908–12. Revised concept for the estimation of insulin sensitivity from a 34. Yudkin JS, Stehouwer CD, Emeis JJ, Coppack SW. C-reactive single sample. Diabetes Care 1999;22:1003–4. protein in healthy subjects: associations with obesity, insulin 21. Anderson RL, Hamman RF, Savage PJ, Saad MF, Laws A, Kades resistance, and endothelial dysfunction: a potential role for cyto- WW, et al. Exploration of simple insulin sensitivity measures kines originating from adipose tissue? Arterioscler Thromb Vasc derived from frequently sampled intravenous glucose tolerance Biol 1999;19:972–8. (FSIGT) tests. The Insulin Resistance Atherosclerosis Study. Am J Epidemiol 1995;142:724–32. 35. Kahn SE, Zinman B, Haffner SM, O’Neill MC, Kravitz BG, Yu D, et 22. Centers for Disease Control and Prevention. NHANES 1999–2000 al. Obesity is a major determinant of the association of C-reactive public data release file documentation. http://www.cdc.gov/nchs/ protein levels and the metabolic syndrome in type 2 diabetes. about/major/nhanes/nhanes99–00.htm (accessed December 15, Diabetes 2006;55:2357–64. 2005). 36. Santos AC, Lopes C, Guimaraes JT, Barros H. Central obesity as 23. Centers for Disease Control and Prevention. NHANES 2001–2002 a major determinant of increased high-sensitivity C-reactive pro- public data release file documentation. http://www.cdc.gov/nchs/ tein in metabolic syndrome. Int J Obes 2005;29:1452–6. about/major/nhanes/nhanes01–02.htm (accessed December 15, 37. McLaughlin T, Allison G, Abbasi F, Lamendola C, Reaven G. 2005). Prevalence of insulin resistance and associated cardiovascular 24. Centers for Disease Control and Prevention. The Third National disease risk factors among normal weight, overweight, and obese Health and Nutrition Examination Survey (NHANES III 1988–94) individuals. Metabolism 2004;53:495–9. reference manuals and reports [CD-ROM]. Bethesda, MD: Na- 38. Romero-Corral A, Somers VK, Sierra-Johnson J, Jensen MD, tional Center for Health Statistics, 1996. Thomas RJ, Squires RW, et al. Diagnostic performance of body 25. World Health Organization. Physical Status: The Use and Interpre- mass index to detect obesity in patients with coronary artery tation of Anthropometry. Report of a WHO Expert Committee. disease. Eur Heart J 2007;28:2087–93. Technical Report Series No. 854. Geneva, Switzerland: World Health Organization, 1995. 39. Goldberg AP, Busby-Whitehead MJ, Katzel LI, Krauss RM, Lumpkin 26. Polonsky KS, Pugh W, Jaspan JB, Cohen DM, Karrison T, Tager M, Hagberg JM. Cardiovascular fitness, body composition, and HS, et al. C-peptide and insulin secretion: relationship between lipoprotein lipid metabolism in older men. J Gerontol A Biol Sci peripheral concentrations of C-peptide and insulin and their Med Sci 2000;55:M342–9. secretion rates in the dog. J Clin Invest 1984;74:1821–9. 40. Reaven GM, Chen YD, Hollenbeck CB, Sheu WH, Ostrega D, 27. Balkau B, Charles MA. Comment on the provisional report from Polonsky KS. Plasma insulin, C-peptide, and proinsulin concentra- the WHO consultation. European Group for the Study of Insulin tions in obese and nonobese individuals with varying degrees of Resistance (EGIR). Diabet Med 1999;16:442–3. glucose tolerance. J Clin Endocrinol Metab 1993;76:44–8. Clinical Chemistry 53:12 2160–2168 (2007) Endocrinology and Metabolism

Frank Giton,1,2* Saı¨k Urien,3 Catherine Born,4 Jean Tichet,4 Je´roˆ me Gue´chot,5 Jacques Callebert,6 Franc¸oise Bronsard,7 Jean Pierre Raynaud,8 and Jean Fiet2

Background: Bioavailable testosterone (BT) is measured was found to be 2-fold higher than for aBT. With ؋ 4 ؍ ؋ 9 ؍ [assayed BT (aBT)] or calculated (cBT) in the diagnosis optimized Ks 1.9 10 L/mol and Ka 2.45 10 of hypogonadism in men. The cBT depends, however, L/mol, cBT values close to aBT were obtained. When on the values of the association constants of total testos- 5-diol was included in the model as a competitive SHBG terone (TT) for sex hormone–binding globulin (SHBG; inhibitor, the correlation between cBT and aBT was Ks) and albumin (Ka), and its use therefore remains better and the cBT:aBT ratios vs 5-diol were less biased. controversial. Conclusion: Lower normal serum aBT concentration in Methods: In 503 selected, untreated healthy men, 20–74 normal men appears to be between 2.30 and 2.72 nmol/L. years old, we measured TT, dihydrotestosterone (DHT), Much higher serum cBT concentrations are associated and androstenediol (5-diol) by GC-MS, SHBG by RIA, with use of different association constants that may be and BT after ammonium sulfate precipitation or by inappropriate. When using the optimized binding con- calculation according to the law of mass action. stants, taking age-related 5-diol values into consider- Results: A slight decrease in TT, significant decreases in ation slightly improves prediction of cBT. BT and 5-diol, no variation in DHT, and an increase in © 2007 American Association for Clinical Chemistry SHBG were observed with age. In young males (<39 years), the lower normal limits were between 2.30 and 9 ؋ Bioavailable testosterone (BT) appears to provide a more ؍ 2.72 nmol/L for aBT and 8.50 nmol/L for TT. For Ks 1 ؋ 4 accurate view of testosterone status than total testosterone ؍ 9 10 L/mol and Ka 3.6 10 L/mol, the lower cBT limit (TT) (1). According to recommendations of the Interna- tional Society for the Study of the Aging Male (ISSAM) (2), the most reliable and widely accepted parameters for 1 Assistance Publique-Hoˆpitaux de Paris (AP-HP), Biological Center of confirming hypogonadism are measurement of BT or, Investigations, University Hospital Group, Sud Henri Mondor, Faculte´de Me´decine, Cre´teil, France. alternatively, calculating free testosterone and BT. How- 2 Institut National de la Sante´ et de la Recherche Me´dicale Unite´ 841 Eq07, ever, the lower limits for demonstrating testosterone Centre Hospitalier Universitaire Henri Mondor, Faculte´deMe´decine, Cre´teil, deficiency in both young and aging men have not been France. 3 Unite´ de Recherche Clinique Paris Centre, Hoˆpital Tarnier, Paris, France. established using reference methods in a sufficiently large 4 Institut Inter-Re´gional pour la Sante´, La Riche, France. population. 5 AP-HP, Laboratoire d’Hormonologie, Hoˆpital St. Antoine, Paris, France. 6 AP-HP, Laboratoire de Biochimie, Hoˆpital Lariboisie`re, Paris, France. 7 Institut Universitaire de Technologie de Cachan, Cachan, France. 8 Universite´ Paris VI, Pierre et Marie Curie, Paris, France. * Address correspondence to this author at: Centre de Recherches Chirur- 9 Nonstandard abbreviations: BT, bioavailable testosterone; TT, total tes- gicales, Faculte´deMe´decine, 8 rue du Ge´ne´ral Sarrail, 94010 Cre´teil Cedex, tosterone; ISSAM, International Society for the Study of the Aging Male; cBT, France. Fax 33-1-49-81-35-52; e-mail [email protected]. calculated BT; SHBG, sex hormone–binding globulin; aBT, assayed BT; It, Received February 10, 2007; accepted September 18, 2007. inhibitor testosterone; DHT, dihydrotestosterone; 5-diol, androstenediol; Previously published online at DOI: 10.1373/clinchem.2007.087312 DHEA, dehydroepiandrosterone.

2160 Clinical Chemistry 53, No. 12, 2007 2161

Various normal lower limits have been proposed for HP5973 (Agilent Technologies) quadrupole mass spec- measured TT and BT, usually based on studies carried out trometer was used for detection. in rather small groups of normal men. In addition, the use The interassay CVs (n ϭ 23) were 6.0%, 2.8%, and 2.0% of a simplified formula for calculating BT [calculated BT for TT (2.50, 10.47, and 20.59 nmol/L, respectively); 1.9%, (cBT)] using specific association constants of testosterone 2.7%, and 3.7% for DHT (0.86, 3.55, and 8.26 nmol/L); and for sex hormone–binding globulin (SHBG) and for albu- 13.7%, 8.2%, and 6.0% for 5-diol (0.90, 3.37, and 6.85 min (3) seems to yield higher calculated than measured nmol/L). The lower limits of quantification for TT, DHT, BT [assayed BT (aBT)] (4). This formula takes into account and 5-diol were 0.17, 0.069, and 0.34 nmol/L, respectively. only binding of testosterone to albumin and SHBG in determining BT, although other steroids with nonnegli- BT assay. We measured the percentage BT (%BT) as gible serum concentrations in males, such as dihydrotes- previously described (4). Briefly, after adding minute tosterone (DHT) and androstenediol (5-diol), also bind to doses of purified tritiated testosterone to serum samples, SHBG. followed by incubation at 37 °C, we used a solution of To take into account the interference of DHT and saturated ammonium sulfate to precipitate the SHBG- 5-diol, we used reference analytical methods to determine bound testosterone. The precipitate was then centrifuged, TT, DHT, aBT, 5-diol, and SHBG in a selected group of 503 and we deduced the percentage of SHBG-unbound triti- men age 20 to 74 years. We then calculated cBT and ated testosterone (BT) from the radioactivity measure- compared it with aBT. To demonstrate that BT did not ments. We then multiplied the %BT by the serum TT depend only on TT and SHBG, we compared aBT in pairs concentration and calculated the concentration of bio- of young and older men in whom the SHBG concentra- available serum testosterone. tions were identical and the TT concentrations were nearly identical. SHBG assay. We used the Schering SHBG-RIACT radio- immunometric kit to measure SHBG. The interassay CVs Materials and Methods for SHBG were 7.1%, 8.2%, and 5.6% (15.8, 46.3, and 68.2 population studied nmol/L). Volunteer men age 20 to 74 years and representing the general French population were recruited in a health Gonadotropins and albumin assays. We measured follicle- center in the city of Tours, France, the Institut Inter- stimulating hormone and luteinizing hormone using ra- Re´gional pour la Sante´. The recruitment procedure was in dioimmunometric kits (Beckman Coulter luteinizing hor- accordance with the Helsinki Declaration, and each par- mone IM 1381; follicle-stimulating hormone IM 2125). We ticipant provided prior written informed consent. All used the bromcresol green method (Olympus AU 800) to volunteers were excluded who were taking drugs known assay the albumin concentration. to influence the hypothalamic-pituitary-gonadal axis function, had a body mass index Ն29 kg/m2, or were suffering from chronic disease. Blood was drawn between calculations 0800 and 1000 AM after a 12-h overnight fast. Serum for Calculation of BT. cBT was calculated according to Ver- hormone study was isolated from the blood, divided into meulen’s formula (3), but applying association constants ϫ 9 ϫ Ϫ of TT for SHBG (Ks) varying between 0.6 10 and 2 several samples, and frozen at 20 °C. Samples of the sera 9 of the 539 men retained for the study were sent to our 10 L/mol and association constants for albumin (Ka) ϫ 4 ϫ 4 hormone biochemistry laboratory for analysis, and we varying between 0.7 10 and 3.6 10 L/mol. The cBT measured their glucose, creatinine, triglycerides, ␥-glu- was then compared with the aBT. Comparison of aBT and tamyl transferase, alanine aminotransferase (ALT), and cBT was based on counting the number of samples in Ͻ aspartate aminotransferase (AST). All volunteers exhibit- which the cBT differed from the aBT by 10%, 20%, and ing abnormal biochemical values were excluded from the 30% (4). The (cBT-aBT):aBT ratio (relative difference be- study. tween cBT and aBT) was determined for each sample. We also calculated the cBT:aBT correlation coefficients (r) and hormone assays cBT:aBT ratios for each sample. TT, DHT, and 5-diol. Serum TT, DHT, and 5-diol were We also carried out an alternative calculation that took measured by GC-MS according to Labrie et al. (5). Briefly, into account the possible inhibitory effect of 5-diol on 0.5 mL serum was extracted after adding deuterated testosterone binding to SHBG, according to published steroid internal standards. The organic extracts were equations (6). Accordingly, new equations were derived purified on conditioned Bond Elut-Si (Varian, ref. for the determination of cBT. 12102037). After derivatization with pentafluorobenzoyl- chloride (Aldrich, ref. 10 377-2), TT, DHT, and 5-diol cBT correction for the presence of 5-diol as a competitive analytes were separated on a GC system (6890N Agilent inhibitor. Protein-bound testosterone (B) and protein- Technologies) using a 50% phenylmethylpolysiloxane bound inhibitor (Ib) concentrations in the presence of capillary column (J&W Scientific, ref. 122-1831). An 5-diol were calculated from the following equations (6): 2162 Giton et al.: Measurement of BT in a Healthy French Population

ϭ ⅐ ⅐ ϩ ⅐ ϩ ⅐ ϩ ⅐ ⅐ B Ks F SHBG / (1 Ks F Ksi If) Ka F A Comparison of aBT in young and older men with the same serum SHBG concentrations and nearly identical TT concen- (1) trations. Classically, BT is considered to depend princi- pally on TT and SHBG (normal slight physiological vari- Ib ϭ K ⅐ If ⅐ SHBG / (1 ϩ K ⅐ F ϩ K ⅐ If) ϩ K ⅐ If ⅐ A si s si a ations in the albumin concentration do not influence cBT (2) results). Thus, sera with the same SHBG and TT will theoretically have the same BT. We hypothesized that BT in which F, If, SHBG, and A stand for serum concentra- could also depend on factors other than TT and SHBG tions of free testosterone, free inhibitor, total SHBG, and that are not taken into account in calculating the cBT. We albumin, respectively. Ks and Ksi are the respective asso- compared aBT in pairs of young and aging men in whom ciation constants of TT and 5-diol for SHBG (7). Ka is the the SHBG was identical and the TT not more than 10% proportional binding constant of testosterone and 5-diol ϭ ϫ 9 different. to albumin. The binding constants were Ks 1.9 10 L/mol, K ϭ 1.5 ϫ 109 L/mol, and K ϭ 2.45 ϫ 104 L/mol. si a Measurement of aBT in charcoal-treated sera overloaded with The total plasma concentrations of testosterone (TT) TT and 5-diol or dehydroepiandrosterone. We measured BT in and inhibitor (It) were as follows: 2 pools of charcoal-treated sera containing the same TT ϭ B ϩ F (3) concentrations of testosterone (13.7 nmol/L) and either 45.5 nmol/L or 17.7 nmol/L of SHBG. Each pool was ϭ ϩ It Ib If . (4) divided into samples overloaded with increasing quantities of 5-diol or of dehydroepiandrosterone (DHEA). We Given the binding constants for testosterone and inhibitor ϭ and the TT and It concentrations, the free concentrations then measured BT as previously described (n 6). of the ligands F and If were determined by iterative minimization of the function f: statistics We used the nonparametric paired Wilcoxon and Spear- f ϭ (TT Ϫ B Ϫ F)2 ϩ (It Ϫ Ib Ϫ If)2 . (5) man regression tests.

This determination was done using the Nelder–Mead Results algorithm implemented in the R program (R Develop- Of the 539 individuals age 20 to 74 years, 36 with ment Core Team). Free testosterone concentrations were abnormally high or low serum gonadotropin concentra- then used to determine cBT. tions were excluded. The remaining 503 serum samples were then stratified into groups of increasing age: 20–39 Simplified solution for cBT determination using the TT and (n ϭ 142); 40–49 (n ϭ 100); 50–59 (n ϭ 135); and 60–74 5-diol concentrations. The free 5-diol fraction in the 503 years (n ϭ 126). patients was characterized by a median of 2.42%, with an interquartile range of 1.94%–2.92%. This narrow range of normal tt, bt, dht, and 5-diol values and shbg free fraction variation suggested that the free 5-diol con- variation expressed in percentiles by age centration could be approximated as follows: group (table 1) If ϭ 0.024 ϫ [5-diol] . (6) The percentiles of total participants and age-stratified subgroups are reported in Table 1. In the group of young Using this approximation, an analytical solution to F can men (20–39), the 1st and 5th percentiles of TT and BT be derived for Eqs. 1 and 3: were 8.49 and 9.67 nmol/L and 2.70 and 3.95 nmol/L, respectively, whereas in the overall population studied aF2 ϩ bF ϩ c ϭ 0 (7) they were 7.45 and 9.33 nmol/L and 2.08 and 2.60 with nmol/L. We observed significant age-related decreases in P Ͻ P Ͻ ϭ ϩ BT and 5-diol ( 0.0001) and TT ( 0.0048), as well as a Ks (1 Ka A) (8) a significant increase in SHBG (P Ͻ0.0001), but no signif- ϭ ϩ ϩ Ϫ icant variation in DHT with age (P ϭ 0.4499). b (1 Ki If) (1 Ka A) TT Ks (9) ϭϪ ϩ c TT (1 Ksi If) (10) comparison of aBT and cBT We calculated cBT according to the association constants 9 r⌬ϭ͌(b2 Ϫ 4ac) (11) of testosterone for SHBG, 1 ϫ 10 L/mol, and for albumin, 3.6 ϫ 104 L/mol (values in the ISSAM website) (3). The F ϭ (Ϫb ϩ r⌬)/(2a) (12) values obtained for cBT were higher (P Ͻ0.0001) than for aBT (Table 2), although aBT and cBT were correlated (r ϭ then 0.831; mean cBT:aBT ratio 1.63; Table 3). Moreover, the cBT ϭ F ϩ Ka ⅐ FA . (13) cBT:aBT ratio significantly increased with age, according Clinical Chemistry 53, No. 12, 2007 2163

Table 1. Normal serum values of TT, BT, DHT, 5-diol, SHBG, and albumin in untreated healthy men according to age. Age range, years Percentile TT, nmol/L BT, nmol/L DHT, nmol/L 5-diol, nmol/L SHBG, nmol/L Albumin, g/L 20–39 (n ϭ 142) 1st 8.49 2.70 0.48 2.27 7.14 37.5 5th 9.67 3.95 0.79 3.00 11.90 40.4 25th 13.17 5.44 1.17 4.58 19.05 42.6 50th 16.71 6.59 1.69 6.06 26.19 44.3 75th 21.63 8.46 2.10 8.33 33.33 46.5 95th 29.16 11.30 2.89 13.53 45.24 48.9 99th 35.74 12.93 3.62 18.11 60.71 50.0 40–49 (n ϭ 100) 1st 8.53 2.77 0.65 2.07 9.52 37.0 5th 9.98 3.19 0.69 2.65 13.10 39.1 25th 13.49 4.47 1.27 4.03 22.62 42.2 50th 17.06 5.82 1.72 4.89 29.76 43.8 75th 20.04 7.07 2.17 5.96 36.90 45.6 95th 30.02 8.98 3.00 7.88 50.00 48.7 99th 39.42 10.57 3.31 9.26 66.67 49.7 50–59 (n ϭ 135) 1st 6.31 1.73 0.62 1.00 13.10 34.2 5th 8.36 2.46 0.86 1.89 19.05 38.6 25th 12.03 3.61 1.45 3.03 25.00 40.5 50th 15.84 4.37 1.86 4.10 34.52 43.1 75th 20.21 5.41 2.48 5.72 44.05 45.0 95th 25.41 8.08 3.31 9.30 55.95 47.7 99th 26.94 9.15 4.41 14.29 86.90 49.6 60–74 (n ϭ 126) 1st 6.55 1.80 0.62 0.93 16.67 34.1 5th 9.71 2.25 0.93 1.55 20.24 37.1 25th 12.41 3.26 1.38 2.79 27.38 39.3 50th 16.09 4.23 1.93 3.86 35.71 41.2 75th 18.55 5.17 2.34 5.13 41.67 43.4 95th 26.38 6.90 3.62 7.33 69.05 46.1 99th 30.65 7.80 4.41 10.33 76.19 47.2 20–74 (n ϭ 503) 1st 7.45 2.08 0.48 1.45 10.71 35.8 5th 9.33 2.60 0.83 2.00 14.29 38.2 25th 12.97 3.95 1.34 3.55 23.81 40.8 50th 16.40 5.06 1.79 4.61 30.95 43.1 75th 17.02 6.66 2.31 6.30 40.48 45.4 95th 26.94 9.53 3.20 10.67 54.76 48.2 99th 35.19 12.48 4.17 15.46 75.00 49.6 to the regression equation cBT:aBT ϭ 0.009 ϫ years ϩ influence of 5-diol on testosterone binding 1.208 (r ϭ 0.598). To find out whether BT depended only on TT and SHBG, we compared BT values in young and aging individuals optimization of Ks and Ka association in whom the SHBG values were identical and the TT constants values were nearly identical (Table 4). Using optimal paired association constants, the number of Comparison of SHBG and TT among those in the 20–39 cBT values differing by Ͻ30% from the aBT was between and 40–49 age groups yielded 55 sample pairs with 421 and 428 (among the 503 normal participants), whereas identical SHBG and TT values in the latter group that applying the constants used in the ISSAM website (3), were Ͻ10% different from those in the former group only 67 of the cBT values differed by Ͻ30% from the aBT (Table 4). The mean aBT (5.83 nmol/L) in the older group (Table 3). was significantly lower (P Ͻ0.0001) than the mean aBT ϭ ϫ 9 We show that for the same Ks 1.9 10 L/mol, the (6.7 nmol/L) in the younger group. Comparisons of pairs corresponding optimal Ka values decreased according to of individuals in other groups yielded similar observa- age range, as reported in Table 3. tions (Table 4). In our overall normal population, the optimal cBT The mean DHT was not significantly different in pairs values (reported as a percentile curve in Fig. 1) could be of younger and aging participants with the same SHBG superimposed over the aBT values, whereas cBT values and nearly identical TT (Table 4). Conversely, under these calculated from the Ks and Ka used in the ISSAM website conditions, a significant decline in mean 5-diol was ob- formula led to a higher shift in the cBT curve (Fig. 1). served in the aging compared with younger participants. 2164 Giton et al.: Measurement of BT in a Healthy French Population

Table 2. aBT and cBT values according to Vermeulen formula and using association constants for SHBG and albumin ؋ 4 ؍ ؋ 9 ؍ (respectively, Ks 1 10 L/mol, Ka 3.6 10 L/mol) and optimal bioavailable testosterone calculated according to

paired optimal Ks and Ka as reported in Table 3 for each age group. Age range, years Percentile aBT, nmol/L cBT, nmol/L Optimal cBT, nmol/L 20–74 (n ϭ 503) 5th 2.60 4.57 2.27 50th 5.06 8.07 4.89 95th 9.53 14.94 10.66 20–39 (n ϭ 142) 5th 3.95 5.44 3.19 50th 6.59 9.76 6.52 95th 11.30 16.36 12.69 40–49 (n ϭ 100) 5th 3.19 4.72 2.84 50th 5.82 8.67 5.14 95th 8.98 15.63 11.42 50–59 (n ϭ 135) 5th 2.46 3.66 1.83 50th 4.37 7.38 4.27 95th 8.08 11.80 7.60 60–74 (n ϭ 126) 5th 2.25 3.82 2.02 50th 4.23 7.24 4.19 95th 6.90 12.07 7.64 relation between in vitro 5-diol and bt in sera containing increasing concentrations of 5-diol concentrations (2.13, 4.27, 8.54, and 17.11 nmol/L) revealed an increase in For the same TT (13.87 nmol/L) and SHBG concentrations serum aBT concentrations compared with sera with no in charcoal-treated pooled serum (older men, 45.5 5-diol (Fig. 2; aging men, 7%, 11%, 20%, and 38%; younger nmol/L; younger men, 17.7 nmol/L), measurement of BT men, 0%, 2%, 6%, and 16%.) Overloading with 3.47, 6.93,

Table 3. Optimization and comparison between aBT and cBT in 503 samples of normal subjects (full population) and

groups of increasing age, using optimal paired and nonoptimal Ks and Ka. ؋ 4 ؋ 9 Ka, 10 L/mol Ks, 10 L/mol Slope r cBT/aBT mean SD RD <0.1 RD <0.2 RD <0.3 Age range, years 0.7 0.6 1.026 0.876 1.018 0.220 178 351 428 Full population 20–74 years (n ϭ 503) 0.8 0.7 1.005 0.875 0.993 0.217 165 341 424 0.9 0.8 0.989 0.875 0.974 0.215 163 326 423 1.1 0.9 1.015 0.876 1.004 0.219 172 346 425 1.2 1.0 1.001 0.875 0.988 0.217 165 338 424 1.3 1.1 0.990 0.875 0.975 0.216 164 326 422 1.5 1.2 1.009 0.876 0.997 0.218 166 342 423 1.7 1.3 1.025 0.876 1.016 0.220 175 349 427 1.7 1.4 0.990 0.875 0.975 0.216 163 325 422 1.9 1.5 1.006 0.875 0.993 0.218 166 339 423 2.1 1.6 1.019 0.876 1.009 0.219 174 346 424 2.1 1.7 0.990 0.875 0.975 0.216 162 326 422 2.3 1.8 1.003 0.875 0.990 0.217 164 340 421 2.5 1.9 1.015 0.876 1.004 0.219 171 344 423 2.6 2.0 1.008 0.876 0.996 0.218 164 340 422 3.6 1.0 1.535 0.831 1.630 0.336 9 26 67 3.0 1.9 1.063 0.865 1.051 0.206 52 95 122 20–39 years (n ϭ 142) 3.6 1.0 1.422 0.847 1.457 0.264 8 21 45 2.6 1.9 1.101 0.868 1.055 0.238 32 64 80 40–49 years (n ϭ 100) 3.6 1.0 1.599 0.861 1.618 0.305 1 3 10 2.5 1.9 0.971 0.832 0.975 0.200 45 104 120 50–59 years (n ϭ 135) 3.6 1.0 1.580 0.796 1.657 0.300 0 2 9 2.3 1.9 0.992 0.831 0.995 0.226 42 87 107 60–74 years (n ϭ 126) 3.6 1.0 1.721 0.784 1.802 0.381 0 0 3 2.45 1.9 1.090 0.862 1.004 0.223 119 250 303 40–74 years (n ϭ 361) 3.6 1.0 1.372 0.849 1.695 0.339 1 5 22 a Slopes of the regression curves between aBT and cBT, correlation coefficients (r), means cBT/aBT, and ͉RD͉. Clinical Chemistry 53, No. 12, 2007 2165

theoretical: y ϭ 1), whereas the cBT:aBT ratio vs the 5-diol plasma concentration showed a significant bias. Indeed, the cBT:aBT ratio rapidly decreased below 1 when the 5-diol concentration was Ͼ5 nmol/L. However, the exact determination of cBT in the presence of 5-diol requires an iterative process. Therefore, assuming a constant median free fraction of plasma 5-diol (2.4%), we were able to derive an analytical solution for the calculation of free testosterone allowing determination of cBT(5-diol; see Eqs. 6–13 in Calculations of BT). Correlation between the analytical cBT(5-diol) and the exact BT(5-diol) was 0.999.

Discussion normal limits of tt, bt, dht, shbg, and 5-diol Normal or usual hormone concentration values depend on several factors, such as accuracy of assay methods Fig. 1. Percentiles of BT normal values in 503 untreated healthy men (whole population) age 20–74 years. used, the population studied and its age, and choice of BT was measured (E ϭ aBT) or calculated according to paired values of statistical expression of normal limits. Because assayed ϫϭ optimized association constants Ks and Ka [ opt (optimized) BT] and hormone plasma concentrations do not have a gaussian nonoptimized (ISSAM website, ϩϭcBT) association constants. distribution, we preferred to express the results of serum concentrations in percentile form. 13.87, and 27.74 nmol/L DHEA revealed an increase in As reported (8–12), we found a slight but significant aBT of 1%, 2%, 3%, and 5% in the aging men’s pooled sera decrease with age in TT and a steeper decrease in BT, with and no increase in younger men. Overloading with the a significant concomitant increase in SHBG. As others same concentrations of 5-diol plus DHEA as above led to have, we found no significant variation in DHT (13–15) an increase in aBT of 6%, 12%, 21%, and 41% in the sera of and a significant decrease in 5-diol (14, 16) with age. the aging men, and 1%, 1%, 6%, and 15% in younger men. The aim of this study was to define the limits of normal serum BT values obtained from the results of accurate influence of 5-diol concentrations on the GC-MS steroid assay in a population of untreated healthy serum cBT and cBT:aBT ratio French men, to best identify a state of hypogonadism. In We also determined the cBT(5-diol; BT calculated taking a group of 142 young men (Ͻ40 years), we observed a into account the 5-diol plasma concentration), assuming mean TT concentration of 9.67 nmol/L, corresponding to competitive inhibition between testosterone and 5-diol on the 5th percentile (Table 1)—that is, Ͻ11 nmol/L, consid- SHBG (see Eqs. 1–4 in Calculations of BT. The cBT(optimal) ered the lower limit of TT calculated from the mean Ϫ2.5 vs aBT correlation was 0.876, whereas the cBT(5-diol) vs SD obtained in a population of healthy young men in aBT correlation was 0.888. In addition, the cBT(5-diol):aBT whom the population distribution was considered to be ratio vs the 5-diol plasma concentration (Fig. 2) did not normal after log transformation (2, 17). These TT limits show any bias (regression equation: y ϭ 1.09 Ϫ 0.002x; are interesting, because a clinically symptomatic thresh-

Table 4. Comparison of aBT, TT, DHT, and 5-diol in pairs of samples from young and older subjects exhibiting the same .SHBG values and TT in older individuals equal to TT in young individuals ؎10% 20–39 40–49 20–39 50–59 20–39 60–74 40–49 50–59 40–49 60–74 50–59 60–74 aBT mean 6.69 5.83 6.35 5.51 6.35 5.10 5.10 4.79 5.24 4.51 4.82 4.23 SD 1.87 1.63 1.63 1.39 1.53 1.32 1.25 1.11 1.21 1.25 1.35 1.21 P Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 0.0026 Ͻ0.0001 Ͻ0.0001 TT mean 16.51 16.35 17.16 17.09 17.23 17.10 15.24 15.42 15.80 15.80 15.94 15.90 SD 3.79 3.57 3.76 3.81 4.04 3.87 3.69 3.71 3.92 3.94 3.90 3.94 P 0.2126 (NS)a 0.5228 (NS) 0.5620 (NS) 0.07 (NS) 0.5489 (NS) 0.7195 (NS) DHT mean 1.93 1.96 2.07 1.89 1.96 1.79 2.00 2.00 2.10 1.93 1.96 1.89 SD 0.96 0.76 0.76 0.72 0.72 0.76 0.79 0.72 0.83 0.79 0.76 0.79 P 0.7597 (NS) 0.0886 (NS) 0.1159 (NS) 0.9778 (NS) 0.3286 (NS) 0.4853 (NS) 5-diol mean 6.99 4.72 6.33 4.65 6.16 3.62 4.79 4.17 5.13 3.72 4.65 3.48 SD 3.20 1.41 3.00 2.20 2.41 1.65 1.72 2.24 1.58 2.07 2.55 1.79 P Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 0.0381 Ͻ0.0001 0.0006 n couples 55 89 89 67 74 97 a NS, not significant. 2166 Giton et al.: Measurement of BT in a Healthy French Population

Fig. 2. aBT gain (%) in charcoal-treated sera overloaded with a constant testosterone concentration and overloaded with increasing amounts of 5-diol and/or DHEA [n ϭ 6 (SD)]. Inset, cBT:aBT ratio in 503 serum samples as a function of 5-diol concentration; 2.5th and 97.5th percentile lines did not show any obvious trend (slope approximately Ϫ0.002) when the 5-diol concentration was taken into account in the cBT:aBT ratio, whereas there was a clear trend (slope approximately Ϫ0.02) when this was not taken into account (plot not shown).

old for androgen deficiency corresponding to a TT of aBT (Table 2). These higher cBT than aBT values (4, 26, 27) approximately 10 nmol/L measured before reimplanta- are supported by several recent validation experiments in tion of a testosterone depot product has recently been which the free testosterone concentration calculated using reported in a population of testosterone-treated hypogo- the ISSAM website calculator was found to be higher than nadal men (18). the measured free testosterone (28–31). Morales and Lunenfeld (2) considered BT measure- In young men, the lower normal aBT limit and the ment to be the most reliable and widely acceptable mean normal (7 nmol/L) and median (6.59 nmol/L) aBT parameter for confirming hypogonadism. In our reference values we found were very similar to those obtained by population of 142 healthy young men, the 1st 2 individual other authors using the same ammonium sulfate SHBG lower BT values were 2.32 and 2.72 nmol/L. Therefore the precipitation for BT measurement (19, 20, 32). However, 1st percentile of our reference population of healthy these aBT results contrast with recently published lower young men was between 2.32 and 2.72 nmol/L. This normal cBT limits of 4.85 nmol/L (33) and 5.27 nmol/L agrees with a cutoff point of 2.50 nmol/L (19) or 2.32 (34) and a mean cBT of 13.28 nmol/L (35). These same nmol/L (20) for hypogonadism, established as a value not recently published normal cBT values (obtained from the found in eugonadal young males. Nelson et al. (21) ISSAM website) are much higher than widely published reported that BT concentrations Ͻ2.43 nmol/L warrant normal limits for measured BT in men and lead to androgen treatment. confusion when used in diagnosing hypogonadism. The interquartile BT values (3.26–5.17 nmol/L; n ϭ 126) we reported in the aging group (60–74 years) may be optimization of Ks and Ka association compared with interquartile BT values (2.77–4.09 constants nmol/L) published for a large group of 547 aging men Results for cBT obtained using the Vermeulen formula between 59 and 89 years (22). As we have, 3 other authors largely depend on the association constant values used for measured BT by the SHBG ammonium sulfate precipita- calculation. A wide range of association constants of tion method, adding minute amounts of 3H-testosterone testosterone for SHBG have been published (3, 36–40).

(4, 23, 24). Our median DHT concentration (1.69 nmol/L; We determined optimal paired Ks and Ka (4), which, Table 1) was similar to that reported by these 3 authors, when used with the Vermeulen formula, yielded cBT and a stable normal mean value, whatever the age, has values that were very close to the aBT values. already been reported (13, 14, 16). Median 5-diol concen- After optimization, 83% to 85% of our cBT values trations in the younger (Ͻ40 years; 6.06 nmol/L) and differed by Ͻ30% from the aBT. In addition, the cBT:aBT aging (60–74 years; 3.86 nmol/L) groups are similar to ratio was nearly 1 (Table 3) and did not vary with age. In those described by others in smaller groups (14, 25). contrast, using the ISSAM website, only 13.3% had cBT values similar to aBT values, and the mean cBT:aBT ratio aBT and cBT was 1.63 and increased with age. Using the Vermeulen formula (3) and association con- We showed that factors other than TT and SHBG ϭ ϫ 9 ϭ ϫ 4 stants (Ks 1 10 L/mol and Ka 3.6 10 L/mol) in contribute to the decrease in the aBT in aging men relative 503 healthy normal men, cBT was found to be higher than to younger men. We hypothesized that 5-diol, whose Clinical Chemistry 53, No. 12, 2007 2167

association constant for SHBG (1.5 ϫ 109 L/mol) (7) is 3. Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of practically the same as that of testosterone for SHBG simple methods for the estimation of free testosterone in serum. (1.6 ϫ 109 L/mol) (7), could modulate SHBG testosterone J Clin Endocrinol Metab 1999;84:3666–72. binding in men. Therefore, besides the well-known age- 4. Giton F, Fiet J, Gue´chot J, Ibrahim F, Bronsard F, Chopin D, et al. Serum bioavailable testosterone: assayed or calculated? Clin related variations in TT and SHBG, the reduction in 5-diol Chem 2006;52:474–81. with age further decreases aBT. 5. Labrie F, Be´langer A, Be´langer P, Be´rube´ R, Martel C, Cusan L, et We demonstrate in vitro that aBT increases as 5-diol al. Androgen glucuronides, instead of testosterone, as the new increases and that this increase is relatively higher in markers of androgenic activity in women. J Steroid Biochem Mol serum with a higher SHBG concentration than in serum Biol 2006;99:182–8. with a lower SHBG concentration (Fig. 2). In addition, we 6. Claudepierre P, Urien S, Chassany O, Tillement JP. Analysis of free show that an overload of serum with DHEA increases the fatty acid effect on methotrexate binding to albumin. Biochem aBT rather moderately in serum with a high SHBG and Pharmacol 1994;47:415–7. 7. Dunn JF, Nisula BC, Rodbard D. Transport of steroid hormones: not at all in serum with a low SHBG. This moderate binding of 21 endogenous steroids to both testosterone-binding interference of DHEA in the aBT compared with 5-diol is globulin and corticosteroid-binding globulin in human plasma. to be expected with the lower association constant of J Clin Endocrinol Metab 1981;53:58–68. 6 DHEA for SHBG (66 ϫ 10 L/mol) (7). 8. Harman SM, Metter EJ, Tobin JD, Pearson J, Blackman MR. The significant effect of 5-diol concentrations on the Longitudinal effects of aging on serum total and free testosterone binding of testosterone to SHBG was confirmed by the levels in healthy men. J Clin Endocrinol Metab 2001;86:724–31. bias observed when the cBT:aBT ratio was plotted against 9. Feldman HA, Longcope C, Derby CA, Johannes CB, Araujo AB, the 5-diol concentration. This clearly showed that cBT:aBT Coviello AD, et al. Age trends in the level of serum testosterone and other hormones in middle-aged men: longitudinal results from ratios decreased more rapidly below 1 as 5-diol concen- the Massachusetts Male Aging Study. J Clin Endocrinol Metab trations increased. When the inhibiting effect of 5-diol on 2002;87:589–98. SHBG was included in the testosterone-binding model 10. Ferrini RL, Barrett-Connor E. Sex hormones and age: a cross- (see Calculation of BT), the cBT (5-diol):aBT ratio vs 5-diol sectional study of testosterone and estradiol and their bioavail- concentration was significantly improved, with no further able fractions in community-dwelling men. Am J Epidemiol 1998; evidence of bias. In Calculation of BT, we propose a cBT 147:750–4. determination based on the total concentrations of testos- 11. Nahoul K, Roger M. Age-related decline of plasma bioavailable terone, 5-diol, SHBG, and albumin. testosterone in adult men. J Steroid Biochem 1990;35:293–9. 12. Nankin HR, Calkins JH. Decreased bioavailable testosterone in aging normal and impotent men. J Clin Endocrinol Metab 1986; In conclusion, using reference assay methods, the lower 63:1418–20. normal limits we found in the younger men in our study 13. Barrett-Connor E, Von Mu¨hlen DG, Kritz-Silverstein D. Bioavailable were 8.50 nmol/L for TT and between 2.30 and 2.72 testosterone and depressed mood in older men: the Rancho nmol/L for BT, similar to values reported by most other Bernardo Study. J Clin Endocrinol Metab 1999;84:573–7. authors, whereas according to the ISSAM website calcu- 14. Belanger A, Candas B, Dupont A, Cusan L, Diamond P, Gomez JL, lator, the corresponding cBT was much higher. Using et al. Changes in serum concentration of conjugated and uncon- optimized association constants of testosterone for SHBG jugated steroids in 40- to 80-year-old men. J Clin Endocrinol Metab and albumin and including 5-diol levels in the calculation 1994;79:1086–90. of BT improved the correlation between cBT and aBT. An 15. Gray A, Feldman HA, McKinlay JB, Longcope C. Age, disease, and changing sex hormone levels in middle-aged men: results of the alternative calculation that takes into account a possible Massachusetts Male Aging Study. J Clin Endocrinol Metab 1994; inhibitory effect of 5-diol on testosterone binding to SHBG 73:1016–25. was also carried out, using previously published equa- 16. Labrie F, Be´langer A, Cusan L, Gomez JL, Candas B. Marked tions (6). We therefore derived new equations for use in decline in serum concentrations of adrenal C19 sex steroid determining cBT (see Calculation of BT). precursors and conjugated androgen metabolites during aging. J Clin Endocrinol Metab 1997;82:2396–402. 17. Vermeulen A, Kaufman JM. Diagnosis of hypogonadism in the aging male. Aging Male 2002;5:170–6. Grant/funding support: None declared. 18. Kelleher S, Conway AJ, Handelsman DJ. Blood testosterone Financial disclosures: None declared. threshold for androgen deficiency symptoms. J Clin Endocrinol Metab 2004;89:3813–7. Acknowledgments: We thank Dr. Noah Hardy for editing the 19. Sih R, Morley JE, Kaiser FE, Perry HM 3rd, Patrick P, Ross C. manuscript and Rene´Be´rube´ for expert technical assistance. Testosterone replacement in older hypogonadal men: a 12 month randomized controlled trial. J Clin Endocrinol Metab 1997;82: 1661–7. References 20. Korenman SG, Morley JE, Mooradian AD, Davis SS, Kaiser FE, 1. Morley JE, Patrick P, Perry HM. Evaluation of assays available to Siver AJ, et al. Secondary hypogonadism in older men: its relation measure free testosterone. Metabolism 2002;51:554–9. to impotence. J Clin Endocrinol Metab 1990;71:963–9. 2. Morales A, Lunenfeld B. Investigation, treatment and monitoring 21. Nelson RE, O’Kane DJ, Heser DW, Klee GG. A simple and rapid assay of late-onset hypogonadism in males. Aging Male 2002;5:74–86. for bioavailable testosterone. Mayo Clinic 2001;47(Suppl):A20. 2168 Giton et al.: Measurement of BT in a Healthy French Population

22. Barrett-Connor E, Goodman-Gruen D, Patay B. Endogenous sex 31. Rinaldi S, Geay A, Dechaud H, Biessy C, Zeleniuch-Jacquotte A, hormones and cognitive function in older men. J Clin Endocrinol Akhmedkhanov A, et al. Validity of free testosterone and free Metab 1999;84:3681–5. estradiol determinations in serum samples from postmenopausal 23. Tremblay RR, Dube JY. Plasma concentrations of free and non- women by theoretical calculations. Cancer Epidemiol Biomarkers TeBG bound testosterone in women on oral contraception. Con- Prev 2002;11:1065–71. traception 1974;10:599–605. 32. Hajjar RR, Kaiser FE, Morley JE. Outcomes of long-term testoster- 24. Loric S, Gue´chot J, Duron F, Aubert P, Giboudeau J. Determination one replacement in older hypogonadal males: a retrospective of testosterone in serum not bound by sex-hormone-binding analysis. J Clin Endocrinol Metab 1997;82:3793–6. globulin: diagnostic value in hirsute women. Clin Chem 1988;34: 33. Kaufman JM, Vermeulen A. The decline of androgen levels in 1826–9. elderly men and its clinical and therapeutic implications. Endocr 25. Nahoul K, Bournique B, Adeline J, Scholler R. Radioimmunoassay Rev 2005;26:833–76. of 5-androstene-3 beta, 17 beta-diol in plasma and in breast cyst 34. Vermeulen A. Hormonal cut-offs of partial androgen deficiency: a fluid. J Steroid Biochem 1986;24:835–42. survey of androgen assays. J Endocrinol Invest 2005;28(Suppl 3):28–31. 26. Emadi-Konjin P, Bain J, Bromberg IL. Evaluation of an algorithm for 35. Mahmoud AM, Goemaere S, El-Garem Y, Van Pottelbergh I, calculation of serum “bioavailable” testosterone (BAT). Clin Bio- Comhaire FH, Kaufman JM. Testicular volume in relation to chem 2003;36:591–6. hormonal indices of gonadal function in community-dwelling el- 27. Dechaud H, Denuziere A, Rinaldi S, Bocquet J, Lejeune H, Pugeat derly men. J Clin Endocrinol Metab 2003;88:179–84. M. Age-associated discrepancy between measured and calculated 36. Vermeulen A, Verdonck L. Studies on the binding of testosterone bioavailable testosterone in men. Clin Chem 2007;53:723–8. to human plasma. Steroids 1968;11:609–35. 28. Van Uytfanghe K, Stockl D, Kaufman JM, Fiers T, De Leenheer A, 37. O’Connor S, Baker HWG, Dulmanis A, Hudson B. The measure- Thienpont LM. Validation of 5 routine assays for serum free ment of sex steroid binding globulin by differential ammonium testosterone with a candidate reference measurement procedure sulphate precipitation. J Steroid Biochem 1973;4:331–9. based on ultrafiltration and isotope dilution-gas chromatography- 38. Rosner W, Smith RN. Isolation and characterization of the testos- mass spectrometry. Clin Biochem 2005;38:253–61. terone-estradiol-binding globulin from human plasma: use of a 29. Ly LP, Handelsman DJ. Empirical estimation of free testosterone novel affinity column. Biochemistry 1975;14:4813–20. from testosterone and sex hormone-binding globulin immunoas- 39. So¨dergard R, Ba¨ckstro¨m T, Shanbhag V, Carstensen H. Calcula- says. Eur J Endocrinol 2005;152:471–8. tion of free and bound fractions of testosterone and estradiol-17␤ 30. Ross HA, Meuleman EJ, Sweep FC. A simple method for estimat- to human plasma proteins at body temperature. J Steroid Bio- ing equilibrium constants for serum testosterone binding resulting chem 1982;16:801–10. in an optimal free testosterone index for use in elderly men. Clin 40. Burke CW, Anderson DC. Sex hormone binding globulin is an Chem Lab Med 2005;43:613–6. estrogen amplifier. Nature 1982;240:38–40. Clinical Chemistry 53:12 2169–2176 (2007) Automation and Analytical Techniques

Metabolomics Identifies Perturbations in Human Disorders of Propionate Metabolism

William R. Wikoff,1 Jon A. Gangoiti,2 Bruce A. Barshop,2* and Gary Siuzdak1*

Background: We applied untargeted mass spectrometry- associated with human disease and shows that this based metabolomics to the diseases methylmalonic aci- method may be useful for disease diagnosis and patient demia (MMA) and propionic acidemia (PA). clinical evaluation. Methods: We used a screening platform that used un- © 2007 American Association for Clinical Chemistry targeted, mass-based metabolomics of methanol- extracted plasma to find significantly different molecu- Inborn errors of metabolism can have a severe impact on lar features in human plasma samples from MMA and human health, so comprehensive diagnostic neonatal PA patients and from healthy individuals. Capillary screening is used for early diagnosis to avoid potentially reverse phase liquid chromatography (4 ␮L/min) was catastrophic physical and neurological effects (1). These interfaced to a TOF mass spectrometer, and data were defects can cause a buildup of toxic metabolites, resulting processed using nonlinear alignment software (XCMS) in serious, often fatal, disease early in life (2, 3). Neonatal and an online database (METLIN) to find and identify blood screening with mass spectrometry (MS)3 is now metabolites differentially regulated in disease. commonly used to test newborns for a wide array of Results: Of the approximately 3500 features measured, inborn errors of metabolism using specific metabolites for propionyl carnitine was easily identified as the best diagnosis (4). We studied 2 diseases using metabolomics ؊ -biomarker of disease (P value 1.3 ؋ 10 18), demonstrat- with nontargeted LC-MS to simultaneously profile thou ing the proof-of-concept use of untargeted metabolom- sands of metabolites and obtain a more comprehensive ics in clinical chemistry discovery. Five additional acyl- metabolic profile of plasma. The platform we developed carnitine metabolites showed significant differentiation (Fig. 1) includes a profiling approach, along with a between plasma from patients and healthy individuals, software package that incorporates new nonlinear time and ␥-butyrobetaine was highly increased in a subset of correction, peak-finding, and integration methods to al- patients. Two acylcarnitine metabolites and numerous low for semiquantitative comparison between healthy unidentified species differentiate MMA and PA. Many individuals and patient populations (5). metabolites that do not appear in any public database, Methylmalonic acidemia (MMA) and propionic aci- and that remain unidentified, varied significantly be- demia (PA) are inborn errors of amino acid metabolism tween normal, MMA, and PA, underscoring the complex affecting 1 in 50 000 to 1 in 100 000 individuals (6, 7).PA downstream metabolic effects resulting from the defect results from a defect in the enzyme propionyl-CoA car- in a single enzyme. boxylase, which catalyzes the biotin-dependent conver- Conclusions: This proof-of-concept study demonstrates sion of propionyl-CoA to methylmalonyl-CoA (8). MMA that metabolomics can expand the range of metabolites results from deficiency of the immediately downstream enzyme methylmalonyl-CoA mutase, which catalyzes the

vitamin B12-dependent conversion of methylmalonyl- CoA to succinyl-CoA (Fig. 2). Patients have considerable 1 Department of Molecular Biology and The Center for Mass Spectrometry, The Scripps Research Institute, La Jolla, CA. variability in symptoms and clinical prognosis, which are 2 Department of Pediatrics, University of California–San Diego, School of correlated with genetic locus (6, 7). Patients with com- Medicine, La Jolla, CA. plete apoenzyme deficiency (mut0) typically present early * Address correspondence to these authors at: Department of Pediatrics, University of California–San Diego, School of Medicine, 9500 Gilman Dr., in life with severe symptoms, including high mortality #0830, La Jolla, CA 92093-0830. Fax 619-543-3565; e-mail [email protected] or Department of Molecular Biology and The Center for Mass Spectrometry, The Scripps Research Institute, 10550 North Torrey Pines Rd., La Jolla, CA 92037. Fax 858-784-9496; e-mail [email protected]. 3 Nonstandard abbreviations: MS, mass spectrometry; MMA, methylma- Received April 12, 2007; accepted October 2, 2007. lonic acidemia; PA, propionic acidemia; CID, collision-induced dissociation; Previously published online at DOI: 10.1373/clinchem.2007.089011 ESI-TOF, electrospray ionization TOF; MS/MS, tandem MS.

2169 2170 Wikoff et al.: Metabolomics of Human Disorders of Propionate Metabolism

Fig. 1. Metabolomics workflow. (A), outline of the methods used for untargeted metabolomics comparing disease and normal plasmas. Human plasma is methanol-extracted and applied to a capillary reverse phase HPLC column, interfaced to an ESI-TOF mass spectrometer. Data are converted and aligned in the time domain using the program XCMS, with 3500–4500 features found. Statistically significant differences are ranked using a t-test. These metabolites are identified by use of the methods outlined in B. (B), accurate mass information from the metabolite is used either directly for a database search or for calculating the possible elemental compositions, which are then used for database searching. Test compounds are selected, and their HPLC retention time and MS/MS fragmentation data are compared with the unknown.

and neurologic impairment, whereas those with partial The power of untargeted metabolomics lies in its apoenzyme deficiency (mut–) typically have less severe potential to broaden our understanding of disease bio- symptoms. Abnormalities in the processing of vitamin B12 chemistry, identify new biomarkers, and provide finer to the active coenzyme adenosylcobalamin (cblA, B, C, disease categorization and treatment. Although originat- and D) are amenable to treatment with B12 supplementa- ing in the disturbance of a single gene, inborn errors of tion, and present later with the best prognosis (9). Within metabolism produce highly diverse phenotypes and com- each of these groups there is substantial variability in plex downstream metabolic effects, the result of the disease severity and prognosis, for which there is no interplay of many biochemical pathways and the individ- biomarker. For example, in the complete apoenzyme ual’s interaction with the environment (17–19).Wehy- deficiency form, both MMA and PA often include varying pothesized that a metabolomics study may reveal addi- degrees of neurological involvement, the underlying bio- tional differences between disease and normal plasma, chemical cause of which is not known (10–14). Perhaps and thus a more complete biochemical profile. This study differences in metabolite concentrations correlating with demonstrated the application of a new toolbox of meth- neurological damage could eventually be identified using ods to clinical chemistry, an approach with the potential untargeted metabolomics. to provide new insights into the biochemical mechanisms MMA and PA were the 1st human diseases to be of disease. diagnosed using MS, originally by GC-MS (15). The level of propionyl carnitine is highly increased in both diseases, Materials and Methods arising from the transesterification of propionyl-CoA (Fig. sample preparation 2), and in MMA methylmalonyl carnitine is also typically Plasma samples were stored frozen at –80 °C until use, at increased. Propionyl carnitine is now used as the primary which point the samples were thawed on ice. We per- biomarker, screened by tandem electrospray MS after formed metabolite extraction and protein precipitation by derivatizing bloodspot methanol extracts by butylation adding 100 ␮L cold methanol to 50 ␮L plasma. Samples (16). Acylcarnitine fragmentation yields a common colli- were then vortex-mixed and stored at –20 °C for 1 h. The sion-induced dissociation (CID) fragment with m/z 85, pellet was removed by centrifuging at 16 000g for 10 min. and thus rapid multiple-reaction monitoring with direct The supernatant was removed to a clean tube and the injection of the butylated methanol extract has become the centrifugation step was repeated, and the samples were method of choice for large-scale neonatal screening (4). dried in a SpeedVac to dryness. We added 50 ␮L water, Clinical Chemistry 53, No. 12, 2007 2171

Fig. 2. Metabolic pathways. (A), simplified pathway illustrating the metabolic blockage points of MMA and PA disease. The mut0 form of MMA results in a complete blockage of methylmalonyl-CoA mutase, whereas PA results from a defect in propionyl-CoA carboxylase. Both produce a buildup of toxic upstream acyl-CoA precursors, and their corresponding acylcarnitines, propionyl-carnitine, and methylmalonyl-carnitine. (B), simplified pathway illustrating conversion of toxic acyl-CoA intermediates to the corresponding acylcarnitines and cleared into the plasma. Long-chain acylcarnitines require carnitine acetyltransferase (CAT) to pass through the membrane, whereas shorter- chain acylcarnitines can pass directly through the cell membrane. The direct precursor of carnitine is ␥-butyrobetaine, which was found in this study to be highly increased in some patients, without a corresponding increase in carnitine concentrations. followed by vortex-mixing and sonication in a bath soni- We used a metabolomics workflow (Fig. 1) to find cator for 5 min. The samples were maintained at 4 °C in significantly different molecular features in human the autosampler and analyzed by LC-MS. plasma, followed by metabolite identification. Briefly, the plasma samples were extracted with methanol to remove patients and metabolomics workflow proteins and extract the maximum number of metabolites Existing samples referred to a clinical laboratory for prior (20). We injected 8 ␮L extracted plasma onto a reverse testing were anonymized, and the investigation was in phase capillary-flow HPLC interfaced to an electrospray accordance with the University of California–San Diego ionization TOF (ESI-TOF) mass spectrometer. Data were Human Research Protection Program. In an initial exper- collected and converted to a common data format (.cdf) iment, plasma samples from individuals with MMA and for MS. We used the program XCMS (5) to integrate the healthy individuals were collected, comparing 9 MMA chromatographic peaks and assign the peaks into groups, samples to 10 normal plasma samples. After finding followed by nonlinear alignment of the grouped data in significantly increased concentrations of several metabo- the time domain. lites in MMA, including various acylcarnitines, the sample size was expanded. Samples from patients with PA chromatography were included, along with healthy adults before and after We injected 8 ␮L processed plasma for each run. Reverse supplementation with l-carnitine (TwinLabs), 1000 mg phase chromatography was performed using either a 150 twice a day for 1 week. The final study included samples by 0.32 mm (diameter) endcapped C18 column (Aquasil, from healthy children (n ϭ 12), healthy adults with (n ϭ Thermo) with 5-␮m particles, at a flow rate of 4 ␮L/min 3) and without (n ϭ 3) carnitine supplementation, healthy or a 150 by 0.5 mm (diameter) Zorbax C18 column children with carnitine supplementation (n ϭ 3), children (Agilent) with 5-␮m particles at a flow rate of 8 ␮L/min. with MMA (n ϭ 15), and children with PA (n ϭ 9). The The LC system was an Agilent 1100 with a capillary patient group was restricted to patients with complete pump. Buffer A was water with 0.1% formic acid, and apoenzyme deficiency (termed mut0 in MMA). buffer B was acetonitrile with 0.1% formic acid. For the 2172 Wikoff et al.: Metabolomics of Human Disorders of Propionate Metabolism

endcapped C18 column, the column was equilibrated in tion patterns of acylcarnitines. This process, in combina- 100% A, and the gradient was 0%–20% B over 20 min, tion with accurate mass data from the ESI-TOF, was used 20%–60% B over 20 min, and 60%–80% B over 5 min. For to identify the compound as propionyl carnitine. In addi- the C18 (Zorbax) column, the column was equilibrated in tion, the retention time of a propionyl carnitine calibrator 5% B, and the gradient was 5%–95% B over 50 min. was checked against the retention time of the endogenous peak in patient plasma. The raw integrated intensity data mass spectrometry from XCMS indicates a mean intensity ratio of approxi- We collected data in continuum mode from m/z 75 to mately 2000 between disease and normal plasma (Table 1, 1000 using an Agilent ESI-TOF. The capillary voltage was Fig. 3). The endcapped Aquasil column provided better 3500 V, with a nebulizer gas flow of 13 L/min. The retention of propionyl carnitine, with a retention time of instrument was calibrated immediately before use. The 16 min, than a standard C18 column, on which it is not data files were converted from the instrument format retained and elutes at 5.5 min. (.wiff) to the common data format, using the PESciEX data translator. We used the XCMS program (5) to align and increase in acylcarnitines in disease analyze the LC-MS data. To confirm the identification of In the initial experiment comparing MMA and normal significant ions, we used a linear ion trap (Thermo LTQ), plasma, a number of compounds were found to be with a custom nanospray interface operated at a voltage significantly increased in MMA plasma, several of of 2 kV. The Aquasil column was used with the identical which were identified as acylcarnitines. To test whether chromatography parameters as for the metabolomics treatment with carnitine produced this effect, samples runs, with a passive postcolumn split, to reduce the flow from 3 healthy children treated with carnitine and from rate from 4 ␮L/min to 300 nL/min for nanospray. Masses 3 healthy adults before and after carnitine supplemen- for CID were targeted from a mass list and fragmented tation were included in the experimental controls. The during the chromatography run. acylcarnitine levels of the 6 supplemented individuals were not significantly increased relative to the un- Results supplemented controls, and levels were within the propionyl carnitine general range of the other normal samples. Comparing The most significant difference between normal and dis- supplemented vs unsupplemented normal samples, ease samples in this study was an ion with m/z 218.14 at only the ion at m/z 244 (identified as pentenoyl or C5:1 a retention time of 16 min, with a t-test P value of 1.3 ϫ carnitine) differed by more than a factor of 2 (2.2), Ϫ 10 18, comparing all disease and normal samples, and an whereas this ion is increased by a factor of 6.6 in disease Ϫ ANOVA P value of 1.8 ϫ 10 18. Tandem MS (MS/MS) vs normal plasma. fragmentation in a linear ion trap produced fragments of After propionyl carnitine and its related ions (isotopes, m/z 159 and 85. The common fragment of 85 and the the sodium adduct, a dimer, and fragments), the most neutral loss of 59 are both characteristic CID fragmenta- significant feature differentiating disease from normal

Table 1. Identification and significance of compounds.a Retention Fold Molecular time Observed increase P value [t-test P value formula Calculated Observed ppm of HPLC CID fragments (disease/ (disease vs [ANOVA (normal, [(Molecule [M؉H] m/z m/z error peak (min) (linear ion trap) normal) normal]) MMA, PA Ϫ ϩ ϫ Ϫ18 ϫ Ϫ18 Propionyl carnitine C10H20NO4 218.1386 218.1380 3.14 16 159 85 2171 1.3 10 1.8 10 (C3)

Fragment of propionyl C4H5O2 85.0284 85.0292 9.34 16 NA NA NA NA carnitine Ϫ ϩ ϫ Ϫ06 ϫ Ϫ5 Hexanoyl carnitine C13H26NO4 260.1856 260.1846 3.98 36.6 201 85 7.0 9.1 10 5.6 10 (C6) Ϫ ϩ ϫ Ϫ4 ϫ Ϫ8 Acetyl carnitine C9H18NO4 204.1230 204.1227 1.64 8.6 145 85 3.0 1.1 10 1.2 10 ϩ ϫ Ϫ3 ϫ Ϫ3 Pentenoyl carnitine C12H22NO4 244.1543 244.1546 1.09 27.7 185 85 7.0 1.4 10 2.1 10 (C5:1) ␥ Ϫ ϩ ϫ Ϫ3 ϫ Ϫ4 -Butyrobetaine C7H15NO2 146.1175 146.1175 0.38 6.9 87 60 13.3 3.3 10 7.2 10 ϩ ϫ Ϫ7 ϫ Ϫ11 Hexenoyl (C6:1) or C13H24NO4 258.1699 258.1693 2.32 33 199 85 36.1 6.4 10 4.0 10 methyl C5:1 carnitine ϫ Ϫ08 ϫ Ϫ9 Myristoyl carnitine C21H42NO4 372.3108 372.3140 8.50 59.2 ND 7.7 6.0 10 3.3 10 ϩ ϩ Carnitine C7H16NO3 162.1124 162.1134 2.35 5.5 162 103 60 0.95 0.7 ND a A longer retention time for the acyl carnitines on the reverse-phase column correlates with a longer acyl chain length. The ratio of the average intensity of all disease samples to normal samples is shown in column 8. Shown are the P value of the t-test between disease and normal samples and the P value of an ANOVA test comparing MMA, PA, and normal samples. NA, not applicable; ND, not determined. Clinical Chemistry 53, No. 12, 2007 2173

Fig. 3. Box plots showing integrated ion intensities (linear, arbitrary scale) for the identified ions that differ most significantly between normal individuals (with and without carnitine supplementation) and MMA and PPA. Individual values are indicated by dots in red (normal), green (MMA), and blue (PA), with SD indicated by the upper and lower extent of the box. Statistics and mean differences are shown in Table 1.

plasma has an m/z of 258, with a retention time of PA patient urines during acute illness (21), including 32.7 min. This ion was identified as an acylcarnitine based 2-methyl-2-pentenoic carnitine, which could correspond on fragmentation pattern (characteristic CID fragments to the C6:1 acylcarnitine compound identified in our with m/z 199 and 85), corresponding to the molecular study. In another recent study, novel branched-chain formula C13H24NO4, which could be either a straight- acylcarnitines were recently identified in the urine of 1 chain hexenoic (C6:1) or a branched-chain methyl-pente- patient with another metabolic disorder, medium-chain noic carnitine; these compounds cannot be differentiated CoA dehydrogenase disorder (22). by mass or fragmentation pattern, because the elemental composition and general acylcarnitine structure are the ␥-butyrobetaine same. Although neither compound is known to be asso- The observed ion with m/z of 146.18 and retention time of ciated with MMA or PA as a biomarker, 2-methyl- 6.9 min was identified as ␥-butyrobetaine, based on branched acylcarnitines have been reported in MMA and accurate mass, retention time, and MS/MS in a linear ion 2174 Wikoff et al.: Metabolomics of Human Disorders of Propionate Metabolism

trap. The average concentration of ␥-butyrobetaine in Agilent) with a total run time of 75 min, 4313 features Ϫ patient plasma was increased 13-fold (P value 3.3 ϫ 10 3), were detected when the equivalent of only 8 ␮L plasma with 1 PA patient having a 55-fold increase over the upper was injected. Using an endcapped C18 column (Aquasil; reference limit value (Table 1, Fig. 3). Some MMA and PA Thermo) with a run time of 75 min, 3473 features were samples were within the reference interval. detected. The simultaneous detection and profiling of many (hundreds to thousands) metabolites increases the differentiating between mma and pa probability of identifying new compounds associated Many ions differed significantly between MMA and PA with disease using untargeted metabolomics. (Table 2). At a t-test ␣ of 0.001, 76 features differed The target compound for newborn screening by between MMA and PA. One of the most significant differ- MS/MS in both MMA and PA is propionyl carnitine entiating features had an m/z of 246.17 and was identified as (Fig. 2) (4, 16). Indeed, the most significant feature in isovaleryl carnitine (C12H23NO4) based on accurate mass the untargeted metabolomic analysis that distinguished and MS/MS data (see Table 1 in the Data Supplement that disease patient vs normal plasma was propionyl carnitine Ϫ accompanies the online version of this article at http:// (P value 1.3 ϫ 10 18). This result validates the untar- www.clinchem.org/content/vol53/issue12).Isovaleryl car- geted metabolomics approach for identification of bio- nitine was increased in MMA relative to normal samples by markers, and is the most important finding, demonstrat- Ϫ a factor of 5.1 (P value 1.66 ϫ 10 9) and was increased ing proof-of-concept application of metabolomics to clin- relative to PA samples, which had isovaleryl carnitine con- ical chemistry. centrations similar to normal samples. The P value differen- Many other compounds were significantly altered in Ϫ tiating MMA and PA was 9.7 ϫ 10 8. Also found to disease vs normal samples, with 317 features differing at differentiate MMA and PA was an ion with m/z 177.10 at a an ␣ level of 0.005 (Table 2), approximately 72% of which retention time of 48.5 min, which was increased 40-fold in represent increased plasma metabolite concentrations. Ϫ PA compared to MMA. The P value was 5.8 ϫ 10 8 for PA Among the identified compounds were C5:1 and C6:1 Ϫ vs normal samples and 1.6 ϫ 10 5 for PA vs MMA samples. carnitine. These compounds do not seem to be the result Methylmalonyl carnitine was not observed in the untargeted of nonspecific transesterification of CoA esters, because metabolomic analysis, although it is usually increased in they are not increased in samples from metabolically MMA. normal children on carnitine supplementation or the adults treated with carnitine. The C5:1 carnitine species overall differences between disease and (m/z 244) may be tiglyl carnitine arising from an inter- normal samples and unidentified ions mediate in the catabolism of isoleucine; increase of tiglyl- We observed several statistically significant differ- glycine occurs in PA (23). ences between disease and normal plasma. At a t-test ␣ This study also suggests that MMA and PA can be of 0.001 (P value threshold), there were 153 features, differentiated, and 2 of these molecules were identified or 4.4% of the total ions detected, that differed be- (Table 2, Supplemental Data Table 1, and Fig. 4). The ion tween disease and normal samples, and at an ␣ of 0.005 with m/z 258.17 and a retention time of 33 min (Table 1) there were 317 features, or 9.1% of the total that differ was identified as a C6:1 (or methyl-C5:1) carnitine ester. It (Table 2). According to either significance criterion, is significantly increased in PA relative to MMA (P value Ϫ about 72% of those differences were the result of an 1.6 ϫ 10 4) (Supplemental Data Table 1). There are many increased concentration of the ion in plasma in the disease more significant differences, however, and C6:1 ranks state vs normal samples, indicating a clearly skewed only at number 54 when sorted using a t-test. More distribution. significantly, isovaleryl carnitine was increased by a fac- Ϫ tor of 2.8 in MMA vs PA, with a P value of 9.7 ϫ 10 8) Discussion (Fig. 4). There are other significant ions differentiating The advantage of capillary chromatography for untar- MMA and PA, which are clearly not acylcarnitines based geted metabolomics is the sensitivity arising from the low on fragmentation pattern but have not yet been identified. flow rate (4 ␮L/min), allowing for the detection of many For compounds such as the ion at m/z 177.1, which is Ϫ compounds. Using a 300-␮m ID C18 column (Zorbax; increased in PA vs MMA (P value 1.6 ϫ 10 5) (Supple-

Table 2. Number and percentage of the features (of 3472 total) differentiating normal, MMA, and PA at 2 significance cutoffs. Features Features differentiating increased disease vs in disease vs t-test criteria normal samples, normal samples, Features differentiating Features increased in (P < value) n (%) n (%) MMA vs PA, n (%) PA vs MMA, n (%) 0.001 153 (4.4) 110 (72) 76 (2.2) 49 (64.5) 0.005 317 (9.1) 225 (71) 154 (4.4) 96 (62.3) Clinical Chemistry 53, No. 12, 2007 2175

Fig. 4. Box plots showing integrated ion intensities for selected ions that differentiate between PA and MMA. (A), isovaleryl carnitine; (B), m/z 177.10 (unidentified); and (C), m/z 386.33 (unidentified). Statistics and mean differences are shown in Supplemental Data Table 1. mental Data Table 1, Fig. 4), identification of this molecule Although there was no significant difference in free may be biochemically informative. The task of de novo carnitine concentrations in patient samples compared to identification of metabolites can be very time-consuming, normal samples (Table 1), each of the patients was receiv- however, and is an open-ended process; identification of ing high doses (100–200 mg/kg/day) of supplemental additional compounds awaits further study and valida- carnitine. ␥-Butyrobetaine was not increased in carnitine- tion with additional samples. By combining multiple ions, supplemented children or adults from this study. Plasma it is possible to clearly differentiate between the 2 dis- ␥-butyrobetaine can increase with carnitine supplementa- eases, but larger numbers of patient samples are needed tion in dialysis patients (26) and in patients with medium- for validation. Methylmalonyl carnitine was not identified chain CoA dehydrogenase deficiency (28) and via carni- in the untargeted analysis, although it was 15-fold in- tine metabolism by gut flora (29, 30). Further study is creased in MMA vs PA and normal individuals in an needed to separate effects of carnitine supplementation independent targeted analysis (data not shown). Methyl- and correlate ␥-butyrobetaine concentrations with clinical malonyl carnitine may not have been observed because of variables. its high polarity and lack of retention on the reverse phase Application of metabolomics to MMA and PA has column. revealed metabolites that differentiate between disease An intriguing result of this study is the significant and normal patients, and between MMA and PA. Propio- increase in ␥-butyrobetaine concentrations in patients nyl carnitine was identified as a biomarker using a with MMA and PA, which has not been previously completely untargeted approach, illustrating the useful- reported. ␥-Butyrobetaine is converted to l-carnitine by ness of metabolomics for identifying biomarkers of dis- ␥-butyrobetaine hydroxylase, localized in liver, kidney, ease. New metabolites continue to be discovered even in and brain in humans (24). Although studies are limited, well-studied diseases (21, 22, 31), and the metabolomics normal plasma concentrations of ␥-butyrobetaine were approach permits analysis of an unprecedented range mean (SD) 1.8 (0.23) ␮mol/L at rest in normal adults (25) of compounds. In this study, metabolites such as un- and 0.98 (0.08) ␮mol/L in long-term hemodialysis pa- saturated acylcarnitines, isovaleryl carnitine, and ␥- tients (26). Increase of ␥-butyrobetaine may arise from butyrobetaine were associated with MMA and PA. The disruption of carnitine synthesis, for example by in- finding of many significant differences underscores the hibition of ␥-butyrobetaine hydroxylase. Alternatively, evolving understanding of metabolic disease as having propionyl carnitine inhibits ␥-butyrobetaine transport complex, diverse phenotypes that transcend simple con- across the plasma membrane in liver (27), and thus cepts of a single-locus genotype (17, 19). high propionyl carnitine concentrations in MMA and PA may affect ␥-butyrobetaine transport dynamics and thus plasma concentrations. The increase of ␥-butyrobetaine Grant/funding support: This work was supported by grants could also be a direct result of carnitine supplementation. from the Department of Energy (DE-AC02-05CH11231) and 2176 Wikoff et al.: Metabolomics of Human Disorders of Propionate Metabolism

National Institutes of Health (R24 EY017540 and P30 16. Chace DH, DiPerna JC, Kalas TA, Johnson RW, Naylor EW. Rapid MH062261) (to W.R.W. and G.S.). diagnosis of methylmalonic and propionic acidemias: quantitative Financial disclosures: None declared. tandem mass spectrometric analysis of propionylcarnitine in filter-paper blood specimens obtained from newborns. Clin Chem Acknowledgments: We thank Andy Gieschen for helpful 2001;47:2040–4. advice on data collection. 17. Lanpher B, Brunetti-Pierri N, Lee B. Inborn errors of metabolism: the flux from Mendelian to complex diseases. Nat Rev Genet References 2006;7:449–60. 1. Clague A, Thomas A. Neonatal biochemical screening for disease. 18. Nicholson JK, Wilson ID. Opinion: understanding ‘global’ systems Clin Chim Acta 2002;315:99–110. biology: metabonomics and the continuum of metabolism. Nat 2. Deodato F, Boenzi S, Santorelli FM, Dionisi-Vici C. Methylmalonic Rev Drug Discov 2003;2:668–76. and propionic aciduria. Am J Med Genet C Semin Med Genet 19. Scriver CR. Translating knowledge into practice in the “post- 2006;142:104–12. genome” era. Acta Paediatr 2004;93:294–300. 3. Kahler SG, Fahey MC. Metabolic disorders and mental retardation. 20. Want EJ, O’Maille G, Smith CA, Brandon TR, Uritboonthai W, Qin C, Am J Med Genet C Semin Med Genet 2003;117:31–41. et al. Solvent-dependent metabolite distribution, clustering, and 4. Chace DH, Kalas TA, Naylor EW. The application of tandem mass protein extraction for serum profiling with mass spectrometry. spectrometry to neonatal screening for inherited disorders of Anal Chem 2006;78:743–52. intermediary metabolism. Annu Rev Genomics Hum Genet 2002; 21. Libert R, Van Hoof F, Thillaye M, Vincent MF, Nassogne MC, de 3:17–45. Hoffmann E, et al. Identification of undescribed medium-chain 5. Smith CA, Want EJ, O’Maille G, Abagyan R, Siuzdak G. XCMS: acylcarnitines present in urine of patients with propionic and processing mass spectrometry data for metabolite profiling using methylmalonic acidemias. Clin Chim Acta 2000;295:87–96. nonlinear peak alignment, matching, and identification. Anal 22. Libert R, Van Hoof F, Thillaye M, Vincent MF, Nassogne MC, Chem 2006;78:779–87. Stroobant V, et al. Identification of new medium-chain acylcar- 6. Coulombe JT, Shih VE, Levy HL. Massachusetts Metabolic Disor- nitines present in urine of a patient with medium-chain acyl-CoA ders Screening Program. II. Methylmalonic aciduria. Pediatrics dehydrogenase deficiency. J Inherit Metab Dis 1999;22:9–18. 1981;67:26–31. 23. Duran M, Bruinvis L, Ketting D, Kamerling JP, Wadman SK, 7. Dionisi-Vici C, Rizzo C, Burlina AB, Caruso U, Sabetta G, Uziel G, et Schutgens RB. The identification of (E)-2-methylglutaconic acid, a al. Inborn errors of metabolism in the Italian pediatric population: new isoleucine metabolite, in the urine of patients with beta- a national retrospective survey. J Pediatr 2002;140:321–7. ketothiolase deficiency, propionic acidaemia and methylmalonic 8. Gompertz D, Storrs CN, Bau DC, Peters TJ, Hughes EA. Localisa- acidaemia. Biomed Mass Spectrom 1982;9:1–5. tion of enzymic defect in propionicacidaemia. Lancet 1970;i: 24. Rebouche CJ, Engel AG. Tissue distribution of carnitine biosyn- 1140–3. thetic enzymes in man. Biochim Biophys Acta 1980;630:22–9. 9. Matsui SM, Mahoney MJ, Rosenberg LE. The natural history of the 25. Inoue F, Terada N, Nakajima H, Okochi M, Kodo N, Kizaki Z, et al. inherited methylmalonic acidemias. N Engl J Med 1983;308:857– Effect of sports activity on carnitine metabolism: measurement of 61. free carnitine, gamma-butyrobetaine and acylcarnitines by tandem 10. Brusque AM, Borba Rosa R, Schuck PF, Dalcin KB, Ribeiro CA, mass spectrometry. J Chromatogr B Biomed Sci Appl 1999;731: Silva CG, et al. Inhibition of the mitochondrial respiratory chain 83–8. complex activities in rat cerebral cortex by methylmalonic acid. Neurochem Int 2002;40:593–601. 26. Vernez L, Dickenmann M, Steiger J, Wenk M, Kra¨henbu¨hl S. Effect 11. Dutra JC, Dutra-Filho CS, Cardozo SE, Wannmacher CM, Sarkis JJ, of l-carnitine on the kinetics of carnitine, acylcarnitines and Wajner M. Inhibition of succinate dehydrogenase and beta- butyrobetaine in long-term haemodialysis. [Erratum in Nephrol hydroxybutyrate dehydrogenase activities by methylmalonate in Dial Transplant 2006 Apr;21:1141] Nephrol Dial Transplant brain and liver of developing rats. J Inherit Metab Dis 1993;16: 2006;21:450–8. 147–53. 27. Berardi S, Stieger B, Wachter S, O’Neill B, Krahenbu¨hl S. Charac- 12. Ko¨lker S, Schwab M, Ho¨rster F, Sauer S, Hinz A, Wolf NI, et al. terization of a sodium-dependent transport system for buty- Methylmalonic acid, a biochemical hallmark of methylmalonic robetaine into rat liver plasma membrane vesicles. Hepatology acidurias but no inhibitor of mitochondrial respiratory chain. J Biol 1998;28:521–5. Chem 2003;278:47388–93. 28. Huidekoper HH, Schneider J, Westphal T, Vaz FM, Duran M, 13. Okun JG, Ho¨rster F, Farkas LM, Feyh P, Hinz A, Sauer S, et al. Wijburg FA. Prolonged moderate-intensity exercise without and Neurodegeneration in methylmalonic aciduria involves inhibition with l-carnitine supplementation in patients with MCAD deficiency. of complex II and the tricarboxylic acid cycle, and synergistically J Inherit Metab Dis 2006;29:631–6. acting excitotoxicity. J Biol Chem 2002;277:14674–80. 29. Kleber HP. Bacterial carnitine metabolism. FEMS Microbiol Lett 14. Wajner M, Coelho JC. Neurological dysfunction in methylmalonic 1997;147:1–9. acidaemia is probably related to the inhibitory effect of methyl- 30. Rebouche CJ, Seim H. Carnitine metabolism and its regulation in malonate on brain energy production. J Inherit Metab Dis 1997; microorganisms and mammals. Ann Rev Nutr 1998;18:39–61. 20:761–8. 31. Libert R, Van Hoof, Thillaye M, Vincent MF, Nassogne MC, 15. Oberholzer VG, Levin B, Burgess EA, Young WF. Methylmalonic Stroobant V, et al. Identification of new medium-chain acylcar- aciduria: an inborn error of metabolism leading to chronic meta- nitines present in normal human urine. Anal Biochem 1997;251: bolic acidosis. Arch Dis Child 1967;42:492–504. 196–205. Clinical Chemistry 53:12 2177–2185 (2007) Automation and Analytical Techniques

Visual Recognition and Efficient Isolation of Apoptotic Cells with Fluorescent-Magnetic- Biotargeting Multifunctional Nanospheres

Er-Qun Song,1 Guo-Ping Wang,1 Hai-Yan Xie,2 Zhi-Ling Zhang,1 Jun Hu,1 Jun Peng,1 Dao-Cheng Wu,3 Yun-Bo Shi,4 and Dai-Wen Pang1*

Background: Fluorescent-magnetic-biotargeting multi- Conclusions: Our results show that fluorescent-magnetic- functional nanospheres are likely to find important biotargeting trifunctional nanospheres can be a power- applications in bioanalysis, biomedicine, and clinical ful tool for rapidly recognizing, magnetically enriching diagnosis. We have been developing such multifunc- and sorting, and simultaneously identifying different tional nanospheres for biomedical applications. kinds of cells. Methods: We covalently coupled avidin onto the sur- © 2007 American Association for Clinical Chemistry faces of fluorescent-magnetic bifunctional nanospheres to construct fluorescent-magnetic-biotargeting trifunc- Rapid recognition and sorting of cells are important in the tional nanospheres and analyzed the functionality and developing fields of cellular therapy and biotechnology. specificity of these trifunctional nanospheres for their The separation and analysis of different types of individ- ability to recognize and isolate apoptotic cells labeled ual cells, such as cancer cells, T cells (the first line of with biotinylated annexin V, which recognizes phos- defense against invading pathogens), and apoptotic cells phatidylserine exposed on the surfaces of apoptotic associated with many kinds of diseases (1–3) are becom- cells. ing increasingly important, not only for basic research but Results: The multifunctional nanospheres can be used also, more importantly, for drug screening and develop- in combination with propidium iodide staining of nu- ment and disease diagnosis. Techniques for recognizing clear DNA to identify cells at different phases of the and separating specific target cells from a mixed-cell apoptotic process. Furthermore, we demonstrate that population are urgently needed for such biomedical anal- apoptotic cells induced by exposure to ultraviolet light yses and clinical applications. Of the methods currently can be isolated simply with a magnet from living cells at used to recognize and isolate types of individual cells for an efficiency of at least 80%; these cells can then be biochemical or functional studies, the most widely used easily visualized with a fluorescence microscope. are fluorescence-activated cell recognition techniques such as fluorescence imaging (4), flow cytometry (5–7), and magnetically activated cell sorting [e.g., immunomagnetic microsphere-based selection (8–10)]. In general, the

1 College of Chemistry and Molecular Sciences and State Key Laboratory of sorting of cells is based on recognition, which requires Virology, Wuhan University, Wuhan, Peoples Republic of China. sufficient random collisions between the cell-specific mi- 2 School of Life Science and Technology, Beijing Institute of Technology, crospheres and the target cells. Evidently, it is difficult to Beijing, Peoples Republic of China. get heavy microspheres to interact with micrometer-sized 3 School of Life Science and Technology, Xi’an Jiaotong University, Peoples Republic of China. target cells; however, the use of nanospheres promises to 4 Section on Molecular Morphogenesis, Program on Cell Regulation and solve such problems, because the lightness and small sizes Metabolism, National Institute of Child Health and Human Development, of nanospheres facilitate their collision and interaction National Institutes of Health, Bethesda, MD. * Address correspondence to this author at: College of Chemistry and with larger objects such as cells. Molecular Sciences and State Key Laboratory of Virology, Wuhan University, Nanomaterials are increasingly being used in biology Wuhan 430072, Peoples Republic of China. Fax 86-27-6875-4067; e-mail and medicine, particularly in the fields of cell recognition [email protected]. Received May 15, 2007; accepted October 2, 2007. and separation. Cell-targeting microspheres conjugated Previously published online at DOI: 10.1373/clinchem.2007.092023 with organic dyes have broad applications in recognizing

2177 2178 Song et al.: Multifunctional Nanosphere Capture of Apoptotic Cells

and imaging target cells (11, 12). Quantum dots (QDs),5 (FITC), which might limit the applications for these bio- relatively new colloidal semiconductor nanocrystals, have conjugates because of the disadvantages of this organic also attracted increasing attention in fluorescence appli- dye. A hallmark of apoptosis is the externalization of cations, including cell imaging and recognition (13, 14), phosphatidylserine, which is recognized by annexin V. because of their unique optical properties (15, 16).In Accordingly, van Tilborg et al. (35) and Prinzen et al. (36) contrast with the use of organic dyes, the ease of gener- used annexin V–conjugated QDs coated with a paramag- ating QDs with different emission wavelengths facilitates netic lipid to detect apoptotic cells with fluorescence and the use of QDs for simultaneously recognizing and imag- magnetic resonance imaging. We recently demonstrated ing several kinds of cells, and this recognition can be that trifunctional avidin-coupled nanospheres are capable modified through the use of different biotargeting mole- of recognizing apoptotic cells preincubated with annexin cules. On the other hand, magnetic nanoparticles offer the V–biotin via the strong biotin-avidin interaction (31).To capability of cell isolation from original or enriched sam- investigate the suitability of such trifunctional nano- ples without the use of centrifugation or filtration (17–22). spheres for use in biomedical research, drug develop- In 1990, Miltenyi et al. (23) developed a method for ment, and/or diagnosis, we have conducted a detailed separating and identifying rare cells that uses magnetic characterization of the properties of these trifunctional nanoparticles coupled to an antibody tagged with a avidin-coupled nanospheres and applied them to visual- fluorescent dye; however, organic dyes have several dis- izing and isolating apoptotic cells from a mixed cell advantageous properties that limit the use of this method. population. More recently, the combination of fluorescent QDs and magnetic nanoparticles into single nanospheres to obtain Materials and Methods fluorescent-magnetic bifunctional nanospheres (FMBNs) reagents has created the potential for broader applications in The TACS Annexin V-Biotin Apoptosis Detection Kit was biomedicine and in clinical diagnosis. We previously purchased from R&D Systems. Avidin, FITC-labeled synthesized such FMBNs by coembedding QDs and biotin (FITC-biotin), and FITC-conjugated streptavidin ␥ nano- -Fe2O3 magnetic nanoparticles into swelling poly (FITC-streptavidin) were obtained from Sigma-Aldrich. (styrene/acrylamide) nanospheres (24). In addition, sev- Coomassie Brilliant Blue G250 and BSA were purchased eral other approaches incorporating the use of QDs and from Shanghai Bio Life Science & Technology. Ultrapure magnetic nanoparticles have recently been applied to the water (18 M⍀⅐cm) was produced with a WaterPro Plus synthesis of fluorescent-magnetic bifunctional nanopar- water-purification system (Labconco). ticles (25–28). More recently, we developed several types of biotargeting trifunctional nanospheres by coembed- preparation of fluorescent-magnetic- ␥ ding QDs and nano- -Fe2O3 particles into poly(styrene/ biotargeting trifunctional nanospheres acrylamide) copolymer nanospheres and then covalently Monodisperse CdSe QDs were first obtained according to coupling different biological reagents, such as folic acid, a published procedure (37). The core/shell CdSe/ZnS biotin, avidin, and antibody, onto the surfaces of the QDs (38) were then prepared by covering the core CdSe nanospheres (24, 29–31). nanocrystals with a thin but higher band-gap material, In this study, we used apoptosis as a model to inves- ZnS, to yield QDs with strong light emission and high ␥ tigate the potential of trifunctional nanospheres for visu- photostability. The nano- -Fe2O3 particles and hydrazine- alizing and efficiently isolating specific cells. Apoptosis, a treated poly(styrene/acrylamide) nanospheres were pre- critical process in the development and homeostasis of pared as previously described (30). FMBNs were subse- multicellular organisms, has been associated with many quently prepared by swelling a 2-mL suspension of the diseases, including AIDS and tumor development (1–3). hydrazine-treated poly(styrene/acrylamide) copolymer ␥ Our understanding of the progression of apoptosis and nanospheres, CdSe/ZnS QDs (3.0 mg), and nano- -Fe2O3 analyses of the associated changes in DNA, proteins, or particles (2.0 mg) in a chloroform/butanol solvent (5:95 organelles would greatly benefit from the ability to visu- by volume). The mixture was then ultrasonicated for alize and isolate cells at various stages of apoptosis. 60 min, centrifuged for 5 min at 2790g, and washed Schellenberger et al. (32, 33) and van Tilborg et al. (34) 3 times with butanol to produce the poly(styrene/ ␥ synthesized a series of functionalized magnetic-fluores- acrylamide) copolymer CdSe/ZnS- -Fe2O3 bifunctional cent nanoparticle conjugates that recognize apoptotic nanospheres. cells. The fluorophores involved in these bioconjugates, We oxidized 0.4 mL of an avidin solution (5 g/L) with however, were not QDs but fluorescein isothiocyanate periodic acid according to a previously published method to create active aldehydes (39). The concentration of aldehyde-containing avidin was approximately 2.8 g/L according to the Bradford assay (40) measured with a 5 Nonstandard abbreviations: QD, quantum dot; FMBN, fluorescent- TU1900 ultraviolet (UV)/visible spectrophotometer magnetic bifunctional nanosphere; FITC, fluorescein isothiocyanate; FITC- biotin, FITC-labeled biotin; FITC-conjugated streptavidin; UV, ultraviolet; PI, (Beijing Purkinje General Instrument). The aldehyde- propidium iodide. containing avidin (240 ␮L) was then mixed with the Clinical Chemistry 53, No. 12, 2007 2179

FMBNs (2.4 mL of a 20.0-g/L suspension), diluted to 3.0 in the 100 ␮L binding buffer containing 20 ␮g FITC- mL, and incubated for6hatroom temperature. The streptavidin and then incubating in the dark for 15 min at nanospheres were washed 5 times with sodium room temperature, washing cells once with 100 ␮L1ϫ phosphate buffer (0.1 mol/L, pH 6.8), and the resulting binding buffer, suspending them in 300 ␮L binding buffer fluorescent-magnetic-biotargeting trifunctional avidin- and then analyzing the cells by flow cytometry (from the coupled nanospheres were stored at 4 °C in sodium protocols provided by the manufacturer). phosphate buffer (0.1 mol/L, pH 6.8). isolation of apoptotic hela cells with measurement of fluorescence spectra trifunctional avidin-coupled nanospheres We measured the fluorescence spectra of QDs in n-hexane UV-irradiated HeLa cells were suspended in 100 ␮L and the FMBNs in ultrapure water by means of a incubation buffer as mentioned above. The cells then were PerkinElmer LS 55 fluorescence spectrometer with a incubated for 15 min at room temperature, washed 3 388-nm laser for excitation and analyzed the results with times with cold 1ϫ PBS, and incubated with the trifunc- FL WinLab 4.00.02 software (PerkinElmer). tional avidin-coupled nanospheres (20.0 g/L in 200 ␮L binding buffer) for 15 min at room temperature. The measurement of the zeta potential of apoptotic HeLa cells bound to the trifunctional avidin- multifunctional nanospheres coupled nanospheres were precipitated with a magnet The zeta potential is a measure of the potential at the and then imaged with the aid of a fluorescence micro- interface between a solid surface and the liquid medium. scope. To demonstrate the nanospheres’ selectivity, we We used a Zetasizer Nano ZS90 (Malvern Instruments) to carried out a control experiment as outlined above except measure the zeta potential of FMBNs before and after that we did not treat the cells with UV light. An additional coupling avidin to their surfaces in 1ϫ PBS (137 mmol/L control consisted of incubating irradiated cells with

NaCl, 2.7 mmol/L KCl, 10 mmol/L Na2HPO4, 1.8 mmol/L avidin-free nanospheres. KH2PO4, pH 7.4). measurement of apoptotic cell isolation measurement of protein on the surfaces of the efficiency with trifunctional avidin-coupled trifunctional nanospheres nanospheres Aldehyde-containing avidin (20 ␮L of a 2.8-g/L solution) The negative control consisted of suspending HeLa was mixed with 50 ␮L, 100 ␮L, 150 ␮L, or 200 ␮Lofa cells (1 ϫ 105) not treated with UV light in 100 ␮L 20.0-g/L suspension of FMBNs, diluted to 500 ␮L with incubation buffer, incubating the cells with 0.5 ␮gPIin sodium phosphate buffer (0.1 mol/L, pH 6.8), and incu- the dark for 15 min at room temperature, incubating the bated for6hatroom temperature. We then centrifuged cells with 20 ␮g FITC-streptavidin in the dark for another the suspensions, measured the amount of unbound avidin 15 min at room temperature, and then analyzing the cells in the supernatants, and washed the nanospheres with by flow cytometry. UV-treated HeLa cells (5 ϫ 105) were sodium phosphate buffer (0.1 mol/L, pH 6.8) until the suspended in 500 ␮L incubation buffer containing 2.5 ␮g Bradford assay detected no residual protein in the super- PI, incubated in the dark for 15 min at room temperature, natants (40). We then estimated the amount of nano- and then divided into 5 equal aliquots. One aliquot of cells sphere-bound avidin by subtracting the amount of un- was resuspended in 100 ␮L1ϫ binding buffer containing bound avidin from that in the starting solution. 20 ␮g FITC-streptavidin, incubated for 15 min in the dark at room temperature, and washed 3 times with cold 1ϫ cell culture PBS. The cells were then collected and analyzed by flow HeLa cells obtained from the China Center of Type cytometry to estimate the fraction of apoptotic cells in the Culture Collection were grown in DMEM with 100 mL/L aliquot of irradiated cells. The other 4 aliquots of cells fetal bovine serum. HeLa cells were irradiated with a were mixed with 50 ␮L, 100 ␮L, 150 ␮L, or 200 ␮L 20-W UV lamp for 10 min to induce apoptosis, which was of the trifunctional nanospheres (20.0 g/L) and incu- verified by flow cytometric analysis on an XL-MCL flow bated in 1ϫ binding buffer for 15 min in the dark. The cytometer (Beckman Coulter) after the cells were stained cells captured by the nanospheres were then isolated with first with propidium iodide (PI) and annexin V-biotin and a magnet for 2 min and imaged with the aid of a then incubated with FITC-streptavidin in a 1ϫ binding fluorescence microscope. The cells remaining in the solu- ␮ buffer (containing 1.8 mmol/L CaCl2, 10 mmol/L HEPES, tion were incubated with 20 g FITC-streptavidin and 150 mmol/L NaCl, 5 mmol/L KCl, 1 mmol/L MgCl2, then analyzed by flow cytometry as described above to pH 7.4) according to the following protocol: washing cells estimate the fraction of apoptotic cells remaining in with cold 1ϫ PBS 3 times, suspending the cells in 100 ␮L each of the samples. The efficiency of the trifunctional incubation buffer [containing 1.0 mg/L annexin V-biotin nanospheres to isolate apoptotic cells was calculated and 10 ␮L of a concentrated solution (10ϫ) of the binding from the difference in the fraction of apoptotic cells in buffer mentioned above] and incubating with 0.5 ␮gPIin cell samples treated and not treated with the trifunctional the dark for 15 min at room temperature, suspending cells nanospheres. 2180 Song et al.: Multifunctional Nanosphere Capture of Apoptotic Cells

imaging by fluorescence microscopy fluorescence spectrum showed that the photolumines- All fluorescence micrographs were taken with the aid of cence properties of the QDs were essentially unchanged an Olympus IX70 inverted fluorescence microscope cou- in the FMBNs (Fig. 1B), although the maximal emission pled with a U-MWB filter cube (450–480 nm/500 nm/ spectra were redshifted by approximately 3–5 nm, com- 515 nm). Samples were spread on glass slides, placed on pared with the free QDs in n-hexane solution. This slight the microscope, visualized with a 100ϫ oil-immersion redshift in the fluorescence-emission spectra might be objective or a 40ϫ common objective, and imaged with a because the QD environments in n-hexane and in the Nikon COOLPIX 5400 digital camera. copolymer are different. Although the nanospheres undergo aggregation with storage, they are monodisperse Results and Discussion and can retain the fluorescence property (Fig. 1, C and D). characterization of the multifunctional To make FMBNs cell specific, we fabricated trifunctional nanospheres avidin-coupled nanospheres by conjugating avidin to the Transmission electron microscopy of the FMBNs (Fig. 1A) surfaces of FMBNs, as we have reported previously (31) ␥ revealed that the QDs and the nano- -Fe2O3 nanoparticles (Fig. 2A). were approximately 2 nm and approximately 8 nm in To understand the status of avidin on the nanospheres diameter, respectively, and were well distributed in the (including the uniformity of coupling, the amount, and poly(styrene/acrylamide) copolymer nanospheres [ap- bioactivity), we 1st studied the uniformity of avidin proximately 200 nm in diameter (30)]. An analysis of the coupling to nanospheres by measuring the zeta potential

Fig. 1. Characterization of FMBNs. ␥ (A), transmission electron microscopy image. Thin arrows and thick arrows indicate the QDs and magnetic nano- -Fe2O3 particles, respectively, embedded inside the nanospheres. (B), normalized fluorescence spectra of free green CdSe (a) and red CdSe/ZnS (c) QDs suspended in n-hexane solution and CdSe (b) and CdSe/ZnS (d) QDs embedded in nanospheres suspended in ultrapure water. Fluorescence microscopy with a 100ϫ oil-immersion objective of green CdSe (C) or red CdSe/ZnS (D) QDs embedded in FMBNs. A.U., arbitrary unit. Clinical Chemistry 53, No. 12, 2007 2181

Fig. 2. Characterization of fluorescent-magnetic-biotargeting trifunctional avidin-coupled nanospheres. (A), schematic diagram of the nanospheres. (B), zeta potential distribution of FMBNs before (green curve) and after (red curve) coupling of avidin to their surfaces. (C), histogram for the amount of avidin molecules coupled to nanosphere surfaces under different conditions: 20 ␮L of a 2.8-g/L avidin solution reacting with 50 ␮L(a), 100 ␮L(b), 150 ␮L(c), or 200 ␮L(d) of the nanosphere suspension; data are presented as the mean (SD). (D–F), fluorescence images of trifunctional avidin-coupled nanospheres after incubation with FITC-biotin: (D), green fluorescence from FITC-biotin binding with avidin. (E), red fluorescence from QDs embedded inside the nanospheres after FITC photobleaching. (F), control reaction consisting of avidin-free FMBNs incubated with FITC-biotin; only a red fluorescence from the nanospheres was observed. K cps, kilo counts per second. of the nanospheres. Avidin has an isoelectric point of cells (31) preincubated with annexin V-biotin via avidin 10.5 and therefore is positively charged at pH 7.4. binding to biotin and annexin V binding to phosphati- Consequently, the zeta potential of the avidin-coated dylserine molecules exposed on the surfaces of the nanospheres will be shifted positively compared with apoptotic cells (Fig. 3). Cells in the early stages of uncoated nanospheres. The results in Fig. 2B show that apoptosis can be distinguished from late-apoptotic/ the FMBNs had peak zeta potentials of Ϫ27.26 mV and necrotic cells by their ability to be labeled with annexin Ϫ 11.15 mV before and after avidin coupling to their V but not with the DNA-binding organic fluorescent surfaces, respectively. The sharp pattern in the zeta dye PI, whereas late-apoptotic cells can be labeled with potential distribution before and after avidin coupling both annexin V and PI. Thus, we investigated the indicates that the nanospheres had similar numbers of potential for such trifunctional nanospheres to identify avidin molecules. We measured the amount of avidin apoptotic cells at different stages of the apoptotic immobilized on the surfaces of the nanospheres with a process and for efficiently isolating apoptotic cells from simple and rapid method based on the difference living cells. The cell in Fig. 4A has a morphology between the protein content of the starting solution and characteristic of cells in early apoptosis, an observation the amount remaining after avidin binding to the in agreement with the intense green fluorescence (from nanospheres. The results from 3 independent experi- the trifunctional nanospheres) at the periphery of the ments are shown in Fig. 2C. Approximately 0.037 mg of cell. The cell also exhibited a weak red fluorescence avidin was coupled per milligram of nanospheres un- (from PI) in the periphery of the nucleolus, because the der our experimental conditions, regardless of the quantity of nanospheres used. These results suggest lack of damage to the nuclear membrane prevented PI that a fixed number of sites on the nanosphere surface from entering the nucleus (Fig. 4B). In contrast, the cell were available for coupling, a finding that agrees with in Fig. 4D shows an intense green fluorescence from the the sharp pattern in the zeta potential distribution, trifunctional nanospheres and an intense red fluores- which shows uniform coupling of avidin on all nano- cence from PI, indicating the presence of phosphatidyl- spheres. We subsequently confirmed that the avidin on serine on the cell surface and the membrane damage the nanospheres specifically interacted with FITC-bi- associated with late-stage apoptosis (Fig. 4C). It is otin, as we previously reported (31) (Fig. 2, D–F). evident that these multifunctional nanospheres in combination with PI staining of nuclear DNA can effec- multifunctional nanosphere recognition of tively distinguish apoptotic cells at different stages of apoptotic cells at different stages of the the apoptotic process. This method may be helpful for apoptotic process understanding the progression of apoptosis and in Our preliminary studies had shown that the avidin- analyzing apoptosis-associated changes in DNA, pro- coupled trifunctional nanospheres could bind apoptotic teins, or organelles. 2182 Song et al.: Multifunctional Nanosphere Capture of Apoptotic Cells

Fig. 3. Microscopy images of apoptotic HeLa cells captured by trifunctional avidin-coupled nanospheres. (A), bright-field image of nanospheres bound to cell surfaces. (B), green fluorescence from QDs embedded inside the nanospheres bound to the cells. (C–F), control experiments: HeLa cells without UV irradiation incubated with trifunctional avidin-coupled nanospheres (C and D); apoptotic HeLa cells incubated with avidin-free FMBNs (E, F). No nanospheres (green fluorescence) were found on cells in these control experiments.

Fig. 4. HeLa cell fluorescence at different apoptotic stages for cells preincubated with annexin V-biotin and PI and then treated with avidin-containing trifunctional nanospheres and magnetic isolation. (A and C), bright-field and (B and D) fluorescence images of the cells irradiated with UV light for 10 min and incubated for 12 h (A and B)or48h(C and D). Clinical Chemistry 53, No. 12, 2007 2183

efficient isolation of target cells with signal (horizontal axis) due to PI binding to nuclear DNA. multifunctional nanospheres Incubating the UV-irradiated cells with the nanospheres As we demonstrated previously (30, 31), the magnetic and precipitating the nanospheres with a magnet pro- property of multifunctional nanospheres can be helpful duced a cell-distribution pattern (Fig. 5C) in the flow for isolating specific cells. We used flow cytometry to cytometry analysis similar to that produced by nonirradi- obtain the fraction of apoptotic cells before and after ated HeLa cells (Fig. 5A), suggesting that nearly all isolation with multifunctional nanospheres to estimate apoptotic cells were removed at all of the nanosphere the efficiency of the multifunctional nanospheres for concentrations used (50–200 ␮L of a 20.0-g/L nanosphere separating apoptotic cells from living cells. Apoptosis- suspension; Fig. 5D). We carried out a quantitative anal- induced HeLa cells were treated as described in Materials ysis by using the equation: and Methods and isolated with the aid of a magnet. The F Ϫ F entire process, including cell recognition and magnetic ϭ b a Es ͑ Ϫ ͒ , isolation, could be accomplished within 35 min. The cells Fb 1 Fa remaining in the solution were then analyzed by flow cytometry after incubation with FITC-streptavidin. The where Es is the isolation efficiency and Fb and Fa are the results showed that in the absence of UV irradiation, few fractions of apoptotic cells in the total sample before and cells were in the apoptotic region (Fig. 5A, upper-right after isolation with multifunctional nanospheres, respec- quadrant), and the vast majority was in the region of live tively. The flow cytometric analysis revealed an isolation cells (Fig. 5A, lower-left quadrant). UV irradiation pro- efficiency for apoptotic cells of approximately 80%; that is, duced an appreciable and increasing proportion of the the trifunctional nanospheres removed approximately cells in the apoptotic region (Fig. 5B, upper-right quad- 80% of the apoptotic cells at all nanosphere concentrations rant), as reflected by the increases in the FITC signal used, with the exception of the highest nanosphere con- (vertical axis) due to FITC-streptavidin binding to an- centration (Fig. 5E). This last result was likely due to the nexin V–biotin bound to the apoptotic cells and in the PI characteristics of the nanospheres themselves. Our tri-

Fig. 5. Flow cytometric analysis of HeLa cells. (A), negative control: typical nonirradiated HeLa cells not incubated with trifunctional nanospheres. (B), positive control: UV-irradiated HeLa cells not incubated with nanospheres. (C), UV-irradiated HeLa cells after isolating the nanosphere-bound apoptotic cells with a magnet. (D), histogram for the fraction of apoptotic cells before (solid columns) and after (open columns) isolation with 50 ␮L(a), 100 ␮L(b), 150 ␮L(c), or 200 ␮L(d) of multifunctional nanospheres; data are presented as the mean (SD) of 3 independent measurements. (E), histogram for the efficiency of isolation of UV-irradiated HeLa cells after incubation with 50 ␮L(a), 100 ␮L(b), 150 ␮L(c), or 200 ␮L(d) of trifunctional nanospheres, isolation with a magnet, and flow cytometric analysis; data are presented as the mean (SD) of 3 independent measurements. 2184 Song et al.: Multifunctional Nanosphere Capture of Apoptotic Cells

functional avidin-coupled nanospheres were made from color labeling and single-excitation/multiple-emission) ␥ poly(styrene/acrylamide) infused with nano- -Fe2O3 par- and magnetic traits. ticles and QDs. The materials sometimes caused un- wanted aggregation. Some factors (i.e., nanosphere size, surface charge level, and concentration) will produce Grant/funding support: This work was supported by the aggregation. For very small nanospheres such as ours, 863 Program (no. 2006AA03Z320), the National Key Sci- maintaining a monodisperse suspension was relatively entific Programs—Nanoscience and Nanotechnology (no. easy at low nanosphere concentrations without the addi- 2006CB933100), the Science Fund for Creative Research tion of a surfactant, but as the nanosphere concentration Groups (no. 20621502), the National Natural Science Foun- in the suspension increases, the likelihood of collisions dation of China (Grant nos. 20505001 and 30570490), the and hydrophobic interactions increases. Such interactions Ministry of Education (nos. 306011 and IRT0543), and the can lead to partial aggregation. Partial nanosphere aggre- Beijing Institute of Technology Fund for Excellent Youth gation can decrease the amount of the immobilized pro- (000Y06-24). This research was also supported in part by the teins accessible to binding by cells in the mixture and Intramural Research Program of the National Institute of decrease Brownian motion, again leading to a reduced Child Health and Human Development, National Institutes probability of the trifunctional nanospheres to bind to of Health. apoptotic cells. In any case, our results show that these Financial disclosures: None declared. multifunctional nanospheres can be used to isolate target Acknowledgments: We thank Wei Dong and Ming-Xi Zhang cells from samples at an efficiency of approximately 80% for their kind help. or greater. References In conclusion, our previous studies demonstrated that 1. Blankenberg FG, Tait JF, Strauss HW. Apoptotic cell death: its the fluorescent-magnetic multifunctional avidin-coupled implications for imaging in the next millennium. Eur J Nucl Med nanospheres could bind to apoptotic cells prebound to 2000;27:359–67. 2. Blankenberg FG, Strauss HW. Will imaging of apoptosis play a role annexin V-biotin. In the present detailed characterization in clinical care? A tale of mice and men. Apoptosis 2001;6:117– of these nanospheres, we have demonstrated that these 23. nanospheres retained the fluorescence spectra of the indi- 3. Kerr JF, Wyllie AH, Currie AR. Apoptosis: a basic biological vidual QDs. We also have demonstrated the easily mono- phenomenon with wide-ranging implications in tissue kinetics. Br J disperse nature of the nanospheres and thus their suit- Cancer 1972;26:239–57. ability for biomedical applications that typically require a 4. Chen FQ, Gerion D. Fluorescent CdSe/ZnS nanocrystal-peptide uniform distribution of nanomaterials. We also have conjugates for long-term, nontoxic imaging and nuclear targeting in living cells. Nano Lett 2004;4:1827–32. shown that a fixed amount of avidin was coupled to per 5. Gorman AM, Samali A, McGowan AJ, Cotter TG. Use of flow milligram nanospheres. Such characteristics are particu- cytometry techniques in studying mechanisms of apoptosis in larly important in many biomedical applications, espe- leukemic cells. Cytometry 1997;29:97–105. cially when quantification is required. Finally and more 6. Jennings CD, Foon KA. Recent advances in flow cytometry: appli- importantly, we have shown in an example of a potential cation to the diagnosis of hematologic malignancy. Blood 1997; use for our nanospheres that the combination of annexin 90:2863–92. V-biotin and PI with the trifunctional nanospheres en- 7. Villas BH. Flow cytometry: an overview. Cell Vis 1998;5:56–61. 8. Medina F, Segundo C, Brieva JA. Purification of human tonsil ables the visualization of apoptotic cells at different stages plasma cells: pre-enrichment step by immunomagnetic selection of apoptosis and that we can isolate apoptotic cells from of CD31(ϩ) cells. Cytometry 2000;39:231–4. living cells with an efficiency of at least 80%. Obviously, 9. Boor PPC, Ijzermans JNM, van der Molen RG, Binda R, Mancham such nanospheres have superior properties of fluores- S, Metselaar HJ, et al. Immunomagnetic selection of functional cence, magnetism, specificity, monodispersity, and agility dendritic cells from human lymph nodes. Immunol Lett 2005;99: in colliding and interacting with large species. We envi- 162–8. sion that the choice of biotargeting molecules other than 10. Medina F, Segundo C, Salcedo I, Garcia-Poley A, Brieva JA. Purification of human lamina propria plasma cells by an immuno- avidin and/or the use of different biotin-coupled mole- magnetic selection method. J Immunol Methods 2004;285:129– cules may make trifunctional nanospheres useful for 35. visually identifying and isolating other target cells at 11. Toublan FJJ, Boppart S, Suslick KS. Tumor targeting by surface- various stages of development, such as tumor cells, acti- modified protein microspheres. J Am Chem Soc 2006;128: vated macrophages, and even bacteria and viruses, for 3472–3. purposes of biochemical, molecular, and genetic analyses, 12. Zhao XJ, Hilliard LR, Mechery SJ, Wang YP, Bagwe RP, Jin SG, et clinical diagnosis, drug development, and so on. In addi- al. A rapid bioassay for single bacterial cell quantitation using bioconjugated nanoparticles. Proc Natl Acad Sci U S A 2004;101: tion, multifunctional nanospheres may become useful for 15027–32. synchronous diagnosis of multiple diseases, the identifi- 13. Hoshino A, Hanaki K, Suzuki K, Yamamoto K. Applications of cation and collection of multiple bacteria or viruses, and T-lymphoma labeled with fluorescent quantum dots to cell tracing simultaneous magnetic resonance and fluorescence imag- markers in mouse body. Biochem Biophys Res Commun 2004; ing owing to the nanospheres’ fluorescence (e.g., multi- 314:46–53. Clinical Chemistry 53, No. 12, 2007 2185

14. Le Gac S, Vermes I, van den Berg A. Quantum dot based probes dual-function microcarriers for optical encoding and magnetic conjugated to annexin V for photostable apoptosis detection and separation. Anal Chem 2006;78:5627–32. imaging. Nano Lett 2006;6:1863–9. 29. Xie HY, Xie M, Zhang ZL, Long YM, Liu X, Tang ML, et al. Wheat 15. Weller H. Colloidal semiconductor Q-Particles: chemistry in the germ agglutinin-modified trifunctional nanospheres for cell recog- transition region between solid state and molecules. Angew Chem nition. Bioconjugate Chem 2007;[Epub ahead of print]. Int Ed Engl 1993;32:41–53. 30. Xie HY, Zuo C, Liu Y, Zhang ZL, Pang DW, Li XL, et al. Cell-targeting 16. Alivisatos AP. Perspectives on the physical chemistry of semi- multifunctional nanospheres with both fluorescence and magne- conductor nanocrystals. J Phys Chem 1996;100:13226–39. tism. Small 2005;1:506–9. 17. Tibbe AG, de Grooth BG, Greve J, Liberti PA, Dolan GJ, Terstappen 31. Wang GP, Song EQ, Xie HY, Zhang ZL, Tian ZQ, Zuo C, et al. LW. Optical tracking and detection of immunomagnetically se- Biofunctionalization of fluorescent-magnetic-bifunctional nano- lected and aligned cells. Nat Biotechnol 1999;17:1210–3. spheres and their applications. Chem Commun 2005;34: 18. Moore A, Marecos E, Bogdanov A Jr, Weissleder R. Tumoral 4276–8. distribution of long-circulating dextran-coated iron oxide nanoparticles in a rodent model. Radiology 2000;214:568–74. 32. Schellenberger EA, Reynolds F, Weissleder R, Josephson L. 19. Berry CC, Wells S, Charles S, Aitchison G, Curtis AS. Cell response Surface-functionalized nanoparticle library yields probes for apo- to dextran-derivatised iron oxide nanoparticles post internalisa- ptotic cells. Chembiochem 2004;5:275–9. tion. Biomaterials 2004;25:5405–13. 33. Schellenberger EA, Sosnovik D, Weissleder R, Josephson L. 20. Wang XB, Yang J, Huang Y, Vykoukal J, Becker FF, Gascoyne PR. Magneto/optical annexin V, a multimodal protein. Bioconjugate Cell separation by dielectrophoretic field-flow-fractionation. Anal Chem 2004;15:1062–7. Chem 2000;72:832–9. 34. van Tilborg GAF, Mulder WJM, Deckers N, Storm G, Reuteling- 21. McCloskey KE, Chalmers JJ, Zborowski M. Magnetic cell separa- sperger CP, Strijkers GJ, et al. Annexin A5-functionalized bimodal tion: characterization of magnetophoretic mobility. Anal Chem lipid-based contrast agents for the detection of apoptosis. Biocon- 2003;75:6868–74. jugate Chem 2006;17:741–9. 22. Su XL, Li YB. Quantum dot biolabeling coupled with immunomag- 35. van Tilborg GAF, Mulder WJM, Chin PTK, Storm G, Reuteling- netic separation for detection of Escherichia coli O157:H7. Anal sperger CP, Nicolay K, et al. Annexin A5-conjugated quantum dots Chem 2004;76:4806–10. with a paramagnetic lipidic coating for the multimodal detection of 23. Miltenyi S, Muller W, Weichel W, Radbruch A. High gradient apoptotic cells. Bioconjugate Chem 2006;17:865–8. magnetic cell separation with MACS. Cytometry 1990;11:231–8. 36. Prinzen L, Miserus RJ, Dirksen A, Hackeng TM, Deckers N, Bitsch 24. Pang DW, Xie HY, Liu Y, Zuo C, Zhang ZL, inventors. Method for NJ, et al. Optical and magnetic resonance imaging of cell death the construction of fluorescent magnetic multifunctional nanopar- and platelet activation using annexin A5-functionalized quantum ticles. China Patent ZL 200310111290.9; 2003. dots. Nano Lett 2007;7:93–100. 25. Wang DS, He JB, Rosenzweig N, Rosenzweig Z. Superparamag- 37. Qu LH, Peng XG. Control of photoluminescence properties of CdSe netic Fe O beads-CdSe/ZnS quantum dots core-shell nanocom- 2 3 nanocrystals in growth. J Am Chem Soc 2002;124:2049–55. posite particles for cell separation. Nano Lett 2004;4:409–13. 26. Yi DK, Selvan ST, Lee SS, Papaefthymiou GC, Kundaliya D, Ying 38. Xie HY, Liang JG, Liu Y, Zhang ZL, He ZK, Lu ZX, et al. Preparation JY. Silica-coated nanocomposites of magnetic nanoparticles and and characterization of overcoated II-VI quantum dots. J Nanosci quantum dots. J Am Chem Soc 2005;127:4990–1. Nanotechnol 2005;5:880–6. 27. Salgueirin˜o-Maceira V, Correa-Duarte MA, Spasova M, Liz-Marza´n 39. Wolfe CAC, Hage DS. Studies on the rate and control of antibody LM, Farle M. Composite silica spheres with magnetic and lumi- oxidation by periodate. Anal Biochem 1995;231:123–30. nescent functionalities. Adv Funct Mater 2006;16:509–14. 40. Bradford MM. A rapid and sensitive method for the quantitation of 28. Sathe TR, Agrawal A, Nie S. Mesoporous silica beads embedded microgram quantities of protein utilizing the principle of protein- with semiconductor quantum dots and iron oxide nanocrystals: dye binding. Anal Biochem 1976;72:248–54. Clinical Chemistry 53:12 2186–2192 (2007) Clinical Immunology

Antibodies against Synthetic Deamidated Gliadin Peptides as Predictors of Celiac Disease: Prospective Assessment in an Adult Population with a High Pretest Probability of Disease

Sonia Niveloni,1 Emilia Sugai,1 Ana Cabanne,1 Horacio Vazquez,1 Julio Argonz,1 Edgardo Smecuol,1 Marı´a L. Moreno,1 Fabio Nachman,1 Roberto Mazure,1 Zulema Kogan,2 Juan C. Gomez,3 Eduardo Maurin˜o,1 and Julio C. Bai1*

Background: Noninvasive serologic tests have shown detecting IgA ؉ IgG a-DGP plus IgA a-tTG had 100% high diagnostic accuracy for celiac disease (CD) in positive and negative predictive values if concentrations selected populations. Our aim was to determine proof both tests in either combination were above or below spectively the performance of CD-related serology in the cutoff. individuals undergoing intestinal biopsy because of Conclusions: In a population with high pretest proba- clinical suspicion of small-bowel disorders. bility, the newly developed a-DGP tests have diagnostic Methods: We enrolled 141 unselected consecutive adult accuracy that is at least equivalent to that of established patients attending a small-bowel disease clinic. Patients assays. underwent endoscopy and biopsy; serum samples were © 2007 American Association for Clinical Chemistry obtained at that time for measurements of anti–tissue transglutaminase (a-tTG), IgA and IgG anti–deamidated Intestinal biopsy is still considered the gold standard for gliadin-related peptide (a-DGP), and IgA antiactin anti- diagnosis of celiac disease (CD)3 (1–4). For more than 30 bodies (AAAs). Characterization of patients was based years, CD-related serology has been used as a valuable on histological criteria (Marsh type II lesion or greater). marker for diagnosis, screening, and noninvasive fol- Results: The prevalence of CD was 42.5%. Sensitivity, low-up of patients (1, 5–9). Over the years, celiac serology specificity, and positive and negative predictive values has evolved, with the identification of newer and more were >90% for most assays. Diagnostic accuracy based specific antibodies and the improvement of technical on ROC curve analysis was similar for all assays [area approaches resulting in more reliable assays (10). under the curve (95% CI): 0.996 (0.967–0.998) for a-tTG, Currently, the well-known endomysial (EmA) and 0.995 (0.964–0.998) for IgA a-DGP, 0.989 (0.954–0.999) anti–tissue transglutaminase (a-tTG) autoantibodies are for IgG a-DGP, 0.996 (0.966–0.998) for blended conju- considered to be among the most reliable noninvasive gated of IgA ؉ IgG a-DGP in a single assay, and 0.967 tools for detecting a disease in all of internal medicine (0.922–0.990) for AAA]. The combinations of 2 tests, IgG (5, 11–15). Although these exhibit very high sensitivity a-DGP plus IgA a-tTG or the single blended conjugate and specificity for CD (16–21), recent investigations have shown that accuracy of tests remains controversial in 2 areas. First, sensitivity has been considered unacceptable 1 Department of Medicine, Hospital de Gastroenterologıá “Dr. Carlos both in patients with minor degrees of mucosal damage Bonorino Udaondo”, Buenos Aires, Argentina. and in cases with a more indolent clinical course (22, 23). 2 Pathology Service, Department of Medicine, Hospital de Gastroenterolo- Second, current antibodies seem to have variable accuracy gia´, “Dr. Carlos Bonorino Udaondo”, Buenos Aires, Argentina. 3 Unidad de Soporte Nutricional y Malabsorcioń, Hospital San Martıń, La in the follow-up of patients on a gluten-free diet (24). Plata, Argentina. * Address correspondence to this author at: Department of Medicine, Hospital de Gastroenterologıá “Carlos Bonorino Udaondo”, Av. Caseros 2061, 1264 Buenos Aires, Argentina. Fax 54-11-4304-1018; e-mail [email protected]. 3 Nonstandard abbreviations: CD, celiac disease; EmA, endomysial; a-tTG, Received October 5, 2006; accepted September 13, 2007. anti–tissue transglutaminase; DGP, deamidated gliadin-related peptide; a- Previously published online at DOI: 10.1373/clinchem.2006.081364 DGP, anti-DGP; AAA, antiactin antibodies; IEL, intraepithelial lymphocyte.

2186 Clinical Chemistry 53, No. 12, 2007 2187

Recently, a series of new antibody tests have demon- scopic findings, and serum samples were obtained from strated very high sensitivity for diagnosis of CD. One of all study patients at the time of the endoscopic procedure. these new assays detecting IgA antibodies to the fibrillar Demographic and clinical data of all patients enrolled are form of the depolymerized actin protein of the cytoskel- reported in Table 1. We based the categorization of eton is promising but not completely defined (25–27). patients and controls on currently accepted histological Very recently, we have reported highly encouraging re- criteria, the presence of a type II or more severe enterop- sults with the use of ELISAs to detect antibodies binding athy (Marsh modified classification) (1, 21). The final synthetic deamidated gliadin-related peptides (DGPs). diagnosis of CD was supported by the additional pres- Both isotypes (IgA and IgG) of the peptide antibodies ence of positive a-tTG antibodies or EmA and/or the [anti-DGPs (a-DGPs)] have been shown to be highly histological response to a gluten-free diet. Patients were sensitive and specific for active CD (28–31). categorized according to the clinical status at the time of The outstanding performance of the newer noninva- diagnosis as presenting with classic symptomatic disease sive CD serology opens the possibility that the tests can be (mainly gastrointestinal symptoms); an atypical form used not only as a marker of the disease but, more (subclinical CD) such as chronic anemia, hypertransami- importantly, as a substitute for intestinal biopsy in se- nasemia, or autoimmune diseases; or a silent clinical lected populations. Interestingly, although intestinal bi- course (asymptomatic CD). The latter group consisted of opsy is still considered the gold standard for diagnosis of patients detected during evaluation of 1st-degree relatives CD, the procedure has disadvantages—for instance, it is of index cases (32). often refused by patients and has inherent difficulties that can lead to improper diagnosis (1, 5). Thus, serology can cd-related serology be a very interesting adjunct or alternative to biopsy. Serum samples were kept frozen at Ϫ30 °C until assays Most prior studies assessing the diagnostic perfor- were performed. Tests were performed by personnel who mance of the available CD serologic tests have suffered did not have knowledge of the diagnoses. The following 5 from several potential design biases, including the use of assays were used. (a) IgA a-tTG (Quanta Lite™ h-tTG IgA, preselected populations. Also important is the inherent Inova Diagnostic) by ELISA (cutoff provided by the interdependence between the assays’ test results, the manufacturer, 20 kU/L) was used. (b) For patients with diagnostic criteria used to define CD, and the selection negative a-tTG test and controls with positive a-tTG bias that results (6, 7). Ideally, a marker for screening of a antibodies, we further tested for IgA EmA by immuno- disorder must have high predictive values (both positive fluorescence on primate esophagus substrate (Inova Di- and negative). The predictive value of a test is determined agnostics) tested at a 1:5 dilution. (c) Newly developed by the sensitivity and specificity of the test and is also ELISAs were used to separately detect IgA and IgG affected by the prevalence or pretest probability of the antibodies reacting with a fully synthetic, selectively disorder in the assessed population. Considering all these glutamine-to-glutamate substituted gliadin-analogous aspects, our main objective has been to assess prospec- peptide that incorporates several B-cell epitopes (IgA tively the performance of serologic tests in populations a-DGP and IgG a-DGP) by use of a reagent set provided with different prevalence. In this study, our aim was to by the manufacturer for research use only (Quanta Lite report on the performance of some established and newer Gliadin IgA and IgG II, Inova Diagnostic; cutoff deter- tools for detecting CD in adult individuals with high mined in a former study, 20 kU/L) (31).(d) A single pretest probability undergoing intestinal biopsy because reagent set was used to assess simultaneously the pres- of clinical suspicion of a small-bowel disorder. ence of both antibody isotypes (IgA ϩ IgG a-DGP; Quanta Lite Celiac DGP Screen; Inova Diagnostic; cutoff provided Materials and Methods by the manufacturer, 20 kU/L). Assays (c) and (d) use as patients antigen the same gliadin-analogous peptide described. From December 2004 to December 2005, we studied a The peptide is constructed so that each epitope is pre- series of serum samples collected prospectively from a group of 141 consecutive adult patients with suspected intestinal disorders during their 1st clinic visit at the Small Table 1. Demographic and clinical data of the overall Bowel Diseases Clinic of the “Dr. Carlos Bonorino Uda- population enrolled. ondo” Gastroenterology Hospital. Patients considered for n 141 study inclusion underwent an upper gastrointestinal en- Median age (range), years 38 (16–80) doscopy and intestinal biopsy. Those with a previous Female/male, n 114/27 2 diagnosis of CD (n ϭ 176), prior treatment with a gluten- Mean body mass index (SD), kg/cm 20.6 (3.2) free diet (n ϭ 32), CD-related serology tests performed Symptom inducing consultation, n before enrollment (n ϭ 235), or diagnosis of dermatitis Chronic diarrhea 114 herpetiformis (n ϭ 3) were excluded from the study. After Weight loss 115 Chronic anemia 95 giving informed consent, all study patients underwent Distension 11 intestinal biopsy irrespective of the clinical and/or endo- 2188 Niveloni et al.: CD Serology

sented in a proper conformational shape. The conjugate is curves and their 95% CIs were determined by use of a blend of both antihuman IgA and IgG, with most of the Medcalc. The performance of different combinations of reactivity biased toward the IgG (approximate IgG vs IgA assays reported (e.g., a-DGPs plus a-tTG) was assessed ratio, 70:30). (e) IgA-type antiactin antibodies (AAAs) taking into account that a given combination of tests is determined by use of a modification of a commercial considered positive if at least 1 of the 2 tests has concen- ELISA assay for IgG type AAAs (Quanta Lite Actin; Inova tration above the cutoff and negative if both tests are Diagnostics) and an antihuman IgA conjugate (Inova below the cutoff. Comparisons were performed using the Diagnostic) were used. Serum samples were studied at Student t-test, Mann–Whitney U-test, ␹2 test, or Fisher 1:101 dilutions. According to our previous study, the exact test, as appropriate. cutoff for normal AAA concentration was 25 kU/L (27). All reagents were generously provided by the Results manufacturer. diagnostic characterization of patients and clinical findings endoscopic procedure and small-bowel With histological criteria used as the gold standard, CD histology enteropathy was diagnosed in 60 patients. The remaining We obtained biopsy samples from the distal duodenum 81 patients had no histological evidence of active CD. As by duodenoscopy, following a standard protocol (31). All expected and according to the design of the protocol, we procedures were performed by the same operator (J.A.) detected CD only in patients with symptomatic disease blinded to the clinical and laboratory data. Endoscopic (patients with classic symptoms or atypical manifesta- examination of the 2nd duodenal portion was reported tions; Table 2). The very high prevalence of CD in the with and without air insufflations. The endoscopic mark- population enrolled was, as expected, very similar to that ers evaluated were scalloped duodenal folds, mosaic reported by our group in previous publications. Despite pattern, and reduction in the number of folds (33). Their the fact that we avoided preselection bias by excluding performance in the suspicion of CD will be reported patients in whom CD was previously suspected or diag- separately. At least 3 samples were obtained using con- nosed, the high prevalence could be due to the fact that ventional endoscopic forceps (no enrichment, open cup: 8 our referral clinic receives only patients with small-bowel mm). Samples were oriented carefully on paper, fixed in and diarrhea disorders. As reported in Table 2, the mean 10% formalin, embedded in paraffin wax, and conven- body mass index of CD patients was significantly lower tionally stained with hematoxylin and eosin. than that of control individuals (P Ͻ0.002). Table 2 also Morphology and quantitative assessments [intraepi- shows the histological characteristics of small-bowel bi- thelial lymphocyte (IEL) density] were performed by 1 of opsies according to the modified Marsh classification. 2 experienced observers (A.C. and Z.K.) who were un- aware of the clinical and laboratory findings of the patients. Morphology was categorized according to the Table 2. Comparison of demographic data, clinical modified Marsh classification (1, 22). Briefly, type 0 is characteristics, and histological features of CD patients normal mucosa; type I is an infiltrative stage marked by a and controls. normal mucosal architecture in which the villous epithe- CD Control lium has intraepithelial lymphocytosis (Ͼ30 IELs/100 n6081 epithelial cells). Type II shows the addition of enlarged Female/male, n 53/7 62/19 crypts (hyperplasic stage), and type III comprises a large Mean age (range), years 36 (19–72) 38 (16–80) 2 a spectrum of changes ranging from minor villous atrophy Mean body mass index (SD), kg/cm 19.3 (3.2) 21.5 (4.2) to complete villous atrophy (subcategorized as Marsh IIIa, Clinical categorization of CD IIIb, and IIIc). Qualitative and quantitative findings are Classic 51 Subclinical 9 listed here as reported by each observer. Asymptomatic 0 Marsh classification, n ethics and statistics Type 0 0 80 The protocol was approved by the Research and Ethical Type I 0 Committees of the Gastroenterology Hospital. Data were Type II 0 analyzed using Statistix 7 for Windows Analytical Soft- Type IIIa 2 1b ware (2000 Analytical Software). According to data distri- Type IIIb 9 bution, results are reported as mean (SD) or median and Type IIIc 49 range, and statistical analyses were used as appropriate. Mean IELs/100 epithelial 40.8 (15.5) 11.2 (4.4)c The diagnostic performance of single tests was deter- cells (SD) mined by comparing sensitivity, specificity, 95% CIs, a P Ͻ0.002. positive and negative predictive values, and likelihood b Control case with a type IIIa histology had a gastrojejunal anastomosis with ratios calculated using conventional formulas (Medcalc, small-intestinal bacteria overgrowth. c Ͻ version 9.2.1.0). The area under the curve for the ROC P 0.00001 (Student t-test). Clinical Chemistry 53, No. 12, 2007 2189

tests expressed in the binary form (positive or negative). Table 3. Patients with a positive test and concentrations Ͼ (kU/L) of serology tests in CD patients and controls. Interestingly, whereas 59 of 60 patients had 1 positive test, 1 case was positive only for IgA a-tTG antibody and CD Control had a Marsh IIIb lesion. In contrast, 10 controls had only n6081 Serology tests 1 positive test, and 1 case had 2 positive tests (IgA a-DGP IgA a-DGP and IgA a-tTG). This individual was IgA EmA negative Positive, n 59 5 and presented with a nonatrophic histology with normal Mean concentration 277.4 (168.9) 6.5 (10.5)a IEL density. Final diagnoses in non-CD patients with (SD) positive serology were undetermined in 3; irritable bowel IgG a-DGP syndrome in 4 (2 of these were 1st-degree relatives of CD Positive, n 58 0 patients); and colorectal cancer, choleretic diarrhea, and Mean concentration 121.1 (54.9) 2.0 (2.8)a chagasic megacolon in 1 patient each. In these tables, we (SD) can also establish the frequency of positive and negative ϩ IgA IgG a-DGP results for combinations of tests in both patients and Positive, n 59 1 controls to determine the best combinations and whether Mean concentration 176.8 (86.2) 2.0 (3.6)a (SD) this approach is better than using a single determination IgA a-tTG of 1 assay. Positive, n 57 2 Table 6 shows the statistical analysis of all serology Mean concentration 142.8 (99.2) 6.9 (6.2)a tests assessed. Both IgA a-DGP and the dual conjugate (SD) a-DGP had a diagnostic sensitivity of 98.3%. Each test AAA missed only 1 patient. The IgG a-DGP was negative in all Positive, n 52 4 controls. In addition, the dual-conjugate test (IgA plus Median concentration 143.3 (12.0–550.0) 11.9 (1.0–64.0)a IgG a-DGP) had only 1 false-positive result. Thus, the (range) positive predictive value was 100% for IgG a-DGP and a P Ͻ0.00001 (Student t-test). 98.3% for the IgA plus IgG a-DGP test. Furthermore, the highest point estimates of positive and negative likeli- Only a minority of patients had mild histological changes hood ratios were produced by the same assays. (Marsh type II and IIIa); most patients showed severe We analyzed the performance of different combina- villous atrophy (Marsh IIIb or IIIc). No CD patient or tions of 2 tests, considering a result to be positive if at least control had histological evidence of Marsh type I lesion, 1 of the assays produced a concentration above the cutoff and the highest IEL density determined in controls was and negative if both were below the cutoff. Thus, the 23.3%. All these features confirm that the newly diag- associations of (a) IgG a-DGP plus IgA a-tTG or (b) the nosed CD population represents a more severe end of the dual-specificity conjugate (IgA plus IgG a-DGP) plus IgA clinical spectrum of the disorder. Three patients had a-tTG exhibited 100% sensitivity and negative predictive negative a-tTG and EmA tests, and the diagnosis of CD value (Tables 4–6). On the other hand, if we consider only was supported by improvement of the intestinal mucosa cases with both assays positive (concentrations above the at rebiopsy. cutoff), the positive predictive value increases to 100% for both combinations, with sensitivities of 93.3% and 91.7%, antibody testing respectively. Our serology assessment did not detect patients or controls with IgA deficiency. Table 3 shows the number of Discussion patients and controls with positive tests and the mean Screening and diagnostic algorithms to detect CD are concentrations for all serology tests assessed. Compared based on the sequential use of serologic tests followed by with controls, CD patients had highly significantly greater a confirmatory intestinal biopsy (20, 21, 34).CDisa absolute values for all assays assessed (P Ͻ0.00001 for all unique autoimmune disorder in which very specific sero- comparisons). Tables 4 and 5 show results of individual logical testing has demonstrated high clinical utility. In

.Table 4. Positive (؉) and negative (؊) results for the different tests in the CD subgroup na IgA a-DGP IgG a-DGP IgA ؉ IgG a-DGP IgA a-tTG IgA AAA Biopsy 50 ϩϩ ϩ ϩϩϩ 5 ϩϩ ϩ ϩϪϩ 2 ϩϩ ϩ ϪϪϩ 1 ϩϩ ϩ Ϫϩϩ 1 ϩϪ ϩ ϩϩϩ 1 ϪϪ Ϫ ϩϪϩ a n, number of patients with the given combination. 2190 Niveloni et al.: CD Serology

.Table 5. Positive (؉) and negative (؊) results for different tests in individuals without CD n IgA a-DGP IgG a-DGP IgA ؉ IgG a-DGP IgA a-tTG IgA AAA Biopsy 70 ϪϪ Ϫ ϪϪϪ 4 ϪϪ Ϫ ϪϩϪ 4 ϩϪ Ϫ ϪϪϪ 1 ϩϪ Ϫ ϩϪϪ 1 ϪϪ ϩ ϪϪϪ 1 ϪϪ Ϫ ϩϪϪ the last 30 years, a series of alimentary antibodies (anti- obtained with the widely used and reliable a-tTG. This gliadin) and autoantibodies (antireticulin, a-tTG, and study confirmed the high diagnostic accuracy for the IgG EmA) have been successfully used for detecting CD. isotype reported in our earlier study (30) and contrasts Recently, newly developed assays such as AAA and with findings reported for other antibodies (including a-DGP have increased interest in the use of these nonin- traditional gliadin antibodies, EmA, and a-tTG). vasive serological tools (25–31). In this prospective study, In the present study, we tested for the 1st time the we aimed to determine the performance of individual performance of a newly developed assay to detect both tests or combinations of these tests that could result in isotypes (IgA and IgG) of the a-DGP simultaneously. This cost-effective selection of cases referred to a specialized new test was designed on the basis of our prior observa- center. Furthermore, although the very high prevalence of tion that some patients are positive only for the IgG the disease in our population imposes some limitations to a-DGP antibody whereas others are positive only for the the interpretation of results, these newer assays have high IgA antibodies against DGP (31). On the basis of the high predictive values, and their use could minimize the often- sensitivity and extremely high specificity of the IgG refused option of endoscopy and intestinal biopsy. isotype of the a-DGP antibody (30), the single conjugate Because EmA is a highly operator-dependent assay set was blended in favor of IgG (70% vs 30% for the IgA). (6, 7) and has a history of a lack of sensitivity (22–24),we As expected, the blended conjugate was very useful, used it only to resolve conflicting results such as a-tTG– correctly identifying 98.3% of CD cases and 98.8% of negative patients or positive controls (19). We also de- individuals without CD. cided to not include standard AGA assays because of their Interestingly, only 11 individuals in the control group well-known lack of sensitivity and specificity. had any false-positive results. Whereas 1 had 2 positive Considering the performance of individual tests, all assays (a-tTG and IgA a-DGP; the final diagnosis was newer serology tests evaluated exhibited results similar to irritable bowel syndrome), 10 presented only 1 positive those reported in studies of other populations. Although test. In our critical analysis of these false-positive cases it without statistically significant differences, the a-DGP seems likely that in 2 or more of these controls the positive tests had especially impressive results, such as those serology might indicate the potential for developing CD.

Table 6. Statistical performance of individual CD serologic tests and some combinations.a Test Sensitivity, % Specificity, % PPV,b % NPV, % PLR NLR AUC/ROC IgA a-DGP 98.3 93.8 92.2 98.7 15.9 0.02 0.995 95% CI 91.0–99.7 86.2–97.9 0.964–0.998 IgG a-DGP 96.7 100.0 100.0 97.6 NC 0.03 0.989 95% CI 88.4–99.5 95.5–100.0 0.954–0.999 IgA ϩ IgG a-DGP 98.3 98.8 98.3 98.8 79.6 0.02 0.996 95% CI 91.0–99.7 93.3–99.8 0.966–0.998 IgA a-tTG 95.0 97.5 96.6 96.3 38.5 0.05 0.996 95% CI 86.1–98.9 91.3–99.6 0.967–0.998 IgA AAA 86.7 95.1 92.9 90.6 17.5 0.14 0.967 95% CI 75.4–95.2 87.8–98.6 0.922–0.990 IgA a-DGP ϩ IgA a-tTG 100.0 92.6 90.9 100.0 13.5 0 95% CI 94.0–100.0 84.6–97.2 IgG a-DGP ϩ IgA a-tTG 100.0 97.5 96.7 100.0 40.0 0 95% CI 94.0–100.0 91.3–99.6 IgA ϩ IgG a-DGP ϩ a-tTG 100.0 96.3 95.2 100.0 27.0 0 95% CI 94.0–100.0 89.5–99.2 a The statistical performance of the 3 combinations assessed is reported considering a result to be positive if at least 1 of the 2 tests has a concentration above the cutoff and negative if both are below the cutoff. b PPV, positive predictive value; NPV, negative predictive value; PLR, positive likelihood ratio; NLR, negative likelihood ratio; AUC, area under the curve. Clinical Chemistry 53, No. 12, 2007 2191

The evaluation of combinations of 2 different assays 4. Bai JC, Zeballos E, Fried M, Corazza GR, Schuppan D, Farthing also shows very interesting results. If we take into account MJG, et al. Celiac Disease. WGO-OMGE Practice Guidelines. World that a given combination of tests is considered positive if Gastroenterology News 2005;10:S1–8. 5. Green PHR, Rostami K, Marsh MN. Diagnosis of celiac disease at least 1 of the 2 tests has a concentration above the cutoff [Review]. Best Pract Res Clin Gastroenterol 2005;19:389–400. and negative if both tests are below the cutoff, all 3 6. Ciclitira PJ. Celiac disease: a technical review. Gastroenterology combinations assessed have absolute sensitivity (100%) 2001;120:1526–40. and specificity comparable to single assays. However, if 7. Hill ID. What are the sensitivity and specificity of serologic tests for we consider as positive a finding that both assays were celiac disease? Do sensitivity and specificity vary in different above the cutoff, the associations of either the blended populations? Gastroenterology 2005;128:S25–32. conjugate of IgA plus IgG a-DGP plus IgA a-tTG or IgA 8. Rostom A, Dube´ C, Cranney A, Saloojee N, Sy R, Garritty C, et al. a-tTG plus IgG a-DGP had a positive predictive value of The diagnostic accuracy of serologic tests for celiac disease: a 100%. This serological approach allowed identification of systemic review. Gastroenterology 2005;128:S38–46. 9. Corazza GR, Biagi F, Andreani ML, Gasbarrini G. Screening test for 93% and 92% of new cases for each option, respectively. In coeliac disease. Lancet 1997;349:325–6. addition, no patients with negative results for both assays 10. Burgin-Wolff A, Berger R, Gaze H, Huber H, Lentze MJ, Nussle D. of the given combination were found to have CD. These IgG, IgA and IgE gliadin antibody determinations as screening test final observations suggest that these combinations could for untreated coeliac disease in children, a multicentre study. Eur obviate the need for intestinal biopsy in 95% of patients in J Pediatr 1989;148:496–502. this clinical scenario. Unfortunately, we could not explore 11. Stern M. Comparative evaluation of serologic tests for celiac the utility of single or combined tests in the identification disease: a European initiative toward standardization. J Pediatr of IgA-deficient CD cases because we did not encounter Gastroenterol Nutr 2000;31:513–9. 12. Maki M, Hallstrom O, Marttinen A, Lipsanen V, Viander M, Holm K, this situation in our study population. et al. Screening tools for use in coeliac disease. In: Aurichio S, Visakorpi JK, eds. Common Food Intolerance I: Epidemiology of In conclusion, this prospective study on a high pretest Celiac Disease. Basel: Krager, 1992;93–104. probability population showed that the newer noninva- 13. McMillan SA, Haughton DJ, Biggart JD, Edgar JD, Porter KG, sive serology tests are highly accurate markers of CD. The McNeill TA. Predictive value for coeliac disease of antibodies to newly developed a-DGP test, alone or in combination gliadin, endomysium, and jejunum in patients attending for jejunal with the more commonly used a-tTG test, can potentially biopsy. BMJ 1991;303:1163–5. be used in many cases to avoid intestinal biopsy, thus 14. Dieterich W, Laag E, Schopper H, Volta U, Ferguson A, Gillett H, et al. Autoantibodies to tissue transglutaminase as predictors of having an impact on cost savings and better acceptance by celiac disease. Gastroenterology 1998;115:1317–21. patients. 15. Sulkanen S, Halttunen T, Laurila K, Kolbo KL, Korponay-Szabo IR, Sarnesto A, et al. Tissue transglutaminase autoantibody enzyme- linked immunosorbent assay in detecting celiac disease. Gastro- enterology 1998;115:1322–8. Grant/funding support: The study was supported by a grant 16. Troncone R, Maurano F, Rossi M, Micillo M, Greco L, Auricchio R, from the Consejo de Investigacio´n en Salud del Ministerio de et al. 19A antibodies to tissue transglutaminase: an effective Salud del Gobierno Auto´nomo de la Ciudad de Buenos Aires, diagnostic test for celiac disease. J Pediatr 1999;134:166–71. Argentina. Reagent sets used in this study were generously 17. Sugai E, Selvaggio G, Vazquez H, Viola M, Mazure R, Pizarro B, et provided by Inova Inc., San Diego, CA. al. Tissue transglutaminase antibodies in celiac disease: assessment of a commercial kit. Am J Gastroenterol 2000;95:2318–22. Financial disclosures: Inova Inc. was not involved in the 18. Sblattero D, Berti I, Trevisiol C, Marzari R, Tommasini A, Bradbury study design, data analysis, interpretation of results, or A, et al. Human recombinant tissue transglutaminase ELISA (an decision to submit the manuscript for publication. innovative diagnostic assay for celiac disease). Am J Gastroen- Acknowledgments: We thank Dr. Christina Surawicz from terol 2000;95:1253–7. the University of Washington for critical reading of our 19. Fabiani E, Catassi C. International Working Group on Eu-tTG. The manuscript and Dr. Walter Binder from Inova Inc. for the serum 19A class anti-tissue transglutaminase antibodies in the generous cooperation. diagnosis and follow up of coeliac disease: results of an international multi-centre study. Eur J Gastroenterol Hepatol 2001;13: 659–65. 20. Tesei N, Sugai E, Vazquez H, Smecoul E, Niveloni S, Mazure R, et References al. Antibodies to human recombinant tissue transglutaminase 1. Marsh MM. Gluten histocompatibility complex and the small may detect coeliac disease patients undiagnosed by endomysial intestine: a molecular and immunobiologic approach to the spec- antibodies. Aliment Pharmacol Ther 2003;17:1415–23. trum of gluten sensitivity (“celiac sprue”). Gastroenterology 1992; 21. Gomez JC, Selvaggio G, Pizarro B, Viola M, La Motta G, Smecuol 102:230–54. E, et al. Value of a screening algorithm for celiac disease using 2. Fasano A, Catassi C. Current approaches to diagnosis and treat- tissue transglutaminase antibodies as first level in a population- ment of celiac disease: an evolving spectrum. Gastroenterology based study. Am J Gastroenterol 2002;97:2785–90. 2001;120:636–51. 22. Rostami K, Kerckhaert J, Tiemessen R, von Blomberg ME, Meijer 3. National Institutes of Health Consensus Development Conference JWR, Mulder CJJ. Sensitivity of antiendomysium and antigliadin Statement on Celiac Disease, June 28–30, 2004. Gastroenterol- antibodies in untreated celiac disease: disappointing in clinical ogy 2005;128:S1–9. practice. Am J Gastroenterol 1999;94:888–94. 2192 Niveloni et al.: CD Serology

23. Tursi A, Brandimarte G, Giorgetti GM, Gigliobianco A, Lombardi D, 29. Aleanzi M, Demonte AM, Esper C, Garcilazo S, Waggener M. Gasbarrini G. Low prevalence of antigliadin and anti-endomysium Celiac disease: antibody recognition against native and selectively antibodies in subclinical/silent celiac disease. Am J Gastroenterol deamidated gliadin peptides. Clin Chem 2001;47:2023–8. 2001;96:1507–10. 30. Schwertz E, Kahlenberg F, Sack U, Richter T, Stern M, Conrad K, 24. Vahedi K, Mascart F, Mary JY, Laberenne JE, Bouhinik Y, Morin et al. Serologic assay based on gliadin-related nonapeptides as a MC, et al. Reliability of antitransglutaminase antibodies as pre- highly sensitive and specific diagnostic aid in celiac disease. Clin dictors of gluten-free diet compliance in adult celiac disease. Am J Chem 2004;50:2370–5. Gastroenterol 2003;98:1079–87. 31. Sugai E, Va´zquez H, Nachman F, Moreno ML, Mazure R, Smecuol 25. Clemente MG, Musu MP, Troncone R, Volta V, Congia M, Ciacci C, E, et al. Accuracy of testing for antibodies to synthetic gliadin- et al. Enterocyte actin autoantibody detection: a new diagnostic related peptides in celiac disease. Clin Gastroenterol Hepatol tool in celiac disease diagnosis: results of a multicenter study. 2006;4:1112–7. Am J Gastroenterol 2004;99:1551–6. 32. Niveloni S, Maurin˜o E, Pedreira S, Vazquez H, Smecuol E, Moreno ML, et al. Clinical picture. In: Catassi C, Fasano A, Corazza GR, 26. Carroccio A, Brusca I, Iacono G, Di Prima L, Teresa S, Pirrone G, eds. The Global Village of Coeliac Disease. Perspectives on et al. Anti-actin antibodies in celiac disease: correlation with Coeliac Disease, Vol. II. Pisa: AJC Press, 2005:23–44. intestinal mucosal damage and comparison of ELISA with immu- 33. Niveloni S, Fiorini A, Dezi R, Pedreira S, Smecuol E, Vazquez H, et nofluorescence assay. Clin Chem 2005;51:917–20. al. Usefulness of videoduodenoscopy and vital dye staining as 27. Pedreira S, Sugai E, Moreno ML, Vazquez H, Niveloni S, Smecuol indicators of mucosal atrophy of celiac disease: assessment of E, et al. Significance of smooth muscle/anti-actin autoantibodies interobserver agreement. Gastrointest Endosc 1998;47:223–9. in celiac disease. Acta Gastroenterol Latinoam 2005;35:83–93. 34. Gomez JC, Selvaggio GS, Viola M, Pizarro B, la Motta G, de Barrio 28. Osman, AA, Gu¨nnel T, Dietl A, Uhlig HH, Amin M, Fleckenstein B, S, et al. Prevalence of celiac disease in Argentina: screening of an et al. B cell epitopes of gliadin. Clin Exp Immunol 2000;121:248– adult population in the La Plata area. Am J Gastroenterol 2001; 54. 96:2700–4. Clinical Chemistry 53:12 2193–2201 (2007) Infectious Disease

Midregional Pro-A-Type Natriuretic Peptide and Carboxy-Terminal Provasopressin May Predict Prognosis in Community-Acquired Pneumonia

Mar Masia´,1* Jana Papassotiriou,2 Nils G. Morgenthaler,2 Ildefonso Herna´ ndez,3 Conrado Shum,4 and Fe´lix Gutie´rrez1

Background: Markers to better assess severity of dis- procalcitonin, C-reactive protein, lipopolysaccharide- ease in patients with community-acquired pneumonia binding protein, CT-proAVP, and MR-proANP concen- (CAP) would help improve medical care of this condi- trations, only CT-proAVP remained an independent Cutoff .(0.007 ؍ tion. The hemodynamic biomarkers carboxy-terminal predictor of death (odds ratio 1.05, P provasopressin (CT-proAVP; copeptin) and midregional values of >18.9 pmol/L for CT-proAVP and >227 proatrial natriuretic peptide (MR-proANP) are increased pmol/L for MR-proANP showed the highest diagnostic under septic conditions, in which MR-proANP has been accuracy to predict mortality. described as a prognostic predictor. We aimed to explore Conclusions: CT-proAVP and MR-proANP may be the diagnostic accuracy of MR-proANP and CT-proAVP used to predict prognosis in patients with CAP. to predict mortality in patients with CAP. © 2007 American Association for Clinical Chemistry Methods: We conducted a prospective observational study of patients with CAP. We measured biomarkers in Assessing severity of community-acquired pneumonia serum samples obtained at diagnosis and performed (CAP)5 still remains a challenge for clinicians (1, 2). In the univariate and multivariate analyses to identify poten- last few years, prognostic scoring systems for CAP have tial predictors of mortality. been developed to stratify patients on the basis of their Results: CT-proAVP and MR-proANP concentrations risk of mortality, but methods are not ideal (3). The most were measured in 173 patients. We found a positive prominent and validated tool for this purpose, the pneu- correlation between pneumonia severity index (PSI) and monia severity index (PSI) (1), is highly complex and can ؍ MR-proANP (rs 0.68, P <0.0001) and between PSI and occasionally underestimate the gravity of CAP, especially ؍ CT-proAVP (rs 0.44, P <0.0001). Median (interquartile in young patients without comorbid illnesses (2, 4), be- range) CT-proAVP and MR-proANP values were 8.2 cause it heavily weights age and comorbidity and does (5.3–16.8) and 73.6 (44.6–144.0) pmol/L, respectively. not directly measure CAP-specific disease severity (3). Nonsurvivors had significantly higher MR-proANP and The use of biomarkers as tools to assess diagnosis, CT-proAVP than survivors (median 259.0 vs 71.8 prognosis, and treatment response in infectious diseases, ,including lower respiratory tract infections and sepsis ,0.03 ؍ and 24.9 vs 8.1 pmol/L, P ,0.01 ؍ pmol/L, P respectively). In multivariate analysis including PSI, constitutes an area of growing interest for investigators (5–9). Biochemical inflammation and infection markers [e.g., procalcitonin (PCT), C-reactive protein (CRP), and lipopolysaccharide-binding protein (LBP)] have been the 1 Infectious Diseases Unit, Internal Medicine Department, Hospital Gen- eral Universitario de Elche, Alicante, Spain. subject of intense investigation, but attention has also 2 Research Department, BRAHMS AG, Hennigsdorf/Berlin, Germany. focused on neurohumoral hemodynamic markers. Pep- 3 Public Health Department, Universidad Miguel Herna´ndez, Elche, Spain. 4 Pneumology Section, Hospital General Universitario de Elche, Alicante, Spain. * Address correspondence to this author at: Unidad de Enfermedades 5 Nonstandard abbreviations: CAP, community-acquired pneumonia; PSI, Infecciosas, Hospital General Universitario de Elche, Camı´ de la Almazara 11, pneumonia severity index; PCT, procalcitonin; CRP, C-reactive protein; LBP, 03203 ELCHE, Alicante, Spain. Fax 00-34-96-667 91 56; e-mail marmasia@ya. lipopolysaccharide-binding protein; AVP, vasopressin; ANP, atrial natriuretic com. peptide; MR-proANP, midregional proANP; ICU, intensive care unit; CT- Received January 12, 2007; accepted September 28, 2007. proAVP, carboxy-terminal proAVP; IQR, interquartile range; NPV, negative Previously published online at DOI: 10.1373/clinchem.2007.085688 predictive value.

2193 2194 Masia´ et al.: Hemodynamic Biomarkers and CAP

tide hormones involved in cardiovascular/osmotic ho- ously described (25, 26). The study was approved by the meostasis, such as members of the natriuretic peptide local ethics committee. All adult patients (Ն15 years) with family and vasopressin (AVP), are molecules within this signs and symptoms compatible with pneumonia occur- class of biomarkers. ring during 2 consecutive periods of 12 months were Natriuretic peptides, such as atrial natriuretic peptide eligible for inclusion in the study. Primary care physicians (ANP), play an important pathophysiological role in were asked to refer to the hospital emergency department cardiovascular diseases (10, 11). Increased concentrations all patients in whom a diagnosis of CAP was suspected. of ANP or ANP prohormone fragment have been re- CAP was defined as an acute illness associated with at ported to indicate cardiovascular dysfunction in septic least 1 of the following signs or symptoms: fever (axillary patients (12–14). In addition, midregional proANP (MR- temperature Ն38 °C), new cough with or without sputum proANP) has been shown to be a valuable tool for risk production, pleuritic chest pain, dyspnea, or altered assessment, with utility similar to that of the Acute breath sound on auscultation, plus a chest radiograph Physiology and Chronic Health Evaluation II score to showing an opacity compatible with the presence of acute predict outcome in septic patients admitted to an inten- pneumonia. Patients were evaluated clinically and by sive care unit (ICU) (14), as well as to categorize patients x-ray at the emergency department, and those with a with CAP from a cohort of patients with lower respiratory provisional diagnosis of CAP were seen by a study tract infection (15). investigator. Patients in whom diagnosis of pneumonia AVP, a hormone released from the posterior pituitary was not finally confirmed, those with a prior hospitaliza- gland, has vasoconstrictor and antidiuretic properties and tion within 2 weeks of a current diagnosis of pneumonia, potency to restore vascular tone in vasodilatory hypoten- and those in whom measurements were not performed sion (16). AVP is derived from a larger precursor because a serum sample was not available or was insuf- (proAVP) along with 2 other peptides of unknown func- ficient were excluded. During the first 12-month study tion, neurophysin II and copeptin, the carboxy-terminal period, from October 15, 1999, to October 14, 2000, pa- part of the precursor (17). AVP contributes to the patho- tients gave informed consent, were enrolled in the inves- genesis of several diseases such as congestive heart failure tigation, and had a blood sample collected within the 1st (18). In recent studies, AVP concentrations have been 24 h after fulfilling the pneumonia criteria for routine shown to be increased in critically ill patients, including blood analysis and measurement of biological markers. those with septic shock (19, 20). Individuals recruited through that time period comprised We hypothesized that patients with CAP who have the cohort included in this study. All patients were poor outcomes might have altered cardiovascular func- followed up for at least 4 weeks or until death. An tion/hemodynamic function biomarkers. In such patients outpatient visit after discharge for collection of follow-up ANP as well as AVP concentrations might be useful for data was automatically scheduled. Patients who did not predicting outcome at the initial assessment. We analyzed turn up for the appointment were interviewed by phone data from a population-based study in which patients about clinical outcome. To calculate the severity of pneu- were prospectively evaluated and an extensive microbio- monia, we used the PSI scoring system, which classifies logical investigation was carried out. Because of the short patients according to age, sex, comorbidity, and clinical half-life of ANP and AVP, as well as instability and and laboratory data. Laboratory work-up for a patient platelet binding of AVP (21–23), precursor fragments of with CAP as well as criteria for etiological diagnosis has both hormones [MR-proANP and the carboxy-terminal been described in detail (25). proAVP (CT-proAVP; copeptin)] were analyzed as alter- Because MR-proANP concentrations are age depen- native diagnostic targets. MR-proANP and CT-proAVP dent (24) and CT-proAVP concentrations have been re- concentrations directly reflect the release of their rapidly ported to be higher in males than females (20),we degraded active hormones ANP and AVP (20, 24). included data from an age- and sex-matched healthy Other potential predictors of mortality were also ex- reference group (n ϭ 37; mean age 58.2 years; age range plored, including age, sex, comorbidity (diabetes, chronic 19–91). We collected serum samples from the members of obstructive pulmonary disease, congestive heart failure, a local health club and employees of a local biotechnology hepatic disease, chronic renal insufficiency, neoplasia, center and their relatives/friends as described in detail immunodepression, altered mental status, and malnutri- (20). Written informed consent was obtained from all tion), arterial blood pressure (measured by aneroid healthy volunteers. There was no difference in age (P ϭ sphygmomanometer), laboratory data (including PCT, 0.76) and sex (P ϭ 0.82) between individuals of the CRP, and LBP concentrations), and etiology of CAP. healthy reference group and CAP patients. For measurement of MR-proANP and CT-proAVP, Materials and Methods serum was separated from blood samples at the time of A prospective population-based investigation of CAP was blood draw and frozen at –80 °C until analysis. Measure- conducted at Hospital General Universitario de Elche, a ments were performed in a blinded fashion as a batch 430-bed university-affiliated teaching hospital covering a analysis by 1 trained laboratory assistant. MR-proANP population of about 275 000 people in Spain, as previ- measurements were performed using an immunolumino- Clinical Chemistry 53, No. 12, 2007 2195

metric sandwich assay (MR-proANP LIA; Brahms AG) as Results described (24), but we changed the calibration of the In experiments comparing MR-proANP concentrations in reported assay from a synthetic peptide to native pro- 100 matched EDTA-plasma and serum samples from ANP in human serum, which increased precision and healthy volunteers (mean age 42 years, range 18–74 years) dynamics of the assay; details on this modification have there was a significant correlation (P Ͻ0.0001) between ϭ been published (14). The functional assay sensitivity MR-proANP in plasma and serum (rs 0.98; 95% CI (defined as the lowest value with an interassay CV Ͻ20%) 0.96–0.98). Additionally, comparison of the 2 groups by was 11 pmol/L. Median MR-proANP of 325 healthy use of the Wilcoxon matched-pairs test revealed no sig- ϭ individuals in previous investigations was 45 pmol/L nificant difference (P 0.13) between MR-proANP values (interassay CV at 45 pmol/L ϭ 10%), and the 97.5th of the matched EDTA-plasma and serum samples. These percentile (upper limit of reference interval) was 163.9 results indicate equivalence of MR-proANP measurements in plasma and serum. The median, 97.5th percen- pmol/L (interassay CV at 163.9 pmol/L ϭ 7.5%) (24). tile, and 99th percentile MR-proANP concentrations in Because measurements were performed in EDTA-plasma serum were 44.1, 161.0, and 198.0 pmol/L, respectively. and no equivalence data in serum have been shown so far, This population did not serve as a reference group in the we tested 100 matched EDTA-plasma and serum samples present CAP study because of lower age and a lower from healthy volunteers (mean age 42 years, range 18–74 percentage of men compared with the group of CAP years). Samples were collected as described (20). patients. CT-proAVP measurements were performed with a We measured serum concentrations of MR-proANP sandwich immunoluminometric assay (CT-proAVP LIA; and CT-proAVP in 173 (72.1%) of 240 CAP patients from Brahms AG), as described (20). Briefly, we used 2 poly- October 15, 1999, to October 14, 2000. In the remaining clonal antibodies to the carboxy-terminal region (covering patients, measurements were not performed because a amino acids 132–164 of preproAVP). In contrast to mea- serum sample obtained within the first 24 h of diagnosis surements of mature AVP, no extraction step before of pneumonia was not available or was insufficient. There measurement was needed, and the analyte showed ex were no differences in age, sex, comorbidity, or PSI scores vivo stability for at least 7 days at room temperature. The between patients in whom MR-proANP and CT-proAVP functional assay sensitivity was 2.25 pmol/L. Median were measured and those in whom it was not (data not CT-proAVP in 359 healthy individuals in previous inves- shown). Baseline characteristics of the patients are de- tigations was 4.2 pmol/L (interassay CV at 4.2 pmol/L ϭ scribed in Table 1. The mean (SD) age of the 173 patients 14%), and the 97.5th percentile was 11.25 pmol/L (inter- was 59.3 (20.6) years and 64.2% were male. In 85 patients assay CV at 11.25 pmol/L ϭ 9%) (20). MR-proANP and (49.1%) there was 1 or more underlying disease, mostly CT-proAVP are stable at 4 °C for at least 48 h and 14 days, diabetes (n ϭ 37, 21.4%) and chronic obstructive pulmo- respectively. No signs of degradation were observed. We nary disease (n ϭ 36, 20.8%). The etiological distribution is also measured LBP, CRP, and PCT in serum as described shown in Table 2. (6, 26). Compared with healthy individuals (age- and sex- We tested distribution with the Kolmogorov–Smirnov matched healthy reference group), median [interquartile test. Categorical data among survivors and nonsurvivors range (IQR)] values of MR-proANP [73.6 (44.6–144.0) vs ϭ were compared with the ␹2 test, and the Mann–Whitney 50.1 (32.1–112.5) pmol/L, P 0.03] and CT-proAVP [8.2 Ͻ U-test was used for comparison of continuous variables in (5.3–16.8) vs 4.3 (2.7–6.6) pmol/L, P 0.0001] were significantly increased in CAP patients, although there was a 2 groups because they were not gaussian distributed. For high degree of overlap between the groups. Patients with multigroup comparisons, we used the Kruskal–Wallis test pneumonia caused by atypical organisms had lower MR- followed by the Dunn posttest. The relative contributions proANP concentrations than any other etiologic group to mortality of the hemodynamic biomarkers PCT, CRP, (P ϭ 0.04), than patients with bacterial pneumonia (P ϭ LBP, and PSI were estimated using a multivariable logis- 0.02), pneumonia of unknown etiology (P ϭ 0.005), and tic regression model. We used a forward-selection proce- mixed pneumonia (P ϭ 0.03) (Table 2). CT-proAVP con- dure: a variable was retained in the model if the LR test P centrations were significantly lower in patients with atyp- Ͻ value was 0.05. CIs (95% CI) for the adjusted odds ratios ical pneumonia than in patients whose pneumonia was were calculated. Continuous variables were investigated bacterial (P ϭ 0.04) or of unknown origin (P ϭ 0.01). for nonlinearity and fitted as having a linear relationship However, for both serum markers, there was again a high with the log-odds ratio of death unless there was clear overlap between etiological groups. evidence of nonlinearity. The median (range) PSI score, which indicates pneu- We calculated the correlation between continuous vari- monia severity, was 70.0 (9–177). There was a positive ables using the Spearman correlation coefficient and used correlation between PSI and MR-proANP concentrations ϭ Ͻ ROC curves to identify the best cutoff concentrations to (rs 0.68, P 0.0001) and between PSI and CT-proAVP ϭ Ͻ predict mortality risk. Statistical analysis was performed concentrations (rs 0.44, P 0.0001). Concentrations of by use of SPSS version 11. MR-proANP and CT-proAVP according to PSI risk class 2196 Masia´ et al.: Hemodynamic Biomarkers and CAP

ϭ Table 1. Baseline characteristics of the patients with CAP. (44.0–129.5) pmol/L, P 0.01] (Fig. 1C). CT-proAVP concentrations were also significantly higher in patients Characteristic who died [24.9 (8.2–105.0) vs 8.1 (5.3–16.2) pmol/L, P ϭ Sex, no. male (%) 111 (64.2) 0.03] (Fig. 1D). In univariate analysis, patients who died Mean age, years (range) 59.3 (15–93) had significantly higher age (P ϭ 0.003), PCT (P ϭ 0.002), Comorbidity, n (%) 85 (49.1) ϭ ϭ ϭ Diabetes 37 (21.4) LBP (P 0.004), CRP (P 0.006), prothrombin time (P Chronic obstructive pulmonary 36 (20.8) 0.03), associated comorbidity (diabetes, chronic obstruc- disorder tive pulmonary disease, congestive cardiac failure, he- Altered mental status 19 (11) patic disease, chronic renal insufficiency, neoplasia, im- Coronary heart disease 11 (6.4) munodepression, altered mental status, or malnutrition) Neoplasia 8 (4.6) (P ϭ 0.006), and PSI score (P ϭ 0.002) and lower systolic Chronic renal insufficiency 6 (3.5) arterial blood pressure (P ϭ 0.005). In multivariate anal- Hepatic disease 5 (2.9) ysis including PSI, PCT, CRP, LBP, and CT-proAVP and Congestive heart failure 3 (1.7) MR-proANP concentrations, the only variable that re- Mitral valve disease 2 (1.2) mained an independent predictor of death was CT- a Other 11 (6.4) proAVP (odds ratio 1.05, 95% CI 1.01–1.09, P ϭ 0.007). More than 1 comorbidity 51 (29.5) Hosmer-Lemeshow c statistic for the model was 5.63 (P ϭ Current smoker, n (%) 50 (28.9) 0.69). When age was included in multivariate analysis, the Previous antibiotic treatment, n 44 (25.4) (%) same results were obtained (data not shown). Physical and laboratory findings, Median values of MR-proANP and CT-proAVP ac- median (IQR) cording to different dichotomic variables were compared Mean arterial blood pressure, 90 (83.3–100) and are summarized in Table 3. There was no relationship mmHg between any of the biomarkers and blood pressure. Systolic arterial blood 130 (120–145) A cutoff point for MR-proANP of Ͼ227 pmol/L pre- pressure, mmHg dicted mortality with a sensitivity of 71.4% (95% CI Heart rate, per minute 80 (76–88) 29.3%–95.5%), specificity of 91% (85.5%–94.9%), positive Tachypnea Ͼ20 respirations 58 (33.5) per minute, n (%) predictive value of 25%, and negative predictive value Leukocyte count, ϫ109/L 13 000 (8750–17 100) (NPV) of 98.7%. Area under the curve was 0.78 (0.71–0.84) Ͼ Serum osmolarity, mmol/Kg 290.5 (285.8–296.4) (Fig. 2). A cutoff concentration for CT-proAVP of 18.9 MR-proANP, pmol/L 73.6 (44.6–144) pmol/L predicted mortality with a sensitivity of 71.4% CT-proAVP, pmol/L 8.2 (5.3–16.8) (29.3%–95.5%), specificity of 79.5% (72.6%–85.4%), posi- PCT, ␮g/L (n ϭ 160) 0.10 (0.10–0.14) tive predictive value of 12.8%, and NPV of 98.5%. The CRP, mg/L (n ϭ 169) 10.9 (3.7–61.7) NPV was 100% for patients with PSI risk classes I–III. LBP, ␮g/mL (n ϭ 171) 13.9 (8.9–28) Area under the curve was 0.75 (0.68–0.81). Although the Thoracic chest x-ray difference did not reach statistical significance, when all characteristics, n (%) the variables (the combination of MR-proANP, CT- Unilateral alveolar infiltrate 148 (85.5) proAVP, and PSI) were analyzed together as predictors of Bilateral alveolar infiltrate 19 (11) mortality, the area under the curve was higher [0.83 Interstitial infiltrate 6 (3.5) (0.76–0.88)] than that of any single biomarker or PSI score Hospital admission, n (%) 137 (79.1) [0.79 (0.72–0.84) for PSI score] separately. ICU admission, n (%) 2 (1.2) PSI points, median (IQR) 70 (45.5, 101.5) Discussion Risk classes I–II, n (%) 89 (51.4) Risk class III, n (%) 31 (17.9) Despite their attractiveness for clinicians, very few bio- Risk classes IV–V, n (%) 53 (30.6) chemical markers add significant information to PSI about Death, n (%) 7 (4.0) prognosis of CAP. Currently, the most useful biomarker in the prediction of outcome of CAP is serum procalcito- a Immunodepression, malnutrition. nin (5, 8, 27). In a previous study, we found that procalcitonin was mainly a marker of poorer outcome in pa- are shown in Fig. 1, A and B. MR-proANP and CT- tients with CAP classified into PSI high-severity risk proAVP values exhibited a gradual increase from risk classes (5). In a preliminary study including 58 patients class I (the lowest risk) to risk class V (highest risk) (P with severe CAP according to PSI score, B-type natriuretic Ͻ0.0001 for both cases). A positive correlation was also peptide was helpful for the risk stratification of patients ϭ Ͻ found for age and both MR-proANP (rs 0.74, P 0.0001) compared with PSI (28). Christ-Crain et al. (29) have ϭ Ͻ and CT-proAVP (rs 0.38, P 0.0001). recently found that proadrenomedullin, a peptide with Seven (4%) of the 173 patients died. Nonsurvivors had immune modulating, metabolic, and vascular actions, significantly higher MR-proANP concentrations than pa- predicted severity and outcome of patients with CAP tients who survived [median (IQR) 259.0 (122–281) vs 71.8 with prognostic accuracy similar to that of the PSI, in Clinical Chemistry 53, No. 12, 2007 2197

Table 2. Distribution of the causative microorganisms in patients admitted to hospital and in outpatients. Patients admitted Median MR-proANP Median CT-proAVP Etiology n (%) to hospital, n (%) Outpatients, n (%) (IQR) (IQR) Classic bacterial pathogens 38 (22.0) 34 (24.8) 4 (11.1) 74.5 (46.3–132.5)a 10.8 (6.0–15.5)b Streptococcus pneumoniae 26 (15.0) 23 (16.8) 3 (8.3) Hemophilus influenzae 6 (3.2) 5 (3.6) 1 (2.8) Pseudomonas aeruginosa 3 (1.7) 3 (2.2) 0 Otherc 3 (1.7) 3 (2.2) 0 Atypical pathogens 35 (20.2) 25 (18.2) 10 (27.8) 56.0 (33.0–77.8)d 6.9 (4.7–10.5) Mycoplasma pneumoniae 15 (8.7) 10 (7.3) 5 (13.9) Legionella pneumophila 11 (6.4) 10 (7.3) 1 (2.8) Chlamydophila spp. 8 (4.6) 5 (3.6) 3 (8.3) Coxiella burnetii 1 (0.6) 0 1 (2.8) Viruses 11 (6.4) 6 (4.4) 5 (13.9) 117.0 (46.8–180.0) 8.0 (3.8–17.1) Influenza/parainfluenza virus 7 (4.0) 4 (2.9) 3 (8.3) Othere 4 (2.3) 3 (2.2) 1 (2.8) Mixed 13 (7.5) 9 (6.6) 4 (11.1) 99.1 (49.5–214.5)f 11.4 (4.9–27.7) Bacterial ϩ atypical 5 (2.9) 3 (2.2) 2 (5.6) Atypical ϩ viral 2 (1.2) 2 (1.5) 0 Two atypical microorganisms 3 (1.7) 1 (0.7) 2 (5.6) Two viruses 1 (0.6) 1 (0.7) 0 Two bacteria 1 (0.6) 1 (0.7) 0 Bacteria ϩ virus 1 (0.6) 1 (0.7) 0 Unknowng 76 (43.9) 63 (46.0) 13 (36.1) 80.6 (46.8–191.5)h 9.9 (6.2–20.1)i Total 173 137 36 73.6 (44.6–144) 8.2 (5.3–16.8) a P ϭ 0.02 for the comparison with atypical pathogens. b P ϭ 0.04 for the comparison with atypical pathogens. c M. catarralis (1 case), Citrobacter spp (1 case), S. aureus (1 case). d P ϭ 0.04 for the comparison with any other etiologic group. e Respiratory syncytial virus (3 cases), adenovirus (1 case). f P ϭ 0.03 for the comparison with atypical pathogens. g In all, the clinical diagnosis of pneumonia was confirmed on follow-up evaluations. h P ϭ 0.005 for the comparison with atypical pathogens. i P ϭ 0.01 for the comparison with atypical pathogens. contrast to other biomarkers such as CRP and leukocyte Other factors might implicate a higher severity of count. disease in CAP, with high PSI score, leukocyte cell counts, According to our results, MR-proANP and CT-proAVP bilateral pulmonary involvement, or age (above 65 years) could be helpful as prognostic tools. Both biomarkers being associated with significantly higher concentrations correlated with PSI, and significantly higher concentra- of MR-proANP and CT-proAVP. We found a correlation tions were found in patients who died compared with between age and MR pro-ANP and CT pro-AVP concen- survivors. In multivariate analysis of predictors of death, trations (data not shown). A trend toward a correlation including procalcitonin concentrations and PSI as risk between AVP and age has been reported (18). Other factors, CT-proAVP turned out to be the only biomarker researchers have described a modest influence of age on that remained as an independent predictor. Although this natriuretic peptides (30), as well as documented hormone finding must be interpreted with caution because of the deficiency with advanced age (31). Despite this associa- small number of patients who died, this marker might tion, when age was included in multivariate analysis, the add complementary information about severity of disease same results were obtained (data not shown), a finding to established prognostic scores or markers such as PSI or that may indicate that at least CT-proAVP is independent procalcitonin, opening a way for future investigation of the role of hemodynamic factors in the prognosis of CAP. from age and thus could be a valuable prognostic factor in Because MR-proANP and CT-proAVP have a very high younger people with CAP. However, no apparent rela- NPV, especially among PSI risk classes I–III, they could be tionship between CT-proAVP or MR-proANP concentra- useful to more accurately identify patients at low risk for tions and arterial blood pressure was found. This discrep- death, who do not need admission to hospital. This ancy may be related to earlier hemodynamic dysfunction information may help physicians to make more rational or vascular tone/myocardial function changes not re- decisions about hospitalization for some patients with flected in arterial blood pressure. We did find a relation- CAP. ship between the biomarkers and other closely related 2198 Masia´ et al.: Hemodynamic Biomarkers and CAP

Fig. 1. Box plots showing MR-proANP (A) and CT-proAVP (B) in CAP patients according to PSI risk class and MR-proANP (C) and CT-proAVP (D) in CAP patients according to survival. The box plots represent 25th, 50th, and 75th percentiles; whiskers, minimum and maximum values. The P value (Kruskal–Wallis test; top of A and B), significant differences between individual PSI classes (Dunn multiple posttest; values above the solid lines), and number of patients (percentage in parenthesis) in each group are shown. The dotted lines indicate the median MR-proANP concentration in the age- and sex-matched healthy reference group. factors such as serum osmolarity, but not with plasma between the different etiologic groups, limiting the useful- sodium. ness of these markers for suggesting etiology. Despite their prognostic value, MR-proANP and CT- The main limitation of our current study was the small proAVP had little utility for predicting etiology in CAP. number of patients who had a poor outcome. Because this Although patients with atypical pneumonia showed signif- was a population-based study, overall prognosis of the icantly lower concentrations of both markers than patients patients was good, and the number of those who died or with pneumonia of other etiologies, there was much overlap developed complications was low. The associations found Clinical Chemistry 53, No. 12, 2007 2199

Table 3. Univariate analysis of MR-proANP and CT-proAVP concentrations. Variable n Median MR-proANP, pmol/L (range) Pa Median CT-proAVP, pmol/L (range) Pa Level of care Admitted to hospital 137 87.3 (19.0–1130) Ͻ0.0001b 11.4 (0.5–113) Ͻ0.0001b Admitted to ICU 2 132.1 39.4 Outpatients 36 40.5 (17.0–288.0) 6.3 (2.0–22.4) Age Ͼ65 years 84 124 (45.9–1130) Ͻ0.0001 13.8 (2.9–113) Ͻ0.0001 Յ65 years 89 46.2 (17–486) 7.4 (0.49–57.5) Comorbidityc Yes 85 105 (26–1130) Ͻ0.0001 11.7 (2.7–113) 0.003 No 88 55.4 (17–842) 7.5 (0.49–45.9) Systolic arterial blood pressure Ͻ100 mmHg 8 67.2 (33–249) 0.7 6.3 (2.8–105) 0.4 Ն100 mmHg 165 73.6 (17–1130) 8.2 (0.49–113) Leukocyte count Ͼ15 000 or Ͻ2000 ϫ 63 83.8 (26–1130) 0.04 11.6 (0.49–105) 0.06 109/L Յ15 000 or Ն2000 ϫ 110 70.4 (17–860) 7.9 (2–113) 109/L Serum osmolarity Ͻ0.0001 0.001 Ͻ295 mmol/Kg 123 59.7 (17–842) 7.6 (0.49–105) Ն295 mmol/Kg 49 129 (21.5–1130) 13.9 (2.7–113) Serum sodium 0.7 0.8 Ն135 mEq/L 144 75.7 (17–1130) 8.2 (0.5–113) Ͻ135 mEq/L 29 70.6 (19–380) 9.1 (2.7–70.1) Thoracic x-ray results Unilateral alveolar infiltrate 148 70.6 (17–1130) 0.02 8 (2–113) 0.03 Bilateral alveolar infiltrates 19 120 (43.4–842) 17.1 (2.7–57.5) Interstitial infiltrates 6 91.3 (27.2–514) 10 (0.49–16.4) Bacteremia Yes 5 171 (26.7–217) 0.2 11.3 (3.6–32.9) 0.7 No 168 72.2 (17–1130) 8.2 (0.49–113) Development of complicationsd Yes 3 171.0 (93–259) 0.1 32.9 (24.9–45.9) 0.01 No 170 72.6 (17–1130) 8.1 (0.9–113) Death Yes 7 259 (27.5–860) 0.01 24.9 (3.6–113) 0.02 No 166 71.8 (17–1130) 8.1 (0.49–70.1) Procalcitonin levels Ն0.5 ␮g/L 15 171 (81.1–860) Ͻ0.0001 32.9 (3.7–113) Ͻ0.0001 Ͻ0.5 ␮g/L 145 68.6 (17–1130) 7.8 (0.49–70.1) a Comparison of 2 and 3 groups (Mann–Whitney U-test and Kruskal–Wallis test), respectively. b Comparison of biomarker levels between hospitalized patients (including ICU patients) and outpatients. c Diabetes, chronic obstructive pulmonary disease, congestive cardiac failure, hepatic disease, chronic renal insufficiency, neoplasia, immunodepression, altered mental status, and malnutrition. d Admission to the ICU, septic shock, empyema, mechanical ventilation requirement. were very strong, however, and goodness of fit of the disease, to the high MR-proANP concentrations of the multivariate model indicates that it was well calibrated patients. The most important factor affecting ANP secre- and fitted the data well. Furthermore, a close relationship tion from the atria is mechanical stretch, such as in heart of these biomarkers with most of the established prognos- failure, but cardiac ischemia also stimulates ANP release tic factors of CAP was also observed, thereby confirming (32). To address this issue, we excluded from data anal- their prognostic value. The low number of deaths, how- ysis patients with congestive heart failure or coronary ever, may limit generalization of the cutoff points identi- heart disease. With those patients excluded, there were no fied. Another limitation of the study might be the contri- changes in the relationships of MR-proANP with PSI bution of the coexisting illnesses, particularly heart score and mortality, and the results of multivariate anal- 2200 Masia´ et al.: Hemodynamic Biomarkers and CAP

4. Angus DC, Marrie TJ, Obrosky S, Clermont G, Dremsizov TT, Coley 100 MR-proANP: 227 pmol/L Sensitivity: 71.4% C, et al. Severe community-acquired pneumonia: use of intensive Specificity: 91.0% care services and evaluation of the American and British Thoracic Society criteria. Am J Respir Crit Care Med 2002;166:717–23. 80 5. Masia M, Gutierrez F, Shum C, Padilla S, Navarro JC, Flores E, et al. Usefulness of procalcitonin levels in community-acquired pneumonia according to the patients outcome research team pneumo- 60 nia severity index. Chest 2005;128:2223–9. CT-proAVP: 18.9 pmol/L MR-proANP Sensitivity: 71.4% CT-proAVP 6. Opal SM, Scannon PJ, Vincent JL, White M, Carroll SF, Palardy JE, Specificity: 79.5% et al. Relationship between plasma levels of lipopolysaccharide Sensitivity 40 (LPS) and LPS-binding protein in patients with severe sepsis and septic shock. J Infect Dis 1999;180:1584–9. 7. Masia M, Gutierrez F, Llorca B, Navarro JC, Mirete C, Padilla S, et 20 al. Serum concentrations of lipopolysaccharide-binding protein as a biochemical marker to differentiate microbial etiology in patients with community-acquired pneumonia. Clin Chem 2004;50:1661–4. 0 8. Christ-Crain M, Stolz D, Bingisser R, Muller C, Miedinger D, Huber 0 20 40 60 80 100 PR, et al. Procalcitonin-guidance of antibiotic therapy in communi- 100-Specificity ty-acquired pneumonia: a randomized trial. Am J Respir Crit Care Fig. 2. ROC plot analysis of the biomarkers with respect to mortality Med 2006;174:84–93. prediction in patients with CAP. 9. Hedlund J, Hansson LO. Procalcitonin and C-reactive protein levels The arrays indicate the cutoff concentrations for MR-proANP (227 pmol/L) and in community-acquired pneumonia: correlation with etiology and CT-proAVP (18.9 pmol/L). prognosis. Infection 2000;28:68–73. 10. Sagnella GA. Measurement and significance of circulating natriuretic peptides in cardiovascular diseases. Clin Sci 1998;95: ysis of predictors of mortality did not change (data not 519–29. shown), suggesting that the prognostic accuracy of MR- 11. Potter LR. Natriuretic peptides, their receptors, and cyclic guanosine proANP in CAP is not due to the presence of coexisting monophosphate-dependent signaling functions. Endocr Rev 2006; heart disease per se. Likewise, the concentrations of AVP 27:47–72. can be increased in many other diseases or situations (33). 12. Hoffmann U, Brueckmann M, Bertsch T, Wiessner M, Liebetrau C, Lang S, et al. Increased plasma levels of NT-proANP and NT- proBNP as markers of cardiac dysfunction in septic patients. Clin In conclusion, MR-proANP and CT-proAVP are predic- Lab 2005;1:373–9. tors of prognosis of CAP that might expand on the 13. Brueckmann M, Huhle G, Lang S, Haase KK, Bertsch T, Weiss C, information of PSI or procalcitonin. These markers may be et al. Prognostic value of plasma N-terminal pro-brain natriuretic useful in the initial management of patients with CAP. peptide in patients with severe sepsis. Circulation 2005;112: Further studies to confirm our results are needed. 527–34. 14. Morgenthaler NG, Struck J, Christ-Crain M, Bergmann A, Muller B. Pro-atrial natriuretic peptide is a prognostic marker in sepsis, Grant/funding support: None declared. similar to the APACHE II score: an observational study. Crit Care 2005;9:R37–45. Financial disclosures: Mar Masia´, Ildefonso Hernańdez, and 15. Muller B, Suess E, Schuetz P, Muller C, Bingisser R, Bergmann A, Fe´lix Gutie´rrez declare that they have no competing interests. et al. Circulating levels of pro-atrial natriuretic peptide in lower Jana Papassotiriou and Nils Morgenthaler are employees of respiratory tract infections. J Intern Med 2006;260:568–76. BRAHMS AG, the manufacturer of the MR-proANP and the 16. Asfar P, Hauser B, Radermacher P, Matejovic M. Catecholamines CT-proAVP assay. and vasopressin during critical illness. Crit Care Clin 2006;22: Acknowledgments: We thank Inmaculada Jarrıń (Public 131–49. Health Department, Universidad Miguel Hernańdez) for her 17. De Bree FM, Burbach JP. Structure-function relationship of the help in data analysis and Carlos Mirete (Hospital General vasopressin prohormone domains. Cell Mol Neurobiol 1998;18: Universitario de Elche) for his help in data collection. We also 173–91. 18. Goldsmith SR, Gheorghiade M. Vasopressin antagonism in heart thank Frank Bonconseil, Angelina Herzberg, and Johanna failure. J Am Coll Cardiol 2002;46:1785–91. Hetzel for excellent technical assistance. 19. Jochberger S, Mayr VD, Luckner G, Wenzel V, Ulmer H, Schmid S, et al. Serum vasopressin concentrations in critically ill patients. References Crit Care Med 2006;34:293–9. 1. Fine MJ, Auble TE, Yealy DM, Hanusa BH, Weissfeld LA, Singer DE, 20. Morgenthaler NG, Struck J, Alonso C, Bergmann A. Assay for the et al. A prediction rule to identify low-risk patients with community- measurement of copeptin, a stable peptide derived from the acquired pneumonia. N Engl J Med 1997;336:243–50. precursor of vasopressin. Clin Chem 2006;52:112–9. 2. Ewig S, De Roux A, Bauer T, Garcia E, Mensa J, Niederman M, et 21. Buckley MG, Marcus NJ, Yacoub MH. Cardiac peptide stability, al. Validation of predictive rules and indices of severity for aprotinin and room temperature: importance for assessing car- community acquired pneumonia. Thorax 2004;59:421–7. diac function in clinical practice. Clin Sci 1999;97:689–95. 3. Niederman MS, Feldman C, Richards GA. Combining information 22. Preibisz JJ, Sealey JE, Laragh JH, Cody RJ, Weksler BB. Plasma from prognostic scoring tools for CAP: an American view on how to and platelet vasopressin in essential hypertension and congestive get the best of all worlds. Eur Respir J 2006;27:9–11. heart failure. Hypertension 1983;5:I129–38. Clinical Chemistry 53, No. 12, 2007 2201

23. Kluge M, Riedl S, Erhart-Hofmann B, Hartmann J, Waldhauser F. evaluation in patients with severe pneumonia. Clin Microbiol Improved extraction procedure and RIA for determination of argi- Infect 2002;8:93–100. nine8-vasopressin in plasma: role of premeasurement sample 28. Mueller C, Laule-Kilian K, Scholer A, Perruchoud AP. B-type treatment and reference values in children. Clin Chem 1999;45: natriuretic peptide for risk stratification in community-acquired 98–103. pneumonia. J Intern Med 2005;258:391–3. 24. Morgenthaler NG, Struck J, Thomas B, Bergmann A. Immunolumi- 29. Christ-Crain M, Morgenthaler NG, Stolz D, Muller C, Bingisser R, nometric assay for the midregion of pro-atrial natriuretic peptide in Harbarth S, et al. Pro-adrenomedullin to predict severity and human plasma. Clin Chem 2004;50:234–6. outcome in community-acquired pneumonia [ISRCTN04176397]. 25. Gutierrez F, Masia´ M, Rodriguez JC, Ayelo A, Soldan B, Cebrian L, Crit Care 2006;10:R96. et al. Evaluation of the immunochromatographic Binax NOW assay 30. Hogenhuis J, Voors AA, Jaarsma T, Hillege HL, Boomsma F, van for detection of Streptococcus pneumoniae urinary antigen in a Veldhuisen DJ. Influence of age on natriuretic peptides in patients prospective study of community-acquired pneumonia in Spain. with chronic heart failure: a comparison between ANP/NT-ANP and Clin Infect Dis 2003;36:286–92. BNP/NT-proBNP. Eur J Heart Fail 2005;7:81–6. 26. Gutie´rrez F, Masia´ M, Rodrı´guez JC, Mirete C, Soldan B, Padilla S, 31. Lamberts SW, van den Beld AW, van der Lely AJ. The endocrinol- et al. Epidemiology of community-acquired pneumonia in adult ogy of aging. Science 1997;278:419–24. patients at dawn of the twenty-first century: a prospective study at 32. Dietz J. Mechanisms of atrial natriuretic peptide secretion from the Mediterranean coast of Spain. Clin Microbiol Infect 2005;11: the atrium. Cardiovasc Res 2005;68:8–17. 788–800. 33. Holmes CL, Patel BM, Russell JA, Walley KR. Physiology of 27. Brunkhorst FM, Al-Nawas B, Krummenauer F, Forycki ZF, Shah PM. vasopressin relevant to management of septic shock. Chest Procalcitonin, C-reactive protein and APACHE II score for risk 2001;120:989–1002. Technical Briefs

Effects of 7 Hemoglobin Variants on the Measurement thus far) have been recognized in single reports of studies of Glycohemoglobin by 14 Analytical Methods, Seung- using 1 or 2 HbA1c determination methods. Nevertheless, Tae Lee,1 Cas W. Weykamp,2 Yong-Wha Lee,3 Jong-Won Kim, 1 studies examining the effects of these variants on various and Chang-Seok Ki1* (1 Department of Laboratory Medicine methods have been restricted to a few HbVAR such as HbS and Genetics, Samsung Medical Center, Sungkyunkwan and HbC, which are common in some regional popula- University School of Medicine, Seoul, Korea; 2 Department tions (2–5). of Clinical Chemistry, Queen Beatrix Hospital, Winterswijk, We previously reported that the HbVAR and high HbF The Netherlands; 3 Department of Laboratory Medicine and in the Korean population are not uncommon, with Hb Genetics, Soonchunhyang University Bucheon Hospital and G-Coushatta and Hb Queens being the 2 main HbVAR Soonchunhyang University College of Medicine, Gyeonggi- found in Koreans (6). Hb G-Coushatta is also one of the do, Korea; * address correspondence to this author at: De- most common HbVAR found in families from the Silk partment of Laboratory Medicine and Genetics, Samsung Road region of China and some North American Indian Medical Center, Sungkyunkwan University School of Med- tribes, as well as in Koreans, Japanese, Thai, Turks, icine, 50 Irwon-Dong, Gangnam-Gu, Seoul 135-710 Korea; Algerians, and Egyptians (6–10). Hb Queens is the most fax 82-2-3410-2719, e-mail [email protected]) common ␣-chain variant in people from the Silk Road region of China, and has also been found in Korean, Background: Hemoglobin variants (HbVAR) are not Japanese, Vietnamese, and Thai families (6, 7, 10). How- uncommon in the Korean population, with Hb ever, there are limited data on the effect of these HbVAR on G-Coushatta and Hb Queens being the 2 most common the different methods for analyzing the HbA1c. Therefore, HbVAR. Hb G-Coushatta is also the most common we examined the possible effects on GHb results for Hb VAR HbVAR in Chinese people from the Silk Road region, as G-Coushatta and Hb Queens and 5 rare Hb analyzed well as in some North American Indian tribes. However, with 14 different methods commonly used to analyze data are scarce on the effect of these HbVAR on the HbA1c. The Institutional Review Board of Samsung Med- ical Center, Seoul, Korea, approved this study. different methods used for analyzing HbA1c. Methods: Specimens from 24 individuals with 7 HbVAR Fresh whole blood specimens were obtained from 24 VAR ␤ (Hb G-Coushatta, Hb Queens, Hb G-Hsi-Tsou, Hb patients with 7 different Hb ; 7 with Hb G-Coushatta [ 22(B4) GluϾAla], 11 with Hb Queens [␣ 34(B15) Ube-4, Hb G-Waimanalo, Hb Inglewood, and Hb LeuϾArg], 2 with Hb G-Hsi-Tsou [␤ 79(EF3) AspϾGly], 1 Bologna-St.Orsola) were collected and tested using the each with Hb Ube-4 [␣ 116(GH4) GluϾAla], Hb G- International Federation of Clinical Chemistry primary Waimanalo [␣ 64(E13) AspϾAsn], Hb Inglewood [␤ reference method as well as 14 routine HbA assay 1c 142(H20) AlaϾThr], and Hb Bologna-St.Orsola [␤ methods. 146(HC3) HisϾTyr]. These variants were identified by Results: Hb G-Coushatta showed a clinically significant direct sequencing of the hemoglobin, alpha 1 (HBA1); effect on the measured HbA1c, particularly when analy- hemoglobin, alpha 2 (HBA2); and hemoglobin, beta (HBB) sis was performed with ion-exchange HPLC methods genes, using previously reported protocols (6). The sam- with short elution times. This interference could be ples were stored below Ϫ70 °C when the assay methods resolved by measuring the HbA1c using other methods were unavailable within 7 days of collection. such as HPLC with a long elution time, immunoassay, We analyzed all the collected specimens using the IFCC boronate affinity chromatography, and enzymatic assay. primary reference method (PRM) with HPLC-capillary Hb Queens showed a clinically significant difference, electrophoresis according to approved IFCC protocols defined as a >10% deviation from regression lines, in (11). This method was chosen as a comparative method results from the 2 HPLC methods but not in the other owing to its excellent precision and accuracy and minimal methods. The remaining 5 rare HbVAR showed different interference with common HbVAR, such as HbS and HbC

HbA1c results in the different assays. (11, 12). Furthermore, the values of IFCC PRM accorded Conclusion: Hb G-Coushatta, Hb Queens, and other very well with those of the boronate affinity method, VAR rare HbVAR can interfere with glycohemoglobin assays, which has been regarded as unaffected by most Hb including ion-exchange HPLC methods with short elu- (1). To compare the values with other methods, the results tion times, but the interference can be resolved using of the IFCC PRM were converted into a National GHb other unaffected methods. It is important to identify Standardization Program (NGSP) value using the follow- these HbVAR through a careful inspection of the chro- ing master equation suggested by Hoelzel et al. (12): matograms and apply other noninterfering methods for Ϫ ϭ Ϫ ϩ NGSP HbA1c 0.915 (IFCC HbA1c) 2.15%. accurate measurements of the HbA1c. © 2007 American Association for Clinical Chemistry We also analyzed the specimens with IFCC secondary reference methods (SRM), including Primus Ultra2 A1c The presence of variant hemoglobins (HbVAR) can inter- and Hemoglobin Variants Analyzer (Ultra2), Roche Uni- fere with some assays used for measuring glycohemoglo- mate HbA1c test reagents on the Modular system (Uni- bin (GHb) (1). Most of these HbVAR (as many as 800 types mate), Tosoh G7 in variant mode (G7-VM), and an Arkray

2202 Clinical Chemistry 53, No. 12, 2007 Clinical Chemistry 53, No. 12, 2007 2203

HA-8160 HbA1c analyzer in diabetic (HA8160-DM) and approximately 1.84 min). On the VII-T instrument, the thalassemia mode (HA8160-TM). Other commercially variant peak might be mistaken for a normal P3 peak, available assays were also used to examine the GHb which is sometimes observed, but the variant peak had a concentrations, including Primus PDQ (PDQ), Bio-Rad slightly shorter retention time (approximately 0.72 min) Variant II System (VII), Bio-Rad Variant II Turbo System than the P3 peak (approximately 0.77 min; see Table 1 in (VII-T), Bio-Rad Variant II System in thalassemia mode the Data Supplement that accompanies the online version (VII-TM), Tosoh G7 in standard mode (G7-SM), Roche of this Technical Brief at http://www.clinchem.org/ TinaQuant II reagent on the Modular system (TinaQuant- content/vol53/issue12). Instruments with long elution II), Olympus AU640 System (AU640), Siemens DCA 2000 times, including VII-TM, G7-VM, and HA8160-TM sepa-

Analyzer (DCA2000), and Daiichi Norudia HbA1c test rated Hb G-Coushatta from the other normal peaks. reagent on the Roche Modular system (Norudia). Inter- From the Deming regression analysis, Hb G-Coushatta method calibration differences were corrected as previ- was predicted to produce clinically significant negative ously reported (5). biases in most ion-exchange HPLC methods at both 6%

We used EP Evaluator Release 7 software (RHOADS) and 7% HbA1c, with the exception of VII-TM, which to perform Deming regression analysis on the 2 common showed acceptable results (Table 1 and see Fig. 1 in the HbVAR (Hb G-Coushatta and Hb Queens) (13). The devi- online Data Supplement). This underestimation of the ations at 6% and 7% HbA1c concentrations, at which the HbA1c concentrations appears to be due to the coelution examined concentrations of the study samples were dis- of the nonglycated HbVAR with the normal A0 peak, tributed, were predicted using the regression lines, and a and the separate elution of the glycated HbVAR from the deviation Ͼ10% of the assessing points (Ϯ0.6% and normal A1c peak. G7-VM, which separated both the Ϯ0.7%, respectively) was defined as a clinically significant glycated and nonglycated Hb G-Coushatta, did not difference (14). automatically subtract the variant portion when calcu-

The chromatogram results obtained with the 7 differ- lating the HbA1c. When a manual calculation was ent HPLC methods for the 2 common HbVAR are shown in performed by subtracting the proportion of the variant Fig. 1. Hb G-Coushatta was unrecognizable in VII and peak areas in the denominator according to the manu- HA8160-DM, possibly owing to coelution with the normal facturer’s instructions, the differences fell into an ac- peaks, whereas the glycated form of this HbVAR was ceptable mean difference (0.3%), but there were wide eluted separately as a small peak between the S-A1c and variations (from Ϫ1.1% to 1.3%) in each sample (see A0 peaks in VII-T and G7-SM. Although no abnormal Table 2 in the online Data Supplement), possibly attrib- peaks occurred on the VII instrument, we noted blunting utable to the close proximity of the retention times of of the normal P3 peak and a subtle shift in the retention HbVAR and HbA (mean 1.15 and 1.09 min, respectively), time of the A0 peak (from approximately 1.78 min to leading to inconsistent separation. Four immunoassays,

VII VII-T VII-TM G7-SM G7-VM HA8160-DM HA8160-TM Normal G-Coushatta Queens

Fig. 1. Chromatograms of the 2 common HbVAR in the 7 different HPLC methods. The solid arrowheads indicate the HbVAR, and the open arrowheads indicate the glycated HbVAR. Hb G-Coushatta was unrecognizable in VII and HA8160-DM and was recognized as only a small glycated peak between the S-A1c and A0 in VII-T and G7-SM, whereas the nonglycated form was separated from the other normal peaks in VII-TM, G7-VM, and HA8160-TM. The nonglycated form of Hb Queens was eluted discretely in all instruments except G7-SM. 2204 Technical Briefs

the 2 boronate-affinity methods and 1 new enzymatic variable degrees of deviation from the IFCC PRM results, assay, were predicted to have good performance. but the number of samples was not enough to draw A few case studies of patients with Hb G-Coushatta conclusions about possible interference (see Table 2 in the have obtained results in accordance with the present data. online Data Supplement). One study of a Japanese Hb G-Coushatta family found Hb Bologna-St.Orsola, which is a rare HbVAR with that unexpectedly low HbA1c values were obtained with increased oxygen affinity and compensatory erythrocy- an ordinary HPLC method (method unknown) compared tosis, showed significant differences in all ion-exchange with the values from a latex agglutination immunoassay HPLC methods, even when it was separated from the and high-resolution weak cation-exchange column chro- other normal peaks (as in VII-TM, G7-VM, HA8160-DM matography (15). Another study found 1 case of a lower and HA8160-TM). It is possible that the glycation rate HbA1c concentration result obtained with the Daiichi or erythrocyte lifespan of this HbVAR might be different Hi-AUTO A1c HA-8150 HbA1c analyzer (2.7%) than with from the normal Hbs. Theoretically, even methods such the DCA2000 (4.7%) and electrospray mass spectrometry as an immunoassay, boronate-affinity, and the IFCC (5.0%) (16). PRM, which provide analytically valid GHb results, Hb Queens is another common HbVAR in the Korean may not provide an accurate assessment of the mean population. Hb Queens was eluted discretely on the 3 plasma glucose with rare HbVAR such as Hb Himeji, Bio-Rad instruments (VII, VII-T, and VII-TM) in the because of the differences in glycation rates or erythro- S-window. This HbVAR was also separated using G7-VM, cyte life spans (16, 17). Interestingly, an abnormal chro- HA8160-DM, and HA8160-TM, but was indistinguishable matogram pattern in this patient suggested the pres- in G7-SM (Fig. 1). Hb Queens was expected to produce no ence of HbVAR and helped identify the cause of the clinically significant effects with most test methods except for G7-SM and VII, for which regression lines predicted a unexplained erythrocytosis through a subsequent ge- considerable underestimation and slight overestimation, netic study. Comparison studies showed that the common HbVAR respectively (Table 1). VAR Chromatograms of the 5 rare HbVAR are shown in Fig. as well as some rare Hb in Koreans can interfere 2 of the online Data Supplement. Hb G-Hsi-Tsou was with the different HbA1c methods, particularly those separated in all instruments. Hb Ube-4 could not be performed on HPLC instruments with a short elution differentiated in either of the 2 modes of the G7 instru- time. This interference can be resolved by measurement ment. Hb G-Waimanalo was separated with all methods with unaffected methods such as HPLC with a long except for G7-SM. Hb Inglewood was recognized only in elution time, immunoassay, boronate affinity chroma- VII and VII-TM (VII showed only a small peak between tography, and enzymatic assay. The chromatograms of A1c and A0). Hb Bologna-St.Orsola had a retention time the ion-exchange HPLC methods also showed variable similar to the A1c peak, a result that led to a gross patterns, a finding that highlights the need for a careful overestimation of the HbA1c concentrations in VII and inspection of the chromatograms, particularly for anal- VII-T, whereas only a slight increase of the L-A1c peak ysis of samples suspected of having an HbVAR or when was observed in G7-SM. The 5 rare HbVAR showed an unexpected result is obtained. Overall, it is impor-

Table 1. Mean differences between the commercial methods and the IFCC PRM for samples containing HbVAR. (11 ؍ Hb Queens (n (7 ؍ Hb G-Coushatta (n

Principle Method 6% 7% 6% 7% Boronate affinity Ultra2 0.14 Ϫ0.08 Ϫ0.02 Ϫ0.05 PDQb 0.09 Ϫ0.09 Ϫ0.08 Ϫ0.03 Ion-exchange HPLC VII Ϫ1.28a Ϫ1.79a 0.83a 0.85a VIIT Ϫ1.49a Ϫ1.99a 0.15 0.29 VII--TM 0.19 Ϫ0.07 0.03 0.08 G7--SM Ϫ1.77a Ϫ2.29a Ϫ1.21a Ϫ1.38a G7--VM Ϫ1.59a Ϫ2.13a 0.05 Ϫ0.03 HA8160--DM Ϫ1.53a Ϫ2.13a Ϫ0.09 Ϫ0.22 HA8160--TM Ϫ1.05a Ϫ1.59a Ϫ0.13 Ϫ0.33 Immunoassay Unimate 0.05 Ϫ0.20 Ϫ0.10 Ϫ0.18 TinaQuant--IIb Ϫ0.03 Ϫ0.44 Ϫ0.11 Ϫ0.28 AU640 Ϫ0.42 Ϫ0.50 Ϫ0.57 Ϫ0.48 DCA2000 0.12 Ϫ0.15 0.00 Ϫ0.10 Enzymatic assay Norudia 0.00 Ϫ0.20 Ϫ0.20 Ϫ0.24 a Clinically significant difference (Ͼ0.6% at 6% and Ͼ0.7% at 7%). b n ϭ 5 for Hb G-Coushatta, n ϭ 9 for Hb Queens. Clinical Chemistry 53, No. 12, 2007 2205

tant to understand the effects of these HbVAR on the Mass Spectrometry–Based Detection of Hemoglobin E various methods. Mutation by Allele-Specific Base Extension Reaction, Jason C.H. Tsang,1 Pimlak Charoenkwan,2 Katherine C.K. Chow,1 Yongjie Jin,3 Chanane Wanapirak,2 Torpong San- guansermsri,2 Y.M. Dennis Lo,1,4 and Rossa W.K. Chiu1,4* Grant/funding support: This work was supported by a (1 Department of Chemical Pathology, The Chinese Uni- Samsung Biomedical Research Institute grant (SBRI versity of Hong Kong, Shatin, New Territories, Hong C-A6-403-2). Kong SAR; 2 Faculty of Medicine, Chiang Mai University, Financial disclosures: None declared. Chiang Mai, Thailand; 3 Centre for Emerging Infectious Diseases and 4 Centre for Research into Circulating Fetal Nucleic Acids, Li Ka Shing Institute of Health Sciences, References The Chinese University of Hong Kong, Shatin, New 1. Bry L, Chen PC, Sacks DB. Effects of hemoglobin variants and chemically Territories, Hong Kong SAR; * address correspondence to modified derivatives on assays for glycohemoglobin. Clin Chem 2001;47: 153–63. this author at: Department of Chemical Pathology, Rm. 2. Schnedl WJ, Liebminger A, Roller RE, Lipp RW, Krejs GJ. Hemoglobin variants 38063, 1/F, Clinical Sciences Building, Prince of Wales and determination of glycated hemoglobin (HbA1c). Diabetes Metab Res Rev Hospital, 30-32 Ngan Shing St., Shatin, Hong Kong SAR; 2001;17:94–8. fax 852 2636 5090, e-mail [email protected]) 3. Sacks DB. Hemoglobin variants and hemoglobin A1c analysis: problem solved? Clin Chem 2003;49:1245–7. 4. Weykamp CW, Penders TJ, Muskiet FA, van der Slik W. Influence of Background: The specific detection of a minor popula- hemoglobin variants and derivatives on glycohemoglobin determinations, as tion of mutant DNA molecules requires methods of investigated by 102 laboratories using 16 methods. Clin Chem 1993;39: high specificity and sensitivity. While the single-allele 1717–23. base extension reaction (SABER) was shown to be 5. Roberts WL, De BK, Brown D, Hanbury CM, Hoyer JD, John WG, et al. Effects of hemoglobin C and S traits on eight glycohemoglobin methods. Clin Chem useful for the detection of certain beta-thalassemia mu- 2002;48:383–5. tations, we encountered problems with false positivity 6. Lee ST, Kim MS, Choi DY, Kim SK, Ki CS. Incidence of variant hemoglobin during development of SABER for the noninvasive (Hb) and increased fetal Hb concentrations and their effect on Hb A1c measurement in a Korean population. Clin Chem 2006;52:1445–6. prenatal diagnosis of the hemoglobin E (HbE) disease. 7. Hardison RC, Chui DH, Giardine B, Riemer C, Patrinos GP, Anagnou N, et Systematic optimization resulted in an alternative pro- al. HbVar: a relational database of human hemoglobin variants and tocol, the allele-specific base extension reaction thalassemia mutations at the globin gene server. Hum Mutat 2002;19: 225–33. (ASBER). 8. Blackwell RQ, Ro IH, Liu CS, Yang HJ, Wang CC, Huang JT. Hemoglobin Methods: An artificial model was established by mixing variant found in Koreans, Chinese, and North American Indians: alpha-2 genomic DNA of HbE carriers and normal individuals. beta-2 22 Glu Ala. Am J Phys Anthropol 1969;30:389–91. Effects of terminator concentration and annealing tem- 9. Blackwell RW, Liu CS, Yang HJ, Wang CC, Huang JT. Hemoglobin variant common to Chinese and North American Indians: alpha-2-beta-22 Glu-Ala. perature on the nonspecificity of SABER were then Science 1968;161:381–2. studied. The use of a single relevant terminator and the 10. Li HJ, Zhao XN, Qin F, Li HW, Li L, He XJ, et al. Abnormal hemoglobins in the other 3 types of dideoxynucleotide as competing termi- Silk Road region of China. Hum Genet 1990;86:231–5. nators were also compared in the development of the 11. Jeppsson JO, Kobold U, Barr J, Finke A, Hoelzel W, Hoshino T, et al. Approved IFCC reference method for the measurement of HbA1c in human ASBER protocol. Thirteen cases of HbE-susceptible blood. Clin Chem Lab Med 2002;40:78–89. pregnancies were tested to compare the SABER and the 12. Hoelzel W, Weykamp C, Jeppsson JO, Miedema K, Barr JR, Goodall I, et al. ASBER protocols. IFCC reference system for measurement of hemoglobin A1c in human blood and the national standardization schemes in the United States, Japan, and Results: Decreasing the single relevant terminator con- Sweden: a method-comparison study. Clin Chem 2004;50:166–74. centration and increasing the annealing temperature in 13. Martin RF. General Deming regression for estimating systematic bias and its SABER were found to improve specificity. The use of confidence interval in method-comparison studies. Clin Chem 2000;46: 100–4. the other 3 types of dideoxynucleotide as competing 14. Roberts WL, Safar-Pour S, De BK, Rohlfing CL, Weykamp CW, Little RR. terminators was shown to offer better detection sensi- Effects of hemoglobin C and S traits on glycohemoglobin measurements by tivity than a single terminator in ASBER. Genotyping eleven methods. Clin Chem 2005;51:776–8. results were all correctly determined by ASBER, except 15. Ogawa K, Bando T, Ogawa M, Miyazaki A, Nakanishi T, Shimizu A. Hemo- globin variant HbG-Coushatta (beta-22 Glu –Ͼ Ala) found by dissociation of one false-negative detection (sensitivity: 80%, specific- blood glucose from values of HbA1C measured by HPLC. Intern Med ity: 100%). 2003;42:781–7. Conclusions: An alternative mass spectrometry–based 16. Nakanishi T, Miyazaki A, Shimizu A, Yamaguchi A, Nishimura S. Assessment protocol for noninvasive prenatal diagnosis, ASBER, of the effect of hemoglobin variants on routine HbA1c measurements by electrospray ionization mass spectrometry. Clin Chim Acta 2002;323:89– has been successfully developed to allow the detection 101. of a minor DNA population with a point mutation. 17. Ohba Y, Miyaji T, Murakami M, Kadowaki S, Fujita T, Oimomi M, et al. Hb © 2007 American Association for Clinical Chemistry Himeji or beta 140 (H18) Ala—Asp. A slightly unstable hemoglobin with increased beta N-terminal glycation. Hemoglobin 1986;10:109–25. The discovery of circulating fetal nucleic acids in maternal Previously published online at DOI: 10.1373/clinchem.2007.093963 plasma has opened up exciting possibilities for noninvasive prenatal diagnosis (1, 2). The recent development of 2206 Technical Briefs

the mass spectrometry-based single-allele base extension previously (3). HotStar Taq polymerase (Qiagen) was reaction (SABER) protocol has enabled sensitive differen- used in the PCR at a final volume of 25 ␮L, containing 10 tiation of fetal-specific alleles down to a single-nucleotide ␮L of plasma DNA and 0.2 ␮mol/L PCR primers (Inte- level (3–5). In this report, we intended to develop a mass grated DNA Technologies). The thermal profile was 95 °C spectrometry–based method for the noninvasive prenatal for 15 min for hot start, 45 cycles of denaturing at 95 °C for diagnosis of the hemoglobin E (HbE) mutation. Unexpect- 20 s, annealing at 58 °C for 30 s, and extension at 72 °C for edly, the lack of specificity of SABER for the HbE muta- 1 min, followed by a final incubation at 72 °C for 3 min. tion was discovered during assay development, and sys- PCR products were then treated with shrimp alkaline tematic optimization on an artificial model has been phosphatase (Sequenom) for 40 min at 37 °C to remove carried out. This development has resulted in an alterna- unreacted dNTPs. Base extension reaction was carried out tive protocol, the allele-specific base extension reaction with thermosequenase (Sequenom) on 10 ␮L of shrimp (ASBER). alkaline phosphatase–treated PCR product in a final reac- HbE disease is an autosomal recessive hemoglobinop- tion volume of 14 ␮L, with 1.54 ␮mol/L of the extension athy caused by a (GAG3AAG) missense mutation in primer (Integrated DNA Technologies) and a terminator codon 26 of the ␤-globin gene (6). It is the most common mix of dideoxy/deoxynucleotides, each at 64 ␮mol/L. thalassemic hemoglobinopathy in Southeast Asia (7). Al- The thermal profile consisted of 94 °C for 2 min, followed though homozygotes of HbE are mildly affected by the by a rapid thermocycling for 75 cycles at 94 °C, 52 °C, and mutation, compound heterozygotes of the HbE and the 74 °C, all for 5 s. Products were then analyzed by the ␤-thalassemia mutation will result in severe anemia MassARRAY™ Analyzer Compact Mass Spectrometer (HbE/␤-thalassemia). Therefore, detection of the HbE (Brucker), a MALDI-TOF system. Details of the PCR mutant allele is critically important in the prenatal diag- primers, termination mix, and extension primers are listed nosis of thalassemia major in Southeast Asia. In noninva- in the Data Supplement that accompanies the online sive prenatal diagnosis of HbE, the absence of the pater- version of this Technical Brief at http://www.clinchem. nally transmitted fetal HbE mutant allele in the plasma of org/content/vol53/issue12. a pregnant female carrier of a ␤-thalassemic mutation Thirteen HbE-negative pregnant participants (gesta- negates the fetal inheritance of HbE/␤-thalassemia. The- tional age: 16–22 weeks) with male partners being HbE oretically, this suggests that 50% of invasive procedures carriers were recruited. Genotypes of the couples and the are unnecessary and can thus be avoided. fetuses were confirmed by analysis of buffy coat and cord All samples in this study were collected with informed blood samples, respectively, by the standard homogenous consent, and approval was granted by the institutional MassEXTEND protocol. Five of the fetuses were shown to ethics committee. Venous blood (6 mL) was collected into be HbE positive. EDTA tubes from each couple referred for prenatal diag- Genotyping results of the SABER protocol on maternal nosis. Plasma and buffy coat were harvested from blood plasma showed nonspecificity (see Fig. 1 in the Data samples after a 1st centrifugation at 1600g for 10 min and Supplement). All 8 informative-negative pregnancies a further microcentrifugation of plasma aliquot at 16 000g were misclassified as positive. To evaluate the important for 10 min as previously described (8). DNA was ex- analytical parameters for the nonspecificity of SABER, an tracted from buffy coat and 800 ␮L of plasma by a artificial model was established. Genomic DNA of a male Nucleon Blood DNA Extraction Kit (GE Healthcare) and a HbE carrier was mixed with wild-type maternal genomic QIAamp Blood Mini Kit (Qiagen) with elution volume of DNA to mimic the plasma DNA of an affected pregnancy. ␮ 50 LH2O, respectively, according to the manufacturers’ Direct dilution of wild-type maternal genomic DNA was recommendations. used to mimic plasma DNA of normal pregnancy. The principle of standard homogenous MassEXTEND Systematic evaluation revealed that the terminator con- (Sequenom) and SABER protocols have been described centration and the annealing temperature in the base

Table 1. Comparing the percentage of false-positive detection and true-positive detection of SABER at different terminator concentrations (A) and annealing temperatures (B). A. Effect of annealing temperature in the base extension step on the percentage of positive call. 52 °C* 54 °C 56 °C 58 °C 60 °C 62 °C 64 °C 66 °C 68 °C 70 °C True positive 100% 100% 100% 100% 100% 100% 87.5% 100% 87.5% 62.5% False positive 100% 100% 100% 87.5% 100% 75% 12.5% 0% 0% 0% B. Effect of terminator concentration on the percentage of positive call. Standard 2؋ 8؋ 20؋ 50؋ 100؋ True positive 100% 100% 100% 87.5% 100% 100% False positive 100% 100% 62.5% 12.5% 0% 0% a The percentage of positive call was calculated similarly as described in Fig. 1B. Clinical Chemistry 53, No. 12, 2007 2207

extension reaction were critical for the reduction of false- (Fig. 1A). Initially, the use of a single relevant terminator positive detection in SABER. A 50-fold dilution of the (ddA only) at standard concentration also revealed nonstandard terminator concentration and annealing temper- specificity (data not shown). Thus, terminator dilution ature at 66 °C were shown to be optimal for both specific and the introduction of competing terminators (ddA, and sensitive detection of the HbE mutation in the artifi- ddC, ddG, and ddT, each at 64 ␮mol/L) in the ASBER cial model (Table 1). protocol were evaluated. Both the use of competing We explored the development of an alternative proto- terminators and 20-fold dilution of the relevant termina- col, ASBER. The 3Ј end of the extension primer for ASBER tor were found to provide specific genotyping results. was engineered to be complementary to the fetal mutant However, the use of competing terminators appears to allele. Hence, primer extension of the maternal wild-type offer better detection sensitivity on the artificial model allele would be inhibited by 3Ј-primer-template mismatch with different percentage mix of mutant DNA (Fig. 1B).

Fig. 1. Comparison between the SABER and ASBER protocols. (A), Schematic diagram comparing the principles of the SABER and ASBER protocols. The site of the point mutation is indicated in capital letters. The boxed letter indicates the type of dideoxynucleotide terminator in the base extension reaction. (B), grouped bar chart comparing detection sensitivity of different protocol designs on artificial plasma model with different concentrations. Eight separate PCRs were performed for a particular concentration of artificial mutant and wild-type allele mix followed by 8 separate base extension reactions. The base extension products were then analyzed in duplicate by the MassARRAY™ system. A reaction was scored positive if 1 of the duplicates was called by the TyperAnalyzer software (Sequenom). The sensitivity was then calculated as the percentage of positive detection out of the 8 separate PCRs. 2208 Technical Briefs

The design of ASBER with competing terminators was minators) is better than dilution of the relevant terminator adopted to reanalyze the 13 maternal plasma samples, (Fig. 1B). and the fetal genotypes were all correctly determined Reanalysis of the 13 pregnancies at risk for HbE/␤- except 1 case of false-negative result. thalassemia with the ASBER protocol showed substantial In this study, we have shown that for certain muta- improvement in specificity compared with the SABER tional context, SABER might generate nonspecific results protocol, with 1 false-negative result. The diagnostic per- due to incorporation of the single terminator in the formance of the HbE ASBER assay needs to be confirmed extension reaction mixture despite it being noncomple- with larger-scale studies. ment to the template PCR product. Although the mecha- In summary, we have successfully developed ASBER, nism behind this nonspecificity or misincorporation is still which confers advantages in terms of specificity and unclear, we hypothesize that the intrinsic specificity of sensitivity over SABER for the detection of the HbE SABER is a subtle balance between the fidelity of the mutation. Using this approach, noninvasive prenatal di- polymerase and the degree of excess of the single relevant agnosis of the HbE mutation has been achieved with a terminator in the primer extension reaction mixture (9). specificity of 100% (8 of 8) and a sensitivity of 80% (4 of 5). The lower the polymerase fidelity and the higher the To further improve the sensitivity of the assay, potential terminator concentration promote misincorporation. fetal allele enrichment steps—for example, PCR clamping Based on this reasoning, optimization can be achieved by by PNA probe and size fractionation—can be included adjusting the terminator concentration as described above (13). We speculate that the application of the ASBER (Table 1A). Others have shown that the polymerase protocol can be extended to other fields of circulating fidelity could be improved by using a proofreading nucleic acids, for instance, detection of circulating tumor- polymerase (9). However, this method did not offer any specific DNA in cancer patients such as KRAS point improvement in the specific detection of the trace amount mutations (14) and detection of circulating donor-specific of fetal DNA in maternal plasma in our preliminary DNA in transplant recipients (15). findings (data not shown). On the other hand, we showed that nonspecificity of SABER was reduced by increasing the annealing temperature for base extension (Table 1B). Grant/funding support: This work was supported by an Earmarked Research Grant (CUHK4395/03M) from the Re- This result might be because the environment becomes search Grants Council of the Hong Kong Special Adminis- less favorable for misincorporation of the noncomplemen- trative Region, China. Y.M.D.L. is supported by the Chair tary terminator. Professorship Scheme of the Li Ka Shing Foundation. To overcome the intrinsic tendency of terminator mis- Financial disclosures: Y.M.D.L. and R.W.K.C. hold patents incorporation in SABER, we explored and developed and have filed patent applications on aspects of the use of ASBER based on the principle of allele-specific primer fetal nucleic acids in maternal plasma for noninvasive pre- extension, which has been reported to offer successful natal diagnosis, a proportion of which has been licensed to mass spectrometry–based mutation genotyping (10, 11). Sequenom, Inc. Y.M.D.L. is a consultant for Sequenom Inc. However, such design has not been applied to specific References fetal allele detection in noninvasive prenatal diagnosis. 1. Lo YMD, Corbetta N, Chamberlain PF, Rai V, Sargent IL, Redman CW, et al. Compared with previous studies (10, 11), specific priming Presence of fetal DNA in maternal plasma and serum. Lancet 1997;350: of the fetal-specific allele in noninvasive prenatal diagno- 485–7. 2. Lo YMD, Chiu RWK. Prenatal diagnosis: progress through plasma nucleic sis is less favorable due to the overwhelming background acids. Nat Rev Genet 2007;8:71–7. of maternal allele and the lack of priming competition 3. Ding C, Chiu RWK, Lau TK, Leung TN, Chan LC, Chan AY, et al. MS analysis of single-nucleotide differences in circulating nucleic acids: application to from alternative allele-specific primer. Yet we expect noninvasive prenatal diagnosis. Proc Natl Acad Sci U S A 2004;101: better specificity in ASBER than the conventional allele- 10762–7. 4. Li Y, Page-Christiaens GC, Gille JJ, Holzgreve W, Hahn S. Non-invasive specific PCR (12), because the ASBER extension products prenatal detection of achondroplasia in size-fractionated cell-free DNA by would not serve as templates for further amplification. MALDI-TOF MS assay. Prenat Diagn 2007;27:11–7. Thus, a misprimed ASBER product, if it occurs, would not 5. Li Y, Wenzel F, Holzgreve W, Hahn S. Genotyping fetal paternally inherited SNPs by MALDI-TOF MS using cell-free fetal DNA in maternal plasma: be exponentially propagated. Nevertheless, false-positive influence of size fractionation. Electrophoresis 2006;27:3889–96. detection is also detectable in ASBER with the single 6. Orkin SH, Kazazian HH Jr, Antonarakis SE, Ostrer H, Goff SC, Sexton JP. Abnormal RNA processing due to the exon mutation of beta E-globin gene. relevant terminator and therefore required optimization. Nature 1982;300:768–9. False-positive detection was not observed in ASBER with 7. Fucharoen S, Winichagoon P. Hemoglobinopathies in Southeast Asia. He- moglobin 1987;11:65–88. competing terminators. This finding further suggested 8. Chiu RWK, Poon LL, Lau TK, Leung TN, Wong EM, Lo YMD. Effects of that an overdominance of any single terminator in the blood-processing protocols on fetal and total DNA quantification in maternal extension mixture promotes misincorporation. Further plasma. Clin Chem 2001;47:1607–13. 9. Di Giusto D, King GC. Single base extension (SBE) with proofreading comparison of detection sensitivity of the 2 ASBER pro- polymerases and phosphorothioate primers: improved fidelity in single- tocols shows that improving specificity by reduction of substrate assays. Nucleic Acids Res 2003;31:e7. 10. Blievernicht JK, Schaeffeler E, Klein K, Eichelbaum M, Schwab M, Zanger the fractional concentration of a terminator (i.e., inclusion UM. MALDI-TOF mass spectrometry for multiplex genotyping of CYP2B6 of 3 other types of dideoxynucleotide as competing ter- single-nucleotide polymorphisms. Clin Chem 2007;53:24–33. Clinical Chemistry 53, No. 12, 2007 2209

11. Higgins GS, Little DP, Koster H. Competitive oligonucleotide single-base ments from an upright starting point, a surprising ob- extension combined with mass spectrometric detection for mutation screening. Biotechniques 1997;23:710–4. servation, because laboratorians are usually trained to 12. Wu DY, Ugozzoli L, Pal BK, Wallace RB. Allele-specific enzymatic amplifica- mix 10 or 20 times. We also observed that, in a frozen tion of beta-globin genomic DNA for diagnosis of sickle cell anemia. Proc overfilled tube, resulting analyte concentrations will be Natl Acad Sci U S A 1989;86:2757–60. 13. Li Y, Di Naro E, Vitucci A, Zimmermann B, Holzgreve W, Hahn S. Detection lower because more concentrated solutes leak from the of paternally inherited fetal point mutations for beta-thalassemia using tube. size-fractionated cell-free DNA in maternal plasma. JAMA 2005;293:843–9. Conclusions: A high-throughput, automated thawing 14. Anker P, Lefort F, Vasioukhin V, Lyautey J, Lederrey C, Chen XQ, et al. K-ras mutations are found in DNA extracted from the plasma of patients with and mixing workcell was successfully built, validated, colorectal cancer. Gastroenterology 1997;112:1114–20. and installed on our automated transport and sorting 15. Lo YMD, Tein MS, Pang CC, Yeung CK, Tong KL, Hjelm NM. Presence of system. donor-specific DNA in plasma of kidney and liver-transplant recipients. Lancet 1998;351:1329–30. © 2007 American Association for Clinical Chemistry

Previously published online at DOI: 10.1373/clinchem.2007.095133 Our laboratory is a high-volume esoteric reference laboratory, accepting approximately 25–30 000 specimens per day. One of our largest laboratory sections, the Auto- mated Core Laboratory, receives approximately 25% of that daily volume, performing more than 140 different Development and Validation of an Automated Thawing chemistry, immunoassay, and specific protein tests, em- and Mixing Workcell, Charles D. Hawker,1,2* William L. phasizing cancer antigens, endocrine testing, and urine Roberts,1,2 Antonio DaSilva,3 Gordon D. Stam,1 William E. chemistry, but not routine serum chemistry. More than Owen,4 DeVirl Curtis,1 Byung-Sang Choi,5† and Terry A. half of these 6–7000 specimens per day are frozen. The Ring5 (1 ARUP Laboratories, Salt Lake City, UT; 2 Department laboratory has been thawing them manually at room of Pathology, School of Medicine, University of Utah, Salt Lake temperature (to prevent degradation of labile analytes) by City, UT; 3 Motoman, Inc., Irvine, CA; 4 ARUP Institute for blowing air from an ordinary electric fan at batches of Clinical and Experimental Pathology, Salt Lake City, UT; tubes, a process requiring more than1htoassure 5 Department of Chemical and Fuels Engineering, College of complete thawing. The specimens were then mixed by Engineering, University of Utah, Salt Lake City, UT; * address manually inverting the tubes 10 times, before decapping correspondence to this author at: ARUP Laboratories, 500 for the various analyses. Chipeta Way, Salt Lake City, UT 84108; fax 801-584-5261, Our long-term objective for this laboratory section is to e-mail [email protected]); † Current address: Department interface analyzers to our automated transport and sort- of Materials Science and Engineering, Gwangju Institute of ing system to achieve total laboratory automation and to Science and Technology (GIST), Republic of Korea) automate other manual activities such as inspecting for adequate specimen volume and for the presence of inter- Background: Working toward a goal of total laboratory fering substances as indicated by hemolysis, lipemia, and automation, we are automating manual activities in our icterus. The development of some form of automated highest volume laboratory section. Because half of all workcell for rapidly thawing specimens at room temper- specimens arriving in this laboratory section are frozen, ature and mixing the thawed specimens was a 1st step we began by developing an automated workcell for toward our overall automation objective. We are not thawing frozen specimens and mixing the thawed spec- aware that such a robotic system has previously been built imens to remove concentration gradients resulting from or described. Our design intention with the workcell was to leave the freezing and thawing. specimen tubes in their track carriers (see Supplemental Methods: We developed an initial robotic workcell that Fig. 1 that accompanies the online version of this Tech- removed specimens from the transport system’s con- nical Brief at http://www.clinchem.org/content/vol53/ veyor, blew high-velocity room temperature air at the issue12). Therefore, we sought to aim the thawing air tubes, mixed them, and replaced them on the conveyor. directly at the tubes in the carrier through a slit normally Aliquots of citrated plasma were frozen with thermo- used for reading bar codes. We developed an experimen- couples immersed in the tubes, and thawing times and tal apparatus with which we could evaluate a variety of temperatures were monitored. Completeness of mixing nozzle designs that would direct air into the carrier slit of thawed specimens was studied by careful removal of (Supplemental Fig. 2). This design included thermocou- small aliquots from the uppermost layer of the upright ples to monitor the thawing times of the tubes. Initially, tubes without disturbing tube contents and analysis of thawing of frozen water was evaluated, but later, when total protein and electrolytes. the conditions of air flow and nozzle design had been Results: High velocity ambient air aimed directly at established, we tested the system with specimens of tubes ranging from 12 ؋ 75 to 16 ؋ 100 mm brought out-of-date blood bank plasma. These studies (shown in specimens to room temperature in a maximum of 23 Supplemental Fig. 3) indicated that a nozzle shaped as a min. Adequate mixing of the specimens by the work- small brass plug with a 2-mm orifice on a beveled edge cell’s robot required only 2 approximate 126° move- (Supplemental Fig. 4) aimed at the bottom of the specimen 2210 Technical Briefs

matic tool that grasps up to 10 carriers, with tubes, at a time for transfers between the conveyor system and deck and for mixing. This tool also uses pneumatic pin cylinders that press tightly against the top of each tube cap. In case a tube is capped with a push-top instead of a screw cap, the pin cylinders assure that the cap will not loosen during mixing and possibly leak. The completed workcell was subjected to a rigorous validation study. In our thawing studies, most of the 760 locations on the thawing deck were populated with labeled tubes of frozen water, except that in key locations we placed plasma specimens in labeled tubes with temperature-recording thermocouples and surrounded those tubes with other plasma specimens in labeled tubes. The intent was to have a full deck of 760 frozen specimens to Fig. 1. Temperatures continuously monitored by thermocouples in determine their impact on the plenum temperature and tubes of frozen, out-of-date blood bank plasma in different locations on the overall thawing rate. In repetitive studies we then the deck of the workcell. adjusted the air flow from the compressor by varying the The temperatures in all tubes reached 20 °C in 24 min or less in this study at a compressor speed, seeking to balance an optimal thawing variable frequency drive setting of 40% (24 cycles per second). The variable frequency drive setting chosen for routine operation was 45% (27 cycles per time against generation of excess heat and compressor second), which gave slightly greater air flow, with all tubes at or above 20 °C in noise. Compressing air generates heat (1), and, at higher 23 min or less. compressor speeds (higher air flow), the plenum became warmer than desired for thawing of specimens, as deter- portion of the false-bottom tube would bring specimens mined by baseline temperature measurements. The re- frozen in dry ice to 20 °C in 22–23 min, at a plenum sults of one of these thawing studies with different air pressure of 8.795 ϫ 104 Pa with an air flow of 0.000434 flows are shown in Fig. 1. All of the thermocouple m3/s (0.434 L/s) per nozzle. temperature readings were above 20 °C in 23 min or less. The Motoman workcell (Supplemental Fig. 5) uses an These rapid thawing times were achieved with a combi- HP3XFC articulated 6-axis robot to remove specimen nation of high air flow (0.0000333 m3/s, or 0.0333 L/s, per carriers from the track and place them on the deck in specimen) and directing the air flow into the slit of the holders that prevent them from falling over, even with carrier so that it wrapped around the tube, thawing it high velocity air blowing at them. A 10-hp, 480-volt, from all sides. We postulated that, without the transport 60-cycle compressor directs the air into a plenum, which carrier, the air would simply have deflected off the front comprises the deck of the system. A total of 760 brass of the tube, and the back of the tube would not have nozzles as described above are attached to the plenum, thawed as quickly. This air flow was obtained with the each designed to aim at the tube directly in front of it variable frequency drive set at 45%, which delivered an through the slit in the carrier. A variable frequency drive alternating current frequency of 27 cycles per second. regulates compressor speed to maintain constant pressure The formation of concentration gradients in serum, in the plenum, thus optimizing the air flow from each urine, and other frozen specimens after the specimens are nozzle. The number of nozzles in the deck design was thawed is known (2, 3), and thawed specimens must be based on the reach of the robot combined with a target well mixed before they can be analyzed. Most laboratori- workcell throughput speed of at least 1000 specimens per ans are trained to mix specimens 10–20 times before hour. The robot was fitted with a custom-designed pneu- testing, but we were unable to find a published reference for 5 ؍ Table 1. Summary of analytical results on the upper layers of thawed specimens after indicated number of mixes (n each group). Sodium, mmol/L Potassium, mmol/L Chloride, mmol/L Albumin, g/L

No. of Mixes Mean SD Mean SD Mean SD Mean SD Baseline 167.6 0.55 3.50 0.05 75.0 0.00 38.0 0.00 0 129.4a 12.46 2.70a 0.26 58.0a 5.83 29.2a 0.26 2 167.0 0.00 3.40 0.09 75.0 0.00 38.0 0.00 4 167.8 1.10 3.40 0.04 75.0 0.45 37.8 0.04 6 167.8 0.45 3.50 0.05 75.2 0.45 37.8 0.04 8 168.0 1.00 3.40 0.08 75.6 0.55 37.8 0.04 10 168.2 0.45 3.50 0.05 75.2 0.45 38.3 0.05 12 167.8 0.84 3.40 0.05 75.4 0.04 38.2 0.04 a Significantly different from baseline value by paired t-test, P Ͻ0.005; all other P values Ͼ0.05. Clinical Chemistry 53, No. 12, 2007 2211

recommending any particular number of mixes. We de- Grant/funding support: This work was supported by ARUP signed the workcell’s 6-axis robot to perform the mixing Laboratories. by maintaining the specimens over a fixed location, so Financial disclosures: None declared. that a stainless steel pan could be positioned to catch drips from any leaking specimens. This design, coupled with References 1. Buse FW, Karassik IJ, Krutzsch WC, Worster AR, Dayton BB, Jorgensen R. the design of the pneumatic tool and its pin cylinders Pumps and compressors. In: Baumeister T, Avallone EA, Baumeister T III, limited the range of rotation of the tubes to 126° in either eds. Marks’ Standard Handbook for Mechanical Engineers, 8th ed. New York: direction. The resulting mixing pattern consisted of a McGraw Hill, 1978(Chapt 14):14-30–14-44. 2. Omang SH, Vellar OD. Concentration gradients in biological samples during rotation of approximately 126° to the left from an upright storage, freezing and thawing. Z Anal Chem 1974;269:177–81. starting position, return to upright, then 126° to the right, 3. Hirano T, Yoneyama T, Matsuzaki H, Sekine T. Simple method for preparing a and return to upright. A single 126° tilt and return to concentration gradient of serum components by freezing and thawing. Clin Chem 1991;37:1225–9. upright constitutes 1 mix cycle in this discussion. 4. Ophardt CE. Elmhurst College, Virtual ChemBook. http://www.elmhurst.edu/ We evaluated 0, 2, 4, up to 12 mix cycles each with 5 ϳchm/vchembook/122Adensityice.html (accessed June 2007). different replicate expired blood bank plasma specimens 5. McGlory DH. Often-overlooked effect of freezing standard solutions [Letter]. Clin Chem 1971;17:1074. of 4.5 mL using our standard false bottom tube. The tubes were thawed on the workcell deck, but without the air Previously published online at DOI: 10.1373/clinchem.2007.094185 blowing, to avoid any vibration or shaking of the tubes. After the tubes had thawed, we carefully sampled 200 ␮L from the uppermost layer of each tube. The 1st set of replicates was sampled without robotic mixing, the 2nd set was sampled after 2 mix cycles, the 3rd set after 4 mix cycles, and so on. The aliquots were analyzed for albumin, Detection of Factor VIII Gene Mutations by High- sodium, potassium, and chloride on a Modular P analyzer Resolution Melting Analysis, Andrew D. Laurie,* Mark P. (Roche Diagnostics) and compared to 5 replicates of Smith, and Peter M. George (Department of Molecular unfrozen plasma that served as baseline or expected Pathology, Canterbury Health Laboratories, Christchurch, values. The results are shown in Table 1. After only 2 mix New Zealand; * address correspondence to this author at: cycles (2 elevations to 126° followed by return to upright), P.O. Box 151, Christchurch, New Zealand; fax 64-3- the levels of all 4 analytes were indistinguishable from the 3640545, e-mail [email protected]) baseline levels. This result was surprisingly fewer than we had expected based on experience. However, because Background: Single base-pair substitution mutations in human mixing motions may not duplicate the uniform the gene for coagulation factor VIII, procoagulant com- speed and angles of our programmed robot (approxi- ponent (hemophilia A) (F8) account for approximately mately 2 seconds to tilt 126° and return to upright), we are 50% of severe cases of hemophilia A (HA), and almost not recommending that laboratorians reduce their speci- all moderate or mild cases. Because F8 is a large gene, men mixes. mutation screening using denaturing HPLC or DNA Knowing that an air bubble was required to achieve sequencing is time-consuming and expensive. specimen mixing, we evaluated overfilling tubes with Methods: We evaluated high-resolution melting analy- plasma in an attempt to determine the minimum size of sis as an option for screening for F8 gene mutations. The air bubble necessary for adequate mixing. The volume of melting curves of amplicons heterozygous for known F8 water expands approximately 9% when it is frozen (4). gene mutations were compared with melting curves of Furthermore, if a tube is filled too full and leaks due to the corresponding normal amplicons to assess whether this expansion, specimen solutes (minerals, proteins, etc.) melting analysis could detect these variants. We exam- preferentially squeeze through the cap threads, as has ined 2 platforms, the Roche LightCycler 480 (LC480) and been reported for frozen, overfilled standard solutions (5) the Idaho Technology LightScanner. and for frozen, overfilled serum and urine specimens (2), Results: On both instruments, 18 (90%) of the 20 F8 gene and leaked specimens will be unacceptable for testing because the concentrations of analytes will have changed. variants we examined were resolved by melting analy- We also learned that the minimum size of air bubble to sis. For the other 2 mutations, the melting curves of the facilitate mixing was 1.0 mL, which was sufficiently below heterozygous amplicons were similar to the correspond- the top of the tube to prevent leakage during freezing. ing normal amplicons, suggesting these variants may not be detected by this approach in a mutation-scanning In summary, we designed, validated, and installed an screen. automated thawing and mixing workcell, which is con- Conclusion: High-resolution melting analysis is an ap- nected to our automated transport system and has a pealing technology for F8 gene screening. It is rapid and throughput to thaw and mix Ͼ1000 specimens per hour. quickly identifies mutations in the majority of HA The 6-axis robot appears able to effectively mix specimens patients; samples in which no mutation is detected with fewer mixes than routinely taught to laboratorians. require further testing by DNA sequencing. The LC480 Overfilled specimens that leak when frozen are unaccept- and LightScanner platforms performed similarly. able for laboratory analysis. © 2007 American Association for Clinical Chemistry 2212 Technical Briefs

Hemophilia A (HA) is a coagulation disorder caused by a In this study we assessed whether 20 different F8 gene lack of normal coagulation factor VIII (FVIII) activity (1). mutations, located across 13 exons, could be detected by This disease is caused by mutations in the coagulation high-resolution melting analysis. These are mutations we factor VIII, procoagulant component (hemophilia A) (F8) have previously identified by dHPLC and DNA sequenc- gene, which encodes FVIII and is located at Xq28. F8 is a ing during our F8 gene-screening program (6). For each large gene, comprising 26 exons across 186 kb, and mutation we assessed whether the melting curve of the produces a 9-kb mRNA transcript. All exons are small heterozygous amplicon differed from that of the corre- (69–262 bp) except exon 14, which is 3.1 kb. sponding normal amplicon to evaluate whether a melting Approximately 45% of severe HA cases are the result of analysis screen of the F8 gene would detect the variant. a large inversion that disrupts the F8 gene in intron 22, We tested 2 different platforms—the Roche LightCycler and a further 1%–5% are caused by an inversion affecting 480 (LC480) and the Idaho Technology LightScanner. intron 1. In the remaining severe cases, and in cases of Genomic DNA from 20 male HA patients, each with a mild or moderate HA, a single-base substitution, small different F8 gene mutation, was mixed with an equal insertion, or deletion is usually the causative mutation. amount of DNA from a healthy male control. Because Such mutations create either missense, nonsense, or males are hemizygous for the F8 gene, this mixing is frameshift mutations, or affect consensus splice sites [re- necessary to ensure the amplified exon is heterozygous viewed in (2)]. Many such variants have been reported in for the F8 mutation, which is a requirement for detection HA patients, and more than 900 different mutations have by melting analysis. Each mixed sample, and a mixed been described (3). normal control, was amplified using primers targeting Many hematology services now offer genetic screening only the F8 exon where the mutation is located. We used to identify the F8 gene mutation in their HA patients. This primer sets previously optimized for dHPLC by Olden- information allows identification of carrier females, pre- burg et al. (5), because the requirements of amplicons for natal testing for affected pregnancies, preimplantation melting analysis are likely to be similar to those for genetic diagnosis, and in some cases prediction of the dHPLC. Samples were amplified in duplicate 15-␮L PCRs likelihood of FVIII inhibitor production in HA patients using the LightCycler 480 Genotyping Master PCR mix- receiving prophylactic FVIII treatment (2). ture (Roche), in the presence of 1ϫ LCGreen PLUS (Idaho Genetic screening of HA patients who do not have Technology), 0.5 ␮mol/L forward and reverse primers, either the intron 1 or intron 22 inversions is problematic and 50 ng DNA template. The thermal cycling protocol because of the size of the F8 gene and the variety of was as follows: polymerase activation (95 °C for 10 min); mutations that can occur. Before capillary-based DNA touchdown cycling step [5 cycles of: denaturation at 95 °C sequencing platforms were widely available it was not for 20 s, annealing at 61 °C for 20 s (decreased by 1 °C per practical to sequence all 26 exons, so mutation-scanning cycle), extension at 72 °C for 20 s]; amplification (40 cycles strategies such as single-strand conformational polymor- of: denaturation at 95 °C for 20 s, annealing at 56 °C for phism analysis were used (4). A more sensitive technique, 20 s, extension at 72 °C for 20 s); and a final extension denaturing HPLC (dHPLC), has also been used success- (72 °C for 5 min). The touchdown cycling step was in- fully for F8 mutation scanning (5). In our laboratory we cluded to improve specificity of the PCR because have used dHPLC combined with DNA sequencing to LCGreen PLUS is known to increase the melting temper- identify 20 different F8 gene mutations in 28 HA patients ature of primers by 2–4 °C (15). For samples analyzed on we have analyzed (6). Although we have found dHPLC the LC480, the melting step was appended to the ampli- to be a sensitive scanning method, it is a low-throughput fication step; samples melted on the LightScanner were technology and has many ongoing costs. Likewise, DNA amplified on the LC480 as above, then transferred to the sequencing of the F8 gene is time-consuming and LightScanner for the melting analysis. expensive. To analyze melting data on both the LC480 and the High-resolution melting analysis represents the next LightScanner software, melting curves were normalized generation of mutation scanning technology and offers by defining regions in the pre- and postdenaturation parts considerable time and cost savings over both dHPLC and of the curve, and the value was set for the melting sequencing. This closed-tube assay is performed on am- temperature shift adjustment. Output plots are in the plicons post-PCR. In the presence of a saturating double- form of normalized temperature-shifted melting curves stranded DNA-binding dye, amplicons are slowly heated that show the decrease in fluorescence (Fl) against in- until fully denatured while the fluorescence is monitored creasing temperature, and difference curves that show the (7). Amplicons heterozygous for a sequence variant yield difference in fluorescence (⌬Fl) between the melting altered melting curves compared with normal control curves of the mutation sample and the normal control, samples. High-resolution melting analysis has recently against temperature. The difference curves provided the been tested in a variety of clinical mutation-scanning best resolution to differentiate mutation and normal sam- ⌬ applications and shown to be a sensitive and cost-effective ples, with the amplitude of the peak ( Flmax) recorded as technique (8–14). a measure of the resolution. On the basis of the observed Clinical Chemistry 53, No. 12, 2007 2213

Table 1. Summary of high-resolution melting analysis of 20 F8 gene variants. a b ⌬ c ⌬ c Nucleotide change Amino acid change Exon Amplicon size (bp) Detected Flmax (LightSscanner) Flmax (LC480) c.1172GϾA R372H 8 338 Yes 0.08 7 c.1409_1418 delCTTTA 9 284 Yes 0.14 12 c.1649CϾA R531H 11 294 Yes 0.17 15 c.1834GϾT R593C 12 230 Yes 0.10 11 c.3864_3870insA 14vd 429 Yes 0.08 10 c.4757GϾA W1567X 14viid 436 Yes 0.05 6 c.5380TϾC F1775L 16 330 No 0.03 2 c.5399GϾA R1781H 16 330 Yes 0.06 Ϫ5 c.5573CϾT S1839F 16 330 Yes 0.16 17 c.5602TϾC S1849P 17 349 Yes 0.10 12 c.6193TϾC W2046R 21 168 Yes 0.12 12 c.6317AϾC Q2087P 22 206 Yes 0.20 18 c.6449AϾT D2131V 23 250 Yes Ϫ0.05–0.02 5 c.6532CϾT R2159C 23 250 Yes 0.18 16 c.6533GϾA R2159H 23 250 Yes 0.18 16 c.6547AϾG M2164V 23 250 Yes 0.13 11 c.6682CϾT R2209X 24 249 Yes 0.25 22 c.6723 ϩ 1GϾC24e 249 Yes 0.17 18 c.6744GϾT W2229C 25 323 Yes Ϫ0.06–0.07 8 c.6901–2AϾG26e 217 No 0.04 2 a Mutation nomenclature is in accordance with the Human Genome Variation Society guidelines (numbering begins at the “A” nucleotide of the initiating “ATG” codon). b Amino acid numbering uses the classical FVIII system, which starts at the first residue of the mature peptide. c ⌬ Flmax values represent the maximum peak amplitude of the mutation sample difference curve compared to the baseline normal control. d Exon 14 is amplified as 8 overlapping amplicons 14i–14viii. e The splicing variants c.6723 ϩ 1GϾC (intron 24) and c.6901-2AϾG (intron 25) were detected in the exon 24 and 26 amplicons, respectively, which include flanking intronic regions. variation in 4 normal controls (see Figs. 1–3 in the Data duplicate) derived using the normal sample melting curve Supplement that accompanies the online version of this as a reference, which appears on the difference plot as the Technical Brief at http://www.clinchem.org/content/ baseline. Each exon 23 variant has a distinct curve profile, ⌬ vol53/issue12), we considered that a Flmax value of at indicating this technique can distinguish different vari- least 5 on the LC480, which is equivalent to 0.05 on the ants from each other, as well as from the normal sample. LightScanner, was sufficient resolution to distinguish Melting curves for these exon 23 variants on the LC480 are the mutant sample from the normal control, although the shown in Fig. 2 of the online Data Supplement, along with shape of the curve is also an important consideration and 4 controls to indicate the variability in normal samples. ⌬ Յ may indicate the presence of a variant even if Flmax The LC480 and LightScanner platforms performed 5/0.05. similarly overall, although each has strengths and weak- Analysis of the data indicated that 18 of 20 F8 gene nesses. The LC480 is a thermal cycler, so samples for mutations included in this study were detected by melt- melting analysis can be amplified on this machine, with ing analysis on both LC480 and LightScanner platforms, the melting step appended to the PCR protocol. The ⌬ with Flmax values of at least 5/0.05 (Table 1). The amplification of samples can be monitored in real time, variants F1775L (c.5380TϾC, exon 16) and c.6901–2AϾG because LCGreen PLUS fluorescence is proportional to ⌬ Ͻ [intron 25 (exon 26 amplicon)] had Flmax of 5/0.05, the amount of double-stranded DNA in the PCR, and any indicating a lower confidence in the ability of the melting samples in which amplification failed can be removed analysis to resolve the mutant sample from the normal from the melting analysis. The LightScanner performs controls (see Figs. 1 and 3 in the online Data Supplement). only the melting step, but in our hands gave greater Even when PCR products from the mutant samples were consistency between replicate samples. Both machines use mixed with products from normal DNA post-PCR, the a 96-well plate format (with the option for a 384-well resolution of these samples was still poor (see Figs. 4 and format), enabling rapid throughput of samples. 5 in the online Data Supplement). Data for the melting analysis of F8 exon 23 variants on In summary, high-resolution melting analysis is a power- the LightScanner is shown in Fig. 1. These curves are ful and cost-effective option for mutation scanning of the representative of the other samples in the study listed in F8 gene. The sensitivity of this technique is comparable to Table 1 and show the difference curve for each variant (in dHPLC, and it offers advantages in the speed and cost of 2214 Technical Briefs

Grant/funding support: We acknowledge Roche Diagnostics (New Zealand) for the use of the Roche LC480, and John Morris Scientific for the use of the Idaho Technology LightScanner. Financial disclosures: None declared. Acknowledgments: We are grateful to Scott Mead and Camp- bell Sheen for helpful comments.

References 1. Online Mendelian Inheritance in Man, OMIM (TM). Johns Hopkins University, Baltimore, MD. MIM Number: {306700}: {5/1/2007} http://www.ncbi.nlm. nih.gov/omim/ (accessed June 2007). 2. Graw J, Brackmann H-H, Oldenburg J, Schneppenheim R, Spannag M, Schwaab R. Haemophilia A: from mutation analysis to new therapies. Nat Rev Genet 2005;6:488–501. 3. Kemball-Cook G, Tuddenham EG, Wacey AI. The factor VIII structure and mutation resource site: HAMSTeRS version 4. Nucleic Acids Res 1998;26: 216–9. 4. Economou EP, Kazazian HH, Antonarakis SE. Detection of mutations in the factor VIII gene using single-stranded conformational polymorphism (SSCP). Genomics 1992;13:909–11. 5. Oldenburg J, Ivaskevicius V, Rost S, Fregin A, White K, Holinski-Feder E, et al. Evaluation of DHPLC in the analysis of hemophilia A. J Biochem Biophys Methods 2001;47:39–51. 6. Laurie AD, Sheen CR, Hanrahan V, Smith MP, George PM. The molecular aetiology of haemophilia A in a New Zealand patient group. Haemophilia 2007;13:420–7. 7. Wittwer CT, Reed GH, Gundry CN, Vandersteen JG, Pryor RJ. High-resolution genotyping by amplicon melting analysis using LCGreen. Clin Chem 2003; 49:853–60. 8. Chou L-S, Lyon E, Wittwer CT. A comparison of high-resolution melting analysis with denaturing high-performance liquid chromatography for mutation scanning: cystic fibrosis transmembrane conductance regulator gene as a model. Am J Clin Pathol 2005;124:330–8. 9. Dobrowolski SF, Ellingson C, Coyne T, Grey J, Martin R, Naylor EW, et al. Fig. 1. High-resolution melting curves for the F8 exon 23 amplicon, Mutations in the phenylalanine hydroxylase gene identified in 95 patients with phenylketonuria using novel systems of mutation scanning and specific showing variants D2131V, R2159C, R2159H, and M2164V, obtained genotyping based upon thermal melt profiles. Mol Genet Metab 2007;91: using the LightScanner. 218–27. The upper chart shows the normalized temperature-shifted melting curves [using 10. Kennerson ML, Warburton T, Nelis E, Brewer M, Polly P, De Jonghe P, et al. the default melting-temperature shift adjustment (0.05)], and the lower chart Mutation scanning the GJB1 gene with high-resolution melting analysis: shows the difference curves, derived using the normal sample as the baseline. implications for mutation scanning of genes for Charcot-Marie-Tooth dis- Data for duplicate samples for each variant is shown. For data from the LC480, ease. Clin Chem 2007;53:349–52. see Fig. 2 in the online Data Supplement. 11. Krypuy M, Newnham GM, Thomas DM, Conron M, Dobrovic A. High resolution melting analysis for the rapid and sensitive detection of mutations in clinical samples: KRAS codon 12 and 13 mutations in non-small cell lung cancer. BMC Cancer 2006;21:6:295. analysis. To support the F8 mutation-scanning screen in 12. Lonie L, Porter DE, Fraser M, Cole T, Wise C, Yates L, et al. Determination of the mutation spectrum of the EXT1/EXT2 genes in British Caucasian the clinical setting, DNA sequencing of amplicons in patients with multiple osteochondromas, and exclusion of six candidate which a putative mutation (or polymorphism) was de- genes in EXT negative cases. Hum Mutat 2006;27:1160. 13. Margraf RL, Mao R, Highsmith WE, Holtegaard LM, Wittwer CT. Mutation tected can be used to confirm and identify the variant and scanning of the RET protooncogene using high-resolution melting analysis. to rescreen any samples in which no mutations were Clin Chem 2006;52:138–41. 14. Willmore-Payne C, Holden JA, Chadwick BE, Layfield LJ. Detection of c-kit detected across all F8 exons. In this context, scanning by exons 11- and 17-activating mutations in testicular seminomas by high- high-resolution melting analysis is an ideal technology resolution melting amplicon analysis. Mod Pathol 2006;19:1164–9. 15. LightScanner Instrument Demonstration Guidelines. Idaho Technology Inc., because it is rapid, economical, and capable of detecting Salt Lake City, Utah. most mutations (90% in this study). Mutations not detected by the melting analysis screen would be identified Previously published online at DOI: 10.1373/clinchem.2007.093781 by subsequent sequencing of all F8 exons for that sample. Letters

The Origin of Circulating Free DNA many cancer cells are resistant to is a common feature, and thus some apoptosis, and to escape the immune kind of correlation ought to be found To the Editor: system proliferating cells lose the that may point to a common mecha- Almost every report on circulating ability to become apoptotic (2, 3),a nism of origin. Although researchers DNA identifies apoptosis or necrosis process that also runs counter to the have been studying the origin of cir- or both as the main source of free notion of apoptosis as the main culating DNA for more than 30 circulating DNA in serum and mechanism for generating free DNA. years, the mechanism of release still plasma. A hallmark of apoptosis is Necrosis, on the other hand, pro- has to be elucidated. A reasonable DNA degradation, in which chromo- duces large DNA fragments. possibility is that more than one somal DNA is 1st cleaved into large Even with no cells dying, the DNA mechanism is involved; if so, the fragments (50–300 kb) and subse- concentration in culture medium in- variables influencing the relative quently into multiples of nucleoso- creases in proportion to the prolifer- contributions and the interactions be- mal units (180–200 bp) (1). This lad- ation of cultured cancer cells. Human tween the mechanisms must be understood for optimal utilization of der pattern is also visible after lymphocytes have also been ob- this very valuable, minimally inva- electrophoresis of circulating DNA served to actively release double- sive biomarker. Thus more work and is frequently considered to be stranded DNA into culture medium to a certain concentration (5), irre- needs to be done to determine the evidence that apoptosis may be the spective of incubation time. A similar mechanism(s) of release and clear- source of the observed DNA frag- observation was made in experi- ance as well as the significance of ments in plasma (2, 3). It has been ments with frog auricles: DNA was circulating DNA in the body. shown, however, that the character- released to the same concentrations istic ladder pattern can also be ob- during successive transfer of auricles served for actively released DNA (3). Grant/funding support: None declared. to fresh medium, purified frog DNA The contents of apoptotic cells Financial disclosures: None declared. did not inhibit release of DNA, and are rapidly ingested by profes- Acknowledgments: We thank Dr. Phiyani damaged auricles did not yield more sional phagocytes or neighboring Lebea for proofreading the final DNA into the medium (5). There- cells through mechanisms that are manuscript. fore, quantities of released DNA are not fully understood (4), and the similar regardless of the proportion DNA is consequently completely References of dying cells (5), with the exception 1. Nagata S, Nagase H, Kawane K, Mukae N, digested by DNase II in lysosomes applicable for cancer cells, which can Fukuyama H. Degradation of chromosomal DNA (1). Thus the possibility exists that during apoptosis. Cell Death Differ 2003;10:108– release more DNA than normal cells 16. DNA fragments released by apopto- (2). Evidence for preferential release 2. Anker P, Mulcahy H, Chen XQ, Stroun M. Detec- sis are removed before appearing in of DNA was found by comparing the tion of circulating tumour DNA in the blood the circulation. If this engulfment of (plasma/serum) of cancer patients. Cancer Me- proportion of Alu repeat sequences tastasis Rev 1999;18:65–73. apoptotic bodies is impaired or cell to ␤-globin in serum and lympho- 3. Stroun M, Lyautey J, Lederrey C, Olson-Sand A, death is increased enough to produce cytes (5). We thus conclude that ap- Anker P. About the possible origin and mechanism substantial amounts of circulating of circulating DNA: apoptosis and active DNA re- optosis and necrosis are not the main lease. Clin Chim Acta 2001;313:139–42. DNA, inflammation would definitely source of circulating DNA, although 4. Viorritto IC, Nikolov NP, Siegel RM. Autoimmu- be a problem and autoimmunity they may contribute. DNA released nity versus tolerance: can dying cells tip the would occur frequently in cancer balance? Clin Immunol 2007;122:125–34. by living cells is a viable alternative. 5. Stroun M, Anker P. Circulating DNA in higher and other conditions involving in- The mechanism of DNA clearance organisms cancer detection brings back to life creased circulating DNA (1, 4). from plasma is poorly understood. an ignored phenomenon. Cell Mol Biol (Noisy-le- grand) 2005;51:767–74. Radiotherapy, chemotherapy, and Increased amounts of circulating other cancer treatments cause cell DNA in the blood of patients may Maniesh van der Vaart death by apoptosis, and less circulat- reflect disturbance of the equilibrium Piet J. Pretorius* ing DNA is found in cancer patients between the release of DNA by liv- after treatment than before treating cells and the clearance of DNA. School of Biochemistry ment, possibly because of the inhibi- The low concentrations of circulating North-West University tory effect of treatment on the prolif- DNA in healthy individuals may re- Potchefstroom Campus eration of cancer cells. Furthermore, flect a lower rate of DNA release by Potchefstroom, South Africa in the early stages of cancer, when cells or a rapid removal of DNA by little cell death seems to occur, circu- the optimal functioning of clearance lating DNA may already be present mechanisms, and when this equilib- * Address correspondence to this au- in higher than normal concentra- rium is disturbed, the amount of thor at: School of Biochemistry, North- West University, Potchefstroom Campus, tions. As the cancer burden in- circulating DNA increases. 11 Hoffmann St., Potchefstroom 2531, creases, so does the rate of cell death Circulating DNA can be found in a South Africa. Fax 27(0)18-299-2316; and the amount of proliferating can- variety of conditions, and although e-mail [email protected]. cer cells, with a concomitant increase these conditions are unrelated, the DOI: 10.1373/clinchem.2007.092734 in circulating DNA. In addition, presence of circulating nucleic acids

Clinical Chemistry 53, No. 12, 2007 2215 2216 Letters

A Receptor-Mediated Mechanism to onstrated in FcRn-deficient mice that mutagenesis studies have been per- Support Clinical Observation of show dramatically reduced circulat- formed to investigate the exact inter- Altered Albumin Variants ing concentrations of IgG (3). Re- action site on domain III, but the cently, the same phenomenon was abnormal HSA variants described To the Editor: shown to be valid for the long-lived lack domain III or have structural Dolcini et al. (1) recently reported in HSA molecule (4). An FcRn-deficient alterations in domain III, which is this journal a novel splice-site muta- strain degraded albumin twice as absolutely crucial for the FcRn inter- tion (denoted Bartin) that causes de- fast as the wild-type strain. This find- action. These characteristics surely will affect the pH-dependent binding letion of exon 11 in the human serum ing indicates that FcRn rescues as to FcRn and rescue of the protein albumin (HSA) gene, ALB. In spite of much albumin from degradation as from degradation. These novel and their uncertain relevance to patho- is produced by the liver. Interest- ingly, individuals with the rare hu- important findings should be taken physiology of diseases, differences in man syndrome familial hypercata- into consideration, and they may HSA sequence have interesting corre- bolic hypoproteinemia show marked contribute to reevaluation of existing lations with functional properties and decreases of both endogenous IgG data and explain clinical observations stability. Bisalbuminemia (or alloalbu- and HSA (4). This clinical observa- regarding altered HSA variants. minemia) is a rare inherited or action was an unresolved question for quired condition characterized by the decades, until Wani and colleagues occurrence of 2 circulating compo- (4) elegantly showed that the syn- nents that are observed, typically, dur- drome is characterized by deficient Grant/funding support: This work ing routine clinical electrophoresis or FcRn expression due to a mutation was supported by grants from the in genetic surveys. affecting ␤ -microglobulin, which to- Steering board for Research in Molec- Over the past decades several 2 gether with the so-called heavy chain ular Biology, Biotechnology and Bio- cases of genetic polymorphisms, de- constitutes the heterodimeric receptor. informatics (EMBIO) at the University of Oslo and The Norwegian Cancer scribed in peer-reviewed papers and Molecular studies reveal that IgG Society. listed in the analbuminemia register as well as HSA bind FcRn heavy Financial disclosures: None declared. (2), have been characterized as rep- chain in a strictly pH-dependent resenting site-specific mutations, fashion, binding at acidic pH and splice-site mutations, or frame-shift releasing at physiological pH (4, 5). mutations. The deletion reported by References The binding sites for the 2 ligands are 1. Dolcini L, Caridi G, Dagnino M, Sala A, Gokce S, Dolcini et al. (1) would give rise to a located distally on the ␣2-domain of Sokucu S, et al. Analbuminemia produced by a C-terminally truncated HSA variant the heavy chain, a position that al- novel splicing mutation. Clin Chem 2007;53: of 410 amino acids rather than the 1549–52. lows simultaneous binding, and the 2. The Albumin Website. http://www.albumin.org 585 amino acids of the wild-type binding of IgG does not affect bind- (accessed August 2007). protein, and would almost com- ing of HSA (5). In the proposed 3. Roopenian DC, Akilesh S. FcRn: the neonatal Fc pletely delete domain III of HSA. regulatory mechanism (3, 4), FcRns, receptor comes of age. Nat Rev Immunol 2007; Furthermore, the authors discuss the 7:715–25. predominantly localized intracellu- 4. Anderson CL, Chaudhury C, Kim J, Bronson CL, impact of the C-terminal end of HSA larly in endothelial cells that cover Wani MA, Mohanty S. Perspective—FcRn trans- on in vivo stability. Interestingly, the bloodstream, capture circulating ports albumin: relevance to immunology and several other reported mutations also medicine. Trends Immunol 2006;27:343–8. IgG and HSA taken up by fluid- 5. Andersen JT, Dee Qian J, Sandlie I. The con- generate altered C-terminal HSA phase pinocytosis, after the ligands served histidine 166 residue of the human neo- variants, resulting from truncation or enter acidified endosomal compart- natal Fc receptor heavy chain is critical for the pH-dependent binding to albumin. Eur J Immunol elongation as well as single muta- ments. The lower pH herein facili- 2006;36:3044–51. tions in the primary sequence. All tates binding to FcRn, and the inter- identified HSA variants with such actions trigger recycling back to the Jan Terje Andersen alteration are present in serum of cell surface and release at pH 7.2–7.4. Inger Sandlie heterozygous carrier individuals in IgG and HSA that escape receptor the range of 2%–30% of total HSA, a binding go to lysosomal degrada- Department of Molecular fact that underscores the significance tion. An alternative cooperative path- Biosciences, University of Oslo of the C-terminal domain III on way may be bidirectional transport Oslo, Norway stability. between the bloodstream and the ex- In light of these observations, it is travascular space guided by the same noteworthy that a widely expressed pH-dependent mechanism. Address correspondence to the authors receptor known as the neonatal Fc In light of the findings of Dolcini et at: Department of Molecular Biosciences, University of Oslo, P.O. Box 1041, 0316 receptor (FcRn) was recently found al. (1), it is noteworthy that this Oslo, Norway. Fax 47 22 85 40 61; e-mail to bind HSA (3). This receptor is FcRn-mediated recycling mechanism [email protected] or inger.sandlie@imbv. known to be the major homeostatic is completely dependent on binding uio.no. regulator of the serum half-life (ap- to domain III of HSA. Unfortunately, DOI: 10.1373/clinchem.2007.097071 proximately 20 days) of IgG, as dem- no cocrystal structure or site-directed Clinical Chemistry 53, No. 12, 2007 2217

Stability of Plasma Homocysteine, S-Adenosylmethionine, and S- Adenosylhomocysteine in EDTA, Acidic Citrate, and Primavette™ Collection Tubes

To the Editor: After blood collection, homocysteine (Hcy) is generated in blood cells and is continuously released into the plasma. Stabilizing Hcy in blood samples requires immediate sample centrifugation and plasma separation from blood cells, sample cooling, or the use of special collection tubes (1, 2). To investigate the stability of Hcy and its precursors S-adenosyl-Hcy (SAH) and S-adenosyl-methionine (SAM), we collected fasting blood samples from healthy individuals (n ϭ 8, age 25–46 years) and renal patients (n ϭ 9, GFR 11–48 mL/min, 78–90 years). Patients were recruited from the Department of Internal Medicine IV-Nephrology and Hy- pertension of the Saarland Univer- sity Hospital. Controls were selected among hospital employees. The local ethics committee approved the usage of blood samples from patients, and all participants gave informed consent. Samples were collected into EDTA, acidic citrate, and Pri- mavette™ tubes. The samples were incubated at 4 °C or at room temperature for 0, 2, 6, and 24 h and were centrifuged afterward. A total of 357 samples were analyzed for Hcy by HPLC (Immundiagnostik) and by fluorescence polarization immunoassay (FPIA; Abbott). Plasma SAM and SAH were determined by a modified liquid chromatography–tandem mass spectrometry method according to Gellekink et al. (3) in 210 samples from a representative subgroup with 10 individuals. Baseline Hcy values (EDTA) ranged from 5 to 16 ␮mol/L in healthy individuals and from 9 to 65 ␮mol/L in Fig. 1. Geometric means of plasma Hcy (n ϭ 17) obtained by HPLC and FPIA and geometric ϭ renal patients. Baseline plasma SAM means of plasma SAM and SAH (n 10) in dependency on incubation time of the blood samples (0, 2, 6, and 24 h), temperature (room temperature and 4 °C), and collection tubes (F, EDTA; and SAH (EDTA) were determined ᭜, acidic citrate; f, Primavette). from 68 to 142 and 9 to 16 nmol/L in The concentrations after 2-, 6-, and 24-h incubation were compared with the corresponding baseline values healthy individuals and from 148 to by applying the Wilcoxon test for paired samples with Bonferoni correction. *, P Ͻ0.013. 392 and 31 to 361 nmol/L in renal patients. The individual increase of higher (P Ͻ0.001) in healthy individu- Fig. 1 presents the geometric Hcy in EDTA samples at room tem- als (8.7–23.6 ␮mol/L) than in renal means of plasma Hcy obtained by perature after 24 h was significantly patients (0.1–9.5 ␮mol/L). HPLC and FPIA and plasma SAM 2218 Letters

and SAH after blood sample incuba- (r ϭϪ0.647; P ϭ 0.002) but not in Ulrich Hu¨ bner1 tion over a period of 0, 2, 6, and 24 h EDTA samples (r ϭ 0.067). Hcy is Heike Schorr1 at room temperature or at 4 °C. In the generated in erythrocytes from its Rudolf Eckert2 Primavette, Hcy obtained by HPLC precursor SAH by catalysis of SAH Ju¨ rgen Geisel1 decreased only slightly (Ϫ4.8%) after hydrolase. Therefore, the inhibition Wolfgang Herrmann1* 6-h incubation at room temperature of SAH hydrolase activity causes an 1 and returned to the baseline value increase of SAH and no or low in- Department of Clinical Chemistry after 24 h. At 4 °C a significant decrease of Hcy. In contrast, in EDTA and Laboratory Medicine crease of Hcy (6.9%) after 24-h incu- samples the increase of SAH is low Faculty of Medicine ϭ bation was found (P 0.003). Apply- because SAH is metabolized to Hcy. University of Saarland ing FPIA for the determination of Furthermore, a leakage of SAH from Homburg, Germany Hcy, no significant changes of Hcy erythrocytes into the plasma might oc- 2 compared with the baseline values cur because cellular SAH concentra- Department of were found. However, the compari- tions are approximately 10-fold higher Geriatric Rehabilitation son between baseline Hcy deter- than in plasma (4). In blood samples St. Ingbert Hospital St. Ingbert, Germany mined by FPIA and HPLC revealed Hcy can be stabilized by the addition that Hcy values obtained by FPIA of an SAH hydrolase inhibitor (5). were significantly higher (median In conclusion, our results indicate difference 11%) than values obtained * Address correspondence to this au- that the stabilizing effect of Pri- thor at: Department of Clinical Chemistry by HPLC (P Ͻ0.001). In EDTA and mavette and acidic citrate on Hcy is and Laboratory Medicine, Central Labo- acidic citrate plasma we observed no due to the inhibition of SAH hydro- ratory, University Hospital of the Saar- significant difference between Hcy land, Bldg. 57, D-66421 Homburg/Saar, lase activity. The inhibition of SAH obtained by FPIA and HPLC. The Germany. Fax 49 68411630703; e-mail hydrolase is more efficient in Pri- [email protected]. positive bias seen with Primavette mavette than in acidic citrate tubes. tubes was likely due to interferences DOI: 10.1373/clinchem.2007.093930 with the FPIA method by its propri- However, in Primavette samples Hcy etary components, which are kept obtained by FPIA was approximately secret by the manufacturer. 11% higher than Hcy obtained by In acidic citrate samples Hcy in- HPLC. creased slowly at room temperature, From Syndrome to Spectrum: What reaching the level of significance af- Evolution Suggests about the Status ter 6 h (FPIA) and 24 h (HPLC), of the Metabolic Syndrome respectively. At 4 °C Hcy was stable Grant/funding support: None declared. Financial disclosures: None declared. over a 24-h period. In EDTA tubes a To the Editor: strong increase of Hcy was observed In 2005 the American Diabetes Asso- that was markedly decreased at 4 °C. ciation and European Association for In the Primavette, SAM was stable at References the Study of Diabetes questioned the room temperature and at 4 °C. At 1. Willems HP, den Heijer M, Lindemans J, Beren- value of the metabolic syndrome as a room temperature SAM decreased in schot HW, Gerrits WB, Bos GM, et al. Measure- ment of total homocysteine concentrations in diagnostic category (1). In the same EDTA and to a smaller extent in acidic citrate- and EDTA-containing tubes by dif- year, Reaven delivered a noted “obit- acidic citrate tubes. This decrease ferent methods. Clin Chem 2004;50:1881–3. uary” for the metabolic syndrome in was clearly decelerated at 4 °C. At 2. Bisse´ E, Epting T, Ko¨gel G, Obeid R, Gempel K, this journal, while defending the in- Huzly D, et al. Clinical validation of a new blood room temperature SAH increased in collection tube for the accuracy of total homo- sulin resistance syndrome as a patho- all 3 collection tubes and reached the cysteine measurement by different methods. physiologic entity (2). But despite highest values in the Primavette after Clin Biochem 2007;40:739–43. these broadsides, the concept of the 24 h. The increase was attenuated at 3. Gellekink H, van Oppenraaij-Emmerzaal D, van metabolic syndrome is attracting Rooij A, Struys EA, den Heijer M, Blom HJ. 4 °C. Stable-isotope dilution liquid chromatography- continued support. PubMed lists In Primavette and acidic citrate electrospray injection tandem mass spectrome- 1775 articles with “metabolic syn- samples we observed 7- and 3-fold try method for fast, selective measurement of drome” in the title or abstract pub- S-adenosylmethionine and S-adenosylhomocys- increases of plasma SAH after 24-h teine in plasma. Clin Chem 2005;51:1487–92. lished during 2005, 2566 during 2006, incubation at room temperature, 4. Becker A, Henry RM, Kostense PJ, Jakobs C, and 1689 for the first 6 months of whereas Hcy was stable and in- Teerlink T, Zweegman S, et al. Plasma homocys- 2007. Clearly, rumors of the syn- creased by 20%, respectively (geo- teine and S-adenosylmethionine in erythrocytes drome’s demise have been greatly as determinants of carotid intima-media thick- metric means). In EDTA samples ness: different effects in diabetic and non-dia- exaggerated. SAH and Hcy increased 1.8-fold. In- betic individuals. The Hoorn study. Atherosclero- There are several reasons why the terestingly, the increase of SAH after sis 2003;169:323–30. metabolic syndrome continues to 24 h correlated negatively with the 5. Martin I, Gibert MJ, Vila M, Pintos C, Obrador A, thrive as a concept. First, there is a Malo O. Stabilization of blood homocysteine in increase of Hcy after 24 h in Pri- an epidemiological setting. Eur J Cancer Prev widespread sense that its compo- mavette and acidic citrate samples 2001;10:473–6. nents are sufficiently coupled mech- Clinical Chemistry 53, No. 12, 2007 2219

anistically to suggest some kind of ical subsystems. And hence too, the based prediction models grow in disease entity. Second, the term and number of pathophysiologic factors sophistication, medicine will need to concept retain utility simply for want implicated in the metabolic syn- move beyond the era of the all- of a fuller description of the under- drome, such as hepatic, hemody- embracing syndrome and adopt a lying pathology. Third, the concept is namic, endothelial, and inflamma- more systematic approach that better self-sustaining, in that the existing tory elements, continues to grow. reflects the nature of biological sys- body of research attracts further re- Human cognition is serial and tems and their evolutionary history. search and comment. Fourth, indus- weighted toward simplicity; cell and try promotes the syndrome in sym- organismal physiology is parallel posia and review articles, alongside and complex. The error in our con- Grant/funding support: None declared. related concepts such as the “cardio- ceptualization of the metabolic syn- Financial disclosures: None declared. metabolic syndrome” and “cardio- drome lies in overintegration, for in Acknowledgments: The author is grate- metabolic risk”. this setting it is to be expected that a ful to an unnamed reviewer for helpful But perhaps most interestingly, the spectrum or range of pathology comments on the manuscript. metabolic syndrome thrives because should exist rather than a single car- the notion of a distinct pathophysio- dinal disease pathway. Perhaps at References logic entity indulges the intuitive es- the top level it is therefore preferable 1. Kahn R, Buse J, Ferrannini E, Stern M. The metabolic syndrome: time for a critical apprais- sentialism and predilections for sim- to envisage not a metabolic syn- al: joint statement from the American Diabetes plification and category formation drome, but a cardiovascular, renal, Association and the European Association for the Study of Diabetes. Diabetes Care 2005;28: that inhere in human cognition as and metabolic pathophysiologic ma- 2289–304. evolutionary legacies (3).ToHomo trix (CRM matrix), within which a 2. Reaven GM. The metabolic syndrome: requies- sapiens, a unified syndrome has gut range of related benign and disease cat in pace. Clin Chem 2005;51:931–8. appeal and even mystique—as exem- states may develop. 3. Pinker S. The Blank Slate. London: Penguin, plified by the allure of the phrase The characteristic range of disease 2002. 4. Reaven GM. Syndrome X. Blood Press Suppl “syndrome X” (4). A closely con- states that emerge within this matrix 1992;4:13–6. nected nosologic issue that reflects might be described as CRM spec- 5. Monod, J. Chance and Necessity. London: Pen- similar inclinations is the tendency to trum disorders. The insulin resis- guin, 1997. dichotomize continuous variables tance syndrome of Reaven might into crude categories of disease and prove to be one such disorder; alter- Alastair Matheson health. natively, several distinct metabolic It is possible, however, that no syndromes, or defined axes of varia- 5 Highcroft Rd. consensus will ever be achieved on a tion within a pathologic continuum, London N19 3AQ, United Kingdom definition for the metabolic syn- might be characterized. That is a Fax 44-20-7272-0282 drome as a diagnostic category, and matter for research. But as research E-mail [email protected]. that no disease entity will be conclu- proceeds, its correct task might be sively delimited. This lack of resolu- framed not as one of unification, but DOI: 10.1373/clinchem.2007.096321 tion, in a different sense, is also a of taxonomy, whose goal is to elabo- matter of our evolutionary legacy. rate the nature and relationships of The systems that transact the busi- the full range of CRM spectrum dis- ness of cells and regulate human orders and identify appropriate in- physiology evolved not with rational terventions for each. Effect of Corticosteroid Therapy on foresight, but through bricolage, the Such an approach might also im- Low-Molecular–Weight Protein selective but ramshackle accretion of pact on the assessment of risk. One Markers of Kidney Function countless innovations of contingent might, for instance, conceive of a origin (5). Consequently the regula- CRM risk profile with an associated To the Editor: ␤ tory systems of higher organisms are scoring system. Serum cystatin C, 2-microglobulin, highly complex, interconnected, and The concepts of a CRM matrix and and ␤-trace protein are endogenous multifarious; riven with redundan- CRM spectrum disorders run counter markers of glomerular filtration rate cies and lacunae; and resistant to to the instinct to overintegrate, but (GFR). Cystatin C, in particular, is a systematization. allow instead for a family of disor- promising alternative to creatinine This form of complexity is the fun- ders or range of pathology to be for the detection of incipient renal damental reason why it is so difficult comprehended within a single ana- failure. However, corticosteroids af- to define the metabolic syndrome, lytic perspective. These concepts also fect the extrarenal metabolism of cys- either as disease entity or diagnostic provide a more realistic frame- tatin C, which limits the use of cys- category. It is also why the metabolic work for ongoing research. As tatin C as a marker of GFR in a syndrome cannot be adequately de- population genetics and molecular variety of clinical settings. Low-mo- scribed on any single level, such as physiology advance and a more me- lecular–weight (LMW) ␤-trace pro- that of genetics, molecular biology, ticulous parsing of disease states be- tein might be a useful alternative in intercellular signaling, or physiolog- comes possible, and as computer- this respect. The present study set 2220 Letters

Table 1. Factors influencing the serum concentrations of the LMW proteins.a

1/Cystatin C, L/mg 1/ß2-microglobulin, L/mg 1/ß-Trace protein, L/mg Whole model r2 0.72 0.67 0.63 b inulin clearance, mL ⅐ minϪ1 ⅐ (1.73 m2)Ϫ1 0.0093, 0.0081 to 0.0104 0.0051, 0.0040 to 0.0061 0.0110, 0.0087 to 0.0133 P value Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 b prednisone dose, mg/m2/d Ϫ0.0023, Ϫ0.0045 to Ϫ0.0001 0.0065, 0.0045 to 0.0085 0.0114, 0.0069 to 0.0159 P value 0.0445 Ͻ0.0001 Ͻ0.0001 a Multiple linear regression analysis between the reciprocal of the respective serum concentrations, inulin clearance, and corticosteroid dose.

out to compare the effect of cortico- highly significant positive relation- opposite was shown in children who steroid therapy on the serum concen- ship (Table 1). There was a strong did not receive glucocorticoids (2).In ␤ trations of cystatin C, 2-microglobu- positive relationship between pred- another study, however, Risch et al. lin, and ␤-trace protein. nisone dose and the reciprocals of (3) followed 6 patients with sub- ␤ ␤ We studied a group of 108 children both 2-microglobulin and -trace arachnoid hemorrhage treated with being treated or followed for malig- protein. Cystatin C showed a weak high-dose methylprednisolon, and nancy (n ϭ 41) or renal disease (n ϭ but statistically significant negative did not find a significant change in 67). In the former group 14 patients relationship with inulin clearance. serum ␤-trace protein concentration. (34%) were treated with glucocorti- The paired analysis revealed that Their results may have been due to coids, in the latter 18 (27%). We com- corticosteroid therapy significantly de- unrecognized changes in GFR or ␤ ␤ pared single-injection inulin clear- creased mean (SD) 2-microglobulin leakage of -trace protein through ance studies in 76 patients not from 1.67 (0.46) mg/L to 1.19 (0.59) the blood-brain barrier. receiving steroids with 32 in patients mg/L (mean difference Ϫ0.48 mg/L, We previously found in paired receiving corticosteroid treatment 95% CI Ϫ0.76 to Ϫ0.21, P ϭ 0.002 analysis of children with steroid-sen- (median dose 33.0 mg prednisone- paired t-test) and ␤-trace protein from sitive nephrotic syndrome that cysta- equivalent per m2 body surface area 0.80 (0.23) mg/L to 0.54 (0.15) mg/L tin C concentrations were unchanged per day, range 1.2–70.4). Mean (SD) (mean difference Ϫ0.26 mg/L, 95% CI on and off corticosteroids (4), a find- age was 9.7 (5.8) years, mean (SD) Ϫ0.37 to Ϫ0.15, P ϭ 0.0001). Cystatin C ing that may be attributable to a Ϫ GFR 92.8 (34.6) mL ⅐ min 1 ⅐ (1.73 concentrations, by contrast, did not parallel increase in GFR that ob- Ϫ m2) 1. Patients included in the study change significantly: 0.862 (0.144) scured increased cystatin C produc- had to be on corticosteroid therapy mg/L without vs 0.840 (0.210) mg/L tion. This theory is supported by the for at least 5 days or had corticoste- with corticosteroids; P ϭ 0.76. Mean slightly lower ratio between esti- roids discontinued for at least 10 (SD) inulin clearance tended to be mated and measured GFR during days. higher with corticosteroids, 127 (25) corticosteroid therapy. Chronic corti- Ϫ Ϫ In 16 patients, a paired analysis mL ⅐ min 1 ⅐ (1.73 m2) 1 vs 118 (25) costeroid administration increases Ϫ Ϫ was performed before and during mL ⅐ min 1 ⅐ (1.73 m2) 1; P ϭ 0.14. The GFR by causing vasodilation of glo- high-dose corticosteroid therapy ratio between a cystatin C-based GFR merular resistance vessels. Although with a median (range) dose of 36.3 estimate and measured inulin clear- this mechanism is maintained in mg/m2 (11.5 to 61.4 mg/m2). This ance tended to be lower during corti- chronic renal failure, it may be analysis included 10 children in the costeroid treatment, but this difference blunted in acute renal injury, possi- reinduction protocol of acute lym- was not statistically significant [mean bly explaining why this phenomenon phoblastic leukemia (ALL-9 protocol difference Ϫ0.030 (CI Ϫ0.133 to 0.074); was not observed in patients treated of the Dutch Childhood Oncology P ϭ 0.54]. for renal allograft rejection (5). Group) and 6 with nephrotic syn- Our data confirm a dose-depen- In conclusion, glucocorticoid ther- ␤ drome. The study was approved by dent decrease in serum 2-micro- apy leads to a dose-dependent un- local and national ethics committees, globulin and a weak but significant derestimation of GFR calculations and written informed consent was increase in cystatin C during cortico- based on cystatin C and overestima- ␤ obtained from the patients and/or steroid treatment. Corticosteroid tion of those based on 2-micro- their guardians. therapy was associated with a strong globulin and ␤-trace protein. In The markers were measured by par- dose-dependent decrease in ␤-trace patients receiving corticosteroids, ticle-enhanced immunonephelometry protein in both the cross-sectional ␤-trace protein offers no advantage on a Behring Nephelometer II (DADE and the paired analyses. Indirect ev- over cystatin C as a marker of GFR. Behring, Marburg, Germany). idence to support these findings As expected for endogenous GFR comes from Po¨geetal.(1), who stud- markers, multiple linear regression ied steroid-treated transplant recipi- Grant/funding support: The study was analysis between the reciprocals of ents and found a stronger increase in supported by a grant from Madeleine the LMW protein serum concentra- cystatin C than in ␤-trace protein Schickedanz Kinder-Krebsstiftung. The tions, and inulin clearance showed a with decreasing GFR, although the immunonephelometric assays were a Clinical Chemistry 53, No. 12, 2007 2221

kind gift from Dade Behring (Marburg, 3 Clinical Epidemiology and carried out at 60 °C for a 1-h period Germany). Biostatistics, VU Medical Center in a moderately alkaline medium Financial disclosures: A.B. received Amsterdam, The Netherlands (pH ϭ 9.5) in the presence of ex- honoraria from Dade Behring (Mar- cess reagent. Moreover, fast oxi- burg, Germany) and DAKO (Glostrup, 4 Department of Clinical dation or degradation of thiols, as Denmark). Biochemistry, Bonn University well as of other species participat- Acknowledgments: We are indebted to Medical Center ing in the reaction, may take place the patients and their parents who Bonn, Germany under these conditions. To circum- agreed to participate in the study. This vent these drawbacks a number of study would have been impossible new fluorogenic reagents possessing without the diligent work of the pedi- * Address correspondence to this au- the benzofurazan structure have atric nursing staff of VU Medical Cen- thor at: Department of Pediatrics, Vrije been proposed (4), but all require ter performing the inulin clearance Universiteit Medical Center, De Boele- studies. We also wish to thank Ma- laan 1117, NL-1081 HV Amsterdam. Fax heating and/or long reaction time. reike Reichelt and Jennifer Roos, who 31-20–4440849; e-mail Bokenkamp@ All of the molecules used in ami- did the immunonephelometric mea- VUmc.nl. nothiol assays are characterized by the presence of 2 substituents in po- surements at Bonn University Hospital DOI: 10.1373/clinchem.2007.094946 and the technicians, who did the inulin sitions 4 and 7 of the 2,1,3-benz- determinations at VU Medical Center. oxadiazole skeleton: the leaving Ingrid Metgod was involved in the group (halogen) and a 2nd group logistics of the study. Lyonne van Ros- whose function is to increase both sum is kindly acknowledged for her Ammonium 5-Bromo-7-fluorobenzo-2- reactivity and solubility in water. In advice during the introduction of the oxa-1,3-diazole-4-sulphonate: the search for new fluorogenic re- single injection inulin clearance tech- A New Fluorogenic Reagent for the agents with benzofurazan structure nique, and Netteke Schouten is recog- along with an additional reactivity- Determination of Aminothiols by HPLC nized for her support and comments enhancing substituent to improve re- on the study protocol. activity, we synthesized ammonium To the Editor: 5-bromo-7-fluorobenzo-2-oxa-1,3- References The measurement of aminothiols such diazole-4-sulfonate (SBD-BF) and 1. Po¨ge U, Gerhardt TM, Stoffel-Wagner B, Palmedo as cysteine (Cys), cysteinyl-glycine H, Klehr HU, Sauerbruch T, et al. beta-Trace characterized this reagent by means protein is an alternative marker for glomerular (Cys-Gly), homocysteine (Hcy), and of spectroscopic and analytical data. filtration rate in renal transplantation patients. glutathione (GSH), as well as the The results of a preliminary kinetic Clin Chem 2005;51:1531–3. corresponding disulfides, has gained 2. Bokenkamp A, Franke I, Schlieber M, Duker G, study, carried out with Cys and Hcy Schmitt J, Buderus S, et al. Beta-trace protein: a high interest within the biomedical nucleophiles by means of the ultra- marker of kidney function in children: “Original community because such molecules violet-visible detection technique, in- research communication-clinical investigation”. Clin Biochem 2007;40:969–75. are important biomarkers for a wide dicated that SBD-BF reacts about 3 3. Risch L, Saely C, Reist U, Reist K, Hefti M, Huber range of diseases. In particular, in- times faster at 25 °C than SBD-F at AR. Course of glomerular filtration rate markers creased total plasma Hcy is now 60 °C. SBD-BF was therefore more in patients receiving high-dose glucocorticoids considered a risk factor for cardio- following subarachnoidal hemorrhage. Clin Chim reactive than SBD-F, and it reacted Acta 2005;360:205–7. vascular disorders (1) as well as with thiols under milder conditions 4. Bokenkamp A, van Wijk JA, Lentze MJ, Stoffel- other degenerative conditions. Al- and during a time period of about Wagner B. Effect of corticosteroid therapy on though various methods for the serum cystatin C and beta2-microglobulin con- 15 min. centrations. Clin Chem 2002;48:1123–6. measurement of aminothiols are This result was supported by time- 5. Risch L, Herklotz R, Blumberg A, Huber AR. available, HPLC coupled with flu- course studies, carried out with re- Effects of glucocorticoid immunosuppression on orometric detection is one of the versed-phase HPLC, on the deriv- serum cystatin C concentrations in renal transplant patients. Clin Chem 2001;47:2055–9. most suitable techniques for determi- atization reactions of SBD-BF with nation of minute amounts of thiols the aminothiols used in the present Arend Bo¨kenkamp1* (2). Most methods require precol- work. Moreover, the spectra of the Ce`leste A.R.C. Laarman1 umn derivatization, and although derivatives of Cys with SBD-F and Katja I. Braam1 different types of fluorogenic re- SBD-BF, recorded with a Perkin- Joanna A.E. van Wijk1 agents for thiols have been proposed, Elmer MPF 44 A spectrofluorimeter, Wijnanda A. Kors1 the most commonly used and sensi- were practically superimposable, sug- Marijke Kool1 tive reagent is the commercially gesting that our new reagent SBD-BF Janneke de Valk1 available ammonium 7-fluorobenzo- is a suitable precolumn derivatization Anna A. Bouman2 2-oxa-1,3-diazole-4-sulfonate (SBD-F) reagent for reversed-phase HPLC flu- Marieke D. Spreeuwenberg3 (3). orometric determination of thiols. Birgit Stoffel-Wagner4 With this method the completion Calibration curves for the deriva- of the reaction takes a very long time tives of Cys, Cys-Gly, Hcy, and GSH Departments of 1 Pediatrics, and the conditions required are quite with SBD-BF, obtained in concentra- 2 Clinical Chemistry, and strict, i.e., derivatization must be tion ranges from 0.125 to 50 ␮mol/L 2222 Letters

(5), may be attributable to oxygen exposure. In conclusion, SBD-BF reacts with thiols under mild conditions, at 25 °C for a time period of approximately 15 min, and seems to be a very promising fluorogenic reagent for amino- thiol derivatization.

Grant/funding support: Financial support was provided by Ministero dell’ Istruzione, dell’Universita` e della Ri- cerca (PRIN2005). Financial disclosures: None declared.

References 1. Mangoni AA, Jackson SHD. Homocysteine and cardiovascular disease: current evidence and future prospects. Am J Med 2002;112: 555–64. 2. Ducros V, Demuth K, Sauvant MP, Quillard M, Causse´ E, Candito M, et al. Methods for homocysteine analysis and biological relevance of the results. J Chromatogr B 2002; 781:207–26. 3. Imai K, Toyo’oka T, Watanabe Y. A novel fluorogenic reagent for thiols: ammonium 7-fluoro- Fig. 1. Chromatograms of (A): derivatized mixture of Cys, Cys-Gly, Hcy, and GSH, each 5 ␮mol/L; benzo-2-oxa-1,3-diazole-4-sulfonate. Anal Biochem and (B): derivatized human plasma. 1983;128:471–3. Conditions: Hypersil ODS RP C18 200 ϫ 2.1 mm (i.d.) column (5 ␮m), thermostated at 23 °C; gradient elution 4. Okabe K, Wada R, Ohno K, Uchiyama S, Santa T, running from 98% eluent A (0.1 mL/L aqueous trifluoroacetic acid) and 2% eluent B (acetonitrile) to 4% B Imai K. Development of hydrophilic fluorogenic (5 min), 6% B (10 min), flow rate 0.45 mL/min; total analysis time 16 min. Fluorescence intensities measured derivatization reagents for thiols: 4-(N-acetyl-ami- with excitation at 385 nm and emission at 515 nm. nosulfonyl)-7-fluoro-2,1,3-benzoxadiazole and 4-(N- trichloroacetylaminosulfonyl)-7-fluoro-2,1,3- benzoxadiazole. J. Chromatogr A 2002;982: 111–8. 5. Krijt J, Vackova M, Kozˇich V. Measurement in borate buffer, exhibited excellent ter with EDTA 1 mmol/L) was then of homocysteine and other aminothiols in linearity between peak areas and added and the sample was centri- plasma: advantages of using tris(2-carboxy- concentrations over the entire range, fuged for 6 min at 14.5g. An aliquot ethyl)phosphine as reductant compared with tri- n-butylphosphine. Clin Chem 2001;47:1821–8. with r2 values of 0.9992, 0.9993, of the resulting solution (100 ␮L) 0.9989, and 0.9991, respectively. The ␮ was mixed with 380 L of a solu- Giorgio Cevasco* lower limit of detection (signal-to- tion of SBD-BF (25 mmol/L in 0.125 Anna Maria Mumot noise ratio of 3:1) for the derivative mol/L borate buffer with EDTA 5 Carlo Scapolla of Hcy was 3 nmol/L, corresponding ␮ mmol/L, pH 9.5) and 20 Lof Sergio Thea to 60 fmol (the upper limit was not NaOH 1 mol/L, incubated at room tested). temperature for 30 min, and then acid- Comparative chromatographic runs Dipartimento di Chimica e ified with 50 ␮L of HCl 1 mol/L. In a of standard solution of aminothiols in Chimica Industriale Universita` final step, 20 ␮L of the sample was borate buffer (Fig. 1A) and human di Genova and Consiglio injected into a Agilent 1100 HPLC sys- plasma (Fig. 1B) after derivatization Nazionale delle Ricerche with SBD-BF demonstrate that this tem equipped with fluorescence and Genova, Italia new fluorogenic reagent is suitable for ultraviolet-visible detectors. thiol determination in biological sam- The samples, as well as the starting ples. Briefly, 100 ␮L of human plasma solutions, were carefully flushed * Address correspondence to this au- ␮ with nitrogen and protected from air thor at: Dipartimento di Chimica e was mixed with 10 L of a solution of Chimica Industriale, Consiglio Nazionale Tris (2-carboxyethyl)phosphine (75 until injection onto the column. Be- delle Ricerche, Via Dodecaneso, 31 mmol/L in borate buffer, pH 7.4) and cause oxygen is indeed a strong flo- I-16146 Genova, Italy. Fax (ϩ39)010 353 allowed to react at room temperature rescence quencher, some previously 6107; e-mail [email protected]. for 15 min. Then 90 ␮L of a solution of reported unexpected results, i.e., de- DOI: 10.1373/clinchem.2007.095406 trichloroacetic acid (100 mL/L in wa- crease in the fluorescence intensities Clinical Chemistry 53, No. 12, 2007 2223

44 Single-Nucleotide Polymorphisms SNP markers were tested for their used 3 chorionic villus samples Expressed by Placental RNA: presence in 1st-trimester plasma and (weeks 8, 11, and 12) to test markers Assessment for Use in Noninvasive their absence in nonpregnant women. for placental expression. Screening Prenatal Diagnosis of Trisomy 21 Peripheral blood samples were of pregnant and control plasma collected from pregnant women at- was done in triplicate. To minimize To the Editor: tending the Prenatal Diagnostic the effect of biological variation of For noninvasive prenatal diagnosis, Centre of the VU University Medical marker levels in plasma, each of markers that directly reflect changes Center. All participants gave in- the 3 screens in plasma was per- in chromosome dosage are preferred formed consent before study inclu- formed on pooled RNA fractions over indirect markers that are associ- sion. The study was approved by the isolated from individual females in ated with epiphenomena (1, 2). The ethics committee at our institution. series of 44. In practice, this process RNA:single-nucleotide polymorphism We collected EDTA blood samples was performed by downstream (SNP) allelic ratio strategy was de- between weeks 9 and 14 of preg- pooling of the concentrated, indi- scribed recently as a means to di- nancy, before invasive diagnostic vidual RNA fractions isolated from ␮ rectly assess fetal chromosome dosage procedures were performed. Sam- plasma (10 L each) after automated in maternal plasma (2). Quantitive ples were processed and RNA ex- extraction of 44 different plasma comparison of the allelic expression tracted as described previously, samples. For the SNPs of most use, ratios of a placentally expressed, chro- with automated isolation (BioRobot the final screen was done by in- mosome 21–encoded gene, placenta- MDX) (1). RNA extraction from dividual analysis of 6 pregnant specific 4 (PLAC4), enabled detection PAX gene tubes was performed us- plasma and 6 control plasma in maternal plasma of the differences ing the BioRobot MDX with a stan- samples. With the use of RNA iso- between 2 (normal) or 3 copies of chro- dardized protocol (Qiagen). For se- lated from EDTA plasma, 5 of 44 mosome 21 (2). The RNA:SNP ratio lected genes, allele frequencies were SNP markers were detectable in strategy is currently limited to a determined by cycle sequencing maternal plasma and absent in non- subset of the population with hetero- with a Big Dye terminator, followed pregnant plasma: rs8130833 (PLAC4), zygosity of the SNP used. Theoreti- by capillary electrophoresis (ABI rs9977003 (PLAC4), rs11554667 cally, an increase in population cover- 3100XL). (COL6A2), rs9637170 (COL6A1), and age can be obtained by inclusion of Within the transcripts of the 7 rs2187247 (C21orf105) (Table 1). In additional SNPs within PLAC4 or genes of interest, 44 SNPs were iden- contrast, in RNA isolated from other chromosome 21–encoded tran- tified (www.hapmap.org) (Table 1). whole blood collected in Paxgene scripts with placental expression and Primers flanking these SNPs were tubes, no SNP markers fulfilled the detectability in maternal plasma (2). designed with similar thermody- criterion of absence in nonpregnant We therefore tested 44 SNPs ex- namic characteristics to permit RT- blood. Identical analysis of hPL pressed by 7 chromosome 21–en- PCR analysis in single runs. All RNA (3) excluded false positivity, coded, placentally expressed genes primers were intron spanning, ex- because in RNA recovered from (2), PLAC4, collagen, type VI, alpha 2 cept for the primers of PLAC4. Using whole blood in Paxgene tubes, this (COL6A2), collagen, type VI, alpha 1 a sensitive, 2-step, 1-tube RT-PCR marker was clearly present and ab- (COL6A1), BTG family, member 3 assay (Superscript II RT-PCR, In- sent, respectively, in samples ob- (BTG3), ADAM metallopeptidase vitrogen) supplemented with 1 mol/L tained from pregnant and nonpreg- with thrombospondin type 1 motif, betaine to increase reverse transcrip- nant females (data not shown). 1(ADAMTS1), chromosome 21 open tase efficiency and enzyme stability We conclude the following: (a) Al- reading frame 105 (C21orf105), and (1), the marker set was tested in though the PAX gene tube reagent amyloid beta (A4) precursor protein placental tissue (positive control), that stabilizes RNA may be bene- (peptidase nexin-II, Alzheimer dis- plasma from nonpregnant women ficial for RNA isolation from whole ease) (APP), for their potential use in (negative control), and pregnant blood, the large contribution of noninvasive prenatal diagnosis. All women. During the initial screen, we intracellular RNA from maternal

Table 1. SNPs expressed by placental RNA and present in maternal plasma but not control plasma. No. Gene Exon SNPa HET FREQb A1c A2c Forward primer Reverse primer 1 PLAC4 1 rs8130833 0.448 A G GGGACTCGCCGCTAGGGTGTCT GGTGGGGATCCCTTATGCATGG 2 PLAC4 1 rs9977003 0.339 A G AACCGTGGGACCAGTGTAGAAGAATG GGGCAAGTGGAAAACACGCAGT 3 COL6A2 28 rs11554667 Ͻ0.1% C G CACAGCAGGTGCGCAACATG AAGCGCCGGGCCTTGTG 4 COL6A1 23 rs9637170 Ͻ0.1% A C CCTATCGGACCTAAAGGCTAC TCCAAAATCTCGCATTCGTC 5 C21orf105 2 rs2187247 0.5 A C GCGCGCTCTCCGGGTTCCAACC GGGGCCTGTCCACTTCGGTGGTAG a Selected from among 44 tested SNPs of the 7 gene selected genes. The additional SNPs tested (n ϭ 39), and their primer sequences are available on request. b Heterozygote frequencies (HET FREQ) are given for white individuals only. c A1 and A2, variant alleles. 2224 Letters

peripheral blood cells prevents wide- prenatal diagnosis should be retested 4. Go AT, Visser A, Mulders MA, Twisk JW, Blanken- stein MA, van Vugt JM, et al. C21orf105, a spread prenatal use. Prenatal PAX with the RNA:SNP allelic ratio strat- chromosome 21-encoded mRNA, is not a dis- gene tube use appears to be limited egy by use of rs2187247. (g) Our data criminative marker gene for prediction of Down to genes with high relative expres- permit an evidence-based selection syndrome in maternal plasma. Prenat Diagn sion differences between placental of target genes and markers to in- 2007;27:146–9. tissue and maternal blood cells, crease the population coverage of the allelic ratio strategy for noninvasive 1 such as hPL.(b) Our data confirm Attie T.J.I. Go the utility and high expression of prenatal diagnosis of trisomy 21. 2 Allerdien Visser PLAC4 (2).(c) The use of SNP mark- Monique A.M. Mulders2 ers is restricted to specific exons for Marinus A. Blankenstein2 genes with complex transcriptional Grant/funding support: Part of this John M.G. van Vugt1 organization, such as COL6A1 and work was supported by the SAFE 2* COL6A2.(d) For the transcripts of network (Project Number LSHB-CT- Cees B.M. Oudejans 2004-503243). COL6A2 and COLA1 with placental 1 specificity (for example encompass- Financial disclosures: None declared. Departments of Obstetrics/ Acknowledgments: We greatly appre- Gynecology, and ing exon 23 in COL6A1), SNPs re- 2 main to be identified for use in ciate the continuous support from Clinical Chemistry, RNA-SNP assays. The heterozygote the Department of Obstetrics and VU University frequencies of rs11554667 (COL6A2) Gynaecology. Medical Center and rs9637170 (COL6A1) are Ͻ0.1% Amsterdam, The Netherlands References in the white population we tested. 1. Oudejans CBM, Go ATJJ, Visser A, Mulders M, (e) Alternatively, for COL6A2 and Westerman BA, Blankenstein MA, et al. Detec- COL6A1, the combined detection of tion of chromosome 21-encoded mRNA of pla- * Address correspondence to this au- cental origin in maternal plasma. Clin Chem exons with specificity (exons 28 and thor at: Department of Clinical Chemis- 2003;49:1445–9. try, VU University Medical Center, De 23 for COL6A2 and -6A1, respec- 2. Lo YM, Tsui NB, Chiu RW, Lau TK, Leung TN, Henung MM, et al. Plasma placental RNA allelic Boelelaan 1117, 1081 HV Amsterdam, tively) with additional exons carry- ratio permits noninvasive prenatal chromosomal The Netherlands. Fax 31-20-444 3895; e- ing SNPS with high heterozygosity aneuploidy detection. Nat Med 2007;13:218–23. mail [email protected]. (rs2839114, rs1053312) might yield 3. Ng EK, Tsui NB, Lau TK, Leung TN, Chiu RW, Panesar NS, et al. mRNA of placental origin is useful combinations. (f) The predic- readily detectable in maternal plasma. Proc Natl DOI: 10.1373/clinchem.2007.093146 tive power of C21orf105 (1, 4) for Acad Sci U S A 2003;100:4748–53. Book, Software, and Web Site Reviews

Tietz’s Applied Laboratory Medicine, and then presents subsequently or- example, on thyroid and cardiac dis- 2nd ed. Mitchell G. Scott, Ann M. dered laboratory tests. The labora- eases). I found the numerous non- Gronowski, and Charles S. Eby, eds. tory findings, given both in conven- standardized abbreviations annoy- Hoboken, NJ: Wiley-Interscience, a tional and SI units and including the ing; an abbreviation list at the John Wiley & Sons, Inc. Publication, corresponding reference ranges, are beginning of each case presentation 2007, 679 pp., $69.95, paperback. explained and discussed. The under- would have been helpful. The index ISBN 978-0-471-71457-6. lying disease with its pathogenesis generally lists the diseases and clini- and treatment is described in detail, cal symptoms, but analytes are noted Applied Laboratory Medicine was 1st and the possible differential diag- published in 1992, and this useful in a rather inconsistent and incom- noses are given. If necessary, general plete manner, a shortcoming that collection of interesting medical methodical problems (preanalytical may prevent the reader from finding cases has since become indispensable and analytical factors) are consid- all important issues relevant to a as teaching material for laboratory ered. As with any multiauthored particular topic, especially because medicine. The new editors of this text, the quality of writing varies; 2nd edition have maintained the however, the editors succeeded in the cases with their clinical and cor- classical case-oriented approach. presenting uniform reports for each responding laboratory data are de- Moreover, by including new diag- case. Most of the contributions dem- scribed under different aspects. For nostic findings based on molecular onstrate that clinical and laboratory example, neither troponin nor creat- biology methods, this new volume data must be strongly intertwined to ine kinases are listed in the index. encompasses the enormous develop- provide an accurate diagnosis. Arti- In summary, this book provides a ments in laboratory medicine since cles of salient information that I en- concise and successful bridge be- the appearance of the 1st edition. joyed reading are those that use the tween clinical and laboratory medi- These timely additions are particu- cases to discuss ongoing topics of cine. I would highly recommend it to larly evident and distinguish this special importance; among them, the all trainees, scientists, and physicians book as a carefully prepared, up-to- validity of the different laboratory in laboratory medicine as well as to date compilation of interesting med- data to diagnose myocardial infarc- all clinicians, who should be pre- ical cases. Not only has there been an tion, the use of the Modification of pared for a fruitful, interactive dis- increase in scope, but also one of size, Diet Disease Study equation for de- cussion on the clinical problems in with a 60% increase in the number of termining the glomerular filtration question. I believe that this book will pages from the 1st edition. rate, metabolic syndrome and nonal- serve as an invaluable resource for The book comprises a total of 92 coholic fatty liver disease, Heliobacter scientists, not only for teaching med- medical cases presented by 94 well- pylori infection, prostate cancer and its ical students and technologists but known authors from leading centers screening using prostate-specific anti- also for refreshing their own knowl- in the US, Canada, Australia, and gen, heparin-induced thrombocytope- edge of the topics involved. Its mod- Israel. Cases are arranged in 21 dis- nia, and the role of pharmacogenomics est price makes it acquirable both for ease-specific (among them, cardiac, in personalized medicine. the reference library and for personal pulmonary, renal, liver, endocrino- There are few shortcomings in logical, inherited, infectious, hemato- content and form in this well- use. logic, hemostaseologic, malignant, constructed book. Updated notices Klaus Jung metabolic, and autoimmune dis- on hemoglobin A1c standardization eases) and laboratory-specific sec- and bone markers should have been tions (examples of analytical errors, included. Although a major advan- Department of Urology pharmacogenomics, and toxicology) tage of this multiauthor book is that University Hospital Charite´ and cover the areas of importance to each case is presented as a complete, Berlin, Germany modern laboratory medicine. Each self-contained chapter, this approach case presentation starts with the leads to some redundant and not DOI: 10.1373/clinchem.2007.088963 health history and clinical symptoms always up-to-date discussions (for

Clinical Chemistry 53, No. 12, 2007 2225 Correction In our Technical Brief “Can Whole-Blood Samples Be Stored over 24 Hours without Compromising Stability of C-Reactive Protein, Retinol, Ferritin, Folic Acid, and Fatty Acids in Epidemiologic Research?” by Manon van Eijsden, Marcel F. van der Wal, Gerard Hornstra, and Gouke J. Bonsel (Clin Chem 2005;51:230–2; DOI: 10.1373/ clinchem.2004.042234), we used the term “folic acid” for the nutrient as measured in serum. However, because folic acid generally refers to the synthetic form (e.g., the form in which the nutrient appears in nutritional supplements), the term “folate”, referring to the active, natural form of the nutrient as measured in serum, would have been more appropriate. Although this translation error does not alter any of our results or conclusions, we regret any confusion that may have occurred as a result.

Manon van Eijsden Municipal Health Service/Cluster EDG Amsterdam, The Netherlands

DOI:10.1373/clinchem.2007.098541

2226 The Clinical Chemist

Compiled by David E. Bruns, Editor ([email protected])

Parting Thoughts from Your Editor impending demise of journals. So I manuscript I submitted. Our legacy will refrain from looking into any emanates from previous chairs of the The time has come for me to sign crystal balls. Editorial Board: Donald Young (who off as your editor, a role that has As for advice to the incoming edi- labored successfully to make this an occupied most of my waking life for torial team, it is clear that they do international journal and who took a the last 18 years. Editing has been a not need it. In fact, they probably chance by appointing me to the Edi- privilege and an ongoing education, did not need the advice I already torial Board’s Executive Committee but all good things must end, and gave in response to their questions. shortly after I was elected to the now I must say my goodbyes. They have been busy for months and Board), Ted Peters (who taught all of Farewells from editors may be ex- us about science and about how to pected to cover 3 areas: (1) a look have already prepared a January is- behave), Jack Ladenson (who was back in time at the editor’s journal, sue that you will recognize as mark- chair when I started as editor and (2) a look forward, and (3) advice to ing the beginning of the journal’s who was and is the journal’s secret the incoming editorial team. move to its next, higher level. In case weapon), and Carl Burtis (who, I will dismiss the 1st issue by indi- they might need advice in the future, among other things, tutored me in cating that there has been too much I will write the following (with apol- analytical chemistry and thus pre- history in 18 years to summarize in 1 ogies to W. Somerset Maugham): vented many embarrassing moments or 2 (or, as we used to write, “one or There are 3 rules for editing a peer- for the journal). two”) paragraphs. I will summarize reviewed journal. Unfortunately, no A peer-reviewed journal, like science, by indicating there has been much one knows what they are. represents a collaboration, supported more pleasure than pain (although My thoughts on a beautiful, sunny, by mutual trust, of the following both were available in abundance) late October day in Virginia turn to people: authors, readers, reviewers, and much satisfaction. the people who have made this jour- Regarding the look forward, pre- nal what it is. We inherited a legacy editors, office editorial staff, associate dictions are notoriously wrong. Even from Harold Appleton, who was the editors, editorial board members, the fabled Stan King, my distin- journal’s 1st editor, and from Stan manuscript editors, copyeditors, pro- guished predecessor, seems to have King, the executive editor who suc- duction editors, printers, electronic erred when he warned in his depart- ceeded Appleton and taught me publication experts, publication direc- ing comments 18 years ago of the about writing by tearing apart each tors, owners (here, AACC), and more. Each part of this team is essential, and none is more important than any one of many people who are unknown to readers. I think of Tommy Sadler and the time when all 10 000-plus copies of a monthly issue were printed with the same page numbers as had already been used in the preceding month’s issue. Such an error is a publishing nightmare. On a Friday afternoon, I was told that the issue could not be reprinted because there was no paper. On Saturday, Tommy Sadler, whose job was to print the journal not to find paper, said he would look into it on Monday. In fact, he found the paper over the weekend, reprinted the issue, and mailed it on time. None of the readers knew. This scene was representative of actions by individuals who went beyond their job descriptions to bring you the best possible journal that they could deliver. I would like to thank individually all members of the team that brought you this journal over the last 18 years, but there are too many to list by name. I must, however, make a Fig. 1. “One more to go.” Dr Bruns surrounded by 17 years and 11 months of Clinical Chemistry. special thanks to the office staff and

Clinical Chemistry 53, No. 12, 2007 2227 2228 The Clinical Chemist

association election. The president- elect, two positions on the Board of Directors, and four positions on the Nominating Committee are to be elected to serve, beginning January 1, 2009. Nominations should be sent in writing to: Dr. Larry J. Kricka, Hos- pital of the University of Pennsylva- nia, 3400 Spruce St., 7103 Founders Pavilion HUP 3400, Philadelphia, PA 19104-4283 (fax 215-662-7529; e-mail [email protected]). You should indicate the office for which the nominee is proposed and include 1 or 2 sentences explaining the nomination. A nomination form will be available on the AACC web- Fig. 2. Images from the beautiful Spirited Garden, Jeju Island, South Korea (Republic of Korea). site (www.aacc.org) January 1. All Photos courtesy of Sung Bum-young of Spirited Garden. nominations must be received by January 31, 2008. The Nominating Committee will obtain the candidates’ consent to run for office. Dr. Jim Boyd in Charlottesville. Two work with them. Where I have failed, According to the association by- of the staff, Sandy Weaver and I apologize. laws, in this election, members of the Donna Brandl, have served the jour- Finally, as some people know, I am Capital, Chicago, Northern Califor- nal almost as long as the office has most indebted to my wife, Liz. She nia, and Southeast Sections are not been here. The staff and Jim have has put up with late and aborted eligible for election to the Nominat- been daily companions and have dinners, working weekends, and ing Committee for the term begin- canceled vacations, all in support of shown an unparalleled dedication to ning January 2009. my attempt to do my part in the clinical chemistry and to Clinical Members of the 2008 Nominating Chemistry. Their positive attitudes production of a journal that would Committee are as follows: Larry have brightened many a day. reflect credit on our profession. She Kricka, DPhil (Chair); Corinne Fantz, I have tried during my tenure to has other plans for me now. PhD; David Grenache, PhD; Patricia meet what I viewed as my primary I wish the journal every success Jones, PhD; Stephen Manzella, PhD; ethical responsibility of providing and look forward to the new features fair, expert, and timely reviews. I that are in store. Joseph McConnell, PhD; Judy Stone, tried to work with authors to im- PhD; and Donald Wiebe, PhD. prove their papers. Where I have call for nominations succeeded, I thank authors for their AACC’s Nominating Committee is DOI: 10.1373/clinchem.2007.099861 cooperation and the opportunity to accepting nominations for the 2008 Invited Reviewers—2007

As an expression of our appreciation, we list here the names of those who have served as invited reviewers between October 1, 2006, and September 30, 2007.

Stephen Abbs Piter J. Bosma Daniel Chiu B.B. Duncan Christian C. Abnet Xavier Bossuyt Tor Chiu Paul N. Durrington Vahid Afshar Roger Bouillon M. Choolani Jacob D. Durtschi Yash Pal Agrawal Edwin Bovill Uwe Christians Charles S. Eby Habibul Ahsan Larry D. Bowers Robert Clarke John Eckfeldt Bjorn Akesson Robert Bowser William Clarke Graeme Eisenhofer Keith Roland Allen Ronald R. Bowsher Judith Ann Clements R.P. Ekins Raymond D. Aller Arthur R. Bradwell Aldo Clerico Lena Ekstrom William S. Allison Mark Brantly Christa Cobbaert Donald L. Elbert Thomas J. Allred Simona Bratu Laurence A. Cole Moses Elisaf Leigh Anderson Lewis E. Braverman Paul O. Collinson J. Elliott Jose Manuel Andrade-Garda Stephen O. Brennan Wayne D. Comper John Ellison Cristina Andres-Lacueva D. Brenner Mario Comporti Mahmoud ElSohly Anne Angelillo-Scherrer John Brick Danila Coradini Kenneth Emancipator Victor W. Armstrong Richard Bringhurst Jean-Benoit Corcuff Hanna Engler David E. Arnot Douglas A. Brooks Emanuela Corsini Maria Erali S. Asher Larry Broussard Catherine Costa Juergen G. Erhardt Kadir Aslan William Brugge David A. Cowan James R. Eshleman Nancy Augustine Carlo Brugnara Diane W. Cox David R. Eyre Neil David Avent Merce Brunet Michael H. Creer Michael Fahie-Wilson Antonio Ayala John Brunstein Massimo Cristofanilli Jim D. Faix Aaron Baggish Mo`nica Bullo Susan Crocker Emmanuel J. Favaloro David Bahler David Bunk R. Croner Con Feighery Eleni Bairaktari Michael E. Burczynski Gyorgy Csako Peter Ferenci Renze Bais Andrew Burd Mohamed R. Daha Uden Fernandez-Real Jaap A. Bakker Raymond F. Burk Marja-Liisa Dahl Maurizio Ferrari Cecilia Balbi John R. Burnett Morten Dahl Soldano Ferrone Don Baldwin Carl A. Burtis Bjorn Dahlback Marco Fichera Marie A. Balikova Anthony W. Butch Shale Dames Heidi Fiegl Cynthia M. Balion Paul Cairns Pierre D’Amour J. Findlay Ulysses J. Balis Daniele Calistri Nadia Dandachi Guy D. Fink Rosamonde E. Banks Veronica Calvin Jacqueline Suk Danik Judith A. Finlay Daniel D. Bankson Handan Ankaraly´ Ian N. Day Delbert A. Fisher Gisela Barbany Camdeviren Gianluca De Bellis George H. Fisher David R. Barnidge Hania Campos Sarah D. De Ferranti Rob Fitzgerald Markus J. Barten Carlos Camps Frank H. De Jong Anne Mentro Fitzpatrick Chris Bates Tom Cantor Rafael Dela Torre Robert Fleming Hannsjorg Baum Louis Caplan James De Lemos Pasquale Florio Christian Baumgartner Joyce A. Carlson R.R. De Winter D. Foran Antoni Bayes-Genis Filippo Carlucci David Deamer Paolo Fortina Olof M.L. Beck Ralph Carmel Chris Defilippi Elizabeth Frank Karsten Becker Kevin Carpenter Pierre Delanaye M.S. Freedman Aditya Bedekar Kenneth J. Carpenter Joris R. Delanghe Hudson H. Freeze O. Bekers Antonio Carroccio Edgard Delvin Jan M. Friedman Alain Belanger Don H. Catlin Laurence M. Demers Simonetta Friso Chantel Be´meur Francesco Cavagnini David T. Denhardt Eric T. Fung Elizabeth Benito-Garcia Etienne Cavalier Ramon Deulofeu Andrea Gaedigk Peter A. Benn Lisa H. Cazares Sridevi Devaraj Shira Gal Marianne Benn Anna Cederholm Koen Deventer Douglas Galasko Anders Berg George S. Cembrowski R. Dhallan Monica Galli Philip Bernard Ferruccio Ceriotti Rosalie Dhonukshe-Rutten Marcello Galvani Mario Berth Geun Sig Cha Eleftherios P. Diamandis Peter Garred Ernest Beutler Donald H. Chace Stuart J. Dickson David C. Gaze Diana W. Bianchi Yvon Chagnon G. Dickstein Adrian Gear Luigi Biasucci Toby C. Chai Dennis J. Dietzen Joerg Geiger Martin Bidlingmaier Krishnendu Chakrabarty Chunming Ding Michael H. Gelb Christian Hans Bieglmayer Dick C. Chan Alexander Dobrovic D. Geraghty Jacek Bielawski Iris H.S. Chan Steve F. Dobrowolski Robert Gerszten Neil Binkley Juliana C.N. Chan Allan Doctor Robert H. Getzenberg D. J. Birch Abalo Chango Gregory G. Dolnikowski Don Giacherio Andrew D. Blann John F. Chapman Gabriel Dorado Evangelos Giannitsis Dennis M. Bleile Caroline J. Chapman Paul A. D’Orazio James M. Gillard Kaj Blennow Dan Chasman Basil T. Doumas J.M. Gimble Jay L. Bock Pierre Chaurand Steven K. Drake Betti Giusti Olaf A. Bodamer Roy Chen H. Drakesmith Paul Glasziou Geza Bodor Yu-Ju Chen Olaf Drummer Hansruedi Glatt Christoph Borchers Sulin Cheng David Duewer Jan Glatz Susan Bortolin Kwai Wa Cheng Michael J. Duffy Paul Glendenning Katrin Borucki Rossa W.K. Chiu D. Robert Dufour Tony E. Godfrey

Clinical Chemistry 53, No. 12, 2007 2229 2230 Invited Reviewers—2007

Marianna Goldrick David A. Herold Julia S. Johansen Malgorzata Krzystek- J. Goldstein Wolfgang Herrmann Rhys John Korpacka Denise Gonzales Paul Hewitson Garry John Mikael Kubista Victor M. Gonzalez Stephen M. Hewitt Patricia Jones Doris Kuehnelt Michel Goossens Erik L. Hewlett Graham Ross Jones David N. Kuhn W. Grady Peter E. Hickman Julia C. Jones David Allan Kulesh Alessandro Granito Trefor N. Higgins T.H. Jones Georg Kurz Julian Charles Gray James A. Higgins Etienne Joosten Mark M. Kushnir Stefan K.G. Grebe Jørgen Hilden Joanne Jordan Dirk R.J. Kuypers Ralph Green Peter G. Hill Saeed A. Jortani Patricia Kyd Sol Green Ulf Hindorf Harald Jueppner Robert F. Labbe Jesse F. Gregory III Toru Hiyama R. Kahn Andre Lacroix David G. Grenache Ilse E.A. Hoffman Anders Kallner Jack Ladenson Andrea Griesmacher Michael E. Hogan Malek Kamoun Fred Lakeman A. Griffith Kerstin Hogg Roman Kandar Ching-Wan Lam Peter R. Griffiths Gudrun Høiseth Daniela Kandioler Rolf Maria Lamerz Wayne W. Grody Stefan Holdenrieder Sandra Kaplan James P. Landers Matthew C. Groll Carla E.M. Hollak Jerry Kaplan Silke Lankiewicz Ann M. Gronowski Bruce W. Hollis Isabella Kardys Kaiqin Lao Anders Grubb David W. Holt Hiroshi Kasai A. Lapolla Scott M. Grundy Paul Holvoet Juan-Carlos Kaski Michael Laposata Li-Qun Gu Wolfgang Holzgreve Ritu Kataky Catherine Larue Valentina Guida Sun Hong Hugo A. Katus Bill L. Lasley Cameron Gundry John W. Honour Jerry A. Katzmann Franc¸oise Lasne Rebekah L. Gundry Andrew N. Hoofnagle Jean M. Kaufman Gordan Lauc Teemu Gunnar James Hooper David A. Kaufman Fulvio Laurentani Ivo G. Gut Bernd Hoppe Rymantas Kazlauskas John Lawry Holger Hackstein Michael Hoppe Jacquie T. Keer J.H. Lazarus Rainer Haeckel Adrian Horgan Charles R. Keese X. Chris Le Sinuhe Hahn Paul S. Horn S. Keiles Thomas Ledue Daniel J. Haisenleder Andrea R. Horvath Martin A. Kennedy Caroline G. Lee Christiane S. Hampe Michael K. Hourfar Marina L. Kennerson Duk-Hee Lee Harald Hampel John Greg Howe Louise Clare Kenny Kelvin H. Lee Xianlin Han R. Rodney Howell Jana Oliveriusova Kent Lynne Lennard William S. Hancock James D. Hoyer David F. Keren Gary L. Lensmeyer Ulf Hannelius A. Hozawa Nigel S. Key Rudolf M. Lequin L. Hansen Frank Hu Omar S. Khalil Ky Leung Helle Ru¨ sz Hansen Xiaoli Hu Thomas Kickler Piotr Lewczuk Nadia Harbeck Andreas R. Huber D. A. Kidwell Lawrence Lewis Robert W. Hardy Barry I. Hudson Paul Kiefer Russell J. Lewis Iain Parry Hargreaves J. Stephen Huff Michael Kiehntopf Giovanni Li Volti Harri Ha¨rma¨ Menno Huisman Jan T. Kielstein Evi S. Lianidou Frank E. Harrell Steve E. Humphries Eric S. Kilpatrick John C. Lieske James H. Harrison Anne-Mette Hvas Mary M. Kimberly Michael Liew Alison Harte S. Ibrahim Pascal Kintz Hans Lilja C.L. Harteveld Mats Ingana¨s Michael Kleerekoper Johan R. Lillehaug Kenji Hashimoto Magnus Ingelman-Sundberg Roger David Klein Bertil Lindahl I. Hatada Steven H. Itzkowitz J. Stacey Klutts Ulf Lindahl Yetrib Hathout Kaisa K. Ivaska Janina Kneipp Gregory Lip Aristides T. Hatjimihail Anne Jaaskelainen Pernille B. Koefoed-Nielsen Giuseppe Lippi Ulrike Haug Craig M. Jackson Wolfgang Koenig Randie R. Little Johan Haux Donald W. Jacobsen John A. Koepke R-M. Liu Dede Haverstick Glenn A. Jacobson Wolfgang J. Koestler John Hamilton Livesey Charles D. Hawker Jaak Jaeken Kenro Koide Stefan Lofas Manzour Hernando Hazbo´n Allan S. Jaffe Wim Th. Kok Lorne Lonie Kevin C. Hazen Hieronim Jakubowski Tomokazu Konishi Per Eystein Lønning Stanley L. Hazen S. Jill James Tae-Woong Koo Andreas L. Lopata Robert P. Heaney Timothy Jang R. Kopelman Abel Lo´pez-Bermejo Mehdi Hedayati Bart Janssen Paul F.J. Koppens Robin G. Lorenz Christopher Heeschen James Januzzi Minna A. Korolainen Anu Loukola Robert A. Hegele Petter Jaremo Gerald J. Kost Johan Lundin August Heidland Petr Jarolim Sibylle A. Kozek-Langenecker Regina Luttge Lutz Heinemann Gary P. Jeffrey J. Kristiansen Andrew W. Lyon Anders Helander Rosalind E. Jenkins Martin Kroll Elaine Lyon Nicole Helbecque Erika Jensen-Jarolim Robert Kronstrand Yun Ma Ilkka Hemmila Per Bendix Jeppesen Jan S. Krouwer Andrew R. Mac Rae Jim Hempe Ishwarlal Jialal Mogens Kruhøffer Piet Maes Luis Hernandez Hao Jiang J. Krupinski Mark J. Magera Clinical Chemistry 53, No. 12, 2007 2231

Per Magnusson Estanis Navarro Thierry Poynard Roger K. Schindhelm Johannes M. Mair Edwin W. Naylor Aruna D. Pradhan Rosemary L. Schleicher Susan Elisabeth Manley Bryant C. Nelson Guy Pratt Hermann J. Schluesener Kenneth G. Mann Arja M. Nenonen Christopher P. Price H. Schmidt Ferdinando Mannello Irina Neverova Tim Prickett I.M. Schmidt R. Mao Valerie L. Ng Ronald L. Prior Warren N. Schmidt Dominique Marcus- Maggie Ng Thomas W. Prior Henrik Schmidt Soekarman L-C. Ng Stanley Prusiner Manfred Schmitt Glenn Markenson James H. Nichols Edward Quadros Gerd Schmitz Rodney S. Markin I. Nieman Hershel Raff Erasmus Schneider Salvatore A.E. Marras M. Nishikawa J. Ragoussis Dale A. Schoeller Sharon Marsh Andreas Nitsche Alex J. Rai Greet Schoeters Sally M. Marshall A. Barry Nix Timothy H. Rainer Jens-Michael Schro¨der Steven C. Martin T. Nolan Petrie Rainey Matthias Schwab Jaume Martorell Frederick S. Nolte Ramani Ramchandran Eve B. Schwartz Catherine Massart Anthony G. Norden H. Randeva Fred Schwartz Alan Mast Hans Nossent David F. Ransohoff Heidi Schwarzenbach Stephen R. Master William E. O’Brien Felix Ratjen Katarina Sebekova Dietrich Matern Sean D. O’Broin Manfred Rauh Brahm H. Segal Robert Matson John F. O’Connor Jan Ravkilde Shamala Devi Sekaran Hans H. Maurer Sandra O’Dell Emmanuelle Reboul Akihiko Sekizawa Alex McAdam Marco R. Oggioni John E. Rectenwald John Semmes Patrick E. McBride Shuji Ogino Thomas Remer J-M. Serot E.R.B. McCabe Daylily S. Ooi Patsy Renard Alessandro Serretti Andrew McCaddon Antone R. Opekun Juan A. Rey Nathalie S. Seta James McCord Jordi Ordonez-Llanos Gavin P. Reynolds Iris Shai Kilmer S. McCully Jose M. Ordovas Erinn Rhodes F. Shakib Stanley McMillan Bjarne Østerud Thomas L. Richie Nathan Sharon Gwen A. McMillin Mehmet Ozsoz Paul M. Ridker Leslie Shaw Helene McNulty Stephanie Padilla Mark J. Rieder Paul Shaw Matthew J. McQueen Bob Palais Eric Rimm Andrew Shennan Peter J. Meikle Darryl Palmer-Toy Piero Rinaldo Ie-Ming Shih Christa Meisinger Murugan R. Pandian Lorenz Risch Zak Shihabi Henri A. Me´nard Mauro Panteghini Harry G. Rittenhouse Terry Shirey D. Merrick Nick J. Parham Rebecca L. Roberts Yehuda Shoenfeld D. Mezzano Walther Parson Robert Roberts Edward K. Shultz Leann Mikesh George H. Parsons Sander J. Robins Lothar Siekmann Walter L. Miller Curtis A. Parvin Alan L. Rockwood K. Silander James J. Miller Brittan L. Pasloske George M. Rodgers Leslie E. Silberstein David T. Miller Marzia Pasquali Diane Roe Lawrence Silverman Alison S. Millson Kenneth A. Pass Angela Rogers David Sinclair Mara Mirasoli Millan S. Patel Thomas Rohan Ravinder J. Singh M. Misra Randi A. Paynter K. Roper Jasbir Singh Masato Mitsuhashi Frank Peacock David S. Rosenblatt Robert Slinger Kiyonori Miura Anna Pelander Roland Rosmond Patrick M. Sluss Martin Moeckel Amir Pelleg Enrico Rossi J. Smerage Anne Louise Mørkbak Silvia Pellegrini Ranieri Rossi B.E. Smink L. Moldawer Giuseppe Penno Paul Rozen David Smith Manfred R. Mo¨ller M. S. Pepe Patrizia A. Russo Brian Richard Smith Carla Monico Luca Persani Anne W. Rutjes Alberto Smith Myeonghee Moon Inga Peter Robert G. Rutledge Steven Sidney Smith Christine Moore Theodore Peters Henk Ruven M. Sohne Samia Mora Per Hyltoft Petersen Michael E. Rybak Steven J. Soldin J.E. Morley Jon Peterson Seungho Ryu Nina J. Solenski Michael Mosesson Martin Pfeffer Ann Saada Gyo¨rgy So¨le´tormos Thomas Mothes John D. Pfeifer Marc Sabatine Govert Somsen David Charles Muddiman John T. Philbrick Ulrich Sack Yiqing Song Utpal K. Mukhopadhyay David Phillips Marianne Sadar Sabine Sonnenberg Mathias M. Muller Simona Pichini Sayed M.H. Sadrzadeh Oswald Sonntag Uwe Muller Tobias Pischon Devereux N. Saller Valeria Sordi Volkmar Mu¨ ller Mario Plebani Osamu Samura Karl Sotlar Manuel Mun˜oz Uwe Poege Sverre Sandberg Pierre-Edouard Sottas S. Muttukrishna Laurent Poirel Johannes Sander Jean-Claude Souberbielle Gary L. Myers Leon Poller R. Sanfordd Anne Soutar Megumi Nagaoka Terence C.W. Poon Remy Sapin Julian E. Spallholz Moon Hea Nahm Leo Poon Andreas Schaffler Paul N. Span Kimberly L. Napoli Mark A. Poritz Mark Schiffman Carole Ann Spencer Matthias Nauck Jonathan D. Powell Ingolf Schimke S. Spivack 2232 Invited Reviewers—2007

P. Sriramarao Jillian R. Tate Teun van Gelder Peter Wilding Christine E. Staatz Paul J. Taylor Fred Van Lente Jochen Wilhelm Anders Stahlberg G.R. Taylor Walther J. van Venrooij S. Wilkening Frank Z. Stanczyk Malcom Taylor Jochem W. van Werkum Walter C. Willett Ana K. Stankovic Dennis Templeton Dirk Vanderschueren George Wilson Noel V. Stanton M. Tena-Sempere Jo Vandesomple Emily S. Winn-Deen J. Steffann Dzung C. Thach Johannes Veldhuis Steve Winters Michael W. Steffes Georgios Theodoridis Jean-Pierre Vendrell Steven H. Wong Werner Steimer Theoharis Theoharides Hein W. Verspaget David T. Wong Evan A. Stein Mario Thevis Marco Vincenti Mark Worwood Steven Steinberg Perumal Thiagarajan Frank Vinicor R. Scott Wright David Stejskal Julian Thomas Margaret A. Vizzard Alan H. Wu Lesley Stevens Edward J. Thompson K. Voelkerding Hugo Stiegler Visith Thongboonkerd Benjamin F. Voight Bai-Lin Wu P. Stockley Michael O. Thorner Nicolas von Ahsen Friedrich Wurst Mark H. Stoler David Timson Angie Wade Aimin Xu James Storhoff Laurence Tiret U. Waje-Andreassen Mototeru Yamane Michael Stowasser E. Tizzano Brian K. Walsh Yoshio Yamaoka Marek Straczkowski John G. Toffaletti Jun Wang Tim Yandle Charles M. Strom Joyce Tombran-Tink Zhouping Wang Kuender D. Yang Maurice Stroun Aaron Tomer Lili Wang Shuzhang Yang Stephen Stroupe Tony Torresani L. Wangh Pan-Chyr Yang Catharine M. Sturgeon Maret Traber Joseph Watine C. Yen Miroslav Styblo R. Trent Michael Wayne Kiang-Teck Jerry Yeo Papasani V. Subbaiah Anubhav Tripathi Stephen G. Weber Shea Ping Yip Pauli Suominen Armando Tripodi Yau-Huei Wei Timothy Tak Chun Yip Denis Sviridov Heinz Troxler Heidrun Weidemann Alice Ylikoski Rama Swaminathan Michael Tsai Steffen Weikert Donald S. Young Fred C.G.J. Sweep F. Tsai Debra Weiner Bingfang Yue Ronald Swerdloff Agathocles Tsatsoulis Mark H. Wener Jack Zakowski Dorine W. Swinkels Ching-Ping Tseng Rick A. Wetsel Ulrich M. Zanger Pawel Szulc Gregory J. Tsongalis Ron A. Wevers Robert Zee Franco Tagliaro Hirokazu Tsukahara Cas Weykamp Barbara K. Zehentner Susan S. Tai Ursula Turpeinen Sharon Whatley Barbara Zehnbauer Joelle Taieb Luis A. Ugozzoli Hayley C. Whitaker Kenji Takagi Edwin F. Ullman David Whitcombe Sichun Zhang Kazuhiko Takahashi S. Usami Helen E. White Xueji Zhang Yoshio Takihana Gerd Utermann John B. Whitfield Zhi Zheng Philippa Talmud Matti J. Va¨lima¨ki Penny Whiting Xiao Yan Zhong Dinesh K. Talwar Richard B. van Breemen Michael P. Whyte Lan Zhou Weihong Tan E. Van Cott Frank H. Wians Zhifeng Zhou Atsushi Tanemura Angela van Daal Mark R. Wick Willem G. Zijlstra Peter H. Tang Danielle A.W. M van der Donald Wiebe Robert Zimmermann Giovanni Targher Windt William R. Wikoff Hongzhi Zou Beth A. Tarini Jennifer E. Van Eyk R. Bruce Wilcox Jan Zuna

DOI: 10.1373/clinchem.2007.099598