Article Urinary Angiotensinogen and Risk of Severe
Total Page:16
File Type:pdf, Size:1020Kb
Article Urinary Angiotensinogen and Risk of Severe AKI Joseph L. Alge,* Nithin Karakala,*† Benjamin A. Neely,* Michael G. Janech,*† James A. Tumlin,‡ Lakhmir S. Chawla,§ | Andrew D. Shaw, ¶ and John M. Arthur,*† for the SAKInet Investigators Summary Background Biomarkers of AKI that can predict which patients will develop severe renal disease at the time of diagnosis will facilitate timely intervention in populations at risk of adverse outcomes. *Medical University Design, setting, participants, & measurements Liquid chromatography/tandem mass spectrometry was used to of South Carolina, identify 30 potential prognostic urinary biomarkers of severe AKI in a group of patients that developed AKI after Charleston, South cardiac surgery. Angiotensinogen had the best discriminative characteristics. Urinary angiotensinogen was Carolina; †Ralph H. subsequently measured by ELISA and its prognostic predictive power was verified in 97 patients who underwent Johnson Veterans cardiac surgery between August 1, 2008 and October 6, 2011. Affairs Medical Center, Charleston, South Carolina; Results The urine angiotensinogen/creatinine ratio (uAnCR) predicted worsening of AKI, Acute Kidney Injury ‡University of Network (AKIN) stage 3, need for renal replacement therapy, discharge .7 days from sample collection, and Tennessee College of Medicine, composite outcomes of AKIN stage 2 or 3, AKIN stage 3 or death, and renal replacement therapy or death. The Chattanooga, prognostic predictive power of uAnCR was improved when only patients classified as AKIN stage 1 at the time of Tennessee; §George urine sample collection (n=79) were used in the analysis, among whom it predicted development of stage 3 AKI or Washington death with an area under the curve of 0.81. Finally, category free net reclassification improvement showed that University, Washington, DC; the addition of uAnCR to a clinical model to predict worsening of AKI improved the predictive power. | Duke University, Durham, North Conclusions Elevated uAnCR is associated with adverse outcomes in patients with AKI. These data are the first to Carolina; and ¶ demonstrate the utility of angiotensinogen as a prognostic biomarker of AKI after cardiac surgery. Durham Veterans Clin J Am Soc Nephrol – Affairs Medical 8: 184 193, 2013. doi: 10.2215/CJN.06280612 Center, Durham, North Carolina Introduction proteomics to identify prognostic urinary biomarkers Correspondence: One of the most important factors underlying the poor of AKI after cardiac surgery. We performed a verifi- Dr. John Arthur, outcomes seen in AKI is our current method of Division of cation of angiotensinogen in patients who developed Nephrology, diagnosis, which is based on either an increase in AKI after cardiac surgery. This is the first study to Department of serum creatinine (SCr) or decreased urine output (1,2). demonstrate the utility of angiotensinogen as a prog- Medicine, Medical SCr and urine output values at the time of diagnosis nostic biomarker of AKI. University of South are of limited prognostic value, making it difficult to Carolina, 96 Jonathan Lucas Street, PO Box discriminate between mild and severe AKI at the time 250623, Charleston, of diagnosis. The need for better biomarkers of AKI Materials and Methods SC 29425. Email: has been recognized as a crucial barrier to improve- Urine Samples [email protected] ment of the outcomes of AKI patients. Newer bio- Urine samples were obtained from 99 consecutively markers of AKI include kidney injury molecule 1, enrolled patients who had cardiac surgery at one of neutrophil gelatinase-associated lipocalin (NGAL), the SAKInet institutions between August 1, 2008 IL-18, and cystatin C (3–9). Many of these biomarkers and October 6, 2011. Informed consent was obtained initially appeared capable of early, accurate detec- in accordance with the institutional review board– tion of AKI, but subsequent verification studies have approved protocol at each institution. Samples were reported lower accuracy (10–17). In addition, the em- collected and stored using a standard operating pro- phasis on early detection has overshadowed investiga- cedure that included centrifugation, addition of pro- tion of their prognostic predictive power. Available tease inhibitors, and storage at 280°C. Urine samples data on the prognostic value of these biomarkers were collected as early as possible after AKIN serum suggest that they are limited predictors of adverse creatinine criteria were met (2). Inclusion criteria outcomes (18,19). The limitations of biomarkers un- were surgery of the heart or ascending aorta and de- derscore the need to discover new prognostic bio- velopment of AKI within 2 days of surgery. Partici- markers. pants with baseline SCr .3 mg/dl were excluded. Approximately 20% of patients who undergo car- Twelve samples were used in the proteomic studies, diac surgery develop AKI. The timing of the injury can 10 of which were also used in a validation set that be readily determined in these patients (20). We used included samples collected from the remaining 87 184 Copyright © 2013 by the American Society of Nephrology www.cjasn.org Vol 8 February, 2013 Clin J Am Soc Nephrol 8: 184–193, February, 2013 Urinary Angiotensinogen Predicts Worsening AKI, Alge et al. 185 participants in the study. Of the samples used in the val- combination of P values from the Wilcoxon rank-sum idation set, 79 were from patients classified as having test and mean fold-change between the experimental AKIN stage 1 at the time of sample collection. groups. The relationship between these two measures was visualized by “volcano plot.” In verification studies, ’ Proteomic Analyses count data were analyzed using the chi-squared or Fisher s A detailed description of the methods is available in the exact test as appropriate. Continuous variables were ana- t U Supplemental Material. HIV protein gp160 (200 ng; Bio- lyzed using the test or Mann Whitney test. ANOVA or – post hoc clone Inc) was spiked into each urine sample. Proteins Kruskal Wallis ANOVA on ranks test and the ’ were denatured, alkylated, and digested with trypsin. Dunn s test for pairwise comparison were used to evaluate Samples were prefractionated by solid phase extraction continuous variables when more than two groups were using a Strata-X SPE cartridge (Phenomenex). Sample frac- compared. Odds ratios (ORs) were used to test the associ- tions were reconstituted in mobile phase A (MPA) (98% ation of uAnCR with selected outcomes. Patients were stratified by uAnCR into quartiles. The effect of uAnCR H20, 0.1% formic acid; 2% acetonitrile). Five microliters of each fraction was injected onto an Acclaim PepMap100 on the risk of developing an outcome was tested by calcu- trap column, washed with 100% MPA, and separated on lating the OR of the upper and lower quartiles and esti- fi an Acclaim PepMap100 analytical column (75 mmID3 15 mating the 95% con dence interval of the OR. Receiver cm, C18, 3 mm, 100 Å; Thermo Scientific) using a 45-min- operator characteristic (ROC) curves were constructed to ute two-step gradient. Tandem mass spectrometry (MS/ determine the prognostic predictive power of uAnCR. MS) was performed using an AB SCIEX Triple TOF 5600 Univariate ROC curves were considered statistically sig- fi mass spectrometer. Acquired spectra were searched ni cant if the area under the curve (AUC) differed from z against the 2011_6 release of the Human UniProtKB/ 0.5, as determined by the test. Optimal cut-offs were de- Swiss-Prot database (20,127 entries) with addition of com- termined by selecting the data point that minimized the mon contaminants (112 entries) using the Mascot search geometric distance from 100% sensitivity and 100% speci- fi engine with trypsin as the specified enzyme. Monoisotopic city on the ROC curve (24). To visualize the relationship masses were used, and the error tolerances were 10 ppm between uAnCR and length of stay, patients were strati- fi – and 0.5 Da for peptides and MS/MS fragments, respec- ed into tertiles by uAnCR. Kaplan Meier curves with tively. Mascot search results were loaded into Scaffold censoring for death were plotted. The log-rank test was (Proteome Software Inc), which used the Peptide Prophet used to compare the curves and the Holm-Sidak test was post hoc and Protein Prophet algorithms to validate peptide and used for pairwise comparison. Category free net fi protein identifications (21,22). The Scaffold quantitative reclassi cation improvement (cfNRI) was used to deter- values of identified proteins were normalized to the inter- mine if addition of uAnCR to a multivariate logistic re- nal standard HIV protein, and the relative abundance of gression model for prediction of risk increased the ability each protein is reported in normalized spectral counts. of the model to predict worsening of AKI (25,26). The risk prediction model consisted of the Cleveland Clinic cardiac surgery risk score and percent change in serum creatinine Angiotensinogen ELISA at the time the urine sample was collected (27). Statistical The Human Total Angiotensinogen Assay Kit (Immuno- tests were performed in Matlab or SigmaPlot. Biologic Laboratories Co. Ltd.) was used according to the manufacturer’s protocol. Values for intra- and interassay variability were 2.4% and 9.9%, respectively. Results Discovery of Candidate Prognostic AKI Biomarkers Urine Creatinine Determination We used LC-MS/MS to compare the urinary proteomic Urine creatinine was measured using the Jaffe assay. profiles of 12 patients who developed AKI after cardiac surgery, 6 of whom later required RRT and 6 of whom did Outcomes not. There were no statistically significant differences fi The primary outcome was worsening of AKI, de ned as between the two groups with respect to the demographic progression to a higher AKIN stage after the time of sample characteristics, sample collection time, use of cardiopul- collection. Secondary outcomes were progression to AKIN monary bypass, bypass time, type of surgery, preoperative stage 3, the need for renal replacement therapy (RRT) SCr, and SCr at the time of sample collection (Supplemental within 10 days of sample collection, progression to AKIN Table 1).