Proteomic Changes in Cerebrospinal Fluid of Presymptomatic and Affected Persons Carrying Familial Alzheimer Disease Mutations

Home , Acetylesterase, CD99, Enolase 2, Osteoglycin

ORIGINAL CONTRIBUTION Proteomic Changes in Cerebrospinal Fluid of Presymptomatic and Affected Persons Carrying Familial Alzheimer Disease Mutations

John M. Ringman, MD, MS; Howard Schulman, PhD; Chris Becker, PhD; Ted Jones, PhD; Yuchen Bai, PhD; Fred Immermann, MA, MStat; Gregory Cole, PhD; Sophie Sokolow, PhD; Karen Gylys, PhD; Daniel H. Geschwind, MD, PhD; Jeffrey L. Cummings, MD; Hong I. Wan, PhD

Objective: To identify cerebrospinal fluid (CSF) pro- Results: Fifty-six proteins were identified, represented by tein changes in persons who will develop familial Alz- multiple tryptic peptides showing significant differences be- heimer disease (FAD) due to PSEN1 and APP muta- tween MCs and NCs (46 upregulated and 10 downregu- tions, using unbiased proteomics. lated); 40 of these proteins differed when the analysis was restricted to asymptomatic individuals. Fourteen proteins Design: We compared proteomic profiles of CSF from have been reported in prior proteomic studies in late- individuals with FAD who were mutation carriers (MCs) onset AD, including amyloid precursor protein, transfer- ␣ ␤ and related noncarriers (NCs). Abundant proteins were rin, 1 -glycoprotein, complement components, afamin pre- depleted and samples were analyzed using liquid chro- cursor, spondin 1, plasminogen, hemopexin, and neuronal matography–electrospray ionization–mass spectro- pentraxin receptor. Many other proteins were unique to metry on a high-resolution time-of-flight instrument. our study, including calsyntenin 3, AMPA (␣-amino-3- Tryptic peptides were identified by tandem mass spectro- hydroxy-5-methyl-4-isoxazolepropionic acid) 4 gluta- metry. Proteins differing in concentration between the mate receptor, CD99 antigen, di-N-acetyl-chitobiase, and MCs and NCs were identified. secreted phosphoprotein 1.

Setting: A tertiary dementia referral center and a pro- Conclusions: We found much overlap in CSF protein teomic biomarker discovery laboratory. changes between individuals with presymptomatic and symptomatic FAD and those with late-onset AD. Our re- Participants: Fourteen FAD MCs (mean age, 34.2 sults are consistent with inflammation and synaptic loss years; 10 are asymptomatic, 12 have presenilin-1 early in FAD and suggest new presymptomatic biomark- [PSEN1] gene mutations, and 2 have amyloid precur- ers of potential usefulness in drug development. sor protein [APP] gene mutations) and 5 related NCs (mean age, 37.6 years). Arch Neurol. 2012;69(1):96-104

LZHEIMER DISEASE (AD) IS Entrez Gene 5663]) or amyloid precur- a growing public health sor protein (APP [NCBI Entrez Gene 351]) Author Affiliations: Mary S. problem, but our ability to genes enables the study of biomarkers in Easton Center for Alzheimer’s diagnose and treat it is in- asymptomatic persons in whom the de- Disease Research (Drs Ringman, adequate. Although bio- velopment of AD can be predicted with es- Cole, Sokolow, Gylys, and markers for AD are known, there is no di- sentially 100% certainty. The study of per- Cummings), Department of A Neurology, David Geffen School agnostic procedure accurate enough to sons at risk for FAD therefore provides of Medicine (Drs Ringman, screen for presymptomatic individuals in insight into biomarkers reflecting com- 1 Cole, Geschwind, and the general population. The AD patho- mon elements of disease pathogenesis. Cummings), and School of logic process is initiated long before overt Cerebrospinal fluid (CSF) offers sev- Nursing (Drs Sokolow and dementia manifests. The associated bio- eral advantages as a biological fluid for AD Gylys), University of California, chemical changes, if better understood, research. One of these advantages is that Los Angeles; Caprion could provide a basis for presymptom- CSF is in direct continuity with the brain Proteomics, Montreal, Quebec, atic diagnosis, be outcome measures for and therefore more directly reflects chemi- Canada (Drs Schulman, Becker, disease-modifying interventions, or help cal changes occurring there. Cerebrospi- and Jones); and Translational to identify novel therapeutic targets. nal fluid has a narrower dynamic range of and Molecular Medicine (Drs Bai and Wan) and Research on presymptomatic AD is protein concentration than does blood, Statistics (Dr Immermann), hampered by difficulties in prospective with fewer dominant high-abundance pro- BioTherapeutics Research and identification of persons who will de- teins complicating analysis of less abun- Development, Pfizer, Inc, velop AD. Familial AD (FAD) due to mu- dant ones. Accordingly, the markers most New Haven, Connecticut. tations in the presenilin-1 (PSEN1 [NCBI diagnostic and predictive of AD diagno-

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 96

Symptomatic Presymptomatic NCs FAD MCs FAD MCs Characteristic (n=5) (n=4) P Valuea (n=10) P Valueb Age (range), y 37.6 (29 to 55) 43.5 (37 to 55) .11 30.5 (23 to 43) .13 Adjusted age (range), yc −5.8 (−18 to −18) 0.3 (−12 to 12) .41 −15.5 (−22 to −5) .13 Female sex, No. (%) 3 (60) 3 (75) .60 8 (80) .41 Genetic alteration in family PSEN1 A431E (n=4) PSEN1 A431E (n=2) PSEN1 A431E (n=7) APP V717I (n=1) PSEN1 L235V (n=1) PSEN1 L235V (n=1) PSEN1 S212Y (n=1) APP V717I (n=2) APOE genotypes (2/3, 3/3, 3/4), No. (0, 3, 2) (0, 4, 0) .28 (3, 6, 1) .22 CSF A␤42 levels, mean (SD) 618.4 (100.1) 277.1 (162.9) .03 387.6 (130.1) .004 CSF t-tau levels, mean (SD), pg/mL 50.5 (9.4) 140.0 (59.9) .02 144.3 (58.6) .001

CSF p-tau181 levels, mean (SD), pg/mL 24.6 (9.2) 67.0 (29.0) .03 71.2 (31.8) .007

Abbreviations: A␤, beta-amyloid; CSF, cerebrospinal fluid; FAD, familial Alzheimer disease; MCs, mutation carriers; NCs, noncarriers. a P value for Mann-Whitney tests and Fisher exact tests comparing symptomatic MCs (Clinical Dementia Rating [CDR], Ͼ0) and all NCs. b P value for Mann-Whitney tests and Fisher exact tests comparing presymptomatic (CDR, 0) MCs and NCs. c Adjusted age indicates the age of participants relative to the median age of dementia diagnosis within their family.

sis to date are measurements of tau and amyloid-␤ (A␤) FAD in a tertiary dementia referral center. These 19 people were in CSF.2 Recent application of multiplexed antibody- from 10 families with or at risk for FAD due to PSEN1 (n=16) or based screens of proteins provides support for CSF bio- APP (n=3) mutations (Table 1). The Clinical Dementia Rating 9 markers beyond tau and A␤ providing additional diag- scale (CDR) was administered by investigators blind to muta- nostic information.3 tion status (except for persons with overt dementia). We compared participants who were mutation carriers (MCs) with those Advancements in mass spectrometry (MS)–based tech- who were related noncarriers (NCs) as the control population. nology have enabled the identification of proteins dif- All study procedures were approved by the University of Cali- ferentially expressed between populations. Such unbi- fornia, Los Angeles, Institutional Review Board, and all partici- ased approaches permit the identification of novel pants or their representatives gave informed consent. molecular changes and have been used to study various Age at onset of disease in FAD tends to be consistent within, illnesses, including AD. Such studies have focused on later but can vary among, families.10,11 Therefore, to compare par- stages of the disease, comparing control groups with in- ticipants relative to the expected onset of clinical disease, their dividuals having a diagnosis of AD. Comparison of 10 ages relative to the median age of disease diagnosis in their fami- or fewer people by 2-dimensional sodium dodecyl sulfate– lies were calculated (adjusted age). polyacrylamide gels or surface-enhanced laser desorption/ Of the 19 participants, 14 were MCs and 5 were NCs (Table 1). The mean ages were 34.2 and 37.6 years, respectively. Mean ad- ionization MS found a few changes in AD.4,5 In a larger 6 justed age for MCs was 11 years younger than the family- study, Finehout et al used 2-dimensional sodium do- specific age of dementia diagnosis. Among the 14 MCs, 10 were decyl sulfate–polyacrylamide gel to identify 23 proteins asymptomatic (CDR,0), 2 were mildly symptomatic but did not that separated groups. More extensive profiling was have dementia (CDR,0.5), 1 had mild dementia (CDR,1; S212Y achieved by Zhang et al,7 who used liquid chromatogra- PSEN1 mutation), and 1 had moderate dementia (CDR,2; A431E phy–mass spectrometry (LC-MS) with isotope labeling PSEN1 mutation). Eleven of the 14 MCs were women. Of the 10 in 17 patients with AD and 16 control participants. The asymptomatic MCs, 8 carried PSEN1 mutations (7 with A431E investigators found changes in many proteins, includ- and 1 with L235V) and 2 carried APP mutations (V717I). Of the ing APP and complement components. 2 MCs with CDR scores of 0.5, one carried the A431E mutation We performed proteomic analyses of CSF using high- and the other carried the L235V PSEN1 mutation. Of the 5 NCs, 3 were women and 2 were men. resolution LC-MS. This label-free method relies on a novel validated approach to quantification of LC-MS data.8 In- GENETIC TESTING dividual samples are neither mixed nor extensively ma- nipulated while sample preparation and MS conditions Extraction of DNA and genotyping of apolipoprotein E were are controlled, ensuring that changes in analyte intensi- performed using standard techniques. The presence of the A431E ties reflect their concentrations in one sample relative to and L235V substitutions in PSEN1 were assessed using restric- another. Using this technique, we studied CSF from per- tion fragment length polymorphism analyses. The participant sons with or at risk for inheriting FAD to identify pro- carrying the S212Y PSEN1 mutation was evaluated using a com- tein changes during the presymptomatic phase and in es- mercial test (Athena Diagnostics) in which the open reading tablished disease. frame of the coding region of the PSEN1 gene was sequenced. The presence of the V717I substitution in APP was assessed with direct sequencing. METHODS CSF ANALYSES POPULATION Cerebrospinal fluid was collected at various times of the day Participants were 19 persons consenting to lumbar puncture from within the same week as clinical assessments. The fluid was as- a total of 41 who were taking part in a comprehensive study of pirated and placed in polystyrene tubes on ice. Within 2 hours,

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 97

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 CSF was centrifuged and aliquotted into 0.5-mL siliconized poly- MCs with NCs. These analyses were performed in R, version propylene tubes. A protease inhibitor mixture containing apro- 2.7.2 (http://www.r-project.org), and later. No corrections for tinin, 1 µg/mL, and sodium azide, 0.05%, was added, and the multiple comparisons were made. samples were centrifuged at 1200g for 15 minutes and frozen Two protein lists were generated from these statistical com- ␤ to −80°C. The CSF was analyzed for A 42, t-tau, and p-tau181 parisons using heuristic protein-level assessment based on tryp- levels, using multiplex assays (INNO-BIA AlzBio3; Innogenet- tic peptides. For a protein to be included in a list, there had to ics) on standardized technology (xMAP; Luminex Corp). Cen- be more than 1 peptide showing a statistically significant change trifugation was repeated at 16 000g for 15 minutes at 4°C to (PϽ.05), substantial agreement between peptides regarding the remove debris after thawing. The CSF was then subjected to direction of change, at least 8% of the tracked molecular ions MS analysis for profiling and protein identification at a pro- for a given protein had to reach statistical significance, and at teomic biomarker discovery laboratory. Abundant proteins were least 2 molecular ions with PϽ.05. ␤ substantially depleted by an antibody-based affinity resin col- Cerebrospinal fluid levels of A 42, t-tau, and p-tau181 were umn (model MARS Hu-14; Agilent Technologies) to increase compared between groups by Mann-Whitney tests using com- the effective dynamic range, the total number of tracked pro- mercial software (PASW 18.0 IBM SPSS). teins, and accuracy of the measurements. The proteins depleted were albumin, IgG, IgA, IgM, transferrin, fibrinogen, PATHWAY ANALYSIS apolipoprotein A-I, apolipoprotein A-II, haptoglobin, ␣ ␣ ␣ 1-antitrypsin, 1-acid glycoprotein, 2-macroglobulin, comple- Weperformedpathwayanalysisinwhichnetworks,canonicalpath- ment C3, and transthyretin. ways, and functional processes for the proteins were identified from The depleted samples were analyzed with a 2-dimensional analysis 1 (Ingenuity Pathways Analysis; Ingenuity Systems). liquid chromatographic separation approach (DeepLook analy- sis12; Biomarker Discovery Sciences, PPD, Inc). Proteins were denatured, disulfide bonds were reduced, and sulfhydryl groups RESULTS were carboxymethylated before digestion by modified trypsin. Four fractions of peptides per sample were obtained by off-line TARGETED ASSAYS OF A␤42, (fraction collected) strong cation-exchange chromatography. The T-TAU, AND P-TAU tryptic peptides were then profiled. Individual molecules were 181 tracked across samples, and their differential expression was quan- ␤ tified by liquid chromatography–electrospray ionization mass Mean CSF A 42 levels were lower and t-tau and p-tau181 spectrometry on a high-resolution (RϾ5000, where R indicates levels were higher among symptomatic and presymp- resolving power, per the International Union of Pure and Applied tomatic MCs relative to NCs (Table 1). Chemistry) time-of-flight (TOF) instrument with an online reverse-phase capillary liquid chromatography using a 1-hour PROTEOMICS RESULTS solvent gradient of increasing acetonitrile in an aqueous solu- tion with formic acid, 0.1%. A total of 49 261 molecular ions were tracked and quantified. Median coefficients of variation (CVs) within the MC IDENTIFICATION OF PEPTIDES AND PROTEINS and NC groups were 21% to 22%. A total of 16 940 ions were linked to 7407 peptide sequences corresponding to For peptide identification, MS-MS spectra were acquired un- 600 proteins, 585 of which had 2 or more peptides re- der identical chromatographic conditions as for the profiling, ported. Although there were not many ions with a low P using a linear ion-trap mass spectrometer (model LTQ; Thermo value relative to the number of ions tracked, the consis- Scientific). Peptide identifications for MS-MS spectra were made, tent results between multiple peptides representing these and a high-confidence protein list was created from them 13,14 proteins revealed clear changes in many. From analysis 1, (ByOnic and ComByne programs; Palo Alto Research Center). a condensed list of 56 proteins with consistent evidence The ByOnic identification searches were performed using a decoy database in which equal numbers of proteins were in- of differences in concentration between all MCs and NCs cluded but their sequences were reversed. The appearance of was produced (Table 2). Although most of the proteins 4-6,16 decoy sequences among the identifications was used to deter- were novel, one-fourth (14 of 56) have been shown to mine an objective false discovery rate and retain only high- have differential levels in patients with AD. These include ␣ ␤ quality identifications. Ions observed in the 2-dimensional APP, transferrin, 1 -glycoprotein, complement compo- LC-MS analysis of the individual samples were linked to MS-MS nents, afamin precursor, spondin 1, and plasminogen. identifications, using accurate mass and chromatographic re- Twelve of the 14 proteins have shown changes in the same tention time, in a computation that included postacquisition 15 direction in at least 2 of these studies. In analysis 2, only nonparametric mass calibration. A differential quantifica- presymptomatic MCs were compared with those who were tion method was used to compare peptide and protein levels NCs. Among the 56 proteins from the list from analysis 1, among samples, as detailed by Wang et al,8 which relies on the changes in analyte intensities directly reflecting their concen- 41 were also found to be different comparing presymp- trations in one sample relative to another. tomatic MCs and NCs, and 38 were not selected in the condensed list for analysis 1 (Table 3). At least 2 of these, neuronal pentraxin receptor and hemopexin, have been STATISTICAL METHODS shown17,18 to be elevated in the CSF of patients with AD. We performed statistical analyses, comparing MCs with NCs for each tryptic peptide’s molecular ion tracked by MS, based PATHWAY ANALYSIS on t tests or Wilcoxon rank sum 2-sample statistics depending on the outcome of a Shapiro-Wilk test for normality. Two analy- A pathway analysis was performed on the 56-protein list ses were performed for each ion: analysis 1 compared all MCs generated from analysis 1. The first level examined the with NCs and analysis 2 compared all asymptomatic (CDR,0) functions linked to these proteins in the literature. Many

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 98

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 Table 2. The List of 56 Proteins Different Between FAD MCs and NCs in the CSF 2D Proteomic Studya

Changed in Fold Effect Presymptomatic Accession No. Protein Name Trend Change Size Participants? NP_001124.1 Afamin precursor Up 1.30 0.85 ϩ

NP_569711.2 ␣1-Type XVIII collagen isoform 3 precursor Down −1.13 −1.38 ϩ

NP_570602.2 ␣1B-Glycoprotein precursor Up 1.35 1.07 ϩ

NP_000005.2 ␣2-Macroglobulin precursor Up 1.23 1.28 − NP_001135750.1 Amyloid ␤(A4) precursorlike protein 2 isoform 4 Up 1.19 0.85 − NP_001129603.1 Amyloid ␤ A4 protein isoform g Up 1.23 1.34 ϩ NP_001636.1 Apolipoprotein C-I precursor Up 1.52 1.11 − NP_000051.1 Biotinidase precursor Up 1.59 1.24 ϩ NP_055207.1 Brain neuron cytoplasmic protein 1 Down −1.34 −1.76 ϩ NP_055533.2 Calsyntenin 3 Down −1.32 −0.91 − NP_666023.1 Cathepsin L1 preproprotein Up 1.35 1.17 − NP_001116370.1 CD99 antigen isoform b precursor Up 1.35 0.99 ϩ NP_000087.1 Ceruloplasmin precursor Up 1.41 0.98 ϩ NP_004379.1 Di-N-acetyl-chitobiase Up 1.30 1.12 ϩ NP_001139375.1 Complement component 2 isoform 2 preproprotein Up 1.49 1.48 ϩ NP_009224.2 Complement component 4A preproprotein Up 1.32 0.98 − NP_001002029.3 Complement component 4B preproprotein Up 1.56 0.97 − NP_001726.2 Complement component 5 preproprotein Up 1.51 1.03 ϩ NP_000056.2 Complement component 6 precursor Up 1.36 1.00 ϩ NP_000057.1 Complement component 8, ␤ polypeptide Up 1.55 1.34 ϩ preproprotein NP_001701.2 Complement factor B preproprotein Up 1.40 1.28 ϩ NP_001966.1 Enolase 2 Up 1.28 1.35 ϩ NP_001121089.1 Fructose-bisphosphate aldolase A Up 1.39 1.19 ϩ NP_001106283.1 Glutamate receptor, ionotrophic, AMPA 4 Up 1.27 0.78 ϩ isoform 3 precursor NP_001138648.1 Hypothetical protein LOC729956 (Homo sapiens) Down −1.18 −1.05 − NP_002169.1 Insulinlike growth factor binding protein 6 Down −1.55 −1.99 ϩ

NP_002206.2 Inter-␣ (globulin) inhibitor H1 Up 1.64 1.10 ϩ NP_001137435.1 L1 cell adhesion molecule isoform 3 precursor Up 1.16 1.00 ϩ

NP_002283.3 Laminin, ␤2 precursor Up 1.90 0.87 − NP_000419.1 Latent transforming growth factor ␤-binding protein 2 Up 1.26 0.82 −

NP_443204.1 Leucine-rich ␣2-glycoprotein 1 Up 1.35 0.99 ϩ NP_006491.2 Melanoma cell adhesion molecule Down −1.19 −1.12 ϩ NP_002490.2 Neogenin homologue 1 Up 1.29 1.39 − NP_776169.2 Neuronal growth regulator 1 Up 1.16 0.85 ϩ NP_148935.1 Osteoglycin preproprotein Down −1.14 −1.08 ϩ NP_443122.3 Peptidoglycan recognition protein 2 precursor Up 1.39 1.02 ϩ NP_000910.2 Peptidylglycine ␣-amidating monooxygenase Up 1.34 1.11 − isoform a preproprotein NP_000292.1 Plasminogen Up 1.50 1.18 ϩ NP_002846.3 Poliovirus receptor–related 1 isoform 1 Up 1.23 1.32 ϩ NP_005304.3 Protein disulfide-isomerase A3 precursor Down −1.37 −1.38 ϩ NP_001035802.1 Protein tyrosine phosphatase, receptor type, D isoform 5 precursor Down −1.21 −1.20 ϩ NP_002837.1 Protein tyrosine phosphatase, receptor type, N precursor Up 1.27 1.05 ϩ NP_005380.1 Protein-L-isoaspartate (D-aspartate) Up 1.31 0.99 − O-methyltransferase NP_872271.1 Pyruvate kinase, muscle isoform M1 Up 1.22 1.21 ϩ

P01860.2 Ig ␥3-chain C region Up 1.56 1.11 ϩ NP_001030024.1 Secreted modular calcium-binding protein 1 isoform 1 Up 1.87 1.57 − NP_001035147.1 Secreted phosphoprotein 1 isoform a Up 1.38 1.07 ϩ NP_000573.1 Secreted phosphoprotein 1 isoform b Up 1.60 1.21 ϩ NP_001035149.1 Secreted phosphoprotein 1 isoform c Up 1.44 1.04 − NP_066938.2 Seizure-related 6 homologue (mouse)–like precursor Up 1.27 1.21 ϩ NP_001076.2 Serpin peptidase inhibitor, clade A, member 3 precursor Up 1.40 0.99 ϩ NP_612449.2 Slitlike 2 Up 1.34 1.04 ϩ NP_006099.2 Spondin 1, extracellular matrix protein Up 1.19 1.35 ϩ NP_003093.2 Superoxide dismutase 3, extracellular precursor Down −1.35 −1.46 ϩ NP_001054.1 Transferrin Up 1.38 0.97 ϩ NP_003369.2 VGF nerve growth factor–inducible precursor Up 1.30 0.89 ϩ

Abbreviations: AMPA, ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; CSF, cerebrospinal fluid; 2D, 2-dimensional; NCs, noncarriers; ϩ, change in presymptomatic (Clinical Dementia Rating [CDR] score, 0) familial Alzheimer disease (FAD) mutation carriers (MCs) as well as all the group of all MCs; −, no change in presymptomatic FAD MCs. a Direction of change (trend), magnitude of change, effect size, and whether the change was evident in asymptomatic (Clinical Dementia Rating, 0) MCs are indicated.

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 99

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 Table 3. The List of 38 Proteins Different Between Presymptomatic (CDR, 0) FAD MC and NC Cohorts Not Found When Affected MCs Were Includeda

Accession No. Protein Name Trend Fold Change Effect Size NP_002070.1 Aspartate aminotransferase 1 Up 1.33 1.79 NP_002071.2 Aspartate aminotransferase 2 precursor Up 1.24 1.83

NP_000713.2 Calcium channel, voltage-dependent, ␣2/delta subunit 1 Up 1.30 1.30 NP_002405.1 CD99 antigen isoform a precursor Up 1.36 1.32 NP_054860.1 Cell recognition molecule Caspr2 precursor Up 1.29 0.96 NP_004377.2 Chondroitin sulfate proteoglycan 3 Up 1.26 1.39 NP_000497.1 Coagulation factor II preproprotein Up 1.33 1.05 NP_000054.2 Complement component 2 isoform 1 preproprotein Up 1.41 1.47 NP_000055.2 Complement component 3 precursor Up 1.33 1.08 NP_733746.1 Cytosolic sialic acid 9-O-acetylesterase homologue Up 1.39 1.35 NP_002014.2 Fibromodulin precursor Down −1.60 −1.63 NP_001987.2 Fibulin 1 isoform C precursor Down −1.35 −1.30 NP_000604.1 Hemopexin Up 1.42 1.09 NP_000590.1 Insulinlike growth factor binding protein 5 Up 1.28 1.07 NP_001001851.1 Inter-␣ trypsin inhibitor heavy chain precursor 5 isoform 3 Up 1.39 1.08 NP_001095886.1 Kininogen 1 isoform 1 Up 1.48 1.05 NP_001008701.1 Latrophilin 1 isoform 1 precursor Down −1.17 −1.46 NP_005568.2 Lipoprotein Lp(a) precursor Up 1.65 1.17 NP_079032.2 Multimerin 2 precursor Up 1.34 1.49 NP_002505.1 Nephroblastoma overexpressed precursor Up 1.60 1.41

NP_620060.1 Neurexin 2 isoform ␣2 precursor Down −1.21 −1.31 NP_877421.2 Neuroblastoma, suppression of tumorigenicity 1 isoform 1 Down −1.27 −1.61 NP_002514.1 Neuronal pentraxin II Up 1.27 1.07 NP_055108.2 Neuronal pentraxin receptor Up 1.26 1.21 NP_001116224.1 Neuroserpin precursor Up 1.18 1.18 NP_775733.3 Papilin Up 1.26 1.80 NP_000883.2 Plasma kallikrein B1 precursor Up 1.38 1.21 NP_065138.2 Plexin domain containing 1 precursor Up 1.37 1.03 XP_001715435.1 PREDICTED: hypothetical protein Up 1.36 1.20 NP_001035519.1 Protocadherin 17 precursor (Homo sapiens) Up 1.22 1.19 Q9Y4C0.4 Neurexin-3-␣ Up 1.41 1.13 NP_006206.2 Serine (or cysteine) proteinase inhibitor, clade A, member 4 Up 1.40 1.62 NP_443142.1 Slit and trk-like 1 protein Down −1.24 −1.16 NP_003246.1 TIMP metallopeptidase inhibitor 2 precursor Down −1.14 −1.46 NP_006455.2 trans-Golgi network protein 2 Down −1.23 −1.58 NP_001129174.1 Tyrosine 3/tryptophan 5-monooxygenase activation protein, zeta polypeptide Up 1.34 1.51 NP_061828.1 Ubiquitin B precursor Up 1.25 1.40 NP_006867.1 UDP-GlcNAc:␤Gal ␤-1,3-N-acetylglucosaminyltransferase 1 Up 1.19 1.18

Abbreviations: CDR, Clinical Dementia Rating; FAD, familial Alzheimer disease; MC, mutation carrier; NC, noncarrier; TIMP, tissue metalloproteinase inhibitor; UDP, uridine diphosphate. a Direction of change (trend), magnitude of change, and effect size are indicated.

proteins are involved in cellular-mediated immune re- COMMENT sponse and cell signaling functions and have been associated with inflammatory and neurologic diseases. The next level of analysis investigated whether functional cel- In this CSF proteomic study of persons inheriting FAD lular networks in the literature could be identified that mutations, most of whom were presymptomatic, we iden- link the proteins. The more nodes of a particular net- tified known and novel candidate biomarkers. Differ- work that are represented by candidate proteins, the more ences were found in 56 CSF proteins between the MC likely this network is relevant to the biological question and NC cohorts based on univariate analysis. The fold being investigated. Our analysis revealed 3 networks, with changes among the 56 proteins were relatively small, rang- the first 2 having substantial numbers of proteins iden- ing from 1.18 to 1.90 (mean, 1.33). The differences in ␤ tified from the present study. Network 1 is related to an- CSF A 42, t-tau, and p-tau181 found in MCs were large tigen presentation, cell-mediated immune response, and and consistent with reported fold changes for these tra- humoral immune response (Figure 1). This network ditional markers in persons with mild cognitive impair- contains 29 proteins, 25 of which were upregulated and ment and AD of later onset. In this study, the changes 4 of which were downregulated in MCs. Network 2 is were detected 10 or more years before diagnosis. associated with metabolic disease, renal disease, uro- Many more of the 56 proteins are upregulated (n=46) logic disease, and molecular transport and contains 15 than downregulated (n=10) in MCs. The finding of both proteins (12 upregulated and 3 downregulated in MCs) upregulated and downregulated proteins suggests that our (Figure 2). results are not due to nonspecific brain degeneration,

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 100

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 PP Fx: Survival of granule cells PP Fx: Quantity of cerebral cortex cells ENO2 L1CAM VASN PP

Fx: Cell viability of neurons L PP L Fx: Outgrowth of neurites CTSL1 Secretase γ PP PP Fx: Alzheimer disease TF PP AFM A, PP, RB, RE, TR L L PP COL18A1 PLG RB TR LO I PP Fx: Survival of microglia TR MB A, E, EC, L, LO, PP A2M A, RB A, L E E PP APLP2 CLSTN3 E, PD A, TR LO PP I, PP Collagen(s) CP E NFκB (complex) MmpL LO PP, RB E, PP PP LO, PP PP TR A, E, RB Fx: Familial Alzheimer disease E, L, LO, M, PP, RB, RE, TR E A, E, T, TR E, LO E IgG A, RB L RB, TR SPON1 E, T, TR APP PD A, E, LO, PP PP E, L, LO, PP, RB E, L, RB, TR A, E RB IL1 E, PP, RBSERPINA3 MB RB E L E, PD LDL E, RB IGHG3 A, PP A, PP E A PP Fx: Neurodegeneration A A, TR E I A, L PP, RE, TR PTPRD Δ TR Complement component 1 LO CFB TR C4 E, LO L, PP NADPH oxidase MB, RE E L, PP E MB A Δ Fx: Corticobasal degeneration RE C4B L, PP A Fx: Cerebral amyloid angiopathy of cerebrum PP L, RE E, L, PP RB C5 C4A PP, RB E, L, PP PP C6 PP PP C2 PP PP C8 C8B MB

Network shapes Path designer shapes Relationships Relationships labels

Cytokine Cytokine/growth factor A B A Activation Binding only Growth factor Drug E Expression (includes metabolism/synthesis Chemical/drug/toxicant Chemical/toxicant A B for chemicals) Inhibits Enzyme Enzyme EC Enzyme catalysis A B I Inhibition G-protein coupled receptor G-protein coupled receptor Acts on L Proteolysis (includes degradation for chemicals) Ion channel Ion channel A B LO Localization Inhibits AND acts on M Biochemical modification Kinase Kinase MB Group/complex membership A B Ligand-dependent nuclear receptor Ligand-dependent nuclear receptor Leads to P Phosphorylation/dephosphorylation

Peptidase Peptidase PD Protein-DNA binding A B PP Protein-protein binding Translocates to Phosphatase Phosphatase RB Regulation of binding Transcription regulator Transcription regulator A B RE Reaction Reaction Translation regulator Translation regulator TR Translocation

Transmembrane receptor Transmembrane receptor Enzyme Catalysis A B Transporter Transporter Reaction Complex/group MicroRNA Direct interaction MicroRNA Mature microRNA

Mature microRNA Complex/group/other Indirect interaction Other

Figure 1. Network 1 (antigen presentation, cell-mediated immune response, and humoral immune response) generated by pathway analysis of the hot list. This network is displayed as nodes (genes or gene products) and edges. “Acts on” and “inhibits” edges may also include a binding event. Lines indicate biological relationships between nodes. Blue proteins are downregulated in individuals who are mutation carriers, and red proteins are upregulated. Uncolored proteins were not found to be different in our study but are linked by pathway analyses. Fx indicates function.

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 101

Fx: Progressive motor neuropathy

PP Fx: First-onset paranoid schizophrenia PP PKM2 ADCYAP1R1 NEO1 PTPRN

E RB E E A LO, RB VGF E RAPGEF3 FSH GRIA4 A PLC A LO A ANG A MCAM A A P PP, TR PP P TR MB A A JnK PDGFPP BB Pkc(s) A A, LO PP, TR A, P, TR E LAMB2 E A PPA Akt A A, LO, PA A E ERO1L A A, P A, LO, P PP PDIA3 A, LO, TR A, P P A, E PI3K P RYR3 Fx: Morphology of synapse ERK A, P Tgf beta A, P A E A A, I, PP P Fx: Tumorigenesis of brain cancer cell lines FPR1 A P PP E LO TR A PP A P E, PPP E E, LO, TR P38 MAPK A, P Δ PP IL 12M (complex)A, E, P A CR1 MHC class I (complex) E LTBP2

M, P M, RB E LO E M ∗ PP SPP1 LO LRG1 Histone h3 PD CD99 (includes EG:4267) P ALDOA PD PAM Fx: Demyelination of spinal cord Ck2 P

Fx: Experimental autoimmune uveoretinitis

Figure 2. Network 2 (metabolic disease, renal and urologic disease, and molecular transport) generated by pathway analysis of the hot list. This network is also displayed as nodes (genes or gene products) and edges. Fx indicates function. *Indicates that components mapped to multiple protein database entries of a single gene. For further explanation of the lines and symbols used, see Figure 1 and its legend.

degradation of the blood-brain barrier, or contamina- be related to neurite growth, synaptogenesis, and syn- tion from blood. Such global changes would be attenu- aptic plasticity. The APP peptides from the soluble ex- ated by the normalization algorithm in the analysis soft- tracellular region were consistently elevated. Our find- ware. ing of elevated CSF APP before diagnosis in the study Fourteen proteins found to be differentially ex- participants with FAD mutations is consistent with mod- pressed in previous studies were also seen in our study, els in which A␤42 aggregates induce a positive feed- indicating that identifiable molecular changes occur 10 back cascade that elevates APP production through BACE years before diagnosis. Transthyretin, albumin, and apo- (beta-site APP cleaving enzyme) activity.20 Both the cur- lipoprotein A15 may also change prior to the appear- rent and previously reported21 data in the same popula- ance of symptoms but were among the abundant pro- tion determined with use of immunoassays found re- teins depleted from our samples. There was sufficient duced levels of A␤42 in the CSF of the MCs, a finding ␣ ␤ residual 2-macroglobulin and complement component thought to represent selective deposition of A 42 in the 3 after depletion to identify it as differentially ex- brain. The short A␤ peptides were likely removed in the pressed, consistent with previous findings.3 Multiple proteomic processing used in this study, which uses a complement components were elevated, consistent with 5-kDa filter cutoff prior to tryptic digestion of these previous studies19 indicating an early role of inflamma- proteins. tion in AD. Some of the proteins showing the most ro- Secreted phosphoprotein 1 (ie, osteopontin), a cyto- bust differences between MCs and NCs (CD99 antigen kine expressed by activated T cells, was elevated in the isoform b, di-N-acetyl chitobiase, and secreted phospho- CSF of MCs. Osteopontin has been found to be elevated protein 1 isoform b) were unique, and their relationship in the pyramidal cells in the brains of persons with AD22 to AD pathogenesis is unclear. and in transgenic mice23; in addition, there is a single re- Numerous peptides derived from APP were detected port24 of it being elevated in the CSF of persons with AD. and quantified. The median concentration change ob- Osteopontin is also elevated in multiple sclerosis,25 where served in these peptides is 23% upward for the MC co- it might serve as a therapeutic target. A similar role might hort. The normal functions of APP are not clear but may be considered in AD.

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 102

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 Superoxide dismutase 3 was decreased in the CSF of the In summary, although the statistical power of the FAD MC group. The activity of superoxide dismutase has present study was limited by small numbers, many novel been found26 to be decreased in the CSF of persons with proteomic differences were found in addition to others AD. Diminished superoxide dismutase activity could con- consistent with prior studies. We identified candidate bio- tribute to oxidative stress and therefore propagate disease markers associated with mutation status that suggest pathologic characteristics. However, consistent with a pro- changes occurring a decade before clinical dementia, in- teomic study18 of CSF in postmortem AD cases, we found cluding increases in APP, increases in inflammatory mark- elevated levels of hemopexin, a protein that binds heme, ers, and changes in proteins related to synaptic plastic- in presymptomatic MCs, thus preventing its pro-oxidant ity. These measures may be useful in diagnosis, patient and proinflammatory effects.27 stratification, and monitoring of response to therapy. Rep- We also observed changes in several proteins related lication in a larger FAD population, longitudinal analy- to excitatory synapses. Neuronal pentraxin 2, neuronal sis of the same individuals, and comparison with changes pentraxin receptor, and the AMPA (␣-amino-3-hydroxy- in mild cognitive impairment and AD of late onset would 5-methyl-4-isoxazolepropionic acid) 4 glutamate recep- provide further validation of our findings. tor were found to be elevated in presymptomatic MCs. Neuronal pentraxin 2 is a secreted protein homologous Accepted for Publication: May 17, 2011. to acute-phase reactants and therefore an indicator of in- Correspondence: John M. Ringman, MD, MS, Mary S. flammation. Neuronal pentraxin receptor was found to Easton Center for Alzheimer’s Disease Research, 10911 be elevated in proteomic studies of CSF in patients with Weyburn Ave, Ste 200, Los Angeles, CA 90095 (jringman AD,6,17 but NP1 and neuronal pentraxin receptor were @mednet.ucla.edu). decreased in another such study.16 Neuronal pentraxin Author Contributions: Study concept and design: Ring- receptor is a transmembrane protein expressed on neu- man, Schulman, Becker, Cummings, and Wan. Acquisi- rons and glia that binds AMPA receptors and plays a role tion of data: Ringman, Becker, Sokolow, Gylys, and in synaptogenesis.28 Elevation of these proteins in CSF Geschwind. Analysis and interpretation of data: Schul- could represent synapse turnover or loss occurring early man, Becker, Jones, Bai, Immermann, Cole, Sokolow, in the course of the disease. Calsyntenin 3, a postsynap- Geschwind, Cummings, and Wan. Drafting of the manu- tic membrane protein with putative Ca2ϩ-binding capac- script: Ringman, Schulman, Jones, Cummings, and Wan. ity associated with ␥-aminobutyric acid–ergic neurons, Critical revision of the manuscript for important intellec- was decreased in the CSF of MCs in our study. The cal- tual content: Ringman, Schulman, Becker, Jones, Bai, Im- sytenins have been linked to APP transport.29 mermann, Cole, Sokolow, Gylys, Geschwind, and Wan. A potential criticism of this study is the apparent lack Statistical analysis: Jones and Immermann. Obtained fund- of control for multiple comparisons. In the present study, ing: Schulman, Cummings, and Wan. Administrative, tech- P values were computed at the ion levels, whereas the nical, and material support: Ringman, Schulman, Becker, tables are protein-level summaries of peptide results and Cole, Sokolow, Gylys, Geschwind, and Cummings. Study consistent changes across peptides representing a single supervision: Ringman, Schulman, Geschwind, and Wan. protein were required to make the list. Using the cur- Financial Disclosure: None reported. rent methods in which hundreds of peptides were pro- Funding/Support: This study was supported by grant PHS filed, no false-positives were found at the protein level K08 AG-22228 (National Institute on Aging), grant 04- in an anonymized comparison of proteomics groups in 35522 (California Department of Health Services), Alzhei- which identical samples were spiked with differing lev- mer’s Disease Research Center grant P50 AG-16570 (Na- els of 12 proteins. tional Institute on Aging), National Institutes of Health grant There are many advantages to the study of persons with 5RC2NS069502 (National Institute of Neurological Dis- FAD mutations. Discovery of biomarkers for late-onset orders and Stroke) (Dr Schulman), the Easton Consor- neurodegenerative diseases that develop over decades is tium for Alzheimer’s Disease Drug Discovery and Biomark- confounded by the fact that disease-associated markers ers, Pfizer Pharmaceuticals, General Clinical Research may also change with normal aging.30 In addition, some Centers Program grant M01-RR00865, the Sidell Kagan healthy control participants in their 70s and 80s may be Foundation, the Shirley and Jack Goldberg Trust, and Cali- presymptomatic for neurodegenerative diseases diag- fornia Alzheimer’s Disease Center grant 09-11408. nosed after the study. Another advantage of studying persons at risk for FAD is that the population is more ho- REFERENCES mogeneous, enabling smaller cohort sizes, and that control participants and those who are presymptomatic carriers 1. Mattsson N, Blennow K, Zetterberg H. CSF biomarkers: pinpointing Alzheimer are younger, with fewer age-dependent neurologic and pathogenesis. Ann N Y Acad Sci. 2009;1180:28-35. somatic changes. There are, however, potential limita- 2. Blennow K, Hampel H, Weiner M, Zetterberg H. Cerebrospinal fluid and plasma tions in the generalizability of findings in FAD to late- biomarkers in Alzheimer disease. Nat Rev Neurol. 2010;6(3):131-144. 3. Hu WT, Chen-Plotkin A, Arnold SE, et al. Novel CSF biomarkers for Alzheimer’s onset AD. It is thought that the pathogenesis of FAD is disease and mild cognitive impairment. Acta Neuropathol. 2010;119(6):669- driven by a relative overproduction of A␤42, whereas late- 678. onset AD may be due to decreased degradation of A␤.A 4. Carrette O, Demalte I, Scherl A, et al. A panel of cerebrospinal fluid potential bio- previous investigation31 in this population revealed a markers for the diagnosis of Alzheimer’s disease. Proteomics. 2003;3(8):1486- 1494. unique pattern of APP-derived peptides in carriers of the 5. Puchades M, Hansson SF, Nilsson CL, Andreasen N, Blennow K, Davidsson P. A431E PSEN1 mutation that was not present in CSF from Proteomic studies of potential cerebrospinal fluid protein markers for Alzhei- persons with late-onset AD. mer’s disease. Brain Res Mol Brain Res. 2003;118(1-2):140-146.

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 103

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 6. Finehout EJ, Franck Z, Choe LH, Relkin N, Lee KH. Cerebrospinal fluid proteomic 20. Tamagno E, Bardini P, Guglielmotto M, Danni O, Tabaton M. The various aggre- biomarkers for Alzheimer’s disease. Ann Neurol. 2007;61(2):120-129. gation states of ␤-amyloid 1-42 mediate different effects on oxidative stress, neu- 7. Zhang J, Sokal I, Peskind ER, et al. CSF multianalyte profile distinguishes Alz- rodegeneration, and BACE-1 expression. Free Radic Biol Med. 2006;41(2): heimer and Parkinson diseases. Am J Clin Pathol. 2008;129(4):526-529. 202-212. 8. Wang G, Wu WW, Pisitkun T, Hoffert JD, Knepper MA, Shen RF. Automated quan- 21. Ringman JM, Younkin SG, Pratico D, et al. Biochemical markers in persons with tification tool for high-throughput proteomics using stable isotope labeling and preclinical familial Alzheimer disease. Neurology. 2008;71(2):85-92. LC-MSn. Anal Chem. 2006;78(16):5752-5761. 22. Wung JK, Perry G, Kowalski A, et al. Increased expression of the remodeling- 9. Morris JC. Clinical dementia rating: a reliable and valid diagnostic and staging mea- and tumorigenic-associated factor osteopontin in pyramidal neurons of the Alz- sure for dementia of the Alzheimer type. Int Psychogeriatr. 1997;9(suppl 1): heimer’s disease brain. Curr Alzheimer Res. 2007;4(1):67-72. 173-178. 23. Wirths O, Breyhan H, Marcello A, Cotel MC, Bru¨ck W, Bayer TA. Inflammatory 10. Fox NC, Kennedy AM, Harvey RJ, et al. Clinicopathological features of familial changes are tightly associated with neurodegeneration in the brain and spinal Alzheimer’s disease associated with the M139V mutation in the presenilin 1 gene: cord of the APP/PS1KI mouse model of Alzheimer’s disease. Neurobiol Aging. pedigree but not mutation specific age at onset provides evidence for a further 2010;31(5):747-757. genetic factor. Brain. 1997;120(pt 3):491-501. 24. Comi C, Carecchio M, Chiocchetti A, et al. Osteopontin is increased in the cere- 11. Murrell J, Ghetti B, Cochran E, et al. The A431E mutation in PSEN1 causing fa- brospinal fluid of patients with Alzheimer’s disease and its levels correlate with milial Alzheimer’s disease originating in Jalisco State, Mexico: an additional fif- cognitive decline. J Alzheimers Dis. 2010;19(4):1143-1148. teen families. Neurogenetics. 2006;7(4):277-279. 25. Braitch M, Nunan R, Niepel G, Edwards LJ, Constantinescu CS. Increased os- 12. Roy SM, Becker CH. Quantification of proteins and metabolites by mass spectro- teopontin levels in the cerebrospinal fluid of patients with multiple sclerosis. Arch metry without isotopic labeling. Methods Mol Biol. 2007;359:87-105. Neurol. 2008;65(5):633-635. 13. Bern M, Cai Y, Goldberg D. Lookup peaks: a hybrid of de novo sequencing and database search for protein identification by tandem mass spectrometry. Anal 26. De Deyn PP, Hiramatsu M, Borggreve F, et al. Superoxide dismutase activity in Chem. 2007;79(4):1393-1400. cerebrospinal fluid of patients with dementia and some other neurological disorders. 14. Bern M, Goldberg D. Improved ranking functions for protein and modification- Alzheimer Dis Assoc Disord. 1998;12(1):26-32. site identifications. J Comput Biol. 2008;15(7):705-719. 27. Tolosano E, Fagoonee S, Morello N, Vinchi F, Fiorito V. Heme scavenging and 15. Becker CH, Kumar P, Jones T, Lin H. Nonparametric mass calibration using hun- the other facets of hemopexin. Antioxid Redox Signal. 2010;12(2):305- dreds of internal calibrants. Anal Chem. 2007;79(4):1702-1707. 320. 16. Abdi F, Quinn JF, Jankovic J, et al. Detection of biomarkers with a multiplex quan- 28. Sia GM, Be´ı¨que JC, Rumbaugh G, Cho R, Worley PF, Huganir RL. Interaction of the titative proteomic platform in cerebrospinal fluid of patients with neurodegen- N-terminal domain of the AMPA receptor GluR4 subunit with the neuronal pen- erative disorders. J Alzheimers Dis. 2006;9(3):293-348. traxin NP1 mediates GluR4 synaptic recruitment. Neuron. 2007;55(1):87-102. 17. Yin GN, Lee HW, Cho JY, Suk K. Neuronal pentraxin receptor in cerebrospinal 29. Ludwig A, Blume J, Diep TM, et al. Calsyntenins mediate TGN exit of APP in a fluid as a potential biomarker for neurodegenerative diseases. Brain Res. 2009; kinesin-1–dependent manner. Traffic. 2009;10(5):572-589. 1265:158-170. 30. Zhang J, Goodlett DR, Peskind ER, et al. Quantitative proteomic analysis of age- 18. Roher AE, Maarouf CL, Sue LI, Hu Y, Wilson J, Beach TG. Proteomics-derived related changes in human cerebrospinal fluid. Neurobiol Aging. 2005;26(2): cerebrospinal fluid markers of autopsy-confirmed Alzheimer’s disease. Biomarkers. 207-227. 2009;14(7):493-501. 31. Portelius E, Andreasson U, Ringman JM, et al. Distinct cerebrospinal fluid amyloid 19. Akiyama H, Barger S, Barnum S, et al. Inflammation and Alzheimer’s disease. beta peptide signatures in sporadic and PSEN1 A431E-associated familial Alzhei- Neurobiol Aging. 2000;21(3):383-421. mer’s disease. Mol Neurodegener. 2010;5:2. doi:10.1186/1750-1326-5-2.

ARCH NEUROL / VOL 69 (NO. 1), JAN 2012 WWW.ARCHNEUROL.COM 104