Genome-Wide Analysis of Genetic Loci Associated with Alzheimer Disease

ORIGINAL CONTRIBUTION

Genome-wide Analysis of Genetic Loci Associated With Alzheimer Disease

Sudha Seshadri, MD; Annette L. Fitzpatrick, PhD; Context Genome-wide association studies (GWAS) have recently identified CLU, M. Arfan Ikram, MD, PhD; Anita L. DeStefano, PICALM, and CR1 as novel genes for late-onset Alzheimer disease (AD). PhD; Vilmundur Gudnason, MD, PhD; Merce Objectives To identify and strengthen additional loci associated with AD and Boada, MD, PhD; Joshua C. Bis, PhD; Albert confirm these in an independent sample and to examine the contribution of recently V. Smith, PhD; Minerva M. Carrasquillo, PhD; identified genes to AD risk prediction in a 3-stage analysis of new and previously pub- Jean Charles Lambert, PhD; Denise Harold, PhD; lished GWAS on more than 35 000 persons (8371 AD cases). Elisabeth M. C. Schrijvers, MD; Reposo Ramirez- Lorca, PhD; Stephanie Debette, MD, PhD; W. T. Design, Setting, and Participants In stage 1, we identified strong genetic asso- Ͻ −3 Longstreth Jr, MD; A. Cecile J. W. Janssens, PhD; ciations (P 10 ) in a sample of 3006 AD cases and 14 642 controls by combining new data from the population-based Cohorts for Heart and Aging Research in Ge- V. Shane Pankratz, PhD; Jean Franc¸ois Dartigues, nomic Epidemiology consortium (1367 AD cases [973 incident]) with previously re- PhD; Paul Hollingworth, PhD; Thor Aspelund, ported results from the Translational Genomics Research Institute and the Mayo AD PhD; Isabel Hernandez, MD; Alexa Beiser, PhD; GWAS. We identified 2708 single-nucleotide polymorphisms (SNPs) with PϽ10−3.In Lewis H. Kuller, MD; Peter J. Koudstaal, MD, stage 2, we pooled results for these SNPs with the European AD Initiative (2032 cases PhD; Dennis W. Dickson, MD; Christophe Tzourio, and 5328 controls) to identify 38 SNPs (10 loci) with PϽ10−5. In stage 3, we com- MD; Richard Abraham, PhD; Carmen Antunez, bined data for these 10 loci with data from the Genetic and Environmental Risk in AD MD; Yangchun Du, PhD; Jerome I. Rotter, MD; consortium (3333 cases and 6995 controls) to identify 4 SNPs with PϽ1.7ϫ10−8. These Yurii S. Aulchenko, PhD; Tamara B. Harris, MD; 4 SNPs were replicated in an independent Spanish sample (1140 AD cases and 1209 Ronald C. Petersen, MD; Claudine Berr, MD, PhD; controls). Genome-wide association analyses were completed in 2007-2008 and the Michael J. Owen, MB, ChB, PhD; Jesus Lopez- meta-analyses and replication in 2009. Arrieta, MD; Badri N. Vardarajan, MS; James Main Outcome Measure Presence of Alzheimer disease. T. Becker, PhD; Fernando Rivadeneira, MD, PhD; Results Two loci were identified to have genome-wide significance for the first time: Michael A. Nalls, PhD; Neill R. Graff-Radford, rs744373 near BIN1 (odds ratio [OR],1.13; 95% confidence interval [CI],1.06-1.21 MD; Dominique Campion, MD, PhD; Sanford per copy of the minor allele; P=1.59ϫ10−11) and rs597668 near EXOC3L2/BLOC1S3/ Auerbach, MD; Kenneth Rice, PhD; Albert MARK4 (OR, 1.18; 95% CI, 1.07-1.29; P=6.45ϫ10−9). Associations of these 2 loci Hofman, MD, PhD; Palmi V. Jonsson, MD; Helena plus the previously identified loci CLU and PICALM with AD were confirmed in the Schmidt, MD, PhD; Mark Lathrop, PhD; Thomas Spanish sample (PϽ.05). However, although CLU and PICALM were confirmed to be H. Mosley, PhD; Rhoda Au, PhD; Bruce M. Psaty, associated with AD in this independent sample, they did not improve the ability of a MD, PhD; Andre G. Uitterlinden, PhD; Lindsay model that included age, sex, and APOE to predict incident AD (improvement in area A. Farrer, PhD; Thomas Lumley, PhD; Agustin under the receiver operating characteristic curve from 0.847 to 0.849 in the Rotter- Ruiz, MD, PhD; Julie Williams, PhD; Philippe dam Study and 0.702 to 0.705 in the Cardiovascular Health Study). Amouyel, MD, PhD; Steve G. Younkin, PhD; Conclusions Two genetic loci for AD were found for the first time to reach genome- Philip A.Wolf, MD; Lenore J. Launer, PhD; wide statistical significance. These findings were replicated in an independent popu- Oscar L. Lopez, MD; Cornelia M. van Duijn, PhD; lation. Two recently reported associations were also confirmed. These loci did not im- Monique M. B. Breteler, MD, PhD prove AD risk prediction. While not clinically useful, they may implicate biological for the CHARGE, GERAD1, pathways useful for future research. and EADI1 Consortia JAMA. 2010;303(18):1832-1840 www.jama.com

NE OF EVERY 5 PERSONS AGED 65 years is predicted to de- The apparent substantial heritability of RefSeq NM_000484]).3 Initial genome- velop Alzheimer disease late-onset AD2 is inadequately ex- wide association studies (GWAS) iden- (AD) in their lifetime, and plained by genetic variation within the tified putative new candidate genes geneticO variants may play an important well-replicated genes (apolipoprotein E 1 Author Affiliations are available online at http://www part in the development of the disease. [APOE; RefSeq NG_007084], preseni- .jama.com. lin-1 [PSEN1; RefSeq NG_007386], pre- Corresponding Author: Monique M. B. Breteler, MD, PhD, senilin-2 [PSEN2; RefSeq NG_007381], Department of Epidemiology Erasmus MC, University For editorial comment see p 1864. Medical Center Rotterdam, PO Box 2040, 3000 CA Rot- and amyloid beta precursor protein [APP; terdam, the Netherlands ([email protected]).

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

(GRB2-associated binding protein consortium15 with GWAS data from the cluding the Mayo AD GWAS, the only [GAB2; RefSeq NG_016171], protocad- Translational Genomics Research In- overlapping study).13 All participants herin 11 x-linked [PCDH11X; RefSeq stitute (TGEN) public release data- (or their authorized proxies) in the con- NG_016251], lecithin retinol acyltrans- base4 and the Mayo AD GWAS.5 The tributing studies gave written in- ferase [LRAT; RefSeq NG_009110], and sample characteristics of the partici- formed consent including for genetic transient receptor potential cation pants contributing to this discovery analyses. Local institutional review channel, subfamily C, member 4– stage are summarized in TABLE 1.In boards approved study protocols. De- associated protein [TRPC4AP; RefSeq stage 2, we combined results for our tails of study sample selection for the NM_015638])4-6 and regions of interest most suggestive findings (single- contributing studies are described in (eg, on chromosomes 14q, 10q, and nucleotide polymorphisms [SNPs] with section 2 of the eAppendix (section 1 12q),7-10 but no locus outside of the APOE PϽ10−3) with corresponding results in lists abbreviations used in this article) region consistently reached genome- the EADI1 consortium.14 In stage 3, we and in eFigure 1, parts A through D wide significance.4,11,12 These disappoint- combined results for the most promis- (available at http://www.jama.com). ing results are most likely explained by ing hits in stage 2 (selecting top SNPs In each study, dementia was defined the modest sample size and, hence, lim- from all loci that reached PϽ10−5) with using the Diagnostic and Statistical ited statistical power of early studies to data from the nonoverlapping studies Manual of Mental Disorders, Third Edi- detect genes with small effects. Re- within the GERAD1 consortium (ex- tion Revised or Fourth Edition (DSM-IV) cently, 2 large GWAS, the United King- dom–led Genetic and Environmental Figure 1. The 3-Stage Approach and the Various Studies Included in the Different Stages Risk in Alzheimer Disease 1 consortium (GERAD1)13 and the European Alzheimer Disease Initiative stage 1 (EADI1),14 reported 3 new genome- Stage 1: initial discovery wide significant loci for AD: within the Studies included in this meta-analysis: CLU gene (GenBank AY341244) encod- CHARGE Consortium15 2.5 million autosomal SNPs from Age, Gene/Environment Susceptibility– ing clusterin (also called apolipopro- HapMap CEU population Reykjavik Study tein J), near the PICALM gene (GenBank Cardiovascular Health Study BC073961) encoding phosphatidylino- Framingham Heart Study Rotterdam Study sitol–binding clathrin assembly pro- Translational Genomic Research Institute4 tein, and within the CR1 (RefSeq Mayo AD GWAS5 NG_007481) gene encoding comple- ment component (3b/4b) receptor 1.13,14 We performed a 3-stage analysis of Stage 2: pooling with additional data GWAS data to identify additional loci 2708 SNPs identified Studies included in this meta-analysis: associated with late-onset AD. More- with P<10-3 Studies from stage 1 over, we sought to replicate genome- European AD Initiative 114 wide significant loci, from both the current analysis and previous reports, in an independent case-control popula- Stage 3: pooling with additional data tion. Finally, we used 2 large, prospec- Studies included in this meta-analysis: tive, population-based studies to as- Studies from stage 2 sess the improvement in incident AD 38 SNPs (10 loci) identified with P<10-5 Genetic and Environmental Risk in risk prediction conferred by the re- AD 1 Consortium (excluding the cently described loci. Mayo AD GWAS; already included in stage 1)13 METHODS

We used a 3-stage sequential analysis Replication stage to identify novel loci associated with 4 SNPs (apart from the APOE locus) reaching P<1.7x10-8 Spanish case-control study (Fundació ACE19) late-onset AD (FIGURE 1). Our initial investigation (stage 1) was a meta- analysis combining new genome-wide association data from white participants in the large, population-based Co- AD indicates Alzheimer disease; APOE, apolipoprotein E; CHARGE, Cohorts for Heart and Aging Research in horts for Heart and Aging Research in Genomic Epidemiology; Fundacio´ ACE, Fundacio´ Alzheimer Centre Educacional; GWAS, genome-wide association study; SNP, single-nucleotide polymorphism. Genomic Epidemiology (CHARGE)

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

criteria.16 Among persons with demen- ies that have used different genotyping study fit an additive genetic model (a tia, all studies used the National Insti- platforms because the platforms differ 1-degree-of-freedom trend test) relat- tute of Neurological and Communica- in the SNPs genotyped. Imputation ing genotype dosage (0, 1, or 2 copies of tive Disorders and Stroke and the methods and quality control filters in the minor allele) to study trait. In the Alzheimer Disease and Related Disor- each sample are described in section 3 CHARGE cohorts, prevalent cases were ders Association (NINCDS-ADRDA) cri- of the eAppendix. compared with controls free of demen- teria to define AD and included persons All analyses were restricted to white tia at the DNA draw date. Participants with definite (diagnosis of AD patho- persons, racial identity being self- were excluded if they declined consent logically confirmed at autopsy), prob- defined by the participants (see section or genotyping failed. For analysis of able, or possible AD.17 2 of the eAppendix). We included white prevalent events in the CHARGE co- The individual studies in stage 1 Hispanics and adjusted for population horts and for the case-control data from were genotyped on different plat- structure. Since only 1 of the CHARGE the TGEN and Mayo cohorts, we used forms, as shown in Table 1. The EADI1 studies, the Cardiovascular Health Study logistic regression models. For the analy- used the Illumina Quad 6.0 (Illumina (CHS), had a small number of African sis of incident events in the CHARGE co- Inc, San Diego, California), and American participants (n=574 with horts, participants who were free of de- GERAD1 was genotyped on various genotyping), this racial subgroup was too mentia entered the analysis at the time Illumina chips. In each of the CHARGE small for independent analysis. Link- of the DNA sample collection and were cohorts and in the TGEN database, we age disequilibrium patterns are very dif- followed up until development of inci- used the genotype data to impute to the ferent in individuals of African heri- dent AD; participants were censored at 2.5 million nonmonomorphic autoso- tage, which leads to greater uncertainty death, at the time of their last follow-up mal SNPs described in HapMap (Utah in imputation, as well as the possibility examination or health status update residents with Northern and Western of false-positive associations if data from when they were known to be free of clini- European ancestry from the Centre 2 racial groups are combined when dis- cal dementia, and when they developed d’Etude du Polymorphisme Humain ease risk differs by race (referred to as dementia from a cause other than AD. collection [CEU population]). Impu- population stratification); hence, Afri- We used Cox proportional hazards mod- tations are needed to meta-analyze ge- can American participants in the CHS els to calculate hazard ratios with cor- nome-wide association data across stud- were excluded from these analyses. Each responding 95% confidence intervals

Table 1. Characteristics of Studies in Stage 1 of the Analysis CHS FHS Rotterdam Study AGES TGEN Mayo AD GWAS

Characteristics Cases Controls Cases Controls Cases Controls Cases Controls Cases Controls Cases Controls Study design Cohort Cohort Cohort Cohort Case-control Case-control Genotype platform Illumina Affymetrix Illumina Infinium Illumina Affymetrix Illumina HumanCNV370 GeneChip HumanHap550 HumanCNV370 GeneChip HumanHap300 Duo Human Chip Version Duo Human Version 2 Duo Mapping 500K 3.0 Mapping 500K BeadChips Array Set ϩ50K Array Set Gene Focused Panel Prevalence studiesa No. of participantsb 93 2429 52 2091 171 5700 78 2684 829 536 810 1202 Women, No. (%) 49 (53) 1506 (62) 42 (81) 1192 (57)128 (75) 3347 (59) 39 (50) 1557 (58) 431 (52) 338 (63) 462 (57) 601 (50) Age, mean (SD), y 80 (6) 75 (5) 87 (6) 76 (7) 84 (9) 69 (9) 81 (5) 76 (5) 81 (10) 80 (7) 73 (4) 74 (5) APOE ε4 positive, 35 (38) 583 (24) 20 (38) 418 (20) 62 (36) 1549 (28) 38 (49) 725 (27) 481 (58) 107 (20) 535 (66) 337 (28) No. (%)c Incidence studiesa Cohort at risk, No.b 2429 806 5700 Women, No. (%) 1506 (62) 484 (60) 3347 (59) Age at start/age at incident 75 (5)/ 82 (6)/ 69 (9)/ dementia, mean (SD), y 82 (5) 88 (5) 82 (7) Incident AD cases, No. 435 76 462 Follow-up, mean (SD), y 6.8 (3.6) 4.8 (3.0) 9.3 (3.2) APOE ε4 positive, No. (%)c 632 (26) 153 (19) 1549 (28) Abbreviations: AD, Alzheimer disease; AGES, Age, Gene/Environment Susceptibility–Reykjavik Study; APOE, apolipoprotein E; CHS, Cardiovascular Health Study; FHS, Framing- ham Heart Study; GWAS, genome-wide association study; TGEN, Translational Genomics Research Institute. a In the prevalence studies, cases were persons who had AD at time of DNA draw. Controls were those who were free of any dementia. In the incidence studies, cases were persons from the cohort at risk who developed dementia during follow-up. Persons who developed another type of dementia were censored at date of onset. b Includes only persons with genotyping who also provided consent for these analyses and had high-quality genotyping (met quality control criteria) (eAppendix). In the FHS, only original cohort participants were included in incident analyses. c Among those with APOE genotyping available.

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

(CIs) after ensuring that assumptions of study’s use of PϽ5ϫ10−8.18 The 3 stages genome-wide significance in previous proportionality of hazards were met. In of meta-analyses were completed in May publications and that we replicated the CHS, Framingham Heart Study to August 2009. nominally within CHARGE: PICALM (FHS), and Rotterdam Study, controls Significant hits from stage 3 of the and CLU (PϽ.05). Moreover, the analy- contributed one set of person-years to the discovery phase were replicated in an sis was restricted to incident AD to prevalent analysis and a second, nonover- independent Spanish case-control avoid survival bias and was restricted lapping set of person-years to the inci- sample (the Fundacio´ Alzheimer Cen- to population-based samples because dent analyses. Under the Martingale tre Educacional [ACE]) of 1140 pa- case-control studies may overestimate property of Cox models, the 2 analyses tients with AD (mean age, 78.8 [SD, the effects of the genes if cases and con- are independent, and their indepen- 7.9] years; 69.9% women) compared trols were not randomly selected from dence was confirmed in simulation stud- with 1209 general population con- the populations in which AD risk pre- ies. Primary analyses were adjusted for trols (mean age, 49.9 [SD, 9.2] years; diction is to be applied.22 The improve- age and sex and any evidence of popu- 52.8% women).19,20 All AD patients ful- ment in risk prediction was investi- lation stratification. Details of the screen- filled DSM-IV criteria for dementia and gated by comparing 3 sequentially ing for latent population substructure in NINCDS-ADRDA criteria for possible incremental AD risk prediction mod- each discovery sample are available in and probable AD.16,17 Both cases and els that first incorporated age and sex section 4 of the eAppendix. In addition, controls were white. Details of the alone, then added data on risk allele sta- the CHS also adjusted for study site and sample are provided in section 6 of the tus at the APOE locus and, finally, data the FHS accounted for familial relation- eAppendix. Genotyping was under- on risk allele status at the CLU and ships (by using a Cox model with ro- taken using real-time polymerase chain PICALM loci. We did not assess the util- bust variance estimator clustering on reaction coupled with fluorescence ity of novel loci uncovered in this ar- pedigree to account for family relation- resonance energy transfer. Effect sizes ticle (using CHARGE as part of the dis- ships) and for whether the DNA had for single markers were calculated by covery sample) to avoid the risk of been whole-genome amplified. unconditional logistic regression analy- overestimating effects by using the same Our stage 1 meta-analysis combined sis using SPSS software, version 13.0 sample for gene discovery and risk pre- results from 9 discrete sources: inci- (SPSS Inc, Chicago, Illinois). Replica- diction.22 Prediction models were con- dent AD in the CHS, FHS, and Rotter- tion was completed in October 2009. structed using Cox proportional haz- dam Study; prevalent AD in the Age, In secondary analyses, we also exam- ards methods using the R package Gene/Environment Susceptibility– ined results for previously reported survcomp. APOE ε4 status was in- Reykjavik Study, CHS, FHS, and Rotter- loci.5,13,14 For these loci, which included as a discrete variable (0, 1, or 2 dam Study; and the TGEN and Mayo AD cluded the recently reported loci by the alleles) and the other 2 genetic loci as GWAS case-control studies. We used in- EADI1 and GERAD1 consortia, we re- dosages; all gene effects were exam- verse-variance weighting (also known as stricted our analysis to the previously un- ined using additive models. The accu- a fixed-effects analysis) for meta- published CHARGE data. We did not as- racy of risk prediction for each model analysis, applying genomic control to sess the association with PCDH11X was assessed as discriminative accu- each study of stage 1. This approach as- because we focused only on autosomal racy, measured by the area under the signs greater weight to more precise SNPs in these analyses. We did exam- receiver operating characteristic curve (study-specific) estimators; thus, greater ine associations with the top 15 candi- (AUC). The AUC theoretically ranges weight is given to studies in which a date genes listed in the Alzgene data- from 0.50 (as predictive as tossing a given SNP was genotyped or more ef- base (http://www.alzgene.org)21 as of coin) to 1.00 (perfect prediction). fectively imputed and to studies with August 12, 2009, including the APOE/ larger sample sizes. Details of the meta- TOMM40/APOC1 locus and 12 genes RESULTS analyses are available in section 5 of the outside that locus. Details of SNPs se- The stage 1 meta-analysis had 8935 de- eAppendix. We retained only SNP- lected and results for these SNPs are pro- mentia-free individuals (mean age, 72 phenotype associations that were based vided in section 7 of the eAppendix and [SD, 7] years), of whom 973 devel- on results from at least 2 of the 9 dis- in eTable 1. oped incident AD over an average fol- covery samples and for which the mi- We sought to estimate the effect of low-up time of 8 [SD, 3] years, and 2033 nor allele frequency was at least 2%. For recently identified loci on 10-year risk prevalent cases of AD who were com- stages 2 and 3, we again used inverse- prediction in the general population pared with 14 642 dementia-free con- variance meta-analysis but without ge- using the data for prospectively ascer- trols. Of these, 1367 AD cases (973 in- nomic control adjustment. We decided tained incident AD in the 2 largest com- cident) were from the CHARGE a priori on a genome-wide significance munity-based cohort studies available cohorts. In this discovery analysis based threshold of PϽ1.7ϫ10−8, which gives, to us (Rotterdam Study and CHS). In on the CHARGE cohorts, TGEN, and for a 3-stage sequential analysis, the same these analyses, we only included SNPs Mayo GWAS, there was no evidence of control of false-positives as a single from the 2 loci that were shown to have spurious inflation of P values or sig-

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

nificant population stratification (see sample- and stage-specific estimates of nificance (eTable 2 and eFigure 3). In eFigure 2 for the quantile-quantile plot relative risks, 95% CIs, and P values for stage 2, 2 additional loci, rs11136000 comparing the observed and expected each of the 38 SNPs selected in stage 2 in CLU and a locus (rs11771145) at P value distributions). eFigure 3 illus- analyses. FIGURE 2 and FIGURE 3 show chromosome 7 in the 5Ј upstream trates the primary findings from the regional association plots for the 2 SNPs promoter/regulatory region of ephrin stage 1 meta-analysis in a Manhattan not previously reported to have reached receptor A1 (EPHA1; GenBank plot showing genome-wide P values for genome-wide significance, rs744373 AH007960) reached genome-wide sig- all interrogated SNPs across the 22 au- and rs597668 on chromosomes 2 and nificance. However, the latter became tosomal chromosomes. After stage 1, 19, respectively. In these figures, we nonsignificant after adding GERAD1 2708 SNPs had a PϽ10−3 and were stud- show the linkage disequilibrium (with data in stage 3, though the effect seen ied in stage 2. In stage 2, pooling these the index SNP) and stage 1, 2, and 3 in GERAD1 was in the same direction results with data from EADI1 (2032 association results for the index SNP in that the same allele was associated cases and 5328 controls), 38 SNPs in and stage 1 results for all SNPs within with an increased (but nonsignifi- 10 loci had a PϽ10−5. Finally, in stage 200 kilobase (kb) on either side of the cant) risk of AD. In stage 3, genome- 3, the most significant SNPs from these index SNP at that locus, as well as gene wide significant evidence for associa- 10 loci were meta-analyzed with the locations and recombination rates in the tion with AD was reached at the APOE nonoverlapping studies from GERAD1 region. Regional association plots (rs2075650; P =1.04ϫ10−295), CLU (3333 cases and 6995 controls). The for the other loci listed in Table 2 are (rs11136000; P =1.62ϫ10−16), and findings of the stages 1, 2, and 3 analy- presented in eFigure 4, eFigure 5, PICALM (rs3851179; P=3.16ϫ10−12) ses at these 10 loci are presented in eFigure 6, eFigure 7, and eFigure 8. loci, as well as for 2 novel loci on chro- TABLE 2. Additional details are pro- In stage 1, 11 SNPs in the APOE/ mosomes 2 (rs744373; P=1.59ϫ10−11) vided in eTable 2, which shows chro- TOMM40/APOC1 region reached our and 19 (rs597668; P =6.45ϫ10−9). mosomal location, adjacent genes, preset threshold for genome-wide sig- Table 2 shows the odds ratios (ORs)

Table 2. Genetic Loci at Which SNPs Were Associated With Alzheimer Disease at PϽ10−5 in Stage 2 Meta-analysis and That Underwent Further Meta-analysis in Stage 3 Stage 1 Meta-analysis Stage 2 Meta-analysis Stage 3 Meta-analysis

Chromosome: Additional Nearest Minor Meta-OR Meta-P Meta-OR Meta-P Meta-OR Meta-P Top SNPa Position SNPsb Genec Alleled MAF (95% CI)e Value (95% CI)e Value (95% CI)e Value rs2075650 19:50087459 18 APOE (RefSeq G 13.7 2.23 (2.04-2.44) 3.18ϫ10−68 2.61 (2.45-2.80) 4.67ϫ10−172 2.53 (2.41-2.66) 1.04ϫ10−295 NG_007084) rs11136000 8:27520436 CLU (GenBank T 39.2 0.89 (0.83-0.94) 4.98ϫ10−4 0.85 (0.81-0.90) 1.49ϫ10−9 0.85 (0.82-0.88) 1.62ϫ10−16 AY341244) rs3851179 11:85546288 PICALM T 37.1 0.86 (0.81-0.92) 1.22ϫ10−5 0.89 (0.84-0.93) 2.81ϫ10−6 0.87 (0.84-0.91) 3.16ϫ10−12 (GenBank BC073961) rs744373 2:127611085 BIN1 (RefSeq G 29.1 1.13 (1.06-1.21) 4.93ϫ10−4 1.14 (1.08-1.20) 1.02ϫ10−6 1.15 (1.11-1.20) 1.59ϫ10−11 NG_012042) rs597668 19:50400728 1 EXOC3L2 C 15.4 1.18 (1.07-1.29) 5.91ϫ10−4 1.18 (1.10-1.26) 2.16ϫ10−6 1.17 (1.11-1.23) 6.45ϫ10−9 (RefSeq NM_138568) rs11771145 7:142820884 EPHA1 A 34.7 0.87 (0.81-0.94) 2.14ϫ10−4 0.86 (0.81-0.90) 1.32ϫ10−8 0.91 (0.87-0.94) 1.70ϫ10−6 (GenBank AH007960) rs2043948 14:74142801 LTBP2 (RefSeq T 7.7 1.25 (1.10-1.42) 6.96ϫ10−4 1.27 (1.16-1.39) 4.44ϫ10−7 1.13 (1.06-1.22) 4.46ϫ10−4 NM_000428) rs2825544 21:19662423 PRSS7 C 34.6 1.13 (1.06-1.21) 2.55ϫ10−4 1.14 (1.08-1.20) 4.85ϫ10−7 1.09 (1.05-1.13) 2.10ϫ10−5 (RefSeq NG_012207) rs7527934f 1:14231011 9 PRDM2 G 25.7 0.86 (0.79-0.93) 3.50ϫ10−4 0.87 (0.82-0.92) 5.87ϫ10−6 0.97 (0.91-1.03) (RefSeq NM_012231) rs4296166f 14:32022118 AKAP6 A 47.8 1.14 (1.07-1.21) 8.36ϫ10−5 1.12 (1.07-1.18) 4.08ϫ10−6 0.98 (0.89-1.08) (RefSeq NM_004274) Abbreviations: CI, confidence interval; MAF, minor allele frequency; OR, odds ratio; SNP, single-nucleotide polymorphism. aAt each locus, the SNP with the lowest (most statistically significant) P value was selected for stage 3 meta-analysis. bNumber of additional SNPs at the locus with PϽ10−5. cColumn shows the Human Gene Organization Gene Nomenclature System symbols for the gene located closest to each SNP. Standardized gene annotations for all SNP results were derived programmatically from the University of California, Santa Cruz Genome Browser RefSeq gene track (hg18). dAlleles were coded on the forward strand of the genome. eThe minor allele was taken as the coded allele. The ORs represent the relative increase of disease risk per increase of 1 copy of the minor allele. fThese SNPs did not pass quality control criteria in the Genetic and Environmental Risk in Alzheimer Disease 1 consortium. Stage 3 results are presented from this database for proxy SNPs: rs2206586 for rs7527934 and rs10146289 for rs4296166.

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

Figure 2. Regional Association Plot for Novel Locus rs744373 Near BIN1 on Chromosome 2 Significantly Associated (PϽ5ϫ10−8) With Alzheimer Disease in Stage 3 Analyses

rs744373

60 Stage 3 (P = 1.59x10-11) 12 55 -6 11 Stage 2 (P = 2.16x10 ) 50 Recombination Rate (cM/Mb) 10 Stage 1 (P = 5.91x10-4) 45 9 Linkage disequilibrium 40 8 0.5

10 5 25

Log 4 − 20 3 15 2 10 1 5 0 0 ERCC3 MYO7B GYPC MAP3K2 LIMS2 BIN1 PROC GPR17 CYP27C1 IWS1

127091 127611 128130 Position on Chromosome 2, Kilobase

Each data marker represents the statistical significance (P value) of each single-nucleotide polymorphism (SNP) plotted on the −log10 scale against its chromosomal position (NCBI Build 36).The blue diamonds show stage 1 P values for the sentinel (top) SNP at each locus, whereas the gray and black diamonds show the P values for the same SNP following stage 2 and stage 3 meta-analyses, respectively. The fine-scale recombination rate is shown by the blue line, which shows the average frequency with which recombination occurs at that site. Genes located in the region shown are shown as green arrows with HUGO gene nomenclature committee gene symbols. The length of the green arrow represents the size/extent of the gene and the arrowhead the direction in which transcription of mRNA occurs.

Figure 3. Regional Association Plot for Novel Locus rs597668 Near BLOC1S3 and MARK4 on Chromosome 19 Significantly Associated (PϽ5ϫ10−8) With AD in Stage 3 Analyses

rs597668

10 60 Stage 3 (P = 6.45x10-9) 9 55 -6 Stage 2 (P = 1.02x10 ) 50 8 Recombination Rate (cM/Mb) Stage 1 (P = 4.93x10-4) 45 7 Linkage disequilibrium 40 Value 6 r 2>0.8 P 0.2

2 15 10 1 5 0 0

BCAM APOC1 RELB NKPD1 MARK4 PPP1R13L RTN2 GPR4 QPCTL

CEACAM16 PVRL2 APOC4 SFRS16 TRAPPC6A CKM CD3EAP FLJ40125 EML2 FBXO46

BCL3 TOMM40 APOC2 ZNF342 BLOC1S3 KLC3 ERCC1 VASP GIPR

CBLC APOE CLPTM1 GEMIN7 EXOC3L2 ERCC2 FOSB OPA3 SNRPD2

49882 50401 50920 Position on Chromosome 19, Kilobase

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

Among the 54 SNPs selected from the Table 3. Replication of Genome-wide Significant Results From an Independent Discovery Sample19 top 12 candidate genes (outside the Gene SNP MAF, Cases/Controls OR (95% CI) P Value APOE/TOMM40/APOC1 locus) listed CLU rs11136000 0.36/0.39 0.82 (0.77-0.99) .03 in the Alzgene Web site, we found evi- PICALM rs3851179 0.30/0.34 0.84 (0.74-0.95) .007 dence for a nominal association of BIN1 rs744373 0.30/0.27 1.17 (1.03-1.33) .02 rs4362 in the angiotensin-converting EXOC3L2 rs597668 0.13/0.11 1.26 (1.05-1.51) .01 enzyme (ACE; RefSeq NC_000017.10) Abbreviations: CI, confidence interval; MAF, minor allele frequency; OR, odds ratio; SNP, single-nucleotide polymorphism. gene and rs1784933 in the sortilin- related receptor L (DLR class A) re- associated with the minor allele for 1.16; P=.05 for analysis 2). We also ex- peats-containing (SORL1; RefSeq each of these SNPs. Locus rs744373 is amined the effect of adjusting for age, sex, NC_000011.9) gene with AD (relative within 30 kb of the gene bridging in- and presence of at least 1 APOE ε4 al- risks associated with each copy of the tegrator 1 (BIN1; RefSeq NG_012042) lele (using a dominant genetic inherit- minor allele were 0.92; 95% CI, 0.85- (Figure 2), while rs597668 is within ance model) in the Spanish replication 0.99; P=.03 for ACE and 1.33; 95% CI, 60 kb of 6 genes including exocyst com- sample. Again, the results were attenu- 1.03-1.72; P=.03 for SORL1) (eTable 1). plex component 3–like 2 (EXOC3L2; ated (OR, 1.24; 95% CI, 1.02-1.51; RefSeq NM_138568), biogenesis of P=.03). These findings are consistent Genetic Risk Prediction lysosomal organelles complex 1, sub- with the moderate to low level of link- We assessed the extent to which APOE unit 3 (BLOC1S3; RefSeq NG_008372), age disequilibrium observed between ε4, PICALM, and CLU can improve pre- and microtubule-associated protein/ rs597668 and SNPs within the APOE and dictive models for risk of incident AD microtubule affinity-regulating ki- TOMM40 region (r2Ͻ0.01 according to in the general population (repre- nase 4 (MARK4; GenBank BC071948) HapMap CEU data) (Figure 3). sented by the cohorts of the Rotter- (Figure 3). dam Study and CHS). The addition of Replication of Previously Reported APOE ε4 carrier status to a prediction Independent Replication Associations in the CHARGE Sample model including age and sex only in- We replicated the 4 associations that In our secondary analyses examining rep- creased the AUC from 0.826 (95% CI, reached our preset genome-wide signifi- lication of published findings in the 0.806-0.846) to 0.847 (95% CI, 0.828- cance threshold (1.7ϫ10−8) in an inde- previously unreported CHARGE data, 0.865) in the Rotterdam Study and from pendent sample of cases and controls 6intronicor3Ј untranslated region 0.670 (95% CI, 0.625-0.723) to 0.702 (TABLE 3). Effect sizes in the replica- SNPs in the APOE/TOMM40/APOC1 re- (95% CI, 0.654-0.754) in the CHS. Fur- tion cohort were similar to those ob- gion (rs6857, rs2075650, rs4420638, ther inclusion of risk allele status for served in the discovery sample; each of rs157582, rs6859, and rs10119) reached CLU and PICALM improved the AUC these associations reached PϽ.05. a genome-wide significance threshold of only minimally to 0.849 (95% CI, PϽ1.7ϫ10−8, and we replicated the most 0.831-0.867) in the Rotterdam Study Conditional Analyses statistically significant SNPs within 2 of and to 0.705 (95% CI, 0.654-0.751) in at Chromosome 19 Locus the 3 recently reported genetic loci as- the CHS. The corresponding receiver Because rs597668 is on chromosome 19, sociated with AD in prior GWAS: CLU operating characteristic curves are fairly close to the APOE locus, we un- (rs11136000; OR, 0.90; 95% CI, 0.82- shown in eFigure 10. dertook conditional analyses to exam- 0.98; P=.02) and PICALM (rs3851179; ine whether its association with AD was OR, 0.90; 95% CI, 0.83-0.99; P=.02) COMMENT independent of APOE ε4. We con- (eTable 2 and eAppendix). We did not We report results of an international ducted 2 analyses with AD (among per- find a significant association with the 3-stage genome-wide analysis to study sons with directly genotyped APOE ε4 most statistically significant CR1 SNP genetic variation underlying late- status) in the CHARGE, TGEN, and (rs3818361) in the CHARGE data. How- onset, sporadic AD. We studied more Mayo samples, adjusting (1) for our ever, 13 SNPs within the gene showed than 35 000 persons (8371 AD cases), strongest association in the APOE/ nominal significance (PϾ.001 but constituting the largest sample ana- TOMM40/APOC1 locus (rs2075650) PϽ.05) (eTable 3). Furthermore, add- lyzed to date. In the gene discovery and (2) for the actual APOE ε4 SNP, ing CHARGE and TGEN data on phase, we showed genome-wide sig- rs429358. In each case, we found that the rs3818361 to the previously reported nificance for 2 loci related to AD, one association was attenuated but a mar- EADI1 and GERAD1 data (Mayo AD on chromosome 2 and a second locus ginal signal remained when adjusting for GWAS data were included in the on chromosome 19, that had not pre- APOE ε4 (OR, 1.18; 95% CI, 1.08-1.24; GERAD1 data for this analysis) showed viously been found to achieve genome- P=3.9ϫ10−4 without adjustment; OR, that results now reached genome-wide wide significance and that appear to be 1.17; 95% CI,1.07-1.23; P=8.7ϫ10−4 for significance (OR, 1.15; 95% CI, 1.11- independent of APOE. BIN1 was pre- analysis 1; and OR, 1.10; 95% CI, 1.00- 1.20; P=1.04ϫ10−11) (eFigure 9). viously identified as having a possible

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

association with AD in the recent vated by hepatitis B virus X antigen based cases and controls and the sub- GWAS from the GERAD1,13 but until (XTP7) gene. sample of prospectively ascertained in- our analysis, this association was not When evaluating the added value of cident AD that permitted the exploration significant at the genome-wide level. the new AD genes in clinical risk pre- of incident risk prediction algorithms. Furthermore, we replicated both these diction, we focused on the 2 recently Alzheimer disease was diagnosed using loci as well as the recently identified reported AD genes13,14 that were repli- standard NINCDS-ADRDA criteria. The loci, CLU and PICALM, in an indepen- cated in our population-based stud- observed associations are unlikely to be dent sample. Although genetic varia- ies, CLU and PICALM, and found that due to population stratification since the tion at the CLU and PICALM loci did they only slightly improved predic- discovery and replication samples were modify the risk of AD in our population of incident AD beyond age, sex, and restricted to white individuals of Euro- tion-based sample, these polymor- APOE ε4 based models. The increase in pean origin and were also investigated phisms added very little to prediction AUC was 0.002 in the Rotterdam Study for latent population substructure. of AD risk. and 0.003 in the CHS, which would not The study also has limitations. De- The locus on chromosome 2q14.3 is be of value in the clinical setting. There spite our large sample size, we had lim- adjacent to the BIN1 gene, which is 1 are 2 reasons for this. First, the asso- ited power to detect associations with of 2 amphiphysins and is expressed ciations of CLU and PICALM with AD small effect sizes and associations with most abundantly in the brain and risk were markedly lower than those of rare variants. Although all studies used muscle.23 Amphiphysins promote age and APOE; therefore, a major im- accepted clinical or pathological crite- caspase-independent apoptosis and also provement was not expected. This fits ria to define dementia and AD, pheno- play a critical role in neuronal mem- with recent insights on polygenic mod- typic heterogeneity between samples brane organization and clathrin- els that assume there are tens of thou- may have limited our ability to detect mediated, synaptic vessel formation,24 sands of risk alleles, each with a small some associations. Moreover, the con- a process disrupted by ␤-amyloid.25 (approximately 5% increase in rela- trols in the Spanish replication sample Knockout mice with decreased expres- tive risk) effect throughout the whole were younger than the cases, and their sion of the amphiphysins have sei- genome, rather than a discrete num- cognitive status had not been formally zures and major learning deficits.26 Al- ber of alleles with moderate effects. examined. However, whereas this could tered expression of BIN1 has been Such models appear to underlie the sus- reduce our power to observe an asso- demonstrated in aging mice, in trans- ceptibility to schizophrenia risk, and a ciation, it would not invalidate the as- genic mouse models of AD, and in per- similar model may be applicable to sociations we did observe. Further- sons with schizophrenia.27,28 AD.34 Second, the extent to which risk more, the frequency distribution of The 19q13.3 locus (rs597668), a site factors improve risk prediction de- minor and major alleles among the distal to and not in linkage disequilib- pends on the predictive performance of Spanish controls was similar to that rium with SNPs in the APOE locus, had the initial risk model. Added risk fac- noted in the discovery sample and in been suspected in an early linkage study tors need to have stronger effects to im- the HapMap CEU sample. to harbor a gene for AD.29 There are 6 prove a risk model with high AUC than In conclusion, this meta-analysis of genes adjacent to this locus, 2 of which to improve a model with lower AUC. GWAS data from several of the largest are part of pathways linked to AD pa- Alzheimer disease risk prediction based AD GWAS studies to date confirms pre- thology. The protein product of on age, sex, and APOE already has very viously known and recently described BLOC1S3, biogenesis of lysosomal or- high discriminative accuracy: the AUC associations (CLU and PICALM) and ganelles complex 1, subunit 3, is ex- was 0.826 in the Rotterdam Study and shows genome-wide significance and pressed in the brain, regulates endo- 0.670 in the CHS, which implies that replication for 2 biologically plau- somal to lysosomal routing,30 and has further improvements require many sible, novel loci on chromosomes 2 and been implicated in schizophrenia.31 The new variants or variants with strong ef- 19. However, the predictive ability of second gene, MARK4, is inducible, ex- fects. Whether such improvements are CLU and PICALM to identify individu- pressed only in the brain, and plays a to be expected will depend in large part als at risk of AD is not clinically sig- role in neuronal differentiation.32 on the ability to unravel the underly- nificant. The value of these associa- MARK4 is a kinase that phosphory- ing genetic architecture and to iden- tions may lie in the insights they could lates tau, is polyubiquitinated in vivo, tify and quantify environmental risk fac- provide for research into the patho- and is a substrate of the aging-related tors, including complex interactions.35 physiological mechanisms of AD. deubiquitinating enzyme USP9X; A next step for genetic research in AD Author Contributions: Drs Seshadri, Ikram, DeStefano, hence, it may play a role in the abnor- will be to further increase the sample and Breteler had full access to all of the data in the study 33 and take responsibility for the integrity of the data and mal tau phosphorylation seen in AD. sizes of GWAS and evaluate further ge- the accuracy of the data analysis. Drs Seshadri, Little is known of the function of the netic models. Fitzpatrick, Ikram, DeStefano, Gudnason, and Boada contributed equally as first authors. Drs Ruiz, Williams, gene closest to rs597668, EXOC3L2, Strengths of this study include the Amouyel, Younkin, Wolf, Launer, Lopez, van Duijn, and also referred to as protein 7 transacti- large sample of clinic- and community- Breteler contributed equally as last authors.

Downloaded From: https://jamanetwork.com/ on 09/24/2021 GENETIC LOCI ASSOCIATED WITH ALZHEIMER DISEASE

Study concept and design: Seshadri, Bis, Rotter, REFERENCES association studies of nearly all common variants. Genet Hofman, Schmidt, Mosley, Lumley, Amouyel, Wolf, Epidemiol. 2008;32(4):381-385. Lopez, Duijn, Breteler. 1. Seshadri S, Wolf PA. Lifetime risk of stroke 19. Antu´ nez C, Boada M, Lopez-Arrieta J, et al. Acquisition of data: Seshadri, Ikram, Gudnason, Boada, and dementia. Lancet Neurol. 2007;6(12):1106- GOLPH2 gene markers are not associated with Carrasquillo, Lambert, Schrijvers, Ramirez-Lorca, 1114. Alzheimer’s disease in a sample of the Spanish Janssens, Pankratz, Dartigues, Hernandez, Beiser, 2. Gatz M, Reynolds CA, Fratiglioni L, et al. Role of population. J Alzheimers Dis. 2009;18(4):751- Kuller, Koudstaal, Dickson, Tzourio, Antunez, Rotter, genes and environments for explaining Alzheimer 754. Berr, Owen, Lopez-Arrieta, Becker, Rivadeneira, Nalls, disease. Arch Gen Psychiatry. 2006;63(2):168- 20. Ramirez-Lorca R, Boada M, Saez ME, et al. GAB2 Graff-Radford, Campion, Auerbach, Jonsson, Lathrop, 174. gene does not modify the risk of Alzheimer’s disease Au, Psaty, Uitterlinden, Farrer, Ruiz, Williams, Amouyel, 3. Ertekin-Taner N. Genetics of Alzheimer’s disease. in Spanish APOE 4 carriers. J Nutr Health Aging. 2009; Younkin, Lopez, Breteler. Neurol Clin. 2007;25(3):611-667. 13(3):214-219. 4. Reiman EM, Webster JA, Myers AJ, et al. GAB2 21. Bertram L, McQueen MB, Mullin K, et al. Sys- Analysis and interpretation of data: Seshadri, ε Fitzpatrick, Ikram, DeStefano, Bis, Smith, Carrasquillo, alleles modify Alzheimer’s risk in APOE 4 carriers. tematic meta-analyses of Alzheimer disease genetic Harold, Debette, Longstreth, Janssens, Hollingworth, Neuron. 2007;54(5):713-720. association studies. Nat Genet. 2007;39(1):17- Aspelund, Beiser, Abraham, Du, Rotter, Aulchenko, 5. Carrasquillo MM, Zou F, Pankratz VS, et al. Ge- 23. Harris, Petersen, Owen, Vardarajan, Rivadeneira, Nalls, netic variation in PCDH11X is associated with suscep- 22. Janssens AC, van Duijn CM. Genome-based pre- Rice, Lathrop, Mosley, Farrer, Lumley, Ruiz, Williams, tibility to late-onset Alzheimer’s disease. Nat Genet. diction of common diseases. Genome Med. 2009; Amouyel, Younkin, Launer, Lopez, Duijn, Breteler. 2009;41(2):192-198. 1(2):20. Drafting of the manuscript: Seshadri, Fitzpatrick, Ikram, 6. Poduslo SE, Huang R, Huang J, Smith S. Genome 23. Wechsler-Reya R, Sakamuro D, Zhang J, et al. DeStefano, Bis, Ramirez-Lorca, Du, Vardarajan, Lopez, screen of late-onset Alzheimer’s extended pedigrees Structural analysis of the human BIN1 gene. J Biol Duijn, Breteler. identifies TRPC4AP by haplotype analysis. Am J Med Chem. 1997;272(50):31453-31458. Critical revision of the manuscript for important in- Genet B Neuropsychiatr Genet. 2009;150B(1): 24. Wigge P, Kohler K, Vallis Y, et al. Amphiphysin Mol Biol Cell tellectual content: Seshadri, Ikram, Gudnason, Boada, 50-55. heterodimers. . 1997;8(10):2003- 7. Bis, Smith, Carrasquillo, Lambert, Harold, Schrijvers, Beecham GW, Martin ER, Li YJ, et al. Genome- 2015. 25. Ramirez-Lorca, Debette, Longstreth, Janssens, wide association study implicates a chromosome 12 Kelly BL, Ferreira A. Beta-amyloid disrupted syn- Am J Hum Pankratz, Dartigues, Hollingworth, Aspelund, risk locus for late-onset Alzheimer disease. aptic vesicle endocytosis in cultured hippocampal Genet Neuroscience Hernandez, Beiser, Kuller, Koudstaal, Dickson, Tzourio, . 2009;84(1):35-43. neurons. . 2007;147(1):60-70. 8. 26. Abraham, Antunez, Rotter, Aulchenko, Harris, Bertram L, Lange C, Mullin K, et al. Genome-wide Di Paolo G, Sankaranarayanan S, Wenk MR, et al. Petersen, Berr, Owen, Lopez-Arrieta, Becker, association analysis reveals putative Alzheimer’s dis- Decreased synaptic vesicle recycling efficiency and cog- APOE Am J Hum Neuron Rivadeneira, Nalls, Graff-Radford, Campion, Auerbach, ease susceptibility loci in addition to . nitive deficits in amphiphysin 1 knockout mice. . Genet Rice, Hofman, Jonsson, Schmidt, Lathrop, Mosley, Au, . 2008;83(5):623-632. 2002;33(5):789-804. 9. 27. Psaty, Uitterlinden, Farrer, Lumley, Ruiz, Williams, Myers A, Holmans P, Marshall H, et al. Suscepti- English JA, Dicker P, Focking M, et al. 2-D DIGE Amouyel, Younkin, Wolf, Launer, Lopez, Breteler. bility locus for Alzheimer’s disease on chromosome 10. analysis implicates cytoskeletal abnormalities in psy- Science. 2000;290(5500):2304-2305. chiatric disease. Proteomics. 2009;9(12):3368- Statistical analysis: Ikram, DeStefano, Bis, Smith, 10. Li H, Wetten S, Li L, et al. Candidate single- 3382. Harold, Schrijvers, Janssens, Hollingworth, Aspelund, nucleotide polymorphisms from a genomewide asso- 28. Yang S, Liu T, Li S, et al. Comparative proteomic Beiser, Abraham, Du, Aulchenko, Vardarajan, ciation study of Alzheimer disease. Arch Neurol. 2008; analysis of brains of naturally aging mice. Neuroscience. Rivadeneira, Nalls, Rice, Lathrop, Farrer, Lumley, Ruiz, 65(1):45-53. 2008;154(3):1107-1120. Duijn. 11. Coon KD, Myers AJ, Craig DW, et al. A high- 29. Poduslo SE, Yin X. A new locus on chromosome Obtained funding: Seshadri, Gudnason, Boada, density whole-genome association study reveals that 19 linked with late-onset Alzheimer’s disease. Dartigues, Kuller, Antunez, Rotter, Harris, Petersen, APOE is the major susceptibility gene for sporadic late- Neuroreport. 2001;12(17):3759-3761. Becker, Jonsson, Schmidt, Uitterlinden, Ruiz, Younkin, onset Alzheimer’s disease. J Clin Psychiatry. 2007; 30. Starcevic M, Dell’Angelica EC. Identification of Launer, Lopez, Breteler. 68(4):613-618. snapin and three novel proteins (BLOS1, BLOS2, and Administrative, technical, or material support: Seshadri, 12. Feulner TM, Laws SM, Friedrich P, et al. Exami- BLOS3/reduced pigmentation) as subunits of biogen- Fitzpatrick, Bis, Carrasquillo, Ramirez-Lorca, Pankratz, nation of the current top candidate genes for AD in esis of lysosome-related organelles complex-1 Hernandez, Kuller, Koudstaal, Antunez, Rotter, Becker, a genome-wide association study. Mol Psychiatry. (BLOC-1). J Biol Chem. 2004;279(27):28393- Nalls, Campion, Auerbach, Mosley, Psaty, Uitterlinden, 2009. 28401. Farrer, Wolf. 13. Harold D, Abraham R, Hollingworth P, et al. Ge- 31. Morris DW, Murphy K, Kenny N, et al. Dysbin- Study supervision: Seshadri, Gudnason, Boada, nome-wide association study identifies variants at CLU din (DTNBP1) and the biogenesis of lysosome- Koudstaal, Tzourio, Antunez, Rotter, Berr, Owen, and PICALM associated with Alzheimer’s disease. Nat related organelles complex 1 (BLOC-1). Biol Psychiatry. Hofman, Lathrop, Uitterlinden, Williams, Amouyel, Genet. 2009;41(10):1088-1093. 2008;63(1):24-31. Duijn, Breteler. 14. Lambert JC, Heath S, Even G, et al. Genome- 32. Moroni RF, De Biasi S, Colapietro P, et al. Dis- Financial Disclosures: None reported. wide association study identifies variants at CLU and tinct expression pattern of microtubule-associated pro- Funding/Support: The funding/support for this study CR1 associated with Alzheimer’s disease. Nat Genet. tein/microtubule affinity-regulating kinase 4 in dif- is described in the eAppendix. 2009;41(10):1094-1099. ferentiated neurons. Neuroscience. 2006;143(1): Role of the Sponsor: The funding organizations and 15. Psaty BM, O’Donnell CJ, Gudnason V, et al. Co- 83-94. sponsors had no role in the design and conduct of the horts for Heart and Aging Research in Genomic Epi- 33. Trinczek B, Brajenovic M, Ebneth A, Drewes G. study; collection, management, analysis, and inter- demiology (CHARGE) Consortium. Circ Cardiovasc MARK4 is a novel microtubule-associated proteins/ pretation of the data; and preparation, review or ap- Genet. 2009;2(1):73-80. microtubule affinity-regulating kinase that binds to the proval of the manuscript. The final version submitted 16. Association AP. Diagnostic and Statistical cellular microtubule network and to centrosomes. was approved without changes by the National Heart, Manual of Mental Disorders (DSM-IV). Wash- J Biol Chem. 2004;279(7):5915-5923. Lung, and Blood Institute and the National Institute ington, DC: American Psychiatric Association; 34. Purcell SM, Wray NR, Stone JL, et al. Common on Aging. 1994. polygenic variation contributes to risk of schizophre- Online-Only Material: The eAppendix, eTables 1 17. McKhann G, Drachman D, Folstein M, et al. Clini- nia and bipolar disorder. Nature. 2009;460(7256): through 3, and eFigures 1 through 10 are available cal diagnosis of Alzheimer’s disease. Neurology. 1984; 748-752. at http://www.jama.com. 34(7):939-944. 35. Janssens AC, van Duijn CM. Genome-based pre- Additional Contributions: Additional contributions are 18. Pe’er I, Yelensky R, Altshuler D, Daly MJ. Estima- diction of common diseases. Hum Mol Genet. 2008; listed in the eAppendix. tion of the multiple testing burden for genomewide 17(R2):R166-R173.

Downloaded From: https://jamanetwork.com/ on 09/24/2021