|Home | About | Journals | Submit | Contact Us | Français|
Recent genome-wide association studies (GWAS) of late-onset Alzheimer disease (LOAD) identified 9 novel risk loci. Discovery of functional variants within genes at these loci is required to confirm their role in Alzheimer disease (AD). Single nucleotide polymorphisms that influence gene expression (eSNPs) constitute an important class of functional variants. We therefore investigated the influence of the novel LOAD risk loci on human brain gene expression.
We measured gene expression levels in the cerebellum and temporal cortex of autopsied AD subjects and those with other brain pathologies (~400 total subjects). To determine whether any of the novel LOAD risk variants are eSNPs, we tested their cis-association with expression of 6 nearby LOAD candidate genes detectable in human brain (ABCA7, BIN1, CLU, MS4A4A, MS4A6A, PICALM) and an additional 13 genes ±100 kb of these SNPs. To identify additional eSNPs that influence brain gene expression levels of the novel candidate LOAD genes, we identified SNPs ±100 kb of their location and tested for cis-associations.
CLU rs11136000 (p = 7.81 × 10−4) and MS4A4A rs2304933/rs2304935 (p = 1.48 × 10−4–1.86 × 10−4) significantly influence temporal cortex expression levels of these genes. The LOAD-protective CLU and risky MS4A4A locus alleles associate with higher brain levels of these genes. There are other cis-variants that significantly influence brain expression of CLU and ABCA7 (p = 4.01 × 10−5–9.09 × 10−9), some of which also associate with AD risk (p = 2.64 × 10−2–6.25 × 10−5).
CLU and MS4A4A eSNPs may at least partly explain the LOAD risk association at these loci. CLU and ABCA7 may harbor additional strong eSNPs. These results have implications in the search for functional variants at the novel LOAD risk loci.
Recent genome-wide association studies (GWAS) of late-onset Alzheimer disease (LOAD) identified 9 novel loci, in addition to APOE.1–6 Despite the success of disease GWAS, there remains a knowledge gap with this approach. First, loci identified by disease GWAS may harbor more than one candidate gene.7 Second, the mechanism of action of the risk variants at the disease loci is not immediately obvious.1,7 Third, variants identified in complex disease GWAS do not explain all their underlying genetic components.8,9
An alternative approach is using biologically relevant, genetically driven, quantitative phenotypes, i.e., endophenotypes. Gene expression levels are a special group of endophenotypes with a substantial genetic component.7,10–17 Identification of variants that influence both gene expression and human disease can discover the actual risk genes at the GWAS loci and their potential mechanism of action.7,10,11,13,14,16
We postulate that LOAD has multiple risk variants, some of which influence gene expression (eSNPs). In this study, we assessed the novel LOAD GWAS loci, first to determine if any of the “top LOAD risk SNPs” influence brain expression of nearby genes in-cis. We arbitrarily defined cis-associations as those between a transcript and a SNP, which is within the gene encoding that transcript or within 100 kb of its 5′ start or 3′ end sites. Second, we aimed to determine if the brain expression levels of any of the “top LOAD candidate genes” at these loci are influenced by any other eSNPs in-cis (cis-eSNPs). In this study, we utilize the term eSNPs to define variants associating with expression levels and not to indicate direct evidence of functionality, as the eSNPs are likely markers of other functional, regulatory variants. To determine whether any of the significant eSNPs that are identified in this second aim also influence AD risk, we utilized meta-analyses results from the LOAD risk GWAS of the Alzheimer's Disease Genetics Consortium (ADGC).6 Our findings have implications about the potential mechanism of action for some of the top LOAD risk SNPs and for functional regulation of some of the top LOAD candidate genes.
We measured gene expression levels from temporal cortex (TCX) of 399 subjects and from cerebellar tissue (CER) of 374 subjects (table 1). A total of 340 subjects had both TCX and CER measurements. All subjects were participants in the published Mayo LOAD GWAS18 as part of the autopsy-based series (AUT_60–80). All subjects had neuropathologic evaluation by D.W.D. All subjects with AD (n = 202 for TCX and 197 for CER) had definite diagnosis of AD according to the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association criteria19 and had Braak scores of ≥4.0. All control subjects (non-AD, n = 197 for TCX and 177 for CER) had Braak scores of ≤2.5, but many had brain pathology unrelated to AD (table e-1 on the Neurology® Web site at www.neurology.org).
Total RNA was extracted from frozen TCX or CER samples using the Ambion RNAqueous kit according to the manufacturer's instructions. The quantity and quality of the RNA samples were determined by the Agilent 2100 Bioanalyzer using the Agilent RNA 6000 Nano Chip.
Transcript levels were measured using the Whole Genome DASL assay (Illumina, San Diego, CA). The RNA samples were randomized across the chips and plates using a stratified approach to ensure balance with respect to diagnosis, age, gender, RNA Integrity Numbers (RINs), and APOE genotype. Replicate samples were utilized for quality control (QC). Raw probe-level mRNA expression data of the 399 TCX and 374 CER samples were exported from GenomeStudio software (Illumina Inc.) for preprocessing with background correction, variance stabilizing transformation, quantile normalization, and probe filtering using the lumi package of BioConductor24,25 (appendix e-2: Methods).
This study was approved by the appropriate institutional review board and appropriate informed consent was obtained from all participants.
All genotypes used in the SNP/transcript level association studies were extracted from the Mayo LOAD GWAS, since all autopsied subjects with TCX or CER gene expression measurements were part of this study.18 The LOAD GWAS genotypes were generated using Illumina's HumanHap300-Duo Genotyping BeadChips analyzed with an Illumina BeadLab Station (Illumina) at the Mayo Clinic Genotyping Shared Resource according to the manufacturer's protocols. The LOAD GWAS QC methods were previously published10,18 (appendix e-2: Methods).
To test the influence of the top LOAD risk SNPs on brain gene expression in-cis, we extracted from the Mayo LOAD GWAS the genotypes for these SNPs or their proxies. The Mayo LOAD GWAS had genotypes for the following 5 top LOAD risk SNPs: BIN1_rs744373, CLU_rs1136000, EPHA1_rs11767557, MS4A6A_rs610932, and PICALM _rs3851179; and the proxies for ABCA7_rs3764650, MS4A4A_rs4938933, MS4A6A_rs610932, MS4A4E_rs670139, and PICALM_rs561655 (table e-2). The top LOAD risk SNPs or their proxies were tested for cis-association with the transcript levels of those genes that resided within ±100 kb of their genomic location according to NCBI Build 36. Of the top LOAD candidate genes nearby these SNPs, the following had detectable brain expression levels in both TCX and CER: ABCA7, BIN1, CLU, MS4A4A, MS4A6A, and PICALM. Thirteen additional nearby genes were also tested for SNP/transcript levels associations. A total of 59 SNP/transcript associations were tested (tables e-2 and e-3).
To test whether the brain expression levels of any of the top LOAD candidate genes are influenced by other eSNPs in-cis (cis-eSNPs), we identified all SNPs in the Mayo LOAD GWAS within ±100 kb flanking region of the gene. ABCA7, BIN1, CLU, MS4A4A, MS4A6A, PICALM, and CD2AP were tested for additional eSNPs. A total of 369 SNP/transcript associations were tested (table e-4 and e-5). Study-wide Bonferroni corrections were done for the total number of SNP/transcript associations tested.
Linear regression analysis to test for SNP/transcript associations were done in PLINK.20 Preprocessed probe transcript levels were utilized as endophenotypes. Each probe was assessed separately, even though 1 gene may have multiple probes. An additive model was assumed, with the minor allele dosage (0, 1, 2) as the independent variable, and APOE ε4 dosage (0, 1, 2), age at death, gender, PCR plate, RIN, and adjusted RIN2 (defined as [RIN − RINmean]2) as covariates. The CER and TCX results were analyzed separately. The AD and non-AD subjects were analyzed both separately and jointly. The joint analyses including diagnosis as an additional covariate (AD = 1, non-AD = 0) were considered to be the main analyses. The separate AD and non-AD only analyses are also reported (tables e-1 through e-6).
The top LOAD risk SNPs that were assessed for their influence on brain gene expression have AD risk association results published in the literature. To determine the AD risk association of the remaining SNPs assessed in this study (i.e., proxies for the top LOAD risk SNPs and the other cis-eSNPs), we obtained meta-analyses results from the ADGC.6 The cohorts that are assessed by ADGC, as well as the methodologic details of the meta-analyses, are described in detail in a recent publication.6 Briefly, the meta-analyses of the ADGC dataset results reported here (tables 2, ,33 and e-2 through e-5) are generated from the combined analyses of stage 1 and stage 2 cohorts (appendix e-2: Methods), with detailed descriptions provided elsewhere.6,21 Stage 1 cohorts are comprised of 8,309 LOAD cases and 7,366 cognitively normal elder controls. Stage 2 has 3,531 LOAD vs 3,565 control subjects.
Each cohort was tested for AD risk association using a logistic regression approach, assuming an additive model and adjusting for age, sex, APOE ε4 dosage, and principal components from EIGENSTRAT.22 The meta-analyses results were generated using the inverse variance method implemented in the software package METAL.23
We evaluated 18 top LOAD risk SNPs or their proxies from 6 novel LOAD loci for associations with TCX and CER expression levels of 6 top LOAD candidate genes and 13 others within ±100 kb of these SNPs. All 19 genes had detectable expression both in TCX and CER of autopsied subjects. Given the assessment of 59 SNP/transcript associations (table e-2), p <8.47 × 10−4 is required to achieve significance after Bonferroni correction. CLU_rs11136000 and 2 SNPs at the MS4A locus (rs2304933 and rs2304935) have significant association with TCX expression levels of CLU (p = 7.81 × 10−4) and MS4A4A (p = 1.48 × 10−4 and 1.86 × 10−4), respectively (table 2 and figures e-1 and e-2). The MS4A4A eSNPs also have nominally significant associations in CER (p = 3.65 × 10−2), but CLU_rs11136000/transcript association is not significant in CER. The significant MS4A4A SNP/transcript associations appear to be driven by both AD and non-AD subjects, however CLU_rs11136000/transcript association achieves significance only in the non-ADs (tables e-2 and e-3).
CLU_rs11136000 minor “A” allele associates with both higher TCX expression levels of this gene and lower AD risk (tables 2, e-2, e-3, and figure 1A). CLU_rs11136000 reached genome-wide significance in published LOAD GWAS2,3 and also has significant association with lower AD risk in the meta-analyses of the ADGC dataset (OR = 0.89, p = 5.23 × 10−7). The minor alleles of MS4A4A SNPs rs2304933 and rs2304935 associate with higher brain expression levels of this and increased AD risk that is nominally significant in the meta-analysis of the ADGC dataset (OR = 1.06, p = 1.11 × 10−2 to 1.13 × 10−2) (table 2, table e-3, figure 1, A and B). The MS4A4A transcript-associating SNPs rs2304933 and rs2304935 are proxies for rs670139, which is intergenic between MS4A4E and MS4A6A and which showed genome-wide significant association with increased AD risk.5
We evaluated expression levels of 7 novel top LOAD candidate genes, at 6 of the recently identified novel LOAD loci, detectable in TCX and CER of autopsied subjects, for their association with cis-eSNPs residing within ±100 kb of their flanking region. Given the 369 SNP/transcript associations tested, p < 1.36 × 10−4 is needed to achieve Bonferroni-corrected study-wide significance. There were 9 significant SNP/transcript associations; 6 for CLU expression in TCX (p = 4.01 × 10−5 to 9.00 × 10−9), and 3 for ABCA7 in both TCX and CER (p = 3.53 × 10−5 to 1.03 × 10−7) (tables 3, e-4, and e-5 and figures e-1 and e-3). ABCA7_rs7247087/transcript associations were significant in both brain regions, and the other 2 ABCA7 SNPs (rs757232 and rs2072102) that were significant in CER had nominal significance in TCX. The direction and magnitude of the effect on ABCA7 levels were similar in the 2 brain regions for all 3 SNPs. None of the 5 CLU SNPs with significant TCX transcript associations showed nominal significance in CER. All significant CLU and ABCA7 SNP/transcript associations showed effects of similar direction and magnitude in both AD and non-AD brains.
CLU_ rs569214 which is associated with higher levels of this gene in TCX has nominally significant association with lower AD risk in the meta-analysis of the ADGC dataset (table 3, table e-5, figure 1A). ABCA7_ rs757232 and rs2072102 show significant association with higher ABCA7 levels in 2 brain regions as well as significantly higher risk of AD (OR = 1.11, p = 6.25 × 10−5 to 6.74 × 10−5) in the meta-analysis of the ADGC dataset (table 3, table e-5, figure 1, A and B). None of the other significant eSNPs showed any association with AD risk in this study.
The novel disease loci identified in the recent large disease GWAS provide the opportunity to uncover the pathophysiology of complex diseases, such as LOAD, which may be instrumental in the discovery of novel drug targets and development of early diagnostic tools. Nonetheless, significant progress awaits the identification of the true disease genes at these loci, their functional risk alleles and mechanisms of action. In this study, we used the brain gene expression endophenotype and disease GWAS results in a combined fashion to test whether any of the top LOAD risk SNPs (referred henceforth as LOAD SNPs for short) are eSNPs and whether the expression of any of the top LOAD candidate genes is influenced by any other cis-eSNPs that are within ±100 kb of their genomic location.
Our approach revealed several important findings. First, the LOAD SNPs at the CLU and MS4A loci also influence brain gene expression of CLU and MS4A4A genes, respectively. Second, there are additional variants within the arbitrary 100 kb cis-region that also influence brain expression of CLU and ABCA7. Third, these additional, strong CLU and ABCA7 cis-eSNPs do not appear to be in strong LD with the LOAD SNPs at their respective loci.
The strongest LOAD SNP at the ABCA7 locus, rs3764650, was assessed by its proxy rs375229 in our study and not found to have a significant association with brain ABCA7 levels. Conversely, 2 of the strong ABCA7 cis-eSNPs (rs757232, rs2072102) also have significant AD risk association (p = 6.25 × 10−5 to 6.74 × 10−5). For the CLU locus, the top LOAD SNP, rs11136000, also associates with CLU levels in the temporal cortex. Of the additional strong CLU cis-eSNPs identified in this study, rs569214 associates with AD risk with modest nominal significance (p = 2.64 × 10−2). That the ABCA7 and CLU loci LOAD SNPs and additional cis-eSNPs, which also associate with AD risk, are not in strong LD with each other suggests that there may be multiple, independent variants at these loci that influence AD risk. Furthermore, at least some of this AD risk is conferred by regulatory variants which affect brain gene expression levels.
The direction of the CLU SNP/transcript and SNP/AD risk associations are biologically congruent. The top LOAD risk SNP rs11136000 and the other CLU transcript-associating SNPs have alleles which associate with increased brain levels of this gene and also reduce AD risk. CLU encodes for clusterin, which is thought to promote neuroprotection in AD via multifaceted functions including Aβ clearance, prevention of excessive inflammation, inhibition of apoptosis, and clearance of neuronal debris.24 Thus, genetic variants that influence higher brain CLU levels could conceivably lower AD risk.
MS4A locus SNPs associate with higher levels of MS4A4A and increased AD risk. The functions of the MS4A locus genes, including MS4A4A, are yet to be established, though they are thought to be members of a family of transmembrane proteins which may be parts of oligomeric cell surface complexes involved in signal transduction.25 Given that the MS4A locus harbors 6 genes in an LD block (MS4A3, MS4A2, MS4A6A, MS4A4E, MS4A4A, MS4A6E), it is not possible to distinguish the actual LOAD risk genes in this region based solely on the disease GWAS findings. Our findings suggest that the MS4A4A gene might be the strongest AD candidate at this locus, though they need to be interpreted with caution, since brain expression levels of only MS4A4A and MS4A6A could be tested. Thus, there may be other MS4A genes that are influenced by SNPs at this locus. Indeed, we found nominally significant SNP/transcript associations with MS4A6A and the downstream MS4A7. Others have previously shown nominally significant rs610932/MS4A6A transcript associations in brains of 143 neurologically normal, European subjects,5 though we were not able to confirm this finding in our larger study.
The ABCA7 transcript-associating SNPs both increase its brain levels and AD risk. ABCA7 is member of the ATP-binding cassette (ABC) family of proteins implicated in lipid metabolism,26 shown in vitro to regulate cholesterol efflux, inhibit amyloid precursor protein processing,27 and also to play a role in the phagocytosis of apoptotic cells.28 ABCA7 has a splice variant with a distinct expression profile and which lacks the lipid metabolism functionality.29 Understanding the implications of ABCA7 expressional regulation in AD risk requires further information about the functions of this protein, and clarification about the brain ABCA7 splice variant that is influenced by the ABCA7 SNPs.
Our results also have implications regarding fine mapping studies aimed at functional variant discovery at the novel LOAD risk loci and suggest that screening only for coding polymorphisms may miss important functional variation. The strongest eSNP associations are in the 5′ or 3′ regions of the transcripts and sometimes in different genes. Bioinformatics focused on transcriptional regulation sites in noncoding regions within and surrounding the CLU, MS4A, and ABCA7 genes, followed by targeted sequencing and in vitro functional efforts, may be fruitful in functional variant discovery at these loci.
It is clear that not all the functional disease variants will be eSNPs. We also note that the strongest LOAD risk SNP was not the strongest eSNP and vice versa in this study. This could be due to differences in the populations tested for the LOAD risk and expression associations, sample sizes, LD patterns, or a combination of these factors. Nonetheless, it is also possible that the LOAD risk SNPs identified in the disease GWAS may be marking multiple different types of functional variants, i.e., both eSNPs and missense coding variants. Our results suggest that for 3 of the novel LOAD GWAS loci, eSNPs account for at least part of the disease risk. Not all novel LOAD candidate genes were detectable in both CER and TCX of our samples with the DASL microarrays. It will be important to investigate SNP associations for brain expression levels of the remaining LOAD candidate genes using alternative approaches. It is also necessary to characterize in detail the influence of the eSNPs on brain expression levels of any splice variants.
The strengths of our study include combined use of the gene expression endophenotype and disease GWAS, use of a relatively large sample size of both AD and non-AD brain tissue from 2 different regions, and investigation of both the top LOAD risk SNPs as well as other cis-eSNPs for their effects on brain expression of 7 top LOAD candidate genes. Our study demonstrates the power and utility of the gene expression endophenotypes; identifies brain expression changes in CLU, MS4A4A, and ABCA7 as one potential mechanism of action at these novel LOAD risk loci; and provides direction about functional variant discovery at these loci.
Liana G. Apostolova, MD, Steven E. Arnold, MD, Clinton T. Baldwin, PhD, Robert Barber, PhD, Michael M. Barmada, PhD, Thomas Beach, MD, PhD, Gary W. Beecham, PhD, Duane Beekly, BS, David A. Bennett, MD, Eileen H. Bigio, MD, Thomas D. Bird, MD, Deborah Blacker, MD, Bradley F. Boeve, MD, James D. Bowen, MD, Adam Boxer, MD, PhD, James R. Burke, MD, PhD, Jacqueline Buros, BS, Joseph D. Buxbaum, PhD, Nigel J. Cairns, PhD, FRCPath, Laura B. Cantwell, MPH, Chuanhai Cao, PhD, Chris S. Carlson, PhD, Regina M. Carney, MD, Steven L. Carroll, MD, PhD, Helena C. Chui, MD, David G. Clark, MD, Jason Corneveaux, BS, Carl W. Cotman, PhD, Paul K. Crane, MD, MPH, Carlos Cruchaga, PhD, Jeffrey L. Cummings, MD, Philip L. De Jager, MD, PhD, Charles DeCarli, MD, Steven T. DeKosky, MD, F. Yesim Demirci, MD, Ramon Diaz-Arrastia, MD, PhD, Malcolm Dick, PhD, Beth A. Dombroski, PhD, Ranjan Duara, MD, William G. Ellis, MD, Denis Evans, MD, Kelley M. Faber, MS, Kenneth B. Fallon, MD, Martin R. Farlow, MD, Steven Ferris, PhD, Tatiana M. Foroud, PhD, Matthew P. Frosch, MD, PhD, Douglas R. Galasko, MD, Paul J. Gallins, MS, Mary Ganguli, MD, Marla Gearing, PhD, Daniel H. Geschwind, MD, PhD, Bernardino Ghetti, MD, John R. Gilbert, PhD, Sid Gilman, MD, FRCP, Bruno Giordani, PhD, Jonathan D. Glass, MD, Alison M. Goate, D.Phil, Robert C. Green, MD, John H. Growdon, MD, Hakon Hakonarson, MD, PhD, Ronald L. Hamilton, MD, John Hardy, PhD, Lindy E. Harrell, MD, PhD, Elizabeth Head, PhD, Lawrence S. Honig, MD, PhD, Matthew J. Huentelman, PhD, Christine M. Hulette, MD, Bradley T. Hyman, MD, PhD, Gail P. Jarvik, MD, PhD, Gregory A. Jicha, MD, PhD, Lee-Way Jin, MD, PhD, Nancy Johnson, PhD, Gyungah Jun, PhD, M. Ilyas Kamboh, PhD, Jason Karlawish, MD, Anna Karydas, BA, John S.K. Kauwe, PhD, Jeffrey A. Kaye, MD, Ronald Kim, MD, Edward H. Koo, MD, Neil W. Kowall, MD, Patricia Kramer, PhD, Walter A. Kukull, PhD, James J. Lah, MD, PhD, Eric B. Larson, MD, MPH, Allan I. Levey, MD, PhD, Andrew P. Lieberman, MD, PhD, Oscar L. Lopez, MD, Kathryn L. Lunetta, PhD, Wendy J. Mack, PhD, Daniel C. Marson, JD, PhD, Eden R. Martin, PhD, Frank Martiniuk, PhD, Deborah C. Mash, PhD, Eliezer Masliah, MD, Wayne C. McCormick, MD, MPH, Susan M. McCurry, PhD, Andrew N. McDavid, BA, Ann C. McKee, MD, Marsel Mesulam, MD, Bruce L. Miller, MD, Carol A. Miller, MD, Joshua W. Miller, PhD, Thomas J. Montine, MD, PhD, John C. Morris, MD, Amanda J. Myers, PhD, Adam C. Naj, PhD, Petra Nowotny, PhD, Joseph E. Parisi, MD, Daniel P. Perl, MD, Elaine Peskind, MD, Wayne W. Poon, PhD, Huntington Potter, PhD, Joseph F. Quinn, MD, Ashok Raj, MD, Ruchita A. Rajbhandary, MPH, Murray Raskind, MD, Eric M. Reiman, MD, Barry Reisberg, MD, Christiane Reitz, MD, PhD, John M. Ringman, MD, Erik D. Roberson, MD, PhD, Ekaterina Rogaeva, PhD, Roger N. Rosenberg, MD, Mary Sano, PhD, Andrew J. Saykin, PsyD, Julie A. Schneider, MD, Lon S. Schneider, MD, William Seeley, MD, Michael L. Shelanski, MD, PhD, Michael A. Slifer, MD, PhD, Charles D. Smith, MD, Joshua A. Sonnen, MD, Salvatore Spina, MD, Peter St George-Hyslop, MD, FRCP, Robert A. Stern, PhD, Rudolph E. Tanzi, PhD, John Q. Trojanowski, MD, PhD, Juan C. Troncoso, MD, Debby W. Tsuang, MD, Vivianna M. Van Deerlin, MD, PhD, Badri Narayan Vardarajan, MS, Harry V. Vinters, MD, Jean Paul Vonsattel, MD, Li-San Wang, PhD, Sandra Weintraub, PhD, Kathleen A. Welsh-Bohmer, PhD, Jennifer Williamson, MS, Randall L. Woltjer, MD, PhD.
M.A.: drafting/revising the manuscript for content, study concept, analysis or interpretation of data. F.Z., H.S.C., C.S.Y., J.C., V.S.P., M.M.C.: drafting/revising the manuscript for content, analysis or interpretation of data. C.N.R., A.A.N., S. Middha, S. Maharjan, T.N., L.M., K.G.M., R.P., S.L., G.B., C.G., D.S., F.R.: analysis or interpretation of data. C.P.K., J.J., J.L.H., R.M., M.A.P.V., L.A.F., G.D.S.: drafting/revising the manuscript for content, analysis or interpretation of data. R.C.P., N.R.G.-R., D.W.D.: drafting/revising the manuscript for content. S.G.Y., N.E.T.: drafting/revising the manuscript for content, study concept, analysis or interpretation of data.
The authors thank the study participants and their families who made this work possible and members of the Mayo Clinic Jacksonville Memory Disorders Clinic and Mayo Clinic AD Research Center for their help in the collection of the samples.
Supported by National Institutes of Health grants: National Institute on Aging (R01 032990 to N.E.-T. and R01 AG018023 to N.R.G.-R. and S.G.Y.); Mayo Alzheimer's Disease Research Center (P50 AG016574 to R.C.P., D.W.D., N.R.G.-R., S.G.Y., and N.E.-T.); Mayo Alzheimer's Disease Patient Registry (U01 AG006576 to R.C.P.); National Institute on Aging (AG025711, AG017216, AG003949 to D.W.D.). Also supported by the Robert and Clarice Smith and Abigail Van Buren Alzheimer's Disease Research Program (to R.C.P., D.W.D., N.R.G.-R., and S.G.Y.), and by the Palumbo Professorship in Alzheimer's Disease Research (to S.G.Y.). N.E.T. is the recipient of National Institutes of Health (KL2 RR024151) and Siragusa Foundation grants.
N. Graff-Radford has served as a consultant to Codman and received grant support from Elan Pharmaceutical Research, Pfizer Pharmaceuticals, Medivation, and Forrest. R.C. Petersen has been a consultant to GE Healthcare and Elan Pharmaceuticals, has served on a data safety monitoring board in a clinical trial sponsored by Elan Pharmaceuticals, and a safety monitoring board for Wyeth Pharmaceuticals. The remaining authors report no disclosures relevant to the manuscript. Go to Neurology.org for full disclosures.