Volume 20, Issue 9 p. 5973-5995
RESEARCH ARTICLE
Open Access

A genome-wide association meta-analysis of all-cause and vascular dementia

The Mega Vascular Cognitive Impairment and Dementia (MEGAVCID) consortium

The Mega Vascular Cognitive Impairment and Dementia (MEGAVCID) consortium

Search for more papers by this author
First published: 24 July 2024

Please contact Bernard Fongang for correspondence at Long School of Medicine, UT Health San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229, USA. Email: [email protected]

[Correction added on July 26, 2024, after first online publication: The byline was revised to list only the consortium name. Bernard Fongang is acting as the contact for correspondence for the consortium and had been incorrectly listed in the byline.]

Abstract

INTRODUCTION

Dementia is a multifactorial disease with Alzheimer's disease (AD) and vascular dementia (VaD) pathologies making the largest contributions. Yet, most genome-wide association studies (GWAS) focus on AD.

METHODS

We conducted a GWAS of all-cause dementia (ACD) and examined the genetic overlap with VaD. Our dataset includes 800,597 individuals, with 46,902 and 8702 cases of ACD and VaD, respectively. Known AD loci for ACD and VaD were replicated. Bioinformatic analyses prioritized genes that are likely functionally relevant and shared with closely related traits and risk factors.

RESULTS

For ACD, novel loci identified were associated with energy transport (SEMA4D), neuronal excitability (ANO3), amyloid deposition in the brain (RBFOX1), and magnetic resonance imaging markers of small vessel disease (SVD; HBEGF). Novel VaD loci were associated with hypertension, diabetes, and neuron maintenance (SPRY2, FOXA2, AJAP1, and PSMA3).

DISCUSSION

Our study identified genetic risks underlying ACD, demonstrating overlap with neurodegenerative processes, vascular risk factors, and cerebral SVD.

Highlights

  • We conducted the largest genome-wide association study of all-cause dementia (ACD) and vascular dementia (VaD).
  • Known genetic variants associated with AD were replicated for ACD and VaD.
  • Functional analyses identified novel loci for ACD and VaD.
  • Genetic risks of ACD overlapped with neurodegeneration, vascular risk factors, and cerebral small vessel disease.

1 BACKGROUND

Traditionally, Alzheimer's disease (AD) is considered the most common dementia subtype, followed by vascular dementia (VaD). The two conditions are considered clinically distinct. VaD is diagnosed based on the presence of stroke or extensive cerebral vascular disease, with atherosclerosis and arteriolosclerosis considered the underlying pathologies.1 However, a wealth of evidence from recent years has emphasized a broad role for brain vascular damage, beyond that of lacunar and larger cerebral infarcts, as a major mechanism for cognitive impairment.2 It is now increasingly recognized that a component of vascular pathology is prominent in all major dementias and acts synergistically with amyloid beta (Aβ), tau, and other neurodegenerative pathologies to affect dementia risk.3 Moreover, a new hypothetical model of dementia dynamics suggests that damage to brain vasculature is an early process in the dementia continuum that precedes brain atrophy, neurodegeneration, and the emergence of amyloid and tau biomarker abnormalities.4 Recent genetic studies using methods that are relatively immune to reverse causation also suggest a putative causal relationship between brain imaging markers of cerebral small vessel disease (SVD) and AD.5

Hence there is a strong rationale to examine the “vascular contributions to cognitive impairment and dementia” (VCID), a term that includes a broad range of vascular mechanisms and phenotypes and represents the multifactorial nature of dementia and related disorders as a pathway for reducing dementia burden.6 In particular, genetic exploration of VCID may highlight important mechanisms across the wide spectrum of pathologies, including vascular pathways, which, in turn, are considered to be a major and modifiable target for the prevention of dementia, including the Alzheimer's type.7

Emerging evidence suggests that VCID is highly heritable.8 Mutations in the NOTCH3 gene known to cause monogenic cerebral SVD and early cognitive impairment also influence later onset polygenic manifestations of VCID by acting through common, less pathogenic variations in the same genes. Other examples are several point mutations in the amyloid precursor protein (APP) gene that lead to cerebral amyloid angiopathy (CAA)9 as well as mutations in HtrA Serine Peptidase 1 (HTRA1) and Collagen Type IV Alpha 1 Chain (COL4A1) or COL4A2 genes.10 Further support for the strong genetic basis of VCID stems from heritability and genome-wide association studies (GWASs) of cerebral SVD endophenotypes that are closely related to VCID, including ischemic stroke (IS),11 and white matter hyperintensities (WMHs).12, In contrast to the over 70 loci identified as being associated with AD genetic variance, the genetic architecture of “sporadic” VCID is largely unknown. Most genetic studies of VCID have utilized a candidate gene approach, which did not yield consistent and replicable findings.13

RESEARCH IN CONTEXT

  1. Systematic review: While findings from genome-wide association studies (GWASs) of Alzheimer's disease (AD) highlighted multiple genetic risk variants, the genetics of all-cause dementia (ACD) and vascular dementia (VaD) has been rarely studied. In this meta-analysis of unpublished GWASs, we utilized data from 21 cohorts and consortia for a total of 46,902 and 8702 cases of ACD and VaD, respectively.

  2. Interpretation: Known genetic variants for AD were identified as risk factors for ACD and VaD. Downstream bioinformatics revealed novel genetic loci functionally associated with ACD and VaD, including SEMA4D, RBFOX1, and SPRY2.

  3. Future directions: These results should be validated in additional datasets. Particularly, studies are warranted to explore the genetic variation of ACD and VaD in non-European individuals.

GWASs of VCID are sparse. In 2012, a GWAS of VaD conducted among the participants of the Rotterdam Study (N = 67 cases and 5700 controls) identified a novel locus associated with VaD, located near the androgen receptor on the X chromosome14; however, this finding could not be replicated.15 More recently, a GWAS of dementia and its clinical endophenotypes was conducted as part of the GR@ACE study.16 This study demonstrated the differential biological pathways associated with clinical AD subgroups based on the degree of vascular burden. It identified a variant near CNTNAP2 associated with probable or possible VCID (N = 373). However, this finding did not reach genome-wide (GW) significance.

The multifactorial nature of VaD and the heterogeneity of the clinicopathological criteria used to define this entity have hampered the identification of genetic polymorphisms underlying VCID. To overcome these limitations, large-scale studies with sufficient power to detect genetic signals specific to VCID are needed.17 In this study, we investigate the genetic predisposition to VCID specifically. Hence, we explored the genetic variability associated with ACD as a broad phenotype, as well as VaD as an extreme phenotype of the dementia continuum characterized by increased vascular burden. Our findings were then analyzed in light of the knowledge already gained from previous large-scale GW and sequencing studies on the genetic determinants of AD, stroke, and additional phenotypes along the VCID spectrum.10

2 METHODS

2.1 Study population

A total of 800,597 participants from 21 cohorts and consortia contributed to 46,902 and 8702 cases of ACD and VaD, respectively. The overall sample included individuals from four different ethnicities (European, African, Asian, and Hispanic) from North America, Europe, and Asia. The mean age ranged between 54 and 80 years, with 54% to 68% females. The summary demographics are described in Table 1 (also detailed in Table S1 of Supplementary File 1). Each study obtained written informed consent from participants or, for those with substantial cognitive impairment, from a caregiver, legal guardian, or other proxy. Study protocols for all cohorts were reviewed and approved by the appropriate institutional review boards.

TABLE 1. Demographics: Data from 17 CHARGE cohorts were included in our meta-analysis, as were the UKBB, ADGC, and EADB for the replication of our VaD results in European ancestry.
Study N/Control ACD VaD Percentage VaD Age (mean) Sex, % (percentage female)
European ancestry
3C 6475 808 162 20.1 74.2 61.0
AGES 5656 1501 118 7.9 76.1 61.0
ARIC 3145 165 36 21.8 75.5 60.0
ASPREE 12,480 319 NA NA 75.0 55.0
CHS 2169 508 156 30.7 74.9 61.5
FVG 804 73 NA NA 58.2 58.3
FHS 4175 679 167 24.5 54.6 54.3
GRACE 12,599 7516 1953 26.0 78.8 68.2
GREAT-AGE 1504 138 7 5.1 73.7 50.3
HUNT 69,633 3982 681 17.1 67.7 57.4
MEMENTO 2050 263 36 13.7
MYHAT 865 50 NA NA 83.7 59.5
ROSMAP 1335 626 NA NA 79.8 69.7
RS (1,2,3) 11,390 1715 178 10.4 63.6 56.8
ADGC-NAJ-2011 15,675 8309 NA NA 75.4 59.5
UKBB 314,278 17,008 332 NA 66.1 63.1
Total European 466,606 44,009 3892
African ancestry
ARIC 905 101 31 30.7 75.5 60.0
CHS 514 194 65 33.5 74.9 61.5
ADGC-Reitz (2013) 5896 1968 NA NA 80.5 63.9
Total African 7315 2263 96
Asian ancestry
HKOS 2373 349 66 18.9 60.1 67.9
Harmonization 385 153 49 32.0 73.6 55.0
Total Asian 2758 502 115
Hispanic ancestry
SALSA 1271 128 35 27.3 68.9 58.6
Total Stage 1 for ACD and VaD 477,950 46,902 4138
Replication of VaD results in EADB
EADB 275,745 NA 4,564
Total 753,695 46,902 8702
  • Note: Overall, 800,597 individuals were included in this study, accounting for 46,902 and 8702 cases of ACD and VaD, respectively. For UKBB, we used the proxy-AD (familial AD) for ACD analysis and assessed VaD cases using ICD10 codes (see Methods). We also used ADGC-NAJ-201118 and ADGC-Reitz-201319 for ACD in European and African ancestry, respectively, to avoid overlap with CHARGE samples. We subsequently replicated our VaD results in EADB.
  • Abbreviations: ACD, all-cause dementia; ADGC, Alzheimer's Disease Genetics Consortium; EADB, the European Alzheimer Disease Biobank; UKBB, the UK Biobank.

2.2 Phenotype definition

The primary study outcomes are ACD and VaD, measured by each participating cohort as described in Supplementary File 2. Briefly, to diagnose ACD and VaD, a neurological evaluation and diagnosis based on validated criteria were required. These criteria, as shown in Table S1 of Supplementary File 1, include the use of International Statistical Classification of Diseases and Related Health Problems (ICD) codes in most cohorts, as well as additional criteria such as the Diagnostic and Statistical Manual of Mental Disorders, Third to Fifth editions (DSM-III to V), National Institute of Neurological Disorders and Stroke-Alzheimer's Disease and Related Disorders Association, National Institute of Neurological Disorders and Stroke-the Association Internationale pour la Recherche et l'Enseignement en Neurosciences, and dementia by proxy for United Kingdom Biobank (UKBB). Additionally, VaD cases were included in ACD, and the proportion of ACD classified as VaD is reported in Table S1 of Supplementary File 1. Moreover, to increase the sensitivity, cohorts were asked to run separate association analyses for (a) incident ACD, (b) prevalent ACD, (c) incident VaD, (d) prevalent VaD, (e) incident probable and definite VaD, and (f) prevalent probable whenever possible. To address the overlap with AD, we included all VCID (including persons with possible VCID) and separately analyzed only cases of “pure” (probable and autopsy-proven definite) VCID and requested that all cohorts provide the most accurate, detailed description of their diagnostic algorithm. Although VaD in UKBB was defined based on ICD-10 codes, we used the family history of dementia GWAS (“imputed dementia”) recently published by Marioni et al.18 for ACD. Imputed dementia was defined as individuals at least 65 years old reporting a history of dementia in one or both parents. As explained in Ghosh et al.,19 the effect sizes and standard errors of the imputed dementia GWASs were doubled to analytically correct for the use of proxy phenotypes.

2.3 Genotyping and imputation

Genotyping was performed using cohort-specific genotyping arrays as described in Supplementary File 2. Genetic variants were imputed using 1000 Genomes Project (1KG), the Haplotype Reference Consortium (HRC),20 and the National Heart, Lung, and Blood Institute Trans-Omics for Precision Medicine (TOPMed). UKBB imputed the genotypes to HRC, 1KG, and UK10K. Details on study-specific quality control (QC) filters and software used for phasing and imputation are provided as supplementary materials (Supplementary File 2). Briefly, rare variants (minor allele frequency [MAF] < 1%) and poorly imputed variants (imputation quality, Rsq < 0.3) were excluded, as were variants mapping to sex chromosomes or mitochondria. Samples with poor genotyping call rate (<95%) and Hardy–Weinberg p values < 1 × 10−6 were removed. All genetic positions are reported in genome build 37 (GRCh37, hg19). Moreover, we used HRC version 1.2 as the main reference panel, and only variants in this panel were subsequently used in the association analyses. Additional details on the genotyping and imputation methods and QC are provided in Supplementary File 2.

2.4 Genome-wide association analyses, QC, and meta-analysis

2.4.1 Study-level association analyses

We conducted study and ethnicity-specific association analyses adjusting for age, sex, sites, and population structure to test the association of each variant with VaD and ACD. Cohorts were asked to run logistic regression and Cox proportional hazard models for prevalent and incident VaD/ACD, respectively, assuming additive allelic effects and imputed dosages. The UKBB association analyses were performed with linear mixed models (LMMs) using the BOLT-LMM software.21 BOLT-LMM has the advantage over other methods in that it accounts for cryptic relatedness and population structure and, thus, allows the inclusion of related individuals in models, which increases the overall sample size. Details on the methods and software used for study-level association analyses are provided in Supplementary File 2.

2.4.2 QC of study-level summary statistics

We performed a stringent QC check of the summary statistics from each cohort using EasyQC.22 We mapped each variant from the non-European ancestry (EA) cohort to the appropriate 1KG project phase 3 reference panel and all EA to HRC (details in Table S1). Then the following steps were performed to ensure proper QC of each file before the meta-analysis: (a) remove all structural variants and INDELs; (b) filter out variants with missing or unusual values (p value < 0 or > 1, effect size > 10, effect allele frequency < 0 or > 1, imputation quality < 0 or > 1); (c) filter out variants with effective allele count (EAC, 2 × minor allele frequency × N × imputation quality) < 10; (d) filter out variants with low imputation quality (eg, INFO scores reported by the imputation software); (e) filter out variants with MAF < 1%; (f) align variants to the main reference panel (HRC for EA, and ethnicity-specific 1KG for others); remove variants with absolute difference between its allele frequencies in the cohort and reference panel greater than 0.2. All variants were assigned a unique identifier as a combination of the chromosome, position, reference, and alternative alleles separated by semi-colons (CHR:POS:REF:ALT) to avoid issues with chromosomal positions mapping to multiple marker IDs. The foregoing steps were repeated until satisfactory results were obtained after visual inspection of the different diagnostic QC plots (AF, P-Z, Q-Q, and SE plots) generated by EasyQC as explained in Winkler et al.22

2.4.3 Meta-analysis of GWAS results

Ancestry-specific meta-analysis

The meta-analyses were conducted using the fixed-effect inverse variance-weighted method implemented in METAL.23 Post-analysis results were filtered to retain only variants present in more than 40% of the overall cohorts and the effective sample size greater than 40% of the study sample size. We evaluated the heterogeneity across cohorts using the I2 statistic provided by METAL, which represents the percentage of variation across studies that is due to heterogeneity rather than chance. We used the standard p value thresholds for GW significance, p < 5e-8, and suggestive p < 1e-6. Since there was no evidence of genomic inflation in the cohort summary statistics (lambda 0.98 to 1.06), no genomic control was applied during the meta-analysis. Genomic loci were defined as the region ±500 kb around the single nucleotide polymorphism (SNP) with the lowest p value, considered as the index SNP. We assessed the heterogeneity across studies using the I2 statistics of METAL (HetPVal output), which represents the percentage of variation across cohorts that is due to genetic heterogeneity rather than chance. Except for the APOE region (defined as SNPs located on chromosome 19 between positions 45,000,000 and 45,800,000 base pairs according to GRCh37 [hg19]), for which the HetPVal was >1e-8, significant SNPs were selected with HetPVal > 0.01. We conducted ancestry-specific meta-analyses of VaD and ACD for EA, African ancestry (AA), Asian ancestry (SA), and Hispanic ancestry (HA). In addition, we used linkage disequilibrium (LD) score regression to quantify the contribution of true polygenicity and biases such as cryptic relatedness and population stratification of the meta-analysis results.

Cross-ancestry meta-analysis

We performed cross-ancestry meta-analyses to assess whether the increase in sample size could lead to adequate power to identify additional GW significant loci associated with ACD and VaD. To this end, we used Meta-Regression of Multi-Ancestry Genetic Association (MR-MEGA) software, which has proven more efficient than others when dealing with genetic heterogeneity.24 MR-MEGA uses a matrix of mean pairwise allele frequency differences to quantify the genetic similarity between studies and estimate the effect of each SNP after adjusting for ancestry principal components. We applied study-specific filters, as previously described in the QC section, with EAC > 20, for studies with small sample sizes to reduce the amount of noise in the results-driven rare SNPs in small cohorts. We fitted three principal components, as suggested by MR-MEGA authors, which proved sufficient to separate the cohorts into self-reported ancestry groups (Figure S1). As in the ancestry-specific meta-analysis, we retained only SNPs that were present in over 40% of cohorts, with >40% total sample size. GW significant SNPs had p < 5e-8 and showed evidence of allelic heterogeneity across populations (MR-MEGA P-Het > 1e-5).

2.5 Shared genetic susceptibility with complex disease traits

A gene-based association test was conducted using MAGMA,25 with p < 2.8e−6 as a genome-wide significance threshold. Gene regions with SNPs not reaching GW significance for ACD or VaD in the primary GWAS analysis and additionally not in LD (r2 < 0.10) with the lead SNP were considered novel.

We first explored the association of lead risk variants with related vascular, neurological traits and metabolic traits, excluding the APOE region. For each related trait, association statistics of SNPs falling in a window of ±250 kb around each lead SNP were queried,26 and SNPs satisfying the GW significance threshold in the original study were retained. Leveraging the polygenicity of ACD (mean chi-squared = 1.1) and VaD (mean chi-squared = 1.06), we systematically explored the genetic overlap of ACD and VaD (in European-only analysis) with (i) neurological and neurodegenerative traits (any stroke [AS], IS, small vessel stroke [SVS], large artery stroke [LAS], cardioembolic stroke [CES], general cognitive function [GCF], and Alzheimer-type dementia [AD]); (ii) common magnetic resonance imaging (MRI) marker of cerebral SVD (WMHs)5; and (iii) vascular risk factors (systolic blood pressure [SBP], diastolic blood pressure [DBP], pulse pressure [PP], high-density lipoprotein [HDL], low-density lipoprotein [LDL]).27 We acquired summary statistics of the largest European-only GWAS for these traits.

Using LD score regression (LDSR) analysis,28 genetic correlation estimates between ACD/VaD and the aforementioned complex traits were obtained. A similar and potentially powerful approach called genetic covariance analyzer (GNOVA)29 was additionally used to study the shared genetic covariance across the genome between a given pair of complex traits. LDSR and GNOVA compute genetic correlation and covariance, respectively, while adjusting for potential sample overlap and accounting for the LD of genetic variants. Though LDSR and GNOVA are substantially similar, differences in the minor allele frequency thresholds may influence genetic correlation estimates and significance to some extent. A p value < 8.3e-3 correcting for six independent phenotypes was considered significant. Additionally, for the traits with significant genetic overlap, we performed causal inference analysis in the Mendelian randomization (MR) framework with ACD/VaD as the outcome. Using the MR-LAP method,30 we addressed potential bias in the causal effects due to sample overlap between the exposure and the outcome variables. Briefly, MR-LAP utilizes the LDSR intercept estimates – a measure of the degree of sample overlap, polygenic architecture, and the heritability of the genetic instruments of the exposures – to account for the sample overlap bias and other biases (weak instrument and winner's curse bias) that push the causal estimates toward the null.

Since GW correlation estimates may miss significant correlations at the regional level (balancing effect),31 a Bayesian pairwise GWAS approach (GWAS-PW) was applied.32 GWAS-PW identifies trait pairs with high posterior probability of association (PPA) with a shared genetic variant (Model 3, PPA3 ≥ 0.90). To ensure that PPA3 is unbiased by sample overlap, fgwas version 0.3.6 was run on each pair of traits, and the correlation estimated from regions with null association evidence (PPA < 0.20) was used as a correction factor.32 We then calculated Spearman's rank correlation for regions showing PPA3 > 0.90, approximating the direction of effect.

Finally, using a Bayesian method – ashR33 – we studied the effect-size distribution for ACD and VaD and related risk factors. Briefly, ashr tests the probability of non-zero effect conferred by SNPs as a function of LD score, measuring the true effect size that is not zero and the underlying polygenic background. Using MTAG,34 traits falling in similar polygenic profile to ACD or VaD are jointly analyzed in a bivariate scheme leveraging the pairwise trait genetic correlation to boost power to discover new loci. The significance threshold in the MTAG analysis is determined based on the number of traits sharing a similar polygenic profile and was additionally restricted to SNPs that also had nominal significance (p < 0.05) for each phenotype separately in the pre-existing univariate GWAS.

2.6 Transcriptome-wide association study and colocalization

We performed transcriptome-wide association studies (TWASs) using the association statistics from the ACD and VaD (European-only) and weights from 22 publicly available gene expression reference panels from blood (Netherlands Twin Registry [NTR], Young Finns Study [YFS]), arterial (genotype-tissue expression [GTEx]), brain (GTEx, CommonMind Consortium [CMC]), and peripheral nerve tissues (GTEx). For each gene in the reference panel, precomputed SNP-expression weights in the 1-Mb window were obtained, including the highly tissue-specific splicing quantitative trait loci (sQTLs) information on gene isoforms in the dorsolateral prefrontal cortex (DLPFC) derived from the CMC. TWAS-Fusion35 was used to estimate the TWAS z-score (association statistic between predicted expression and ACD or VaD), derived from the SNP-expression weights, SNP-trait effect estimates, and the SNP correlation matrix. Transcriptome-wide (TW) significant genes (eGenes) and the corresponding QTLs (expression QTLs [eQTLs]) were determined using Bonferroni correction in each reference panel, based on the average number of features (4235 genes) tested across all the reference panels.35 eGene regions with eQTLs not reaching GW significance in association with ACD or VaD and not in LD (r< 0.01) with the lead SNP for GW significant risk loci were considered novel. Finally, a colocalization analysis (COLOC)36 was carried out at each locus to estimate the posterior probability of a shared causal variant (PP4 ≥ 0.75) between the gene expression and trait association, using a prior probability of 1.1 × 10−5. Furthermore, functional validation of the eGenes was performed by testing for positional overlap of the best eQTLs from TWAS with enhancer (H3K4me1, H3K27ac) and/or promoter (H3K4me3/H3K9ac) elements across a broad category of relevant tissue types (blood [BLD], brain/neurological [BRN]) using Haploreg version 4.1.37

2.7 Identification of independent case–case loci with case–case GWAS

Leveraging summary statistics from our GWAS of ACD and VaD, as well as from publicly available existing GWASs of AD38 and stroke,39 we examined genetic uniqueness between these highly correlated though distinct disorders using case–case GWAS (CC-GWAS), a method that tests for differences in allele frequency between cases of two disorders without individual-level data.40 By allowing for sample overlap between the two case-control GWASs, CC-GWAS can increase the power to detect signals otherwise missed in case-control GWASs. We used a LD threshold of 0.2 (r2 < 0.2) to distinguish CC-GWAS-specific loci from genome-wide significant variants identified in the input case-control GWAS.

3 RESULTS

Our analysis included 800,597 individuals comprising 46,902 and 8702 cases of ACD and VaD, respectively. They were recruited from the 19 Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) cohorts, the Alzheimer's Disease Genetics Consortium (ADGC), the European Alzheimer Disease Biobank (EADB), and the UK Biobank (UKBB), encompassing four different reported ancestries: European (98.5%), African (1.0%), Asian (0.4%), and Hispanics/Latino (0.1%). Association analyses were performed in each cohort following a predefined analysis plan, using logistic regression and Cox proportional hazards models for prevalent and incident cases, respectively. We performed study-specific QC of the summary statistics data, followed by ancestry-specific meta-analyses and cross-ancestry meta-analyses of ACD and VaD, as described in the Methods section. For each cohort, a description, association analysis method, QC parameters, and cutoffs are provided in Supplementary File 2.

3.1 Meta-analyses of ACD and VaD GWAS in European ancestry populations replicated known AD loci

We conducted fixed-effects inverse variance-weighted meta-analyses of the 14 European ancestry cohorts (N = 466,606, NACD = 44,009, NVaD = 3892) from the CHARGE consortium, ADGC, and the UKBB. Furthermore, we replicated significant and suggestive signals from our VaD GWAS in the EADB consortium VCID data (N = 275,745, NVaD = 4564). The complete list of cohorts included in this study is provided in Table S1.

A total of 11,596,629 and 9,878,961 SNPs passed the study-level QC criteria and were tested for association with ACD and VaD, respectively. After post-meta-analysis QC, we identified 10 GW significant loci associated with ACD (GWS, p < 5 × 10−8), all of which had been previously associated with AD (Table 2, extended results in Table S2). Significant loci associated with ACD included signals in or around known AD genes such as APOE, BIN1, MS4A6A, PICALM, CR1, CD2AP, ABCA7, PILRB, SLC24A4, and ACE. For VaD, only one variant located near the APOE gene reached GW significance (Table 3, extended results in Table S3). The genomic inflation coefficients (lambda) were 1.05 and 1.07 for ACD and VaD, respectively. The lambda intercept computed with the LDSC software was 1.01 for both analyses, suggesting no systematic inflations of association statistics. The Manhattan and quantile-quantile (QQ) plots for both analyses are provided in Figures 1-3 (Forest and locusZoom plots for prominent signals are provided in Figures S2 to S65).

TABLE 2. Genome-wide significant (p < 5 × 10−8) and suggestive (p < 1 × 10−6) variants associated with all-cause dementia in European populations.
rsID Nearest gene CHR POS EA/NEA EAF BETA p value HetISq HetChiSq HetPVal
rs429358 APOE 19q13.32 45411941 C/T 0.1515 1.1728 6.87E-305 98.3 1104.478 2.534E-222
rs4663105 BIN1 2q14.3 127891427 C/A 0.4151 0.156 5.856E-34 63.4 51.892 0.00006868
rs12453 MS4A6A 11q12.2 59945745 C/T 0.3998 −0.0933 1.116E-16 13.2 23.037 0.287
rs10792832 PICALM 11q14.2 85867875 A/G 0.3585 −0.0857 8.234E-14 61.8 52.41 0.0000992
rs10948367 CD2AP 6p12.3 47585615 G/A 0.2712 0.0886 4.041E-13 4.2 20.88 0.4042
rs4295 ACE 17q23.3 61556298 C/G 0.3897 −0.0784 1.014E-11 10.8 22.43 0.3176
rs4844610 CR1 1q32.2 207802552 A/C 0.1846 0.0961 1.245E-11 71.4 66.468 3.519E-07
rs2906644 PILRB 7q22.1 99956290 G/C 0.1294 −0.1016 7.627E-09 0 8.219 0.9616
rs3764650 ABCA7 19p13.3 1046520 G/T 0.0957 0.1141 1.125E-08 35 24.604 0.07712
rs9323877 SLC24A4 14q32.12 92934269 G/A 0.2521 0.0699 3.804E-08 0 19.881 0.4654
rs1532278 CLU 8p21.1 27466315 T/C 0.3819 −0.0629 6.262E-08 66.2 59.185 9.516E-06
rs7912495 USP6NL 10p14 11718713 G/A 0.4612 0.0684 6.774E-08 0 15.343 0.7006
rs17125924 FERMT2 14q22.1 53391680 G/A 0.0928 0.1009 9.38E-08 2.6 19.508 0.4247
rs11767557 EPHA1 7q34 143109139 C/T 0.1998 −0.0742 9.721E-08 31.2 29.052 0.08675
rs1854554 SEMA4D 9q22.2 92155871 A/G 0.409 0.0598 1.096E-07 4 20.83 0.4072
rs7118826 ANO3 11p14.2 26195535 G/C 0.4585 0.0686 1.238E-07 18.7 23.368 0.2215
rs79832570 SPATC1 8q24.3 145097720 C/T 0.086 0.1433 1.385E-07 37.3 19.147 0.08504
rs13010870 RBM43 2q23.3 151765163 C/T 0.2162 −0.0732 1.649E-07 0 18.774 0.5366
rs1354106 CD33 19q13.41 51737991 G/T 0.3441 −0.0614 1.673E-07 6.4 21.373 0.3755
rs6014724 CASS4 20q13.2 54998544 G/A 0.0874 −0.1071 1.863E-07 0 18.603 0.4825
rs897150 TRIB1 8q24.13 126576702 A/G 0.3016 −0.0634 2.33E-07 4 20.842 0.4065
rs11168036 HBEGF 5q31.3 139707439 T/G 0.4933 0.0572 2.336E-07 0 18.21 0.5735
rs17269688 NCK2 2q12.2 106469267 G/A 0.025 −0.1937 2.389E-07 41.8 27.478 0.03647
rs677649 RNU6-11P 7 123439244 T/G 0.1756 0.0851 2.505E-07 9.9 19.987 0.3336
rs8081878 ZNF652 17q21.32 47436812 T/A 0.4634 0.057 2.609E-07 9.9 22.19 0.3303
rs4654450 RP1-37J18.2 1 4667378 G/A 0.3314 −0.0697 2.684E-07 16.2 22.679 0.2518
rs11218343 SORL1 11q24.1 121435587 C/T 0.0364 −0.1658 2.909E-07 20.7 21.432 0.2076
rs2297508 SREBF1 17p11.2 17715317 C/G 0.3659 −0.0585 4.604E-07 31.9 29.367 0.08078
rs442495 ADAM10 15q21.3 59022615 C/T 0.3245 −0.0599 4.722E-07 41 33.895 0.02684
rs13316744 AC091493.2 3 16742711 C/G 0.4544 −0.0549 7.609E-07 0 15.72 0.7338
rs7068231 ANK3 10q21.2 61784928 T/G 0.4005 −0.0579 7.833E-07 11.1 22.503 0.3139
rs834398 GABRB2 5q34 160528276 G/A 0.1971 0.0793 8.951E-07 0 11.885 0.8905
rs62013908 RBFOX1 16p13.3 5991314 G/C 0.2547 0.0724 9.422E-07 3.9 19.767 0.4087
  • Note: Genome-wide significant variants are highlighted in orange.
TABLE 3. Genome-wide significant (p < 5 × 10−8) and suggestive (p < 1 × 10−6) variants associated with vascular dementia in European populations.
rsID Nearest gene CHR POS EA/NEA EAF

BETA

CHARGE

p value

CHARGE

p value

EADB

p value

COMBINED

Direction
rs429358 APOE 19q13.32 45411941 C/T 0.1549 0.8794 2.67E-86 5.66E-113 2.9E-196 ++
rs11911 SPRY2 13q31.1 80910851 C/A 0.3756 −0.1466 2.60E-06 0.0653 0.0000335
rs7101996 GALNT18 11 11259298 T/C 0.4198 −0.1362 3.06E-06 0.48 0.00452 −+
rs2845990 LINC02113 5 98907502 C/T 0.352 −0.1383 3.24E-06 0.872 0.00131 −+
rs117904289 FOXA2 20p11.21 22782154 G/A 0.0858 0.2516 3.35E-06 0.987 0.00112 +−
rs838941 SCARB1 12q24.31 125183316 A/G 0.4261 0.1344 3.56E-06 0.33 0.0000654 ++
rs17418160 ERBB4 2q34 213119022 C/T 0.0396 0.3341 3.58E-06 0.882 0.00171 +−
rs77542509 TRPC6 11q22.1 101415824 C/T 0.058 −0.3038 4.77E-06 0.137 0.0000213
rs6127311 DOK5 20q13.2 53501017 C/T 0.0532 −0.3141 4.83E-06 0.294 0.0144 +−
rs35945091 LCN1P2 9 136185411 C/T 0.2227 0.164 5.15E-06 0.855 0.00176 +−
rs17059857 ZNF236 18q23 74469493 C/T 0.0403 0.358 5.41E-06 0.417 0.00934 +−
rs143750890 AJAP1 1p36.32 4602505 C/T 0.0273 0.4283 5.56E-06 0.163 0.0000334 ++
rs9510987 SPATA13 13 24575243 G/T 0.2525 0.1466 6.28E-06 0.291 0.0000744
rs55709546 PHACTR3 20q13.32 58261107 C/A 0.0484 0.3044 6.53E-06 0.645 0.000496 ++
rs12667855 TMEM106B 7p21.3 12124166 T/G 0.0981 0.225 6.59E-06 0.542 0.000299 ++
rs281219 SEMA6D 15q21.1 47711652 A/G 0.1954 0.1641 6.65E-06 0.316 0.0129 +−
rs138352554 GBP1 1p22.2 89517105 G/A 0.0359 0.3761 6.95E-06 0.573 0.000406 ++
rs16967121 RASGRP1 15q14 38923007 G/A 0.0658 0.2605 7.31E-06 0.916 0.00126 ++
rs11007123 WAC 10p12.1 28763005 C/T 0.2804 0.1395 7.70E-06 0.292 0.0139 +−
rs4794009 GIP 17q21.32 47051955 A/G 0.4412 0.1274 8.17E-06 0.797 0.000746 ++
rs35448830 PRKCE 2p21 46080762 C/T 0.0368 0.3346 8.48E-06 0.631 0.000579 ++
rs2233754 PSMA3 14q23.1 58755574 C/A 0.07 0.2575 9.39E-06 0.601 0.00604 +−
  • Note: The meta-analysis includes 11 cohorts from the CHARGE consortium and the UK Biobank (UKBB) GWAS. Direction denotes the direction of association in CHARGE and EADB. The genome-wide significant variant is highlighted in orange.
  • Abbreviations: ADGC, Alzheimer's Disease Genetics Consortium; CHARGE, Cohorts for Heart and Aging Research in Genomic Epidemiology; EADB, European Alzheimer Disease Biobank; UKBB, UK Biobank.
Details are in the caption following the image
Manhattan plot of ACD GWAS. In addition to variants in APOE region, we identified five new genetic loci associated with VaD. Blue and red lines correspond to p value of 5e−7 and 5e−8 for genome-wide suggestive and significant SNPs, respectively. Manhattan plots for the cross-ancestry meta-analysis. Each dot represents a SNP, the x-axis shows the chromosomes where each SNP is located, and the y-axis shows −log10 p value of the association of each SNP with ACD in the cross-ancestry meta-analysis. The red horizontal line shows the genome-wide significant threshold (p value = 5e-8; −log10 p value = 7.30). The nearest gene to the most significant SNP in each locus has been labeled.
Details are in the caption following the image
Manhattan plot of VaD GWAS. In addition to variants in the APOE region, we identified five new genetic loci associated with VaD. Blue and red lines correspond to a p value of 5e−7 and 5e−8 for genome-wide suggestive and significant SNPs, respectively. Manhattan plots for cross-ancestry meta-analysis. Each dot represents a SNP, the x-axis shows the chromosomes where each SNP is located, and the y-axis shows the −log10 p value of the association of each SNP with VaD in the cross-ancestry meta-analysis. The red horizontal line shows the genome-wide significant threshold (p value = 5e-8; −log10 p value = 7.30). The gene closest to the most significant SNP in each locus has been labeled.
Details are in the caption following the image
Q-Q plots of ACD (left) and VaD (right) GWASs. The expected p values (x-axis) are plotted against the observed p values (y-axis). The units of the axes are the −log10 of the p value. The red and blue curves represent the plots with MAF ≥ 0.05 and 0.01, respectively. The diagonal line of the null hypothesis and its 95% confidence interval are plotted in gray based on the p values without the previously reported SNPs. The red dotted line represents the cutoff for genome-wide significance. MAF, minor allele frequency.

For the VaD trait, we selected all variants with a p value less than 1 × 10−5 and meta-analyzed with EADB summary results using a weighted sum of z-scores approach. Only one variant near the APOE gene was statistically significant and had the same direction of effect in both studies (Table 3).

The meta-analyses of ACD and VaD GWAS in African, Asian, and Hispanic/Latino ancestries did not provide new GW significant variants.

We replicated our VaD signals using EADB data (Table 3). Overall, we replicated an association within the APOE region. The suggestive variant near SPRY2 also has the lowest p value in the EADB GWAS with the same direction of effect in both studies.

3.2 Cross-ancestry meta-analysis of ACD and VaD GWAS

Next, we performed a cross-ancestry meta-analysis using MR-MEGA, first to assess whether the increased sample size could lead to the identification of additional loci associated with ACD and VaD and to identify loci that are relevant in other ancestries. Most of the cross-ancestry meta-analyses included individuals of European ancestry and smaller samples from African, Asian, and Hispanic/Latino ancestries. The total number of variants included was 17,054,226 and 11,595,061 for ACD and VaD, respectively. The Manhattan plots of the SNP-wide meta-analyses for both traits are provided in Figures S66 and S67. Significant and suggestive signals for ACD and VaD are presented in Tables 4 and 5, and the extended results are in Tables S4 and S5. We identified novel signals reaching GW significance at 20q11.21 (CHD6, an oxidative DNA damage response factor previously associated with neurological phenotype),41 2q14.1 (DAW1, involved in cerebrospinal fluid circulation and cilia motility during development),42 and 15q15.1 (PWRN2, previously associated with tauopathy and Prader–Willi syndrome)43 for ACD and 17q21.1 (MARCHF10) for VaD.

TABLE 4. Genome-wide significant (p < 5 × 10−8) variants associated with all-cause dementia in cross-ancestry meta-analysis.
rsID Nearest gene Chr Pos EA/NEA P value MAF Beta SE
rs10402524 BCAM 19p11 45329344 T/C 1.21E-17 0.2336 −0.168 0.045
rs744373 BIN1 2q14.3 127894615 A/G 1.90E-17 0.358 −0.139 0.031
rs2278867 MS4A6A 11q13.1 59943109 A/T 1.72E-15 0.2897 0.113 0.020
rs10792832 PICALM 11q13.1 85867875 A/G 3.77E-12 0.3135 −0.074 0.036
rs10948367 CD2AP 6q14.3 47585615 A/G 1.67E-11 0.2328 −0.042 0.017
rs1408077 CR1 1q11.1 207804141 A/C 4.75E-10 0.1412 0.088 0.055
rs4295 ACE 17q21.1 61556298 C/G 1.60E-09 0.3666 −0.066 0.018
rs2208524 CHD6 20q11.21 40423299 T/C 1.66E-09 0.1268 −0.103 0.027
rs11691153 DAW1 2q14.1 228780072 T/C 1.83E-09 0.1536 0.099 0.025
rs6853262 LPHN3 4q22.1 61221892 C/T 6.22E-09 0.06989 0.208 0.112
rs2677386 PWRN2 15q15.1 24432053 T/C 7.18E-09 0.3612 −0.083 0.017
rs7006786 ARHGEF10 8q13.2 1792639 G/A 8.81E-09 0.08986 0.097 0.045
rs35483531 DEGS2 14q21.3 100653772 C/T 1.35E-08 0.2993 −0.004 0.024
rs170084 PMFBP1 16q11.2 72178483 T/A 2.79E-08 0.107 −0.068 0.029
rs10940421 SNX18 5q14.3 54036059 A/G 3.34E-08 0.372 0.040 0.017
rs138908633 EPB41L4A 5q14.3 111649017 G/A 3.76E-08 0.03095 −0.029 0.050
rs74435987 DUSP6 12q14.1 89152253 G/T 4.20E-08 0.08766 0.078 0.130
rs11225924 DDI1 11q13.1 103493165 C/T 4.27E-08 0.1034 0.152 0.105
rs113747850 MAPK9 5q14.3 179710663 T/C 4.93E-08 0.123 0.071 0.024
  • Note: The meta-analysis includes European, African, Asian, and Hispanic/Latino ancestries. Three new variants at 20q11.21, 2q14.1, and 15q15.1 reached genome-wide significance (highlighted in orange).
  • Abbreviation: MAF, minor allele frequency.
TABLE 5. Genome-wide significant (p < 5 × 10−8) and suggestive (p < 1 × 10−6) variants associated with vascular dementia in cross-ancestry meta-analysis.
rsID Nearest gene Chr Pos EA/NEA p value MAF Beta SE
rs10119 TOMM40 19 45406673 G/A 1.21E-17 0.2476 −0.327 0.054
rs4380108 MARCHF10 17 60893485 C/T 9.59E-09 0.3127 −0.172 0.031
rs55747619 ITSN2 2 24530447 C/G 8.05E-08 0.08706 −0.384 1.969
rs9379092 CAGE1 6 7344531 G/A 9.80E-08 0.1172 −0.336 0.077
rs3757193 RPS6KA2 6 166923463 C/T 1.09E-07 0.08347 2.151 0.636
rs3871399 CMTM7 3 32496413 C/G 2.61E-07 0.124 0.550 0.640
rs17315346 BRINP2 1 177282235 C/T 2.67E-07 0.01538 −2.412 5.17
rs1738249 DNAH8 6 38753960 C/T 2.86E-07 0.3013 −0.050 0.040
rs12095469 OSBPL9 1 52206082 G/A 3.60E-07 0.05292 3.654 3.513
rs4820650 ADRBK2 22 25925358 T/C 3.82E-07 0.2468 0.050 0.054
rs61859886 MGMT 10 131353192 T/G 4.45E-07 0.1528 −0.274 0.057
rs9857196 RYK 3 133830660 T/A 5.09E-07 0.01997 3.438 2.670
rs637924 PCDH7 4 31465610 T/C 6.77E-07 0.2564 −0.056 0.051
rs35810115 ZNF675 19 23780763 C/T 6.81E-07 0.04992 −1.169 2.339
rs115331896 CRBN 3 3204942 T/G 6.95E-07 0.01218 3.139 2.706
rs4401880 SLC18A1 8 19946066 C/T 7.20E-07 0.3249 0.019 0.051
rs4823298 FBLN1 22 45915987 T/C 7.96E-07 0.4581 0.029 0.0439
rs17335455 NXPH1 7 8853946 T/G 8.20E-07 0.1633 −0.096 0.042
rs517484 RP11-6N13.1 5 104490130 T/C 8.51E-07 0.1965 −0.120 0.046
rs12814413 RBMS2 12 56916614 T/C 8.77E-07 0.3514 0.050 0.051
rs4665372 CGREF1 2 27325837 T/A 9.19E-07 0.3948 −0.104 0.043
  • Note: The meta-analysis includes European, African, Asian, and Hispanic/Latino ancestries. Genome-wide significant variants are highlighted in orange.
  • Abbreviation: MAF, minor allele frequency.

3.3 Functional characterization of GW suggestive signals for ACD and VaD meta-analyses

3.3.1 Shared genetic susceptibility with complex disease traits

The substantial shared genetic susceptibility of ACD and VaD with risk factors and complex disease traits is evident across different genomic scales (single variant, regional, and the global level). ACD exhibits genetic pleiotropy with vascular risk factors (hypertension, WMH burden), hematological traits (neutrophil, lymphocyte count), and blood-based biomarkers indicative of inflammation (C-reactive protein levels), hemostasis (fibrinogen, factor-VII levels), and neurodegeneration (soluble TREM2 levels) (Table S6). This shared genetic susceptibility is primarily driven by the MS4A gene family (membrane-spanning 4A; MS4A6A, MS4A4A). The sharing of common genetic variation between ACD and vascular risk factors (blood pressure traits [DBP, SBP, PP], and T2D) at the ACE and PILRB locus (Figure S68, Table S6) is further supported by our regional Bayesian pairwise (GWAS-PW) analysis highlighting the high probability of harboring a shared causal variant (Table S7). Interestingly, the GWAS-PW approach additionally reveals the shared genetic susceptibility of VaD with IS and WMH at the PRPF8 and PRDM6 locus. In support, global-level genetic overlap analysis (excluding the APOE region) using GNOVA showed statistically robust evidence for the association of increased levels of WMH with increased risk of VaD (Table S26, Figure 4). Additionally, we observed an inverse association of high levels of HDL (protective) with ACD risk and high levels of DBP and LDL with VaD risk. As expected, a strong genetic correlation between poorer cognitive performance (GCF) and ACD was also observed. Our causal inference analysis, using MR-LAP, confirmed the putative causal association of increased DBP and WMH levels with VaD risk. However, the genetic correlation between AD and related risk factors using Kunkle 2019 GWAS did not show this causal association (Tables S26–S27, Figure 4).

Details are in the caption following the image
Shared genetic contribution between ACD/VaD and related risk factors. Contributions determined by LD score regression analysis (LDSR) (top), and Genetic Covariation Analyzer (GNOVA) (bottom). Effect sizes (rg) and significance levels (logp) are represented by color and symbol size. AD, Alzheimer's disease; GCF, general cognitive function; all stroke (AS) and its subtypes (ischemic, IS; cardioembolic, CES; small vessel, SVS; large artery, LAS); WMH, white matter hyperintensity burden; DBP, diastolic blood pressure; SBP, systolic blood pressure; PP, pulse pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein. Diamond shape: non-significant. Cross-significant causal effect estimates from MR-LAP analysis.

3.4 Polygenicity and multi-trait analysis

To identify additional SNPs conferring susceptibility to ACD and acting through related risk factors, we jointly studied the GW distribution of genetic effects for ACD and its closely related traits. We first prioritized those traits that have a polygenic background similar to ACD using a Bayesian approach (ashR). The ashR analysis showed that certain traits (ACD, stroke and its subtypes, WMH, coronary artery disease [CAD]) had specific, possibly overlapping, pathophysiological processes compared to other ACD risk factors (SBP, smoking [SMK], body mass index) that involved multiple biological pathways (Figure S69, Table S9). Next, using multitrait GWAS analysis (MTAG, see Methods) on ACD and the prioritized traits (CAD, stroke, WMH), we identified intronic SNPs in SMG6 and ABCG8 to be GW significant (pMTAG < 1.67E-08, for three phenotypes) for ACD (Table S10). Interestingly, SMG6 may also have a role in tau biology.44 Finally, we explored the genetic difference between ACD/VaD and related disorders using case–case GWAS (CC-GWAS, see Methods). Specifically, we compared (1) ACD/VaD with AD and (2) ACD with stroke. VaD was not compared with stroke because the two disorders are highly correlated. Here we report signals that were not GW significant in both respective case-control GWASs. For AD, we identified two loci associated with ACD-AD status, including the known APOE region and the IQUB gene (p < 2e-08) on chr7 (Table S11). No GW significant loci were associated with VaD-AD status, although we observed some suggestive association (Table S12). For ACD-stroke, we identified 56 variants mapping to 10 genes on chromosomes 17, 8, 11, 15, 4, and 12, most of which are located at the SREBF1/TOM1L2 locus (p < 1e-10) on chr17 (Table S13).

3.5 Functional prioritization using molecular profile (gene expression)

To functionally characterize and prioritize individual ACD and VaD genomic risk loci, we performed TWASs using TWAS-Fusion, ACD, and VaD association statistics and weights from 23 gene-expression reference panels from blood, arterial, and brain tissues (see Methods). We identified 29 trait-associated (ACD/VaD) SNPs functioning as eQTLs, regulating the expression of 22 genes (eGenes) in disease-relevant tissue types (Table S14). To explore whether the observed associations are real or merely reflect the random overlap between eQTLs and non-causal risk variants for the dementia traits, a colocalization analysis was performed at each significant locus estimating the posterior probability of a shared causal variant (PP4 ≥ 75%) between the gene expression and trait association. Overall, 30% of the eQTL-eGene satisfied the colocalization threshold for a shared causal variant between the ACD or VaD and gene expression. In addition to fine mapping functional genes (RP11-385F7.1, CR1, MS4A6A, ACE, APOC4) in the loci exhibiting GW association with ACD/VaD, the TWASs identified putative novel (CLU-ACD, PIKFYVE-VaD, SH3D21-ACD) genes satisfying transcriptome-wide significance threshold (pTWAS < 1.18E-05) and the colocalization probability threshold. Most (91%) of the eGenes are supported by the positional overlap of corresponding eQTLs with regulatory marks (enhancer and promoter binding sites) for active transcription in relevant tissue types.

3.6 Protein–protein interaction (PPI) evidenced SEMA4D, RBFOX1, and SPRY2 as hub genes for ACD and VaD

To determine the functional interactome of genes near genome-wide significant (excluding APOE region) and suggestive loci (p < 1e-6) associated with ACD and VaD, we performed a PPI analysis using the STRING database. The analysis comprised 82 ACD and 21 VaD GW significant and suggestive genes that were successfully mapped to the human genome. Evidence of interaction between proteins was based on “experiments,” “co-occurrence,” “database,” and “co-expression,” with a minimum score of 0.15. Non-connected proteins were removed from the network. To further determine how suggestive genes will fit in the network of known AD genes, we used kmeans to cluster the proteins based on validated interaction. ACD genes formed two main clusters (Figure S70). The first cluster was enriched in known AD genes, including BIN1, CLU, ABCA7, and CR1, but also suggestive genes, including SEMA4D, CHD18, and APH1B, with more than two types of connection evidence. RBFOX1 appears to be a major hub gene for the second cluster, which includes other suggestive genes like AJAP1, ANO3, and TRIB1. RBFOX1 and SEMA4D strongly (>2 evidence of connection) interact with known AD genes, suggesting their potential role in ACD. The PPI network of VaD (Figure S71) genes highlights the potential role of SPRY2 as it functionally connects other genes, including ERBB4, RASGRP1, and FOXA2.

3.7 Pathway and functional enrichment analysis

We conducted several analyses (pathways, gene ontology, disease enrichment) to obtain functional and biological contexts of genes (near variants with p < 1e-6, excluding the APOE region) associated with ACD and VaD.

3.7.1 Pathway analysis

Pathway analyses (Tables S15 and S16) revealed enrichment in several pathways, including “SREBF and miR33 in cholesterol and lipid homeostasis,” “Hypertrophy model,” and “Cholesterol metabolism with Bloch and Kandutsch-Russell pathways” for ACD.

3.7.2 Gene Ontology (GO) analysis

GO analysis for ACD (Figure S72 and Table S17) focusing on the biological processes (GO-BP) were enriched in terms related to amyloid-beta, “amyloid-beta metabolic process,” “amyloid precursor protein catabolic process,” and “negative regulation of amyloid precursor protein catabolic process” for ACD. For VaD (Figure S73 and Table S18), GO-BP analysis was enriched in several terms, including “response to glucose,” “response to hexose,” “response to monosaccharide,” “mesenchymal cell differentiation,” and “response to carbohydrate.”

3.7.3 Disease enrichment and association analysis

(Figures S74 and S75, Tables S19–S25) revealed that ACD genes were previously connected to AD, tauopathy, nephritis, and central nervous system disease. It also highlighted previous associations of SEMA4D and RBFOX1 with diseases of the central nervous system. Besides the AD connection, VaD genes were previously related to cancer, diabetes, and colorectal carcinoma. Finally, we used Framingham Heart Study data to estimate the heritability of VaD and the genetic correlation with ACD. We found the heritability of VaD to be 6.1%, with a 95% confidence interval of [3.2%, 21%]. The genetic correlation of VaD and ACD was 0.48 (SE = 0.84).

4 DISCUSSION

Our findings expand the current knowledge base of dementia genetics by focusing on both ACD and VaD. Our GWAS of ACD replicated several genes previously associated with AD, and GWAS of VaD identified SNPs in the APOE region. Using functional PPI and TW analyses, we identified novel genes underlying ACD that have been implicated in recovery from vascular injury and in neurotrophin signaling. On the basis of LD score regression analysis, we suggest that certain vascular risk factors may not have a causal role not in both ACD and VaD pathogenesis.

In our ACD analysis of European ancestry, we identified 10 GW significant loci, including APOE, BIN1, MS4A6A, PICALM, CR1, CD2AP, ABCA7, PILRB, SLC24A4, and ACE, all of which have been linked with AD risk in prior studies.45 In addition, our analyses highlighted 24 suggestive risk loci, of which 13 are novel. Among them are variants located near ANO3, a gene that encodes anoctamin-3, a transmembrane protein that belongs to a family of calcium-activated chloride channels and is implicated in focal dystonia, particularly craniocervical.46 Another suggestive locus was located near SEMA4D, a gene that encodes Semaphorin 4D and is known to modulate various processes related to neuroinflammation and neurodegeneration, including the initiation of inflammatory microglial activation.47 Indeed, SEMA4D is critical in regulating the transition between homeostatic and reactive states of various types of glial cells. Antibody blockade of SEMA4D is being explored as a potential disease-modifying strategy to slow cognitive decline in patients with early Huntington's disease48 and may be beneficial in other ACD. We have also identified a prominent signal near RBFOX1, a gene that encodes the RNA binding fox-1, which has been shown to have a role in alternative splicing of the amyloid precursor protein. Genetic variation in this gene has been associated with brain amyloid burden in preclinical and early AD and with the risk of clinical AD in African Americans.49 This gene may also impact dementia risk through non-amyloidogenic pathways as it additionally regulates neuron development and neuronal excitability, including brain-derived neurotrophic factor (BDNF)-dependent long-term potentiation in the hippocampus and has been implicated in brain development, essential tremor, and schizophrenia.50

Other suggestive loci are located in the ZNF652 gene, a transcriptional repressor involved in nucleic acid binding that has diverse effects, including determining the risk of hypertension. Hypertension is the most important risk factor for stroke and WMH and may be the most important modifiable risk factor for population prevention of dementia.51 We additionally identified a variant near Heparin Binding EGF like growth factor (HBEGF), a growth factor implicated in the pathobiology of cerebral autosomal dominant arteriopathy with sub-cortical infarcts and leukoencephalopathy (CADASIL),52 the major Mendelian prototype of VaD. HBEGF also has an effect on angiogenesis, expression of vascular endothelial growth factor A (VEGF-A), inflammation, and oxidative stress and has been implicated in hydrocephalus.53

APOE was strongly associated with both ACD and VaD in our meta-analysis. While AD could drive the association of APOE with ACD, the relationship with VaD is less established but has been demonstrated in some population studies and candidate-gene analyses54 and in a recent GWAS among the GR@ACE project participants.16 The link of APOE with VaD is in line with recent literature suggesting that the pathogenesis of APOE extends beyond Aβ peptide aggregation and clearance.55 Indeed, APOE also influences microglia and the blood-brain barrier (BBB)56 and is associated with intracranial atherosclerosis,57 WMH burden, and the presence of cerebral microbleeds,58 as well as with cerebral hypertensive angiopathy, which is common in individuals with VaD.59

In addition to a significant association of APOE with VaD in our sample, we identified several suggestive variants also associated with VaD. These include variants near the SPPRY2 protein-coding gene as well as GALNT8, FOXA1, ERBB4, PSMA3, and SEMA6D with consistency across samples in the direction of effect and many SNPs in LD with the lead SNP. Our downstream analyses supported a highly plausible causal link between variants, including SEMA4D, HBEGF, PIKFYVE, and RBFOX1 with ACD and SPRY2 with VaD. These genes collectively emphasize a possible role for novel pathological mechanisms in ACD and VaD. Our findings highlight a crucial mechanism underlying ACD: recovery after vascular injury. For example, SEMA4D, a member of the semaphorin family, is upregulated in the neurovascular unit after IS, where it exerts multiple neuroprotective effects.60 Moreover, this gene has been additionally highlighted in our PPI analysis as strongly associated with known AD genes. Another example is SPRY2, highlighted in our study as a suggestive gene for VaD, with strong functional associations with known AD and related dementias genes.

In the replication analysis of VaD signals in the EADB dataset, SPRY2 has the lowest p value and a consistent direction of association. This gene has also been suggested as a possible pharmacological target for stroke patients, as it promotes angiogenesis and glial scarring around the ischemic injury, preventing an increase in lesion size and secondary damage to brain tissue.61 Also, SPRY2 may exert neuroprotective effects as its expression regulates BDNF-induced signaling pathways.62 Similarly, PIKFYVE is an essential regulator of platelet lysosome homeostasis, which in turn may promote recovery after IS.63 Another hub gene in our analyses is RBFOX1, which, in addition to having a role in amyloid accumulation as discussed earlier, mediates ischemic damage by enhancing neuronal survival and BBB integrity after stroke.64 This gene is a neuron-specific splicing factor implicated in intellectual disability, epilepsy, autism, and Parkinson's disease. Its downregulation has been associated with destabilizing mRNAs encoding for synaptic transmission proteins, which may contribute to the loss of synaptic function in AD.65 Furthermore, RBFOX1 upregulation was shown to influence neuronal expression levels of the BDNF receptor, TrkB, which in turn may affect the risk for ACD.66

We found that the MS4A gene cluster drove genetic pleiotropy that involves vascular risk factors, inflammation, hemostasis, and soluble TREM2 levels. These findings align with preclinical studies67 and emphasize the critical role and multifactorial contribution of this gene cluster to ACD pathogenesis. Although previous literature pointed to an association of ACE with AD but not VaD,68 we herein show that this gene underlies both ACD and vascular risk factors. A recent study supports this finding by showing that overexpression of ACE on macrophages reduces vascular amyloid and GFAP+ astroglial reactivation, indicating its role in the protection of the neurovascular unit.69 Moreover, our pairwise analysis highlighted a locus at the PRDM6 that explained a shared genetic susceptibility of VaD with IS and WMH. Low levels of leukocyte DNA methylation of the PRDM6 gene have been associated with an increased risk of IS and worse outcomes 3 months after an IS.70 Moreover, PRDM6 acts as an epigenetic regulator of vascular smooth muscle cell plasticity.71

Despite evidence showing an inverse relationship between plasma HDL levels and risk of incident AD, results are conflicting, with some studies pointing to higher dementia risk in individuals with high HDL levels, as was also the case in our study.72 It should be acknowledged that HDL represents a class of lipoproteins that are heterogeneous in structure and function, which is not reflected by a simple measurement of HDL plasma levels. High HDL levels can be deleterious under certain conditions.73 Vascular risk factors and the presence of cardiovascular disease can alter HDL functionality by changing the structure of HDLs and converting them into pro-inflammatory, pro-oxidant, prothrombotic, and proapoptotic compounds. Our observation aligns with recent Mendelian randomization data implicating an elevated HDL in risk of AD.74

The following limitations should be considered when interpreting the results of this study. First, the multifactorial nature and heterogeneous clinical manifestations of ACD and VaD have led to various attempts to develop diagnostic criteria, which were differentially applied across the participating cohorts. ACD has been ascertained using DSM-IV in some studies. In contrast, others have used ICD-9/10 codes alone or in combination with autopsy or death certificate information, which can result in a varying proportion of persons identified as having dementia. The various cohorts also used different diagnostic criteria to define VaD. In all cohorts, a key requirement for VaD diagnosis remains the demonstration of a cognitive deficit and the presence of cerebrovascular disease, consistent with the most recent consensus criteria for VCID.75 Whereas these criteria differ in sensitivity and specificity, thereby introducing statistical noise, this heterogeneity does not diminish the importance of the loci identified despite the constraints. A second limitation is the limited power to identify associations with VaD in ancestries other than European.

Our study identified several putative genetic variants and biological pathways associated with ACD and VaD and added additional support for the involvement of vascular mechanisms in dementia pathogenesis.

ACKNOWLEDGMENTS

We thank the many study participants, researchers, and staff for collecting and contributing to the data. This project was conducted within the neurology working group of the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium. The CHARGE cohorts are supported in part by the National Heart, Lung, and Blood Institute (NHLBI) infrastructure grants R01HL105756 (Psaty), RC2HL102419 (Boerwinkle) and the neurology working group is supported by the National Institute on Aging (NIA) R01 grant AG033193. Additional funding sources include the UT Health San Antonio Center for Biomedical Neuroscience (CBN) and grants from the NIA (AG059421, AG054076, AG049607, AG033090, AG066524, P30 AG066546, 5P30AG059305-03, RF1 AG061729A1, 5U01AG052409-04) and NINDS (NS017950, UF1NS125513, K01NS126489). Funding sources for each cohort are listed in Supplementary File 2.

    CONFLICT OF INTEREST STATEMENT

    Agustin Ruiz and Itziar de Rojas acknowledge research support from Grifols SA (Spain), Fundacion Bacaria LaCaixa (Spain), Instituto de Salud Carlos III Ministry of Health (Spain), Roche, and Janssen. Agustin Ruiz received consulting fees and honoraria from Landsteiner Genmed SL, Grifols SA, and Janssen; support for attending meetings from Grifols SA; and stock options from Landsteiner Genmed SL. All authors report no conflicts of interest. Additional author disclosures are available in the supporting information (Supplementary File 2).

    CONSENT STATEMENT

    All participants provided written informed permission, or, for those with substantial cognitive impairment, consent was provided by a caregiver, legal guardian, or other proxy. Author disclosures are available in the supporting information.

    SOFTWARE AVAILABILITY

    CC-GWAS: https://github.com/wouterpeyrot/CCGWAS

    Gene expression weights for TWAS: http://gusevlab.org/projects/fusion/

    HESS: https://huwenboshi.github.io/hess/local_hsqg/

    LDSR: https://github.com/bulik/ldsc

    GWAS-PW: https://github.com/joepickrell/gwas-pw

    Radial-MR: https://github.com/WSpiller/RadialMR

    GREP: https://github.com/saorisakaue/GREP

    EPIGWAS: https://immunogenomics.hms.harvard.edu/code

    Magma.Celltyping: https://github.com/NathanSkene/MAGMA_Celltyping

    MR-MEGA: https://www.geenivaramu.ee/en/tools/mr-mega

    DEFINITIONS

    GWAS: Genome-Wide Association Analysis Study

    ACD: All-cause dementia

    VaD: Vascular dementia

    VCID: Vascular Cognitive Impairment and Dementia

    CHARGE: Cohorts for Heart and Aging Research in Genomic Epidemiology

    ADGC: Alzheimer's Disease Genetics Consortium

    UKBB: UK Biobank

    EADB: European Alzheimer Disease DNA BioBank

    APPENDIX

    1 APPENDIX COLLABORATORS

    Bernard Fongang1,2,3,#,*, Muralidharan Sargurupremraj1,3,*, Xueqiu Jian1,3, Aniket Mishra4, Vincent Damotte5, Itziar de Rojas6,7, Olivia Skrobot8, Joshua C. Bis9, Kang-Hsien Fan10, Erin Jacobsen11, Gloria Hoi-Yee Li12, Jingyun Yang13, Bizzarro Alessandra14, Lauria Alessandra14, Saima Hilal15,16, Joyce Ruifen Chong15, Yuek Ling Chai15, M. J. Knol17, Maria Pina Concas18, Girotto Giorgia18,19, Moeen Riaz20, Chenglong Yu20, Alexander Guojonsson21, Paul Lacaze20, Adam C Naj22, Monica Gireud-Goss1, Yannick N. Wadop1, Aicha Soumare4, Vincent Bouteloup4,23, Vilmundur Gudnason21,24, Petronilla Battista25, Aurora Santin19, Beatrice Spedicati19, Rodolfo Sardone26,27, Lenore Launer28, Jan Bressler29, Rebecca F Gottesman30, Quentin Le Grand31, Ilana Caro31, Gennady V. Roshchupkin32,33, Hampton L. Leonard34,35,36, Chaojie Yang37,38, Traci M. Bartz39,40, Constance Bordes31, Paul M. Ridker41,42, Mirjam I. Geerlings43,44,45,46, Natalie C. Gasca40, Ani Manichaikul37, Mike A. Nalls34,35,36, Stephen S. Rich37, Carsten O. Schmidt47, Stella Trompet48,49, Jessica van Setten50, Marion van Vugt50, Hans J. Grabe51,52, J Wouter Jukema49,53,54, Ina L. Rissanen55, Sylvia Wassertheil-Smoller56, M. Arfan Ikram32, Eleanor M. Simonsick57, W T. Longstreth58,59, Daniel I. Chasman41,42, Jerome I. Rotter60, Naveed Sattar61, David J Stott62, Eric J Shiroma63, Sigurdur Sigurdsson24, Mohsen Ghanbari32, Ulf Schminke64, Eric Boerwinkle29,65, Hugo J Aparicio66,67, Alexa S Beiser66,68, Jose R Romero66,67, Vasileios Lioutas66,69, Ruiqi Wang66,68, Chloe Sarnowski70,71, Alexander Teumer51,72, Uwe Völker72,73, Thomas H. Mosley74, Marta Marquié6,7, Pablo García-González6,7, Clàudia Olivé6, Raquel Puerta6, Amanda Cano6,7, Oscar Sotolongo-Grau6,7, Sergi Valero6,7, Vanesa Veronica Pytel6, Maitée Rosende-Roca6,7, Montserrat Alegret6,7, Lluís Tàrraga6,7, Mercè Boada6,7, Ángel Carracedo75,76, Emilio Franco-Macías7,77, Gerard Piñol-Ripoll78,79, Guillermo Garcia-Ribas7,80,81,82, Jordi Pérez-Tur7,82, Jose Luís Royo83, Jose María García-Alberca84, Luis Miguel Real85,86, María Eugenia Sáez87, María J. Bullido7,88,89,90, Miguel Calero7,91,92, Miguel Medina7,93, Pablo Mir7,94,95, Pascual Sánchez-Juan7,96, Pau Pastor97,98, Victoria Álvarez99,100, Benjamin Grenier-Boley5, Fahri Küçükali101,102,103, Sven Van der Lee104,105,106, Oliver Peters107,108, Anja Schneider109,110, Martin Dichgans111,112,113, Dan Rujescu114, Jürgen Deckert115, Emrah Düzel116,117, Jens Wiltfang118,119,120, Michael Wagner121,122, Timo Grimmer123, Nikolaos Scarmeas124,125, Fermin Moreno7,126,127, Raquel Sánchez-Valle128, Luis M Real85,129, Eloy Rodriguez-Rodriguez7,130, Adolfo Lopez de Munain7,126,131, Alexandre de Mendonça132, Jakub Hort133,134, Caroline Graff135, Goran Papenberg136, Vilmantas Giedraitis137, Børge G. Nordestgaard138,139, Hilkka Soininen140, Miia Kivipelto141,142,143,144,145, Annakaisa Haapasalo146, Gael Nicolas147, Florence Pasquier148, Olivier Hanon149, Edna Grünblatt150,151,152, Daniela Galimberti153,154, Beatrice Arosio155,156, Patrizia Mecocci157, Alessio Squassina158, Lucio Tremolizzo159, Innocenzo Rainero160, Davide Seripa161, Julie Williams162, Philippe Amouyel163, Frank Jessen109,164,165, Tsolaki Magda166, Ruth Frikke-Schmidt167,168, Kristel Sleegers101,102,169, Sebastiaan Engelborghs170,171, Rik Vandenberghe172,173, Martin Ingelsson174,175,176, Giacomina Rossi177, Mikko Hiltunen178, Rebecca Sims162, Magdalena Gugała-Iwaniuk179, Mitchell K. P. Lai15, Venketasubramanian N180, Boon-Yeow Tan181, Angelo Baldassare Cefalù182, Nicola J Armstrong183, Roberta Baschi184,185, Regis bordet 186,187, Anne-Marie Bordet186,187, Henry Brodaty188, Srdjan Djurovic189,190, Grazia D'Onofrio191, Margaret Esiri192, Patrick Gelé186,187, Teresa Juarez-Cedillo193, Raj Kalaria194,195, Pekka Karhunen196, Jan LACZO133, Ondrej LERCH133,134, Carlo Masullo197, Karen A Mather188,198, Vaclav MATOSKA199, Susanna Melkas200, Roberto Monastero184,185, Katya Numbers188, Francesco Panza201,202,203, Tuomo M Polvikoski195,204, Joe Quinn205, Arvid Rongve206,207, Perminder S Sachdev188,208, Michela Scamosci209, Anbupalam Thalamuthu188, Anne Tybjærg-Hansen210, Martin VYHNALEK133,134, Shawn K. Westaway211, Amy E Martinsen212,213,214, Anne Heidi Skogholt214, Cristen J Willer215, Eystein Stordal216,217, Geir Bråthen218,219,220, Jonas Bille Nielsen214,215, Lars G Fritsche221, Laurent F Thomas214,222,223,224, Linda M Pedersen212, Maiken E Gabrielsen214, Ole Kristian Drange216,225, Sigrid Botne Sando214,218,226, Tore Wergeland Meisingset218,226, Genevieve Chene4,23, Wei Zhou227,228, Christophe Tzourio4,229, Adrienne Tin230, Oscar L Lopez231, Haan Mary232, Allison E Aiello233, Sigrid Børte213,214,234, Ingunn Bosnes216,217, Cornelia van Duijn235,236,237, Ching-Lung Cheung238, David A Bennett13, Christopher Chen15, M. Ilyas Kamboh10, Claudia Satizabal1,3, M. Kamran Ikram17,239, Hieab Adams240,241,242, Yang Qiong68, Gerard D. Schellenberg22, Geir Selbæk213,243,244, Kristian Hveem214,245,246, Ole A Andreassen247,248, Alfredo Ramirez109,249,250,251, Carole Dufouil4,23, Wiesje van der Flier252, John-Anker Zwart212,213,214, Stéphanie Debette4,253, Myriam Fornage29,254, Bendik Winsvold214,255,256, Jean-Charles Lambert5, Agustin Ruiz6,7, Patrick G. Kehoe257, Galit Weinstein258,#, and Sudha Seshadri1,259,260,261,#

    1 Glenn Biggs Institute for Alzheimer's & Neurodegenerative Diseases, University of Texas Health Science Center, San Antonio, TX, USA

    2 Department of Biochemistry and Structural Biology, University of Texas Health Science Center, San Antonio, TX, USA

    3 Department of Population Health Sciences, University of Texas Health Science Center, San Antonio, TX, USA

    4 University of Bordeaux, Inserm, Bordeaux Population Health Research Center, UMR 1219, F-33000 Bordeaux, France

    5 Univ. Lille, Inserm, CHU Lille, Institut Pasteur de Lille, U1167-RID-AGE facteurs de risque et déterminants moléculaires des maladies liés au vieillissement, Lille, France

    6 Research Center and Memory Clinic, ACE Alzheimer Center Barcelona. Universitat Internacional de Catalunya, Spain

    7 Network Center for Biomedical Research in Neurodegenerative Diseases (CIBERNED), Instituto de Salud Carlos III, Madrid, Spain

    8 Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, UK

    9 Cardiovascular Health Research Unit, Department of Medicine, University of Washington, Seattle, WA

    10 Department of Human Genetics, School of Public Health, University of Pittsburgh, Pittsburgh, PA, USA

    11 Department of Psychiatry and Neurology, School of Medicine, University of Pittsburgh, Pittsburgh, PA, USA

    12 Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hung Hom, Hong Kong

    13 Rush Alzheimer's Disease Center and Department of Neurological Sciences, Rush University Medical Center, Chicago, IL, USA

    14 Geriatrics Unit, Policlinico Universitario Fondazione Agostino Gemelli IRCCS, Largo a Gemelli, 8−00168 Rome, Italy

    15 Department of Pharmacology, National University of Singapore, Singapore

    16 Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore

    17 Department of Epidemiology, Erasmus MC, University Medical Center, Rotterdam, the Netherlands

    18 Institute for Maternal and Child Health, IRCCS Burlo Garofolo, 34127 Trieste, Italy

    19 Department of Medicine, Surgery and Health Sciences, University of Trieste, 34139 Trieste, Italy

    20 Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, VIC, Australia

    21 Faculty of Medicine, University of Iceland, Reykjavik, Iceland

    22 Department of Biostatistics and Epidemiology/Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA

    23 Pôle de Santé Publique Centre Hospitalier Universitaire (CHU) de Bordeaux, Bordeaux, France

    24 Icelandic Heart Association, Kopavogur, Iceland

    25 Istituti Clinici Scientifici Maugeri IRCCS, Laboratory of Neuropsychology, Bari Institute, Italy

    26 Department of Translational Biomedicine and Neuroscience, University of Bari “Aldo Moro,” Bari, Italy

    27 Unit of Statistics and Epidemiology, Local Healthcare Authority of Taranto, Taranto, Italy

    28 Laboratory of Epidemiology and Population Sciences, Intramural Research Program, National Institute of Aging, National Institutes of Health, Bethesda, MD, USA

    29 Human Genetics Center, School of Public Health, The University of Texas Health Science Center at Houston, Houston TX, USA

    30 Stroke Branch, National Institute of Neurological Disorders and Stroke Intramural Program, National Institutes of Health, Bethesda, MD, USA

    31 University of Bordeaux, Inserm, Bordeaux Population Health Research Center, team ELEANOR, UMR 1219, F-33000 Bordeaux, France

    32 Department of Epidemiology, Erasmus MC University Medical Center Rotterdam, Rotterdam, the Netherlands

    33 Department of Radiology and Nuclear Medicine, Erasmus MC University Medical Center, the Netherlands

    34 Center for Alzheimer's and Related Dementias, National Institutes of Health, Bethesda, MD, USA

    35 Laboratory of Neurogenetics, National Institute on Aging, National Institutes of Health, Bethesda, MD, USA

    36 Data Tecnica International LLC, Glen Echo, MD, USA

    37 Center for Public Health Genomics, University of Virginia, Charlottesville, VA, USA

    38 Department of Biochemistry and Molecular Genetics, University of Virginia, Charlottesville, VA, USA; Department of Biochemistry and Molecular Genetics, University of Virginia, Charlottesville, VA, USA

    39 Cardiovascular Health Research Unit, Department of Medicine, University of Washington, Seattle, WA, USA

    40 Department of Biostatistics, University of Washington, Seattle, WA, USA

    41 Division of Preventive Medicine, Brigham and Women's Hospital, Boston, MA, USA

    42 Harvard Medical School, Boston, MA, USA

    43 Department of General Practice, Amsterdam UMC, location University of Amsterdam, Meibergdreef 9, Amsterdam, the Netherlands

    44 Amsterdam Public Health, Aging & Later Life and Personalized Medicine, Amsterdam, the Netherlands

    45 Amsterdam Neuroscience, Neurodegeneration and Mood, Anxiety, Psychosis, Stress, and Sleep, Amsterdam, the Netherlands

    46 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht and Utrecht University, Utrecht, the Netherlands

    47 University Medicine Greifswald, Institute for Community Medicine, SHIP/KEF, Greifswald, Germany

    48 Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands

    49 Department of Cardiology, Leiden University Medical Center, Leiden, the Netherlands

    50 Division Heart & Lungs, Department of Cardiology, University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands

    51 Department of Psychiatry and Psychotherapy, University Medicine Greifswald, Germany

    52 German Center for Neurodegenerative Diseases (DZNE), Site Rostock/Greifswald, Rostock, Germany

    53 Netherlands Heart Institute, Utrecht, the Netherlands

    54 Einthoven Laboratory for Experimental Vascular Medicine, LUMC, Leiden, the Netherlands

    55 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands

    56 Department of Epidemiology and Population Health, Albert Einstein College of Medicine, New York, NY, USA

    57 Longitudinal Studies Section, Translational Gerontology Branch, National Institute on Aging, Baltimore, MD, USA

    58 Department of Epidemiology, University of Washington, Seattle, WA, USA

    59 Department of Neurology, University of Washington, Seattle, WA, USA

    60 Institute for Translational Genomics and Population Sciences, Department of Pediatrics, Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center, Los Angeles, CA, USA

    61 BHF Glasgow Cardiovascular Research Centre, Faculty of Medicine, Glasgow, UK

    62 Institute of Cardiovascular and Medical Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow, UK

    63 Laboratory of Epidemiology and Population Sciences—National Institutes of Health, Bethesda, MD, USA

    64 University Medicine Greifswald, Department of Neurology, Greifswald, Germany

    65 Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX, USA

    66 Framingham Heart Study, Framingham, MA, USA

    67 Department of Neurology, Boston University School of Medicine, Boston, MA, USA

    68 Department of Biostatistics, Boston University School of Public Health, Boston, MA, USA

    69 Department of Neurology, Beth Israel Deaconess Medical Center, Boston, MA, USA

    70 Department of Epidemiology, Human Genetics and Environmental Sciences, University of Texas Health Science Center at Houston, School of Public Health, Houston, TX, USA

    71 Department of Epidemiology, Human Genetics and Environmental Sciences, The University of Texas School of Public Health, Houston, Texas, USA

    72 DZHK (German Centre for Cardiovascular Research), Partner Site Greifswald, Greifswald, Germany

    73 Interfaculty Institute for Genetics and Functional Genomics, University Medicine Greifswald, Greifswald, Germany

    74 Memory Impairment and Neurodegenerative Dementia (MIND) Center and Department of Medicine, University of Mississippi Medical Center, Jackson, MS, USA

    75 Grupo de Medicina Xenómica, CIBERER, CIMUS. Universidade de Santiago de Compostela, Santiago de Compostela, Spain

    76 Fundación Pública Galega de Medicina Xenómica-IDIS, Santiago de Compostela, Spain

    77 Unidad de Demencias, Servicio de Neurología y Neurofisiología. Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocío/CSIC/Universidad de Sevilla, Seville, Spain

    78 Unitat Trastorns Cognitius, Hospital Universitari Santa Maria de Lleida, Lleida, Spain

    79 Institut de Recerca Biomedica de Lleida (IRBLLeida), Lleida, Spain

    80 Hospital Universitario Ramon y Cajal, IRYCIS, Madrid, Spain

    81 Unitat de Genètica Molecular, Institut de Biomedicina de València-CSIC, Valencia, Spain

    82 Unidad Mixta de Neurologia Genètica, Instituto de Investigación Sanitaria La Fe, Valencia, Spain

    83 Departamento de Especialidades Quirúrgicas, Bioquímica e Inmunología. School of Medicine. University of Málaga, Málaga, Spain

    84 Alzheimer Research Center & Memory Clinic, Instituto Andaluz de Neurociencia, Málaga, Spain

    85 Unidad Clínica de Enfermedades Infecciosas y Microbiología, Hospital Universitario de Valme, Seville, Spain

    86 Departamento de Especialidades Quirúrgicas, Bioquímica e Inmunología, Facultad de Medicina, Universidad de Málaga, Málaga, Spain

    87 CAEBI, Centro Andaluz de Estudios Bioinformáticos, Sevilla, Spain

    88 Centro de Biología Molecular Severo Ochoa (UAM-CSIC)

    89 Instituto de Investigacion Sanitaria “Hospital la Paz” (IdIPaz), Madrid, Spain

    90 Universidad Autónoma de Madrid

    91 CIEN Foundation/Queen Sofia Foundation Alzheimer Center/Instituto de Salud Carlos III

    92 UFIEC, Instituto de Salud Carlos III, Madrid, Spain

    93 CIEN Foundation/Queen Sofia Foundation Alzheimer Center, Madrid, Spain

    94 Unidad de Trastornos del Movimiento, Servicio de Neurología y Neurofisiología. Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocío/CSIC/Universidad de Sevilla, Seville, Spain

    95 Departamento de Medicina, Facultad de Medicina, Universidad de Sevilla, Seville, Spain

    96 Alzheimer's Centre Reina Sofia-CIEN Foundation, Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas (CIBERNED), Madrid, Spain

    97 Unit of Neurodegenerative Diseases, Department of Neurology, Hospital Germans Trias i Pujol, Badalona, Barcelona, Spain

    98 Neurodegenerative Diseases Research Laboratory, Germans Trias i Pujol Research Laboratory, Badalona, Barcelona, Spain

    99 Laboratorio de Genética, Hospital Universitario Central de Asturias, Oviedo, Spain

    100 Instituto de Investigación Sanitaria del Principado de Asturias (ISPA), Oviedo, Spain

    101 Complex Genetics of Alzheimer's Disease Group, VIB Center for Molecular Neurology, VIB, Antwerp, Belgium

    102 Laboratory of Neurogenetics, Institute Born—Bunge, Antwerp, Belgium

    103 Department of Biomedical Sciences, University of Antwerp, Neurodegenerative Brain Diseases Group, Center for Molecular Neurology, VIB, Antwerp, Belgium

    104 Alzheimer Center Amsterdam, Neurology, Vrije Universiteit Amsterdam, Amsterdam UMC Location VUmc, Amsterdam, the Netherlands

    105 Amsterdam Neuroscience, Neurodegeneration, Amsterdam, the Netherlands

    106 Section Genomics of Neurodegenerative Diseases and Aging, Human Genetics, Vrije Universiteit Amsterdam, Amsterdam UMC location VUmc, Amsterdam, the Netherlands

    107 German Center for Neurodegenerative Diseases (DZNE), Berlin, Germany

    108 Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Institute of Psychiatry and Psychotherapy, Hindenburgdamm 30, 12203 Berlin, Germany

    109 German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany

    110 Department for Neurodegenerative Diseases and Geriatric Psychiatry, University Hospital Bonn, Venusberg-Campus 1, 53127 Bonn, Germany

    111 Institute for Stroke and Dementia Research (ISD), University Hospital, LMU Munich, Munich, Germany

    112 German Center for Neurodegenerative Diseases (DZNE), Munich, Germany

    113 Munich Cluster for Systems Neurology (SyNergy), Munich, Germany

    114 Martin-Luther-University Halle-Wittenberg, University Clinic and Outpatient Clinic for Psychiatry, Psychotherapy and Psychosomatics, Halle (Saale), Germany

    115 Department of Psychiatry, Psychosomatics and Psychotherapy, Center of Mental Health, University Hospital of Würzburg, Germany

    116 German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany

    117 Institute of Cognitive Neurology and Dementia Research (IKND), Otto-von-Guericke University, Magdeburg, Germany

    118 Department of Psychiatry and Psychotherapy, University Medical Center Goettingen, Goettingen, Germany

    119 German Center for Neurodegenerative Diseases (DZNE), Goettingen, Germany

    120 Medical Science Department, iBiMED, Aveiro, Portugal

    121 Department of Neurodegenerative Diseases and Geriatric Psychiatry, University of Bonn, Bonn, Germany

    122 German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany

    123 Technical University of Munich, School of Medicine, Klinikum rechts der Isar, Department of Psychiatry and Psychotherapy

    124 Taub Institute for Research in Alzheimer's Disease and the Aging Brain, The Gertrude H. Sergievsky Center, Department of Neurology, Columbia University, New York, NY, USA

    125 1st Department of Neurology, Aiginition Hospital, National and Kapodistrian University of Athens, Medical School, Greece

    126 Department of Neurology, Hospital Universitario Donostia, San Sebastian, Spain

    127 Neurosciences Area. Instituto Biodonostia. San Sebastian, Spain

    128 Alzheimer's Disease and Other Cognitive Disorders Unit, Service of Neurology, Hospital Clinic of Barcelona, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona, Barcelona, Spain

    129 Depatamento de Especialidades Quirúrgicas, Bioquímica e Inmunología, Facultad de Medicina, Universidad de Málaga, Málaga, Spain

    130 Neurology Service, Marqués de Valdecilla University Hospital (University of Cantabria and IDIVAL), Santander, Spain

    131 Department of Neurosciences, Faculty of Medicine and Nursery, University of the Basque Country, San Sebastián, Spain

    132 Faculty of Medicine, University of Lisbon, Portugal

    133 Memory Clinic, Department of Neurology, Charles University, Second Faculty of Medicine and Motol University Hospital, Czech Republic

    134 International Clinical Research Center, St. Anne's University Hospital Brno, Brno, Czech Republic

    135 Unit for Hereditary Dementias, Theme Aging, Karolinska University Hospital-Solna, 171 64 Stockholm, Sweden

    136 Aging Research Center, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet and Stockholm University, Stockholm, Sweden

    137 Department of Public Health and Carins Sciences/Geriatrics, Uppsala University, Sweden

    138 Department of Clinical Biochemistry, Copenhagen University Hospital – Herlev Gentofte, Denmark

    139 Department of Clinical Medicine, University of Copenhagen, Denmark

    140 Institute of Clinical Medicine—Neurology, University of Eastern Finland, Finland

    141 Division of Clinical Geriatrics, Center for Alzheimer Research, Care Sciences and Society (NVS), Stockholm, Sweden

    142 Karolinska Institutet, Stockholm, Sweden

    143 Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland

    144 Neuroepidemiology and Ageing Research Unit, School of Public Health, Imperial College London, London, UK

    145 Stockholms Sjukhem, Research & Development Unit, Stockholm, Sweden

    146 A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland

    147 University of Rouen Normandy, Normandy University, Inserm U1245 and CHU Rouen, Department of Genetics and CNRMAJ, Rouen, France

    148 University of Lille Inserm 1171, CHU Clinical and Research Memory Research Centre (CMRR) of Distalz, Lille, France

    149 Université de Paris, EA 4468, APHP, Hôpital Broca, Paris, France

    150 Department of Child and Adolescent Psychiatry and Psychotherapy, University Hospital of Psychiatry Zurich, University of Zurich, Zurich, Switzerland

    151 Neuroscience Center Zurich, University of Zurich and ETH Zurich, Switzerland

    152 Zurich Center for Integrative Human Physiology, University of Zurich, Switzerland

    153 Neurodegenerative Diseases Unit, Fondazione IRCCS Ca’ Granda, Ospedale Policlinico, Milan, IT

    154 Department of Biomedical, Surgical and Dental Sciences, University of Milan, Milan, Italy

    155 Department of Clinical Sciences and Community Health, University of Milan, Milan, Italy

    156 Geriatric Unit, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy

    157 Institute of Gerontology and Geriatrics, Department of Medicine and Surgery, University of Perugia, Italy

    158 Department of Biomedical Sciences, University of Cagliari, Cagliari, Italy

    159 Neurology, “San Gerardo” Hospital, Monza and University of Milano-Bicocca, Italy

    160 Department of Neuroscience “Rita Levi Montalcini,” University of Turin, Turin, Italy

    161 Department of Hematology and Stem Cell Transplant, Vito Fazzi Hospital, Lecce, Italy

    162 Division of Psychological Medicine and Clinical Neuroscience, School of Medicine, Cardiff University, Wales, UK

    163 Univ. Lille, Inserm, CHU Lille, Institut Pasteur de Lille, U1167-RID-AGE—LabEx DISTALZ, facteurs de risque et déterminants moléculaires des maladies liés au vieillissement, Lille, France

    164 Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany

    165 Cluster of Excellence Cellular Stress Responses in Aging-Associated Diseases (CECAD), University of Cologne, Cologne, Germany

    166 1st Department of Neurology, Medical school, Aristotle University of Thessaloniki, Thessaloniki, Makedonia, Greece

    167 Department of Clinical Biochemistry, Copenhagen University Hospital—Rigshospitalet, Copenhagen, Denmark

    168 Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark

    169 Department of Biomedical Sciences, University of Antwerp, Antwerp, Belgium

    170 Department of Neurology, Universitair Ziekenhuis Brussel and NEUR (Neuroprotection & Neuromodulation) Research Group, Center for Neurosciences, Vrije Universiteit Brussel (VUB), Brussels, Belgium

    171 Reference Center for Biological Markers of Dementia (BIODEM), Institute Born-Bunge, University of Antwerp, Antwerp, Belgium

    172 Laboratory for Cognitive Neurology, Department of Neurosciences, University of Leuven, Leuven, Belgium

    173 Neurology Department, University Hospitals Leuven, Leuven, Belgium

    174 Department of Public Health and Carins Sciences/Geriatrics, Uppsala University, Sweden

    175 Krembil Brain Institute, University Health Network, Toronto, Canada

    176 Tanz Centre for Research in Neurodegenerative Diseases, Departments of Medicine and Laboratory Medicine & Pathobiology, University of Toronto, Toronto, Canada

    177 Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy

    178 Institute of Biomedicine, University of Eastern Finland, Kuopio, Finland

    179 Institute of Psychiatry and Neurology, First Department of Neurology, Warsaw, Poland

    180 Raffles Neuroscience Center, Raffles Hospital, Singapore

    181 St Luke's Hospital, Singapore, Singapore

    182 Department of Health Promotion Sciences, Maternal and Infant Care (PROMISE), University of Palermo, Palermo, Italy

    183 Department of Mathematics and Statistics, Curtin University, Perth, Australia

    184 Department of Biomedicine, Neuroscience and Advanced Diagnostics (BIND), University of Palermo, Palermo, Italy

    185 Dementia and Parkinson's disease Center, University Hospital, “Paolo Giaccone,” Palermo, Italy

    186 Univ Lille, Inserm, CHU Lille, France

    187 Lille Neuroscience & Cognition, Lille, France

    188 Centre for Healthy Brain Ageing, Discipline of Psychiatry & Mental Health, School of Clinical Medicine, Faculty of Medicine and Health, University of New South Wales, Sydney, Australia

    189 NORMENT Centre, University of Bergen, Bergen, Norway

    190 Dept of Medical Genetics, Oslo University Hospital, Oslo, Norway

    191 Clinical Psychology Service, Health Department, Fondazione IRCCS Casa Sollievo della Sofferenza, San Giovanni Rotondo (FG), Italy

    192 Nuffield Department of Clinical Neurosciences, Oxford University, Oxford, UK

    193 Unidad de Investigación en Epidemiología y Servicos de Salud Área Envejecimiento, Centro Medico Nacional Siglo XXI, Instituto Mexicano del Seguro Social. Ciudad de Mexico, Mexico

    194 Translational and Clinical Research Institute, Newcastle University

    195 Campus for Ageing and Vitality, Newcastle upon Tyne NE4 5PL, UK

    196 Faculty of Medicine and Health Technology, Tampere University, and Department of Clinical Chemistry, Fimlab Laboratories. Tampere, Finland

    197 Institute of Neurology, Catholic University of the Sacred Heart, School of Medicine, Largo Agostino Gemelli, Rome, Italy

    198 Neuroscience Research Australia, Sydney, Australia

    199 Department of Clinical Biochemistry, Hematology and Immunology, Na Homolce Hospital, Prague, Czech Republic

    200 Helsinki University Hospital, University of Helsinki, Helsinki, Finland

    201 Neurodegenerative Disease Unit, Department of Basic Medicine, Neuroscience, and Sense Organs, University of Bari Aldo Moro, Policlinico, Piazza Giulio Cesare 11, 70124 Bari, Italy

    202 Geriatric Unit & Laboratory of Gerontology and Geriatrics, Department of Medical Sciences, IRCCS “Casa Sollievo della Sofferenza,” San Giovanni Rotondo, Viale Cappuccini 1, 71013 San Giovanni Rotondo, Foggia, Italy

    203 Unit of Research Methodology and Data Sciences for Population Health, National Institute of Gastroenterology Saverio de Bellis, Research Hospital, Castellana Grotte, Bari, Italy

    204 Translational and Clinical Research Institute, Newcastle University, Newcastle, UK

    205 Department of Neurology, Oregon Health & Science University, Portland, Oregon, USA

    206 Department of Research and Innovation, Helse Fonna, Haugesund, Norway

    207 Department of Clinical Medicine (K1), University of Bergen, Bergen, Norway

    208 Neuropsychiatric Institute, Euroa Centre, Prince of Wales Hospital, Sydney, Australia

    209 Institute of Gerontology and Geriatrics, Department of Medicine, University of Perugia, Perugia, Italy

    210 Department of Clinical Biochemistry, Copenhagen University Hospital – Rigshospitalet, Copenhagen Denmark & Department of Clinical Medicine, Copenhagen, Denmark

    211 Department of Neurology, Oregon Health & Science University, Portland, Oregon, USA

    212 Department of Research and Innovation, Division of Clinical Neuroscience, Oslo University Hospital, Oslo, Norway

    213 Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway

    214 K. G. Jebsen Center for Genetic Epidemiology, Department of Public Health and Nursing, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    215 Department of Internal Medicine, Division of Cardiovascular Medicine, University of Michigan, Ann Arbor, MI, USA

    216 Department of Mental Health, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    217 Department of Psychiatry, Hospital Namsos, Nord-Trøndelag Health Trust, Namsos, Norway

    218 Department of Neuromedicine and Movement Science, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    219 Department of Neurology and Clinical Neurophysiology, St. Olav's Hospital, Trondheim University Hospital, Trondheim, Norway

    220 K G Jebsen Centre for Alzheimer's Disease. Kavli Institutes of Systems Neuroscience, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    221 Center for Statistical Genetics, Department of Biostatistics, University of Michigan, Ann Arbor, MI, USA

    222 Department of Clinical and Molecular Medicine, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    223 BioCore—Bioinformatics Core Facility, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    224 Clinic of Laboratory Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway

    225 Division of Mental Health Care, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway

    226 Department of Neurology and Clinical Neurophysiology, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway

    227 Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, MI, USA

    228 Analytic and Translational Genetics Unit, Massachusetts General Hospital, Boston, MA, USA

    229 Bordeaux University Hospital, Department of Medical Informatics, Bordeaux, France

    230 Memory Impairment and Neurodegenerative Dementia (MIND) Center and Department of Medicine, University of Mississippi Medical Center, Jackson, MS, USA

    231 Department of Neurology, School of Medicine, University of Pittsburgh, PA, USA

    232 Department of Epidemiology & Biostatistics, University of California, San Francisco, CA, USA

    233 Department of Epidemiology, Carolina Population Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

    234 Research and Communication Unit for Musculoskeletal Health (FORMI), Department of Research and Innovation, Division of Clinical Neuroscience, Oslo University Hospital, Oslo, Norway

    235 Department of Epidemiology, Erasmus Medical Center, Rotterdam, the Netherlands

    236 Nuffield Department of Population Health, University of Oxford, Old Road Campus, Headington, Oxford, UK

    237 Big Data Institute, Li Ka Shing Centre for Health Information and Discovery, Old Road Campus, Headington, Oxford, UK

    238 Department of Pharmacology and Pharmacy, Centre for Genomic Sciences, University of Hong Kong, Hong Kong

    239 Department of Neurology, Erasmus University Medical Centre, Rotterdam, the Netherlands

    240 Department of Clinical Genetics, Erasmus MC, Rotterdam, the Netherlands

    241 Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, the Netherlands

    242 Department of Psychology, Latin American Brain Health (BrainLat), Universidad Adolfo Ibáñez, Santiago, Chile

    243 Norwegian National Advisory Unit on Ageing and Health, Vestfold Hospital Trust, Tønsberg, Norway

    244 Department of Geriatric Medicine, Oslo University Hospital, Oslo, Norway

    245 HUNT Research Center, Department of Public Health and Nursing, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

    246 Department of Research, Innovation and Education, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway

    247 Division of Mental Health and Addiction, Oslo University Hospital, Oslo, Norway

    248 NORMENT, University of Oslo, Oslo, Norway

    249 Division of Neurogenetics and Molecular Psychiatry, Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany

    250 Department of Neurodegenerative diseases and Geriatric Psychiatry, University Hospital Bonn, Medical Faculty, Bonn, Germany

    251 Department of Psychiatry & Glenn Biggs Institute for Alzheimer's and Neurodegenerative Diseases, San Antonio, TX, USA

    252 Alzheimer Center Amsterdam, Department of Neurology, Amsterdam Neuroscience, Vrije Universiteit Amsterdam, Amsterdam UMC, Amsterdam, the Netherlands

    253 CHU de Bordeaux, Department of Neurology, Institute for Neurodegenerative Diseasese, Bordeaux, France

    254 Institute of Molecular Medicine, McGovern Medical School, The University of Texas Health Science Center at Houston, Houston, TX, USA

    255 Department of Research, Innovation and Education, Division of Clinical Neuroscience, Oslo University Hospital, Oslo, Norway

    256 Department of Neurology, Oslo University Hospital, Oslo, Norway

    257 Translational Health Sciences, Bristol Medical School, University of Bristol, Bristol, UK

    258 School of Public Health, Faculty of Social Welfare and Health Sciences, University of Haifa, Haifa, Israel

    259 Framingham Heart Study, Framingham, MA, USA

    260 Department of Neurology, University of Texas Health San Antonio, San Antonio, TX, USA

    261 Department of Neurology, Boston University School of Medicine, Boston, MA, USA

    DATA AVAILABILITY STATEMENT

    The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. https://ctg.cncr.nl/software/summary_statistics, https://pubmed.ncbi.nlm.nih.gov/36180795/