Genome-Wide Association Study of Vascular Dementia
Background and Purpose—Most studies investigating the genetics of dementia have focused on Alzheimer disease, but little is known about the genetics of vascular dementia. The aim of our study was to identify new loci associated with vascular dementia.
Methods—We performed a genome-wide association study in the Rotterdam Study, a large prospective population-based cohort study in the Netherlands. We sought to replicate genome-wide significant loci in 2 independent replication samples.
Results—In the discovery analysis of 5700 dementia-free individuals, 67 patients developed incident vascular dementia over a mean follow-up time of 9.3±3.2 years. We showed genome-wide significance for rs12007229, which is located on the X chromosome near the androgen receptor gene (OR, 3.7; 95% CI, 2.3–5.8, per copy of the minor allele; P=1.3×10−8). This association was further confirmed in 2 independent populations (probability value of combined replication samples=0.024).
Conclusions—Our study shows a novel genetic locus for vascular dementia on the X chromosome. Further replication of this finding is required.
Vascular dementia is the second most common type of dementia, accounting for approximately 15 to 20% of all cases of dementia.1 Vascular dementia is defined as the loss of cognitive function resulting from ischemic, ischemic-hypoxic, or hemorrhagic brain lesions as a result of cerebrovascular disease and cardiovascular pathological changes.2 Most studies that investigated the genetics of dementia have focused on Alzheimer disease, the most common type of dementia. Candidate gene studies on vascular dementia often investigated genes implicated in Alzheimer disease3 or focused on rare subtypes of vascular dementia like cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL).4 Recently, several genome-wide association studies have been performed on Alzheimer disease, leading to the discovery of new loci.5,6 To identify new loci associated with vascular dementia, we performed a genome-wide association study in the Rotterdam Study, a large prospective population-based cohort study. We sought to replicate genome-wide significant loci in 2 independent populations.
Incident Vascular Dementia in the Rotterdam Study
The Rotterdam Study is conducted among inhabitants age 55 years and older of Ommoord, a district of Rotterdam, the Netherlands (N=7983, virtually all white). Details of the study have been described elsewhere.7 The medical ethics committee at Erasmus University Rotterdam approved the study and written informed consent was obtained from all participants. Baseline examinations were conducted in 1990 to 1993, with follow-up examinations in 1993 to 1994, 1997 to 1999, and 2002 to 2004.
Participants were screened for dementia at baseline and follow-up visits using a 3-step protocol.8 In addition, the cohort was continuously monitored for incident dementia through medical records from general practitioners.8 The diagnoses of dementia and vascular dementia were made in accordance with internationally accepted criteria for dementia (Diagnostic and Statistical Manual of Mental Disorders; DSM-III-R), and vascular dementia (National Institute of Neurological Disorders and Stroke and Association Internationale pour la Recherché et l'Enseignement en Neurosciences; NINDS-AIREN).9 Patients with mixed-type dementia (eg, Alzheimer disease with cerebrovascular disease) were not included in the vascular dementia group, but to maximize power were included as controls. Follow-up for dementia was virtually complete as of January 1, 2005.
All patients attending the baseline examination in 1990 to 1993 consented to genotyping and had DNA extracted. Genotyping was attempted in patients with high-quality extracted DNA (n=6449) and was performed in 2007 to 2008 using the Illumina Infinium II HumanHap550chipv3.0 array according to the manufacturer's protocols. Samples with low call rate (<97.5%, n=209), with excess autosomal heterozygosity (>0.336, n=21), with sex-mismatch (n=36), or if there were outliers identified by the identity-by-state (IBS) clustering analysis (>3 standard deviations from population mean, n=102 or IBS probabilities >97%, n=129) were excluded from the study population with some patients meeting more than one exclusion criterion.6 In total, 5974 samples were available with good-quality genotyping data, 42 patients were excluded because they did not undergo cognitive screening, and 232 patients were excluded because they suffered from dementia at baseline. This yielded a discovery population of 5700 patients, who were followed for incident vascular dementia. Because imputation was unavailable for the X chromosome, unimputed data were used. X chromosomal single-nucleotide polymorphisms (SNP) with impossible heterozygotes in men were excluded (8608 SNPs).
We used PLINK genetics software to perform 1-degree-of-freedom allelic tests to obtain odds ratios with corresponding 95% CI per copy increase of the coded allele. For analysis of the X chromosome, the allelic test in PLINK considers male carriers equivalent to heterozygote women, (ie, the obtained β reflects the effect size per allele increase). We used a threshold of 5×10−8 for genome-wide significance. We used marker inclusion thresholds of minor allele frequency >0.01, SNP call rate >0.90, and Hardy-Weinberg P>0.001. There were a total of 515 461 autosomal and 12 666 X chromosomal SNPs in the analysis.
Prevalent Vascular Dementia in the Rotterdam Study
At baseline, 35 patients had vascular dementia. Genotyping of the significant hit from the discovery phase, rs12007229, was unsuccessful in 74 patients without dementia, leaving 35 prevalent cases and 5626 patients without dementia at baseline for the initial replication. Because mixed dementia cases were not collected in the German sample (see below), we also excluded prevalent mixed dementia cases from the Rotterdam Study for comparability across replication samples.
German Case-Control Sample
Patients with vascular dementia and controls were recruited at the Memory Clinics, Department of Psychiatry, University of Bonn, and Central Institute of Mental Health, Mannheim (total, 160 cases and 160 controls). Additional patients (27 cases and 29 controls were recruited at 6 German gerontopsychiatric university departments in the context of the German Dementia Competence Network (http://www.kompetenznetz-demenzen.de), and another 46 cases and corresponding controls were recruited through primary care in the scope of the German Competence Network of Degenerative Dementias (http://www.knd-demenzen.de/) from 6 German recruitment sites. Vascular dementia diagnosis was guided by the NINDS-AIREN criteria and supported by general physical, neurological, and psychiatric examinations; standard laboratory testing; and neuroimaging (although imaging was not performed in the latter 46 cases and controls). Patients with mixed-type dementia were excluded from the study. All patients or their legal guardians gave informed consent for participation in the study. The ethics committees of all the participating universities approved the study.
DNA was shipped to Rotterdam for genotyping. The significant hit from the discovery phase, rs12007229, was genotyped using Applied Biosystems Taqman allelic discrimination assay. For comparability across the discovery and replication study, we excluded 5 cases and 6 controls age <55 years, and 7 cases and 16 controls on whom data on genotyping were missing. This resulted in a study population of 221 cases and 213 controls for the replication in an independent sample.
For the replication analyses, we used 1-degree-of-freedom allelic tests with 1-sided P<0.05 as threshold for significance. In the Rotterdam Study, the incident discovery analysis is independent from the prevalent replication analysis because the controls contribute different sets of person-years to the 2 analyses, which are therefore independent from each other. The independence of these analyses was confirmed in simulation studies and has been successfully applied previously.10
Table 1 shows the age and sex distribution of the different samples. In the discovery analysis of 5700 dementia-free individuals, 67 patients developed incident vascular dementia during an average follow-up time of 9.3±3.2 years. We identified 2 SNPs that reached genome-wide significance (Figure A): rs12007229 on chromosome X with an odds ratio of 3.7 (95% CI, 2.3–5.8; P=1.3×10−8) per copy increase of the minor allele ‘A' compared with ‘C' and minor allele frequency of 7%; and rs10491487 on chromosome 5 with an odds ratio of 3.5 (95% CI, 2.2–5.4; P=3.7×10−8) per copy increase of the minor allele ‘C' compared with ‘A' and minor allele frequency of 5%. Supplemental Table S1, http://stroke.ahajournals.org, shows all SNPs that reached P<10−4 in the discovery analysis. The overall genomic inflation factor λ was 1.01. In a sensitivity analysis, we excluded 45 patients from the control group who developed mixed dementia during follow-up, which yielded largely unchanged results: OR for rs12007229 in the total group was 3.7 (2.3–6.0) with P=1.2×10−8; OR in men, 3.2 (1.4–7.5); P=4.4×10−3; and OR in women, 3.9 (2.2–7.0) with P=7.7×10−8.
We first performed preliminary replication analyses in the prevalent cases of the Rotterdam Study, which in itself was underpowered as a full replication study. Because we found no association between rs10491487 and vascular dementia (OR, 1.1; 1-sided probability value, 0.42), and we did observe an association between rs12007229 and vascular dementia (OR 1.7, 1-sided probability value, 0.09), we decided to genotype only rs12007229 in the independent German sample. This approach was also required because financial and logistic constraints provided the possibility to genotype only one SNP.
Table 2 shows the odds ratios for this SNP in the discovery phase and the replication studies separately and combined. We found a significant association of rs12007229 with vascular dementia when we combined both replication samples (OR 1.5; 1-sided probability value, 0.02). Given differences between men and women in number of alleles for X chromosomal SNPs, we also stratified the analyses by sex. However, we found similar effect estimates for men and women, especially for the discovery analysis and meta-replication, which are the 2 most powered results (Table 2). We note that had we sought formal replication of both genome-wide significant SNPs the replication of rs12007229 would not have survived Bonferroni correction. Figure B plots probability values from the discovery analysis of SNPs in the vicinity of rs12007229 and shows this locus to be close to the androgen-receptor gene (AR).
In this study of vascular dementia, we identified a new locus related to vascular dementia on the X chromosome. We replicated this locus in independent populations.
The locus on the X-chromosome (rs12007229) is located close to the AR gene, which is composed of 8 exons.11 Exon 1 of the AR-gene encodes the amino terminal domain, exons 2 and 3 encodes the DNA-binding domain, and exons 4 to 8 encode the steroid-binding domain. The amino terminal domain contains 2 polymorphic trinucleotide repeats: a CAG repeat encoding for polyglutamine, and a GGN repeat encoding for polyglycine.12 To our knowledge, there is no LD between rs12007229 and any known SNP within exon 1 of the AR gene. Still, it is interesting to note that an expansion of the CAG repeat is responsible for a rare X-linked motor neuron disorder in men: spinal and bulbar muscular atrophy or Kennedy's disease.13 The main neurological manifestation of Kennedy's disease is a slowly progressive weakness and wasting of bulbar, facial, and limb muscles.14 Although cognitive impairment or dementia is not one of the main features of Kennedy's disease, several case reports and a family study suggest that cognitive dysfunction might be part of the disease's clinical manifestation.15
Although the AR gene has not been described in vascular dementia thus far, there are some reports on associations of the CAG repeat length within the normal range with cognitive function and Alzheimer disease.16,17 Whether the locus we found reflects an association of the AR gene, and possibly the CAG repeat length, with vascular dementia it is unclear and needs additional investigation. Another possibility is that this SNP reflects an association with the ectodysplasin A2 receptor gene (EDA2R), although the distance between rs12007229 and the EDA2R gene is larger than with the AR gene.18
Some methodological issues need to be addressed. Strengths of our study are the large sample size from both a population-based cohort and an independent clinic-based sample. The diagnosis of vascular dementia can be difficult, especially in the subtype of poststroke vascular dementia. If a stroke is fatal or too severe (eg, leading to severe aphasia or hemiplegia), the diagnosis of vascular dementia cannot be made, thus leading to underestimation of the effect size. Moreover, vascular dementia is a heterogeneous syndrome that can be caused by various distinct underlying pathologies, (ie, large and small vessel disease). In a population-based study such as ours, stringent case-finding procedures allow for inclusion of patients in nursing homes or who were not hospitalized after stroke. However, this results in inclusion of patients who do not undergo neuro-imaging and therefore cannot be further classified according to subtypes of vascular dementia. Therefore, future studies need to clarify the role of rs12007229 in the pathologies underlying vascular dementia.
Another phenomenon we encountered in our study is the so-called winner's curse, where the effect estimates in the discovery set are overestimated resulting in much greater effect sizes than in the replication analyses.19 Another explanation for the difference in effect estimates between discovery and replication could be the difference in number of cases, with smaller sample sizes yielding more extreme effect estimates. Finally, the difference in effect size could be attributed to chance with OR=3.7 and OR=1.5 reflecting an overestimation and underestimation, respectively, and the true effect estimate lying somewhere in between. Future replication studies will need to establish the true effect estimate of this association.
Because of relatively small numbers, a few words on power are needed. Given our sample size and the minor allele frequency of 7%, we had >80% power to detect an OR of 3.0 in the discovery at α 5×10−8, and >80% power to find an OR of 1.4 in the replication at α 0.05.20
In conclusion, we report a novel putative locus for vascular dementia on the X chromosome close to the AR gene, which we replicated in an independent sample. However, additional replication is necessary to further establish this association.
We thank Birgitt Wiese for her support.
Sources of Funding
The generation and management of genome-wide genotype data for the Rotterdam Study is supported by the Netherlands Organisation of Scientific Research (Nederlandse Organisatie voor Wetenschappelijk Onderzoek [NOW] Investments) grants 175.010.2005.011 and 911-03-012. This study was funded by the Research Institute for Diseases in the Elderly (RIDE2) grant 014-93-015; the Netherlands Genomics Initiative (NGI)/Netherlands Organisation for Scientific Research (NWO) project number 050-060-810. M.A.I. received funding from the Netherlands Heart Foundation (Nederlandse Hartstichting, 2009B102) and Internationaal Parkinson Fonds. The Rotterdam Study was funded by Erasmus Medical Center and Erasmus University, Rotterdam, Netherlands Organization for the Health Research and Development (Zorgonderzoek Nederland/Medische Wetenschappen [ZonMw]), the Research Institute for Diseases in the Elderly (RIDE), the Ministry of Education, Culture and Science, the Ministry for Health, Welfare and Sports, the European Commission, and the Municipality of Rotterdam. The study was supported by the German Dementia Competence Network (DCN) and by the German Competence Network on Degenerative Dementias (CNDD), which was funded by the German Federal Ministry for Education and Research (Bundesministerium für Bildung und Forschung; grants: 01GI0420, DCN and 01GI0711, CNDD). H.K., R.H., and F.H. received funding from the German Research Foundation (Deutsche Forschungsgemeinschaft, He 2748/5 and KO2327/2-1).
The online-only Data Supplement is available at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.628768/-/DC1.
- Received June 9, 2011.
- Accepted October 14, 2011.
- © 2012 American Heart Association, Inc.
- Ott A,
- Breteler MM,
- van Harskamp F,
- Stijnen T,
- Hofman A
- Román GC,
- Tatemichi TK,
- Erkinjuntti T,
- Cummings JL,
- Masdeu JC,
- Garcia JH,
- et al
- Kuiper GG,
- Faber PW,
- van Rooij HC,
- van der Korput JA,
- Ris-Stalpers C,
- Klaassen P,
- et al
- Purcell S,
- Cherny SS,
- Sham PC