Skip to main content
UKPMC Funders Author Manuscripts logoLink to UKPMC Funders Author Manuscripts
. Author manuscript; available in PMC: 2009 Jun 30.
Published in final edited form as: Nat Genet. 2008 Sep;40(9):1056–1058. doi: 10.1038/ng.209

Collaborative genome-wide association analysis supports a role for ANK3 and CACNA1C in bipolar disorder

Manuel A R Ferreira 1,2,3,4,5,6, Michael C O’Donovan 7, Yan A Meng 1,2,3,4,5, Ian R Jones 7, Douglas M Ruderfer 1,3,4,5, Lisa Jones 8, Jinbo Fan 1,3,4,5, George Kirov 7, Roy H Perlis 1,2,3,4,5, Elaine K Green 7, Jordan W Smoller 1,2,3,4,5, Detelina Grozeva 7, Jennifer Stone 1,2,3,4,5, Ivan Nikolov 7,9, Kimberly Chambert 4,5, Marian L Hamshere 7,9, Vishwajit L Nimgaonkar 10, Valentina Moskvina 7,9, Michael E Thase 11,12, Sian Caesar 8, Gary S Sachs 1,2, Jennifer Franklin 5, Katherine Gordon-Smith 7,8, Kristin G Ardlie 5, Stacey B Gabriel 5, Christine Fraser 7, Brendan Blumenstiel 5, Matthew Defelice 5, Gerome Breen 13,14, Michael Gill 15, Derek W Morris 15, Amanda Elkin 14, Walter J Muir 16, Kevin A McGhee 16, Richard Williamson 14, Donald J MacIntyre 16, Alan W MacLean 16, David St Clair 13, Michelle Robinson 17, Margaret Van Beck 16, Ana C P Pereira 17, Radhika Kandaswamy 17, Andrew McQuillin 17, David A Collier 14, Nicholas J Bass 17, Allan H Young 18,19, Jacob Lawrence 17, I Nicol Ferrier 18, Adebayo Anjorin 17, Anne Farmer 14, David Curtis 17, Edward M Scolnick 4,5,20, Peter McGuffin 14, Mark J Daly 5,21,22,23, Aiden P Corvin 15, Peter A Holmans 7,9, Douglas H Blackwood 16; Wellcome Trust Case Control Consortium24, Hugh M Gurling 17, Michael J Owen 7, Shaun M Purcell 1,2,3,4,5,25, Pamela Sklar 1,2,3,4,5,20,25, Nick Craddock 7,25
PMCID: PMC2703780  EMSID: UKMS27163  PMID: 18711365

Abstract

To identify susceptibility loci for bipolar disorder, we tested 1.8 million variants in 4,387 cases and 6,209 controls and identified a region of strong association (rs10994336, P = 9.1 × 10-9) in ANK3 (ankyrin G). We also found further support for the previously reported CACNA1C (alpha 1C subunit of the L-type voltage-gated calcium channel; combined P = 7.0 × 10-8, rs1006737). Our results suggest that ion channelopathies may be involved in the pathogenesis of bipolar disorder.


Recent genome-wide association studies (GWAS) have identified genetic variants that show consistent association with common, complex diseases, such as type 2 diabetes1, prostate cancer2 and Crohn’s disease3. In many cases, a critical component for success involved combining results and data across multiple, smaller studies to provide adequate power to detect common variants of modest relative risk. Given that a single study will often not be sufficient, here we present the results of combining two previously published4,5 and one new (N 2,365) GWAS of bipolar disorder.

Three groups have performed independent GWAS of bipolar disorder, with little agreement among the most associated regions. Baum et al.6, using DNA pooling, identified a SNP located in DGKH (diacylglycerol kinase eta) that was associated with bipolar disorder in 1,233 cases and 1,439 controls with a P value of 1.5 × 10-8. The Wellcome Trust Case Control Consortium (WTCCC)4 analyzedm 1,868 cases and 2,938 controls and identified a locus in a gene-rich region of high linkage disequilibrium (LD) on chromosome 16p12 with P = 6.3 × 10-8. Recently, Sklar et al.5 performed a GWAS of 1,461 cases and 2,008 controls (the STEP-UCL study), finding the strongest single SNP result in MYO5B (myosin 5B), with P = 1.7 × 10-7. Although there was no obvious correspondence among the studies’ top few hits, a broader comparison of the WTCCC and STEP-UCL studies identified CACNA1C as showing the strongest consistent signal5.

We genotyped a new sample of 1,098 individuals with bipolar disorder and 1,267 controls on a similar platform as used by the WTCCC and STEP-UCL studies (Supplementary Table 1 online), with 331,786 SNPs passing quality control (Supplementary Methods online). This dataset, referred to as ED-DUB-STEP2, included additional samples from the STEP-BD study5 as well as cases and controls from the University of Edinburgh and Trinity College Dublin. Approval for each study was obtained from the appropriate review boards of the participating institutions and informed consent was obtained from all subjects. SNPs were tested for association using logistic regression co-varying for sample collection site and quantitative indices of ancestry based on a multi-dimensional scaling analysis; the genomic inflation factor (see Supplementary Methods) was λ1,000 = 1.056 (Supplementary Fig. 1 online).

Analysis of the ED-DUB-STEP2 samples identified 14 chromosomal regions associated at P < 5 × 10-5 (Supplementary Table 2 online). No region exceeded our threshold × for genome-wide significance of 5 × 10-8, correcting for an effective number of one million independent, common SNPs in the genome1,7. However, one of these regions spanned the CACNA1C association identified previously in the WTCCC and STEP-UCL studies5 (Supplementary Fig. 2 online), thus providing further independent support that this gene is associated with risk of bipolar disorder. In this region, the best single SNP across the WTCCC and STEP-UCL studies showed a consistent association in ED-DUB-STEP2 (rs1006737, P = 0.011, allele A, OR 1.21). Furthermore, the SNP with strongest association in ED-DUB-STEP2 (rs10774037, P = 4.6 × 10-5, allele G, OR 1.43) also had a consistent effect (P = 0.0008, OR = 1.14) in the WTCCC and STEP-UCL samples and was in LD with rs1006737 (r2 = 0.39).

Next, we combined individual genotyping data=from the WTCCC, STEP-UCL and ED-DUB-STEP2 studies, resulting in an overall sample of 4,387 cases and 6,209 controls genotyped on 325,690 overlapping SNPs, which substantially improves power to detect risk alleles of modest effect (Supplementary Table 1). The case sample comprised the following diagnoses: 81% bipolar 1, 16% bipolar 2, 2% schizoaffective manic and 1% bipolar NOS (Supplementary Table 3 online).

Genomic coverage was increased by imputing the genotypes for 1,444,258 additional HapMap SNPs using PLINK8 (Supplementary Methods), resulting in a total set of 1,769,948 SNPs. Applying a leave-one-out procedure for every genotyped SNP, we estimated concordance between imputed and true genotypes as 0.987 (Supplementary Table 4 online). Similar results were obtained using MACH1 (ref. 9; data not shown).

Our primary analysis was a logistic regression of disease state on single SNP allelic dosage, either directly genotyped or imputed, controlling for study site and two quantitative indices of ancestry (Supplementary Methods). The genomic inflation factor was λ1,000 = 1.024 when considering all SNPs, and λ1,000 = 1.021 for directly genotyped SNPs (Supplementary Fig. 1 and Supplementary Fig. 3 online).

Results for the primary analysis are shown in Figure 1a and available for download (http://pngu.mgh.harvard.edu/purcell/bpwgas2/). Thirty-nine SNPs showed a P < 10-6 and were located in three distinct chromosomal regions (Table 1). An additional 18 regions had at least one SNP with a P < 10-5 (Supplementary Fig. 4 and Supplementary Table 5 online). Cluster plots for these SNPs are shown in Supplementary Figure 5 online.

Figure 1.

Figure 1

Association results for the combined analysis of the WTCCC, STEP-UCL and ED-DUB-STEP2 studies. (a) Genome-wide results (-log10P) are shown in chromosomal order for directly genotyped (N = 325,690) and imputed (N = 1,444,258) SNPs that were tested for association in the overall sample of 4,387 bipolar cases and 6,209 controls. Horizontal lines indicate a P value of 5 × 10-8 (dashed) and 1 × 10-6 (dotted). (b) Plots for the three regions of strongest association. Results (-log10P) are shown for directly genotyped (diamonds) and imputed (circles) SNPs. The most associated SNP for each region is shown in blue, and the color of the remaining markers reflects the linkage disequilibrium (r2) with the top SNP in each panel (increasing red hue associated with increasing r2). The recombination rate (second y axis) is plotted in light blue and is based on the CEU HapMap population. The dashed horizontal line indicates a P = 5 × 10-8. Exons for each gene are represented by vertical bars, based on all isoforms available from the Mar 2006 UCSC genome browser assembly.

Table 1. Chromosomal regions with at least one SNP associated with bipolar disorder at P < 10-6 in the combined analysis of WTCCC, STEP-UCL and ED-DUB-STEP2 datasets.

Combined analysis
WTCCC
STEP-UCL
ED-DUB-STEP2
MA frequency
MA frequency
MA frequency
MA frequency
Type SNP, minor allele P OR Cases Controls P Cases Controls P Cases Controls P Cases Controls
Chromosome 10q21 (ANK3)
Imputed rs10994336,T 9.1 × 10-9 1.450 0.070 0.053 0.0006 0.070 0.054 0.0004 0.073 0.056 0.0002 0.070 0.049
Genotyped rs1938526,G 1.3 × 10-8 1.395 0.075 0.056 0.0014 0.074 0.058 0.0008 0.076 0.059 0.0002 0.073 0.047
Chromosome 12p14 (CACNA1C)
Genotyped rs1006737,A 7.0 × 10-8 1.181 0.356 0.0324 0.0015 0.357 0.324 0.0003 0.357 0.315 0.0108 0.353 0.337
Imputed rs1024582,A 1.7 × 10-7 1.180 0.382 0.351 0.0019 0.381 0.349 0.0012 0.384 0.348 0.0056 0.379 0.359
Chromosome 15q14
Imputed rs12899449,G 3.5 × 10-7 0.836 0.246 0.276 0.0140 0.253 0.277 0.0005 0.237 0.276 0.0013 0.247 0.273
Genotyped rs2172835,T 7.5 × 10-7 0.853 0.281 0.312 0.0314 0.292 0.315 0.0004 0.267 0.309 0.0011 0.281 0.312

The most associated directly genotyped and imputed SNPs are shown for each region. MA, minor allele.

The strongest association was in ANK3 on chromosome 10q21 for the imputed SNP rs10994336, with P = 9.1 × 10-9 (Fig. 1b). This result was supported by the following observations. First, multiple SNPs were associated across a 195-kb region. Second, three genotyped SNPs showed P values similar to rs10994336 (for example, rs1938526, P = 1.3 × 10-8, r2 = 0.40 with rs10994336). Third, haplotype analysis of genotyped SNPs supported the single SNP results (most associated haplotype P 2.0 × 10-9, Supplementary Table 6 online). Fourth, the association remained (rs10994336, P = 5.2 × 10-9) after including the six additional WTCCC disease panels as extra controls for the WTCCC dataset (‘expanded reference group’ analysis). Finally, individual Sequenom genotyping confirmed that rs10994336 was imputed with high accuracy (concordance = 0.983, N = 3,293).

The second strongest region of association was located in the third intron of CACNA1C on chromosome 12p13 (rs1006737, P = 7.0 × 10-8). The third region was 3.3 kb away from an uncharacterized gene (C15orf53) on chromosome 15q14 (rs12899449, P = 3.5 × 10-7), in the same LD block with the brain-expressed RASGRP1 (RAS guanyl releasing protein 1) gene. The association with both regions remained in the expanded reference group analysis (rs1006737, P = 4.0 × 10-8; rs12899449, P = 9.3 × 10-6).

Genotypic and heterogeneity analyses for these three regions suggested that the effect at each locus was consistent with an additive model (Supplementary Table 7 online) and of similar magnitude across studies (Supplementary Table 8 online). There was no clear indication of differential association across bipolar subtypes, psychosis, age-at-onset and sex (Supplementary Table 9 online). No significant treatment response effects were observed for these three loci (data not shown).

Finally, we carried out a two-locus genome-wide search to identify epistatically acting loci of moderate-to-large effect that single-locus analysis may have missed10,11. From over 4 × 1010 pairwise tests, we identified 61 interactions with a P < 5 × 10-10, representing seven distinct pairs of regions (Supplementary Table 10 online). No single interaction exceeded a Bonferroni corrected-threshold of 1.2 × 10-12, in spite of reasonable power to detect large interaction effects (Supplementary Methods).

In summary, we present evidence that variation in ANK3 confers risk of bipolar disorder in three independent datasets. ANK3 is an adaptor protein found at axon initial segments that has been shown to regulate the assembly of voltage-gated sodium channels12,13. We also provide independent support that CACNA1C is associated with bipolar disorder. In addition, we have recently shown that both ANK3 and subunits of the calcium channel are downregulated in the mouse brain in response to lithium14, one the most effective bipolar pharmacotherapies. Taken together, these results point to the possibility that bipolar disorder is in part an ion channelopathy15.

Supplementary Material

Supplementary data

ACKNOWLEDGMENTS

Study supported by US National Institute of Mental Health (MH062137, MH067288, P.S.; MH063445, J.W.S.; MH63420, V.L.N.; N01MH80001, G.S.S., M.E.T. and M.S. Bauer), NARSAD (R.H.P., S.P., J.F., P.S.); Johnson & Johnson Pharmaceutical Research & Development, Sylvan C. Herman Foundation, Stanley Medical Research Institute, Dauten Family, Merck Genome Research Institute (E.M.S.); and National Health Medical Research Council (MRC), Australia (M.A.R.F.). Broad Institute Center for Genotyping and Analysis supported by U54 RR020278, National Center for Research Resources. UCL sample collection supported by MDF-The Bipolar Organization, Neuroscience Research Charitable Trust, central London NHS Blood Transfusion Service, Priory Hospitals and UK MRC (G9623693N, G0500791). Dublin sample collection supported by the Health Research Board and Science Foundation Ireland; controls supplied by J. McPartlin, Trinity College Biobank. Edinburgh cohort collection supported by the Wellcome Trust, London, Chief Scientist Office of the Scottish Executive and MRC, London. WTCCC genotype analyses were funded by the Wellcome Trust and undertaken within the context of the WTCCC. WTCCC members are listed in Supplementary Note.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary data

RESOURCES