In their invited review on the genetics of bipolar affective disorder (BP), Keers et al. (2009) reported multiple association signals across the CACNA1C gene and suggested that because ‘these association signals were located in three blocks of largely distinct regions of linkage disequilibrium (LD)’ they ‘may therefore be considered as three relatively independent associations between CACNA1C and BP’.
However, it should be noted that non-negligible LD often exists between non-contiguous single nucleotide polymorphisms (SNPs), separated by tens to sometimes hundreds of kilobases and within separate so-called ‘LD blocks’, and that in the specific case reported by Keers et al., the moderate associations (7.38×10−5⩽p⩽3.88×10−4) between 15 SNPs and BP across CACNA1C can be completely explained by a single effect.
As shown in Fig. 1, substantial LD (r 2) (calculated using release 22 HapMapI+II CEU data; http://www.hapmap.org) exists between the 15 SNPs listed in table 1 of Keers et al. (2009), thus indicating that these SNPs are expected to provide similar evidence for association. Indeed, this can be easily demonstrated through simulating the results of Keers et al. by splitting the HapMap CEU samples into two groups to approximate the evidence for association (p≅7×10−5) of the most significant SNP (rs2238054) reported by Keers et al. and performing association analysis with and without conditioning on rs2238054.
Results from allelic association analyses utilizing logistic regression within the PLINK program (Purcell et al. 2007) (Table 1), clearly demonstrate that after conditioning on rs2238054, none of the 14 remaining SNPs listed in table 1 of Keers et al. show evidence for association.
OR, Odds ratio; Stat, t statistic; n.a. indicates rs2238054 completely accounted for the association signal at this locus.
a SNP rs2238054 produced the most significant association signal in Keers et al. (2009).
As a consequence, while the reported association(s) between CACNA1C and BP remain an interesting, although non-genome-wide significant finding, there is currently no evidence for multiple independent effects and further studies are therefore required to confirm involvement of CACNA1C variants with BP susceptibility.
Declaration of Interest
The authors reply
We welcome Dr Nyholt's further analyses of our findings from the WTCCC data and any further discussion regarding this important candidate for bipolar disorder (BP). We agree that Dr Nyholt's analysis suggests that the association signals we detected between variants in the CACNA1C and BP may not be entirely independent. This should not, however, detract from the body of evidence now linking CACNA1C with BD or from the focus in our original article, that this candidate also represents a potential drug target (Keers et al. 2009).
In addition to the positive findings reported by Sklar et al. (2002, 2008), a large collaborative study combining data from three genome-wide association datasets identified a genome-wide significant association signal in CACNA1C (Ferreira et al. 2008). The top hit from this study has subsequently been associated with schizophrenia and major depression (Green et al. 2009) and, consistent with previous studies of both disorders and BP, shown to confer specific verbal fluency deficits in a population sample (Krug et al. 2009).
CACNA1C encodes the alpha subunit of the calcium channel Cav1.2. The association between calcium dysregulation and BP is well documented (Warsh et al. 2004), as is the overlap between BP and other ‘channelopathies’ such as migraine and epilepsy (Sheftell & Atlas, 2002). Moreover, drugs which affect interneuronal calcium ion activity by targeting Cav1.2 have been shown to be effective in the treatment of BP (Levy & Janicak, 2000).
CACNA1C remains a biologically plausible drug target associated with BP. More in-depth genetic and pharmacogenetic studies of CACNA1C and BP may yet provide a greater understanding of the aetiology and treatment of the disorder.