A genome-wide association study of suicide severity scores in bipolar disorder

Highlights

• We performed a genome-wide association study (GWAS) of suicide behaviour severity.

• Our GWAS was conducted in three samples of bipolar disorder patients.

• Suggestive associations were found in chromosomes 8p23-p24 and 10p11.

Abstract

Background

Suicide claims one million lives worldwide annually, making it a serious public health concern. The risk for suicidal behaviour can be partly explained by genetic factors, as suggested by twin and family studies (reviewed in (Zai et al. 2012)). Recently, genome-wide association studies (GWASs) of suicide attempt on large samples of bipolar disorder (BD) patients from multiple sites have identified a number of novel candidate genes. GWASs of suicide behaviour severity, from suicidal ideation to serious suicide attempt, have not been reported for BD.

Methods

We conducted a GWAS of suicide behaviour severity in three independent BD samples:212 small nuclear families with BD probands from Toronto, Canada, 428 BD cases from Toronto, and 483 BD cases from the UK. We carried out imputation with 1000 Genome Project data as reference using IMPUTE2. Quality control and data analysis was conducted using PLINK and R. We conducted the quantitative analyses of suicide behaviour severity in the three samples separately, and derived an overall significance by a meta-analysis using the METAL software.

Results

We did not find genome-wide significant association of any tested markers in any of the BD samples, but we found a number of suggestive associations, including regions on chromosomes 8 and 10 (p < 1e-5).

Conclusions

Our GWAS findings suggest that likely many gene variants of small effects contribute collectively to the risk for suicidal behaviour severity in BD. Larger independent replications are required to strengthen the findings from the GWAS presented here.

Introduction

Suicide claims 1 million lives each year worldwide, and for each completed suicide, there are twenty suicide attempts, making it an important public health issue. Over 90% of suicide victims have at least one psychiatric diagnosis, including bipolar disorder (BD) (Mann, 2002), where as much as 8% of BD followed for up to 40 years committed suicide (Angst et al., 2002, Nordentoft et al., 2011).

Suicide has a prominent genetic component (reviewed in (Zai et al., 2012)). Suicide attempts tend to occur more often within families (Brent et al., 2002, Johnson et al., 1998). Greater concordance was observed between monozygotic twins than between dizygotic twins (Roy and Segal, 2001, Statham et al., 1998). The concordant phenotype includes both completed and attempted suicides (Roy et al., 1995). A review of twin studies estimated the heritability of suicidal behaviour to be up to 55% (Voracek and Loibl, 2007).

A number of linkage studies have been conducted on suicide starting with Zubenko and coworkers (Zubenko et al., 2004). Their findings on the short arm of chromosome 2 were later replicated in 162 BD families (Willour et al., 2007), and more recently confirmed in the regional linkage study meta-analysis of 2p12 (Butler et al., 2010). Recent technological advances have permitted the high-throughput genotyping of hundreds of thousands of single-nucleotide polymorphisms across the genome. While no genome-wide significant findings (at significance levels of less than 5 × 10⁻⁸) have been reported to date, a number of suggestive findings have emerged (Perroud et al., 2011, Schosser et al., 2011). Recently, a genome-wide association study (GWAS) was reported on samples of 2698 BD patients of which 1201 had a previous suicide attempt. After meta-analysis of markers with p < 1 × 10⁻³ from their discovery sample (GAIN, TGEN, German) with their replication BD sample (STEP-BD, WTCCC, UCL), the most significantly associated marker was rs300774 in an intergenic region at chromosomal region 2p25, which contains the SH3YL1, ACP1, and FAM150B genes. The association finding was supported by post-mortem prefrontal cortical gene expression analysis, where suicide completers were found to have significantly higher ACP1 expression than non-suicide victims (Willour et al., 2012). The strongest association signal from another report of a GWAS of suicide attempt on the BD (STEP-BD, WTCCC, UCL) came from the intergenic chromosome 10 marker rs1466846; this finding was not replicated in the replication sample (GAIN, TGEN, German) (Perlis et al., 2010). Part of the reason for both lack of a strong association signal and robust replication could be that the samples are underpowered for analysis due to dichotomizing of the suicide attempt as the outcome variable. A GWAS on suicidality scores, which are derived from the Schedule for Clinical Assessment in Neuropsychiatry (SCAN) interview, was conducted with a major depression sample from the RADIANT study (Schosser et al., 2011). The suicidality score captures suicide severity from suicide ideation to attempt. The most significant findings from the RADIANT sample failed to replicate in the German replication sample.

In the present study, to address the issue of power with the dichotomous variable, we carried out a similar GWAS using the quantitative variable of suicide severity in three samples of BD patients.