Genes, Brain and Behavior (2010) 9: 673–680
doi: 10.1111/j.1601-183X.2010.00601.x
A genome-wide association study of bipolar disorder and comorbid migraine K. J. Oedegaard∗,†,‡,§,1 , T. A. Greenwood†,1 , S. Johansson¶,∗∗ , K. K. Jacobsen¶,∗∗ , A. Halmoy¶ , O. B. Fasmer‡,§ , H. S. Akiskal†,††,1 The Bipolar Genome Study (BiGS)1 , J. Haavik¶,‡‡ and J. R. Kelsoe†,††,1 † Department of Psychiatry, University of California San Diego, La Jolla, CA, USA, ‡ Department of Clinical Medicine, Section for Psychiatry, University of Bergen, § Moodnet, Division of Psychiatry, Haukeland University Hospital, ¶ Department of Biomedicine, University of Bergen, ** Center for Medical
Genetics and Molecular Medicine, Haukeland University Hospital, Bergen, Norway, †† Department of Psychiatry, Veterans Affairs San Diego Healthcare System, La Jolla, CA, USA, and ‡‡ Division of Psychiatry, Haukeland University Hospital, Bergen, Norway 1 Bipolar
Genome Study Coauthors are listed in the Appendix.
*Corresponding author: K. J. Oedegaard, MD, PhD, Division of Psychiatry, Haukeland University Hospital, 5021 Bergen, Norway. E-mail:
[email protected]
Both migraine and bipolar affective disorder (BPAD) are complex phenotypes with significant genetic and nongenetic components. Epidemiological and clinical studies have showed a high degree of comorbidity between migraine and BPAD, and overlapping regions of linkage have been shown in numerous genome-wide linkage studies. To identify susceptibility factors for the BPAD/migraine phenotype, we conducted a genomewide association study (GWAS) in 1001 cases with bipolar disorder collected through the NIMH Genetics Initiative for Bipolar Disorder and genotyped at 1 M single-nucleotide polymorphisms (SNPs) as part of the Genetic Association Information Network (GAIN). We compared BPAD patients without any headache (n = 699) with BPAD patients with doctor diagnosed migraine (n = 56). The strongest evidence for association was found for several SNPs in a 317-kb region encompassing the uncharacterized geneKIAA0564 {e.g. rs9566845 [OR = 4.98 (95% CI: 2.6–9.48), P = 7.7 × 10−8 ] and rs9566867 (P = 8.2 × 10−8 )}. Although the level of signficance was significantly reduced when using the Fisher’s exact test (as a result of the low count of cases with migraine), rs9566845 P = 1.4 × 10−5 and rs9566867 P = 1.5 × 10−5 , this region remained the most prominent finding. Furthermore, marker rs9566845 was genotyped and found associated with migraine in an independent Norwegian sample of adult attention deficit hyperactivity disorder (ADHD) patients with and without comorbid migraine (n = 131 and n = 324, respectively), OR = 2.42
(1.18–4.97), P = 0.013. This is the first GWAS examining patients with bipolar disorder and comorbid migraine. These data suggest that genetic variants in the KIAA0564 gene region may predispose to migraine headaches in subgroups of patients with both BPAD and ADHD. Keywords: ADHD, bipolar disorder, GAIN, KIAA0564, migraine, whole-genome association study
Received 2 March 2010, revised 1 May 2010 and 3 May 2010, accepted for publication 6 May 2010
Bipolar affective disorder (BPAD [MIM 125480]) affects about 1% of the population, increasing morbidity and mortality (APA 2000). It is characterized by episodes of depression and mania, and there is strong support for a genetic basis with heritability of at least 60% (Craddock & Jones 1999; Smoller & Finn 2003). Although BPAD is highly heritable, the identification of specific genetic variations has yielded limited findings (Baum et al . 2008a,b; Sklar et al . 2008; WTCCC 2007), and subgrouping patients with BPAD according to clinical subphenotypes has been suggested as a fruitful approach for further genetic studies in BPAD (McQueen et al . 2005). Migraine (MIM 157300) affects approximately 18% women, 6% men and 4% children, and is characterized by recurrent headache, nausea, emesis, phonophobia and photophobia (Headache classification committee 2004). The heritability in migraine is estimated between 40% and 65% (Gervil et al . 1999; Honkasalo et al . 1995; Larsson et al . 1995). Linkage studies have not been consistent, and no genes associated with common forms of migraine have been recognized (Fasmer et al . 2009a; Oedegaard et al . 2010). However, three genes for the autosomal dominant disorder Familial Hemiplegic Migraine (FHM1–3, MIMs 141500, 602481 and 609634, respectively) have been identified. These genes are involved in ion-transport: CACNA1A (MIM 601011) (Ophoff et al . 1996), SCN1A (MIM 182389) (Dichgans et al . 2005) and ATP1A2 (MIM 182340) (De Fusco et al . 2003), and the identified mutations can lead to increased extra cellular glutamate and potassium levels, and through that mechanism affect neuronal membrane thresholds and cause hyperexcitability of neurons (van de Ven et al . 2007). There is a substantial comorbidity between migraine and bipolar disorders documented in clinical and epidemiological studies. Both disorders have been linked to disturbances in the serotonergic, dopaminergic and glutamatergic systems, and alterations in ion channels may be involved in the pathophysiology of migraine and bipolar disorders (Fasmer
© 2010 The Authors
Genes, Brain and Behavior © 2010 Blackwell Publishing Ltd and International Behavioural and Neural Genetics Society
673
Oedegaard et al.
et al . 2009a; Oedegaard et al . 2009). Population studies have showed that there is a two- to threefold increased prevalence of migraine in patients with bipolar disorders (Hirschfeld et al . 2003; McIntyre et al . 2006). The relationship might be particularly strong in patients with bipolar II disorder (Fasmer 2001; Low et al . 2003), and an increased risk of having migraine has been associated with having a family history of bipolar disorder (Dilsaver et al . 2009). Some pharmacological treatments are successful in the prevention of both disorders, most notably valproate (Bowden et al . 2000; Mulleners & Chronicle 2008). We recently used BPAD with comorbid migraine as an alternative phenotype definition in a genome-wide linkage study (a reanalysis of the NIMH Bipolar Genetics Initiative wave 4 data set) (Oedegaard et al . 2009) and identified a locus on chromosome 20p11 with overlapping elevated LOD scores for both migraine (LOD = 1.95) and BPAD (LOD = 1.67) phenotypes. In order to validate our results, we used genetic information from an available sample of patients with attention deficit hyperactivity disorder (ADHD). This a neuropsychiatric disorder strongly associated with bipolar disorder (Halmoy et al . 2010), and the sample had also been assessed for migraine headache. We hypothesized that the BPAD/migraine symptom complex could be a meaningful alternative phenotype/endophenotype for the identification of genetic susceptibility regions in a genome-wide association study (GWAS). To the best of our knowledge, this study is also the first GWAS in migraine in any sample (National Human Genome Research Institute catalogue of published GWASs: http:// www.genome.gov/gwastudies/ (updated 16 February 2010).
Methods and materials The GAIN study The study population and subject characteristics are described elsewhere (Smith et al . 2009). In brief, this is a genome-wide association case–control study including an European American sample of 1001 bipolar cases and 1033 controls. The final dataset (after study subjects, genotyping and quality control) consisted of 724 067 single-nucleotide polymorphisms (SNPs). The patients were recruited at 11 data collection sites across the United States. Patients were interviewed with the Diagnostic Interview for Genetic Studies (DIGS) (APA 2000; Nurnberger et al . 1994). Medical records were obtained, and a consensus final best estimate diagnosis was made using the Diagnostic and Statistical Manual of Mental Disorders-IV criteria. (APA 2000).
Bipolar cases Bipolar cases were selected from those collected and characterized by the NIMH Genetics Initiative for Bipolar Disorder Consortium over the past 18 years. These subjects were collected in 5 waves. Waves 1–4 were families collected for linkage studies, whereas wave 5 was a large set of primarily unrelated cases collected for largescale association studies. The study sample includes 175 unrelated subjects from waves 1 and 2 and 396 unrelated subjects from waves 3 and 4. In addition, the sample includes 430 subjects from the wave 5 data collection. The subjects abstracted from waves 1–5 totalled 1041 individuals. Forty subjects were removed because of a non-BPAD I/SABP (Schizo-Affective Bipolar Disorder) best estimate diagnosis or low diagnostic confidence. The final sample included 1001 cases of which 951 had a diagnosis of BPAD 1 and 50 had a diagnosis of SABP.
674
Migraine cases In waves 1–2 information regarding doctor diagnosed migraine was not assessed. Patients were asked whether they suffered from migraine headaches, but not whether the diagnosis had been affirmed by a medical doctor. In waves 3–5, all patients were asked whether they suffered from migraine headaches and also if the diagnosis had been confirmed by a medical doctor. As there is a substantial overlap between symptoms of tension type headache and migraine headaches (Davidoff 2002), we only included those patients who reported that their diagnosis had been verified by a medical doctor in the current analyses. Furthermore, in order to avoid possible contamination of migraine headaches by other types of headache, all patients with self-reported migraine only were excluded from this GWAS.
Genotyping Genotyping was carried out by The Broad Institute Center for Genotyping and Analysis (http://www.broad.mit.edu/node/306), using the Affimetrix Genome-Wide Human SNP Array 6.0 (Mccarroll et al . 2008). Samples for which fewer than 86% of the quality control (FQC) SNPs produced genotypes were rerun. Allele calling was performed using the BIRDSEED algorithm (Korn et al . 2008) AFFYMETRICS POWER TOOLS version apt-1.8.6 (http://www.affymetrix.com/partners programs/programs/developer/ tools/powertools.affx#1 3) and cluster models file (http://www. broad.mit.edu/ dbmirel/ncbi/affy6.0 birdseed 1.31.priors.tsv). Further quality control was performed on SNPs to remove duplicate samples, poorly genotyped and/or contaminated samples, and poorly performing SNPs, as part of standard quality control measures used for the whole-genome association study of the same sample (Smith et al . 2009). The genotype data are available through dbGaP (phs000017.v1.p1).
Statistical analyses All genetic analyses were conducted using PLINK version 1.07 (Purcell et al . 2007) for the BPAD with migraine vs. BPAD without migraine phenotypes. The distribution of expected P values under the null hypothesis (i.e. no association) and genomic inflation value (λ) was calculated in PLINK using the adjust command.
Power and statistical significance
In the current study, we have employed a significance of 5 × 10−8 threshold, as this has been commonly used in previous GWA studies. One region had two SNPs that were close to this threshold, but because the strongest test of association lies in follow-up replication studies we conducted a study in order to replicate this BPAD/migraine finding in an available sample of patients with ADHD and comorbid migraine. In addition, we list all SNPs with P < 10 × 10−5 (Table S2).
The ADHD study The patients were recruited in a genetic study using a national registry of adults diagnosed with ADHD in Norway during 1997 through 2005, as recently described (Halmoy et al . 2010). The diagnostic evaluation of the patients in the registry was made by one of three national expert commissions for ADHD/hyperkinetic disorder by systematic assessment of information records (including information from informants) provided by the referring psychiatrists. The diagnosis of ADHD was made according to the ICD-10 research criteria. In addition, to increase recruitment and to also include patients diagnosed later than May 2005, clinicians nationwide were asked to recruit formally diagnosed adult patients with ADHD. These patients were assessed by specialists in clinical psychiatry according to national guidelines based on the criteria described above, though without the mandatory evaluation of the committees yielding a total sample of 466 adults with persistent ADHD diagnosis. Information on comorbid migraine and bipolar disorder was extracted from questionnaire data obtained in the primary study. In the ADHD sample only self-reported migraine data were available. Genomic DNA was extracted either from whole blood or from saliva using the Oragene™ Genes, Brain and Behavior (2010) 9: 673–680
GWAS of bipolar disorder/migraine and KIAA0564 DNA Self-Collection Kit from DNA Genotek (DNA Genotek, Inc., Ontario, Canada) at the HUNT biobank (Levanger, Norway).
without migraine (58.8% vs. 42.9%), and also here the mean age was similar in both genders. In the ADHD sample, 11 subjects were excluded as as result of genotyping errors.
BPAD and migraine comorbidity in the ADHD study A lifetime history of bipolar disorder was reported by 12.3% of the adult ADHD patients and 1.7% of controls (P < 0.001). Bipolar disorder in first degree family members was reported by 11.3% and 2.7% of patients and controls, respectively (P < 0.001). Using the Mood Disorder Questionnaire (MDQ), 50.6% of the ADHD patients screened positive for a bipolar spectrum disorder (BSD)(Halmoy et al . 2010). Among adult ADHD patients, 27.9% reported a lifetime history of migraine compared with 18.6% in the control group (P = 0.001). Migraine in first degree family members was reported by 35.2% of the patients and 28.1% of controls (P = 0.001). The ADHD study included 466 patients with ADHD and 133 of these had a comorbid migraine diagnosis. This represented the replication sample.
Genotyping Marker rs9566845 which tagged the entire associated haplotype in the GWA-analysis was selected for replication in the ADHD samples. Genotyping was carried out by the multiplex MassARRAY®iPLEX ™ System (SEQUENOM, Inc., San Diego, CA, USA) at the Norwegian national technology platform CIGENE as part of a bigger ADHDgenotyping study including 55 markers. Genotyping call rate in the 484 patients was 99.6% and concordance rate was 100% (38 duplicates) for marker rs9566845. A total of 18 ADHD patients were subsequently excluded based on strict QC measures in the total data set (>10% missing genotypes) yielding a final data set of 466 ADHD patients.
Statistical analyses Data analyses were performed using the PLINK software, and the single point allelic association test was performed with the χ 2 test and Fisher’s exact test.
Results Descriptive statistics of the BPAD and ADHD samples Demographic variables for the study population, for the migraine phenotypes and for the replication sample are presented in Table 1. The study population included 699 individuals without migraine and 56 individuals who reported having a doctor diagnosed migraine; 233 individuals were excluded from the study as they reported having headache that had not been diagnosed by a doctor. The demographic data showed that the proportion of women was significantly higher in the BPAD/migraine group compared with the group without migraine (76.8% vs. 43.9%). The mean age was similar for both groups in both men and women. In the ADHD group, the proportion of women was also higher in the ADHD/migraine group compared with the ADHD group
BPAD GWA analyses Figure 1 is a Manhattan plot of all SNPs in the analyses, displaying the P values of the comparisons between the BPAD with migraine and those without migraine. SNPs with P values below 10 × 10−5 (blue line) are highlighted (117 SNPs). We report a list of these top associated SNPs in Table S1. The SNP with the lowest P value (rs11031481; P < 9.4 × 10−9 ) is located in a gene desert and surrounding SNPs were not genotyped in the current study. The most noticeable peak was located in a 317-kb region on chromosome 13 with several SNPs approaching the proposed 5 × 10−8 threshold for GWA-significance, e.g. rs9566845 [OR = 4.98 (95% CI: 2.6–9.48), P = 7.7 × 10−8 ] and rs9566867 (P = 8.2 × 10−8 ), with additional support from numerous surrounding SNPs. The rare allele of rs9566845 was significantly more frequent in patients with migraine headache compared with patients without migraine headaches (12.5% vs. 2.8%). This region harbours an uncharacterized gene, KIAA0564 (chr13:41,191,473-41,433,221) (Fig. 2). An in-depth analyses of this region showed that the association is best explained by one relatively rare haplotype which involves the top 10 associated SNPs of the region (see Tables S4–S6 and Fig. S1). The haplotype has an allele frequency of 2.3% among bipolar patients without migraine and 12.7% of patients with doctor diagnosed migraine (the haplotype is indicated by a blue line in Fig. 2). This haplotype spans at least 315 kb and covers the entire associated region and is fully tagged by markers rs9566845 and rs9566876. Furthermore, we include a report on all chromosomal regions that had at least two SNPs with P values
