Abstract
Background: Migraine is a prevalent neurological disorder with a complex genetic background characterized by recurrent episodes of headache. The disease is subclassified into migraine with aura (MA) and migraine without aura (MO). Many association studies have been performed to date to identify genetic risk variants for common migraine, most of them focusing on selected candidate genes, with variable and often inconsistent results. Recently, a clinic-based genome-wide association study for migraine reported a functionally relevant risk variant (SNP rs1835740), involved in glutamate homeostasis, which showed a significant association with MA. We aimed to replicate this finding in a clinic-based study of a Spanish cohort with MA and MO patients.
Methods: We genotyped SNP rs1835740 in a Spanish sample of 1521 patients and 1379 screened controls and performed a case-control association study.
Conclusion: No association was found between the assayed SNP and any of the clinical groups considered.
Introduction
Migraine is an episodic neurological disorder that affects between 10% and 20% of the population, being around three times more prevalent in women than in men (1,2). Common forms of migraine present a complex genetic basis, with heritability values ranging from 33% to 57% and the likely participation of multiple genes that interact with environmental factors (1). The International Classification of Headache Disorders (ICHD) distinguishes two clinical forms of the disorder: migraine with aura (MA) and migraine without aura (MO). Both presentations share clinical features such as the presence of recurrent headaches, generally accompanied by photophobia, phonophobia, nausea and aggravation by physical activity. They are distinguished by the presence of aura, characterized by transient neurological symptoms that accompany the headache.
Genetic studies in migraine include genome-wide hypothesis-free approaches that aim to identify new susceptibility variants. To date, three genome-wide association studies (GWASs) have reported four variants associated with common migraine (3–5).
The first risk variant, rs1835740G > A (8q22), was identified in the European population in a clinic-based study carried out by the International Headache Genetics Consortium (IHGC) (3) in 2731 migraine cases and 10,747 controls from three countries, and replicated in four additional population cohorts, with a genome-wide significant overall meta-analysis p-value of 1.69 × 10−11. The ‘MA only’ group, including individuals that had suffered MA but not MO episodes, displayed the most significant association, as compared with the ‘Both MA and MO’ and ‘MO only’ groups. SNP rs1835740 is located near the MTDH gene (AEG-1, 3D3, LYRIC), encoding metadherin, and the CPQ gene (PGCP, LDP), encoding carboxypeptidase Q, both involved in the homeostasis of glutamate. eQTL analysis suggested that this genetic variation is a cis-regulator of MTDH expression in lymphoblastoid cell lines, with higher transcript levels associated with the risk allele (3). MTDH downregulates SLC1A2 (EAAT2, GLT-1) (6), encoding the solute carrier family 1 (glial high affinity glutamate transporter), member 2, thus leading to accumulation of glutamate in the synaptic cleft and potentially enhancing cortical spreading depression (CSD), the underlying mechanism of migraine aura (7).
Recently, three additional variants, rs2651899 (PRDM16, PR domain containing 16), rs10166942 (TRPM8, transient receptor potential cation channel, subfamily M, member 8) and rs11172113 (LRP1, low density lipoprotein receptor-related protein 1), were identified as nominally significant in a population-based GWAS from the Women’s Genome Health Study, consisting of around 5000 migraineur women and a control group of 18,000 non-migraineur women (4). Interestingly, these single nucleotide polymorphisms (SNPs) met genome-wide significance in a meta-analysis combining the discovery cohort and three replication sets (4): the Dutch Genetic Epidemiology of Migraine study (GEM), the German Study of Health in Pomerania (SHIP) and the IHGC discovery stage cohort used in the first migraine GWAS (3). None of the three associations was preferential for MA or MO. LRP1 modulates neuronal glutamate signalling, whereas TRPM8 has been related to neuropathic pain.
Finally, a meta-analysis of migraine GWAS was performed in a population-based study of six European cohorts, by the Dutch–Icelandic migraine genetics consortium (DICE), which included >2000 cases and >8000 controls (5). No genome-wide significant association was found, although 32 SNPs reached a p-value below 1 × 10−5. A replication study with a selection of 19 of these SNPs in two Dutch and one Australian cohort did not show any association.
We aimed to replicate the results of the only reported migraine GWAS that focuses on clinic-based samples, which rendered the first variant associated with common forms of migraine. We applied a clinic-based case-control association study design to test the rs1835740 SNP in a Spanish cohort of 1521 patients and 1379 non-migraineur controls.
Materials and methods
The sample consists of 1521 Caucasian Spanish patients (78.2% women) and 1379 ethnically-matched migraine-free controls (79.2% women), recruited after a direct clinical interview and a physical exam at Hospital Universitari Vall d’Hebron (Barcelona, Spain), Hospital Sant Joan de Déu (Manresa, Barcelona, Spain) and at Fundación Pública Galega de Medicina Xenómica (Santiago de Compostela, Spain) between 2002 and 2010. According to the International Classification for Headache Disorders: 2nd edition (8), patients were classified as MO (n = 951, 62.5%) or MA (n = 570, 37.5%). The average age at assessment was 41 years (standard deviation, SD = 14.12; range 5 to 86) for patients and 55 years (SD = 16.85; range 19 to 96) for controls. The study was approved by the local ethics committees and informed consent was obtained from all the participants or their legal representatives according to the Helsinki Declaration.
DNA was extracted from peripheral blood lymphocytes (9). Genotyping of the rs1835740 SNP was performed with a Taqman® SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA). Three CEPH samples were included in every plate and a 100% concordance with HapMap data was observed.
Power calculation for analysis under a dominant model, α = 0.05, lifetime risk of 1.3, disease prevalence of 0.15 and minor allele frequency (MAF) found in controls (0.241), performed with Genetic Power Calculator (pngu.mgh.harvard.edu/∼purcell/gpc) (10), displayed values of 95.5% in the MO group, 87.2% in the MA group and 98.5% in the ‘All migraine’ group. Statistical analysis for the association study was carried out using the SNPassoc R library (11). First, Hardy–Weinberg equilibrium (HWE) was tested in our control sample, which was found to be in equilibrium (p = 0.12). Then, both allele and genotype frequencies were compared between cases and controls, considering codominant, dominant, recessive, overdominant and log-additive models and adjusting by sex. Three main phenotypic groups were considered in the analysis: MO, MA and ‘All migraine’. Additionally, in order to better match the comparisons performed in the original GWAS that we aimed to replicate, the MA group was divided into two subcategories: ‘MA only’, which includes MA patients that suffered mostly MA episodes, and ‘Both MA and MO’, including MA patients that suffered MA and MO episodes.
Results and discussion
Association study of single nucleotide polymorphism (SNP) rs1835740 with several migraine phenotypes in 1521 cases and 1379 screened controls from Spain.
MO: migraine without aura; MA: migraine with aura; MAF: minor allele frequency; Ca: cases; Co: controls; OR: odds ratio; CI: confidence interval.
Number of individuals (percentage).
These subgroups are the result of splitting the MA category (see Materials and methods).
Several reasons may explain the discrepant results between our study and the original one by Anttila et al.: 1) Differences in the study design: first, in the original GWAS, the male:female ratio was different in cases (1:4) and controls (1:1), whereas in our study it was approximately 1:4 in the two groups; second, the controls in the present study were negative for migraine history, whereas they were population-based in the original work; third, the MAF values reported in the two studies are comparable for the cases, but they differ in controls; and fourth, sample size is larger in the GWAS (about 6000 cases and 50,000 controls) compared with the present study (1500 cases and 1300 controls), although the statistical power of our sample exceeds 87%. 2) It is also possible that, the original association being correct, the rs1835740 may not be the actual susceptibility variant, although the fact that it seems to be functionally relevant (3) and its definition as a singleton SNP according to CEU HapMap data (www.hapmap.org, release #28) work against this possibility. 3) Finally, the associated genetic variant may exert population-specific effects depending on genetic/epigenetic background or variable environmental exposures.
Although divergent from the original GWAS (3), our results are in line with previous attempts to replicate the initial finding. Thus, a GWAS in a population-based cohort of migraineurs from the Women’s Genome Health Study failed to identify association of rs1835740 with overall migraine (p = 0.22) nor with MA or MO (4). The inspection of SNPs covering the entire genomic region did not reach locus-wide significance either. In a second GWAS (5), rs1835740 and two nearby SNPs (rs982502, rs2436046) did not show significant association with migraine in a meta-analysis of six cohorts (p = 0.64). An association study of SNPs within the CPQ and MTDH genes, located proximal and distal from rs1835740, respectively (Figure 1), identified a gene-based association with MTDH that overcame the Bonferroni correction for multiple testing (p = 0.002) (5). Furthermore, the Dutch replication cohort inspected in the original GWAS (3) also displayed negative results (p > 0.25 with ‘All migraine’, ‘MA only’ and ‘Both MA and MO’), whereas there was association in the combined discovery sample from Germany, Finland and the Netherlands (p = 5.38e-09, ‘All migraine’), and nominal associations in three out of four replication samples (p-values between 0.010 and 0.038, with different migraine phenotypes). Finally, a recent study tested pre-existing genotype data from one of the replication cohorts of the original GWAS for association between rs1835740 and several clinical features of MA, displaying non-significant results (12).
Location of rs1835740 on chromosome 8q22.1, close to the genes CPQ (carboxypeptidase Q), TSPYL5 (TSPY-like 5) and MTDH (metadherin). Arrows indicate direction of transcription. Allelic variation of the rs1835740 SNP modulates the transcriptional activity of MTDH. Adapted from ENSEMBL (www.ensembl.org, release 63, June 2011).
It is possible that the discordant results of the studies by Chasman et al. (4) and Ligthart et al. (5) with respect to the first migraine GWAS (3) may relate to differences in migraine ascertainment, as the original study used a clinic-based cohort, whereas the subsequent works focused on migraineurs from the general population. In this regard, our study is the only one that uses a clinic-based cohort to inspect the genome-wide association signal on rs1835740, with negative results. The patient samples, as described above, were recruited from three Spanish hospitals, and the controls, originating from the same geographical areas as the patients, were screened for migraine and lacked any family history of migraine in their first degree relatives.
In sum, although the association described between rs1835740 and MA in the original clinic-based GWAS remains a robust finding, sustained by a meta-analysis in seven cohorts (3), our study displays a clear negative result in the Spanish population. The in-depth inspection of the MTDH genomic region in other clinic-based cohorts is warranted to gain more insight into the possible involvement of this gene in migraine.
Footnotes
Funding
This work was supported by Ministerio de Ciencia e Innovación (grant nos SAF 2009-13182-C01 and SAF 2009-13182-C03) and Agència de Gestió d'Ajuts Universitaris i de Recerca (grant nos 2009SGR-078, 2009SGR0971). MJS and PC received support for this work from the Ministerio de Ciencia y Tecnología (SAF2005-07978), Consellería de Educación-Xunta de Galicia (REGENPSI network 2009/019), Fundación Pedro Barrié de la Maza (DIANA program) and FEDER funds.
Acknowledgements
We are grateful to patients and controls for their participation in the study. We thank R Corominas, E Cuenca-León and M Vila for technical assistance.