GENOME-WIDE ANALYSIS OF 102,084 MIGRAINE CASES IDENTIFIES 123 RISK LOCI AND SUBTYPE-SPECIFIC RISK ALLELES 179 8 Discussion We conducted the largest GWAS meta-analysis on migraine thus far by combining genetic data on 102,084 cases and 771,257 controls. We identified 123 migraine risk loci, of which 86 are novel since the previous migraine meta-analysis that yielded 38 loci.13 This shows that we have now reached the statistical power for rapid accumulation of new risk loci for migraine, in line with the progress of GWAS seen with other common diseases79, and as expected for a highly polygenic disorder like migraine.80 Migraine subtypes MO and MA were defined as separate disease entities some 30 years ago, and since then, the debate has continued as to what extent they are biologically similar. Over the years, arguments in favor6 and against5 have been presented, but convincing genetic evidence to support subtype-specific risk alleles has been lacking in genetic studies with smaller sample sizes.18, 81, 82 Here we increased considerably the evidence for subtype specificity of some risk alleles by including new migraine subtype data at the 123 migraine risk variants. We observed that, with a probability of > 95%, three lead variants (in HMOX2, in CACNA1A and in MPPED2) are associated with MA but not MO. Of them, CACNA1A is a well-known gene linked to familial hemiplegic migraine, a rare subform of MA.83, 84 The observation that CACNA1A seems involved in both monogenic and polygenic forms of migraine provides the first gene-based support for the increased sharing of common variants between the two disorders.80 We find no evidence that any of the seven loci, previously reported as GWS in MO but not in MA,13 would be specific for MO, while four of them (LRP1, FHL5, near FGF6 and near TRPM8) are among the nine loci shared by both subtypes with a probability over 95%. Loci (e.g., LRP1 and FHL5) that are strongly associated with both subtypes provide convincing evidence for a previous hypothesis that the subtypes partly share a genetic background.13, 85 In accordance with our analysis, effects in both subtypes were suggested before at the TRPM8 and TSPAN2 loci while, in contrast to our results, the LRP1 locus was previously reported to be specific for MO.81 Finally, we also detected four lead variants (including LRP1) that do not appear specific for MO but do confer a higher risk for MO than for MA. It has been long debated whether migraine has a vascular or a neuronal origin, or whether it is a combination of both.8, 9, 86, 87 Here we found genetic evidence for the role of both vascular and central nervous tissue types in migraine from several tissue enrichment analyses, which refined earlier analyses based on smaller sample sizes.13, 14 With respect to a vascular involvement in the pathophysiology of migraine, both gene expression and chromatin annotation data from LDSC-SEG showed that migraine signals are enriched for genes and cell type-specific annotations that are highly expressed in aorta and tibial and coronary arteries. The involvement of arteries was also proposed by our DEPICT tissue enrichment analysis. In addition, cardiovascular disease and blood pressure phenotypes were among the top
RkJQdWJsaXNoZXIy MTk4NDMw