Rare variants with large effects provide functional insights into the pathology of migraine subtypes, with and without aura

Abstract

Migraine is a complex neurovascular disease with a range of severity and symptoms, yet mostly studied as one phenotype in genome-wide association studies (GWAS). Here we combine large GWAS datasets from six European populations to study the main migraine subtypes, migraine with aura (MA) and migraine without aura (MO). We identified four new MA-associated variants (in PRRT2, PALMD, ABO and LRRK2) and classified 13 MO-associated variants. Rare variants with large effects highlight three genes. A rare frameshift variant in brain-expressed PRRT2 confers large risk of MA and epilepsy, but not MO. A burden test of rare loss-of-function variants in SCN11A, encoding a neuron-expressed sodium channel with a key role in pain sensation, shows strong protection against migraine. Finally, a rare variant with cis-regulatory effects on KCNK5 confers large protection against migraine and brain aneurysms. Our findings offer new insights with therapeutic potential into the complex biology of migraine and its subtypes.

Fig. 1. Manhattan plot of GWAS meta-analysis results for all studied phenotypes.

The graph shows data for migraine (n_case/control = 74,495/1,259,808), MA (n_case/control = 16,603/1,336,517), MO (n_case/control = 11,718/1,330,747), VD (n_case/control = 30,297/86,134) and BRH (n_case/control = 51,803/123,732). See Supplementary Table 1 for n_case/control for each cohort. On the x axis, variants are plotted along the 22 autosomes and the X chromosome. On the y axis is the statistical significance of their association with the respective phenotypes from meta-analyses using a fixed-effects inverse-variance method based on effect estimates and s.e. under the additive model, in which each dataset was assumed to have a common OR but allowed to have different population frequencies for alleles and genotypes. Gray dots are not significant variants. Variant associations that reach the P threshold weighted by variant annotation are represented by color-coded dots. Adjacent chromosomes are presented in different shades of gray. Known migraine loci are represented by gene names in black text, and new loci are represented by gene names in blue text.

Fig. 2. Effects of SNPs associated with self-reported headache-related VD in clinically defined MA, overall migraine and MO.

The x axis (VD, n_case/control = 30,297/86,134) and the y axis (MA, n_case/control = 16,603/1,336,517; migraine, n_case/control = 74,495/1,259,808 and MO, n_case/control = 11,718/1,330,747) show the logarithmic estimated odds ratios, log(OR), for the associations with the respective phenotypes from meta-analyses using a fixed-effects inverse-variance method based on effect estimates and s.e. under the additive model, in which each dataset was assumed to have a common OR but allowed to have different population frequencies for alleles and genotypes. All effects are shown for the VD risk allele, and black crosses indicate 95% CIs. The dashed red lines represent slope (s.d.) based on a simple linear regression through the origin using 1/s.e. as weights. Effect estimates are 73%, 29% and 0% of VD effect estimates for MA, migraine and MO, respectively.

Fig. 3. Subtype classification of lead variants.

Effect plots for all lead variants except the MA variant in PRRT2. Effects are from meta-analyses using a fixed-effects inverse-variance method based on effect estimates and s.e. under the additive model, in which each dataset was assumed to have a common OR but allowed to have different population frequencies for alleles and genotypes. Data are presented as additive effect estimates (center) with 95% CI (crosses) for the annotated variants. a, Axes show logarithm of odds ratios (log(OR)) for MO (x axis; n_case/control = 11,718/1,330,747) and MA (y axis; n_case/control = 16,603/1,336,517). b, Axes show MO (x axis; n_case/control = 11,718/1,330,747) and VD (y axis; n_case/control = 30,297/86,134). log(OR) is calculated for the effect allele. The effects of variants that have been colored and annotated with gene names differ between the migraine subtypes at a significance threshold of 0.0012 = 0.05/43. The 95% CIs for the log(ORs) are shown for annotated variants. Effects are adjusted with sample overlap (r_ij) estimated from counts of cases, controls and the counts of overlaps in these groups between phenotypes from all cohorts except FinnGen (for which we only have summary statistics). The parameter representing sample overlap between MO and MA is r_ij = 0.023 and MO and VD is r_ij = 0.012. Dashed lines show the coordinate axes, the diagonal and a line through the origin with slope = 1 (Methods; see Supplementary Tables 13 and 14 and Supplementary Fig. 4 for VD versus MA plot).

Fig. 4. Rare variant rs72854118 in regulatory region targeting KCNK5 associates with BRH.

Effect–effect plot of clinically defined migraine (n_case/control = 74,495/1,259,808) vs. self-reported BRH (n_case/control = 51,803/123,732) effects for 42 lead variants identified in this study (excluding high-impact variants in PRRT2 and A3GALT2; see Supplementary Table 7 for their associations with the respective phenotypes). Effects are from meta-analyses using a fixed-effects inverse-variance method based on effect estimates and s.e. under the additive model, in which each dataset was assumed to have a common OR but allowed to have different population frequencies for alleles and genotypes. The x axis and the y axis show the logarithmic estimated ORs for the associations with the respective phenotypes. Error bars represent 95% CI. The dashed red lines represent slope (s.d.) based on a simple linear regression through the origin using 1/s.e. as weights. Cohort descriptions are in Supplementary Table 1. Variants are colored according to their primary associations in this study. The red dot outlier depicts the variant rs72854118-G near KCNK5, its effects on BRH exceeding its effects on all migraine. Pheno, phenotype; Migr, migraine.