Unveiling migraine subtype heterogeneity and risk loci: integrated genome-wide association study and single-cell transcriptomics discovery.

Abstract

Background: Migraine, a debilitating neurological disorder with distinct subtypes (migraine with aura [MA] and migraine without aura [MO]), exhibits genetic and spatial heterogeneity that remains poorly understood. While genetic correlations between subtypes are established, spatially resolved molecular mechanisms driving their divergent clinical phenotypes-particularly in tissue microenvironments-are unclear, limiting targeted therapeutic development.

Methods: We integrated genome-wide association study (GWAS) data from FinnGen R11 and international cohorts with transcriptomic, epigenomic, and spatially resolved single-cell spatial transcriptomics (sc-ST) profiles. Genetic correlations and functional annotations were assessed using Linkage Disequilibrium Score Regression (LDSC), High-Definition Likelihood (HDL), and partitioned heritability analyses. A multi-omics framework combined Summary Mendelian Randomization (SMR) for expression and methylation quantitative trait loci (eQTL/mQTL), Functional Summary-based Imputation (FUSION), Multi-marker Analysis of GenoMic Annotation (MAGMA), Joint-Tissue Imputation Enhanced PrediXcan Analysis (JTI-PrediXcan), and the Polygenic Priority Score (PoPS) to systematically prioritize genes based on methodological robustness (≥ 2 analytical approaches) and cross-subtype consistency. Tissue-enriched specificity was validated via genetically informed spatial mapping of cells for complex traits (gsMap), a novel algorithm integrating sc-ST and GWAS data to map subtype-associated cellular architectures at single-cell resolution across embryonic tissues.

Results: LDSC and HDL confirmed strong genetic correlations between MA and MO. But they showed divergent functional architectures in functional genomic annotations, with MA enriched in conserved regulatory elements (e.g., Backgrd_Selection_StatL2_0, enrichment = 1.38, P = 5.47 × 10^-6) and MO in vascular pathways (e.g., GERP.NSL2_0, enrichment = 2.12, P = 1.04 × 10^-6). Sc-ST revealed spatially divergent niches: MA showed prenatal enrichment in neural crest-derived tissues (jaw primordium, p = 0.0039) and hypothalamic microglial adjacencies, aligning with neuroimmune regulation, while MO exhibited peripheral tropism in vascular smooth muscle and gut-brain interfaces, corroborated by LDSC-SEG/MAGMA vascular pathways. Multi-omics integration identified high-confidence cross-subtype genes (LRP1 [PoPS: Overall = 3.67, MO = 0.80], PHACTR1 [PoPS: Overall = 2.65, MA = 0.33, MO = 1.28], STAT6 [PoPS: Overall = 3.00, MO = 2.29], RDH16, TTC24, ZBTB39, FHL5, MEF2D, NAB2, UFL1, and REEP3) supported by ≥ 2 methods. Subtype-specific genes included MA-associated neuronal regulators (CACNA1A, KLHDC8B) and MO-specific vascular/metabolic genes (e.g., ACO2, BCAR1, CCDC134).

Conclusion: Our study delineates spatially constrained mechanisms underlying migraine heterogeneity: MA arises from neuroimmune-epigenetic dysregulation, while MO is driven by vascular-metabolic perturbations. Key genes and pathways provide actionable targets for subtype-specific therapies. By bridging genetic architecture with spatial biology, we redefine migraine pathogenesis and precision intervention strategies.