Proteome-Wide Mendelian Randomization Identifies Therapeutic Targets for Ischemic Stroke

Ischemic stroke carries substantial residual risk despite acute revascularization and secondary prevention, and small-vessel disease (SVD) lacks specific preventive therapy. Mendelian randomization (MR) using cis-acting protein quantitative trait loci (pQTLs) identifies circulating proteins whose lifelong modulation alters stroke risk.

We screened 589 conditionally independent cis-pQTLs from the INTERVAL study against four GIGASTROKE outcomes: any ischemic stroke (AIS), large-artery atherosclerosis (LAA), cardioembolic stroke (CES), and SVD. Wald ratio estimates were corrected by Benjamini-Hochberg false discovery rate (FDR) within each outcome, with a pooled cross-outcome correction as a sensitivity analysis. FDR-significant associations underwent colocalization (approximate Bayes factor; H4 > 0.80), replication in the independent deCODE pQTL cohort, cell-type expression validation in a human brain vascular single-nucleus atlas (65,479 nuclei), and phenome-wide safety screening.

Ten protein-outcome associations reached FDR < 0.05; nine replicated in deCODE and eight colocalized at H4 > 0.80, with two further pairs (ABO-LAA, cyclin H-CES) in the suggestive range (H4 0.70-0.80). For SVD, higher SPATA20 raised risk (OR 1.47, 95% CI 1.24-1.73, p = 5.3 × 10⁻⁶), and higher KDELC2 lowered risk (OR 0.82, 95% CI 0.74-0.91, p = 2.2 × 10⁻⁴). A protective cyclin H signal for CES passed outcome-wise FDR (q = 0.046) but not the pooled correction (q = 0.059) and is reported as preliminary. ABO, MMP-12, and SWAP70 confirmed earlier reports; SCARA5 failed colocalization and replication and was excluded. Colocalization confirmed the lead liabilities for MMP-12 (atopic dermatitis, two cohorts) and SWAP70 (rheumatoid arthritis) as on-target effects; two further liability associations (allergic asthma, ER-negative breast cancer) did not meet the colocalization threshold and remain hypothesis-generating.

Proteome-wide MR prioritizes two SVD-associated proteins: SPATA20 (risk-raising; now independently confirmed across screens) and KDELC2 (protective; newly reported). Both require replication on orthogonal proteomic platforms and functional validation before therapeutic development.