Abstract Background and aims APOE ε4 carriers have a higher risk of intracerebral hemorrhage (ICH) than non-carriers. APOE ε4 also substantially increases the likelihood of amyloid-related imaging abnormalities and ICH in Alzheimer’s disease patients on anti-amyloid therapy. The molecular intermediaries of this genetic predisposition are unknown. We aim to identify proteins mediating the effect of APOE ε4 on ICH risk using proteogenomic integration. Methods Data from 52,560 UK Biobank participants (mean age 56.8, 46% female) were analyzed. The stepwise analysis comprised: (1) linear regression of APOE ε4 against 2,923 plasma proteins; (2) testing resulting proteins for ICH association; (3) two-sample Mendelian Randomization (MR) to assess causality; (4) mediation analysis to quantify their contribution to the APOE ε4-ICH relationship; (5) MR with other ICH-related outcomes. Analyses adjusted for age, sex, and genetic principal components and controlled for false-discovery rate. Results Of 11 proteins passing steps 1-2, only Paralemmin-1 (PALM) satisfied both MR (inverse-variance weighted OR 1.47, 95% CI 1.18-1.80) and mediation criteria. PALM accounted for 16% of the total effect of APOE ε4 on ICH risk (P = 0.007). In further MR analyses, PALM was also associated with microbleeds, white matter hyperintensities, and white matter integrity measures (fractional anisotropy, mean diffusivity). Conclusions This study pinpointed PALM as potential mediator linking APOE ε4 to ICH. PALM organizes neuronal and endothelial inner membrane cytoskeleton scaffold supporting junctional integrity and tissue resilience. Dysregulation of PALM-dependent scaffolding offers a mechanism for small vessel fragility and hemorrhage. These findings suggest PALM-centered pathways as actionable targets mitigating ICH risk in APOE ε4 carriers. Conflict of interest Shufan Huo: received funding from the German Research Foundation and NIH StrokeNet. Santiago Clocchiatti-Tuozzo: nothing to disclose. Srikant Rangaraju: nothing to disclose. Santosh Murthy: nothing to disclose. Seyedmehdi Payabvash: nothing to disclose. Cyprien A. Rivier: nothing to disclose. Guido J. Falcone: nothing to disclose. Figure 1 - belongs to Methods Figure 2 - belongs to Results
Huo et al. (2026) studied this question.