Dynamic Expression Quantitative Trait Locus Mendelian Randomization in Activated CD4-Positive T Cells Nominates SMDT1 and CMTM7 as Candidate Genes in Pancreatic Cancer.

Pancreatic ductal adenocarcinoma (PDAC) is largely refractory to immune checkpoint blockade. Conventional Mendelian randomization (MR) using steady-state expression quantitative trait loci (eQTLs) may miss regulatory effects that emerge during CD4 + T-cell activation. We used a two-stage genetic framework. Fixed-effect meta-analysis across four PDAC GWAS served as an exploratory cross-cohort discovery screen, followed by GCST90041814 as the ancestry-matched European primary analysis. Associations were assessed by colocalization, summary-data-based MR with HEIDI testing, and DICE/eQTLGen replication. Paired sample-level pseudobulk analysis evaluated tumor-adjacent CD4 + T cells. Bulk transcriptomic, mediation, virtual perturbation, survival and phenome-wide analyses were treated as exploratory. The pooled discovery screen identified 67 genes at FDR < 0.05; 65 also showed concordant effect directions across all four cohorts and constituted the main discovery set. In the European primary analysis, 66 of 1247 genes had nominal p < 0.05, but none survived cohort-wide FDR correction. SMDT1 (European OR = 0.603, 95% CI 0.412-0.884, p = 0.009, FDR = 0.754) and CMTM7 (European OR = 1.208, 95% CI 1.026-1.424, p = 0.023, FDR = 0.754) retained directionally consistent nominal associations and were prioritized as exploratory candidates. In six paired GSE212966 samples, pseudobulk changes were directionally consistent but not significant (SMDT1 log2FC = -0.381, p = 0.250; CMTM7 log2FC = 0.493, p = 0.199). Dynamic CD4 + T-cell eQTL MR nominates SMDT1 and CMTM7 as candidate genes associated with PDAC susceptibility. The pooled discovery evidence and directionally consistent nominal European associations support prioritization, but no gene survived FDR correction in the European analysis. Transcriptomic and downstream functional findings remain exploratory and require experimental validation.