Inferring genetic associations of programmed cell death genes with retinitis pigmentosa through multimodal Mendelian randomization

BACKGROUND: Retinitis pigmentosa (RP), a leading cause of inherited blindness, lacks effective treatment, necessitating the elucidation of its pathogenesis and identification of novel therapeutic targets. This study aimed to identify the genetic role of programmed cell death (PCD)-related genes in RP pathogenesis via an integrative multimodal framework. METHODS: We integrated summary-level data from genome-wide association studies (GWAS), expression/protein quantitative trait loci (eQTL/pQTL), and processed transcriptomic datasets. The workflow encompassed causal inference mainly using two-sample Mendelian randomization (MR), with complementary validation by summary-data-based MR (SMR); specificity validation via Bayesian colocalization, multivariable MR (MVMR), phenome-wide association study (PheWAS), and MR examining tissue specificity using GTEx eQTL data; and functional annotation through bulk and single-cell RNA-seq analyses. The study used publicly available genetic and transcriptomic databases. Primary analyses used GWAS summary statistics for RP (ebi-a-GCST90018912, ebi-a-GCST90013904, and ebi-a-GCST90013954). Validation was performed in bulk (GSE62020) and single-cell (GSE183206) retinal transcriptomic datasets. Exposures were genetically instrumented expression (cis-eQTL) or protein abundance (cis-pQTL) of PCD-related genes. Primary outcome was RP. Secondary outcomes included gene expression dysregulation, functional enrichment, immune cell infiltration, and cell-type-specific expression. RESULTS: Integrative two-sample and SMR analyses identified eight core PCD-related genes with robust genetic evidence linking to RP (e.g., AIM2, odds ratio = 2.90, 95% confidence interval: 1.81-4.63), which showed consistent effects across multiple tissues and remained nominally significant in MVMR models adjusting for proxy phenotypes of smoking and physical activity (nominal P < 0.05). PheWAS showed no evidence of pleiotropy for these associations (all P > 5 × 10⁻⁸). In retinal transcriptomes, these genes were dysregulated and enriched in immune pathways; CIBERSORT analysis revealed correlated immune cell infiltration, including a strong negative correlation between H13 and M1 macrophages (r = -0.625, P < 0.05). Single‑cell RNA‑seq further pinpointed Müller glia as the most reprogrammed population (area under the curve score: 0.07). CONCLUSIONS: This multimodal study provides genetic evidence linking a core set of PCD-related genes to RP pathogenesis. The findings suggest a mechanistic model wherein dysregulation of these genes may contribute to immune dyshomeostasis and Müller glia dysfunction, emphasizing a network of promising therapeutic targets for this blinding disorder.