子宫内膜癌中肿瘤特异性剪接异构体的系统性发现与验证

Endometrial carcinoma remains an area of substantial unmet clinical need with >65,000 new diagnoses and ~14,000 deaths annually in the United States. Despite its high incidence, response to PD-1-based immunotherapy remain modest: single-agent PD-1 inhibitors show objective response rates of ~30-40% overall, with notably lower activity in mismatch repair-proficient tumors and limited options beyond immunotherapy or chemotherapy, underscoring a major unmet need for new targeted modalities. Large-scale genomic studies indicate that aberrant RNA splicing is a defining feature of this disease, with many tumors harboring mutations or dysregulated expression in core splicing factors, suggesting a rich but underexplored source of therapeutic targets.Here, we applied the NeoSplice platform to systematically identify and validate tumor-specific splice isoforms in endometrial cancer. Long-read RNA sequencing on 20 tumor and normal endometrial tissue samples enabled full-length isoform resolution. Approximately 40% of detected isoforms were previously unannotated, including ~15% containing novel splice junctions, many of which were recurrent across patients. Differential expression analysis identified more than 10,000 isoforms significantly altered between tumor and paired normal tissue, including 768 upregulated isoforms encoding predicted transmembrane surface proteins.To assess cohort-level, we applied NeoSplice-AI, a proprietary isoform discovery and validation framework, to mine public short-read RNA-seq data from TCGA and GTEx, confirming many novel junctions across large endometrial cancer cohorts while demonstrating minimal expression in healthy tissues. We further developed a novel proprietary approach to resolve isoform junction usage from archival tissues at single-cell resolution, enabling quantification of isoform percent expression across cell types, and revealing multiple tumor-associated splice variants selectively enriched in stromal or immune cell populations.Finally, mass spectrometry-based proteomics confirmed peptide-level evidence for multiple isoforms, providing orthogonal validation of their protein-level expression. These orthogonal datasets were combined into a multi-parameter scoring framework to prioritize therapeutically actionable isoforms based on expression prevalence, cell-type and tissue specificity, protein evidence, localization, and druggability, with AlphaFold used to model select aberrant splice-derived protein structures.Together, these results establish aberrant splicing as a scalable source of actionable targets in endometrial cancer, enabling antibody-based modalities such as T-cell engagers targeting surface neoantigens, and TCR-mimetic strategies directed against intracellular splice variants, with broad applicability across many other solid tumors.