A comprehensive endometrial cancer organoid biobank reveals subtype-specific transcriptional programs and therapeutic targets

Background : Endometrial cancer (EC) is a clinically and biologically heterogeneous disease with unequal outcomes, particularly among high-grade tumors that disproportionately affect racially diverse populations. Existing preclinical models incompletely capture this heterogeneity. We established a large, racially diverse EC patient-derived organoid (PDO) biobank with matched normal to define subtype-specific molecular programs and actionable vulnerabilities. Methods : Tumors and matched normal endometrium were prospectively collected (n=380). Morphologic fidelity was assessed by histology and immunophenotyping. Multi-omic profiling evaluated driver retention and defined subtype-specific transcriptional states. High-throughput drug screening identified candidate vulnerabilities. Results : A total of 319 samples were included, with a ~93% PDO establishment rate, and matched tumor-normal PDOs generated across major EC histologies. PDOs recapitulated the key architectural and protein-expression features of their parental tumors, including TP53-abnormal serous phenotypes and mixed epithelial-mesenchymal organization in carcinosarcoma (CS). PDOs also preserved the key genomic characteristics, including TP53, PTEN, and PIK3CA alterations. Transcriptomic profiling revealed distinct subtype programs. Endometrioid PDOs showed Wnt-associated epithelial renewal with reduced apoptosis, p53, mTORC1, and inflammatory signaling. Serous PDOs displayed TP53-abnormal, MYC-and cell-cycle-enriched programs with uniformly suppressed interferon signaling. CS exhibited hybrid epithelial-mesenchymal features with activation of Hedgehog and myogenic pathways and broad suppression of NF-kB and interferon signaling. Stratification by p53 stability and MSI status defined opposing hyperproliferative versus immune-active axes. High-throughput drug screening identified class I HDAC inhibitor (romidepsin, RD) as a consistent activity signal. RD-perturbation RNA-seq in CS showed extensive remodeling at 10 nM, with downregulation of core G2-M/E2F mitotic regulators and suppression of CS-associated secreted factors, accompanied by restoration of antigen-presentation and interferon-responsive genes. Pathway analysis demonstrated coordinated suppression of mitotic modules and reactivation of cytokine and immune signaling. RD inhibited the TPX2-CDK1-KIF11 mitotic module, indicating a druggable spindle-dependency circuit in CS. Conclusions : This large, racially diverse EC PDO biobank provides a robust preclinical platform that captures EC heterogeneity and enables controlled, subtype-resolved analyses. Integrated multi-omics and functional perturbation identify a spindle-dependency axis in CS, supporting subtype-specific and ancestry-aware therapy development.