Castration-resistant prostate cancer (CRPC) is largely dependent on the androgen receptor (AR) for growth and often exhibits hyperactive PI3K signaling, most frequently because of PTEN loss. Therapeutic pressure from anti-AR therapies can induce transdifferentiation toward an AR-independent phenotype. Recently, different subtypes of AR-independent CRPC have been redefined, with the stem cell–like (SCL) subtype emerging as one of the most prevalent. Elucidation of the epigenetic mechanisms controlling the maintenance of these distinct CRPC cell states could pave the way for effective combinatorial therapies for CRPC. In this study, we identified a key role for the histone methyltransferase KMT2D in establishing the chromatin competence necessary for the recruitment of AR and FOXA1 transcription factors (TF) that are essential for the AR transcriptional output in AR-dependent CRPC cell lines, patient-derived organoids, and patient samples. Unexpectedly, KMT2D maintained the identity of the AR-low CRPC-SCL subtype and controlled activity of AP-1 TFs such as FOSL1, which acts as a master regulator of this subtype. Single-cell transcriptomics and chromatin assays underscored the role of KMT2D in sustaining a mixed lineage cell state via AP-1 and FOXA1. The combined suppression of PI3K/AKT and KMT2D reduced cell proliferation in prostate cancer cells and patient-derived organoids in both CRPC-AR and CRPC-SCL subtypes. Altogether, these results unveil KMT2D as a major mediator of the epigenetic landscape in subtype-specific CRPC, contributing to tumor growth and therapeutic response. Significance: KMT2D is a critical regulator of chromatin accessibility and transcriptional landscapes in castration-resistant prostate cancer that drives both AR-dependent and AR-independent subtypes, highlighting KMT2D as a potential therapeutic target.
The understanding of how transcriptional programs give rise to cellular morphology, and how morphological features reflect and influence cell identity and function remains limited. This is due in part to the lack of large-scale datasets pairing the two modalities as well as the absence of computational frameworks capable of modeling their cross-modal structure. Here, we introduce COSMIC, a bidirectional generative framework that enables quantitative decomposition of transcriptional variance reflected in morphology and morphological variance explained by gene expression. COSMIC builds on a foundation model trained on over 21 million segmented nuclei and couples it with existing transcriptomic embeddings. To enable cross-modal learning, we leveraged a newly generated multimodal dataset acquired using IRIS, a technology that captures high-resolution images and transcriptomes from the same single cells at scale. COSMIC accurately modeled cell type identity, as well as continuous dynamics such as cell-cycle progression, establishing a quantitative link between morphological phenotypes and underlying gene expression. In prostate cancer cells, COSMIC identified morphological and transcriptomic differences between chemotherapy drug treatment-responsive and -resistant cells, and revealed morphology-associated genes linked to tumor state. Together, these results demonstrate that generative modeling powered by paired single-cell measurements can capture the bidirectional flow of information between cellular form and gene expression, opening new avenues for mechanistic discovery and predictive modeling in both basic and translational cell biology.