Purpose: Phosphatase and tensin homolog (PTEN) loss of function occurs in approximately 50% of patients with metastatic castrate-resistant prostate cancer (mCRPC), and is associated with poor prognosis and responsiveness to standard-of-care therapies and immune checkpoint inhibitors. While PTEN loss of function hyperactivates PI3K signaling, combinatorial PI3K/AKT pathway and androgen deprivation therapy (ADT) has demonstrated limited anticancer efficacy in clinical trials. Here, we aimed to elucidate mechanism(s) of resistance to ADT/PI3K-AKT axis blockade, and to develop rational combinatorial strategies to effectively treat this molecular subset of mCRPC. Experimental Design: Prostate-specific PTEN/p53-deficient genetically engineered mice (GEM) with established 150–200 mm3 tumors, as assessed by ultrasound, were treated with either ADT (degarelix), PI3K inhibitor (copanlisib), or anti–PD-1 antibody (aPD-1), as single agents or their combinations, and tumors were monitored by MRI and harvested for immune, transcriptomic, and proteomic profiling, or ex vivo co-culture studies. Single-cell RNA sequencing on human mCRPC samples was performed using 10X Genomics platform. Results: Coclinical trials in PTEN/p53-deficient GEM revealed that recruitment of PD-1–expressing tumor-associated macrophages (TAM) thwarts ADT/PI3Ki combination–induced tumor control. The addition of aPD-1 to ADT/PI3Ki combination led to TAM-dependent approximately 3-fold increase in anticancer responses. Mechanistically, decreased lactate production from PI3Ki-treated tumor cells suppressed histone lactylation within TAM, resulting in their anticancer phagocytic activation, which was augmented by ADT/aPD-1 treatment and abrogated by feedback activation of Wnt/β-catenin pathway. Single-cell RNA-sequencing analysis in mCRPC patient biopsy samples revealed a direct correlation between high glycolytic activity and TAM phagocytosis suppression. Conclusions: Immunometabolic strategies that reverse lactate and PD-1–mediated TAM immunosuppression, in combination with ADT, warrant further investigation in patients with PTEN-deficient mCRPC.
Purpose: Phosphatase and tensin homolog (PTEN) loss-of-function/PI3K pathway hyperactivation is associated with poor therapeutic outcomes and immune checkpoint inhibitor resistance across multiple malignancies. Our prior studies in Pb-Cre;PTENfl/flTrp53fl/fl genetically engineered mice (GEM) with aggressive-variant prostate cancer (AVPC) demonstrated tumor growth control in 60% mice following androgen deprivation therapy/PI3K inhibitor (PI3Ki)/programmed cell death protein 1 (PD-1) antibody combination, via abrogating lactate cross-talk between cancer cells and tumor-associated macrophages (TAM), and suppression of histone lactylation (H3K18lac)/phagocytic activation within TAM. Here, we targeted immunometabolic mechanism(s) of PI3Ki resistance, with the goal of durable tumor control in AVPC. Experimental Design: Pb-Cre;PTENfl/flTrp53fl/fl GEM were treated with PI3Ki (copanlisib), MEK inhibitor (trametinib) or Porcupine inhibitor (LGK'974) singly or their combinations. MRI was used to monitor tumor kinetics and immune/proteomic profiling/ex vivo coculture mechanistic studies were performed on GEM tumors or corresponding tumor-derived cell lines. Results: Given our proteomic profiling showing persistent MEK signaling within tumors of PI3Ki-resistant GEM, we tested whether addition of trametinib to copanlisib enhances tumor control in GEM, and we observed 80% overall response rate via additive suppression of lactate within TME and H3K18lac within TAM, relative to copanlisib (37.5%) monotherapy. The 20% resistant mice demonstrated feedback Wnt/β-catenin activation, resulting in restoration of lactate secretion by tumor cells and H3K18lac within TAM. Cotargeting Wnt/β-catenin signaling with LGK'974 in combination with PI3Ki/MEKi, demonstrated durable tumor control in 100% mice via H3K18lac suppression and complete TAM activation. Conclusions: Abrogation of lactate-mediated cross-talk between cancer cells and TAM results in durable ADT-independent tumor control in PTEN/p53-deficient AVPC, and warrants further investigation in clinical trials.
Artificial intelligence models have been increasingly used in the analysis of tumor histology to perform tasks ranging from routine classification to identification of molecular features. These approaches distill cancer histologic images into high-level features, which are used in predictions, but understanding the biologic meaning of such features remains challenging. We present and validate a custom generative adversarial network—HistoXGAN—capable of reconstructing representative histology using feature vectors produced by common feature extractors. We evaluate HistoXGAN across 29 cancer subtypes and demonstrate that reconstructed images retain information regarding tumor grade, histologic subtype, and gene expression patterns. We leverage HistoXGAN to illustrate the underlying histologic features for deep learning models for actionable mutations, identify model reliance on histologic batch effect in predictions, and demonstrate accurate reconstruction of tumor histology from radiographic imaging for a “virtual biopsy.”
Precision medicine in cancer treatment depends on deciphering tumor phenotypes to reveal the underlying biological processes. Molecular profiles, including transcriptomics, provide an information-rich tumor view, but their high-dimensional features and assay costs can be prohibitive for clinical translation at scale. Recent studies have suggested jointly leveraging histology and genomics as a strategy for developing practical clinical biomarkers. Here, we use machine learning techniques to identify de novo latent transcriptional processes in squamous cell carcinomas (SCCs) and to accurately predict their activity levels directly from tumor histology images. In contrast to analyses focusing on pre-specified, individual genes or sample groups, our latent space analysis reveals sets of genes associated with both histologically detectable features and clinically relevant processes, including immune response, collagen remodeling, and fibrosis. The results demonstrate an approach for discovering clinically interpretable histological features that indicate complex, potentially treatment-informing biological processes.