Mechanisms of pathogenesis in patient derived organoid models of prostate cancer

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

I. A cancer stem cell population underlies a multi-lineage phenotype and drug resistance in prostate cancer: We have performed an in-depth analysis on an organoid culture established from a CCR patient biopsy sample exemplifying a resistance mechanism that bypasses an AR requirement via lineage plasticity, resulting in the emergence of phenotypes spanning luminal epithelial and neuroendocrine phenotypes (NEPC). This work addresses the molecular phenotypes and lineage relationships as well as the therapeutic sensitivity of cells within a mixed lineage mPC tumor population. Some key accomplishments include: 1) The definition of multiple states of differentiation and their developmental hierarchy originating from bipotential and clonal stem/progenitor subpopulations. 2) A determination that the program driving the multi-lineage phenotype of the organoids included concurrently active luminal epithelial, NE, and stem cell-determining transcription factors. 3) The identification of specific molecular vulnerabilities in the stem/progenitor subpopulations to block tumor growth. The AURKA inhibitor, alisertib caused a selective depletion of the stem/progenitors, and combined inhibition of AURKA and AR blocked the growth of xenograft tumors. This work contributes significantly to our understanding of treatment-induced lineage plasticity in mPC by illuminating underlying cellular and molecular drivers of the phenomenon in near patient models. Importantly, it demonstrates a potential treatment for certain types of CRPC. This work has been submitted for publication. II. A genotoxic antibody drug conjugate targeting CD276/B7H3 demonstrates efficacy across multiple biomarker defined classes of treatment refractory metastatic prostate cancer: We have analyzed the therapeutic response by our extensive organoid/PDX library of models to an antibody-drug conjugate that targets a tumor-specific antigen, B7H3. B7H3 is a cell surface immunomodulatory protein often correlated with higher tumor grade and poor patient prognosis in several solid tumors. It is an attractive therapeutic target due to its frequent upregulated protein expression relative to normal tissue. We interrogated response to an anti-B7H3 targeted antibody conjugated to the pyrrolobenzodiazepine (PBD) genotoxic agent and identified predictive biomarkers of response. This analysis exemplifies the utility of comprehensively evaluating a large, heterogeneous cohort. In addition to demonstrating broad efficacy for treatment-resistant forms of mPC, we have identified multiple, partially-overlapping, specific biomarker classes of responsiveness, which contribute to both a mechanistic understanding and also provide in-depth information for patient stratification. RB1-deficiency/ replication stress, SLFN11 expression, and chronic tumor intrinsic interferon (IFN) signaling are partially-overlapping, significant predictors of B7H3-PBD response, independent of tumor lineage. In addition, models compromised in specific aspects of double-stranded break repair, CHD1 and ATR loss of function, were responsive even in the absence of the other biomarkers. We performed preclinical trials using four PDX models based on their RB1, SLFN11, and IFN signature status. B7H3-PBD eradicated large established subcutaneous tumors and metastasis and improved long-term overall survival in RB1-deficient tumors, with or without SLFN11 expression, but showed no response in an RB1+/SLFN11NEG tumor model. Collectively, these findings support the potential of B7H3-PBD-targeted therapeutics for mPC. Of note, at least one of the multiple above biomarkers is expressed in about 70% of CRPC clinical samples, suggesting broad potential for mechanistically complementary treatment of tumors escaping current regimens. This work has been submitted for publication. III. High throughput drug screening: Because mPC/CRPC is a highly phenotypically and genotypically heterogenous disease, our hypothesis has been that the effective translation of novel treatments requires a representative cohort of models in order to discover rational response classes and to effectively predict efficacy in treatment-resistant mPC patient populations. We have implemented a high throughput screening (HTS) approach in order to efficiently, uniformly, and comparatively assay a large number of drugs encompassing various mechanisms of action. Several conclusions from the HTS are summarized below. Unfortunately, in vivo confirmatory studies were delayed due to COVID, which has postponed the submission of a manuscript. High-throughput drug screening on a cohort of 30 patient-derived organoid models covering 5 histological phenotypes and 6 sites of metastasis has been completed (Fig. 1). Full transcriptomic sequencing of generation 1 organoids was conducted on the entire model cohort to aid in future biomarker discovery based on baseline transcriptomics, shown in a principal component analysis plot (PCA) in Fig. 2. Adenocarcinoma (ARPC) models cluster together and with their matched recurrence ARPC models that were previously experimentally castrated (designated EXP-CR). Neuroendocrine models (NE) cluster together; amphicrine models (AMP), expressing both ARPC and NE markers, reside in the space between ARPC and NE clusters, and the single double negative (DNPC) model shows no overlap with any other histological phenotype. An important feature of this HTS was the screening in 3D solid Matrigel, which enables a closer resemblance to physiological tumor biology. Each of the 80 compounds was tested at 10 doses, resulting in the ability to generate entire drug response curves and enabling the use of Area Under Curve (AUC) as a continuous variable of response in organoid-drug pairs, which maximized resolution in assigning responder and non-responder categorizations accurately. The drug response patterns demonstrated that the majority of compounds tested showed either single model-specific responses or a general lack of efficacy. However, roughly 20 compounds demonstrated a range of responses providing opportunity for further analysis into characteristics governing resistance and sensitivity. Docetaxel (DTX) is a relatively widely used and proven efficacious drug for mPC and provided the largest observed range in AUC responses. Strikingly, for a proportion of drugs, including DTX, we observed significant correlations between and among particular drugs targeting cell cycle, apoptosis, and cell replication pathways. This cluster of therapeutics demonstrated common response and nonresponse patterns across multiple models, suggesting a similarity in the biology governing resistance and sensitivity among these correlated response patterns. Importantly, removing RB1/TP53 null models largely conserved this cluster of co-efficacious compounds suggesting this observation holds true for ADPC and is not driven by replication stress or RB1-loss-dependent lineage/phenotypic differences. Because DTX responses overlapped with the majority of correlated compounds in the cluster of co-efficacious compounds, it was used as the basis for building transcriptomic contrasts from sensitive versus resistant models in order to investigate the underlying mechanisms of vulnerability. Overrepresentation of differential gene analysis between responsive and nonresponsive, RB1/TP53- intact models revealed strong enrichment in nonresponsive models for glucuronidation and HNF1A transcription factor driven activity. Responsive models demonstrated an enrichment of cell-cycle, YAP/TAZ, TGFbeta, *TRUNCATED*