Mechanisms of pathogenesis in patient derived organoid models of prostate cancer

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

The focus of on-going projects is the use of tractable genetic models to mechanistically investigate therapeutic vulnerabilities of CRPC. Historically, the ability to grow prostate cancer in culture has been extremely limited, leading to a lack of in vitro models that recapitulate the diversity of human prostate cancer. Therefore, although extensive genomic characterization exists for primary and metastatic prostate cancer patient samples, the ability to interrogate mutational mechanisms in the naturally-occurring genomic context has been severely limited. Also, the dearth of models has impeded interrogating the generality of many interesting and potentially translatable findings from studies using individual prostate cancer models. Mechanistic understanding is key to defining highly clinically relevant information such as predictive mutations, novel therapeutic combinations, and diagnostic or predictive biomarkers. Recent advances in growing metastatic prostate cancer biopsies from the Clevers/Sawyers labs have used methodology adapted from conditions developed for the growth of intestinal stem cells and human colon cancer, often referred to as organoid growth conditions. The success reported by Sawyers and colleagues for the establishment of cultures from metastatic prostate cancer biopsies is 20%. This low success rate is a significant impediment to biobank establishment due to the relatively infrequent biopsy rate for this patient population. However, metastatic castrate-resistant prostate cancer (mCRPC) patient-derived xenografts (PDXs) recapitulate the genetic and phenotypic diversity of the disease. We have invested significant effort to set up the infrastructure and adapt the University of Washington mCRPC PDX collection (so-called LuCaP models) to growth in organoid culture. Genetically characterized PDX cohorts in other solid cancers have been useful and predictive of clinical outcome. The LuCaP models are particularly valuable in representing a range of genotypes and phenotypes of acquired resistance to ADT. Although LuCaP models were derived from CRPC patients, the PDXs demonstrate variable sensitivity to growth in castrated hosts. This likely represents differences in the hormone physiology and microenvironment of the mouse in comparison to human and also suggests that resistance to ADT is a continuum. AR is not a typical oncogene since it drives multiple pathways that encompass growth, differentiation, and lineage selection that may be plastic and differentially expressed in the tumor population. The full utility of mCRPC PDX models for high throughput and mechanistic analyses has not been realized due to their lack of robust growth in vitro. Using 23 representatives from the LuCaP mCRPC cohort, we developed optimized organoid conditions that allow growth in a majority of PDX models, and also improved establishment of CRPC patient biopsies over current methods. Adenocarcinoma organoids demonstrated continued dependence on AR signaling, replicating a dominant characteristic of CRPC. Unexpectedly, we found that p38 MAPK activity was frequently essential and led to increased AKT activity. GSK3-beta inactivation was one functional AKT target, although not entirely sufficient to replace p38 activity. Genomic analyses revealed considerable subclonal heterogeneity with high conservation between LuCaP PDX tumors and organoid cultures. The LuCaP PDX/organoid platform provides an expansive, manipulable platform to investigate predictive and mechanistic questions for mCRPC. There are immediate implications for prostate cancer drug screening as well as long-term potential to advance the understanding of fundamental mechanisms driving the disease and resistance to therapy, previously not possible due to insufficient numbers of relevant models. In addition, we have succeeded in establishing a small number of mCRPC biopsy-derived organoid cultures. These cultures allow us to apply principles of precision medicine, including thorough genomic characterization and drug sensitivity screening, in order to discover efficacious treatment options defined by genomic biomarkers.We have previously established patient-derived metastatic prostate cancer organoid models of treatment-induced lineage plasticity in order to study the states of plasticity. They have been characterized by single-cell RNA-seq (scRNA-seq) and immunofluorescence to show distinct subpopulations with respect to lineage (adenocarcinoma and neuroendocrine) and proliferative capacity. Over the past year we have performed experiments to validate these results. We have established a protocol for single-molecule RNA-FISH in organoids and used it to validate the scRNA-seq results with selected lineage and proliferation markers. We have performed experiments to define the states of transition in our organoid models and interrelationships between these states. To wit, we have developed a protocol to use a method of CellTag lineage-tracing in our organoid models in collaboration with Samantha Morris at Washington University who created the CellTag technology. We have run a lineage-tracing experiment in a neuroendocrine mCRPC organoid model as a control. We have run a preliminary lineage-tracing experiment in the organoid models of plasticity and found it to be consistent with a model of lineage-distinct differentiation from progenitor cells. We have modified to improve, and then repeated the experiment. Importantly, this data identifies cancer stem cell populations that give rise to more differentiated, quiescent cells. Because epigenetic regulation of chromatin accessibility is an important feature of differentiation, we have performed single-cell ATAC-sequencing (scATAC-seq) on the organoid models of lineage-plasticity. The data have been analyzed and is consistent with the scRNA-seq results. We have found differential accessibility at regulatory regions of lineage-specific genes that separate into distinct subpopulations. These data identify enriched transcription factor activity in distinct cell subpopulations. Over this year, we have screened the final patient-derived models against a databank of 100 compounds curated by their capacity for translation into treatment of prostate cancer, completing a biobank of 30 models containing common prostate cancer phenotypes consisting of adenocarcinoma, castrate-resistant adenocarcinoma, neuroendocrine, amphicrine, and double negative, along with one non-cancer prostate tissue derived organoid model to serve as a baseline for visualizing the drug toxicity for non-tumor tissue. In order to demonstrate the value of screening a heterogeneous cohort of models, we added LnCAP and RWPE-1 cells to represent cell models of diseased and normal tissue, screened as both 2D monolayers and 3D spheroids, to allow for determining whether any compounds in the library would be missed when screening using common cell lines, and to create context for 2D vs 3D model screening. Having completed aggregating drug responses, we obtained full transcriptome data for all screened models, and we have generated transcriptomic signatures to predict sensitivity or resistance to individual compounds or compound classes, as well as gene pathway enrichments which can reveal mechanisms of resistance. Importantly, we have determined the clustering of several compounds including chemotherpeutics, cell cycle modulators, apoptosis inducers, and DNA repair modulators together, demonstrating broadly sensitive and resistant models with common vulnerabilities. With profiled compound sensitivity patterns across models, we are currently in the process of verifying which compounds or compound combinations can translate to in vivo efficacy using patient-derived xenografts mice models and will further test promising candidates in metastatic organoid mouse models.