Prostate Cancer

Project 1 leverages tumor specimens from patients with high-risk prostate cancer treated with neoadjuvant therapies to understand how tumors respond and resist acute potent androgen receptor blockade and to develop novel strategies to improve cure rates and combat resistance.

Localized high-risk prostate cancer (PCa) comprises ~15% of newly diagnosed localized PCa, and the cure rate for these tumors after primary surgical treatment by radical prostatectomy (RP) is suboptimal, with recurrent disease in up to ~50%. Neoadjuvant trials for these localized high-risk PCa may provide valuable opportunities to increase cure rates and our understanding of response and resistance. However, previous studies of neoadjuvant androgen deprivation using GnRH agonists alone have generally shown infrequent pathological complete responses (pCRs) and have not shown evidence of clinical efficacy. We have hypothesized that responses to neoadjuvant ADT have been limited due to substantial residual androgen in the prostate, and that more intensive neoadjuvant androgen signaling inhibition (ASI) therapy, in addition to increasing pCR rates, will translate into an improvement in disease free survival (DFS) and ultimately overall survival (OS).

We have been testing this hypothesis in a series of phase 2 trials examining the efficacy of neoadjuvant intensive ASI using leuprolide in combination with abiraterone (ABI) and/or enzalutamide (ENZ) or apalutamide (APA) for 6 months prior to radical prostatectomy (RP) in men with localized high-risk localized PCa. Responses appear to be improved relative to previous neoadjuvant trials, with pCR rates of 10-12%, and with an additional 10-20% having minimal residual disease (MRD). Moreover, compared to expected relapse rates, patients with pCR/MRD have had much lower than predicted rates of PSA failure and progression to metastatic PCa. However, it remains unclear whether the intensive ASI therapies are preventing the emergence of metastatic disease (presumably by treating microscopic metastatic disease) versus the possibility that the cases with pCR/MRD reflect tumors that have less metastatic potential. Moreover, the molecular basis for residual disease in prostate, and its relationship to metastatic disease in patients who progress, remain to be determined.

To address the genomic basis of response versus resistance, we propose comprehensive genomic analyses of tumors with exceptional responses (pCR/MRD) to intensive neoadjuvant ASI therapy, and of residual disease in RP specimens in men who do not achieve pCR/MRD, including determining whether oncogenic alterations in the residual disease are driving the subsequent emergence of metastatic PCa (Aim 1). These studies will leverage samples from our previous trials, as well as a large multicenter phase 3 study of neoadjuvant GnRH agonist/antagonist combined with APA in comparison with GnRH agonist/antagonist alone (PROTEUS trial, supported by Janssen, NCT03767244). Aim 2 will identify actionable acute nongenomic adaptations that mediate initial resistance to intensive ASI therapy, and specifically test the hypothesis that these adaptations converge on expression of D cyclins and activation of CDK4/6 to drive proliferation. Aim 3 is a randomized phase 2 trial of intensive ASI (leuprolide plus darolutamide, DARO) alone or in combination with the CDK4/6 inhibitor abemaciclib.

Our long-term goal is to establish neoadjuvant trials as a platform for assessing the efficacy of novel combination therapies in castration-sensitive PCa. Bringing effective systemic therapies to the neoadjuvant setting has the potential to increase cure rates for men with localized high-risk disease, who may be rendered disease free by the therapy or who may benefit from subsequent adjuvant therapy guided by findings in the RP. Moreover, validating that pathological responses after neoadjuvant therapy are predictive of DFS and ultimately OS has the potential to markedly accelerate testing of novel agents and combinations through neoadjuvant trials, which may then also be effective in metastatic castration-sensitive PCa. Finally, insights gained into tumor biology and clonal dynamics can be exploited to develop new therapeutic strategies. The Specific Aims and subaims are as follows:

Aim 1. Identify molecular features of tumors that correlate with pathological responses to neoadjuvant intensive androgen signaling inhibition (ASI) and its clinical significance.

• Determine molecular features of tumors that are associated with pathological complete responses.
• Identify features predictive of improved disease free survival.
• Identify genomic mechanisms of resistance to neoadjuvant intensive androgen signaling inhibition and their contribution to metastatic disease.

Aim 2. Identify actionable nongenomic mechanisms of resistance to neoadjuvant intensive androgen signaling inhibition.

• Identify acute adaptations to intensive androgen signaling inhibition.
• Assess efficacy and mechanism of action of CDK4/6 inhibitors combined with castration in PDX models.

Aim 3. Assess the efficacy of neoadjuvant androgen signaling inhibition combined with abemaciclib.

• Assess efficacy of neoadjuvant intensive ASI therapy combined with abemaciclib.
• Assess target engagement and mechanisms of intrinsic or acquired resistance.

Project 2 will develop innovative strategies to target the epigenome in later stages of advanced castration resistant prostate cancer and will develop a first-in-field clinical trial focused on co-targeting EZH2 and PARP.

Despite significant advances in prostate cancer therapy, available treatment options particularly in later stages of the disease, are still limited and the treatment-resistant setting represents a serious unmet medical need. Resistance to androgen receptor (AR) therapies in prostate cancer is predominantly AR-driven. Non-AR driven castration resistant prostate cancer (CRPC) represents approximately 15-20% of advanced disease and is more predominant after potent AR-targeted therapies; one emerging mechanism is through lineage plasticity with reprogramming of prostate adenocarcinoma tumor cells towards a neuroendocrine phenotype. Epigenetic dysregulation, including changes in DNA methylation and overexpression of the enhancer of zeste homolog 2 (EZH2) also drive castration resistance, though much remains to be understood regarding the impact and contextual timing of these epigenetic changes in mediating AR-driven versus non-AR-driven resistance. EZH2 is a member of a polycomb repressor complex 2 that plays an important role in development through transcriptional repression via histone methylation. EZH2 is overexpressed in CRPC, both in adenocarcinoma and neuroendocrine prostate cancer (NEPC), along with EZH2 activity measured by histone H3 lysine 27 trimethylation. Beyond its canonical role, we previously identified a non-canonical role of EZH2 in activating AR signaling (mediated in part by phosphorylation) as well as in driving lineage plasticity and NEPC (in part through cooperation with N-myc). Importantly, EZH2 is targetable and EZH2 inhibitors are currently in clinical trials. The EZH2 inhibitor tazemetostat has shown clinical activity in non-Hodgkin lymphoma, synovial sarcoma, mesothelioma, and was recently FDA approved for the treatment of epithelioid sarcoma. The prostate cancer patient subset most likely to benefit from EZH2 inhibition has not been established. Current data suggests that single agent activity may be limited and that combination strategies should be pursued. We recently discovered a novel interaction between EZH2 and DNA repair processes and a synergy between EZH2 and PARP inhibition in CRPC. We hypothesize that overexpression of EZH2 during CRPC progression drives downstream molecular programs through both canonical and non-canonical mechanisms, and that these downstream effects are context dependent. We hypothesize that combination therapy with EZH2 and PARP inhibition will have differential biologic impact in CRPC based on the underlying genomic, epigenomic, and phenotypic context.

Aim 1: Elucidate the crosstalk between EZH2 and DNA repair pathways and mechanisms of synthetic lethality. We found that DNA repair genes are induced by EZH2 and acutely repressed by EZH2i, and the combination of EZH2 with PARP inhibition is synergistic. We will examine the canonical regulation by EZH2 of DNA repair pathways (NHEJ, HR) and the effects of EZH2 inhibition on de-repression of MAD2L2 and SLFN11 as a means to sensitize to PARP inhibition. We will also investigate non-canonical regulation of DNA repair genes through EZH2 interaction with cofactors such as E2F and AR. We will further evaluate response to EZH2i and EZH2i+ PARPi and downstream effects in homologous recombination (HR) intact and HR-deficient patient derived xenograft models. Results from this Aim will provide a novel molecular route for PARPi synthetic lethality in HR proficient prostate cancer.

Aim 2: Define the biologic impact of EZH2 + PARP inhibition across the heterogeneous spectrum of CRPC including both AR-driven and non-AR-driven tumors. We have observed synergy when co-targeting EZH2 and PARP1 in AR-driven CRPC models. We will extend this combination to models that represent the CRPC spectrum, including NEPC. We will assess consequences of EZH2i and effects of combination therapy on tumor growth, DNA damage, downstream gene expression profiles, DNA methylation, and chromatin accessibility. We will specifically evaluate EZH2 targets and PRC dependence, changes in AR and neuroendocrine programs, as well as heterogeneity at the single cell level. Results from this Aim will provide insights into the downstream consequences of EZH2i and PARPi therapy across distinct subsets of CRPC.

Aim 3: Identify biomarkers of response and resistance to the combination of EZH2 + PARP inhibition in patients. We will conduct investigator-initiated Phase 1/2 clinical trials of the combination of the EZH2i tazemetostat with the PARPi talazoparib to assess the safety, efficacy, and biomarkers of response in patients with CRPC/NEPC. Pre-treatment and on-therapy biopsies will be assessed for changes in g-H2AX on therapy, as a primary pharmacodynamic marker supporting mechanism. Baseline tumors will also be classified based on their underlying genomic (DNA repair/AR/RB) and epigenetic (NE/adeno) status and correlated with response and duration of therapy. We will evaluate other specific DNA damage readouts (eg., RAD51, SLFN11, MAD2L2, phospho-RPA) and downstream changes in signaling/expression profiles (eg., NHEJ/HR repair genes, PRC targets, AR signaling, NE programs) on therapy. Results from this Aim will provide first-in-field clinical data of combined EZH2 and PARP inhibitor therapy in patients with advanced prostate cancer.

Project 3 delves deep into biomarkers in localized prostate cancer, leveraging innovations in computation and biologically-guided deep learning, to deliver on precision cancer medicine in this disease state. The ability to understand why some localized prostate cancers are phenotypically aggressive and predict which localized prostate cancers will behave in this manner addresses a large clinical unmet need.

Over 90% of the 268,490 estimated cases of prostate cancer (PCa) in 2022 will present as localized disease, with a majority defined as intermediate or high-risk for recurrence based on traditional pathologic features1-4. Despite standard of care therapy commonly involving radical prostatectomy (RP) or combined radiation and hormonal therapy, the risk of recurrent disease is as high as 50% for subsets of men with localized, high-risk PCa, and more than half of PCa deaths occur in men with initially localized or locally advanced disease. In the preceding decade, multiple molecular determinants of lethal prostate cancer have been identified, including germline and somatic mutations in DNA repair genes that impact genome-wide and immune microenvironmental molecular signatures in retrospective cohorts5-14. Understanding the biological relevance of these properties and their co-occurring interactions, and as well as their relationship to existing tissue-based and clinical biomarkers for determining at the point-of-care which patients are likely to progress to metastasis and death despite typical management, is essential. Similarly, the extent to which these features imprint spatial patterns in PCa histopathology, and whether prognostic patterns can be learned from these data for clinical use, is uncertain but may dramatically impact worldwide practice patterns given the ubiquity of histopathology. Broadly, two major activities would significantly advance localized PCa care: (i) understanding the convergent biological underpinnings behind phenotypically aggressive but clinically localized PCa to inform new therapeutic paradigms, and (ii) predicting which localized PCa will behave in this manner to enrich for adjuvant studies geared towards improving survival for those at highest risk of progression to lethal disease. Taken together, tackling these challenges would significantly advance basic prostate cancer research and impact the development of personalized escalation and de-escalation treatment strategies for patients. Innovations in explainable artificial intelligence, paired with the advent of harmonized and representative clinicogenomic PCa cohorts, has created a new opportunity to engage in hypothesis-driven reframing of localized PCa. Our overarching hypothesis is that interacting molecular and histopathologic features determine prognostic outcomes in intermediate and high-risk localized PCa after RP, and that biologically guided deep learning models of molecular, pathologic, and phenotypic data can distinguish these clinical states. Our interdisciplinary team will leverage diverse cohorts to address this hypothesis through these Specific Aims: 

Aim 1: Define the interacting germline and somatic properties and co-occurring alterations that mediate aggressive localized prostate cancer using biologically guided neural networks
Hypothesis: Co-occurring known and novel germline and somatic alterations discriminate aggressive localized primary PCa after RP, and these patterns can be revealed through biologically guided neural networks.
1A: Identify germline mutation patterns in DNA repair and other canonical PCa pathways in large and ancestrally diverse prostate cancer cohorts using deep learning models
1B: Determine the co-occurring molecular properties that mediate metastasis and overall survival (MFS/OS) outcomes using biologically guided interpretable deep learning
1C: Functionally validate candidate co-mutation events identified through molecular deep learning.

Aim 2: Determine the histopathologic latent representations of molecularly and clinical distinct localized prostate cancer through digital pathology deep learning
Hypothesis: The spatial histopathologic architecture of intermediate/high risk localized PCa and its surrounding microenvironment reflect specific underlying molecular states, and deep learning models of histopathology images can predict these molecular states and learn graphical patterns that discriminate MFS/OS outcomes.
2A: Determine the tumor and microenvironmental spatial characteristics of distinct PCa molecular subtypes
2B: Develop computer vision models that predict MFS/OS in localized PCa after RP from histopathology images of RP specimens
2C: Implement histopathology-based deep learning predictive models in PCa pathologist clinical settings

Aim 3: Evaluate condensed deep learning genomic models for predicting MFS/OS in retrospective and clinical trial prostate cancer cohorts
Hypothesis: Deep learning genomic models developed with clinically validated and widely available NGS panels can stratify MFS/OS outcomes in intermediate and high risk localized PCa after RP beyond of established histopathologic and clinical prognostic risk factors.
3A: Compare the prognostic performance of a condensed P-NET model with existing histopathologic and clinical models in retrospective real-world cohorts
3B: Evaluate the prognostic performance of a condensed P-NET PCa model in the Phase 3 CALGB 90203 trial Successful execution of this proposal will transform our biological understanding of PCa and our ability to develop future surveillance and adjuvant strategies for men at highest risk of metastatic relapse and death from PCa.
Broadly, this project will transform precision cancer medicine for localized PCa and serve as a model for the creation, development, and application of these emerging methodologies across the disease continuum.