Transcriptional profiling of matched patient biopsies clarifies molecular determinants of enzalutamide-induced lineage plasticity

Abstract

The androgen receptor (AR) signaling inhibitor enzalutamide (enza) is one of the principal treatments for metastatic castration-resistant prostate cancer (CRPC). Several emergent enza clinical resistance mechanisms have been described, including lineage plasticity in which the tumors manifest reduced dependency on the AR. To improve our understanding of enza resistance, herein we analyze the transcriptomes of matched biopsies from men with metastatic CRPC obtained prior to treatment and at progression (n = 21). RNA-sequencing analysis demonstrates that enza does not induce marked, sustained changes in the tumor transcriptome in most patients. However, three patients’ progression biopsies show evidence of lineage plasticity. The transcription factor E2F1 and pathways linked to tumor stemness are highly activated in baseline biopsies from patients whose tumors undergo lineage plasticity. We find a gene signature enriched in these baseline biopsies that is strongly associated with poor survival in independent patient cohorts and with risk of castration-induced lineage plasticity in patient-derived xenograft models, suggesting that tumors harboring this gene expression program may be at particular risk for resistance mediated by lineage plasticity and poor outcomes.

Introduction

Androgen deprivation therapy (ADT) is the principal treatment for metastatic prostate cancer, but progression to castration-resistant prostate cancer (CRPC) is nearly universal. In recent years, potent inhibitors of the androgen receptor (AR)—a luminal lineage transcription factor—have been developed, including the AR antagonist enzalutamide (enza)^1,2,3,4,5. Enza improves progression-free survival and overall survival in patients with CRPC; further, enza also increases overall survival in patients with hormone-naive prostate cancer who are beginning ADT for the first time^6,7,8. However, one-third of patients do not respond, and those with de novo resistance have a significantly increased risk of death compared to responders^6,7,8.

Despite intense study, clinical enza resistance remains poorly understood. Several studies examined mechanisms of de novo or acquired enza resistance in clinical samples and implicated: AR amplification^9,10, AR splice variants^11,12, increased Wnt/β-catenin signaling^13,14,15, increased TGF-β signaling^14,16, epithelial to mesenchymal transition or increased stemness^14,17, and lineage plasticity¹⁴. However, these prior studies were largely restricted to DNA mutational profiling, compared baseline and progression samples from different patients, used limited numbers of matched samples, or did not focus on transcriptional changes.

Prior work suggests that most CRPC tumors resistant to AR signaling inhibitors (ARSIs) continue to depend on the AR^17,18. However, we now appreciate that lineage plasticity—most commonly exemplified by loss of AR signaling and a switch from a luminal to an alternate differentiation program—is a resistance mechanism that appears to be increasing in the era of more widespread use of ARSIs¹⁹. The emergence of tumors with features of lineage plasticity may occur through diverse mechanisms: selection of a pre-existing clone that has already undergone differentiation change, acquisition of new genetic alterations that promote differentiation change, or transdifferentiation of tumor cells through epigenetic mechanisms^17,20,21,22.

Lineage plasticity is a continuum, ranging from tumors with persistent AR expression but low AR activity, those that lose AR expression but do not undergo neuroendocrine differentiation (double negative prostate cancer [DNPC]), and those that lose AR expression and do undergo neuroendocrine differentiation (neuroendocrine prostate cancer [NEPC])²³. Importantly, CRPC tumors that have undergone lineage plasticity are associated with a much shorter survival than CRPC tumors that have persistent AR activity and a luminal lineage program, demonstrating an urgent need to understand treatment-induced lineage plasticity in prostate cancer²⁴.

In this work, we hypothesized that comparing gene expression profiles between matched CRPC tumor biopsy samples prior to enza and at the time of progression would identify pre-treatment and treatment-induced resistance mechanisms in individual patients. We describe results from 21 matched samples. We find evidence of lineage plasticity occurring in three of 21 progression tumors and define pathways and transcription factors that are highly activated in the baseline samples from patients whose tumors undergo lineage plasticity after enzalutamide treatment. Finally, we identify a gene signature associated with risk of therapy-induced lineage plasticity and poor patient survival.

Results

Heterogenous effects of enzalutamide treatment on the tumor transcriptome across matched biopsy samples

By examining the Stand Up to Cancer Foundation/Prostate Cancer Foundation West Coast Dream Team (WCDT) prospective cohort, we identified 21 patients with CRPC who underwent a metastatic tumor biopsy prior to enza and a repeat biopsy at the time of progression and whose tumor cells underwent RNA-sequencing after laser capture microdissection. All progression biopsies were performed prior to discontinuing enza, enabling us to identify resistance mechanisms induced by continued enza treatment.

The study design is shown in Fig. 1a. Patient demographic information and prior treatments are shown in Supplementary Table 1. Bone was the most common site for both pre-treatment and progression biopsies. Eighteen of 21 patients had the same tissue type biopsied at progression. In eight patients, the exact same lesion was biopsied at baseline and progression (Fig. 1b, Supplementary Table 2). The median time on enza treatment was 226 days, shorter than previous trials conducted in this same disease state^6,25. PSA response at 12 weeks and the time between biopsies for each patient are shown in Fig. 1c.