Emerging Mechanisms of Resistance to Androgen Receptor Inhibitors in Prostate Cancer

Introduction

Prostate cancer (PCa) is one of the most commonly diagnosed cancers among men in Western industrialized nations and a leading cause of cancer-related deaths ¹ (http://seer.cancer.gov/statfacts/html/prost.html). In the era of routine screening of serum prostate specific antigen (PSA) levels, the majority of cancers are detected while organ-confined and hence potentially curable, while the incidence of lethal metastatic disease at the time of diagnosis has declined ². Androgen synthesis is under the physiological regulation of the hypothalamic-pituitary-testicular axis (Figure 1A), with contributions from the adrenal glands resulting from de novo steroidogenesis (Figure 1B). Pioneering work by Charles Huggins in 1941 demonstrated the remarkable benefit of androgen deprivation therapy (ADT) via surgical castration for men with advanced metastatic PCa ³, establishing a clinical paradigm that continues to this day. Contemporary first-line ADT for PCa recurring after prostatectomy or radiotherapy typically involves chemical castration through the chronic use of gonadotropin-releasing hormone (GnRH) agonists or antagonists (Table 1) that lower testosterone levels by stable suppression of androgen secretion from the testes (Figure 1A). Combined androgen blockade incorporates the additional use of a competitive androgen receptor (AR) antagonist (an antiandrogen) (Table 1) to further impede AR signaling within the PCa cell (Figure 1C) and mitigate the effects of acute systemic testosterone surges resulting from the initial use of GnRH agonists ⁴. Although nearly all patients respond to hormonal therapy, response duration varies from months to years followed by uniform progression to a lethal stage of the disease, termed castration resistant prostate cancer (CRPC) (Figure 2A).

Just over a decade ago, it was generally believed that AR signaling was dispensable to the biology of CRPC. This led to the frequent usage of terms such as “androgen independent” or “hormone refractory” to describe this stage of the disease. An abundance of data acquired since then, however, has made it overwhelmingly clear that residual androgens remaining after castration and AR itself remain central both to the progression to CRPC and to its continued growth. An early indication of the possible contribution of AR to the progression towards CRPC came from the observation that 30% of CRPC patients harbored genomic amplification of the AR locus in late stage tumors but not in patient-matched tumor samples obtained prior to ADT ⁵. In in vitro and in vivo studies using the preclinical PCa models LNCaP and LAPC4 (Box 1), our laboratory established that AR overexpression was indeed a sufficient and principal driver of progression to CRPC, with these cells exhibiting acquired resistance to both ADT and to the primary antiandrogen in clinical usage at that time, bicalutamide ⁶. These findings provided the rationale for a drug discovery screen of novel antiandrogens that would maintain the ability to inhibit AR signaling in the setting of receptor overexpression, which led to the identification of enzalutamide (formerly MDV3100) ⁷. In parallel, others developed the CYP17A1 inhibitor, abiraterone acetate (hereafter, simply abiraterone), which targets this central enzyme in de novo steroidogenesis (Figure 1B) ⁸.

Abiraterone and enzalutamide are both approved for treatment of CRPC in chemotherapy-naive and chemotherapy refractory patients, based on a series of phase III trials showing overall improved survival for either agent used alone versus placebo ^9–12. Despite the success of these second-generation AR targeted therapies, inherent or acquired resistance remains a major clinical challenge. Here we will review the current understanding of resistance to contemporary AR inhibition strategies, which we group into the three general categories of restored AR signaling, bypass of AR and complete AR independence (Figure 2B). We also discuss implications for the development of the next generation of molecularly targeted therapies for PCa.

Restored Androgen Receptor Signaling

Androgen Receptor Bypass Signaling

Recent work reveals a novel AR pathway resistance mechanism analogous to one originally described for kinase inhibitors, in which signaling downstream of the targeted kinase is restored by activation of a related kinase not targeted by the inhibitor ⁷¹. In the context of kinase inhibitors, the clinical impact of this escape mechanism is well established in epidermal growth factor receptor (EGFR)-mutant lung cancer and BRAF-mutant melanoma and is widely referred to as “bypass” signaling to emphasize the sustained importance of the initial oncogenic pathway now activated by a different driver. Two groups have now documented an analogous mechanism for hormone receptors ^{72, 73}. [To avoid confusion, we note that the term “bypass” was used in earlier reviews of castration resistance to describe mechanisms completely independent of AR ^{74, 75}. In light of the contemporary analogy with kinase inhibitors, we suggest that “bypass” in this context is better suited to refer to mechanisms in which downstream hormone receptor pathway signaling remains relevant but through activation by a different hormone receptor, as described below.] In the LNCaP xenograft model with exogenous AR overexpression (LNCaP-AR) ⁶, acquired resistance to enzalutamide or ARN-509 correlated with upregulation of the glucocorticoid receptor (GR) as revealed by transcriptome analysis ⁷³. A LNCaP-AR subline termed LREX′, with acquired resistance to enzalutamide, was shown to be dependent on GR expression for enzalutamide-resistant growth. In VCaP cells, glucocorticoid mediated activation of the comparatively lower level of endogenous GR was sufficient to confer enzalutamide resistance. ChIP-seq and mRNA expression analysis for AR and GR revealed highly overlapping cistrome and transcriptome profiles for both receptors ^{73, 76}. In the resistant LREX′ tumors, GR induction was associated with restored expression of a restricted subset of AR target genes that are presumed to mediate resistance. Analysis of bone marrow biopsies from patients treated with enzalutamide supported a role for GR induction in clinical resistance to enzalutamide ⁷³. Recent data presented at the 2015 ASCO Annual Meeting suggested that GR bypass may occur in earlier stages of disease. Tumor cells in men with high-risk localized PCa with early resistance to neoadjuvant chemical castration plus abiraterone also expressed significant levels of GR ⁷⁷. It is worth highlighting that active AR inhibition is necessary to maintain high levels of GR expression in preclinical models, due to active repression of GR expression by AR binding to the GR locus. For this reason, it may be important to obtain clinical specimens from patients undergoing active antiandrogen treatment to fully evaluate the importance of GR as a resistance mechanism ⁷³.

At first glance, the hypothesis that GR can confer resistance may seem inconsistent with clinical evidence that glucocorticoid administration can be beneficial to some CRPC patients. This apparent paradox is explained by the fact that glucocorticoids inhibit adrenocorticotropic hormone (ACTH) production by the pituitary which results in reduced androgen levels (Figure 4A) ⁷⁸. This androgen lowering activity explains declines in serum PSA level observed in men taking prednisone alone, which was documented in the comparator arm of the phase III clinical trial that led to abiraterone approval for chemotherapy naive CRPC ⁷⁹. However, in men whose prostate cancers express high levels of GR, this androgen lowering benefit would be counteracted by GR activation in tumor cells (Figure 4B). In this setting, a more effective treatment strategy could be combined inhibition of AR and GR, as is currently being explored in an early phase clinical trial of enzalutamide with the GR antagonist mifepristone (NCT02012296). One potential confounder of this study is the fact that mifepristone also has a high binding affinity for AR and can function as an AR agonist ⁸⁰. Therefore, mifepristone treatment could unintentionally result in AR activation through agonism by displacing the potent antagonism of enzalutamide. Also of concern is the fact that mifepristone treatment resulted in higher androgen levels in an earlier single agent phase II study ⁸¹, likely due to GR inhibition in the pituitary gland, with a subsequent increase in ACTH and adrenal androgens (the opposite effect of glucocorticoid administration). It will be important to document combined AR and GR inhibition in tumor cells from patients treated in the ongoing combination therapy trial; otherwise, a lack of clinical benefit could be due to AR re-activation by mifepristone.

In addition to GR, the progesterone receptor (PGR) and the mineralocorticoid receptor (MR) are also steroid hormone nuclear receptor family members structurally related to AR, sharing substantial homology within the DNA binding domain ⁸². As with GR, it is possible that PGR or MR could transcriptionally regulate a subset of AR target genes in PCa, and thereby bypass AR. PGR expression has been demonstrated in prostate tumor cells in some ^{83, 84} although not all ^{85, 86} studies. Interestingly, high PGR staining in primary PCa was associated with clinical recurrence in a recent, large retrospective analysis ⁸³. There is less evidence implicating MR, but it is worth noting that MR was among the top upregulated genes in VCaP xenograft tumors with acquired resistance to abiraterone ⁴⁷.

Complete Androgen Receptor Independence

It has long been known that metastatic CRPC displays a remarkable degree of inter- and intra-patient molecular heterogeneity ^{19, 87–89}. Heterogeneity also extends to the distribution and intensity of AR expression, as revealed by immunohistochemical studies of CRPC bone metastases ^{88, 90–92}. For example, in a study with 44 CRPC bone metastases, 58.1% of patients had moderate (30.4% of cases) to intense (69.6% of cases) AR staining in 76–100% of the tumor cells, whereas 8.8% of the patients in the study had AR staining in just 1–25% of tumor cells ⁹². Thus, metastatic CRPC can exist as a mixture of cells displaying a range of AR expression levels.

With the growing clinical use of abiraterone and enzalutamide, it is increasingly appreciated that some men can relapse with clinically aggressive variants of PCa with reduced or absent AR expression. While the precise histological classification of these subtypes continues to be refined, they are often found to express markers of neuroendocrine differentiation (NED) (chromogranin A, synaptophysin and neural cell adhesion molecule) and may also show histological features of small cell carcinoma (SCC), a rare variant of AR-negative primary PCa also displaying NED ^93–95. Adding further complexity, tumor cells with NED can often be found mixed with usual adenocarcinoma cells ⁹⁵, but the relevance of these subpopulations to the disease course is unclear. Molecular profiling of PCa with NED has revealed loss of RB1, PTEN, and TP53 mutations as well as amplification of MYCN and Aurora kinase A (AURKA) ^{96, 97}. MYCN overexpression in LNCaP resulted in the induction of NED features concurrent with downregulation of AR and AR target genes ⁹⁷. Conditional deletion studies in mice provide strong evidence of a causal role for loss of both RB1 and TP53 in the genesis of metastatic CRPC with NED, although tumors in these mice retained heterogeneous expression of AR and luminal epithelial cytokeratins ⁹⁸. On the other hand, RB1, PTEN, and TP53 deletions and mutations are observed in CRPC with usual adenocarcinoma histology, so the association with NED is not absolute ¹⁹.

It is unclear whether AR-negative PCa arise directly from typical AR-positive adenocarcinomas by a process of transdifferentiation or instead from a population of AR-negative neuroendocrine cells present in the normal prostate. Evidence supporting transdifferentiation comes from multiple studies showing the presence of the AR-regulated TMPRSS2-ERG genomic translocation by fluorescence in situ hybridization in AR-negative SCC ^99–101 at a frequency akin to that seen in AR-positive adenocarcinoma ¹⁰². These findings are consistent with earlier observations that LNCaP cells can acquire expression of some neuroendocrine markers upon prolonged ADT ¹⁰³.

Since CRPC undergoes temporal clonal selection in response to treatment ³³, there is concern that the widespread and long-term usage of next generation AR inhibitors could increase the prevalence of PCa with loss of AR and NED. At present, it is still too early to know whether this concern is justified. One recent genomic landscape study of 150 CRPC patients, including many with prior exposure to abiraterone or enzalutamide, suggests this may not be the case because more than 96% of cases had usual adenocarcinoma histology. Subtypes of adenocarcinoma with NED comprised just 2.9% of cases, while SCC was only 0.7% ¹⁹. However, another study presented at the 2015 ASCO Annual Meeting of 101 cases of CRPC resistant to abiraterone or enzalutamide reported that only 33% displayed typical adenocarcinoma histology, with 12% SCC and an intermediate histology distinct from either in 27% of cases ¹⁰⁴. Further analyses of contemporary CRPC cohorts, with careful attention to potentially new emerging histological subtypes, will be critical in resolving this issue.

Future Therapeutic Options for CRPC

Despite the clinical success of abiraterone and enzalutamide as single agent therapy for metastatic CRPC, it is clear that additional therapeutic advances are needed. One possibility is that the efficacy of both drugs may increase substantially when they are used much earlier in the disease course, as has been seen routinely with kinase inhibitors for various cancers. Clinical trials addressing this question are underway in men with pre-metastatic CRPC (also called M0 disease) and as front line therapy in men with hormone sensitive disease (NCT02058706, NCT01664923, NCT02003924, NCT01927627, NCT01715285, NCT01957436, NCT02064582, NCT02023463, NCT02028988, NCT01751451, NCT01023061, NCT02203695). In addition, it is possible that combination therapy with abiraterone and enzalutamide may be more effective than either drug used alone since they attack AR signaling by distinct mechanisms. This hypothesis is also being tested in several clinical trials (NCT01949337, NCT01650194, NCT02268175, NCT02125357). Sequential studies using one of these agents, followed by second-line treatment with the other, have shown responses in some patients but the overall impact has been modest ^105–110. Recent findings that a metabolite of abiraterone, Δ⁴-abiraterone, is also a potent AR antagonist on par with enzalutamide may provide an explanation for cross-resistance between enzalutamide and abiraterone ¹¹¹.

Although sequential or combined abiraterone and enzalutamide therapy may be insufficient to control CRPC, recent genomic landscape studies employing either tumor biopsies or circulating cell-free DNA sampling underscore the sustained importance of AR in late stage disease, with AR amplification present in 45–52% of the cancers ^{19, 29}. These results justify continued endeavors to discover novel AR focused treatment strategies. The most advanced efforts are with compounds that continue to be directed against the LBD of AR and CYP17A1, but it remains to be seen whether these agents can overcome the cross-resistance seen after prior abiraterone or enzalutamide treatment. Several novel AR targeting methods are under investigation that bypass the conventional approach of interfering with ligand-mediated AR activation. These include small molecule inhibitors directed against the amino terminal transactivation domain ¹¹² or the DNA binding domain ¹¹³. These molecules offer the additional advantage of efficacy against all isoforms of AR, including ARVs. Another strategy is to pharmacologically trigger AR degradation, given the analogous success with the estrogen receptor degrader fulvestrant, used for the treatment of metastatic breast cancer progressing after failure of first-line antiestrogen therapy ¹¹⁴. AZD3514, a purported AR degrader, showed PSA declines in 13% of CRPC patients but clinical development was halted due to gastrointestinal toxicity ¹¹⁵.

In addition to the sustained role of AR, another important insight from CRPC genomic landscape studies is the number of molecular alterations in other actionable pathways, including PI3K-AKT-PTEN, RAF, WNT, DNA repair and the cell cycle ¹⁹ (cBioPortal; www.cbioportal.org). Numerous inhibitors of the PI3K-AKT-mTOR pathway have shown activity in preclinical models, often in combination with next generation AR therapy, and are under study in clinical trials in CRPC ^116–118. Perhaps the most unexpected finding is the high frequency of germline and somatic mutations in genes encoding proteins in DNA repair pathways, such as BRCA1, BRCA2 and ATM. BRCA1 and BRCA2-deficient tumor cells are deprived of normal repair of DNA breaks through homologous recombination, and as such, become highly sensitized to inhibitors of the DNA repair enzyme, poly(ADP-ribose) polymerase (PARP1) ^{119, 120}. One such inhibitor, olaparib, is now approved for treatment of advanced ovarian cancer in women with germline mutations in BRCA1 or BRCA2 and others are in late stage clinical development. The ovarian cancer experience has sparked analogous PARP inhibitor trials in CRPC ^121–124. Although still early, overall response rates have been encouraging including several exceptional responses in men with BRCA mutations.

Looking forward, it seems likely that the standard of care for CRPC will evolve toward distinct molecular subclasses with individualized therapies, as has already occurred in other tumor types such as lung adenocarcinoma. We base this prediction on the unexpectedly high percentage of patients with potentially actionable mutations ¹⁹ together with the now demonstrable feasibility of testing for these mutations in biopsies of metastatic lesions as well as in circulating tumor cells and circulating cell-free DNA ^{29, 125}. Analogous to lung cancer and EGFR inhibitors, this shift in clinical practice will be driven by a compelling new treatment option. Assuming the early clinical data are confirmed, this will likely be treatment with PARP inhibitors, and perhaps cisplatin chemotherapy, in patients with BRCA1 or BRCA2 mutations. We anticipate this approach will rapidly expand beyond traditional single gene tests (companion diagnostics) to multi-gene sequencing platforms capable of identifying an array of mutations that will steer patients to appropriate clinical trials. Based on the high frequency of mutations observed in PI3K pathway genes (PTEN, PIK3CA and PIK3CB) in CRPC, we predict increased focus on clinical trials of PI3K inhibitors, particularly those selective for the p110α and p110β isoforms, in this subset of patients. Patients with complete AR independence are more challenging as there is currently little insight into actionable mutations for this subset. With growing evidence that lineage plasticity or transdifferentiation (and associated changes in chromatin landscape) may precipitate the transition to complete AR independence, the focus may shift to preventing this transition with drugs targeting chromatin modifying enzymes. Whether AR-focused treatment will remain the backbone for all men with metastatic prostate cancer remains to be determined.

Summary

The survival advantage seen from treatment with either enzalutamide or abiraterone in patients with metastatic CRPC has further solidified the importance of AR even in late stage disease. However, inherent or acquired resistance to these agents remains a major clinical obstacle and a greater understanding of biomarkers of response as well as mechanisms of resistance is urgently needed. Multiple mechanisms of resistance to AR targeted therapies have been identified (AR overexpression with or without amplification, AR mutations, ARVs, intratumoral DHT synthesis, GR overexpression and loss of AR) and others undoubtedly remain to be discovered. Due to the multiclonal and heterogeneous nature of PCa, it is probable that multiple mechanisms of resistance may be operating concurrently in any given patient, and these may also change temporally in response to sequential treatments. Clinical trials with biopsies of metastases before the onset of a new treatment and again at emergence of resistance, coupled with integrative genomic analysis, should help to identify these evolving resistance mechanisms, which could then ideally be acted upon to improve patient outcomes. In addition, insights gained through the ongoing efforts to classify molecular subtypes of PCa according to their genomic profiles should continue to identify candidate driver mutations that will better inform clinical trial design. In the long term, success will most likely come from early and aggressive treatment of high risk patients with combinations tailored to prevent resistance before tumors evolve to genomically complex stages.

Supplementary Material

Biographies