Targeting the androgen receptor pathway in prostate cancer

Summary of recent advances

When prostate cancers progress following androgen depletion therapy, there are currently few treatment options with only one, docetaxel, that has been shown to prolong life. Recent work has shown that castration resistant prostate cancers (CRPC) continue to depend on androgen receptor (AR) signaling which is reactivated despite low serum androgen levels. Currently available AR targeted therapy, including GnRH agonists and antiandrogens, cannot completely shut down AR signaling. Several mechanisms that enhance AR signaling in an androgen depleted environment have been elucidated. These include AR mutations that allow activation by low androgen levels or by other endogenous steroids, AR overexpression, increased local intracrine synthesis of androgens, and upregulation of tyrosine kinase pathways. This has led to the development of a number of novel agents targeting AR signaling pathway, including more effective antiandrogens, inhibitors of CYP17, an enzyme required for androgen synthesis, inhibitors of 5α-reductase, inhibitors of HSP90 which protects AR from degradation, inhibitors of histone deacetylases which is required for optimal AR mediated transcription, as well as inhibitors of tyrosine kinase inhibitors. Many of these strategies are currently being tested in clinical trials in CRPC.

Introduction

The androgen receptor (AR), located on Xq11-12, is a 110 kD nuclear receptor that, upon activation by androgens, mediates transcription of target genes that modulate growth and differentiation of prostate epithelial cells. AR signaling is crucial for the development and maintenance of male reproductive organs including the prostate gland, as genetic males harboring loss of function AR mutations and mice engineered with AR defects do not develop prostates or prostate cancer [1,2]. This dependence of prostate cells on AR signaling continues even upon neoplastic transformation, leading to the seminal discovery by Huggins and Hodges in 1941 that orchiectomy produced prostate cancer regression [3]. Androgen depletion (now using GnRH agonists) continues to be the mainstay of prostate cancer treatment. However androgen depletion is usually effective for a limited duration and prostate cancer evolves to regain the ability to grow despite low levels of circulating androgens [4]. Treatment options for castration resistant prostate cancer (CRPC) are an unmet need with docetaxel being the only agent that has been shown to prolong survival [5,6]. Interestingly, while a small minority of CRPC does bypass the requirement for AR signaling [7], the vast majority of CRPC, though frequently termed “androgen independent prostate cancer” or “hormone refractory prostate cancer,” retains its lineage dependence on AR signaling. Over the past several years, several important mechanisms of enhanced AR signaling in low serum androgen levels in CRPC have been elucidated. This has led to novel therapeutic strategies targeting AR signaling which offer promising potential in future treatment of CRPC (Figure 1).

Figure 1.

Schematic of therapeutic targets of the AR pathway. 1) AR is bound to the molecular chaparonin HSP90 which prevents its degradation. HSP90 inhibitors, such as 17-AAG, cause AR degradation and decrease AR levels. 2) In men treated with GnRH agonists to shut down testicular androgen synthesis, residual serum androgens are synthesized by the adrenal glands. In additional, evidence suggests intratumoral androgen synthesis. Both can be inhibited by the non-specific p450 inhibitor ketoconazole and the 17-lyase inhibitor abiraterone. 3) Testosterone is converted to the more potent dihydrotestosterone (DHT) by 5α-reductase, which is inhibited by finasteride and dutasteride. 4) Ligands, such as DHT bind to AR and this is inhibited by antiandrogens such as bicalutamide and novel agents MDV-3100 and BMS641988. Mutation of AR as well as AR overexpression can convert endogenous steroids (e.g., progestins, estrogens, corticosteroids) and some antiandrogens into agonists. MDV-3100 was designed to suppress AR function even when AR is overexpressed. 5) Activation of receptor tyrosine kinases, in particular HER2, can lead to downstream AR activation. Two downstream kinases that directly phosphorylate AR on tyrosine are Ack1 and SRC. Other downstream pathways of receptor tyrosine kinases, including the AKT and MAP kinase pathways, are also implicated. Antibodies such as trastuzamab and pertuzumab and small molecular TKI inhibitors such as erlotinib and lapatinib target HER2. Dasatinib target SRC. 6) Proper transcription mediated by AR requires the proper chromatin state. HDAC inhibitors inhibit transcription of AR target genes by disruption of chromatin structure and inhibition of recruitment of coactivators and RNA polymerase II.

Castration resistant prostate cancer requires AR signaling

Several clinical observations have long offered clues that AR signaling is active and required in most CRPC. PSA, an exquisitely AR dependent gene, is widely used as a marker for disease activity. PSA declines after initiation of hormone depletion therapy and a subsequent rise is commonly the first sign of disease progression. This indicates that reactivation of AR signaling accompanies the development of CRPC. Both the relative and absolute level of PSA decline—markers of the degree of AR inhibition—after initial androgen depletion is predictive of outcome [8]. After development of castration resistance, further hormonal manipulations targeting AR can elicit response while treatment with exogenous androgens usually results in tumor flare. First demonstrated for flutamide, treatment with any currently available antiandrogen may result in agonist conversion, and tumor response can be observed upon andiandrogen withdrawal (AAWD) [9]. Gene expression studies of clinical prostate cancer specimens show that AR activated genes (defined as genes down-regulated after neoadjuvant androgen deprivation prior to prostatectomy) were reactivated in CRPC despite continued androgen deprivation [10]. In the laboratory, knockdown of AR results in cell death in both human and murine castration resistant prostate cancer cell lines [11–13].

Recently, it was discovered that up to 90% of all prostate cancers overexpress an ets oncogene, including ERG, ETV1, ETV5 and ETV6 via a variety of mechanisms. The most common mechanism of overexpression is fusion of the ets gene (in particular ERG) to the 5’ untranslated region of highly AR regulated TMPRSS2 gene [14,15]. Thus, in addition to the lineage dependence of prostate cells on AR signaling, prostate cancer has additional selection pressure to maintain TMPRSS2 expression and AR activity.

Mechanisms of AR activation in CRPC

Numerous mechanisms have been implicated in reactivation of AR in the castrate environment and have been extensively reviewed (Figure 1) [16–18]. Most directly, mutations of AR that allow other steroids such as corticosteroids and anti-androgens are detected in ~10% of CRPC in a CALGB clinical trial, though the actual incidence may be more frequent [19]. Other mechanisms including activation of kinase pathways that can both stabilize AR and enhance its transcriptional activity and upregulations of AR coactivators that increase AR mediated transcription. These sensitize AR to lower levels of ligand. Here, we focus on three crucial and druggable mechanisms—increased level of AR and increased level of ligand and activation of kinase pathways.

CRPC utilize several sources of ligands

Hormone ablation therapy does not completely eliminate serum androgens. Serum testosterone is reduced to a mean of 15 ng/ml from a normal range of >200 ng/ml while serum levels of adrenal androgens such as dehydroepiandrosterone (DHEA) and androsteindione are unaffected. Intraprostatic androgen concentration is reduced much less dramatically by only ~75%, and is sufficient to activate AR [20,21]. While this level of reduction is sufficient to induce response in untreated prostate cancer, cellular alterations that sensitize the AR pathway induce resistance and confer growth.

One possible source of increased intratumoral androgens are the tumor cells. Two expression profiling studies comparing metastatic CRPC with primary tumors show that enzymes involved in androgen synthesis are upregulated in CRPC. Holzbeierlein et. al. found overexpression of enzymes involved in synthesis of cholesterol, the common steroid precursor, from acetyl-CoA [10] and Stanbrough et. al. found overexpression of enzymes involved in synthesis of testosterone and the more potent androgen DHT from cholesterol [22] (Figure 2). Using mass spectrometry to directly measurement of intratumoral androgens, Montgomery and colleagues found that castration resistant metastatic tumors in men treated with GnRH have higher levels of testosterone but not dihydrotestosterone than primary tumors in untreated men. They also corroborated overexpression of enzymes in androgen synthesis measured by real-time PCR. Castration resistant xenografts similarly maintained elevated intratumoral testosterone levels in castrate mice [23]. These data suggest that intracrine androgen synthesis may allow tumors to grow despite low serum androgen levels.

Figure 2.

Androgen synthesis pathway and therapeutic targets. Cholesterol is synthesized from Acetyl-CoA and enzymes in this pathway were found to be upregulated by Holzbeierlein et. al. (red) [10]. Subsequently, the weak androgen androstenedione, testosterone, and the potent DHT are synthesized from cholesterol and enzymes in this pathway were found to be upregulated by Stanbrough et. al. (blue) and Montgomery et. al. (orange) [22,23]. Abiraterone blocks CYP17 which contain both 17α-hydroxylase/C17,20-lyase activities (boxed orange). Dutasteride blocks both SRD5A1 and SRD5A2 (boxed green).

AR overexpression in CRPC

Overexpression of AR is common in CRPC. Compared to localized disease, CRPC has higher expression of AR based on immunochemical staining with genomic amplification seen in ~20% of cases [24,25]. Expression profiling studies consistently identify AR to be overexpressed in CRPC [10,22,26,27].

Laboratory data indicate that AR overexpression is necessary and sufficient to induce CRPC in xenograft models. To identify genes important for development of castration resistance, a panel of 7 prostate cancer xenografts was selected for castration resistance by passage in castrated mice. Comparison of the expression profiles between the 7 isogenic pairs reveal that AR is the only gene overexpressed in all resistant xenografts. Further, forced modest overexpression of AR in the parental LnCaP and LaPC4 xenografts conferred castration resistance by sensitizing cells to residual levels of androgens restoring expression of AR regulated genes. Intriguingly, overexpression of AR also converted the anti-androgen bicalutamide into a weak agonist, indicating that AR overexpression can underlie AAWD [12].

Activation of kinase pathways in CRPC

The HER2 receptor tyrosine kinase is progressively overexpressed in more advanced, castrate resistance prostate cancers, though it is seldom, if ever, amplified as seen in breast cancers [28]. In experimental systems, forced overexpression of HER2 results in increased AR activity and stability while pharmacologic inhibition or knockdown of the protein results in growth suppression [29]. One possible downstream target of HER2 activation is the Cdc42-associated tyrosine kinase Ack1. Activated Ack1 mediates AR activation through tyrosine phosphorylation of Y267 and knockdown of Ack1 or mutation of Y267 to phenylalanine abrogates HER2 mediated AR activation and growth [30].

In addition to HER2, increased signaling by a number of other growth factor receptors (e.g, EGFR, IGF-1R, IL-6R) can enhance AR signaling and confer castration resistance in preclinical models [31–33]. These receptors induce downstream activation of critical growth and survival pathways, including the AKT, MAPK, and STAT pathways. Expression of both activated AKT and BRAF, also results in castration resistance [13]. While these mechanisms of CRPC are frequently referred to as “ligand independent”, it is unknown whether AR is truly activated without ligand binding or whether AR is sensitized to lower levels of ligands since experimental systems to decrease AR ligands, such as in vitro growth in charcoal stripped serum or castration of mice in vivo leave residual ligands. This distinction between true ligand independent and hypersensitized AR is not just semantic since therapies designed to further reduce AR ligands would be active only if ligand is still required. Furthermore, AR alleles containing mutations that impair ligand binding can no longer confer resistance to castration.

Another kinase implicated in AR crosstalk is SRC. Upon ligand binding, AR binds and activates SRC and downstream events within 5 minutes in a “non-genomic” mechanism [34]. SRC can in turn tyrosine phosphorylate AR augmenting its transcriptional activity. SRC activity is substantially increased in models of CRPC [35,36]. A 10 amino acid peptide that blocks the AR-SRC interaction inhibits androgen mediated proliferation in tissue culture and xenografts [37].

Evolving Treatments for CRPC

Antiandrogens

Upon development of castration resistance, an antiandrogen, such as flutamide, bicalutamide, nilutamide, and cyproterone acetate, is typically added if it was not included in initial treatment. In some cases, this can result in prolonged disease control. However, in the majority of patients who have progressed despite androgen ablation and especially in symptomatic patients, the time to progression is usually short [38,39]. In addition, all these compounds have clinically been observed to convert to agonists [9]. A number of novel antiandrogens are currently in preclinical development.

MDV-3100, a novel antiandrogen, was rationally designed utilizing the AR crystal structure, modeling and cell based screening. Since bicalutamide is converted to an agonist in LnCaP cells that overexpress AR, candidate compounds were screened for their ability to inhibit growth and PSA secretion in these cells. Of the resulting hits, MDV-3100 has the best pharmacological properties of good oral availability and long half-life. MDV-3100 binds AR in cells with 10-fold higher affinity that bicalutamide in competition studies and similarly, it inhibits PSA secretion at 10-fold lower concentrations. Subsequent preclinical studies showed that MDV-3100 completely inhibits growth of both castration-resistant xenografts. Unlike bicalutamide, MDV-3100 impairs AR nuclear translocation and blocks DNA binding. MDV-3100 is currently undergoing Phase I/II clinical trial in CRPC. Early data is highly promising (Table 1).

Table 1.

Clinical trials AR targeting agents in CRPC

Agent	Phase	RESIST Response	PSA Response	Reference
Androgen and DHT lowering compounds
Abiraterone	II (chemo naive)	12/21	18/30 (60%)	JS De Bono, abs 5005, 2008 ASCO Annual Meeting
	II (chemo refractory)	3/8	14/35 (40%)	DC Danila, abs 5019, 2008 ASCO Annual Meeting
	II (chemo refractory)	4/28	16/31 (52%)	JS De Bono, abs 5005, 2008 ASCO Annual Meeting
	III (open)
Dutasteride	II (open)
Dustasteride + ketoconazole	II (keto sensitive)		30/57 (53%)	M Taplin, abs 5068, 2008 ASCO Annual Meeting
Dustasteride + ketoconazole	II (keto refractory)		0/10 (8/10 with PSA decline)	AO Sartor, abs 257, 2007 Prostate Cancer Symposium
Antiandrogens
MDV-3100	I/II		13/14	HI Scher, abs 5006, 2008 ASCO Annual Meeting.
BMS-641988	I
EGFR and HER2 Targeting Agents
Erlotinib	II (closed)
Erlotinib + docetaxel	II (closed)	0/8	5/22	[65]
Gefitinib	II (closed)	0	0/23	[66]
Lapatinib	II (open)		1/21	YE Whang, abs 16037, 2008 ASCO Annual Meeting
Trastuzumab	II (closed)			[67]
Pertuzumab	II	0/30 (5/30 SD >23 wk)	0/41	[68]
SRC Inhibitors
Dasatinib	II	67% disease control	1/36	EY Yu, abs 5156, 2008 ASCO Annual Meeting
HDAC inhibitors
Depsipeptide		1/21, (6/21 SD > 4m)	2/31	C Parker, abs in J Clin Oncol (2007) 25(18_suppl) 15507
LBH589	I		0/8	DE Rathkopf, abs 5152, 2008 ASCO Annual Meeting
SAHA+DOC	I	2/8		DE Rathkopf, abs 5152, 2008 ASCO Annual Meeting
HSP90 Inhibitors
17-AAG	II		0/15	EI Heath, abs in J Clin Oncol (2007) 25(18_suppl) 15553

BMS-641988 is a novel antiandrogen with 20-fold greater affinity for AR than bicalutamide and is effective in two xenograft models that have progressed on bicalutamide. It is currently undergoing Phase I clinical trial in CRPC.

Androgen Lowering Therapies

Since residual serum androgens as well as upregulated intracrine androgen synthesis may be sufficient to promote CRPC growth in patients on hormone ablation therapy, especially when the pathway is sensitized by AR mutation or overexpression, strategies to further lower androgen levels are under investigation. Treatment with aminoglutethimide and ketoconazole, both nonspecific p450 inhibitors that target numerous enzymes in the steroid synthesis pathway (Figure 2), results in 50% PSA reduction in 25–50% of patients with CRPC for a median duration of ~6 months [40]. Analysis of the CALGB 9583 comparing AAWD with AAWD plus ketoconazole showed that patients with higher baseline serum levels of the adrenal androgen androstenedione (Figure 2) had higher response rates and prolonged overall survival with ketoconazole treatment [40,41]. Non-selective p450 inhibitors have significant constitutional side effects, limiting quality of life and treatment duration. Additionally, they interfere with the metabolism of multiple drugs.

Abiraterone is a selective, high affinity (IC50=2nM), irreversible inhibitor of the p450 enzyme, CYP17, which contain both 17α-hydroxylase/C17,20-lyase activities, targeting both adrenal and tumor intracrine androgen synthesis (Figure 2). At tolerable doses, serum testosterone concentration fell to castrate levels in intact mice without significantly affecting serum cortisol levels, whereas ketoconazole suppressed cortisol more than testosterone [42]. While single agent abiraterone can suppress testosterone to castrate levels in the short term, there is some testosterone recovery with long term treatment due to compensatory increase in gonadotropin release. Addition of abiraterone to GnRH treatment results in substantial decrease of both testosterone and adrenal androgens (Figure 2) [43]. Preliminary data from phase I/II trials of patients with CRPC show significant activity with PSA response exceeding 60% in chemotherapy naïve and 40% in chemotherapy treated patients with median time to progression of 8 and 5.5 months respectively (JS deBono, abs 5005 and DC Danila abs 5019 presented in 2008 ASCO Annual Meeting). This has motivated an international randomized phase III randomized trial comparing abiraterone with placebo in patients CRPC who have progressed on docetaxel.

Testosterone is converted to the more potent DHT by two isoforms of 5α-reductase, SRD5A1 and SRD5A2, in peripheral androgen dependent tissues (Figure 2). SRD5A2 is the predominant isoform in the benign prostate and finasteride, a specific inhibitor of SRD5A2 reduces PSA and prostate weight and is approved for use in benign prostatic hypertrophy. However, progressive prostate cancer is characterized by increased SRD5A1 and decreased SRD5A2 [22,44]. Dutasteride, a potent inhibitor of both SRD5A1 and SRD5A2, inhibits cancer growth in the R-3327H rat prostate cancer model and the probasin large-T antigen (TRAMP) mouse prostate tumor model [45,46]. Because dutasteride causes feedback increase in testosterone levels, simultaneous castration may increase its efficacy. In the LnCaP xenograft model, dutasteride but not finasteride inhibits tumor growth more than castration alone [46]. Thus, dutasteride is in phase II clinical trials of various stages of prostate cancer, including CRPC (Table 1).

HSP90 Inhibitors

HSP90, a molecular chaparonin, is required for the refolding of denatured proteins. In addition, it is required to maintain the proper baseline folding of several important proteins in oncogenesis, including AR, Her2, AKT, and mutagenic BRAF. Geldanamycin, a natural compound produced by Streptomyces hygroscopicus, binds the ATP-binding pocket of HSP90 and causes degradation of client proteins [47]. Tanespimycin (17-AAG,), a more stable geldanamycin derivative, inhibits AR positive prostate cancer xenografts without significant toxicity to the murine hosts. Treatment resulted in 80% loss of AR expression and 97% loss of HER2 expression [48]. In an independent unbiased validation that HSP90 inhibition targets AR activity, Hieronymus el. al. screened a library of chemical compounds that decreased expression in AR regulated genes. Two compounds that scored in this screen, celastrol and gedunin, were subsequently discovered to be HSP90 inhibitors and cause the same gene expression perturbations as 17-AAG [49].

In a recent phase I trial, 17-AAG plus trastuzumab (Herceptin) showed significant clinical activity in patients with HER2 positive metastatic breast cancer that have progressed on trastuzumab based therapy, offering a proof of principle that clinically achievable levels HSP90 inhibitors can be active [50]. However, 17-AAG did not show significant clinical activity in CRPC (EI Heath, abstract in J Clin Oncol 2007, 18_suppl:15553). One limitation with 17-AAG is its insolubility in water and thus difficulty obtaining high serum concentrations. More water-soluble geldanamycin analogs, such as 17-DMAG (alvespimycin) as well as combination strategies are currently under clinical investigation.

HDAC Inhibitors

Gene transcription depends on a complex interplay between transcription factors, such as AR and a chromatin state that support active gene transcription [51]. The fundamental unit of chromatin, the nucleosome, consists of 146 base pairs of DNA wrapped around an octamer of highly conserved core histone proteins (two each of H2A, H2B, H3, and H4). All four core histones have a number of positively charged lysine residues in their amino-terminal tail that form tight interactions with negatively charged DNA. The majority of these lysines can undergo acetylation. Acetylation of lysine residues is regulated by the balance of histone aceyltransferases (HAT) and histone deacetylases (HDAC), akin to the regulation of phosphorylation by kinases and phosphatases. More recently, numerous other proteins have been found to be acetylated, including p53, HSP90 and AR.

Naturally occurring antibiotics including trichostatin A, depsipeptide and the synthetic polar compound suberoylanilide hydroxamic acid (SAHA) were found to induce differentiation and to be selective more toxic in tumor cells. These agents were found to inhibit HDAC, and as expected, induce hyperacetylation of histones and transcriptional changes. HDAC inhibitors have demonstrated encouraging anti-tumor activity and SAHA has been approved in the treatment of cutaneous T-cell lymphomas [52].

HDAC inhibitors have been noted to have greater anti-proliferative effects on AR-positive prostate cancer cells than their AR-negative counterparts, and inhibit xenograft growth in both castration sensitive and resistant models [53,54]. One proposed mechanism is that HDAC inhibitors target HDAC6 which deacetylates HSP90 and decreases AR stability [55]. Furthermore, HDAC inhibitors directly suppress AR transcription [54,56].

While these mechanisms are important, we have further observed that HDAC inhibitors directly inhibit the transcription of approximately half of AR target genes, even in CRPC models where AR is overexpressed and present at high levels that cannot be decreased by HSP90 or AR transcriptional inhibition. These genes include the TMPRSS-ERG fusion implicated in prostate oncogenesis as well as well characterized AR response genes such as PSA and KLK2 (DS Welsbie et. al., unpublished).

Several HDAC inhibitors, including depsipeptide, SAHA, and LBH589 are in phase I/II clinical trials in CRPC. So far, the observed activity is modest at best (Table I). The low activity may be due to suboptimal dosing as low concentration of HDAC inhibitor may paradoxically activate AR in tissue culture models [57]. This has motivated trials designed to increase HDAC inhibitor concentration with shorter exposure as well as combination trials with chemotherapy.

Kinase Inhibitors

Agents targeting EGFR or HER2, including small molecule kinase inhibitors (gefitinib, erlotinib, lapatinib) and monoclonal antibodies (trastuzumab, pertuzamab) have been studied clinically in CRPC but showed disappointing results (Table 1). Since these inhibitors are most active in tumors where the target is mutated or amplified (e.g., EGFR mutation in lung cancer, HER2 amplification in breast cancer), one explanation is that EGFR/HER2 is not a relevant target in CRPC despite showing activity in preclinical models. Alternatively, there is the mounting evidence that loss of PTEN mediates resistance to EGFR and HER2 targeted therapies in both breast cancer and glioblastoma [58–61]. PTEN loss is common in high grade and metastatic CRPC and mediates early development of castration-resistance in mouse models [62,63]. Therefore, combination treatment with novel PI3K inhibitors with ERFR/HER2 inhibitors may be warranted.

Dasatinib, currently approved as an ABL kinase inhibitor for treatment of chronic myelogenous leukemia [64], is also a namomolar SRC inhibitor. Phase II clinical trials of dasatinib in CRPC are ongoing.

Conclusion

With increasing appreciation that AR signaling remain crucial in CRPC, a number of new therapeutic strategies have evolved. Phase I–II clinical data show that some have very promising activity while others have been disappointing. As expected from the fact that multiple mechanisms can underlie AR activation, no single therapeutic agent is active in all patients. The ability to dissect AR pathway aberrations in patients would allow individualized therapy targeting the particular aberration.