Blog

By Jeffrey D. Thomas; Charles G. Longen; Halley M. Oyer; Nan Chen; Christina M. Maher; Joseph M. Salvino; Blase Kania; Kelsey N. Anderson; William F. Ostrander; Karen E. Knudsen; and Felix J. Kim

Excerpt from the article published in Cancer Res (2017) 77 (9): 2439–2452. DOI: https://doi.org/10.1158/0008-5472.CAN-16-1055

Editor’s Highlights

• Androgen receptor (AR) is the primary driver of prostate cancer progression, and at all stages of disease prostate cancer cells are dependent upon AR signaling for growth and survival.
• Sigma1 is required for prostate cancer cell growth and survival
• Negative modulators of Sigma1 (inhibitors) disrupt protein homeostasis in cancer cells and thus suppress their growth and survival.
• Sigma1 inhibitor prevents nuclear transport of cytoplasmic and induces ubiquitin proteasome–mediated degradation of AR.
• Sigma1 inhibitor, induced protein elimination and eliminates AR and AR splice variants and suppresses ARV transcriptional activity

Abstract

Suppression of androgen receptor (AR) activity in prostate cancer by androgen depletion or direct AR antagonist treatment, although initially effective, leads to incurable castration-resistant prostate cancer (CRPC) via compensatory mechanisms including resurgence of AR and AR splice variant (ARV) signaling. Emerging evidence suggests that Sigma1 (also known as sigma-1 receptor) is a unique chaperone or scaffolding protein that contributes to cellular protein homeostasis. We reported previously that some Sigma1-selective small molecules can be used to pharmacologically modulate protein homeostasis pathways. We hypothesized that these Sigma1-mediated responses could be exploited to suppress AR protein levels and activity. Here we demonstrate that treatment with a small-molecule Sigma1 inhibitor prevented 5α- dihydrotestosterone-mediated nuclear translocation of AR and induced proteasomal degradation of AR and ARV, suppressing the transcriptional activity and protein levels of both full-length and splice-variant AR. Consistent with these data, RNAi knockdown of Sigma1 resulted in decreased AR levels and transcriptional activity. Furthermore, Sigma1 physically associated with ARV7 and AR^v567es as well as full-length AR. Treatment of mice xenografted with ARV-driven CRPC tumors with a drug-like small-molecule Sigma1 inhibitor significantly inhibited tumor growth associated with elimination of AR and ARV7 in responsive tumors. Together, our data show that Sigma1 modulators can be used to suppress AR/ARV–driven prostate cancer cells via regulation of pharmacologically responsive Sigma1-AR/ARV interactions, both in vitro and in vivo. 

Introduction

Androgen receptor (AR) is the primary driver of prostate cancer progression, and at all stages of disease prostate cancer cells are dependent upon AR signaling for growth and survival (1, 2). First-line therapy involves suppression of AR activity by androgen depletion or direct AR ligand treatment. However, incurable castration-resistant tumors develop as a result of resurgent AR activity attributable to a series of molecular alterations including: increased expression of AR or AR cofactors, mutations that facilitate promiscuous AR activation by nonandrogen steroid hormones, or constitutively active AR splice variants that are nonresponsive to androgen ablation or AR ligand treatment (1–4). Furthermore, intratumoral androgen biosynthesis and compensatory activation of multiple kinase pathways have been implicated in reactivation of the AR axis in castration-resistant prostate cancer (CRPC; refs. 2, 3). The adaptive nature of the AR axis in prostate cancer cells underscores the importance of discovering and developing novel approaches to overcome the inevitable resistance to current AR-targeted therapies by targeting AR and the networks on which it depends.

In prostate cancer cells, the cascade of events leading to AR transcriptional activity is initiated by binding to dihydrotestosterone (DHT), which is converted from testosterone by the enzyme 5α-reductase (2, 4). DHT binding to the carboxy-terminal androgen binding domain results in AR conformational changes, dimerization, and transport from the cytoplasm to the nucleus, wherein it interacts with transcriptional cofactors and is assembled into DNA-bound protein complexes that produce AR-driven transcriptional activity (2). In the absence of bound androgens, newly synthesized AR is held in the cytoplasm by multichaperone complexes, including HSPs, which are thought to stabilize cytoplasmic AR. Protein chaperones, including HSP90, bind to and facilitate the processing, assembly, and transport of AR and associated proteins. Disruption of AR-chaperone associations can destabilize AR and result in subsequent degradation by the ubiquitin proteasome system (UPS; refs. 5, 6).

Multiple AR splice variants (ARV) have been identified in CRPC tumors and cell lines (2, 4). The best characterized ARVs are the ARV7 and AR^v567es splice variants, both of which lack much of the carboxy-terminal region of AR, including the ligand-binding domain (LBD; refs. 2, 4). Therefore, the activities of these ARV are hormone independent, and LBD-targeted therapeutic agents, such as enzalutamide, are ineffective against these variants (2, 4).

Sigma1 (traditionally known as the sigma-1 receptor) is a unique 26 kDa integral membrane protein that more recently has been described as a molecular chaperone (7–9). Sigma1 is highly expressed in various cancer cell lines and can be detected throughout the cytoplasm, primarily as an endoplasmic reticulum (ER) protein with a long cytoplasmic tail (7, 8, 10). The physiologic role of Sigma1 in cancer cells is still unclear. However, emerging evidence from our laboratory and others (9, 11) suggest that Sigma1 may in fact function as a novel molecular chaperone or a scaffolding protein in cancer cells. We have found that Sigma1 is involved in the aspects of cancer cell protein homeostasis including protein synthesis, folding, transport, and degradation (12, 13).

Furthermore, we have found that these Sigma1 protein homeostasis regulating functions can be pharmacologically modulated by selective small-molecule compounds. Negative modulators of Sigma1 (inhibitors) disrupt protein homeostasis in cancer cells and thus suppress their growth and survival. Here we asked whether we could exploit these small-molecule Sigma1 modulator responses to suppress AR levels and activity in hormone-sensitive and CRPC cells.

…

Results

Sigma1 is required for prostate cancer cell growth and survival

First, we confirmed that RNAi or pharmacologic inhibition of Sigma1 suppresses prostate cancer cell growth and survival. In these studies, we evaluated a panel of clinically relevant prostate cancer cell lines, including a hormone therapy–sensitive model of prostate cancer, LNCaP, which exhibits ligand-dependent AR activity in response to dihydrotestosterone (DHT). LNCaP harbors a gain-of-function AR mutation (T877A) that occurs with significant frequency in human disease (15). We also tested the CRPC model cell lines, C4-2 (16, 17), VCaP (18), and 22Rv1 (19–21). The VCaP and 22Rv1 cells express splice variant-derived forms of the AR, including the AR^v567es and ARV7 splice variants that are constitutively active and refractory to AR-targeted therapeutics (19–21).

Using an anti-Sigma1 antibody generated in our laboratory (12, 13), we first confirmed Sigma1 protein expression in the cytoplasm of AR-positive prostate adenocarcinoma cell lines by confocal immunofluorescence microscopy (data shown for C4-2 in Fig. 1A and B). We confirmed specificity of the Sigma1 immunostain by Sigma1-blocking peptide preadsorption (Fig. 1A). We then confirmed the ability of treatment with Sigma1 siRNA to significantly reduce Sigma1 protein expression (Fig. 1B and C).

Next, we performed siRNA-mediated knockdown experiments to determine whether Sigma1 is required for prostate cancer cell growth and survival. Clonogenic growth was suppressed in Sigma1 knockdown cell cultures, shown for C4-2 in Fig. 1D and quantified for C4-2, LNCaP, and 22Rv1 using a crystal violet staining assay in Fig. 1E. Consistent with the RNAi results, we found that treatment with a prototypic, selective small-molecule inhibitor of Sigma1, IPAG, suppressed prostate cancer cell growth and survival in a dose-responsive manner in both 3D soft agar growth and 2D monolayer growth assays (Fig. 1F and G, respectively). Interestingly, IPAG inhibited prostate cancer cell growth more potently in 3D culture compared with 2D monolayer culture (Fig. 1F and G), with an IC₅₀ of 3 μmol/L for all cell lines under 3D growth conditions compared with IC₅₀ values of 6, 7, and 10 μmol/L for LNCaP, C4-2, and 22Rv1, respectively, under 2D growth conditions. The difference in response may be due to the growth conditions as well as the time of treatment: the 2D assay measures inhibition of cell survival after 72 hours of treatment with IPAG, whereas the 3D assay measures inhibition of cell survival after approximately 3–4 weeks of treatment. The Sigma1-selective actions of IPAG have been confirmed elsewhere (13, 22).

Importantly, small-molecule inhibition of Sigma1 produced time-dependent cell death. Significant cell death was observed between 24 and 48 hours after initiation of treatment of LNCaP cells with 10 μmol/L IPAG (Fig. 1H). On the basis of this data, we performed all subsequent mechanism-focused experiments well within the 24-hour time point to minimize potential confounds associated with cell death.

Androgen receptor transcriptional activity is suppressed by RNAi-mediated knockdown and small-molecule inhibition of Sigma1

Subsequently, we asked whether the growth- and survival-suppressing effects of Sigma1 inhibitor treatment corresponded with changes to AR transcriptional activity. We used qRT-PCR to evaluate changes in the transcript levels of the AR target genes KLK3/PSA and TMPRSS2. For these studies, we evaluated the same panel of prostate cancer cell lines described above, with the addition of VCaP cells, which harbor an amplification of the AR locus and the TMPRSS2-ERG fusion, which renders ERG expression under AR control (23).

We found significant reduction of KLK3/PSA and TMPRSS2 transcript levels in both LNCaP and VCaP cells following treatment with 20 μmol/L IPAG for 16 hours under steady-state, standard culture conditions, that is, in FBS-containing growth medium (Fig. 2A).

To evaluate the effects of Sigma1 inhibition specifically on androgen-induced AR transcriptional activity, we maintained LNCaP cells in culture medium supplemented with CSS to mimic androgen deprivation and treated these cells with DHT in the absence or presence of IPAG (Fig. 2B). After 3 days of growth in CSS medium, treatment with 1 nmol/L DHT for 3 hours induced KLK3/PSA and TMPRSS2 transcript levels (Fig. 2B). Treatment with IPAG alone for 3 hours reduced KLK3/PSA and TMPRSS2 transcript levels below control levels (Fig. 2B). Cotreatment with IPAG blocked DHT induction of KLK3/PSA and TMPRSS2 transcripts, maintaining transcripts close to control levels (Fig. 2B).

To measure effects on direct AR transcriptional activity, we used a luciferase-based AR reporter assay. In this assay, 16-hour treatment with IPAG dose responsively inhibited an AR agonist, 6αFIT, with an IC₅₀ of 6 μmol/L (Fig. 2C).

Consistent with small-molecule inhibition of Sigma1, siRNA knockdown of Sigma1 in LNCaP cells significantly suppressed KLK3/PSA and TMPRSS2 transcript levels (Fig. 2D). Two different commercial Sigma1 siRNA produced decreases in KLK3/PSA and TMPRSS2 transcript levels (data shown for SCBT siRNA in Fig. 2D).

Altogether, these data demonstrate that both RNAi-mediated knockdown and pharmacologic inhibition of Sigma1 suppress AR transcriptional activity.

Treatment with a Sigma1 inhibitor prevents nuclear transport of cytoplasmic AR

As Sigma1 is present throughout the cytoplasm of these prostate cancer cells (Fig. 1), and as nascent AR is stabilized by its cognate chaperones in the cytoplasm, we asked whether the decreased transcriptional activity in response to pharmacologic inhibition of Sigma1 could be explained by inhibition of DHT-induced translocation of AR to the nucleus. To evaluate this effect, we generated an LNCaP cell line stably transfected with a GFP-tagged androgen receptor (GFP-AR) construct. Using this LNCaP (GFP-AR) cell line, we monitored and evaluated DHT-induced nuclear localization of GFP-AR in the absence and presence of IPAG (Fig. 2E and F). LNCaP (GFP-AR) cells were cultured in CSS-supplemented medium for 3–5 days then treated for 3 hours with 1 nmol/L DHT alone or in the presence of 10 μmol/L IPAG (Fig. 2E and F). Treatment with DHT alone for 3 hours resulted in an increase from 34% ± 1% to 68% ± 2% nuclear GFP-AR (Fig. 2E and F). Cotreatment with 10 μmol/L IPAG completely abrogated this effect, maintaining the level of nuclear GFP-AR at 33% ± 1% (Fig. 2E and F). Thus, the small-molecule Sigma1 inhibitor prevented DHT-induced nuclear translocation of AR.

Sigma1 inhibitor induces ubiquitin proteasome–mediated degradation of AR

Subsequently, we investigated the fate of AR following extended treatment (16 hours) with IPAG or knockdown with Sigma1 siRNA. Knockdown of Sigma1 in LNCaP cells suppressed AR protein levels, under standard culture conditions (Fig. 3A). Consistent with Sigma1 siRNA, treatment with IPAG decreases steady-state AR protein levels in all five cell lines examined (Fig. 3B). In LNCaP cells, treatment with IPAG resulted in a >75% decrease in AR protein levels (Fig. 3B). IPAG treatment suppressed AR to similar levels in LAPC4 (wild-type AR expressing) cells (24) and castration-resistant C4-2 and 22Rv1 cells and clearly suppressed AR in VCaP cells as well (Fig. 3B). Interestingly, treatment with IPAG also eliminated splice variant AR (ARV), including ARV7, as well as full-length AR in ARV-bearing 22Rv1 and VCaP cells under CSS medium culture conditions (Fig. 3B). Densitometry of immunoblots in Fig. 3B are presented in Supplementary Fig. S2.

We previously found that extended treatment with IPAG could suppress unfolded protein response (UPR)-associated protein translation (12, 13). Therefore, we evaluated the potential and relative contribution of protein synthesis and degradation by treating with IPAG in the presence of the translation inhibitor, cycloheximide (CHX). Whereas translation arrest by CHX alone resulted in a time-dependent decrease in AR protein levels from 95% to 11% over 24 hours, the combination of CHX and IPAG resulted in decreases that ranged from 80% to 1% of control levels (Fig. 3C). These data suggest that although translation arrest may contribute to decreased AR levels in response to IPAG treatment, protein degradation plays a significant role in the elimination of AR.

We previously found that extended treatment with IPAG could also induce autophagy (13). Therefore, we tested whether the IPAG-mediated decrease in steady-state AR protein levels was due to autophagosomal degradation by combining IPAG with bafilomycin A1 (inhibitor of autolysosomal degradation) or treating cells in which ATG5 had been knocked down with siRNA. ATG5 is a protein product of the essential autophagy gene ATG5, which is required for autophagosome formation. Inhibition of autophagosomal degradation or autophagosome formation did not block IPAG-induced degradation of AR (Fig. 3D and Supplementary Fig. S3, respectively).

In parallel, we examined whether the ubiquitin proteasome system (UPS) is the operative pathway in Sigma1 inhibitor-induced AR degradation. Cotreatment of LNCaP cells with the proteasome inhibitor MG132 blocked IPAG-mediated degradation of AR (Fig. 3D), providing compelling evidence in support of the hypothesis that Sigma1 inhibitor–mediated degradation of AR is UPS-mediated.

By 3 hours of treatment with IPAG steady-state AR protein levels did not yet change significantly; however, DHT-induced nuclear localization was blocked (Fig. 2E and F). Subsequently, we asked whether, during this time frame, treatment with IPAG resulted in sequestration and targeting of cytoplasmic AR to the ubiquitin proteasome system (UPS). We used an isopycnic centrifugation technique to isolate subcellular fractions of LNCaP cells. In this assay, IPAG induces cofractionation of AR and Sigma1 along with UPS components including PSMB5, polyubiquitinated proteins, and p97/VCP (Fig. 3E). In response to treatment with IPAG, 90% of Sigma1, 98% of PSMB5, 84% of p97/VCP, 66% of polyubiquitin immunoreactivity, and 46% of AR cofractionated in fractions 4–6, from initially disparately distributed fractions (Fig. 3E). These results further support the idea that the Sigma1 inhibitor induces accumulation of AR to ubiquitin-enriched cell compartments to be targeted for proteasomal degradation. It is noteworthy that a portion of the total cellular AR within these cells shifts to Sigma1-enriched fractions, suggesting that a subpopulation of AR interacts with Sigma1 at a given time and emphasizing that IPAG-induced AR degradation is time dependent.

Selectivity of Sigma1 inhibitor induced protein elimination

Treatment with the Sigma1 inhibitor appears to be selective. Indeed, in response to IPAG treatment the levels of several proteins decrease; however, the levels of some proteins remain steady and a few increase (Fig. 4A). Because of its features, Sigma1 regulation of protein homeostasis draws comparison with HSP90. Here, we demonstrate that pharmacologic modulation of Sigma1 and HSP90 involve distinct mechanisms.

Sigma1 modulators and HSP90 inhibitors induce the degradation of some overlapping but also distinct sets of client proteins. HSP90 chaperones a range of client proteins important in cancer cell growth and survival. We tested the effects of IPAG treatment on steady-state levels of several well-defined HSP90 client proteins that play important roles in this disease: ErbB2/HER2, ErbB3/HER3, and Akt. We found that the Sigma1 inhibitor affects some but not all of these client proteins. Treatment with IPAG elicited a similar dose-related decrease in steady-state ErbB2/HER2 and ErbB3/Her3 as the geldanamycin analogue 17-AAG (Fig. 4B and C). However, in contrast to elimination of Akt produced by 17-AAG, we observed no significant change in Akt protein levels in response to IPAG treatment (Fig. 4B and C).

HSP90 is the best-characterized chaperone of glucocorticoid receptor (GR) as well as AR. In VCaP cells, treatment with 17-AAG resulted in dose-related elimination of GR (Fig. 4B). Interestingly, we observed no measurable change in GR protein levels in response to IPAG treatment (Fig. 4C). Considering the conservation and overlap in AR and GR mechanisms and support machinery, this result is somewhat unexpected. However, it further highlights the specificity and some key differences in mechanisms governing the actions of HSP90 and Sigma1.

Compensatory upregulation of HSP70 and HSP27 are hallmarks of HSP90 inhibition (6, 25). We confirmed that treatment with 17-AAG results in salient induction of both HSP70 and HSP27 (Fig. 4D). In clear contrast, treatment with IPAG did not induce HSP70 and HSP27 (Fig. 4E).

Together, these data demonstrate a degree of selectivity of Sigma1 inhibitor–mediated actions. They demonstrate that Sigma1 and HSP90 have some overlapping but also distinct sets of client proteins, and they suggest that pharmacologic modulation of Sigma1 and HSP90 are governed by distinct mechanisms of action.

Sigma1 inhibitor eliminates AR and AR splice variants

Nascent cytoplasmic AR is bound and stabilized by multichaperone protein complexes that include HSP90. DHT binding to AR stimulates transport of this complex to the nucleus, and HSP90 is thought to play a role in the transport as well as stabilization of AR (6, 26). However, HSP90 does not bind ARVs (14, 20, 21, 27, 28), and HSP90 inhibitors such as 17-AAG are unable to suppress ARV-driven transcriptional activity (14, 29).

Consistent with previously published reports, treatment with 17-AAG resulted in decreased full-length AR levels compared with DMSO, whereas ARV levels did not significantly change compared with vehicle-treated control levels (Fig. 5A). In contrast, treatment with the Sigma1 inhibitor, IPAG, resulted in a significant decrease of both full-length AR and ARV levels (Fig. 5B).

Among the multiple ARVs that have been identified in CRPC tumors and cell lines (2, 20, 21, 27, 28) the best characterized are the ARV7 and AR^v567es (2, 28). As a result of epitopes produced during splicing, these ARV contain unique amino acids that have enabled the generation of antibodies specific to these variants (2, 20, 21, 27, 28). The 22Rv1 cell line expresses ARV7 (14). Using a validated antibody specifically raised against ARV7 (30, 31), we demonstrate that IPAG suppresses ARV7 compared with vehicle-treated control levels in 22Rv1 cells, whereas 17-AAG does not (Fig. 5C and D).

We asked whether proteasomal degradation contributes to the elimination of ARVs in response to treatment with IPAG. Cotreatment of 22Rv1 cells with IPAG and MG132 blocked the degradation of both full-length AR and ARVs (Fig. 5E), which we confirmed specifically for ARV7 as well (Fig. 5F).

Sigma1 inhibitor suppresses ARV transcriptional activity

We confirmed that IPAG-mediated elimination of AR and ARV corresponds with the suppression of AR and ARV transcriptional activity. When cultured in CSS medium, ARV transcriptional activity continues as evidenced by detectable KLK3/PSA and TMPRSS2 transcripts under this hormone-depleted culture condition. In this experiment, we set our baseline relative KLK3/PSA and TMPRSS2 transcript value at 1.0 (Fig. 5G). Treatment with DHT induces full-length AR transcriptional activity, increasing KLK3/PSAand TMPRSS2 transcript levels above DHT-naïve cells (Fig. 5G). When treated with IPAG, KLK3/PSAand TMPRSS2 transcript levels decreased (Fig. 5G). This is consistent with the decreased ARV protein levels in response to treatment with IPAG (Figs. 3 and 5). Cotreatment of DHT and with IPAG maintained KLK3/PSA and TMPRSS2 transcript levels below baseline (Fig. 5G).

In contrast to IPAG, treatment of 22Rv1 cells with 17-AAG did not decrease AR target gene transcript levels below the baseline value, with KLK3/PSA and TMPRSS2 transcript levels (Fig. 5G). Therefore, constitutive ARV transcriptional activity in 22Rv1 cells is not altered by 17-AAG. These data support previous reports demonstrating that the HSP90 inhibitor, geldanamycin, does not suppress ARV transcriptional activity (14, 29). Cotreatment with DHT and 17-AAG resulted in increased KLK3/PSA and TMPRSS2 transcript levels (Fig. 5G).

These data demonstrate that pharmacologic inhibition of Sigma1 suppresses ARV as well as AR transcriptional activity. However, consistent with recent reports, HSP90 inhibitors suppress AR but not ARV transcriptional activity (Fig. 5; refs. 14, 29).

Finally, we asked whether ARV transcriptional activity would be altered by Sigma1 RNAi. Consistent with the small-molecule inhibition of Sigma1, siRNA knockdown of Sigma1 in 22Rv1 cells grown in CSS medium (Fig. 5H) suppressed KLK3/PSA and TMPRSS2 transcript levels (Fig. 5I).

Sigma1 physically associates with AR and AR splice variants

The inability of HSP90 inhibitors to degrade ARV may be attributed to the lack of physical association between the two proteins (14, 29). This is consistent with HSP90 binding to AR CTD at amino acid residues, which are absent in the ARVs (14, 20, 21, 27, 28). However, as the Sigma1 inhibitor decreased the levels of both AR and ARV, we asked whether Sigma1 physically associates with ARV7 and AR^v567es. To address this question, we performed coimmunoprecipitation assays using FLAG epitope–tagged AR, ARV7, and AR^v567es constructs (described elsewhere; refs. 14, 29) along with an affinity-tagged Sigma1 construct. The Sigma1-HA plasmid construct contains a carboxy-terminal hemagglutinin (HA) epitope tag (32). The high degree of amino acid homology between murine and human Sigma1 make them functionally interchangeable (32). We transiently transfected both constructs into PC3 cells (AR-negative prostate adenocarcinoma cell line) and immunoprecipitated postnuclear cell lysates with an anti-FLAG antibody. Whereas HSP90 binds full-length AR, but not ARV7 or AR^v567es (14), we demonstrate here that Sigma1 physically associates with both AR variants (Fig. 6B) as well as full-length AR (Fig. 6A). Together, our data suggest that Sigma1 may play a role in the support machinery of the AR axis and contribute to the regulation of ARV as well as full-length AR. Our data also suggest that Sigma1 may physically associate with the N-terminal domain of AR; however, whether this association is direct or indirect remains to be determined.

In vivo efficacy of CT-189, a novel Sigma1 modulator with drug-like properties

Although it is a useful chemical probe, IPAG has insufficient drug-like properties, and we were unable to use this compound in our in vivo efficacy studies. Therefore, we focused on modifications of IPAG as starting points to develop more drug-like probes to study the function of Sigma1 in vivo. We synthesized 1-(4-chlorophenyl)-3-(3-(4-fluorophenoxy)propyl)guanidine, which we named CT-189 (Fig. 7A and Supplementary Table). CT-189 binds Sigma1 with a K_i = 38 nmol/L using a [³H](+)-pentazocine binding competition assay described elsewhere (33), maintains AR and ARV eliminating properties (Fig. 7B and C), and AR and ARV transcriptional activity suppressing activities (Fig. 7D–F).

CT-189 has drug-like properties. It is well tolerated by mice at efficacious doses. Importantly, mice treated with CT-189 did not lose weight (Fig. 7H), and we observed no behavioral abnormalities compared with vehicle-treated mice.

In vivo, CT-189 inhibited the growth of xenografted 22Rv1 tumors by 60% following approximately 36 days of treatment (Fig. 7G). A dose of 30 mg/kg CT-189 was administered by oral gavage every other day during this period. We performed biochemical analysis of the harvested tumors by immunoblot to confirm that AR and ARV7 elimination was associated with tumor growth inhibition. Compared with treatment with drug vehicle, tumors from CT-189–treated mice had decreased levels of ARV7 (79% ± 7% decrease) as well as AR (67% ± 8% decrease) protein levels (Fig. 7I and J, respectively).

Discussion

Sigma1 and the AR axis

The dual goals of this study were to better understand the role of Sigma1 with regard to AR processing and function and to determine whether modulation of its activity may have therapeutic potential for the treatment of CRPC. The concept of the role of Sigma1 is rapidly evolving, and emerging evidence suggests that either Sigma1 may function as an allosteric cofactor protein associated with bona fide receptor systems or as a novel chaperone (34) or scaffolding protein. We demonstrate that treatment with a prototypic Sigma1 modulator, IPAG, results in cytoplasmic sequestration of AR and proteasome-mediated degradation. Consistent with these data, RNAi knockdown of Sigma1 results in decreased AR levels and transcriptional activity. Furthermore, we demonstrate that a drug-like analogue of IPAG, which we named CT-189, maintains Sigma1-mediated AR/ARV–eliminating activities, suppresses AR transcriptional activity, and importantly, significantly inhibits ARV driven tumor growth in a 22Rv1 xenograft model. To our knowledge, this work represents the first demonstration of a direct interaction between Sigma1 and the AR axis, and it is the first evidence of in vivo efficacy of Sigma1 modulator compounds through this mechanism. An indirect connection between the AR axis and Sigma1 can be inferred from parallel lines of research showing that the endogenous steroid hormone dehydroepiandrosterone (DHEA) has affinity for both AR and Sigma1 (34–37). DHEA has been shown to upregulate AR levels and activity in the central nervous system (38) and it has been proposed to act as a Sigma1 agonist or activator (34). In the context of prostate cancer, circulating adrenal androgens, DHEA and DHEA sulfate (DHEA-S), are taken up by prostate tumors and converted into testosterone and other steroids that bind and activate AR (39). Elevated intratumoral androgens are characteristic of castration-resistant tumors (39, 40). DHEA and DHEA-S contribute to the persistent synthesis of intratumoral androgens following androgen deprivation therapy, which drives CRPC (39, 40). It is tempting to speculate that Sigma1 interaction with DHEA may also play a role in prostate tumor growth. Our data, along with these indirect lines of evidence, suggest a role for Sigma1 in feedback mechanisms that regulate AR-associated networks.

Sigma1 as a novel ligand-operated scaffolding protein that engages in selective protein associations

Sigma1 is expressed throughout the cytoplasm of prostate cancer cells and has been described as a primarily endoplasmic reticulum (ER) protein. Based on the recently resolved crystal structure (8), Sigma1 is an ER bound integral membrane protein with a relatively short ER luminal component and a relatively long cytoplasmic tail. The ER extends throughout the cytoplasmic space of the cell and associates with essentially all organelles (41). Therefore, it is not surprising that Sigma1 can physically associate with cytoplasmic and cytosolic proteins as well as ER luminal proteins. Here we have found that Sigma1 physically and functionally interacts with AR.

HSP90 is the most extensively characterized AR chaperone. Along with co-chaperones, HSP90 binds and stabilizes nascent AR in the cytoplasm, and upon DHT binding to AR, AR is thought to dissociate from this chaperone complex and subsequently undergo transport to the nucleus (reviewed in reference 6). In our pull-down experiments we show that AR and Sigma1 physically associate (Fig. 6). Together, our experiments demonstrate that although Sigma1 and HSP90 may share some common associated proteins and may engage in multi-protein complexes, they do not necessarily interact directly and they have distinct mechanisms of action. Three lines of evidence exemplify that pharmacological modulation of Sigma1 and HSP90 are driven by distinct mechanisms of action.

First, Sigma1 modulators such as IPAG and CT-189 eliminate both AR and ARV in 22Rv1 cells whereas the HSP90 inhibitor 17-AAG eliminates only the full-length AR (Fig. 5, 7, and (14)). Among the multiple ARVs that have been identified in CRPC tumors and cell lines (2, 20, 21, 27, 28) the best characterized are the ARV7 and AR^v567es splice variants (2, 28). Recent reports demonstrate that HSP90 inhibitors suppress AR but not ARV transcriptional activity (14, 29). Whereas HSP90 binds full-length AR but not ARV7 or AR^v567es (14), Sigma1 physically associates with both AR variants as well as full-length AR (Fig. 6). Furthermore, full-length AR and ARV are eliminated by IPAG but not 17-AAG (14, 29). This is consistent with HSP90 binding to AR CTD at amino acid residues that are absent in the AR variants (14, 20, 21, 27, 28). Our results with 22Rv1 cells suggest that Sigma1 binds to AR, either directly or indirectly, however, by a mechanism distinct from HSP90.

Second, our discovery that IPAG destabilizes and eliminates some, but not all, HSP90 client proteins provides further evidence that Sigma1 and HSP90 have distinct physiological roles and mechanisms of action. Notably, glucocorticoid receptor (GR) protein levels do not change in response to treatment with IPAG under conditions wherein AR is nearly eliminated (Fig. 4B and C). This was a surprising finding considering the degree of overlap between AR and GR mechanisms. It highlights the selectivity of Sigma1 modulator actions and further distinguishes Sigma1 and HSP90 mechanisms of action. Treatment with IPAG also eliminates ErbB2/HER2 and ErbB3/HER3, whereas the level of other HSP90 clients such as Akt is not altered (Fig. 4B and C).

The third finding providing evidence that Sigma1 and HSP90 function through distinct mechanisms is the lack of HSP70 and HSP27 induction in response to treatment with Sigma1 modulators (Fig. 4E). Inhibition of HSP90 with small molecule inhibitors results in heat shock factor 1 (HSF1)-dependent transcriptional up-regulation of HSP70 and HSP27 (6, 25). This cytoprotective, compensatory induction of HSP70 and HSP27 is a pharmacodynamic marker of HSP90 inhibition. This response is absent in Sigma1 modulation and thus further distinguishes HSP90 and Sigma1 mediated effects (Fig. 4D and E).

Our data suggest that Sigma1 engages in a number of multi-protein complexes, some of which include AR. Whether Sigma1 directly or indirectly binds AR remains to be determined. Furthermore, it remains to be determined whether Sigma1 modulators directly alter protein-protein associations or the intracellular transport and localization of Sigma1-associated complexes. Evidence of physical association with AR along with convergence of Sigma1, UPS components, and a subpopulation of AR to the same subcellular fractions suggest that both contribute to the pharmacologic mechanism of Sigma1 modulator actions.

Implications for discovery of novel therapeutic agents and strategies to treat advanced prostate cancer

Resistance to first- and second-generation AR–targeting agents is invariably associated with reactivation of the AR axis, either via induction of intratumoral steroidogenesis or increased expression of AR and truncated AR splice variants (2). This is further complicated by compensatory upregulation or feedback regulation of associated pathways, including ErbB receptor and PI3K activation in PTEN-deficient prostate cancers (42). These examples demonstrate the importance of discovering and developing novel approaches to cotargeting the AR axis and the networks on which it depends (2).

It is noteworthy that Sigma1 inhibition suppresses growth of PTEN-mutant LNCaP and C4-2 and PTEN-null PC3 cells (Fig. 1 and (12). These data suggest that Sigma1 inhibitors engage mechanisms downstream of PTEN or mechanisms that cooperate with but are distinct from canonical PI3K/Akt and ERK growth and survival signaling pathways. PTEN deficiency, by mutation or loss of PTEN, has a significant impact on prostate cancer progression. Indeed, over 50% of advanced prostate cancers are PTEN deficient (42–44). Suppression of growth signaling in PTEN-deficient prostate cancers (12, 13) as well as suppression of pathways that compensate for AR-targeted inhibition demonstrate that Sigma1 modulators may be used more broadly than just to target the AR axis.

Although Sigma1 modulators alter multiple processes and systems in cancer cells, our data provide evidence that Sigma1 is a specific but multifunctional target. Our comparison of IPAG and 17-AAG provides evidence in support of this notion. In addition, in LNCaP and C4-2 cells, the steady-state protein levels of other nuclear hormone receptors including retinoic acid receptors and retinoid X receptors were not affected by treatment with IPAG (Supplementary Fig. S4). We are just beginning to learn how selective Sigma1 modulator compounds can be used to regulate specific but multiple cellular processes. Here we demonstrate that Sigma1-mediated protein homeostasis can be pharmacologically regulated to inhibit AR transport and stability. The three CRPC lines (C4-2, VCaP, and 22Rv1) evaluated here were all responsive to small-molecule Sigma1 inhibition. The Sigma1 inhibitor also reduced ErbB2/HER2 and ErbB3/HER3 protein levels (Fig. 4C), thus abrogating compensatory upregulation of ErbB2/HER2 and ErbB3/HER3 in response to AR axis–targeted treatments (2, 42, 45). The ability to pharmacologically modulate multifunctional targets such as Sigma1 is advantageous in cancer as it imposes a barrier to compensatory response mechanisms to targeted therapies without the broad and often toxic effects of chemotherapy.

Here, we demonstrate in vivo efficacy with a drug-like Sigma1 compound, CT-189, in an ARV-driven CRPC tumor xenograft model (Fig. 7 and Supplementary Table). Importantly, this Sigma1 inhibitor produced no detectable side-effects at efficacious doses—no weight loss and no behavioral abnormalities were observed under these study conditions. Altogether, our data provide evidence in support of the potential value of this therapeutic approach and Sigma1 as a drug target in the treatment of AR-driven cancers.