Second-Generation HSP90 Inhibitor Onalespib Blocks mRNA Splicing of Androgen Receptor Variant 7 in Prostate Cancer Cells

Prostate cancer is the most common cancer in men and a leading cause of male cancer mortality in Western societies (1). Androgen deprivation remains the mainstay of treatment for patients with advanced stage disease. However, the response to suppression of gonadal androgens is not durable, and the transition of metastatic disease to a lethal castration-resistant prostate cancer (CRPC) is inevitable. It is now widely recognized that CRPC progression results from many different alterations of the androgen/androgen receptor (AR) axis that promote continued AR transcriptional activity. Novel androgen biosynthesis inhibitors (e.g., abiraterone acetate) and AR antagonists (e.g., enzalutamide) overcome several of the mechanisms underlying persistent AR transcriptional activity and have been shown to extend overall survival of patients with CRPC (2, 3).

AR belongs to the nuclear receptor superfamily of transcription factors, whose members are ligand-responsive transcription factors. The AR gene is located on Xq11-12; it contains eight exons that encode a protein of 919 amino acids (AR-FL). It is composed of four functional domains: the amino-terminal transcription activation domain (NTD), which possesses a ligand-independent transcriptional function (AF-1); the DNA-binding domain (DBD), which binds to specific DNA sequences named androgen-responsive elements (ARE); the hinge region, which includes a part of the nuclear localization signal; and the carboxy-terminal ligand-binding domain (LBD), which has a ligand-dependent transcriptional activity (AF-2). Multiple alternatively spliced AR variants (AR-V) have been identified and translated from alternatively spliced AR mRNAs (4). These variants typically contain exons 1–3, but lack portions of the carboxy-terminal LBD, and may be constitutively active (5, 6). AR-Vs have been shown to drive androgen-independent cell proliferation in a manner that is resistant to antiandrogens, including enzalutamide (7). The expression of AR-Vs is higher in CRPC compared with primary hormone–naïve prostate cancer and is reported to be mostly lacking in normal prostate epithelium (5, 6, 8–11). Constitutively active AR-Vs have been implicated as a mechanism of resistance to current hormonal therapies targeting the LBD (7, 12–15).

One of the best characterized variants, and the most frequently detected in CRPC to date (11), is the AR-V7 (also termed AR3), which contains exons 1–3 and 16 unique amino acids that arise from cryptic exon 3 (6). AR-V7 localizes constitutively to the nucleus and facilitates AR-FL nuclear localization in the absence of androgen (14). AR-V7 upregulation mitigates the ability of the AR antagonist enzalutamide to inhibit AR-FL nuclear trafficking (14). AR-V7 may also direct a distinct transcriptional program to AR-FL, including the transcription of several cell cycle–related genes in addition to canonical androgen-responsive genes (11, 16). Recent clinical studies have linked AR-V7 expression in circulating tumor cells (CTC) and CRPC tissue with resistance to enzalutamide and abiraterone (17, 18). This highlights an urgent need for alternative treatment strategies targeting AR-Vs.

HSP90 is a molecular chaperone essential for the late stage maturation, stability, and function of a large number of client proteins involved in key signal transduction pathways in prostate cancer, including AR, AKT, and the glucocorticoid receptor [GR (NR3C1) approved name; refs. 19–21]. A key requirement for the transcriptional activity of steroid receptors is the interaction of unbound receptor with the HSP90 chaperone machinery. This interaction occurs exclusively through the LBD via direct binding to HSP40 and HSP70 (22, 23). In prostate cancer cells, inhibition of HSP90 chaperone activity impairs the nuclear translocation of AR-FL and enhances the degradation of AR-FL protein, resulting in decreased AR-FL transcriptional output (20, 24, 25) and prostate cancer cell proliferation (26, 27). Conversely, recent studies have suggested that AR-V7 nuclear translocation and transcriptional activity does not require HSP90 activity (28, 29), although HSP90 has been proposed as a therapeutic target in CRPC.

Several HSP90 inhibitors are undergoing clinical trial (29, 30). In this study, we investigated the effect of HSP90 inhibition in prostate cancer cells expressing AR-V7. We demonstrate that HSP90 inhibition reduces expression of AR-V7. Interestingly, we show that AR-V7 downregulation is not due to a traditional client–chaperone mechanism but results from disruption of AR splicing following HSP90 inhibition.

22Rv1 (#CRL-2505), VCaP (#CRL-2876), and PC3 (#CRL-1345) cells were obtained from the ATCC and grown in their recommended culture medium, containing 10% FBS at 37°C in an atmosphere of 5% CO₂. LNCaP95, an androgen-independent cell line derived from long-term continuous culture of LNCaP cells, were cultured in red-phenol free RPMI1640 supplemented with 10% charcoal-stripped FBS (CSS; Invitrogen). Tanespimycin (17-AAG; LC Laboratories), alvespimycin (17-DMAG; LC laboratories), R1881 (Steraloids), geldanamycin (Stratech), MG132 (Sigma-Aldrich), actinomycin D (Sigma), and cycloheximide (Sigma) were obtained from commercial sources. Lactate salt of onalespib (AT13387) was synthesized by Astex Pharmaceuticals according to Woodhead and colleagues. (31).

Cell growth inhibition was assessed using the sulforhodamine B assay (32). The GI₅₀ was calculated as the drug concentration that inhibits cell growth by 50% compared with control growth using GraphPad Prism software.

Cell cycle distribution and sub-G₁ populations were analyzed by flow cytometry using propidium iodide staining (33). Data were analyzed with the FlowJo Software.

Whole cells were lysed in RIPA buffer (Pierce) supplemented with protease inhibitor cocktail (cOmplete, Roche). The nuclear and cytosolic subcellular fractions were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Pierce) according to the manufacturer's instructions. To obtain detergent-soluble and detergent-insoluble fractions, cells were lysed in TNESV buffer (1% NP40 detergent, 0.5% deoxycholate, 50 mmol/L Tris-Cl (pH 7.4), 150 mmol/L NaCl, 0.5 mmol/L EDTA) for 20 minutes on ice. The NP40 detergent–insoluble pellet fraction was resuspended in RIPA lysis buffer with protease inhibitors by brief sonication. Protein extracts (20 μg) were separated on 4% to 12% NuPAGE Bis-Tris gel (Invitrogen) by electrophoresis and subsequently electrotransferred onto Immobilon-PPVDF Membranes of 0.45-μm pore size (Millipore). Details of primary antibodies used are provided in Supplementary Table S1. Chemiluminescence was detected on Amersham Hyperfilm ECL (GE Healthcare Life Sciences). In some experiments, the FluorChem Q Imager and Software was used for annotation and analysis of images, including densitometry.

Total RNA was extracted from cells using the RNeasy Plus Mini Kit (Qiagen) and from formalin-fixed and paraffin-embedded (FFPE) tissue using the RNeasy FFPE Kit (Qiagen) as per manufacturer's instructions. RNA (5 μL) was reverse transcribed into cDNA using the First Strand cDNA Synthesis Kit (Roche). qRT-PCR was carried out on ViiA 7 System Real-Time PCR System (Life Technologies) using the TaqMan Universal PCR Master Mix (Applied Biosystems). The assays used in this study are listed in Supplementary Table S2. Fold change in mRNA expression levels was calculated by the comparative C_t method, using the formula 2^−ΔΔCt and human RPLPO (large ribosomal protein) as calibrator. The vehicle-treated sample was set to 1.

Whole-cell lysates were obtained using Pierce IP Lysis Buffer (Thermo Scientific). Protein extracts (approximately 1 mg of total protein) were precleaned by protein A/G PLUS-Agarose (Millipore), and extracts were incubated with the respective antibodies (1 μg) overnight at 4°C. Immune complexes were collected by protein A/G PLUS-Agarose beads, analyzed by immunoblotting, probed with the protein A/G HRP-conjugated secondary antibody (Thermo Scientific), and visualized using an ECL chemiluminescence (Thermo Scientific).

PC3 cells (3 × 10⁵) were plated on 6-well plates and reverse transfected using the X-tremeGENE HP DNA Transfection Reagent (3 μl/well; Roche; ref. 34). The plasmid constructs (1 μg DNA/well) used were pcDNA3.1-AR-V7 (encoding the V7-truncated AR isoform, a gift from Dr. Jun Luo, Johns Hopkins University, Baltimore, MD) and F527-AR (encoding the full-length AR; ref. 34).

Subcutaneous 22Rv1 solid tumor xenografts were prepared by serial passaging in male NMRI nu/nu mice (Harlan). Tumors were measured twice weekly with microcalipers and tumor volumes determined using the formula: (length × breadth² × 0.5). Drug treatment was started when tumors were well established (mean tumor volume, 120 mm³). Onalespib was dissolved in 17.5% (w/v) hydroxypropyl-β-cyclodextrin and administered intraperitoneally at 70 mg/kg twice a week for 2 weeks or 80 mg/kg once weekly (35). For pharmacodynamic analyses, tumor-bearing animals were sacrificed at indicated time points after a single dose of onalespib at 80 mg/kg (typically 4/time point).

IHC for AR N-terminus domain (AR-NTD), HSP72, Ki67, and cleaved caspase-3 was performed on 4-μm FFPE tumor sections. Details of the IHC assays are provided in Supplementary Table S3. Slides were counterstained with hematoxylin (TCS Biosciences Ltd), dehydrated, and mounted with DPX mountant (Sigma). All sections were scored by a pathologist blinded to treatment information.

Total cellular RNA was extracted and purified using the RNeasy Mini Kit (Qiagen). RNA integrity was determined by Bioanalyzer 2100 (Agilent Technologies), and only samples with RNA integrity numbers ≥9.5 were used. RNA (500 ng) was prepared into a library using the TruSeq RNA Sample Prep Kit v2 (Illumina) and then clustered and sequenced on a Hi-Seq 2000 (Illumina) using TruSeq v3. An average of around 40 million paired end reads were sequenced per sample. The experiment was repeated three times. The paired end raw reads in FASTQ format were aligned to the reference human genome (hg19) using TopHat v2.0.7. The library and mapping quality were estimated using Picard tools (http://broadinstitute.github.io/picard). The alternative splicing events (SE, RI, MXE, A5SS, and A3SS), based on Ensembl v61 annotation, were accessed using MATS v3.0.8 (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3333886/).

Effects of different treatments were compared using the Student t test, Mann–Whitney test, or ANOVA. Survival analysis was performed by Kaplan–Meier methods and log-rank analysis using quadruplication of the tumor as surrogate endpoint. In the RNA sequencing (RNA-Seq) analysis, the supportive reads of each alternative splicing event were counted, with the inclusion level being estimated using a Bayesian model as described by Shen and colleagues, 2012 (36). Pairwise comparisons of each splicing event were performed among treatments using the Markov chain Monte Carlo method coupled with a simulation-based adaptive sampling procedure to calculate the P value (36). Sample replicates were used to control variability within-treatment groups. The FDR was obtained by the Benjamini–Hochberg method with an FDR < 0.05 being employed for the differential alternative splicing significance assessment.

We initially examined the effects of the first-generation HSP90 inhibitors tanespimycin (17-AAG) and alvespimycin (17-DMAG) and the second-generation HSP90 inhibitor onalespib (AT13387) on cell proliferation in human prostate cancer cell lines expressing the AR-V7, including 22Rv1, VCaP, and LNCaP95. Previous studies had established the expression status for both AR-FL and AR-Vs in these cell lines (5, 6, 28). HSP90 inhibition reduced cell proliferation in a dose-dependent manner (data not shown). Onalespib and alvespimycin potently inhibited the growth of the androgen-dependent cell line VCaP and also the growth of the androgen-independent 22Rv1 and LNCaP95 cell lines, which are resistant to the AR antagonist enzalutamide (Table 1; refs. 7, 11). Our results confirmed previous observations that AR-V7 expression does not confer resistance to HSP90 inhibitors in prostate cancer cells (29).

We then investigated the effect of HSP90 inhibition on AR-FL and AR-V7 protein levels in prostate cancer cells expressing both the AR-FL and AR-V7. Onalespib reduced AR-FL protein irrespective of AR-FL status (i.e., wild type or mutant) in a concentration- and time-dependent fashion in VCaP and 22Rv1 cell lines (Fig. 1). Notably, onalespib caused similar depletion of AR-V7 protein in the 3 cell lines tested (Fig. 1). AR-FL and AR-V7 depletion was accompanied by a concentration-dependent (data not shown) and time-dependent depletion of various HSP90 clients, including AKT, ERBB2, CRAF (RAF1), and GR (Supplementary Fig. S1). HSP72 (HSPA1) and HSP27 (HSPB1) were induced by HSP90 inhibition, associated with the heat shock factor1 (HSF1)–mediated stress response, confirming HSP90 target engagement (37). Similar results were observed with first-generation HSP90 inhibitors (geldanamycin, tanespimycin, and alvespimycin); however, tanespimycin resulted only in modest depletion of AR-FL and AR-V7 at the concentrations used in VCaP and 22Rv1 cells (Supplementary Figs. S2 and S3). As DT-diaphorase/NQO1 plays a role in primary and acquired resistance to tanespimycin (32, 38), we evaluated constitutive protein expression levels in prostate cancer cell lines to clarify whether lower sensitivity to tanespimycin was due to different expression of the enzyme. Low levels of DT-diaphorase/NQO1 protein were detected in 22Rv1 and VCaP cells, respectively, by immunoblotting, possibly explaining the differences in effect seen with tanespimycin compared with alvespimycin and onalespib (data not shown). We should note that all the HSP90 inhibitors we had tested inhibit both α and β HSP90 isoforms. We confirmed depletion of AR-V7 protein when the expression of both the HSP90 isoforms was reduced simultaneously by siRNA knockdown (Supplementary Fig. S4).

Importantly, in time-course experiments, the kinetics of AR-V7 protein loss in response to HSP90 inhibition differed from that for AR-FL (Fig. 1C and D). Onalespib induced a reduction of AR-FL protein within 4 hours. In VCaP cells, AR-FL was reduced in a time-dependent fashion, with maximal depletion reached at 48 to 72 hours, but in LNCaP95 AR-FL, protein level remained constant. Depletion of AR-V7 was first noted after 12 to 24 hours, reaching maximal depletion at 48 to 72 hours when AR-V7 was hardly detectable. The timing and degree of AR-V7 depletion was similar in VCaP and LNCaP95. These results confirmed AR-FL depletion following HSP90 inhibition and also indicated for the first time that AR-V7 protein level is also decreased when HSP90 is inhibited.

HSP90 inhibition results in protein client destabilization and degradation by the ubiquitin–proteasome pathway (39). Ubiquitinated proteins first accumulate in the cytosol and become aggregated and relocated in the detergent-insoluble fraction of cells when proteasomes are inhibited. To evaluate whether AR-V7 protein depletion was due to increased proteasomal degradation after HSP90 blockade, we examined the impact of the proteasome inhibitor MG132 on AR-V7 depletion by onalespib. VCaP cells were treated with onalespib in the presence or absence of proteasome inhibitor MG132 for 24 hours. As expected, addition of MG132 to onalespib treatment induced accumulation of high-molecular weight ubiquitinated AR-FL forms that were redistributed to the detergent-insoluble fraction (Fig. 2A). Such accumulation of high-molecular weight AR-V7 forms was not observed, suggesting the lack of involvement of the proteasomal pathway in AR-V7 depletion by onalespib (Fig. 2A).

To determine whether the observed decrease in AR-V7 protein in onalespib-treated cells was the result of altered protein stability, we treated VCaP with the protein synthesis inhibitor cycloheximide in the presence or absence of onalespib (Fig. 2B and C). When protein synthesis was inhibited by cycloheximide, AR-V7 stability was greater than AR-FL (Fig. 2B), suggesting that AR-V7 has a longer half-life than AR-FL. Onalespib treatment significantly reduced AR-FL protein levels, indicating increased degradation of AR-FL (Fig. 2B and C, top). Conversely, AR-V7 protein stability was not affected as demonstrated by similar levels of AR-V7 in the presence of onalespib (Fig. 2B and C, bottom). To further confirm that the effect of onalespib was not due to protein destabilization and increased protein degradation, we also investigated the effects of onalespib in the AR-negative PC3 cell line transiently transfected with plasmid constructs carrying AR-FL and AR-V7 cDNA. Onalespib treatment in this transient transfection system resulted in the depletion of transiently expressed AR-FL, but not AR-V7 protein, confirming that the effect of HSP90 inhibition on AR-V7 was not due to destabilization or a direct effect on AR-V7 protein (Supplementary Fig. S5).

A key requirement for steroid receptor activity is the interaction of unliganded receptors with the HSP90 chaperone machinery (40). To further examine the relationship between AR-V7 and HSP90, we determined by coimmunoprecipitation assays whether AR-V7 and HSP90 protein complexes exist in prostate cancer cells that endogenously express the variant (Fig. 2D). As expected, AR-FL formed a complex with HSP90 when the cells were cultured in androgen-depleted media (Fig. 2D, top). In contrast, AR-V7 did not coimmunoprecipitate detectable amounts of HSP90 (Fig. 2D, bottom). This was in agreement with results of Gills and colleagues in PC3 cells exogenously expressing the AR-V7 (29). Also, in keeping with a previous report (7), using subcellular fractionation we confirmed that AR-V7 entered the nucleus independently of HSP90 activity (data not shown). Taken together, these results indicated that AR-V7 protein, unlike AR-FL, does not require the HSP90 chaperone.

To evaluate whether AR-V7 depletion was due to the downregulation of AR mRNA expression, we next examined the levels of AR transcripts following onalespib treatment (Fig. 3A). In VCaP, 22Rv1, and LNCaP95 cells treated for 24 hours with onalespib, AR-FL mRNA levels were not significantly altered (data not shown). AR-V7 mRNA levels, however, were significantly decreased by onalespib in a concentration-dependent fashion (Fig. 3A and Supplementary Fig. S6). Onalespib treatment at 0.6 μmol/L for 48 hours decreased AR-V7 mRNA level by 72 % (Fig. 3A). To distinguish between transcriptional and posttranscriptional effects on AR-V7 mRNA, we examined the effect of onalespib treatment on mRNA stability. VCaP cells were treated with onalespib or vehicle in combination with the transcriptional inhibitor actinomycin D (ACTD). Onalespib treatment did not significantly alter the AR-V7 (Fig. 3B) and AR-FL (data not shown) mRNA levels in the presence of ACTD, suggesting that the decrease in AR-V7 mRNA in onalespib-treated cells was not due to reduced stability of the mRNA. These results, coupled with the lack of depletion of AR-FL mRNA, suggested that onalespib reduced the generation of AR-V7 mRNA. To test this, we evaluated levels of unspliced AR pre-mRNA in 22Rv1 cells using three different sets of primers/probes (AR-5, AR-6, and AR-7; Supplementary Table S2) directed against different regions of the unspliced AR cDNA (intron 8–exon 9, intron 3, and exon 2–intron 2). Onalespib did not reduce levels of unspliced AR and AR-FL but only AR-V7 mRNA levels (Fig. 3C), suggesting that HSP90 inhibition specifically altered pre-mRNA splicing and impaired the generation of AR-V7 mRNA.

HSP90 inhibition elicits a dual effect in cells: the depletion of client proteins and the activation of a heat shock response (HSR) mediated by HSF1 activation (41). It has been reported that the splicing of mRNA precursors is broadly repressed on heat shock (42–44). To evaluate global changes in splicing pattern associated with HSP90 inhibition, we treated VCaP cells with onalespib (0.3 and 0.6 μmol/L) for 24 hours and performed RNA-Seq analysis. The two different concentrations of onalespib (0.3 and 0.6 μmol/L) resulted in largely overlapping splicing changes (Supplementary Table S4; Fig. 4A). Onalespib treatment introduced 743 alternative splicing events involving 557 genes (FDR < 0.05; Supplementary Table S4). In keeping with data following heat shock, we observed a significant increase in the retention of introns in transcripts and decreased exon skipping (Fig. 4C; refs. 42, 43). Furthermore, RNA-Seq analyses confirmed a reduction of AR-V7 cryptic exon expression (Fig. 4B). Pathway analyses showed that genes involved in splicing, including processing of capped intron–containing pre-mRNA(R) and spliceosome(K), as well as heatshock-related pathways, were significantly enriched (FDR < 0.05) in the alternative splicing events induced by HSP90 inhibition (Supplementary Table S5).

We next evaluated the effect of onalespib on tumor growth and AR-V7 levels in 22Rv1 subcutaneous solid tumor xenografts. Onalespib was administered intraperitoneally at 70 mg/kg in a twice a week schedule (35). Onalespib reduced tumor growth significantly when administered twice a week (P < 0.001 against the vehicle control; Fig. 5A) and significantly prolonged survival (Fig. 5B). Weekly treatment at 80 mg/kg did not show significant benefit, suggesting that more frequent dosing is required in this fast growing model (data not shown). A single administration of onalespib at 80 mg/kg caused increases in cleaved caspase-3 and wide areas of necrosis together with reduced Ki67 expression were observed 48 and 72 hours after drug administration (Fig. 5C). Reduction in total AR as assessed by an antibody directed against the AR-NTD was observed (Supplementary Fig. S8). After a single dose of onalespib AR-V7, mRNA was reduced by 20% after 6 and 24 hours (P < 0.05; Fig. 5D), whereas AR-FL levels were not significantly altered (Fig. 5E). Levels of mRNA at 48 and 72 hours after drug administration were not evaluated due to the low amount and poor quality of RNA extracted from these tumors, in which significant tumor cell killing and necrosis were observed (Fig. 5C). Onalespib treatment also significantly downregulated PSA expression in tumors 24 hours following treatment (Supplementary Fig. S7). Upregulation of HSP72 mRNA and protein by IHC confirmed HSP90 target engagement (Fig. 5F and Supplementary Figs. S8 and S9).

In this study, we examined the effects of HSP90 inhibition in prostate cancer cell lines expressing the AR-V7 variant. In vitro, the first-generation HSP90 inhibitors tanespimycin and alvespimycin and the second-generation HSP90 inhibitor onalespib induced cell growth inhibition and degradation of client proteins, including AR-FL, AKT, and GR. For the first time, we demonstrated that the truncated splice variant AR-V7 protein is depleted when HSP90 is inhibited. Our results confirmed that AR-V7 protein, unlike AR-FL, does not require the HSP90 chaperone for its function and suggested that the mechanism underlying AR-V7 depletion upon HSP90 blockade was not directly linked to interaction of AR-V7 protein with the HSP90 chaperone. We demonstrated that AR-V7 mRNA levels were downregulated when HSP90 was inhibited in a concentration- and time-dependent fashion. Previous studies have linked heat shock to widespread, but selective, repression of splicing (42, 43). We therefore hypothesized that the heat shock response induced by HSP90, through the release of HSF1, can similarly broadly impact alternative splicing. We have shown that onalespib treatment introduced 743 alternative splicing events involving 557 genes. In keeping with published data evaluating the impact of heat stress, we observed a significant increase in retention of introns in transcripts and decreased exon skipping (42, 43). Pathway analyses showed that genes critically important to splicing, including processing of capped intron–containing pre-mRNA and spliceosome genes, as well as heat shock–related pathways, were significantly enriched in the alternative splicing events induced by HSP90 inhibition.

Further studies are required to elucidate the complex and multiple mechanisms that can underlie the alternative splicing disruption observed upon HSP90 inhibition. A previous study identified the serine/arginine–rich (SR) splicing factors U2AF65 and ASF/SF2 as critical to AR-V7 splicing (45). The function of SR splicing factors is regulated by their complex interaction with kinases, including the SR protein kinases (SRPK) and CLK, as well as phosphatases that regulate their phosphorylation status (46). SRPK1 interacts with various chaperones and cochaperones, including HSP70, HSP40, and AHA1 (47). Alternative splicing is also extensively regulated by signal transduction pathways, whereby signaling cascades can link the splicing machinery to the exterior environment (48, 49). HSP90 is heavily involved in the function of key nodes in these pathways. For instance, AKT can directly phosphorylate SR proteins (50) or indirectly regulate SR protein by activating SRPK (51). In addition, repression of splicing following heat stress has been related to the SR protein SRP38 (SRSF10) dephosphorylation by the phosphatase PP1 (PPP1CA; ref. 43).

Other groups have evaluated the effect of HSP90 inhibition on AR-V7 function and activity. Shafi and colleagues used LNCaP stably transfected with AR-V7 cDNA and demonstrated no effect on AR-V7 protein levels (52). Gillis and colleagues demonstrated that AR-V7 transcriptional activity is not altered by HSP90 inhibition in PC3 cells transfected with prespliced AR-V7 driven by an ARR2-probasin luciferase (29). They also demonstrated AR-V7 does not bind HSP90 by coimmunoprecipitation using a FLAG-tagged AR-V7 transfected in PC3 cells. Our results in cell lines with endogenous AR-V7 support these observations. However, Gills and colleagues also reported that treatment of VCaP cells with the second-generation HSP90 inhibitors, luminespib (AUY922) and HSP990, increased AR-V7 protein (29), and Shafi and colleagues reported no effect of geldanamycin on AR-V7 in 22Rv1 cells (52). Their results are incongruent with what we have observed with different HSP90 inhibitors in three prostate cancer cell lines expressing the AR-V7 (VCaP, LNCaP95, and 22Rv1).

In conclusion, our observations confirm that AR-V7 protein functions independently of HSP90 and that AR-V7 is not a direct HSP90 client (29). However, we have demonstrated for the first time that HSP90 inhibition leads to the depletion of AR-V7 protein by downregulating AR-V7 mRNA splicing. AR-V7 splicing was decreased by HSP90 inhibition in prostate cell lines resistant to androgen depletion and enzalutamide, suggesting that HSP90 inhibition remains a viable approach to block the generation and possibly the upregulation of AR-V7 in CRPC. Our data demonstrate a novel mechanism of action for this class of drugs in prostate cancer cells. Onalespib and other HSP90 inhibitors are now being tested in proof-of-concept clinical trials (53, 54). These have included a phase I study of onalespib in combination with abiraterone acetate in patients no longer responding to abiraterone (ClinicalTrials.gov identifier: NCT01685268; ref. 55). The impact of HSP90 inhibitors on not only AR-FL and AR-V7, but also AKT and GR, makes HSP90 inhibitors a class of drugs that merits further evaluation against metastatic prostate cancer.

R. Ferraldeschi is a director of Translational Research at Astex Pharamceuticals. J. Lyons is Vice President at Astex Pharamceuticals. S.R. Plymate is a staff physician at GRECC/VAPSHCS and a consultant/advisory board member for Astellas and ESSA. G. Attard reports receiving commercial research grants from AstraZeneca and Janssen, has received speakers bureau honoraria from Astellas, Janssen, Roche/Ventana, Sanofi-Aventis, and Takeda, has ownership interests (including patents) in The Institute of Cancer Research (Rewards to Inventors list of abiraterone acetate), is a consultant/advisory board member for Abbott laboratories, Astellas, Bayer HealthCare Pharmaceuticals, Essa Pharmaceuticals, Janssen, Medivation, Millennium Pharma, Novartis, Roche/Ventana, and Veridex, and has provided expert testimony for Abbott Laboratories (travel and accommodation), Astellas (travel and accommodation), Bayer HealthCare Pharmaceuticals (travel and accommodation), Janssen (travel and accommodation), Medivation (travel and accommodation), and Roche/Ventana (travel and accommodation). P. Workman is a consultant/advisory board member for Astex, Novartis, Nuevolution, and Vernalis and has provided expert testimony for ICR Rewards to Inventors Scheme (could benefit from future commercialization through Vernalis and Chroma). J.S. de Bono is a consultant/advisory board member for Astex. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Ferraldeschi, J.F. Lyons, N. Wallis, P. Workman, J.S. de Bono

Development of methodology: R. Ferraldeschi, M.V. Powers, R. Riisnaes, S. Hedayat, M. Crespo, D.N. Rodrigues, S. Miranda, J.S. de Bono

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Ferraldeschi, J. Welti, M.V. Powers, T. Smyth, S. Hedayat, D.N. Rodrigues, I. Figueiredo, S. Miranda, S.R. Plymate, G. Attard, J.S. de Bono

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Ferraldeschi, W. Yuan, T. Smyth, G. Seed, D.N. Rodrigues, S. Carreira, S.R. Plymate, G. Attard, P. Workman, J.S. de Bono

Writing, review, and/or revision of the manuscript: R. Ferraldeschi, J. Welti, M.V. Powers, W. Yuan, T. Smyth, S. Carreira, S. Sharp, S.R. Plymate, G. Attard, N. Wallis, P. Workman, J.S. de Bono

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Ferraldeschi, W. Yuan, S. Hedayat, H. Wang, D.N. Rodrigues, S. Sharp, S.R. Plymate, J.S. de Bono

Study supervision: J.F. Lyons, P. Workman, J.S. de Bono

R. Ferraldeschi was funded through The Wellcome Trust PhD Programme on Mechanism-Based Drug Discovery directed by P. Workman. J. Welti and S.R. Plymate were supported by a grant from the Department of Defense (DoD) Prostate Cancer Research Program Transformative Impact Award #PC121152. M. Powers and S. Sharp were funded by a Cancer Research UK Programme Grant to P. Workman (C309/A11566) and P. Workman is a Cancer Research UK Life Fellow (C309/A8992). G. Attard is supported by a Cancer Research UK Clinician Scientist Fellowship. The authors acknowledge the support from Cancer Research UK Centre at The Institute of Cancer Research and Royal Marsden and an Experimental Cancer Medical Centre (ECMC) grant from Cancer Research UK and the Department of Health (ref: C51/A7401). The authors also acknowledge NHS funding to the NIHR Biomedical Research Centre at the Royal Marsden and The Institute of Cancer Research.