Abiraterone switches castration‐resistant prostate cancer dependency from adrenal androgens towards androgen receptor variants and glucocorticoid receptor signalling

2.1. Cell culture

VCaP and VCaP CRPC subclones (DCC‐E, BIC‐B, FLU‐D ¹⁴ ) cells were cultured in RPMI 1640 supplemented with 10% fetal calf serum (FCS) or steroid dextran‐coated charcoal stripped FCS (DCC) with or without addition of 1 µM antiantiandrogen for CRPC clones and antibiotics. H295R cells—a human adrenocortical tumour cell line which expresses CYP17A1 and subsequently produces high amounts of DHEA and androstenedione ²⁵ —were grown in DMEM/F12, supplemented with 5% FCS and antibiotics. For growth experiments, 5 × 10³ VCaP cells were plated in 96‐well culture plates (Corning) in RMPI supplemented with 10% dextran‐coated charcoal stripped FCS (DCC) and antibiotics and left to attach overnight. Cultures were incubated with steroids similar at serum levels found in 60‐year‐old men and men ondergoing ADT ²⁶ , ²⁷ with 0.1 nM DHT as a positive control based on our experiences in quadruplicate (Table S1). After 10 days, cell numbers were measured by 3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT)‐assay (Sigma) as published in, ²⁸ with each experiment performed in triplicate. For coculture experiments, 5 × 10⁴ VCaP cells were plated in 24‐well culture plates (Corning) and in parallel, 1 × 10⁴ H295R cells were plated in 24 well culture inserts with 0.4 μm pores (Corning), both in RPMI supplemented with 10% DCC and antibiotics. After separate overnight attachment, cultures were combined and incubated with indicated compounds and left to grow for 9 days, after which cell number was assessed by MTT assay for each separate compartment, with each condition in duplicate.

2.2. Animal studies

All animal studies were performed in accordance with local animal welfare legislation and with permission of the local animal experimentation ethics committee. Intact male NMRI^nu/nu mice were subcutaneously inoculated with 5 × 10⁶ VCaP cells in 100 μl matrigel (BD Biosciences) with or without concurrent injection of 5 × 10⁶ H295R cells in PBS in the contralateral flank (+ADR→ORx). Tumour volume (TV) was measured twice a week using callipers and estimated by [(height × width)^1.5 × 0.5236]. At VCaP TV >150 mm³, animals bearing only VCaP tumours were randomly assigned to sham castration (intact), castration with or without concomitant H295R inoculation (ORx + ADR and ORx, respectively) and all +ADR→ORx animals were castrated. Orchiectomy was surgically performed under isoflurane anaesthesia. When tumours progressed after castration (defined as a twofold increase in tumour volume after castration), animals were fed daily with 20% HP‐β‐cyclodextrin (Sigma), adjusted to pH 4.0 (vehicle) or vehicle with 10 mg/ml abiraterone acetate (kindly provided by Janssen, Beerse, Belgium) to reach a final dose of 175 mg/kg. At indicated time points, 50 μl blood was drawn via mandibular punction and stored at −80°C until further use. When single tumour volume reached >1500 mm³ or combined TV was >2500 mm³, animals were anaesthetised and blood was drawn via orbital bleeding, after which animals were euthanized. Blood and tumours were snap frozen and stored at −80°C until further use, with a part of each tumour stored separately in paraffin.

2.3. PSA response

Blood was collected in heparin tubes and plasma was separated by centrifugation at 1500g for 5 min. PSA was quantified by enzyme‐linked immunoassay, using the Elecsys total PSA immunoassay on a Cobas e602 analyzer (Roche Diagnostics). PSA response after castration was the calculated as the % increase or decrease vs that of PSA at time of (sham) castration, PSA response after treatment was calculated as the % increase or decrease versus that of PSA at time of treatment initiation.

2.4. Messenger RNA (mRNA) expression analysis

RNA was isolated using RNA‐Bee (TEL‐TEST, Inc.) from frozen tumour pieces. Reverse transcriptase and quantitative polymerase chain reaction (qPCR) runs were performed as described previously ¹³ using an ABI Prism 7900 Sequence Detection System under standard conditions. Complementary DNA (cDNA) (20 ng) was amplified in SYBR Green PCR Master Mix (Applied Biosystems) or TaqMan Universal Master Mix (Applied Biosystems). PCR efficiency was verified by cDNA dilution curves and exceeding 90% for all assays. Primer/probe sets used were AR (forward 5′‐CATCAAGGAACTCGATCGT‐3′, reverse 5′‐GAACTGATGCAGCTCTCTC‐3′, probe 5′‐ACATCCTGCTCAAGACGCTCCT‐3′), PSA (forward 5′‐CCCTCAGAAGGTGACCA‐3′, reverse 5′‐ACCACCTTGGTGTACAGG‐3′, probe 5′‐TATCACGTCATGGGGCAGTG‐3′) CYP17A1 (Hs01124136_m1), CYP11A1 (Hs00167984_m1), AKR1C3 (Hs00366267_m1), HSD17B6 (Hs06955149_cn). Gene expression was calculated as fold expression over housekeeping genes GAPDH (Hs99999905_m1) and PBGD (forward 5′‐CATGTCTGGTAACGGCAATG‐3′, reverse 5′‐GTACGAGGCTTTCAATGTTG‐3′) and normalised to the mean expression level from tumours growing in uncastrated placebo treated animals.

2.5. LC/MS‐MS quantification of tumour steroid concentration

Steroids were extracted from 200 µl of murine plasma or xenograft tissues homogenized in 1 ml milliQ using an Ultra‐Turrax T25 (IKA). Following the addition of deuterated steroid controls, liquid‐liquid extraction of steroids was performed using tert‐methyl‐butyl‐ether (Sigma‐Aldrich). The organic phase was removed after vortexing and evaporated under N₂ at 55°C. Samples were reconstituted in 50/50 methanol/water and analysed by liquid chromatography/tandem mass spectrometry (LC–MS/MS) employing a Waters Xevo mass spectrometer with Acquity uPLC system according to conditions previously described. ²⁹ , ³⁰ Steroids were identified through matching retention times and two mass transitions relative to authentic steroid standards. Quantification was performed relative to a calibration series (range: 0.5–500 ng/ml).

2.6. Immunohistochemistry

Formalin‐fixed paraffin embedded tumour pieces were cut, and stained with haematoxylin‐eosin. Vital tumour areas were selected for TMA construction with two cores included from each individual tumour. TMA slides were dewaxed, rehydrated, and blocked for endogenous peroxidase activity. Antigen retrieval was performed in sodium citrate buffer (pH 6.0) for 15 min at 100°C. Antibodies against AR (SP107, 1:200; Spring Bioscience) and GR (611227, 1:500; BD Transduction Laboratories) were diluted in normal antibody diluent (Immunologic) and incubated overnight at 4°C. Detection of antigens was performed using the DAKO REAL Envision Rabbit/Mouse HRP kit (DAKO) according to standard manufacturer's protocol, after which slides were counter stained with haematoxylin. Images were taken with an Olympus BX41TF microscope equipped with a Olympus Colorview IIIu camera (Olympus). AR expression was quantified by multiplying the fraction of AR positive cells with the staining intensity (0 = no staining, 1 = nuclear, 2 = nuclear + cytoplasmic), GR staining was scored as the percentage of positive cells per field at ×20 magnification.

2.7. Statistical analysis

Raw data (TV, serum PSA, PSA response) were compared with the appropriate nonparametrical tests. MTT data were analysed by a one‐way analysis of variance (ANOVA), followed by a Bonferroni corrected post hoc test to assess significant growth induction by steroids and inhibition by abiraterone and enzalutamide per steroidal condition. qPCR results levels were log₁₀ transformed to normalize for difference in variance and subsequently compared using a one‐way‐ANOVA with a Bonferroni posttest. Survival curves were analysed using the Log‐rank Test. Endpoint for progression‐free survival (PFS) after castration was a fivefold increase in tumour volume (mean + 800 mm³), censoring animals started on abiraterone before this endpoint was met. Endpoint for PFS after treatment with abiraterone or placebo was a twofold increase in tumour volume (mean + 450 mm³), censoring animals that were euthanized before this endpoint was met. Growth rates were fitted on the %growth for each individual tumour using an exponential growth curve with Y = 100 for Y = 100*e ^(k*x) where Y is the %TV, x is the days postintervention and k is the tumour dependent variable defining the steepness of the curve, correlating to growth rate. All statistical analyses and graphs were made using Graphpad Prism 9.0.

3.1. Castration‐naïve and CRPC VCaP cell lines are stimulated by physiological levels of adrenal androgens

VCaP cells were incubated with testosterone precursor steroids to test the proliferative effects of physiological steroid levels found in prostate cancer patients (Table S1). CYP17A1 substrates pregnenolone and progesterone failed to induce castration‐naïve parental VCaP proliferation (Figure 1A). In contrast, adrenal androgens DHEA and androstenedione induced growth comparable to DHT, which was abolished by direct AR antagonism with 1 µM enzalutamide, but not by 0.1 µM abiraterone. The abiraterone concentration of 0.1 µM was chosen to allow efficient CYP17A1 inhibition, while limiting the direct anti‐androgen effect of abiraterone, observed at concentrations >1 µM. ¹³ , ³¹ A similar response was observed for the VCaP CRPC clones DCC‐E, BIC‐B and FLU‐D (Figure 1A). Of note, VCaP parental cells treated with vehicle did not divide during the 10‐day incubation period, while DCC‐E, BIC‐B and FLU‐D multiplied 3‐fold. This growth was unaffected by CYP17A1 inhibition by abiraterone.

3.2. Human adrenal tissue drives castration‐resistant growth of VCaP, which can be blocked by CYP17A1 inhibition

To simulate the endocrine system present in men under ADT, human adrenal H295R cells were cocultured with parental and CRPC VCaP cells. H295R stimulated VCaP growth (Figure 1B) similar to DHT. Addition of 0.1 µM abiraterone reduced the H295R‐induced proliferation of castration‐naïve VCaP and its CRPC clone DCC‐E, while not affecting proliferation of VCaP monocultures.

To take this further, we inoculated parental hormone‐naïve VCaP cells in intact NMRI nu/nu mice whilst simultaneously inoculating H295R cells contralaterally (Figure 2A). Consistent with the in vitro data, castration of VCaP tumour‐bearing mice resulted in a significant reduction in VCaP growth rate (Figure 2B) and serum PSA (Figure 2C). Co‐inoculation of H295R cells significantly supported VCaP growth upon castration (+ADR→ORx) compared to VCaP growth in castrate animals without H295R (ORx) (Figure 2B). Despite the supportive effects of H295R on VCaP growth after castration, a clear castration response was still observed in the co‐inoculated group, demonstrated by reduced growth rates compared to sham‐castrated animals (Figure 2B) and a reduction in serum PSA 1 week postorchiectomy (Figure 2C). Moreover, castration strongly reduced prostate and seminal vesicle weights in castrate animals, with no observed differences were between castrate animals with or without H295R, further demonstrating that H295R cells did not interfere with the systemic state of castration of the host animals (Figure S1).

A PSA rise from nadir was observed in 14/18 castrate animals concomitant with tumour growth, in castrate animals both with and without H295R (Figure 2D). Subsequent Abiraterone treatment at disease progression after castration had a significant impact on VCaP growth and PSA production, but only in castrate animals concomitantly bearing H295R tumours (Figure 2E–G and Figure S2), whilst we were unable to detect a significant difference in tumour growth or PSA production in castrate animals without H295R (ORx). Consistent with these results, in a parallel group of animals co‐inoculated with H295R tumours at the time of castration, abiraterone was equally effective in reducing H295R‐stimulated VCaP growth and PSA production (Figure S3). Of note, H295R growth was unaffected by castration or abiraterone treatment (Figure S4).

3.3. Endocrine profiles in castration‐ and abiraterone‐resistant VCaP xenografts

Levels of all steroids depicted in Figure 3A were investigated in tumour homogenates and plasma by LC‐MS/MS. Testosterone was detectable in homogenates of four out of five VCaP tumours progressive under castration and coinoculated with H295R tumours. Surprisingly, testosterone could still be detected in homogenates of two out of five H295R‐unsupported progressive VCaP tumours and three out of five abiraterone‐treated sham‐castrated VCaP tumours. One out of five castrate animals, but none of the abiraterone‐treated intact animals, showed detectable testosterone plasma levels (Figure 3B). In plasma, DHEA was detectable in four out of five H295R‐bearing animals, whilst in tumours, DHEA was found in four out of five placebo‐treated H295R tumours and not in any VCaP tumour nor in any abiraterone‐treated H295R tumour. Combined, these steroid levels suggest that testosterone found in H295R‐supported VCaP tumours could be derived from DHEA secreted by H295R.

We further analysed mRNA expression for the key steroidogenic enzymes CYP11A1, CYP17A1, AKR1C3, and HSD17B6 (Figure 3A) in VCaP and H295R tumours. As expected, CYP11A1 mRNA was abundantly expressed in H295R while it was below the limit of detection for qPCR (ct > 40) in 14 out of 19 VCaP tumours (Figure 3C). CYP17A1 expression was detected in all VCaP tumors at a magnitude (on average: 1100 fold) lower than in H295R, with no alteration by castration or by abiraterone treatment. In contrast, AKR1C3 mRNA was abundantly expressed in VCaP and H295R tumours with slightly higher (fivefold) expression levels in H295R tumours. AKR1C3 expression in VCaP was higher in castrated than in noncastrated hosts, with no additional effect on expression levels by abiraterone treatment. HSD17B6 mRNA was detected in all VCaP tumours at levels above those found in H295R, with no effect of abiraterone treatment on expression levels.

We further evaluated serum levels of intermediates of the alternative androgen synthesis pathway. Although these intermediates accumulated upon blocking CYP17A1 with abiraterone (Figure 3D), this did not result in detectable intratumoural DHT levels (Figure 3B) nor did it result in measurable levels of CYP17A1‐dependent downstream metabolites of the alternative androgen synthesis pathway (data not shown).

Finally, serum analysis revealed higher corticosterone levels in castrate animals, either with or without H295R tumours, regardless of abiraterone treatment (Figure 3E), suggesting increased glucocorticoid production in mice upon castration. Moreover, in castrate mice, CYP17A1‐hydroxylase‐dependent 11‐deoxycortisol could be detected, which was unaffected by treatment with abiraterone, except in H295R‐bearing mice, where 11‐deoxycortisol levels were significantly reduced by abiraterone.

3.4. Castration and abiraterone resistance is mediated via expression of ligand‐independent AR (variants) and GR signalling

VCaP tumours progressively growing in castrate hosts showed higher mRNA expression levels of full length AR and AR variants (AR‐V1 and AR‐V7) as compared to VCaP tumours growing in sham‐castrated animals, with no significant differences in expression level between abiraterone‐ and placebo‐treated animals. In placebo‐treated H295R‐supported VCaP tumours, AR and AR variant expression was significantly higher compared to intact animals, but lower than in castrate hosts without H295R. In this group, abiraterone significantly increased AR and AR‐variant expression, similar to levels found in progressive VCaP tumours from castrate animals without H295R (Figure 4A).

PSA mRNA expression was similar for all VCaP tumours irrespective of H295R support and regardless of abiraterone treatment (Figure 4A). Further analysis of AR target genes, such as FKBP5, TMPRSS2, TIPARP, and NDRG1, revealed no statistically significant differences between groups or treatments (Figure 4B), pointing towards restored AR signalling, despite limited detection of intratumoural testosterone and androgen synthesis intermediates in abiraterone‐treated tumours. Moreover, expression of the AR target genes TMPRSS2, PSA and FKBP5 was correlated with increased expression of full length AR and AKR1C3, but not with increased expression of CYP17A1 (Figure S5).

In parallel with AR, GR expression was elevated in progressive H295R‐unsupported VCaP tumours compared to tumours grown in sham‐castrated animals irrespective of abiraterone treatment. In H295R‐supported VCaP tumours, abiraterone treatment increased GR expression (Figure 4C). Similarly, mRNA for the GR‐regulated gene SGK1 could be detected in all VCaP tumours and was elevated in all castration‐resistant and abiraterone‐treated H295R‐supported tumours, whilst remaining unaltered in the intact arm. Upregulation of GR was significantly correlated with the expression of SGK1 (Figure 4C). Subgroup analysis indicated that in VCaP tumours with increased expression of GR, SGK1 was only upregulated in castration‐resistant tumours grown without H295R (open and closed red dots) and in abiraterone‐treated tumours grown with H295R (purple filled dots) (Figure 4C).

Expression of AR and GR was further validated by immunohistochemistry. In noncastrate animals, AR staining in VCaP tumours was only nuclear. In castrated animals, both with and without H295R tumours, AR staining was intense and also cytoplasmatic, indicating these tumours were growing in an androgen‐deprived environment (Figure 4D). Immunohistochemistry further revealed that in progressive VCaP tumours under castrate conditions, GR localised to the nucleus, whereas GR protein was absent in VCaP tumours grown in sham‐castrated hosts (Figure 4E).