AKR1C3 Promotes AR-V7 Protein Stabilization and Confers Resistance to AR-Targeted Therapies in Advanced Prostate Cancer

Recently approved androgen receptor signaling inhibitors (ARSI) such as enzalutamide, abiraterone, and apalutamide improved the therapeutics for advance prostate cancer including castration-resistant prostate cancer (CRPC) patients (1–3). These ARSIs surpass the effect of conventional androgen-deprived therapy using GnRH receptor agonists and antagonists to circumvent androgen production to aid tumor progression (4). However, drug resistance still emerges with amplification of AR variants as the leading cause for the escape mechanism. Constitutively active AR variants can activate distinct transcriptional program and confers enzalutamide and abiraterone resistance (5–7). AR variants can be generated either by gene rearrangement or by mRNA splicing (8–10). However, regulation of AR variants at the protein level is still incompletely understood.

Previous studies demonstrate that AKR1C3 promotes enzalutamide and abiraterone resistance by activating the androgen biosynthesis pathway and AR signaling (11, 12). AKR1C3, also called HSD17B5, is a critical gene in androgen synthesis. Upregulation of AKR1C3 was readily observed in abiraterone- and enzalutamide-resistant prostate cancer cells. Not only does this play a central role in intracrine androgen biosynthesis, its expression is also closely correlated to disease stage (13–15). Moreover, intracrine steroids, including androgens, were elevated in castration and ARSI-resistant cells, which could be due to overexpression of steroidogenic genes such as AKR1C3 (11, 16, 17). AKR1C3 binds with full-length AR and coactivates AR in prostate cancer cells (18). In addition to the enzyme activity of AKR1C3, other functions of AKR1C3 have not been fully investigated yet.

In this study, we demonstrate that AKR1C3 reprograms AR/AR-V7 signaling in enzalutamide-resistant cells. AKR1C3 induces AR-V7 overexpression and stabilizes AR-V7 protein in resistant cells through alteration of the ubiquitin proteasome system. Targeting AKR1C3 by indomethacin activates UPR and p53 pathways but suppresses AR/AR-V7 signaling. Orally administrated indomethacin significantly enhances enzalutamide treatment through AKR1C3/AR-V7 signaling suppression. Our results highlight the role of AKR1C3/AR-V7 complex in enzalutamide and abiraterone resistance.

LNCaP, C4-2B, and CWR22Rv1 were maintained in RPMI1640 supplemented with 10% FBS, 100 units/mL penicillin, and 0.1 mg/mL streptomycin. Note that 293 cells were maintained in DMEM supplemented with 10% FBS, 100 units/mL penicillin, and 0.1 mg/mL streptomycin. All experiments with cell lines were performed within 6 months of receipt from the ATCC or resuscitation from cryopreservation. C4-2B cells were kindly provided and authenticated by Dr. Leland Chung, Cedars-Sinai Medical Center. The resistant cells were developed and are referred to as C4-2B MDVR (C4-2B enzalutamide resistant) and C4-2B AbiR (C4-2B abiraterone resistant) as previously described (19). C4-2B MDVR and C4-2B AbiR cells were maintained in 20 μmol/L enzalutamide-containing medium and 10 μmol/L abiraterone acetate–containing medium, respectively. Parental C4-2B cells were passaged alongside the resistant cells as an appropriate control. All cells were maintained at 37°C in a humidified incubator with 5% carbon dioxide. Enzalutamide, abiraterone acetate, and indomethacin were purchased from Selleck Chemicals.

For transfection of siRNA, cells were seeded at a density of 0.5 × 10⁵ cells per well in 12-well plates or 2 × 10⁵ cells per well in 6-well plates and transfected with 20 nmol/L of siRNA targeting the AR-V7 sequence (GUAGUUGUGAGUAUCAUGA) or control siRNA (Catalog# 12935300) using Lipofectamine-iMAX (Invitrogen). The effect of siRNA-mediated gene silencing was examined using qRT-PCR and Western blot 2 to 3 days after transfection. Lentiviral plasmids encoding shRNA targeting AKR1C3 (TRCN0000026561) were purchased from Sigma-Aldrich. pLenti-GFP Lentiviral control vectors were used as control. Lentiviral particles were produced in 293T cells after cotransfection of the lentivirus vectors, psPAX2 and pMD2.G, in 10 cm dishes. The lentivirus-containing medium was collected, and cells were infected.

Whole-cell protein extracts were resolved on SDS-PAGE, and proteins were transferred to nitrocellulose membranes. After blocking for 1 hour at room temperature in 5% milk in PBS/0.1% Tween-20, membranes were incubated overnight at 4°C with the indicated primary antibodies AR (441), AR (N-20), Ubiquitin (P4D1 and FL76; 1:1,000 dilution, Santa Cruz Biotechnology), AR-V7 (AG10008, Mouse monoclonal antibody, 1:1,000 dilution, precision antibody), FLAG M2 monoclonal antibody (F1804, 1:1,000 dilution, Sigma-Aldrich), AKR1C3 (A6229, 1:1,000 dilution, Sigma-Aldrich), AKR1C3 (11194-1-AP, 1:1,000 dilution, Proteintech Group, Inc.), and Tubulin (T5168, Monoclonal Anti-α-Tubulin antibody, 1:5,000 dilution, Sigma-Aldrich). Tubulin was used as loading control. Following secondary antibody incubation, immunoreactive proteins were visualized with an enhanced chemiluminescence detection system (Millipore).

LNCaP-neo or LNCaP-AKR1C3 cells were seeded on 12-well plates at a density of 0.3 × 10⁵ cells/well in phenol red–free RPMI 1640 media containing 10% charcoal-stripped FBS (CS-FBS). Total cell number was determined at 0, 3, and 5 days. LNCaP-AKR1C3 or C4-2B-AKR1C3 cells were transiently transfected with siRNA targeting AR-V7 or a control siRNA, and then treated with enzalutamide for 3 days, and total cell number was counted.

Total RNA was extracted using TriZOL reagent (Invitrogen). cDNA was prepared after digestion with RNase-free RQ1 DNase (Promega) and then subjected to real-time reverse transcription-PCR (RT-PCR) using Sso Fast Eva Green Supermix (Bio-Rad) according to the manufacturer's instructions and as described previously (20). Each reaction was normalized by coamplification of actin. Triplicates of samples were run on default settings of a Bio-Rad CFX-96 real-time cycler. The Primers used for RT-PCR were: AR-full length: Forward (F)-AAG CCA GAG CTG TGC AGA TGA, Reverse (R)-TGT CCT GCA GCC ACT GGT TC; AR-V7: F-AAC AGA AGT ACC TGT GCG CC, R-TCA GGG TCT GGT CAT TTT GA; KLK3: F-GCC CTG CCC GAA AGG, R-GAT CCA CTT CCG GTA ATG CA; FKBP5: F-AGA ACC AAA CGG AAA GGA GA, R-GCC ACA TCT CTG CAG TCA AA; UBE2C: F-TGG TCT GCC CTG TAT GAT GT, R-AAA AGC TGT GGG GTT TTT CC; Myc: F-TGA GGA GAC ACC GCC CAC, R-CAA CAT CGA TTT CTT CCT CAT C; FHL2: F-ACC AAG AGT TTC ATC CCC, R-TCA GGC AGT AGG CAA AGT; MED12: F-CTG GAC GAA GAT CGC GTC TG, 3-ATT CAA GCA GCT ATG GGA TTC AA; FKBP4: 5-GTC ATC AAG GCT TGG GAC AT, 3-CCC TCA TTG GGC TTA GCA TA; NCOA3: 5-GGT AGG CGG CAT GAG TAT GTC, 3-TGT TAC TGG AAC CCC CAT ACC T; Actin: F-AGA ACT GGC CCT TCT TGG AGG, R-GTT TTT ATG TTC CTC TAT GGG.

Equal amounts of cell lysates (1,500 μg) were immunoprecipitated using 1 μg of AKR1C3 antibody with 50 μL of protein A/G agarose with constant rotation overnight. The immunoprecipitants were washed with 10 mmol/L HEPES (pH 7.9), 1 mmol/L EDTA, 150 mmol/L NaCl, and 1% Nonidet P-40 twice with 1 mL each. The precipitated proteins were eluted with 30 μL of SDS-PAGE sample buffer by boiling for 10 minutes. The eluted proteins were electrophoresed on 8% SDS-PAGE, transferred to nitrocellulose membranes, and probed with indicated antibodies.

C4-2B MDVR cells were treated with vehicle or the AKR1C3 inhibitor indomethacin for 24 hours before RNA extraction. RNA-seq libraries from 1 μg total RNA were prepared using Illumina Tru-Seq RNA Sample, according to the manufacturer's instructions. mRNA-Seq paired-end library was prepared through Illumina NGS on HiSeq 4000: 2 × 150 cycles/bases (150 bp, PE). Around 30 M reads/sample were generated. Data analysis was performed with a Top Hat-Cufflinks pipeline and sequence read mapping/alignment using HISAT. StringTie Data were mapped to and quantified for 60,658 unique genes/transcripts. Gene and transcript expression is quantified as FPKM (Fragments Per Kilobase of transcript per Million mapped reads). Principal component analysis was conducted on the FPKM gene-level data for all genes/transcripts passing filter (Filtered on Expression > 0.1) in the Raw Data. The genes regulated by indomethacin treatment were clustered with Hierarchical Clustering algorithm by StrandNGS software. The RNA sequence data in the present study have been deposited in Gene Expression Omnibus with the accession number GSE129596.

Gene set enrichment analysis (GSEA) was performed using the Java desktop software (http://software.broadinstitute.org/gsea/index.jsp), as described previously (21). Genes were ranked according to the shrunken limma log² fold changes, and the GSEA tool was used in “pre-ranked” mode with all default parameters. KEGG-Ubiquitin–mediated proteolysis pathway was used in the GSEA analysis.

Note that 1 × 10⁴ 293 cells were plated in 4-well Nunc Lab-Tek II Chamber Slides and transfected with AR-V7, AKR1C3, for 3 days. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and incubated with 1% BSA to block nonspecific binding. Cells were incubated with anti-AR (N20, Santa Cruz Biotechnology) and anti-AKR1C3 antibodies (A6229, Sigma-Aldrich) overnight. Intracellular AR-V7 was visualized with FITC-conjugated secondary antibodies, AKR1C3 was visualized with Texas red–conjugated secondary antibodies, and nuclei were visualized with DAPI by All-in-One Fluorescence Microscope (BZ-X700).

All experimental procedures using animals were approved by the Institutional Animal Care and Use Committee of UC Davis. CWR22Rv1 cells (4 million) were mixed with Matrigel (1:1) and injected s.c. into the flanks of 5- to 6-week-old male SCID mice. Tumor-bearing mice (tumor volume around 50–100 mm³) were randomized into six groups (6 mice per group) and treated as follows: (1) vehicle control [15% Cremophor EL, 82.5% PBS, and 2.5% dimethyl sulfoxide (DMSO), i.p.], (2) enzalutamide (25 mg/kg, p.o.), (3) Indomethacin (3 mg/kg, p.o.), or (4) enzalutamide (25 mg/kg, p.o.) + Indomethacin (3 mg/kg, p.o.). Tumors were measured using calipers twice a week, and tumor volumes were calculated using length × width²/2. Tumor tissues were harvested and weighed after 3 weeks of treatment.

To assess the effect of enzalutamide on the growth of LNCaP-AKR1C3 tumors, 3- to 4-week-old SCID C.B17 mice were orthotopically injected with 2 million LNCaP-AKR1C3 cells into the prostate. After the mice plasma PSA reached to 5 to 10 ng/mL, the mice were surgically castrated. Two weeks later, the mice were randomized into two groups (4 mice per group) and treated as follows: (1) vehicle control (15% Cremophor EL, 82.5% PBS, and 2.5% DMSO, i.p.) and (2) enzalutamide (20 mg/kg, p.o.). Mice PSA level was monitored every week. Tumor tissues were harvested and weighed after 4 weeks of treatment.

The steroid extraction and analysis have been described previously (11, 22). Briefly, tumor samples (eight each of C4-2B tumors and MDVR tumors) were ground and suspended in 4 mL of a 1:1 water/methanol mixture. The suspension was homogenized, and the resulting homogenate was cooled on ice. The precipitated material was removed by centrifuging at high speed for 5 minutes, and the supernatant was removed and evaporated in a SpeedVaac (Labconco Inc.) followed by lyophilizer (Labconco Inc.). The residue was suspended in 150 μL of CH3OH/H2O (1:1), filtered through a 0.2 μm ultracentrifuge filter (Millipore Inc.), and subjected to UPLC/MS-MS analysis. Samples were run in duplicate during UPLC-MS/MS analysis. Samples were placed in an Acquity sample manager which was cooled to 8°C to preserve the analytes. Pure standards were used to optimize the UPLC-MS/MS conditions prior to sample analysis. The standard mixture was run before the first sample to prevent errors due to matrix effect and day-to-day instrument variations. In addition, immediately after the initial standard and before the first sample, two spiked samples were run to calibrate for the drift in the retention time of all analytes due to the matrix effect. After standard and spiked sample runs, a blank was injected to wash the injector and remove carry over effect.

Tumors were fixed by formalin- and paraffin-embedded tissue blocks and were dewaxed, rehydrated, and blocked for endogenous peroxidase activity. Antigen retrieval was performed in sodium citrate buffer (0.01 mol/L, pH 6.0) in a microwave oven at 1,000 W for 3 minutes and then at 100 W for 20 minutes. Nonspecific antibody binding was blocked by incubation with 10% FBS in PBS for 30 minutes at room temperature. Slides were then incubated with anti-AR (N20, at 1:200; Santa Cruz Biotechnology) at 4°C overnight. Slides were subsequently washed and incubated with biotin-conjugated secondary antibodies for 30 minutes, followed by incubation with avidin DH-biotinylated horseradish peroxidase complex for 30 minutes (Vectastain ABC Elite Kit, Vector Laboratories). The sections were developed with the diaminobenzidine substrate kit (Vector Laboratories) and counterstained with hematoxylin. Nuclear staining of cells was scored and counted in 5 different vision fields. Images were taken with an Olympus BX51 microscope equipped with DP72 camera.

All data are presented as mean ± SD of the mean from three independent experiments. Statistical analyses were performed with SPSS16.0. Differences between individual groups were analyzed by two-tailed Student t test or ANOVA followed by the Scheffé procedure for comparison of means. P < 0.05 was considered statistically significant.

It is crucial to understand how prostate cancer evolves to the drug-resistant stage and identify early intervention strategies for treating CRPC patients. Toward this goal, we have developed enzalutamide-resistant C4-2B MDVR cells (11). To understand potential resistance mechanisms, RNA sequence data from the C4-2B MDVR and parental cells were analyzed. As shown in Fig. 1A, a significant enrichment of steroid hormone biosynthesis signaling was revealed in C4-2B MDVR cells (P < 0.05) as well as in mouse LNCaP/AR xenograft tumors derived from enzalutamide-resistant LREX' cells (P < 0.001; ref. 23) by the GSEA. Data analysis showed that genes such as AKR1C3 involved in androgen biosynthesis were upregulated in the enzalutamide-resistant xenograft tumors (Fig. 1B; Supplementary Fig. S1). Classically, AKR1C3 is one of the most important enzymes catalyzing androstenedione conversion to testosterone, and it also facilitates the 5′ dione and back door synthesis pathways by catalyzing 5α′ androstenedione and androsterone to 5′ androstanediol, respectively. Both products are then converted to Dihydrotestosterone (DHT). Our previous study showed that C4-2B MDVR cells had significantly higher androgen production compared with the parental line (11). In the present study, we confirmed these results in xenograft tumors by LC-MS. As shown in Fig. 1C, C4-2B MDVR xenograft tumors expressed significantly higher testosterone levels compared with C4-2B parental xenograft tumors (P = 0.0301). IHC staining of AR in C4-2B MDVR xenografts was predominately in the nucleus compared with C4-2B parental tumors (Fig. 1D). To examine the AR response to androgen stimulation, C4-2B parental or C4-2B MDVR cells were cultured in CS-FBS conditions for 3 days and treated with 10 nmol/L DHT overnight. As shown in Fig. 1E, C4-2B MDVR cells expressed significantly higher AR levels by DHT stimulation compared with C4-2B parental cells. Levels of AR-V7, AKR1C3, and c-Myc were upregulated in C4-2B MDVR cells. To further determine the AR/AR-V7 signaling pathway regulation in enzalutamide-resistant cells, expression of genes involved in AR signaling was analyzed. As shown in Fig. 1F and G, the expression of several AR/AR-V7 coactivators was significantly upregulated in C4-2B MDVR cells, such as FHL2, MED12, FKBP4, and NCOA3, but that of classical AR target genes, such as PSA was decreased (Supplementary Fig. S2), which suggested a distinct AR signaling is constitutively active in C4-2B MDVR cells. Taken together, the results showed that steroid hormone biosynthesis pathway and a distinct AR signaling are activated in resistant C4-2B MDVR cells.

Our previous study found that knockdown of AKR1C3 downregulated AR and AR-V7 expression in resistant cells (12). To understand how AKR1C3 regulates AR/AR-V7 expression, we constructed stable AKR1C3-overexpressing LNCaP and C4-2B cells and determined AR/AR-V7 expression. As shown in Fig. 2A and Supplementary Fig. S3, AKR1C3-overexpressing LNCaP-AKR1C3 and C4-2B-AKR1C3 cells express higher levels of AR-V7 and AR-FL protein compared with LNCaP-neo and C4-2B-neo cells. Other AR variants were also increased by the pan-AR antibody determination; however, AKR1C3 overexpression did not affect AR-FL or AR-V7 mRNA levels in either LNCaP-AKR1C3 or C4-2B-AKR1C3 cells (Fig. 2B). Next, we determined if AKR1C3 binds with AR-V7 in resistant cells. As shown in Fig. 2C, AKR1C3 and AR-V7 formed complex in both C4-2B MDVR and CWR22Rv1 cells. The results were also confirmed by dual immunofluorescence staining (Fig. 2D).

To further investigate the role of AKR1C3 in castration resistance, we first determined that overexpression of AKR1C3 significantly promoted the LNCaP cell growth in CS-FBS condition (Fig. 3A). Then we used an in vivo xenograft model to determine the effects of enzalutamide in the LNCaP-AKR1C3 stable cell line. As shown in Fig. 3B–C, LNCaP-AKR1C3 cells slightly responded to the castration; within 1 to 2 weeks, the tumors were starting to relapse, and the relapsed LNCaP-AKR1C3 tumors were resistant to enzalutamide treatment. We also determined the AR-V7 level in LNCaP-AKR1C3 tumors. As shown in Fig. 3D, LNCaP-AKR1C3 tumors expressed significantly higher AR-V7 compared with the LNCaP-neo cells. To further determine if AKR1C3-induced AR-V7 overexpression confers resistance to enzalutamide, AR-V7 was knocked down in LNCaP-neo and LNCaP-AKR1C3 cells and then treated with enzalutamide. Knockdown of AR-V7 slightly reduced growth of LNCaP-neo cells but enzalutamide itself significantly suppressed their growth. Knockdown of AR-V7 in combination with enzalutamide did not further reduce cell growth, indicating that AR-V7 knockdown had no effect on the sensitivity of LNCaP-neo cells to enzalutamide (Supplementary Fig. S4). In LNCaP-AKR1C3 cells with higher levels of AR-V7, knockdown of AR-V7 slightly decreased cell proliferation; however, AR-V7 knockdown combined with enzalutamide significantly suppressed cell growth (Fig. 3E). The results were also confirmed in C4-2B AKR1C3 cells (Fig. 3F). These data suggest that AKR1C3 induced AR-V7 overexpression and thus confers resistance to enzalutamide treatment.

To further dissect the underlying mechanisms of AKR1C3-mediated AR-V7 protein upregulation, we determined AR/AR-V7 protein stability in both C4-2B-neo and C4-2B-AKR1C3 cells. As shown in Fig. 4A, overexpression of AKR1C3 significantly extended the half-life of AR-V7 protein in C4-2B-AKR1C3 cells (over 8 hours) compared with the control cells (less than 2 hours). AR-FL protein half-life was also extended in C4-2B-AKR1C3 cells. Next, we determined if AKR1C3 knockdown affected AR/AR-V7 protein stabilization in CWR22Rv1 cells. As shown in Fig. 4B, AKR1C3 knockdown significantly shortened the half-life of AR-V7 protein (around 2 hours) in CWR22Rv1 cells compared with the vector-transfected controls (> 8 hours). AR-FL protein half-life was also shortened by AKR1C3 knockdown in CWR22Rv1 cells. The results were also confirmed in C4-2B MDVR cells (Fig. 4C). Previous studies showed that indomethacin is a potent AKR1C3 inhibitor that enhanced enzalutamide treatment by decreasing the testosterone level in resistant cells (11). Here, we found indomethacin suppressed AR and AR variants expression in C4-2B MDVR cells (Fig. 4D). Inhibiting proteasome activity by MG132 rescued AR/AR-V7 protein expression in C4-2B MDVR cells (Fig. 4E). In addition, indomethacin significantly enhanced AR/AR-V7 protein ubiquitination in C4-2B MDVR cells (Fig. 4F). These data suggest that AKR1C3 stabilizes AR/AR-V7 protein levels through the ubiquitin–proteasome pathway.

To further address the inhibitory effects of indomethacin in resistant prostate cancer cells, we analyzed the RNA sequencing data from C4-2B MDVR cells treated with indomethacin and identified gene programs that are regulated by AKR1C3 inhibition. Analyzing by GSEA, we identified that the unfolded protein response (UPR), p53 signaling, apoptosis, and hypoxia pathways were the top pathways upregulated by indomethacin treatment. The pathways most downregulated include E2F targets, cell-cycle, and Myc targets (Fig. 5A). Using hierarchical clustering to plot the heatmap with the genes found regulated by indomethacin, 4,941 genes were upregulated or downregulated by indomethacin (Fig. 5B, left). At the individual gene level, we determined that genes involved in UPR and ER stress (CHAC1, DDIT4, CEBPB, and ATF6) and genes in the p53 and apoptosis pathways (TP53, CDKN1A, and SOCS1) were upregulated in indomethacin-treated cells. We also found that Myc and cell-cycle pathway genes (Myc, MCM4, and CCNE2) and AR/AR-V7–regulated genes (KLK3, FKBP5, and UBE2C) were downregulated by indomethacin treatment (Fig. 5B, middle and right). Further GSEA revealed AR and AR-V7 pathways were significantly blocked by indomethacin treatment in C4-2B MDVR drug-resistant cells (Fig. 5C). The AR and AR-V7 target genes were verified by qRT-PCR. As shown in Fig. 5D, KLK2, KLK3, NKX3-1, FKBP5, UBE2C, and Myc were decreased by indomethacin treatment. Notably, AR-V7 preferentially regulated genes such as UBE2C and Myc were significantly suppressed by indomethacin treatment (Fig. 5D). These results suggested that indomethacin treatment significantly disrupts the drug resistant gene programs and suppresses AR/AR-V7 signaling.

Previous data suggested indomethacin enhanced enzalutamide treatment when administered through i.p. injection (11). To further identify the potential activity of indomethacin in vivo, we determined its tumor inhibition effects though oral administration. As shown in Fig. 6A–C, CWR22Rv1 tumors were completely resistant to enzalutamide treatment, orally administered indomethacin significantly reduced tumor growth, and indomethacin combined with enzalutamide treatment further suppressed tumor growth. However, all treatments did not affect the mice body weight (Fig. 6D). The intratumoral testosterone level of each treatment group was also determined; as shown in Fig. 6E, enzalutamide slightly decreased the testosterone level, however, indomethacin and indomethacin plus enzalutamide treatment group significantly decreased the tumor testosterone level. We also extracted tumor proteins and found that indomethacin alone and the combination treatment groups significantly decreased AR/AR-V7, c-Myc, and Bcl-2 expression in these tumors. These results suggest that targeting AKR1C3 with small-molecule indomethacin enhances enzalutamide treatment in vivo through suppressing both intratumoral testosterone and AR-V7 expression.

Our study discovered a novel mechanism by which AKR1C3 activates the ubiquitin–proteasome pathway and thus induces AR-V7 protein stabilization. We show that the AKR1C3-associated steroid hormone biosynthesis pathway is activated in enzalutamide-resistant models. AR and AR-V7 signaling is reprogramed in resistant cells, possibly through AKR1C3 upregulation. We further determined that AKR1C3 promotes AR and AR-V7 stabilization in enzalutamide-resistant cells. Targeting AKR1C3 with indomethacin significantly enhances enzalutamide treatment in vivo through inhibition of AR-V7. Our data highlight that AKR1C3 and AR-V7 are major contributors promoting prostate cancer resistance to ARSI. The main function of AKR1C3 is to catalyze the reduction of 5α-dihydrotestosterone to 5α-androstane-3α,17β-diol with its 3α-HSD activity and conversion of androstenedione to testosterone by its 17β-HSD activity (13). On the other hand, AKR1C3 also catalyzes the formation of prostaglandin (PG) F_2α and 11β-PGF_2α from PGH2 and PGD2, respectively. The PGF_2α and 11β-PGF_2α can inactivate peroxisome proliferator-activated receptor gamma and has antiproliferative effects (24). Emerging evidence shows that AKR1C3 not only functions as an enzyme, but also plays important roles in regulating oncogenic protein expression and activity. Several AKR1C3 oncogenic regulators have been identified previously. ERG upregulates AKR1C3 expression and leads to biochemical reduction of 5α-Androstenedione to DHT in prostate cancer cells (25). AKR1C3 also regulates E3 ligase SIAH2 protein stability and thus enhances SIAH2-dependent AR activity in prostate cancer cells (26). In the present study, we found that AKR1C3-involved steroidogenesis pathway plays a central role in enzalutamide resistance. Two independent databases both indicated that androgen biosynthesis pathway was activated in enzalutamide-resistant cells and xenograft tumors. In addition, we discover a new function of AKR1C3 that AKR1C3 promotes AR-V7 stabilization at the protein level. Our findings highlight the importance of AKR1C3 in next-generation antiandrogen resistance. Targeting AKR1C3 with indomethacin leads to AR/AR-V7 degradation and may be a valid strategy to overcome the resistance.

As an NSAID, indomethacin is used for reducing pain and inflammation by inhibiting COX-2. Emerging evidence suggests that indomethacin inhibits AKR1C3 activity and enhances ARSI treatments (11, 12). Indomethacin significantly promotes AR/AR-V7 protein degradation through activation of the ubiquitin–proteasome system and affects the stability of other proteins as well. Currently, the clinical trial of indomethacin and enzalutamide combination treatment in CRPC patients is ongoing (NCT02935205; ref. 27). Our study provides further support that targeting AKR1C3 by indomethacin significantly improves enzalutamide treatment through AR/AR-V7 inhibition. The second-generation ARSI brought a new revolution to advanced prostate cancer therapy. Unfortunately, the initial responders will inevitably develop resistance. AR variants which lack the ligand-binding domain, particularly AR-V7, are involved in resistance to ARSI. They are frequently expressed in CRPC patients and correlate with poor survival (28, 29). These variants constitutively activate AR target genes and promote the androgen/AR transcriptional pathway in the absence of ligand. We previously reported that C4-2B cells treated chronically with enzalutamide (C4-2B MDVR) had significantly enhanced AR variants expression, which suggested AR variants–mediated distinct AR activation is sufficient to support cell proliferation and drug resistance (19). AR deregulation has been found in 80% CRPC patients (30). AR overexpression in prostate cancer cells confers AR hypersensitivity to low levels of androgen and might result the drug resistance (31). In this study, we found that upregulation of AR expression induced by androgen treatment was more profound in enzalutamide-resistant cells than parental cells. Bioinformatic analysis revealed that AR was deregulated and distinct AR pathway was reactivated in C4-2B MDVR cells. Several AR coactivators, such as FHL2, MED12, FKBP4, and NCOA3, were upregulated. Notably, FHL2 was defined not only as a new AR coregulator but also bound with AR variants especially AR-V7 in prostate cancer cells. Nuclear localization of FHL2 in CRPC cells may promote constitutive AR-V7 activation (32). Our data suggested that chronic enzalutamide treatment in prostate cancer cells deregulated AR/AR-V7 signaling in C4-2B MDVR cells, possibly through the AKR1C3-mediated steroid hormone biosynthesis pathway activation.

Previous studies established that AR variants are generated through structure rearrangement of AR gene or RNA splicing. AR-V7 can be regulated by various modulators of RNA splicing, such as U2AF65, ASF/SF2 (8), and JMJD1A (33). It can also be regulated through cistromes, such as Hoxb13 (34) and FOXA1 (35). However, mechanisms of AR variant protein regulation have not yet been fully investigated. We recently reported that HSP70/STUB1complex–mediated AR-V7 protein homeostasis controls next-generation antiandrogen resistance. HSP70 binds AR-V7 and forms a complex assisting AR-V7 protein maturation. STUB1 induces AR-V7 protein degradation by blocking HSP70/AR-V7 complex formation (36). Li and colleagues reported that PP-1 and Akt signaling regulates AR-V7 phosphorylation on serine 213 through E3 ligase MDM2. MDM2 recognizes the AR-V7 phosphorylation site and induces AR-V7 ubiquitination and protein degradation (37). A previous study found that AKR1C3 is an AR coactivator that interacts with AR in prostate cancer cells, xenograft tumor samples, and CRPC patient samples (18). Our data demonstrating that AKR1C3 interacts with AR-V7 reinforces the novel function of AKR1C3. AKR1C3 not only controls the full-length AR-required activation via ligand-binding but also regulates the truncated AR that bypasses this activation. Our data indicated that targeting AKR1C3 by shRNA or the small-molecule inhibitor indomethacin significantly suppressed AR-V7 protein expression through the ubiquitin–proteasome pathway alteration. Notably, c-Myc and Bcl-2 which are oncogenic proteins possibly regulated by AR-V7 were also suppressed by AKR1C3 inhibition. Myc can deregulate the AR transcriptional program and drive prostate cancer progression (38, 39). Our data suggest that AKR1C3 plays an important role in controlling androgen signaling and prostate cancer progression.

Collectively, our results identify a new role for AKR1C3 in next-generation antiandrogen resistance. We demonstrated that the AKR1C3/AR-V7 complex collaboratively confers resistance to AR-targeted therapies. Targeting AKR1C3 not only can inhibit intracrine androgen synthesis, but also can further suppress AR/AR-V7 signaling by altering its protein stability.

C. Liu has an ownership interest (including patents) in a patent application covering the use of indomethacin. No potential conflicts of interest were disclosed by the other authors. A.C. Gao is also a Research Career Scientist at VA Northern California Health Care System.

Conception and design: C. Liu, C.P. Evans, A.C. Gao

Development of methodology: C. Liu, A.C. Gao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Liu, J.C. Yang, C.M. Armstrong, L. Liu, B. Zou, A.P. Lombard, L.S. D'Abronzo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Liu, J.C. Yang, A.P. Lombard, C.P. Evans, A.C. Gao

Writing, review, and/or revision of the manuscript: C. Liu, J.C. Yang, C.M. Armstrong, A.C. Gao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Liu, J.C. Yang, W. Lou, X. Qiu, A.C. Gao

Study supervision: C. Liu, C.P. Evans, A.C. Gao

We sincerely thank the Dr. Clifford G. Tepper and Genomics Shared Resource (GSR) at the UC Davis Comprehensive Cancer Center for their assistance in our study. This work was supported in part by grants NIH/NCI CA168601, CA179970, DOD PC150229, and the U.S. Department of Veterans Affairs, Office of Research and Development BL&D grant number I01BX0002653 (A.C. Gao), and a Research Career Scientist Award (A.C. Gao).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.