Loss of a Negative Feedback Loop between IRF8 and AR Promotes Prostate Cancer Growth and Enzalutamide Resistance

Androgen and androgen receptor (AR) signaling pathway play a critical role in the carcinogenesis and progression of prostate cancer, which is the second leading cause of cancer-related deaths in North America (1). Consequently, androgen depletion therapy (ADT) has been the first-line therapy for primary prostate cancer for decades. Despite the initial efficacy of ADT, regeneration of tumor will occur eventually in almost every patient, leading to alleged castration-resistant prostate cancer (CRPC).

AR-targeted therapy is a gold standard therapy for CRPC (2–4). Enzalutamide is approved for the treatment of patients with CRPC, based on its ability to block androgen binding to AR in a competitive manner, inhibiting AR nuclear translocation and DNA fixation (5–7). Despite the success of enzalutamide in improving the overall survival of patients with CRPC, inherent or acquired enzalutamide resistance (ENZR) remains a major clinical challenge (8–10). AR deregulation, including overexpression (OE) of AR full-length (ARfl) and AR variants (ARv), has been identified as a unique factor consistently associated with the progression of prostate cancer to CRPC and ENZR (9); therefore, studying the mechanisms of AR deregulation is critical to improve the efficacy of current treatments. AR undergoes degradation mainly by the ubiquitin/proteasome system, as well as modification of protein stability triggered by ubiquitin-like signaling pathways, such as ISGylation (Interferon stimulated gene; refs. 11, 12). AR is a type I IFN-regulated protein and disruption of IFN system genes plays a novel function in malignant transformation of prostate cancer (12–15). However, the mechanism for AR maintaining its stability under the influence of IFN system remains unknown.

Activity of the IFN system is mainly regulated by IFN regulatory factor 8 (IRF8), a member of the IRF family (IRF1-9; ref. 16). Loss of IRF8 in immune cells leads to the occurrence of chronic myelogenous leukemia and aberrant methylation of IRF8 gene in nonhemopoietic cells play an increasingly important role in tumorigenesis (17–25). Among all the IRFs, only IRF8 facilitates a protein–protein interaction model, which interacts with proteins containing the PEST motif, a region that plays an important role in protein degradation by the proteasome system, including AR degradation (26, 27). Emerging evidences suggest that IRF8 may function in the cytosol ubiquitylation system (16, 28), but the exact role and the regulation mechanism of IRF8 in cancers, especially in prostate cancer tissues, have not been explored. More importantly, the relationship between IRF8 with AR in prostate cancer and whether IRF8 interacts with AR (containing the PEST motif) and functions in the regulation of AR stability are still unknown.

This study provides evidence to support an advantageous role for IRF8 in CRPC and ENZR, and this role of IRF8 is likely mediated through regulating AR stability. We then evaluated the potential of IFNα targeting IRF8 to improve the therapeutic efficacy of hormonotherapy in prostate cancer, suggesting that IFNα combined with enzalutamide is an attractive therapeutic strategy for CRPC and ENZR.

FBS and charcoal-stripped, dextran-treated FBS (CSS, depleted androgen, and any other steroid) were purchased from Biological Industries. Enzalutamide (MedChemExpress), R1881, DHT (Meilunbio), EGF, and IGF-1 were commercially obtained (Peprotech). Serial dilutions of all drugs were made using DMSO.

PC3 cells were cultured in F12 Medium, HEK293T were cultured in DMEM, and 22RV1 and LNCaP cells were cultured in RPMI1640 Medium. PC3, 22RV1, LNCaP, and HEK293T cells were purchased from Cell Bank of the Chinese Academy of Sciences (Shanghai, China), two stable LNCaP-sh-IRF8-Puro (LNCaP-shIRF8) and LNCaP-sh-negative control (LNCaP-shNC)-Puro cells were purchased from GenePharma Technology Co. Ltd. (GenePharma). All cells were authenticated by the short tandem repeat DNA profiling (Cobioer) and confirmed Mycoplasma free using GMyc-PCR Mycoplasma Test Kit (YeSen, 40601ES10) after last experiment, and used within 15 cell passages after thawing. All cell lines were cultured in medium supplemented with 10% FBS, 1% penicillin-streptomycin (Gibico), and 5% CO₂ at 37°C.

Plasmids including pcDNA3.1 (Vector), pcDNA3.1-hIRF8 (IRF8), and pcDNA3.1-hAR (AR), containing the whole CDS domains were constructed by Genscript; double luciferase reporter plasmid pEZX-FR03-hIRF8-luc, carrying the IRF8 promoter (-1106-+166 bp), was constructed by FulenGen; 3xFlag-Ub plasmid was generously provided by Dr. Guo-qiang Xu (Soochow University, Suzhou, China). For transient transfections, cells were seeded into 6-well plates at 150,000 cells per well and transfected with IRF8 or AR expression vectors using empty vector (pvector) as control. The total plasmid DNA was adjusted to the same with empty vector.

siRNA and shRNA targeting IRF8 are as follows: siNC (shNC): TTC TCC GAA CGT GTC ACG TTT C; siRNA1# (shIRF8A): GCA GTT CTA TAA CAG CCA GGG; siRNA2# (shIRF8B): GGG AAG AGT TTC CGG ATA TGG.

LNCaP-shNC and LNCaP-shIRF8B cells were plated in the 6-well plates in triplicate with 500 cells every well and cultured with complete medium for 14 days. For enzalutamide sensitivity experiment, once cells were attached, enzalutamide (10 μmol/L) were added to the media using 0.1%DMSO as control. The complete medium or culture medium containing enzalutamide or DMSO was replaced every 2 days, and cells were cultured for 10 days. Crystal violet was used to stain the colonies. Colony number and the inhibition effects of enzalutamide were calculated according the number of the cloning formation measured by ImagePro plus software.

Briefly, whole cell lysates were prepared with ice-cold RIPA lysis buffer or nuclear and cytoplasmic protein extraction lysis buffer with protease inhibitors (Roche). Approximately, 30 μg of total protein was separated by SDS-PAGE and transferred to polyvinylidene difluoride transfer membrane and the membranes were incubated with specific antibodies for IRF8 (ab28696, Abcam), AR (ab74272, Abcam), p-STAT1(Tyr701), p-STAT1 (Ser727), STAT1 (Cell Signaling Technology), and β-actin (bsm-33036M, Bioss). Each experiment was repeated at least twice with similar results. The densitometry data presented below the bands are “fold change” as compared with control normalized to respective β-actin from two independent Western blot analyses. Values are expressed as mean.

Total RNA was prepared by using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. RNA was reverse transcribed into cDNA using random hexamers, which was used for qPCR using gene-specific primers. Data were normalized by the level of GAPDH expression in each sample. The primer sequences used in this study are as follows: IRF8 forward 5′- TCG GAG TCA GCT CCT TCC AGA CT -3′, reverse 5′- TCG TAG GTG GTG TAC CCC GTC A -3′; AR forward 5′- CTA CTC TTC AGC ATT ATT CCA G -3′, reverse CAT GTG TGA CTT GAT TAG CA- 3′; GAPDH, forward 5′- CAT GAG AAG TAT GAC AAC AGC CT -3′, reverse 5′- AGT CCT TCC ACG ATA CCA AAG T -3′.

Cells treated with the appropriate stimuli were lysed with Western blot analysis and immunoprecipitation (IP) lysis buffer (Beyotime). Aliquots of 600 μg protein from each sample were precleared by incubation with 20 μL of Pierce Protein A/G magnetic beads (88803; Thermo Fisher Scientific) for 1 hour at 4°C and then incubated with anti-IRF8 antibody (5628s, Cell Signaling Technology), or anti-AR (ab74272, Abcam) in lysis buffer at 4°C overnight. Protein A/G beads were added and incubated for 2 hours at 4°C. The beads were washed five times with PBS and once with lysis buffer, boiled, separated by 10% SDS-PAGE, and analyzed by Western blotting as described above. For in vitro binding assays, purified recombinant Flag-tagged AR protein was incubated with recombinant His-tagged IRF8, and anti-flag agarose beads (Bimake) in buffer C. Bound immunocomplexes were washed three times with buffer B for nuclear extracts or buffer C-400 mmol/L NaCl for total protein extracts IP and in vitro binding assays.

The chromatin immunoprecipitation (ChIP) assay was performed using the EZ-Magna ChIP A/G Chromatin Immunoprecipitation Kit (Millipore) according to the manufacturer's protocol. Anti-AR antibodies and normal IgG were used to precipitate DNA. The normal IgG was used as a negative control and the PSA enhancer region was used as a positive control. The precipitated DNA was subjected to PCR to amplify the PSA and IRF8 promoter regions with the primers listed as follows, and were quantified with the qPCR using SYBR Green. PSA Enhancer+: TGG GAC AAC TTG CAA ACC TG, PSA Enhancer−: CCA GAG TAG GTC TGT TTT CAA TCC A; IRF8 (-872/-863 bp) F: 5′- TGT GTG ATT CTC TAC TGG GCA A -3′, R: 5′- CTG GAA ACG GAA AAA GAA GCG T -3′; IRF8 (-852/-843 bp) F: 5′-TTT CTT TTT CCA GTG TCG TTC TCC -3′, R: 5′- GGG CGT TAA GAT GTC CCC T -3′; IRF8 (-896/-783 bp) F: 5′- GGA TAG AAC GCG GAA ACG CT -3′, R: 5′- CCT GGG TGT GCA CTG ACA TTT A -3′.

A total of 13 benign prostatic hyperplasia (BPH), 20 untreated primary prostatic cancer tissues, and 13 CRPC paraffin-embedded tissues were collected from the Departments of Urinary surgery at Jiangsu Province Hospital of Traditional Chinese Medicine (TCM; Jiangsu, China), which is approved by the Institutional Review Board of the Jiangsu Province Hospital of TCM for IHC and MSP-PCR analysis.

Protein expression of IRF8 and AR in clinical prostate cancer tissues were determined using tissue microarray analysis (TMA, Shanghai OUTDO Biotech), containing 64 paired human primary prostate cancer/adjacent noncancerous lesion tissues. Tissue microarray were stained with hematoxylin and eosin to verify histology and the IHC staining of IRF8 (ab28696; Abcam) and AR (ab74272) were performed. The positive staining rate was estimated in three fields with different staining intensity by pathologists. The staining of IRF8 and AR in the TMA was scored independently by two pathologists blinded to the clinical data using the following criterion: the intensity of immunostaining was scored from 0 to 3 (0, negative; 1, weak; 2, moderate; and 3, strong); the percentage of immunoreactive was deemed as 0 (0%–5%), 1 (6%–25%), 2 (26%–50%), 3 (51%–75%), and 4 (76%–100%). The final score was calculated using the percentage score × staining intensity score as described previously (29). For tissue samples from patient with clinical prostate cancer and animal xenografts, the primary antibodies of the anti-IRF8 antibody (Abcam, 1:200), anti-AR antibody (Abcam, 1:200) or anti-PCNA antibody (Santa Cruz Biotechnology, 1:200), were used for staining at 4°C overnight. Immunostaining pictures were acquired using an inverted fluorescence microscope (Leica) and the integral optical density (IOD) sum was calculated using ImagePro plus software. The average diaminobenzidine staining intensity of the selected two cores represents the quantitative protein expression level in these tissues.

Nuclear extracts and affinity binding assays were prepared as described previously (30). Briefly, LNCaP cells were treated with 10 nmol/L DHT or 2,000 IU/mL IFNα2a for 48 hours, using ethanol and medium as control, respectively, and lysed with ice-cold nuclear protein extraction lysis buffer. 4.5-μg aliquots DNA formed by 8 pairs of the ISRE and ISRE-like primers [2 wild-type (WT) and 6 mutants, Mut] were conjugated to streptavindin M280 magnetic beads (Dynal) in the presence of 20 μg of salmon sperm DNA (Sigma) for 10 minutes at room temperature. DNA-coupled magnetic beads were incubated with 1 mg of nuclear extracts for 1 hour at 4°C. Beads were washed three times with washing buffer and bound materials were eluted in 1 × SDS sample buffer. Samples were separated by 8% SDS-PAGE for immunoblot detection with anti-AR or other antibodies as indicated at 4°C overnight, using 30 μg nuclear protein as input and Lamin B as negative control.

HEK293T cells (1 × 10⁷/100 mm dish) were precultured with 5% CSS overnight and transfected with pcDNA3.1-AR (1 μg). Cells were cotransfected with an artificial promoter luciferase reporter with both the proximal IRF8 5′ flank (flanking regions of IRF8 (−1106/+166 bp) and the firefly/renilla luciferase (pEZX-FR03-hIRF8, 3 μg) in the same vector. The empty reporter vector (pEZX-FR03) was used as a control. The transfected cells were cultured with 5% CSS for 24 hours and then seed into 96 wells (1 × 10⁴ cells/well) and treated with DHT (1 nmol/L, 10 nmol/L); R1881(1 nmol/L, 10 nmol/L); IGF-1(10 nmol/L); EGF (10 nmol/L) with or without enzalutamide (10 μmol/L). DMSO was used as a control. After incubation with these stimuli for 24 hours, the transfected cells were collected and AR transcriptional activity was detected by dual luciferase assay.

LNCaP cells (1 × 10⁷/100 mm dish) were precultured with 5% CSS overnight and transfected with various combinations of vectors to express IRF8 or control (0.75, 3 μg), IRF8-specific shRNA or scrambled control (3 μg). Cells were cotransfected with an androgen-dependent firefly luciferase reporter (pMMTV-Luc, 3 μg, for AR activity as described previously; ref. 31) and a renilla luciferase promoter (pRL-SV40, 0.04 μg, for transfection efficiency). The transfected cells were cultured with 5% CSS for 24 hours and then seeded into 96 wells (1 × 10⁴ cells/well) and treated with or without DHT (10 nmol/L).

Luciferase activities were measured 24 hours after treatment using the dual luciferase reporter system according to the manufacturer's instructions (Promega). All samples were tested in triplicate.

All athymic nu/nu BALB/c (BALB/c nude) and nonobese diabetic severely combined immunodeficient (NOD-SCID) mice, ages 4–6 weeks, were purchased from Lingchang biotech and housed in a specific pathogen-free facility and maintained in a standard temperature and light-controlled animal facility for 1 week before used. For transgenic animal experiment, the Hi-myc mice [FVB-Tg (ARR2/Pbsn-MYC) 7Key, strain number: 01XK8] were obtained from the Mouse Repository of the NCI. All animal procedures were performed according to the guidelines of the U.S. NIH on Animal Care and appropriate institutional certification, with the approval of the Committee on Animal Use and Care of Center for New Drug Evaluation and Research, China Pharmaceutical University (Nanjing, China).

For tumor growth of LNCaP cell xenograft (growing in intact BALB/C nude mice), 1×10⁷ LNCaP-shNC or LNCaP-shIRF8A cells (with higher KD efficiency, indicated as LNCaP-shIRF8) were resuspended in 100 μL medium containing 50% matrigel (BD Biosciences) and 50% growth media and subcutaneously injected in the flank of BALB/c nude mice. For the tumor growth of LNCaP-CRPC xenograft (growing in surgically castrated BALB/C nude mice with low serum androgen), BALB/c nude mice were anesthetized using 10% chloral hydrate. Testes were excised distal to the ligature, and the incision was closed with sterile dissolvable sutures and disinfected with betadine solution and ampicillin. Two weeks later, LNCaP-shNC or LNCaP-shIRF8 cells were inoculated as described above.

In the enzalutamide sensitivity study, orchiectomized BALB/c or NOD/SCID male mice were inoculated subcutaneously with 1 × 10⁷ LNCaP or LNCaP-shNC or LNCaP-shIRF8 cells. Once tumors were established, mice were randomly assigned to vehicle (1% carboxymethyl cellulose, 0.1% Tween-80, 5% DMSO), and enzalutamide (10 mg/kg) treatment groups. In the sensitivity study of enzalutamide combined IFNα, mice were treated with vehicle alone, enzalutamide (10 mg/kg) alone, IFNα (1.5 × 10⁷ IU/kg) alone or enzalutamide (10 mg/kg) combined with IFNα (1.5 × 10⁷ IU/kg). Enzalutamide and IFNα were administered by oral gavage and subcutaneously daily, respectively. Tumor volume measurements were performed every 2 days and were calculated by the formula: length × width²/2. At experimental endpoint, tumors were harvested and tumor inhibition ratio were calculated and compared with the average tumor weight in the vehicle control.

Three advanced and untreated patients with prostate cancer with bone metastasis were enrolled at the Department of Urinary Surgery at Jiangsu Province Hospital of TCM (Jiangsu, China), and written informed consent was received from participants prior to inclusion in the study, in accordance with the guidelines of the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), with the approval of the Institutional Ethics Review Boards of Jiangsu Province Hospital of TCM (number 2017NL-054-02). Inclusion criteria were histologically confirmed prostate adenocarcinoma, untreated, cT3-cT4 M0 and M1, serum PSA ≥ 150 ng/mL, Gleason score ≥ 9 (4+5 or 5+4), age ≤ 80 years, and normal liver function not suitable for definitive treatment. For case 1 clinical study, 2 patients received Goserelin Acetate Sustained-Release Depot (3.6 mg/28 days) plus bicalutamide (50 mg/days; Casodex) as maximal androgen blockade (MAB) therapy, one patient received MAB therapy plus recombinant IFNα2a (3×10⁶ IU/3 days; Yintefen). For case 2 clinical study, one advanced and untreated patients with prostate cancer with continued PSA level rise for 6 months were enrolled and received MAB therapy plus recombinant IFNα2a (3×10⁶ IU/3 days; Yintefen). Patients in both cohorts were discontinued when there was evidence of biochemical progression, defined as an increase of the PSA level ≥ 0.2 ng/mL on two successive occasions at least 1 month apart.

To reveal the relationship between IRF8 and AR protein expression in prostate cancer, prostate cancer tissue microarrays were immunohistochemically assessed using optimized anti-IRF8 and anti-AR antibodies. The representative primary prostate cancer specimens showed high IRF8 expression and high AR accumulation compared with the adjacent nontumor tissue (Fig. 1A; Supplementary Fig. S1A and S1B). Tumor samples with high expression of AR protein exhibited more IRF8 accumulation and a positive correlation was observed between protein levels of IRF8 versus AR in primary prostate cancer specimens (Supplementary Fig. S1C and S1D). Similar findings confirmed that IRF8 and AR were increased in tumor tissues of Hi-myc transgenic (TG) murine prostate cancer models during the prostate cancer progression (≥6 months; Fig. 1B; Supplementary Fig. S1E and S1F). Moreover, in LNCaP-CRPC xenografts models, IRF8 expression was significantly decreased in tumors treated with enzalutamide (Supplementary Fig. S1G). To clarify this point, we quantified IRF8 and AR protein levels using IHC of paraffin-embedded clinical BPH, primary prostate cancer and CRPC tissues. The results showed that IRF8 and AR were increased in primary prostate cancer tissues compared with BPH, while low IRF8 but high AR were observed in the representative CRPC specimens (Fig. 1C; Supplementary Fig. S1H). A positive correlation between IRF8 and AR in primary prostate cancer was confirmed. In contrast, statistically significant negative correlation was observed between protein levels of IRF8 versus AR in CRPC specimens (Fig. 1D). Similarity to CRPC, IRF8 was decreased but AR was increased in ENZR xenograft tissues and an inverse correlation between IRF8 with AR protein was observed (Fig. 1E; Supplementary Fig. S1I and S1J).

Previous findings revealed that epigenetic mechanisms including mutation and promoter methylation play a crucial role in the IRF8 expression (19, 25, 32–35). However, IRF8 was not frequently mutated and methylated in clinical prostate cancer tissues and cells, using methods of ref.1 in the Supplementary Table (Supplementary Fig. S1K; Supplementary Table). Lots of AR CHIP-seq data reveal that IRF8 is a putative target of AR upon DHT or R1881 treatment (36–38). Furthermore, we found several AR-binding sites at the promoter of IRF8 using JASPAR database. We next tested whether IRF8 expression is regulated upon AR activation and repression. AR activation by DHT increased IRF8 mRNA and protein levels in prostate cancer cells (Fig. 1F; Supplementary Fig. S2A–S2C). However, AR repression by enzalutamide treatment decreased IRF8 expression (Fig. 1G; Supplementary Fig. S2C).

In CRPC, AR can be bypass-activated by growth factors (GF) such as EGF, IGF-1, at low levels of androgen after ADT (39–42). In the experiments, we treated LNCaP cells with these factors and found that AR activation by GFs reduced IRF8 protein levels (Supplementary Fig. S2D). We further identified whether these factors regulate IRF8 transcription by AR signaling, and the results showed that DHT and R1881-induced AR activation promoted IRF8 transcription but bypass AR activation by GFs inhibited IRF8 transcription. In contrast, enzalutamide reduced IRF8 transcription, reversed androgen-mediated IRF8 upregulation and enhanced GFs-mediated IRF8 downregulation (Fig. 1H).

IRF8 transcription is mainly regulated by the binding of p-STAT1 to the IFN-stimulated response element motif (ISRE, TTTC A/G G/C TTTC; ref. 16). Predictably, one canonical ISRE motif and one ISRE-like motif were identified in the IRF8 promoter (Supplementary Fig. S2E). In our experiments, the amounts of AR bound to the ISRE and ISRE-like motifs were markedly increased after DHT treatment, whereas binding by AR were not increased upon GFs stimulation and only p-AR^Ser81 were activated by DHT (Supplementary Fig. S2F and S2G). Furthermore, we found mutations of the GAAA motif greatly diminished binding of AR to the ISRE-like motif while AR binding to ISRE mutant was not reduced (Fig. 1I), indicating the ISRE-like motif has stronger affinity for AR. We further determined whether the IRF8 is a direct AR-targeted gene using ChIP-qPCR assays. The results showed that AR was recruited to the ISRE and ISRE-like sites regardless of DHT treatment, whereas occupancy of the ISRE-like site was markedly increased and occupancy of the ISRE site was not changed in LNCaP cells upon DHT stimulation (Fig. 1J; Supplementary Fig. S2H and S2I). Together, these results identified AR as a factor that is recruited to the ISRE-like motif upon DHT stimulation, which promotes IRF8 transcription in an androgen–AR dependent pathway.

To functionally demonstrate the potential role of IRF8 in tumorigenesis of prostate cancer, we generated two stable IRF8-knockdown (IRF8-KD) LNCaP cell lines (Fig. 2A; Supplementary Fig. S3A). The proliferation ratios of LNCaP-shIRF8 cells were higher than those in LNCaP-shNC cells (Fig. 2B; Supplementary Fig. S3B). Furthermore, increased clonogenic ability, number, and diameter of tumor cell spheres were also observed in LNCaP-shIRF8 cells (Fig. 2C and D; Supplementary Fig. S3C and S3D). In vivo xenograft experiments showed that LNCaP-shIRF8 results in increased tumor growth and proliferating cell nuclear antigen (PCNA) protein expression in intact or castrated BALB/c nude mice (Fig. 2E–G; Supplementary Fig. S3E–S3G), whereas 22RV1-IRF8- OE inhibited the proliferation, clone formation, and tumor growth (Fig. 2H and I; Supplementary Fig. S3H). Moreover, IRF8-KD also increased P38/MAPK/ERK phosphorylation and CD133 expression compared with LNCaP-shNC cells (Fig. 2J), indicating that intrinsic IRF8 plays an important role in inhibiting prostate cancer progression.

To elucidate the mechanisms responsible for the growth-accelerating effects of IRF8-KD in prostate cancer, we assessed the effects of IRF8 on AR protein expression. IRF8-KD increased and IRF8-OE decreased endogenous AR including ARvs expression, especially in the nucleus (Fig. 3A and B; Supplementary Fig. S4A and S4B). Moreover, we demonstrated that IRF8 can reduce ectopic AR protein levels in AR⁻ HEK293T and PC3 cells transfected with AR alone or combined with various amounts of IRF8 expression plasmids (Fig. 3C and D).

Androgen binding to AR results in homodimerization and nuclear translocation, and subsequent induction of target gene expression, which is regulated by the ubiquitin/proteasome system (27). Native polyacrylamide gel electrophoresis of lysates from HEK293T cells transfected with AR and IRF8 showed that decreased level of AR protein mediated by IRF8 resulted in decreased expression of AR dimers in a high molecular weight complex to a degree equal to the reduction of ubiquitin-mediated AR dimers (Fig. 3E), indicating that IRF8-mediated AR protein degradation could affect AR homodimerization. So we next detected effect of IRF8 on AR transcriptional activity. The AR-targeted PSA is the most important marker for prostate cancer assessment. As expected, we found that PSA and AR protein were increased by IRF8 siRNA (Fig. 3F). We next explored AR activity using the pMMTV-luc luciferase reporter in IRF8-KD and IRF8-OE LNCaP cells. IRF8-KD increased and IRF8-OE decreased the DHT-dependent AR activity after androgen stimulation (Fig. 3G; Supplementary Fig. S4C and S4D), which was associated with nuclear AR expression mediated by IRF8 in LNCaP cells.

IRF8-mediated AR reduction could not be reversed by DHT treatment in both the cytosol and nucleus, suggesting that IRF8 has no effect on AR translocation (Supplementary Fig. S4E and S4F), while the presence of IRF8 accelerated AR degradation and shortened the half-life of AR (Fig. 4A and B; Supplementary Fig. S5A and S5B). IRF8-mediated AR degradation was reversed by proteasome inhibitor MG132, but MG101, calpastatin, and lysosomal enzyme inhibitor leupeptin had no effect on IRF8-mediated AR degradation (Fig. 4C and D; Supplementary Fig. S5C). These results indicated that the ubiquitin/proteasome involved in the IRF8-mediated AR degradation.

Cytoplasmic AR undergoes proteasome-mediated degradation in the absence of ligand. Our results showed that ubiquitin or IRF8-mediated AR degradation could be reversed by MG132 and DHT (Supplementary Fig. S5D and S5E), indicating that IRF8 and ubiquitin regulate the same cellular machinery responsible for cytoplasmic AR degradation. The subcellular colocalization of IRF8 and AR were both observed by immunofluorescent staining in LNCaP cells and ectopically overexpressed HEK293T cells (Fig. 4E; Supplementary Fig. S5F). We further found that IRF8 could precipitate with AR and vice versa (Fig. 4F). To confirm this interaction, we proved that IRF8 could be eluted with AR in vitro binding assays (Supplementary Fig. S5G). To identify region within AR responsible for its binding to IRF8, we cotransfected IRF8 with several N-terminal Flag-tagged AR-truncated mutants, including ARfl, AR-NTD, AR-DBD, AR-Hinge, and AR-LBD into HEK293T cells (Supplementary Fig. S5H). Co-IP assay showed that IRF8 was immunoprecipitated with ARfl, AR-NTD but not with AR-LBD, AR-Hinge, and AR-DBD (Fig. 4G). These results indicated that IRF8 physically interacts with the N-terminal domain of AR. We next analyzed the effect of IRF8 on AR ubiquitylation. Results showed that IRF8 enhanced AR polyubiquitination as visualized by staining with Flag-ubiquitin (Fig. 4H). There are two common types of polyubiquitination, which are distinguished by the lysine residue through which formation of K48-chain and K63-chain occurs. Immunoblotting (IB) revealed that both WT and K48 ubiquitin led to polyubiquitinated AR, but polyubiquitination was impaired in the presence of K63 ubiquitin (Fig. 4I), suggesting that IRF8-mediated AR polyubiquitination is mainly via the K48 branch. Moreover, the K48-chain of AR polyubiquitination is mainly mediated by mouse double minute 2 homolog (MDM2; ref. 43). IP experiments confirmed that AR interacted with MDM2 (Supplementary Fig. S5I). In addition, we detected the interaction between IRF8 and MDM2 (Supplementary Fig. S5J). Together with previous result that AR interacted with IRF8, we tested whether IRF8 could influence the interaction between AR and MDM2. IP experiments further supported that the interaction between AR and MDM2 was enhanced upon the expression of IRF8 (Fig. 4J).

Together with the ubiquitination and protein interaction experiments, our results suggested that IRF8 enhanced the interaction between AR and its E3 ligase MDM2 and thus promoted the ubiquitination and degradation of AR.

We further explored the effects of IRF8 on sensitivity to enzalutamide therapy in vitro and in vivo. LNCaP viability treated with enzalutamide was decreased, whereas LNCaP-shIRF8 cells were resistant to enzalutamide treatment with increased AR expression (Fig. 5A and B; Supplementary Fig. S6A and S6B). Moreover, the IRF8-OE 22RV1 cell was more sensitive to enzalutamide treatment (Supplementary Fig. S6C). Furthermore, the colony number with enzalutamide treatment was significantly decreased in LNCaP-shIRF8 cells compared with controls (P < 0.05; Fig. 5C; Supplementary Fig. S6D and S6E). Notably, IRF8-KD decreased and IRF8-OE promoted caspase-3 activity induced by enzalutamide treatment (Fig. 5D and E). We further confirmed that IRF8-KD could induce ENZR in castrated BALB/c nude and NOD-SCID mice. The enzalutamide treatment inhibited tumor growth of LNCaP-shNC (P < 0.05), while it did not significantly alter the growth of LNCaP-shIRF8 tumors (Fig. 5F; Supplementary Fig. S6F and S6G).

Because decreased IRF8 induces ENZR during ADT, we next explored a novel strategy to attenuate the resistance. The IRF8 expression could be induced by IFNs through phosphorylation and dimerization of STAT1 (p-STAT1^Tyr701 and p-STAT1^Ser727; refs. 19, 44, 45). IFNγ receptor is frequently inactivation but expression of IFNA receptors is much higher than IFNγ receptors in human prostate cancer cells (1, 3). Fundamental studies have shown IFNα phosphorylates STAT1 (p-STAT1^Tyr701 and p-STAT1Ser⁷²⁷) and activates IRF8 gene expression by binding to the GAS element in human NK and T cells (4–6). We explored whether IFNα could improve enzalutamide efficiency in prostate cancer by targeting the IRF8–AR axis. We found that IFNα increased IRF8 and decreased AR including ARvs expression in prostate cancer cells (Fig. 6A; Supplementary Fig. S7A). DNA pull-down assays showed that IFNα treatment promotes AR binding to both WT and mutant ISRE-like motif of the IRF8 promoter rather than affecting only the canonical STAT1 pathway (Fig. 6B and C; Supplementary Fig. S7B). Furthermore, AR degradation by the ubiquitin pathway was accelerated upon IFNα treatment, and the IFNα-induced AR reduction was inhibited by MG132; however, this was greatly diminished when IRF8 was KD in 22RV1 cells (Fig. 6D–F). These results demonstrated that IFNα-mediated IRF8 expression promotes AR degradation by the ubiquitin pathway. We next examined the synergy between IFNα and enzalutamide in inhibiting the cancer cells growth. While IFNα alone had limited effect on cancer cell death (Supplementary Fig. S7C), its combination with enzalutamide markedly dropped the IC₅₀ value of enzalutamide in prostate cancer cells (Fig. 6G).

We further assessed the therapeutic activity of IFNα and enzalutamide combination in CRPC mice. IFNα combined with enzalutamide showed the better curative effects compared with IFNα or enzalutamide alone, whereas there were no significant changes among all the treatments in IRF8-KD groups (Fig. 6H; Supplementary Fig. S7D–S7F), indicating that IRF8 may have an important role in the therapeutic effect of IFNα- enzalutamide combination therapy.

We retrospectively designed a clinical case study to evaluate the effects of IFNα combined with MAB therapy on advanced metastatic patients with prostate cancer. In the case 1 clinical study, after 3 months of MAB alone therapy, both patients had reduced serum PSA levels, and after 6 months, serum PSA levels were increased. In the MAB and IFNα combination treatment group, the patient serum PSA level continued to decrease with treatment and remained below the biological progression line during follow-up treatment for 11 months (Supplementary Fig. S7G, case 1). In the case 2 clinical study, the PSA level of the one untreated patient with prostate cancer was continuously increased over the course of 6 months. Once the patient received MAB combined with IFNα therapy, the PSA level declined continuously for 3 months. The combination therapy was stopped when the patient developed a fever, and the PSA level subsequently increased slightly (Supplementary Fig. S7G, case 2). These clinical studies indicated that the combination of IFNα and hormonal therapies may enhance the treatment efficacy.

In this work, we demonstrate the precise mechanisms of how IRF8 was involved in prostate cancer progression and ENZR. We report that IRF8 could reduce AR protein levels and block AR activation via the ubiquitin/proteasome systems (Fig. 6I). IRF8 is induced by IFNs as a transcription factor in the IFN system, it is thought to mainly function in the nucleus. We found IRF8 protein levels are upregulated in primary prostate cancer and its expression is elevated upon DHT stimulation; in contrast, its levels are downregulated in CRPC and ENZR tissues upon bypassing AR activation and AR repression by ENZ. In CRPC after ADT, IGF-1 and EGF can also enhance the bypass of AR activation via AR phosphorylation at sites different than that phosphorylated by DHT treatment (p-AR^Ser81; refs. 46–48) and we found that p-AR^Ser81 is responsible for the different effects on IRF8 expression caused by androgen–AR activation and bypass of AR activation (Supplementary Fig. S2F and S2G). Moreover, we found that decreased IRF8 expression in CRPC, resulted in tumor proliferation even under ADT and ENZR, accompanied by a significant upregulation of AR. Accordingly, IRF8-OE decreased AR expression. We found that IRF8 promoted AR degradation dependent on enhanced AR ubiquitylation in the cytosol. The function of IRF8 in the cytosol and ubiquitylation is less understood (49, 50). We report here that AR interacts with IRF8 in the cytosol through its AR-NTD and regulates its ubiquitylation. IRF8-mediated AR polyubiquitination is mainly branched through K48, and we verified MDM2 as the E3 ligase responsible for IRF8-mediated cellular AR polyubiquitination and degradation. Although we found that IRF8 promoted the interaction of AR and its E3 ligase MDM2 in the cytosol, the detailed mechanisms underlying the involvement of IRF8 in MDM2-mediated AR degradation need to be further investigated. We suggest that IRF8 is part of the signaling complex that includes AR and an E2 ubiquitin-conjugating enzyme, such as the MDM2–Ubc5 complex, altering the activity of the MDM2–Ubc5 complex (43). There may be additional E3 ligases that modulate IRF8-mediated AR responses, and further work is required to fully reveal the roles of IRF8 in the regulation of AR signaling pathway.

For translational medicine, we found that IFNα increased IRF8 expression and promoted AR degradation through both STAT1 and AR pathways activation (Fig. 6). Without androgen, treatment with IFNα alone promoted AR binding to both the WT and mutant ISRE-like domains in the IRF8 promoter, while the canonical phosphorylation and dimerization of the STAT1 pathway (p-STAT1^Tyr701 and p-STAT1^Ser727) did not compete with binding to the same site. In vivo and in vitro studies, although AR could be activated by IFNα, treatment with IFNα alone had limited effect on prostate cancer cell proliferation. It is possible that activation of the AR pathway by IFNα only promotes AR bound to ISRE-like motifs in the IRF8 promoter rather than enhancing AR bound to the androgen response element in the proliferation-related target gene promoter. Treatment of IFNα combined with enzalutamide significantly improved efficiency in suppressing tumor growth of LNCaP-CRPC, while the enhanced antitumor activity was attenuated in LNCaP-shIRF8 CRPC. Most importantly, in our clinical case study, IFNα combined with hormonotherapy MAB was more efficacious than MAB alone. Unfortunately, although the potential for combination treatment is obvious, the side effects of IFNα therapy, such as fever and excessive perspiration, may limit its clinical use as an antitumor therapy. Drug-resistant CRPC is current common and a major challenge in the management of prostate cancer, and targeting the AR axis by disrupting androgen–AR interactions remains the primary treatment for CRPC. We suggest that alternative drugs can induce IRF8 expression could be combined with enzalutamide therapy to target the IRF8–AR axis. In conclusion, we showed that decreased IRF8 in prostate cancer cells impair its interaction with AR and promotion on AR degradation via the ubiquitin/proteasome pathway, accompanied AR activation, AR-mediated CRPC progression, and ENZR, suggesting that IRF8 could be a promising target to overcome enzalutamide resistance in CRPC.

No potential conflicts of interest were disclosed.

Conception and design: H. Wu, J.-R. Zhou, Y. Yang

Development of methodology: H. Wu, L. You, Y. Li, Z. Zhao, G. Shi, Z. Chen, Z. Wang, X. Li, S. Du, W. Ye, X. Gao, J. Duan, W. Tao, Q. Zhu, Y. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Wu, L. You, Y. Li, S. Du, W. Tao, Y. Yang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Wu, J.-R. Zhou, Y. Yang

Writing, review, and/or revision of the manuscript: H. Wu, Y. Cheng, J. Bian, J.-R. Zhou, Y. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Chen, Q. Zhu, Y. Yang

Study supervision: Y. Yang

We thank Dr. Lutz Birnbaumer (NIEH) for critical comments and assistance with preparation of the manuscript. We thank pathologists Mrs. Ning Su and Mr. Hongbao Yang for H&E to verify histology and the IHC staining of IRF8 studies. These studies were supported by the National Natural Science Foundation of China (grant numbers 81903656 to H. Wu; 81772732 to Q. Zhu; 81673468 to Y. Yang; 81672752 to Z. Chen), “Double First-Class” University project (numbers CPU2018GF10 and CPU2018GY46 to Y. Yang), Natural Science Foundation of Jiangsu Province (number BK20180560 to H. Wu), and China Postdoctoral Science Foundation (number 2018M632430 to H. Wu).

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.