Introduction

Prostate cancer is a common malignancy among men1,2,3. The androgen/androgen receptor (AR) signaling is the defining and driving force in prostatic carcinogenesis4,5and androgen deprivation therapy (ADT) via surgical or chemical castration is thus widely used to treat locally advanced or metastatic prostate cancer6. For example, abiraterone acetate7,8 and enzalutamide (MDV3100)9,10 are clinically approved AR signaling inhibitors for prostate cancer treatment. However, most ADT-treated patients ultimately develop resistance and progress to lethal castration‐resistant prostate cancers (CRPCs), which usually maintain AR activity after 12–24 months of ADT5,11. A better understanding of the molecular basis of androgen/AR signaling should improve the development of ADT for CRPCs.

The zinc finger homeobox 3 (ZFHX3, also known as ATBF1) is a large transcription factor containing 4 homeodomains, 23 zinc-finger domains, and multiple other motifs12. Previous studies indicate that ZFHX3 plays a tumor suppressor role in prostate cancer. The ZFHX3 gene undergoes frequent loss-of-function mutation in metastatic or high-grade human prostate cancers13,14,15. The rs8052683 SNP, located in an intron of ZFHX3, reduces ZFHX3 expression in the prostate, and its expression level is significantly associated with prostate cancer16. In mice, specific deletion of Zfhx3 in the prostate induces prostatic intraepithelial neoplasia and promotes the development of prostate tumors induced by Pten deletion17,18. It is currently unknown whether loss of ZFHX3 function interacts with AR signaling during prostatic carcinogenesis.

In cultured AR-positive human prostate cancer cells (C4-2B), ZFHX3 is necessary for estrogen receptor beta (ERβ) to inhibit cell proliferation by downregulating MYC19. In addition, ZFHX3 also interacts with ERα, PR, and PRLR to modulate their functions in mammary gland development and breast cancer growth20,21,22,23all of which belong to hormone receptors along with AR. Taken together with our previous finding that the AR upregulates the transcription of ZFHX3 via its binding to the androgen-responsive elements (AREs) of ZFHX3 promoter in both mouse and human prostate cells15we hypothesized that the proper function of AR likely depends on the existence of ZFHX3 in prostate epithelial cells. Supporting this hypothesis are the correlations between ZFHX3 deletion and reduced AR activities and the upregulation of androgen-responsive genes in prostate cancer samples15.

In this study, we examined whether ZFHX3 modulates the function of AR signaling in prostate cancer cells. We found that ZFHX3 is integral to AR’s transcriptional activities, as the loss of ZFHX3 upregulated some and downregulated more of the AR target genes. In addition, ZFHX3 loss compromised but did not eliminate AR’s binding to gene promoters in their transcription. ZFHX3 is physically associated with AR via multiple regions. ZFHX3 loss attenuated the inhibitory effect of an AR antagonist on cell proliferation and colony formation in cultured cells, and higher AR expression levels were significantly correlated with worse disease-free survival only in prostate cancers with lower ZFHX3 levels. These findings suggest that ZFHX3 is integral to AR functions in gene transcription, cell proliferation, and prostatic carcinogenesis.

Materials and methods

Cell culture and transfection

Human embryonic kidney 293T cells were purchased from ATCC (Manassas, VA) and cultured in DMEM medium supplemented with 10% fetal bovine serum (FBS; Gibco, Waltham, MA). Human prostate cancer cell lines LNCaP and 22Rv1 were purchased from ATCC and cultured in RPMI-1640 medium supplemented with 10% FBS. The C4-2B line, a derivative of LNCaP, was a gift from Dr. Leland Chung of Cedar Sinai Medical Center. Wt is the vector control, while KO3 and KO8 are two ZFHX3-null clones of C4-2B cells generated in a previous study using the CRISPR/Cas9 system19. All cell lines were cultured in a humidified incubator (37oC and 5% CO2). For experiments involving the treatment of R1881, a synthetic androgen (Melonepharma, Dalian, China), cells were incubated in phenol red-free medium containing 5% charcoal-stripped FBS for 24 h.

Lipofectamine RNAiMAX and Lipofectamine 2000 (Invitrogen, Carlsbad, CA) were used to transfect siRNA and plasmids into cells. Sequences of siRNAs against human ZFHX3 and AR were 5’-AGAAUAUCCUGCUAGUACA-3’ and 5’-CAAGGGAGGUUACACCAAA-3’, respectively, which were validated in previous studies20,22,24.

Construction of expression plasmids

Mammalian expression plasmids for AR (pSG5-AR), HA-tagged ZFHX3 (HA-ZFHX3), HA-tagged ZFHX3 fragments A-F (HA-ZFHX3-A to HA-ZFHX3-F), and the pKLK3-Luc luciferase reporter plasmid containing ARE-rich enhancer and promoter region of the KLK3 gene were generated in our previous studies20,24. Additional HA-tagged ZFHX3-A fragments, ZFHX3-A1 to ZFHX3-A5, were created using PCR-based approaches. Primer sequences are listed in Table 1.

Table 1 PCR primer sequences used in the study.
Full size table

Immunoprecipitation (IP) and western blotting

Total proteins were extracted using the Cell Lysis Buffer for Western and IP (Beyotime, Shanghai, China), containing PMSF (Sangon Biotech, Shanghai, China). Cell lysates were incubated with anti-HA-agarose beads at 4oC overnight (Sigma, St Louis, MO). Beads were washed and eluted, and supernatants were analyzed by western blotting. Nuclear and cytoplasmic proteins were extracted using the Nuclear Protein Extraction Kit (Solarbio, Beijing, China) containing PMSF.

Proteins were subjected to 4% (for ZFHX3) or 10% (for all other proteins) SDS-PAGE and then transferred to PVDF transfer membranes with 0.45 μm pore size. After incubation with 5% nonfat milk for 30 min at room temperature, the membranes were probed with primary antibodies overnight at 4oC. Membranes were incubated with the secondary antibodies for 1 h at room temperature. The WesternBright ECL HRP Substrate (Advansta, Menlo Park, CA) was then used with the ChampChemi 610 Plus Imaging System (SAGECREATION, Beijing, China) to capture images.

Antibodies used in western blotting included ZFHX3 (1:800, prepared in our previous study [20]), AR (1:2000, 5153 s, Cell Signaling, Danvers, MA), KLK3 (1:3000, 10679-1-AP, Proteintech, Wuhan, China), HA-tag (1:3000, 3724 s, Cell Signaling), Lamin B1 (1:1000, 13435 s, Cell Signaling), and α-tubulin (1:10000, T1699, Sigma, St Louis, MO).

The western blotting images presented in the figures have been cropped to standardize the presentation. The absence of images of adequate length in the figures is due to the fact that, after protein transfer, the membranes were cut to enable separate detection and avoid cross-reactivity or signal interference between proteins of similar molecular weights. This prevents membrane damage from repeated stripping/reprobing and ensures detection specificity. The original blots can be found in the Supplementary Information (Figs. S1–S4).

RNA extraction and real-time qPCR

Total RNA was extracted using the Eastep Super Total RNA Extraction Kit (Promega, Madison, WI), and cDNA was synthesized using the M-MLV Reverse Transcriptase (Promega). Real-time qPCR was performed with the 2x SYBR qPCR Mix (KT Life, Shenzhen, China) using the qTOWER3/G system (Analytik Jena, Jena, Germany). Primer sequences are listed in Table 1.

Luciferase reporter assay

Wt and KO8 cells were transiently transfected with pGL3-Basic or pKLK3-Luc and pRL-TK (Renilla luciferase plasmid, Promega, internal control). SiRNA transfection was carried out six hours after plasmid transfection. Forty-eight hours after plasmid transfection, including 24-hour R1881 or enzalutamide treatment, cells were lysed in 5 × lysis buffer (Promega) for 30 min, and luciferase activities were determined using the Dual-Luciferase Reporter Gene Assay Kit (Promega). The luciferase activity was normalized to the Renilla luciferase activity in each reaction. Experiments were performed in triplicate.

Chromatin Immunoprecipitation (ChIP) assay

The SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads; Cell Signaling) was used for ChIP according to the manufacturer’s instructions. Briefly, C4-2B cells grown in complete medium were cross-linked with 1% formaldehyde at room temperature for 10 min, quenched with glycine for 5 min, collected, digested with micrococcal nuclease for 20 min at 37oC, and sonicated. ChIP was performed with an anti-AR antibody (06–680-AF488, Millipore, Billerica, MA) or IgG. ChIP products were detected using regular PCR, and primer sequences are listed in Table 1. The agarose gel electrophoresis images of the PCR products shown in Fig. 1 have been cropped to standardize the presentation. The original gel images can be found in the Supplementary Information (Fig. S1).

Fig. 1
figure 1

Loss of ZFHX3 decreases the expression of AR target genes in prostate cancer cells. (A) Enrichment of hallmark gene sets in Wt and KO8 clones of C4-2B cells, with the x-axis representing normalized enrichment scores (NES). (B) Heatmap of the 25 detected genes in 27 genes indicative of AR activities (GLRA2 and CD200 were not detected). The genes marked in red and blue are significantly upregulated and downregulated in the ZFHX3-null KO8 clone based on the RNA-seq data. (C-E) Knockout or knockdown of ZFHX3 downregulated KLK3, FKBP5 and TMPRSS2 mRNA expression and KLK3 protein expression in C4-2B and LNCaP cells, as detected by real-time qPCR (C-D) and western blotting (E), along with GAPDH as a control in qPCR and α-tubulin as a control in western blotting. (F) Knockout of ZFHX3 in C4-2B cells reduced the KLK3 protein level in both cells and culture supernatants. Culture supernatants were collected after centrifugation at 1500 rpm for 10 min. The blank group corresponds to a complete medium not exposed to cells. (G) C4-2B cells were co-transfected with ZFHX3 siRNA and HA-tagged ZFHX3 (HA-ZFHX3) plasmid. Forty-eight hours after transfection, the cells were harvested and analyzed using western blotting and real-time qPCR. (H) Knockout of ZFHX3 in C4-2B cells decreased the binding of AR to enhancer or promoter regions of KLK3, FKBP5, and TMPRSS2 as detected by ChIP-PCR, along with the TATA box as a negative control in PCR. The original blots and gel images can be found in Fig. S1 (Supplementary Information). ns, not significant; **, P < 0.01; ***, P < 0.001.

Full size image

Cell proliferation assay

Following previously published procedures25C4-2B clones and LNCaP cells 24 h after siRNA transfection were seeded into 24-well tissue culture plates at 1.2 × 104 cells and 2 × 104 cells per well, respectively. After 2 days, cells were treated with enzalutamide and collected every two days. The cells were fixed in 10% TCA (Sigma), stained with 0.4% SRB (Sigma), and washed with 1% acetic acid. Absorbance was recorded with BioTek Synergy HTX Multimode Reader (Agilent, Santa Clara, CA) at 490 nm. Experiments were performed in triplicate.

Soft agar colony formation assay

C4-2B and LNCaP cells were cultured in RPMI-1640 medium with 10% FBS. 5 × 103 C4-2B clones or 1 × 104 LNCaP cells 24 h after siRNA transfection were suspended in 0.35% agar with or without enzalutamide, and laid on top of 1.5 mL of RPMI-1640 solidified with 0.6% agar in each well of a 6-well tissue culture plate. After incubation at 37 °C in a CO2 incubator for 2 weeks (for C4-2B) or 3 weeks (for LNCaP), colonies with a diameter > 70 μm were imaged and counted with the ImageJ program. Experiments were performed in triplicate.

RNA sequencing and bioinformatic analyses

RNA extraction from Wt and KO8 cells and next-generation sequencing of mRNA-derived cDNA libraries using the PE100 strategy on the DNBSEQ platform were performed at the Beijing Genomics Institute (Wuhan, China). Clean reads were aligned to the human GRCh38 genome. DESeq2 was applied to identify the differentially expressed genes between Wt and KO8 cells, and Q values < 0.05 were considered statistically significant. Based on log2 fold changes of the transcripts per million (TPM) values, gene set enrichment analysis (GSEA) was performed using the OmicStudio tools at https://www.omicstudio.cn/tool26and gene sets with an adjusted P value < 0.05 were considered significantly enriched. A heatmap was generated using the OmicStudio tools based on the TPM values.

Survival and statistical analyses

Two prostate cancer datasets, TCGA (PanCancer Atlas)27 and MSKCC (MSK Cancer Cell, 2010)28were downloaded from cBioPortal (https://www.cbioportal.org/ accessed on 28 June 2023)29,30 and respectively contained 333 and 140 prostate cancers that had both mRNA expression data and disease-free survival status. Survival analyses were prepared by using the Kaplan–Meier method and the log-rank test.

All in vitro experiments were repeated at least twice. Statistical analysis was based on 3 replicates of one experiment. All experimental readings are expressed as mean ± SD. A two-tailed Student’s t-test was performed for statistical comparison of groups. All statistical analyses were conducted using GraphPad Prism 6. P values < 0.05 were considered statistically significant.

Results

ZFHX3 loss compromises ar’s transcriptional activities by reducing its promoter binding

We first carried out RNA-seq analysis in C4-2B prostate cancer cells to detect which genes and signaling pathways are caused by the deletion of ZFHX3. Gene set enrichment analysis (GSEA) revealed that among the top altered hallmark gene sets after ZFHX3 deletion, only the one for androgen response was decreased, whereas those for TNFα, interferon γ, and inflammatory response were enriched (Fig. 1A). For the 27 genes indicative AR activities, as defined in a previous study31expression levels for 18 of the 27 genes were significantly changed by ZFHX3 loss (Fig. 1B). Fifteen of these 18 genes were downregulated, including classical AR target genes KLK3, FKBP5, and TMPRSS2 (Fig. 1B).

In AR-positive and androgen-responsive prostate cancer cells, real-time qPCR confirmed that ZFHX3 deletion in C4-2B cells or knockdown in LNCaP cells significantly reduced expression levels of KLK3, FKBP5, and TMPRSS2 (Fig. 1C and D). Western blotting also confirmed that ZFHX3 loss reduced the protein expression of KLK3 (also known as PSA) (Fig. 1E). ZFHX3 did not affect AR expression in these cell lines, though (Fig. 1C-E). Secreted KLK3 protein was also decreased in the culture medium of ZFHX3-null C4-2B cells (Fig. 1F). In C4-2B cells with RNAi-mediated ZFHX3 silencing, KLK3 expression was reduced, and ectopic expression of ZFHX3 rescued the effect to certain extent, as detected by western blotting and qPCR for protein and mRNA, respectively (Fig. 1G).

We also carried out ChIP-PCR to evaluate whether ZFHX3 affects AR’s binding to the promoters of its target genes. In AR antibody-precipitated DNA, the ARE-containing DNA elements were detected for the KLK3 enhancer, KLK3 promoter, FKBP5 promoter, and TMPRSS2 enhancer (Fig. 1H). Knockout of ZFHX3 reduced but did not eliminate AR’s binding to these promoters/enhancers (Fig. 1H). AR did not bind to the TATA box in KLK3’s promoter, which is consistent with a previous finding4. These findings indicate that ZFHX3 is required for AR-mediated gene transcription.

Loss of ZFHX3 compromises but does not eliminate AR activity in KLK3 transcription

The gene encoding kallikrein-related peptidase 3 (KLK3), more commonly known as prostate-specific antigen (PSA), is a classical downstream gene of the androgen/AR signaling pathway. To further define the role of ZFHX3 in androgen/AR signaling, we tested the effect of ZFHX3 loss on AR activity in KLK3 transcription. The synthetic androgen R1881 was used to activate AR in C4-2B and LNCaP cells in which ZFHX3 was knocked out or down. Whereas R1881 treatment at various concentrations significantly increased the expression of KLK3 at both mRNA and protein levels (Fig. 2A), loss of or decreased ZFHX3 expression reduced KLK3 expression at each of the R1881 concentrations (Fig. 2A).

Fig. 2
figure 2

Loss of ZFHX3 compromises but does not eliminate AR’s activity in KLK3 transcription. (A-B) Detection of KLK3 protein and KLK3 mRNA by western blotting and real-time qPCR, respectively, in C4-2B and LNCaP cells cultured in the hormone-free medium for 24 h and then treated with R1881 at indicated concentrations for 24 h (A) or cultured in complete medium for 48 h (B). The transfection of siRNAs was done for 48 h before cell harvesting. (C-E) Expression plasmids of pGL3 vector control (pGL3-basic), pKLK3-Luc, and the pRL-TK reporter were transfected into C4-2B cells incubated in phenol red-free medium containing 5% charcoal-stripped FBS for 24 h (C) or cultured in complete medium (D-E). After 24 h, cells were treated with R1881 (C) or enzalutamide (D) at the indicated concentrations for 24 h. The transfection of siRNAs was done for 42 h before cell harvesting. Relative luciferase activities were then determined. The original blots can be found in Fig. S2 (Supplementary Information). ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Full size image

When AR was knocked down in C4-2B and LNCaP cells cultured in a complete medium, KLK3 expression was dramatically reduced as expected (Fig. 2B). Meanwhile, AR knockdown and ZFHX3 knockdown/knockout demonstrated an additive effect on lowering KLK3 expression (Fig. 2B). Interestingly, AR knockdown increased the protein level of ZFHX3 (Fig. 2B).

We also analyzed the promoter activity of the ARE-rich promoter regions of KLK3 using the pKLK3-Luc expression construct in C4-2B cells, with or without the R1881, enzalutamide, or AR silencing. While R1881 treatment significantly increased and enzalutamide or AR silencing significantly decreased the luciferase activity, ZFHX3 loss further reduced the luciferase activity (Fig. 2C-E). These findings indicate that loss of ZFHX3 compromises but does not eliminate AR activity in KLK3 transcription.

ZFHX3 physically associates with AR via specific regions in human epithelial cells

Considering that both ZFHX3 and AR are transcription factors, it is likely that ZFHX3 and AR physically interact with each other to regulate gene transcription. To test this prediction, we transfected HA-tagged ZFHX3 (HA-ZFHX3) with pSG5-AR into 293 T cells and performed IP with an anti-HA antibody. Western blotting detected AR in the ZFHX3 precipitate (Fig. 3A). We also transfected HA-ZFHX3 into C4-2B cells, which express AR endogenously. In the HA-ZFHX3 complexes, endogenous AR was also detected (Fig. 3B). Treatment of C4-2B cells with the enzalutamide androgen antagonist reduced ZFHX3-bound AR (Fig. 3B). These findings indicate that ZFHX3 and AR physically associate with each other, and the association is influenced by AR status.

Fig. 3
figure 3

ZFHX3 physically associates with AR in human epithelial cells. (A) 293 T cells were transiently transfected with expression plasmids of vector control or HA-ZFHX3 and pSG5-AR and then subjected to co-IP and western blotting with the indicated antibodies. (B) C4-2B cells transfected with HA-ZFHX3 for 42 h were treated with 10 µM enzalutamide for 6 h and then subjected to co-IP and western blotting with the indicated antibodies. (C) Schematic of full-length ZFHX3 protein (3703 residues, horizontal bar) with 23 zinc fingers (ovals) and 4 homeodomains (black rectangles). Dark and light bars indicate positive and negative interactions with AR, respectively, with the name of each mutant and the residues spanned shown to the right of each bar. Blue vertical lines mark four LXXXL variants of the LXXLL motif in mutant A. (D-G) Co-IP and western blotting results for the interaction of pSG5-AR and each of the HA-tagged ZFHX3 mutants in 293 T cells (D, F-G) or the interaction of endogenous AR mutants and HA-tagged ZFHX3-A (residues 1–800) in 22Rv1 (E) cells. Arrows indicate HA-tagged ZFHX3 mutants. The original blots can be found in Fig. S3 (Supplementary Information).

Full size image

To map the ZFHX3 domains interacting with AR, we constructed 6 HA-tagged overlapping fragments of ZFHX3 (Fig. 3C) and expressed each with pSG5-AR in 293 T cells. IP and western blotting with HA antibody demonstrated that fragments A, B, C, D, and F of ZFHX3 interacted with AR, and fragment A pulled down more AR compared to other fragments (Fig. 3D), suggesting a higher affinity in the interaction of fragment A with AR. When fragment A was expressed in 22Rv1 cells, in which a full-length AR with an additional zinc finger in the DNA-binding ___domain is expressed along with several AR splice variants lacking the ligand-binding ___domain (AR-Vs, e.g., AR-V7)32,33fragment A’s interaction was detected for both the full-length AR and the AR variants (Fig. 3E).

We then evaluated the peptide sequence of fragment A for the presence of the LXXLL motif and its variant LXXXL, which could mediate a protein’s interaction with the ligand-binding domains of nuclear receptors, including AR34,35. Four LXXXL but no LXXLL motifs were identified. We also carried out deletion mutants of fragment A of ZFHX3 to create smaller fragments and determined the smallest region of fragment A that interacts with AR. Only the region spanning residues 1–223 in fragment A interacted with AR, although this region does not contain any LXXLL or LXXXL motif (Fig. 3F and G). These findings support the ZFHX3-AR interaction and define the regions of ZFHX3 involved in the interaction.

ZFHX3 loss also attenuates enzalutamide’s inhibitory effect on cell proliferation

Considering that the loss of ZFHX3 compromises AR’s activity in KLK3 transcription, it is reasonable to propose that AR function is influenced by ZFHX3 status, including AR’s oncogenic activity. In testing this hypothesis, we treated C4-2B and LNCaP cells with enzalutamide at 10 µM and found that enzalutamide significantly inhibited cell proliferation and decreased the number and average size of colonies regardless of ZFHX3 levels (Fig. 4A-D), knockout or knockdown of ZFHX3 in C4-2B or LNCaP cells promoted cell proliferation, increased the number of colonies, but not changed the average size of colonies (Fig. 4A-D). When the AR activity was inhibited by enzalutamide treatment, ZFHX3 loss still promoted cell proliferation (Fig. 4A-D). These findings suggest that ZFHX3 loss attenuates enzalutamide’s inhibitory effect on AR activity.

Fig. 4
figure 4

The role of ZFHX3 deficiency for AR signaling in prostate cancer cells and patients. (A-D) C4-2B (A, C) and LNCaP (B, D) cells were used for both cell proliferation assay (A, B) and colony formation assay (C, D). Enzalutamide was added to inhibit the AR activity. (E-F) Kaplan-Meier analysis of disease-free survival of prostate cancer patients in TCGA (E) and MSKCC (F) datasets with different ZFHX3 and AR expression statuses. (G) Detection of AR and ZFHX3 in the nucleus and the cytoplasm by western blotting in C4-2B cells cultured in a complete medium. Lamin B1 and α-tubulin serve as nuclear and cytoplasmic markers, respectively. The original blots can be found in Fig. S4 (Supplementary Information). ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Full size image

ZFHX3 loss appears to make AR more oncogenic in prostate cancer

We evaluated the relationship between ZFHX3 loss and AR activity in human prostate cancer. In two prostate cancer datasets where both mRNA expression data and disease-free survival (DFS) status are available, we stratified patients into four groups according to their median levels of ZFHX3 and AR: ZFHX3 high/AR high, ZFHX3 high/AR low, ZFHX3 low/AR high, and ZFHX3 low/AR low. Interestingly, higher AR levels significantly correlated with worse DFS only in patients with lower ZFHX3 levels but not in those with higher ZFHX3 levels (Fig. 4E and F). These findings suggest that ZFHX3 loss makes AR more oncogenic in prostate cancer.

Western blotting demonstrated that ZFHX3 loss in C4-2B cells did not alter AR’s cellular localization, as ZFHX3 protein was primarily detected in the nucleus, while AR protein was detected in both the cytoplasm and the nucleus before and after ZFHX3 knockout (Fig. 4G).

Discussion and conclusions

Both AR and ZFHX3 are transcription factors that modulate prostatic development and carcinogenesis, as the androgen/AR signaling is essential for these processes, and loss of ZFHX3 function causes prostatic intraepithelial neoplasia in mouse prostates4,17. Whether and how ZFHX3 and AR functionally relate to each other is thus an important question to address. Previous studies demonstrated that in AR-positive LNCaP and C4-2B prostate cancer cells, androgen induces the ZFHX3 transcription via the binding of AR to the ZFHX3 promoter15and the enzalutamide antiandrogen prevents this induction15. As discussed below, the current study demonstrated that the induction of ZFHX3 transcription is necessary for AR to function properly.

First, ZFHX3 is crucial for AR to activate or repress the transcription of its target genes. In the AR-positive C4-2B prostate cancer cells, deletion of ZFHX3 significantly altered the expression of 18 of the 27 AR target genes indicative of AR activities31including classical AR target genes KLK3, FKBP5, and TMPRSS2 (Fig. 1). Consistently, ZFHX3 loss reduced but did not eliminate AR’s binding to the promoters/enhancers of KLK3, FKBP5, and TMPRSS2 (Fig. 1). In addition, AR knockdown and ZFHX3 loss showed an additive effect on the reduction of KLK3 expression (Fig. 2B).

AR regulates KLK3 transcription by binding to the AREs in KLK3’s proximal promoter and enhancer36and the KLK3 enhancer is androgen-responsive and prostate-specific37. Many transcription factors have been identified for their direct interactions with AR and their binding to the KLK3 promoter38,39including c-Jun40,41,42EGR143, and Sp144. Using the AR-mediated KLK3 promoter activity assay, we further found that ZFHX3 loss compromised AR’s transcriptional activities by reducing its binding to the KLK3 promoter (Fig. 1). However, ZFHX3 loss did not eliminate such an activity (Fig. 2). In addition, loss of ZFHX3 still reduced the expression and luciferase activity of KLK3 when the AR signaling was blocked (Fig. 2), suggesting that ZFHX3 is not a typical cofactor for AR’s transcriptional activities.

ZFHX3 physically associates with AR via the region spanning residues 1–223, which could underlie ZFHX3-modulated AR activity. AR is composed of an N-terminal transactivation ___domain, a DNA-binding ___domain, a hinge region, and a conserved C-terminal LBD45and the LBD often mediates AR’s interactions with other cofactors via motifs like LXXLL and LXXXL34,35. Immunoprecipitation and western blotting demonstrated that ZFHX3 physically interacted with AR. The ZFHX3-AR interaction could be attenuated by the enzalutamide treatment (Fig. 3B). Whereas multiple regions of ZFHX3 could be involved in the ZFHX3-AR interaction, fragment A of ZFHX3 showed a higher binding affinity than others, particularly the fragment spanning residues 1–223 (Fig. 3). It should be noted that exogenous expression may elevate protein levels beyond physiological ranges, potentially inducing artificial binding or non-specific interactions. Future validation in endogenous systems (e.g., CRISPR-based tagging or proximity assays) is warranted.

The ZFHX3-AR interaction does not involve AR’s LBD. The ZFHX3-AR binding occurred not only with the full-length AR protein but also with LBD-less AR splicing variants such as the AR-V7 (Fig. 3). In addition, the smallest region of ZFHX3 mediating a more potent ZFHX3-AR interaction (residues 1–223) does not contain any LXXXL or LXXLL motifs, which mediate a protein’s interaction with AR’s LBD [34, 35]. The reduction of ZFHX3-AR interaction by the enzalutamide treatment (Fig. 3B) is likely due to reduced AR protein in the nucleus because enzalutamide is a potent AR antagonist that binds to the LBD of AR to prevent its nuclear translocation9. The ZFHX3-AR association does not alter AR’s cellular localization, nor does it change AR’s expression levels (Figs. 1, 2 and 4G).

The exact mechanisms for how ZFHX3 modulates AR function remain to be clarified. We noticed that ZFHX3 loss compromised but did not eliminate AR’s function, which suggests that the AR-ZFHX3 partnership is different from a classic transcriptional complex, in which one member’s loss usually removes the function of the entire transcription complex. ZFHX3 has 23 zinc fingers and numerous other domains, and ZFHX3 likely provides a scaffold for AR and other members of its transcriptional partners to act.

Whereas ZFHX3 is crucial for AR to modulate gene transcription, the ZFHX3-AR partnership plays a more complicated role in cell proliferation and tumorigenesis. The androgen/AR signaling is a driving force in prostatic carcinogenesis, and increased AR expression plays a significant role in prostate cancer progression4. However, the ZFHX3-AR association appears to inhibit AR-mediated cell proliferation. Previous studies have demonstrated that ZFHX3 is frequently inactivated in prostate cancer, and its loss not only promotes AR-mediated cell proliferation in culture but also causes neoplastic lesions in mouse prostates13,17,19. Consistently, the current study further demonstrated that increased AR expression was associated with worse disease-free survival only in prostate cancers with reduced ZFHX3 expression but not in those with higher ZFHX3 expression levels (Fig. 4E and F). Therefore, the ZFHX3-AR association could suppress AR’s oncogenic function, even though ZFHX3 is crucial for AR to regulate gene transcription.

On the other hand, it is well-established that AR also has a tumor suppressor activity in prostate cancer, which can involve the recruitment of the RB1 tumor suppressor46,47,48,49,50. Some AR coactivators, such as TBL1X/Y related 1, are associated with AR-mediated tumor suppression51. The ZFHX3-AR partnership could thus be crucial for AR’s tumor suppressor activity. ZFHX3 is likely necessary for AR to balance its oncogenic and tumor-suppressive functions by maintaining the transcription of a specific set of genes. When ZFHX3 is inactivated by genetic deletion or mutation, the genes responsible for AR’s tumor suppressor activity are dysregulated, and AR’s oncogenic function becomes dominant. This hypothesis remains to be tested. Tumor suppressors Rb52 and BRCA153 have been shown to interact with AR to affect AR transcriptional activity and exert their tumor suppressor roles, and ZFHX3 could act similarly.

ZFHX3 loss also attenuates the inhibitory effect of the enzalutamide antiandrogen on cell proliferation and colony formation in prostate cancer cells (Fig. 4A-D), which could have therapeutic implications. Enzalutamide is a potent AR inhibitor and an effective therapeutic agent in the ADT of prostate cancer. Prostate cancers initially respond to ADT well, but most of them develop resistance to ADT after a period of treatment, and metastatic castration-resistant prostate cancer represents a significant clinical problem. Loss of ZFHX3 significantly attenuated enzalutamide’s inhibitory effect on proliferation and colony formation, and loss of ZFHX3 still promoted cell proliferation even in the presence of enzalutamide (Fig. 4A-D). It is thus likely that the inactivation of ZFHX3 by chromosomal deletion or mutation, which frequently occur in advanced prostate cancer13,14could make prostate cancer cells less responsive to ADT and thus contribute to CRPC development. On the other hand, prostate cancer cells lacking ZFHX3 could be useful for developing therapeutic strategies against CRPC.

The ZFHX3 status could impact the application of the PSA test in prostate cancer patients. The test of blood PSA levels is used to screen for prostate cancer in men, but PSA levels vary significantly among men in this regard. For example, PSA levels of > 4.0 ng/mL were once widely claimed to have predictive value for prostate cancer diagnosis54but about 15% of men at 62–91 years with PSA levels ≤ 4 ng/mL were diagnosed with prostate cancer within 7 years55. PSA levels are also used to indicate the recurrence of prostate cancer27,28. Loss of ZFHX3 decreases KLK3 expression at both mRNA and protein levels in prostate cancer cells (Fig. 1), and ZFHX3 is frequently inactivated by mutations in advanced prostate cancer13,14which implies that patients with ZFHX3 loss could have lower PSA levels but advanced tumors and PSA screening is thus not suitable for this group of men.

In summary, we found that ZFHX3 is integral to the androgen/AR signaling activity in prostate cancer cells, which involves ZFHX3 binding to AR via multiple domains. Loss of ZFHX3 compromises but does not eliminate AR’s binding to the promoters of AR target genes in their transcription. Loss of ZFHX3 also attenuates the inhibitory effect of enzalutamide, an AR antagonist, on cell proliferation and colony formation. Furthermore, a correlation between higher AR expression levels and worse patient survival is only detectable in prostate cancers with reduced ZFHX3 expression. These findings not only provide novel insights into the functional relationship between ZFHX3 and the androgen/AR signaling, but they also implicate ZFHX3 loss in the diagnosis and ADT treatment of prostate cancer.