AR imposes different effects on ZFHX3 transcription depending on androgen status in prostate cancer cells

Abstract Both androgen receptor (AR) and the ZFHX3 transcription factor modulate prostate development. While AR drives prostatic carcinogenesis, ZFHX3 is a tumour suppressor whose loss activates the PI3K/AKT signalling in advanced prostate cancer (PCa). However, it is unknown whether ZFHX3 and AR are functionally related in PCa cells and, if so, how. Here, we report that in AR‐positive LNCaP and C4‐2B PCa cells, androgen upregulates ZFHX3 transcription via androgen‐induced AR binding to the androgen‐responsive elements (AREs) of the ZFHX3 promoter. Androgen also upregulated ZFHX3 transcription in vivo, as castration dramatically reduced Zfhx3 mRNA and protein levels in mouse prostates, and ZFHX3 mRNA levels correlated with AR activities in human PCa. Interestingly, the binding of AR to one ARE occurred in the absence of androgen, and the binding repressed ZFHX3 transcription as this repressive binding was interrupted by androgen treatment. The enzalutamide antiandrogen prevented androgen from inducing ZFHX3 transcription and caused excess ZFHX3 protein degradation. In human PCa, ZFHX3 was downregulated and the downregulation correlated with worse patient survival. These findings establish a regulatory relationship between AR and ZFHX3, suggest a role of ZFHX3 in AR function and implicate ZFHX3 loss in the antiandrogen therapies of PCa.

metastatic or high-grade human PCa, and the mutations are mainly loss-of-function mutations. 11,12 Tissue-specific deletion of Zfhx3 induces mouse prostatic intraepithelial neoplasia (mPIN) and promotes prostatic tumorigenesis initiated by Pten deletion. 13,14 In cultured human prostate cancer cells, ZFHX3 is indispensable for oestrogen receptor beta (ERβ) to inhibit cell proliferation via MYC downregulation. 15 These findings indicate that ZFHX3 plays a role in androgen-driven prostatic development and that ZFHX3 is a tumour suppressor in PCa.
In PCa cells, activation of ERβ by diarylpropionitrile (DPN) upregulates ZFHX3, and ZFHX3 is crucial for ERβ to inhibit cell proliferation. 15 In mammary epithelial cells, activation of ERα by oestrogen or PR by progesterone also upregulates ZFHX3 expression, and ZFHX3 is also involved in the functions of ER and PR. [16][17][18] The upregulation is primarily at the transcriptional level. AR, ER and PR belong to the steroid receptor family and have structural similarities; AR and PR are even more similar than ER. 19 In addition, androgens, oestrogens and progesterones share some synthetic routes. 20 We, therefore, hypothesize that androgen/AR signalling also regulates ZFHX3 expression to regulate prostatic development and tumorigenesis.
This study examined the regulatory relationship between androgen/AR signalling and ZFHX3 in PCa cells. We demonstrated that the androgen/AR signalling upregulates the transcription of ZFHX3 via the binding of AR to AREs of ZFHX3 promoter in PCa cells, which appeared to occur in both mice and humans. Interestingly, binding to AR to another ARE of ZFHX3 promoter occurred in the presence of androgen, and the binding repressed ZFHX3 promoter activity.
Furthermore, an inhibition of AR signalling by an AR antagonist pre- For gene silencing by RNA interference (RNAi), cells were transiently transfected with siRNAs using the Lipofectamine RNAiMAX reagent (Invitrogen). The siRNA sequence against human AR (siAR) was 5′-CAAGGGAGGUUACACCAAA-3′, validated in a previous study. 21

| Western blotting
Cultured cells were lysed in the cell lysis buffer for Western and IP from Beyotime. Cell lysates were subjected to 4% (for ZFHX3) or 10% (for all other proteins) sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE), and proteins were transferred to polyvinylidene fluoride (PVDF) membranes according to a standard protocol. The membranes were blocked with 5% nonfat milk for 30 min at room temperature and probed with primary antibodies overnight at 4°C. On the next day, the membranes were incubated with the secondary antibodies for 1 h at room temperature; and the WesternBright ECL HRP substrate (Advansta) was used with the Luminescent Image Analyzer (JUNYI, Beijing, China) to capture images.
Antibodies used in Western blotting included ZFHX3 (1:800, prepared in our previous study 22

| RNA extraction and real-time qPCR
Total RNA was extracted using the Eastep Super Total RNA Extraction Kit (Promega), and cDNA was synthesized with the Moloney murine leukaemia virus reverse transcriptase (Promega).

| Luciferase reporter assay
C4-2B cells were transiently transfected with indicated ZFHX3 plasmids and pRL-TK Renilla luciferase plasmid (Promega) using the Lipofectamine 2000 reagent (Invitrogen). Forty-eight hours after transfection, including 24-h R1881 treatment before cell collection, luciferase activities were determined using the Dual-Luciferase Reporter Gene Assay Kit (Promega). The luciferase activity was normalized by the Renilla luciferase activity in each sample.
Experiments were performed in quadruplication and repeated twice.

| Chromatin immunoprecipitation (ChIP) assay
C4-2B cells were grown in a hormone-free medium for 1 day

| CHX assay
C4-2B cells were seeded into 12-well plates at a density of 2 × 10 5 cells per well. After 24 h, cells were treated with 100 μg/ml of cycloheximide (CHX) with and without 10 μM enzalutamide treatment. At 0, 2, 4 and 6 h after CHX treatment, total proteins were collected, separated by SDS-PAGE and subjected to Western blotting for ZFHX3 and α-tubulin. Band intensities were quantified using the ImageJ program (NIH, Bethesda, MD). ZFHX3 protein level was normalized by α-tubulin level for each time point CHX treatment.

| Detection of Zfhx3 expression in mouse prostates after castration
Mouse experiments were approved by the Institutional Animal Care and Use Committee at the Southern University of Science and Technology. Twelve-week-old male C57BL/6 mice were purchased from Charles River, and castration was performed as described in our previous study. 23 One and 14 days after castration, mice were euthanized, and prostates were surgically isolated as previously  Image-Pro Plus (NIH) was used to determine the integral optical density (IOD) of the positively stained signal (brown) and area of haematoxylin-stained signal (blue) for each slide. The IOD/mm 2 of the haematoxylin-stained area was then calculated and compared between groups.

| Bioinformatic and statistical analyses
The TCGA prostate adenocarcinoma dataset was downloaded from the UCSC Xena public data hub (http://xenab rowser.net/). A heatmap was built using the R package, in which the PCa samples were divided into two groups (i.e., ZFHX3 higher and ZFHX3 lower) by the median ZFHX3 expression level. AR activities were defined by the sum of expression levels of 27 genes indicative of AR activities, as established in a previous study. 24 The expression level of a gene was indicated by the Z-score of log 2 (RSEM normalized count +1) relative to that in all PCa samples.
The Pearson correlation analysis was conducted to determine the linear correlation between ZFHX3 mRNA levels and AR activity scores using GraphPad Prism 6 (GraphPad software).
The differences between rates of ZFHX3 deletion at one and both copies in PCa samples with different Gleason scores were tested using Fisher's exact test. The disease-free survival was analysed using the Kaplan-Meier analysis, and the statistical parameters were calculated using a log-rank test.
All in vitro experiments were repeated at least twice. Statistical analysis was based on 3 or 4 replicates of one experiment. All experimental readings are expressed as mean ± SD. For statistical comparison between two groups, a two-tailed Student's t-test was performed.
For statistical comparison of more than two groups, one-way ANOVA or two-way ANOVA was performed. All statistical analyses were conducted using GraphPad Prism 6. p values <0.05 were considered statistically significant.

| Androgen-AR signalling increases ZFHX3 expression at both protein and mRNA levels
To test whether androgen modulates ZFHX3 expression, we de- To further evaluate the role of the androgen/AR signalling in ZFHX3 expression, we knocked down the AR by siRNA in C4-2B cells treated with varying concentrations of R1881 for 24 h in a hormonefree medium and analysed ZFHX3 expression. AR silencing, which was confirmed by western blotting, eliminated the induction of ZFHX3 by R1881 at both protein and mRNA levels ( Figure 1E).
Interestingly, AR knockdown increased the mRNA level of ZFHX3 in the absence of R1881 ( Figure 1E).
We also applied enzalutamide, an AR antagonist that binds to AR to block its nuclear translocation and transactivation activity, 26,27 to C4-2B cells cultured in R1881-containing complete medium and analysed ZFHX3 expression ( Figure 1F). Enzalutamide significantly reduced R1881-induced ZFHX3 expression at both protein and mRNA levels ( Figure 1F). As a control, the mRNA level of KLK3 was increased by R1881 treatment and decreased by enzalutamide treatments ( Figure 1F).
These results suggest that androgen induces ZFHX3 expression via the binding of AR to ZFHX3 promoter and subsequent transcriptional activation.

| Androgen-induced ZFHX3 expression depends on the binding of AR to the ZFHX3 promoter
To determine whether androgen-induced ZFHX3 expression is indeed mediated by the binding of AR to the ZFHX3 promoter, we analysed the promoter sequence of ZFHX3 from −2954 to +597 relative to its transcriptional initiation site by the online software JASPAR CORE (http://jaspar.gener eg.net/), in which the consensus AR binding sequences are based on ChIP-Seq studies. 28 A total of six potential AR binding sites were identified based on the binding scores, which were named ARE1-ARE6 ( Figure 2A).
Then, the pZFHX3-Luc promoter-luciferase reporter plasmid for ZFHX3, which contains the promoter sequence of ZFHX3 in  unchanged, as the AR-bound P2 fragment was increased by R1881, P8 and P10 fragments were decreased, and other fragments were not affected ( Figure 2D). This finding confirms that different AREs in the ZFHX3 promoter have different responses to androgen-induced AR binding and suggests that ARE1 is likely responsible for R1881induced ZFHX3 upregulation, but ARE6 likely has an opposite effect.
The binding of AR to P8 and P10 fragments of the ZFHX3 promoter without androgen indicates that some AR molecules exist in the nucleus even without androgen. To test whether this is true, we grew LNCaP and C4-2B cells in hormone-free medium or complete medium for 24 and 72 h, extracted cytoplasmic and nuclear proteins and analysed the expression of AR and ZFHX3 by Western blotting. In the absence of androgen, AR protein was indeed detected in both the cytoplasm and the nucleus. In contrast, ZFHX3 protein was primarily detected in the nucleus ( Figure 2E,F), although AR protein was increased in the cytoplasm and decreased in the nucleus ( Figure 2E,F). Therefore, even without androgen, some AR proteins still exist in the nucleus in PCa cells.

| AR-bound ARE1 and ARE6 have opposing effects on ZFHX3 transcription
We further characterized ARE1 and ARE6 for their opposite effects on R1881-induced ZFHX3 transcription. In the pZFHX3-Luc reporter plasmid, ARE1, ARE6, or both ARE1 and ARE6 were mutated at the consensus ARE nucleotides, and the mutant reporter plasmids were then analysed for promoter activities ( Figure 2G).
Mutation of ARE1 not only decreased ZFHX3 promoter activity but also prevented the induction of promoter activity by R1881 ( Figure 2H). It appeared that R1881 further decreased the promoter activity of pZFHX3-Luc-mARE1. On the other hand, mutation of ARE6 dramatically increased ZFHX3 promoter activity while maintaining the promoter's response to R1881 ( Figure 2H). When both ARE1 and ARE6 were mutated, the promoter's activities were somewhat similar to that of the wildtype, maintaining a modest promoter activity and androgen responses ( Figure 2H). The results further indicate that AR-bound ARE1 and ARE6 have opposite effects on ZFHX3 transcription, with ARE1 increasing but ARE6 decreasing ZFHX3 transcription in AR-positive PCa cells.

| Inhibition of AR activity by enzalutamide causes ZFHX3 protein degradation without apparent change in ZFHX3 mRNA level
To further test whether enzalutamide decreases ZFHX3 protein level without significantly affecting ZFHX3 mRNA expression, as suggested by the result in Figure 1F, we treated LNCaP and C4-2B To test whether enzalutamide decreases ZFHX3 protein level via protein degradation, we measured the half-life of ZFHX3 protein in C4-2B cells treated with cycloheximide (CHX) for different times.
The CHX assay demonstrated that enzalutamide significantly decreased the half-life of ZFHX3 protein ( Figure 3E), suggesting that enzalutamide causes ZFHX3 protein degradation.

| Androgen ablation in mice decreases Zfhx3 expression in mouse prostates
To determine whether the regulation of ZFHX3 expression by the androgen-AR signalling also occurs in vivo, we castrated 12-week-old Real-time qPCR demonstrated that at both d1 and d14 after castration, Zfhx3 mRNA level decreased, and Ar mRNA level was increased, which is considered negative autoregulation. 29,30 As expected, the mRNA levels of known AR target genes Fkbp5 and Tmprss2 were decreased by castration ( Figure 4C). These results further indicate that the androgen-AR signalling induces the expression of ZFHX3 in prostate cells.

| ZFHX3 mRNA levels positively correlate with AR activities in human PCa specimens
To evaluate whether transcriptional regulation of ZFHX3 by the androgen-AR signalling also occurs in human prostate cells, we collected RNA-sequencing data of PCa samples from the TCGA database. We then analysed ZFHX3 expression levels and AR activities defined by the sum of expression levels of 27 genes that showed robust activation or inhibition of expression upon androgen stimulation. 24 The 498 PCa samples with mRNA data were stratified into two groups according to the median level of ZFHX3 expression: ZFHX3-high and ZFHX3-low. The heatmap demonstrated that tumours with a higher ZFHX3 level also had a higher AR activity ( Figure 4D, left). Pearson correlation analysis further confirmed the positive correlation between ZFHX3 expression levels and AR activity scores regardless of tumour stage, grade, metastasis or age ( Figure 4D, right; data not shown). Therefore, induction of ZFHX3 transcription by the androgen-AR signalling also occurs in human prostatic epithelial cells.

| Alterations of ZFHX3 in human PCa and their potential impacts
Using the TCGA PCa dataset, we evaluated the status of ZFHX3 and its potential clinical relevance in PCa. We found that deletion of one or both copies of ZFHX3 was more frequent in PCa samples with higher Gleason scores ( Figure 5A). Interestingly, deletion of ZFHX3 significantly correlated with reduced AR activities defined by the sum of 27 genes' expression levels that show robust expression activation or inhibition upon androgen stimulation 24 (Figure 5B), supporting a role of ZFHX3 in AR function.
The ZFHX3 mRNA levels, as indicated by the Z-scores of log 2 (RSEM normalized count +1) relative to that of all PCa samples, were lower in PCa samples, particularly those with Gleason scores greater than 7 ( Figure 5C). We also tested the potential relationship between ZFHX3 expression and disease-free survival in human PCa.
Kaplan-Meier curve with log-rank analysis demonstrated that patients with lower ZFHX3 expression levels had poorer disease-free survival ( Figure 5D), implicating that decreasing ZFHX3 expression could aggravate the progression of PCa.

C4-2B
To explore the biological functions of ZFHX3, we divided the 498 PCa samples with mRNA data into two groups, ZFHX3-high and ZFHX3-low, according to the median ZFHX3 mRNA level, as indicated by the Z-score of log 2 (RSEM normalized count +1). The gene set enrichment analysis (GSEA) was then performed using the preranked list of genes based on fold changes. The top five enriched hallmarks included androgen response genes, genes downregulated in response to UV radiation, protein secretion, mitotic spindle and early response to oestrogen. Those enriched in the ZFHX3-low group included interferon-gamma response, allograft rejection, E2F targets, MYC-regulated genes, DNA repair, interferon-alpha response, and oxidative phosphorylation ( Figure 5E). The hallmark for androgen response included 100 genes defining response to androgens. Among the ZFHX3-high group, the androgen-responsive genes had the highest normalized enrichment score (NES), further supporting a role of ZFHX3 in AR function ( Figure 5E).

| DISCUSS ION
Our findings in this study establish a regulatory relationship between the androgen/AR signalling and the ZFHX3 transcription In addition, downregulation of ZFHX3 was correlated with worse disease-free survival in patients with PCa ( Figure 5D), further supporting the role of ZFHX3 in human PCa. Transcriptional activation of ZFHX3 by the androgen/AR signalling thus suggests that the proper function of AR depends on the existence of ZFHX3 in prostate epithelial cells. The role of ZFHX3 in AR function is further supported by the correlation between ZFHX3 deletion and reduced AR activities ( Figure 5B) and the enrichment of androgenresponsive genes in PCa samples expressing higher levels of ZFHX3 ( Figure 5E). Whether and how AR and ZFHX3 functionally cooperate is a meaningful question that has not been addressed.
ZFHX3 could be involved in the balance between cell proliferation and differentiation induced by androgen in normal AR-positive epithelial cells.
Autoregulation of AR mRNA and protein is an evolutionary conserved regulatory mechanism. 33 While the autoregulation of AR appears to be negative in most AR-expressing rat tissues and human cell lines, 29,30,34 some human cells appear to have a positive AR autoregulation, including PC-3 and DU 145 PCa cells ectopically expressing AR and osteoblastic cells of Saos-2. [35][36][37][38] The multiple exonic AREs could mediate such a contrasting pattern of AR autoregulation in the AR gene. 30,36 In the current study, we found that activation of the androgen/AR signalling by R1881 increased, while inhibition of the signalling by enzalutamide or castration decreased, AR protein levels in LNCaP and C4-2B cells and mouse prostates (Figures 1,3A-D,4B). In contrast, castration increased AR mRNA levels in mouse prostates ( Figure 4C). The opposite pattern of autoregulation of AR may correspond to its different functions, that is, enhancing cellular responsiveness to androgen and maintaining homeostasis.
There are multiple potential AREs in the ZFHX3 promoter. Two of them, ARE1 and ARE6 (Figure 2A), are critical for AR-mediated ZFHX3 transcription, as demonstrated by promoter-reporter assay and ChIP-PCR ( Figure 2). ARE1 acts as a classic ARE for an AR target gene, as androgen induced the binding of AR to ARE1 of ZFHX3 promoter, the binding activated ZFHX3 transcriptional activity, and mutation of the ARE1 eliminated androgen-induced ZFHX3 promoter activity ( Figure 2). However, ARE6 acts in an opposing manner compared to ARE1. AR bound to ARE6 without androgen, and androgen treatment interrupted the binding ( Figure 2C,D). In addition, the androgenindependent binding of AR to ARE6 represses the ZFHX3 promoter activity, as mutation of ARE6 dramatically increased ZFHX3 promoter activity both with and without androgen treatment ( Figure 2H).
Consistently, AR knockdown in the absence of androgen upregulated ZFHX3 transcription in the same cell lines ( Figure 1E). Therefore, it appears that at least in some AR-positive prostate cancer cells, AR is present in the nucleus even in the absence of androgen, and such androgen-independent AR binds to the ZFHX3 promoter to repress its transcription.
Such a unique pattern of androgen-induced reversal in gene transcription is unusual. Still, it has been reported that the prostatespecific membrane antigen (PSMA) gene (also known FOLH1) is transcriptionally upregulated without androgen but repressed when androgen is present due to an enhancer element. 39,40 Western blotting of cytoplasmic and nuclear proteins showed that some AR molecules were localized in the nucleus without androgen binding ( Figure 2E,F). Such a ligand-independent nuclear localization for AR is somewhat unexpected but demonstrated previously in LNCaP and LNCaP-derived C4-2 cell lines by Western blotting and immunofluorescence staining. [41][42][43] In general, the nuclear receptor AR depends on androgen as its ligand to enter the nucleus and bind to gene promoters, so whether and how some AR molecules can enter into and function in the nucleus without androgen binding is interesting and worth further investigation.
We noticed that when both ARE1 and ARE6 were mutated, androgen treatment still induced a detectable ZFHX3 promoter activity ( Figure 2H), suggesting that one or more AREs within the cloned ZFHX3 promoter (Figure 2A) also play a role in androgen-induced ZFHX3 transcription.
Enzalutamide, however, while preventing the activation of ZFHX3 transcription by androgen ( Figure 3A-D), did not decrease ZFHX3 transcription as in the transcription of KLK3, a classic target gene of AR ( Figure 3A-D). Lack of enzalutamide-induced ZFHX3 downregulation could be attributed, at least in part, to AR's binding to ARE6 without androgen. For example, the binding of enzalutamide to AR, while preventing cytoplasmic AR from entering the nucleus and reducing AR protein level, could release androgen-free nuclear AR from the ARE6 promoter DNA, eliminating the repression of AR activity on ARE6. This speculation remains further tested.
While not decreasing ZFHX3 mRNA level further as expected, enzalutamide treatment significantly decreased ZFHX3 protein level via excess protein degradation ( Figure 3E). Enzalutamide downregulated AR protein as expected (Figure 3). Simultaneous downregulation of both AR and ZFHX3 proteins further suggests the necessity of ZFHX3 for AR function. Such a necessity is also supported by the observation that AR activity scores were lower in PCa samples with ZFHX3 deletion and higher ZFHX3 expression correlated with genes indicative of AR activities ( Figures 4D,5B).
ZFHX3 undergoes frequent loss-of-function mutation in advanced prostate cancer. 11 Its deletion is more frequent in PCa specimens with higher Gleason scores ( Figure 5A), and loss of Zfhx3 induces or promotes neoplastic lesions in the prostate. 13,14 In addition, downregulation of ZFHX3 correlated with lower AR activities and worse disease-free survival in PCa patients ( Figures 4D,5D).
On the other hand, ectopic expression of ZFHX3 decreases colony formation in the 22Rv1 CRPC cell line. 11 It is thus likely that loss or downregulation of ZFHX3 enhances the development of castration resistance in PCa. In this regard, whether the downregulation of ZFHX3 protein by enzalutamide affects therapy resistance becomes a valid question. For example, while enzalutamide is effective in treating advanced and metastatic prostate cancer, 27,44 therapeutic resistance and subsequent recurrence eventually occur after enzalutamide treatment, 45,46 and over-activation of the PI3K/Akt pathway is often involved. 47,48 On the other hand, loss of Zfhx3 in mouse prostates activates the Akt signalling pathway. 13,14 It thus should be meaningful to determine whether the downregulation of ZFHX3 protein by enzalutamide contributes to its therapeutic resistance.
If downregulation of ZFHX3 indeed contributes to the development of castration resistance, the AR-ZFHX3 interaction and its downstream target genes could provide an opportunity for developing agents that could overcome resistance to ADT in PCa.
In summary, we found that in prostatic epithelial cells, androgen upregulates the transcription of ZFHX3 via the binding of AR to specific AREs in the ZFHX3 promoter, and the regulatory relationship between androgen/AR and ZFHX3 is valid in both mouse prostates and human prostate cancer specimens. When androgen is absent, some AR molecules in the nucleus bind to an ARE in the ZFHX3 promoter, and the binding represses ZFHX3 transcription.
Furthermore, the enzalutamide antiandrogen prevents AR from activating ZFHX3 transcription while causing excess ZFHX3 protein degradation. A dynamic regulatory relationship between the androgen/AR signalling and ZFHX3, both of which modulate prostate cancer development and progression, could have therapeutic implications.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.