Androgen Signaling Represses Homeobox C9, an Inhibitor of Androgen Receptor, in Prostate Cancer Cells

TL;DR

This study shows how androgen signaling and vitamin D3 interact to regulate HOXC9, a gene that inhibits prostate cancer growth.

Contribution

The study reveals a novel regulatory mechanism where HOXC9 is repressed by androgen signaling and activated by vitamin D3.

Findings

HOXC9 is a common target gene regulated oppositely by androgen and vitamin D3.

Androgen receptor preferentially suppresses HOXC9 through stronger binding and DNA methylation.

Forced HOXC9 expression inhibits androgen-dependent prostate cancer proliferation.

Abstract

Because prostate cancer proliferates in an androgen-dependent manner, various inhibitors of androgen production and antagonists of the androgen receptor (AR) are used as therapeutic agents. However, the emergence of castration-resistant prostate cancer has prompted the development of additional treatment strategies. In this study, we focused on the antiprostate cancer effects of vitamin D3 and examined novel antiproliferative effects through the crosstalk with androgen signaling. In human prostate cancer LNCaP cells, homeobox C9 (HOXC9) was identified as a common regulated target gene by dihydroxytestosterone and 1α,25-dihydroxyvitamin D3, but in opposite directions. Ligand-stimulated AR and vitamin D receptor competitively shared binding sites in the HOXC9 regulatory region, but dihydroxytestosterone stimulation preferentially suppressed HOXC9 expression due to the stronger binding properties of AR and the induction of DNA methylation. Forced expression of HOXC9 inhibited androgen signaling to eliminate the androgen-dependent proliferation by associating with the AR transcription complex, in part due to interference with AR binding to some of its targets in LNCaP cells. In summary, this study provides evidence for the involvement of HOXC9 in antiproliferative effects through a regulatory mechanism mediated by a crosstalk between vitamin D receptor and AR.

Figures

Full text

1. Introduction

Prostate cancer is one of the most commonly diagnosed cancers in developed countries [1]. Prostate cancers commonly exhibit an androgen receptor (AR), a member of the nuclear receptor (NR) family. Androgens, specifically testosterone and dihydroxytestosterone (DHT), serve as AR ligands, and androgen–AR signaling drives the progression of prostate cancer. Hence, an exhaustive target gene regulation by AR stimulated by androgens is the true nature of prostate cancer progression. Androgen-deprivation therapy (ADT) is frequently used for treatment [2,3].

For prostate cancer, ADT involves drugs that inhibit androgen production, such as CYP17A1 inhibitors [4] and LHRH agonists/antagonists [5], and a series of antiandrogens, including enzalutamide [6]. Although these therapies are initially effective, patients often become resistant, and prostate cancer reappears as castration-resistant prostate cancer (CRPC). Most CRPCs reactivate AR under androgen-depleted conditions; in other words, ADT drugs lose efficacy [7]. During the pathogenesis of CRPC, various AR-related alterations occur, including expression of excessive AR, emergence of point mutations, and alternations of coregulators [8]. The involvement of AR splicing variants, typified by the AR-V7, AR-v567es, has been argued [9,10]. These AR splice variants result from frame-shift mutations through the use of cryptic exons or by exon skipping, leading to defects in the ligand-binding domain and promoting the expression of an androgen-independent constitutively active form. Constitutively active ARs have no target site for ADT. Thus, there is a need to develop molecular inhibitors that inhibit the function of ARs lacking the ligand-binding domain.

The reciprocal crosstalk between steroid hormone–NR signaling pathways modifies the actions of steroid hormones, exerting anticancer effects or contributing to the development of drug resistance in sex steroid-dependent cancers. The development and proliferation of most types of breast cancer depend on estrogen–estrogen receptor (ERα) signaling [11]. Upon coactivation of the progesterone receptor (PR) and ERα by their agonists, PR associates with ERα and induces modification in the ERα cistrome, inhibiting breast cancer growth [12,13]. Glucocorticoid receptor (GR) induces chromatin remodeling by facilitating the binding of ERα to the target regulatory region [14]. Glucocorticoid signaling inhibits ERα activation by inducing Sulfotransferase Family 1E Member 1, an estrogen-inactivating metabolic enzyme [15]. Thus, PR and GR exert anticancer proliferative effects in breast cancer. However, in prostate cancer, long-term treatment with enzalutamide induced GR expression, leading to the recurrence of CRPC through the hijacking of some AR cistromes by GR [16]. Notably, in prostate cancer, crosstalk mechanisms by other NRs that are inhibitory toward cancer proliferation have not been fully analyzed.

The hormone 1α,25-dihydroxyvitamin D3 (referred to as D3) is an active form of vitamin D3 that exerts its activity through the NR of vitamin D receptor (VDR). D3 is of interest for its preventive effect on the development of various cancer types, including prostate cancer [17,18,19,20]. In prostate cancer cell culture, D3 induced apoptosis and inhibited cell proliferation by regulating gene expression [21]. Because the antiproliferative effect of D3 was reported to require androgen signaling in prostate cancer cells, unforeseen interactions between D3 signaling and androgen signaling was expected [22]. In this study, novel antiproliferative regulatory mechanisms were examined using human prostate cancer cells, focusing on the crosstalk between androgen and D3 signaling. Comprehensive microarray analysis using human prostate cancer LNCaP cells identified homeobox C9 (HOXC9) gene that is commonly regulated by DHT and D3, although in the opposite direction. HOXC9 expression was forcibly repressed by DHT-AR signaling with DNA methylation, whereas D3 could induce its expression only under androgen-deprived conditions. HOXC9 function was examined to evaluate the significance of the forced repression of HOXC9 in the presence of androgen signaling.

2. Results

2.1. HOXC9 Is Regulated by DHT and D3 in Opposite Directions in LNCaP Cells

To investigate how the DHT-induced proliferation of LNCaP cells is inhibited by D3, we performed microarray analysis and evaluated putative target genes involved in cell proliferation. We isolated DHT or D3 target genes, satisfying the conditions of the Z-score (≥2.0 or ≤−2.0) and ratio (≥2.0 or ≤0.5). In total, 579 and 83 target genes were selected for DHT and D3, respectively (Figure 1a). Among them, 35 probes were shared as common target genes—5 and 30 genes were regulated in the opposite and same directions by DHT and D3, respectively (Figure 1a and Table S5). Of the five genes, we focused on the homeobox C9 (HOXC9) gene, which encodes the transcription factor HOXC9, because aberrant expression of several homeobox genes is involved in the development and progression of cancer [23]. qRT-PCR analysis revealed that DHT repressed and D3 promoted HOXC9 gene expression (Figure 1b). The inhibition of HOXC9 gene expression by DHT was verified in VCaP cells, another androgen-responsive human prostate cancer cell line [24]; by contrast, it was repressed by D3, indicating that only the suppression of HOXC9 by DHT was common to LNCaP cells (Figure 1c). siRNA experiments showed that siAR and siVDR completely abrogated DHT-induced suppression and D3-induced stimulation of HOXC9 at both mRNA and protein levels, respectively, suggesting that HOXC9 gene expression was regulated through genuine DHT-AR and D3-VDR pathways (Figure 1d–f). Culturing LNCaP cells in an androgen-deprived medium also confirmed an increase in the expression of HOXC9 (Figure 1g). Interestingly, when DHT and D3 were administered simultaneously, DHT completely abolished the inductive effect of D3 (Figure 1e,f, see siCT). At concentrations of >10^−9^ M, DHT completely diminished the induction effect of 10^−7^ M D3 (Figure 1h), suggesting the dominant inhibitory effect of DHT.

To elucidate the mechanism of dominant HOXC9 repression by DHT, we performed chromatin immunoprecipitation (ChIP). Using the ChIP-Atlas database (https://chip-atlas.org/, accessed on 7 July 2023), we identified an AR-binding region containing two AR-binding half-sites located at −1219/−1215 and −1150/−1146 from the HOXC9 transcription start site. There was no information about the VDR binding region around the HOXC9 gene (Figure 1i). ChIP-qPCR using a primer set covering these two AR-binding half-sites revealed that both DHT-AR and D3-VDR bind to these regions, in other words, both NRs share the same binding region in the upstream region of the HOXC9 gene. Interestingly, when DHT and D3 were administered simultaneously, AR continued to bind to the same extent with DHT alone, whereas VDR binding was significantly decreased compared with D3 alone (Figure 1j). Taken together, AR and VDR competitively bound to the same binding site, and AR had a higher binding affinity than VDR when DHT and D3 were coadministered. These results indicate that DHT inhibits HOXC9 with a predominance compared with D3.

2.2. Involvement of DHT-Induced DNA Methylation in HOXC9 Repression

HOXC9 was one of the HOXC gene clusters consisting of nine family genes (Figure S1a). Microarray analysis revealed opposing gene regulation by DHT and D3 in not only HOXC9 but also other HOXC family genes (Figure S1b). qRT-PCR confirmed opposing regulation in HOXC13 (Figure S1c). In HOXC10 and HOXC12, the same tendency was observed, although their basal expression was low (Figure S1c). Taken together, the HOXC9 gene, whose expression levels and range of variation in regulation were sufficient, seemed to be the main target of DHT and D3 in LNCaP cells.

Because repression by DHT spanned a broad genomic region from HOXC9 to HOXC13, 5-azacytidine, a pan-inhibitor of DNA methyltransferases, was utilized to determine whether genomic methylation was involved in repression. Preadministration of 10^−5^ M 5-azacytidine abolished the DHT-dependent repression of HOXC9 (Figure 2a,b) but had no effect on D3-mediated stimulation (Figure 2c). Similar abolition was confirmed for HOXC13, 12, and 10, indicating that their DHT-dependent repression was mediated by DNA methylation (Figure S1d–f). Interestingly, even when the DHT-dependent repression of HOXC genes was abolished by 5-azacytidine, coadministration of D3 with DHT failed to exert its stimulatory effects, supporting our hypothesis that AR occupied the binding region and inhibited VDR from binding to the same region (Figure 2d and Figure S1d–f). Pretreatment with 5-azacytidine did not substantially alter AR or VDR binding to the AR-binding region of HOXC9 (Figure 1i,j), except for enhanced AR binding under vehicle-treated conditions (Figure 2e).

In VCaP cells, 5-azacytidine abolished the DHT-induced repression of some HOXC genes, especially in HOXC9, as in LNCaP cells (Figure S2a–i). By contrast, in the other two human prostate cancer cell lines that exhibit less androgen-dependent properties, 22Rv1 [25] and PC-3 [26], DHT did not repress HOXC9 expression (Figure S2j,k). Basal expression of HOXC9 was lower in 22Rv1 and PC-3 cells than in LNCaP and VCaP cells, and 5-azacytidine-mediated DNA demethylation upregulated their basal expression (Figure S2l). These results indicate that the genomic region regulating HOXC9 expression is in a hypomethylated state in androgen-responsive prostate cancer cells and can be targeted for androgen-induced HOXC9 suppression with increased DNA methylation.

PC-3 cells, which express VDR [27] but lack an endogenous AR, were exogenously transfected with wild-type AR (AR-WT) at an efficiency of ~60% (Figure S2m,n) and subsequently exposed to DHT or D3. A modest, non-significant trend toward suppression of HOXC9 expression by DHT was observed in PC-3 cells expressing AR-WT, unlike D3, which did not affect HOXC9 expression (Figure S2o). DHT-mediated repression of HOXC9 was observed in multiple prostate cancer cell lines analyzed in this study, whereas the opposite impact was observed with D3 only in LNCaP cells.

Whole-genome bisulfite sequencing revealed that multiple methylated CpG sites activated by DHT were scattered throughout the HOXC locus (Figure 2f and Table S6). We could not identify any methylated CpG sites within the ARB located ~1.2 kb upstream of the HOXC9 transcription start site, corresponding to the DNA region presented in Figure 1i,j and Figure 2e. Therefore, we focused on methylated CpG sites around HOXC9 gene (Figure S3a). The SunTag method was used to induce CpG region-specific demethylation by region-specific recruitment of multiple catalytic domains of TET1s (TETCDs) using CRISPR–dCas9 [28]. However, this approach failed to abolish the DHT-dependent repression of HOXC9 (Figure S3a–d). The recruitment of VP64, which is a transcription activator [29], induced the basal expression of HOXC9, indicating that designed functional gRNAs recruited the component to respective target regions (Figure S3e). These results suggest that DNA methylation-mediated repression of HOXC9 involves a broader range of genomic region.

To assess the involvement of DNMTs in androgen-dependent repression of HOXC9, DNMT1, DNMT3A, and DNMT3B were knocked down using siRNA (Figure S4). Among them, DNMT1 and DNMT3B exhibited the highest and lowest expression, respectively, in LNCaP cells (Figure S4a). At 48 h post-transfection, expression was efficiently reduced to ~11%, 40%, and 41% for DNMT1, DNMT3A, and DNMT3B, respectively (Figure S4c). However, these values declined by 72 h, the time point used for HOXC9 expression analysis, likely due to the limited half-life of these siRNAs (Figure S4c). Comparable effects were observed upon the simultaneous knockdown of all three genes (si1 + 3A + 3B) (Figure S4c). Under these conditions, treatment with 10^−9^ M DHT modestly attenuated DHT-dependent repression of HOXC9 expression in the siDNMT1, siDNMT3A, and si1 + 3A + 3B groups compared with siCT (Figure S4d). However, this influence was modest and was completely abolished with 10^−8^ M DHT, likely due to incomplete DNMT knockdown by the siRNAs used (Figure S4d). Collectively, these results indicate a potential involvement of DNMT1 and DNMT3A in the androgen-dependent repression of HOXC9.

2.3. HOXC9 Inhibited the Transcriptional Activity of AR and Its Splice Variants, Potentially by Associating with AR-Containing Complexes Involving the N-Terminal Activation or DNA-Binding Domains

The expression of the DHT-dependent inducible target gene KLK3, which encodes a prostate-specific antigen, decreased in LNCaP cells when 5-azacytidine was administered (Figure 3a). We hypothesize that the suppression of DHT-induced KLK3 expression in 5-azacytidine-treated cells is due to the maintenance of HOXC9 expression. Therefore, we transiently expressed HOXC9 using the Neon electroporation system. HOXC9 expression significantly decreased the DHT-dependent induction of KLK3 expression (Figure 3b,c).

To evaluate whether HOXC9 inhibited the transcription activity of AR, we performed the luciferase assay using a reporter vector with multiple consensus AR response sequences in LNCaP cells. The structures of a series of ARs used in the luciferase assay are shown in Figure 3d. AR-V7′ skipping exons 4–7, which is a similar mutant to AR-V7, was used as one of the constitutively active ARs (Figure 3d). HOXC9 significantly repressed DHT-induced promoter activity through the endogenously expressed AR(T877A) (Figure 3e) and the exogenously coexpressed ARs (Figure 3f,g). Constitutively active AR-v567es and AR-V7′ without DHT exhibited strong promoter activity and were repressed by HOXC9 coexpression (Figure 3h,i).

To investigate whether HOXC9 forms a transcriptional complex with ARs, we performed co-immunoprecipitation analyses using LNCaP cells transiently expressing HOXC9 and AR-WT, AR-v567es, or AR-V7′ isoforms. Protein extracts from LNCaP cells were immunoprecipitated with αHOXC9, followed by immunoblotting with αAR or αHOXC9. The expected molecular weights of AR were identified, and full-length AR was found to be expressed endogenously in LNCaP under all conditions (Figure 3j). These results indicate that exogenously expressed HOXC9 associates with transcriptional complexes containing various exogenously introduced AR constructs, in the presence or absence of DHT, potentially involving the N-terminal activation and/or DNA-binding domains of AR. This association may underlie the suppressive effects of HOXC9 on AR-v567es and AR-V7, which were ineffective for the AR antagonist enzalutamide (Figure 3k–n).

2.4. Stably Expressed HOXC9 Was Associated with AR-Containing Transcriptional Complexes upon DHT Administration

To evaluate the intracellular function of HOXC9, we established LNCaP-C9 cells that stably expressed exogenous HOXC9 under the CMV promoter. LNCaP-C9 cells maintained HOXC9 mRNA expression under DHT administration (Figure 4a). Immunoblotting using whole protein lysates and fractionated proteins from LNCaP-C9 showed that HOXC9 accumulated in the nucleus and increased in amount not only with D3 but also with DHT administration (Figure 4b). Immunofluorescence revealed an increase in the nuclear signal of αHOXC9 in response to DHT administration, similar to the enhancement of αAR (Figure 4c,d). However, the immunofluorescence intensities of AR and HOXC9 in individual LNCaP or LNCaP-C9 cells did not correlate, either in the presence or absence of DHT (Figure S5). Co-immunoprecipitation comparing LNCaP and LNCaP-C9 cells was performed to investigate whether the stably expressed HOXC9 is associated with AR-containing transcription complexes. In LNCaP cells, IP with αHOXC9 followed by blotting with αAR showed weak bands detected when either hormone was administered. Conversely, no band was observed with the reverse combination of antibodies (Figure 4e). By contrast, in LNCaP-C9 cells, the results of IP with αAR followed by blotting with αHOXC9 showed a dramatic increase in bands in response to DHT and D3 administration. Similar results were observed when the combination of the antibodies was reversed (Figure 4e). These results suggest that, when DHT is administered, stably expressed HOXC9 is associated with AR-containing transcription complexes.

2.5. DHT-Mediated Stable Expression of HOXC9 Impeded AR Binding to Its Target Regulatory Region, Abrogating the Regulation of a Part of AR Target Genes

To further evaluate the inhibitory action of stably expressed HOXC9 on androgen signaling, we performed a series of experiments comparing LNCaP and LNCaP-C9 cells. The luciferase assay revealed that the transcription activity mediated by ARs in response to DHT decreased in LNCaP-C9 cells compared with that in LNCaP cells (Figure 5a,b). LNCaP-C9 cells exhibited a decrease in DHT-dependent proliferation compared with LNCaP cells (Figure 5c), while maintaining the D3-induced decrease in cellular proliferation (Figure 5d). These results indicate that the stable expression of HOXC9 reduced responsiveness to androgen signaling.

To clarify the reduction in androgen signaling by HOXC9, a comprehensive microarray analysis was performed. We identified 434 genes as DHT-responsive genes in LNCaP cells, satisfying Z-score conditions (≥2.0 or ≤−2.0) and ratio criteria (≥2.0 or ≤0.5). We selected genes with less than two-thirds change in responsiveness to DHT in LNCaP-C9 cells compared with LNCaP cells. We identified 81 probes, including 33 genes regulated positively and 48 genes regulated negatively by DHT (Figure 5e and Table S7). Functional annotation using DAVID revealed that the regulation of genes related to some categories, including “sphingolipid biosynthesis process,” decreased in LNCaP-C9 cells (Figure 5f). Interestingly, the fatty acid elongation enzymes of ELOVL5 and ELOVL7 have important roles in androgen-dependent growth of prostate cancer [30,31,32].

qRT-PCR analyses of selected AR target genes (KLK3, TMPRSS2, ELOVL7, and ELOVL5) revealed a significant decrease in basal expression and DHT-induced expression in LNCaP-C9 cells compared with LNCaP cells (Figure 5g–i and S6a). From the ChIP-Atlas database, respective AR-binding regions were confirmed to be in the flanking regions of these genes. Comparative ChIP-qPCR analyses utilizing αAR revealed that DHT-dependent AR binding to the regulatory regions of KLK3, TMPRSS2, and ELOVL7 significantly decreased in LNCaP-C9 cells, implying that HOXC9 disturbed AR binding (Figure 5j–l). Similar ChIP-qPCR analyses utilizing αHOXC9 revealed HOXC9 binding to the respective AR-binding regions even in the absence of DHT in LNCaP-C9 cells (Figure 5m–o). These results suggest that stably expressed HOXC9 autonomously binds to the HOXC9-binding site in the vicinity of the AR-binding sequence in LNCaP-C9 cells, because the HOXC9 consensus sequence (T/ATTAT) [33] was identified near each AR half-site (AGAACT/A) [34] (Figure S7). In addition, such binding was further increased by DHT (Figure 5m–o), whereas siRNA-based knockdown of AR in LNCaP-C9 cells abolished these DHT-induced enhancements (Figure 5p–r). Together, these results suggest that the DHT-induced increase in HOXC9 binding to the AR-binding region is a result of HOXC9 being recruited with AR, possibly interfering with the authentic transcription complex of AR (Figure 5m–r and Figure S6e). In this manner, HOXC9 disrupts AR function by association with AR-containing transcriptional complexes and binding to HOXC9-binding sites in the vicinity of AR regulatory regions in LNCaP-C9 cells (Figure 6).

3. Discussion

In this study, D3-VDR was shown to crosstalk with DHT-AR signaling and act in a repressive manner in prostate cancer LNCaP cells. HOXC9 was identified as the key factor, competitively sharing the same regulatory region but being regulated in opposite directions by DHT-AR and D3-VDR. HOXC9 was strictly repressed by androgen-dependent DNA methylation, but when persistently forced, exerted its antiproliferative effect by inhibiting part of the AR transcriptional machinery. These findings indicated that the true role of HOXC9 in LNCaP cells would be to strictly suppress its expression, accompanied with DNA methylation in an androgen-dependent manner to prevent its inhibiting role against intracellular androgen signaling. The antiandrogenic effect of vitamin D3 in inducing HOXC9 expression was masked in the presence of androgens.

3.1. D3 Crosstalks with DHT Signaling to Regulate HOXC9 Expression in the Opposite Direction

Comprehensive gene expression analysis revealed that HOXC9 is a common target gene regulated by DHT-AR and D3-VDR (Figure 1a and Table S5). HOXC9 is one of the HOXC cluster genes, a group of homeobox genes [35]. Because disruption of homeobox gene expression is closely associated with carcinogenesis in many organs, including the prostate [23], we focused on HOXC9 in this study. Among other target genes, AGR2 promoted bone metastasis and proliferation of prostate cancer [36,37]. Its suppression by D3, as opposed to DHT, may explain some of the antiproliferative effects of D3 (Figure 1a and Table S5). The TRIB2 gene was reported to be hypermethylated in prostate cancers with the TMPRSS2:ERG fusion gene [38], confirming its androgen-dependent suppression in this study (Figure 1a and Table S5). Functional analysis of these opposite target genes by DHT and D3 other than HOXC9 would also be interesting. A limitation of this study is that the opposite regulation of HOXC9 by DHT and D3 was observed only in LNCaP cells, but not in other prostate cancer cells, such as VCaP or AR-overexpressing PC-3 cells. Thus, based on the current data, this regulatory mechanism cannot be generalized across prostate cancers.

3.2. Epigenome Editing of HOXC9 as a Novel Target Factor for Prostate Cancer Therapy

The correlation between homeobox genes and prostate cancer indicates homeobox factors, especially HOXC9, as new therapeutic targets for prostate cancer, because the DHT-dependent repression of HOXC9 was mediated by genomic DNA methylation. Recently developed epigenome editing technology can induce DNA demethylation specifically at any genomic locus, with potential application as therapeutic targets [28]. However, epigenome editing has a limitation in inducing demethylation in only a ~1000 bp region. Unfortunately, it was not possible to restore HOXC9 repression after androgen administration by only designing four gRNAs at methylation sites in the regions surrounding HOXC9 (Figure S3a–d). This is because the DHT-dependent DNA methylation sites were distributed extensively across the HOXC cluster. Future development of more broadly targeted epigenome editing tools may solve this problem. This study could not fully clarify the molecular mechanisms underlying DHT-dependent DNA methylation-mediated repression of HOXC9. Although siRNA-mediated knockdown suggests the involvement of DNMT1 and DNMT3A, the effects were partial, potentially due to incomplete suppression (Figure S4). Moreover, it remains unclear whether these DNMTs are directly recruited to the regulatory regions of HOXC9 or interact with AR-containing transcription complexes. Thus, while DNMT1 and DNMT3A are most likely involved in androgen-dependent HOXC9 repression, the precise molecular mechanisms remain unrevealed by this study.

3.3. HOXC9 as a Novel AR-Associated Factor Exhibiting Inhibitory Action

Our co-immunoprecipitation experiments revealed the nature of HOXC9 as a novel AR-associating factor in LNCaP cells (Figure 3j and Figure 4e). Stably expressed HOXC9 accumulated in the nucleus in response to DHT (Figure 4b–d), like other AR coregulators [39,40,41]. Although the immunofluorescence intensities of AR and HOXC9 did not correlate during single-cell-level analyses (Figure S5), HOXC9 increased in a DHT-dependent manner in LNCaP-C9 cells, similar to AR (Figure 4b–d). AR exhibited molecular mechanisms to enhance protein stability as well as nuclear translocation in response to androgens [42,43,44,45,46]. Stably expressed HOXC9 may utilize these mechanisms to avoid protein degradation and increase stability in an androgen-dependent manner. Several AR-interacting coregulators have been identified that interact through the interaction motifs of FxxLF, WxxLF, and LxxLF [47,48]. Because HOXC9 does not contain these motifs, the interaction between HOXC9 and AR may be indirect and mediated by other factors expressed in LNCaP cells. Our experiments using AR splice variants suggested that this association involves the N-terminal activation domain and/or the DNA-binding domain of AR (Figure 3).

The primary action of HOXC9 was thought to be the inhibition of AR binding to target regions, because both androgen-dependent stimulatory and repressive target genes reduced DHT-dependent responsiveness in LNCaP-C9 cells (Figure 5e and Table S7). Actually, DHT-dependent AR binding to the regulatory regions of KLK3, TMPRSS2, and ELOVL7 decreased significantly in LNCaP-C9 cells (Figure 5j–l). By contrast, among the several target genes, some were in different regulatory situations. The molecular mechanism for ELOVL5 impairment was slightly different, because the binding of AR on the ELOVL5 regulatory region was comparable between LNCaP and LNCaP-C9 cells (Figure S6c). Notably, AR-associated HOXC9 bindings were confirmed for ELOVL5 in LNCaP-C9 cells (Figure S6e), suggesting that HOXC9 disrupts the AR transcription complex. Because HOXC8, which belongs to the same family as HOXC9, inhibited the recruitment of SRC-3 and CBP to the target locus of AR [49], it is possible that the stably expressed HOXC9 could be affecting the authentic AR transcription complex. In DEGS1, a control gene whose regulation by DHT was unaffected by stably expressed HOXC9 (Figure S6b), the binding of AR was comparable (Figure S6d), and the additional recruitment of HOXC9 by DHT was not detected (Figure S6f). Taken together, the reduction in DHT-dependent regulation in LNCaP-C9 cells may result from declined binding of AR to regulatory regions and/or functional disruption of the AR transcription complex, possibly mediated by the presence of a stably expressed HOXC9 within the AR-associated complex (Figure 6).

3.4. Fatty Acid Elongation Enzymes as One of the Potential Target for the Inhibitory Action of HOXC9

This study demonstrated that the sustained stable expression of HOXC9, even under androgenic conditions, attenuates androgen-dependent proliferation of LNCaP cells (Figure 5c). However, these findings were derived solely from in vitro experiments using cultured cell systems; a lack of validation employing in vivo models, such as xenograft tumors, represents one limitation of this study.

Nevertheless, stably expressed HOXC9 exerted its inhibitory effect on only a subset of 81 AR target genes (Figure 5e), and the regulatory loss of its limited number of target genes was sufficient to prevent androgen-dependent cell proliferation (Figure 5). We hypothesized that, among these 81 genes, many were responsible for androgen-dependent cell proliferation. To identify such candidate genes, functional enrichment analysis was performed. Based on the DAVID analysis results, we selected ELOVL5 and ELOVL7, as they play critical roles in androgen-dependent prostate cancer growth. ELOVL5 and ELOVL7 are known as enzymes involved in unsaturated fatty acid and saturated fatty acid elongation, respectively. In recent years, it has become clear that lipid metabolism contributes to the growth of prostate cancer and the acquisition of therapeutic resistance [50,51]. ELOVL5 and ELOVL7 are androgen target genes in prostate cancer, whereas depletion of ELOVL5 and ELOVL7 induced suppression of its cellular proliferation [30,31,32]. The androgen-dependent induction of ELOVL5 was observed in prostate cancer cell lines and human prostate cancer clinical specimens; ELOVL5 suppression causes mitochondrial dysfunction, increasing reactive oxygen species production and attenuating androgen-dependent prostate cancer proliferation [31]. ELOVL7 promotes prostate cancer growth through the elongation of saturated very-long-chain fatty acids and subsequent enhancement of de novo androgen synthesis [32]. These results indicated that fatty acid elongation through ELOVL5 and ELOVL7 has an important role in the androgen-dependent growth of prostate cancer. Given that the androgen-responsive induction of these critical genes is impaired in LNCaP-C9 cells, this defect may underlie the loss in androgen-dependent proliferative capacity. Although the involvement of other genes cannot be excluded, the insufficiency of these enzymes may be pivotal for the antiandrogenic influence of HOXC9 observed in this study.

4. Materials and Methods

4.1. Cell Culture and Hormones

LNCaP cells (RRID:CVCL_0395), 22Rv1 (RRID:CVCL_1045), and PC-3 cells (RRID:CVCL_0035) were obtained from American Type Culture Collection (Manassas, VA, USA) and maintained in RPMI-1640 phenol red-free medium (Life Technologies, Carlsbad, CA, USA) supplemented with 5% (v/v) fetal bovine serum and antibiotics (Life Technologies, Carlsbad, CA, USA). VCaP cells (RRID:CVCL_2235) were obtained from the European Collection of Authenticated Cell Cultures and maintained in DMEM (Life Technologies, Carlsbad, CA, USA) supplemented with 10% (v/v) fetal bovine serum and antibiotics (Life Technologies, Carlsbad, CA, USA). Cells were maintained in humidified 5% CO_2_–95% air at 37 °C. All cells used in this study were derived from male human prostate cancer. Cells used in this study were employed in the experiment from the time of acquisition to the 10 passage and verified to be free of mycoplasma contamination through PCR analysis using a Mycoplasma Detection kit (SouthernBiotech, Birmingham, AL, USA). At 24 h before hormone or chemical treatment, the culture media was replaced with serum-free media, and the cells were exposed to the indicated hormones with serum-free media for 24 h, as described [52]. The cells were used for subsequent experiments. The following chemicals were obtained from commercial sources: DHT (Sigma Aldrich, St. Louis, MO, USA), 1α,25-dihydroxyvitamin D3 (Merck KGaA, Darmstadt, Germany), and 5-azacytidine (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan). Proliferation assay was performed using LNCaP cells, as described [53].

4.2. Microarray Analysis

Total RNA from LNCaP and LNCaP-C9 cells treated with 1.0 × 10^−7^ M DHT or D3 for 24 h was used for cDNA microarray analysis by APRO Life Science (APRO Life Science, Tokushima, Japan) and Cell Innovator (Cell Innovator, Fukuoka, Japan), as described [53]. We selected the probes that called the “P” flag and the signal cut-off threshold was <100. To identify regulated genes, we calculated the Z-score and the ratio (non-log-scaled fold-change) for comparisons between each vehicle- and hormone-treated sample.

4.3. qRT-PCR

Total RNA isolation from cells, synthesis of cDNA, and quantitative real-time PCR (qRT-PCR) were performed as described [53]. The primer sets used for qRT-PCR are shown in Table S1. Data for gene expression analyses were normalized to TATA-binding protein (TBP) expression.

4.4. siRNA Transfection

Non-targeting control siRNA pool and a combination of four sets of siRNAs designed for AR, VDR, DNMT1, DNMT3A, and DNMT3B were obtained from Dharmacon (Thermo Fisher Scientific, Waltham, MA, USA). siRNA sequences are shown in Table S2. siRNA transfection was performed as described [54].

4.5. Immunoblotting and Immunocytochemistry

Immunoblotting of whole-cell protein [53] and fractionated cytoplasmic and nuclear proteins was performed as described [52]. Immunocytochemistry was performed as described [53]. The intensity of immunofluorescence staining for individual cells were quantified using ZEN 2.1 software (Leica Microsystems GmbH, Wetzlar, Germany). Antibodies against AR (αAR) diluted at 1:1000 (H-280, RRID:AB_633881; Santa Cruz Biotechnology, Santa Cruz, CA, USA), VDR (αVDR) diluted at 1:1000 (D2K6W, RRID:AB_2637002; Cell Signaling Technology, Danvers, MA, USA), HOXC9 (αHOXC9) diluted at 1:1000 (ab50839, RRID:AB_880494; Abcam, Cambridge, UK), α-tubulin (αTubulin) diluted at 1:5000 (T5168, RRID:AB_477579; Sigma Aldrich, St. Louis, MO, USA), and lamin A/C (αLamin A/C) diluted at 1:2000 (N-18, RRID:AB_648152; Santa Cruz Biotechnology, Dallas, TX, USA) were used. Uncropped versions of the blots are shown in Figure S8.

4.6. Chromatin Immunoprecipitation

The fixation of LNCaP and LNCaP-C9 cells treated with each reagent for 24 h followed by cell lysis, sonication, immunoprecipitation, and DNA purification was performed as described [54]. Specific antibodies against AR (αAR) (H-280, RRID:AB_633881; Santa Cruz Biotechnology, Santa Cruz, CA, USA), VDR (αVDR) (D2K6W, RRID:AB_2637002; Cell Signaling Technology, Danvers, MA, USA), and αHOXC9 (ab50839, RRID:AB_880494; Abcam, Cambridge, UK) were used. Quantification of each ChIP-DNA was performed similar to that for qRT-PCR with the primer sets listed in Table S3. Normal IgG was used as a negative control. Data are shown relative to the results using each purified DNA that was not immunoprecipitated as input DNA.

4.7. Gene Transfer by Neon Electroporation

The Neon electroporation system was used for high-transfection-efficiency gene transfer in LNCaP and PC-3 cells. Electroporation conditions to introduce 8 μg of plasmid DNA into 5 × 10^7^ cells using a 100 μL tip were as follows—pulse voltage: 1250 V, pulse width: 20 ms, and pulse number: 2. Control experiments utilizing GFP expression vectors demonstrated a transfection efficiency of 70–80% in LNCaP cells and 60% in PC-3 cells, respectively. LNCaP-C9 cells stably expressing HOXC9 in LNCaP cells were established by selection using 1000 μg/mL G-418 after transfecting with the HOXC9/pTargeT Mammalian Expression Vector (GenBank Accession Number: AY540613) using Neon electroporation.

4.8. Whole-Genome Bisulfite Sequencing

Genomic DNA from LNCaP cells (RRID:CVCL_0395) treated with 10^−7^ M DHT for 24 h was utilized for whole-genome bisulfite sequencing. The analysis was performed in a commissioned study by Macrogen Japan (Macrogen Japan, Tokyo, Japan) and APRO Life Science. Briefly, library construction was performed using Accel-NGS Methyl-Seq DNA Library Kit EZ DNA Methylation-Gold Kit (Zymo Research, Irvine, CA, USA) according to the protocol of Accel-NGS Methyl-Seq DNA Library Kit for Illumina platforms (Integrated DNA Technologies, Inc., Coralville, IA, USA). Statistical analysis was performed using the differences between the DHT- and vehicle-treated groups (delta_mean) per comparison pair. Significant results were selected on conditions of “delta_mean ≥ 0.2”.

4.9. Plasmid Construction and Luciferase Assay

Coding sequences of AR-WT and AR(T877A) were ligated into the pcDNA3.1zeo+ expression vector (Life Technologies, Carlsbad, CA, USA) and three copies of the AR response element (ARE; GGAACAgtaTGTTCT) were ligated into the pGL4-TK vector (Promega, Madison, WI, USA, GenBank Accession Number; AY738230), as described [52]. AR-v567es and AR-V7′ expression vectors were constructed by inverse PCR utilizing AR-WT/pcDNA3.1zeo as template with respective primer sets. To construct a human HOXC9 expression vector, the coding sequences of HOXC9 were amplified by PCR from LNCaP cells and ligated to the pTARGET mammalian expression vector (Promega, Madison, WI, USA, GenBank Accession Number; AY540613). The primer sets used for vector construction are shown in Table S4. Transient transfection, administration of DHT and/or enzalutamide, and measurement of luciferase activity in LNCaP and/or LNCaP-C9 cells were performed as described [52].

4.10. Immunoprecipitation

LNCaP and/or LNCaP-C9 cells treated with Neon electroporation and/or hormones were lysed in NP-40 buffer (20 mM HEPES [pH 8.0], 0.1 mM EDTA [pH 8.0], 150 mM NaCl, 1% NP-40, and protease inhibitor cocktail [Merck]) for 30 min on ice. The supernatant was incubated with the rabbit polyclonal antibody against human AR (αAR) (H-280, RRID:AB_633881; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or mouse monoclonal antibody against human HOXC9 (αHOXC9) (ab50839, RRID:AB_880494; Abcam, Cambridge, UK) for 24 h and then incubated with Dynabeads Protein G (Thermo Fisher Scientific, Waltham, MA, USA) for 24 h. Dynabeads collected using a magnet bar were washed five times with NP-40 buffer. The immunoprecipitated samples were eluted with SDS buffer used for immunoblotting. The immunoprecipitated samples were used for immunoblotting using a mouse monoclonal antibody against human AR (αAR) (441, RRID:AB_307266; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or mouse monoclonal antibody against human HOXC9 (αHOXC9) (ab50839, RRID:AB_880494; Abcam, Cambridge, UK), respectively. The steps for incubation with secondary antibody and detection of chemiluminescence using HRP were similar to those described for immunoblotting.

4.11. Statistical Analysis

All figures were created using GraphPad Prism 9.0 (RRID:SCR_002798; GraphPad Software, Boston, MA, USA). The number of repetitions and errors for each experiment are given in each figure legend. Student’s t-test (Figure 1d, Figure 2a,e, Figure 3a–c, Figure 4d, Figure 5a,b,g–i, Figures S2l,m, S3c and S6a,b), one-way analysis of variance (ANOVA) and Dunnett’s multiple comparisons test (Figure 1b,c,g,j, Figure 2b,c, Figure 3e–i,k–n, Figure 4a, Figure 5c,d, Figures S1c, S2a–k,o and S4d), one-way ANOVA and Tukey’s test (Figure 1e,h, Figure 2d, and Figure S1d–f), or two-way ANOVA and Tukey’s test (Figure 5j–r, Figures S3d,e and S6c–f) were performed to analyze statistical significance. Analyses with p > 0.05 in one-way ANOVA and two-way ANOVA were indicated as not significant (ns).

4.12. Targeted Demethylation and Activation Using the CRISPR–dCas9–SunTag Platform

Targeted DNA demethylation by recruiting the catalytic domain of TET1 (TET1CD) and target gene activation by recruiting VP64 using the CRISPR–dCas9–SunTag platform was performed [28,29]. The plasmid vectors used in the CRISPR–dCas9–SunTag platform of pCAG-dCas9-5xPlat2Af1D (RRID:Addgene_82560), pCAG-scFvGCN4sfGFPTET1CD (RRID:Addgene_82561), and pCAG-scFvGCN4sfGFP-VP64-GB1 (RRID:Addgene_141417) were obtained from Addgene. pCAG-dCas9-5xPlat2Af1D, pCAG-scFvGCN4sfGFPTET1CD, and pCAG-scFvGCN4sfGFP-VP64-GB1-expressed dCas9-multiGCN4, scFv-GFP-TET1CD, and scFv-GFP-VP64 fusion proteins, respectively. The gRNA expression vectors to recruit these components to specific genomic loci were constructed using VectorBuilder (VectorBuilder Japan, Kanagawa, Japan). These plasmid vectors were introduced into LNCaP cells by Neon electroporation. After 48 h, LNCaP cells were used for gRNA quality checks or DHT administration experiments. The quality of the designed gRNAs was evaluated using GeneArt Genomic Cleavage Detection Kit (Thermo Fisher Scientific, Waltham, MA, USA). gRNA sequences and primer sets designed using GeneArt Genomic Cleavage Detection Kit are listed in Supplementary Tables S8 and S9, respectively.

5. Conclusions

In this study, we identified HOXC9 as a vitamin D target that exerts antiandrogenic effects masked by androgen signaling in LNCaP cells. While androgen-dependent HOXC9 repression was consistent across multiple prostate cancer cell lines, vitamin D-induced HOXC9 upregulation was detected only in LNCaP cells. Thus, the bidirectional regulation of HOXC9 seen in prostate cancer cells could not be generalized, representing a limitation of the present study. Although vitamin D itself has a potent impact on calcium metabolism and is unlikely to be useful as a treatment for prostate cancer, the identification of one of the genes that exerts anticancer effects through vitamin D may provide an approach for the treatment of prostate cancer. In particular, induction of HOXC9 expression by DNA demethylation through epigenome editing can be used as a therapeutic approach against prostate cancer. We hypothesize that androgen signaling masks its own inhibition factor, HOXC9, by activating genomic DNA methylation, and can be used as a novel antiprostate cancer factor. Further studies employing in vivo models are required to clarify the physiological relevance of HOXC9 and determine its therapeutic potential in prostate cancer.