Definition of a FoxA1 Cistrome That Is Crucial for G₁ to S-Phase Cell-Cycle Transit in Castration-Resistant Prostate Cancer

The FoxA subfamily of winged helix/forkhead box (Fox) transcription factors, which consists of 3 members, FoxA1, FoxA2, and FoxA3, have been found to play important roles in multiple stages of development, metabolism, differentiation, and proliferation (1). FoxA proteins function as “pioneer factors” that engage chromatin before other transcription factors (2–5). Recent studies have further shown that FoxA1 functions as a pioneer factor for steroid hormone receptors (SR), including androgen receptor (AR) in prostate cancer cells and estrogen receptor (ER) in breast cancer cells, directing AR- and ER-regulated hormone (androgen and estrogen)-responsive genes (6–8). These studies further suggested that FoxA1 acts upstream of AR and ER to regulate their target genes in hormone-dependent prostate and breast cancers.

In both androgen-dependent prostate cancer (ADPC) and fatal castration-resistant prostate cancer (CRPC), AR expression and functionality have been well documented (9), but the receptor seems to play different roles in the two diseases. For example, in ADPC, AR functions primarily to promote G₁–S cell-cycle progression, by transcriptional and/or posttranscriptional regulation of CDKN1A, CCND1, and CDKN1B (10). However, in CRPC, the primary function of AR seems to be regulation of G₂–M transition. Thus, the receptor seems to be “reprogrammed” to direct transcriptional regulation of G₂–M phase-specific genes, including UBE2C and CDK1 (11), by mechanisms that are not well understood. Consistent with the critical role of FoxA1 in assisting SR binding, FoxA1 silencing in CRPC leads to decreased AR binding to enhancers of G₂–M phase genes and lower gene expression levels (11). Hence, in CRPC, FoxA1 seems to be mainly involved in AR-regulated G₂–M cell-cycle progression. However, whether this pioneer factor can also function independently in CRPC, in addition to collaborating with AR, has not been investigated.

In this study, we examined the role of FoxA1 in CRPC cell-cycle progression by silencing FoxA1 in unsynchronized CRPC cells. We found that in the absence of FoxA1, CRPC experienced a G₁–S block and, unexpectedly, not G₂–M arrest. Our comprehensive integrated analysis of gene expression and FoxA1 cistrome data further revealed direct upregulation of CCNE2 by FoxA1 binding sites specific to CRPC, as well as indirect upregulation of CCNA2 by E2F1. We further established a requirement for CREB1 and MYBL2 in CRPC-specific FoxA1 binding, through histone H3 acetylation-facilitated nucleosome disruption, resulting in upregulation of CCNE2 and CCNA2 expression and enhanced growth of CRPC. Collectively, these data indicate that FoxA1, together with CREB1 and MYBL2, drive G₁–S progression in CRPC, which is distinct from the classic role of FoxA1 as an AR cofactor.

The ADPC cell line LNCaP was purchased from the American Type Culture Collection (ATCC) and cultured in RPMI 1640 media (Invitrogen) supplemented with 10% FBS. The CRPC cell line abl was provided by Zoran Culig (Innsbruck Medical University, Austria). A second CRPC cell line, CWR22Rv1, was provided by Steven P. Balk (Harvard Medical School, Boston, MA). LNCaP was authenticated by ATCC, and abl and CWR22Rv1 were authenticated as described previously (12, 13). All 3 cell lines were passaged in our laboratory for less than 6 months after resuscitation.

siRNA transfections were done by Lipofectamine 2000 (Invitrogen). The sequences for siRNAs were listed in Supplementary Table S1.

Western blot analyses were carried out as previously described (14). Antibodies used are listed in the Supplementary Table S2.

Cell proliferation was measured by the WST-1 kit (Roche), as previously described (15).

Cells were collected, and DNA contents were analyzed by a FACS Calibur cell flow cytometer (Becton Dickinson Biosciences) as previously described (14, 15).

Real-time reverse transcriptase (RT)-PCR was carried out as previously described (14). Primers used are listed in Supplementary Table S1.

The ChIP-on-chip experiments were carried out in biological triplicates as previously described (11). The raw data of FoxA1 ChIP-on-chip have been submitted to the Gene Expression Omnibus (GEO) repository under the accession number GSE26329. Chromatin immunoprecipitation (ChIP) and re-ChIP assays were carried out as previously described (16). Antibodies for ChIP and re-ChIP assays are listed in Supplementary Table S2.

A meta-analysis was carried out by using 3 clinical ADPC/CRPC gene expression microarray data sets (17–19) from Oncomine (20). The overexpressed and underexpressed genes in CRPC versus ADPC, and randomly selected genes were correlated with 3 FoxA1 binding groups. Details are available in the Supplementary Materials and Methods.

Formaldehyde-assisted isolation of regulatory elements (FAIRE) was carried out as previously described (21, 22). All primer sequences are listed in Supplementary Table S1.

MDmodule (23) was used to conduct a de novo search for motifs significantly enriched in abl-specific binding sites with LNCaP-specific binding regions as background. Motif length varied from 6 to 15 nucleotides. Significantly enriched motif was annotated using STAMP (24).

Our previous findings that FoxA1 binding at the enhancer regions of several cell-cycle G₂–M phase genes (e.g., UBE2C and CDK1) is required for overexpression of these genes in CRPC (11) prompted us to further examine the functional role of FoxA1 in CRPC. We first tested the effect of FoxA1 silencing on cell proliferation in abl, a CRPC cell line that closely models clinical CRPC (11, 12, 25). siRNAs targeting FoxA1 (2 independent siRNAs) were transfected into abl. Both siRNAs dramatically reduced cellular FoxA1 levels (Fig. 1A) and significantly decreased androgen-independent cell proliferation of abl (Fig. 1A). The inhibitory effect of FoxA1 silencing on cell proliferation was also observed in CWR22Rv1 (Supplementary Fig. S1A), another CRPC cell model (26). We next examined the effect of FoxA1 silencing on CRPC cell-cycle progression by using a thymidine-nocodazole block to enrich abl cells at mitosis (15). Consistent with the role of FoxA1 in the upregulation of cell-cycle G₂–M phase genes in CRPC cells (11), FoxA1 silencing caused a G₂–M accumulation in the treated abl cells (Fig. 1B). However, FoxA1 knockdown resulted in a G₁–S block in unsynchronized abl and CWR22Rv1 cells (Fig. 1C; Supplementary Fig. S1B), suggesting that FoxA1 promotes G₁–S and G₂–M progression in CRPC cells. As a major role of AR in CRPC is to upregulate G₂–M phase gene expression and promote the G₂–M transition (11), these data further indicate that the ability of FoxA1 to promote G₁–S progression is an AR-independent phenomenon. In agreement with our previous findings (16), silencing of FoxA1 had no effect on dihydrotestosterone (DHT)-stimulated LNCaP G₁–S progression and cell proliferation (Supplementary Fig. S1C and D). Taken together, these data suggest that FoxA1 is required for androgen-independent but not androgen-dependent growth of prostate cancer.

To identify FoxA1-regulated genes that contribute to FoxA1-mediated G₁–S transition in CRPC cells, abl cells were transfected with siFoxA1 and a control siRNA (siControl), cultured for 72 hours, and analyzed for expression of the G₁ phase genes CCNE2, CCNA2, CCNE1, CDK2, CDK4, CDK6, CCND1, CCND2, CCND3, CDKN1A, and CDKN1B and RB. LNCaP cells transfected with these siRNAs and treated with DHT (10 nmol/L) for 4 hours and 24 hours served as controls. Among the 11 genes examined by real-time RT-PCR, expression levels of CCNE2, CCNA2, and CDK2 were higher in abl cells transfected with siControl than similarly transfected, untreated or DHT-treated LNCaP (Fig. 2A), although CCNE2, CCNA2, and CDK2 expression levels were increased after DHT treatment of LNCaP cells (Fig. 2A; ref. 27). Silencing of FoxA1 markedly decreased CCNE2 and CCNA2 but not CDK2 expression in abl cells, but not LNCaP cells (Fig. 2A). Importantly, reanalysis of gene expression profiles from 3 independent clinical studies (17–19) confirmed that expression of CCNE2 and CCNA2 was greater in cases of CRPC than in cases of ADPC (Fig. 2B). Western blot analysis showed that CCNE2 and CCNA2 protein levels were also greater and were FoxA1-dependent in abl compared with LNCaP cells, in the presence or absence of DHT (Fig. 2C). Consistent with previous reports showing that CCNE2 and CCNA2 binding and activation of CDK2 lead to RB phosphorylation (28, 29), FoxA1-enhanced expression of CCNE2 and CCNA2 protein increased FoxA1-dependent expression of phosphorylated CDK2 and phosphorylated RB in abl versus LNCaP (Fig. 2C). The inhibitory effect of FoxA1 depletion on mRNA and protein expression levels of CCNE2, but not CCNA2, was also observed in CWR22Rv1 cells (Supplementary Fig. S2A).

We next examined the functional role of CCNE2 and CCNA2 in ADPC and CRPC cell growth. In cell proliferation assays, CCNE2 or CCNA2 silencing markedly decreased abl and CWR22Rv1 growth (Fig. 2D; Supplementary Fig. S2B), suggesting that CCNE2 and CCNA2 play an essential role in CRPC cell proliferation. To test whether differences in expression of CCNE2 and CCNA2 were responsible for androgen-independent and androgen-dependent phenotypic differences between abl and LNCaP cells, we generated LNCaP cell lines stably expressing CCNE2 or CCNA2. Consistent with a previous report that overexpression of G₁ cyclins was not sufficient to induce LNCaP growth in the absence of androgen (30), no effect of CCNE2 or CCNA2 overexpression on cell proliferation and G₁–S progression was observed for vehicle-treated LNCaP cells (Fig. 2E; Supplementary Fig. S2C). However, while DHT treatment (1 nmol/L or higher) increased growth of vector transfected LNCaP, a growth response in LNCaP cells overexpressing CCNE2 or CCNA2 was observed at a lower (0.1 nmol/L) concentration of DHT (Fig. 2E), suggesting that CCNE2 and CCNA2 overexpression increases LNCaP androgen sensitivity.

To investigate the mechanism underlying FoxA1-regulated CCNE2 and CCNA2 gene expression, we mapped the FoxA1 cistromes in LNCaP and abl cells by combining ChIP with Affymetrix human whole genome tiling arrays (ChIP-on-chip). Using the MAT (model-based analysis of tiling-array) algorithm (31) with a P value cut-off of 1E-4 or less, we identified 14,965 and 18,110 FoxA1 binding sites in LNCaP and abl cells, respectively. Overlapping analysis of FoxA1 binding in LNCaP and abl identified 14,248 common FoxA1 binding regions between the 2 cell lines, 717 LNCaP-specific FoxA1 binding sites and 3,862 abl-specific FoxA1 binding sites (Fig. 3A; Supplementary Fig. S3A). The ChIP-on-chip results were confirmed by using direct ChIP for FoxA1 on a subset of common FoxA1 binding regions, LNCaP-specific FoxA1 binding regions, abl-specific FoxA1 binding regions, and negative regions (Supplementary Fig. S3B). To determine the functional significance of common LNCaP-specific and abl-specific FoxA1 binding sites, these 3 types of FoxA1 binding sites were correlated to gene expression profiles from 3 clinical studies on ADPC and CRPC cases (17–19). Interestingly, abl-specific FoxA1 binding sites were significantly enriched within 20 or 40 kb of the transcription start sites (TSS) of overexpressed (e.g., CCNE2; Fig. 2B) and underexpressed genes (but more notably overexpressed genes) in CRPC versus ADPC, but not randomly selected genes (Fig. 3B; Supplementary Fig. S3C and D). By contrast, no obvious enrichment of LNCaP-specific FoxA1 binding sites and common FoxA1 binding sites was observed near overexpressed, underexpressed, or randomly selected genes (Fig. 3B; Supplementary Fig. S3C and D). These results suggest that abl-specific FoxA1 binding sites may, in general, directly regulate differentially expressed genes in CRPC versus ADPC.

On the basis of strong correlation between global FoxA1 binding and differential gene expression profiles, we next investigated regulation of CCNE2 by FoxA1 in CRPC-abl and ADPC-LNCaP cells. ChIP-on-chip analysis identified an abl-specific FoxA1 binding site at the CCNE2 promoter region, 3 abl-specific FoxA1 binding sites located −24.0, −9.8, and +23.5 kb away from the TSS of CCNE2, and a common FoxA1 binding site (defined as MAT-score ≥ 3.72 in both cell lines; see the Supplementary Materials and Methods) 14.3 kb downstream of TSS of CCNE2 (Fig. 4A). The common site displayed stronger FoxA1 binding in abl (MAT-score = 9.87) compared with LNCaP (MAT-score = 4.37). No overlap was seen between these abl-specific FoxA1 binding sites and AR binding regions in abl cells (ref. 11; data not shown). Direct ChIP analysis showed higher FoxA1 occupancy at the CCNE2 promoter region and the 4 putative CCNE2 enhancer regions in abl versus LNCaP cells (Fig. 4B). To further characterize the CCNE2 promoter and the putative CCNE2 enhancer regions, ChIP assays were carried out in LNCaP and abl cells, using antibodies against an enhancer histone mark H3K4 mono-methylation (H3K4me1; ref. 32), phosphorylated RNA polymerase II at serine 5 (p-Pol II), and transcription coactivators CREB binding protein (CBP) and Mediator 1 (MED1). Enrichment of H3K4me1 was observed at the 4 putative CCNE2 enhancer regions compared with the CCNE2 promoter region in abl, and the level of H3K4me1 was higher at putative CCNE2 enhancers 1, 2, and 4 in abl compared with LNCaP (Fig. 4C), indicating that these 4 distal FoxA1 binding regions function as enhancers in abl cells. Greater p-Pol II level and increased CBP and MED1 binding at the CCNE2 enhancer and promoter regions in abl versus LNCaP cells (Fig. 4C) further supported the hypothesis that these FoxA1 binding sites may play more important transcriptional regulatory roles in abl than in LNCaP cells. In addition, H3K4me1 levels were higher, and greater recruitment of FoxA1, p-Pol II, CBP, and MED1 at UBE2C and/or CDK1 enhancers was observed in abl versus LNCaP cells (Supplementary Fig. S4A), in agreement with our previous findings (11).

As transcriptionally active cis-regulatory elements always reside within nucleosome-depleted regions (NDR; ref. 33) and the FAIRE technique has been used to successfully identify NDRs (21, 22), we used FAIRE to analyze the local chromatin structure of the FoxA1 binding regions. LNCaP and abl cells were transfected with siControl or siFoxA1 followed by FAIRE. In siControl transfected cells, a higher FAIRE signal was observed at the CCNE2 enhancer and promoter regions in abl cells than in LNCaP cells (Fig. 4D), and importantly, FoxA1 silencing decreased FAIRE enrichment at the CCNE2 enhancer and promoter regions only in abl cells (Fig. 4D). Taken together, these data strongly indicate that increased FoxA1 binding induces nucleosome depletion at the CCNE2 enhancer and promoter regions. We obtained essentially similar results at the UBE2C and CDK1 enhancers (Supplementary Fig. S4B).

Although, in general, FoxA1 binding was significantly enriched near overexpressed genes in CRPC compared with ADPC (Fig. 3B), we were unable to identify, using the ChIP-on-chip technique, abl-specific FoxA1 binding sites near CCNA2 (Fig. 5A), indicative of an indirect mechanism of FoxA1-mediated CCNA2 upregulation in abl cells (Fig. 2A). As previous studies in other systems showed direct regulation of CCNA2 by E2F1 transcription factor binding to the CCNA2 promoter region (34, 35) and our FoxA1 ChIP-on-chip and direct FoxA1 ChIP analyses identified and confirmed 2 putative E2F1 enhancers reside 15.8 and 33.4 kb downstream of the TSS of E2F1 (Fig. 5A and B), we hypothesized that FoxA1 may regulate CCNA2 expression through a direct upregulation of E2F1. As expected, increased levels of H3K4me1 and p-Pol II, higher occupancy of CBP and MED1, and higher FAIRE enrichment at E2F1 enhancers, were observed in abl compared with LNCaP cells (Fig. 5C and D), resulting in FoxA1-dependent increased mRNA and protein levels of E2F1 in abl versus LNCaP cells in the absence of androgen (Fig. 5E). Similar to CCNE2 regulatory regions, no AR binding was observed at these 2 E2F1 enhancers in abl cells (data not shown). We next performed ChIP to examine whether E2F1 directly regulates CCNA2 in abl cells, and expression of CCNA2 mRNA following E2F1 silencing was also examined in LNCaP and abl cells. As shown in Figure 5F and G, E2F1 binding at the CCNA2 promoter was increased, and E2F1-dependent, CCNA2 mRNA expression was greater in abl versus LNCaP in the absence of DHT, showing that CCNA2 is a direct E2F1 target gene in abl but not LNCaP cells. Thus, FoxA1, through a direct transcriptional regulation of E2F1, indirectly upregulates (induces) CCNA2 expression in abl cells.

We next investigated the mechanisms responsible for abl-specific FoxA1 binding resulting in abl-specific FoxA1 target gene expression. Given that previous studies have reported that cooperation among transcription factors can result in altered chromatin binding activity (7, 11, 16), we hypothesized that this may be the case for FoxA1 binding in abl but not in LNCaP cells. Thus, to examine whether other transcription factors may affect FoxA1 binding, we conducted a de novo transcription factor motif search within the abl- and LNCaP-specific FoxA1 binding regions. As expected, Forkhead motifs were significantly enriched within both abl- and LNCaP-specific FoxA1 binding sites compared with the whole genome background (abl HyperGeometric P values (36): 5.7E-27 for abl and 1.0E-27 for LNCaP). Interestingly, MYB and CREB motifs were significantly enriched within abl-specific FoxA1 binding regions compared with LNCaP-specific FoxA1 binding regions (HyperGeometric P values: 5.9E-20 for MYB and 1.1E-15 for CREB; Fig. 6A), indicating that transcription factors recognizing MYB and CREB motifs may play a “cooperative role” in FoxA1 binding and abl-specific FoxA1 target gene regulation. As ubiquitous expression of MYBL2 (within the MYB family) and CREB1 (within the CREB family) has been reported (37, 38), we examined expression and chromatin binding of MYBL2 and CREB1 in abl and LNCaP cells. MYBL2 and CREB1 protein levels were higher in abl versus LNCaP cells (Fig. 6B), and ChIP assays showed greater recruitment of MYBL2 and CREB1 to regulatory regions of the abl-specific FoxA1 target genes CCNE2, E2F1, UBE2C, and CDK1 (Fig. 6C; Supplementary Fig. S5A) in abl versus LNCaP cells.

We next examined whether MYBL2 and CREB1 affect FoxA1 binding and p-Pol II loading at regulatory regions of abl-specific FoxA1 target genes. LNCaP and abl cells were transfected with siRNAs targeting MYBL2 or CREB1, and Western blot analyses and ChIP assays were carried out by using an anti-FoxA1 antibody. No effect of MYBL2 or CREB1 silencing on FoxA1 protein expression levels was observed; however, silencing of either transcription factors decreased FoxA1 binding at the enhancers and promoters of CCNE2 and E2F1, as well as the UBE2C enhancers and the CDK1 enhancer in abl but not LNCaP cells (Fig. 6D; Supplementary Fig. S5B and C). In addition to decreased FoxA1 binding, ChIP analyses further showed that MYBL2 and CREB1 silencing reduced p-Pol II levels on most enhancers and promoters of abl-specific FoxA1 target genes, similar to the effect of FoxA1 silencing itself (Supplementary Fig. S5D). Collectively, these results show that MYBL2 and CREB1 significantly alter FoxA1 binding and p-Pol II loading on chromatin.

To reveal the hierarchical relationship among MYBL2, CREB1 and FoxA1 binding, we silenced FoxA1 and examined MYBL2 and CREB1 binding on chromatin. ChIP analyses showed that FoxA1 silencing decreased MYBL2 and CREB1 recruitment to most regulatory sites of CCNE2, E2F1, UBE2C, and CDK1 in abl but not LNCaP cells (Fig. 6E; Supplementary Fig. S6A). Serial ChIP (re-ChIP) analyses of FoxA1/MYBL2 and FoxA1/CREB1 further showed a stronger FoxA1-MYBL2 or -CREB1 interaction on abl-specific FoxA1 target gene loci in abl versus LNCaP cells (Supplementary Fig. S6B). These data suggest that FoxA1-MYBL2 or -CREB1 codependently bind to the same regulatory elements of abl-specific FoxA1 target genes.

To further investigate the molecular mechanisms for the codependent binding of FoxA1-MYBL2 and -CREB1 on chromatin, we examined the effect of their silencing on histone acetylation. Altered chromatin structure can greatly influence transcription factor access to chromatin, and histone acetylation has been shown to have profound effects on chromatin architecture leading to a decrease in internucleosomal interaction (33, 39). LNCaP and abl cells were transfected with siRNAs targeting FoxA1, MYBL2, or CREB1, and ChIP assays were carried out by using an antibody against acetylated histone H3 (AcH3). Silencing of FoxA1, MYBL2, or CREB1 reduced AcH3 levels at regulatory sites of abl-specific FoxA1 target genes in abl but not LNCaP cells, and the level of AcH3 reduction was similar to CBP [a potent histone acetyltransferase (HAT); positive control] silencing (Fig. 6F; Supplementary Fig. S6C), suggesting that these 3 transcription factors, possibly via recruitment of CBP and other HATs, can significantly induce histone acetylation, leading to nucleosome disruption at FoxA1 target gene regulatory regions specifically in abl cells (Figs. 4D, 5D, and 6G; Supplementary Figs. S4B and S6D).

Finally, we examined the effect of MYBL2 and CREB1 silencing on abl-specific FoxA1 target gene expression. Expression of CCNE2, E2F1, CCNA2, UBE2C, and CDK1 after MYBL2 or CREB1 silencing in LNCaP and abl cells were assessed by real-time RT-PCR. Although silencing of MYBL2 and/or CREB1 decreased CCNE2, E2F1, CCNA2, UBE2C, and CDK1 mRNA levels in LNCaP cells to some extent, in abl cells, knocking down of these 2 transcription factors greatly reduced mRNA expression of these abl-specific FoxA1 target genes (Fig. 6H; Supplementary Fig. S6E). As anticipated, slower growth of siMYBL2- or siCREB1-transfected CRPC cells was observed versus siControl (Fig. 6I; Supplementary Fig. S6F), presumably due to decreased expression of the 5 essential cell-cycle genes, indicating that MYBL2 and CREB1 are critical for abl-specific FoxA1 target gene expression and CRPC cell growth.

In this study, we found that in addition to its known role as an AR collaborator in regulating CRPC-specific AR target G₂–M gene transcription and thus a driver of G₂–M cell-cycle progression (ref. 11; Fig. 1; Supplementary Figs. S4–S6), FoxA1 directs CRPC G₁–S cell-cycle progression through direct regulation of CCNE2, and indirect regulation of CCNA2 via E2F1 (Figs. 2, 4, and 5). These findings establish a previously undescribed yet essential role for FoxA1 as a master cell-cycle regulator, required for G₁–S and G₂–M progression in CRPC, whereas AR is mainly involved in promoting cell-cycle G₂–M but not G₁–S transition in CRPC (Fig. 7; refs. 11, 40).

The findings that an abl-specific FoxA1 cistrome promotes G₁–S and G₂–M cell-cycle progression, through transcriptional regulation of non-AR target genes (CCNE2 and E2F1; Figs. 4 and 5) and AR target genes (UBE2C and CDK1; Supplementary Fig. S4; ref. 11), raise the question: what are the mechanisms controlling differential FoxA1 binding in abl and LNCaP cells? Although it has been established that H3K4me1 and H3K4 dimethylation (H3K4me2) levels determine differential FoxA1 binding in different cell types (refs. 6, 11; Figs. 4C and 5C), whether differential expression and binding of other transcription factors affect FoxA1 binding is unknown. By employing an integrated computational and experimental approach, we identified transcription factors MYBL2 and CREB1 as potential regulators of abl-specific FoxA1 binding. Furthermore, by combining ChIP, siRNA-ChIP, and re-ChIP analyses, we showed codependent FoxA1/MYBL2 and FoxA1/CREB1 binding to the same DNA fragments in regulatory regions of abl-specific FoxA1 target genes CCNE2, E2F1, UBE2C, and CDK1 (Fig. 6; Supplementary Figs. S5 and S6). This codependent chromatin binding is determined by the nonredundant function of FoxA1, MYBL2, and CREB1 in inducing histone H3 acetylation and thus facilitating nucleosome depletion at abl-specific FoxA1 binding regions (Fig. 6F and G; Supplementary Fig. S6C and D). Our findings are consistent with previous in vitro findings that activator-dependent histone H3 acetylation leads to an open chromatin structure through disruption of both inter- and intrafiber internucleosome interactions (41, 42). Thus, although histone modifications and nucleosome depletion are not required for FoxA1 to open in vitro reconstituted condensed chromatin (4), our findings suggest that in vivo FoxA1 binding requires both active histone H3K4 methylation and other collaborating transcription factors capable of inducing histone acetylation and/or nucleosome disruption.

Interestingly, the expression of MYBL2 and CREB1 is not only increased in abl compared with LNCaP (Fig. 6B), but, more importantly, in clinical cases of CRPC versus ADPC (43, 44). By contrast, strong FoxA1 protein expression persists in early and late phases of prostate cancer (Fig. 1A; refs. 45, 46), suggesting that increased expression and binding of MYBL2 and CREB1 during prostate cancer progression alter FoxA1 genomic binding, leading to CRPC-specific upregulation of critical G₁–S and G₂–M cell-cycle genes by FoxA1 (Fig. 7). Therefore, MYBL2 and CREB1 may serve as new therapeutic targets for CRPC.

No potential conflicts of interest were disclosed.

This work was supported by NIH grants R00 CA126160 (to Q. Wang and W. Li) and U54 CA113001 (to Q. Wang, K.P. Nephew, and T.H.M. Huang), DOD grant PC094421 (to W. Li), and the Ohio State University Comprehensive Cancer Center (to Q. Wang and T.H.M. Huang).

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