BAZ2A-mediated repression via H3K14ac-marked enhancers promotes prostate cancer stem cells.

Prostate cancer (PCa) is one of the most prevalent cancers in men. Cancer stem cells are thought to be associated with PCa relapse. Here, we show that BAZ2A is required for PCa cells with a cancer stem-like state. BAZ2A genomic occupancy in PCa cells coincides with H3K14ac-enriched chromatin regions. This association is mediated by BAZ2A-bromodomain (BAZ2A-BRD) that specifically binds H3K14ac. BAZ2A associates with inactive enhancers marked by H3K14ac and repressing transcription of genes frequently silenced in aggressive and poorly differentiated PCa. BAZ2A-mediated repression is also linked to EP300 that acetylates H3K14ac. BAZ2A-BRD mutations or treatment with inhibitors abrogating BAZ2A-BRD/H3K14ac interaction impair PCa stem cells. Furthermore, pharmacological inactivation of BAZ2A-BRD impairs Pten-loss oncogenic transformation of prostate organoids. Our findings indicate a role of BAZ2A-BRD in PCa stem cell features and suggest potential epigenetic-reader therapeutic strategies to target BAZ2A in aggressive PCa.

Introduction

Tumors are composed of heterogeneous populations of cells, which differ in their phenotypic and genetic features (Saygin et al, 2019). It has been suggested that cellular heterogeneity within a tumor is organized in hierarchical manner with a subpopulation of cancer stem cells (CSC) that give rise to heterogeneous cancer cell lineages and undergo self‐renewal to maintain their reservoir (Pattabiraman & Weinberg, 2014). Furthermore, non‐stem, differentiated cancer cells can transit into a dedifferentiated, CSC‐like phenotype (Plaks et al, 2015; Rich, 2016). CSCs were shown to have an enhanced capacity for therapeutic resistance, immune evasion, invasion, and metastasis (Prager et al, 2019). Thus, efficient targeting of CSCs in cancer treatment is critical for developing effective therapeutics.

Prostate cancer (PCa) is the second most common epithelial cancer and the fifth leading cause of cancer‐related death in men worldwide (Bray et al, 2018). PCa displays a high heterogeneity that leads to distinct histopathological and molecular features (Li & Shen, 2018). This heterogeneity posits one of the most confounding and complex factors underlying its diagnosis, prognosis, and treatment (Yadav et al, 2018). Current therapeutic approaches for PCa remain insufficient for some patients with progressive disease. Targeting of the androgen receptor (AR) axis through androgen ablation therapy and/or AR antagonists (castration) in patients with PCa relapse was shown to be efficient only for a short period of time (Denmeade & Isaacs, 2002). These treatments are not curative; a population of cells resistant to androgen‐deprivation therapy emerges and PCa becomes unresponsive and progresses to a castrate‐resistant prostate cancer (CRPC) and metastasis, with limited treatment options (Linder et al, 2018). Due to the enticing possibility that PCa aggressiveness and relapse arises from PCa stem cells (Colombel et al, 2012; Seiler et al, 2013; Mayer et al, 2015), the development of stem cell‐specific anticancer drugs constitutes an attempt to innovate in the treatment of PCa (Leão et al, 2017).

Targeting of bromodomain (BRD)‐containing proteins is a promising therapeutic approach, currently undergoing clinical evaluation in CRPC patients (Welti et al, 2018). Bromodomain‐containing proteins regulate gene expression primarily through recognition of histone acetyl residues, leading to the recruitment of protein complexes that modulate gene expression (Fujisawa & Filippakopoulos, 2017). Bromodomain‐containing proteins are frequently deregulated in cancer, and the development of BRD inhibitors as anticancer agents is now an intense area of research (Fujisawa & Filippakopoulos, 2017). JQ1, an inhibitor of proteins belonging to bromodomain and extraterminal domain (BET) family, has an effect in AR‐signaling‐competent human CRPC cell lines, whereas in AR‐independent cell lines, such as PC3 cells, these inhibitors do not display any effect (Asangani et al, 2014).

BAZ2A (also known as TIP5) is a bromodomain‐containing protein and subunit of the nucleolar‐remodeling complex NoRC (Santoro et al, 2002; Zhou & Grummt, 2005). Recent work implicated BAZ2A in aggressive PCa. BAZ2A is highly expressed in metastatic tumors compared with primary and localized tumors, required for cell proliferation, viability, and invasion, and represses genes frequently silenced in aggressive PCa (Gu et al, 2015). Furthermore, high BAZ2A levels in tumors associate with poor prognosis and disease recurrence. Finally, it was shown that BAZ2A is required for the initiation of PCa driven by PTEN loss, one of the most commonly lost tumor suppressor genes in PCa (Pietrzak et al, 2020).

BAZ2A contains a bromodomain that binds to acetylated histones (Zhou & Grummt, 2005; Tallant et al, 2015). In this work, we show that BAZ2A genomic occupancy in PCa cells coincides with H3K14ac‐enriched chromatin regions. This association is mediated by BAZ2A‐bromodomain, an epigenetic reader of H3K14ac. In PCa cells, BAZ2A‐bromodomain is required for the interaction with a class of inactive enhancers that are marked by H3K14ac and represses the expression of genes implicated in developmental and differentiation processes that are linked to aggressive and dedifferentiated PCa. BAZ2A‐mediated repression of these genes is also linked to the histone acetyltransferase EP300 that acetylates H3K14ac and displays the highest positive correlation with BAZ2A expression in both metastatic and primary tumors compared with the other H3K14 acetyltransferases. Mutations of BAZ2A‐bromodomain or treatment with chemical probes that abrogate BAZ2A‐bromodomain association with H3K14ac impair PCa stem cells. Furthermore, pharmacological inactivation of BAZ2A‐bromodomain impairs the oncogenic transformation mediated by Pten loss in PCa organoids. Our findings indicate that BAZ2A is a key player in PCa stem cell features and suggest potential epigenetic‐reader therapeutic strategies to target BAZ2A‐bromodomain in aggressive, poorly differentiated PCa.

Results

BAZ2A associates with chromatin regions marked by H3K14ac

To determine whether BAZ2A‐bromodomain is implicated in BAZ2A‐mediated regulation of PCa, we investigated its genomic occupancy and the correlation with histone marks in metastatic PCa PC3 cells. We performed ChIPseq analysis of PC3 cells transfected with plasmids expressing HA‐tagged BAZ2A and found 96,351 regions that were bound by BAZ2A (Fig 1A). We validated these results by ChIP‐qPCR through the measurement of the binding of endogenous BAZ2A with two selected genes (AOX1, that as described below is a gene directly regulated by BAZ2A, and EVX2) using a PC3 cell line that expresses endogenous BAZ2A with a FLAG‐HA (F/H) tag at the N‐terminus (Figs 1B and C, and EV1A and B). The correlation between ChIPseq signal enrichment for BAZ2A binding and histone marks in PC3 cells indicated that BAZ2A has the strongest association with H3K14ac whereas its interaction with H3K27ac regions was much weaker (Fig 1A, B and D). To determine whether BAZ2A‐bromodomain specifically recognizes H3K14ac, we used histone peptide arrays containing 384 unique histone modification combinations that allow the analysis of not only individual modified residues but also the effects of neighboring modifications (Figs 1E and EV1C). We incubated purified recombinant human BAZ2A‐bromodomain fused to GST tag with the histone array and monitored the binding to histone peptides using anti‐GST antibodies. Consistent with previous results (Zhou & Grummt, 2005; Tallant et al, 2015), BAZ2A‐bromodomain did not interact with unmodified histone peptides and showed a preferential binding to a specific set of acetylated histone peptides. In line with the ChIPseq analyses, the acetylation per se was not sufficient for the binding of BAZ2A‐bromodomain as evident by the lack of interaction with acetylated histone peptides such as H3K27ac. Interestingly, H3K14ac peptide was the mono‐acetylated histone peptide displaying the highest interaction with BAZ2A‐bromodomain, a result that is consistent with the ChIPseq analysis. Phosphorylation of H3S10 or acetylation of H3K9 improved the interaction of BAZ2A with H3K14ac. Poly‐acetylated histone H4 containing K5ac, K8ac, K12ac, and K16ac was also a good target for BAZ2A‐bromodomain whereas the interaction with these single H4 acetylated residues was much weaker or even absent as in the case of H4K5ac. To further support the specificity of BAZ2A‐bromodomain as reader of H3K14ac, we mutated BAZ2A‐Y1830 and N1873, which recent structural analyses indicated as critical residues for the interaction of BAZ2A‐bromodomain with acetylated histones (Tallant et al, 2015). Accordingly, the interaction of both BAZ2A‐BRDY1830F and BAZ2A‐BRDN1873L with modified histone peptides was strongly reduced, particularly for the mono‐acetylated H3K14ac peptide (Fig 1E).

Figure 1

BAZ2A‐bromodomain is an epigenetic reader of H3K14ac that mediates the BAZ2A association with H3K14ac‐chromatin domains in PC3 cells

Heatmap profiles of BAZ2A‐bound regions in PC3 cells and the corresponding signals of input, BAZ2A‐bromodomain (BRD) mutant (Y1775F), H3K14ac, and H3K27ac. Data were ranked to BAZ2A binding in PC3 cells.

Validation of ChIPseq results by anti‐HA ChIP‐qPCR of PC3 wild‐type (WT) and a PC3 cell line that expresses endogenous BAZ2A with a FLAG‐HA tag (PC3‐H/F‐BAZ2A). GAPDH promoter represents a region not bound by BAZ2A. Average values of three independent experiments. Data were normalized to input and to AOX1 levels. Statistical significance (P‐values) was calculated using two‐tailed t‐test (*< 0.05, **< 0.001, ns: not significant). Error bars represent SD.

Wiggle tracks displaying BAZ2A‐bound regions, H3K14ac, H3K27ac, and input at AOX1, EVX2, and GAPDH.

Pearson correlation heat map of BAZ2A‐bound regions and histone marks. Data for BAZ2A, H3K4me3, H3K27me3, H3K27ac, and H3K14ac are from this work. H3K9me3, H3K9me2, and H3K4me1 were obtained from ENCODE.

Images of histone peptide arrays probed with 10 nM of recombinant BAZ2A‐BRD wild‐type (GST‐BAZ2A‐BRDwt) and mutants (GST‐BAZ2A‐BRDY1830F or GST‐BAZ2A‐BRDN1873L). Visualization of binding was performed by incubation with anti‐GST antibodies and imaged on Odyssey Infrared Imaging System. Blue circles mark peptides recognized by BAZ2A‐BRD (i.e., H3K14ac), whereas black circles show some of the acetylated peptides not recognized by BAZ2A‐BRD (H3K27ac and H3K9ac). Right panel shows heatmap of the relative binding intensity of BAZ2A‐BRD against modified histone peptides. Binding intensity was calculated as average of fold change from peptides signal over background controls of two different arrays.

BAZ2A read coverage quantification at ± 1 Kb from peak summit of BAZ2A‐bound regions identified in PC3 cells and the corresponding reads for BAZ2A‐BRD mutant (Y1775F) in PC3 cells. Statistical significance (P‐values) was calculated using two‐tailed t‐test (***< 0.001).

Regions bound by BAZ2A in PC3 cells are enriched in H3K14ac. Read coverage quantification for H3K14ac levels at regions bound by BAZ2A in PC3 cells (BAZ2Awt and BAZ2A‐BRD mutant [Y1775F]) at ± 1 Kb from BAZ2A peak summits. Statistical significance (P‐values) was calculated using two‐tailed t‐test (***< 0.001).

Data information: Boxplots in Fig 1 show the median and quartiles with the whiskers extending to the most extreme data point within 1.5 times the interquartile range.

Figure EV1

BAZ2A‐bromodomain is an epigenetic reader of H3K14ac

Schematic representation of the strategy to generate PC3 cell line with FLAG‐HA (F/H) tag insertion at BAZ2A locus by CRISPR/Cas9 method. Lower panel. PCR genotyping to measure the insertion of F/H sequences into BAZ2A locus. Arrows represent the primers used for PCR genotyping.

Immunoblot of PC3 wild‐type (wt) cells and the PC3 cell line expressing endogenous BAZ2A with F/H tag showing comparable levels of BAZ2A. SNF2h and STAT1 serve as protein loading control.

Representative protein gel showing the purity of recombinant GST‐BAZ2A‐BRD wild‐type [WT]) and mutants (GST‐BAZ2A‐BRDY1830F or GST‐BAZ2A‐BRDN1873L). Proteins are visualized by Coomassie staining.

Western blot showing equal expression of ectopically expressed HA‐BAZ2Awt and HA‐BAZ2A‐BRD mutant (Y1775F) in transfected PC3 cells and endogenous levels of BAZ2A in untransfected PC3 cells. Whole cell lysates from equivalent amounts of cells were analyzed. Nucleolin is shown as a protein loading control.

Average density plots of ChIPseq read counts of BAZ2A (BAZ2Awt and BAZ2A‐BRD mutant, Y1775F) and H3K14ac at ± 1 Kb from BAZ2A peak summits in PC3 cells.

Source data are available online for this figure.

To determine whether BAZ2A‐bromodomain mediates the association of BAZ2A with chromatin, we performed ChIPseq analysis of PC3 cells expressing murine BAZ2A mutated at the bromodomain HA‐BAZ2AY1775F, an orthologous to the human BAZ2A‐BRDY1830F unable to bind to acetylated histones (Zhou & Grummt, 2005). Equal expression of HA‐BAZ2Awt and HA‐BAZ2AY1775F was assessed by Western blot (Fig EV1D). We found that only 1.3% (1534 peaks) of BAZ2A‐bound sites were also occupied by BAZ2AY1775F, indicating that a functional bromodomain domain is required for BAZ2A binding to chromatin in PC3 cells (Figs 1A and F, and EV1E). Furthermore, the few DNA regions that retained binding with BAZ2A‐bromodomain mutant contained low H3K14ac levels (Figs 1G and EV1F). Together, these results suggest that BAZ2A is an epigenetic reader of H3K14ac. Moreover, they show that in PC3 cells BAZ2A preferentially binds to regions marked by H3K14ac and that this interaction is mediated by BAZ2A‐bromodomain.

BAZ2A binds to inactive enhancers marked by H3K14ac and mediates the repression of genes linked to developmental and differentiation processes

Previous studies showed that acetylation of H3K14 is mediated by histone acetyltransferases (HATs) EP300, KAT2A (GCN5), or KAT6A (Myst3) (Lee & Workman, 2007; Jin et al, 2011). Interestingly, we found that all these HATs are highly expressed in metastatic prostate cancer (Fig EV2A). In PC3 cells, however, only EP300 and KAT6A are expressed whereas KAT2A sequences are deleted (Seim et al, 2017) (Fig EV2B). The expression profile of these HATs in a large cohort of primary PCa and metastatic CRPC (Cancer Genome Atlas Research Network, 2015; Robinson et al, 2015) revealed that the levels of EP300 had the highest positive correlation with BAZ2A expression in both metastatic and primary tumors compared with KAT2A and KAT6A, suggesting that EP300 might be functionally related to BAZ2A binding to H3K14ac chromatin and the regulation of gene expression in PCa (Figs 2A and EV2C). Treatment of PC3 cells with A‐485, a selective catalytic EP300 inhibitor (Lasko et al, 2017), induced a decrease in H3K14ac levels, confirming the role of EP300 in acetylating H3K14 (Fig 2B). To test the functional link between BAZ2A and EP300, we initially performed genomic annotation analyses of the BAZ2A‐bound sites obtained with the BAZ2A‐ChIPseq analysis. We found that BAZ2A peaks were mainly located in intronic or intergenic regions, whereas only a small fraction (11%) was located at gene promoters (Fig 2C). To determine which of the genes with BAZ2A‐bound promoter transcriptionally depend on BAZ2A, we performed RNAseq of PC3 cells depleted of BAZ2A by siRNA. Consistent with previous results (Gu et al, 2015), BAZ2A‐KD affects gene expression in PC3 cells (log2 fold change > ± 1; Padj < 0.05; 535 upregulated genes, 819 downregulated genes; Fig 2D, Dataset EV1). We applied high stringency parameters to define genes with BAZ2A‐bound promoter (527 genes with BAZ2A peak within ± 2 kb from transcription start and BAZ2A fold change over input ≥ 3). However, only a minority of these genes were affected upon BAZ2A depletion (30 up‐ and 25 downregulated; Dataset EV2). Although the interaction of BAZ2A with the promoter does not appear the most prominent mechanism for BAZ2A‐mediated regulation in PC3 cells, we found that depletion of EP300 by siRNA significantly reactivated the expression of AOX1, DDB1, MANB2, and DMPK (Fig 2E and F). Interestingly, these are the only upregulated genes upon BAZ2A‐KD with BAZ2A‐bound promoter that are significantly lower expressed in metastatic PCa compared with normal prostate tissue (Figs 2E and EV2D). Furthermore, low expression of AOX1 has recently been correlated with shorter time to biochemical recurrence of PCa (Li et al, 2018). These results indicate that EP300 is required for BAZ2A‐mediated repression of a set of genes with BAZ2A‐bound promoter that are implicated in metastatic PCa.

Figure EV2

BAZ2A binds a class of inactive enhancers marked by H3K14ac and represses genes implicated in aggressive PCa

The H3K14ac histone acetyltransferases EP300, KAT2A, and KAT6A are highly expressed in metastatic prostate cancer tissues. Boxplots showing expression profiles from Gene Expression Omnibus (GEO) data sets GSE6919 (gene expression microarray data) (Yu et al, 2004; Chandran et al, 2007). *< 0.05, **< 0.01, ****P < 0.0001, two‐tailed Student's t‐test; ns, not significant. Boxplots show the median and quartiles with the whiskers extending to the most extreme data point within 1.5 times the interquartile range.

Wiggle tracks showing RNAseq expression profiles of EP300, KAT2A, and KAT6A in PC3 cells. KAT2A is not expressed in PC3 cells due to deletion of the entire locus.

Scatter plot showing the expression of KAT2A and KAT6A vs BAZ2A in two large cohorts of metastatic and primary PCa. Data were from (Cancer Genome Atlas Research Network, 2015; Robinson et al, 2015).

Boxplots showing expression profiles of AOX1, MANB2, DMPK, and DDB1 from GEO data sets GSE6919 (gene expression microarray data) (Yu et al, 2004; Chandran et al, 2007). **< 0.01, ****P < 0.0001, two‐tailed Student's t‐test; ns, not significant. Boxplots show the median and quartiles with the whiskers extending to the most extreme data point within 1.5 times the interquartile range.

Boxplots showing expression of BAZ2A‐bound C2‐enhancer genes KRT8, EVC, ITPKB, SH3BP4, and RASD1 in dedifferentiated and aggressive primary tumors scored with Gleason 9 and 10 and indolent tumors assigned with Gleason 6. Data are from TCGA_PRAD (Cancer Genome Atlas Research Network, 2015). Statistical significance (P‐values) was calculated using two‐tailed t‐test (*< 0.05, ***< 0.001). Boxplots show the median and quartiles with the whiskers extending to the most extreme data point within 1.5 times the interquartile range.

Figure 2

EP300‐mediated acetylation of H4K14 mediates the repression of a set of genes with BAZ2A‐bound promoter

Scatter plot showing the expression of EP300 and BAZ2A in two large cohorts of metastatic and primary PCa. Data were from Cancer Genome Atlas Research Network (2015); Robinson et al (2015).

EP300 acetylates H3K14 in PC3 cells. Western blot showing H3K14ac levels in PC3 cells treated for 3 days without and with the selective catalytic EP300 inhibitor A‐485 (10 μM). Histone H3 serves as protein loading control.

Genomic annotations of BAZ2A‐bound regions in PC3 cells.

Volcano plot showing fold change (log2 values) in transcript levels of PC3 cells upon BAZ2A knockdown. Gene expression values of two replicates were averaged and selected for 1.5‐fold changes and P < 0.05.

Table showing expression changes in genes with BAZ2A‐bound promoter upregulated upon BAZ2A‐KD in PC3 cells in metastatic PCa vs normal prostate tissue and in PC3 cells depleted of EP300 via siRNA.

qRT–PCR showing increased expression levels of genes with BAZ2A‐bound promoter and downregulated in metastatic tumors (AOX1, MANB2, DMPK, and DDB1—see (E)) in PC3 cells treated with siRNA‐EP300.

Data information: Data were from three independent experiments. Values were normalized to GAPDH mRNA. Statistical significance (P‐values) was calculated using two‐tailed t‐test (*<0.05, ****< 0.0001). Error bars represent SD.

Source data are available online for this figure.

Given that a large fraction of BAZ2A‐bound regions are intronic and intergenic, we asked whether BAZ2A could regulate gene expression through its association with enhancers. First, we classified annotated enhancers (Plaschkes et al, 2017) at intergenic regions according to histone marks and chromatin accessibility and found that a large fraction of H3K14ac peaks in PC3 cells are located at active enhancers marked by H3K27ac, H3K4me1, and DNase I hypersensitive sites (from here on named as enhancer class 1, C1; Fig 3A). However, this class of active enhancers was not bound by BAZ2A. Interestingly, we found that BAZ2A only associates with a class of enhancers (named as C2) that relative to C1 enhancers had higher H3K14ac and H3K27me3 levels, lower content of H3K27ac, and lacked H3K4me1 and DNase I hypersensitive sites (Fig 3A and B). These results suggest that in PC3 cells, BAZ2A binds to a class of inactive enhancers that is enriched in H3K14ac. To determine whether the binding of BAZ2A to C2 enhancers affects gene expression, we analyzed the expression levels of genes in the nearest linear proximity to BAZ2A‐bound C2 enhancers upon BAZ2A‐KD in PC3 cells (Fig 3C and D, Dataset EV3). We found that 56% of these genes (417 out of 747) were significantly affected in their expression upon BAZ2A depletion. Furthermore, the majority of these BAZ2A‐regulated genes (254 genes, 61%) showed upregulation upon BAZ2A‐KD, suggesting a major role of BAZ2A in repressing gene transcription through its interaction with C2 enhancers. Interestingly, the top 10 gene ontology (GO) terms of these BAZ2A‐regulated genes were strongly linked to developmental and differentiation processes (Fig 3E). Histological grading from well‐differentiated to poorly differentiated PCa measured by Gleason score marks the progression from low‐ to high‐grade cancer (Ahmed et al, 2012). We set to identify BAZ2A‐bound C2 enhancer regulated genes that were upregulated upon BAZ2A‐KD and lower expressed in Gleason 9 and 10 tumors (advanced) compared with Gleason 6 tumors (indolent). Using these criteria, we found 25 genes (Figs 3F and EV2E). Since GO terms associated with these genes were linked to developmental and cellular growth pathways, we selected five genes (KRT8, EVC, ITPKB, SH3BP4, and SUN2) that were implicated in these processes and analyzed their expression upon EP300 depletion in PC3 cells (Fig 3G and H). We found that EP300‐KD increased the expression of all the selected genes, a result that further supports the functional link between BAZ2A and EP300 in the regulation of gene expression in PCa cells.

Figure 3

The association of BAZ2A with a class of inactive enhancers marked by H3K14ac represses genes implicated in aggressive PCa

BAZ2A binds to a class of inactive enhancers marked by H3K14ac. Heat maps of BAZ2A, H3K14ac, H3K27ac, H3K4me1, and DNAseq signal at annotated intergenic enhancers are shown. Enhancer regions were clustered in 4 groups based on the presence (+) or absence (−) of H3K27ac and H3K4me1. C1 (H3K27ac+/H3K4me1+), C2 (H3K27ac+/H3K4me1−), C3 (H3K27ac−/H3K4me1+), and C4 (H3K27ac−/H3K4me1−).

Inactive enhancers (C2) are enriched in H3K14ac and BAZ2A levels. Read coverage at enhancer classes C1 and C2 for BAZ2A, H3K14ac, H3K27ac, H3K4me1, and H3K27me3 at ± 1 Kb from annotated enhancer regions. Statistical significance (P‐values) was calculated using two‐tailed t‐test (***< 0.001). Boxplots show the median and quartiles with the whiskers extending to the most extreme data point within 1.5 times the interquartile range.

Pie charts showing the number of genes in the nearest linear genome proximity to BAZ2A‐bound C2 enhancers and their expression changes upon BAZ2A‐KD in PC3 cells.

Wiggle tracks showing RNA levels in PC3 cells treated with siRNA‐BAZ2A and occupancy of BAZ2A, H3K14ac, H3K27ac, and H3K4me1 at C2 enhancer and its nearest gene CXCR4.

Top 10 biological process gene ontology (GO) terms as determined using DAVID 6.8 for genes in the nearest linear proximity to BAZ2A‐bound C2 enhancers that are differentially expressed upon BAZ2A‐KD in PC3 cells.

Heatmap showing expression changes in BAZ2A‐bound C2‐enhancer genes that were both significantly upregulated upon BAZ2A‐KD and lower expressed in Gleason 9 and 10 tumors (advanced) compared with Gleason 6 tumors (indolent). Values from tumors were from Cancer Genome Atlas Research Network (2015).

GO terms associated with the genes listed in (F). Genes labeled in bold were analyzed in (H).

qRT–PCR showing increased expression levels of BAZ2A‐bound C2‐enhancer genes in PC3 cells treated with siRNA‐EP300.

Data information: Data were from three independent experiments. Downregulation of EP300 expression is shown in Fig 2F. Values were normalized to GAPDH mRNA. Statistical significance (P‐values) was calculated using two‐tailed t‐test (*< 0.05, **< 0.001). Error bars represent SD.

Collectively, these results suggest that BAZ2A binds to inactive C2 enhancers containing H3K14ac and mediates the repression of genes linked to developmental and differentiation processes that are frequently silenced in aggressive and poorly differentiated tumors.

BAZ2A‐bromodomain is required for PCa stem cells

Several studies have reported that non‐stem, differentiated cancer cells can transit into a dedifferentiated, CSC‐like phenotype (Plaks et al, 2015; Rich, 2016). The role of BAZ2A in modulating the expression of genes linked to developmental processes and frequently silenced in aggressive and poorly differentiated tumors prompted us to determine whether BAZ2A plays a role in PCa stem cells and whether this is mediated by BAZ2A‐bromodomain. To study this, we analyzed PCa stem cells in PC3 cells by measuring the amounts of tumorspheres generated using serum‐free medium and low attachment culture conditions (Li et al, 2008; Rajasekhar et al, 2011; Sheng et al, 2013; Portillo‐Lara & Alvarez, 2015). The PCa sphere‐formation assay is an in vitro method commonly used to identify cancer stem cell. In particular, PC3 cells were shown to contain a small fraction (∼1%) of cells that can generate tumorspheres (Li et al, 2008; Sheng et al, 2013) and become significantly more clonogenic, chemoresistant, and invasive compared with the heterogeneous PC3 cell population (Civenni et al, 2013; Portillo‐Lara & Alvarez, 2015). Tumorspheres isolated from PC3 cells (PC3‐CSCs) could be serially passaged in vitro, and when plated in adherent growth conditions, they could differentiate into monolayer cultures morphologically indistinguishable from the original PC3 cells (Fig EV3A). To further characterize the obtained PC3‐CSCs, we performed RNAseq analysis and found that 1,906 genes display transcriptional changes (log2 fold change > 1; P < 0.05; 922 upregulated and 984 downregulated in PC3‐CSCs vs PC3 cells; Fig 4A, Dataset EV4). GO analysis indicated that genes downregulated in PC3‐CSCs were associated with developmental processes (i.e., HOXA1, HOXA4, SOX7), reflecting a more dedifferentiated state compared with the parental PC3 cells (Fig 4B and C). Furthermore, gene set enrichment analysis (GSEA) using predefined embryonic stem cell (ESC)‐like gene signatures showed that genes upregulated in PC3‐CSCs were enriched in adult stem cell signatures (Wong et al, 2008) and displayed a transcription profile similar to genes highly expressed in human embryonic stem cells compared with differentiated cells (Ben‐Porath et al, 2008; Wong et al, 2008) (Fig EV3B). Importantly, among the upregulated genes we found markers associated with CSCs such as LGR5, CD24, Nestin, EpCAM, and ALDH1A2 (Raha et al, 2014; Cao et al, 2017; Li et al, 2017) (Figs 4C and EV3C). Taken together, these results indicated that tumorspheres derived from PC3 cells have a gene signature that resemble the molecular signature of a cancer stem cell‐like phenotype.

Figure EV3

PC3‐CSCs display the molecular signature of a cancer stem cell‐like phenotype

Representative images showing PC3 cells, the derived tumorspheres PC3‐CSCs and PC3 cells generated from PC3‐CSCs under adherent cell culture conditions.

Genes upregulated in PC3‐CSCs are enriched in stem cell‐like signature. GSEA analysis was obtained by comparing expression values of tumorspheres vs PC3 cells of genes upregulated in PC3‐CSCs (P < 0.05, log2FC > 1). P was calculated over 100,000 permutations based on the phenotype, and FDR was < 25%. NES: normalized enrichment score.

Representative wiggle tracks from RNAseq of PC3 and PC3‐CSCs of cancer stem cell markers that are upregulated in PC3‐CSCs compared with the parental PC3 cells.

Figure 4

BAZ2A‐bromodomain is required for PCa stem cells

RNAseq showing differential gene expression between PC3‐CSCs and the parental PC3 cells (P < 0.05, Log2 fold change 1).

Top 10 biological process gene ontology (GO) terms as determined using DAVID 6.8 for genes downregulated and upregulated in PC3‐CSCs relative to parental PC3 cells.

qRT–PCR of genes differentially expressed in PC3‐CSCs compared with PC3 cells showing the upregulation of CSC markers (ALDH1A2, LGR5, CD24) and downregulation of genes related to developmental processes (HOXA1, HOXA4, SOX7). Data are from two independent experiments. mRNA levels were normalized to GAPDH.

BAZ2A is required for the dedifferentiation of PC3 into PC3‐CSCs. Left panel: Representative images from two independent experiments showing the impairment of PC3 cells to form tumorspheres upon stable depletion of BAZ2A. Right panel: BAZ2A mRNA levels in three independent PC3 cell lines stably expressing shRNA‐BAZ2A were measured by qRT–PCR and normalized to GAPDH mRNA from 2 independent experiments.

Left panel: qRT–PCR showing BAZ2A mRNA levels in PC3 cells depleted of endogenous BAZ2A with siRNA specifically targeting human BAZ2A sequences and expressing mouse BAZ2Awt and BAZ2A‐BRD mutant (BAZ2AY1775F). Values were normalized to MRPS7 mRNA. Data are from three independent experiments. Right panel: Quantification of tumorspheres from three independent experiments. Statistical significance (P‐values) was calculated using two‐tailed t‐test (*< 0.05; **< 0.01). Error bars represent SD.

To determine whether BAZ2A is implicated in CSCs, we established three PC3 cell lines stably expressing shRNA against BAZ2A sequence (Fig 4D). Remarkably, all shRNA‐BAZ2A PC3 cell lines were unable to generate PC3‐CSCs. Similarly, treatment of PC3 cells with siRNA‐BAZ2A reduced the number of tumorspheres (Fig 4E). This phenotype could be rescued upon ectopic expression of mouse BAZ2A, which is not targeted by the siRNA‐BAZ2A. Importantly, the expression of the bromodomain‐mutant mBAZ2AY1775F abrogated the formation of tumorspheres, indicating that a functional BAZ2A‐bromodomain is critical for the CSC‐like properties of PC3 cells (Fig 4E). Collectively, these results indicate that a functional BAZ2A‐bromodomain is required for PC3‐CSCs and supported a role of BAZ2A in driving aggressive and poorly differentiated PCas.

Pharmacological inhibition of BAZ2A‐bromodomain impairs PCa stem‐like cells and the oncogenic transformation mediated by Pten‐loss

The requirement of a functional BAZ2A‐bromodomain for PC3‐CSCs prompted us to test the possibility to inhibit BAZ2A‐bromodomain for targeting CSCs as a therapeutic strategy. Recently, two chemical probes, GSK2801 and BAZ2‐ICR, have been generated for specific targeting of bromodomains of BAZ2A and the closely related BAZ2B (Chen et al, 2015; Drouin et al, 2015), which display 65% sequence similarity (Tallant et al, 2015). Both compounds are highly specific for BAZ2A‐ and BAZ2B‐bromodomains and show half‐maximal inhibitory concentration (IC50) in the nanomolar range (Chen et al, 2015; Drouin et al, 2015). To determine whether BAZ2‐ICR or GSK2801 destabilizes BAZ2A‐bromodomain interaction with histones, we incubated recombinant BAZ2A‐bromodomain with both compounds and measured BAZ2A‐bromodomain binding to modified histones with the histone peptide array (Figs 5A and EV4A). Both BAZ2‐ICR and GSK2801 abolished the interaction of BAZ2A‐bromodomain with histones, although GSK2801 inhibition was weaker than BAZ2‐ICR.

Figure 5

Pharmacological targeting of BAZ2A‐bromodomain impairs the transition of PCa cells into a dedifferentiated, cancer stem‐like state

Heatmaps showing the relative BAZ2A‐BRD binding intensity at modified histone peptides upon incubation with BAZ2A‐BRDi BAZ2‐ICR or GSK2801 displayed in Fig EV4A. Binding intensity was calculated as average of fold change from peptides signal over background controls of two different arrays.

Cell proliferation curves of PC3 cells treated with siRNA‐BAZ2A. Error bars represent the standard deviation from three independent experiments.

Cell proliferation curves of PC3 cells treated with 5 μM BAZ2‐ICR and GSK2801. Error bars represent the standard deviation from three independent experiments.

Representative images of tumorspheres generated from PC3 cells treated with BAZ2‐ICR (1 μM) or GSK2801 (1 μM).

Half‐maximal effective concentration (EC50) of BAZ2‐ICR. Measurements were performed 5 days after treatment of PC3 cells and initiation of tumorspheres using the indicated BAZ2‐ICR concentration. The amount of tumorspheres was assessed by measurement of cell viability. Error bars represent SD. Values are from three independent experiments.

Representative images showing tumorspheres derived from the metastatic PCa cell line DU‐145 treated with DMSO or BAZ2‐ICR (1 μM) inhibitor for 5 days.

Figure EV4

BAZ2A‐bromodomain inhibitors impair PCa cells with a cancer stem‐like state

Images of histone peptide arrays incubated with 10 nM of recombinant GST‐BAZ2A‐BRDwt in the presence of 50 nM of BAZ2A‐BRDi, BAZ2‐ICR, or GSK2801. Visualization of binding was performed by incubation with anti‐GST antibodies and imaged on Odyssey Infrared Imaging System. For a better visualization of the data, the image of GST‐BAZ2A‐BRDwt without BAZ2A‐BRDi shown in Fig 1D has been included. Blue circles mark peptides recognized by BAZ2A‐BRD (i.e., H3K14ac), whereas black circles show some of the acetylated peptides not recognized by BAZ2A‐BRD (H3K27ac and H3K9ac).

qRT–PCR showing increased expression levels of BAZ2A‐regulated genes in PC3 cells treated with DMSO or 1 μM BAZ2‐ICR for 5 days. Values were normalized to GAPDH mRNA levels. Average values of two independent experiments. Gray and black dots represent the values obtained from each single experiment.

Western blot showing similar levels of BAZ2A and H3K14ac in PC3 cells treated without or with 1 μM BAZ2‐ICR for 5 days. Histone H3 was used to ensure similar protein levels.

Cell proliferation curves of PC3 cells treated with 5, 10, and 50 μM BAZ2‐ICR. Error bars represent the standard deviation from three independent experiments.

Representative wiggle tracks from RNAseq of PC3 and PC3‐CSCs of BAZ2A and BAZ2B.

BAZ2B depletion in PC3 cells does not affect the dedifferentiation into PC3‐CSCs. Left panel: BAZ2B mRNA levels were measured by qRT–PCR in PC3‐CSCs and normalized to GAPDH mRNA. Right panel: Representative figures showing PC3‐CSCs generated from PC3 cells transfected with siRNA‐control or siRNA‐BAZ2B.

Source data are available online for this figure.

Next, we tested the possibility to impair BAZ2A function in PCa cells through pharmacological targeting of BAZ2A‐bromodomain. Treatment of PC3 cells with BAZ2‐ICR upregulates the expression of several BAZ2A‐regulated genes such as AOX1, DDB1, SH3BP4, and EVC without affecting BAZ2A and H3K14ac levels (Fig EV4B and C). Previous results showed that BAZ2A depletion in several metastatic PCa cells, including PC3 cells, impairs cell proliferation (Fig 5B) (Gu et al, 2015). Indeed, all our attempts to isolate BAZ2A‐KO PC3 cell lines through CRIPS/Cas9 failed, indicating the requirement of BAZ2A expression in metastatic PCa cells. Surprisingly, treatment of PC3 cells with BAZ2‐ICR or GSK2801 at concentrations up to 50 μM did not affect cell proliferation (Figs 5C and EV4D), suggesting that BAZ2A‐bromodomain is not functionally relevant in the proliferation of the heterogeneous PC3 cell population. In contrast, treatment with BAZ2‐ICR or GSK2801 from the beginning of the culture of PC3 cells in low adhesion plate drastically reduced the number of tumorspheres (EC50 0.83 μM; Fig 5D and E), indicating that pharmacological targeting of BAZ2A‐bromodomain impairs PC3‐CSCs. The different effects of BAZ2A inhibitors did not depend on different BAZ2A expression levels since BAZ2A is expressed at similar levels in both PC3 and tumorspheres (Fig EV4E). The abrogation of tumorsphere formation upon treatment with BAZ2A‐bromodomain inhibitors could also be observed in another metastatic PCa cell line, DU‐145 (Fig 5F). These results are consistent with the data showing the tumorspheres require a functional BAZ2A‐bromodomain (Fig 4E). Since BAZ2‐ICR or GSK2801 can also target BAZ2B‐bromodomain (Chen et al, 2015; Drouin et al, 2015), we tested the contribution of BAZ2B in PC3‐CSCs by transfecting PC3 cells with siRNA directed against BAZ2B that is similarly expressed in both PC3 and PC3‐CSCs (Fig 4E and F). We found no detectable defects compared with control cells, indicating that pharmacological targeting of BAZ2A‐bromodomain has a major function in PCa cells with CSC‐like state.

Next, we analyzed the role of BAZ2A‐bromodomain in PCa initiation using mouse prostate organoids to model PCa driven by the loss of Pten, one of the most commonly lost tumor suppressor gene in PCa (Yoshimoto et al, 2006; Berger et al, 2011; Phin et al, 2013; Karthaus et al, 2014; Cancer Genome Atlas Research Network, 2015) (Fig 6). PTEN was also shown to have a pivotal role in CSCs by regulating pathways critical for stem cell maintenance, homeostasis, self‐renewal, and migration (Wang et al, 2006; Mulholland et al, 2009; Luongo et al, 2019). Furthermore, recent results showed that BAZ2A is required for the initiation of PCa driven by PTEN‐loss (Pietrzak et al, 2020). Downregulation of Pten was achieved by infection of prostate organoids with lentivirus encoding shRNA‐Pten, which allowed 80% global reduction in Pten levels (Fig EV5A). Consistent with previous reports (Karthaus et al, 2014; Pietrzak et al, 2020), compared to control organoids, shRNA‐Pten organoids displayed altered morphology as evident by the solid and compact structure and elevated expression of nuclear proliferation marker Ki67, without altering the number of organoids (Fig 6A–E). Remarkably, treatment with BAZ2A‐bromodomain inhibitors BAZ2‐ICR or GSK2801 followed by transduction with shRNA‐control or shRNA‐Pten impaired Pten‐loss mediated transformed phenotype as evident by the translucent phenotype, the intact bilayer structure, and low Ki67 signal whereas EZH2 inhibitor GSK126 did not show any significant effect (Fig 6A–D). These results are consistent with previous data showing that BAZ2A is required for the initiation of PCa driven by Pten‐loss (Pietrzak et al, 2020) and highlighted an important role of BAZ2A‐bromodomain in this process. Collectively, these data indicate that pharmacological targeting of BAZ2A‐BRD abolishes PCa cells with CSC‐like state and the oncogenic transformation mediated by Pten‐loss.

Figure 6

Pharmacological targeting of BAZ2A‐bromodomain impairs the initiation of PCa driven by Pten loss in a model of PCa organoids

Representative brightfield images of organoids derived from mouse prostate cells treated with DMSO, BAZ2‐ICR (5 μM), GSK2801 (5 μM), or GSK126 (5 μM) inhibitors followed by transduction with shRNA‐control or shRNA‐Pten.

Quantification of shRNA‐control and shRNA‐Pten organoid morphology upon treatment with DMSO, EZH2 (GSK126), and BAZ2A‐BRD inhibitors (BAZ2‐ICR and GSK2801).

Representative images of Ki67‐immunostained mouse prostate organoid sections expressing shRNA‐control and shRNA‐Pten and treated with DMSO or BAZ2‐ICR (5 μM).

Quantification of Ki67+ cells in mouse prostate organoids shown in (C). **< 0.01, ****P < 0.0001, two‐tailed Student's t‐test; ns, not significant. The median values are represented by a line.

Quantification of mouse prostate organoids per field expressing shRNA‐control and shRNA‐Pten and treated with DMSO, GSK126 (5 μM), BAZ2‐ICR (5 μM), or GSK2801 (5 μM). Values were normalized to shRNA‐control samples. Average values of two independent experiments.

Figure EV5

BAZ2A‐bromodomain inhibitors impair the initiation of PCa driven by Pten‐loss in a model of PCa organoids

qRT–PCR showing Pten mRNA levels in mouse prostate organoids expressing shRNA‐control and shRNA‐Pten and treated with DMSO, GSK126 (5 μM), BAZ2‐ICR (5 μM), or GSK2801 (5 μM). Values are from two experiments. Data were normalized to Hprt mRNA and organoids treated with siRNA‐Control.

Discussion

Previous work showed that BAZ2A, a component of the chromatin remodeling complex NoRC, is implicated in aggressive PCa (Gu et al, 2015; Pietrzak et al, 2020). In this study, we showed that BAZ2A‐bromodomain is required for PCa cells with a cancer stem‐like state and the initiation of PCa driven by PTEN‐loss. We showed that BAZ2A‐bromodomain is an epigenetic reader of H3K14ac and is required for the association of BAZ2A with a class of inactive enhancers that are marked by H3K14ac. We showed that the expression of a large fraction of genes in the nearest linear genomic proximity to these BAZ2A‐bound enhancers depends on BAZ2A. These genes are implicated in developmental and differentiation pathways, which are usually altered in aggressive and dedifferentiated PCa. The expression analysis of a set of genes that are repressed by BAZ2A and frequently silenced in advanced and poorly differentiated tumors relative to indolent tumors showed the requirement of EP300, one of the writers of H3K14ac. The link of H3K14ac with gene silencing is also consistent with previous studies showing that the transcription regulator ZMYND8, the tumor suppressor PBRM1, and H3K9 methyltransferase SETDB1 can recognize H3K14ac and repress gene transcription (Li et al, 2016; Jurkowska et al, 2017; Liao et al, 2019). The interaction of BAZ2A‐bromodomain with H3K14ac is also of particular interest since H3K14ac was found to be the most prominent histone modification in regulating SNF2H/ISWI nucleosome remodeling and activating only the NoRC complex (Dann et al, 2017). Thus, it is possible that H3K14ac is not only acting for the recruitment of BAZ2A to chromatin but also for the remodeling of nucleosomes that might play a role in the regulation of gene expression. A BAZ2A‐mediated remodeling at nucleosomes marked with H3K14ac might therefore be part of the mechanisms to silence gene expression and future studies will address this point.

Cancer stem cells represent a subpopulation within tumors that has an enhanced capacity for therapeutic resistance and thought be implicated in relapse of many tumors' types, including PCa (Flemming, 2015; Mayer et al, 2015; Zhang et al, 2017). Further, cancer cells can transition between stem and differentiated states in response to therapeutic insults or other stimuli within the microenvironment (Plaks et al, 2015; Rich, 2016). Thus, efficient targeting of cancer stem cells in cancer treatment is critical for developing effective therapeutics. In this work, we used an established method to measure the ability of PCa cells to form tumorspheres (Rajasekhar et al, 2011; Portillo‐Lara & Alvarez, 2015). We showed that BAZ2A‐bromodomain is required for PCa cells with cancer stem‐like properties, suggesting that targeting BAZ2A with BAZ2A‐bromodomain inhibitors in tumors should be beneficial in impairing cancer cells with a stem‐like state, thereby decreasing tumor relapse potential.

Current therapeutic approaches for PCa remain insufficient for some patients with progressive disease. Targeting AR axis can manage PCa relapses only for a limited time period since PCas become unresponsive and progress to a castrate‐resistant form with limited treatment options (Linder et al, 2018). The development of PCa stem cell‐specific anticancer drugs constitutes an attempt to innovate the treatment of PCa (Leão et al, 2017). Previous results showed that AR‐signaling‐competent human CRPC cell lines are preferentially sensitive to BET inhibitors JQ1 (Asangani et al, 2014). However, inhibition of BET‐BRD does not have an effect in AR‐independent cell lines such as PC3 cells (Asangani et al, 2014). Our study demonstrated that BAZ2A‐bromodomain function can be pharmacologically targeted through BAZ2A‐BRD inhibitors in CRPC cells and proposed that inhibition of BAZ2A‐bromodomain can be a potential strategy to use against cancer stem cells in PCa.

Materials and Methods

Culture of PC3 and PCa spheres

All the indicated cell lines were purchased from the American Type Culture Collection. PC3 cells were cultured in RPMI 1640 medium and Ham's F12 medium (1:1; Gibco) containing 10% FBS (Gibco) and 1% penicillin–streptomycin (Gibco). DU‐145 were cultured in RPMI 1640 medium containing 10% FBS and 1% penicillin–streptomycin. PCa spheres were cultured in serum‐free DMEM/F12 medium (1:1, Gibco) supplemented with 20 ng/ml of basic human FGF (Sigma), 20 ng/ml of EGF (Sigma), 3 μg/ml of Insulin (Sigma), and 1× B27 (Gibco). 1 × 104 cells were seeded into low adhesion petri dishes (Greiner) and cultured for 6–8 days. About 20% of media was changed every 2 days. All cells were regularly tested for mycoplasma contamination. PC3 cells were treated with 1 μM of BAZ2A‐BRD inhibitors BAZ2‐ICR (SML1276, Sigma) and GSK2801 (SML0768, Sigma), EZH2 inhibitor GSK126 (S7061‐5MG, Lubio), BET inhibitors JQ1 (SML1524, Sigma), or DMSO as control. Fresh medium supplemented with the drugs was added every 2 days. PCa spheres were culture for 6 days. Cells were collected and transferred to a 24‐well plate to image and quantify the number of PCa spheres upon drug treatments.

Culture of prostate organoids

Isolation of prostate epithelial cells was performed as previously described (Karthaus et al, 2014; Pietrzak et al, 2020). Murine prostate lobes were isolated from 10‐ to 11‐week‐old mice under dissection microscope, macerated with blade, and transferred into tube containing 1.5 ml solution of collagenase/hyaluronidase (Stem Cell Technologies) in ADMEM/F12 (Gibco Life Technologies) containing 100 μg/ml Primocin (InvivoGen). Subsequently, prostate tissue was incubated at 37°C rocking for 2 h. After collagenase digestion, samples were centrifuged at 350 g for 5 min, supernatant was discarded and cell pellet was washed with HBSS (Life Technologies). To dissociate prostate tissue into single cells, cell pellet was mixed with 5 ml TrypLE (Gibco Life Technologies) with the addition of 10 μM ROCK inhibitor Y‐27632 (STEMCELL Technologies) and incubated for 15 min at 37°C followed by pipetting up and down several times. Subsequently, HBSS + 2% FBS (Gibco; equal to 2× volume of TrypLE) was added to dissolve TrypLe and quench the reaction. Trypsinized prostate tissue was centrifuged at 350 g for 5 min, supernatant was discarded, and the pellet was washed with HBSS and centrifuged again. 1 ml of pre‐warmed Dispase (STEMCELL Technologies)/DNaseI (STEMCELL Technologies) solution was added, and samples were vigorously and continuously pipetted up and down until solution was homogenously translucent with no visible tissue fragments. Samples were centrifuged to remove Dispase solution, and cell pellet was washed with ADMEM/F12 containing 2% FBS (equal to 5× volume of Dispase). To obtain single cells, suspension was filtered through a 40‐um cell strainer and centrifuged, and then, supernatant was removed. Cells were resuspended in ADMEM/F12 + 2% FBS, and viable cells were counted using hemocytometer and trypan blue.

Generation of organoids was performed as previously described (Pietrzak et al, 2020). Prostate cells were resuspended in cold ADMEM/F12 5+ medium at density 5,000 cells per 40 μl for total organoid culture. Cell suspension was mixed with 60 μl of pre‐thawed growth factor reduced Matrigel (Corning) and plated in a form of drop into the middle of a well from 24‐well plate (TPP). Organoids were cultured in 500 μl of ADMEM/F12 supplemented with B27 (Gibco Life Technologies), GlutaMax (Gibco Life Technologies), 10 mM HEPES (Gibco Life Technologies), 2% FBS, 100 μg/ml Primocin (ADMEM/F12 5+ medium), and contained following growth factors and components: mEGF (Gibco Life Technologies), TGF‐β/Alk inhibitor A83‐01 (Tocris), dihydrotestosterone (DHT; Sigma‐Aldrich) with the addition of 10 μM ROCK inhibitor Y‐27632 for the first week after seeding (organoid culture complete medium). Medium was changed every 2–4 days by aspiration of the old one and addition of fresh 500 μl of culture organoid medium. Sequences encoding control shRNA (CACAAGCTGGAGTACAACTAC) and shRNA targeting Pten (CGACTTAGACTTGACCTATAT) were cloned into lentiviral vectors. Lentiviral supernatants were concentrated 100 times using Lenti‐X Concentrator (Clontech) according to manufacturer's protocol. 40 μl of concentrated virus was mixed with 100.000 cells suspended in transduction medium (ADMEM/F12, B27 GlutaMax, 10 mM HEPES, 100 μg/ml Primocin, 2 μg/ml Polybrene (Sigma‐Aldrich), 10 μM ROCK inhibitor). Cells were incubated with virus for 20 min at RT, centrifuged for 40 min at 800 g at RT, and placed in incubator for 3.5 h to recover. Cells were plated in Matrigel at 5,000 cells/well density, and after 3 days postseeding, 1 μg/ml puromycin (Gibco) was applied to select organoids stably expressing shRNA‐control or shRNA‐Pten.

Organoids were cultured for 22 days in the presence of 5 μM of BAZ2A inhibitors BAZ2‐ICR (SML1276, Sigma), GSK2801 (SML0768, Sigma), EZH2 inhibitor GSK126 (S7061‐5MG, Lubio), BET inhibitors JQ1 (SML1524, Sigma) or DMSO as control. Medium was replaced every 2 days with 500 μl of fresh medium supplemented with inhibitors.

Establishment of FLAG‐HA‐BAZ2A PC3 cell line

In order to introduce the FLAG‐HA tag in BAZ2A locus, a donor plasmid was synthetized by IDT containing the genomic sequence of the FLAG‐HA tag flanked by ± 1 KB homology arm to BAZ2A locus. Generation of stable cell was achieved by Alt‐R™ CRISPR‐Cas9 system from IDT as per manufacturer protocol. A crRNA guide sequence (CCTTCTCTCCCAGTTCTCGG) was chosen to target the BAZ2A locus on exon 2 three base pairs upstream of the ATG start codon. Equimolar concentrations of RNA oligos synthetized by IDT (sgRNA and transactivating crRNA) were mixed in Nuclease‐Free Duplex Buffer (IDT) to a final concentration of 1 μM and were hybridized by heating the mix at 95°C 5 min and cooling it down to RT. 1 μl of hybridized RNA oligos was incubated with 1 μM of Cas9 protein (IDT) in Opti‐MEM I Reduced‐Serum Medium (Gibco) and incubated at RT for 15 min to form a ribonucleoprotein (RNP) complex. RNP complex was transfected for 48 h with Lipofectamine RNAiMax (Invitrogen) to 4 × 104 PC3 previously transfected with donor plasmid carrying the desired inclusion sequence. Single cell clones were isolated and grew until colonies were big enough to be genotyped. To assess the integration of FLAG‐HA tag in BAZ2A locus, two oligos were designed to amplify a genomic region that spans the FLAG‐HA tag and a region 150 bp outside of the left homology arm that is not encoded in the donor plasmid. Genotyping of clones was performed using primers amplifying the region that spans the insertion site of FLAG‐HA tag within BAZ2A locus, where a product of 230 bp represents the wild‐type allele and a product of 300 bp indicates the integration of the FLAG‐HA tag (primers are listed in Table 1).

Table 1

List of primers used in this study.

Gene	Forward (5′–3′)	Reverse (5′–3′)
Cloning
hBAZ2A‐BRD_mutant (Y1830F)	ggtcagcggctTtcgccgtatcatcaaaaacc	ggtttttgatgatacggcgaAagccgctgacc
hBAZ2A‐BRD_mutant (N1873L)	gtcagaccttcCTtgaagatgacagc	gctgtcatcttcaAGgaaggtctgac
mBAZ2A‐BRD_mutant (Y1775F)	ggtgagtggaTTCcgacgtgtcatcaagaacc	ggttcttgatgacacgtcgGAAtccactcacc
Genotyping
BAZ2A_FLAG‐HA‐tag_insertion	CTCAGGTGAGCCACCGTAC	TCGAGCTTGTCGTCATCGTC
BAZ2A_FLAG‐HA‐tag_genotyping	TGCTCAGGCTAGAGGCATATT	CAGGGGGAAGGCCAGTAAAG
ChIP_qPCR
AOX1	GTTCGTCCCATCCTTTCGT	CCTGAAAACGCAGCCTTTAG
GAPDH	CGGGATTGTCTGCCCTAATTAT	GCACGGAAGGTCACGATGT
EVX1	TACACAGCATCTGGGGAGTG	GTGTGCTGGGTTAAGGGAGA
qRT–PCR
ALDH1A2	AGCTGGGACTGTTTGGATCA	ACTCCCGCAAGCCAAATTCT
LGR5	AGCATTCACTGGCCTTTACAGT	CATCCAGACGCAGGGATTGAA
CD24	CTGCTGGCACTGCTCCTAC	GGCATTAGTTGGATTTGGGGC
HOXA1	CACGCCAGCCACCAAGA	ACTTTCCCTGTTTTGGGAGGG
HOXA4	ATGTCAGCGCCGTTAACCC	CCTTCTCCAGCTCCAAGACC
SOX7	TTCATGGTTTGGGCCAAGGA	CCACGACTTTCCCAGCATCT
MRPS7	AGCCTATGATTGGGCTGGTA	CGGCACTCAGTGATCATCC
RPS3A	AATCATGACCCGAGAGGTGC	TCCAATGCTGTCTGGAATCAATTT
BAZ2A	AAGATGTGTGGCTACAATGG	TCTGCACCCATCAGCTCCG
KRT8	TAAGGATGCCAACGCCAAGT	AGACTCCAGCCGGCTCTC
EVC	ACGCGACACCAGAAAGACG	TATTGCTGTTGGAAGGCGGC
ITPKB	TTCAAGTGCCACGAGGTCAT	GGTTATTGAGCCCCGAGAGG
SH3BP4	GAGCTTGTGATGGCCCTACT	GTCCATCTGCTGCTTGGAGA
SUN2	CGAGTCATCCTCCAGCCAGA	GCTGCAGGTCTTCGTCAAAC
EP300	CCTGGGTCCTATGCCAACAG	TGGCTTGGACGAGTTTGTGA
AOX1	CGACAAGCCCAGCGACAGG	GTGGCTGGACCAACGCCTCC
MAN2B2	CTGCAGATCCTGAGCATCCC	AGCACAGCCTCCAGATTCAC
DMPK	GCCTTTTGTGGGCTACTCCT	GCCACTTCAGCTGTTTCATCC
DDB1	TCAACGGCATGATAGGGCTG	CGCTCGGTGTGAAAGGATCT

Expression and purification of recombinant proteins

Sequence corresponding to BAZ2A‐bromodomain (residues 1,797–1,899) was codon optimized for robust expression in E. coli and cloned into pGex4T‐1. BAZ2A‐bromodomain mutants (Y1830F, N1873L) were achieved by single point mutation PCR and sequenced to ensure accuracy of the point mutations (primers are listed in Table 1). The expression of recombinant proteins was achieved by transforming E. coli competent BL21 bacteria with 600ng of plasmids expressing BAZ2A‐bromodomain. Bacteria were grown at 37°C in Terrific Broth (24 g/l yeast extract, 20 g/l tryptone, 4 ml/l glycerol, 0.017 M KH2PO4, and 0.072 M K2HPO4) until reached an OD of 0.6–0.7 at 600 nm of absorbance. Induction of the expression of recombinant proteins was done at 16°C overnight by supplementing the inoculated culture with 0.5 mM IPTG. Bacteria were harvested by centrifugation, and pellet was resuspended in 4× (w/v) lysis buffer (20 mM Tris pH7.5, 300 mM NaCl, 1 mM DTT, 20% glycerol, 1 mM PMSF) and sonicated 3× 30 s on/off. After sonication, lysates were treated with DNase I (50 U) and RNase‐A (50 μg) 30 min at 4°C with constant rotation. Lysates were then spun down, and supernatant was collected and filtered (0.22 μM). GST‐fusion proteins were purified using Glutathione‐Sepharose 4B beads (GE Healthcare). Purified proteins were eluted with 20 mM Glutathione and dialyzed overnight (20 mM Tris pH 7.5, 100 mM NaCl, 2 mM EDTA, 20% Glycerol). Purity of eluted proteins was assessed by SDS–PAGE.

Histone peptide arrays

MODified Histone Peptide Arrays were acquired from Active Motif. Briefly, arrays were blocked for 4 h with Odyssey® Blocking Buffer diluted in PBS (1:1). Binding of recombinant proteins to histone peptide array was done by diluting 10 nM of recombinant GST or GST‐BAZ2A proteins in 3ml of HEMG buffer (25 mM HEPES pH 7.6, 12.5 mM MgCl2, 0.5 m EDTA, 1 mM DTT, 0.2 mM PMSF, 10% glycerol) and incubated overnight at 4°C. Pharmacological inhibition of GST‐BAZ2A‐BRD was achieved by incubating recombinant proteins with 50 nM of BAZ2‐ICR or GSK2801 overnight in HEMG buffer. Detection of binding was determined by 2‐h incubation (room temperature) with anti‐GST (SC‐459, 1:1,000 in Odyssey® Blocking Buffer), followed by 1‐h incubation with goat anti‐Rabbit (IRDye® 800CW, 1:10,000 in Odyssey® Blocking Buffer). Between incubation periods, histone peptide arrays were washed 3X in PBST to reduce background levels. Arrays were scanned on Odyssey Infrared Imaging System. Active Motif's Array Analyzer software was used to identify peptides bound by recombinant proteins, and intensity of binding was measured as fold changes over background level.

Plasmid and siRNA transfections

cDNA corresponding to mouse BAZ2A wild type was cloned into AASV1 donor vector (DC‐DON‐SH01, GeneCopoeia). Site‐directed mutagenesis was performed to create BAZ2A‐bromodomain mutant (Y1775F), and sequencing of plasmid was performed to ensure fidelity of sequences. 1 × 106 of PC3 cells were transfected for 48 h with 8 μg of plasmids with X‐tremeGENE HP DNA Transfection Reagent (Roche). siRNA against BAZ2A (Cat. # SI00102389) and siRNA‐control (Cat. # 1027281) was obtained from Qiagen. EP300 siRNA (Cat. # J‐003486‐11‐0002) and BAZ2B siRNA (Cat. # D‐020487‐02‐0005) were obtained from Dharmacon. Transfections of siRNA were performed with Lipofectamine RNAiMax (Invitrogen) in Opti‐MEM Reduced‐Serum Medium (Gibco).

Chromatin immunoprecipitation

ChIP experiments were performed as previously described (Leone et al, 2017). Briefly, 1% formaldehyde was added to cultured cells to cross‐link proteins to DNA. Isolated nuclei were then lysed (50 mM Tris–HCl pH 8.1, 10 mM EDTA and 1% SDS) and sonicated using a Bioruptor ultrasonic cell disruptor (Diagenode) to shear genomic DNA to an average fragment size of 200 bp. 20 μg of chromatin was diluted tenfold with ChIP buffer (16.7 mM Tris–HCl pH 8.1, 167 mM NaCl, 1.2 mM EDTA, 0.01% SDS, and 1.1% Triton X‐100) and precleared for 2 h with 10 μl of packed Sepharose beads for at least 2 h at 4°C. Immunoprecipitation was done overnight with the indicated antibodies. Pulldowns were done with Dynabeads protein‐A (or ‐G, Millipore) for 4 h at 4°C. For HA‐BAZ2A ChIPs, preclearing and pulldowns were done with beads block with BSA (50 μg) and ytRNA (50 μg), per 400 μl of Dynabeads protein‐A. After washing, elution, and reversion of cross‐links, eluates were treated with RNAse A (1 μg). DNA was purified with phenol–chloroform, ethanol precipitated, and quantified by quantitative PCR. Primer sequences and antibodies are listed in Tables 1 and 2.

Table 2

List of antibodies used in this study.

	Company	Cat. number
Ab used for ChIP
H3K27me3	Active Motif	39155
H3K4me3	Merck	07‐473
H3K27ac	Abcam	ab4729
H3K14ac	Active Motif	61433
HA	Abcam	ab9110
Ab used for immunohistochemistry
Anti‐Ki67	Abcam	ab15580

ChIPseq analysis

The quantity and quality of the isolated DNA was determined with a Qubit® (1.0) Fluorometer (Life Technologies, California, USA) and a Bioanalyzer 2100 (Agilent, Waldbronn, Germany). The NEBNext® Ultra™ II DNA Library Prep Kit from NEB (New England Biolabs, Ipswich, MA) was used in the succeeding steps. ChIP samples (1 ng) were end‐repaired and adenylated. Adapters containing the index for multiplexing were ligated to the fragmented DNA samples. Fragments containing adapters on both ends were enriched by PCR. The quality and quantity of the enriched libraries were measured using Qubit® (1.0) Fluorometer and the TapeStation (Agilent, Waldbronn, Germany). The product is a smear with an average fragment size of approximately 300 bp. The libraries were normalized to 10 nM in Tris‐Cl 10 mM, pH8.5 with 0.1% Tween 20. The TruSeq SR Cluster Kit v4‐cBot‐HS (Illumina, Inc, California, USA) was used for cluster generation using 8 pM of pooled normalized libraries on the cBOT. Sequencing was performed on the Illumina HiSeq 2500 single end 125 bp using the TruSeq SBS Kit v4‐HS (Illumina, Inc, California, USA). Obtained ChIPseq reads from Illumina HiSeq 2500 were aligned to the human hg38 reference genome using Bowtie 2 (version 2.2.5). Read counts were computed and normalized using “bamCoverage” from deepTools (version 2.0.1 (Ramirez et al, 2014)) using a bin size of 50 bp. deepTools was used to generate all heat maps, Pearson correlation plots. Peaks from ChIPseq experiments were defined using MACS2 (version 2.1.0 (Zhang et al, 2008)) comparing ChIP signal against input. For all histone modifications, default settings were used. BAZ2A peaks were determined using a P‐value cutoff as < 0.01. Data sets from H3K4me1 (GSM2534170) and DNAse I (GSM2400265) were obtained from ENCODE (ENCODE Project Consortium, 2012) (see also Table 3). Integrative Genome Viewer (IGV, version 2.3.92 (Robinson et al, 2011) was used to visualize and extract representative ChIPseq tracks. Genomic annotation of ChIPseq peaks was done with ChIPseeker (Yu et al, 2015), and promoter regions were defined as ± 5 Kb from TSS. Analysis of overlapping genomic regions was done with bedtools (version 2.27.1 (Quinlan & Hall, 2010)). Enhancer coordinates were obtained from GeneHancer V4.4 (Plaschkes et al, 2017), and the read coverage ± 1 Kb from the center of peak was obtained using “computeMatrix” from deepTools with a bin size of 50 bp. Mean read coverage was then used to classify the regions based on the presence or absence of H3K27ac and H3K4me1.

Table 3

List of datasets used in this study.

H3K4me1	https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE96243	ENCODE Project Consortium (2012)
DNase I	https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE90286	ENCODE Project Consortium (2012)

RNA extraction, reverse transcription, and quantitative PCR

RNA was purified with TRIzol reagent (Life Technologies). 1 µg total RNA was primed with random hexamers and reverse‐transcribed into cDNA using MultiScribe™ Reverse Transcriptase (Life Technologies). Amplification of samples without reverse transcriptase assured the absence of genomic or plasmid DNA. The relative transcription levels were determined by normalization to GAPDH mRNA levels. qRT–PCR was performed with KAPA SYBR® FAST (Sigma) on Rotor‐Gene RG‐3000 A (Corbett Research). Primer sequences are listed in Table 1.

RNAseq and data analysis

Total RNA from two independent samples form PC3 cells and PCa cells, and from three independent experiments on PC3 cells treated with siRNA control or siRNA‐BAZ2A, was purified with TRIzol reagent (Life Technologies) as stated above. DNA contaminants were removed by treating RNA with 2 U Turbo‐DNaseI (Invitrogen) for 1 h at 37°C, and the RNA samples were re‐purified using TRIzol. The quality of the isolated RNA was determined by Bioanalyzer 2100 (Agilent, Waldbronn, Germany). Only those samples with a 260 nm/280 nm ratio between 1.8 and 2.1 and a 28S/18S ratio within 1.5–2 were further processed. The TruSeq RNA Sample Prep Kit v2 (Illumina, Inc, California, USA) was used in the succeeding steps. Briefly, total RNA samples (100–1,000 ng) were poly A enriched and then reverse‐transcribed into double‐stranded cDNA. The cDNA samples were fragmented, end‐repaired, and polyadenylated before ligation of TruSeq adapters containing the index for multiplexing fragments containing TruSeq adapters on both ends were selectively enriched with PCR. The quality and quantity of the enriched libraries were validated using Qubit® (1.0) Fluorometer and the Caliper GX LabChip® GX (Caliper Life Sciences, Inc., USA). The product is a smear with an average fragment size of approximately 260bp. The libraries were normalized to 10 nM in Tris‐Cl 10 mM, pH 8.5 with 0.1% Tween 20. The TruSeq SR Cluster Kit HS4000 (Illumina, Inc, California, USA) was used for cluster generation using 10 pM of pooled normalized libraries on the cBOT. Sequencing was performed on the Illumina HiSeq 2500 single end 100 bp using the TruSeq SBS Kit HS2500 (Illumina, Inc, California, USA). Reads were aligned to the reference genome (hg38) with Subread (i.e., subjunc, version 1.4.6‐p4; (Liao et al, 2013)) allowing up to 16 alignments per read (options: –trim5 10 –trim3 15 ‐n 20 ‐m 5 ‐B 16 ‐H –allJunctions). Count tables were generated with Rcount (Schmid & Grossniklaus, 2015) with an allocation distance of 10 bp for calculating the weights of the reads with multiple alignments, considering the strand information, and a minimal number of 5 hits. Variation in gene expression was analyzed with a general linear model in R with the package edgeR (version 3.12.0; (Robinson & Oshlack, 2010)) according to a crossed factorial design with two explanatory factors (i) siRNA against BAZ2A and a mock sequence and (ii) expression pattern of PCa spheres vs PC3 cells. Genes differentially expressed between specific conditions were identified with linear contrasts using trended dispersion estimates and Benjamini–Hochberg multiple testing corrections. Genes with a P‐value below 0.05 and a minimal fold change of 1.5 or 2 were considered to be differentially expressed. These thresholds have previously been used characterizing chromatin remodeler functions (de Dieuleveult et al, 2016). Gene ontology analysis was performed with David Bioinformatics Resource 6.8 (Huang et al, 2009).

Gene set enrichment analysis

GSEA (Subramanian et al, 2005) was obtained by comparing expression values of PCa spheres and PC3 cells of genes upregulated in PCa spheres (P < 0.05, log2FC > 1). GSEA P‐value was calculated over 100,000 permutations based on the phenotype, and FDR was < 25%.

Immunohistochemistry

Organoids were fixed using 4% paraformaldehyde (PFA) overnight and then washed with 70% ethanol. Fixed organoids were pre‐embedded in HistoGel (Thermo Fisher Scientific) to pellet organoids together. Tissues and organoid pellets embedded in HistoGel were processed with an automated tissue processor and subsequently embedded in paraffin wax according to standard techniques. Samples were sectioned at 4 μm onto SuperFrost® Plus (Thermo Scientific) microscope slides and air‐dried 37°C overnight. Immunohistochemistry was performed using BrightVision+ histostaining kit (Immunologic) according to manufacturer's protocol. Briefly, organoid sections were dewaxed in xylene (3× 3 min), followed by rehydration steps in descending percentages of EtOH (3 min in 100, 95, 90, 70, 50%), and washed twice in PBS. Slides were boiled in sodium citrate pH 6.0 (10 mM sodium citrate, 0.05% Tween 20) for 15 min and then allowed to cool to room temperature for 30 min. Slides were then washed in PBS and blocked for at least 1 h in PBS + 2% BSA. Samples were incubated with anti‐Ki67 primary antibody (1:100, Abcam ab15580) overnight in blocking reagent at 4°C and then washed three times with PBS. Post‐antibody blocking solution was added for 15 min, and samples were incubated at RT. Sections were washed twice with PBS and then were incubated for 30 min at RT with Poly‐HRP‐Goat anti‐Mouse/Rabbit IgG. After incubation, sections were washed twice with PBS and were incubated with 3,3′ diaminobenzidine (DAB) mixture (Bright‐DAB, Immunologic) for 8 min according to manufacturer protocol. After incubation, organoid sections were rinsed in deionized water and mounted with Fluoro‐Gel.

Author contributions

RP‐H and RS designed the study, made the figures, and wrote the article. RP‐H performed the majority of the experiments and analyzed data. KP, SS, and RA performed analyses with organoids. SCF analyzed tumorspheres in the presence of inhibitors. MR and RA performed expression analyses with EP300 and BAZ2A inhibitors. JB analyzed BAZ2B in tumorspheres. LL provided bioinformatic support. RP‐H, RA, and RS organized the structure of the manuscript. RS supervised and directed the study. All authors commented on the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Expanded View Figures PDF

Click here for additional data file.

Dataset EV1

Click here for additional data file.

Dataset EV2

Click here for additional data file.

Dataset EV3

Click here for additional data file.

Dataset EV4

Click here for additional data file.

Source Data for Expanded View

Click here for additional data file.

Source Data for Figure 2

Click here for additional data file.