The Pioneering Role of GATA2 in Androgen Receptor Variant

Published OnlineFirst March 4, 2019; DOI: 10.1158/1541-7786.MCR-18-1231

Cancer Genes and Networks Molecular Cancer Research The Pioneering Role of GATA2 in Androgen Receptor Variant Regulation Is Controlled by Bromodomain and Extraterminal Proteins in Castrate-Resistant Prostate Cancer Lewis Chaytor, Matthew Simcock, Sirintra Nakjang, Richard Heath, Laura Walker, Craig Robson, Dominic Jones, and Luke Gaughan

Abstract

The androgen receptor (AR) is a key driver of prostate (BET) proteins and is codependent for DNA binding. cancer development. Antiandrogens effectively inactivate GATA2 activity is compromised by BET inhibitors, which the AR, but subsequent AR reactivation progresses the dis- attenuates the pioneering role of GATA2 in CRPC. In all, ease to castrate-resistant prostate cancer (CRPC). Constitu- this study indicates that GATA2 is a critical regulator of tively active AR splice variants (AR-V) that function unchal- AR-V–mediated transactivation and is sensitive to BET inhi- lenged by current AR-targeted therapies are key drivers of bitors, signifying these agents may be efﬁcacious in patients CRPC. Currently, very little is known about the regulation of with CRPC which overexpress GATA2. AR-Vs at the chromatin level. Here, we show that the pioneer factor GATA2 is a critical regulator of AR-Vs. Furthermore, Implications: We have deﬁned novel mechanisms of AR-V and we demonstrate that the GATA2 cistrome in CRPC shares GATA2 regulation in advanced prostate cancer that could be considerable overlap with bromodomain and extraterminal therapeutically exploited.

Introduction approximately 50% (5, 6) and development of resistance limits their efficacy in the advanced setting. A major contributing factor Prostate cancer is the second most commonly diagnosed cancer to the compromised activity of second-generation antiandrogens in men globally with 1.3 million annual diagnoses worldwide in CRPC is the expression of AR splice variants (AR-V), which are (World Health Organization, 2018). At presentation, prostate bone fide drivers of AR signaling programs in CRPC. Critically, cancer growth is androgen dependent; hence, the mainstay of AR-Vs, such as the clinically relevant AR-V7 and AR-V3 iso- treatment is androgen deprivation therapy (ADT) in combination forms (7), lack the ligand binding domain, but retain the tran- with antiandrogens. Although initially effective, patients eventu- scriptionally potent N-terminal domain and DNA-binding ally relapse with more aggressive disease termed castrate-resistant domain, allowing AR-Vs to drive gene expression independently prostate cancer (CRPC), which is largely fatal. of hormone stimulation, and critically evade all current direct The androgen receptor (AR) is a member of the nuclear hor- AR-targeting therapies (8–10). This poses a major clinical chal- mone receptor family of transcription factors and is the primary lenge as the most recent observations indicate approximately 80% target of ADT. Androgens, in the form of testosterone, and its more of CRPC patients express these constitutively active forms of the potent metabolite dihydrotestosterone (DHT), provide andro- receptor (11). genic signals that are transmitted via the AR and ultimately drive Without the availability of direct AR-V antagonists, recent prostate cancer progression (1). Importantly, AR signaling persists research has focused on establishing if AR-V coregulatory proteins in CRPC due to several well-characterized molecular mechan- represent tractable targets to attenuate AR-V function (12–14). isms (2), including AR gene amplification and mutation, which AR-V chromatin occupancy at both proximal promoters and distal has justified the development of second-generation antiandro- enhancers of target genes is required for canonical AR gene gens, such as enzalutamide and apalutamide (3, 4). These com- transactivation (15). Proteins that control the enrichment of pounds are used clinically to treat CRPC, but response rates of AR-Vs at these sites may therefore represent novel therapeutic Northern Institute for Cancer Research, Newcastle University, Newcastle Upon targets. Consistent with this, GATA2 and FOXA1 are pioneer Tyne, UK. factors that have been shown in multiple studies to engage and reconfigure condensed chromatin (16, 17) to facilitate DNA Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/). binding of AR-FL (15). Together, GATA2 and FOXA1 co-occupy approximately 55% of the AR-FL cistrome in prostate cancer (15). Corresponding Author: Luke Gaughan, Newcastle University, Paul O'Gorman FOXA1 and GATA2 are critical for AR-FL recruitment to Building, Framlington Place, Newcastle Upon Tyne, Newcastle NE2 4HH, UK. cis fi Phone: 191-208-4360; E-mail: [email protected] -regulatory elements of AR-target genes and promote ef cient AR-FL transcriptional activity (15, 18). However, our understand- Mol Cancer Res 2019;17:1264–78 ing of how AR-Vs are controlled by pioneer factors is ill-defined, doi: 10.1158/1541-7786.MCR-18-1231 particularly with respect to GATA2. Importantly, our previous 2019 American Association for Cancer Research. data indicated that only 41% of the AR-V transcriptome was

1264 Mol Cancer Res; 17(6) June 2019

AR-V Dependency on GATA2

dependent upon FOXA1 (12), suggesting that GATA2 may con- study. Gene-expression values were downloaded and plotted tribute to regulation of AR-V–target genes in CRPC. according to clinical disease type: benign prostatic hyperplasic The GATA family of transcription factors, which consists of (BPH), localized prostate cancer, and metastatic castrate-resistant GATA1-6, are differentially expressed across diverse cell lineages prostate cancer (mCRPC). Outliers are shown as *, &, and ~, and play numerous important tissue-specific roles in a range of respectively, (, P 0.05; ns, P 0.05). physiologic processes, including erythrocyte differentiation and proliferation, as well as cardiovascular and gastrointestinal devel- Plasmids opment (19–23). GATA transcription factors contain two zinc- pFLAG-GATA2 (Addgene plasmid #1418), GFP-BRD2 fingers, which enable DNA and cofactor interactions (24), and are (Addgene plasmid #65376), and GFP-BRD3 (Addgene plasmid directly acetylated which controls DNA-binding capacity and #65377) were purchased from Addgene. pFLAG-CMV-BRD4- target gene transactivation (25). GATA2 is the most highly short (1-700) was created by cloning the truncated BRD4 expressed member of the GATA family in the prostate, and its sequence, amplified by PCR using the primers BRD4F: AGG AGA expression correlates with Gleason score, metastasis, and inva- TAT ACC ATG AAG CTT ATG TCT GCG GAG AGC GGCCC and siveness of prostate cancer (26, 27). Our understanding of the role BRD4SR: CAG CAC TAG TGG ATC CTC AGG ACG AGA AGC CCT of GATA2 in prostate transcriptional regulation is largely limited TCA TCT into Bam H1 and Hind III sites of pFLAG-CMV2 (Sigma). to its ability to act as a pioneer factor for AR-FL. Importantly, The QuikChange II Site-Directed Mutagenesis Kit (Agilent Tech- GATA2 and AR are involved in an autoregulatory loop in prostate nologies) was used to introduce mutations into mammalian cancer cells. GATA2 gene expression is elevated by ADT and expression vectors by PCR according to the manufacturer's antiandrogens, which in turn upregulates GATA2-mediated trans- instructions. Mutagenized plasmid DNA was used to transform activation of AR expression via binding to transcriptional regu- competent E. coli. Colonies were picked, propagated, and DNA latory sites approximately 5.5 and 4.6 kb upstream of the AR was extracted. Colonies were screened for desired mutations using gene (15, 26). Moreover, by engaging and priming regulatory Sanger sequencing (Genewiz). elements of canonical AR-target genes, GATA2 enhances AR-FL activity by facilitating chromatin deposition of the active receptor Cell culture, siRNA transfections, and DNA transfections at target loci. Importantly, depletion of GATA2 compromises CWR22Rv1 (ATCC CRL-2505), VCaP (ATCC CRL-2876), and activation of the AR-FL transcriptome in the LNCaP prostate HEK293T (ATCC CRL-3216) authenticated cell lines were cancer cell line (26). Consistent with these observations, the purchased from ATCC and maintained at 5% CO2 and at 37 C. GATA2 inhibitor K7174 was shown to diminish in vivo growth Each cell line was tested for Mycoplasma every 3 months and of AR- and GATA2-expressing LNCaP-abl xenografts (26). grown to a maximum of 35 passages. CWR22Rv1 and HEK293T Critically, outside of the demonstration that GATA2 is impor- were maintained in RMPI-1640 (R5886, Sigma-Aldrich) supple- tant for facilitating ectopically expressed AR-V7–mediated FKBP5 mented with 10% (v/v) fetal bovine serum (HyClone) and transactivation in LNCaP cells, very little is known about the 2 mmol/L L-glutamine (Sigma-Aldrich). VCaP cells were main- global requirement of GATA2 for AR-V–mediated transactivation tained in Dulbecco's Modified Eagle's Medium (R6171, Sigma- in physiologically relevant models. Furthermore, our knowledge Aldrich) supplemented with 10% (v/v) fetal bovine serum and of what governs GATA2 chromatin occupancy to enable it to 2 mmol/L L-glutamine. For experiments requiring steroid-deplet- function as an AR-V pioneer factor is extremely limited and needs ed conditions, CWR22Rv1 cells were cultured in RPMI-1640 addressing. To this end, we show that approximately 36% of the supplemented with 10% (v/v) dextran-coated charcoal-stripped AR-V transcriptome is dependent upon GATA2, and depletion of fetal bovine serum (HyClone) and 2 mmol/L L-glutamine. VCaP the pioneer factor diminishes expression of an AR-V–driven gene cells were cultured in Dulbecco's Modified Eagle's Medium signature important for CRPC progression. Using chromatin (R6171, Sigma-Aldrich) supplemented with 10% (v/v) dextran- immunoprecipitation (ChIP)-sequencing, we show that GATA2 coated charcoal-stripped fetal bovine serum (HyClone) and chromatin occupancy overlaps with multiple members of the 2 mmol/L L-glutamine. For siRNA transfections, siRNAs (Supple- bromodomain and extraterminal (BET) family of transcriptional mentary Table S1) were transfected using Lipofectamine RNAi- regulators, including BRD2, BRD3, and BRD4. Furthermore, we MAX transfection reagent (Thermo) to a final concentration of show that GATA2 is acetylated and interacts with BET proteins in a 25 nmol/L (as described in ref. 29). For DNA transfections, bromodomain-dependent manner. Consistent with this mode of plasmid DNA was transfected using TransIT-LT1 transfection interaction, we demonstrate that the pan-BET inhibitor JQ1 reagent (Mirus). attenuates GATA2–BET interaction and compromises GATA2 chromatin binding and transcriptional competency, suggesting RT-qPCR and microarray gene-expression analysis that GATA2 is dependent on BET proteins for its role as a pioneer RNA exactions were performed using TRIzol Reagent (Life factor. Overall, our data indicate a novel mechanism for regulat- Sciences, Invitrogen) according to the manufacturer's recommen- ing GATA2 binding to AR-V target genes that involve BET proteins, dations and resultant RNA was reverse transcribed using the and that this dependence may sensitize GATA2-driven CRPC to M-MLV reverse transcription kit (Promega). Changes in gene BET inhibitors. expression were quantified using quantitative RT-PCR incorpo- rating cDNA and AR- and GATA2-target gene-specific primers (Supplementary Table S2). RPL13A was used as a housekeeping Materials and Methods gene. Data are presented as mean normalized gene expression of 3 Online gene-expression profiling independent experiments SEM (, P 0.05; ns, P 0.05). PRIM1, TRIM14, and RAP1GAP mRNA expression was assessed For microarray studies, RNA was extracted as described above by analyzing preprocessed gene-expression microarray/RNA-seq and subject to differential gene-expression analysis using the data sets from the Taylor and colleagues (ref. 28; GSE21032) Illumina HT-12 BeadChip array (AROS); the GEO accession

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1265

Chaytor et al.

number for these data is GSE124999. Array processing, back- sample) on an Illumina NextSeq 500 System (performed by ground correction, normalization, and quality control checks Genomics4Life). ChIP-Seq data are publicly accessible with the were performed using the R package Lumi. Probe intensity values accession number GSE125236. Fastq files for BRD2, BRD3, BRD4, þ were converted to variance stabilized data (VSD) using variance and AR / DHT were downloaded from Asangani and collea- stabilizing transformation. The robust spline normalization was gues (14), c-Myc and E2F1 fastq files were downloaded online used as an array normalization method. Poor quality probes from Barfeld and colleagues (30) and Xu and colleagues (34), (detection threshold < 0.01), and probes that are not detected respectively, and the LNCaP GATA2 BED file was a kind gift from at all in the remaining arrays, were removed prior to downstream Nicholas Mitsiades (26). The raw read data were aligned to the analysis. The remaining probe (22,380) normalized intensity was human genome hg38 assembly using Bowtie2 with default para- used in the differential expression analysis. Differentially meters. Nonuniquely aligned reads and aligned reads with more expressed genes were analyzed using the Gene Set Enrichment than two mismatches were removed using an in-house script. Analysis (GSEA) tool (Broad Institute). Differentially expressed Potential PCR duplicates were removed using Picard Mark Dupli- genes were ranked from highest to lowest by log fold change cate. Regions of genomic enrichment were identified using (LogFC). GSEA was performed using classic weighting of pre- MACS2 with default settings (q < 0.05). Bioconductor package ranked genes. Normalized enrichment scores (NES) and false DiffBind was used for differential binding analysis between discovery rates (q-value) were computed using 1,000 random knockdown/treated and control samples. Bioconductor package permutations of preranked genes. Using GSEA, the GATA2 gene ChipPeakAnno was used to annotate the identified binding signature was compared with the AR-V gene signature from our regions. For motif analysis, the peak summits returned from previous study (12) to assess the correlation between signatures. MACS2 peak calling were extended 100 bp upstream and down- The GATA2 gene signature was also assessed against the Gene stream for the strongest 100 GATA2 peaks. DNA sequences were Ontology (GO) and Oncogenic signatures data sets using GSEA. retrieved and submitted to MEME-suite (http://meme-suite.org/). Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 was also used to identify pathways affected by Live-cell imaging and cell counts GATA2 depletion. The GATA2-LNCaP gene signature was a For live-cell imaging, the IncuCyte ZOOM was used to measure kind gift from Nicholas Mitsiades (26), the c-Myc gene signature cell confluence over a period of 120 hours. Cells were seeded on to was downloaded from Barfeld and colleagues (30), and the an appropriate vessel 24 hours prior to measuring cell confluence. 0.5 mmol/L I-BET762 LNCaP gene signature was downloaded Following seeding, the appropriate cellular manipulation was from Wyce and colleagues (31). applied (siRNA transfection) to cells. Images of the vessel were taken every 6 hours for 120 hours. Proliferation was measured as Immunoprecipitation, chromatin immunoprecipitation cell confluence. Each experiment was performed in triplicate, with (ChIP), and Western blotting 3 technical replicates per experimental arm. Percentage of con- Immunoprecipitation was performed as in ref. 32 using an anti- fluence was normalized to the 0 hour time point and presented as GATA2 antibody (Sigma-Aldrich HPA005633) or an anti-Flag a mean normalized fold change in confluence SEM (, P 0.05; antibody (Sigma-Aldrich) and resultant immunoprecipitates sub- ns, P 0.05). For cell counts, cells were seeded into a 12-well ject to Western blot analysis using the following antibodies: anti– culture vessel for 96 hours. Cells were washed, trypsinized, and a-tubulin (Sigma-Aldrich), anti-AR (N20; Santa Cruz Biotechnol- counted using a hemocytometer. A mean cell count for each ogy; sc-816), anti-AR (BD; BD Biosciences; #9441), anti-GATA2 sample was calculated by counting the number of cells in each (Sigma-Aldrich; HPA005633), anti-BRD4 (Bethyl; A301- experimental arm 5 times. Each experimental arm was performed 985A100), antiacetylated-lysine (Cell Signaling Technology; in triplicate (technical replicates), and each experiment was #9441), antihistone H1 (Santa Cruz Biotechnology; sc-8030), performed independently in triplicate. Data represent mean nor- anti-V5 (Santa Cruz Biotechnology; sc-271944), and anti-p300 malized cell counts SEM (, P 0.05; ns, P 0.05). (Abcam; ab14984). ChIP was performed as in ref. 33, and re-ChIP was performed as Statistical analysis in ref. 32. Cells were seeded at a density of 5 106 per 150-mm Statistical analysis was performed in Prism 7 (GraphPad) dish and typically cultured in steroid-depleted conditions for using a one-way ANOVA followed by a Bonferroni multiple 72 hours in total. For ChIP experiments involving GATA2 knock- correction test or an unpaired two-tailed Student t test (, P down, GATA2 was depleted from VCaP and CWR22Rv1 cells for 0.05; ns, P 0.05). 48 hours prior to ChIP. For JQ1-treated ChIP experiments, VCaP and CWR22Rv1 cells were treated with 1 mmol/L JQ1 for 24 hours prior to ChIP. Typically, 80 mg of chromatin was incorporated into Results each ChIP and 2 mg of the following antibodies: anti-AR (N20), GATA2 regulates AR-V transcriptional activity in models of anti-AR (BD), anti-GATA2, anti-BRD4 (Abcam; A301-985A100), CRPC anti-FOXA1 (Abcam; ab128874), and isotype controls We first assessed the effect of GATA2 depletion on a number of (Diagnode; C15410206). DNA fragments retrieved by ChIP were canonical AR-V–target genes in the AR-V–expressing CWR22Rv1 analyzed by quantitative PCR using the primers described in CRPC cell line. To discriminately assess AR-V–mediated transac- Supplementary Table S3. Data are presented as mean normalized tivation, CWR22Rv1 cells were grown in steroid-depleted condi- percentage input of 3 independent experiments SEM (, P tions supplemented with 10 mmol/L enzalutamide to mimic 0.05; ns, P 0.05). maximal castrate conditions (12). As expected, enzalutamide did For ChIP-sequencing experiments in VCaP cells, 5 ng total not affect AR-target gene expression in CWR22Rv1 cells (Fig. 1A). DNA from replicate GATA2 ChIPs was incorporated into library In contrast, compared with a nonsilencing control (siScr), deple- preparation and subsequently sequenced (30 million reads per tion of GATA2 (siGATA2) for 48 hours markedly reduced the

1266 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

Figure 1. AR-V chromatin occupancy and transcriptional activity is dependent upon GATA2. A, CWR22Rv1 cells cultured in steroid-depleted conditions were transiently transfected with either control (siScr) or GATA2 (siGATA2) siRNAs for 24 hours prior to treatment 10 mmol/L enzalutamide (Enz) for an additional 24 hours before RT-qPCR analysis of AR-target genes. Graphs represent mean normalized gene expression (n ¼ 3) SEM (, P 0.05; ns, P 0.05). B, CWR22Rv1 cells were cultured as above, and the effect of GATA2 knockdown on AR protein levels was assessed by Western analysis. C, CWR22Rv1 cells were cultured in steroid- depleted conditions for 72 hours and subject to chromatin immunoprecipitation using an anti-GATA2 antibody or isotype control (IgG) antibody. Graphs represent mean normalized percentage input (n ¼ 3) SEM (, P 0.05; ns, P 0.05). D, CWR22Rv1 cells cultured in steroid-depleted conditions were transiently transfected with either control (siScr) or GATA2 (siGATA2) siRNAs for 24 hours prior to treatment 10 mmol/L enzalutamide (Enz) for an additional 24 hours prior to chromatin immunoprecipitation using an anti-AR (AR N-20) or isotype control (IgG) antibody. Graphs represent mean normalized percentage input (n ¼ 3) SEM (, P 0.05; ns, P 0.05). expression of AR-target genes PSA, KLK2, and TMPRSS2 in targeting siRNA (Supplementary Fig. S1A). Western analysis of CWR22Rv1 cells grown in the presence and absence of enzaluta- CWR22Rv1 cells depleted of GATA2 demonstrated reduced AR-V mide (Fig. 1A). This was conﬁrmed using a second GATA2 protein levels (Fig. 1B), which is consistent with previous ﬁndings

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1267

Chaytor et al.

from CWR22Rv1 cells (26), indicating GATA2 regulates both 41% were coregulated by AR-Vs, suggesting that GATA2 plays an expression and transcriptional competency of AR-Vs. Important- important role in controlling a CRPC-associated AR-V gene sig- ly, expression of the AR-V–target gene UBE2C was unaffected by nature in advanced disease (Fig. 2D; Supplementary Table S2). knockdown of GATA2 (Fig. 1A; Supplementary Fig. S1A), suggest- GO also indicated that GATA2 drives expression of clinically ing that GATA2 regulates discriminate AR-V–mediated transacti- relevant genes involved in cell-cycle regulation independently of vation and the effect of GATA2 knockdown on PSA, KLK2, and AR-Vs (Supplementary Fig. S4A). To identify cellular processes TMPRSS2 is not exclusively a consequence of diminished AR-V that are perturbed by GATA2 depletion, GSEA was performed protein levels. Furthermore, interrogating ChIP-sequencing data using the 2,333 GATA2-regulated genes against the GO GSEA data from VCaP cells demonstrated GATA2 enrichment at PSA, KLK2, sets. Two hundred ninety-one GO data sets were significantly and TMPRSS2 cis-regulatory elements, but not at UBE2C, con- deregulated upon GATA2 depletion (Supplementary Table S1; firming that GATA2 controls AR-V function at specific target loci Fig. S2E, top 5 negatively enriched GO biological processes (Supplementary Fig. S1B). shown), including a negative enrichment of transcripts involved Given that GATA2 facilitates AR-FL chromatin deposition, we in the cell cycle (Fig. 2F). Concordant with these findings, GATA2 next assessed the impact of GATA2 depletion on AR-V chromatin depletion in CWR22Rv1 cells grown in castrate conditions binding. CWR22Rv1 cells grown in maximal castrate conditions resulted in reduced cell proliferation (Fig. 2G; Supplementary transfected with control (siScr) or GATA2 (siGATA2) siRNAs Fig. S4B), which is consistent with compromised AR-V signaling. for 48 hours were subject to ChIP using anti-GATA2 and –N-terminal-specific AR antibodies. As shown in Fig. 1C and GATA2 regulates the transcriptional activity of other oncogenic D, GATA2 and AR-Vs were enriched at PSA and TMPRSS2 enhan- transcription factors cers and knockdown of GATA2 significantly reduced AR-V chro- Given that fewer than half of the GATA2-regulated CRPC matin occupancy at these cis-regulatory elements (Fig. 1D). genes were coregulated by AR-Vs, we speculated that GATA2 Immunoblotting of the ChIP chromatin extractions demonstrat- controls activity of other oncogenic transcription factors to ed that GATA2 was successfully depleted and, as expected, facilitate disease progression. To this end, we performed GSEA this modestly reduced global AR-V chromatin binding (Supple- using the 2,333 GATA2 regulated genes against the Oncogenic mentary Fig. S2). Importantly, AR-FL was not detected in the signatures GSEA data set. Fifty-five oncogenic signatures were chromatin extractions, which is a consequence of the maximal deregulated upon GATA2 depletion (Supplementary Table S1 castrate growth conditions and supports the concept that the AR and Fig. 3A, top 10 negatively enriched shown), including species detected by ChIP are AR-Vs and not AR-FL. c-Myc and E2F1. Transcripts induced by c-Myc and E2F1 were significantly negatively enriched for transcripts regulated by GATA2 depletion downregulates a distinct subset of GATA2 (Fig. 3B), indicating a potential regulatory role of AR-V–driven target genes GATA2 for both transcription factors. To further investigate To examine the role of GATA2 in controlling global AR-V– the dependency of c-Myc on GATA2, a c-Myc–induced gene mediated transactivation, we depleted GATA2 in CWR22Rv1 signature (30) was compared with a GATA2 gene signature (26), cells for 48 hours in maximal castrate conditions and per- both of which were derived from LNCaP cells. Genes induced formed transcriptomic analysis. Genes (2,333) were significant- by c-Myc expression were found to be significantly negatively ly altered upon GATA2 knockdown, with 1,193 and 1,140 enriched for genes that are regulated by GATA2 (Fig. 3C). genes demonstrating upregulation and downregulation, respec- Moreover, interrogating publicly available c-Myc, E2F1, and tively. Comparing the GATA2 transcriptome with our AR-V and GATA2 ChIP-seq data sets from LNCaP cells, we identified FOXA1 gene signatures derived from CWR22Rv1cells(12),we respective overlaps of 11% and 10% between the cistromes of demonstrate that GATA2 regulates 36% (856/2,367) of genes GATA2 and those of c-Myc and E2F1 (Fig. 3D), validating the controlled by AR-Vs and 30% (522/1,722) of genes regulated regulatory role of GATA2 on other oncogenic transcription by FOXA1 (Fig. 2A). When considering genes which are altered factors outside of AR-Vs. by 1.5-fold or greater, we show that GATA2 knockdown upregulates 168 genes and downregulates 113 genes (Supple- Bromodomain and extraterminal family proteins interact with mentary Fig. S3). Overlapping these genes with our in-house GATA2 in a bromodomain-dependent manner AR-V transcriptome, we demonstrate that GATA2 regulates Acetylation of GATA family members at multiple lysine resi- approximately 10% of highly induced AR-V genes and 12% dues enhances DNA binding (25) and, in the case of GATA1, also of highly repressed AR-V genes. To confirm the role of GATA2 facilitates interaction with the BET protein BRD3 in erythroid in AR-V regulation, two AR-V–induced genes identified by our cells (36). Whether GATA2 acetylation occurs in nonerythroid microarray analysis, CDC25A and EXO1, were validated by cells and affects interactions with BET proteins is ill-defined. To RT-qPCR (Supplementary Fig. S3). this end, ectopically expressed Flag-tagged GATA2 immunopre- GSEA demonstrated significant enrichments between the cipitated from HEK293T cells, grown in the presence and absence GATA2 and AR-V global gene signatures; genes downregulated of the histone deacetylase (HDAC) inhibitor TSA, was subject to upon AR-V depletion were negatively enriched for genes repressed Western analysis using a pan–acetyl-lysine antibody. As shown by GATA2 knockdown and vice versa (Fig. 2B), indicating that in Fig. 4A, we detected an acetylated form of GATA2 in the GATA2 depletion affects strongly on AR-V transcriptional activity. presence of TSA, suggesting that, like GATA1, GATA2 is acetylated, Interestingly, 42% (181/426) of the genes that are induced by and this process is under tight control by the reversible effects of GATA2 (LogFC < 0.4) were overexpressed in two CRPC gene sets histone acetyltransferases (HAT) and HDACs. reanalyzed by Urbanucci and colleagues (28, 35), indicating that a We hypothesized that GATA2 and BET family members large proportion of GATA2-regulated genes are clinically relevant interact, and this interaction is dependent upon the bromodo- (Fig. 2C). Furthermore, of the 181 GATA2-regulated CRPC genes, mains of GATA2. HEK293T cells were transfected with a

1268 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

Figure 2. GATA2 regulates oncogenic transcriptional activity in CRPC. A, CWR22Rv1 cells cultured in steroid-depleted media þ 10 mmol/L enzalutamide were depleted of AR-Vs, FOXA1, or GATA2 for 48 hours. Differential gene expression was quantified by gene-expression microarray analysis. Venn diagram shows the overlap of significantly deregulated genes in response to depletion of each protein. B, Gene signatures for GATA2 and AR-Vs were compared using GSEA. Enrichment plots showing genes that are downregulated in response to AR-V depletion are negatively enriched for genes that are regulated GATA2 (left) and genes that are upregulated in response to AR-V depletion are positively enriched for genes that are regulated GATA2 (right). NES, normalized enrichment score; q, false discovery rate q-value. C, Genes that are induced by GATA2 (LogFC < 0.4) are overexpressed in two CRPC data sets as identified by Urbanucci et al. (37). D, Genes that are induced by GATA2 (LogFC < 0.4) and overexpressed in CRPC (LogFC > 0.4) are AR-V regulated (LogFC < 0.4). E, The GATA2 gene signature was compared with GO data sets using GSEA, and the top 5 most negatively enriched GO gene sets are shown. F, Enrichment plot shows the effect of GATA2 depletion on the GO_CELL_CYCLE gene set. NES, normalized enrichment score; q, false discovery rate q-value. G, CWR22Rv1 cells were cultured in steroid- depleted conditions and transfected with control (siScr) or GATA2 (siGATA2) siRNAs. Confluence was measured every 6 hours for 120 hours. Graph represents mean normalized confluence (n ¼ 3) SEM (, P 0.05; ns, P 0.05).

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1269

Chaytor et al.

Figure 3. GATA2 regulates the transcriptional activity of other oncogenic transcription factors. A, The GATA2 gene signature was compared with Oncogenic signatures data sets using GSEA. Graph shows the top 10 most negatively enriched oncogenic signatures. B, Enrichment plots show that the MYC_UP.V1_UP (left) and E2F1_UP.V1_UP (right) gene sets are negatively enriched for genes that are regulated by GATA2 in CWR22Rv1 cells. C, Enrichment plot shows genes that are induced by c-Myc overexpression in LNCaP cells are signiﬁcantly negatively enriched for genes that are regulated by GATA2 in LNCaP cells. D, c-Myc and E2F1 genomic binding sites in LNCaP cells overlapped with GATA2 genomic binding sites in LNCaP cells.

Flag-tagged BRD4 construct encoding a truncated form of acetylated-lysine residues. To examine the bromodomain require- BRD4, termed BRD4S [amino acids 1–700 containing both bro- ments for the BRD4–GATA2 interaction, HEK293T cells were modomains (BD) 1, BD2, and the extraterminal domain] and transfected with wild-type, BD1 (BRD4SmBD1)-mutant or BD2 subject to endogenous GATA2 immunoprecipitation. Immuno- (BRD4SmBD2)-mutant Flag-tagged BRD4S derivatives for 48 blotting of resultant immunoprecipitates detected the truncated hours and then subjected to GATA2 immunoprecipitation. form of BRD4, suggesting that the BET protein and GATA2 interact Immunoblotting GATA2 immunoprecipitates with an anti-Flag (Fig. 4B). Importantly, a 24-hour treatment of BRD4S-transfected antibody demonstrated that mutation of BD1, but not BD2, cells with 1 mmol/L of the BET inhibitor JQ1 attenuated the severely diminished the GATA2–BRD4S interaction, suggesting GATA2–BRD4S interaction, suggesting that this interaction is that BD1 mediates the BRD4S–GATA2 interaction (Fig. 4C). mediated by the BDs of BRD4S (Fig. 4B). BD1 and BD2 of BET Interestingly, mutation of key residues in BD2 was found to proteins differ vastly in their substrate-binding capacities, but increase the association between BRD4S–GATA2, suggesting that together enable simultaneous, bivalent interaction with distinct BD2 may counteract binding of BRD4 to GATA2 (Fig. 4C). In

1270 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

Figure 4. GATA2 is acetylated and interacts with BET proteins in a BD-dependent manner. A, HEK293T cells were transfected with empty vector (EV) or Flag-GATA2 for 48 hours and treated with a DMSO control or 5 mmol/L trichostatin A (TSA) for 6 hours before immunoprecipitation (IP) with an anti-Flag antibody. Immunoprecipitates were subject to Western blot analysis with anti-Flag or anti-pan acetylated-lysine antibodies. B, HEK293T cells were transfected with an empty vector or Flag-BRD4-short (BRD4S) for 48 hours and treated with a DMSO control or 1 mmol/L JQ1 for 24 hours before IP with an anti-GATA2 antibody. Immunoprecipitates were analyzed by Western blot using an anti-Flag antibody. C, HEK293T cells were transfected with an empty vector, Flag-BRD4S BD1-mutant (BRD4S mBD1), Flag-BRD4S BD2 mutant (BRD4S mBD2) for 48 hours before IP using an anti-GATA2 antibody. Immunoprecipitates were subject to Western blot using an anti-Flag antibody. D, HEK293T cells were transfected with an empty vector, V5-BRD2, or V5-BRD3 for 48 hours before IP with an anti- GATA2 antibody. Immunoprecipitates were analyzed by Western blot using an anti-V5 antibody. For each IP, input samples were probed for Flag and a-tubulin. E, VCaP cells cultured in steroid-depleted conditions for 72 hours were subject to chromatin immunoprecipitation (ChIP) using an anti-BRD4 antibody and then re-ChIP using either anti-GATA2 or isotype control (IgG) antibodies prior to quantitative PCR analysis. F, VCaP cells cultured in steroid-depleted conditions were transiently transfected with control (siScr) or GATA2 (siGATA2) siRNAs for 48 hours prior to ChIP using anti-BRD4 or isotype control (IgG) antibodies and quantitative PCR analysis. G, VCaP cells were cultured as in E and treated for 24 hours with either vehicle or 1 mmol/L JQ1 prior to chromatin immunoprecipitation using either an anti-GATA2 or isotype control (IgG) antibody. Graphs represent mean normalized percentage input (n ¼ 3) SEM (, P 0.05; ns, P 0.05).

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1271

Chaytor et al.

addition to BRD4, we also demonstrate that GATA2 interacts with by GATA2 were less sensitive to JQ1 treatment and vice versa BRD3 and, to a lesser extent, BRD2 (Fig. 4D), indicating possible (Supplementary Fig. S6D). We hypothesized that GATA2 would redundancy between BET family members in potentiating GATA2 be affected by JQ1 at genomic regions co-occupied by the BET regulation. family (BRD2, BRD3, and BRD4) and therefore examined the genome-wide overlap of GATA2 and BET proteins. Fifty-six per- BRD4 and GATA2 are codependent on each other for chromatin cent (3,836/6,901) of the GATA2 cistrome was co-occupied by at occupancy least one BET protein (Fig. 5D; Supplementary Fig. S7A). In Given that both GATA2 and BRD4 are chromatin-associated addition, we found that 64% (659/1,022) of the JQ1-sensitive proteins, we next investigated the interplay between the two GATA2 bound sites were co-occupied by either BRD2, BRD3, proteins at the chromatin level. Re-ChIP experiments performed BRD4, or a combination of BET proteins, such as at the PSA in VCaP cells using consecutive BRD4 and GATA2 immunopre- (KLK3), KLK2, and TMPRSS2 enhancers; Fig. 5E; Supplementary cipitations demonstrated that GATA2 and BRD4 are closely asso- Fig. S7B), suggesting that BET protein occupancy represents ciated with the gene enhancers of PSA and TMPRSS2 (Fig. 4E). a large determinant for GATA2 chromatin binding sensitivity Moreover, depleting GATA2 (siGATA2) for 48 hours in VCaP to JQ1. cells markedly diminished BRD4 chromatin occupancy at PSA To investigate the potential clinical impact of JQ1-mediated and TMPRSS2 cis-regulatory elements compared with control GATA2 loss at chromatin, the most proximal genes were assigned (siScr; Fig. 4F), without affecting total BRD4 protein levels (Sup- to the 1,022 genomic binding sites from which GATA2 was lost in plementary Fig. S5A), which is consistent with previous findings response to BET inhibition. We identified 916 genes that are describing a dependency of BRD3 on GATA1 for chromatin potentially affected by JQ1-mediated loss of GATA2. Importantly, binding (36). We next investigated if a reciprocal relationship 245 of these genes have been demonstrated to be overexpressed in existed between GATA2 and BET proteins by assessing the sensi- CRPC (28, 35, 37), suggesting that JQ1 treatment may attenuate tivity of GATA2 chromatin occupancy to BET inhibition using CRPC-associated gene expression by preventing GATA2 binding 1 mmol/L JQ1. As shown in Fig. 4G, GATA2 enrichment at the PSA to chromatin (Supplementary Fig. S8A; left). One hundred fifty- and TMPRSS2 enhancers in VCaP cells was significantly reduced seven genes were identified which possessed a differentially upon JQ1 treatment compared with the DMSO-treated control bound GATA2 site within a gene promoter (10 kb from TSS). (NT) and was not a consequence of reduced total GATA2 levels in Thirty-one percent (49/157) of these genes were overexpressed response to BET inhibition (Supplementary Fig. S5B). Important- in CRPC, reaffirming that JQ1 treatment may attenuate CRPC- ly, the dependency of the GATA2 chromatin association with BET associated gene expression by preventing GATA2 binding to proteins was confirmed using GATA2 ChIP experiments in chromatin (Supplementary Fig. S8A; right). CWR22Rv1 cells treated with 1 mmol/L JQ1 (Supplementary Fig. S5C). Moreover, unaltered FOXA1 chromatin binding in BET inhibitors affect global GATA2-regulated gene expression response to JQ1 in VCaP cells (Supplementary Fig. S5D) suggests To investigate the effect of BET inhibitors on global GATA2- that BET inhibition does not indiscriminately downregulate regulated gene expression, we assessed the correlation between association of all DNA-binding proteins and confirms codepen- GATA2-regulated genes (26) and those sensitive to the BET dency of GATA2 and BET proteins for chromatin occupancy at inhibitor I-BET762 (31) from two publicly available data sets specific genomic loci. derived from LNCaP cells. Using GSEA, we determined that To understand how JQ1 affects global GATA2 chromatin asso- transcripts downregulated by 0.5 mmol/L I-BET762 were signifi- ciation, we performed GATA2 ChIP-seq in VCaP cells treated with cantly negatively enriched for transcripts that were regulated and without 1 mmol/L JQ1. We chose VCaP cells to conduct these by GATA2 and vice versa (Fig. 6A). Furthermore, 13% of all experiments because ChIP-seq data sets for the BET family mem- significantly deregulated transcripts upon GATA2 depletion bers BRD2, BRD3, and BRD4 were publicly available and hence were strongly (2 LogFC) deregulated by I-BET762 in LNCaP would enable a more robust assessment of the GATA2–BET cells (Supplementary Fig. S8B), suggesting that treatment with protein chromatin interplay in CRPC. As expected, motif analysis I-BET762 affects global GATA2-regulated gene expression. of the 6,901 GATA2 peaks present in both GATA2 ChIP replicates As proof of concept that GATA2 activity is sensitive to BET (termed the consensus peak set; Supplementary Fig. S6A), inhibitors in CRPC, we investigated the effect of JQ1 on expres- derived from the control experiment, identified the GATA2 sion of PRIM1, TRIM14, and RAP1GAP, which are all significantly response element as the most prominent feature of the sequenced overexpressed in advanced prostate cancer (Fig. 6B). Our ChIP- peaks (Fig. 5A). As shown in Supplementary Fig. S6B, GATA2 sequencing data demonstrate that GATA2 is lost from cis-regula- was enriched at cis-regulatory elements upstream from the PSA tory elements of the PRIM1, TRIM14, and RAP1GAP genes in (KLK3), KLK2, and TMPRSS2 genes in both experimental repli- response to JQ1, which are co-occupied by BET family members cates and was reduced in response to 1 mmol/L JQ1 treatment. (Fig. 6C). We validated this finding using ChIP-qPCR in VCaP Furthermore, consistent with previous reports (26), the distribu- cells (Fig. 6D). Furthermore, GATA2 knockdown in VCaP and tion of GATA2-binding sites was varied, but distal intergenic and CWR22Rv1 cells significantly reduced PRIM1 mRNA levels intronic loci were the most abundant genomic sites associated (Fig. 6E; Supplementary Fig. S8C), indicating that PRIM1 is with GATA2 (Fig. 5B) comprising approximately 36% and 26% of directly regulated by GATA2. To investigate if loss of GATA2 the total described peaks, respectively. Comparing the consensus chromatin occupancy by JQ1 affects PRIM1 expression, VCaP GATA2 cistromes between control and 1 mmol/L JQ1 treatment, cells cultured in steroid-depleted conditions were treated with we identified 1,040 GATA2 binding sites (15% of total) that were 1 mmol/L JQ1 for 24 hours. As shown in Fig. 6F, JQ1 treatment significantly differentially bound in response to JQ1 treatment significantly reduced expression of the CRPC-associated gene (Fig. 5C; Supplementary Fig. S6C). Visualization of differentially compared with control. Additionally, PRIM1, TRIM14, and RAP1- bound sites indicated that sites that are more strongly bound GAP mRNA was significantly downregulated in response to

1272 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

Figure 5. JQ1 reduces GATA2 chromatin deposition genome-wide. A, Motif analysis on the 100 most strongly bound GATA2 ChIP-seq peaks. B, GATA2 ChIP-seq peaks annotated to gene features. C, Visualization of the read intensity across the 1,022 sites (2 kb) at which GATA2 binding afﬁnity is decreased upon JQ1 treatment. D, GATA2 genomic binding sites identiﬁed in VCaP cells overlapped with BRD2, BRD3, and BRD4 genomic binding sites (14). E, Visualization of AR/þ DHT, GATA2, GATA2 þ JQ1, BRD2, BRD3, and BRD4 peaks upstream of PSA (KLK3), KLK2,andTMPRSS2.

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1273

Chaytor et al.

Figure 6. BET protein inhibition regulates GATA2-regulated gene expression. A, Gene signatures for GATA2 and 0.5 mmol/L I-BET762 were compared using GSEA with enrichment plots showing genes that are downregulated in response to 0.5 mmol/L I-BET762 treatment are negatively enriched for genes that are regulated GATA2 (left) and genes that are upregulated in response to 0.5 mmol/L I-BET762 treatment are positively enriched for genes that are regulated GATA2 (right). NES, normalized enrichment score; q, false discovery rate q-value. B, PRIM1, TRIM14,andRAP1GAP expression were assessed in 29 BPH samples, 131 localized prostate cancer (PCa) samples and 19 mCRPC samples (28). Outliers are shown as *, &, and ~, respectively (, P 0.05; ns, P 0.05). C, Visualization of GATA2, GATA2 þ JQ1, BRD2, BRD3, and BRD4 enrichment upstream of PRIM1, TRIM14,andRAP1GAP. D, VCaP cells cultured in steroid-depleted conditions were treated for 24 hours with control (NT) or 1 mmol/L JQ1 prior to chromatin immunoprecipitation using an anti-GATA2 or isotype control (IgG) antibody. Graph represents normalized % input of at least n ¼ 2 SEM (, P 0.05; ns, P 0.05). E, VCaP cells cultured in steroid-depleted conditions were transiently transfected with control (siScr) or GATA2 (siGATA2) siRNAs for 72 hours prior to RT-qPCR assessing PRIM1 mRNA expression. Graphs represent mean normalized gene expression; n ¼ 3 SEM (, P 0.05; ns, P 0.05). F, VCaP cells were cultured as in D before RT-qPCR assessing PRIM1 mRNA expression. Graphs represent mean normalized gene expression; n ¼ 3 SEM (, P 0.05; ns, P 0.05). G, PRIM1, TRIM14, and RAP1GAP mRNA expression in response to a 24-hour 0.5 and 10 mmol/L I-BET762 treatment (31). Graph represents gene expression; n ¼ 2 SEM (, P 0.05; ns, P 0.05).

1274 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

0.5 mmol/L and 10 mmol/L I-BET762 in VCaP cells (Fig. 6G), with this observation, we identified that 62% of AR-V–regulated supporting the concept that GATA2-mediated regulation of genes are controlled by both FOXA1 and GATA2 in CWR22Rv1 CRPC-associated genes can be attenuated by BET protein cells, supporting the concept that reconfiguration of chromatin blockade. architecture by multiple pioneer factors is critical for AR-V func- In all, our findings indicate that GATA2 is a key regulator of tion at a large proportion of target loci. AR-V transcriptional activity in CRPC and depletion attenuates Further interrogation of our CWR22Rv1-derived transcrip- cell growth. Furthermore, we show that GATA2 is important for tomic data also provided evidence that GATA2 regulates the controlling other key oncogenic transcription factors which drive activity of other oncogenic transcription factors, including aberrant transcriptional programs in CRPC. Critically, we show c-Myc and E2F1. This was confirmed in silico using transcrip- that GATA2 chromatin occupancy is regulated by BET proteins tomic and cistromic data sets derived from LNCaP cells (30, 34), and treatment of CRPC cells with JQ1 attenuates GATA2 recruit- which indicated considerable DNA-binding site and target ment to chromatin and expression of GATA2-induced CRPC gene overlaps between GATA2 and both c-Myc and E2F1. genes. These data advocate the use of BET inhibitors in GATA2- Consistent with GATA2 regulating the transcriptional compe- overexpressed prostate malignancy. tency of AR-Vs and other key CRPC-associated transcription factors, depletion of GATA2 markedly reduced expression of genes involved in the cell cycle and attenuated proliferation of Discussion CWR22Rv1 cells. Although persistent AR signaling is a key driver of disease Interestingly, after 48-hour knockdown of GATA2 in progression (3, 38), only 50% of CRPC patients will CWR22Rv1 cells, AR-V protein levels appeared to be diminished respond (5, 6) to the current gold-standard AR-targeting treat- to a greater extent than AR-FL (Fig. 1B; Supplementary Fig. S9A), ments and even this is short-lived in the majority of cases. One which is in contrast to prolonged GATA2 depletion which leads to reason for this failure is the expression of numerous alterna- a reduction in both AR-FL and AR-Vs (15, 26). GSEA of our GATA2 tively spliced forms of the receptor termed AR-Vs in approxi- transcriptome identifies several spliceosome-related transcripts mately 80% of CRPC (11). These shortened forms of the AR-FL downregulated in response to GATA2 knockdown, including the drive an androgenic signaling program without requiring characterized AR-FL/AR-V splicing factor SF3B3 (ref. 40; Supple- ligand stimulation and go unchallenged by all clinically rele- mentary Fig. S9B-E), supporting the concept that GATA2 may vant antiandrogens (39), hence pose a major clinical problem. control synthesis of AR-Vs by enabling appropriate spliceosome Although considerable progress has been made in recent years composition. How this would affect AR-Vs preferentially over to understand how AR-Vs facilitate prostate cancer progression, the AR-FL in the short-term, however, is currently unclear, but our knowledge of what AR-Vs interact with and how they are given splicing is tightly linked to transcriptional elongation regulated at the level of chromatin is still lacking. Given that rates (41–43), GATA2 depletion may enable synthesis of AR-FL AR-FL transcriptional activity is highly dependent upon the at the expense of AR-Vs. Further investigation is required to pioneer factors FOXA1 and GATA2 (15, 18, 26), it is likely that determine the mechanism by which this sequential downregula- AR-Vs are also subject to this level of regulation. Indeed, several tion occurs. It is likely that GATA2 regulates AR expression by studies have indicated that FOXA1 facilitates chromatin asso- multiple mechanisms which are not mutually exclusive. Taken ciation of AR-Vs and is important for controlling approximately together, we have conclusively shown that GATA2 is a pioneer 41% of the AR-V transcriptome (12). In contrast, the involve- factor for AR-V–mediated transactivation and functions to control ment of GATA2 in global AR-V regulation is considerably less other transcriptional regulators to enable maintenance of the well characterized with only one study demonstrating ectopi- CRPC phenotype. cally expressed AR-V7–mediated transactivation of FKBP5 is The notion that GATA2 represents an important therapeutic enhanced by GATA2 in LNCaP cells (26). target in CRPC is supported by our data, but selective inhibitors of We therefore comprehensively assessed the role of GATA2 in GATA2 are currently not available for both preclinical and clinical regulating AR-Vs in the physiologically relevant CWR22Rv1 studies. Although GATA2 blockade in vitro and in vivo has been CRPC cell line. Consistent with GATA2-mediated regulation of achieved with the small-molecule inhibitor K7174 (26), how AR-FL (15, 18, 26), we have shown that GATA2 is present at efficacious this compound is in the clinical setting remains cis-regulatory promoter and enhancer elements of the canonical unknown. Therefore, it is important to identify other modes of AR-target genes PSA, KLK2, and TMPRSS2 in maximal castrate GATA2 regulation that can be exploited therapeutically with conditions. Moreover, siRNA-mediated GATA2 depletion dimin- clinically relevant drugs. Recently, inhibitors of the BET family ished AR-V chromatin occupancy at these sites and in turn reduced of proteins, including BRD2, BRD3, and BRD4, have shown utility AR-V–driven PSA, KLK2, and TMPRSS2 expression, indicating in downregulating both AR-FL and AR-V chromatin occupancy GATA2 regulates the activity of AR-Vs at these loci. To assess the and target gene expression in models of CRPC (13, 14). The effect involvement of GATA2 in global AR-V–mediated transactivation, of BET inhibitors on AR-Vs is seemingly multifaceted, with evi- we defined the GATA2 transcriptome in CWR22Rv1 cells and dence indicating impact on both expression and transcriptional compared it to our previously published AR-V gene set (12). potency of AR-Vs (13). Critically however, the interplay between GATA2 regulated approximately 36% of the AR-V transcriptome BET proteins and pioneer factors has not been investigated in the and GSEA indicated significant enrichments between the two gene context of CRPC. Considering that both BRD3 and BRD4 interact sets. Importantly, we observed that a considerable number of with acetylated GATA1 in the GIE erythroid cell line (36), we genes were downregulated by both GATA2 and AR-V knockdown, speculated that GATA2 and members of the BET family would validating GATA2 as a bona fide pioneer factor for AR-Vs. In the interact in CRPC cell lines. We demonstrated that GATA2 was context of AR-FL, both GATA2 and FOXA1 have been shown to co- acetylated and interacted with BRD2, BRD3, and a physiologically occupy 55% of AR-FL cis-regulatory elements (15). Consistent relevant short derivative of BRD4, termed BRD4S. Importantly,

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1275

Chaytor et al.

we found that GATA2 interaction was mediated by BD1 of BRD4S the potential clinical association of the genes neighboring and was sensitive to JQ1 treatment indicating a dependency GATA2-binding peaks that were sensitive to JQ1 treatment. on bromodomain function for the GATA2–BRD4S interaction. Nine hundred sixteen genes were identified as being adjacent to Given the high sequence homology between BRD4S and the GATA2 peaks from our cistromics data. Importantly, from this longer BRD4L isoform, it is likely that the same degree of interplay gene set, 245 were classified as upregulated in CRPC, suggesting exists between GATA2 and BRD4L. that GATA2 and BET family members control an important Close association between GATA2 and BRD4 at cis-regulatory CRPC-associated gene signature that can be attenuated with enhancer elements of the PSA and TMPRSS2 genes, as demon- BET inhibitors. We next assessed the functional dependency of strated by GATA2/BRD4 re-ChIP experiments, suggested that an GATA2-mediated transactivation on BET proteins by examining interplay between GATA2 and BET proteins may be required for two LNCaP transcriptomes: one derived from depletion of their chromatin occupancy in CRPC. ChIP-seq of GATA2 indi- GATA2 and another treated with the BET inhibitor I-BET762. cated that a large proportion of GATA2 genomic binding sites GSEA of the two gene sets demonstrated that transcripts down- in VCaP cells were co-occupied by BRD2, BRD3, and BRD4. regulated by treatment with I-BET762 were significantly nega- Furthermore, JQ1 diminished GATA2 chromatin occupancy at tively enriched for genes regulated by GATA2, suggesting that a subset of GATA2-regulated genes, indicating that there is a BET proteins control a considerable cohort of GATA2-regulated dependency on BET proteins for GATA2 DNA binding at a genes. Given that our cistromics data demonstrated reduced cohort of genomic loci. How JQ1 is causing GATA2 chromatin GATA2 chromatin occupancy at only 15% of global GATA2 disassociation is presently unclear, but widespread changes to binding sites in response to JQ1, dissociation of the pioneer the local chromatin environment, as a consequence of altered factor from chromatin is likely to be one of several mechanisms histone acetylation levels, may drive indiscriminate displace- of diminished GATA2 transcriptional competency in response ment of all transcriptional regulators from target loci (44). to BET inhibitors. Loss of histone acetylation, leading to ele- However, analysis of FOXA1 deposition at AR-V–target gene cis- vated chromatin compaction, or compromised binding/activity regulatoryelementssuggeststhisisnotthecase,withchroma- of other transcriptional coregulators in response to BET inhi- tin occupancy of FOXA1 refractory to BET inhibition at the bition will likely contribute to attenuated GATA2 function analyzed loci. Instead, we speculate that blockade of the BD1- without physical dissociation from DNA. Given that AR-FL dependent interaction between GATA2 and BET proteins by and AR-V activity is sensitive to BET protein inhibition, it is JQ1 untethers GATA2 from co-occupied sites and is sufficient to possible that the effects of JQ1/I-BET762 on global GATA2 diminish GATA2-chromatin binding. Importantly, experiments activity in CWR22Rv1 and VCaP cells could also be attributed in VCaP and CWR22Rv1 cells depleted of GATA2 demonstrated to loss of AR from co-occupied GATA2 and AR bound genes. It the existence of a reciprocal relationship between the two is currently difficult to examine GATA2 transcriptional activity proteins whereby GATA2 is required for BRD4 chromatin independently of AR because of the lack of AR-negative, association. This interdependency between the two proteins GATA2-positive cell lines. To gain a more clear, precise readout for cis-regulatory element binding may, again, be driven by a of how BET inhibition affects GATA2 transcriptional activity, bromodomain-dependent interaction between BRD2/BRD3/ these experiments will have to be conducted in an AR-negative BRD4 and GATA2 or a consequence of diminished access of system. BET proteins to cis-regulatory elements through loss of pioneer Our data demonstrate that GATA2 is a critical regulator of factor activity. Interestingly, we identified a subset of GATA2 AR-Vs, and targeting GATA2 directly may offer new therapeutic genomic binding sites which are co-occupied by BET proteins strategies in the long term. Furthermore, we have uncovered a and are not affected by JQ1. Visualization of differentially critical interdependency between GATA2 and BET family mem- bound GATA2 genomic binding sites in response to JQ1 indi- bers and demonstrate therapeutic sensitivities of GATA2 signaling cated that GATA2 binding is more refractory to JQ1 treatment at to BET inhibitors that could be exploited in the short term. In all, sites that are most strongly bound by GATA2. It is possible that we have demonstrated that GATA2 drives global AR-V activity and GATA2 presence at these strongly bound sites is mediated by have uncovered a CRPC-associated GATA2-driven gene-expres- other factors, and BET proteins are less critical for GATA2 sion signature that is sensitive to JQ1, suggesting that GATA2- and binding. However, it is also possible that there are subsets of AR-V–overexpressing CRPC will be sensitive to clinically relevant BET genomic binding sites that are not affected by JQ1, which BET inhibitors. results in persistent GATA2 genomic binding. Additionally, we fi identi ed 363 GATA2 genomic binding sites that are altered in Disclosure of Potential Conflicts of Interest response to JQ1 treatment, but are not co-occupied by a BET No potential conflicts of interest were disclosed. protein. This may be explained by inhibition of other bromodomain-containing proteins, such as p300 and ATAD2, with Authors' Contributions JQ1, which may also mediate GATA2 genomic binding, but fi Conception and design: L. Chaytor, L. Gaughan have not been identi ed by this study. Future studies using Development of methodology: L. Chaytor, R. Heath, D. Jones Assay for Transposase-Accessible Chromatin (ATAC)-sequenc- Acquisition of data (provided animals, acquired and managed patients, ing and rapid immunoprecipitation mass spectrometry of provided facilities, etc.): L. Chaytor, D. Jones endogenous proteins (RIME) will help to further assess the Analysis and interpretation of data (e.g., statistical analysis, biostatistics, interplay between GATA2 and BET proteins on chromatin and computational analysis): L. Chaytor, M. Simcock, S. Nakjang, L. Walker Writing, review, and/or revision of the manuscript: L. Chaytor, C. Robson, identify other GATA2 interacting bromodomain-containing D. Jones, L. Gaughan proteins. Administrative, technical, or material support (i.e., reporting or organizing To gain an insight into the functional significance of the data, constructing databases): L. Chaytor, R. Heath interplay between GATA2 and BET proteins, we firstly examined Study supervision: C. Robson, L. Gaughan

1276 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

AR-V Dependency on GATA2

Acknowledgments advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate L. Chaytor, M. Simcock, and S. Nakjang were funded by CR UK. D. Jones this fact. was funded by the Medical Research Council (MR/P009972/1).

The costs of publication of this article were defrayed in part by the Received November 16, 2018; revised January 25, 2019; accepted February 28, payment of page charges. This article must therefore be hereby marked 2019; published ﬁrst March 4, 2019.

References 1. Lonergan PE, Tindall DJ. Androgen receptor signaling in prostate cancer 19. Zheng J, Kitajima K, Sakai E, Kimura T, Minegishi N, Yamamoto M, et al. development and progression. J Carcinogen 2012;10:20. Differential effects of GATA-1 on proliferation and differentiation of 2. Harris WP, Mostaghel EA, Nelson PS, Montgomery B. Androgen depriva- erythroid lineage cells. Blood 2006;107:520–7. tion therapy: progress in understanding mechanisms of resistance and 20. Martin DIK, Zon LI, Mutter G, Orkin SH. Expression of an erythroid optimizing androgen depletion. Nat Rev Urol 2009;6:76. transcription factor in megakaryocytic and mast cell lineages. Nature 3. Evans CP, Higano CS, Keane T, Andriole G, Saad F, Iversen P, et al. 1990;344:444. The PREVAIL study: primary outcomes by site and extent of baseline 21. Brewer A, Pizzey J.GATA factors in vertebrate heart development and disease for enzalutamide-treated men with chemotherapy-naive disease. Expert Rev Mol Med 2006;8:1–20. metastatic castration-resistant prostate cancer. Eur Urol 2016;70: 22. Battle MA, Bondow BJ, Iverson MA, Adams SJ, Jandacek RJ, Tso P, et al. 675–83. GATA4 is essential for jejunal function in mice. Gastroenterology 2008; 4. Smith MR, Saad F, Chowdhury S, Oudard S, Hadaschik BA, Graff JN, et al. 135:1676–86. Apalutamide treatment and metastasis-free survival in prostate cancer. 23. Beuling E, Baffour-Awuah NY, Stapleton KA, Aronson BE, Noah TK, N Engl J Med 2018;378:1408–18. Shroyer NF, et al. GATA factors regulate proliferation, differentiation, and 5. Scher HI, Fizazi K, Saad F, Taplin ME, Sternberg CN, Miller K, et al. Increased gene expression in small intestine of mature mice. Gastroenterology 2011; survival with enzalutamide in prostate cancer after chemotherapy. N Engl J 140:1219–29. Med 2012;367:1187–97. 24. Liew CK, Simpson RJ, Kwan AH, Crofts LA, Loughlin FE, Matthews JM, 6. Rathkopf DE, Morris MJ, Fox JJ, Danila DC, Slovin SF, Hager JH, et al. Phase et al. Zinc fingers as protein recognition motifs: structural basis for the I study of ARN-509, a novel antiandrogen, in the treatment of castration- GATA-1/friend of GATA interaction. Proc Natl Acad Sci 2005;102: resistant prostate cancer. J Clin Oncol 2013;31:3525. 583–8. 7. De Laere B, van Dam PJ, Whitington T, Mayrhofer M, Diaz EH, Van den 25. Hayakawa F, Towatari M, Ozawa Y, Tomita A, Privalsky ML, Saito H. Eynden G, et al. Comprehensive profiling of the androgen receptor in Functional regulation of GATA-2 by acetylation. J Leukoc Biol 2004;75: liquid biopsies from castration-resistant prostate cancer reveals novel intra- 529–40. AR structural variation and splice variant expression patterns. Eur Urol 26. He B, Lanz RB, Fiskus W, Geng C, Yi P, Hartig SM, et al. GATA2 facilitates 2017;72:192–200. steroid receptor coactivator recruitment to the androgen receptor complex. 8. Li Y, Hwang TH, Oseth LA, Hauge A, Vessella RL, Schmechel SC, et al. AR Proc Natl Acad Sci U S A 2015;111:18261–6. intragenic deletions linked to androgen receptor splice variant expression 27. Chiang YT, Wang K, Fazli L, Qi RZ, Gleave ME, Collins CC, et al. GATA2 as a and activity in models of prostate cancer progression. Oncogene 2012;31: potential metastasis-driving gene in prostate cancer. Oncotarget 2014;5: 4759. 451. 9. Hu R, Dunn TA, Wei S, Isharwal S, Veltri RW, Humphreys E, et al. Ligand- 28. Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS, et al. independent androgen receptor variants derived from splicing of cryptic Integrative genomic profiling of human prostate cancer. Cancer Cell 2010; exons signify hormone-refractory prostate cancer. Cancer Res 2009;69: 18:11–22. 16–22. 29. O'Neill D, Jones D, Wade M, Grey J, Nakjang S, Guo W, et al. Development 10. Dehm SM, Tindall DJ. Alternatively spliced androgen receptor variants. and exploitation of a novel mutant androgen receptor modelling strategy Endocr Relat Cancer 2011;18:R183–96. to identify new targets for advanced prostate cancer therapy. Oncotarget 11. Sharp A, Coleman I, Yuan W, Sprenger C, Dolling D, Rodrigues DN, et al. 2015;6:26029. Androgen receptor splice variant-7 expression emerges with castration 30. Barfeld SJ, Urbanucci A, Itkonen HM, Fazli L, Hicks JL, Thiede B, et al. c-Myc resistance in prostate cancer. J Clin Invest 2019;129:192–208. antagonises the transcriptional activity of the androgen receptor in prostate 12. Jones D, Wade M, Nakjang S, Chaytor L, Grey J, Robson CN, et al. FOXA1 cancer affecting key gene networks. EBioMedicine 2017;18:83–93. regulates androgen receptor variant activity in models of castrate-resistant 31. Wyce A, Degenhardt Y, Bai Y, Le B, Korenchuk S, Crouthame MC, et al. prostate cancer. Oncotarget 2015;6:29782. Inhibition of BET bromodomain proteins as a therapeutic approach in 13. Chan SC, Selth LA, Li Y, Nyquist MD, Miao L, Bradner JE, et al. prostate cancer. Oncotarget 2013;4:2419. Targeting chromatin binding regulation of constitutively active AR variants 32. Gaughan L, Stockley J, Coffey K, O'Neill D, Jones DL, Wade M, et al. to overcome prostate cancer resistance to endocrine-based therapies. KDM4B is a master regulator of the estrogen receptor signalling cascade. Nucleic Acids Res 2015;43:5880–97. Nucleic Acids Res 2013;41:6892–904. 14. Asangani IA, Dommeti VL, Wang X, Malik R, Cieslik M, Yang R, et al. 33. Schmidt D, Wilson MD, Spyrou C, Brown GD, Hadfield J, Odom DT. Therapeutic targeting of BET bromodomain proteins in castration-resistant ChIP-seq: using high-throughput sequencing to discover protein-DNA prostate cancer. Nature 2014;510:278. interactions. Methods 2009;48:240–8. 15. Wu D, Sunkel B, Chen Z, Liu X, Ye Z, Li Q, et al. Three-tiered role of the 34. Xu H, Xu K, He HH, Zang C, Chen CH, Chen Y, et al. Integrative analysis pioneer factor GATA2 in promoting androgen-dependent gene expression reveals the transcriptional collaboration between EZH2 and E2F1 in the in prostate cancer. Nucleic Acids Res 2014;42:3607–22. regulation of cancer-related gene expression. Mol Cancer Res 2016;14: 16. Wang Q, Li W, Liu XS, Carroll JS, J€anne OA, Keeton EK, et al. A hierarchical 163–72. network of transcription factors governs androgen receptor-dependent 35. Ylipaa A, Kivinummi K, Kohvakka A, Annala M, Latonen L, Scaravilli prostate cancer growth. Mol Cell 2007;27:380–92. M, et al. Transcriptome sequencing reveals PCAT5 as a novel ERG- 17. Lupien M, Eeckhoute J, Meyer CA, Wang Q, Zhang Y, Li W, et al. FoxA1 regulated long noncoding RNA in prostate cancer. Cancer Res 2015; translates epigenetic signatures into enhancer-driven lineage-specific tran- 75:4026–31. scription. Cell 2008;132:958–70. 36. Lamonica JM, Deng W, Kadauke S, Campbell AE, Gamsjaeger R, Wang H, 18. Zhao JC, Fong KW, Jin HJ, Yang YA, Kim J, Yu J. FOXA1 acts upstream of et al. Bromodomain protein Brd3 associates with acetylated GATA1 to GATA2 and AR in hormonal regulation of gene expression. Oncogene promote its chromatin occupancy at erythroid target genes. Proc Natl Acad 2016;35:4335. Sci U S A 2011;108:E159–68.

www.aacrjournals.org Mol Cancer Res; 17(6) June 2019 1277

Chaytor et al.

37. Urbanucci A, Barfeld SJ, Kytol€ €a V, Itkonen HM, Coleman IM, Vodak D, et al. cancer by RNA-binding protein PSF. Proc Natl Acad Sci U S A 2017;114: Androgen receptor deregulation drives bromodomain-mediated chroma- 10461–6. tin alterations in prostate cancer. Cell Rep 2017;19:2045–59. 41. Nogues G, Kadener S, Cramer P, Bentley D, Kornblihtt AR. Transcriptional 38. Hussain M, Fizazi K, Saad F, Rathenborg P, Shore ND, Demirhan E, et al. activators differ in their abilities to control alternative splicing. J Biol Chem PROSPER: A phase 3, randomized, double-blind, placebo (PBO)-con- 2002;277:43110–14. trolled study of enzalutamide (ENZA) in men with nonmetastatic 42. Liu LL, Xie N, Sun S, Plymate S, Mostaghel E, Dong X. Mechanisms of the castration-resistant prostate cancer (M0 CRPC). Am Soc Clin Oncol androgen receptor splicing in prostate cancer cells. Oncogene 2014;33: 2018;36:3. 3140. 39. Li Y, Chan SC, Brand LJ, Hwang TH, Silverstein KA, Dehm SM. Androgen 43. Kornblihtt AR, de la Mata M, Fededa JP, Munoz MJ, Nogues G. Multiple receptor splice variants mediate enzalutamide resistance in castration- links between transcription and splicing. RNA 2004;10:1489–98. resistant prostate cancer cell lines. Cancer Res 2013;73:483–9. 44. Gonzales-Cope M, Sidoli S, Bhanu NV, Won KJ, Garcia BA. Histone H4 40. Takayama KI, Suzuki T, Fujimura T, Yamada Y, Takahashi S, Homma Y, acetylation and the epigenetic reader Brd4 are critical regulators of plur- et al. Dysregulation of spliceosome gene expression in advanced prostate ipotency in embryonic stem cells. BMC Genom 2016;17:95.

1278 Mol Cancer Res; 17(6) June 2019 Molecular Cancer Research

The Pioneering Role of GATA2 in Androgen Receptor Variant Regulation Is Controlled by Bromodomain and Extraterminal Proteins in Castrate-Resistant Prostate Cancer

Lewis Chaytor, Matthew Simcock, Sirintra Nakjang, et al.

Mol Cancer Res 2019;17:1264-1278. Published OnlineFirst March 4, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1541-7786.MCR-18-1231

Supplementary Access the most recent supplemental material at: Material http://mcr.aacrjournals.org/content/suppl/2019/03/02/1541-7786.MCR-18-1231.DC1

Cited articles This article cites 43 articles, 12 of which you can access for free at: http://mcr.aacrjournals.org/content/17/6/1264.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://mcr.aacrjournals.org/content/17/6/1264.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mcr.aacrjournals.org/content/17/6/1264. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.