CHD4 Promotes Breast Cancer Progression as a Coactivator of Hypoxia-Inducible Factors

Hypoxia, a hallmark of the microenvironment of breast tumors, contributes to the pathogenesis of human breast cancer and has been associated with increased risk of mortality in patients with breast cancer (1). The transcription factor hypoxia-inducible factor (HIF), consisting of HIFα and HIFβ subunits (2), is a master regulator of response to hypoxia in breast tumors and activates its target genes to promote angiogenesis, cell proliferation/survival, metabolic reprogramming, epigenetic reprogramming, pH homeostasis, immune evasion, tumor invasion, and metastasis (3). Three HIF family members have been identified in mammals, and HIF1 and HIF2 are highly expressed in breast tumors (4–6). The transcriptional activity of HIF1 and HIF2 is regulated indirectly by the ubiquitin-dependent protein degradation of α subunit (7). Several ubiquitin E3 ligases including cullin 2, CHIP, HAF, and Parkin have been shown to regulate the protein stability of HIF1α and/or HIF2α (8–11). Accumulating studies uncover an essential role of epigenetic regulators in transactivation of HIF1 and HIF2 in cancer cells (12). Multiple histone acetyltransferases/deacetylases, histone/DNA methyltransferases/demethylases, and histone readers, including p300, CBP, ZMYND8, BRD4, JMJD2C, CDK8, Tip60, JMJD1A, G9a, HDAC7, and TET1, are known to function as HIF coregulators (13–19). We and others recently found that RNA polymerase II (RNA Pol II) is preloaded and paused at the promoter of HIF target genes under normoxia and then released upon hypoxia to mediate transcriptional elongation of HIF target genes in breast cancer cells (14, 20). HIF coactivator ZMYND8 controls RNA Pol II–mediated elongation of HIF target genes in breast cancer cells under hypoxia by recruiting BRD4 (14). However, the fundamental mechanism underlying RNA Pol II recruitment to HIF target genes for its pausing under normoxia remains unknown.

Chromodomain helicase DNA-binding protein 4 (CHD4) is a key catalytic subunit of the ATP-dependent chromatin remodeler nucleosome remodeling and deacetylase (NuRD) complex, which plays a critical role in gene regulation by modulating chromatin structure and assembly (21). It has been well documented that CHD4 is involved in gene repression in normal and cancer cells. CHD4 together with the NuRD complex controls the accessibility of the chromatin and mediates downregulation of many genes, whose products are involved in DNA damage response, tumor malignancy, cardiac development, and stem cell fate (22–25). However, the role of CHD4 in gene activation is less studied.

In this study, we found that CHD4 physically interacts with HIF1α and HIF2α and enhances HIF downstream target gene expression by increasing recruitment of HIF and RNA Pol II in breast cancer cells. CHD4 increases breast tumor growth in vitro and in mice through coactivating HIF. Together, these studies uncover a new fundamental mechanism of HIF activation in breast cancer.

Full-length CHD4 cDNA was generated by PCR using mouse CHD4 cDNA (Harvard plasmid, MmCD00320214) as a template and cloned into HpaI/NheI-linearized pcFUGW-3XFLAG vector. FLAG-CHD4 (K750R) was generated by site-directed PCR mutagenesis. Human MBD3 and MTA2 cDNAs were amplified from MDA-MB-321 cell cDNA by PCR and cloned into pcDNA3.1-V5 vector. cDNAs encoding human truncated HIF2α domains were cloned to pGEX-6P-1 vector. DNA oligonucleotides of the short hairpin RNAs (shRNA) targeting CHD4 (Supplementary Table S1) were annealed and ligated into AgeI/EcoRI-linearized pLKO.1 vector (Addgene, catalog no. 8435) or Teton-pLKO (Addgene, catalog no. 21915). Other constructs have been described previously (14, 15, 26). The DNA sequences of all recombinant plasmids were confirmed by nucleotide sequence analysis.

MDA-MB-231 (from Rolf Brekken at UT Southwestern in 2015), MCF7 (from ATCC in 2014), T47D, HeLa, HEK293T (from Gregg L. Semenza at Johns Hopkins in 2014), and HEK293FT (from Thermo Fisher Scientific in 2014) cells were cultured in DMEM or RPMI1640 supplemented with 10% heat-inactivated FBS at 37°C in a 5% CO₂/95% air incubator. Hypoxic cells were placed in a modular incubator chamber (Billups-Rothenberg) flushed with a gas mixture of 1% O₂, 5% CO₂, and balanced N₂. Cells within six passages were used for experiments. Cells were transfected using PolyJet (SignaGen) or FuGENE6 (Promega). All cell lines are Mycoplasma-free and have been authenticated by short tandem repeat DNA profiling analysis in 2016–2017. Lentiviruses encoding scrambled control shRNA (shSC), shCHD4#1, or shCHD4#2 were generated by transfecting HEK293FT cells with the transducing vector and packaging vectors pMD2.G and psPAX2. 48 hours later, virus particles in the medium were harvested, filtered, and transduced into cancer cells. Forty-eight hours after infection, cells were selected for 4 days using 2.5 μg/mL puromycin. CHD4 protein levels were verified by immunoblot assay.

Total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific), treated with DNase I (Thermo Fisher Scientific), and reverse transcribed using an iScript cDNA Synthesis Kit (Bio-Rad). Real-time PCR was performed by a CFX-96 real-time system (Bio-Rad) using iTaq universal SYBR green supermix (Bio-Rad) with primers listed in Supplementary Table S2. The fold change in target mRNA expression was calculated on the basis of the threshold cycle (C_t) as 2^−Δ(ΔC_t), where |$\Delta {C_{\rm{t}}}\, = \,{C_{\rm{t}}}_{_{{\rm{target}}}} - \,{C_{\rm{t}}}_{_{{\rm{18S}}\,{\rm{rRNA}}}}$| and |$\Delta (\Delta {C_{\rm{t}}})\, = \,\Delta {C_{\rm{t}}}_{_{{\rm{1\% }}{{\rm{O}}_{\rm{2}}}}} - \,\Delta {C_{\rm{t}}}_{_{{\rm{20\% }}{{\rm{O}}_{\rm{2}}}}}$|⁠.

MDA-MB-231-shSC and -shCHD4 cells were exposed to 20% or 1% O₂ for 24 hours. Total RNA was isolated using a RNeasy Mini kit and treated with DNase (Qiagen). The quality of total RNA was confirmed with a RNA integrity number score 8.5 or higher by the Agilent Tapestation 4200. A total of 200 ng of DNA-free total RNA was used for library preparation with a Kapa mRNA HyperPrep Kit (Roche). Briefly, mRNA with poly(A) was purified and fragmented before strand-specific cDNA synthesis. cDNAs were end repaired, A-tailed, and ligated with the indexed sequencing adapters. After cDNAs were amplified by PCR and purified with Kapa pure beads, they were sequenced on the Illumina NextSeq500 with the read configuration as 75 bp, single end. Each sample had approximately 25 million to 35 million reads. The Fastq files were subjected to quality check using fastqc (version 0.11.2, http://www.bioinformatics.babraham.ac.uk/projects/fastqc) and fastq_screen (version 0.4.4, http://www.bioinformatics.babraham.ac.uk/projects/fastq_screen), and trimmed using fastq-mcf (ea-utils/version 1.1.2-806, https://github.com/ExpressionAnalysis/eautils). Trimmed fastq files were mapped to hg19 (UCSC version from igenomes) using Tophat. Duplicates were marked using picard-tools (version 1.127, https://broadinstitute.github.io/picard/). Read counts were generated using featureCounts and the differential expression analysis was performed using edgeR with a FDR <0.05 and mRNA fold change >1.5 as a cutoff.

Cells were exposed to 20% or 1% O₂ for 24 hours, cross-linked with 1% formaldehyde for 10 minutes at room temperature and quenched in 0.125 mol/L glycine. Cells were lyzed in lysis buffer (50 mmol/L Tris-HCl, 10 mmol/L EDTA, 1% SDS, protease inhibitor cocktail), sonicated, and subjected to immunoprecipitation overnight in the presence of salmon sperm DNA/protein A beads with antibodies against CHD4 (Proteintech, 14173-1-AP), HIF1α (homemade), RNA Pol II (Abcam, ab817), or IgG at 4°C. Precipitated chromatin DNA was extensively washed, eluted with freshly prepared elution buffer (0.1 mol/L NaHCO₃, 1% SDS), decross-linked at 65°C for 4 hours followed by treatment with proteinase K at 45°C for 45 minutes, purified with phenol/chloroform/isoamyl alcohol (25:24:1, v/v), and quantified by real-time qPCR assay. The primers used for chromatin immunoprecipitation (ChIP)-qPCR are listed in Supplementary Table S2. Fold enrichment was calculated on the basis of C_t as 2^−Δ(ΔC_t), where |$\Delta {C_{\rm{t}}}\, = \,{C_{\rm{t}}}_{_{{\rm{IP}}}} - \,{C_{\rm{t}}}_{_{{\rm{Input}}}}$| and |$\Delta ( {\Delta {C_{\rm{t}}}} )\, = \,\Delta {C_{\rm{t}}}_{{\rm{antibody}}} - \,\Delta {C_{\rm{t}}}_{_{{\rm{IgG}}}}$|⁠.

Cells were lysed in NETN lysis buffer (150 mmol/L NaCl, 1 mmol/L EDTA, 10 mmol/L Tris-HCl, pH 8.0, 0.5% NP-40 and protease inhibitor cocktail) and subjected to IP with indicated antibodies at 4°C for 2 hours, and then protein G magnetic beads (Bio-Rad) were added for further incubation overnight. Next day, proteins bound on the beads were washed four times with NETN lysis buffer, boiled in 1× Laemmli buffer for 6 minutes, and fractionated by SDS-PAGE, followed by immunoblot assay using indicated antibodies.

GST-HIF1α and GST-HIF2α domain proteins were expressed in Escherichia coli BL21-Gold (DE3) and purified by binding to glutathione Sepharose beads (GE Healthcare). The purified GST-HIF1α and GST-HIF2α domains and GST proteins were incubated overnight at 4°C in the presence of glutathione Sepharose beads with lysates from HEK293T cells transfected with FLAG-CHD4 vector. Proteins bound on the beads were washed, boiled in 1× Laemmli buffer for 6 minutes, and subsequently subjected to SDS-PAGE and immunoblot assay.

Cells were plated on 48-well plates and transiently transfected with HIF luciferase reporter plasmid p2.1 containing a HRE (5′-ACGTG-3′) from human ENO1 gene (27); control reporter plasmid pSV-Renilla; shSC, shCHD4#1, shCHD4#2, pcFUGW-3xFLAG-CHD4 [wild-type (WT) or K750R], pcDNA3.1-MBD3, pcDNA3.1-MTA2, or empty vector (EV). Twenty-four hours later, cells were exposed to 20% or 1% O₂ for 24 hours. The firefly and Renilla luciferase activities were measured using the Dual Luciferase Reporter Assay System (Promega).

A total of 100 cells were seeded on a 6-well plate and cultured under 20% or 1% O₂ for 14 days. Subsequently, the colonies were stained with 0.1% crystal violet (Sigma). Only positive colonies (diameter >50 μm) were counted.

A total of 5 × 10⁴ of cells were resuspended in serum-free medium and seeded in a Matrigel-coated transwell insert. Cell culture medium with 10% FBS will be placed at the bottom chamber. Cells were exposed to 20% or 1% O₂ for 24 hours. Cells that invade to the lower side of the transwell insert will be fixed with methanol, stained with 0.1% crystal violet, and counted.

Animal experiments were approved by the Animal Care and Use Committee at UT Southwestern Medical Center (Dallas, TX). A total of 2 × 10⁶ of cells in 100 μL PBS/Matrigel (1:1, Corning) were injected into the second left mammary fat pad of female NSG or NOD/SCID mice (6–8 weeks old). Mice that were implanted with MCF7 cells were administrated subcutaneously with a slow-release 17β-estradiol pellet (0.72 mg/60-day release/pellet) one day before cell implantation. Tumor volume was measured with a caliper every 3 days starting day 14–16 after cell implantation and calculated according to the formula: V = 0.52 × L × H × W (V: volume, L: length, H: height, W: width). Primary tumors were harvested, photographed, and weighed when their volume reached to about 1,500 mm³.

Statistical analysis was performed by two-tailed Student t test between two groups, and one-way or two-way ANOVA with multiple testing correction within multiple groups. Kaplan–Meier survival curve was analyzed by log-rank test. RNA-sequencing (RNA-seq) was repeated twice, and other experiments were repeated at least three times. Data were expressed as mean ± SEM. P < 0.05 is considered significant.

The RNA-seq data were deposited at the Gene Expression Omnibus database with the accession number GSE129293.

To determine the role of CHD4 in HIF transactivation in human breast cancer, we investigated whether CHD4 physically interacts with HIF1α and HIF2α in human breast cancer cells. MDA-MB-231 cells were exposed to 1% O₂ for 6 hours and the whole cell lysates were subjected to coimmunoprecipitation (co-IP) assay with anti-HIF1α antibody or control antibody IgG, followed by immunoblot analysis. HIF1α IP pulled down endogenous CHD4 (Fig. 1A), indicating the physical interaction of HIF1α and CHD4. Likewise, co-IP assay showed that endogenous HIF2α interacted with endogenous CHD4 in hypoxic MDA-MB-231 cells (Fig. 1B). We further detected the physical interaction between endogenous HIF1β and CHD4 in hypoxic MDA-MB-231 cells (Fig. 1C). Similar results were also observed in another breast cancer cell line T47D (Supplementary Fig. S1A–S1C). We next performed a reciprocal co-IP assay in MDA-MB-231 cells exposed to 1% O₂ for 6 hours. Anti-CHD4 antibody, but not control IgG, coprecipitated endogenous CHD4, HIF1α, and HIF2α in hypoxic MDA-MB-231 cells (Fig. 1D). These data indicate that CHD4 physically interacts with both HIF1 and HIF2 in human breast cancer cells under hypoxia.

Next, we mapped the HIF1α and HIF2α domains binding to CHD4. HEK293T cells were transfected with FLAG-CHD4 vector and the whole cell lysates were incubated with a series of truncated GST-HIF1α or GST-HIF2α proteins that were expressed and purified from bacteria, followed by immunoblot assay with anti-CHD4 antibody or Ponceau S staining. CHD4 had a robust interaction with the transactivation domain of HIF1α (531–826 amino acids) or HIF2α (450–870 amino acids), but not other domains of HIF1α or HIF2α and GST (Fig. 1E and F). These data indicate that CHD4 solely binds to the C-terminal transactivation domain of HIF1α and HIF2α.

To determine whether CHD4 increases HIF transcriptional activity, we first performed HIF luciferase reporter assay. HeLa cells were selected for the assay because of their high transfection efficiency. Cells were cotransfected with a HIF luciferase reporter (p2.1), pSV-Renilla, and EV or FLAG-CHD4 expression vector, and exposed to 20% or 1% O₂ for 24 hours. Ectopic expression of WT FLAG-CHD4 significantly increased HIF luciferase reporter activity in hypoxic HeLa cells, compared with EV (Fig. 2A). HIF1/2α double knockout (DKO) completely abolished CHD4's effect on HIF transcriptional activity in hypoxic HeLa cells (Fig. 2B), indicating specific coactivation of HIF1/2 by CHD4. To determine whether the NuRD complex is involved in CHD4-mediated HIF activation, we studied the effect of helicase-dead CHD4 mutant (K750R) and other subunits of the NuRD complex on HIF transcriptional activity. CHD4 (K750R) significantly increased HIF transcriptional activity, similar to WT CHD4 (Fig. 2A). In line with this, CHD4 (K750R) interacted with FLAG-HIF1α as efficiently as WT CHD4 in hypoxic HEK293T cells (Fig. 2C). The other two subunits of the NuRD complex MBD3 and MTA2 both failed to increase HIF transcriptional activity in HeLa cells under hypoxia (Fig. 2D), further suggesting that the NuRD complex does not control HIF transcriptional activity. To complement gain-of-function studies, we performed loss-of-function studies using two independent CHD4 shRNAs. Knockdown (KD) of CHD4 by either of the shRNAs significantly decreased HIF transcriptional activity in hypoxic HeLa cells, compared with shSC (Fig. 2E). Note that the incomplete suppression of HIF transcriptional activity in CHD4 KD cells is possibly due to the KD efficiency of CHD4 by transient transfection of CHD4 shRNAs as well as the redundancy of other HIF coactivators. CHD4 KD had no effect on protein levels of HIF1α or HIF2α in MDA-MB-231 cells (Fig. 2F), indicating that HIF activation by CHD4 is not caused by HIF1/2α protein stability. Together, these data indicate that CHD4 enhances HIF transcriptional activity independent of its helicase activity and the NuRD complex.

Next, we analyzed the effect of CHD4 on HIF downstream target gene expression in breast cancer cells. To this end, we generated two independent CHD4 KD MDA-MB-231 cell lines by transducing lentivirus carrying CHD4 shRNA#1 or #2. Either of the CHD4 shRNAs robustly reduced CHD4 protein levels in MDA-MB-231 cells under 20% or 1% O₂ (Fig. 2F). RNA-seq was performed in shSC and CHD4 KD#1 MDA-MB-231 cells exposed to 20% or 1% O₂ for 24 hours. We found that 220 genes were downregulated in CHD4 KD cells, compared with shSC cells under hypoxia (mRNA fold change >1.5, FDR <0.05; Fig. 2G). Likewise, we identified 269 HIF1/2–inducible genes in MDA-MB-231 cells (Fig. 2G). Intriguingly, 81 genes were upregulated by both CHD4 and HIF1/2, which represented 30% of HIF-inducible genes (Fig. 2G). Gene ontology analysis showed that these overlapped genes participated in several HIF-dependent biological processes, including angiogenesis, response to hypoxia, extracellular matrix organization, collagen catabolic process, apoptosis, and cell proliferation (Fig. 2H; Supplementary Table S3). qRT-PCR analysis validated that the transcription of HIF target genes, VEGFA, ANGPTL4, LOX, and NDNF, was significantly decreased by CHD4 KD#1 or #2 in hypoxic MDA-MB-231 cells (Fig. 2I). However, mRNA transcription of the non-HIF target gene RPL13A was not altered by CHD4 KD#1 or #2 in MDA-MB-231 cells (Fig. 2I), indicating specificity of activation of HIF target genes by CHD4 under hypoxia. These data indicate that CHD4 enhances expression of a subset of HIF downstream target genes in breast cancer cells under hypoxia.

To study the role of CHD4 in HIF1 binding to the HRE, we first examined whether CHD4 localizes on the chromatin near the HRE by ChIP-qPCR assay in MDA-MB-231 cells exposed to 20% or 1% O₂ for 24 hours. CHD4 was associated with the chromatin regions near the HRE of VEGFA, LOX, and ANGPTL4, as compared with control IgG in nonhypoxic MDA-MB-231 cells, and hypoxia significantly increased CHD4 occupancy on these HIF target genes by 2–3.7 folds, but not on RPL13A (Fig. 3A–D). To determine whether HIF1/2α are required for hypoxia-induced CHD4 binding to HIF target genes, we compared CHD4 occupancy in parental and HIF1/2α DKO MDA-MB-231 cells exposed with 20% or 1% O₂ for 24 hours (Fig. 3E). HIF1/2α DKO abolished hypoxia-induced CHD4 occupancy on VEGFA, LOX, and ANGPTL4 but not RPL13A (Fig. 3A–D). These findings indicate that HIF1/2α control CHD4 binding to HIF target genes in breast cancer cells under hypoxia.

We next generated doxycycline-inducible CHD4 KD MDA-MB-231 cells (Fig. 3F) and studied whether CHD4 KD attenuates HIF1 binding to the HRE. ChIP-qPCR results showed that HIF1α strongly bound to the HRE of VEGFA, LOX, and ANGPTL4 genes in hypoxic MDA-MB-231 cells, as expected (Fig. 3G–I). CHD4 KD almost completely abolished HIF1α occupancy at these HREs under hypoxia (Fig. 3G–I). Neither hypoxia nor CHD4 KD affected HIF1α enrichment on the non-HIF target gene RPL13A (Fig. 3J), indicating the specific effect of CHD4 on HIF1 recruitment to the HRE. These findings indicate that CHD4 promotes HIF1 binding to the HRE of its target genes in hypoxic breast cancer cells.

To investigate whether CHD4 regulates RNA Pol II recruitment to HIF target genes, we first studied the physical interaction between CHD4 and RNA Pol II by co-IP assay. Anti-CHD4 antibody, but not IgG, precipitated RNA Pol II in MDA-MB-231 cells under 20% O₂, and hypoxia had no effect on CHD4-RNA Pol II interaction (Fig. 4A). We further found by ChIP-qPCR assay that CHD4 KD#1 or #2 significantly decreased RNA Pol II enrichment at the promoter of the VEGFA and LOX genes in MDA-MB-231 cells under 20% O₂ (Fig. 4B and C). Although reduced RNA Pol II occupancy at VEGFA and LOX was also found in CDH4 KD MDA-MB-231 cells under 1% O₂, it was less robust as compared with that under normoxic conditions (Fig. 4B and C). These data indicate that CHD4 increases recruitment of RNA Pol II to HIF target genes to enhance their transcription in breast cancer cells.

RNA Pol II recruitment to HIF target genes under normoxia is a prerequisite for HIF-mediated transactivation. Thus, we investigated the molecular mechanism of CHD4-mediated RNA Pol II recruitment under normoxia. It was recently reported that the homolog of p300/CREB-binding protein regulates RNA Pol II pausing and release in Drosophila (28). Given that p300 is a well-known HIF coactivator (13), we hypothesized that p300 modulates CHD4-RNA Pol II interaction for RNA Pol II recruitment under normoxia. To address this, we studied whether CHD4 interacts with p300 in breast cancer cells under normoxia. Co-IP assay showed that anti-CHD4 antibody pulled down endogenous p300 in MCF7 cells (Fig. 4D). Conversely, anti-p300 antibody also precipitated endogenous CHD4 in MCF7 cells (Fig. 4E). KD of p300 completely abolished CHD4-RNA Pol II interaction in MCF7 cells (Fig. 4F). These data indicate that p300 is required for CHD4 binding to RNA Pol II in breast cancer cells.

To determine the oncogenic role of CHD4 in breast cancer cells in vitro, we first performed clonogenic assay. Single cell suspension of MDA-MB-231-shSC, -CHD4 KD#1, and -KD#2 cells were cultured under 20% or 1% O₂ for 14 days. Hypoxia markedly increased colony formation of MDA-MB-231-shSC cells (Supplementary Fig. S2A and S2B). CHD4 KD#1 or #2 almost eliminated colony growth under 20% and 1% O₂ (Supplementary Fig. S2A and S2B). To determine specificity of CHD4 KD on colony growth, we generated CHD4-rescued MDA-MB-231 cells (Fig. 5A). Reexpression of FLAG-CHD4 in MDA-MB-231-CHD4 KD#1 cells by lentiviral transduction completely rescued cell growth under 20% and 1% O₂ (Fig. 5B and C). These findings demonstrate that CHD4 increases breast cancer cell colony formation in vitro.

Next, we performed Boyden chamber invasion assay to investigate the effect of CHD4 on breast cancer cell invasion in vitro. Hypoxia significantly increased invasion of MDA-MB-231-shSC cells (Fig. 5D and E; Supplementary Fig. S2C and S2D). CHD4 KD#1 or #2 significantly decreased cell invasion under 20% and 1% O₂ (Supplementary Fig. S2C and S2D). Importantly, reintroduction of FLAG-CHD4 to MDA-MB-231-CHD4 KD#1 cells rescued cell invasion under 20% and 1% O₂ (Fig. 5D and E). These findings indicate that CHD4 increases breast cancer cell motility in vitro.

We next investigated the role of CHD4 in tumor growth in an orthotopic breast cancer xenograft mouse model. CHD4 KD in MDA-MB-231 cells by either of the CHD4 shRNAs significantly attenuated breast tumor growth in NSG mice (Supplementary Fig. S3A–S3C). Immunoblot assay confirmed the CHD4 KD efficiency in CHD4 KD breast tumors as compared with shSC tumors (Supplementary Fig. S3D). Reexpression of FLAG-CHD4 reversed tumor inhibition conferred by CHD4 KD#1 (Fig. 6A–C). Consistently, CHD4 KD#1 in MCF7 cells also significantly suppressed tumor growth in NSG mice implanted with slow-released 17β-estradiol pellets (Supplementary Fig. S3E–S3H). Expression of HIF target genes ANGPTL4, NDNF, and LOX, but not the non-HIF target gene RPL13A, seemed to be reduced in CHD4 KD MCF7 tumors (Fig. 6D). To study whether HIF is required for CHD4-mediated breast tumor progression, we generated parental and HIF1/2α DKO MDA-MB-231 cells overexpressing FLAG-CHD4 or EV (Fig. 6E) and these cells were implanted into the mammary fat pad of female NOD/SCID mice. Overexpression of FLAG-CHD4 in MDA-MB-231 cells significantly increased breast tumor growth in mice, which was abolished by HIF1/2α DKO (Fig. 6F–H). Therefore, these data indicate that CHD4 promotes breast tumor growth in mice through coactivating HIF.

To determine clinical relevance of CHD4 in human breast cancer, we retrieved the CHD4 expression data from The Cancer Genome Atlas (TCGA) breast cancer dataset and found that CHD4 was amplified along with the significantly high mRNA expression in a subset of human breast tumors (Fig. 7A). Consistently, CHD4 mRNA levels were significantly elevated in both primary and metastatic breast tumors as compared with adjacent normal breast tissues (Fig. 7B). CHD4 mRNA upregulation was observed in all breast cancer subtypes including luminal A, luminal B, HER2⁺ and basal-like, and stages 1–4 of human breast tumors (Fig. 7C and D). Remarkably, CHD4 upregulation was mutually exclusive to other known HIF coactivators ZMYND8, KDM4C (JMJD2C), CDK8, CREBBP (CBP), and KAT5 whose mRNAs were also upregulated in human breast tumors, but not with EP300 (p300) that was not upregulated in breast tumors (Fig. 7E), suggesting heterogeneity of epigenetic regulation of HIF in breast cancer. Furthermore, we found that CHD4 expression was positively correlated with expression of HIF target genes LOX and VEGFA in human breast tumors (Fig. 7F and G), which supports our findings about upregulation of LOX and VEGFA by CHD4 in hypoxic breast cancer cells (Fig. 2I). Finally, Kaplan–Meier analysis of the TCGA dataset showed that high CHD4 mRNA levels were positively correlated with poor overall survival of patients with breast cancer (Fig. 7H). Collectively, these data indicate that CHD4 is upregulated and positively associated with HIF target genes in human breast cancer, and that CHD4 is an independent risk factor for women with breast cancer.

In this study, we identified CHD4 as a novel HIF coactivator in breast cancer cells. CHD4 is enriched at the chromatin regions near the HRE and induces a subset of HIF target genes in breast cancer cells under hypoxia and in xenograft tumors. These genes are involved in angiogenesis, collagen remodeling, apoptosis, and cell proliferation, which are the classical HIF-mediated biological functions involved in breast tumor progression. CHD4 is also positively correlated with HIF target genes LOX and VEGFA in human breast tumors, strongly supporting HIF transactivation by CHD4 in vivo. CHD4 is known to mediate gene repression through the NuRD complex. A recent study showed that CHD4 represses hypoxia-induced expression of Ripk3 in murine embryonic endothelial cells (29). Although we did not observe the same regulation for Ripk3 as Ripk3 expression is almost undetectable in human breast cancer cells, we found by RNA-seq that several CHD4-repressed genes can be induced by hypoxia. Neither hypoxia nor HIF is involved in CHD4-mediated gene repression, indicating specificity of CHD4 coactivator function for HIF under hypoxia. Our current studies showed that the coactivator function of CHD4 is independent of its helicase activity and the NuRD complex in hypoxic cells. These findings support an idea that CHD4 is a peripheral component of the NuRD complex and can function independently (21). A recent study also reported coactivation of PAX3-FOXO1 target genes by CHD4 in alveolar rhabdomyosarcoma, although it remains unclear about the role of the NuRD complex in this context (30). Importantly, we found that CHD4 physically interacts with p300, suggesting that free CHD4 outside the NuRD complex is associated with the transcriptional coactivator complex to exert its coactivator function in breast cancer cells.

Our mechanistic studies uncover two layers of HIF regulation by CHD4 in breast cancer cells: (i) increased HIF binding to the HRE and (ii) increased recruitment of RNA Pol II. We showed that CHD4 occupies at the promoter of HIF target genes under normoxia. Previous studies reported that the plant zinc finger homeodomain mediates CHD4 binding to histone H3 and their interaction is enhanced by acetylation of lysine 9 of histone H3 (H3K9ac; ref. 31). We have previously shown H3K9ac at the promoter of HIF target genes (26), suggesting that H3K9ac may be responsible for the recruitment of CHD4 to HIF target genes under normoxia. Interestingly, CHD4 and HIF1/2α are mutually recruited to HIF target genes in breast cancer cells under hypoxia. Our biochemical studies showed that CHD4 binds to the transactivation domain of HIF1/2α. This physical interaction may mediate HIF recruitment to the HRE, thereby enhancing transcription of HIF downstream target genes. Identification of the amino acid residues of HIF1/2α and CHD4 involved in their protein–protein interaction would help address this hypothesis.

RNA Pol II is preloaded and paused on HIF target genes under normoxia (14, 20), which offers a fine and rapid regulatory mechanism of HIF target gene transcription in response to hypoxia. Our recent studies showed that ZMYND8 triggers RNA Pol II–mediated transcriptional elongation of HIF target genes to enhance their expression in breast cancer cells (14). Unlike ZMYND8, CHD4 is required for RNA Pol II loading to HIF target genes in breast cancer cells and this effect is more robust under normoxia as compared with that under hypoxic conditions, indicating its critical role in RNA Pol II recruitment for pausing under normoxia.

Previous studies have discovered many transcriptional coactivators for HIF in breast cancer cells (14, 16, 17, 32). These coactivators are often upregulated in breast tumors and few gene mutations have been identified. Moreover, upregulation of CHD4 and other known HIF coactivators including ZMYND8, KDM4C (JMJD2C), CDK8, CREBBP (CBP), and KAT5 occurs in the mutually exclusive subset of patients with breast cancer according to the TCGA breast cancer dataset, suggesting that the individual HIF coactivator may have a dominant function for HIF transcriptional activity in the particular population of patients with breast cancer. Although some of HIF coactivators (e.g., CHD4 and ZMYND8) have the redundant function for induction of HIF target genes (e.g., LOX and VEGFA), the heterogenous pattern highlights a great importance of comprehensive identification of HIF coactivators including CHD4 toward understanding of regulation of hypoxia responses in breast cancer cells.

Our in vitro functional studies here showed that CHD4 KD decreases colony growth and cell invasion under both normoxia and hypoxia. CHD4's oncogenic functions under normoxia may also require HIF as MDA-MB-231 cells express the low but detectable levels of HIF1/2α under normoxia, but we cannot exclude a possible involvement of the gene repression and DNA damage functions of CHD4 in colony growth and cell invasion under normoxia (33–35). It is worth pointing out that our in vivo xenograft studies showed that HIF is required for CHD4-mediated breast tumor growth in mice. As breast tumor is hypoxic (36), the CHD4-HIF axis appears to play a key role in tumor progression in vivo.

In conclusion, our biochemical, cancer biology, and bioinformatics studies in human breast cancer cells, breast cancer mouse models, and human patients with breast cancer suggest a working model (Fig. 7I): CHD4 interacts with p300 and mediates the recruitment of RNA Pol II to HIF target genes for pausing under normoxia. Upon hypoxia, CHD4 interacts with HIF1/2α and increases their recruitment to the HRE, thereby enhancing transcription of a subset of HIF downstream target genes that promote breast cancer progression. CHD4 is highly expressed in breast tumors and a potential prognostic factor of survival of patients with breast cancer. Importantly, CHD4 is positively correlated with HIF target genes LOX and VEGFA in human breast tumors. Therefore, the CHD4-HIF pathway has clinical relevance and disrupting this signaling pathway may be an invaluable approach for the treatment of breast cancer.

No potential conflicts of interest were disclosed.

Y. Wang: Formal analysis, investigation, writing-original draft, writing-review and editing. Y. Chen: Formal analysis, investigation, writing-original draft, writing-review and editing. L. Bao: Formal analysis, investigation, writing-review and editing. B. Zhang: Formal analysis, investigation, writing-review and editing. J. Wang: Investigation, writing-review and editing. A. Kumar: Formal analysis, writing-review and editing. C. Xing: Formal analysis, writing-review and editing. Y. Wang: Conceptualization, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing. W. Luo: Conceptualization, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

This work was supported by grants from the Susan G. Komen (CCR16376227), the NIH (R01CA222393 and R00CA168746), the CPRIT (RR140036 and RP190358), the Mary Kay Foundation (08-19), and the Welch Foundation (I-1903) to W. Luo, and the NIH (R00NS078049, R35GM124693, and R01AG066166), the CPRIT (RP170671), and the Welch Foundation (I-1939) to Y. Wang. W. Luo is a CPRIT Scholar in Cancer Research.

Y. Wang (CPRIT); Y. Chen (Susan G Komen); L. Bao (NIH, Welch); B. Zhang (CPRIT); J.E. Wang (NIH); A. Kumar (UTSW); C. Xing (UTSW); Y. Wang (NIH, CPRIT, Welch, UTSW); W. Luo (NIH, CPRIT, Welch, Susan G Komen, Mary Kay, UTSW).

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