Functional inversion of circadian regulator REV-ERBα leads to tumorigenic gene reprogramming

Previously, we found that one of the NR CR regulators, RORγ, plays an important role in activation of key tumor drivers and pathways (16, 17). In our continued examination of the function of NR CR regulators in cancer, we observed that like RORγ, elevated NR1D1 expression is significantly associated with poor survival of patients in several different cancers, including prostate cancer, hepatocellular carcinoma (HCC), colon cancer, and gastric cancer (SI Appendix, Fig. S1 A–D↗). In prostate cancer, elevated NR1D1 expression and nuclear REV-ERBα protein overexpression is significantly associated with metastasis (SI Appendix, Fig. S1 E and F↗). We then performed CRISPR knockout or shRNA/siRNA knockdown of the REV-ERBα/NR1D1 gene and found that its depletion significantly diminished the cell growth of castration-resistant prostate cancer (CRPC) (C4-2B, 22RV1, and LAPC4), HCC (Hepa1-6, HepG2, and Hep3B), and colon cancer (SW480, HCT116, and DLD-1) (Fig. 1 A and B and SI Appendix, Fig. S1 G and H↗). As previously reported (13, 20) and in this study (SI Appendix, Fig. S1I↗), those cancer cells displayed no detectable rhythmicity in key CR regulator expression. Interestingly, in cancer cells that retain the rhythmicity (21–23), knockdown of REV-ERBα did not cause significant effects on their growth (SI Appendix, Fig. S1J↗). As expected, the “normal” human prostate epithelial cells (RWPE1) and human hepatocytes (THLE-2) displayed rhythmic patterns of expression of the CR genes, including NR1D1 (SI Appendix, Fig. S1I↗). REV-ERBα depletion in those cells did not result in significant growth inhibition (SI Appendix, Fig. S1G↗).

In our further examination of the function of REV-ERBα, we observed that REV-ERBα knockdown or knockout potently reduced cell viability in organoids of PDX tumors and cell lines (Fig. 1 C and D and SI Appendix, Fig. S1 K and L↗). The knockdown significantly increased cell apoptosis (SI Appendix, Fig. S1M↗). Overexpression of REVE-RBα significantly enhanced colony formation of the CRPC cells (SI Appendix, Fig. S1N↗). Consistent with the effects on the cancer cell growth and survival, knockdown REV-ERBα strongly decreased expression of proteins important for growth, proliferation, and survival including MYC, CDK4, CDK6, CCND2/D3, and BCL2 (Fig. 1E and SI Appendix, Fig. S1O↗).

Next, we examined REV-ERBα functions using a synthetic small-molecule SR8278 that was designed to bind to its ligand binding domain to antagonize its function (24). Consistent with the effects of genetic depletion, SR8278 displayed a significant growth inhibition of the CRPC (C4-2B and 22RV1) and HCC cells (Hepa1-6 and HepG2) (Fig. 1F and SI Appendix, Fig. S1P↗). Similar potent inhibitory effects on the survival of the cancer cells were observed (SI Appendix, Fig. S1Q↗). In line with the effects of REV-ERBα gene silencing, SR8278 strongly elevated cleaved PARP-1 protein level and the activity of caspase3/7 enzymes, and markedly reduced the expression of key proliferation and survival proteins (SI Appendix, Fig. S1 R and S↗). Combined with the results from REV-ERBα knockdown or knockout, the data together suggest that REV-ERBα plays an important role in promoting cancer cell growth and survival where CR is lost.

In normal tissues, REV-ERBα represses the rhythmic induction of key CR regulators (e.g., ARNTL, PER1/2, CRY1) and metabolic regulators and enzymes (e.g., LXR, Insig2, ACACB, and SREBP) (11, 12, 25). Consistently, in “normal” human prostate epithelial cells, REV-ERBα knockdown or its antagonist treatment up-regulated the expression of genes enriched in programs of CR and lipid metabolism (SI Appendix, Fig. S2 A and B↗). Such upregulation of clock and metabolic genes was also observed in “normal” human hepatocytes (SI Appendix, Fig. S2C↗). However, in the cancer cells, REV-ERBα knockdown or its antagonist did not consistently up-regulate the CR regulator genes and key lipid metabolism genes (SI Appendix, Fig. S2D↗), suggesting that REV-ERBα has lost its canonical CR regulator function in cancer cells.

To better understand the functional change of REV-ERBα, we performed RNA-seq analysis of two different CRPC cells treated with REV-ERBα antagonist SR8278 and/or REV-ERBα knockdown. Analysis of genes with expression significantly up-regulated by either antagonist or knockdown did not reveal CR or lipid metabolism genes as the major gene programs. Except p53 signaling pathway, no major tumorigenic programs are highly enriched in the up-regulated genes (SI Appendix, Fig. S2 E and F↗). Intriguingly, neuronal signaling programs and calcium signaling pathway are enriched in the up-regulated genes (SI Appendix, Fig. S2 G and H↗). In contrast, genes with expression significantly down-regulated by the treatments showed a high degree of concordance with around 50% of genes altered by both treatments in the two models (Fig. 2A). KEGG and GO analysis of the 1,789 transcripts commonly down-regulated revealed that genes involved in signaling pathways of MAPK and PI3K-Akt, cell cycle, and DNA replication were among the most highly enriched (Fig. 2B and SI Appendix, Fig. S2I↗). GSEA analysis also showed that pathways of MAPK and PI3K-Akt signaling, KRAS signaling, and cell cycle were highly enriched (SI Appendix, Fig. S2J↗). Analysis of RNA-seq data from HCC cells also revealed that SR8278 and REV-ERBα knockdown down-regulated the expression of genes enriched in the same programs (SI Appendix, Fig. S2 K and L↗). Paradoxical to its function in normal cells, the treatments also resulted in downregulation of many genes in lipid metabolism (Fig. 2B and SI Appendix, Fig. S2 D and I↗).

Our pathway-focused analysis demonstrated that a major proportion of genes in the signaling pathways of MAPK (132 genes out of 301 genes in the pathway) and PI3K-Akt (158 genes out of 359 genes in the pathway) were significantly down-regulated by the REV-ERBα antagonists and REV-ERBα knockdown (Fig. 2B). They include receptor tyrosine kinases (RTKs) (e.g., VEGFR3, EGFR, FGFR2/3/4, and IGF1R), HRAS, RAF, and many members of the MAPK family (e.g., ERK2/3 and JNK1/2), AKT1/3, and PIK3R2-6 (Fig. 2C). Notably, many growth factors (e.g., EGF, IGF1/2, VEGFA/D, FGF2/8/11/19/21/23, and PDGFB/D) were also down-regulated. Consistently, treatment of cells with the REV-ERBα antagonist resulted in a strong downregulation of the signaling proteins and their phosphorylated forms including EGFR, RAS, RAF, AKT, and the MAPKs (e.g., ERK1/2, p38, and MEK) (Fig. 2D and SI Appendix, Fig. S2 M and N↗). Analysis of clinical tumor datasets revealed that REV-ERBα expression is significantly correlated with the expression of MAPK signaling programs in different cancers (Fig. 2E and SI Appendix, Fig. S2O↗). Moreover, cancer cells with a larger number of genes down-regulated tend to be more sensitive to SR8278 treatment in their growth based on the EC₅₀ measurement (SI Appendix, Fig. S2P↗). Together, these results suggest that in certain cancer cells, REV-ERBα has lost its function in repression of clock and metabolic genes, and instead, it plays a previously unrecognized role in stimulation of multiple tumorigenic programs including PI3K-Akt and MAPK signaling pathways, cell cycle, and lipid metabolism.

The unanticipated, paradoxical function of REV-ERBα in stimulating tumorigenic programs prompted us to examine how REV-ERBα switches its gene regulation activities. We first performed REV-ERBα ChIP-seq analysis with the normal cells and cancer cells. Comparison of REV-ERBα cistromes revealed that only a small fraction of sites in cancer cells (2.1% in prostate cancer cells or 1.3 % in the mouse liver cancer cells) are the same as in the corresponding normal cell or tissue (SI Appendix, Fig. S3A↗). This finding indicates clearly that REV-ERBα genome-wide occupancy is dramatically reprogrammed in the cancer cells. Integrated analysis of REV-ERBα ChIP-seq and RNA-seq data revealed that among the 3,649 activated genes in the cancer cells, very small fraction is also activated in the normal cells (23 out of 1,390) (SI Appendix, Fig. S3B↗). Likewise, a small fraction of genes that are directly repressed by REV-ERBα in the cancer cells are also repressed in the normal cells. In contrast, a large number of genes (732 out of 3,649) that are directly activated by REV-ERBα in cancer cells are actually repressed in the normal cells (SI Appendix, Fig. S3B↗). Consistent with previous findings on its function in liver tissue (11) (SI Appendix, Fig. S3C↗), REV-ERBα directly represses genes in CR and lipid metabolism in the “normal” prostate cells (SI Appendix, Fig. S3 D and E↗). Other major programs directly repressed by REV-ERBα include signaling by GnRH and insulin and cell cycle. Intriguingly, REV-ERBα does not directly control programs of PI3K-Akt and MAPK signaling pathways in the normal tissue and cells (SI Appendix, Fig. S3 F↗).

Next, we performed REV-ERBα ChIP-seq with cancer cells. In prostate cancer cells, we found that REV-ERBα bound to a large proportion of genes in the tumorigenic programs as described above, including 168 genes in PI3K-Akt signaling, 140 genes in MAPK signaling, 27 genes in DNA replication, and 105 genes in the cell cycle (Fig. 3 A and B, SI Appendix, Fig. S3 G and H↗, and Dataset S1↗). For example, over 40 % of genes in the MAPK signaling pathway are direct targets of REV-ERBα, including all of the genes listed above in Fig. 2C that were down-regulated by REV-ERBα knockdown and SR8278, thus indicating that they are directly activated by REV-ERBα. Similarly, in HCC cells, REV-ERBα bound to many genes in the tumorigenic programs, including 81 genes in PI3K-Akt signaling, 76 genes in MAPK signaling, and 52 genes in cell cycle, as well as 52 genes in hippo signaling (SI Appendix, Fig. S3 I and J↗ and Dataset S1↗). It is important to note that REV-ERBα antagonist SR8278 strongly disrupted REV-ERBα genome-wide binding and its binding at genes of PI3K-Akt and MAPK signaling programs (Fig. 3 C–E). Interestingly, REV-ERBα directly represses genes enriched in lysosome, focal adhesion, p53 signaling, and ErbB signaling (SI Appendix, Fig. S3 K and L↗).

Next, to understand how REV-ERBα activates, we examined the effects of REV-ERBα antagonist on gene activation-associated histone mark H3K27ac by ChIP-seq and on chromatin openness by ATAC-seq. Our results demonstrated that the antagonist strongly decreased H3K27ac and local chromatin accessibility at both promoter and enhancer regions of the tumorigenic program genes (Fig. 3F and SI Appendix, Fig. S3M↗). K-means analysis of the H3K27ac ChIP-seq data revealed that the effects of the antagonist on H3K27ac mark are closely linked to its effects on REV-ERBα bindings, with decreased H3K27ac mark linked to decreased REV-ERBα bindings in cluster I and increased H3K27ac mark linked to increased REV-ERBα bindings in cluster III (SI Appendix, Fig. S3 N and O↗ and Fig. 3G). Similarly, the antagonist-decreased ATAC-seq peaks were also linked to decreased REV-ERBα bindings (SI Appendix, Fig. S3 N and O↗ and Fig. 3G). Together, these results indicate that REV-ERBα switches from a repressor in normal tissues to a protumorigenic activator to directly stimulate PI3K-Akt and MAPK signaling and other tumorigenic programs. The results also suggested that REV-ERBα activation function involves promoting H3K27ac modification and opening local chromatin at both promoters and enhancers.

In line with a previous study (27), we demonstrated in PLA and ChIP assays that in the “normal” human prostate cells, REV-ERBα associates with NCoR1/NCoR2/HDAC3 corepressor complex in its repression of CR genes (SI Appendix, Fig. S4 A and B↗). Such association was significantly disrupted by its antagonist. Knockdown of the corepressors caused a significantly increased expression of the CR genes, indicating that the corepressors mediate the repression function of REV-ERBα in the prostate cells (SI Appendix, Fig. S4C↗).

As shown above in cancer cells, REV-ERBα activation of genes involves increasing the chromatin accessibility and histone hyperacetylation. We thus hypothesized that proteins that mediate the activation function of REV-ERBα would possess those activities, such as bromodomain-containing proteins. Among them, BRD4 can both bind to acetylated histone tails and disrupt histone DNA interactions at nucleosomes (28) and has been implicated in control of MAPK signaling genes (29, 30). We thus examined the potential involvement of BRD4. Our co-IP and PLAs clearly demonstrated that REV-ERBα and BRD4 associates in the nucleus (Fig. 4 A and B) and that the association was disrupted by SR8278 and AZD5153, suggesting that their interaction involves the LBD of REV-ERBα and BD1 domain of BRD4. Next, we performed BRD4 ChIP-seq in the cancer cells and found that more than 50% of REV-ERBα ChIP-seq peaks overlapped with those of BRD4 (Fig. 4C). The overlapped peaks are enriched in the same tumorigenic programs of REV-ERBα (Dataset S2↗). Consistent with the notion that BRD4 mediates REV-ERBα activation, occupancy of BRD4 at tumorigenic program gene loci was significantly reduced upon SR8278 treatment (Fig. 4D). As expected, treatment of cells with BETi/BRD4 bromodomain inhibitor AZD5153 significantly down-regulated MAPK/PI3K signaling genes (SI Appendix, Fig. S4D↗). Interestingly, the BRD4 inhibitor also potently inhibited REV-ERBα binding at tumorigenic programs (Fig. 4 E and F), suggesting a functional codependence between the two proteins. BRD4 often functions together with protein acetylases. Indeed, our further analysis demonstrated that bromodomain-containing acetylase p300 co-occupied with REV-ERBα at the tumorigenic target and that p300 acetylase inhibitor A485 and SR8278 significantly disrupted the co-occupancy (Fig. 4G and SI Appendix, Fig. S4E↗).

To examine whether in cancer cells, the corepressor complex could mediate the activation function of REV-ERBα, we knocked down NCoR1 and NCoR2 and found that their knockdown did not cause any significant effect on the expression of REV-ERBα activated genes in PI3K and MAPK signaling programs (SI Appendix, Fig. S4F↗). As previously reported (31, 32), their depletion resulted in a significantly decreased or increased expression of their target genes involved in drug resistance. Consistent with a lack of involvement in control of the kinase signaling genes by the corepressors, no significant occupancy was detected at the REV-ERBα binding sites (SI Appendix, Fig. S4G↗). Moreover, no significant association between REV-ERBα and the corepressors was detected by PLA analysis (SI Appendix, Fig. S4H↗). Together, these results suggest that in cancer cells, REV-ERBα switches its cofactor association from the corepressors to coactivator BRD4 and p300 to directly activate tumorigenic gene programs.

REV-ERBα represses genes through binding to sequences characterized as DR2-like REV-REs (11, 12). Consistently, in the normal human epithelial cells, REV-ERBα binds primarily to DR2-like motif (SI Appendix, Fig. S5A↗). However, in cancer cells, REV-ERBα genome-wide binding sites are enriched primarily with REV-RE-like, half-site motif (CA/CA/GAGGTCA) (SI Appendix, Fig. S5B↗), which is in agreement with the notion that REV-ERBα genome occupancy is dramatically reprogrammed in cancer cells. To understand further how REV-ERBα activates the tumorigenic programs, we posited that REV-ERBα may cooperate with a strong transcriptional activator. Interestingly, we found that FOXA1 is one of the top-ranking TFs with motifs enriched (Fig. 5A) and is highly expressed in the different cancer cells. We thus examined whether FOXA1 cooperates with REV-ERBα. Our PLA analysis detected a marked endogenous association between REV-ERBα and FOXA1 in the different cancer cells (Fig. 5B and SI Appendix, Fig. S5C↗). Notably, REV-ERBα antagonist SR8278 significantly decreased the interactions. Our co-IP analysis demonstrated a similar interaction and effects by the antagonist (SI Appendix, Fig. S5D↗).

We then performed FOXA1 cistrome analysis. Interestingly, approximately 35% of FOXA1 bindings are at promoters (SI Appendix, Fig. S5E↗). Analysis of ChIP-seq peaks revealed that over 30% of REV-ERBα peaks overlapped with FOXA1 peaks in different cancer cells (Fig. 5C and SI Appendix, Fig. S5F↗). KEGG analysis showed that the overlapped peaks are linked primarily to genes of PI3K-Akt and MAPK signaling (e.g., pathways of MAPK, VEGF, PI3K-Akt, and Ras) (Fig. 5D, SI Appendix, Fig. S5G↗, and Dataset S3↗). Interestingly, genes in circadian entrainment program including primarily neuronal signaling genes, but not CR program, are also targets of REV-ERBα and FOXA1. Indeed, knockdown of FOXA1 significantly decreased the expression of genes in the PI3K-Akt, MAPK signaling pathways as well as in AR signaling programs as previously reported (33) (Fig. 5E and SI Appendix, Fig. S5 H and I↗). Integrated analysis of ChIP-seq and RNA-seq data revealed that FOXA1 directly stimulates the expression of a large number of genes in PI3K-Akt signaling (57 out of 144, such as FLT1/4, FGFR1/3/4, PIK3R4/5, and VEGFB/D), MAPK cascade/signaling (94 out of 237, such as MAPK1/3/4/6), and Ras protein signal transduction (148 out of 333, such as HRAS, NRAS, and RAF1) (Dataset S4↗). As expected, FOXA1 knockdown strongly decreased the protein expression of the signaling components and their phosphorylated forms including EGFR, RAS, RAF, AKT, and the MAPKs (SI Appendix, Fig. S5J↗). Interestingly, REV-ERBα antagonist SR8278 decreased FOXA1 genome-wide binding and its binding at most of the genes in MAPK, PI3K-Akt, and Ras signaling programs, such as MAPK8/JNK1 and KRAS (Fig. 5 F–H). Close inspection of ChIP-seq peak regions revealed that REV-ERBα and FOXA1 bind closely at promoters (e.g., 60 bp or 160 bp apart in PIK3C2B and KRAS, respectively) to sequences that are consistent with their known binding motifs (SI Appendix, Fig. S5K↗). Together, these results suggest that REV-ERBα cooperates closely with FOXA1 in directly stimulating the expression of the tumorigenic programs in cancer cells.

Our finding that REV-ERBα antagonist markedly alters FOXA1 bindings raises the possibility that REV-ERBα plays an important role in facilitating FOXA1 chromatin occupancy. To further examine how REV-ERBα regulates FOXA1 cistrome, we ectopically expressed REV-ERBα in the cancer cells. Subsequent FOXA1 ChIP-seq analysis revealed that elevated REV-ERBα strongly increased FOXA1 genome-wide binding (Fig. 6A and SI Appendix, Fig. S6A↗). Comparison of FOXA1 ChIP-seq peaks from REV-ERBα-overexpressing (OE) cells with the peaks from control cells revealed dramatic changes in FOXA1 genome-wide bindings. Specifically, 6,468 of FOXA1 ChIP-seq peaks were lost and 21,401 peaks were gained in the OE cells (SI Appendix, Fig. S6B↗). A major proportion of gained peaks are at regions that are close to TSS (approximately 30% at promoters, and the rest at potential enhancers, SI Appendix, Fig. S6C↗). Our KEGG analysis of gained peaks and peaks with increased heights revealed that MAPK and PI3K-Akt signaling pathway genes were among the top enriched ones (Fig. 6B, SI Appendix, Fig. S6 D and E↗, and Dataset S5↗). Through comparing genes that were down-regulated by FOXA1 knockdown in REV-ERBα OE cells versus control cells, we identified 272 genes that displayed newly gained FOXA1 peaks and were down-regulated only in the OE cells. We thus define those genes as new targets of FOXA1 in REV-ERBα OE cells (Dataset S6↗). They include MAP3K6, MECOM, and RASA3 that are involved in MAPK signaling, and FMN2, CS, PLPP4, IDH2, and ACSS1 that are involved in metabolism (Fig. 6C). As expected, REV-ERBα OE significantly increased the expression of MAPK and PI3K-Akt signaling pathway genes (SI Appendix, Fig. S6F↗). Knockdown of FOXA1 significantly decreased the induction by REV-ERBα OE (SI Appendix, Fig. S6F↗). Motif analysis of the newly gained FOXA1 binding sites on the tumorigenic programs showed that the FOXA1 motif and motifs for other FKHD TFs are among the highly enriched ones (SI Appendix, Fig. S6G↗), suggesting that REV-ERBα promotes FOXA1 binding to the preexisting FOXA1 sites.

Pioneer factors are known to scan chromatin for transient association of their potential binding sites. Their stable binding to the nucleosome DNA sites is modulated by many factors including the alteration of local chromatin structure (34). Recent studies also show that pioneer factors can also function in a nonpioneering way (35). Indeed, studies showed that a subset of FOXA1 genome binding sites can be reprogrammed either by steroid hormone-activated GR and ER in breast cancer cells (36) or by neuroendocrine prostate cancer driver TFs ASCL1 and NKX2-1 (37). We speculated that REV-ERBα reprogramming of FOXA1 binding involves changes in local chromatin openness. We thus performed ATAC-seq analysis and found that REV-ERBα OE caused a genome-wide increase of chromatin accessibility (SI Appendix, Fig. S6H↗). K-means analysis of the ATAC-seq data revealed that ATAC-seq peaks gained by REV-ERBα OE corresponded to the group of ChIP-seq peaks with gained and increased FOXA1 peaks (Fig. 6D and SI Appendix, Fig. S6I↗). Motif analysis of gained and increased ATAC-seq peaks showed that the FOXA1 motif is the most highly enriched (SI Appendix, Fig. S6J↗). These results suggest that overexpressed REV-ERBα can reprogram a subset of FOXA1 genome-wide bindings to tumorigenic programs through increasing local chromatin accessibility.

Given that BDR4 mediates REV-ERBα activator function, we examined whether BRD4 is involved in FOXA1 reprogramming by REV-ERBα. Indeed, comparison of BRD4 ChIP-seq peaks from REV-ERBα OE cells with the peaks in control cells revealed that REV-ERBα OE caused a strong reprogramming of BRD4 genome-wide binding (SI Appendix, Fig. S6K↗). Specifically, 18,205 of BRD4 peaks were lost and 28,715 peaks were gained in the OE cells. BRD4 occupancy was significantly enhanced in tumorigenic program genes in REV-ERBα OE cells (SI Appendix, Fig. S6L↗). We then performed ChIP-seq analysis of FOXA1 in REV-ERBα OE cells treated with BETi/BRD4 inhibitor AZD5153. In line with our hypothesis that BRD4 plays a crucial role, AZD5153 erased 70 % of the FOXA1 peaks that were induced by the overexpressed REV-ERBα while it had only marginal effects on the FOXA1 peaks that were unchanged or decreased by REV-ERBα (Fig. 6E and SI Appendix, Fig. S6M↗). Further analysis indicated that MAPK and PI3K-Akt programs are the most severely affected programs (SI Appendix, Fig. S6 N–P↗). Moreover, treatment of cells with AZD5153 significantly down-regulated MAPK/PI3K signaling genes in REV-ERBα OE cells (Fig. 6F). Together, these results suggest that BRD4 plays a crucial role in mediating REV-ERBα reprogramming of FOXA1.

To examine whether the REV-ERBα, FOXA1, and BRD4 regulatory axis plays a function in vivo in control of tumorigenic gene programs, we first performed FOXA1 and REV-ERBα ChIP-seq analysis with a PDX tumor model (38, 39). We found that over 70% of FOXA1 peaks were overlapped with REV-ERBα peaks (Fig. 7A). Consistent with the results from the cell models, the overlapped peaks-linked genes were highly enriched in MAPK and PI3K-Akt signaling programs (Fig. 7B). Strikingly, majority of genes in MAPK signaling (175 out of total 301 in KEGG) and PI3K-Akt signaling (180 out of total 359 in KEGG) were cotargeted by FOXA1 and REV-ERBα (Dataset S7↗). Like the effects observed in the cell culture model, treatment of mice with REV-ERBα antagonist SR8278 markedly decreased FOXA1 and REV-ERBα bindings and chromatin accessibility at both genome-wide and specific programs such as MAPK and PI3K-Akt signaling in the PDX tumors (Fig. 7 C–E and SI Appendix, Fig. S7 A and B↗).

We then treated mice bearing C4-2B-derived xenograft tumors or the PDX tumors with SR8278 and BRD4 inhibitor JQ1 either alone or in combination. Treatment of mice carrying the C4-2B tumors with two different doses of SR8278 for over 3 wk showed a significant, dose-dependent inhibition of the tumor growth in both tumor volume and mass (SI Appendix, Fig. S7 E and F↗). Treatment of mice bearing the PDX tumors with a relatively low dose of either SR8278 (10 mg/kg) or JQ1 (30 mg/kg) for 3 wk showed a significant inhibition of the tumor growth (Fig. 7F and SI Appendix, Fig. S7C↗). Their combined treatment caused a sustained inhibition of the PDX tumor growth (Fig. 7F and SI Appendix, Fig. S7C↗). Importantly, the different treatments had no significant effects on the mouse body weight and their blood levels of ALT, AST, ALP, BUN, and other measurements (SI Appendix, Fig. S7 D, G, and H↗).

Transcriptome analysis revealed that consistent with its tumor inhibition effects, SR8278 treatment significantly decreased expression of genes that are highly enriched in cell growth, cell cycle, and cell proliferation (SI Appendix, Fig. S7I↗). Likewise, JQ1 treatment decreased expression of genes that are highly enriched in the same programs (SI Appendix, Fig. S7J↗). Notably, about 40% of genes in MAPK (123 out of total 301) and about 30% of genes in PI3K-Akt (105 out of total 359) signaling pathways were also inhibited. Interestingly, their combined treatment down-regulated over 50% of the genes in the two signaling pathways (SI Appendix, Fig. S7 K and L↗). Indeed, they include many RTKs (e.g., NGFR, KIT, RET, VEGFR3, EGFR, FGFR1/3/4, and IGF1R), KRAS, RRAS, RAF, members of the MAPK family (e.g., MAPK3/6/8/9 and JNK1/3), AKT1/2/3, PIK3CA, and PIK3R2/3. Notably, many growth factors (e.g., EGF, IGF2, VEGFA/B, FGF11/18, PDGFC, BDNF, and GDNF) were also down-regulated. Taken together, these results strongly indicate that REV-ERBα and FOXA1 cooperates to directly control the tumorigenic programs such as MAPK/PI3K signaling with BRD4 as a mediator and suggest that SR8278 and JQ1 can be effective in treatment of tumors featuring the REV-ERBα, FOXA1, and BRD4 regulatory axis.

Y.Y., R.M.E., and H.-W.C. designed research; Y.Y., X. Zhang, D.C., X. Zheng, J.Z., and J.X.Z. performed research; X. Zhao, A.D.B., M.A.D., E.C., N.M., X.C., and H.-W.C. contributed new reagents/analytic tools; Y.Y., X. Zhang, H.-J.K., J.J.L., and H.-W.C. analyzed data; and Y.Y., M.D., R.M.E., and H.-W.C. wrote the paper.

The authors declare no competing interest.

All raw data and processed information for RNA-seq (GSE260725↗) (53), ChIP-seq (GSE260726↗) (54), and ATAC-seq (GSE260727↗) (55) in this study have been deposited in the Gene Expression Omnibus (GEO). All other data are included in the manuscript and/or supporting information↗.