Stromal induction of BRD4 phosphorylation Results in Chromatin Remodeling and BET inhibitor Resistan

BRD4, a Bromodomain and Extraterminal (BET) protein family member, is a promising anti-cancer drug target. However, resistance to BET inhibitors targeting BRD4 is common in solid tumors. Here, we show that cancer-associated fibroblast (CAF)-activated stromal signaling, interleukin-6/8-JAK2, induces BRD4 phosphorylation at tyrosine 97/98 in colorectal cancer, resulting in BRD4 stabilization due to interaction with the deubiquitinase UCHL3. BRD4 phosphorylation at tyrosine 97/98 also displays increased binding to chromatin but reduced binding to BET inhibitors, resulting in resistance to BET inhibitors. We further show that BRD4 phosphorylation promotes interaction with STAT3 to induce chromatin remodeling through concurrent binding to enhancers and super-enhancers, supporting a tumor-promoting transcriptional program. Inhibition of IL6/IL8-JAK2 signaling abolishes BRD4 phosphorylation and sensitizes BET inhibitors in vitro and in vivo. Our study reveals a stromal mechanism for BRD4 activation and BET inhibitor resistance, which provides a rationale for developing strategies to treat CRC more effectively.

BET family proteins, in particular, the BRD4, which are important transcriptional and epigenetic regulators, play critical roles during cancer development by modulating chromatin activity to facilitate oncogenic transcription1, thus elicited a great level of interest in developing BET inhibitors (BETi) for cancer treatment2. Currently, multiple BETi has shown the promise to induce cancer cell apoptosis and tumor regression and have entered multiple clinical trials including CRC and numerous other solid cancers. However, BET inhibitor resistance often emerges in various cancer types. In particular, BETi in solid tumors are not as effective as in hematological malignance. Moreover, although BETi is responsive in vitro it shows only a modest effect in vivo xenograft model3. These indicate a possible extrinsic resistance mechanism associated with tumor microenvironment (TME).

Current studies have been focusing on the cancer cell-autonomous mechanism of resistance to BETi including “adaptive kinome reprogramming”4,5, “transcriptional remodeling”6,7. Moreover, it has been shown that inactivation mutation of SPOP, an E3 ubiquitin ligase of BRD4, or dysregulation of the deubiquitinase DUB3 result in BRD4 protein stabilization causing intrinsic resistance to BETi8,9,10,11. Recently, several BET inhibitors such as dBET1, MZ1, which employed proteolysis-targeting chimeras (PROTACs) technology to degrade BRD412,13,14, have been developed to conquer BET resistance caused by BRD4 stabilization. However, resistance to BET-PROTACs has also been reported15, although the mechanism need to be further investigated. Of note, there are several key residues of BRD4 in the binding pocket of JQ1 including Tyrosine 97 which are important for JQ1 to inhibit BRD4 activity16,17. The modification of these residues may change the binding affinity of BETi to BRD4, which results in treatment failure with current BET inhibitors including BET-PROTACs.

Pro-inflammatory factors secreted from the tumor microenvironment are important hallmarks of cancer18, which contributes to almost every aspect of tumorigenesis, metastasis and therapeutic resistance19,20,21,22. Although pro-inflammatory factors are well known to promote tumorigenesis via activating vital signaling pathways, it is less investigated whether and how they crosstalk with cancer epigenetic landscape including chromatin modulation to shape the biological behaviors of the tumor. Here, we sought to examine the role of TME-secreted inflammatory factors in the epigenetic regulation of CRC. Using patient-derived CRC cells, we profiled the chromatin regulators of CRC induced by inflammatory cytokines and identified bromodomain protein BRD4 as a key epigenetic modulator which can be activated through a paracrine IL6- and IL8-mediated JAK2 signaling. We further show the TME-induced BRD4 activation provides a previously unrecognized mechanism for chromatin remodeling and resistance to BRD4 inhibitors. Our study provides insights into TME regulation of tumor activity and a combination strategy to overcome therapy resistance.

Pro-inflammatory cytokines interleukin 6 and interleukin 8 induce BRD4 protein expression in cancer cells

To investigate the effect of the inflammatory cytokines on chromatin activity in CRC, we treated the patient-derived colorectal cancer cell line (PDC1) with a panel of 14 recombinant cytokines, which are well-known to be involved in the inflammatory tumor microenvironment (Fig. 1a). The protein levels of major epigenetic regulators, including histone readers, writers, and erasers and associated histone makers, were analyzed by immunoblotting (Figs. 1a and S1a). Among them, we found that the recombinant interleukin 6 (rIL6) and interleukin 8 (rIL8) markedly induced the expression of histone reader protein BRD4, but did not affect the other BET family members BRD2 and BRD3 nor the other chromatin regulators (Fig. S1a). A further time course analysis using two patient-derived CRC lines (PDC1 and PDC2) confirmed that the rIL6 or rIL8 could induce BRD4 expression as early as 4 h, without affecting BRD2 and BRD3 (Fig. 1b, c). Of note, Janus kinase 2 (JAK2), which is the downstream kinase of IL6/8, was activated as early as 2 h prior to the BRD4 induction (Fig. 1c), raising a possibility that JAK2 might be a potential upstream kinase responsible for BRD4 induction. In alignment with this hypothesis, co-treatment with a JAK2 inhibitor pacritinib abolished the rIL6/8-induced BRD4 expression (Fig. 1d), but this was rescued by proteasome inhibitor MG132 (Fig. 1e). This finding indicates that increased BRD4 expression is associated with increased protein stability. Indeed, we found that rIL6 or rIL8 treatment did not increase BRD4 mRNA levels (Fig. S1b). The rIL6/8-induced BRD4 protein expression was also confirmed in several commercial colorectal cancer cell lines (Fig. 1f). Interestingly, it was also seen in cell lines of other cancer types, including breast, lung, and prostate (Fig. S1c). These findings indicate that IL6/8-induction of BRD4 is common and well conserved in human cancers.

a Illustration depicting a screening strategy to investigate the tumor microenvironment-derived cytokines and their proposed function on epigenetic remodeling in CRC. b, c Representative western blot analysis of indicated proteins in patient-derived cancer cells PDC1 (b, n = 3) and PDC2 (c, n = 2) after treatment with 20 ng/ml rIL6 or 20 ng/ml rIL8 for indicated time. d Representative western blot analysis (n = 2) of indicated proteins in patient-derived cancer cells treated with/without 20 ng/ml rIL6 (20 ng/ml), 20 ng/ml rIL8, or 2.5 μM pacritinib. e Representative western blot analysis (n = 2) of indicated proteins in patient-derived cancer cells treated with/without 20 ng/ml rIL6, 20 ng/ml rIL8, 2.5 μM pacritinib, or 20 μM MG132. f Representative western blot analysis (n = 2) of indicated proteins in colon cancer cell lines (HCT116, HT-29, SW480) treated with 5 ng/ml rIL6 or rIL8.

Full size image

JAK2 induces BRD4 protein stabilization through phosphorylation at tyrosine 97 and tyrosine 98

To validate the BRD4 stabilization by JAK2, we transfected 293T cells with a flag-tagged BRD4, together with a wild-type JAK2 (JAK2-WT) or a constitutively activated JAK2 (JAK2-V617F) and examined the BRD4 protein turnover in the presence of protein synthesis inhibitor cycloheximide. The result showed that the active JAK2-V617F, but not the JAK2-WT, induced a remarkable induction of flag-BRD4 (Fig. 2a). Upon cycloheximide treatment, JAK2-V617F-induced BRD4 retained its levels over time, indicating a prolonged half-life, while the BRD4 in mock or JAK2-WT-transfected cells showed rapid degradation (Figs. 2aand S2A). We further confirmed that JAK2-V617F but not the kinase-dead JAK2 (JAK2-K882R), was also able to induce the endogenous BRD4 (Fig. S2b).

a Representative western blot analysis (n = 2) of indicated proteins in 293T cells transfected with indicated constructs, treated with 50 μg/ml cycloheximide (CHX) and harvested at different time points. b Representative western blot analysis (n = 3) of whole-cell lysate (WCL) and samples from co-IP with anti-BRD4 antibody in PDC1 cells treated with 20 ng/ml rIL6 or 2.5 μM pacritinib as indicated. c Scheme depicting functional domains of BRD4 and prediction of BRD4 phosphorylation sites by JAK2. The tyrosine residues predictively phosphorylated by JAK2 were highlighted in red. d Representative western blot analysis (n = 3) of indicated proteins in 293T cells transfected to express the indicated BRD4 mutants together with wild-type JAK2, active JAK2-V617F, or inactive JAK2-K882R. e Representative western blot analysis (n = 2) of tyrosine phosphorylation on BRD4-BD1 and its mutants in 293T cells transfected to express BRD4-BD1 and indicated BRD4-BD1 mutants together with wild-type JAK2 or active JAK2-V617F. f Representative western blot analysis (n = 3) of indicated proteins in 293T cells transfected with BRD4 and its mutants together with active JAK2-V617F. g Representative western blot analysis (n = 2) of phosphorylation of BRD4 peptides (BD1 and BD2) from in vitro kinase assay by active JAK2-V617F or inactive JAK2-K882R. hMass spectrum analysis of phosphorylated residues of BRD4 peptide by active JAK2-V617F. iRepresentative western blot analysis (n = 3) of anti-Flag immunoprecipitates derived from 293T cells transfected with the indicated constructs followed by incubation with active JAK2-V617F in in vitro kinase assay. j Representative western blot analysis (n = 2) of WCL from PDC1 cells treated with indicated cytokines, pacritinib, and 20 μM MG132. k Representative western blot analysis (n = 2) of immunoprecipitates by anti-BRD4 antibody from SW480 cells treated with indicated cytokines, pacritinib, and 20 μM MG132.

Full size image

To investigate whether BRD4 can be phosphorylated by tyrosine kinase JAK2, we detected tyrosine phosphorylation on BRD4 after immunoprecipitation. Immunoprecipitated BRD4 exhibited increased tyrosine phosphorylation in cells treated with rIL6, but not in cells co-treated with JAK2 inhibitor pacritinib (Fig. 2b), supporting that BRD4 is phosphorylated by IL6-JAK2 signaling. To predict the potential tyrosine phosphorylation sites on BRD4 by JAK2, we employed the Group-based Prediction System23, which lead to the prediction of four tyrosine residues (Y97, Y98, Y390, and Y599) that were potentially phosphorylatable by JAK2 (Fig. 2c). Notably, Y97/98 and Y390 are located within the bromodomains BD1 and BD2, respectively, which are known to be essential for the epigenetic reader activity of BRD416,17. Mutagenesis analysis through tyrosine to non-phosphorylated alanine substitutions (Y to A) found that Y97/98 in BD1, but not Y390 or Y599, was required for JAK2-induced BRD4 stabilization (Figs. 2d and S2c), and adding proteasome inhibitor MG132 restored the stability of BRD4-Y97/98A mutant (Fig. S2d). Further dissection of Y97 and Y98 individually indicated that both tyrosine residues were required for JAK2-induced BRD4 phosphorylation (Fig. 2e) and stabilization (Fig. 2f). Interestingly, Y97 but not Y98 was found to be conserved among BET family members including BRD2 and BRD3 (Fig. S2e). This may explain that only BRD4 could be stabilized by active JAK2. Moreover, the Y97/98A mutant showed much higher levels of ubiquitination compared to other mutants (Fig. S2f). By contrast, the tyrosine-to-glutamic acid (Y to E) mutant Y97/98E that mimics constitutive phosphorylation of tyrosine residues showed an increased protein level, which was compatible with the wild-type BRD4 induced by active JAK2 (Fig. S2g). Consistently, the phosphor-mimic Y97/98E mutant showed a much lower level of ubiquitination when compared with the Y97/98A mutant (Fig. S2h). Together, these data stressed that the JAK2-induced phosphorylation of BRD4 at Y97/98 determined the stabilization of BRD4.

To confirm that BRD4 is a direct substrate of JAK2, we next performed the in vitro kinase assay. We show that the active JAK2 (V617F) but not the kinase-dead JAK2 (K882R) can induce strong tyrosine phosphorylation of the synthetic peptide of BRD4-BD1, but slight phosphorylation of BRD4-BD2 (Fig. 2g). Mass-Spectrum analysis of the phosphorylated BD1 domain peptide confirms the abundant phosphorylation at Y97, though the Y98 phosphorylation was modest (Fig. 2h). We then developed a polyclonal antibody raised specifically against the Y97/98 phosphorylated peptide of BRD4 (Fig. S3a), and confirmed that the phosphor-Y97/98 BRD4 antibody recognized specifically the JAK2-induced BD1 domain phosphorylation in the in vitro kinase assay (Fig. S3b, c). The phosphor-Y97/98 antibody also recognized the wild-type BRD4 but not the BRD4-Y97/98A mutant (Figs. 2iand S3d); it also detected the phosphorylation of endogenous BRD4 in cells treated rIL6 and rIL8, which was diminished by pacritinib (Fig. 2j, k). Taken together, we concluded that the BRD4 is a bona fide substrate of JAK2, and the phosphorylation at tyrosine 97/98 is critical for the stabilization of BRD4.

UCHL3 deubiquitinase interacts with and stabilizes phosphorylated BRD4

Having shown that increased BRD4 stabilization is associated with reduced ubiquitination, we set out to investigate the de-ubiquitination events to enable JAK2-mediated BRD4 stabilization. To this end, we performed a deubiquitinase (DUB) library screening in which ubiquitinated BRD4 as a substrate was incubated with a set of DUBs, followed by measurement of the ubiquitination level of BRD4 by western blotting (Fig. S4a). This experiment leads to the identification of 16 DUBs that showed marked de-ubiquitin activity towards BRD4 (Figs. 3a and S4a). To determine which of these DUBs are relevant in CRC, we compared their expression in a gene expression dataset of CRC and adjacent normal controls24. We found that OTUB2, UCHL3, and USP7 are the top DUBs overexpressed in CRC as compared to their normal adjacent tissues with high ectopic expression frequency and significant p value (Frequency > 60%, p value < 0.01) (Fig. 3b). To determine their involvement in regulating BRD4, we performed knockdown of these DUBs in the presence of rIL6, together with USP6 as a negative control and DUB3, which has been previously shown to regulate BRD48. The result shows that UCHL3 knockdown was most efficient in reducing BRD4 expression (Figs. 3c and S4b, c) and shortened BRD4 half-life (Fig. S4d, e), and this effect can be restored by proteasome inhibitor MG132 treatment (Figs. 3d and S4f). Consistently, ubiquitination assay showed knockdown of UCHL3 promoted the poly-ubiquitination of BRD4 (Fig. 3e). We also performed UCHL3 knockdown when active JAK2 was overexpressed. The induction of BRD4 by active JAK2 was significantly diminished by UCHL3 knockdown (Fig. S4g). Whereas, neither protein level nor mRNA level of UCHL3 itself was not affected by JAK2 activation (Fig. S4g, h). It has been recently reported that the E3 ligase SPOP induces BRD4 degradation9,10,11. Interestingly, we found that the concomitant knockdown of SPOP with UCHL3 rescued the effect of UCHL3 knockdown on BRD4 protein (Fig. S4i), indicating that UCHL3 antagonizes SPOP to maintain BRD4 stability. Conversely, ectopic expression of UCHL3 prolonged the half-life of BRD4 (Fig. S4j); It also increased the protein stability of WT BRD4 but not BRD4-Y97/98A mutant (Fig. 3f). Conversely, the UCHL3knockdown also abolished JAK2-induced BRD4 stabilization (Fig. 3g). Of note, our experiments were performed in the presence of IL6/8. whether other deubiquitinases are preferential in other contexts need to be further investigated.

a Identification of deubiquitinases of BRD4. Purified unbiquitinated BRD4 was incubated with indicated deubiquitinases according to the manufacturer’s instructions (DUB Scan kit, Cat. No. 67-0006-001). After that, western blot analysis was performed to probe ubiquitination of BRD4 (Fig. S4a). Quantification of relative intensity of ubiquitin of BRD4 normalized to BRD4 incubated with buffer. b Heat map showing frequency of high expression of candidate DUBs in 50 paired CRC and adjacent normal mucosa tissues. c Representative western blot analysis (n = 2) of indicated proteins in PDC1 and HCT116 cells transfected with indicated siRNAs with 50 ng/ml rIL6. dRepresentative western blot analysis (n = 2) of indicated proteins in PDC1 cells transfected with indicated siRNAs with/without 20 μg/ml MG132 under rIL6 (50 ng/ml) treatment. e In vitro ubiquitination assays (n = 3) of WCL and immunoprecipitates from 293T cells transfected with indicated siRNA or constructs in the presence of 20 μM MG132 and 50 ng/ml rIL6. f Representative western blot analysis (n = 3) of WCL from 293T cell transfected with Flag-BRD4 or Flag-BRD4-Y97/98A together with HA-UCHL3 constructs. g Representative western blot analysis (n = 2) of indicated proteins in 293T cells transfected with indicated siRNAs or constructs. h Representative western blot analysis (n = 2) of WCL and immunoprecipitates by anti-HA antibody from 293T cells transfected with indicated constructs with 20 μM MG132 or 2.5 μM pacritinib as indicated. iRepresentative western blot analysis (n = 2) of WCL and immunoprecipitates by anti-HA antibody from 293T cells transfected with indicated constructs with 20 μM MG132. j Representative western blot analysis (n = 2) of WCL and immunoprecipitates by anti-BRD4 antibody from PDC1 cells treated with 50 ng/ml rIL6 and/or 2.5 μM pacritinib as indicated.

Full size image

To investigate if UCHL3 binds to BRD4, we performed reciprocal co-immunoprecipitation assays in cells transfected with BRD4, alone or with UCHL3 and JAK2-V617F. The result showed that UCHL3 indeed interacted with BRD4, which was greatly enhanced in the presence of JAK2-V617F, but inhibited by pacritinib (Figs. 3h and S4k). This indicates that UCHL3 preferentially binds to a phosphorylated BRD4. Indeed, the non-phosphorylation BRD4-Y97/98A mutant showed a reduced interaction with UCHL3, while the phosphorylation mimic BRD4-Y97/98E presented an increased interaction with UCHL3 (Figs. 3i and S4l). Further investigation also showed the endogenous BRD4 binding with UCHL3 which was enhanced by rIL6 but inhibited by pacritinib (Figs. 3j and S4m). Thus, we confirm that UCHL3 is a deubiquitinase for JAK2-phosphorylated BRD4, whose overexpression in CRC may provide a mechanism to maintain the BRD4 protein abundance in an inflammatory tumor microenvironment.

CAF-secreted IL6/IL8 induces JAK2-BRD4 signaling to confer tumor aggressiveness

Cancer-associated fibroblasts (CAFs) are one of the major sources in the tumor microenvironment to secrete IL6 and IL825,26,27. Indeed, through analyzing the colorectal cancer dataset GSE39396, we found that CAFs express higher levels of IL6 and IL8 than that in other cell types in the TME such as endothelial cells and immune cells and cancer cells25(Fig. S5). We thus sought to use CAFs as a cellular model to investigate if CAFs can induce BRD4 expression in CRC via IL6/IL8 secretion. To do this, we isolated CAFs from resected tumors as well as the adjacent normal fibroblasts (NFs) from various CRC patients for CAF/CRC coculture experiments (Fig. 4a). Enzyme-linked immune sorbent assay (ELISA) analysis confirmed that the patient-derived CAFs secreted much higher levels of IL6 and IL8 compared to normal fibroblasts (NF) as well as PDC cells (Fig. 4b). Similar to rIL6/8, CAFs, but not NFs, induced BRD4 expression along with JAK2 activation in PDC1 and PDC2, but not BRD2 and BRD3 (Fig. 4c). Likewise, the treatment of the PDC cells directly with the conditioned medium of CAF (CM) also induced the BRD4 expression (Fig. 4c). CAF also yielded BRD4 Y97/98 phosphorylation, which was abolished by pacritinib (Fig. 4d), JAK2knockdown (Fig. 4e), or IL6 and IL8 neutralizing antibody (Fig. 4f). Of note, a simultaneous blocking of both IL6/IL8 appears to be necessary to warrant a more efficient ablation of BRD4 induction (Fig. 4f).

We further show that CAF can promote PDC growth, invasion, and migration, which can be abolished by knockdown of JAK2 or BRD4 (Figs. 4g, h, and   S6a, b). Treating the PDC cells with both the IL6 receptor neutralizing antibody tocilizumab and IL8 receptor inhibitor reparixin also ablated the above phenotype, though the single-agent only had modest effects (Fig. S6c–g). Furthermore, the in vivo assay showed that co-engraftment CAF with PDC1 or HT-29 strongly promoted xenograft tumor growth compared with CRC cells alone (Fig. S6h), which was in line with immunohistochemistry (IHC) detection of higher levels of pJAK2 and pBRD4 in CAF/CRC-derived xenograft tumors (Fig. S6i).

Finally, we validated the pJAK2 correlation with BRD4 or pBRD4 in CRC clinical samples. We performed IHC analysis of α-SMA (a marker for CAF), pJAK2, and pBRD4 in a tissue microarray (TMA) consisting of 248 CRC tumors with clinical information28 (Fig. 4i). As expected, a higher CAF infiltration (identified by α-SMA expression) in CRC was significantly correlated with pBRD4 (Pearson’s r = 0.418, P < 1.0 × 10−11) (Fig. 4j). A strong positive correlation between pJAK2 and pBRD4 was also confirmed (with Pearson’s r = 0.607, P < 1.0 × 10−15) (Fig. 4j). Moreover, an examination of a set of paired primary CRC and metastatic CRC samples showed higher levels of pBRD4 and pJAK2, along with a higher level of CAF infiltration in metastatic tumors compared to primary tumors (Fig. 4k). Disease-free survival (DFS) analysis of these 248 CRC patients reveals that the higher expressions of pJAK2 and pBRD4 were associated with poor survival outcomes in CRC patients (Fig. 4l). These data support that CAF-mediated JAK2-BRD4 activation in tumor microenvironment plays an important role in CRC progression.

CAFs confer resistance to BET inhibitors through the IL6/8-JAK2-BRD4 cascade

BET inhibitors (BETi) targeting BET family of BD-containing proteins, including BRD4, have been considered as promising anti-cancer agents, but solid tumors, including CRC, are commonly resistant to BETi29,30. Previous studies have shown an elevated BRD4 level confers resistance to BETi8,9,10,11. Given that CAF induces BRD4, we next sought to determine whether CAF causes BETi resistance. To this end, we used three different small molecule BET inhibitors JQ1, MZ1, and dBET1, which are potent inhibitors of BRD413,14,17. Although all the three BETi can effectively inhibit the CRC proliferation, coculture with CAFs largely diminished their effects (Fig. 5a). In vivo xenograft tumor engrafted with CRC alone showed an inhibitory response to JQ1, but this effect was diminished when CRC was co-engrafted with CAFs (Fig. 5b). These findings support the role of CAFs in promoting CRC resistance to BETi. Consistent with this hypothesis, resistance to BETis in CRC/CAF cocultures were sufficiently reversed by co-treatment with pacritinib (Fig. 5c). Similarly, blockade of IL6/8 receptors with Tocilizumab and Reparixin also abolished the resistance to BETi (Fig. 5d).

Moreover, in several xenograft tumor mouse models co-engrafted with CRC and CAF, we see JQ1 in combination with pacritinib or Tocilizumab/Reparixin resulted in remarkable suppression of the tumor growth, though the JQ1, pacritinib or Tocilizumab/Reparixin alone only produced a modest effect (Fig. 5e). The treatment showed modest effect on CAFs and the production of IL6 and IL8 (Fig. S7a, b). Consistent with the tumor suppression, the combination treatment resulted in much-reduced pJAK2 and pBRD4 in the xenograft tumors as compared with the single-agent treatment (Fig. 5f). Together, these findings demonstrate that blockade of IL6/8-JAK2 cascade could be a potential strategy to overcome BETi resistance caused by the tumor microenvironment.

BRD4 phosphorylation at Y97/98 results in reduced binding to BETi but increased binding to chromatin

Bromodomains (BD1 and BD2) of BRD4 is required for the recognition and binding of the acetylated lysine residues of histones and other cellular proteins to execute its function as a chromatin regulator. It is interesting to notice that Y97 has been previously reported to reside in the binding pocket of JQ1, which is important for JQ1 to inhibit BRD4 activity16,17. Our finding thus raised an intriguing possibility that JAK2-induced BRD4 phosphorylation at Y97/98 may affect JQ1 binding to BRD4, thus resulting in resistance to JQ1. To test this hypothesis, we performed a NanoBRET Reporter assay in which JQ1 binding affinity to Nanoluc-BRD4 fusion protein can be measured by a BRET luminescent tracer in a cellular system31. We found that 293T cells transfected with Nanoluc-BRD4 alone showed a high binding activity to JQ1, which was markedly reduced in cells that were co-transfected with JAK2-V617F (Fig. 6a). Moreover, in contrast to the Nanoluc-BRD4, Nanoluc-BRD4-Y97/Y98E mutant transfected in the cells showed a complete loss of its binding to JQ1 (Fig. 6b). Together, these data demonstrated the Y97/98 is required for JQ1 binding, whose phosphorylation by JAK2 ablated its binding to JQ1. To further consolidate this finding, we also used a BRD4 degrader, dBET1, which is a proteolysis-targeting Chimera (PROTAC) version of JQ1 that binds to BRD4 and induces BRD4 degradation13. As expected, though dBET1 with increasing concentrations can cause BRD4 degradation, it failed to do that in cells co-transfected with the JAK2-V617F (Fig. 6c). Likewise, BRD4 phosphorylation mimic BRD4-Y97E/Y98E also showed resistance to dBET1-induced degradation (Fig. 6d). Together, these findings using different approaches have demonstrated that JAK2-induced BRD4 phosphorylation at Y97/98 abolished its binding to BETi.

a NanoBRET report assay determining the binding ability of JQ1 to BRD4 in 293T cells transfected with BRD4 and active JAK2-V617F or empty vector. b NanoBRET report assay determining the binding ability of JQ1 to BRD4 in 293T cells transfected with BRD4-wild-type or BRD4-Y97/98E. cRepresentative western blot analysis (n = 3) of indicated proteins in 293T cells transfected with indicated constructs and treated with dBET1(1.25 μM, 2.5 μM). d Representative western blot analysis (n = 3) of indicated proteins in 293T cells transfected with flag-BRD4 or flag-BRD4-Y97/98E followed by dBET1 treatment. e NanoBRET™ BRD4/Histone H3.3 Interaction Assay in 293T cells determining the interaction between BRD4 and histone H3.3 under different conditions. 293T cells were transfected with BRD4-NanoLuc and histone H3.3-HaloTag with/without active JAK2-V617F as indicated, then were treated with indicated inhibitors. f Representative western blot analysis (n = 2) of indicated proteins from cytosol and nucleus of PDC1 cells treated with 20 ng/ml rIL6 for indicated time. g Representative western blot analysis (n = 3) of indicated proteins from WCL, chromatin, and supernatant of PDC1 cells treated with 50 ng/ml rIL6/8 and/or 2.5 μM pacritinib. hRepresentative western blot analysis (n = 2) of indicated protein from chromatin of PDC1 cells treated with 20 ng/ml rIL6 followed by JQ1 treatment. In e, data is presented as the mean ± SD (error bars) of three biological replicates. The p values were determined using two-tailed Student’s t tests.

Full size image

BRD4 regulates chromatin activity through binding to acetylated histone32. Next, we performed the NanoBRET assay to monitor the proximity of BRD4 and Histone 3.3 (H3.3) in the context of JAK2 activation. As expected, the interaction between BRD4 and H3.3 was reduced by JQ1 through competitive binding but increased by JAK2-V617F (Fig. 6e). In the presence of JAK2-V617F, the BRD4-Histone 3 interaction was no longer sensitive to JQ1, but this was reversed by pacritinib (Fig. 6e). Interestingly, cell fractionation analysis showed that rIL6 induced nuclear accumulation of pJAK2 (Fig. 6f); It also caused the enrichment of pJAK2, along with BRD4 and pBRD4 in the chromatin fraction, which can be diminished by pacritinib (Fig. 6g). Moreover, although JQ1 was able to displace the BRD4 protein from the chromatin as expected, it could not replace the phosphorylated BRD4 from the chromatin (Fig. 6h). These data demonstrated that JAK2-induced phosphorylation of BRD4 results in a favored binding to chromatin while reducing binding to JQ1. These observations have provided a plausible explanation for the BETi resistance in tumors enriched with an inflammatory tumor microenvironment.

pBRD4 interacts with STAT3 to confer increased oncogenic enhancer activity

To identify the key transcriptional program induced by CAF-activated JAK2-BRD4 signaling, we performed a transcriptome analysis in PDC cocultured with or without CAF and treated with pacritinib, JQ1 or both. We identified 371 genes whose expression were upregulated by CAF but reduced by either JQ1, pacritinb or both (using 2-fold cutoff, p ≤ 0.05) (Fig. 7a). Consistent with the combination effect of JQ1 and pacritinib in CRC growth, we found that the majority of this gene set showed synergistic downregulation upon the combination treatment, indicating that this gene set is regulated cooperatively by JAK2 and BRD4. Not to our surprise, ingenuity analysis of upstream regulators reveals that STAT3 is the top upstream regulator of this gene set (Fig. 7b). Given that STAT3 has been previously shown to associate with other epigenetic regulators to modulate chromatin activity33, we asked whether STAT3 acts as a transcriptional effector of JAK2 to coordinate with BRD4 to regulate chromatin activity. For this reason, we performed reciprocal co-immunoprecipitation and uncovered a steady interaction between pSTAT3 and pBRD4 in PDC cells treated with rIL6/8, which can be disrupted by pacritinib (Fig. 7c). These data suggest that STAT3 activation by IL6/8 leads to association with pBRD4 to regulate the chromatin activity coordinately.

BRD4 is known to bind to gene enhancers and super-enhancers to regulate gene expression34,35. To understand the association between BRD4 and pSTAT3 at the chromatin level, we performed chromatin immunoprecipitation coupled with deep sequencing (ChIP-seq) for the total BRD4, pBRD4, pSTAT3, and the active enhancer marker H3K27ac in PDC cells treated with or without rIL6/8. Although rIL6/8 treatment did not induce ChIP-seq signal of total H3K27ac mark (Fig. 7d), it significantly increased the pBRD4 and pSTAT3 binding to chromatin. In contrast, there was only a modest change of the total BRD4 binding to chromatin (Fig. 7e–g). This observation indicates that in response to IL6 and IL8 signals, pBRD4 and pSTAT3 regulate chromatin activity distinctly from BRD4.

We further characterized the rIL6/8 response on typical enhancers and super-enhancers36(Fig. S8a). Similar to the overall H3K27ac signals (Fig. 7d), rIL6/8 treatment did not seem to induce apparent changes in the total super-enhancers and typical enhancers (Fig. S8a, b). However, rIL6/8 treatment induced much-increased super-enhancers or typical enhancer signals bound by pBRD4, pSTAT3, but to a lesser degree BRD4 (Fig. 7h–j). Further analysis identified 176 sites that are commonly regulated by BRD4, pBRD4, pSTAT3, and H3K27ac in response to rIL6/8 treatment, suggesting a set of enhancers concurrently regulated by BRD4, pBRD4, and pSTAT3 (Fig. 7k). Signal distribution analysis of the associate genes showed that although BRD4 seemed to bind mainly to the promoter region of these genes, pBRD4 and pSTAT3 showed widespread binding to enhancers across the gene bodies (Figs. 7l and S8c). Interestingly, several well-known oncogenes linked to aggressive and metastatic cancer phenotype, including CXCL1, CXCL2, MYC, and BMP4, were included in this gene get, suggesting that BRD4 and STAT3 can coordinately regulate the oncogenic enhancer activity to promote tumor progression. Of note, these oncogenes were occupied by extended enhancer clusters showing H3K27ac, pBRD4, and pSTAT3 peaks across the gene bodies, while BRD4 is more found in the promoter regions (Fig. 7m).

Thus, these findings suggest that paracrine inflammatory signals promote pBRD4 and pSTAT3 interaction to augment enhancer activity towards more robust oncogenic transcription. Consistent with this notion, RT-PCR validation of several of these oncogenes showed that CAFs induced their expression, which required a combination of JQ1 and pacritinib to abolish their expression (Fig. 7n) effectively. ChIP-qPCR analysis showed that the binding of pBRD4 on the promoters of CXCL1, CXCL2, MYC was more resistant to JQ1 treatment comparing to that of BRD4, and the combination of JQ1 and pacritinib largely diminished the binding (Fig. S8d). Taken together, these data indicate tumor microenvironment provides a mechanism to promote the pBRD4 and pSTAT3 association to increase enhancer activity towards oncogenic transcription and a co-targeting strategy is necessary to abolish or revert this process.

We demonstrate a previously unrecognized tumor microenvironment mechanism by which paracrine IL6/IL8-JAK2 signaling induces BRD4 activation in CRC, leading to chromatin remodeling and resistance to BETi treatment. Previous studies have documented the role of stromal infiltration, as well as CAF-secreted IL6/IL8 and associated JAK2-STAT3 signaling in CRC relapse and poor prognosis20,37,38,39,40,41. Our finding highlights important crosstalk between the inflammatory signaling in the tumor microenvironment and the chromatin regulatory machinery in cancer cells, which provides insights into the understanding of the CRC progression.

Several BETis are under clinical development, but the clinical trial results have been less encouraging in most solid tumors. Several recent studies have reported mechanisms of BETi resistance, but these studies have focused on intrinsic cancer mechanisms. Among them, multiple serine phosphorylation of BRD4 modulated by CK2, CDK9, or PP2A phosphatase at nearby the BD2 domain has been shown to increase BRD4 chromatin binding capacity or promotes JQ1 resistance by facilitating bromodomain-independent chromatin recruitment7,42,43. In addition, BRD4 stabilization due to genetic loss of function of E3 ubiquitin ligase SPOP has been associated with resistance to BETi in prostate cancer9,10,11. Our study, however, provides a different mechanism by which the tumor microenvironment induces BRD4 stabilization through tyrosine phosphorylation at BD1 domain at Y97/98.

Interestingly, Y97 has been previously shown to be within the binding pocket of JQ116. Our finding that Y97/98-phosphorylated BRD4 has increased activity binding to chromatin but reduced binding to JQ1 has provided at least partially a mechanistic explanation for the lack of response to BETi in solid tumors. Considering that BETi resistance is common in solid tumors that are featured with an inflammatory TME, we envision that this mechanism is widely applicable to other cancer as well.

Mechanistically, through comprehensive biochemical screening and tumor samples profiling, we identified UCHL3 as a crucial deubiquitinase that binds explicitly to JAK2-phosphorylated BRD4 to maintain its stabilization. Significantly, UCHL3 is overexpressed in CRC, supporting it being a relevant DUB to maintain a high level of BRD4 protein expression in CRC. Of note, new generation of BETi, BET-PROTACs have been designed to degrade BRD4 avoiding resistance induced by BRD4 stabilization through hijacking the ubiquitin E3 ligase. However, the resistance to BET-PROTACs has also emerged. We thus propose that UCHL3 may serve as an additional target to block BRD4 signaling or re-sensitize CRC response to BETi, including BET-PROTACs, which warrant a further investigation. Interestingly, a recent study shows that a proficient expression of deubiquitinase DUB3 in prostate cancer can promote BRD4 stabilization and resistance to BETi8. Therefore, the deubiquitinases involved in the regulation of BRD4 stability appear to be cancer-dependent.

An important finding in this study is the demonstration that IL6/8-JAK2-induced pBRD4 physically associates with pSTAT3. Consistent with this biochemical finding, we show that pBRD4 and pSTAT3 co-localize on the enhancers across the entire gene bodies of some of the key oncogenes such as MYC, CXCL1, and CXCL2, which exhibit increased enhancer-binding activity in response to IL6 and IL8. By contrast, we found that BRD4 is mainly found on the promoters of these genes and is less responsive to IL6 and IL8 treatment. These findings suggest that phosphorylated BRD4 displays a gain of distinct enhancer activity through association with pSTAT3. This feature of TME-induced chromatin remodeling establishes pBRD4 and JAK2-STAT3 signaling to convergence on a set of crucial oncogenes whose hyperactive transcription state is believed to be necessary for disease progression. Indeed, we show that either increased JAK2 or pBRD4 in CRC is associated with metastasis and poor survival of CRC patients.

The corroborative regulation of some of the important oncogenes known to be important in cancer progression by both pBRD4 and pSTAT3 suggests that a co-targeting strategy is necessary to inhibit the oncogenic transcription. Indeed, we show that combined inhibition of BRD4 and IL6/8-JAK2 signaling is required to inhibit oncogene expression more completely, reversed the BETi resistance in vitro, and suppress the CRC xenograft tumor growth empowered by CAF in vivo. In our pre-clinical models, the combination of JQ1 with pacritinib or IL6/IL8 targeting agents did not have observed adverse effects, with no obvious toxicity in mice. Indeed, pacritinib is currently in late-stage clinical development, and tocilizumab has already been in clinical use. Therefore, the pharmacologic blockage of JAK2-induced BRD4 activity through the aforementioned combination approach has the potential of being translatable into a clinical setting. In this clinical setting, assessing the expression of pJAK2, pBRD4, and UCHL3 in CRC may be useful to guide the patient selection for an appropriate treatment strategy. Taken together, we have demonstrated a tumor microenvironment-mediated mechanism that induces chromatin reprogramming and contributes to resistance to BETi, illustrating the potential of developing a therapeutic strategy to target the resistance mechanism.