The MAPK Pathway Regulates Intrinsic Resistance to BET Inhibitors in Colorectal Cancer

Aberrant activation of WNT signaling through loss-of-function of adenomatous polyposis coli (APC) or mutation of β-catenin (CTNNB1) is the key oncogenic event in colorectal cancer (1). The pluripotent transcription factor MYC is arguably the most important target gene of the WNT/β-catenin pathway in colorectal cancer (2). Data from the Cancer Genome Atlas suggest that MYC-dependent transcription program is activated in nearly all colorectal cancer (1). Depletion or reduction of MYC impairs tumorigenesis in APC loss-driven mouse models of colorectal cancer (3, 4). Therefore, targeting MYC has the potential to disrupt key oncogenic functions in colorectal cancer. Although direct pharmacologic intervention for transcription factors like MYC remains difficult, small-molecule compounds targeting the bromodomain and extra-terminal domain (BET) family epigenetic readers recently emerged as alternative approaches to suppress oncogenic transcription factors, including MYC (5).

The BET family includes ubiquitously expressed BRD2, BRD3, BRD4, and a testis-specific BRDT. These proteins contain two evolutionarily conserved bromodomains that recognize acetylated lysine residues on histone tails (5). Through this interaction, BET proteins direct assembly of nuclear macromolecular complexes, such as the mediator complex and the transcription elongation complex, to acetylated chromatin (6). Hence, BET proteins have important roles in transcription initiation and elongation. Small-molecule BET bromodomain inhibitors, such as JQ1, PFI-1, MS417, and i-BET762, exhibit promising antineoplastic activities in a range of preclinical models of human cancers (7–13). The antineoplastic activities of BET bromodomain inhibitors are associated with suppression of prominent oncogenic transcription programs frequently activated in human cancers, such as MYC (9–11), MYCN (12, 14), GLI1/2 (15), and NF-κB (16). As such, BET bromodomain inhibitors represent an appealing therapeutic option for cancers dependent on oncogenic transcription factors. Oncogenic transcription programs driven by β-catenin and MYC are key players in the molecular pathogenesis of colorectal cancer. Hu and colleagues recently reported that BET inhibition by MS417 attenuated colorectal cancer liver metastasis, although the impact on xenograft tumor growth was limited (17). In addition, McCleland and colleagues showed that knockdown of BRD4 induced MYC downregulation, differentiation, and growth inhibition in colorectal cancer xenograft models (18). This study also suggested that colorectal cancer with the CpG island methylator phenotype (CIMP) was preferentially sensitive to BET inhibitors (18), although this link was not confirmed by a more recent study (19). Nevertheless, response to BET inhibitors in colorectal cancer is modest in general, suggesting that colorectal cancer tumors are intrinsically resistant to BET inhibition. In this study, we assessed the therapeutic potential of BET inhibitors in colorectal cancer and interrogated mechanisms conferring resistance to BET inhibitors. Our results demonstrated that blockade of the RAS/RAF/MAPK pathway rendered colorectal cancer cells significantly more sensitive to BET inhibitors. Concurrent inhibition of BET proteins and the MAPK pathway was necessary and sufficient to effectively downregulate MYC and to induce significant tumor regression in xenograft models. Collectively, our findings suggest that the combination of BET inhibitors and MEK inhibitors is a promising therapeutic strategy for colorectal cancer.

Colorectal cancer cell lines were purchased from the American Type Culture Collection (ATCC). MC38 mouse colon adenocarcinoma cells and DLD1 derivative lines, DKO1 and DKS8, were provided by Dr. Robert Coffey, Vanderbilt University, Nashville, TN. RKO derivative lines were provided by Dr. Daniel Liebeler, Vanderbilt University, which were originally purchased from Horizon Discovery. All cells used were amplified from early passages and maintained in cultures for no more than 2 months. These lines were not further authenticated. Mycoplasma tests were performed regularly using the PCR detection kit from the ATCC (ATCC #30-1012K). All cell lines were maintained in RPMI-1640 medium supplemented with 10% FBS and 100 U/mL penicillin–streptomycin (Life Technologies) at 37°C in 5% CO₂. The immortalized nontumorigenic immortalized young adult mouse colon (YAMC) cells were obtained from Dr. Robert Whitehead, Vanderbilt University. YAMC cells were maintained and analyzed at 33°C to activate a temperature-sensitive mutant of SV40 large T antigen (20). Additional supplements were added in the growth medium for YAMC cells as previously described (20). The integrity of cell lines used in this study is described in our recent publications (21). The patient-derived xenograft line, CR-IGR-0034P, was purchased from Oncodesign. This line is serially passaged in immunocompromised mice without in vitro culture.

The pLKO.1 lentiviral vectors directing expression of shRNA sequences specific to BRD2 (2a: TRCN000006309, 2b: TRCN000006310), BRD3 (3a: TRCN0000021374; 3b: TRCN0000021376), and BRD4 (4a: TRCN0000196576; 4b: TRCN0000199427) have been described in our previous publications (13). KRAS knockdown constructs (sh-KRAS-1: TRCN000003260 and sh-KRAS-2L: TRCN000003262) were purchased from Sigma. Lentivirus was produced by cotransfection of the lentiviral vectors with the packaging vectors psPAX2 and pCI-VSVG (Addgene) into 293FT cells. Viral supernatants were collected 2 days after transfection and applied to cells at an approximate multiplicity of infection of 5. Cells were selected with 1 μg/mL puromycin for at least 48 hours prior to experiments. Most antibodies used in this study are described in our recent publications, except that KRAS-specific antibody was purchased from Santa Cruz Biotechnology (#sc-30; ref. 13). Immunohistochemical staining of xenograft tumor sections was performed with primary antibody against MYC (#sc-40; Santa Cruz Biotechnology) at a 1:500 dilution, Ki67 (#VP-K451; Vector Laboratories) at a 1:2,000 dilution, or cleaved Caspase-3 (# 9664; Cell Signaling Technology) at a 1:300 dilution. Staining was visualized by the Bond Polymer Refine detection system.

Total RNA was isolated using the Illustra RNAspin Kit (GE Healthcare) and reverse transcribed using the iScript cDNA synthesis Kit (Bio-Rad). Gene expression was measured by real-time PCR (RT-PCR) using universal SYBR-Green Mastermix (Bio-Rad). The reaction comprises 40 cycles of 95°C for 20 seconds and 60°C for 45 seconds. Primers are listed in Supplementary Table S1. Beta-actin was used as the loading control.

Colorectal cancer cells were treated with DMSO or 1 μmol/L JQ1 for 24 hours. Total RNA was extracted using the Illustra RNAspin Kit with on-column DNA digestion according to the manufacturer's instructions. RNAseq was performed by Vanderbilt Technologies for Advanced Genomics. All RNA samples were quantified using the QuBit RNA assay Kit. RNA quality was checked using Agilent Bioanalyzer. The Illumina TruSeq RNA Sample Preparation Kit (Illumina) was used for library preparation. Samples were sequenced on HiSeq 2500 to obtain at least 30 million paired end (2 × 50 bp) reads per sample. The RNAseq data went through multiple stages of thorough quality control as recommended (22). Raw data and alignment quality control were performed using QC3, and gene quantification quality control was conducted using MultiRankSeq. Raw data were aligned with TopHat2 against human transcript genome HG19. Gene expression was quantified using Cufflinks (23). Differentially expression analysis is performed using Cuffdiff command from Cufflinks package. FDR < 0.05 was used as the significant threshold.

We employed the gene set enrichment analysis (GSEA) method (http://www.broadinstitute.org/gsea) for statistical analysis of changes in gene expression patterns (24). GSEA determines whether a defined set of genes shows statistically significant differences between two phenotypes. The complete list of genes and their scores were used in GSEA with a focus on the C6 MSigDB collection (oncogenic signatures). According to the developer's instructions, the FDR q value represents “the estimated probability that a gene set with a given normalized enrichment score represents a false positive finding,” and the nominal P value estimates “the statistical significance of the enrichment score for a single gene set.”

Cells were treated for 24 hours, fixed in 75% ethanol, treated with RNase A, and stained with 10 μg/mL propidium iodide. Cell-cycle distribution was measured by flow cytometry on a BD LSRFortessa cell Analyzer. The percentage of cells in each cell-cycle phase was assessed by the ModFit LT software.

To measure caspase activity, 2,500 cells were plated in 96-well plates and treated for 72 hours prior to analysis. Activities of caspase-3/7 were measured using a Caspase-Glo 3/7 assay kit (Promega) following the manufacturer's instructions. Relative caspase 3/7 activities were calculated by normalizing the values of caspase-3/7 activities to the values of cell titers measured by a CellTiter-Glo assay kit (Promega) in replicated wells.

Colorectal cancer cells were plated at 200 cells per well in 6-well plates in triplicates and treated the next day with 500 nmol/L JQ1 ± 25 nmol/L trametinib. However, RKO was treated with 100 nmol/L JQ1 and 25 nmol/L trametinib, because this line appeared to be highly sensitive to BET inhibitors alone in vitro. Drugs were withdrawn after 5-day incubation. Cells were allowed to grow for 7 additional days prior to be fixed and stained with 0.5% crystal violet. Colonies were counted using the Image J software.

Colorectal cancer cells were plated 2,500/well in 96-well plates and treated with BET inhibitors with or without the MAPK pathway inhibitors following a 2-fold serial dilution. Five days later, cell viability was measured using a Sensolyte Cell Viability and Proliferation Assay kit (Anaspec). The dose–response curves were fitted, and IC₅₀ values were calculated using GraphPad Prism following a nonlinear regression (least squares fit) method.

All animal experiments were performed under protocols approved by the Vanderbilt University Institutional Animal Care and Use Committee. The xenograft assays used female athymic nude mice 6 to 8 weeks old. Colorectal cancer cells were trypsinized, suspended in PBS, and mixed with equal volume of growth factor reduced Matrigel (BD Biosciences). Both flank sites of mice received injection of 2 million tumor cells. Tumor size was measured by a digital caliper and calculated following a formula of size = length × width × width/2. Treatment began when the median tumor size reached approximately 100 mm³. Prior to treatment, tumor-bearing mice were randomized into 4 arms that the median tumor sizes of each arm were roughly the same. JQ1 was first dissolved in DMSO and diluted to 9 volumes of 10% β-cyclodextrin in sterile saline and administrated through intraperitoneal injection. Trametinib was prepared in 1% β-cyclodextrin and administrated through oral gavage. Gefitinib was prepared in 1% Tween-80 and administrated through oral gavage. Tumor-bearing mice were treated daily with vehicle, 50 mg/kg/day JQ1, trametinib (0.5 mg/kg for SW480 and Difi, or 1 mg/kg for RKO), or both. Mice bearing the patient-derived xenograft tumors were treated with 100 mg/kg/day JQ1 and 1 mg/kg trametinib, of which JQ1 was given twice a day and trametinib was given once a day. In addition, a group of mice bearing Difi xenograft tumor were treated with 50 mg/kg gefitinib daily. Tumors were measured 2 to 3 times a week. Animals were monitored for significant adverse effects. Mice accidentally died of injury caused by oral gavage or intraperitoneal injection was excluded from analyses.

GraphPad Prism 5.0 was used to determine statistical significance. P values of less than 0.05 were considered significant. The combination index (CI) values were calculated using the Chou–Talalay method with the Compusyn software following the developer's instructions.

It has been increasingly documented that inhibition of BET proteins results in potent downregulation of MYC in hematologic cancers, such as multiple myeloma, Burkitt lymphoma, and acute myeloid leukemia (9–11). However, regulation of MYC by BET proteins in solid cancers appears to be heterogeneous and context-dependent (13, 14, 25). MYC is commonly activated in colorectal cancer and plays a critical role in disease initiation and progression. Therefore, we asked whether MYC expression in colorectal cancer was sensitive to BET inhibition. Our results showed that BET inhibition by JQ1 commonly downregulated MYC at both protein and mRNA levels in colorectal cancer cell lines (Fig. 1A and B). However, the reductions of MYC at protein levels were moderate and less significant compared with changes at mRNA levels, suggesting implications of posttranscriptional regulations. Following BET inhibition, expression of MYC targets, such as Bcl-2, Bcl-xL, and hTERT, was decreased, whereas the MYC-repressed tumor suppressor, p21 (a.k.a. CDKN1A), was increased, albeit to various extents in different lines (Fig. 1C and Supplementary Fig. S1A–S1C). Selective depletion of individual BET family members by RNA interference suggested that BRD4 was the primary MYC regulator within the family, but BRD2 and BRD3 might also contribute to MYC regulation (Supplementary Fig. S1D–S1G). We further assessed the impact of JQ1 on gene expression profiles in colorectal cancer cells using RNA seq and employed GSEA to analyze the changes in MYC-dependent gene signatures. In DLD1, KM12C, and RKO cells, JQ1 significantly suppressed MYC-dependent transcription modules (Fig. 1D; Supplementary Fig. S1H). However, JQ1-induced repression of MYC signature was not statistically significant in SW480 cells (FDR q = 0.127; normalized enrichment score = –1.27). These results collectively suggest that BET proteins are commonly implicated in regulation of MYC expression and activity in colorectal cancer. However, in heterogeneous solid caners like colorectal cancer, MYC regulation by BET proteins can be context-dependent.

We next assessed activities of BET inhibitors in a subset of colorectal cancer cell lines harboring genetic alterations commonly found in colorectal cancer (26, 27). JQ1 dose-dependently decreased cell growth in all of the tested colorectal cancer cells (Fig. 2A; Supplementary Table S2; refs. 28, 29). Similar response was observed using distinct BET bromodomain inhibitors, such as PFI-1 (Supplementary Fig. S2A). The antigrowth effects of BET inhibitors were partially phenocopied by knockdown of BRD4, whereas knockdown of BRD2 and BRD3 had less significant impact (Fig. 2B; Supplementary Fig. S2B and S2C). Inhibition of BET proteins was associated with cell-cycle arrest in G₁–G₂ phases and moderate caspase activation (Fig. 2C and D; Supplementary Fig. S2D–S2F). Although these findings are consistent with recently published studies and underscore the potential of BET proteins as drug targets in colorectal cancer (17–19), BET inhibitors alone were largely cytostatic in colorectal cancer cultures, raising the concern whether they may be clinically effective as monotherapy.

The modest response to BET inhibitors and the lack of significant cell death suggest that intrinsic resistance to BET inhibitors is common in colorectal cancer, which can be a critical obstacle for successful translation. The RAS/RAF/MAPK pathway is activated in most colorectal cancer tumors via activating mutations of KRAS, NRAS, BRAF, amplification of EGFR, in addition to other mechanisms (1). Among the colorectal cancer cell lines used in our study, phosphorylation levels of ERK were the highest in BRAF^V600E-expressing RKO and HT29, whereas KRAS mutations did not always result in hyperactivation of the MAPK pathway (Supplementary Fig. S3). Activation of RAS/RAF and the downstream MAPK pathway through mutations or other mechanisms is a common key oncogenic event in colorectal cancer. Introduction of an oncogenic KRAS mutant drastically promotes tumor formation and progression in APC-deficient mouse model of intestinal tumors (30). RAS activation augments WNT signaling (31) and also promotes MYC protein stability, through mechanisms such as ERK-mediated MYC phosphorylation (32). Hence, we asked whether blockade of the MAPK pathway might affect colorectal cancer response to BET inhibitors. We treated colorectal cancer cells using a FDA-approved MEK inhibitor, trametinib, with or without JQ1 at a fixed molar ratio of 1:100 following a 2-fold serial dilution. This combination treatment induced significantly more potent reduction of cell growth than either agent alone in all tested lines across a wide range of concentrations (Fig. 3A–C; Supplementary Fig. S4A and Supplementary Table S2). Among the 14 colorectal cancer cell lines that we have tested, the median IC₅₀ value of JQ1 was approximately 330 nmol/L. The median IC₅₀ value of JQ1 was reduced to 130 nmol/L when combined with trametinib (Fig. 3C; Supplementary Table S2). The median IC₅₀ value of trametinib was approximately 11 nmol/L as monotherapy, whereas it was reduced to merely 1 nmol/L in the presence of JQ1 (Fig. 3C; Supplementary Table S2). The interaction between JQ1 and trametinib was further assessed by the Chou–Talalay statistical method (33). In KM12C, RKO, and SW480, the CI values of JQ1 and trametinib were substantially less than 1 at concentrations across multiple orders of magnitude, indicating strong drug synergism (Fig. 3A and B; Supplementary Fig. S4A, green cross, right Y axis). Similar synergistic interactions between JQ1 and the BRAF inhibitor vemurafenib were shown in BRAF^V600E HT29 cells (Supplementary Fig. S4B). In addition, combinations of JQ1 with an ERK inhibitor SCH772984, or i-BET762 with an MEK inhibitor selumetinib also produced synergistic response (Supplementary Fig. S4C and S4D). In contrast, nontumorigenic YAMC cells were refractory to JQ1 and trametinib, either alone or in combination (Supplementary Fig. S4E), whereas MC38, a cell line derived from genetic mouse models of colon cancer, was highly sensitive to the combination (Supplementary Fig. S4F). Clonogenic assays showed that transient exposure to JQ1 and trametinib almost completely abolished the ability of colorectal cancer cells to form colonies at concentrations that single agent had limited impact (Fig. 3D; Supplementary Fig. S5), suggesting that this combination therapy induced prolonged damage to the tumorigenic potential of colorectal cancer cells. To further validate the specificity of actions of chemical probes, we showed that selective depletion of BRD4 sensitized KM12C and RKO cells to trametinib (Supplementary Fig. S6). Conversely, knockdown of BRD2 or BRD3 did not significantly alter response to MEK inhibition, suggesting that BRD4 plays a major role in regulation of the crosstalk with the MAPK pathway. In addition, knockdown of KRAS decreased phosphorylation levels of ERK in both KRAS-mutant SW480 cells and KRAS–wild-type KM12C cells and rendered cells more sensitive to JQ1 (Supplementary Fig. S7). These results collectively suggest that the synergism between BET inhibitors and MAPK pathway inhibitors is primarily mediated via their respective on-target effects.

Our results showed that the combination of BET inhibitors and MEK inhibitors induced synergistic response in all tested colorectal cancer cell lines, irrespective of KRAS and BRAF mutations. These findings suggest an intrinsic role of MAPK signaling in mediating resistance to BET inhibitors that is not dependent on the mechanisms by which the MAPK pathway is activated. To further test this hypothesis, we investigated two sets of isogenic colorectal cancer lines that were genetically modified to deplete either the mutant allele of KRAS/BRAF or the wild-type one. DLD1 cells have one copy of wild-type KRAS and one copy of KRAS^G13D. This line had low MAPK activity compared with other KRAS-mutant colorectal cancer lines (Supplementary Fig. S3). The derivative line DKS8 lost the mutant KRAS^G13D allele. Interestingly, phosphorylation of ERK in DKS8 was moderately higher than the parental line, despite loss of the mutant KRAS (Supplementary Fig. S8A). The response to JQ1, trametinib, or the combination was similar between DKS8 and the parental DLD1 (Supplementary Fig. S8B and S8C). In contrast, the other derivative line depleted of the wild-type KRAS allele, DKO1, showed much more potent activation of ERK (Supplementary Fig. S8A). DKO1 appeared to be more resistant to JQ1 than two other isogenic lines, but the resistance was essentially abolished by trametinib (Supplementary Fig. S8D and Supplementary Table S2). In BRAF-mutant RKO cells, loss of the V600E allele significantly reduced phosphorylation of ERK (Supplementary Fig. S9A), but did not affect the synergistic interaction between JQ1 and trametinib (Supplementary Fig. S9B–S9D). However, RKO-BRAF^wt cells had lower cell viability in the presence of JQ1 at micromolar concentrations (Supplementary Fig. S9E). Taken together, these results suggest that modulations of MAPK activity may affect colorectal cancer cell sensitivity to BET inhibitors. However, at least in colorectal cancer, the synergistic interaction between BET inhibitors and MEK inhibitors is not genotype-dependent.

We next assessed changes in MYC expression when BET proteins, MAPK, or both were targeted. Our results showed that MYC protein was more effectively decreased following exposure to the combination of JQ1 and trametinib compared with either agent alone (Fig. 4A). However, the combination did not result in a more effective reduction of MYC mRNA (Fig. 4B; Supplementary Fig. S10A). Hence, the ability of trametinib to augment JQ1-induced downregulation of MYC protein appeared to be primarily mediated through posttranscriptional mechanisms. Expression of a constitutively active MEK1 mutant, MEK1^C121S, resulted in accumulation of MYC protein and counteracted JQ1-induced MYC repression (Fig. 4C). MEK1^C121S also attenuated the effects of JQ1 and trametinib in Difi and SW480 cells, either alone or in combination (Fig. 4D; Supplementary Fig. S10B).

Although either JQ1 or trametinib alone was largely cytostatic in colorectal cancer cells, we noticed that targeting both BET and the MAPK pathway induced massive cell death and potent caspase activation (Fig. 5A and B). Several antiapoptosis genes, Bcl-2 in particular, have been found to be targets of BET inhibitors in various cancers (9–11). Bcl-2 may play a primary role in neutralizing apoptotic stresses in some hematologic cancers, as the selective Bcl-2 inhibitors navitoclax and venetoclax generate significant response in patients with chronic lymphocytic leukemia (34, 35). However, the apoptosis regulatory network in solid cancers is more complex and heterogeneous. Our results showed that BET inhibition induced downregulation of multiple antiapoptotic Bcl-2 family members, Bcl-2 and Bcl-xL, as well as the inhibitor of apoptosis (IAP) family member, survivin (Fig. 5C and Supplementary Fig. S11A). In SW480 cells, BET inhibition also increased expression of BIM, a proapoptotic member of the Bcl-2 family (Fig. 5D). However, BET inhibition resulted in downregulation of certain proapoptotic genes, such as BIK in SW480 cells and BIK and PUMA in KM12C cells, which are expected to mitigate the impact of loss of antiapoptotic genes (Fig. 5D; Supplementary Fig. S11B). Trametinib exhibited context-dependent effects on expression of these apoptosis regulators. Combining trametinib with JQ1 resulted in more effective suppression of antiapoptotic genes, as exemplified by survivin in SW480 cells (Fig. 5C), and further increased BIM expression (Fig. 5D; Supplementary Fig. S11B). Interestingly, JQ1-induced suppression of BIK and PUMA was neutralized by trametinib (Fig. 5D; Supplementary Fig. S11B). Although changes of individual apoptosis regulators following treatments of JQ1 and/or trametinib were not identical across colorectal cancer lines with different genetic backgrounds, our results showed a pattern that BET inhibitors and MEK inhibitors in combination induced a more effective downregulation of antiapoptosis genes and upregulation of proapoptosis genes compared with monotherapies, leading to strong apoptotic cell death.

We next sought to assess the in vivo efficacy of the combination of JQ1 and trametinib using subcutaneous xenograft models. A few recent studies showed that BET inhibitors, JQ1 or MS417, had limited impact on colorectal cancer xenograft tumor growth (17, 19). Our data demonstrated that either JQ1 or trametinib alone moderately reduced the rate of tumor growth. In contrast, the combination of these two agents decreased median tumor size by 45% in KRAS^G12V SW480 xenografts and 60% in BRAF^V600E RKO xenografts following 2 weeks of treatment (Fig. 6A and B). Tumor regression was associated with decreased proliferation and increased apoptosis as shown by Ki67 and cleaved caspase 3 staining (Supplementary Fig. S12A). Animals treated with the combination experienced approximately 15% loss of body weight in the RKO experiment, whereas changes in the SW480 experiment were minor (Supplementary Fig. S12B and S12C). Using a patient-derived BRAF^V600E-expressing colorectal cancer xenograft model (36), we further showed that JQ1 or trametinib alone decreased tumor growth, whereas the combination induced a 75% reduction in median tumor size (Fig. 6C). Anti-EGFR therapy has been approved for the management of KRAS- and BRAF–wild-type colorectal cancer. We asked whether the efficacy of the combination of BET inhibitors and MEK inhibitors was comparable with anti-EGFR agents in EGFR-dependent colorectal cancers, such as Difi. Daily administration of 50 mg/kg gefitinib over 15 days decreased growth of Difi xenografts from 176% to 20%. In comparison, the combination of JQ1 and trametinib reduced median tumor volume by approximately 40% (Fig. 6D). Immunohistochemical staining of MYC in tumors treated for 5 days showed that JQ1 or trametinib alone only moderately decreased MYC-expressing cells in RKO tumors and marginally affected MYC expression in SW480 cells. However, MYC-expressing cells were drastically decreased in tumors treated with both compounds (Fig. 6E; Supplementary Fig. S12D and S12E). Taken together, our results strongly suggest that a combination of BET bromodomain inhibitors and MEK inhibitors has broad therapeutic potential for treating colorectal cancer, including those therapeutically refractory tumors that harbor BRAF and KRAS mutations.

BET bromodomain inhibitors are a class of emerging drug candidates with attractive anticancer and anti-inflammatory activities through modulation of aberrantly activated transcription factors. Favorable clinical response was recently reported in hematologic cancers (37). However, the efficacy of BET inhibitors in preclinical models of advanced solid cancers is generally moderate. Mechanisms mediating acquired resistance to BET inhibitors just began to emerge. Acute myeloid leukemia may reactivate MYC expression through WNT signaling that confers resistance to BET inhibitors (38, 39). Alternatively, triple-negative breast cancer may gain resistance through increasing phosphorylation of BRD4 that renders BRD4 maintain its transcription regulatory activities independent of bromodomains (40). Mechanisms mediating intrinsic resistance to BET inhibitors remain elusive. Identification of these mechanisms may direct development of more effective combination therapy and expedite translation of BET inhibitors as viable therapeutic options for advanced solid cancers. In the present study, we showed that activation of the MAPK pathway conferred intrinsic resistance to BET inhibitors. Thus, a combination of BET inhibitors and the MAPK pathway inhibitors generated synergistic response in colorectal cancer cell cultures and xenograft models. This approach is particularly appealing for colorectal cancer tumors carrying mutations in the RAS/RAF/MAPK pathway, which is highly resistant to currently available therapeutics.

A hallmark of colorectal cancer is prevalent coactivation of β-catenin/MYC-driven transcription and the RAS/RAF/MAPK pathway that act synergistically to promote colorectal cancer pathogenesis (1, 30, 41). The RAS/RAF/MAPK pathway may augment β-catenin and MYC activity through a multitude of mechanisms (32, 42, 43). Alternatively, activation of the WNT/β-catenin/MYC pathway is implicated in resistance to BRAF and MEK inhibitors (44–47). As such, accumulating evidence support the rationale to cotarget both pathways for the treatment of colorectal cancer and a range of other cancers. However, the master regulators of these two pathways, KRAS and β-catenin, are both difficult drug targets. Our findings suggest that BET bromodomain proteins represent a druggable signaling hub in the oncogenic network driven by the WNT pathway and the RAS/RAF/MAPK pathway. As a result, the combination comprising a BET inhibitor and an inhibitor blocking the MAPK pathway holds great promise to treat cancers affected by aberrations of these two pathways, such as colorectal cancer.

In our study, we observed synergistic interaction between BET inhibitors and MAPK inhibitors in all tested colorectal cancer cell lines, irrespective of mutational status of KRAS and BRAF. In addition, our data did not suggest a linear correlation between sensitivity to JQ1 and activation of the MAPK pathway. Although the broad response in colorectal cancer may raise concerns over off-target effects, it is in concordance with prevalent activation of the MAPK pathway in this disease, which can be induced through not only mutations of KRAS/BRAF but also many other mechanisms. Using two sets of isogenic colorectal cancer cell lines, we showed that depletion of the mutant KRAS or BRAF alleles might alter JQ1 sensitivity according to the changes in MAPK activation. However, the synergistic interaction between JQ1 and trametinib was not compromised. Thus, the crosstalk between MAPK signaling and BET proteins appears to be an intrinsic phenotype of colorectal cancer that is independent of the mechanisms through which the MAPK pathway is activated or the degree of MAPK activation. On the contrary, nontumorigenic YAMC cells were spared by this combination therapy. Several genetic approaches, such as selective depletion of individual BET genes or KRAS and expression of a constitutively active MEK1 mutant, also underscored the on-target actions of BET inhibitors and MEK inhibitors. It is an interesting question if this paradigm can be applied to additional cancers transformed by MYC and oncogenic KRAS.

The mechanisms mediating the synergistic effects of the combination of BET inhibitors and MEK inhibitors remain to be fully elucidated in different colorectal cancer subtypes. Although many downstream signaling molecules are affected by the combination therapy, MYC is a particularly important target. Overexpression of MYC is extensively documented in colorectal cancer and associated with poor survival (48). The Cancer Genome Atlas data indicate that MYC-dependent transcription is activated in nearly all colorectal cancer samples (1). We found that trametinib did not augment JQ1-induced downregulation of MYC mRNA, but further decreased MYC protein levels in the presence of JQ1, suggesting that trametinib modulated MYC levels through posttranslational mechanisms. The detailed mechanisms through which the MAPK pathway regulates MYC protein levels remain an open question and can be context-dependent. Nonetheless, our findings suggest that targeting both BET proteins and the MAPK pathway is necessary to effectively downregulate MYC proteins in colorectal cancer, in contrast to observations in MYC-amplified hematopoietic cancers. The resistance of MYC to BET inhibition is even more obvious in vivo, as most cells in colorectal cancer xenografts remained positively stained for MYC following administration of JQ1. In contrast, the combination of JQ1 and trametinib drastically reduced the percentage of MYC-positive cells. Hence, this combination strategy has important indications to pharmacologically target oncogenic MYC in solid cancers.

Another phenotype consistently shown in colorectal cancer cells treated with both BET inhibitors and MEK inhibitors was the shift of the balance between antiapoptotic genes and proapoptotic genes. BET inhibition unexpectedly reduced expression of certain proapoptotic genes, such as BIK and PUMA. As such, despite reduced expression of antiapoptotic genes, the overall impact of BET inhibitors on apoptosis regulatory network might not be sufficient to induce significant cell death. Addition of trametinib neutralized JQ1-induced downregulation of BIK and PUMA, augmented induction of BIM, and further decreased survivin. Hence, blockade of both pathways appears to be necessary and sufficient to effectively shift the balance between proapoptotic genes and antiapoptotic genes to activate apoptosis. A recent study shows that concurrent upregulation of NOXA/BIK and inactivation of Bcl-2/Bcl-xL is necessary to induce apoptosis in KRAS-mutant colon cancer cells, underscoring the importance of coordinated changes in the apoptotic regulatory network (49). In addition, Conery and colleagues recently reported that activity of BET inhibitors in preclinical cancer models largely correlated with apoptotic response (50). Collectively, the coordinated actions of BET inhibitors and MEK inhibitors on expression of apoptosis regulators may have an important role in mediating the synergistic response.

To date, BET inhibitors alone exhibit generally moderate preclinical activities in solid tumor models. To realize the therapeutic potential of BET inhibitors, it is crucial to identify the mechanisms of resistance and to develop rational combinations that can produce meaningful response in a preclinical setting. In the current study, we identified a role of the MAPK pathway in mediating intrinsic resistance to BET inhibitors in colorectal cancer. As a result, combinations of BET inhibitors with inhibitors targeting the MAPK pathway induced synergistic response in vitro and in vivo. These findings propose a combination strategy that may substantially promote the uses of BET inhibitors and FDA-approved BRAF and MEK inhibitors. Crosstalks between signal transduction pathways and epigenetic regulation have recently emerged as a key factor to determine therapeutic response to BET inhibitors (40, 51–53). As such, combining with additional targeted agents is likely essential to optimize BET-targeted therapy.

No potential conflicts of interest were disclosed.

Conception and design: Y. Ma, L. Wang, R.D. Beauchamp, J. Wang

Development of methodology: Y. Ma, L. Wang, J. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Ma, L. Wang, L.R. Neitzel, S.N. Loganathan, E.E Crispi, R.D. Beauchamp, E. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Ma, L. Wang, L.R. Neitzel, Y. Guo, J. Wang

Writing, review, and/or revision of the manuscript: Y. Ma, L. Wang, S.N. Loganathan, L. Qin, S. Knapp, R.D. Beauchamp, J. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Ma, L. Wang, N. Tang, S. Knapp

Study supervision: Y. Ma, L. Wang, J. Wang

We want to thank Drs. Robert Coffey, Daniel Liebler, and Robert Whitehead for providing important cell lines.

J. Wang is supported by the NIH grant (R01CA166492) and the Voice Against Brain Cancer Foundation. E. Lee is supported by NIH grants R01GM103926 and R01GM081635, and R.D. Beauchamp is supported by R01CA069457. S. Knapp is grateful for support by the Structural Genomics Consortium, a registered charity (No. 1097737) that receives funds from AbbVie, Bayer, Boehringer Ingelheim, Genome Canada through Ontario Genomics Institute Grant OGI-055, GlaxoSmithKline, Janssen, Lilly Canada, the Novartis Research Foundation, the Ontario Ministry of Economic Development and Innovation, Pfizer, Takeda, and Welcome Trust Grant 092809/Z/10/Z. L.R. Neitzel is a recipient of NIH training grant 5T32HD007502. S.N. Loganathan is supported by NIH training grant T32GM07628.

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.