Bromodomain inhibition overcomes treatment resistance in distinct molecular subtypes of melanoma

We aimed to explore the preclinical utility of bromosporine (bromo), a bromodomain inhibitor with demonstrated affinity for BPTF (16). Bromo treatment of several established BRAF-mutant melanoma cell lines in culture produced broad-based and dose-dependent cytotoxicity, with IC₅₀s ranging from 2 to 8 µM (SI Appendix, Fig. S1A). Bromo administration to multiple melanoma cell lines resulted in significant suppression of cell-cycle progression, as evidenced by reduced S phase and G1 arrest (SI Appendix, Fig. S1 B–D) and was accompanied by induction of apoptosis (SI Appendix, Fig. S1 E–G). Bromo treatment of melanoma cells resulted in down-regulation of pERK, BCL2, and BCL-XL (SI Appendix, Fig. S1H). Finally, bromo administration produced significant in vivo anti-tumor effects against LOX human melanoma cells (SI Appendix, Fig. S1I). Taken together, these studies demonstrate that bromo is active against BRAF-mutant melanoma. In addition, bromo treatment recapitulates several key effects of BPTF silencing in human melanoma cells. Specifically, the down-regulation of pERK following bromo treatment suggested its potential to resensitize to or synergize with MAP kinase pathway inhibitors (i.e., BRAFi or MEKi).

Next, we assessed the effects of bromo administration alone or in combination with MAPK pathway inhibitors against a panel of metastatic melanoma (MM) PDX lines harvested from patients who progressed on BRAFi therapy (Dataset S1). For our pharmacological analyses, we selected vemurafenib (vem) and cobimetinib (cobi) to represent standard-of-care BRAFi and MEKi therapy, respectively. Initially, we assessed whether administration of bromo could resensitize to BRAFi therapy using the MM-302 line, collected from a patient following progression on dabrafenib. While there was evidence of modest activity upon treatment with either vem or bromo alone in MM-302 cells, there was markedly improved activity of the vem/bromo combination (SI Appendix, Fig. S2A). In vivo drug treatment of MM-302 tumors demonstrated resistance to vem, whereas single agent bromo exhibited significant anti-tumor activity (Fig. 1A). Importantly, addition of bromo to vem restored the anti-tumor efficacy of BRAF inhibition.

We then aimed to confirm the activity of bromo in additional PDX models obtained from patients progressing on BRAFi-based therapy. In these in vivo analyses, we included assessment of the bromo/cobi combination, as well as comparison with the vem/cobi combination. MM-300 is a PDX line obtained from another patient progressing on dabrafenib (Dataset S1). Studies in culture indicated that neither vem alone nor combined with cobi showed significant anti-proliferative activity, whereas all bromo-containing regimens produced marked cytotoxic effects (SI Appendix, Fig. S2B). The in vivo drug treatment of MM-300 cells revealed resistance to single-agent vem administration, while displaying evidence of single agent activity for bromo (Fig. 1B). The bromo/vem combination was more active than either agent alone, similar to that observed with MM-302 cells (P < 0.00005, ANOVA). MM-300 cells were sensitive to the standard vem/cobi combination. However, while both bromo-containing regimens were superior to the standard regimen, the bromo/cobi combination emerged as the most active treatment, with an 87% and 61% reduction in tumor volume when compared to vehicle and BRAFi/MEKi treatment, respectively (P < 0.00005 for each comparison).

Subsequently, we assessed the activity of bromo-based therapy against PDX lines obtained following progression on combined BRAFi/MEKi therapy using MM-337 and MM-358 (Dataset S1). When assessed in culture, BRAFi alone or with MEKi did not produce a high level of cell killing, whereas all bromo-containing regimens exhibited robust anti-proliferative activity (SI Appendix, Fig. S2 C and D). In vivo studies in the MM-337 (Fig. 1C, P < 0.00005, ANOVA) and MM-358 (Fig. 1D, P < 0.001, ANOVA) lines consistently demonstrated the superior performance of the bromo/cobi combination. In the MM-337 model, while BRAFi/MEKi produced marked antitumor activity, bromo/cobi treatment was significantly more effective, with a 93% and 59% reduction in tumor volume when compared to vehicle (P < 1.3 × 10⁻⁹) or BRAFi/MEKi (P < 1.4 × 10⁵), respectively, including evidence of tumor regression (Fig. 1C). In the MM-358 model, bromo monotherapy was as effective as BRAFi/MEKi therapy, with further improvements in anti-tumor efficacy observed with the bromo/cobi regimen (Fig. 1D).

Having established the activity of the bromo/cobi combination in the setting of resistance to BRAFi-based therapy, we aimed to assess its efficacy in the treatment-naive setting using the MM-313 line, which was obtained from an untreated BRAF-mutant metastatic melanoma patient. Studies in culture established sensitivity of MM-313 cells to bromo alone or in combination with cobi or vem (SI Appendix, Fig. S2E). In vivo analysis showed the superior anti-tumor activity of both bromo-containing regimens when compared with vem/cobi, which, as expected, was highly active in this untreated line (P < 0.0003, ANOVA, Fig. 1E). The bromo/cobi combination proved the most active in this setting, with a 37% reduction in tumor volume when compared with the standard combinatorial regimen (P < 0.006). Furthermore, given that the bromo/cobi combination does not require presence of a BRAF mutation for its anti-tumor activity, we assessed its efficacy in melanoma PDX lines harboring other mutational drivers, obtained following progression on PD-1 blockade. Bromo/cobi administration was highly active against both MM-386 (an NRAS-mutant line, SI Appendix, Fig. S2F and Fig. 1F) and MM-505 (a NF1-mutant line, SI Appendix, Fig. S2G and Fig. 1G) and proved superior to treatment with MEKi alone [P < 0.002 (Fig. 1F) and P = 0.001 (Fig. 1G)]. We also assessed the potential synergy of the bromo/cobi combination in several of these PDX lines using the combination index (CI) (17), where CI values <1, 1, and >1 indicate synergism, additivity, and antagonism, respectively. We found multiple synergistic interactions for inhibition of cell viability across concentration ranges in several melanoma lines (Fig. 1H). In the multiple in vivo studies, the bromo/cobi regimen was well tolerated, and laboratory analysis of complete blood count and serum chemistries conducted at the completion of a study in the MM-358 model did not identify any abnormalities in white blood cell count, neutrophil count, or liver or kidney function (Dataset S2).

In addition, we aimed to elucidate the signaling pathways altered in melanoma cells following bromo/cobi treatment. Studies in culture indicated that combinatorial therapy of MM-337 and MM-358 cells resulted in marked suppression of cell-cycle progression (involving reduced S phase and G1 arrest) (Fig. 2A and SI Appendix, Fig. S3 A and B), accompanied by induction of apoptosis (Fig. 2B and SI Appendix, Fig. S3 C and D). In addition, bromo/cobi treatment resulted in suppression of pERK, BCL2, and BCL-XL (Fig. 2C and SI Appendix, Fig. S3E), similar to that observed with bromo administration alone. In vivo analysis of treated tumors similarly revealed down-regulation of pERK, BCL2, and BCL-XL (Fig. 2D and SI Appendix, Fig. S3F) and was accompanied by markedly decreased Ki-67 immunostaining (Fig. 2 E and F and SI Appendix, Fig. S3 G and H).

In order to identify genome-wide transcriptomic alterations associated with this effective drug combination, we compared the RNA sequencing (RNA-seq) profiles of in vivo MM-300 tumors treated with vehicle or with bromo/cobi. Supervised analysis of the RNA-seq profiles of the two groups revealed 523 significantly down-regulated and 506 up-regulated genes (Fig. 3A). Gene ontology analysis identified several biological processes altered following bromo/cobi treatment (Dataset S3). Strikingly, plotting the Z-score against P value of the down-regulated genes within this analysis indicated suppression of key biological processes, including cell division and cell cycle (including the M phase), microtubule organization and sister chromatid segregation, as well as hypoxia and angiogenesis (Fig. 3B). The differentially expressed genes included several cyclins (e.g., CCNB1), kinesins (e.g., KIF4A), kinetochore components (e.g., CENPM), as well as key regulators of mitosis (e.g., CDK1, AURKA, and PLK1), cytokinesis (e.g., CDC20), angiogenesis (e.g., HIF1A and VEGFA), and translational elongation (e.g., EEF1A2). The down-regulation of each of these genes was confirmed using quantitative reverse transcription–PCR (qRT-PCR) analysis in drug-treated in vivo specimens (Fig. 3C). Prompted by this gene expression profile, we assessed whether combinatorial drug treatment resulted in DNA damage and mitotic arrest. Bromo/cobi treatment of both MM358 and LOX cells resulted in activation of the DNA damage marker γH2Ax (Fig. 3D and SI Appendix, Fig. S4 A–C), as well as significantly elevated comet tail (Fig. 3E and SI Appendix, Fig. S4D). This was accompanied by profound activation of two histone modifications, methylated H3K4 and H3K27, which are recruited to stabilize stalled replication forks (Fig. 3 F and G and SI Appendix, Fig. S4 E–J). These changes would usually be accompanied by activation of RPA and RAD51, which represent key mediators of stalled replication fork protection and restart. Importantly, however, bromo/cobi treatment resulted in marked suppression of both RPA (Fig. 3H and SI Appendix, Fig. S5 A–C) and RAD51 (Fig. 3I and SI Appendix, Fig. S5 D–F). qRT-PCR analysis extended the profound down-regulation of RPA and RAD51 expression following bromo/cobi administration to the transcriptional level (SI Appendix, Fig. S5 G and H). Thus, the transcriptional repression mediated by a drug combination involving a bromodomain inhibitor results in exhaustion of two key DNA damage sensors, whose depletion has been shown to induce replication catastrophe (18, 19).

As replication stress itself has been shown to result in mitotic arrest (20), we further explored the consequences of drug treatment on the mitotic apparatus. As indicated, RNA-seq analysis identified significant down-regulation of several genes key to mitotic entry and progression, including CDK1, CCNB1, PLK1, and AURK1. The differential expression of these genes following combinatorial drug therapy was confirmed in multiple melanoma cell lines using qRT-PCR (Fig. 4A and SI Appendix, Fig. S6 A and B), and Western analysis (Fig. 4B and SI Appendix, Fig. S6C). In addition, bromo/cobi treatment resulted in activation of p21, which also plays an important role in inhibition of mitotic entry, in part by regulating activity of CCNB1 and CDK1 (Fig. 4C and SI Appendix, Fig. S6 D and E) (21, 22). Accordingly, analysis of in vivo drug-treated tumors revealed significantly suppressed expression of the mitotic marker phosphohistone H3 (pHH3) in melanoma PDX lines treated with bromo/cobi in vivo (Fig. 4D and SI Appendix, Fig. S6F).

Finally, also prompted by the RNA-seq analysis, we assessed the effects of bromo/cobi treatment on the angiogenic potential of melanoma cells. Combinatorial treatment of melanoma cells in culture was shown to suppress HIF1A expression (by immunofluorescence analysis) (Fig. 4E and SI Appendix, Fig. S6 G–I) and VEGFA secretion (using ELISA analysis) (Fig. 4F). Analysis of microvessel density using CD31 immunofluorescence analysis of in vivo PDX tumors treated with bromo/cobi demonstrated significant anti-angiogenic effects (Fig. 4G and SI Appendix, Fig. S6J).

In this study, we assessed the consequences of bromodomain inhibition with bromosporine as a rational strategy for the therapy of melanoma, both in the treatment-refractory and treatment-naïve setting. Bromosporine monotherapy was effective in culture against multiple established melanoma cell lines and PDX models. The combination of bromo with either BRAFi (vem) or MEKi (cobi) resensitized BRAFi-resistant PDX lines to MAPK pathway targeting. The bromo/cobi combination proved consistently superior in multiple treatment-refractory PDX models to the standard BRAFi/MEKi regimen, a phenomenon that was also demonstrated in a previously untreated BRAF-mutant melanoma line. The activity of bromo/cobi was also demonstrated in immunotherapy-resistant NRAS- and NF-1-mutant PDX lines. Finally, RNA-seq analysis identified key genes and pathways altered following bromo/cobi administration, providing evidence of profound DNA damage and cell-cycle arrest, resulting in apoptosis induction.

Several prior studies have examined the consequences of acquired resistance to BRAFi in efforts to identify therapeutic vulnerabilities to target treatment-resistant melanoma in vivo. These include the use of compounds that inhibit RAF, PAK, or histone deacetylation, as well as those triggering ferroptosis, autophagy, ER stress, or antioxidation (23–30). Nevertheless, the findings presented here are significant and unprecedented for a number of reasons. To begin with, they systematically demonstrate the ability of bromodomain inhibition to resensitize to MAPK pathway inhibition in vivo in multiple PDX models of acquired BRAFi resistance. This was a key hypothesis that prompted our analysis, and studies of multiple PDX lines derived following progression on BRAFi in cell culture and in vivo provided consistent evidence of down-regulation of pERK levels following bromo/cobi treatment, providing a mechanism for resensitization to MAPK pathway inhibition. It is important to note that two other studies have shown the potential to sensitize to BRAFi therapy in vivo. Girotti et al. (31) demonstrated the improved activity of gefitinib plus vemurafenib in an A375 subline with acquired BRAFi resistance, while Vargas et al. (32) reported the activity of BETi and MEKi in a PDX model of acquired resistance to dabrafenib.

Beyond the resensitization strategy, our study of MM-313 cells provided evidence of superiority of the bromo/cobi combination to standard BRAFi/MEKi therapy in a treatment-naïve PDX line. These studies uniquely demonstrate the significantly improved activity of this targeted approach in vivo when compared with standard-of-care BRAFi/MEKi therapy in melanoma. In addition, our studies of BRAFi resistance are unique in their assessment of PDX lines that were also resistant to standard immunotherapeutic approaches in melanoma, including antibodies to either CTLA-4 or PD-1 (or both). Accordingly, we showed the activity of the bromo/cobi combination in the setting of BRAF-, NRAS-, and NF-1-mutant melanoma failing therapy with PD-1 blockade. Thus, our studies suggest bromo/cobi as a promising therapeutic approach with broad-based activity against distinct molecular subtypes of melanoma, both in the treatment-refractory and treatment-naive setting. As our studies assessed the consequences of bromo/cobi treatment in various PDX models for up to 38 d, whether these anti-tumor effects can persist beyond this interval awaits future studies of this combination.

While our analysis was motivated in part by prior studies on the role of BPTF in melanoma progression, it is important to stress that bromo targets several bromodomains, and as such, its anti-tumor effects go beyond those driven by BPTF inhibition alone. Future studies with more specific inhibitors of BPTF would address the issue of anti-tumor activity of targeting BPTF alone for the therapy of melanoma and other malignancies in which BPTF is activated. Nonetheless, our studies demonstrate compelling activity for bromosporine-based therapy in the setting of targeted therapy resistance. In addition, in contrast to BET inhibitors, which have been associated with significant toxicity in the clinic (33), the bromo/cobi regimen was well tolerated in NSG mice utilized for the PDX models.

Beyond the demonstration of effects on the MAPK pathway, our analysis identified previously unreported mechanisms of action for the bromo/cobi combination. Specifically, RNA-Seq analysis of drug-treated tumors in vivo identified numerous differentially expressed genes involving fundamental cellular processes such as cell division, chromosome segregation, and mitotic progression. Analyses in melanoma cell lines confirmed the differential expression of these genes following combinatorial drug therapy and enabled a broader understanding of the cellular effects of bromo/cobi administration. Specifically, bromo/cobi treatment produced profound DNA damage, ultimately resulting in exhaustion of cellular pools of RPA and RAD51, which protect replication forks from breakage, and whose depletion results in replication catastrophe (18, 19).

In addition to the effects of drug treatment on DNA replication, RNA-seq analysis of bromo/cobi-treated tumors also indicated down-regulation of several proteins essential to mitotic entry and progression, including CCNB1, CDK1, AURKA, and PLK1, accompanied by activation of p21. These results were consistent with the cell cycle analyses, in which drug treatment prevented mitotic entry. The CCNB1-CDK1 complex is critical for mitotic entry, whereas AURKA and PLK1 play key roles during chromosome segregation. In addition, given the regulation of BCL2 by CDK1 (34), suppression of CDK1 can directly result in inactivation of BCL-2, tipping the balance in favor of proapoptotic mechanisms. Taken together, the repressive transcriptional cascade mediated by bromo/cobi treatment (SI Appendix, Fig. S7) results in induction of DNA damage, cell-cycle arrest, and apoptosis, as well as reduced angiogenic potential, that together account for its significant anti-tumor activity.

In summary, our results identify bromodomain inhibition with bromosporine, either alone or in combination with MAPK pathway inhibitors, as a promising therapeutic approach for multiple molecular subtypes of melanoma, including in the setting of acquired resistance to both targeted and immunotherapeutic interventions.

Experimental details and methods can be found in SI Appendix, including sources of cell lines and cell culture conditions, assays of cell cycle, viability, apoptosis and animal studies. Methods for RNA-seq, Western blot, qRT-PCR, ELISA, and comet assays, as well as immunofluorescence and immunostaining of tumor specimens and statistical analyses are also described in SI Appendix. The PDX models utilized were developed after informed consent obtained from patients under the auspices of a study protocol approved by the Institutional Review Board at California Pacific Medical Center.

These studies were supported by National Institutes of Health (CA114337 and CA215755 to M.K.S.). We thank the California Pacific Medical Center Foundation for their support of the Cancer Avatar Program.