Dual MAPK Inhibition Is an Effective Therapeutic Strategy for a Subset of Class II BRAF Mutant Melanomas

BRAF is a constituent of the MAPK signaling pathway and is one of the most commonly mutated oncogenes in human tumors (1). The most prevalent BRAF mutations occur at codon V600, constitutively activating BRAF's kinase domain and enhancing MAPK signaling (2). Given the importance of this hyperactivated pathway in cancer, several MAPK inhibitors have been developed for targeted treatment of V600 BRAF mutant tumors, including BRAF inhibitors (BRAFi; vemurafenib, dabrafenib, and encorafenib) and MEK inhibitors (MEKi; cobimetinib, trametinib, and binimetinib; refs. 3, 4). BRAFi and MEKi used as single agents, or in combination, have been shown to improve survival in BRAF V600 mutant melanoma and non–small cell lung cancer (NSCLC; refs. 4–6).

Data from large-scale sequencing efforts have identified many additional hotspot BRAF mutations existing outside of the V600 codon (1, 7). Recently, a new classification system of BRAF mutations has been proposed (8, 9). V600 mutations are referred to as class I BRAF mutations and signal constitutively as RAS-independent monomers. Class II mutations are also BRAF-activating, but signal as RAS-independent dimers (9–11). Herein, we draw a distinction between class II BRAF mutations based on their location; class IIa mutations occur within the activation segment (i.e., L597, K601), and class IIb mutations occur within the glycine-rich p-loop (i.e., G464, G469; Fig. 1A). Class III is comprised of “low activity” or kinase-dead BRAF mutations (9, 12).

It has been previously reported that only tumors with class I BRAF mutations are sensitive to approved BRAFi (10). However, several other studies report that cell lines endogenously expressing non-V600 BRAF mutants are sensitive to BRAFi (13–15). This evidence, combined with case reports of patients with BRAF non–V600-expressing tumors responding to BRAFi suggests that the established paradigm for non-V600 BRAF mutants may be incomplete (8, 13, 16, 17).

In this study, we use cell lines and patient-derived xenograft (PDX) models and report on clinical responses in 2 patients to demonstrate that dual MAPK pathway inhibition (dMAPKi) with approved BRAFi + MEKi is an effective therapeutic strategy for some patients with class II BRAF mutant melanoma. These results provide the rationale for clinical trials to assess the efficacy of dMAPKi in these patients.

A next-generation sequencing–based test was performed by the CANCERPLEX assay (18). The CANCERPLEX data analysis pipeline was applied to report single-nucleotide variants, insertions, deletions, structural variants, and copy-number variations. For each patient tumor, the reported mutations in the primary metastatic tumor sample and the PDX sample were intersected to identify common variants. Variant allele frequencies (VAF) were compared and plotted using R (www.R-project.org). Variants of interest were manually reviewed in BAM files using IGV (19).

For long-term growth assays, cells were seeded into 12-well plates and treated with inhibitors at the following concentrations for 10- and 15-day assays, respectively: vemurafenib (1,500 and 2,000 nmol/L), dabrafenib (150 and 300 nmol/L), encorafenib (150 and 300 nmol/L), LY3009120 (100 nmol/L), cobimetinib (25 nmol/L), trametinib (5 nmol/L), and binimetinib (50 and 100 nmol/L). Media with drug were replaced every 4 to 5 days. At experimental endpoint (10 days for Fig. 2A; Supplementary Fig. S1C, 15 days for Fig. 2B and C), cells were fixed in 10% formalin, incubated in crystal violet (Sigma-Aldrich; Cat # HT90132-1L), and washed in water. Five representative images were taken of each well and quantified using Scion Image Software. Positive pixel count was acquired from these images, representing the area covered by tumor cells. Experiments were repeated in 3 wells per experiment and performed in triplicate for a total of 9 wells.

For subcutaneous tumor growth experiments, 5 × 10⁵ tumor cells were injected bilaterally. For cranial tumor growth experiments, 1 × 10⁵ tumor cells were injected into the right frontal lobe using a guide screw technique (20). All in vivo subcutaneous and cranial PDX experiments were performed with passage 2 or earlier, or passage 5 or earlier, respectively. For subcutaneous xenografts, tumors were measured with calipers (ASICSA; cat # 19600). For brain metastasis measurements, lesions were measured with IVIS Spectrum (Perkin Elmer). For each mouse prior to imaging, 50 μL of luciferin was injected intraperitoneally. Quantification of signal intensity was performed with Living Image software. For cranial injection experiments, treatment was initiated when all mice exhibited clear detectable lesions by IVIS Spectrum imaging. Mice were treated by daily oral gavage with vehicle of hydroxypropyl methylcellulose, trametinib (LC Laboratories T-8123) at 0.5 mg/kg mouse body weight, dabrafenib at 25 or 50 mg/kg, as indicated in the figure legends (LC Laboratories D-5699), encorafenib (Array Biopharma) at 75 mg/kg, and binimetinib (Array Biopharma) at 15 mg/kg. All animal studies and protocols were preapproved by the McGill Comparative Medicine and Animal Resources Centre.

Patient clinical information and tissue were received after obtaining written-informed consent from patients in accordance with the Declaration of Helsinki and after studies were approved by an Institutional Review Board.

To assess the prevalence of class II mutations across tumor types, we accessed the AACR GENIE project (21). Within this dataset, among tumor types with at least 5 BRAF mutant tumors present, the prevalence of BRAF mutations varied substantially from 0.4% in breast cancer to 40.4% melanoma (Fig. 1B). Among melanoma samples, class I mutations comprised 65.9% of all BRAF mutations, whereas class II and III comprised 11.4% and 9.5%, respectively (Fig. 1C). A further 13.2% were mutants of unknown function that did not belong to any of the three classes. A similar distribution of class II and III BRAF mutations were observed in The Cancer Genome Atlas (TCGA) melanoma dataset (22). Class II mutations occurred within the activation segment (i.e., L597, K601; class IIa) and in the glycine-rich P-loop (i.e., G464, G469; class IIb; Fig. 1A). An additional subset of class II mutations is comprised of BRAF fusions (class IIc) that have also been reported to signal as RAS-independent BRAF dimers (10, 23, 24). All non-V600 mutations identified in the AACR GENIE dataset with known function are indicated in Supplementary Table S1.

It has been reported that class III BRAF mutations are commonly associated with RAS mutations in melanoma (9). Indeed, we found that 47% of class III mutant melanoma within the GENIE dataset coexpressed activating RAS mutations (Supplementary Table S1; Fig. 1D). In contrast, we found that class II mutant tumors were similar to class I mutant tumors, in that they rarely coexpressed activating RAS mutations (2.8% and 1.4%, respectively). These data support the notion that BRAF class II mutations, like class I mutations, are kinase activating in a RAS-independent manner.

In datasets published before the widespread approval of BRAFi and MEKi, melanoma patients with BRAF V600 mutations who did not receive MAPKi had worse prognosis than those with BRAF wild-type (WT) tumors (25). We asked whether melanoma patients with other, potentially targetable mutations also experienced poor prognosis. Indeed, metastatic melanoma patients with class II/III and/or NRAS mutations in the TCGA dataset experienced inferior overall survival compared with patients with class I mutations (Fig. 1E). The improved survival of melanoma patients with class I BRAF mutations due to the development of targeted therapies highlights the need for the identification of similarly effective targeted therapy strategies for patients with class II/III BRAF mutant and NRAS mutant melanoma (4).

We established PDXs from 4 patients with metastatic melanoma, including 2 with class II BRAF mutations (both BRAF L597S; Fig. 3A). All PDXs retained similar genomic landscapes compared with the tumor from which they were derived. Genomic analyses included copy-number alterations (CNA; Fig. 3B; Supplementary Fig. S2A), somatic missense variants (Fig. 3C and D), and VAFs (Supplementary Fig. S2B). An exception to this trend was the expected discrepancy in the CNA and VAF between the clinical specimen and GCRC2073 PDX (Supplementary Fig. S2A and S2B). This was due to a low purity of the patient sample that can be seen in the representative hematoxylin and eosin (H&E) from this specimen (Fig. 3E). Importantly, the known driver mutations that result in gain (BRAF, NRAS, RET) or loss of function (PTEN, ARID2, and CDKN2A) were conserved in the corresponding PDX models (Fig. 3D). Immunohistochemical staining of three PDX models and corresponding patient tissues revealed that PDXs maintain similar expression of melanoma markers (Melan-A, BRAF V600E, HMB-45) compared with their tumor of origin (Fig. 3E). Taken together, these profiles demonstrate the high fidelity of these PDX model systems to the metastatic tumor from which they were derived.

We also obtained a variety of cancer cell lines bearing WT, class I mutant, and class II mutant BRAF. Among these cell lines and PDX models, cooccurring RAS mutations were present in 2 of 7 melanoma cell lines that expressed class II BRAF mutations (Supplementary Table S2). This is consistent with the notion that activating RAS mutations are commonly found in class III but less frequently in class I or class II BRAF mutant melanomas (Fig. 1C; ref. 9). Both class II BRAF mutant melanoma cell lines that coexpressed activating RAS mutations were of the class IIb type. Class IIb–activating mutations have been reported to enhance mutant BRAF:CRAF dimerization (26), and therefore, the presence of an activating RAS mutation may facilitate their signaling capacity in this manner.

Clinically indicated BRAFi, such as vemurafenib and dabrafenib, cause paradoxical activation of the MAPK pathway in cells with WT BRAF (27, 28), but it is unclear whether the same is true for class II BRAF mutant tumors (7, 10, 11). Therefore, we used cells derived from the aforementioned PDX and cell line models (Supplementary Table S2) to determine whether cells with class II mutations are responsive to BRAFi or MEKi in vitro.

Short-term treatment with MEKi (cobimetinib and trametinib) universally inhibited the MAPK pathway, irrespective of BRAF class (Fig. 2A; Supplementary Fig. S3A). Short-term treatment with BRAFi (vemurafenib, dabrafenib) induces paradoxical activation of the MAPK pathway in BRAF WT cells, whereas class I (BRAF V600) and IIa (K601E and L597S) mutant cells exhibit marked inhibition of the MAPK pathway (Fig. 2B; Supplementary Fig. S3A). In contrast, the class IIb mutant cancer cells tested were neither paradoxically activated nor inhibited by single-agent BRAFi (Fig. 2B). This result highlights the marked difference between class IIa and class IIb cells with respect to their biochemical response to BRAFi. These differences may be based on the location of the mutation within the BRAF protein or by the RAS mutation status of the cell lines tested (Supplementary Table S2).

One of the key determinants of BRAFi efficacy is the speed of pERK recovery following drug treatment (10). We sought to compare the dynamics of pERK recovery between melanoma cells of different BRAF mutant classes treated with physiologically relevant doses of encorafenib. Encorafenib is an emerging BRAFi that is a promising candidate to become a front-line targeted therapy for class I BRAF mutant melanoma (3). Cells were treated with encorafenib for 1 hour, washed with drug-free media, and then lysed at defined time points after washout. In WM3918 BRAF WT cells, we observe paradoxical activation of the MAPK pathway at 1 hour on treatment, which returns to baseline levels within minutes after treatment (Fig. 2C). In A375 class I BRAF mutant melanoma cells, we observe strong pERK inhibition that recovers to baseline levels by 8 hours after removal of drug. In class IIa mutant cells, pERK levels returned to baseline at earlier time points (1–2 hours) following drug removal. By contrast, pERK levels in class IIb HMV-II and M619 melanoma cells are not significantly decreased by encorafenib after 1 hour of treatment. These data indicate that class I and, to a lesser extent, class IIa mutant BRAF dimers are effectively inhibited by single-agent encorafenib, whereas WT and class IIb BRAF dimers are not.

Another important indicator of MAPKi efficacy is the extent to which the inhibitory signal is propagated to downstream effector molecules. Such signals include cell-cycle regulators such as Cyclin D1 (CCND1; ref. 29), which in turn phosphorylates the retinoblastoma (Rb) tumor-suppressor protein to promote cell survival and proliferation (30). In addition to these transcriptionally regulated targets of the MAPK pathway, ERK is itself a kinase that phosphorylates and stabilizes a number of effector proteins with critical functions, including FRA-1 (31).

We examined the effects of either MEKi (trametinib; Fig. 2D) or BRAFi (encorafenib; Fig. 2E) on these downstream effectors of the MAPK pathway. Trametinib inhibited phosphorylation of ERK, FRA-1, and Rb. Total levels of CCND1, FRA-1, and Rb were also diminished in all cell lines treated with trametinib. Encorafenib similarly inhibited these downstream signaling components to a comparable extent in class I and class IIa mutant melanoma cells. This demonstrates that cell proliferation and survival pathways are inhibited in class IIa mutant cells treated with either BRAFi or MEKi.

Next, we sought to determine whether class II BRAF mutant cells were growth inhibited by these targeted therapies using standard BRAFi or MEKi doses that achieved >50% growth inhibition of BRAF V600 mutant cells in short-term proliferation assays (Supplementary Fig. S3B and S3C). In clonogenic growth assays, MEKi effectively inhibited the growth of class I and IIa mutant cancer cells and, to a lesser extent, BRAF WT and class IIb cells (Fig. 4A). BRAFi inhibited growth of class I and IIa BRAF mutant melanoma cells but did not significantly impair the growth of WT and class IIb mutant cancer cells (Fig. 4A). Representative images of clonogenic assays from each class are shown in Supplementary Fig. S4A. Although class I BRAF mutant cells responded similarly to all 3 BRAFi, we observed a marked contrast in class IIa mutant cells between the marginal efficacy seen with vemurafenib and stronger inhibition of cell proliferation in the presence of dabrafenib and encorafenib. Class IIb mutant cells were not significantly growth inhibited by single-agent BRAFi (Fig. 4A).

LY3009120 (LY), a pan-RAF and BRAF dimer inhibitor that is in early stage clinical development, was also tested to assess its efficacy in class II BRAF mutant cells. LY inhibited pERK in class I and II cell lines at low doses, but induced modest paradoxical activation in the BRAF WT cell line, WM3918 (Supplementary Fig. S1A). Using short-term cell growth assays, we determined the dose of LY3009120 that achieved >50% growth inhibition of BRAF V600 mutant cells at 2 days (Supplementary Fig. S1B). In clonogenic growth assays, LY3009120 at this dose (100 nmol/L) moderately inhibited growth of WM3918 but substantially inhibited growth of class I and II BRAF mutant cells, including the class IIb cell line, HMV-II (Supplementary Fig. S1C and S1D). These data demonstrate that although LY3009120 is only marginally effective in BRAF WT cells, it may also be effective for patients with class II BRAF mutant tumors.

To assess efficacy of a combined therapeutic strategy using BRAFi and MEKi, cells were treated with the standard clinical BRAFi/MEKi combinations (vemurafenib/cobimetinib, dabrafenib/trametinib, encorafenib/binimetinib). We observe augmented growth inhibition when either dabrafenib, encorafenib, or LY3009120 were added to a MEKi in class I or IIa BRAF mutant cells (Fig. 4A; Supplementary Fig. S1C and S1D). Although dMAPKi did significantly inhibit the growth of BRAF WT cells compared with DMSO, combined BRAFi + MEKi was less effective than single-agent MEKi in most WT cells. In particular, we observed significantly enhanced growth of BRAF WT cells treated with vemurafenib (SkMel2 P = 0.009; WM3918 P = 0.005; CHL1 P = 0.024) or dabrafenib (SkMel2 P = 0.047, WM3918 P = 0.001) in addition to a MEKi, compared with MEKi alone (Fig. 4A). Encorafenib did not significantly enhance the growth of any BRAF WT cells treated with a MEKi. Conversely, encorafenib significantly inhibited the growth of binimetinib-treated triple WT CHL1 cells (P = 0.047). In class IIb mutant cell lines, we consistently observed further growth inhibition only with encorafenib, but not with vemurafenib, when added to a MEKi (Fig. 4A).

Next, we sought to directly compare the effects of specific BRAFi when added to the same MEKi. To do so, we tested each BRAFi in combination with either trametinib or binimetinib. In long-term growth assays where all cells were grown in the presence of trametinib ± BRAFi (Fig. 4B; Supplementary Fig. S4B), vemurafenib potentiated the growth of BRAF WT, NRAS mutant SkMel2 and GCRC1987 cells, and all class IIb mutant cells tested. Meanwhile, vemurafenib modestly augmented growth inhibition of class I A375 and class IIa mutant cells. Dabrafenib potentiated the growth of trametinib-treated SkMel2 and GCRC1987 cells but inhibited the growth of all trametinib-treated class I, IIa, and IIb cells, with the sole exception of class IIb mutant MDA-MB-231 breast cancer cells, which were unaffected by the addition of dabrafenib. When added to trametinib, encorafenib potently inhibited the growth of BRAF WT, NRAS mutant SkMel2 cells, and all class I, IIa, and IIb BRAF mutant cells tested (Fig. 4A). Similar BRAFi effects were observed when binimetinib was used as the MEKi (Fig. 4C; Supplementary Fig. S4C).

In all classes of cells, 48-hour treatment with trametinib led to sustained inhibition of ERK phosphorylation (Fig. 4D). In class I BRAF mutant melanoma, dMAPKi further impairs the MAPK pathway (32, 33). Therefore, we asked if the addition of BRAFi will have a similar effect in MEKi-treated class II cells. In class I and class IIa mutant cells, the addition of any BRAFi further exacerbated ERK inhibition. Meanwhile, in class IIb mutant cells, only encorafenib led to more profound ERK inhibition than trametinib alone (Fig. 4D).

To determine whether BRAFi + MEKi combinations cause tumor regression in vivo, we used our melanoma PDX models bearing class IIa, BRAF L597S mutations (GCRC2015 and GCRCMel1). Tumors were implanted subcutaneously and treated with vehicle or MAPKi.

In the GCRCMel1 model, treatment with either single-agent dabrafenib or single-agent trametinib was insufficient to induce shrinkage in any of the tumors, but 17 of 19 (89%) tumors treated with dabrafenib + trametinib had shrunk by day 4 (Fig. 5A). This result was corroborated with immunoblots that demonstrated decreased pERK in dabrafenib + trametinib 4-day treated tumors, compared with vehicle-, dabrafenib-, or trametinib-treated tumors (Fig. 5E). Both dabrafenib and trametinib, when used as single agents, were capable of delaying the growth of GCRCMel1 tumors over time. Meanwhile, tumors treated with dMAPKi were significantly more growth inhibited than tumors treated with either single agent (Fig. 5C). By day 15, all tumors in the study had begun to progress, implying that they had acquired resistance to MAPKi (Fig. 5C). Phosphorylated ERK levels were uniform between all arms at experimental endpoint (Fig. 5E). Resistant dabrafenib + trametinib-treated tumors demonstrated increased expression of the HER3 receptor tyrosine kinase (RTK), as well as increased pAKT and pCRAF (Fig. 5E), implying potential mechanisms of resistance to dMAPKi in class II mutant tumors.

In the GCRC2015 model, after 4 days of treatment, 83.3% (10/12) of vehicle-treated tumors were progressively growing. Trametinib monotherapy induced tumor shrinkage in 75% (8/12) of subcutaneous tumors. Meanwhile, dMAPKi with dabrafenib and trametinib induced tumor shrinkage in 100% (13/13) of tumors (Fig. 5B). Immunoblot analysis revealed that early into treatment, dabrafenib augmented the inhibitory effect of trametinib on the MAPK pathway (Fig. 5F). All treatment groups eventually began to acquire MAPK inhibitor resistance, as evidenced by the reactivation of pERK (Fig. 5F) and increasing tumor growth (Fig. 5D) at the experimental endpoint of 14 days. However, 88.9% (8/9) of dabrafenib + trametinib compared with 0% (0/8) of trametinib-treated tumors maintained an overall reduction in tumor size at endpoint. Immunoblots from GCRC2015-resistant tumors demonstrate the same resistance mechanisms as the GCRCMel1 model, in that RTKs (HER2 and HER3) were upregulated, coinciding with increased pAKT in all MAPKi-treated tumors. In both PDX models, we only observed enhanced pCRAF in tumors that had acquired resistance to dMAPKi with dabrafenib + trametinib (Fig. 5E and F). Together, this suggests that activation of CRAF is a mechanism of resistance that is unique to dMAPKi in class II BRAF mutant melanoma.

Treatment with encorafenib, binimetinib, or encorafenib + binimetinib produced similar results to dabrafenib + trametinib, causing shrinkage of 8% (1/12), 25% (3/12), and 67% (8/12) of GCRC2015 tumors, respectively, whereas all of the vehicle tumors were progressively growing by day 4 (Supplementary Fig. S5A). Immunoblot analysis of tumors treated for 4 days demonstrated that both encorafenib and binimetinib robustly inhibit ERK phosphorylation as single agents, whereas the encorafenib + binimetinib combination further inhibited pERK compared with either agent alone (Supplementary Fig. S5B). Both encorafenib and binimetinib, when used as single agents, delayed GCRC2015 tumor growth. Combined encorafenib + binimetinib elicited tumor shrinkage and more significant tumor growth delay compared with either single agent (Supplementary Fig. S5C).

Importantly, the patient from whom the GCRCMel1 PDX was derived presented with stage IV (M1a) metastatic melanoma, with disease involving the inguinal lymph nodes, muscle, and adjacent soft tissues. This patient was treated with dabrafenib + trametinib and achieved an objective radiographic response, with a 34% reduction in tumor size at 2 months on treatment (Fig. 5G). After several months of treatment, the patient began to experience drug toxicity (pyrexia, hepatotoxicity) despite dose reductions and was switched to immunotherapy. These observations of an objective partial response provide proof-of-principle demonstrating that dabrafenib + trametinib has clinical activity in class IIa BRAF mutant melanoma.

GCRC2015 was derived from a melanoma brain metastasis. In light of recent data indicating that dMAPKi can effectively shrink brain metastases in patients with BRAF V600E mutant melanoma (34), we asked whether dabrafenib + trametinib would be similarly effective in class II BRAF mutant brain metastases. The GCRC2015 PDX model was propagated as an intracranial xenograft and infected with pHIV-Luc-ZsGreen virus to allow longitudinal bioluminescence imaging in vivo (Fig. 6A). We observed that both trametinib monotherapy and dabrafenib + trametinib slowed growth of GCRC2015 intracranial tumors. At the experimental endpoint, the change in in vivo bioluminescence from the time treatment was initiated was significantly smaller in the dabrafenib + trametinib group compared with trametinib or vehicle (Fig. 6B; Supplementary Fig. S6A). Indeed, after 9 days of treatment, the average size of the intracranial tumors from dMAPKi was smaller than tumors treated with trametinib alone or vehicle (Supplementary Fig. S6B and S6C). Furthermore, immunohistochemistry for pERK revealed decreased staining in dabrafenib + trametinib, but not in trametinib-treated brain metastases (Fig. 6C and D). In longer-term survival analyses, trametinib alone did not significantly prolong survival compared with vehicle treatment of mice bearing class II BRAF mutant melanoma brain metastases (Fig. 6E). However, dMAPKi treatment did significantly improve survival of mice compared with either trametinib monotherapy or vehicle. Finally, we retrospectively identified a patient with class II BRAF (L597S) mutant melanoma with brain metastases. This patient received treatment with dabrafenib + trametinib and experienced a dramatic response in metastases in the brain, lung, liver, and adrenal gland (Fig. 6F). After 4 months of treatment, this patient developed progressive brain metastases. She went on to receive additional brain radiotherapy and immunotherapy but eventually died of her disease. These observations provide further validation that dMAPKi can induce objective responses in visceral and brain metastases of patients with class II BRAF mutant melanoma.

We initiated this study after encountering 2 melanoma patients with BRAF L597S mutations in clinical practice. At the time of their presentation, little was known about class II BRAF mutations and their responsiveness to targeted therapy. We sought to better characterize this emerging class of BRAF mutant tumors and to inquire whether patients with class II BRAF mutant melanoma are responsive to therapeutic intervention with approved targeted therapies.

Data from colorectal cancer (35) or NSCLC (36) indicate that patients with non-V600 BRAF mutations tend to experience improved overall survival than those with V600 BRAF mutations. In contrast, we report here that advanced melanoma patients with potentially targetable NRAS and/or class II/III BRAF mutations experience worse survival than those with class I BRAF mutations. This finding highlights the need for improved therapeutic strategies for melanoma patients with non-V600 BRAF mutations.

We draw the distinction between class IIa mutations within the activation segment and class IIb mutations within the glycine-rich p-loop. Class IIa and IIb mutations have been shown to engender enhanced kinase activity that is RAS-independent and dimerization-dependent (10). However, the class IIa and IIb mutant cells tested are unique in terms of their sensitivity to single-agent BRAFi: class IIa BRAF mutant cells were sensitive to single-agent BRAFi, whereas class IIb BRAF mutant cells were not. The differential sensitivities of class IIa and IIb BRAF mutants to approved BRAFi may be due, in part, to the ability of BRAFi to more effectively inhibit the second protomer of a class IIa BRAF mutant dimer than that of IIb BRAF dimers. Alternatively, although it is clear that BRAF L597S and K601E signal predominantly as dimers (10), class IIa mutant BRAF may harbor some degree of monomeric signaling capacity. It is also possible that class IIa mutants more readily form BRAF homodimers, whereas class IIb mutants form BRAF:CRAF heterodimers, rendering them less sensitive to BRAFi (26). Interestingly, we found that activated CRAF is a common resistance mechanism to dMAPKi in our two class II BRAF L597S mutant PDX models; this result suggests that acquired resistance to dMAPKi with BRAFi and MEKi in class II BRAF mutant melanoma results from a shift from primarily BRAF homodimer-driven MAPK signaling toward BRAF:CRAF heterodimer or CRAF homodimer-driven MAPK signaling.

Although all three BRAFi augmented MEKi-mediated growth inhibition in class IIa mutant cells, encorafenib was the only BRAFi that consistently augmented MEKi-mediated growth inhibition in class IIb mutant cells. The contrast between vemurafenib, dabrafenib, and encorafenib in this context may be due to differences in eliciting paradoxical activation of the MAPK pathway. This results from differential ability of each inhibitor to bind and inhibit the second protomer of a BRAF dimer. It has been shown that significantly higher concentrations of vemurafenib are required to inhibit BRAF dimers, compared with encorafenib and dabrafenib (10, 37). Moreover, recent findings have demonstrated the efficacy of encorafenib when used in combination with MEKi in NRAS mutant melanoma through an ER stress pathway (38). This may explain the sensitivity we observe in NRAS mutant SKMel2 cells treated with MEKi + encorafenib, highlighting unique properties of encorafenib that may support its broader utility among melanoma patients. The encorafenib + binimetinib combination has been shown to provide a significant survival advantage over vemurafenib in BRAF V600E/K mutant melanoma, with a favorable safety profile (3). Therefore, this combination is promising for patients with class II BRAF mutations and perhaps even some NRAS mutant melanoma patients.

It has been proposed that non-V600 BRAF mutant melanoma are sensitive to single-agent MEKi, prompting an ongoing trial recruiting non-V600 mutant melanoma patients for treatment with trametinib (39, 40). Since these initial observations, several clinical trials investigating single-agent MEKi have failed to yield sustained clinical benefit in a variety of indications (41–43). In BRAF V600 mutant melanoma, trametinib has a much lower overall response rate (22%) than single-agent BRAFi (48%–51%; ref. 4). These data suggest that the more effective therapeutic approach of approved agents for class II BRAF mutant melanoma would be combination therapy including a clinically viable BRAFi (i.e., encorafenib) plus a MEKi. Moreover, in addition to the potential for enhanced efficacy with dMAPKi, these combination regimens are frequently better tolerated than either BRAFi or MEKi alone (3, 44).

In this study, we established two PDX models of BRAF L597S mutant melanoma in order to assess their sensitivity to MAPKi. Importantly, the PDX models established herein adequately retain the genetic features of their tumors of origin. We demonstrate in both class II BRAF mutant PDX models that dMAPKi augments inhibition of the MAPK pathway and impairs tumor growth of class II BRAF mutant melanoma compared with single-agent therapy. Single-agent MEKi produced only short-lived stable disease in the GCRC2015 BRAF L597S subcutaneous PDX model and only modestly slowed progression of the GCRCMel1 model, whereas the addition of BRAFi to MEKi resulted in sustained partial responses in the majority of tumors in both models. As an important proof of principle, we show that single-agent encorafenib is able to robustly inhibit the MAPK pathway and slow tumor growth in the GCRC2015 PDX model. Furthermore, we report a partial response to dabrafenib + trametinib in patient GCRCMel1, corroborating the results from the aforementioned PDX studies.

Although our results indicate that dMAPKi is superior to single-agent MAPKi for class II BRAF L597S mutant melanomas, the duration of growth inhibition with dMAPKi was less than we have observed in class I BRAF mutant melanoma (45). This implies that melanoma patients with class II BRAF mutations may be less responsive to dMAPKi than those with class I BRAF mutations. Moreover, both of our PDX models bore a BRAF L597S mutation, and as such it is unknown at this point whether other common class IIa mutations (i.e., L597Q/R/V, K601N/T) would derive equivalent benefit from dMAPKi in vivo. As such, investigation into combinations of dMAPKi with antibody drug conjugates (46), ERK inhibitors (47), and immunotherapies is also warranted for class II BRAF mutant melanoma. Further investigation of therapies targeting the resistance pathways we identified in both PDX models, such as RTKs, PI3K/AKT signaling, and CRAF, may also be beneficial in preventing or delaying resistance to dMAPKi.

The patient from whom the GCRC2015 PDX was derived presented with brain metastases. Cytotoxic chemotherapies have minimal effect in intracranial metastatic disease, in part due to limitations of the blood–brain barrier (48). However, emerging data espouse the efficacy of systemic immunotherapies and MAPKi in the management of brain metastatic melanoma (34, 48). In this study, dMAPKi provided a significant survival advantage to mice with class II BRAF L597S mutant brain metastases, whereas single-agent MEKi did not. Therefore, we speculate that a similar approach can be applied for melanoma patients with class II BRAF mutant tumors, including those with brain metastases. This is further supported by a patient with BRAF L597S brain-metastatic melanoma who experienced a major intracranial response to dabrafenib + trametinib.

In summary, we have provided in vitro, in vivo, and clinical evidence indicating that dMAPKi effectively impairs the growth of subsets of non-V600, class II BRAF mutant melanoma. These data provide intriguing preclinical rationale to support the development of clinical trials to investigate BRAFi + MEKi combinations in patients with class II BRAF mutations.

D. Hogg reports receiving speakers bureau honoraria from EMD Serono, and is a consultant/advisory board member for Bristol-Myers Squibb, EMD Serono, Merck, Novartis, and Roche. C. Mihalcioiu reports receiving speakers bureau honoraria from Novartis, and is a consultant/advisory board member for Bristol-Myers Squibb, Merck, Novartis, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Dankner, C. Mihalcioiu, A.A.N. Rose

Development of methodology: M. Dankner, D. Vuzman, A.A.N. Rose

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Dankner, D. Moldoveanu, T.-T. Nguyen, P. Savage, S. Rajkumar, M. Lvova, A. Protopopov, D. Hogg, M. Park, M.-C. Guiot, K. Petrecca, C. Mihalcioiu, A.A.N. Rose

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Dankner, M. Lajoie, S. Rajkumar, X. Huang, D. Vuzman, I.R. Watson, A.A.N. Rose

Writing, review, and/or revision of the manuscript: M. Dankner, M. Lajoie, D. Moldoveanu, S. Rajkumar, X. Huang, D. Vuzman, D. Hogg, M. Park, M.-C. Guiot, K. Petrecca, C. Mihalcioiu, I.R. Watson, P.M. Siegel, A.A.N. Rose

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Dankner, M. Park, C. Mihalcioiu, A.A.N. Rose

Study supervision: M. Park, C. Mihalcioiu, I.R. Watson, P.M. Siegel, A.A.N. Rose

The authors are grateful to the patients who donated the tissues studied in this work. The authors acknowledge technical assistance from the McGill/GCRC Histology core facility and the McGill Comparative Medicine and Animal Resources Centre (CMARC). The authors acknowledge the support and assistance of Valentina Muñoz Ramos and Margarita Souleimanova in biobanking and sample collection. The authors are thankful to Array Biopharma for providing encorafenib and binimetinib used in in vivo studies. The authors are thankful to Juan Canale, Karen Stone, Vasilios Papavasiliou, Matthew Annis, and William Muller for animal support. The authors are grateful to Nicholas Hayward, Antoni Ribas, Wilson Miller, and David Dankort for providing cell lines used in this study. The authors thank members of the Siegel laboratory for thoughtful discussions and critical reading of the article.

This research has been supported by a grant from the DOD (CA-140389; to P.M. Siegel). M. Dankner acknowledges support from the McGill University MD/PhD program and the Brain Tumour Foundation of Canada. I.R. Watson is funded by grants from the Melanoma Research Alliance (MRA—Grant #412429), the V Foundation (Grant #2016-023), and the Canadian Institute of Health Research (CIHR—Grant # PJT-152975). P.M. Siegel is a McGill University William Dawson Scholar. A.A.N. Rose acknowledges a David Cornfield Melanoma Fund Award.

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.