Targeting the MAPK Signaling Pathway in Cancer: Promising Preclinical Activity with the Novel Selective ERK1/2 Inhibitor BVD-523 (Ulixertinib)

Through aberrant activation of ERK signaling, genetic alterations in RAS or RAF family members result in rapid tumor growth, increased cell survival, and resistance to apoptosis. Activating mutations of RAS family members KRAS and NRAS are found in approximately 30% of all human cancers, with particularly high incidence in pancreatic and colorectal cancer (1, 2). Constitutively activating mutations in BRAF (codon for V600) have been observed primarily in melanoma (∼48%), papillary thyroid carcinoma (∼44%), colorectal cancer (∼15%), hairy cell leukemia (∼100%), and non–small cell lung cancer (∼4%; ref. 3). Although rare, cancers bearing genetic mutations that result in changes of the downstream components ERK (MAPK1 and MAPK3) and MEK (MAP2K1 and MAP2K2) have also been reported (4, 5). Furthermore, alterations known to activate the MAPK (RAS–RAF–MEK–ERK) pathway are also common in the setting of resistance to targeted therapies, for example inhibitors of BRAF, MEK, and ALK (6). Thus, targeting the MAPK pathway terminal master kinases (ERK1/2) is a promising strategy for the treatment of a broad spectrum of tumors harboring pathway-activating alterations.

Clinical precedent for targeting the MAPK pathway already exists; three MAPK pathway-targeting drugs have been approved by the FDA for single-agent treatment of nonresectable or metastatic cutaneous melanoma with BRAF^V600 mutations: the BRAF inhibitors vemurafenib and dabrafenib and the MEK inhibitor trametinib (7, 8). An additional MEK inhibitor, cobimetinib, is approved in this indication as part of a combination regimen with vemurafenib; furthermore, the combination of dabrafenib and trametinib is also approved in this indication (9, 10). Approved BRAF inhibitors provide the advantage of selectivity for BRAF^V600-mutant proteins, by virtue of ability to signal as monomers, unlike wild-type RAF proteins which require dimerization to activate signaling, therefore providing a greater therapeutic index.

Importantly, the concomitant use of BRAF- plus MEK-targeted therapies has demonstrated simultaneous targeting of different nodes in the MAPK pathway can enhance the magnitude and duration of response (11–13).

Despite improvements in clinical outcomes seen with BRAF-/MEK-inhibitor combination therapies, durable benefit is limited by the eventual development of acquired resistance and subsequent disease progression, with median progression-free survival (PFS) ranging from approximately 9 to 11 months (11, 13–15). Genetic mechanisms of acquired resistance to single-agent BRAF inhibition have been intensely studied; some examples of identified resistance mechanisms include splice variants of BRAF (16), BRAF^V600E amplification (17), MEK mutations (18), NRAS mutations, and RTK activation (19, 20). Resistance mechanisms in the setting of BRAF/MEK inhibitor combination therapy are beginning to emerge and mirror that of BRAF single-agent resistance (21, 22). These genetic events all share the ability to reactivate ERK signaling. Indeed, reactivated MAPK pathway signaling as measured by ERK transcriptional targets is common in tumor biopsies from BRAF inhibitor–resistant patients (23). Moreover, ERK1/2 reactivation has been observed in the absence of a genetic mechanism of resistance (24, 25). The quest to achieve durable clinical benefit has led researchers to focus on evaluating additional agents that target the downstream MAPK components ERK1/2. Inhibiting ERK may provide important clinical benefit to patients with acquired resistance to BRAF/MEK inhibition. ERK family kinases have shown promise as therapeutic targets in preclinical cancer models, including those resistant to BRAF or MEK inhibitors (26–28); however, the potential for ERK inhibitors expands beyond acquired resistance in melanoma.

Targeting ERK1/2 is a rational cornerstone therapy in any tumor type harboring known drivers of MAPK, not only BRAF/MEK therapy–relapsed patients. As ERK1/2 reside downstream in the pathway, they represent a particularly attractive treatment strategy within the MAPK cascade that may avoid upstream resistance mechanisms. Here, we report the preclinical characterization of BVD-523 (ulixertinib) in models of MAPK pathway–dependent cancers, including drug-naïve and BRAF/MEK therapy acquired–resistance models. Phase I clinical studies of BVD-523 are ongoing (NCT01781429, NCT02296242, and NCT02608229).

Compounds were sourced from the following vendors; SCH772984 (SelleckChem), pyrazolylpyrrole (Santa Cruz Biotechnology), Temozolomide (Temodar, Schering Corporation), dabrafenib (Chemietek), trametinib (LC Laboratories), vemurafenib (Focus Synthesis LLC), and selumetinib (Chemietek). BVD-523 was synthesized as per method described in US patent number 7,354,939 B2.

The inhibitory activity of BVD-523 against ERK2 was determined using a radiometric assay, with final concentration of the components being 100 mmol/L HEPES (pH 7.5), 10 mmol/L MgCl₂, 1 mmol/L dithiothreitol (DTT), 0.12 nmol/L ERK2, 10 μmol/L myelin basic protein (MBP), and 50 μmol/L ³³P-γ-ATP. All reaction components, with the exception of ATP and MBP, were premixed and aliquoted (33 μL) into a 96-well plate. A stock solution of compound in DMSO was used to make up to 500-fold dilutions; a 1.5-μL aliquot of DMSO or inhibitor in DMSO was added to each well. The reaction was initiated by adding the substrates ³³P-γ-ATP and MBP (33 μL). After 20 minutes, the reaction was quenched with 20% (w/v) tricholoracetic acid (TCA; 55 μL) containing 4 mmol/L ATP, transferred to the GF/B filter plates, and washed 3 times with 5% (w/v) TCA. Following the addition of Ultimate Gold scintillant (50 μL), the samples were counted in a Packard TopCount. From the activity versus concentration titration curve, the K_i value was determined by fitting the data to an equation for competitive tight binding inhibition kinetics using Prism software, version 3.0.

ERK2 activity was assayed by a standard coupled-enzyme assay. The final concentrations were as follows: 0.1 mol/L HEPES (pH 7.5), 10 mmol/L MgCl₂, 1 mmol/L DTT, 2.5 mmol/L phosphoenolpyruvate, 200 μmol/L NADH, 50 μg/mL pyruvate kinase, 10 μg/mL lactate dehydrogenase, 65 μmol/L ATP, and 800 μmol/L peptide (ATGPLSPGPFGRR). All of the reaction components except ATP were premixed with ERK and aliquoted into assay-plate wells. BVD-523 in DMSO was introduced into each well, keeping the concentration of DMSO per well constant. BVD-523 concentrations spanned a 500-fold range for each titration. The assay plate was incubated at 30°C for 10 minutes in the plate reader compartment of the spectrophotometer (Molecular Devices) before initiating the reaction by adding ATP. The absorbance change at 340 nm was monitored as a function of time; the initial slope corresponds to the rate of the reaction. The rate versus concentration of the BVD-523 titration curve was fitted either to an equation for competitive tight-binding inhibition kinetics to determine a value for K_i or to a 3-parameter fit to determine the IC₅₀ using Prism software, version 3.0.

The cell lines RKO, SW48, A375, MIAPaCa-2, HCT116, Colo205, HT-29, ZR-75-1, and AN3Ca were obtained from ATCC, confirmed mycoplasma negative, and authenticated by STR genotyping. Cell line G-361 was obtained from ECACC and confirmed mycoplasma negative and authenticated by STR genotyping. For each cell line, upon thawing, experiments were performed when required cell number was achieved to avoid long-term passaging of cells.

Xenografts were initiated with A375 cells maintained by serial subcutaneous transplantation in female athymic nude mice. Each test mouse received an A375 tumor fragment (1 mm³) implanted subcutaneously in the right flank. Once tumors reached target size (80–120 mm³), animals were randomized into treatment and control groups, and drug treatment was initiated. BVD-523 in 1% (w/v) carboxymethylcellulose (CMC) was administered orally twice daily at doses of 5, 25, 50, 100, or 150 mg/kg.

The efficacy of BVD-523 in combination with dabrafenib was evaluated in mice randomized into 9 groups of 15 and 1 group (Group 10) of 10. Dabrafenib was administered orally at 50 or 100 mg/kg once daily and BVD-523 was administered orally at 50 or 100 mg/kg twice daily, alone and in combination, until study end; vehicle-treated control groups were also included. Combination dosing was stopped on Day 20 to monitor for tumor regrowth. Animals were monitored individually and euthanized when each tumor reached an endpoint volume of 2,000 mm³ or on the final day (day 45), whichever came first, and median time to endpoint (TTE) calculated. The combination was also evaluated in an upstaged A375 model where larger tumors in the range 228–1,008 mm³ were evaluated. Here, mice were randomized into 1 group (Group 1) of 14 and 4 groups (Groups 2–5) of 20. Dosing was initiated on day 1 with dabrafenib plus BVD-523 (25 mg/kg dabrafenib + 50 mg/kg BVD-523 or 50 mg/kg dabrafenib + 100 mg/kg BVD-523), with each agent given orally twice daily until study end. The study included 50 mg/kg dabrafenib and 100 mg/kg BVD-523 monotherapy groups as well as a vehicle-treated control group. Tumors were measured twice weekly. Combination dosing was stopped on day 42 to monitor for tumor regrowth through study end (day 60). Treatment outcome was determined from % tumor growth delay (TGD), defined as the percent increase in median TTE for treated versus control mice, with differences between groups analyzed via log-rank survival analysis. For tumor growth inhibition (TGI) analysis, % TGI values were calculated and reported for each treatment (T) group versus the control (C). Mice were also monitored for complete regression (CR) and partial response (PR). Animals with a CR at the end of the study were additionally classified as tumor free survivors (TFS). All in vivo studies were conducted in accordance with generally recognized good laboratory practices, all procedures were in compliance with the Animal Welfare Act Regulations (9 CFR 3). All in vivo studies were reviewed and approved by the appropriate Institutional Animal Care and Use Committees (IACUC).

Human Colo205 cells were cultured in RPMI1640 supplemented with 10% (v/v) FBS, 100 U/mL penicillin, 100 μg/mL streptomycin (Invitrogen), and 2 mmol/L l-glutamine. Cells were cultured for <4 passages prior to implantation. Female athymic nude mice (19–23 g) were injected subcutaneously with 2 × 10⁶ Colo205 cells into the right dorsal axillary region on day 0.

Mice with an approximate tumor volume of 200 mm³ were randomized unblinded into 6 experimental groups. Vehicle control, 1% CMC (w/v), was prepared weekly. BVD-523 was suspended in 1% (w/v) CMC at the desired concentration and homogenized on ice at 6,500 rpm for 50 minutes. BVD-523 suspensions were prepared weekly and administered orally twice daily at total daily doses of 50, 100, 150, and 200 mg/kg (n = 12/group) on an 8- or 16-hour dosing schedule for 13 days. The vehicle control (n = 12) was administered using the same dosing regimen.

BVD-523 was evaluated in a patient-derived xenograft (ST052C) representing melanoma from a BRAF^V600E patient who had become clinically refractory to vemurafenib. Immunocompromised mice between 5–8 weeks of age were housed on irradiated corncob bedding (Teklad) in individual HEPA ventilated cages (Sealsafe Plus, Techniplast USA) on a 12-hour light–dark cycle at 70–74°F (21–23°C) and 40–60% humidity. Animals were fed water ad libitum (reverse osmosis, 2 ppm Cl₂) and an irradiated standard rodent diet (Teklad 2919) consisting of 19% protein, 9% fat, and 4% fiber. Tumor fragments were harvested from host animals and implanted into immunodeficient mice. The study was initiated at a mean tumor volume of approximately 170 mm³, whereby the animals were randomized unblinded into 4 groups, including a control [1% (v/v) CMC orally, twice daily × 31] and 3 treatment groups [BVD-523 (100 mg/kg), dabrafenib (50 mg/kg), or BVD-523/dabrafenib (100/50 mg/kg), n = 10/group]; All treatment drugs were administered orally on a twice daily × 31 schedule.

Following extensive optimization of leads originally identified using a high-throughput, small-molecule screen (29), a novel adenosine triphosphate (ATP)-competitive ERK1/2 inhibitor, BVD-523 (ulixertinib) was identified (Fig. 1A). BVD-523 is a potent ERK inhibitor with a K_i of 0.04 ± 0.02 nmol/L against ERK2. It was shown to be a reversible, competitive inhibitor of ATP, as the IC₅₀ values for ERK2 inhibition increased linearly with increasing ATP concentration, with dependence of enzyme rate on BVD-523 concentration (Fig. 1B and C). The IC₅₀ remained nearly constant for incubation times ≥10 minutes, suggesting rapid equilibrium and binding of BVD-523 with ERK2 (Fig. 1D). BVD-523 is also a tight-binding inhibitor of recombinant ERK1 (30), exhibiting a K_i of <0.3 nmol/L.

Binding of BVD-523 to ERK2 was demonstrated using calorimetric studies and compared with data generated using the ERK inhibitors SCH772984 (26) and pyrazolylpyrrole (31). All compounds bound and stabilized inactive ERK2 with increasing concentration, as indicated by positive ΔTm values (Fig. 1E). The 10- to 15-degree change in ΔTm observed with BVD-523 and SCH772984 is consistent with compounds that have low nanomolar binding affinities (32). BVD-523 demonstrated a strong binding affinity to both phosphorylated active ERK2 (pERK2) and inactive ERK2 (Fig. 1F), with a stronger affinity to pERK2 compared with inactive ERK2. BVD-523 did not interact with the negative control protein p38α MAP kinase (Fig. 1F).

BVD-523 demonstrated excellent ERK1/2 kinase selectivity based on biochemical counter-screens against 75 kinases, in addition to ERK1 and ERK2. The ATP concentrations were approximately equal to the K_m in all assays. Kinases inhibited to greater than 50% by 2 μmol/L BVD-523 were retested to generate K_i values (or apparent Ki; Fig. 1G). Twelve of the 14 kinases had a K_i of <1 μmol/L. The selectivity of BVD-523 for ERK2 was >7,000-fold for all kinases tested except ERK1, which was inhibited with a Ki of <0.3 nmol/L (10-fold). Therefore, BVD-523 is a highly potent and selective inhibitor of ERK1/2.

The antiproliferative impact of BVD-523 treatment on sensitive cancer cell lines was characterized by FACS analysis on BRAF^V600E-mutant melanoma cell line, UACC-62, following 24-hour treatment with BVD-523. BVD-523–treated cells were arrested in the G₁ phase of the cell cycle in a concentration-dependent manner (Fig. 2A; Supplementary Fig. S1A). The percentage of cells arrested in G₁ was also demonstrated to be time dependent, as shown by increasing proportion of KRAS^G12C-mutant pancreatic cell line MiaPaca-2 in G₁ following 24 and 48 hours treatment with BVD-523 (Supplementary Fig. S1B).

In addition, caspase-3/7 activity was analyzed as a measure of apoptosis in multiple human cancer cell lines (Fig. 2B). A cell-line–dependent increase in caspase-3/7 was observed following treatment with BVD-523 for 48 hours. BVD-523 treatment resulted in pronounced caspase-3/7 induction (greater than 3-fold) in a subset of cell lines with sensitivity to BVD-523 (<2 μmol/L IC₅₀).

To further characterize the mechanism of action and effects on signaling elicited by BVD-523, the levels of various effector and MAPK-related proteins were assessed in BVD-523–treated BRAF^V600E-mutant A375 melanoma cells (Fig. 2C). Phospho-ERK1/2 levels increased in a concentration-dependent manner after 4 and 24 hours of BVD-523 treatment. Despite prominent concentration-dependent increases in pERK1/2 observed with 2 μmol/L BVD-523 treatment, phosphorylation of the ERK1/2 target RSK1/2 was reduced at both 4 and 24 hours, which is consistent with sustained inhibition. Total protein levels of DUSP6, transcription of which is a target of ERK1/2, were also attenuated at 4 and 24 hours. Following 24 hours of treatment with BVD-523, the apoptotic marker BIM-EL increased in a dose-dependent manner, while cyclin D1 and pRB were attenuated at 2 μmol/L. All effects are consistent with on-target ERK1/2 inhibition.

To examine the effects of BVD-523 on signaling relative to other known ERK1/2 inhibitors (SCH772984, GDC-0994, and Vx-11e), a large-scale reverse-phase protein array (RPPA) of 39 proteins was employed in a variety of cell lines with sensitivity to ERK inhibition (Supplementary Fig. S2). Cell lines with common alterations in BRAF and RAS were assayed: BRAF^V600E-mutant lines A375, Colo205, and HT29; KRAS^G12C-mutant cell line MIAPACa-2; KRAS^G13D-mutant cell line HCT116; and AN3Ca with atypical HRAS^F82L mutation. Changes in protein levels are shown as a percentage change from dimethyl sulfoxide (DMSO)-treated control (Supplementary Fig. S2). All ERK inhibitors elicited qualitatively similar protein effects, with the exception of phosphorylation of ERK1/2 [pERK1/2 (ERK1/2 -T202, -Y204)]; SCH772984 inhibited pERK1/2 in all cell lines, while BVD-523, GDC-0994, and Vx-11e markedly increased pERK1/2 (Fig. 2D). Phospho-p90 RSK (pRSK1) and cyclin D1, which are proximal and distal targets of pERK1/2, respectively, were similarly inhibited by all inhibitors tested regardless of the degree of ERK1/2 phosphorylation (Fig. 2D). These independent findings for BVD-523 are consistent with studies showing that phosphorylation of ERK1/2 substrates RSK1/2 remained inhibited despite dramatically elevated pERK1/2 by Western blot analysis in A375 cells (Fig. 2C), in addition to protein-binding studies demonstrating BVD-523 binding and stabilization of pERK1/2 and inactive ERK1/2 (Fig. 1E and F). Therefore, measuring increased pERK1/2 levels could be considered as a pharmacodynamic biomarker for BVD-523, while quantifying inhibition of ERK1/2 targets such as pRSK or DUSP6 could serve this purpose more directly.

Additional protein changes are of note in this RPPA dataset (Supplementary Fig. S2). Decreased pS6-ribosomal protein is indicated to be another pharmacodynamic marker of ERK1/2 inhibition, as evidenced in all cell lines with all inhibitors (Supplementary Fig. S3A). Furthermore, prominent induction of pAKT appears to be a cell line–dependent observation, where each ERK1/2 inhibitor induced pAKT in cell lines A375 and AN3CA cells (Supplementary Fig. S3A). Interestingly, the degree of inhibition of survival marker pBAD appears to differ between compounds, with only modest inhibition of pBAD by GDC-0994 compared with the other ERK1/2 inhibitors tested (Supplementary Fig. S3B).

Next, we investigated how BVD-523 affects cellular localization of ERK1/2 and downstream target pRSK in a BRAF^V600E-mutant RKO colorectal cell line (Supplementary Fig. S3C). In resting cells, ERK1/2 localizes to the cytoplasm, and once stimulated, pERK1/2 migrates to target organelles, particularly the nucleus where transcriptional targets are activated (33). In DMSO-treated control cells, pERK1/2 is evident in both nuclear and cytoplasmic fractions, which is likely reflective of MAPK pathway activity due to the presence of BRAF^V600E in this cell line. Treatment with BVD-523 resulted in elevated pERK1/2 in the nucleus and cytoplasm, as well as a modest increase in nuclear total ERK1/2 compared with DMSO-treated cells, suggesting that compound-induced stabilization of pERK1/2 stimulates some nuclear translocation. Despite increased pERK1/2 in both compartments, pRSK levels are lower in the cytoplasmic and nuclear compartments compared with DMSO control. Comparator MAPK signaling inhibitors (i.e., trametinib, SCH772984, dabrafenib) inhibited phosphorylation of ERK1/2 and RSK, as reflected by lower levels in the nuclear and cytoplasmic compartments. These data again suggest that BVD-523–associated increases in pERK1/2 are evident in both the cytoplasm and nucleus; however, this does not translate to activation of target substrates. This is consistent with data presented in Fig. 2C and D.

On the basis of our in vitro findings that BVD-523 reduced proliferation and induced apoptosis in a concentration-dependent manner, BVD-523 was administered by oral gavage to demonstrate its in vivo antitumor activity in models with MAPK/ERK pathway dependency. Xenograft models of melanoma (cell line A375), and colorectal cancer (cell line Colo205), were utilized, both of which harbor a BRAF^V600E mutation.

In A375 cell line xenografts, 5, 25, 50, and 100 mg/kg twice daily doses of BVD-523 were compared following 18 days of treatment. For all doses, no significant body weight changes were observed (Supplementary Fig. S4A). BVD-523 demonstrated significant dose-dependent antitumor activity starting at 50 mg/kg twice daily (Fig. 3A). Doses of 50 and 100 mg/kg twice daily significantly attenuated tumor growth, with tumor growth inhibition (TGI) of 71% (P = 0.004) and 99% (P < 0.001), respectively. Seven partial regressions (PR) were noted in the 100 mg/kg twice daily group; no regression responses were noted in any other group.

BVD-523 demonstrated antitumor efficacy in a Colo205 human colorectal cancer cell line xenograft model (Fig. 3B). No significant body weight changes were observed (Supplementary Fig. S4B). BVD-523 again showed significant dose-dependent tumor volume regressions at doses of 50, 75, and 100 mg/kg twice daily, yielding mean tumor regressions T/T_i (T = end of treatment, T_i = treatment initiation) of −48.2%, −77.2%, and −92.3%, respectively (all P < 0.0001). Regression was not observed at the lowest dose of BVD-523 (25 mg/kg twice daily); however, significant tumor growth inhibition, with a T/C (T = Treatment, C = Control) of 25.2% (P < 0.0001), was observed.

In addition, BVD-523 resulted in dose-dependent antitumor activity in KRAS^G12C-mutant pancreatic cell line xenograft model, MIAPaCa2, (Supplementary Fig. S4C). Here, animals were dosed with 10, 25, 50, 75, and 100 mg/kg twice daily BVD-523 for 15 days. No significant body weight loss was observed in any treatment group (Supplementary Fig. S4D). At the highest dose administered for BVD-523 (100 mg/kg twice daily), significant tumor growth inhibition compared with vehicle-treated controls was observed, with a T/C of 5.3% on the last day of treatment (P < 0.0001). As a negative control, xenografts of mammary tumor cell line ZR-75-1 were also tested for BVD-523 antitumor activity. This model was not anticipated to respond to ERK inhibition, as it is not known to harbor alterations in the MAPK/ERK pathway. Indeed, at each dose of BVD-523 tested (10, 30, 60, 100 mg/kg twice daily), no antitumor activity was observed (Supplementary Fig. S4E), and no significant body weight loss was incurred (Supplementary Fig. S4F).

To establish the relationship between pharmacokinetics and pharmacodynamics, BVD-523 tumor concentrations were compared with DUSP6 mRNA and pERK1/2 protein levels over a 24-hour period following a single 100 mg/kg oral dose of BVD-523 (Fig. 3C and D). Phosphorylation of ERK1/2 was low in untreated tumors. Following treatment with BVD-523, ERK1/2 phosphorylation steadily increased from 1 hour postdose to maximal levels at 8 hours postdose, then returned to predose levels by 24 hours. This increase in pERK1/2 correlated with BVD-523 drug tumor concentrations. The in vivo observation of increased pERK1/2 with BVD-523 treatment is consistent with earlier in vitro findings (Fig. 2C and D). Conversely, DUSP6 expression decreased with increasing BVD-523 tumor concentration (Fig. 3D). Maximal DUSP6 inhibition was observed 3 hours post-BVD-523 dose (−96% change from baseline), and similar to pERK, returned to baseline by 24 hours.

Patients with BRAF-mutant cancer may acquire resistance to combined BRAF/MEK therapy (21), warranting consideration of other combination approaches within the MAPK pathway. We assessed the antiproliferative effects of combining BVD-523 with the BRAF inhibitors dabrafenib or vemurafenib in the BRAF^V600E-mutant melanoma cell lines G-361 and A375. As anticipated, single agents BVD-523, dabrafenib, and vemurafenib were each active, and modest synergy was observed with a BVD-523 plus BRAF inhibitor combination (Supplementary Fig. S5A and S5B). This indicated that BVD-523 combined with BRAF inhibitors were at least additive and potentially synergistic in melanoma cell lines carrying a BRAF^V600E mutation. Furthermore, generating acquired resistance in vitro following continuous culturing of BRAF^V600E-mutant cell line (A375) in BRAF inhibitor plus BVD-523 was challenging. In contrast, generating resistance to dabrafenib alone occurred relatively rapidly (Supplementary Fig. S5C). Even resistance to combined dabrafenib and trametinib emerged before dabrafenib plus BVD-523.

The benefit of combined BRAF and ERK inhibition may not be fully realized in in vitro combination studies where concentrations are not limited by tolerability. To understand the benefit of the combination, efficacy was assessed in vivo utilizing xenografts of the BRAF^V600E-mutant human melanoma cell line A375. Because of the response of combined dabrafenib and BVD-523 treatment, dosing in the combination groups was stopped on day 20 to monitor for tumor regrowth, and was reinitiated on day 42 (Fig. 4A). Tumors were measured twice weekly until the study was terminated on day 45. The median time to endpoint (TTE) for controls was 9.2 days, and the maximum possible tumor growth delay (TGD) of 35.8 days (end of study) was defined as 100%. Temozolomide treatment resulted in a TGD of 1.3 days (4%) and no regressions. The 50- and 100-mg/kg dabrafenib monotherapies produced TGDs of 6.9 days (19%) and 19.3 days (54%), respectively, a significant survival benefit (P < 0.001), and 1 PR in the 100 mg/kg group. The 100-mg/kg BVD-523 monotherapy resulted in a TGD of 9.3 days (26%), a significant survival benefit (P < 0.001), and 2 durable complete responses. The combinations of dabrafenib with BVD-523 each produced the maximum possible 100% TGD with statistically superior overall survival compared with their corresponding monotherapies (P < 0.001). The lowest dose combination produced a noteworthy 7/15 tumor-free survivors (TFS), and the 3 higher dosage combinations produced a total of 43/44 TFS, consistent with curative or near-curative activity (Fig. 4B). In summary, the combination of dabrafenib with BVD-523 produced a greater number of TFS and superior efficacy to either single agent.

On the basis of the activity of BVD-523 plus dabrafenib in A375 xenograft models with a starting tumor volume of approximately 75–144 mm³, a follow-up experiment was conducted to determine the efficacy of combination therapy in “upstaged” A375 xenografts (average tumor start volume, 700–800 mm³; Fig. 4C). The median TTE for controls was 6.2 days, establishing a maximum possible TGD of 53.8 days, which was defined as 100% TGD for the 60-day study. BVD-523 100 mg/kg monotherapy produced a negligible TGD (0.7 day, 1%) and no significant survival difference from controls (P > 0.05). The distribution of TTEs and 2 PRs suggested there might have been a subset of responders to treatment with BVD-523 alone. Dabrafenib 50-mg/kg monotherapy was efficacious, yielding a TGD of 46.2 days (86%) and a significant survival benefit compared with controls (P < 0.001). This group had 5 PRs and 5 CRs, including 3 TFS, among the 11 evaluable mice (Fig. 4D). Both combinations of dabrafenib with BVD-523 produced the maximum 100% TGD and a significant survival benefit compared with controls (P < 0.001). Each combination produced 100% regression responses among evaluable mice, though there were distinctions in regression activity. The 25 mg/kg dabrafenib and 50 mg/kg BVD-523 combination had 2 PRs and 8 CRs, with 6/10 TFS, whereas the 50-mg/kg dabrafenib and 100-mg/kg BVD-523 combination had 11/11 TFS on day 60 (Fig. 4D). Overall, these data support the rationale for frontline combination of BVD-523 with BRAF-targeted therapy in BRAF^V600-mutant melanoma, and this is likely to extend to other tumor types harboring this alteration.

Emergence of resistance to BRAF and MEK inhibitors limits their clinical efficacy. Here, we sought to model and compare the development of resistance to BRAF (dabrafenib), MEK (trametinib), and ERK1/2 (BVD-523) inhibition in vitro. Over several months, BRAF^V600E-mutant A375 cells were cultured in progressively increasing concentrations of each inhibitor. Drug-resistant A375 cell lines were readily obtained following growth in high concentrations of trametinib or dabrafenib, while developing cell lines with resistance to BVD-523 proved challenging (Fig. 5A). Overall, these in vitro data suggest that at concentrations yielding similar target inhibition, resistance to BVD-523 is delayed compared with dabrafenib or trametinib, and may translate to durable responses in the clinic.

Reactivation and dependence on ERK1/2 signaling is a common feature of acquired resistance to BRAF/MEK inhibition (26, 27); therefore, we evaluated the activity of BVD-523 in in vitro models of acquired resistance. First, a dabrafenib and trametinib combination-resistant A375 population was obtained using the increased concentration method (as described in detail in Supplementary Materials and Methods section of this manuscript). The IC₅₀-fold change from parental A375 for dabrafenib, trametinib, and BVD-523 in the BRAF/MEK combination-resistant population is shown in Fig. 5B. BVD-523 IC₅₀ was modestly shifted (2.5-fold), while dabrafenib and trametinib were more significantly shifted (8.5-fold and 13.5-fold, respectively; Fig. 5B). The cytotoxic agent paclitaxel was tested as a control with only a modest shift in potency observed. These data support the investigation of BVD-523 in the setting of BRAF/MEK therapy resistance, although the mechanism of resistance in this cell population remains to be characterized.

To further investigate the tractability of ERK1/2 inhibition in a model with a known mechanism of BRAF inhibitor resistance, we used AAV-mediated gene targeting to generate a pair of RKO BRAF^V600E-mutant cell lines isogenic for the presence or absence of an engineered heterozygous knock-in of MEK1^Q56P-activating mutation. MEK1/2 mutations, including MEK1^Q56P, have been implicated in both single-agent BRAF and combination BRAF/MEK targeting therapy-acquired resistance in patients (18, 21, 34–36). Single-agent assays demonstrated that relative to the parental BRAF^V600E::MEK1^wt cells, the double-mutant BRAF^V600E::MEK1^Q56P cells displayed a markedly reduced sensitivity to the BRAF inhibitors vemurafenib and dabrafenib and the MEK inhibitor trametinib (Fig. 5C). In contrast, response to BVD-523 was essentially identical in both the parental and MEK^Q56P-mutant cells, indicating that BVD-523 is not susceptible to this mechanism of acquired resistance. These results were confirmed in 2 independently derived double-mutant BRAF^V600E::MEK1^Q56P cell line clones, thus validating that results were specifically related to the presence of the MEK1^Q56P mutation rather than an unrelated clonal artifact. Similar results were also observed with a second mechanistically distinct ERK1/2 inhibitor (SCH772984), supporting the expectation that these observations are specifically related to mechanistic inhibition of ERK1/2 and not due to an off-target compound effect.

To further characterize the mechanistic effects of BVD-523 on MAPK pathway signaling in BRAF^V600E::MEK1^Q56P cell lines, protein levels were assessed by Western blot analysis (Fig. 5D). In the parental BRAF^V600E RKO cells, a reduced level of pRSK1/2 was observed following 4-hour treatment with BRAF (vemurafenib), MEK (trametinib), or ERK1/2 (BVD-523) inhibitors at pharmacologically active concentrations. In contrast, isogenic double-mutant BRAF^V600E::MEK1^Q56P cells did not exhibit reduced RSK phosphorylation following BRAF or MEK inhibitor treatment, while BVD-523 remained effective in inhibiting pRSK1/2 to a level comparable with parental RKO. Similarly, pRB is reduced, indicating G₀–G₁ arrest by 24 hours of BVD-523 treatment in both parental RKO and BRAF^V600E::MEK1^Q56P. Notably, as introduction of the MEK-Q56P mutation alone increased the level of pERK compared with the RKO parental cell line, BVD-523 treatment did not appear to significantly further boost pERK levels in the engineered cells.

Acquired KRAS mutations are also known drivers of resistance to MAPK pathway inhibitors. To understand the susceptibility of BVD-523 to this mechanism of resistance, an isogenic panel of clinically relevant KRAS mutations in colorectal cell line SW48 was utilized. Sensitivity to BVD-523 was compared with MEK inhibitors selumetinib (37) and trametinib (Fig. 5E and F). Sensitivity to paclitaxel was unaltered (Supplementary Fig. S6). While several mutant KRAS alleles conferred robust to intermediate levels of resistance to MEK inhibition, sensitivity to BVD-523 was unaltered by the majority of alleles, and where a shift in sensitivity was observed, it was not to the extent observed with trametinib or selumetinib. Overall, these data suggest that BVD-523 is more efficacious in this context than MEK inhibitors.

To confirm and extend the antitumor effects of BVD-523 observed in in vitro models of BRAF-/MEK-acquired resistance, we utilized a BRAF-resistant xenograft model derived from a patient with resistance to vemurafenib. BVD-523 was dosed by oral gavage at 100 mg/kg twice daily for 28 days, both alone and in combination with dabrafenib at 50 mg/kg twice daily (Fig. 6). With all regimens, no significant change in body weight was observed (Supplementary Fig. S7). As expected, minimal antitumor activity was demonstrated for single-agent dabrafenib (22% TGI). BVD-523 activity was significant compared with vehicle control (P ≤ 0.05), with a TGI of 78%. In this model, combining BVD-523 with dabrafenib resulted in a TGI of 76% (P ≤ 0.05); therefore, further benefit was not gained for the combination compared with single-agent BVD-523 in this particular model of BRAF-acquired resistance. Whole-exome sequencing was performed on both the patient biopsy and the PDX model. A number of alterations with unknown significance were identified in addition to the known driver mutation PIK3CA-H1047L. We speculate that the presence of this mutation may actually attenuate response to BVD-523, as stasis rather than regression was observed.

BVD-523 is a potent, highly selective, reversible, small-molecule ATP-competitive inhibitor of ERK1/2 with activity in in vitro and in vivo cancer models. In vitro, BVD-523 demonstrated potent inhibition against several human tumor cell lines, particularly those harboring activating mutations in the MAPK signaling pathway, consistent with its mechanism of action. BVD-523 elicited changes in downstream target and effector proteins, including inhibition of direct substrate of ERK1/2, pRSK, and total DUSP6 protein levels (transcriptional target of ERK1/2). These findings are in line with those of previous studies of other ERK1/2 inhibitors, which demonstrated effective suppression of pRSK with ERK1/2 inhibition (26, 27). Interestingly, BVD-523 treatment resulted in a marked increase in ERK1/2 phosphorylation in vitro and in vivo. Similar to our findings, an increase in pERK1/2 has been reported with the ERK1/2 inhibitor Vx11e; conversely, pERK1/2 increases are not observed with SCH772984 (26). Although differences in pERK1/2 levels were observed among the various ERK1/2 inhibitors tested, downstream effectors (i.e., pRSK1 and total DUSP6) were similarly inhibited. These findings suggest quantifying ERK1/2 target substrates, such as pRSK1, may serve as reliable pharmacodynamic biomarkers for BVD-523–mediated inhibition of ERK1/2 activity. The differential effect of BVD-523 and SCH772984 upon the levels of pERK is striking. The significance of this to efficacy and tolerability in the clinic remains to be determined.

While BRAF (dabrafenib, vemurafenib) and MEK (trametinib, cobimetinib) inhibitors validate the MAPK pathway as a therapeutic target, particularly in patients with BRAF^V600 mutations, the antitumor response is limited by the emergence of acquired resistance and subsequent disease progression (38). Reactivation of the ERK1/2 pathway is one common consequence of acquired resistance mechanism. When introduced into the BRAF^V600E-mutant colorectal cell line RKO, MEK^Q56P conferred resistance to MEK and BRAF inhibition. In contrast, BVD-523 retained its potent inhibitory activity in the engineered MEK^Q56P cell line, indicating that ERK1/2 inhibition is effective in the setting of upstream activating alterations, which can arise in response to BRAF/MEK treatment. As further evidence of a role for BVD-523 in the context of acquired resistance, efficacy of BVD-523 was evident in a xenograft model derived from a patient whose disease progressed on vemurafenib; the BRAF inhibitor dabrafenib was not effective in this model. These data support a role for targeting ERK1/2 in the setting of BRAF/MEK resistance, and complement previously published findings (26, 27).

To further characterize resistance to inhibitors of the MAPK pathway, the emergence of resistance to BVD-523 itself was investigated. We found that single-agent treatment of cancer cells with BVD-523 was durable, and it was more challenging to develop resistance against BVD-523 than other agents targeting upstream MAPK signaling components (i.e., dabrafenib, trametinib). This may suggest that acquiring resistance to ERK1/2-targeting agents is harder to achieve than acquiring resistance to BRAF or MEK therapy. However, in vitro studies with other ERK1/2 inhibitors have identified specific mutants in ERK1/2 that drive resistance (39, 40); these specific mutations have yet to be identified in clinical samples from ERK1/2 inhibitor–relapsed patients.

The potential clinical benefit of ERK1/2 inhibition with BVD-523 extends beyond the setting of BRAF/MEK therapy–resistant patients. As ERK1/2 is a downstream master node within this MAPK pathway, its inhibition is attractive in numerous cancer settings where tumor growth depends on MAPK signaling. Approximately 30% of all cancers harbor RAS mutations; therefore, targeting downstream ERK1/2 with BVD-523 is a rational treatment approach for these cancers. Furthermore, results from a study by Hayes and colleagues indicate that prolonged ERK1/2 inhibition in KRAS-mutant pancreatic cancer is associated with senescent-like growth suppression (41). However, a combination approach may be required for maximal and durable attenuation of MAPK signaling in the setting of RAS mutations. For example, MEK inhibition in KRAS-mutant colorectal cancer cell results in an adaptive response of ErbB family activation, which dampens the response to MEK inhibition (42). Similar context-specific adaptive responses may occur following ERK1/2 inhibition with BVD-523. The optimal treatment combinations for various genetic profiles and cancer histologies are the subject of ongoing research. In addition to BRAF and RAS mutations, other alterations which drive MAPK are emerging. For example, novel RAF fusions and atypical (non-V600) BRAF mutations which promote RAF dimerization activate the MAPK pathway (43). BRAF inhibitors such as vemurafenib and dabrafenib which inhibit BRAF^V600-mutant monomer proteins have been shown to be inactive in atypical RAF alterations which drive MAPK signaling in a dimerization-dependent manner (43). However, treatment with BVD-523 to target downstream ERK1/2 in these tumors may be a novel approach to addressing this unmet medical need; indeed patients harboring such atypical BRAF alterations are included in an ongoing clinical trial of BVD-523 (NCT01781429).

In the setting of BRAF^V600-mutant melanoma tumors, combined BRAF and MEK inhibition exemplifies how agents targeting different nodes of the same pathway can improve treatment response and duration. Our combination studies in BRAF^V600E-mutant xenografts of human melanoma cell line A375 provide support for combination therapy with BVD-523 and BRAF inhibitors. The combination demonstrated superior benefit relative to single-agent treatments, including results consistent with curative responses. The clinical efficacy and tolerability of combined BRAF/BVD-523 therapy remains to be determined. It would not be unreasonable to expect that a BRAF/BVD-523 combination will at least be comparable in efficacy to a targeted BRAF/MEK combination. Furthermore, the in vitro observation that acquired resistance to BVD-523 is more challenging to achieve compared with other MAPK pathway inhibitors suggests that the BRAF/BVD-523 combination has the potential to provide a more durable response.

It would be remiss to not highlight that significant progress has also been made using immunotherapy for the treatment of melanoma. The FDA has approved various immune checkpoint inhibitors for the treatment of advanced melanoma, including the cytotoxic T-lymphocyte antigen-4 targeted agent ipilimumab, and the programmed cell death protein-1 (PD-1) inhibitors pembrolizumab and nivolumab. Emerging data suggest MEK inhibitors may have benefit in combination with anti-PD-L1 therapy atezolizumab (44). Combining BVD-523 with such immunotherapies is an attractive therapeutic option; further investigation is warranted to explore dosing schedules and to assess whether synergistic response can be achieved.

On the basis of the preclinical data, BVD-523 may hold promise for treatment of patients with malignancies dependent on MAPK signaling, including those whose tumors have acquired resistance to other treatments. Phase I clinical studies of BVD-523 are ongoing (NCT01781429, NCT02296242, and NCT02608229).

U.A. Germann has ownership interest (including patents) in Vertex Pharmaceuticals Incorporated. J.J. Roix is an associate at BioMed Valley Discoveries. D.A. Sorrell has ownership interest (including patents) in Horizon Discovery. M. Fitzgibbon is a senior research scientist and has ownership interest (including patents) in Vertex Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: U.A. Germann, W. Markland, R.R. Hoover, A.M. Aronov, J.J. Roix, M. Hale, R. Samadani, G. DeCrescenzo, S. Saha, D.J. Welsch

Development of methodology: B.F. Furey, W. Markland, R.R. Hoover, J.J. Roix, D.A. Sorrell, R. Samadani, G. DeCrescenzo, M. Namchuk, S. Saha, D.J. Welsch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.F. Furey, W. Markland, R.R. Hoover, D.M. Boucher, D.A. Sorrell, M. Fitzgibbon, P. Shapiro, M.J. Wick, R. Samadani, K. Meshaw, C.M. Emery, S. Saha, D.J. Welsch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): U.A. Germann, B.F. Furey, W. Markland, R.R. Hoover, J.J. Roix, D.M. Boucher, D.A. Sorrell, P. Shapiro, R. Samadani, A. Groover, G. DeCrescenzo, M. Namchuk, C.M. Emery, S. Saha, D.J. Welsch

Writing, review, and/or revision of the manuscript: U.A. Germann, B.F. Furey, W. Markland, A.M. Aronov, J.J. Roix, M. Hale, D.M. Boucher, D.A. Sorrell, G. Martinez-Botella, M.J. Wick, R. Samadani, A. Groover, G. DeCrescenzo, M. Namchuk, C.M. Emery, S. Saha, D.J. Welsch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.J. Roix, M. Fitzgibbon, A. Groover, C.M. Emery, S. Saha, D.J. Welsch

Study supervision: U.A. Germann, J.J. Roix, D.A. Sorrell, M. Namchuk, C.M. Emery, S. Saha, D.J. Welsch

Other (his group while at Vertex designed and made the compound in the study and holds patents on the chemical matter): M. Hale

The authors thank Corinne Ramos and Nicholas Hoke for performing the large-scale reverse phase protein array. They also thank Sharon Maguire and Paul A. Russell for assay support and Amy Smith and Mark Stockdale for the construction of the MEK1^Q56P cell line.

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.