Response and Resistance to Paradox-Breaking BRAF Inhibitor in Melanomas In Vivo and Ex Vivo

Melanoma is the most aggressive form of cutaneous malignancy with a short time to metastasis and high mortality rate. Enhanced MEK–ERK1/2 signaling occurs in most, if not all, cutaneous melanomas and is frequently activated by a valine to glutamic acid mutation at residue 600 (V600E) in the v-Raf murine sarcoma viral oncogene homolog B (BRAF) protein (1). Recent targeted therapies have focused on selectively targeting BRAF V600E in mutant BRAF-harboring melanomas. Selective BRAF inhibitors, vemurafenib (PLX4032) and dabrafenib (GSK'436), have high response rates and provide remarkable improvements in patients with mutant BRAF melanoma; however, the majority of patients develop resistance within one year (2, 3). In addition, a frequent side effect of vemurafenib and dabrafenib is the induction of squamous cell carcinomas (SCC) and keratoacanthomas, which generally require surgical removal (4, 5). Although vemurafenib favors the V600E form of BRAF (6), its binding to wild-type (WT) BRAF induces heterodimerization with CRAF and ERK1/2 activation (7, 8). This "paradoxical activation" of ERK1/2 likely mediates vemurafenib induction of SCCs and keratoacanthomas (4, 5), leukemia (9, 10), and mutant KRAS pancreatic adenocarcinoma (11). Vertical targeting of the ERK1/2 pathway in melanoma with BRAF plus MEK inhibitor combinations achieves a 64% to 76% response rate, extends median progression-free survival to over 9 months, and reduces the adverse events associated with paradoxical ERK1/2 activation (12–14). However, the BRAF plus MEK inhibitor combination does not prevent relapse and can cause significant toxicities that may result in treatment discontinuation (15). Checkpoint inhibitor agents, such as ipilimumab, nivolumab, and pembrolizumab, act to relieve immunosuppressive signals and often elicit durable responses; however, they do not elicit response rates as high as targeted small-molecule inhibitors (11%–57.6% vs. 48%–69.6%, respectively; ref. 16). Furthermore, immunotherapy approaches are generally not suitable for patients with bulky disease that require rapid intervention (16), and an initial clinical trial combining vemurafenib with ipilimumab (a CTLA-4–targeting agent) resulted in significant hepatotoxicity (17).

New targeted therapies that efficiently inhibit the ERK1/2 pathway with fewer and less serious side effects would be clinically beneficial. Recently, next-generation mutant BRAF inhibitors have been designed that elicit strong efficacy in mutant BRAF melanoma cells but do not elicit paradoxical ERK1/2 activation in mutant RAS–expressing keratinocytes (18–24). Further examination of PLX8394 as a targeted agent is warranted as this agent enters clinical trials, as it may elicit fewer high-grade toxicities than previous generations of mutant-selective BRAF inhibitors and the combination of BRAF plus MEK inhibitors. Targeted inhibitors produce heterogeneous effects in mutant BRAF patients due to intrinsic mechanisms of resistance and adaptive drug responses. There is an important need for targeted agents to be tested in a personalized manner. Patient-derived xenograft models have been developed but typically take several months to be propagated in mice (25–27). Here, we describe the use of a patient-derived melanoma biopsy explant system (PDeX) and in vivo ERK1/2 reporter models to show that PLX8394 is a potent BRAF inhibitor and does not elicit paradoxical activation of ERK1/2 in vivo and ex vivo.

Vehicle-treated and PLX8394-resistant tumors were harvested for genomic DNA with Wizard Genomic DNA Purification Kit (Promega). Samples were barcoded and sequenced using the Ion PGM 200 Sequencing Kit (Life Technologies). For full details, see Supplementary Data.

Western blot analysis was performed as in ref. 28 with volumetric analysis in Quantity One (Bio-Rad). Antibodies were purchased from Cell Signaling Technology, Santa Cruz Biotechnology Inc., Biosciences Inc., Enzo, and Sigma-Aldrich Co. For full details, see Supplementary Data.

Vemurafenib, dabrafenib, and trametinib (GSK'212) were purchased from Selleck Chemicals LLC. PLX8394 was provided by Dr. Gideon Bollag (Plexxikon Inc.). PLX8394 for in vivo experiments was sent to Research Diets Inc. for the production of chow.

1205LuTR GAL4-ELK1 reporter cells [modified cell line – the parental was a gift from Dr. Meenhard Herlyn (2005, Wistar Institute, Philadelphia, PA)], PRT #3 (26), Paradox breaker resistant tumor (PBRT) #15, and #16 cells [in vivo derived resistant cells of 1205LuTR GAL4-ELK1 (2013)] were grown in MCDB 153 medium containing 20% Leibovitz-L15 medium, 2% FBS, 0.2% sodium bicarbonate, and 5 μg/mL insulin. In addition, PRT #3 cells were cultured in 1 μmol/L PLX4720, and PBRT #15 and #16 cells were cultured in 0.5 μmol/L PLX8394. BOWES cells (2013) were grown in MEM containing 10% FBS, 1% nonessential amino acids, 1% sodium pyruvate, and 1% HEPES buffer. B6, MeWo, [gifts from Dr. Barbara Bedogni (2013, Case Western Reserve University, Cleveland, OH)], and CHL-1 cells (purchased from ATCC in 2013) were cultured in DMEM with 10% FBS. Penicillin/streptomycin (1%) was added to all media. All cells were grown at 37°C in a humidified incubator supplemented with 5% CO₂. Cells are routinely assayed for mycoplasma contamination with MycoScope Kit (Genlantis). Cells were assayed in April, May, and September 2016. Cell line authentication via STR analysis was completed in April 2015 for BOWES, MeWo, B6, and CHL-1, and in February 2017 for 1205LuTR GAL4-ELK1 reporter cells and PBRTs. B6 cells produced a unique profile, whereas all other cells matched to known profiles.

Tissue was fixed in formalin and paraffin embedded. Sections were stained for ERK1/2 phosphorylation (Thr202/Tyr204, #4370, Cell Signaling Technology). Staining was scored using the digital Aperio ScanScope GL system in a blinded fashion by a pathologist (A. Goldberg).

Cells (1.4 × 10⁴) were seeded in individual wells of 6-well plates in regular culture medium (containing 0.5 μmol/L PLX8394 for PBRTs). The next day, plates were washed and medium was replaced with medium supplemented with drugs of interest. Medium and drugs were changed every 2 days. After 9 days, cells were fixed in buffered formalin with 0.2% crystal violet. Plates were then scanned for quantitation via ImageJ.

Cells (2 × 10³) were seeded in triplicate in wells of a 96-well plate in regular culture medium (containing 0.5 μmol/L PLX8394 for PBRTs). On the next day, cells were washed twice with PBS, and drug-laced media were added. After 4 days (including one medium change), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reagent (Sigma-Aldrich Co.) was added for 3 hours. Solubilized formazan was analyzed at 450 nmol/L in a Multiskan Spectrum spectrophotometer (Thermo Fisher Scientific). Results are normalized to DMSO conditions and are a composite of three independent experiments.

Unless noted otherwise, significant values (indicated by an asterisk) were considered to have P ≤0.05 as determined by a two-tailed Student t test assuming unequal variance and error bars are ± SEM. The effects of drug treatment on BRAF homodimers was modeled by considering the treatment and experimental replicate (n = 4) as predictors of log(Myc/FLAG). ANOVA analysis was then performed with these considerations. IC₅₀ calculations for ERK1/2 phosphorylation were performed using GraphPad Prism.

Cells (2.0 × 10⁵) were seeded in 6-well plates. Cells were treated with drug of interest for 48 hours. The thymidine analogue, EdU, was added at a final concentration of 10 μmol/L for the final 16 hours. EdU incorporation was measured using the Click-it EdU Alexa Flour 647 Flow Cytometry Assay Kit and was utilized as per the manufacturer's instructions (Molecular Probes). EdU staining was quantified on BD FACSCalibur, and data were analyzed with FlowJo software. Data points are shown as averages of three experimental replicates.

Tumors were collected following informed patient consent at Thomas Jefferson University Hospital (Philadelphia, PA) under an IRB-approved protocol (#10D.341). Less than 16 hours postsurgery, excess adipose and stromal tissue was removed and tumors were cut into 1-mm³ pieces. Vetspon absorbable hemostatic gelatin 1-cm³ sponges (Novartis) were presoaked in 12-well plates for 15 minutes at 37°C in 500 μL of DMEM/10% FBS containing drugs or DMSO as a vehicle control. To avoid concerns of intratumoral heterogeneity, up to three 1-mm³ pieces from different locations of the original tumor were placed per sponge per treatment condition. Similarly, xenograft tumors were dissected into 1-mm³ pieces and placed on medium/drug-soaked sponges. Medium was replaced every 24 hours. Tumor pieces for Western blotting were homogenized in modified RPPA lysis buffer (29) with phosphatase and inhibitors (PhosSTOP and cOmplete tablets Roche). Laemmli sample buffer was added, and samples were heated at ≥95°C for 5 minutes. For IHC analysis, tumor pieces were fixed in formalin for 24 hours. Two of the samples (TJU-MEL-27A and TJU-MEL-27B) were different lesions from the same patient, and combination treatment was not assayed for TJU-MEL-30.

Seven-week-old female nude mice (The Jackson Laboratory, stock# 007850) were injected with 1 × 10⁶ 1205LuTR GAL4-ELK1 reporter cells. Tumors were allowed to form to approximately 100 mm³, at which point the mice were randomly divided into 2 cohorts and fed either vehicle or PLX8394-laced chow. Tumor volumes and ERK1/2 reporter activity via firefly luciferase measurements were recorded every 3 to 4 days. All mouse experiments were performed at Thomas Jefferson University (Association for Assessment and Accreditation of Laboratory Animal Care-accredited) and approved by the Institutional Animal Care and Use Committee. For full details, see Supplementary Data.

1205LuTR GAL4-ELK1 parental reporter cells and PB-resistant tumor (PBRT) #15 and #16 cells (2.5 × 10⁵/per condition) were seeded in 6-well plates in normal growth media (containing 500 nmol/L PLX8394 for PBRTs). The next day, cells were treated with either DMSO or 0.5 μmol/L PLX8394 for 24 hours. Lysates from three independent experiments were processed and analyzed as described previously (29), producing triplicates for each. For details of the analysis, see Supplementary Data section.

1205LuTR cells expressing both Myc and FLAG-tagged BRAF V600E ΔEx 2–8 were seeded (1.0 × 10⁶) on 10-cm plates overnight. Cells were then dosed with 100 ng/mL doxycycline to induce both splice variant expression for 48 hours. Plates were treated with DMSO, PLX4720, or PLX8394 for an additional 4 hours. Cells were PBS washed and lysed in an NP40-based lysis buffer. Twenty microliters of prewashed anti-FLAG Affinity Gel (#A2220 Sigma-Aldrich) was used to immunoprecipitate FLAG-tagged target during an overnight incubation at 4°C. The affinity gel was then washed 3 times with cold TBS, resuspended in Laemmli sample buffer, and boiled for 5 minutes. Equal volume was loaded on acrylamide gels for Western blot analysis.

PLX8394 is a mutant BRAF-selective inhibitor that potently blocks ERK1/2 signaling in BRAF V600E/D–harboring melanoma cells in vitro (18, 19, 24). The structure of PLX8394 has been published previously (18). As an initial assessment of the cellular response to PLX8394, we performed reverse-phase protein array (RPPA) analysis on 1205LuTR GAL4-ELK1 reporter cells (Fig. 1A; ref. 28). RPPA allows for quantitative assessment of >200 targets involved in growth factor signaling, cell-cycle progression, apoptosis, and histone modification (29). To allow for cell-cycle and apoptotic changes to take place, we assayed samples at a 24-hour time point compared with an acute time point, which would most likely only affect signaling. PLX8394 treatment significantly (P < 0.05 and fold change ≥1.5) altered 13 targets, including downregulation of phosphorylated MEK and ERK1/2, upregulation of the proapoptotic protein BIM, and downregulation of cyclin B1 (Fig. 1B). We also observed upregulation of the growth factor receptor, ERBB3, consistent with previous findings with vemurafenib (30, 31). Phosphorylation of Raf-1/CRAF and Src, which are implicated in paradoxical ERK1/2 signaling and are suppressed by pan-RAF inhibitors, remained unaffected by PLX8394 treatment (Supplementary Fig. S1). To quantitatively measure the effects of PLX8394 in vivo, we utilized xenografts from BRAF V600E melanoma cells expressing an ERK1/2 luciferase-based reporter. This model permits quantitative and temporal analysis in a noninvasive manner (28). ERK1/2 reporter luciferase levels (adjusted for tumor volume) were significantly reduced within 7 days of PLX8394 treatment compared with vehicle controls (Fig. 1C and D). PLX8394 also significantly reduced tumor growth compared with vehicle-treated mice (Fig. 1E). Together, these results show that PLX8394 inhibits ERK1/2 signaling in vitro, in vivo, and reduces tumor growth in mutant BRAF melanoma xenografts.

An ex vivo explant model has been previously utilized in prostate cancer (32, 33). These systems are advantageous for preclinical testing as they contain a stromal component and, thus, more closely mimic the tumor microenvironment. We established and validated this model in melanoma, using explants derived from xenograft tumors of 1205LuTR cells (partially sensitive to PLX4720, the tool compound for vemurafenib) and 1205LuTR-PRT #3 cells, which express a BRAF V600E splice variant and are resistant to PLX4720 (28, 34). Tumor tissue was treated ex vivo with vemurafenib at 1 μmol/L, a standard concentration for in vitro experiments (6, 7, 34) and 3D melanoma systems (35) or with 1 μmol/L dabrafenib/16 nmol/L trametinib combination (Supplementary Fig. S2A–S2D). The dabrafenib and trametinib combination at the given concentration is a clinically relevant molar ratio of the two drugs that was found to have significant effect on downstream signaling (Supplementary Fig. S2F); this treatment served as a positive control for ERK1/2 pathway suppression to demonstrate the range of response in the ex vivo explant system. These results provided proof of concept for the patient-derived explant (PDeX) system. Next, we extended PDeX analysis to fresh human melanomas. Sequence-validated, mutant BRAF V600E melanoma biopsy explants (Supplementary Fig. S3A; Table S1) were treated either with vemurafenib, dabrafenib/trametinib combination, or PLX8394 for 48 to 72 hours. By IHC staining of paraffin-embedded tumor sections and quantitative analyses, vemurafenib inhibited ERK1/2 phosphorylation, but inhibition was partial and variable (Fig. 2A and B). This observation is consistent with others who report a modest response to vemurafenib in 3D tumor systems (35), and in stroma/melanoma coculture settings (36–38). In contrast, both combination and PLX8394 treatment consistently and significantly inhibited ERK1/2 phosphorylation in PDeX (Fig. 2A and B). By Western blot analysis, vemurafenib inhibition of ERK1/2 phosphorylation was again variable, but statistically significant compared with vehicle treatment (Fig. 2C and D; Supplementary Fig. S3C). Importantly, both PLX8394 and the dabrafenib/trametinib combination inhibited ERK1/2 phosphorylation by >60% (Fig. 2B and D). Furthermore, PARP cleavage is associated with ERK1/2 inhibition following combination and PLX8394 treatments (Fig. 2C; Supplementary Fig. S3). To better understand pathway alterations, RPPA analysis was performed on BRAF V600E human melanoma PDeX treated with targeted inhibitors. Pathway analysis of the programmed cell death/apoptosis and cell-cycle arrest Gene Ontology pathways by gene set enrichment analysis showed that vemurafenib predominantly enriched a cell-cycle arrest response, whereas PLX8394 and combination treatment induced an apoptotic/cell death response (Fig. 2E; Supplementary Fig. S3F). In a tumor with sufficient sample to assay the effects of a dose response to PLX8394, we observed a dose-dependent increase of PARP cleavage, as well as suppression of ERK1/2 phosphorylation. RPPA analysis of this sample demonstrated a dose-dependent decrease of ERK1/2 pathway targets, and increase in the proapoptotic protein BIM (Supplementary Fig. S3D and S3E). Taken together, these data suggest that PLX8394 is a potent inhibitor of ERK1/2 phosphorylation in human BRAF V600E melanomas and elicits effects comparable with the current FDA-approved dabrafenib/trametinib combination in an explant model.

To determine the efficacy of PLX8394 compared with vemurafenib in the explant system, xenografts were generated with 1205LuTR GAL4-ELK1 parental cells and two different in vivo derived RAF inhibitor resistant lines, PRT #3 and PRT #4 (28). Xenograft tumors were excised and used in the explant system to assay dose responses. After 48 hours of treatment, Western blotting was used to measure ERK1/2 phosphorylation and PARP cleavage. We found that PLX8394 more efficiently suppressed ERK1/2 phosphorylation (IC₅₀ 0.01 μmol/L vs. 1.39 μmol/L) and elicited PARP cleavage compared with vemurafenib in parental 1205LuTR (Fig. 3A; Supplementary Fig. S4A). Importantly, although vemurafenib treatment was largely ineffective (ERK1/2 phosphorylation IC₅₀ is undefined for PRT #3 and 4.05 mmol/L for PRT #4), PLX8394 inhibited ERK1/2 phosphorylation (0.97 μmol/L and 0.096 μmol/L for PRT #3 and #4, respectively) and induced PARP cleavage in BRAF splice variant–expressing tumors, PRT #3 and PRT #4 (Fig. 3B and C; Supplementary Fig. S4B and S4C). It is noteworthy that although the PRT tumors were sensitive to PLX8394, both required a higher dose of PLX8394 than parental cells to suppress ERK1/2 phosphorylation. This result is consistent with other RAF inhibitor–resistant cells treated with potential second-line RAS–RAF–MEK–ERK pathway targeting agents (34, 39). As constitutive BRAF splice variant homodimerization has been linked to vemurafenib resistance (39), we investigated whether PLX8394 affects homodimerization of BRAF splice variants. 1205Lu cell lines were created to inducibly coexpress both Myc-tagged and FLAG-tagged versions of BRAF splice variants lacking exons two through eight (1205LuTR FLAG/Myc BRAF ΔEx 2–8). BRAF ΔEx 2–8 is equivalent to the BRAF splice variant expressed in PRT #4. Similar to the PRT tumors, this cell line demonstrated a dose-dependent reduction of ERK1/2 pathway signaling from PLX8394 but not PLX4720 treatment (Fig. 3D). Parallel lysates were then used to query homodimerization by immunoprecipitation of the FLAG-tagged BRAF splice variant and probing for the association of its Myc-tagged binding partner. Although both drugs impaired homodimerization, PLX8394 treatment elicited a more profound reduction than PLX4720 (Fig. 3E and F). Interestingly, PLX8394 dosing had little effect on homodimerization status (Supplementary Fig. S4D), yet higher doses of PLX8394 were able to inhibit MEK phosphorylation. These observations suggest that although PLX8394 blocks dimerization and ERK1/2 pathway more efficiently than PLX4720, the extent of BRAF splice variant homodimerization itself may not be wholly responsible for vemurafenib resistance. Together, these data demonstrate superior efficacy of PLX8394 as a single-agent RAF inhibitor in comparison with vemurafenib and that PLX8394 can overcome mutant BRAF splice variants, a common mechanism of BRAF inhibitor resistance.

An important goal in the design of PLX8394 is to reduce hyperactivation of ERK1/2 in WT BRAF-containing tissues. To demonstrate the “paradox-breaking” ability of PLX8394, we treated WT BRAF/WT NRAS (WT/WT) melanoma cell lines with vemurafenib, PLX8394, and trametinib in 2D culture conditions. Vemurafenib significantly increased ERK1/2 signaling in CHL-1, BOWES, MeWo, and B6 WT/WT melanoma cell lines compared with DMSO control (Fig. 4A and B). In contrast, PLX8394 treatment did not produce a statistically significant increase in paradoxical activation, and treatment with trametinib strongly reduced ERK1/2 phosphorylation. Extending these studies into explants derived from xenografts from the WT/WT BOWES and B6 cells, vemurafenib enhanced ERK1/2 activation in WT/WT xenograft explants similar to experiments in 2D (Fig. 4C). In comparison with vemurafenib, PLX8394 did not induce strong paradoxical ERK1/2 activation in these samples. Furthermore, we tested paradoxical activation by RAF inhibitors using the PDeX system in a WT/WT patient sample (Supplementary Fig. S3B). Western blot analysis demonstrated a strong paradoxical phosphorylation of ERK1/2 induced by vemurafenib, but comparatively weak ERK1/2 phosphorylation in response to PLX8394 (Fig. 4D). As expected, the MEK inhibitor trametinib suppressed ERK1/2 signaling (Fig. 4D). Taken together, these data show that when using doses effective in suppressing ERK1/2 signaling in mutant BRAF tumor, PLX8394 does not elicit strong paradoxical signaling in WT BRAF tissue, representing an improvement over the previous generation of BRAF inhibitors.

Treatment with targeted therapies is invariably associated with acquired resistance; therefore, we investigated the duration of PLX8394 effects on BRAF V600E melanomas in vivo. Mice bearing mutant BRAF xenografts were continued on PLX8394 treatment until progression (≥1,000 mm³ tumor size or displayed signs of ulceration). Progressing tumors were excised and two PBRT cell lines, #15 and #16, were propagated. PBRT #15 and PBRT #16 were isolated at days 45 and 35 after drug treatment, respectively. In 2D colony formation assays, PLX8394 suppressed the growth of 1205LuTR parental cells in a dose-dependent manner (Fig. 5A and B). Conversely, PBRT #15 maintained growth in PLX8394 and PBRT #16 exhibited addiction to PLX8394, similar to a phenomenon observed in vemurafenib-resistant cells (34). In MTT assays, PLX8394 potently inhibited the viability of parental cells, but both PBRT #15 and #16 cell lines were highly resistant to the inhibitor (Supplementary Fig. S5A).

To understand pathway alterations associated with resistance to PLX8394, we performed RPPA analysis on PBRT #15 and #16 compared with parental 1205LuTR cells. Compared with parental 1205LuTR cells, the levels of 34 proteins significantly changed (1.5 fold; P ≤ 0.01) in either the DMSO control or PLX8394 conditions (Fig. 5C). Western blotting confirmed the RPPA results showing maintenance of ERK1/2 phosphorylation in PBRT #15 and #16 cells treated with PLX8394 (Supplementary Fig. S5B and S5C). In addition, PBRT cells treated with PLX8394 exhibited significantly higher levels of Rb phosphorylation than PLX8394-treated parental cells (Supplementary Fig. S5D), reflecting the ability of PBRT cells to overcome PLX8394-mediated cell-cycle inhibition. Interestingly, both PBRT cell lines displayed increased AKT phosphorylation, enhanced PDGFR levels, and reduced β-catenin expression compared with 1205LuTR parental cells (Supplementary Fig. S5B and S5E–S5G). These alterations have been previously implicated in resistance to vemurafenib (40–44), but did not appear to be primary drivers of resistance in the PBRTs (Supplementary Fig. S6). Furthermore, BRAF V600E splice variants, which drive resistance to vemurafenib, were not detected in PBRT cell lines (Supplementary Fig. S7A). Thus, as with other RAF inhibitors and MEK–ERK1/2 regimens, prolonged exposure to PLX8394 results in acquired resistance associated with ERK1/2 pathway reactivation and compensatory pathway alterations.

We tested whether vertical targeting of the ERK1/2 pathway would overcome the acquired resistance to PLX8394, as it does in other resistant melanoma models (27, 39). Individual treatments of PLX8394, vemurafenib, trametinib, and the ERK1/2 inhibitor SCH772984 (SCH772) suppressed ERK1/2 signaling and reduced Rb phosphorylation in parental cells (Fig. 6A). Although trametinib reduced phospho-ERK1/2 in both PBRT #15 and #16, Rb phosphorylation was only affected in PBRT #15 (Fig. 6A). Similarly, SCH772 treatment reduced Rb phosphorylation in PBRT #15 but not PBRT #16 (Fig. 6A). Dose escalation of SCH772 was associated with an increase in PARP cleavage and BIM levels, as well as a reduction of both Rb phosphorylation and cyclin A expression in PBRT #15 (Supplementary Fig. S7B). However, these changes were not observed in PBRT #16 (Supplementary Fig. S7B). Similarly, crystal violet growth and EdU incorporation assays demonstrated that both parental and PBRT #15 cell lines were more sensitive to ERK1/2 inhibitor treatment than PBRT #16 (Fig. 6B and C; Supplementary Fig. S7C).

Ion Torrent sequencing results of PBRT #15 and PBRT #16 did not yield any missense mutations that would be indicative of ERK1/2 inhibitor resistance (Supplementary Tables S2 and S3). Consequently, we postulated that transcriptional alterations may contribute to resistance. Therefore, we utilized epigenetic agents, the bromodomain and extraterminal domain (BET) bromodomain (BRD) inhibitor, JQ1. Western blot analysis demonstrated that JQ1 treatment increased levels of the cyclin-dependent kinase inhibitors, p21 and p27, in PBRT #16 (Fig. 6D). This correlated with increased sensitivity of PBRT #16 to BET/BRD inhibitors in crystal violet growth assays and EdU incorporation (Fig. 6E and F; Supplementary Fig. S7D). Although PBRT #16 was more sensitive to BET/BRD inhibitor treatment, these agents also suppressed growth of PBRT #15, suggesting a potential universal second-line therapy option (Fig. 6E; Supplementary Fig. S7D).

FDA-approved mutant BRAF-selective inhibitors have markedly improved the treatment options and outcomes for BRAF V600E/K melanoma patients but are limited by the occurrence of adverse events associated with paradoxical ERK1/2 activation. In this study, we show the efficacy of the next-generation “paradox-breaking” BRAF inhibitor, PLX8394. Overall, our data support that PLX8394 is a viable treatment option to efficiently inhibit ERK1/2 signaling while limiting paradoxical effects in WT BRAF cells and highlight the use of tumor explants to rapidly assess the utility of targeted agents.

Using an in vivo GAL4-ELK1 reporter system to quantitatively and temporally measure ERK1/2 signaling in melanoma xenografts (28), we show that PLX8394 effectively inhibits ERK1/2 signaling and reduces mutant BRAF melanoma growth. This approach is complemented by an ex vivo explant system that demonstrates PLX8394 effectively inhibits ERK1/2 phosphorylation in patient tumors. Vemurafenib treatment only elicited an approximately 20% reduction in ERK1/2 phosphorylation in explants consistent with others' observations of minimal vemurafenib efficacy in 3D tumor mimics (35) and stroma/melanoma coculture systems (36–38). Variability in the response of patient tumors to vemurafenib may also be due to different treatment histories of the patients (Supplementary Table S1) and/or a high stromal component, which displays paradoxical ERK1/2 activation (35). In contrast, PLX8394 consistently inhibited ERK1/2 signaling in all patient biopsies independent of treatment history and was comparable with dabrafenib/trametinib combination therapy.

PLX8394 demonstrated enhanced efficacy when directly compared with vemurafenib in parallel ex vivo dosing of xenograft tissue. This may not be surprising as the PLX8394 IC₅₀ for ERK1/2 phosphorylation is approximately 10-fold lower than vemurafenib (18). In contrast, in the explant system, the dose required to reach ERK1/2 phosphorylation IC₅₀ of parental mutant BRAF tumor tissue treated with vemurafenib was much higher (∼44× IC₅₀) compared with PLX8394 (∼3× IC₅₀; Fig. 3A). One explanation for the larger difference of vemurafenib effectiveness is the potential paradoxical activation of ERK1/2 signaling in the stromal component present in the explant tissue, which is not present in 2D culture systems. Vemurafenib has been shown to paradoxically elicit tumor-protective responses from stromal components in ex vivo systems, as well as stimulate production of mitogenic growth factors from WT BRAF tumors in vivo (18, 35), a phenomenon that should be attenuated in PLX8394 treatment. The moderate increase in ERK1/2 phosphorylation IC₅₀ for PLX8394 treatment in the explant system, compared with 2D culture, may reflect ERK1/2 signaling present in stromal cells that should not be inhibited or paradoxically activated.

PLX8394 was effective in suppressing ERK1/2 signaling and eliciting PARP cleavage in BRAF splice variant–expressing, vemurafenib-resistant samples. This may be in part due to PLX8394's ability to better suppress splice variant homodimerization, thereby facilitating efficient inhibition of monomeric mutant BRAF kinase. Homodimerization of mutant BRAF splice variant–expressing cells has been linked to RAF inhibitor resistance (28, 39); however, the dimerization status of these splice variants in the presence of RAF inhibitors has not been tested. Vemurafenib and PLX4720 have previously been shown to destabilize homodimerization of full-length mutant BRAF (45) and heterodimerization of WT BRAF kinase domain with CRAF (46). On the other hand, reports indicate that these drugs may enhance BRAF/CRAF heterodimers (23). In the current study, we utilize differentially tagged V600E BRAF splice variants to measure homodimerization status during drug treatment to model the setting in which splice variants are expressed. We found that although both PLX4720 and PLX8394 reduced homodimerization of V600E BRAF splice variants, PLX8394 was significantly more effective. It is possible that the residual dimerized splice variant in PLX4720-treated cells may be adequate for drug resistance as further blocking of homodimerization by PLX8394 is associated with ERK1/2 pathway inhibition. Alternatively, these data may indicate that although splice variant homodimerization contributes to signaling in the presence of vemurafenib, it may not be wholly responsible for RAF inhibitor resistance.

The explant system was also used to show that vemurafenib, but not PLX8394, induces a strong paradoxical activation of ERK1/2 in WT/WT melanoma. Other groups have reported on pan-RAF/Src inhibitors that do not elicit paradoxical activation properties (47); however, the mutant BRAF selectivity of PLX8394 may afford a higher therapeutic index than agents that broadly inhibit RAF kinases. Furthermore, the mutant BRAF-specific targeting properties of PLX8394 may enable its use in combinatorial regimens with immune therapies. As suppression of ERK1/2 signaling is associated with increased melanoma antigen presentation (48), it is advantageous to use targeted inhibitors as an adjuvant to improve immunotherapy efficacy. However, there are conflicting reports of how systemic pathway inhibition (i.e., MEK inhibitor treatment) affects the antitumor immune response (49–51). Our results indicate that at doses that would inhibit mutant BRAF tumors, PLX8394 minimally affects ERK1/2 status in WT/WT cells, suggesting it will not alter normal T-cell activation. Thus, PLX8394 may be an appropriate partner with immune-based therapies, such as CTLA-4 and PD-1/PD-L1 inhibitors (52).

With an increasing number of available therapies for the treatment of mutant BRAF melanomas, identifying the best therapy for an individual patient is increasingly important. Patient-derived tumor xenograft models accurately reproduce a patient's response to therapy (53); however, these models are associated with long generation times and high cost. As an alternative, we describe a patient-derived explant system, PDeX, to test multiple treatment strategies using a single patient biopsy that accounts for intratumoral heterogeneity by assaying multiple sample pieces from different parts of the lesion. The ability of PDeX to test the efficacy of small-molecule inhibitors and mAbs in a short time period offers an inexpensive and rapid assay that is individualized and may inform patient treatment options.

As with other targeted therapies, acquired resistance to PLX8394 eventually occurs in our preclinical studies. Phospho-proteomic analysis implicated well-known BRAF inhibitor resistance markers in the PLX8394-resistant cell lines (40, 43, 54). However, initial experiments suggest that enhanced AKT activity and upregulation of PDGFR are not sole drivers of resistance in these cells (Supplementary Fig. S6). Rather, it is likely that they work in coordination with reactivation of the ERK1/2 pathway. It is possible that resistance mechanisms to PLX8394 will be unique from vemurafenib and dabrafenib and is underscored by the finding that PBRT #16 is resistant to vertical targeting of the ERK1/2 signaling pathway but is sensitive to BET/BRD inhibitor treatment. As enrollment for PLX8394 phase I/IIa study (ClinicalTrials.gov #NCT02428712) has only recently started, resistance mechanisms in patients remain unknown.

In summary, PLX8394 is a promising next-generation mutant BRAF-selective inhibitor that does not elicit strong paradoxical ERK1/2 activation in nonmutant BRAF cells. PLX8394 monotherapy is entering clinical trials with hope that it will prevent side effects associated with paradoxical ERK1/2 activation while simultaneously reducing grade 4 toxicities and permanent discontinuations associated with dual inhibitor therapies (14). Our parallel dosing experiments of tumor tissue suggest that PLX8394 is more effective than vemurafenib at suppressing ERK1/2 signaling even when considering vemurafenib's lower biochemical potency (18). In addition, the PDeX system utilized in this study provides a rapid and quantitative method to determine the efficacy of PLX8394 (and other targeted therapies) in patient tissues. As a result, our data suggest that PLX8394 is a promising new therapy for the treatment of mutant BRAF melanomas refractive to vemurafenib and that the PDeX system can potentially be used to guide the treatment of patients in a personalized manner.

M.A. Davies reports receiving commercial research grants from AstraZeneca, BMS, GSK, Oncothyreon, Roche/Genentech, and Sanofi Aventis and is a consultant/advisory board member for BMS, Novartis, Roche/Genentech, Sanofi Aventis, and Vaccinex. A.E. Aplin reports receiving a commercial research grant from Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E.J. Hartsough, C.H. Kugel III, M.J. Vido, A.E. Aplin

Development of methodology: E.J. Hartsough, M.J. Schiewer, A.E. Aplin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.J. Hartsough, C.H. Kugel III, M.J. Vido, A.C. Berger, A.E. Aplin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.J. Hartsough, C.H. Kugel III, M.J. Vido, A.C. Berger, T.J. Purwin, A. Goldberg, M.A. Davies, K.E. Knudsen, A.E. Aplin

Writing, review, and/or revision of the manuscript: E.J. Hartsough, C.H. Kugel III, M.J. Vido, A.C. Berger, A. Goldberg, M.A. Davies, A.E. Aplin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.J. Hartsough, A.C. Berger, G. Bollag, A.E. Aplin

Study supervision: A.E. Aplin

We are grateful to Sheera Rosenbaum for generating the WT/WT melanoma xenografts, Dr. Kevin Basile for help with mutant BRAF xenografts, Dr. Meenhard Herlyn (Wistar Institute, Philadelphia, PA) for the WM melanoma cell lines, and Dr. Barbara Bedogni (Case Western Reserve University, Cleveland, OH) for the B6 and MeWo cells.

This work was supported by the National Cancer Center and National Institutes of Health K99 grant CA207855 (E.J. Hartsough) and F30 fellowship CA203314 (M.J. Vido), National Institutes of Health R01 grants, CA182635 and CA196278 (A.E. Aplin), and grants from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (A.E. Aplin and M.A. Davies). The Sidney Kimmel Cancer Center core facilities at Thomas Jefferson University are funded by National Cancer Institute Support Grant P30 CA56036. The Reverse Phase Protein Array Core Facility at M.D Anderson Cancer Center is supported by the NCI Cancer Center Support Grant, CA16672.

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.