ErbB3 Targeting Enhances the Effects of MEK Inhibitor in Wild-Type BRAF/NRAS Melanoma

Effective targeted therapy for wild-type BRAF/wild-type NRAS (WT/WT) cutaneous melanoma remains an unmet clinical need (1). WT/WT tumors account for approximately 30% of cutaneous melanoma (2, 3). Approximately one third of WT/WT melanoma harbor NF1 alterations; cKIT, cyclin D1, and CDK4 amplifications are also detected (2). Despite elevated MEK–ERK1/2 signaling in most WT/WT melanoma, the response of these patients to MEK inhibitors is poor (4). This contrasts with the efficacy of MEK inhibitors in V600E/K BRAF-harboring melanoma, for which trametinib is FDA-approved alone (5) and in combination with BRAF inhibitor (6). Additionally, the MEK inhibitor, binimetinib, showed improved response rates and progression-free survival compared with dacarbazine in patients with mutant NRAS melanoma (7). The standard of care for patients with WT/WT melanoma is immune checkpoint inhibitor therapy with anti–CTLA-4 (ipilimumab) and/or anti–PD-1 (pembrolizumab or nivolumab; ref. 8). Because many WT/WT patients do not respond to immune checkpoint inhibitors and immune-related adverse effects are common, there is a critical clinical need for additional therapeutic options.

Tumors that have primary/intrinsic resistance to ERK1/2 pathway inhibitors frequently show adaptive upregulation of expression of receptor tyrosine kinases (RTK), including IGF1R, EGFR, PDGFRβ, AXL, MET, and ErbB3, and consequential activation of ERK1/2-independent survival pathways (9–12). In mutant BRAF melanoma and thyroid carcinoma, ErbB3/HER3 (v-erb-b2 erythroblastic leukemia viral oncogene homolog 3/human epidermal receptor 3) upregulation is a cellular adaptation to BRAF/MEK inhibition (9, 13). ErbB3 has weak kinase activity compared with other EGFR family members (14) and is activated via heterodimerization with other family members. ErbB2 is the most potent signaling partner for ErbB3 (15–17). The stroma has been strongly implicated in melanoma progression (18). Growth factors that activate RTKs can be derived from the tumor microenvironment, suggesting a potential role for cells such as cancer-associated fibroblasts (CAF) in mediating resistance to targeted therapies (9, 19–23). Although upregulated expression of ErbB3 and other RTKs is linked to resistance to BRAF and MEK inhibitors in BRAF-mutant melanoma, the mechanisms underlying resistance in WT/WT melanoma remain unclear.

Here, we studied adaptive responses to MEK inhibitor in WT/WT cutaneous melanoma. Our results demonstrate that fibroblast-derived NRG1 promotes resistance to MEK inhibitor therapy in these tumors. Resistance is not dependent upon ErbB3 upregulation, but rather is mediated via enhanced phosphorylation of ErbB2 and ErbB3 and increased ErbB2 cell-surface levels. Targeting adaptive ErbB3/ErbB2 signaling with neutralizing antibodies to either ErbB2 or ErbB3 enhances the effects of MEK inhibitors in WT/WT melanoma in vitro and in vivo. These findings provide preclinical data to support the use of anti-ErbB2 and/or anti-ErbB3 antibodies to maximize the effects of MEK inhibitors in patients with WT/WT cutaneous melanoma.

Details of the cell lines utilized in this study are provided in Supplementary Materials.

Human melanoma biopsies (TJUMEL40 and TJUMEL45) were obtained from Thomas Jefferson University Hospital under IRB-approved protocol (#10D.341) that include written informed consent and was in accordance with recognized ethical guidelines. Postsurgery, excess adipose tissue was removed. Tumors were cut into small pieces and digested with collagenase (Sigma-Aldrich Co) in complete medium at 37°C for 2 to 4 hours. Samples were washed with PBS and resuspended in DMEM supplemented with 10% FBS containing 5 μg/mL insulin. CAFs cultures were maintained for 5 passages. CAFs were authenticated by morphology and by the expression of α-smooth muscle actin (α-SMA), fibroblast activation protein (FAP) and PDGFRα, and by the absence of CD45.

Patient-derived xenografts (PDX) were generated in NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. Two human melanoma biopsies (TJUMEL40 and WM4279) were processed, as described above, except samples were incubated with collagenase in complete medium at 37°C for 30 minutes. For TJUMEL40, the sample was divided and implanted into the back of two NSG mice with Matrigel (Corning). Tumors were harvested when they reached 1,000 mm³ in size (passage 1) and reimplanted in mice (passage 2). The WM4279 PDX was obtained from the Wistar Institute as passage 3 (24, 25). Tumor pieces from either passage 1 or 2, for TJUMEL40, and passage 4 for WM4279 were collected, cut into small pieces and digested with collagenase in complete medium at 37°C for 2 to 4 hours. Cells were then centrifuged at 4000 rpm for 4 minutes; the pellet was washed with PBS and centrifuged again. The pellet was resuspended and cells were cultured in DMEM supplemented with 10% FBS for a single passage (TJUMEL40) or used directly for experiments (WM4279).

Recombinant human NRG1 and insulin-like growth factor 1 (IGF1) were purchased from Cell Signaling Technology. Recombinant human epidermal growth factor (EGF), platelet-derived growth factor beta (PDGFbb), and hepatocyte growth factor (HGF) were purchased from Lonza, R&D Systems, and Proteintech, respectively. LJM716 was produced and purified by Novartis. Trametinib and PD0325901 were purchased from Selleck Chemicals.

Proteins were extracted with Laemmli sample buffer, resolved by SDS–PAGE and transferred to PVDF membranes. Immunoreactivity was detected using HRP-conjugated secondary antibodies (CalBioTech) and chemiluminescence HRP-recognizing substrates (Thermo Scientific) on a VersaDoc multi-Imager. Primary antibodies used are listed in Supplementary Materials and Methods.

Crystal violet assays were performed and analyzed as previously described (20). Plates were scanned and cell density was quantified using ImageJ software. The intensity mean was calculated for each well. Pictures were taken with a Nikon Eclipse Ti inverted microscope with NIS-Elements AR 3.00 software (Nikon).

WT/WT cells (B6, FEMX, MeWo, YUROL, and WM3928) were treated, as indicated, and proteins were analyzed, as previously described (26). Samples were normalized as described at https://www.mdanderson.org/research/research-resources/core-facilities/functional-proteomics-rppa-core/faq.html.

For Western blot analysis, conditioned medium was prepared, as previously described (20). For growth assays, conditioned medium was prepared as previously described (20) except that cells were cultured in 18 mL of serum-free medium. For normalization, fibroblasts were counted afterward. Cell counts per plate were: 3.34 ± 0.55 × 10⁶ for HFF cells, 3.37 ± 0.7 × 10⁶ for HTERT cells, and 1.44 ± 0.24 × 10⁶ for CAFs.

Cells were treated, detached using 2 mmol/L EDTA, and washed twice with PBS. Samples were stained with Zombie NIR (BioLegend), for 10 minutes at room temperature in the dark. Following two washes with FACS buffer (PBS with 1% FBS and 0.5% sodium azide), cells were stained with either ErbB3-PE or ErbB2-PE (BioLegend) for 30 minutes at room temperature. Cells were fixed with cytofix/cytosperm and analyzed on an LSR II (BD Biosciences) using FlowJo software (TreeStar).

Animal experiments were performed at a Thomas Jefferson University facility that is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. The Institutional Animal Care and Use Committee approved these studies. FEMX (1 × 10⁶) or MeWo (2 × 10⁶) cells were injected intradermally into the backs of athymic mice (NU/J, homozygous, Jackson, 6–8 weeks, 20–25 g). When tumors were palpable, mice were randomly sorted into four cohorts. For FEMX xenografts, mice were treated with trametinib diet (1 mg/kg) and/or injected intraperitoneally with LJM716 (400 μg/mouse). AIN-76A diet and PBS injection (100 μL) were used as vehicle. Treatments were discontinued from day 9 to day 12, and diet gel 76A (Scanbur) was given to minimize weight loss. Drug treatments were resumed on day 12 and were continued to the end of the experiment. In the MeWo experiment, PD0325901 diet (7 mg/kg), instead of trametinib, was utilized and LJM716 injected (400 μg/mouse to day 9, 200 μg/mouse after day 9). For TJUMEL40 PDX experiments, tumors were processed as described in short-term culture except that when tumors were palpable, mice were randomly sorted into three cohorts. Five mice (4 female, F; 1 male, M) were treated with PD0325901 diet (7 mg/kg; Research Diets, Inc.), five mice (3F and 2M) were treated with PD0325901 diet (7 mg/kg) and injected intraperitoneally with LJM716 (200 μg/mouse for F and 300 μg/mouse for M). AIN-76A diet and PBS injection (500 μL) were used as vehicle in three mice (2F and 1M). Digital caliper measurements of the tumors were taken every 3 days, and tumor volumes were calculated using the formula: volume = (length × width²) × 0.52.

Tissue microarrays (TMA) were generated at The University of Texas MD Anderson Cancer Center containing 345 tumor samples from 164 patients with WT/WT melanoma. Of these, 110 samples were from 43 stage III patients and 235 samples were from 121 stage IV WT/WT patients. Expression of phospho-ErbB2 in the TMA was run on the Ventana Benchmark Discovery staining platform. Discovery CC1 epitope retrieval was followed by antibody (pErbB2 Y1221/Y1222, 6B12, Cell Signaling Technology, at a 1:200 dilution) incubation for 44 minutes at ambient temperature. Phospho-ErbB2 was visualized with the Ventana ultra View Universal Alkaline Phosphatase Red Detection Kit and lastly counterstained with hematoxylin for 8 minutes. The intensity of staining and the percentage of positive cells were evaluated with the VisioPharm system. Samples were first separated by stage. Samples with no detectable level were classified as negative, and a mean intensity cutoff value of 50 was used to separate the remaining samples into high and low expression groups. The highest scoring call was kept for patients with multiple samples.

The Cancer Genome Atlas (TCGA) cutaneous melanoma RNA sequencing (RNA-seq) V2 normalized gene-expression and mutation call data were retrieved from the latest Broad GDAC Firehose data run (stddata__2016_01_28). Similarly, the TCGA cutaneous melanoma replicates-based normalized reverse-phase protein array (RPPA) data (v 4.0) were collected from The Cancer Proteome Atlas (TCPA; ref. 27). Gene set variation analysis (GSVA; ref. 28) was used to calculate a CAF signature score for each sample from the list of 88 CAF signature genes in Tirosh and colleagues (29). ESTIMATE, LUMP, hematoxylin and eosin (H&E) staining, and CPE tumor purity data were collected from Aran and colleagues (30). ESTIMATE, LUMP, and H&E staining purity values were calculated using RNA-seq V2, methylation, H&E staining data, respectively. CPE purity values were calculated from the combination of ESTIMATE, LUMP, and H&E staining values. ABSOLUTE tumor purity data were gathered from TCGA Network (2). ABSOLUTE tumor purity values were calculated using copy-number segment data. Eight of the CAF signature genes were removed from the list due to their use in calculating ESTIMATE purity values. Metastatic samples were retained when there were multiple samples from the same patient. In total, 74 melanoma patient samples had an estimated purity level, RNA-seq data, and a WT/WT mutation profile, with 54 also having RPPA data. Gene-expression levels were log₂ transformed, then median-centered, and clustered. Samples were then sorted based on CPE value. Colorblind-safe color schemes were derived from ColorBrewer (v2.0; ref. 31).

Crystal violet quantification was performed using Student two sample t tests, assuming unequal variance. For RPPA analysis, the two-sample t test with 1,000 permutations and assumed unequal variance was used to compare protein expression levels between NRG1 and ± trametinib groups at different time intervals for each cell line. Additional details are described in Supplementary Data.

To evaluate the percentage of WT/WT melanoma patient samples that coexpress pErbB2 and pErbB3, the TCGA Skin Cutaneous Melanoma replicates-based normalized (v 4.0) RPPA data set was collected from TCPA (27). Correlation analysis was performed between NRG1 (heregulin), pErbB2, and pErbB3 data for WT/WT samples (n = 54). Z-score values were calculated for pErbB2 and pErbB3 expression levels using all samples (n = 354). Samples with both pErbB2 and pErbB3 z-scores >1 and >−1 were classified as high and mid, respectively. The remaining samples were classified as low.

Associations between NRG1, pErbB2, and pErbB3 protein expression levels were determined by correlation analysis of RPPA data. For the association between NRG1 mRNA expression and CAF levels, NRG1 RNA-seq and CAF GSVA scores were used for correlation analysis. Tumor purity and NRG1 expression-level associations were determined by correlation analysis of NRG1 RNA-seq and ABSOLUTE, ESTIMATE, LUMP, H&E staining, and CPE tumor purity estimates plus NRG1 RPPA and CPE tumor purity estimates data. For FEMX and MeWo in vivo studies, the repeated over time log-transformed tumor volumes were modeled as a low order polynomial function of day using a linear mixed effects model adjusting for the random effects of animal and allowing for animal-specific growth trajectories. Additional details are described in Supplementary Materials. For MeWo xenograft and TJUMEL40 PDX tumors, volume day-to-day comparisons were performed using Student two sample t tests, assuming unequal variance.

It is important to provide the rationale for targeted therapy strategies in WT/WT melanoma. TCGA SKMC data set analysis, performed according to Liu and colleagues guidelines (32), showed that ErbB3 expression level is inversely correlated with stage III WT/WT melanoma patient survival before treatment (Supplementary Fig. S1A). Furthermore, expression of NRG1, the ligand for ErbB3, positively correlated with the phosphorylation state of both ErbB3 and its coreceptor ErbB2 in WT/WT melanoma samples within the SKMC data set (Fig. 1A). Analysis of the SKMC replicates-based normalization RPPA data set using the z-score classification methods showed that 78% (42 of 54) of patients with WT/WT melanoma coexpressed medium–high levels of both phosphorylated ErbB3 and ErbB2 (Fig. 1B). A comparable coexpression was observed in BRAF-mutant melanoma samples and similar results were also observed in NRAS-mutant melanoma samples (Supplementary Fig. S1B). Similar findings were evident when the z-score was calculated based on the Pan-Cancer 19 RPPA data collected from TCPA with nearly 60% of patients coexpressing phospho-ErbB3 and phospho-ErbB2 (Supplementary Fig. S1C). Based on these findings, we stained a TMA of WT/WT melanoma patient samples containing 43 stage III tumors and 121 stage IV tumors. Due to the suitability of antibodies for IHC, phospho-ErbB2 was analyzed. Seventy-four percent of stage III and IV WT/WT melanoma patient samples expressed phosphorylated ErbB2. Thirty percent and 43% of 43 stage III and 121 stage IV WT/WT melanoma had high levels of phosphorylated ErbB2, respectively (Fig. 1C). These data indicate that ErbB3 and ErbB2 are activated in a high proportion of WT/WT melanoma patient tumors.

Although autocrine production of NRG1 mediates phosphorylation of ErbB3/ErbB2 in a small cohort WT/WT melanoma (33), ErbB3 may be a relevant target across WT/WT melanoma, possibly in combinatorial approaches. To assess tumor versus stromal source of NRG1, we analyzed the correlation between tumor purity estimates and NRG1 mRNA across 74 WT/WT patient samples from the SKMC TCGA data set (2). Tumor purity estimates (ESTIMATE, LUMP, H&E, ABSOLUTE, and CPE) derived from RNA-seq, H&E staining, methylation (30), and somatic copy-number data (2) all showed statistically significant inverse correlations with NRG1 RNA level (Fig. 2A). CPE, which incorporates multiple tumor purity estimates, showed the strongest negative correlation with NRG1 RNA (Fig. 2A, bottom) and similar results with NRG1 protein levels within the TCGA data set (Fig. 2B). Comparing RNA levels, NRG1 showed a statistically significant positive correlation with CAF signature genes (Fig. 2C and D; ref. 29). The low level of endogenous NRG1 expression was confirmed in multiple WT/WT melanoma cell lines in both basal conditions and following MEK inhibition (Supplementary Fig. S2A and S2B). These data show that NRG1 RNA levels in human melanoma patient samples are associated with a high CAF content.

Because the MEK–ERK1/2 pathway is activated in WT/WT melanoma but MEK inhibitors are ineffective in patients, we analyzed the ability of NRG1 to mediate resistance to the MEK inhibitor, trametinib. We also tested the effect of four additional growth factors (EGF, HGF, IGF, and PDGFbb), which have been implicated in resistance to targeted therapies in mutant BRAF melanoma (9–12, 19). FEMX and MeWo cell lines were utilized, which are wild-type for BRAF and NRAS but differ in their NF1 status. FEMX retain NF1 expression but MeWo are negative for NF1 (Supplementary Fig. S3A and S3B; ref. 34). In the absence of trametinib, all growth factors elicited no to modest increases in colony formation (Fig. 3A; Supplementary Fig. S3C). In trametinib-treated cells, NRG1 strongly rescued growth (Fig. 3B; Supplementary Fig. S3D and S3E). Contrary to the effects observed in BRAF-mutant melanoma, EGF, HGF, IGF, and PDGFbb had little to no protective effect in MEK-inhibited cells (Fig. 3B; Supplementary Fig. S3D and S3E). NRG1-mediated protection was generalizable as we observed similar effects in B6 and YUROL cells (Fig. 3C; Supplementary Fig. S3F). More modest effects were observed in WM3928. NRG1 effects were independent of NF1 expression. Notably, NRG1 stimulated high-level phosphorylation of ErbB3 and ErbB2 in trametinib-treated cells compared with untreated cells (Fig. 3D). MET and IGFR were phosphorylated following treatment with HGF and IGF, respectively; however, phosphorylation of EGFR and PDGFRβ was weak following growth factor stimulation, likely due to low receptor expression (Supplementary Fig. S3G and S3H). These data show that NRG1 modulates the response to MEK inhibitors in WT/WT melanoma cells.

To investigate NRG1-mediated signaling, we performed RPPA in five WT/WT melanoma cell lines treated with/without MEK inhibitor (Supplementary Fig. S4A). NRG1-mediated phosphorylation of ErbB2 and ErbB3 was enhanced in trametinib-treated cells (Fig. 4A and B; Supplementary Fig. S4B), although enhanced ErbB3 phosphorylation did not meet statistical significance in all cell lines. Analysis of individual time points showed that ErbB2 phosphorylation was transiently enhanced in MEK-inhibited, NRG1-stimulated cells and was associated with prolonged AKT activation (quantitated in Fig. 4B, bottom; Supplementary Fig. S4B). In contrast to some findings on ERK1/2 pathway reactivation in mutant BRAF melanoma (21), NRG1 elicited minimal effects on ERK1/2 phosphorylation in MEK-inhibited WT/WT melanoma cells (Fig. 4C; Supplementary Fig. S4C). Western blot analysis corroborated the RPPA data showing enhanced NRG1-stimulated ErbB3/ErbB2 and AKT phosphorylation in MEK-inhibited cells (Fig. 4C; Supplementary Fig. S4D and S4E). The level of NRG1-mediated ErbB2 phosphorylation correlated with the ability of NRG1 to rescue MEK inhibitor effects on cell growth and to induce a higher and prolonged AKT activation (compare Fig. 4B with Fig. 3B and C). NRG1-stimulated AKT phosphorylation was lower and transient in WM3928 compared with the other cell lines, consistent with the lower activation of ErbB2 (Fig. 4C). This result correlated with lower level of protection afforded by NRG1 on cell growth following trametinib treatment (Fig. 3C). ErbB3 partners with other ErbB family members, such as EGFR and ErbB4, as well as MET (15–17, 35, 36). EGFR and ErbB4 were lowly expressed/undetectable and MET phosphorylation was unchanged following NRG1 treatment (Supplementary Figs. S3H, S3G, and S4B). These data indicate that adaptive activation of ErbB3/ErbB2 and AKT signaling is associated with protection against MEK inhibition in WT/WT melanoma. Because patients with NRAS-mutant melanoma showed similar coexpressed medium–high levels of both phosphorylated ErbB3 and ErbB2 compared with WT/WT (Supplementary Fig. S1B), we next analyzed the effect of NRG1 in MEK-inhibited NRAS-mutant cell lines. Of the four cell lines, just one expressed ErbB3 and ErbB2 and showed adaptive activation of ErbB3/ErbB2 pathways in the presence of trametinib and NRG, confirming that NRG1 effects rely on the expression of ErbB3 and its coreceptor, ErbB2 (Supplementary Fig. S4F).

Next, we tested for changes in the expression of ErbB3 and ErbB2 that would underlie enhanced signaling in MEK-inhibited cells. Increased NRG1-initiated ErbB3 and ErbB2 phosphorylation occurred within 6 hours of MEK inhibition in WT/WT melanoma cells, suggesting a nontranscription-based mechanism (Fig. 5A). Total cellular levels of ErbB3 were either unaltered or only slightly upregulated following MEK inhibition (Supplementary Fig. S4B and S4E). Both total and cell surface levels of ErbB2 expression also remained unchanged (Fig. 5A; Supplementary Fig. S5A). These data suggest that short-term MEK inhibition renders ErbB3/ErbB2 complexes more responsive to NRG1.

We also considered effects at later time points. Cell surface levels of ErbB3 after NRG1 stimulation were comparable in the absence and presence of MEK inhibitor (Supplementary Fig. S5B). By contrast, cell surface ErbB2 levels were consistently upregulated by 24 hours of MEK inhibition but were subsequently downregulated by NRG1 stimulation (Fig. 5B and C; Supplementary Fig. S5C and S5D). Importantly in NRG1-stimulated conditions, cell surface ErbB2 was increased in trametinib-treated cells versus control cells between 6 and 24 hours (Fig. 5C; Supplementary Fig. S5C). ErbB3 cell surface expression levels were also reduced following NRG1 stimulation (Fig. 5D; Supplementary Fig. S5E), consistent with previous studies (37, 38). The level of reduction was equivalent at most time points in MEK-inhibited versus vehicle cells and ErbB3 levels were restored by 24 hours of stimulation, making ErbB3 available for dimerization. Together, these data suggest the presence of a second mechanism that prolongs cell surface levels of ErbB2 following MEK inhibition to maintain durable activation of downstream signaling. These findings contrast with the mechanism identified in BRAF-mutant cell lines in which NRG1 effects are mediated by the upregulation of FOXD3 and its ability to induce ErbB3 transcript levels (9, 13).

To address the translational potential of our studies, we tested the ability of ErbB3 and ErbB2 neutralizing antibodies to block the protective effects of NRG1 in MEK-inhibited cells. We utilized LJM716, an ErbB3-neutralizing antibody that selectively binds an epitope created by domains II and IV of the ErbB3 extracellular domain and pertuzumab, a clinical-grade antibody that binds to ErbB2 subdomain II (39–41). LJM716 and pertuzumab individually reduced NRG1-induced phosphorylation of ErbB3, ErbB2 and AKT (Fig. 6A; Supplementary Fig. S6A). In colony growth assays, neither LJM716 nor pertuzumab affected cell growth in normal conditions (Supplementary Fig. S6B) but both antibodies partially reversed the protective effect of NRG1 in MEK-inhibited cell lines (Fig. 6B; Supplementary Fig. S6C). Similarly, the β-sparing phosphoinositide 3-kinase (PI3K) inhibitor, GDC0032 (42), reduced NRG1-induced phosphorylation of ErbB3 and AKT and reversed NRG1 effect on cell growth in MEK-inhibited cells (Supplementary Fig. S6D and S6E). These data corroborate the notion that ErbB3 effects, in response to MEK inhibitor, are mediated through the activation of the PI3K–AKT pathway.

Given the poor response to MEK inhibitors in patients with WT/WT melanoma (4), there are few on-treatment WT/WT melanoma patient samples available for analysis. As an alternative, we generated short-term cultures from WT/WT PDXs. Melanoma cultures from TJUMEL40 and WM4279 were validated by SOX10 expression and the absence of the fibroblast marker, PDGFRα (Supplementary Fig. S6F). Consistent with the established melanoma cell lines, TJUMEL40 cultures did not show activation of ErbB3 in basal conditions (Fig. 6C) and LJM716 did not affect their proliferation in the presence of NRG1 (Supplementary Fig. S6G). Furthermore, MEK inhibition increased phospho-ErbB3 and phospho-ErbB2 levels following NRG1 treatment, effects that were reversed by LJM76 and pertuzumab antibodies (Fig. 6C). LJM716 and pertuzumab partially reversed the protective effect of NRG1 on cell growth in MEK-inhibited TJUMEL40 cultures (Fig. 6D; Supplementary Fig. S6H). Thus, increased ErbB3/ErbB2 phosphorylation following MEK inhibition is targetable by ErbB3/ErbB2-neutralizing agents.

To test the role of the stromal TME, we evaluated the effects of fibroblast-derived conditioned medium. MEK-inhibited MeWo and FEMX cells showed robust phosphorylation of ErbB3, ErbB2, and AKT after treatment with conditioned medium from two different fibroblast cell lines (HFF and HTERT; Fig. 7A; Supplementary Fig. S7A). Additionally, we detected NRG1 in isolated CAF cultures, which were verified by expression of αSMA, FAP, and PDGFRα, from two WT/WT patient-derived melanomas (Supplementary Fig. S7B–S7D). Conditioned media derived from CAF40 and CAF45 stimulated phosphorylation of ErbB3, ErbB2, and AKT in MEK-inhibited WT/WT melanoma cells (Fig. 7B; Supplementary Fig. S7E). ErbB3, ErbB2, and AKT phosphorylation induced by fibroblast and CAF-conditioned media was reversed by LJM716 or pertuzumab treatment (Fig. 7A and B; Supplementary Fig. S7A and S7E). Notably, CAF40-conditioned medium was able to induce ErbB3 and ErbB2 phosphorylation in the matched melanoma cells extracted from the same patient (TJUMEL40; Fig. 7C). Conditioned media from fibroblasts and CAFs rescued the inhibitory effects of trametinib on cell growth (Fig. 7D and E; Supplementary Fig. S7F and S7G). Consistently, these effects were partially reversed by LJM716 and pertuzumab (Fig. 7D and E; Supplementary Fig. S7F and S7G). These data support the notion that paracrine NRG1 production promotes resistance to trametinib in WT/WT melanoma cells.

Our results provide the rationale to test the combination of MEK inhibitors with anti-ErbB3 antibodies in WT/WT melanoma in vivo. To assess the effects of MEK inhibitor and the anti-ErbB3 antibody, LJM716, in a fibroblast-rich microenvironment, we generated FEMX and MeWo intradermal xenografts. Treatment of xenografts with LJM716 alone did not alter tumor growth, supporting the notion that ErbB3 activation is an adaptive response to MEK inhibition (Fig. 7F and G). MEK inhibitor reduced tumor growth in both models, which was significantly further delayed when LJM716 was combined with MEK inhibitor treatment (Fig. 7F and G; Supplementary Fig. S7H). Furthermore, similar results were observed utilizing in vivo PDXs derived from TJUMEL40 (Fig. 7H). Together, these data suggest that the ErbB3-blocking agents significantly enhance the growth reduction effect of MEK inhibitors of WT/WT melanoma.

Our findings demonstrate that the ErbB3/ErbB2 pathway is adaptively activated in MEK-inhibited WT/WT melanoma by stromal NRG1 and that targeting this compensatory pathway with clinical-grade antibodies increases the efficiency of MEK inhibitors. Our findings underscore the influence of the tumor microenvironment on mediating resistance to targeted agents and support testing of MEK inhibitors and in combination with ErbB3/ErbB2 targeting antibodies in WT/WT cutaneous melanoma.

Our studies address an important clinical need. Major advances have been made for the treatment of V600-mutant BRAF melanoma. By contrast, targeted inhibitor trials in nonmutant BRAF melanoma have elicited poor response rates. In a study from Falchook and colleagues, a 20% response rate to the MEK inhibitor, trametinib, was observed in WT/WT (although two of these samples harbored atypical BRAF mutations; ref. 4). Thus, new strategies are needed for the treatment of this subgroup of melanoma. Our findings may extend to mutant NRAS melanoma. Although bioinformatic analysis showed strong basal pErbB3 and pErbB2 levels in mutant NRAS melanoma, cell-based studies showed various levels of ErbB3 adaptive responses. These data reflect the high level of heterogeneity present in NRAS-mutant melanoma and need further investigation to clarify the role of NRG1 in driving resistance to MEK inhibitor in this subgroup.

In the mutant BRAF setting, multiple growth factors and their cognate receptors have been shown to mediate resistance to BRAF inhibitors (9, 12, 19–22, 43). WT/WT melanomas are frequently sensitive to MEK inhibitors in monocultures (data within and refs. 34 and 44). We show that NRG1 protects against MEK inhibitors in this subset of melanoma. By contrast, other growth factors linked to resistance to BRAF inhibitors in mutant BRAF melanoma elicit little to no reversal of growth inhibition. Nevertheless, we do not rule out the possible involvement of other growth factors in protecting WT/WT melanoma from growth blockade mediated by MEK inhibitors. Indeed, LJM716 and pertuzumab partially, but not completely, reversed the effects of CAF-conditioned medium on cell growth in MEK-inhibited cells. PI3K and AKT inhibitors may broadly block signaling downstream of multiple RTKs; however, the combination of MEK inhibitors and either PI3K or AKT inhibitors has been challenging with high toxicity and poor response rate issues (45). The use of bi- and multivalent antibodies may represent a more efficient alternative to block the compensative activation of other RTKs (46). Clinical-grade anti-ErbB3–targeting agents are being developed and tested in clinical trials for many solid malignancies (NCT02387216, NCT02167854, NCT01602406, and NCT02980341; refs. 47–49). Considering the high percentage of tumors coexpressing pErbB3 and pErbB2, drug-conjugated ErbB3/ErbB2 antibodies may increase the cytotoxic effect and the efficacy of treatment. Previous studies have shown that ErbB2 antibody–drug conjugates have a favorable safety profile compared with other treatments and a notable survival benefit in heavily pretreated patients, including patients treated with pertuzumab or lapatinib, with less severe toxic effects than treatment of physician's choice (50).

We show upregulation of ErbB3/ErbB2 phosphorylation in MEK-inhibited WT/WT melanoma. In contrast to previous studies published in the context of mutant BRAF melanoma and mutant KRAS lung and colon cancer (9, 13, 20–22, 51, 52), NRG1 effects in MEK-inhibited WT/WT cells are likely driven by an increase in ErbB2 phosphorylation that occurs within hours of stimulation, suggestive of a nontranscriptional mechanism, and by the retention of ErbB2 protein at the cell surface. Recent studies in breast cancer cells demonstrate ErbB2 is inactivated by ERK1/2-dependent phosphorylation of threonine 677 in the juxtamembrane region. By contrast, in MEK-inhibited cells, dephosphorylation of threonine 677 enables ErbB2 to be a more effective coreceptor for ErbB3 (52, 53). In our studies, phosphorylation of ErbB2 at T677 was not detected in basal conditions (Supplementary Fig. S7I).

We previously showed that a small subset of WT/WT melanoma express NRG1, which acts in an autocrine manner. These melanomas are highly sensitive to ErbB3 monotherapy (33) and show less sensitivity to MEK inhibition, in vitro, compared with the subgroup tested in this study. Through analysis of patient database samples and use of conditioned medium derived from CAFs isolated from tumors, we show that ErbB3/ErbB2 activation is mediated via a paracrine NRG1 mechanism in the majority of WT/WT melanoma, including our PDX-derived cells, TJUMEL40 and WM4279. In our studies, these tumors do not respond to ErbB3 monotherapy. The use of PDX models as well as patient-matched CAFs and tumor cell cultures strengthens the notion that stromal NRG1 drives adaptive resistance in WT/WT melanoma. The mechanism(s) regulating NRG1 expression in CAFs remains to be determined. We show that phosphorylated ErbB3 and ErbB2 are coexpressed in nearly 80% of WT/WT melanoma, suggesting that a high percentage of patients could potentially benefit from the combination of MEK inhibitor plus ErbB3/ErbB2 antibodies. The ability of NRG1 to rescue MEK inhibitors effect on cell growth is linked to its ability to phosphorylate ErbB2, identifying phospho-ErbB2 as a possible biomarker for testing the combinatory effects of MEK inhibitors and ErbB3/ErbB2 targeting antibodies in WT/WT cutaneous melanoma.

Heterogeneity in melanoma is an important issue and moving forward it will be critical to identify biomarkers for patients who are likely to respond to MEK inhibitor/ErbB3 targeting treatment combinations. Given the intratumor heterogeneity, this should ideally be performed at the single-cell level. The transcription factor SOX10 has been described as a direct regulator of ErbB3 in neural crest-derived cells (54); however, SOX10 in mutant BRAF melanoma cells is associated with repression of EGFR and PDGFR (11). Additionally, depletion of SOX10 is associated with a slow-growing phenotype that may contribute to resistance to targeted inhibitors (11). A future direction will be to determine the effect of MEK inhibitors on intratumor heterogeneity in WT/WT melanoma.

M.A. Davies reports receiving a commercial research grant from Roche/Genentech, GSK, Astrazeneca, and Sanofi-Aventis and is a consultant/advisory board member for Novartis, Roche-Genentech, and Sanofi-Aventis. Jeffrey E. Gershenwald is a consultant/advisory board member for Merck, Novartis, and Bristol-Myers Squibb. 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: C. Capparelli, A.E. Aplin

Development of methodology: C. Capparelli, T.J. Purwin, C. Krepler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Capparelli, S.A. Heilman, A.C. Berger, J.E. Gershenwald, C. Krepler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Capparelli, T.J. Purwin, S.A. Heilman, I. Chervoneva, P.A. McCue, M.A. Davies, C. Krepler, A.E. Aplin

Writing, review, and/or revision of the manuscript: C. Capparelli, T.J. Purwin, S.A. Heilman, P.A. McCue, A.C. Berger, M.A. Davies, J.E. Gershenwald, C. Krepler, A.E. Aplin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.J. Purwin, A.C. Berger

Study supervision: A.E. Aplin

We thank Dr. Barbara Bedogni (University of Miami), Dr. Ruth Halaban (Yale University), Dr. Meenhard Herlyn (Wistar Institute), and Dr. David Solit (Memorial Sloan-Kettering Cancer Center) for generously providing cell lines, Dr. Andrea Morrione (Thomas Jefferson University) for positive control lysates, and Dr. Timothy L. Manser and Justin Walker (Thomas Jefferson University) for the NSG mice colony. We are grateful to Dr. Hiroaki Sakurai (University of Toyama) for supplying the anti-phosphoT677 ErbB2 antibody and Novartis Pharmaceutical Corp. for the supply of LJM716.

This work is supported by grants from NIH/NCI (R01 CA196278) and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation to A.E. Aplin. The Sidney Kimmel Cancer Center Flow Cytometry, Translational Pathology and Meta-Omics core facilities are supported by NIH/NCI Support Grant (P30 CA056036). The RPPA studies were performed at the Functional Proteomics Core Facility at The University of Texas MD Anderson Cancer Center, which 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.