Type I Cytokines Synergize with Oncogene Inhibition to Induce Tumor Growth Arrest

More than 50% of patients with metastatic melanoma have tumors driven by oncogenic BRAF^V600E mutations, which promote aberrant cell growth through constitutive activation of the MAPK pathway (1, 2). Inhibition of the BRAF^V600E oncoprotein can mediate profound tumor regression in a majority of patients, but the duration of these benefits is usually transient due to the emergence of drug-resistant tumors (3–5). Adoptive cell transfer (ACT) of autologous tumor-infiltrating lymphocytes can induce tumor regression, but complete responses are only observed in a subset of patients with melanoma (6, 7). Treatment of patients with combinations of these two therapeutic modalities is conceptually promising but requires a fundamental understanding of potential mechanisms by which the two strategies might cooperate to induce tumor regression.

The combined use of the selective BRAF^V600E inhibitor vemurafenib with ACT results in augmented antitumor response (8, 9), but investigations into the mechanisms by which the two therapeutic modalities may synergize have focused on augmentation of host immune responses in the presence of BRAF inhibition (10–14). There is a strong precedent established by Kaplan and colleagues (15) indicating an essential role for the effector cytokine IFNγ in the prevention of tumorigenesis (15, 16). In addition, recent observations indicate that the T-cell–expressed cytokines IFNγ and TNFα exert direct antiproliferative effects on cancer cells through induction of cellular senescence (17). This led us to ask whether T-cell effector cytokines produced by adoptively transferred cells could synergize with vemurafenib to induce growth arrest.

To evaluate this, we constructed a melanoma cell line model designated SB-3123, which enables investigation of the interplay between oncogene inhibition and adaptive immunity both in vitro and upon orthotopic transfer into syngeneic immunocompetent hosts. Using this model, we found that ACT cooperates with vemurafenib to cause enhanced regression of melanoma, but this effect was not dependent upon enhanced infiltration or function of endogenous or adoptively transferred cells within tumors. Instead, we observed that the T-cell effector cytokines IFNγ and TNFα synergized with vemurafenib to directly induce cell-cycle arrest of SB-3123 melanoma cells in vitro. The combination treatment regimen of vemurafenib and effector cytokines reduced proliferative capacity beyond single-agent treatment also in human melanoma–derived cell lines and was restricted to cancers bearing a BRAF^V600E mutation. This mechanism, thus, may not be exclusively model specific and could be applicable in a broad variety of BRAF^V600E-mutant melanoma tumors. Mechanistically, molecular profiling of treated SB-3123 indicated that the provision of vemurafenib promoted the sensitization of SB-3123 to the antiproliferative effects of T-cell cytokines. The unexpected finding that immune cytokines synergize with oncogene inhibitors to induce growth arrest has major implications for understanding cancer biology at the intersection of oncogenic and immune signaling and provides a basis for design of combinatorial therapeutic approaches for patients with metastatic cancer.

The SB-3123_p cell line was derived from spontaneously arising melanoma in a female Tyr::CreER; BRAF^CA/CA; PTEN^{lox4-5/lox4-5}; CTNNB1^{loxex3/loxex3} transgenic mouse (18–22). The tumor was initially divided into small pieces and then implanted onto C57BL/6 female mice. Growing tumors were serially implanted onto C57BL/6 mice and after the second in vivo passage were minced and seeded under tissue culture conditions to derive the SB-3123_p cell line. B16 (H-2^b) is a BRAF wild-type (WT) murine melanoma cell line, and A375 is a BRAF^V600E-mutant human melanoma cell line, both obtained from the NCI tumor repository. The BRAF^V600E-mutant human melanoma UACC-62 cell line was a gift from Dr. Susan Bates (Medical Oncology Branch, NCI, Bethesda, MD). MC38 (H-2^b) is a colon cancer murine cell line obtained from the NCI tumor repository. Mouse Melan-a cells were a gift from Dr. Thomas Hornyak (University of Maryland School of Medicine, Baltimore, MD). Patient-derived, pathology-confirmed melanoma cell lines used in this study were generated from patients with metastatic, pathology-confirmed melanoma receiving treatment under Institutional Review Board–approved clinical protocols in the Surgery Branch of the NCI. Informed consent was obtained from all subjects. Melanoma cell lines grew from enzymatically or mechanically dispersed cells or from 1- to 3-μm tumor fragments that were cultured in 24-well plates (Corning 3524), one fragment or 1 × 10⁶ cells/mL in 2 mL/well of RPMI-1640 medium (Lonza), supplemented with 10% heat-inactivated FBS (Hyclone, Defined) and 100 U/mL of penicillin, 100 μg/mL of streptomycin, and 10 μg/mL of Gentamicin (Lonza). The established cell lines grew as monolayer cultures. Genomic characterization of patient-derived melanoma cell lines was performed through exome sequencing, as previously described (22). The SB-3123, A375, B16, and UACC-62 cells were maintained in culture media composed of DMEM (Life Technologies) with 10% heat-inactivated FBS (Sigma), 1% GlutaMAX (Life Technologies), 1% (v/v) penicillin/streptomycin (Life Technologies), 1% Eagle's Minimum Essential Medium (MEM) nonessential amino acids (Life Technologies), 1% sodium pyruvate (Life Technologies), and 0.1% 2-mercaptoethanol (55 mmol/L; Life Technologies) in 5% CO₂ at a constant temperature (37°C) and humidity. Trophic factor–deficient media consisted of DMEM supplemented only with 1% GlutaMAX, 1% (v/v) penicillin/streptomycin, 1% MEM nonessential amino acids, 1% sodium pyruvate and 0.1% 2-mercaptoethanol. Melan-a cells were cultured in RPMI-1640 culture media (Life Technologies) with 5% heat-inactivated FBS, 0.1% phorbol 12-myristate 13-acetate (PMA; Sigma), 1% (v/v) penicillin/streptomycin and 1% GlutaMAX. All cell lines used were confirmed to be Mycoplasma free. No additional validation assay was performed.

Western blot analysis was performed using standard protocols. Whole-cell lysates were prepared using RIPA lysis buffer (Thermo Fisher Scientific). Proteins were separated by SDS-PAGE, followed by standard immunoblot analysis using phosphor-Erk 1/2, total Erk 1/2, PTEN (all from Cell Signaling Technology), GAPDH (Millipore), V5 (Life Technologies), and β-actin (Santa Cruz Biotechnology). The expression of gp100 was detected using α-PEP13 antisera (gift from Dr. Vincent Hearing, NCI, Bethesda, MD).

Cells were seeded in a 96-well flat-bottom plate at a density of 2.0 to 2.5 × 10³ cells per well and allowed to adhere for 24 hours in complete media. The following day, cells were treated with DMSO vehicle control and increasing concentrations of vemurafenib (PLX4032; Chemie Tek) in triplicate and cells were incubated for 72 hours. In all 8 × 8 matrix experiments, SB-3123 were treated with 0.1% BSA vehicle control, vemurafenib (0–16 μmol/L), mouse IFNγ (0.8–51.2 ng/mL), TNFα (0.08–5.12 ng/mL), human IFNγ (0.4–100 ng/mL), TNFα (0.04–10 ng/mL; PeproTech) and cells were incubated for 96 hours. Cells were analyzed using the CellTiter-Glo Luminescent Cell Viability Assay according to the manufacturer's protocol (Promega).

To generate antigen-specific CD8⁺ T cells for ACT, splenocytes from Pmel-1 (Thy1.1⁺) T-cell receptor (TCR) transgenic mice were depleted of erythrocytes by ACK lysis and cultured in complete media with 30 IU/mL of recombinant human IL2 (rhIL2; Novartis) in the presence of 1 μmol/L hgp100_25-33 of peptide and expanded for 5 days.

Full-length mouse gp100 cDNA was amplified from B16 F10 melanoma cells, and cDNA was cloned into the gamma-retrovirus vector pMSGV1 V5 IRES (internal ribosome entry site) Bsr (blasticidin S resistance) so that the V5 epitope tag was added to the C-terminus of gp100. To retrovirally transduce SB-3123, the retrovirus producer cell line 293gp was transiently transfected with pMSGV1 gp100 V5 IRES Bsr and pMDG that encoded the VSVG envelope gene. Transduced cells were selected by culture in Blasticidin S, and the gp100 expression was confirmed by qRT-PCR, Western blot analysis using anti-V5 tag antibody.

C57BL/6 female mice (n = 5 for all groups), ages 6 to 8 weeks and weighing approximately 19 to 20 g, were injected s.c. with 3 to 5 × 10⁵ SB-3123_p, B16, or SB-3123 melanoma cells in 100 μL of 1× PBS into the abdomen. Tumors were allowed to reach approximately 60 mm² before the initiation of treatment. Vemurafenib tablets were crushed using a mortar and pestle, and the powder was suspended at the desired concentration in an aqueous vehicle containing 2% Klucel LF (Hydroxypropylcellulose; Ashland) and adjusted to pH 4 with dilute HCL. Vemurafenib was administered uninterruptedly by oral gavage on a daily basis.

C57BL/6 female mice (n = 5 for all groups), ages 6 to 8 weeks, were injected s.c. with 3 to 5 × 10⁵ gp100-transduced SB-3123 cells in 100 μL of 1× PBS into the abdomen. Mice with established tumors were treated with i.v. injections of 2.5 × 10⁶in vitro–activated Pmel-1 (Thy1.1⁺) splenocytes for 5 days and vaccinated by i.v. injection of 2 × 10⁷ plaque-forming units of recombinant vaccinia virus encoding hgp100 (rVVhgp100) followed by 2 daily i.p. injections of rhIL2 (150,000 IU rhIL2 in ×1 PBS, for 3 days). Collectively, this treatment is designated as ACT. Treated mice received 600 cGy of sublethal irradiation before transfer of Pmel-1 CD8⁺ T cells.

All FACS antibodies were purchased from BD Biosciences, except antibody to mouse CD8α (eBioscience). A leukocyte activation cocktail containing phorbol myristate acetate and ionomycin (BD Biosciences) was used to stimulate T cells for intracellular cytokine staining. Flow-cytometry acquisition was performed on a BD LSR II flow cytometer.

After treatment, SB-3123 proliferation and cell-cycle analysis was measured by the APC-BrdUrd Flow Kit according to the manufacturer's protocol (BD Pharmingen). The following cell-cycle phases were determined as a percentage of the total population: sub-G₁ (apoptotic cells), G₁–G₀ (2n, BrdUrd-negative), S (2n to 4n, BrdUrd-positive), and G₂–M phase (4n, BrdUrd-negative). Flow-cytometry acquisition was performed on a BD Canto II flow cytometer.

The Loewe synergy analysis was performed on data shown in Supplementary Fig. S7. The analysis was carried out using the Combenefit software version 1.22 (http://www.cruk.cam.ac.uk/research-groups/jodrell-group/combenefit), which relies on the Loewe additivity model (24).

Total RNA (100 ng), extracted as previously described, was amplified using the Ovation Pico WTA System V2 (NuGEN), according to the manufacturer's instructions. Briefly, first-strand cDNA was synthesized using the SPIA-tagged random and oligo-dT primer mix in 10 μL reactions after denaturation and incubated at 65°C for 2 minutes, and priming at 4°C was followed by extension at 25°C for 30 minutes, 42°C for 15 minutes, and 77°C for 15 minutes. Second-strand cDNA synthesis of fragmented RNA was performed using DNA polymerase at 4°C for 1 minute, 25°C for 10 minutes, 50°C for 30 minutes, and 80°C for 20 minutes. 5′ double-stranded cDNA was used as the template for isothermal single-strand cDNA amplification following a cycle of DNA/RNA primer binding, DNA replication, strand displacement, and RNA cleavage at 4°C for 1 minute, 47°C for 75 minutes, and 95°C for 5 minutes in total 100 μL reaction. Samples were fragmented and biotinylated using the Encore Biotin Module (NuGEN) according to the manufacturer's instructions. Biotinylated cDNA was then hybridized to Mouse Gene 1.0 ST arrays (Affymetrix) overnight at 45°C and stained on a Genechip Fluidics Station 450 (Affymetrix), according to the respective manufacturer's instructions. Arrays were scanned on a GeneChip Scanner 3000 7G (Affymetrix). Global gene-expression profiles rank ordered by relative fold-change values were analyzed by using Gene set enrichment analysis software (Broad Institute, MIT). P values were calculated using the Student t test using Partek Genomic Suite after Robust Multiarray Average (RMA) normalization.

In vitro assays were repeated at least three times and in vivo studies at least two times. Mean comparisons were conducted by using an unpaired t test. For in vivo studies, the products of perpendicular tumor diameters were plotted as the mean ± SEM for each data point, and groups were compared using the Wilcoxon rank-sum test. The log-rank test was used to analyze survival curves. P values of ≤0.05 were considered statistically significant. All statistical analyses were calculated using Prism 5 GraphPad software (GraphPad Software Inc.)

The microarray data are available in the Gene Expression Omnibus (GEO) under accession number GSE62249.

To better model human melanoma tumors, we derived a syngeneic transplantable cell line from tumors arising in a BRAF^V600E-mutant autochthonous tumor model that has been previously described (25, 26). We observed that this model, consistent with its molecular architecture, was sensitive to vemurafenib treatment (Supplementary Fig. S1A and S1B). The resultant murine melanoma cell line (SB-3123_p) was transplantable into syngeneic hosts and exhibited key molecular characteristics consistent with BRAF^V600E melanoma (Supplementary Fig. S1C–S1H) as assessed by mutation-specific primers (Supplementary Fig. S2). ERK activity was inhibited in a vemurafenib dose-dependent manner in SB-3123_p cells at levels comparable with human BRAF^V600E-mutant melanoma cells, whereas it remained unaffected in WT BRAF-expressing B16 murine melanoma cells (Fig. 1A). The 50% inhibition concentration (IC₅₀) of SB-1323_p was similar to the highly sensitive A375 human melanoma cell line (Fig. 1B). Furthermore, the antitumor effect of vemurafenib was BRAF^V600E specific because treatment of SB-3123_p–bearing immunocompetent mice resulted in substantial inhibition of tumor growth (Fig. 1C and Supplementary Fig. S3A and S3B) in contrast with treatment of implanted B16 melanoma tumors (Supplementary Fig. S4A and S4B).

Objective tumor responses in patients with BRAF^V600E melanomas treated with BRAF inhibitors are short-lived, and tumors eventually progress due to the development of drug-resistant disease (26). This phenomenon was recapitulated in SB-3123_p melanomas because tumor recrudescence emerged in all mice continuously treated with vemurafenib following approximately 1 month of therapy (Fig. 1D). Thus, SB-3123 is a novel murine orthotopic melanoma model that recapitulates key molecular characteristics of a majority of human melanoma tumors, including initial sensitivity to vemurafenib therapy and emergence of drug resistance in vivo.

Recrudescence of SB-3123_p melanomas led us to inquire whether vemurafenib would synergize with ACT to induce more durable tumor regression. We therefore generated an SB-3123_p–derived cell line called SB-3123 that expresses the model self-antigen murine gp100 that is recognized by Pmel-1 TCR-transgenic CD8⁺ T cells (Supplementary Fig. S5). We found that ACT using tumor-reactive Pmel-1 CD8⁺ T cells cooperated with vemurafenib administration to delay the emergence of drug-resistant melanoma (Fig. 2A and Supplementary Fig. S6) and translated into a significant survival advantage (Fig. 2B and Supplementary Fig. S6), as has been demonstrated in other model systems (8, 9). Because the superior combined effect of concurrent vemurafenib administration has previously been attributed to augmentation of CD8⁺ T-cell–intrinsic antitumor immunity, we evaluated adoptively transferred and endogenous CD8⁺ T-cell responses in SB-3123 tumor–bearing mice treated or untreated with vemurafenib. Surprisingly, we found no significant differences in the function or number of adoptively transferred (Thy1.1⁺) or endogenous splenic and tumor-infiltrating CD8⁺ T cells (Fig. 3A–E). Thus, in the SB-3123 tumor model, ACT cooperates with vemurafenib to induce enhanced disease regression, but this is not attributable to the potentiation of systemic endogenous or transferred immune responses.

The dramatic outcome of concurrent vemurafenib and ACT in delaying SB-3123 tumor progression, despite causing no observable differences in T-cell recruitment and effector function in this model, suggested an alternative mode of cooperation between vemurafenib and ACT. Multiple reports have indicated that T-cell effector cytokines can directly induce tumor cell-cycle arrest (17, 28, 29). In addition, the antitumor effects of vemurafenib administration require intact host IFNγ production, as IFNγ depletion reduced tumor responsiveness to vemurafenib (13). We thus hypothesized that the provision of T-cell effector cytokines from adoptively transferred cells synergized with vemurafenib at a tumor cell–intrinsic level to promote reduced proliferation and superior tumor growth control. To test this hypothesis, we evaluated whether effector cytokines synergize with vemurafenib to inhibit SB-3123 growth in vitro. Although both vemurafenib and IFNγ/TNFα treatment inhibited tumor cell viability and proliferation when individually added to SB-1323 cell cultures at high doses, their effect was strikingly synergistic when IFNγ/TNFα and vemurafenib were combined at lower doses, which individually produced relatively minimal effects (Fig. 4A–D and Supplementary Figs. S7 and S8). Upon cell-cycle analysis, 84% of untreated SB-3123 cells were in S phase and 12% in G₁–G₀, indicative of rapid proliferative growth. This was mildly altered at selected suboptimal doses of cytokines and vemurafenib, but altered significantly when these agents were combined at identical doses, such that 37% of cells were arrested in G₁–G₀ and 31% were in S phase, with a significant increase in the G₁–S ratio upon combination treatment (Fig. 4E and F and Supplementary Fig. S9).

To evaluate whether the synergistic relationship of vemurafenib and effector cytokines in inducing SB-3123 growth arrest was dependent upon the presence of a tumor-specific BRAF^V600E mutation, we treated BRAF^WT (B16) and BRAF-mutant (SB-3123) murine melanoma cells with each agent alone or with a combination of vemurafenib and IFNγ/TNFα in an in vitro dose-escalation matrix. Although both cell lines were moderately sensitive to T-cell effector cytokines at the highest doses evaluated, vemurafenib inhibited cell growth exclusively in the BRAF^V600E-mutant SB-3123. Importantly, synergy between vemurafenib and IFNγ/TNFα in inhibiting cell growth was observed only in BRAF^V600E-mutant SB-3123 cells and not in BRAF-WT B16 cells. We concluded that the cooperative effect of vemurafenib and IFNγ/TNFα was mediated in a BRAF^V600E mutation–dependent manner (Fig. 5A and Supplementary Fig. S10).

To determine whether the combinatorial effect of vemurafenib and effector cytokine treatment was restricted to the SB-3123 model or represents a more generally applicable mechanism in human melanoma, we treated low-passage BRAF^WT and BRAF^V600E patient-derived melanoma cell lines in vitro and observed superior growth inhibition with combination treatment relative to single-agent treatment. This cooperation was seen only in the human BRAF^V600E-mutant melanoma cell lines but not in the BRAF WT cells (Fig. 5B and Supplementary Fig. S11). This indicates that the biologic cooperation between vemurafenib and effector cytokines is BRAF^V600E mutation–dependent and not restricted to murine melanoma.

We next investigated the molecular mechanisms by which vemurafenib and effector cytokines synergized to promote cell-cycle arrest of SB-3123. T-cell effector cytokines IFNγ and TNFα have been shown to have direct antiproliferative effects in tumors (17, 28, 29). Considering the surprising finding that in the SB-3123 model concurrent vemurafenib administration did not affect cytokine production in adoptively transferred CD8⁺ T cells, we were led to hypothesize that the provision of vemurafenib, rather than enhancing T-cell cytokine production, rendered SB-3123 more sensitive to T-cell cytokines. To gain insight into this hypothesis, we performed whole transcriptome analysis on SB-3123 cells grown in vitro under control, vemurafenib only, IFNγ/TNFα only, or combination vemurafenib + IFNγ/TNFα treatment conditions. Each treatment caused a unique gene-expression signature in SB-3123 with differential expression of greater than 1,000 transcripts relative to untreated cells (Fig. 6A; Supplementary Tables S1–S5, P < 0.01). The combination of vemurafenib + IFNγ/TNFα promoted the differential expression of 1,954 gene transcripts normalized to those of untreated cells that were distinct from transcripts induced in either of the single-agent treatment groups (Fig. 6B; Supplementary Tables S2–S5). This indicated that the combination treatment promoted genetic changes beyond the summation of changes induced by either single-agent therapy.

To investigate the genetic changes that occurred upon combination treatment beyond the changes induced by either vemurafenib or IFNγ/TNFα alone, we performed gene set enrichment analysis (GSEA) on the transcripts from the combination treatment group rank ordered by relative fold-change to transcripts from either IFNγ/TNFα-only or vemurafenib-only treatment groups [(Vem + IFNγ/TNFα vs. IFNγ/TNFα only) or (Vem + IFNγ/TNFα vs. Vem only)]. Interestingly, both IFN- and TNF-responsive gene sets showed significant positive enrichment with the transcriptional changes specific to the combination treatment, even relative to IFNγ/TNFα alone–induced gene-expression changes (Fig. 7A; Supplementary Table S6 and S7). Vemurafenib treatment alone did not enrich IFNγ or TNFα gene sets (Supplementary Table S8). Taken together, this suggested that the provision of vemurafenib in the context of IFNγ/TNFα treatment sensitized SB-3123 to cytokine-induced genetic changes.

The GSEA of the combination treatment normalized to IFNγ/TNFα single-agent therapy showed negative enrichment, to a significant degree, of multiple cell-cycle–promoting gene sets (Fig. 7B; Supplementary Table S7). This result is consistent with the observed increase in the G₁–S ratio and reduction in proliferative capacity in SB-3123 treated with vemurafenib + IFNγ/TNFα in vitro. It is also consistent with delayed SB-3123 recrudescence observed in vivo with vemurafenib and ACT cotreatment.

We next directly investigated whether vemurafenib indeed sensitized SB-3123 melanoma cells to IFNγ/TNFα–induced gene-expression changes, as the GSEA data suggested. First, we identified 77 genes that showed significant upregulation in IFNγ/TNFα–treated SB-3123 relative to control-treated cells (FC > 2.0, P < 0.01). We then tested for positive enrichment of these genes in SB-3123 cells receiving combination treatment relative to cells treated with IFNγ/TNFα alone. These analyses revealed positive enrichment with significant cytokine-responsive genetic changes, indicating an increased induction of the IFNγ/TNFα gene signature in SB-3123 upon vemurafenib + IFNγ/TNFα combination treatment (Supplementary Fig. S12).

To further demonstrate this, we evaluated under all treatment conditions the expression of the 20 most positively induced genes upon IFNγ/TNFα treatment relative to untreated SB-3123. Consistent with the enrichment results, we observed increased expression of cytokine-responsive genes in SB-3123 under vemurafenib + IFNγ/TNFα treatment conditions beyond the level of IFNγ/TNFα treatment alone (Fig. 7C).

We next performed a similar analysis to evaluate the expression of the most differentially expressed genes in SB-3123 following single-agent treatment with vemurafenib (Supplementary Table S10). Interestingly, the expression levels of the 20 most differentially expressed transcripts were not significantly different between vemurafenib single-agent treatment and vemurafenib + IFNγ/TNFα combination treatment (Fig. 7D). These data suggest that, although provision of vemurafenib sensitized SB-3123 to effector cytokine–induced genetic changes, there was no enhancement of vemurafenib-induced changes by combination with IFNγ/TNFα.

The lack of correspondence of traditional immunodeficient murine models of melanoma to clinical trial results (30, 31) and logistical challenges of solely using autochthonous tumor models to test synergistic responses between vemurafenib and ACT (including through characterization using defined in vitro conditions) led us to develop the SB-3123 BRAF^V600E PTEN^−/− murine melanoma transplantable model that emulates a common genetic landscape in human melanoma. The presence of a heterozygous BRAF^V600E mutation occurring in a PTEN-deficient background in SB-3123 recapitulates one of the most common genetic profiles in human melanoma, accounting for approximately 44% of patients (32). Compared with other recently developed murine melanoma syngeneic transplantable cell lines (9, 33), the SB-3123 model system closely resembles the human circumstance, given its high in vitro and in vivo sensitivity to BRAF inhibition at clinically relevant drug exposures (34). More importantly, this model system permits the development of tumor resistance in vivo with prolonged vemurafenib exposure, recapitulating the clinical experience in patients treated with BRAF inhibitors (Fig. 1). Furthermore, the emergence of vemurafenib-resistant melanoma at different rates of progression in the SB-3123 model (Fig. 1D) recapitulates the heterogeneity observed in times to disease progression among patients treated with BRAF inhibitors and served to preclinically model a more realistic efficacy endpoint.

Using this model, we found that vemurafenib cooperated with ACT to induce enhanced tumor regression that was independent of an effect on endogenous and transferred immune responses as reported by others (35, 36). A potential explanation for this discrepancy may relate to the high affinity interaction between the transduced melanoma differentiation antigen (MDA) gp100 in SB-3123 and adoptively transferred Pmel-1 TCR-transgenic CD8⁺ T cells. In previous work, the upregulation of MDA by BRAF inhibition is the basis of enhanced T-cell recognition. However, mouse gp100 is not expressed under its physiologic promoter, making it insensitive to MITF (Microphthalmia-associated transcription factor) regulation (37). To evaluate vemurafenib and ACT cooperation, we engineered SB3123 to express gp100, and the resulting supraphysiologic expression of gp100 may lead to increased interaction with the adoptively transferred Pmel-1 T cells. This increased antigen–T-cell interaction may potentially mitigate any enhanced cytokine production with concurrent vemurafenib treatment, which has been observed when targeting antigens expressed at physiologic levels.

Nevertheless, the enhanced tumor regression observed in SB-3123 treated with vemurafenib and ACT prompted us to explore alternative mechanisms by which these two modalities cooperated, namely, by evaluating the molecular responses in SB-3123 upon exposure to vemurafenib and T-cell effector cytokines as single agents or in combination. We observed in vitro that the combination of vemurafenib + IFNγ/TNFα was more effective than single-agent therapy exclusively in BRAF^V600E-mutant mouse and human cell lines. This led us to postulate that vemurafenib may be rendering BRAF-mutant tumors more susceptible to the effects of T-cell cytokines. This hypothesis was supported by evaluation of the expression levels of IFNγ/TNFα treatment–responsive genes in SB3123 as revealed by whole-transcriptome analysis. These experiments indicated an increased IFNγ/TNFα genetic signature in SB-3123 treated with the combination of vemurafenib + IFNγ/TNFα, even beyond that with IFNγ/TNFα treatment alone. GSEA of the gene-expression profile of vemurafenib + IFNγ/TNFα–treated SB-3123 revealed a negative enrichment of the cell-cycle progression genes, which is consistent with the reduced growth kinetics and distinct cell-cycle profile in SB-3123 treated with this combinatorial regimen. Taken together, we propose that the provision of vemurafenib sensitized SB-3123 to T-cell effector cytokines and mediated reduced growth in vitro.

IFNγ has been well characterized as having antiproliferative effects and is a critical factor in controlling the induction and proliferation of cancer (15, 16, 28, 29, 38, 39). Recently, it has been established that the antitumor effects of vemurafenib require intact host IFNγ production (13). Our studies offer a potential explanation as to how vemurafenib co-opts host IFNγ production to mediate tumor regression, namely through increasing tumor sensitivity to the effects of this cytokine.

Despite extensive genetic evidence indicating that provision of vemurafenib enhances SB-3123 sensitivity to a distinct IFNγ/TNFα–driven genetic signature, the specific cytokine-induced genetic changes that mediate the phenotypic observation of reduced SB-3123 growth kinetics remain to be completely elucidated. One specific effector cytokine–induced gene of interest may be Casp1, the gene encoding caspase-1, which is known to be induced by IFNγ in a STAT1-dependent manner and is involved in the induction of apoptosis in multiple tumor types, including breast and pancreatic cancers (38, 39). Casp1 expression is induced in SB-3123 to a greater level by vemurafenib + IFNγ/TNFα combination treatment relative to IFNγ/TNFα alone (FC = 2.12, P = 5.56E−05; Supplementary Table S11). Another cytokine-responsive gene expressed at a higher level upon combination treatment is the cell-surface death receptor Fas (FC = 1.8, P = 2.17E−04; Supplementary Table S11). Increased Fas expression may promote tumor cell apoptosis through an auto- or para-Fas:FasL interaction (40). It remains to be determined whether the differential expression of these select genes upon vemurafenib + IFNγ/TNFα treatment is sufficient to mediate the significant reduction in growth kinetics observed in SB-3123. The dramatic phenotypic effects observed with vemurafenib + IFNγ/TNFα provide a basis for further study of the intersection between immunologic and oncogenic signaling in cancer cells and for the rational design of combinatorial approaches to treat patients with melanoma.

As our findings have indicated that BRAF^V600E inhibition synergized with T-cell effector cytokines to reduce proliferation of BRAF-mutant human and murine melanomas, it would be very interesting to determine whether the combination of targeted inhibition of oncogenic pathways and T-cell–based immune therapies are synergistic in other cancer histologies. For example, the Btk inhibitor ibrutinib is in development for the treatment of multiple B-cell malignancies (41–44); however, its use as a single agent has been limited because of acquired resistance in the tumor (45, 46). CD19 chimeric antigen receptor immunotherapy is also being developed rapidly as a treatment for CD19-positive B-cell malignancies (47, 48). Perhaps the combination of targeted therapy with ibrutinib + CD19 CAR therapy could mediate superior tumor regressions in comparison with either single strategy for multiple B-cell cancer histologies.

The surprising observation that vemurafenib therapy synergizes with type I cytokines in a BRAF^V600E mutation–dependent manner to induce G₁–G₀ cell-cycle arrest has substantial therapeutic implications for the design of clinical trials, as it suggests that optimal synergy between these two therapeutic modalities may depend upon the concurrent long-term maintenance of BRAF oncogenic suppression when combined with T-cell–based immunotherapy. Even though our preclinical findings suggest that in the context of effective antitumor T-cell immunity only concurrent vemurafenib administration may translate into superior clinical efficacy, careful consideration should be given when combined with other immunotherapeutic modalities such as checkpoint inhibitors, which are known to improve endogenous T-cell cytokine production capacity (49). However, clinical data from a recent trial combining the anti–CTLA-4 antibody ipilimumab with vemurafenib strongly argue against a concurrent dosing strategy due to the occurrence of high-grade overlapping toxicities (50). Novel preclinical designs should be considered to better establish combinatorial schemes to lessen coinciding toxicities without necessarily compromising the overall effectiveness of BRAF inhibition, and the SB-3123 model provides an important experimental tool for such preclinical investigations.

No potential conflicts of interest were disclosed.

Conception and design: N. Acquavella, D. Clever, D.C. Palmer, Y. Ji, C.A. Klebanoff, R. Roychoudhuri, N.P. Restifo

Development of methodology: N. Acquavella, D. Clever, Z. Yu, M. Roelke-Parker, D.C. Palmer, L. Xi, K.-i. Hanada, F. Marincola, R. Roychoudhuri, N.P. Restifo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Acquavella, D. Clever, Z. Yu, M. Roelke-Parker, L. Xi, H. Pflicke, A. Gros, Z. Franco, F. Marincola, C.-C.R. Lee, M.W. Bosenberg, N.P. Restifo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Acquavella, D. Clever, L. Xi, H. Pflicke, Y. Ji, A. Gros, I.S. Goldlust, G.U. Mehta, C.A. Klebanoff, J.G. Crompton, M. Sukumar, J.J. Morrow, L. Gattinoni, E. Wang, F. Marincola, C.-C.R. Lee, M. Raffeld, M.W. Bosenberg, R. Roychoudhuri, N.P. Restifo

Writing, review, and/or revision of the manuscript: N. Acquavella, D. Clever, L. Xi, H. Pflicke, Y. Ji, G.U. Mehta, C.A. Klebanoff, J.G. Crompton, J.J. Morrow, L. Gattinoni, E. Wang, D.F. Stroncek, C.-C.R. Lee, M. Raffeld, R. Roychoudhuri, N.P. Restifo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Acquavella, H. Liu, N.P. Restifo

Study supervision: R. Roychoudhuri, N.P. Restifo

The authors thank J.R Wunderlich, S. Xi, and M. Hollingshead (Frederick, MD) for technical advice and thoughtful discussions, E. Tran and D. Jones (Bethesda, MD) for experimental help, T. Hornyak (Baltimore, MD) for providing Melan-a cells, V. Hearing (Bethesda, MD) for providing α-PEP13 antisera, and S. Bates (Bethesda, MD) for providing UACC-62 melanoma cells. The authors especially thank A. “Tito” Fojo, W.L. Dahut, and S.A. Rosenberg (Center for Cancer Research, NIH, Bethesda, MD) for continued support of this project.

The authors thank the Milstein Family Foundation for their generous support. This study was also funded by a gift from Li Jinyuan, Chairman of the Tiens Group, NIH-Center for Regenerative Medicine and by the Intramural Research Program of the NCI (ZIA BC010763), Center for Cancer Research, NIH (Bethesda, MD).