MEK Inhibition Modulates Cytokine Response to Mediate Therapeutic Efficacy in Lung Cancer

Activating mutations in the RAS pathway comprise one of the most common oncogenic abnormalities in cancer. Mutations in the RAS effector BRAF kinase are present in approximately half of all melanoma patients. Pharmacologic targeting of BRAF and the downstream MEK–ERK kinases results in significant benefit in BRAF-mutant melanoma patients (1, 2). This vulnerability is due to addiction of BRAF-mutant melanoma to the BRAF–MEK pathway (1, 3). RAS pathway mutations are also common in epithelial tumors, e.g., ∼30% of lung adenocarcinoma patients have mutations in KRAS. Unlike BRAF-mutant melanoma, RAS pathway inhibitors, such as MEK inhibitors (MEKi), have shown limited efficacy in lung cancer treatment (2, 4). This is likely because KRAS mutations are not always associated with KRAS pathway addiction (5, 6) as well as the redundancy in the function of downstream effector pathways for cancer cell survival, e.g., the MEK and PI3K pathways (7–10). New strategies therefore need to be developed for treating KRAS-mutant cancers.

A variety of approaches are being pursued to target the RAS pathway, including the development of inhibitors that directly target RAS proteins (10). An approach that has been pursued for some time is the simultaneous targeting of multiple arms of the RAS pathway, such as PI3K–AKT and MEK–ERK (8). However, there is concern about the toxicity of combinations of inhibitors of these pathways (2). Immunotherapeutics, especially those targeting checkpoint receptors on T cells, have revolutionized treatment of many cancer types. Interestingly, analysis of melanoma patient biopsies after BRAF and MEK inhibitor treatment indicate increased presence of tumor-infiltrating lymphocytes (TIL; refs. 11–13). Because patient benefit from immunotherapy is associated with high tumor expression of immune surveillance genes and T-cell infiltration (14–16), it has been proposed that MEKi may help generate a tumor microenvironment that enhances response to immunotherapy (17–20). Indeed, combining MEKi with immunotherapy (e.g., T-cell checkpoint blockade) in the preclinical setting substantially improved efficacy (17–20). Therefore, MEKi may find use in combination with immunotherapies in tumor types that are otherwise resistant to MEKi.

In previous studies, we found that NF-κB regulates tumor immune surveillance (21). We hypothesized that MEKi may activate NF-κB to generate a tumor microenvironment that is more permissive to immunotherapy. In both human tumors and in established cell lines, we show that MEKi enhances expression of NF-κB target genes. This was mediated by MEKi-induced upregulation of cell-surface expression of TNFR1, which strongly potentiated gene-expression responses by TNFα as well as genes jointly regulated by TNFα and IFNγ. Furthermore, MEKi cooperated with PD-1 blockade immunotherapy in curtailing lung tumor growth. A key and unexpected finding was the synergy between MEKi and TNFα + IFNγ in inducing cancer cell growth arrest and apoptosis across a broad array of human and mouse lung cancer cell lines. Furthermore, cancer cell knockout of TNFR1 impaired the antitumor activity of MEKi. Such cross-talk between MEKi and cytokine signaling pathways indicates a novel mechanism of action for an anticancer agent that could be used to enhance therapeutic efficacy against cancer types that are minimally responsive to MEKi.

Lung cancer and melanoma cell lines were provided by the Moffitt Lung Cancer Center of Excellence Cell Line Core and Dr. Keiran Smalley, respectively. Cell lines tested negative for Mycoplasma contamination (PlasmoTest, Mycoplasma Detection Kit from InvivoGen) and have been authenticated by STR analysis. All lung cancer and melanoma cell lines were maintained in RPMI-1640 with 10% fetal bovine serum and passaged for 2 to 4 times before use in experiments. NIH-3T3 cells were obtained from ATCC and maintained in DMEM with 10% fetal bovine serum. All cells were cultured with 10% fetal calf serum at 37°C in a 95% air, 5% CO₂ humidified incubator.

The COCTAIL library of 289 different agents was plated on cultures of A549 cells in 96-well plates (10,000 cells/well) using the Biotek robotic system. Twenty-four hours later, two different concentrations of each drug were added: 0.1 μmol/L and 1 μmol/L. TNFα and IFNγ were added to final concentrations of 0.2 ng/mL and 1 ng/mL, respectively. All conditions were in duplicate and the average was used in Fig. 2B. Cell supernatants were removed to determine amounts of secreted chemokines CCL5 and CXCL10 24 hours later. Cell viability was also determined using the CellTiter-Glo Reagent. Chemokine levels in A549 supernatant were detected using Bead-Based Multiplex Assays (Millipore Inc.) with Luminex technology.

All mice were bred and housed in the animal facility at Moffitt Cancer Center under specific pathogen-free conditions. 129S4/SvJaeJ mice were originally obtained from The Jackson Laboratory and were used for LKR tumor studies. Immunodeficient SCID mice (CB17.Cg-Prkdc^scidLyst^bg-J/Crl) were purchased from Charles River and used for A549 tumor growth studies. All animal experiments were approved by the Institutional Animal Care and Use Committee.

Cells were incubated for 5 minutes at room temperature with Fc block (BD Biosciences). Staining was performed in 1% BSA/PBS for 15 minutes at room temperature with Fluorochrome-conjugate monoclonal antibodies, and DAPI was added prior to analysis to assess viability. Flow-cytometric analysis was performed on an LSR II cytometer (BD Biosciences). Data were acquired using FACSDiva software (BD Biosciences) and analyzed using FlowJo software (Tree Star). The antibodies used from Miltenyi Biotec were CD120a (TNFR1)-APC, human (clone: REA252); CD120b (TNFαRII)-PE, human (clone: REA520). PE anti-mouse CD120a (TNFαR Type I/p55) antibody was from BioLegend.

Human TNFR1CRISPR/Cas9 knockout plasmids were purchased from Santa Cruz Biotechnology. Specifically, cells were kept at 40% to 80% confluency in a 6-cm plate, and replaced with fresh antibiotic-free growth medium prior to transfection. One microgram of plasmid DNA was incubated in 25 μL FuGENE Transfection Reagent (Promega) for 5 minutes and then added to 200 μL serum-free Opti-MEM for less than 20 minutes. After 72 hours of transfection, green fluorescent protein–positive (GFP+) cells were single-cell sorted into 96-well plates. Single clones were then expanded and screened by flow cytometry and Western blot analysis of TNFR1 expression.

Cells were harvested in logarithmic growth phase after being cultured for less than 2 weeks, washed once in injection medium (phenol-free DMEM supplemented with 2% FBS) and counted. Cells were injected subcutaneously into the right flank of mice and measured every 4 days. For LKR tumors, 10⁶ LKR cells were injected in phenol-free DMEM injection medium. The tumor volume was determined as length × length × width/2. Trametinib (1 mg/kg or 3 mg/kg) or vehicle was oral gavaged daily, αPD-1 (clone RMP1-14) or isotype control (Bio X Cell) was injected intraperitoneally (250 μg/mouse per injection). To deplete T cells, 300 μg/mouse anti-mCD4 (clone GK1.5), mCD8 (clone 2.43), or isotype control (Bio X Cell) were injected intraperitoneally 1 day prior to antitumor treatment and repeated every 3 days until experiment endpoint. For intratumoral cytokine injection, 100 μL of TNFα and IFNγ (0.5 μg each) or 100 μL PBS were injected directly into the tumor tissue for 3 consecutive days (days 11 to 13 after tumor cell inoculation), trametinib dosing was from days 10 to 14. Mice were sacrificed at the end of experiment or when tumor volume exceeded 2,000 mm³.

A549 cells (2.5 × 10⁵) were injected with PBS plus 1:1 volume Matrigel (Corning) percutaneously into the left lateral thorax in mice anesthetized with isoflurane. Fourteen days later, trametinib (1 mg/kg) or vehicle was oral gavaged daily for 14 days, and tumors were monitored using live imaging system. At the end of the study, lungs were collected for histologic analysis. For bioluminescence imaging (BLI), firefly luciferase-expressing A549 or LKR were used. The IVIS Imaging system was used to capture bioluminescence following i.p. injection of 4 mg luciferin (Gold Biotechnology).

For two group comparisons, Student t test or paired t test (two-sided) was applied. In time-course experiments that involve repetitive measurements, two-way ANOVA was applied to determine group differences and post hoc multiple-comparisons (Sidak or Tukey method, see below) were applied to individual time points. Significance for multiple condition experiments was determined using one-way ANOVA. For type I corrections for multiple-comparisons, the Sidak method was applied for comparisons between specific pairs of conditions, the Tukey method was applied for pair-wise comparisons among all conditions. For in vivo experiments, the Dunnett method was adopted to compare each treatment group with the control group. Correlations between two numerical metrics were determined using Pearson r. Cell-cycle data of each individual experiment were represented using a contingency table, χ² test with Bonferroni correction was applied to determine the significance. Differences between tumor growth curves were determined by two-way ANOVA. In the figures, significance was noted using *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant. All data shown in the bar graphs are the mean ± SD of at least three biological replicates. Statistical analysis was conducted using the Prism7 (GraphPad) software.

Inhibitors of BRAF (BRAFi, e.g., dabrafenib) and MEK (e.g., trametinib) prevent ERK activation and are the standard of care for the treatment of BRAF-mutant melanoma. To determine whether BRAFi/MEKi affects NF-κB signaling, we used RNA-sequencing data from pretreatment and on-treatment biopsies of patients in a recent study (22). We used an NF-κB gene-expression signature (21) to determine potential changes in NF-κB activity after BRAFi/MEKi treatment. Importantly, NF-κB pathway activity was significantly enhanced after treatment (Fig. 1A). Importantly, on-treatment NF-κB activity was significantly associated with depth of patient response to treatment (Fig. 1B). TNFα is a master activator of NF-κB, including in the tumor microenvironment, and previous studies have shown increased TNFα levels in melanoma patients undergoing BRAFi treatment (23–25). We hypothesized that MEKi may increase TNFα expression and/or TNFα-induced NF-κB activation, leading to elevated target gene expression. We tested this in vitro by using BRAF-mutant WM164 and 1205Lu human melanoma cell lines to determine changes in mRNA expression of TNF and the TNFα target gene TNFAIP3 (aka A20). Consistent with previously established autocrine regulation of TNFα, we found that TNFα enhanced expression of TNF in addition to TNFAIP3. Notably, MEKi trametinib enhanced TNFα-induced expression of both genes in these melanoma cell lines (Fig. 1C and D). These results suggest that MEKi potentiates TNFα signaling in melanoma cells, which may be linked to enhancement of NF-κB activity after BRAFi/MEKi treatment.

Unlike melanoma, MEKi have shown limited efficacy in lung cancer treatment (2, 4). With the goal of defining strategies to increase vulnerability of lung cancer to MEKi, we next determined whether a similar potentiation of TNFα induced gene expression by MEKi was evident in lung cancer cells. To broadly assess such a role for MEKi, we performed exome-wide gene-expression studies in the KRAS-mutant human lung adenocarcinoma cell line A549. Interestingly, MEKi trametinib slightly enhanced expression of several TNFα target genes, including TNF and TNFAIP3, but the highest levels of expression were achieved after combined MEKi and TNFα treatment (Fig. 1E; Supplementary Table S1). qRT-PCR confirmed ability of MEKi to enhance expression of TNF and TNFAIP3 in A549 and when combined with TNFα (Fig. 1F) and in the H2122 lung cancer cell line (Supplementary Fig. S1). TNFα and IFNγ synergize in regulating expression of a host of immune function genes, such as the chemokine CXCL10, which was found to be upregulated by TNFα + MEKi (Fig. 1E). TNFα + IFNγ synergistically enhanced CXCL10 expression, which was further enhanced in the presence of MEKi (Fig. 1G). The same results were obtained when expression levels of CXCL10 protein were determined (Fig. 1H).

We next determined whether MEKi stimulation of TNFα-induced gene expression also occurred in response to other anticancer agents. We assembled an in-house customized library of 289 targeted compounds that cover all major target classes, such as epigenetic enzymes, hedgehog, HSP90, and Notch, but has a stronger focus on protein and lipid kinases that reflects the current landscape of targeted agents in clinical use and development (26). This library consists of more than 70% of compounds that are either FDA approved or in clinical development (thus the name COCTAIL—Collection of Clinical Targeted Agents In Lung cancer; Supplementary Fig. S2; Supplementary Table S2). In addition, we incorporated target redundancy so that most targets are covered by several compounds. Both CXCL10 and CCL5 are synergistically regulated by TNFα + IFNγ. We used this library to identify anticancer agents that can enhance TNFα + IFNγ-induced expression of CXCL10 and CCL5 in A549. The combination of TNFα + IFNγ was used based on the reasoning that agents capable of enhancing the already high expression induced by these cytokines will yield the most robust hits. Using the basic scheme outlined in Fig. 2A, we determined the ability of two different concentrations (0.1 and 1 μmol/L) of each of these agents to increase TNFα + IFNγ-induced secretion of CXCL10 and CCL5 (Supplementary Table S2). Using a 2-fold cutoff for the ability to increase expression in one of the four tested conditions over TNFα + IFNγ alone, 16 agents were identified as potential hits. Remarkably, 13 of these agents were either MEK (including trametinib) or ERK inhibitors (Fig. 2B). The additional agents included two SMAC mimetics, which target cIAP proteins and activate TNFα signaling (27), and a phosphodiesterase inhibitor (Fig. 2B). We conclude the MEK/ERK inhibitors can function to enhance TNFα + IFNγ-induced signaling responses. Furthermore, the multitude of MEK/ERK inhibitors identified suggests that on-target effects of these drugs are responsible for the enhancement of TNFα + IFNγ-induced responses.

To investigate how TNFα and TNFα + IFNγ gene-expression responses might be enhanced by MEKi, we next determined its impact on the expression of the two TNFα receptors: TNFR1 and TNFR2. Interestingly, MEKi led to dramatic cell-surface upregulation of TNFR1, but not TNFR2, in a dose-dependent manner in A549 (Fig. 3A–C). However, the total levels of TNFR1 were not affected (Fig. 3C), suggesting that MEKi increases membrane localization of TNFR1. TNFα treatment itself did not enhance TNFR1 expression (Fig. 3A and B). In contrast to MEKi, cytotoxic agents and drugs targeting PI3K or HDACs failed to upregulate TNFR1 expression (Supplementary Fig. S3A). We next determined whether MEKi-induced increase in TNFR1 affected NF-κB activation by TNFα and IFNγ as determined by RelA phosphorylation and increase in nuclear translocation of RelA. MEKi alone activated NF-κB and its target genes TNF and TNFAIP3 in A549 (Supplementary Fig. S3B and S3C), but in combination with TNFα NF-κB activation was synergistically enhanced (Fig. 3D). Interestingly, the combination of MEKi + IFNγ also enhanced NF-κB activation, suggesting that MEKi enhancement of TNFR1 expression may also potentiate IFNγ responses (Fig. 3D). Finally, the highest activation of NF-κB was seen in the presence of MEKi and both cytokines.

MEKi also increased TNFR1 surface expression in multiple additional human lung cell lines such as H2122, H1437, HCC44, and PC9 (Fig. 3E) and the BRAF-mutant WM164 and 1205Lu human melanoma cell lines by both MEKi and BRAFi (Supplementary Fig. S3D). TNFR1 upregulation was seen in both KRAS-mutant (A549, H2122, HCC44, and H2009) and wild-type (H1437 and PC9) human lung cancer cell lines (Fig. 3E). We believe this likely reflects the common utilization of MEK–ERK signaling by cancer cells regardless of oncogene mutation. Similar results were also obtained in mouse lung cancer LKR, which harbors Kras mutation (Fig. 3E). We next tested three additional MEKi included in the above screening (cobimetinib, binimetinib/MEK162, and AZD8330) for their ability to inhibit pERK, upregulate TNFR1 expression, and induce gene expression (Supplementary Fig. S4A–S4C). In a dose-dependent manner, these MEKi inhibited pERK and upregulated cell-surface TNFR1 but not TNFR2 expression (Supplementary Fig. S4A and S4B). Moreover, as observed in Fig. 3C for trametinib, these MEKi did not affect total TNFR1 levels (Supplementary Fig. S4A). Importantly, we found a strong association between pERK inhibition and surface expression of TNFR1 and target gene expression (Supplementary Fig. S4D–S4F). As seen with trametinib, these MEKi also enhanced TNFα-induced TNF expression in a dose-dependent manner (Supplementary Fig. S4C). Furthermore, TNF expression strongly correlated with pERK inhibition (Supplementary Fig. S4E) and surface expression of TNFR1 (Supplementary Fig. S4F). We determined whether enhanced activation of MEK/ERK could conversely reduce TNFR1 expression. Importantly, ectopic expression of mutant KRAS in NIH-3T3 fibroblasts enhanced ERK activation and reduced cell-surface TNFR1 expression (Fig. 3F and G). These results indicate the ability of MEKi to enhance surface expression of TNFR1, which may in turn be responsible for the enhancement of TNFα-induced gene-expression responses.

To determine the role in MEKi-induced stimulation of gene-expression responses, TNFR1 was knocked out (KO) in A549 using CRISPR/Cas9 technology. TNFR1′s absence was confirmed by flow cytometry and Western blotting (Supplementary Fig. S5A and S5B), consistent with which TNFα-induced gene expression was diminished in TNFR1 KO A549 (Supplementary Fig. S5C). To determine the impact of TNFR1′s absence on TNFα and IFNγ gene-expression responses, we assessed the impact on CXCL10 as it is dually regulated by both cytokines (Fig. 1G). The absence of TNFR1 significantly reduced IFNγ and MEKi + IFNγ-induced CXCL10 expression, which was rescued by TNFR1 reexpression (Fig. 3H). Thus, TNFR1 mediates responses to IFNγ + MEKi, likely through MEKi upregulation of TNFα–TNFR1 autocrine signaling, which synergized with IFNγ. TNFα alone did not appreciably induce CXCL10 but a notable increase was seen when TNFα was combined with MEKi. TNFα + IFNγ induced high-level CXCL10 expression, which was further enhanced by MEKi in a TNFR1-dependent manner (Fig. 3H). These results suggest that upregulation of TNFR1 by MEKi plays a key role in enhancing TNFα and IFNγ target gene expression. To test whether MEKi may also directly affect the IFNγ pathway, we determined cell-surface IFNγ receptor 1 and 2 expression, which together make the IFNγ receptor heterodimer, as well as IFNγ-induced STAT1 phosphorylation in the presence of MEKi (Supplementary Fig. S5D–S5G). Unlike TNFR1, we did not see an increase in either IFNγ receptor chain expression (Supplementary Fig. S5D). Furthermore, MEKi did not affect levels of IFNγ-induced pSTAT1 across several IFNγ concentrations (Supplementary Fig. S5E). At 0.5 and 5 ng/mL IFNγ treatment at different time points, MEKi did not increase pSTAT1 levels but substantially increased IFNγ-induced CXCL10 expression (Supplementary Fig. S5F and S5G). We conclude that upregulation of both TNFα and IFNγ-induced gene expression in A549 is mediated by MEKi-induced increase in cell-surface TNFR1.

In addition to gene-expression functions, TNFα is a known and potent inducer of cell death. Furthermore, TNFα and IFNγ synergize in induction of cell death and cell-cycle arrest (28, 29). We next tested the possibility that MEKi-induced increase in TNFR1 expression may also affect cell death and cell-cycle arrest responses. Importantly, we observed that the growth-suppressive effects of MEKi were partly attenuated in TNFR1 KO A549, but the reexpression of TNFR1 in these cells resensitized them to MEKi (Fig. 4A). These results suggest that activation of autocrine TNFR1/TNFα signaling by MEKi could enhance growth suppression. We next tested the impact of exogenous addition of TNFα and IFNγ. TNFα alone and TNFα + IFNγ were found to modestly reduce viable cell numbers. As expected, MEKi reduced cell numbers; however, the combination of MEKi with TNFα + IFNγ resulted in the most reduction in viable cell numbers (Fig. 4B). Treatment with MEKi in the presence of both cytokines led to the largest percentage of G₁-phase cells and the lowest percentage S-phase cells (Fig. 4C) and the highest activation of apoptosis marker cleaved caspase-3 (CC3; Fig. 4D). These results suggest that reduction in viable cell numbers after MEKi and cytokine treatment is mediated through both cell-cycle arrest and cell death induction (also see sections below).

Utilizing TNFR1 KO A549, we next tested the in vivo role of TNFR1 signaling in MEKi antitumor response in an orthotopic lung tumor model using immunodeficient SCID (Prkdc^scid) and beige (Lyst^bg) mice. Unlike vehicle-treated mice, dosing of mice with 1 mg/kg trametinib was associated with minimal tumor burden (Fig. 4E and F). In contrast, no reduction in tumor burden was noticed after MEKi treatment in mice bearing TNFR1 KO tumors (Fig. 4E and F). Importantly, pERK was strongly inhibited by MEKi in KO tumors, suggesting that trametinib retains MEK targeting ability in these tumors (Fig. 4G). To longitudinally assess the impact of MEKi on tumor growth, we utilized BLI in parental and TNFR1 KO A549. Although WT tumor growth was inhibited, TNFR1 KO tumors continued to grow after MEKi treatment (Supplementary Fig. S6A and S6B). Together with in vitro findings, these results indicate a key role for TNFR1 signaling in the MEKi antitumor response. Furthermore, because mouse TNFα cannot signal through human TNFR1, the antitumor effects of MEKi are likely mediated by stimulation of autocrine TNFα signaling.

We next tested the impact of TNFα and IFNγ on MEKi growth suppression in the mouse lung cancer LKR cell line, which also underwent increase in TNFR1 cell-surface expression after MEKi treatment (Fig. 3E). As in A549 cells, the combination of MEKi with both TNFα and IFNγ led to the strongest reduction in cell numbers in LKR cells (Fig. 5A). However, individual cytokine treatment with IFNγ + MEKi decreased cell numbers, whereas TNFα + MEKi treatment increased cell numbers compared with MEKi alone (Fig. 5A). To better understand these findings, we next proceeded to define the individual and combined roles of TNFα and IFNγ in LKR cell-cycle and cell death regulation. However, TNFα + MEKi or TNFα + IFNγ + MEKi did not appreciably affect S-phase cells compared with MEKi alone (Fig. 5B). This was distinct from above results with A549, and suggests that decrease in LKR numbers after TNFα + IFNγ + MEKi (Fig. 5A) could be mediated by enhanced cell death.

To assess the role of apoptosis in the reduction of viable cells after cytokine and MEKi treatments, we determined the levels of CC3. CC3 comprises a 19 kd (p19) and a 17 kd (p17) fully activated form, both of which were detected in LKR cells. We found that in a dose-dependent manner, TNFα promotes the formation of the p19 fragment and a small amount of p17 (Fig. 5C). This increase was not evident after either MEKi or IFNγ treatment (Fig. 5C). Although the combination of TNFα or IFNγ with MEKi led to distinct and complex patterns of p19 and p17 activation, the combination of all three agents led to the highest level of p17 activation (Fig. 5C). To determine the degree of enhancement of CC3 after the combination treatment, we treated LKR cells with 10 nmol/L MEKi and a range of cytokine concentrations (2, 10, and 50 ng/mL). CC3 p17 generation was detected at the higher cytokine concentrations (10 and 50 ng/mL), consistent with reduction in viable cell numbers (Fig. 5D). However, CC3 p17 generation was dramatically enhanced at all three cytokine concentrations in the presence of MEKi (Fig. 5D), indicating that MEKi enhances sensitivity to cytokine-induced apoptosis. Collectively, these results suggest that cell death induction rather than cell-cycle arrest is the primary mechanism of cell number reduction in LKR cells after cytokine and MEKi treatment.

We next determined whether cytokine + MEKi effect on cell viability seen in vitro was also evident in vivo. A tumor stasis effect of MEKi was observed in subcutaneous (s.c.) LKR tumors at 1and 3 mg/kg daily dosing in an LKR (Supplementary Fig. S7). Owing to similarity of response at these two concentrations, further studies were performed at the 1 mg/kg dose. Consistent with in vitro results, upregulation of TNFR1 expression in tumor cells was also observed after MEKi treatment (Supplementary Fig. S8). We next determined whether direct intratumoral injection of cytokines could affect the MEKi antitumor effect. BLI imaging of LKR tumors was used to determine treatment impact on tumor cell viability. Because TNFα + IFNγ have the potential to make tumor cells more susceptible to T-cell killing, e.g., by upregulation of MHC expression, we specifically depleted T cells to better assess the tumor cell–intrinsic effects of cytokine treatment. Consistent with in vitro results, these results indicate that combining MEKi and cytokines results in the most significant loss of tumor cell viability (Fig. 5E). PD-1 blockade is a known inducer of IFNγ and TNFα secretion by T cells, including in the LKR model (30, 31). PD-1 blockade induced tumor regression in a subset of LKR tumor-bearing mice (Fig. 5F). However, we found that the combination of PD-1 blockade and MEKi treatment resulted in profound tumor regression in all treated mice (Fig. 5F). Although these results do not directly implicate roles for TNFα or IFNγ, they nonetheless suggest that immunotherapeutics known to induce their expression can synergize with MEKi in induction of antitumor responses.

We showed above that multiple human lung cancer and melanoma cell lines upregulate TNFR1 expression upon MEKi treatment. We next tested whether cytokine and MEKi effects on viable cell numbers observed in A549 and LKR were generalizable to the additional human lung cancer cell lines PC-9, H1437, HCC44, and H23. Viable cell numbers were reduced most substantially in the combined presence of TNFα + IFNγ and MEKi (Fig. 6A). Unexpectedly, TNFα and IFNγ did not repress cell numbers in melanoma cell lines (Fig. 6B), and cytokines combined with MEKi did not enhance the MEKi inhibitory effect (Fig. 6B and C). The combination of cytokines and MEKi synergistically reduced viable cell numbers as determined by the Bliss score in lung cancer lines (Fig. 6C). These results indicate that sensitivity to TNFα + IFNγ and synergy with MEKi occurs in lung but not in melanoma cells. As observed with trametinib, treatment of A549 and HCC44 with additional MEKi and TNFα + IFNγ also significantly reduced viable cell numbers in A549 and HCC44 (Supplementary Fig. S9).

Consistent with above results, cytokines combined with MEKi treatment led to a significant reduction in S-phase cells in the two of the three lung cancer cell lines tested (Fig. 6D). We next determined apoptotic sensitivity of these lung cancer cells to MEKi and cytokine treatment. In the highly MEKi-sensitive melanoma cell line WM164, a dramatic increase in CC3 generation was evident after 10 nmol/L MEKi trametinib treatment (Fig. 6E). In contrast, 10 nmol/L trametinib induced little CC3 in lung cancer cell lines, which was moderately increased by 100 nmol/L trametinib. Treatment of lung cancer cells with TNFα + IFNγ also induced little CC3 generation but the combination of cytokines and MEKi dramatically enhanced CC3 levels (Fig. 6E; see boxed lanes), consistent with decrease in viable cells seen after this combination treatment (Fig. 6A). In these cell lines, the overall CC3 levels, rather than specifically the p19 and p17 forms, were associated with loss of viable cells. Together with above findings in A549 and LKR, these results indicate that although lung cancer cells are largely resistant to apoptosis induction by MEKi, their apoptotic sensitivity is substantially increased in the presence of TNFα + IFNγ.

The studies presented here demonstrate a novel and unexpected link between MEK inhibitors and TNFα signaling in lung cancer cells. We show that the upregulation of cell-surface TNFR1 expression is a general and widespread effect of MEK inhibition as it was evident across an array of human and mouse cancer cells from different tumor types and with distinct driver oncogene mutations. The upregulation of TNFR1 not only enhanced TNFα-induced responses but also those jointly controlled by TNFα and IFNγ. Crucial among them were gene expression, cell-cycle arrest, and cell death. In multiple human lung cancer cell lines, we show that the combination of cytokines and MEKi induced the highest levels of cell-cycle arrest and apoptosis. In vivo studies performed with human A549 showed that TNFR1 deficiency attenuated the MEKi antitumor response. Interestingly, TNFα and IFNγ did not induce growth suppression or enhance MEKi-induced growth suppression in melanoma cells. The underlying reason for this difference is not presently clear but may indicate intrinsic resistance of melanoma cells to growth suppression and cell death induction by these cytokines. A recent study showed relatively low trametinib treatment response (12%) in non–small cell lung cancer with KRAS mutations (4). Our results suggest that combining MEKi with agents that promote TNFα and IFNγ expression, such as checkpoint blockade or T-cell adoptive cell therapy, may help achieve greater benefit in lung cancer patients. In this setting, distinct mechanisms may cooperatively promote antitumor responses through synergistic increase in expression of immune function genes, and most important, the potentiation of growth suppression and cell death. Furthermore, TNFα and IFNγ expression and/or pathway activation could provide a predictive and prognostic biomarker of MEKi treatment response.

Our results indicate that a crucial aspect of cross-talk between MEKi and TNFα/IFNγ is the augmentation of cell death, which is mediated by caspase-induced apoptosis. At trametinib concentrations achieved in patients (∼10–12 nmol/L; ref. 32), this agent essentially induced no CC3 activation in lung cancer cells. However, at the same concentrations, trametinib combined with TNFα + IFNγ induced robust CC3 activation coincident with the loss of viable cells. In vivo studies indicate that high levels of these cytokines (e.g., by cytokine injection or potentially induced by PD-1 blockade) enhance tumor regression compared with individual treatments. This is consistent with our findings showing a cytokine dose-dependent increase in the apoptotic marker CC3, suggesting that synergy between MEKi and cytokines will be enhanced in proportion to cytokine levels. In all lung cancer cell lines tested, we have observed that the combination of MEKi with both TNFα and IFNγ is required for maximal loss of viable cells. However, several aspects of these findings require further investigation: first, in LKR cells, these two cytokines had distinct effects on CC3 generation individually and when combined with MEKi. Thus, the contribution of individual cytokines and synergy between them in apoptosis induction with MEKi requires further study in LKR as well as in additional cell types in order to better define underlying mechanisms. Second, we have surmised that increase in TNFα responses through TNFR1 upregulation enhance gene expression, cell-cycle arrest, and cell death responses jointly regulated by TNFα + IFNγ. However, it remains to be determined whether MEKi also affects the IFNγ pathway, irrespective of effect on the TNFα pathway. Although MEKi did not affect STAT1 activation by IFNγ in A549, it is nonetheless possible that more subtle cross-talk exists between MEKi and the IFNγ pathway. Finally, TNFR1 deficiency in A549 led to partial resistance to MEKi. Although exogenous cytokine addition universally synergizes with MEKi to reduce lung cancer cell viability, it remains to be determined whether TNFR1 deficiency is also sufficient for conferring resistance to MEKi treatment. The direct systemic administration of TNFα and/or IFNγ is expected to be quite toxic. Type 1 IFN (e.g., IFNα) share many key features with type II IFN (IFNγ), and importantly, IFNα has been in clinical use for cancer treatment for many years. Interestingly, we found that similar to IFNγ, IFNα also synergized with TNFα and MEKi in enhancing gene expression and decreasing cell viability (Supplementary Fig. S10A and S10B). Although this combination needs to be first tested in preclinical models, if effective, it can be explored in the clinical setting as a novel combination treatment for cancer.

Sensitivity to TNFα induced apoptosis is controlled by the balance of prodeath and prosurvival pathways. NF-κB functions in suppressing both TNFα and IFNγ induced apoptosis and necroptosis cell death pathways (33, 34). However, MEKi enhances NF-κB activation while concomitantly promoting cell death. One possibility is that the magnitude of cell death pathway enhancement by MEKi and cytokines is such that it cannot be controlled by NF-κB prosurvival functions. In human lung cancer cell lines, TNFα and IFNγ cooperatively promoted cell-cycle arrest with MEKi. Therefore, loss of viable human lung cancer cells after combined treatment with cytokines and MEKi likely results from both cell-cycle arrest and cell death. In contrast, MEKi combined with cytokines did not enhance cell-cycle arrest in mouse LKR cells where increase in cell death appears to be the dominant pathway. Previous studies showed that these cytokines synergistically induce cancer cell growth suppression through cell-cycle arrest and senescence induction that is mediated by the CDK/Rb pathway (28). Although we did not observe a clear cell-cycle arrest phenotype in human lung cancer cell lines after treatment with TNFα and IFNγ, it is possible that MEKi-induced increase in cytokine signaling enhances growth suppression through the CDK/Rb pathway.

The mechanism by which MEKi may enhance cell-surface TNFR1 expression is not clear. Our results indicate that total protein expression of TNFR1 was not affected by distinct MEKi (Fig. 3C; Supplementary Fig. S4A). Consistent with this, we did not detect any substantial change in TNFR1 mRNA expression after MEKi treatment (Supplementary Fig. S11). These results indicate that enhancement of surface expression is not due to increased de novo synthesis but rather due to increased cell-surface localization of TNFR1. Although the mechanism by which MEKi may enhance cell-surface TNFR1 expression is not clear, a previous study implicated a role for ERK sites in the TNFR1 transmembrane domain in preventing membrane localization (35). To investigate whether MEKi enhances cell-surface TNFR1 expression through this mechanism, we mutated two candidate sites (T244A and S278A) in TNFR1 and reexpressed in TNFR1 knockout A549 and PC-9. Although TNFR1 expression varied among these cell lines, we did not observe differences in baseline or MEKi-induced membrane expression of wild-type versus mutated receptors (Supplementary Fig. S12). These results suggest that either individual mutation of these sites is not sufficient to affect membrane localization or that potentially distinct MEKi-induced mechanisms may control TNFR1 surface expression.

One of our key findings is that MEKi-induced upregulation of TNFR1 enhances TNFα-induced NF-κB activation and TNFα target gene expression, including immune surveillance genes such as T-cell chemokines that are jointly regulated by TNFα and IFNγ. The increase in IFNγ-induced gene expression is likely to be mediated by increase in MEKi-induced cell-surface TNFR1 expression, which enhances TNFα autocrine signaling and synergy with IFNγ. Our findings suggest that the upregulation of TNFR1 expression may provide a mechanistic explanation for BRAFi/MEKi mediated increase in tumor immunogenicity seen in melanoma patients (11–13). Thus, the increase in NF-κB activity evident after BRAFi/MEKi treatment could be mediated through a greater response to TNFα and IFNγ following increase in TNFR1 expression. Because loss of viable cells was not increased by cytokines and MEKi in melanoma cells, it is possible that the combined immune stimulatory effect of MEKi and cytokines augments antitumor activity. Such differential impact of MEKi in melanoma versus lung tumor models requires further study. Previous studies have implicated the RAL-TBK1 signaling arm of RAS in expression of proinflammatory cytokines with tumor-promoting functions (e.g., IL6; ref. 36). Based on our findings, it will also be interesting to determine whether potential cross-talk between TBK1 and/or other oncogenic pathways and TNFα/IFNγ also affects gene expression and cell death pathways.

E.B. Haura reports receiving commercial research grants from Forma Therapeutics and Incyte Pharmaceuticals and is a consultant/advisory board member for Janssen Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Xie, R. Madan-Lala, W.D. Cress, U. Rix, A.A. Beg

Development of methodology: M. Xie, R. Madan-Lala, A.A. Beg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Xie, H. Zheng, R. Madan-Lala, N.T. Gimbrone, F. Kinose, S.A. Blackstone

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Xie, R. Madan-Lala, N.T. Gimbrone, Z. Chen, S.A. Blackstone, K.S.M. Smalley, W.D. Cress, E.B. Haura, A.A. Beg

Writing, review, and/or revision of the manuscript: M. Xie, R. Madan-Lala, Z. Chen, K.S.M. Smalley, W.D. Cress, E.B. Haura, U. Rix, A.A. Beg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Xie, H. Zheng, W. Dai, Z. Chen, E.B. Haura, A.A. Beg

Study supervision: M. Xie, A.A. Beg

We thank colleagues from H. Lee Moffitt Cancer Center: Drs. Jae-Young Kim for providing cell lines, reagents, and technical support, and Dr. Jose Conejo-Garcia for thoughtful comments on this manuscript. We thank Dr. Dung-Tsa Chen and Ram Thapa for help with statistical analysis, and Dr. Manali S. Phadke for help with melanoma cell lines. This work was supported by funds from the H. Lee Moffitt Cancer Center's Miles for Moffitt, Lung Cancer Center of Excellence, Prelude to a Cure, and the DoD Lung Cancer Research Program (grant #LC140306 to A.A. Beg), the Moffitt Skin SPORE P50 CA168536 (to A.A. Beg and K.S.M. Smalley), and a James and Esther King Biomedical Research Program Grant (5JK06) from the Florida Department of Health (to N.T. Gimbrone and W.D. Cress). We would like to acknowledge the Molecular Genomics, Cancer Informatics, Tissue Core, Analytic Microscopy, and flow cytometry shared facilities at H. Lee Moffitt Cancer Center, an NCI designated Comprehensive Cancer Center supported by NIH P30-CA076292.

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.