Context-Dependent Immunomodulatory Effects of MEK Inhibition are Enhanced with T-cell Agonist Therapy

Abstract

MEK inhibition (MEKi) is proposed to enhance antitumor immunity but has demonstrated mixed results as an immunomodulatory strategy in human clinical trials. MEKi exerts direct immunomodulatory effects on tumor cells and tumor-infiltrating lymphocytes, but these effects have not been independently investigated. Here we modeled tumor-specific MEKi through CRISPR/Cas-mediated genome editing of tumor cells (MEK1 KO) and pharmacologic MEKi with cobimetinib in a RAS-driven model of colorectal cancer. This approach allowed us to distinguish tumor-mediated and tumor-independent mechanisms of MEKi immunomodulation. MEK1 KO tumors demonstrated upregulation of JAK/STAT signaling; enhanced MHCI expression, CD8⁺ T-cell infiltration and T-cell activation; and impaired tumor growth that is immune-dependent. Pharmacologic MEKi recapitulated tumor-intrinsic effects but simultaneously impaired T-cell activation in the tumor microenvironment. We confirmed a reduction in human peripheral lymphocyte activation from a clinical trial of anti-PD-L1 (atezolizumab) with or without cobimetinib in biliary tract cancers. Impaired activation of tumor-infiltrating lymphocytes treated with pharmacologic MEKi was reversible and was rescued with the addition of a 41BB agonist. Collectively, these data underscore the ability of MEKi to induce context-dependent immunomodulatory effects and suggest that T cell–agonist therapy maximizes the beneficial effects of MEKi on the antitumor immune response.

Introduction

The mitogen-activated protein kinases (MAPK) pathway is one of the most frequently dysregulated signaling cascades in human cancers, and has been targeted with inhibitors of mitogen-activated protein kinase kinases (MEK1/2) (1–3). Although multiple MEK inhibitors have been developed and can sometimes induce rapid tumor regression, responses to such therapies are rarely durable due to the acquisition of clones containing mutations that confer resistance to oncogene-targeted therapies (4, 5). Although immune checkpoint inhibitors have revolutionized the treatment of cancer over the past decade, only approximately 15% of patients experience clinical benefit. This highlights the urgent need to develop novel drug combinations that can extend the benefit of immune checkpoint inhibitor therapy to immune-resistant tumor types (6). In theory, the use of oncogene-targeted agents in combination with systemic immunotherapies may offer the potential to transform short-lived responses into more durable clinical responses (7). Multiple combinations of targeted therapies and systemic immunotherapies are under preclinical and clinical investigation (8).

Several preclinical studies demonstrate the potential of systemic MEK inhibitors to enhance the efficacy of existing immunotherapeutic agents. These studies show MEK inhibitors can increase tumor cell immunogenicity through an increase of MHC class I and antigen expression on tumors (9–14), reduce immunosuppressive cell populations (15–17), and protect T cells against chronic TCR-driven apoptosis (10). The compatibility of MEK inhibition (MEKi) with immune checkpoint inhibitors is supported by multiple studies using preclinical models of KRAS-mutant colorectal cancer (CRC) (10, 12, 15), triple-negative breast cancer (13), and BRAF^V600E mutant melanoma (11). Collectively these preclinical studies infer a potential for synergy between MEKi and immune checkpoint inhibitors on the basis of depth and durability of immune responses and positive immune changes in the tumor microenvironment with MEKi.

Despite encouraging preclinical findings, MEKi has disappointed as an immunomodulatory therapeutic strategy when administered in combination with immune checkpoint inhibitors in some human clinical trials. Although human serial biopsy studies of MEK inhibitors confirm the ability of MEK inhibitors to enhance CD8⁺ T-cell tumor infiltration in patients (18), an increase in CD8⁺ T-cell infiltration has not translated into meaningful clinical activity in combination with immune checkpoint inhibitor therapy. In immune-resistant tumor types in which inhibitors of programmed death-ligand 1 (PD-L1) have little or no activity as monotherapy, including mismatch-repair proficient CRC and biliary tract cancers, combining the MEK inhibitor cobimetinib with the PD-L1 inhibitor atezolizumab did not show compelling tumor response rates (19, 20). Likewise, in BRAF^V600 wildtype cutaneous melanoma, an immune-sensitive tumor type for which immune checkpoint inhibitors have clinical activity as monotherapy, the combination of cobimetinib and atezolizumab did not improve upon the activity of an immune checkpoint inhibitor as monotherapy in a large phase 3 clinical trial (21). Despite MEKi demonstrating activity in prior studies in this disease subset (22), objective response rates were numerically lower with the combination of MEKi plus PD-L1 inhibition (PDL1i) (26%) versus the checkpoint inhibitor monotherapy arm (32%), raising the possibility that MEKi might impair the T-cell responses that immunotherapy seeks to harness. These data highlight that the effects of MEKi on antitumor immune responses are complex and poorly understood, and may be context dependent (9, 21–24). Thus, a deeper mechanistic understanding of how MEK inhibition impacts antitumor immunity is needed to develop rational and effective therapeutic combinations.

Here, we utilize the RAS-driven CT26 CRC model to investigate the tumor-intrinsic and tumor-extrinsic immunomodulatory effects of MEK1 inhibition using CRISPR/Cas-mediated genome editing of tumor cells and systemic MEKi with the MEK1 inhibitor cobimetinib. We show that MEK1 ablation in tumor cells led to enhanced T-cell infiltration and activation, whereas systemic MEKi impaired T-cell activation. In addition, MEK1 KO in CT26 cells resulted in enhanced survival in an immune-dependent context, but the primary effect on tumor growth with systemic MEKi was driven by the inhibition of oncogenic signaling rather than by enhanced antitumor immunity. Finally, we found that an 4-1BB agonist reversed MEKi-mediated T-cell immunosuppression, leading to significant improvements in survival.

Methods

Cell Culture

The MCF-10A cell line (ATCC Cat# CRL-10317, RRID:CVCL_0598) was maintained in DMEM/F12 (1:1) (Thermo Fisher Scientific, cat. #11320033) supplemented with 5% horse serum (Thermo Fisher Scientific, cat #26050088), 20 ng/mL EGF (Sigma-Aldrich, cat. #11376454001), 10 μg/mL insulin (Thermo Fisher Scientific, cat. #12585014), 0.5 μg/mL hydrocortisone (Sigma-Aldrich, cat. #H0888) 0.1 μg/mL cholera toxin (Sigma-Aldrich, cat. #C8052), and 1% Penicillin-Streptomycin (Thermo Fisher Scientific, cat. #10378016) (25). Prior to in vitro experiments, exponentially growing cells were switched to EGF-free conditions for 24 hours. Cells were then washed with HBSS twice and seeded in DMEM/F12 medium without EGF, using 1% charcoal dextran-treated fetal bovine serum (FBS) (Thermo Fisher Scientific, cat. #12676029), insulin at 10 μg/ml, hydrocortisone at 0.5 μg/ml, and cholera toxin at 0.1 μg/ml, at a density of 1×10⁵ cells/well of a 6-well tissue culture dish on day 0. On day 1, 1uM of cobimetinib or vehicle was added, followed by 50ng/ml of recombinant human IFNγ (PeproTech, cat. #AF-300-02) and the given concentration of EGF. On day 3, cells were trypsinized, washed, and stained for flow cytometry analysis. A list of flow cytometry antibodies, isotopes, and concentrations used for profiling is listed in Supplementary Table S1.

CT26 cells were purchased from ATCC (ATCC Cat# CRL-2638, RRID:CVCL_7256) and maintained in RPMI (Thermo Fisher Scientific, cat. #11875–085) supplemented with 10% FBS (Gemini, cat. #100–106), 1% penicillin/streptomycin (Thermo Fisher Scientific, cat. #15140–122) and 0.5% L-glutamine (Thermo Fisher Scientific, cat. #25030–081). Panc02 cells (NCI-DTP Cat# PAN 02, RRID:CVCL_D627) were obtained from prior Jaffee laboratory stocks and maintained in DMEM (Thermo Fisher Scientific, cat. #11-320-033) supplemented with 10% FBS (Gemini, #100–106), 1% L-glutamine (Thermo Fisher Scientific, cat. #25030–081), and 0.5% penicillin/streptomycin (Thermo Fisher Scientific, cat. #15140–122).

Cell lines were tested for Mycoplasma every 3 months in accordance with laboratory policy. MEK1 KO CT26 cells were periodically tested for continued MEK1 absence by western blot analysis. Prior to in vivo implantation, all cell lines were cultured from frozen stocks and passaged 3 times before use.

Drug Preparation

Clinical grade cobimetinib (GDC-0973, XL-518) was manufactured by Genentech, Inc. and acquired from an outpatient clinical pharmacy. A 10mM Cobimetinib stock was made for in vitro experiments by dissolving one 20mg tablet in vehicle consisting of 20% DMSO and water. For in vivo experiments, a 1.9mM cobimetinib stock solution was made by dissolving one 20 mg cobimetinib tablet in vehicle consisting of 20% DMSO and water. Each mouse received 200ul of cobimetinib solution (approximately 7.5 mg/kg of cobimetinib) by intraperitoneal injection.

For in vitro experiments, ruxolitinib was purchased from Selleck Chemical (cat. #S1378). A stock concentration of 10mM was prepared by reconstituting 5mg of drug in 1.632mL of DMSO. Drug was added at the indicated concentration to CT26 MEK1 KO cells 48 hours prior to flow cytometry analysis. For in vivo experiments, ruxolitinib (phosphate salt) was purchased from LC Laboratories (cat. #INCB018424). A 15mg/ml stock was prepared in vehicle consisting of 15% DMSO, 20% captisol by volume and 58nM citrate buffer. Each mouse received 100ul (60mg/kg) by oral gavage.

Anti-mouse PD-1 (Bio X Cell, cat. #BE0146, Clone RMP1–14) or rat IgG2a isotype control (Bio X Cell, cat. #BE0089, Clone 2A3) was prepared in PBS at a concentration of 2.5mg/mL (0.25mg/mouse or approximately 10mg/kg). Anti-mouse 41BB (Bio X Cell, cat. #BE0239, Clone 3H3) or rat IgG2a isotype control (Bio X Cell, cat. #BE0089, Clone 2A3) was prepared in PBS at a concentration of 0.25mg/ml (0.025mg/mouse or approximately 1mg/kg). Each mouse received 100ul of antibody or isotype by intraperitoneal injection based on the treatment schemes outlined below.

CT26 MEK1 KO generation

The CRISPR-Cas 9 system was used to knockout MEK1 (MAP2K1) in the CT26 cells. Guide RNA sequences were chosen from the GeCKov2 library pool (MGLibA_30220, MGLibA_30222, MGLibB_30213) (26). Vectors were created by cloning the guide RNAs into the lentiCRISPR v2 one vector system backbone (Addgene, cat. #52961). 1×10⁶ CT26 cells were plated in T25 flasks in antibiotic-free media 24 hours prior to transfection. Cells were then transiently transfected with the vectors using lipofectamine (Thermo Fisher Scientific, cat. # L3000008) according to the manufacturer’s instructions. Post transfection, CT26 cells were recovered for 24 hours after which they were selected using puromycin (Sigma-Aldrich, cat. #P8833) at 40ug/ml for 48 hours. Post selection, the cells were single-cell diluted and individual colonies were sequenced at the locus to identify clones with insertions/deletions at the cut site. Several clones were then analyzed by western blot to verify complete knockout of MEK1. The clone selected for in vivo experiments was generated using the guide sequence MGLibA_30222.

Western Blot Analysis

CT26 WT cells and individual MEK1 KO clones were lysed in RIPA buffer (Sigma-Aldrich, cat. #R0278) with added 1μM DTT, 1μM PMSF, and 1:100 protease/phosphatase inhibitor cocktail (Cell Signaling, cat. #5872S) and quantified by BCA (Thermo Fischer Scientific, cat. #23227). 100 μg of protein was run in 4–12% Bis-Tris gels under denaturing conditions. The LiCor Odyssey developing and imaging system was used, and the following primary antibodies were diluted in Odyssey Blocking Buffer (TBS) (cat. #92760001) plus 0.2% Tween® 20 at the indicated concentrations: Abcam recombinant MEK1-specific monoclonal rabbit antibody (cat. #ab32091, 1:2000), Cell Signaling phosphor-p44/42 MAPK–specific (Erk1/2 Thre202/Tyr204) polyclonal rabbit antibody (cat. #9101, 1:1000), Cell Signaling vinculin-specific polyclonal rabbit antibody (cat. #4650, 1:500), Cell Signaling β-actin-specific monoclonal rabbit antibody (cat. #4970, 1:1000), Cell Signaling Jak1-specific monoclonal rabbit antibody (cat. #3344, 1:1000), and Cell Signaling Stat1-specific polyclonal rabbit antibody (cat. #9172, 1:1000). Membranes were blocked with Odyssey blocking buffer TBS for 1hr and then incubated with primary antibody solution overnight at 4°C with gentle shaking. Membranes were washed 3 times with 1X TBS-T and subsequently incubated with the secondary antibody 800CW donkey anti-rabbit IgG (1:10,000) in Odyssey Blocking Buffer (TBS) plus 0.2% Tween® 20 for an hour at room temperature with gentle shaking. Membranes were protected from light during incubation with secondary antibody. Membranes were subsequently washed 3 times with 1X TBS-T and imaged with the Odyssey® imaging system.

Tumor Treatments and Tumor Measurements

For CT26 WT vs MEK1 KO experiments, adult female BALB/c mice (Jackson, #000651) (or RAG1^−/− BALB/c mice for survival experiments; Jackson #003145) were inoculated with 5×10⁵ CT26 WT colon cancer cells or 5×10⁵ CT26 MEK1 KO cells into the lower left flank at 6–8 weeks of age. Mice were sacrificed and tissues were collected 21 days after tumor implantation. For ruxolitinib experiments with CT26 WT vs MEK1 KO cells, adult female BALB/c mice were inoculated with 5×10⁵ CT26 WT colon cancer cells or 5×10⁵ CT26 MEK1 KO cells into the lower left flank at 6–8 weeks of age. Mice were treated with ruxolitinib (60mg/kg) or vehicle by oral gavage once daily beginning on day 1. Mice were sacrificed and tumors were processed 21 days after tumor implantation.

For systemic MEKi in vivo experiments with CT26 cells, adult female BALB/c mice (or RAG1^−/− BALB/c mice for survival experiments) were inoculated with 5×10⁵ CT26 WT cells into the lower left flank at 6–8 weeks of age. For systemic MEKi in vivo experiments with Panc02 pancreatic cancer cells, adult male C57BL/6 mice were inoculated with 5×10⁶ Panc02 WT cells into the lower left flank at 6–8 weeks of age. To mimic the immediate effects seen on the tumor cells with the MEK1 KO cells, mice were treated on day 1 with cobimetinib on Monday, Wednesday, and Friday (MWF), whereas the control mice received vehicle. For cobimetinib and agonist combination treatment, adult BALB/c mice were inoculated with 5×10⁵ CT26 WT colon cancer cells into the lower left flank at 6–8 weeks of age. Tumors were left to establish for 7 days post-injection, at which point they were palpable but not clearly measurable. Mice were then treated with cobimetinib MWF, anti-PD-1 (0.25mg/mouse) and/or anti-41BB (0.025mg/mouse) on Tuesday and Thursday (TR). Mice not receiving treatment on the designated days received vehicle or isotype control antibody (see Drug preparation for details). For all treatment schemes, mice were sacrificed and tissues were collected 21 days after tumor implantation for processing.

To measure tumors for survival experiments, tumor length and width were assessed three times weekly using caliper measurements, with the length assigned to the longest cross-sectional tumor diameter. Tumor volume was calculated as (tumor volume = (length*width²)/2). Tumor volume was assessed until tumors reached 18×18mm, at which point the mice were euthanized.

Hemi-splenectomy in vivo Studies

CRC is frequently diagnosed at an advanced stage, and most often metastasizes to the liver. To model metastatic CRC, we utilized a preclinical murine model of hepatic metastases via direct injection of tumor cells into the hemispleen, as previously described (27). Briefly, we performed intrasplenic injection of 5×10⁵ CT26 WT or MEK1 KO cells in 6–8-week-old syngeneic female BALB/c mice. CT26 cells were cultured from frozen stocks and passaged twice before injection. Livers were processed for flow cytometry analysis 9 days post-injection or monitored for survival.

Tumor Processing

Tumors or livers were finely chopped and processed on a gentleMACS^™ Octo Dissociator using the pre-set 37C_m_TDK_2 or 37C_m_LDK_1 program. Following dissociation, tumors were quenched with 5ml of RPMI media supplemented with 10% FBS, 1% penicillin/streptomycin and 0.5% L-glutamine, and passed through 100mm cell strainers (Falcon, cat. #08-771-19) to further separate cells. An additional 5ml wash was performed on the C-tube and passed through the cell strainer, followed by a 5ml wash over the cell strainer to flush any remaining cells. The flow through was spun and ACK lysis (Quality Biological, cat. #118-156-101) was performed with 4mls of ACK lysis buffer followed by a 4-minute incubation on ice. Cells were quenched with RPMI complete media and spun down. For the formation of a Percoll gradient, a stock Percoll solution was prepared by diluting the supplied Percoll (GE Healthcare Life Sciences, cat. #17-0891-01) with 10X PBS. This was further diluted with 1X PBS to form 80% and 40% Percoll solutions. Cell pellets were resuspended in 40% Percoll and underlayed with 80% Percoll and then spun at 3200 rpm (no brake) for 25 minutes at room temperature. The immune cell layer was then removed and 1×10⁶ cells were plated per well in a 96-well plate for flow cytometry analysis.

CD8⁺ T-cell Isolation

Following initial tumor processing, an equivalent number of cells from each of the five mice were combined within a given treatment group. CD8⁺ T cells were isolated using the EasySep Mouse CD8⁺ Isolation Kit (StemCell Technologies, cat. #19853) per the manufacturer’s instructions. Following isolation, 2.5×10⁶ cells were plated per well in a 6-well plate in quadruplicate to provide technical replicates. Cells were stimmed overnight with anti-CD3/CD28 Dynabeads beads (Thermo Fisher Scientific, cat. #11452D) at a 1:1 bead to cell ratio per the manufacturer’s instructions for T-cell activation. A protein transport inhibitor cocktail (eBioscience, cat. #00-4980-03) was introduced at 1X concentration during the last 4 hours of stimulation since samples were being analyzed for cytokine expression by flow cytometry. Beads were removed after 20 hours at 37°C and samples were plated in a 96-well plate for flow cytometry staining and analysis.

Flow Cytometry

Isolated single-cell suspensions (from in vitro or in vivo experiments) were washed with refrigerated PBS and stained with Live/Dead fixable Aqua or Near-IR (Invitrogen, cat. #L34966 and cat. #L10119, respectively) for 30 minutes at 4°C in the dark. Cells were washed 3X with cold PBS followed by a 30-minute incubation with surface marker antibodies diluted in FACS buffer (PBS with 1% FBS and 0.1% NaN3) for 30 minutes at 4°C in the dark (see supplementary Table S1 for a list of antibodies, dilutions, and isotype controls). For samples being stained for intracellular markers, cells were fixed and permeabilized using the cytofix/cytoperm fixation/permeabilization solution kit (BD Biosciences, cat. #555028) and then incubated at room temperature with intracellular antibodies diluted in the provided 1X fixation/permeabilization buffer for 30 minutes. Samples were washed 3X, resuspended in 200ul of FACS buffer and run immediately on a CytoFLEX Flow Cytometer (Beckman Coulter). Flow cytometry analysis was performed using CytExpert Software (Beckman Coulter) and FlowJo 10.5.3 (FlowJo, LLC).

Tumor Gene Expression Profiling and Analysis

To determine an underlying mechanism explaining the immunomodulatory changes seen in the CT26 MEK1 KO tumors vs CT26 WT tumors, adult BALB/c mice were inoculated with 5×10⁵ CT26 WT or 5×10⁵ CT26 MEK1 KO cells into the lower left flank at 6–8 weeks of age. On day 21, RNA was extracted from whole tumors using the RNeasy extraction kit (Qiagen, cat. #74104). RNA was analyzed on a Nanodrop 2000 for quantity and the Agilent 4200 TapeStation for RNA quality, and 100ng of total RNA was used for hybridization per reaction. The Nanostring Cancer Immune codeset (XT-CSO-MIP1-12) was used to perform the nCounter Gene-Expression Assay, and no custom probes were added. Samples were processed on the NanoString Technologies nCounter Sprint Profiler according to NanoString protocols in the The Sidney Kimmel Comprehensive Cancer Center in the Bloomberg~Kimmel Institute for Cancer Immunotherapy. Data from the NanoString nCounter system were normalized to the internal positive controls and housekeeping gene (Abcf1, Alas1, Edc3, Eef1g, Eif2b4, G6pdx, Gusb, Hdac3, Hprt, Nubp1, Oaz1, Polr1b, Polr2a, Ppia, Rpl19, Sap130, Sdha, Sf3a3, Tbp, Tubb5) using the recommended settings in the nSolver 4.0 software normalization module (NanoString Technologies). Normalized data was exported and TMM preprocessed using the EdgeR statistical package in R. Preprocessed data were then analyzed using the DeSeq2 statistical package in R to obtain fold change and adjusted p values for each gene. Within the DeSeq2 package, statistical significance was calculated using the Wald Test and was corrected for multiple testing using the Benjamini and Hochberg method.

Study approval

All animal studies were reviewed and approved by the Johns Hopkins Institutional Animal Care and Use Committee (ACUC) and Biohazards Committee. All efforts were made to limit animal pain and discomfort. Mice were monitored twice-daily by the Johns Hopkins Institutional Animal Care, and at least three times a week by the investigators to assess for animal suffering or distress such as but not limited to ruffled fur, weight loss, ascites, hunched posture, labored respiration, cyanosis, or signs of morbidity due to the growth of tumor or drug treatment. Feed and water were provided ad libitum, and mice were euthanized by CO₂ narcosis.

Human study approval

Human clinical samples from 56 patients treated with atezolizumab both with (22 samples) and without cobimetinib (34 samples) were obtained from a phase 2 trial in biliary tract cancers (ClinicalTrials.gov Identifier: NCT03201458) that was coordinated by the National Cancer Institute (NCI) Experimental Therapeutics Clinical Trials Network (ETCTN) (19). The study was approved by the NCI’s Cancer Therapy Evaluation Program (CTEP) Institutional Review Board as well as the Institutional Review Board or ethics committee at each participating institution and was conducted in accordance with the provisions of the Declaration of Helsinki and the International Conference on Harmonization guidelines for Good Clinical Practice. All the patients provided written informed consent.

Patient sample collection method and analysis

Blood samples were collected prior to dosing at baseline (day 1), and prior to dosing on cycle 1 (day 15). Blood was collected in 10 mL tubes containing EDTA as an anti-coagulant and shipped on ice to a central site (Lesinski Laboratory, Winship Cancer Institute of Emory University) for processing. Upon arrival to the central site, blood samples were centrifuged at room temperature at 805 × g for 10 minutes. Following aspiration of the plasma layer, peripheral blood mononuclear cells (PBMC) were isolated from these same blood samples via standard methods using density gradient centrifugation with ficoll-paque. Following isolation, PBMCs were resuspended in PBS, counted via trypan blue exclusion and aliquoted into cryovials. The samples were stored in the vapor phase of liquid nitrogen.

Frozen PBMCs were thawed in RPMI and washed with PBS containing 3% FBS and 0.05mM EDTA. Single-cell suspensions were washed and stained with Live/Dead ghost dye 450 (Tonbo Biosciences, cat #13-0863) for 30 minutes at 4°C in the dark. Cells were then washed 3X followed by a 30-minute incubation at 4°C with surface marker antibodies diluted in FACS buffer (see supplementary Table S1 for a list of antibodies, dilutions, and isotype controls). Cells were washed and then permeabilized and fixed using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience, cat. #00-5523-00) for flow cytometric analysis on a Cytek Aurora (Cytek Biosciences, Fremont, CA).

Statistical Analysis

For survival data, results were plotted using a Kaplan-Meier curve and statistical significance was determined via a log-rank test. All data are plotted as mean and standard error of the mean. Differences between two groups were tested using unpaired two-tailed t-tests using GraphPad Prism 8 (GraphPad Software, Inc.). For differences between 3 or more groups, significance was determined by a one-way ANOVA with Tukey’s multiple comparisons test using GraphPad Prism. Differences were considered significant when the P-value was <0.05. Statistically significant p values are abbreviated as follows: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. All experiments were repeated at least 2 times.

Results

MEK1 KO impairs CT26 tumor growth in an immune-dependent manner

To characterize the role of the MAPK pathway in cell-intrinsic immune evasion, we utilized MCF10A cells, an epithelial cell line with low baseline MAPK pathway activation, as measured by pERK levels (28). We induced MAPK pathway activation by increasing the concentration of epidermal growth factor (EGF), and measured cell surface expression of MHCI and MHCII by flow cytometry after 48 hours (29). The addition of increasing concentrations of EGF markedly decreased MHCI and MHCII levels, confirming the MAPK pathway as a negative regulator of MHCI and MHCII expression (Figure 1A). Concurrent treatment with the MEK1 inhibitor cobimetinib markedly increased the expression of MHCI and MHCII in the setting of EGF. The greatest magnitude of increase in MHCI with MEKi was seen at the highest concentration of EGF, indicating that upregulation of MHC by MEKi may be most pronounced in the setting of MAPK pathway activation.

Figure 1. MEK1 KO impairs CT26 tumor growth in an immune-dependent manner.

(A) MCF10A cells exposed to increasing concentrations of EGF in vitro for 48 hours resulted in suppression of MHCI and MHCII expression, which was rescued by treatment with the MEKi cobimetinib (1uM) (B) CT26 MEK1 KO clones were generated using CRISPR/Cas9, and MEK1 ablation was validated through western blot analysis. pERK1/2 activation was maintained in all MEK1 KO clones. The MEK1 KO D12 clone was selected for in vivo experiments. (C) All CT26 MEK1 KO clones showed enhanced MHCI and MHCII baseline expression in vitro as measured by flow cytometry. (D) MEK1 KO clones grew at the same rate as CT26 WT cells in vitro. (E) MEK1 KO tumor growth was significantly delayed in immune-competent mice but unaffected in immune deficient RAG1^−/− mice. WT and RAG1^−/− BALB/c mice were challenged subcutaneously with either CT26 WT or MEK1 KO cells and tumor volumes were monitored. MEK1 tumor growth was only delayed in the WT BALB/c background, confirming the role of the immune system in driving a survival benefit of MEK1 KO cells. One-way ANOVA followed by a post Tukey’s test was used to determine statistical significance between multiple groups. For survival data, results were plotted using a Kaplan–Meier curve, and statistical significance was determined via a log-rank test; n=10 mice/group.

To directly elucidate the role of MEK1 on tumor cell immunogenicity in vitro and in vivo, we used CRISPR/Cas-mediated genome editing to knock out MEK1 through transfection of a RAS-activated CRC cell line, CT26 (Figure 1B). Consistent with the phenotype of pharmacologic MEKi with cobimetinib, MHCI and MHCII were significantly upregulated in multiple independent MEK1 KO clones compared to WT CT26 cells (Figure 1C). MEK1 KO cells did not demonstrate a decrease in pERK levels, which is downstream of MEK1, potentially due to compensation from MEK2 and/or loss of feedback control from ERK1/2 through the phosphorylation of MEK1 (30). Thus, MEK1 is a key regulator of MHCI and MHCII expression and does so independently of ERK signaling. Loss of MEK1 signaling had no appreciable impact on cell growth kinetics in vitro (Figure 1D) implying that MEK1 is dispensable for tumor growth and suggesting that any effects on tumor size in the KO model following implantation in vivo would not simply be a function of impaired growth rate.

To determine whether MEK1 KO impacted tumor progression in an immune-dependent manner, we implanted CT26 WT and MEK1 KO cells subcutaneously into WT syngeneic BALB/c mice as well as RAG1^−/− mice, which have no mature B and T lymphocytes (Figure 1E). While CT26 WT and MEK1 KO tumors grew similarly in the RAG1^−/− mice, the growth of MEK1 KO tumors was markedly impaired in immune-competent BALB/c mice. Consistent with the observations in our in vitro studies of cell growth, these studies demonstrate that MEK1 is dispensable for tumor growth kinetics, but loss of MEK1 critically impairs tumor growth in an immune-dependent manner.

MEK1 KO tumor cells show enhanced antigen presentation and JAK/STAT pathway activation

After establishing a role for MEK1 signaling in modulating tumor immunity, we investigated mechanisms underlying the observed prolongation of survival in mice bearing MEK1 KO tumors. Following implantation of either CT26 WT or CT26 MEK1 KO cells, tumors were established for three weeks before tumor harvest and tumor-infiltrating lymphocyte (TIL) analysis by flow cytometry. Consistent with the increase in survival, MEK1 KO tumors were significantly smaller than WT tumors (Figure 2A). Although no difference was seen in the overall percentage of CD4⁺ T cells among CD45⁺ cells, a significant increase in CD8⁺ T cells was seen in the MEK1 KO tumors (Figure 2B). Representative flow cytometry gating schemes are shown in Supplemental Figure S1A. Further, these CD8⁺ T cells had decreased co-expression of exhaustion makers PD-1 and LAG-3 (Supplemental Figure S2A). There was a small but significant increase in regulatory T cells (FoxP3⁺ CD4⁺) in the KO tumors versus WT tumors, potentially reflecting enhanced immune activation and compensatory reestablishment of tolerance. Within both the CD4⁺ and CD8⁺ compartments, we observed an increase in the percentage of central memory (CD44⁺ CD62L⁺) and effector memory (CD44⁺ CD62L^–) T cells and a decrease in the percentage of naïve cell (CD44^– CD62L⁺), consistent with enhanced priming and memory T-cell responses (Figure 2, C and D). An increase in the proliferation of both CD4⁺ and CD8⁺ T cells was also noted (Figure 2E). We repeated these analyses in a metastatic model of CRC in which hepatic metastases result from direct injection of tumor cells into the hemispleen. This model mimics the liver immune microenvironment of patients who develop CRC liver metastases (27). Consistent with the results in our subcutaneous model, BALB/c mice injected with CT26 MEK1 KO cells in the hemispleen showed a significant increase in survival as compared with mice injected with CT26 WT cells, with liver metastasis developing in 100% of mice in the CT26 WT group and only 50% of mice with CT26 MEK1 KO tumors. Similar to our findings in the subcutaneous model, mice bearing CT26 MEK1 KO tumors demonstrated an increased proportion of activated and proliferating CD8⁺ T cells, which likely mediated tumor clearance (Supplemental Figure S3A–F) (30).

Figure 2. Ablating MEK1 in tumor cells leads to intratumoral CD8⁺ T-cell infiltration and enhanced effector function.

BALB/c mice were challenged subcutaneously with either 5×10⁵ CT26 WT or 5×10⁵ CT26 MEK1 KO cells. Tumors were resected at day 21 for TIL analysis by flow cytometry. Each dot represents one mouse, and each bar represents mean ± SEM (n=4–5 mice/group). Two-tailed unpaired T tests were used to compare WT and MEK1 KO group weights and cell phenotypes. (A) Tumor weights at the time of resection demonstrated significantly decreased tumor weights in the MEK1 KO group. (B) MEK1 KO tumors showed an increase in the proportion of CD8⁺ T cells and T regulatory cells (FoxP3⁺ as a percentage of CD4⁺ T cells) with no significant change in the proportion of CD4⁺ T cells. (C) MEK1 KO tumors had a significant increased proportion of CD4⁺ central memory and effector memory cells, and a concomitant decrease in the proportion of naïve cells. (D) MEK1 KO tumors showed an expansion in the proportion of CD8⁺ central memory and effector memory cells, and a significant decrease in the proportion of naïve cells. (E) MEK1 KO tumors had an increase in proliferating CD8⁺ T cells and CD4⁺ T cells. (F) Equivalent number of cells were combined from each mouse within a treatment group, and CD8⁺ cells were isolated from and stimulated with anti-CD3/anti-CD28 beads overnight. CD8⁺ T cells isolated from MEK1 KO tumors had increased markers of activation. Error bars are representative of technical replicates.

To characterize the effects of MEK1 signaling on T-cell activation and function within the tumor immune microenvironment, we isolated CD8⁺ T cells from either CT26 WT or MEK1 KO subcutaneous tumors and stimulated these T cells with CD3/CD28 magnetic Dynabeads to assess cell-mediated cytotoxicity. A significantly higher percentage of IFNγ and GZMB expression was seen in CD8⁺ T cells that were isolated from the KO tumors compared to the WT tumors, indicating enhanced effector function (Figure 2F). Of additional interest, an increase in the percentage of B cells (CD19⁺) as well as enhanced activation of the B cells, as measured by CD86⁺ CD19⁺ and MHCII⁺ CD19⁺ cells, was also seen in the KO tumors (Supplemental Figure S2B).

To investigate the potential molecular mechanism(s) mediating lymphocyte infiltration and activation in the setting of genetic ablation of MEK1 in tumors, we examined the transcriptional changes in whole tumor lysate of WT tumors compared with MEK1 KO tumors. We identified the top 25 differentially expressed genes and found a large proportion of the genes were upregulated in the KO tumors and were involved in the antigen processing pathway (Figure 3A). Using a curated gene set (Nanostring), we observed marked upregulation of multiple pathways including the antigen processing and presentation pathway and the JAK/STAT pathway in the KO tumors (Figure 3B). Consistent with our downstream findings, several groups have shown that upregulation of antigen processing components can lead to an increase in tumor-infiltrating CD8⁺ T cells and enhanced lysis (14, 31). Upregulation of JAK1 and STAT1 in the KO cells compared with the WT CT26 cells was confirmed at the protein level by western blot (Figure 3C). We hypothesized that this pathway was upregulated in the MEK1 KO tumor cells and that upregulation of JAK1/STAT1 in the MEK1 KO tumors was at least partially driving the increase in antigen processing and resulting in enhanced CTL infiltration and activation. To determine if enhanced MHCI expression in the MEK1 KO cells was modulated by the JAK signaling axis, we treated MEK1 KO cells with the JAK1/JAK2 inhibitor ruxolitinib in vitro. Treatment with ruxolitinib partially abrogated MHCI expression in these cells, confirming that antigen presentation in the MEK1 KO cells was dependent in part on JAK1/JAK2 signaling (Figure 3D).

Figure 3. MEK1 KO tumors have increased JAK/STAT pathway activation resulting in enhanced antigen presentation.

BALB/c mice were challenged subcutaneously with 5×10⁵ CT26 WT or 5×10⁵ CT26 MEK1 KO cells. Tumors were resected at day 21 for whole tumor RNA extraction (n=6 per group). Data was analyzed using the DeSeq2 statistical package in R to obtain fold change and adjusted p values for each gene. Statistical significance was calculated using the Wald Test and corrected for multiple testing using the Benjamini and Hochberg method. (A) Volcano plot of differentially regulated genes in MEK1 KO tumors revealed that genes involved in the antigen processing pathway, including H2-Aa, H2-Ab1, and Tap1, were among the most upregulated in the MEK1 KO tumors. (B) Heatmap of gene expression changes revealed upregulation of the antigen processing pathway and JAK/Stat pathways in MEK1 KO tumors. Red indicates upregulation and yellow indicates downregulation. (C) CT26 WT and MEK1 KO cells (clone D12) were analyzed via western blot for expression of Jak1 and Stat1. (D) Treatment of CT26 MEK1 KO cells with the JAK1/JAK2 inhibitor ruxolitinib decreased expression of MHCI in vitro, revealing a partial dependence of the KO cells on the JAK/STAT pathway for MHC processing. (E and F) BALB/c mice were challenged subcutaneously with either 5×10⁵ CT26 WT or 5×10⁵ CT26 MEK1 KO cells. Mice were treated with either vehicle or ruxolitinib (60mg/kg) daily by oral gavage beginning on day 1. Tumors were resected on day 21. (E) Equivalent numbers of cells were combined from each of the 5 mice within a treatment group and CD8⁺ T cells were isolated and stimulated overnight with anti-CD3/anti-CD28 beads. Significance was determined by a one-way ANOVA with Tukey multiple comparisons test. CD8⁺ T cells isolated from MEK1 KO tumors treated with ruxolitinib showed decreased expression of markers of activation relative to vehicle MEK1 KO tumors. Error bars are representative of technical replicates. (F) Tumor weights on day 21. Each bar is an average of 5 mice/treatment group and error bars indicate mean ± SEM.

To further demonstrate that intratumoral CTL infiltration and activation in the setting of MEK1 inhibition was at least partially mediated by increased JAK1/STAT1 signaling, we implanted CT26 WT or MEK1 KO cells subcutaneously in BALB/c mice and treated daily with either vehicle or the JAK1/JAK2 inhibitor ruxolitinib (60mg/kg) by oral gavage. After three weeks, CD8⁺ T cells were isolated from the tumors and stimulated overnight, at which point markers of activation were measured by flow cytometry. As with earlier experiments, CD8⁺ T cells isolated from KO tumors showed enhanced expression of GZMB and IFNγ relative to CD8⁺ T cells isolated from WT CT26 tumors (Figure 3E). However, although treatment of the WT tumors with ruxolitinib led to a significant increase in these markers relative to vehicle-treated WT tumors, CD8⁺ T cells isolated from ruxolitinib-treated KO tumors showed a significant decrease in GZMB and IFNγ expression. Aligning with this observation, there was a trending decrease in tumor volume in WT CT26 tumors in mice treated with ruxolitinib but a trending increase in volume of MEK1 KO tumors in mice treated with ruxolitinib (Figure 3F). These results demonstrate that the JAK/STAT pathway is a critical mediator of some of the immunomodulatory changes observed in MEK1 KO tumors.

Systemic MEK inhibition impairs effector T-cell function

After confirming that the changes on the tumor cells alone induced by MEK1i are significant enough to lead to increased CD8⁺ T cell frequency and function, we next assessed whether systemic MEKi, which incorporates not only the changes on the tumor cells but also the direct effects on immune cells, affects the tumor microenvironment. We implanted WT CT26 cells subcutaneously in Balb/c mice and treated the mice with either DMSO or cobimetinib, an FDA approved MEK inhibitor, three times a week. We then harvested and processed the tumors after three weeks for TIL analysis. A significant decrease was seen in tumor burden with cobimetinib treatment, but no difference was seen in the percentage of CD4⁺ or CD8⁺ T cells (Figure 4A). Additionally, no significant changes were seen within the CD4⁺ or CD8⁺ T-cell compartment or in proliferation, except for an increase in the percentage of naïve CD8⁺ T cells in cobimetinib-treated tumors (Supplemental Figure S4A–C). Systemic MEKi with cobimetinib induced similar tumor changes in the JAK/STAT pathway at the level of RNA and in the antigen presentation pathway at both the RNA and protein level as had been observed with MEK1 KO (Supplemental Figure S4D–F). These findings led us to ask if pharmacological MEKi was directly impairing T-cell function in our model. As in the KO experiments, CD8⁺ T cells were isolated from WT or cobimetinib-treated tumors after three weeks, stimulated overnight and analyzed by flow cytometry. Of significance, cobimetinib treatment led to a significant decrease in IFNγ, GZMB, and TNFα, indicating cobimetinib was impairing T-cell activation at the given timepoint (Figure 4B). To confirm generalizability of this finding, we repeated this experiment using the Panc02 cell line, a model of pancreatic cancer. As with the CT26 tumors, CD8⁺ T cells from cobimetinib-treated Panc02 tumors showed a significant decrease in IFNγ, GZMB, TNFα, and Ki67 (Figure 4C).

Figure 4. Systemic MEKi impairs CD8⁺ T-cell function, diminishing the efficacy of an antitumor immune response.

BALB/c mice were challenged subcutaneously with 5×10⁵ CT26 WT cells and treated with cobimetinib (MWF, 7.5mg/k) or DMSO by intraperitoneal injection beginning on day 1. Tumors were resected on day 21 for TIL analysis by flow cytometry and CD8⁺ T-cell isolation. Each dot represents one mouse, and each bar represents mean ± SEM (n=5 mice/group). Two-tailed unpaired T test were used to compare DMSO and cobimetinib treated mice group weights and cell phenotypes. (A) Tumor weights at the time of resection demonstrated significantly decreased tumor weights in the cobimetinib treated group, but there was no significant difference in CD8⁺ T-cell infiltration. (B) CD8⁺ T cells isolated from cobimetinib-treated tumors and stimulated with anti-CD3/anti-CD28 beads overnight showed impaired effector function compared with mice treated with vehicle, including decreased GZMB, IFNγ, and TNFα. Error bars are representative of technical replicates. (C) C57BL/6 mice were injected subcutaneously with 5×10⁶ Panc02 cells and treated following the same treatment and resection scheme in A. CD8⁺ T cells isolated from cobimetinib treated tumors had decreased markers of activation, including GZMB, IFNγ, TNFα and Ki67 after overnight stimulation. (D) WT and RAG1^−/− BALB/c mice were challenged subcutaneously with CT26 WT cells and treated with cobimetinib (7.5mg/kg) or vehicle by intraperitoneal injection beginning on day 1. In both the WT BALB/c background and RAG1^−/− background, mice treated with cobimetinib had similar improvements in survival relative to vehicle-treated mice, indicating that the effects of cobimetinib on tumor growth were not mediated by the adaptive immune system. For survival data, results were plotted using a Kaplan–Meier curve, and statistical significance was determined via a log-rank test; n=10 mice/group.

Given the lack of significant changes in the immune cell populations and the impaired function, we next wanted to determine if the adaptive immune system was playing any role in decreasing the tumor burden in cobimetinib-treated CT26 tumors compared with DMSO-treated tumors. As before, we implanted CT26 cells into either WT BALB/c mice or RAG1^−/− mice and treated the mice with either cobimetinib or DMSO. A significant decrease in tumor growth and increase in survival was noted with cobimetinib treatment in both the WT BALB/c mouse background and the RAG1^−/− (Figure 4D), suggesting a less prominent role for adaptive immunity in the efficacy of pharmacologic MEKi. These data suggest that the direct negative effects of pharmacologic MEKi on T cells may be overshadowing the positive effects on T cells that result from MEKi-mediated enhancement of tumor cell immunogenicity.

Addition of cobimetinib impairs T-cell activation in patients receiving concurrent PD-L1 inhibition

As a validation of our preclinical findings, we analyzed peripheral blood patient samples from a recent phase 2 study of the PD-L1 inhibitor atezolizumab alone (Arm A) and in combination with cobimetinib (Arm B) in metastatic biliary tract cancers (19). This study enabled us to specifically isolate the effects of systemic MEK inhibition in the context of anti-PD-L1 (Figure 5A). Using peripheral blood samples, we analyzed the proportion of CD8⁺ T cells expressing CD38, a marker of human T–cell activation, at baseline and compared this to the proportion expressing CD38 at day 15, by which time all patients would have received 1 dose of atezolizumab and patients in Arm B would have received 14 days of cobimetinib treatment (32). Given that PD1 expression can also define T cells that are in a more heightened state of T-cell activation in addition to its role as a biomarker for T-cell exhaustion, we also analyzed the proportion of CD8⁺ T cells that were PD1⁺ but did not express the exhaustion marker TIM3 (33). We observed that the percentage of CD8⁺ T cells expressing CD38 as well as the percentage of CD8⁺ T cells that were PD1⁺ TIM3^– increased in Arm A between cycle 1 day 1 (C1D1) and cycle 1 day 15 (C1D15) (mean fold change between the two time points was 1.58 and 1.631 respectively), whereas the percentage of CD8⁺ T cells expressing CD38 and that were PD1⁺ TIM3^– remained constant between the two time points in Arm B (mean fold change was 1.01 and 1.114 respectively) (Figure 5B and C). The difference in fold change between Arm A and Arm B was statistically significant in both cases (p=0.002 and 0.009). While circulating T cells typically may have a low frequency of tumor antigen specific clones, this clinical trial material was in agreement with our preclinical observations that systemic MEK inhibition globally impairs T-cell activation in the clinical setting.

Figure 5. Addition of cobimetinib to atezoluzimab in a phase 2 clinical trial leads to a decrease in T-cell activation.

(A) Treatment scheme for the multicenter randomized phase 2 trial of atezolizumab as a monotherapy or in combination with cobimentinib (NCT03201458). (B) Example flow cytometry gating strategy used to analyze activated T cells from patient peripheral blood samples taken at baseline and at day 15 of treatment. (C) Fold change of the proportion of activated CD38⁺ CD8⁺ T cells and PD1⁺ TIM3^– CD8⁺ T cells between peripheral blood patient samples taken at cycle 1 day 15 (C1D15) and cycle 1 day 1 (C1D1) patient samples. Each bar represents mean ± SEM (n=26 for Arm A and n=21 for Arm B). Two-tailed unpaired T test were used to compare the two arms.

Addition of agonist therapy overcomes T-cell functional impairment

After demonstrating that systemic MEKi enhances tumor immune responsiveness, but simultaneously impairs T-cell activation, we next asked if we could rescue T-cell activation in the setting of cobimetinib through the addition of an agonist therapy, specifically anti-41BB. 41BB is a costimulatory receptor expressed on T cells following activation that plays a role in T-cell proliferation, activation-induced cell death prevention, and amplification of cytotoxic T-cell responses (34, 35). 41BB-specific antibodies can activate T cells independently of MEK1/2 signaling, instead relying on the NF-κB pathway (36). To understand at a basic level how 41BB impacts T-cell function in the setting of MEK inhibition, we isolated splenic CD8⁺ T cells from healthy mice, added either anti-41BB (5ug), cobimetinib (50nM) or both, and stimulated the cells with anti-CD3- and anti-CD28-coated magnetic Dynabeads for 48 hours. Treatment with cobimetinib led to a significant decrease in TNFα and GZMB expression; the addition of anti-41BB led to a significant increase relative to cobimetinib treatment alone (Figure 6A). Though not significant, similar trends were seen with IFNγ.

Figure 6. Addition of agonist therapy overcomes T-cell functional impairment in MEK inhibitor treated mice.

(A) Splenic CD8⁺ T cells isolated from healthy BALB/c mice were stimulated with anti-CD3/anti-CD28 beads and treated with DMSO, cobimetinib (10nM) and/or α41BB (5ug) for 48 hours. Cobimetinib exposed T cells had impaired activation, which was rescued upon the addition of anti-41BB. (B–E) BALB/c mice were inoculated with 5×10⁵ CT26 WT cells and treated with vehicle/isotype, cobimetinib (7.5mg/kg, MWF), anti-PD1 (0.25mg/kg, TR), and/or anti-41BB (0.025mg/kg, TR) by intraperitoneal injection beginning on day 7 when tumors were approximately 5mm×5mm. Tumors were resected on day 21 and processed for TIL analysis or CD8⁺ T-cell isolation. T cells were isolated and stimulated as above. Each dot represents one mouse and each bar represents mean ± SEM (n=5 mice/group). Significance was determined by a one-way ANOVA with Tukey multiple comparisons test (B) CD8⁺ T cells isolated from the cobimetinib treated group had a significant or near significant decrease in markers of activation. The addition of anti-PD1 to the treatment scheme only improved GZMB. Addition of anti-41BB to anti-PD1 and cobimetinib led to a significant increase in all markers except CD69 relative to cobimetinib alone, and significantly improved CD69 expression relative to cobimetinib plus anti-PD1. (C) Significant or trending increases were observed in the proportion of CD4⁺ T cells with cobimetinib plus anti-PD1 treatment and cobimetinib plus anti-PD1 plus anti-α41BB relative to control mice. A significant increase was also seen in the proportion of CD8⁺ T cells in the triple treatment group. (D) Although cobimetinib plus anti-PD1 decreased the overall proportion of CD8⁺ Ki67⁺ T cells, the triple treatment rescued this decrease. The triple combination significantly increased the overall percentage of GZMB⁺ CD8⁺ T cells (as a percent of CD45⁺ cells) relative to all treatment groups but anti-PD1 alone. (E) Cobimetinib plus anti-αPD1 plus anti-41BB significantly improved survival relative to all other treatment groups, with a subset of triple-treated mice reliably achieving complete tumor regression. For survival data, results were plotted using a Kaplan–Meier curve, and statistical significance was determined via a log-rank test; n=6 mice/group. Note, significance is denoted relative to the triple combination group (MEKi+αPD1+41BB).

We subsequently examined the effects of triple therapy with cobimetinib, anti-PD-1, and anti-41BB in vivo to examine if this combination would result in additive or synergistic antitumor activity. Using the CT26 model, we implanted cells and began treatment with either DMSO, cobimetinib (MWF, 7.5mg/kg), anti-PD-1 (TR, 10mg/kg), anti-PD-1 plus cobimetinib, or anti-PD-1 plus cobimetinib plus anti-41BB (TR, 1mg/kg) when tumors reached approximately 5mm × 5mm. We decided to add anti-41BB to anti-PD-1 given the strong clinical interest that still exists in the cobimetinib plus anti-PD-1/PD-L1 combination. Sequential treatment was started, with cobimetinib treatment first, based on findings from a prior study suggesting blockade of oncogenic MAPK signaling by MEKi is critical and effective to prime and synergize tumors in response to immunotherapy and is not as effective when dosed simultaneously (12). After 3 weeks, we harvested tumors and both isolated CD8⁺ T cells for overnight stimulation and processed the tumors for TIL analysis. Consistent with prior experiments, treatment with cobimetinib led to a significant (or near significant) decrease in IFNγ, GZMB, TNFα, and CD69 (Figure 6B). Addition of anti-PD-1 to cobimetinib helped increase GZMB expression, but it had no effect or led to a larger decrease in IFNγ, TNFα, and CD69 (Figure 6B). However, addition of anti-41BB to cobimetinib and anti-PD-1 led to a significant increase in GZMB, IFNγ, TNFα and a trending increase in CD69 relative to cobimetinib treatment alone. Additionally, although this triple therapy did not increase the percentage of CD4⁺ T cells, it led to a significant increase in CD8⁺ T cells relative to all treatment groups (Figure 6C). This increase in CD8⁺ T cells also drove an overall increase in the percentage of both GZMB⁺ CD8⁺ T and Ki67⁺ CD8⁺ T cells as a function of total CD45⁺ cells (Figure 6D). No meaningful changes were observed within the CD4⁺ and CD8⁺ compartments, although the triple therapy resulted in a significant decrease in PD-1⁺ Lag3⁺ double positive CD8⁺ T cells relative to the monotherapy groups (Supplemental Figure S5A–C). Finally, the triple therapy led to a significant increase in survival, indicating the importance of these positive immunomodulatory changes (Figure 6E).

In conclusion, MEK1i has complex effects on tumor immunity, both potentiating antitumor immunity through direct effects on tumor cells, and simultaneously impairing T-cell responses through off-target effects on lymphocytes. The immunosuppressive effects of MEKi are potentially reversible with agonists of T-cell costimulatory molecules (Figure 7).

Figure 7. Summary of the tumor and systemic immunomodulatory effects seen with MEKi and combination agonist treatment.

MEK1 ablation in the tumor cell increases antigen processing and presentation in a JAK/STAT dependent manner, leading to enhanced T-cell activation. Pharmacologic MEKi leads to impaired T-cell activation and the direct negative effects of MEKi on T cells mask the positive immunomodulatory changes induced on tumor cells, largely ablating the role of the immune system. Addition of the agonist anti-41BB leads to improved T-cell activation and overall survival in the context of pharmacologic MEKi.

Discussion

MEKi has pleiotropic effects on both tumor cells and lymphocytes, and the relative contribution of these components and their interactions have been undefined. Here we have dissociated the effects of MEK1 immunobiology in tumor cells versus systemic effects of pharmacologic MEKi. We demonstrated that changes resulting from MEK1 ablation were sufficient to reprogram the tumor immune microenvironment, enhancing antitumor immunity and prolonging survival in an immune-dependent manner. However, pharmacologic MEKi had detrimental effects on the functional capacity of CD8⁺ T cells, preventing the full realization of immunomodulation obtained from MEKi in tumor cells alone. Our results build upon other preclinical studies that together underscore the ability of pharmacologic MEKi to both enhance and impede immune responses. For example, in some contexts, MEKi increases CD8⁺ T-cell proportion in tumor cells (10, 37), whereas in other similar tumor types or dosing schemes MEKi results in no change or a decrease in the CD8⁺ T-cell proportion (15, 16, 23). These contrasting results may be a manifestation of opposing effects of MEKi, and in certain contexts one may dominate over the other. In the present study, cobimetinib dosing was evaluated immediately after tumor implantation to parallel the MEK1 KO tumor, potentially exacerbating the impairment of T-cell priming and activation as compared to models in which a critical level of lymphocyte priming or recruitment has already occurred (10, 37). The genetic background and baseline activation of the antigen processing pathway in the tumor may also impact this balance.

Given the high frequency of tumors with altered MAPK pathway activation, and the potential for MEKi as an immunomodulatory strategy for overcoming MAPK immunosuppression, we sought to determine if the addition of a T-cell agonist could restore the functional capacity of CD8⁺ T cells in the setting of pharmacologic MEKi. Here, we demonstrated that the impairment of T-cell function observed with pharmacologic MEKi could be overcome both in vitro and in vivo through the addition of anti-4-1BB. Anti-PD-1 in combination with cobimetinib did not lead to the same degree of improvement, highlighting the importance of the addition of the agonist therapy to overcome resistance to PD-L1. Further underlining the functional importance of overcoming the inhibition of T-cell function, anti-4-1BB combined with anti-PD-1 and cobimetinib significantly enhanced survival relative to both cobimetinib alone and cobimetinib in combination with anti-PD-1.

Verma et al also observed an increase in the percentage of naïve T cells with MEKi, but this was attributed to an increase in stem cell memory T cells (T_SCM) that enhanced the antitumor response (38). Our findings do not preclude this discovery, as we did not specifically look at T_SCM markers within the naïve cell population, although it should be noted that in our study, cobimetinib treatment in RAG1^−/− mice showed an equivalent improvement in survival, suggesting a limited role of the adaptive immune system in mediating a response. Additionally, the improvement in an antitumor T-cell response seen with MEKi by Verma et al was observed when T cells were treated ex vivo and then adoptively transferred rather than with systemic in vivo treatment.

MEK1/2 is usually observed in its canonical role in the ERK cascade, and notably ERK1/2 are the sole substrates to be activated by MEK1/2 kinase activity (39). However, we observed that MEK1 could influence MHCI expression independently of ERK1/2 activation, indicating that the regulation of MHCI by MEK1 may be mediated by protein-protein interactions. MEK1 is necessary for phosphatase and tensin homolog deleted on chromosome ten (PTEN) membrane recruitment through the formation of a complex with MAGI1, and ablation of MEK1 can stabilize Akt (protein kinase B) activation (30). However, given that complex formation of MEKI and MAGI1 is dependent on phosphorylation of MEK1 by activated ERK, this is applicable to the use of MEK inhibitors as well and not just full ablation. This is noteworthy given concurrent inhibition of p-Erk and upregulation of p-Akt results in the most prominent increase in MHCI expression, and treatment of cells with lapatinib, which downregulates both pErk and pAkt did not alter MHCI expression, suggesting a potential role for Akt in MAPK regulation of MHC (14). As we also saw a concomitant increase in the JAK/STAT pathway in the MEK1 KO tumors that was partially responsible for the enhanced T-cell activation seen compared to the WT CT26 tumors, this also suggests that interplay between these two pathways, mediated by protein-protein interactions, influences MHC regulation. This is supported by the STAT1 dependent upregulation of HLA-A seen with MEKi (9).

Our findings show that MEK1 ablation alone was sufficient to increase tumor cell immunogenicity, but it is interesting to speculate whether MEK2 ablation has a similar consequence. MEK1 and MEK2 are often considered to be redundant kinases, but MEK1 and MEK2 signaling pathways are not fully overlapping (40). ShRNA-mediated MEK2 knockdown has a stronger inhibitory effect on cell proliferation, whereas MEK1 but not MEK2 over-activation is required for experimentally-induced benign epidermal papilloma formation even though both induce pERK, further highlighting the importance of MEK protein interactions (41, 42). Future studies should investigate the relevant contributions of MEK1 vs MEK2 in enhancing immunogenicity.

Our study demonstrates strong efficacy with the combination of MEKi plus PD-1i plus anti-41BB. It is possible that other agonist therapies could similarly rescue T-cell activation in the context of MEKi plus anti-PD-1/anti-PD-L1 therapy, for example anti-CD27. CD27 is a costimulatory receptor expressed on naïve and memory T cells and can promote T-cell activation and proliferation through activation of NF-κB and MAPK8/JNK (43). Additionally, in the CT26 tumor model, treatment with a MEK inhibitor leads to a significant upregulation of CD27 relative to control mice, whereas 41BB is downregulated, providing a rationale for the alternate selection of anti-CD27 as the additive agonist therapy (10). The combination of MEKi plus PDL1i combined with the T-cell agonist anti-CD27 to overcome MEKi induced T-cell suppression is being explored by our group in a clinical trial in patients with pre-treated unresectable biliary tract cancers.

Synopsis:

The authors distinguish tumor-mediated and tumor-independent mechanisms of MEK inhibitor immunomodulation. Pharmacologic MEK inhibition enhances tumor immunogenicity but impairs T-cell activation. Antitumor activity with pharmacologic MEK inhibition is enhanced by combination with T-cell agonist therapy.