Intermittent MEK inhibition with GITR co-stimulation rescues T-cell function for increased efficacy with CTLA-4 blockade in solid tumor models

Abstract

MEK inhibitors (MEKis) have shown limited success as a treatment for MAPK/ERK pathway–dependent cancers due to various resistance mechanisms tumor cells can employ. CH5126766 (CKI27) is an inhibitor that binds to MEK and prevents release of RAF, reducing the relief of negative feedback commonly observed with other MEKis. We observed that CKI27 increased MHC expression on tumor cells and improved T cell–mediated killing. Yet, CKI27 also decreased T-cell proliferation, activation, and cytolytic activity by inhibiting the MAPK/ERK pathway that is activated downstream of T cell–receptor signaling. Therefore, we aimed to balance the positive and negative immunomodulatory effects of MEKis for optimal combination with immunotherapy. Intermittent administration of CKI27 allowed T cells to partially recover and co-stimulation via GITR and OX-40 agonist antibodies completely alleviated inhibition of function. In Kras mutant lung and colon tumor mouse models, intermittent CKI27 and anti-GITR significantly decreased tumor growth and prolonged survival when further combined with CTLA-4 immune checkpoint blockade. Moreover, this triple combination increased CD8⁺ and CD4⁺ T-cell proliferation, activation, and effector/memory subsets in the tumor draining lymph nodes and tumors and led to intratumoral regulatory T cell (Treg) destabilization. These data, collectively, will allow for more informed decisions when optimizing combination regimens by overcoming resistance, reducing toxicity, and generating long-term immune responses.

Introduction

MAPK/ERK is the most frequently mutated signaling pathway in cancer, with mutations found in up to 66% of melanomas, 35% of lung cancer, and 45% of colorectal cancers (1). There have been many efforts to target this pathway in the last few decades, including the development of MEK inhibitors (MEKis). Currently, MEKis are FDA-approved for pediatric plexiform neurofibromas (2), histiocytic neoplasms (3), and BRAF^V600 mutant cancers in combination with BRAF inhibitors (BRAFis) (4). However, clinical trials with MEKis as monotherapy or in combination with other therapies have shown low response rates in KRAS mutant cancers including non-small cell lung cancer (NSCLC) (5). This has led to the recent development of direct KRAS^G12C inhibitors with promising response rates (6). Despite this success, the majority of patients did not respond and there are other more commonly occurring RAS mutations in cancer that still represent an unmet clinical need (7). This requires exploration of alternative drugs, such as novel MEKis, in RAS mutant cancers. The limitations of MEKis in clinical trials up to now can be explained by 1) the narrow therapeutic index of MEKis and 2) the intrinsic, adaptive, and acquired resistance mechanisms exploited by tumor cells (8).

In tumor cells with wild-type BRAF, including those with mutant KRAS, MEKis, such as selumetinib, relieve ERK driven negative feedback. This causes dephosphorylation and reactivation of CRAF, which then induces MEK phosphorylation (9). To address this problem of adaptive resistance, we utilized a recently reported MEKi, CKI27, (RO5126766, CH5126766, R-7304, RG-7304, VS-6766, avutometinib) that is currently in phase II clinical trials (10). CKI27 is a non–adenosine triphosphate (ATP)–competitive allosteric inhibitor that binds to MEK to inhibit its kinase activity but is novel relative to other MEKis due to its interaction with MEK S218 and 228, which blocks their phosphorylation by RAF. CKI27-bound, dephosphorylated MEK binds more tightly to RAF to create MEK–RAF complexes, acting as a dominant-negative inhibitor. Thus, CKI27 and another newer MEKi, trametinib, can prevent relief of negative feedback (9,11). Another strategy that can be utilized to delay resistance mechanisms while also increasing the therapeutic index is to implement an intermittent dosing regimen. An intermittent dosing schedule of the BRAFi vemurafenib in a BRAF mutant xenograft model and selumetinib in a genetically engineered KRAS^G12C mutant lung cancer mouse model delayed tumor progression and prolonged survival compared to continuous dosing (12,13). Furthermore, a phase I clinical trial exploring different regimens of CKI27 in human patients with solid tumors recommended a 4 days on/3 days off (4on/3off) regimen to allow a higher dose to be delivered without additional toxicity, increasing the therapeutic index of the drug (14). Following these examples, we chose a similar dosing regimen to examine antitumor efficacy in our KRAS mutant mouse models.

In addition to their effects on tumor cells, MEKis have been shown to have a wide array of effects on different immune cell populations that are largely model and context dependent (15). We previously found that, consistent with many other studies, MEK inhibition increased antigen presentation on tumor cells resulting in increased T cell–mediated killing (16–19). MEK inhibition can also induce immunogenic cell death (20). These findings suggest a rationale for combining MEKis with immunotherapy. The MAPK/ERK signaling pathway is upregulated downstream of T-cell receptor (TCR) signaling and, accordingly, other groups have described that MEKis reduced cytokine production, priming, and proliferation of T cells (18,21). We have previously shown that MEK inhibition with selumetinib decreased T-cell activation, although pulsatile treatment partially rescued this effect by allowing the T cells to recover. Pulsatile selumetinib led to improved survival compared to continuous dosing when combined with immune checkpoint blockade (ICB) targeting CTLA-4 in a transplantable Kras^G12C mutant Lewis lung carcinoma (LLC) model. However, the survival improvement was minimal, hence we hypothesized that the T cells may still be partially inhibited, leading to sub-optimal combination efficacy (13). Consequently, we explored other methods to completely rescue inhibition.

Members of the tumor necrosis factor receptor (TNFR) family provide key co-stimulatory and co-inhibitory signals that regulate innate and adaptive immunity. TNFR-associated factors (TRAFs) downstream of these receptors activate several signaling pathways, including MAPKs, NF-κB, and PI3K/Akt/mTOR, that contribute to T-cell function (22,23). Co-stimulatory agonist antibodies targeting TNFRs such as 4–1BB, OX-40 and GITR activated these pathways to restore T-cell function after MEK inhibition (24,25). Thus, we proposed that an intermittent dosing regimen of CKI27 combined with GITR co-stimulation would rescue T-cell function and improve antitumor responses with CTLA-4 ICB. In this study, we showed that this triple combination decreased tumor growth and prolonged long-term survival in murine Kras mutant tumor models, and this was dependent on the adaptive immune system. Surviving mice also rejected tumors upon re-challenge, suggesting induction of a memory response. Indeed, CD8⁺ and CD4⁺ effector T cells from the tumor-draining lymph nodes (TDLNs) and tumors of treated mice were more activated, proliferative, and skewed toward a memory phenotype, which we confirmed by single-cell RNA-sequencing (scRNA-seq) analysis. In contrast, intratumoral regulatory T cells (Tregs) were inhibited and destabilized. Together, these data show that using an intermittent MEKi dosing regimen combined with T-cell co-stimulation is an efficient strategy to improve combination effects with ICB.

Materials and Methods

Cell lines and Reagents

The B16F10 mouse melanoma line (male, RRID:CVCL_XH27) was originally obtained from I. Fidler (MD Anderson Cancer Center, Houston, TX) in 1989. B16 cells expressing YFP (B16-YFP) were generated as previously described (26). B78H1 (RRID:CVCL_8341) is a clone of B16 that was obtained from A. Albino (Sloan-Kettering Institute) in 1992. WG492 is a melanoma cell line derived from a tumor from a female BRAF^V600E/PTEN^−/− transgenic mouse in house in 2016. LLC Lewis lung carcinoma (female, RRID:CVCL_4358) and CT26 colorectal carcinoma (female, RRID:CVCL_7254) were obtained from ATCC in 2017 and 2021, respectively. HKP1 lung cancer cell line (female) was obtained from Dr. Vivek Mittal at Weill Cornell Medicine (13) in 2015. A375 melanoma (female, RRID:CVCL_A375), HCT-116 colorectal carcinoma (male, RRID:CVCL_0291), NCI-H292 (female, RRID:CVCL_0455), NCI-H358 lung cancer (male, RRID:CVCL_1559), and AsPC-1 pancreatic carcinoma (female, RRID:CVCL_0152) human cell lines were obtained from ATCC in 2023. All cell lines were authenticated and tested for mycoplasma in 2023. STR (short tandem repeat) profiling analysis was used to authenticate cell lines. Either the monoclonal core facility at MSK or the MycoStrip Mycoplasma detection kit (Invitrogen, #rep-mysnc-50) were used for routine mycoplasma testing. B16, B78H1, WG492, CT26, H292, H358, and AsPC-1 were cultured in RPMI-1640 medium containing 7.5% fetal bovine serum (FBS) (Corning, #35–011-CV) supplemented with 2mM L-glutamine and penicillin/streptomycin (MSK Media Preparation Core). HKP1, LLC, and A375 were cultured in DMEM containing 10% FBS supplemented with L-glutamine, sodium pyruvate, and pen/strep (MSK Media Preparation Core). HCT-116 was cultured in McCoy’s 5A medium containing 10% FBS supplemented with L-glutamine, sodium pyruvate, and pen/strep (Gibco). Cell lines were passaged 1–4 times after thawing before being used for described experiments. Cells were detached using 0.25% trypsin/EDTA (MSK Media Preparation Core). For cell surface staining and killing assays, cells were detached non-enzymatically using Cellstripper (Corning, #25–056-CI). Selumetinib (#S1008), trametinib (#S2673), and CKI27 (#S7170) were obtained from Selleckchem and dissolved in DMSO for use according to instructions.

In vitro assays

For viability and MHC flow cytometry analyses: B16, WG492, LLC, HKP1, CT26, H2292, A375, H358, HCT-116, and AsPC-1 were all seeded at 20,000–40,000 cells/well in a 6 well plate and treated with DMSO or various concentrations of CKI27 for 72hr with 5–10ng/mL of IFN-γ (PeproTech, #315–05 or 300–02) added for the last 24hr. For mouse T cell–proliferation assays: T cells isolated from the spleens of naïve C57BL/6 mice (JAX, #00664) were purified with the EasySep Mouse T cell Isolation Kit (Stemcell, #19851) and CD5⁺ T cells were purified with CD5⁺ Microbeads positive selection (Miltenyi, #130-049-301). Healthy human donor buffy coats were purchased from New York Blood Center and PBMCs were isolated by density gradient centrifugation. Mouse T cells were then labelled with CellTraceViolet (CTV, Invitrogen, #C34571), activated using Dynabeads Mouse T-Expander CD3/CD28 (ThermoFisher, #11456D) at a 1:2 bead to cell ratio, and treated with DMSO or CKI27 for either 72hr or 16hr followed by a 56hr washout. For agonist antibody experiments, either 10ug/mL anti-GITR (DTA-1, Bio X Cell, #BE0063), isotype control (LTF-2, Bio X Cell, #BE0090), 50ug/mL anti-OX-40 (OX-86, Bio X Cell, #BE0031), or isotype control (HRPN, Bio X Cell, #BE0088) was added with DMSO or CKI27. Human T cells were labelled with CTV, activated using Dynabeads Human T-Expander CD3/CD28 (ThermoFisher, #11161D) at a 1:25 or 1:100 bead to cell ratio, and treated with DMSO or CKI27 for either 96hr or 24hr followed by a 72hr washout. For GITR engagement experiments, 2.5ug/mL anti-HA (R&D Systems, #MAB060) and 1ug/mL human GITR ligand (R&D Systems, #6987-GL) were crosslinked to a plate prior to adding T cells.

In vivo tumor models

All animal studies were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at Memorial Sloan Kettering Cancer Center. For syngeneic tumor experiments, 8–10 week old female C57BL/6 (JAX, #00664) mice or Rag1^−/− (JAX, #002216) mice were subcutaneously (sq) implanted with 0.5×10⁶ LLC cells on the right flank. 8–10 week old BALB/cJ (JAX, #00651) mice or Rag1^−/− (JAX, #003145) mice were sq implanted with 0.5×10⁶ CT26 cells on the right flank. Tumor diameter was measured by calipers and tumor volume was calculated using volume=(L*W²)/2, where L is tumor length and W is tumor width. For LLC, mice were treated when the tumor became palpable (4-22mm³) and for CT26, mice were treated when tumors were 50–400mm³. Prior to treatment, mice were randomized based on tumor size. 2–5mg/kg CKI27 was administered via oral gavage in 10% 2-hydroxypropyl-beta-cyclodextrin (Sigma-Aldrich, H107). A single dose of 1mg anti-GITR (DTA-1, Bio X Cell, #BE0063) or isotype control (LTF-2, Bio X Cell, #BE0090) was administered via intraperitoneal (ip) injection on day 1 of treatment. 100ug anti-CTLA-4 (9D9, Bio X Cell, #BP0164) or isotype control (Bio X Cell, #BP0086) was administered via ip injection 2x/week starting from day 1 of treatment. 100–200ug anti-CD8 (2.43, Bio X Cell, #BE0061) was administered via ip injection 2x/week for 3 weeks. Treatment durations were for 4 weeks unless otherwise indicated. Survival was analyzed based on the approved humane endpoints (distress and tumor size limit). For re-challenge experiments, mice with cured tumors and mostly age-matched naive mice were implanted with 1×10⁶ of LLC or CT26 sq on the opposite (right) flank.

Immunoblot analysis

Tumor cells were seeded at 20,000–50,000 cells/well and treated with drugs at various timepoints. CD8⁺ T cells isolated from the spleens of naïve C57BL/6 mice were purified with the EasySep Mouse CD8+ T cell Isolation Kit (Stemcell, #19853). 1×10⁶ T cells/well were stimulated for 16hr with 1ug/mL plate bound anti-CD3 (145-2C11, MSKCC antibody core facility), then 0.5 ug/mL soluble anti-CD28 (37.51, Tonbo Bioscience, #50-210-3637) and then drugs were added. Cells were lysed in RIPA buffer (Cell Signaling, #9806S) plus protease (Thermo Scientific, #87786) and phosphatase inhibitors (Thermo Scientific, #78420) for 30 min on ice, briefly sonicated, and centrifuged at 14,000×g for 15 min at 4 °C. Lysates were quantified using the BCA method (Thermo Scientific, #23227), prepared using LDS Sample Buffer (Invitrogen, NP0008) and 2-Mercaptoethanol (BME), and boiled. 20–50ug of lysates were loaded onto NuPage 4%–12% Bis-Tris Gels (Thermo Scientific, #NP0321BOX) followed by transfer to PVDF membranes (Invitrogen, #LC2005). Membranes were blocked with 5% BSA/PBST for 1 hr at 25 °C, incubated with the relevant primary antibodies at 4 °C overnight, washed, incubated for 1 hr at 25 °C with anti-rabbit or anti-mouse HRP secondary antibodies and washed again. Primary antibodies used for immunoblotting were purchased from Cell Signaling Technologies: p-ERK (Thr202/Tyr204; #4370, 1:2000, 42/44 kDa), ERK (#4696, 1:2000, 42/44 kDa), p-MEK (Ser217/221, #9154, 1:1000, 45 kDa), MEK (#4694, 1:1000, 45 kDa), p-p38 (Thr180/Tyr182, #9211, 1:1000, 43 kDa), p38 (#9212, 1:1000, 40 kDa), p-JNK (Thr183/Tyr185, #4668, 1:1000, 46/54 kDa), JNK (#9252, 1:1000, 46/54 kDa), p-p70S6k (Thr389, #9234, 1:1000, 70/85 kDa), p70S6k (#2708, 1:1000, 70/85 kDa), p-p65 (Ser536, #3033, 1:1000, 65 kDa), p65 (#8242, 1:1000, 65 kDa). Loading controls used were beta Actin (Santa Cruz, sc-47778, 1:1000, 45 kDa) or vinculin (Santa Cruz, sc-73614, 1:1000, 117 kDa). Secondary antibodies used were purchased from Cell Signaling Technologies, anti-rabbit HRP IgG (#7074, 1:10,000) and anti-mouse HRP IgG (#7076, 1:10,000) for chemiluminescent signal detection. Bands were detected using an ECL kit (Cell Signaling Technologies, #12757) on an Invitrogen iBright FL1500 imaging system and densitometric results were analyzed using ImageJ software (NIH).

Real-Time PCR

Tumor cells were treated with drugs for 72hr with or without 5ng/mL of IFNγ added for the last 24hr. Cells were detached with Cellstripper, pelleted, and total RNA was extracted using the RNeasy Mini Kit (Qiagen, #74106). cDNA was prepared using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, # 4368814) and 250ng was used as template. The following TaqMan real-time PCR assays (Applied Biosystems, #4444556) were run in triplicate and conducted using H2-K1 (Mm01612247_mH), H2-D1 (Mm00833934_g1), H2-Ab1 (Mm00439216_m1), CD274 (Mm00452054_m1), B2m (Mm00437762_m1), Tap1 (Mm00443188_m1), and GAPDH (Mm99999915_g1) on an Applied Biosystems 7500 Real Time PCR machine. The relative quantification of transcript levels was performed by the comparative Ct method (2^-(ΔΔCt)).

Killing assay

Single-cells suspensions of splenocytes were isolated from OT-1 TCR transgenic mice (JAX, #003831) and Pmel-1 TCR transgenic mice obtained from N. Restifo (National Institutes of Health). Splenocytes were primed with OVA (SIINFEKL 257–264, AnaSpec, #AS-60193–1) or human gp100 peptide (25-33, AnaSpec, #AS-62589) in RPMI media supplemented with 10% FCS and 50μM BME as described in (26). For experiments investigating drug effects on T cells: splenocytes were primed with peptide and treated with DMSO or CKI27 for 72hr or 16hr followed by a washout. Agonist antibodies anti-GITR, anti-OX-40, or isotype controls were added during the entire duration of priming. For experiments investigating drug effects on tumor cells: Pmel or OT-1 splenocytes were primed normally. B16-YFP or B78H1 target cells were pre-treated with drugs for 72hr, detached with Cellstripper, and either unpulsed or pulsed with 0.01ug/mL OVA for 2hr at 37C in media. Target cells were washed 2x before being counted and 10,000 were plated per condition in a 24 well plate. For OT-1 co-cultures, 50ug/mL anti-MHC-I (H-2Kb, Bio X Cell, #BE0172) or isotype control (MPC-11, Bio X Cell, #BE0086) was added to B16-YFP 30min prior to the addition of T cells. 100,000 Pmel T cells were added for a 10:1 E:T ratio and 10,000 OT-1 T cells were added for a 1:1 E:T ratio for the co-culture. 24–48 hr later, B16-YFP cells were quantified on a Celigo Image Cytometer (Nexcelom) to calculate the percent of target cells killed. For the colony formation assay after 48hr, B16-YFP and B78H1 cells were trypsinized, diluted, plated in 6 well plates for 7 days, stained with crystal violet, and colonies were counted to calculate the percent of target cells killed.

ATP secretion and HMGB1 release assays

LLC or CT26 tumor cells were seeded at 50,000 cells/well in a 96 well white/clear bottom plate and were treated with various concentrations of CKI27 for 16–24hr. HMGB1 release was detected with the Lumit HMGB1 Human/Mouse Immunoassay (Promega, #WG110) according to manufacturer’s protocol. Extracellular ATP release was measured using the RealTime-Glo Extracellular ATP Assay (Promega, #GA5010) according to manufacturer’s protocol. Luminescence was detected using the SpectraMax iD5 microplate reader (Molecular Devices).

Cytokine profiling analysis

Supernatants were collected from CD5⁺ T cells purified from the spleens of naïve C57BL/6 mice, sub-optimally stimulated with 1:2 Dynabeads Mouse T-Expander CD3/CD28, and treated with drugs and/or antibodies as described above. Supernatants were collected from human PBMCs sub-optimally stimulated with 1:25 or 1:100 Dynabeads Human T-Expander CD3/CD28 and treated with conditions as described above. Cytokines were quantified using the MILLIPLEX MAP Mouse Cytokine/Chemokine Magnetic Bead 32 Plex Panel or MILLIPLEX Human Cytokine/Chemokine/Growth Factor Panel A 38 Plex Premixed Magnetic Bead Panel according to the manufacturer’s instructions (Millipore) and subjected to Luminex cytokine analysis using the Luminex MAGPIX system or Luminex xMAP INTELLIFLEX system. Concentrations [pg/ml] of each protein were derived from 5-parameter curve fitting models. Fold changes relative to the control were calculated and plotted as log₂FC.

Flow Cytometry analysis

For flow cytometry analysis of in vitro assays described above: tumor cells were detached with Cellstripper or T cells were collected and stained with Zombie NIR viability dye for 15min in PBS, washed, and stained with fluorophore conjugated surface antibodies for 30min on ice in FACS buffer (PBS + 0.5% BSA + 2mM EDTA). For flow cytometry analysis of drug-treated tumor-bearing mice: spleens, TDLNs and tumors were isolated. Tumors were weighed, chopped with scissors, and digested in Liberase (Sigma, #5401119001) and DNase I (Sigma, #11284932001) in RPMI at 37°C for 30 min. Single-cell suspensions were prepared by mechanical dissociation through 40 μm filters for spleens and TDLNs and 100 μm filters for tumors. Tumors were further purified using a 40% Percoll gradient centrifugation. Red blood cells were removed from spleens using ACK lysis buffer. Cells were plated and pelleted in 96 well V-bottom plates and stained with Zombie NIR viability dye for 15 minutes in PBS on ice then washed with FACS buffer. Cells were then blocked in 5 mg/ml Fc-block antibody (2.4G2, MSKCC antibody core facility) for 15 minutes on ice in FACS buffer. Cells were then stained with half of the surface antibodies in FACS buffer for 30 mins on ice, washed, stained with the other half of surface antibodies in FACS buffer, and washed 2x with 200 uL FACS buffer. All intracellular staining was conducted using the Foxp3 fixation/permeabilization staining buffer set (eBioscience, 00-5523-00) according to the manufacturer’s protocol. The blocking buffer was supplemented with 1% mouse serum (Thermo Scientific, #24–5544), 1% rat serum (Thermo Scientific, #24–5555), 1% human serum (Thermo Scientific, #BP2525100), and 100U/mL Heparin (Sigma, H3393). All antibodies and the viability dye used for flow cytometry analysis are detailed in Supplementary Table S1. Flow cytometry was performed on a BD LSRII or Cytek Aurora. FlowJo v10 and FCS Express v7 softwares were used for all flow cytometry analyses.

Single-cell transcriptome sequencing

TDLNs isolated from LLC tumor–bearing mice treated with drugs were processed under the same conditions as for flow cytometry analysis described above. Single-cell suspensions of the TDLNs were individually hash-tagged, pooled into treatment groups, labeled with the TotalSeq-C mouse universal cocktail (BioLegend, #199903), and stained with 0.2ug/mL DAPI (Sigma, D9542) for viability and PE-Cyanine anti-mouse CD45 (I3/2.3, BioLegend, #147704). Live CD45⁺ cells for each treatment group were sorted by flow cytometry, adjusted to the same concentrations, quantified, and loaded onto Chromium Next GEM Chip K (10X Genomics PN 1000286). GEM generation, cDNA synthesis, cDNA amplification, and library preparation of 6,700–11,400 cells proceeded using the Chromium Next GEM Single Cell 5’ Kit v2 (10X Genomics PN 1000263) according to the manufacturer’s protocol. cDNA amplification included 13 cycles and 9.4-43 ng of the material was used to prepare sequencing libraries with 14–16 cycles of PCR. Indexed libraries were pooled equimolar and sequenced on a NovaSeq 6000 in a PE28/88 run using the NovaSeq 6000 S1 or S2 Reagent Kit (100 cycles) (Illumina). An average of 44 thousand reads was generated per cell.

Amplification products generated using the methods described above included both cDNA and feature barcodes tagged with cell barcodes and unique molecular identifiers (UMI). Smaller feature barcode fragments were separated from longer amplified cDNA using a 0.6X cleanup using aMPure XP beads (Beckman Coulter catalog # A63882). Libraries were constructed using the 5’ Feature Barcode Kit (10X Genomics PN 1000256) according to the manufacturer’s protocol with 8 cycles of PCR. Indexed libraries were pooled equimolar and sequenced on a NovaSeq 6000 in a PE28/88 run using the NovaSeq 6000 S1 Reagent Kit (100 cycles) (Illumina). An average of 58 million reads was generated per sample.

Single-cell transcriptome analysis

Single-cell sequencing data was aligned to the Genome Reference Consortium Mouse Build 38 (mm10) using Cell Ranger (v7.1.0; 10X Genomics) in order to obtain T-cell clonotypes, feature barcoding, CITE-seq antibody detection and gene expression profiles associated with individual single cells. Each data type was matched to create a UMI matrix and cells were filtered out based on three metrics: (1) cells with less than 200 detectable genes; (2) cells with more than 3,000 detectable genes; (3) cells that had less than 2.25% percentage of counts related to mitochondrial genes. Clusters of immune subsets were manually annotated based on the known markers listed in Supplementary Table S2. Data normalization, Principal Component Analysis, Uniform Manifold Approximation and Projections (UMAP), downstream statistics and figure generation were all performed on the dataset using the R package Seurat v.4.3.0 (https://github.com/satijalab/seurat). Differential expression comparisons were generated using the DESeq2 package with selected genes (FDR < 0.05). Gene set enrichment analysis (GSEA) and visualization were performed using the R package fgsea v.1.24.0 (https://github.com/ctlab/fgsea).

Data Availability Statement

The data generated in this publication have been deposited in the NCBI’s Gene Expression Omnibus (GEO) and are scheduled to release on Apr 19, 2025 through GEO accession number GSE264426. Additional data or additional information are available upon request from the corresponding author.

Results

MEK inhibition increases tumor antigen presentation and T cell–mediated killing

Allosteric MEKis with dual inhibitory effects on RAF/MEK such as trametinib and CKI27 have been shown to inhibit the proliferation of human KRAS mutant cancer cell lines and xenografts more potently than other MEKis (9,11). To extensively investigate the immunomodulatory effects of CKI27 on tumor cells, we utilized a panel of five murine and five human tumor cell lines harboring different Kras or Braf mutations, or wild-type alleles for both. A phase I dose-escalation trial reported the pharmacokinetics for the maximum tolerated dose of once daily dosing (2.25mg) was 106.3ng/mL (225nM) during run-in and 343.8ng/mL at steady state (729nM) (14). Therefore, we utilized a 10nM-1000nM range of CKI27 in all in vitro experiments, while largely focusing on using 250nM. The viabilities of all cell lines were similarly reduced by CKI27 (Fig. 1A). We then assessed whether CKI27 prevented the induction of MEK phosphorylation observed with the relief of negative feedback in Kras mutant cell lines. LLC and CT26 cell lines were treated with selumetinib, trametinib, or CKI27 for different durations and MAPK/ERK pathway phosphorylation was assessed. While all drugs inhibited ERK phosphorylation (pERK) at 2hr, only CKI27 sustained MEK phosphorylation (pMEK) and pERK inhibition. pMEK was induced at 2hr with selumetinib and 24hr with trametinib and pERK levels rebounded at 24hr with selumetinib and at 48hr with trametinib (Fig. 1B). MEK inhibition has been shown to increase MHC-I expression resulting in improved peptide/MHC target recognition and killing by T cells (19). CKI27 dose dependently increased MHC-I and -II surface expression in several murine and human tumor cell lines tested (Fig. 1C, Supplementary Figs. S1A, S2, S3A). MEK inhibition can also modulate the expression of immune checkpoint ligands PD-L1, CD80, and CD86 on tumor cells and dendritic cells (DCs) (27,28). CKI27 increased PD-L1 in all murine and human cell lines except LLC, CT26, H292, HCT-116, and AsPC-1, where it was decreased instead. CD80 was increased in B16, A375, H358, HCT-116, and AsPC-1 and CD86 was variably increased in several cell lines (Supplementary Fig. S1B, S3B).

Figure 1. MEK inhibition with CKI27 increases tumor antigen presentation, release of immune stimulants, and T cell–mediated killing.

(A) Murine and human tumor cell lines were treated with DMSO, 250nM or 1000nM CKI27 for 72hr. Viability was determined by FACS analysis of live cells/uL and fold change was calculated by normalizing to DMSO; n=3. (B) LLC or CT26 tumor cell lines were treated with DMSO, 100nM of selumetinib, trametinib, or CKI27 for the indicated times. Lysates were assayed by immunoblotting to determine the level of MEK and ERK phosphorylation. (C-D) Tumor cell lines were treated with DMSO or CKI27 for 72hr, or with IFNγ (5-10ng/mL) for the last 24hr; n=3. (C) FACS analysis of MHC-I (H-2Kb/Kd and H-2Db/Dd), MHC-II (I-A/I-E), and HLA-ABC surface expression. Median fluorescence intensity (MFI) values were normalized to DMSO and log transformed. (D) Gene expression of MHC related genes and CD274. (E) LLC or CT26 were treated with various concentrations of CKI27 for 16–24hr; n=3. HMGB1 and extracellular ATP release were measured using luminescence-based assays. (F) Pmel primed T cells were co-cultured with unpulsed B16-YFP or B78H1 and OT-1 primed T cells were co-cultured with SIINFEKL/OVA peptide pulsed B16-YFP that were pre-treated with various concentrations of CKI27 and/or αMHC-I or isotype control. Killing was assessed after 24–48hr using a Celigo Imaging Cytometer or colony formation assay; n=3–5. Experiments were repeated 2–3 times. Data are shown as mean±SEM. One-way ANOVA test with Tukey’s correction for multiple comparisons was used. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001).

Gene expression of MHC-I (H2-K1), MHC-II (H2-Ab1), PD-L1 (CD274) and essential components of the antigen presentation pathway, beta-2-microglobulin (β₂M) and transporter associated with antigen processing 1 (Tap1) were similarly altered upon CKI27 treatment (Fig. 1D). Increases in both protein and gene expression were further augmented with the addition of IFNγ (Fig. 1C–D), which mimics a pro-inflammatory tumor microenvironment. MEK inhibition has also been shown to modify the tumor microenvironment via a form of immunogenic cell death, pyroptosis (20). LLC and CT26 tumor cells treated with CKI27 released immune stimulants such as extracellular ATP and a damage-associated molecular pattern (DAMP), HMGB1 (Fig. 1E). To determine whether these changes had a functional impact, we assessed T cell–mediated killing of CKI27-treated B16-YFP melanoma target cells. We pre-treated gp100 peptide–presenting or SIINFEKL/OVA peptide–pulsed B16-YFP cells with CKI27 and co-incubated them with primed Pmel (gp100-specific TCR) or OT-1 (OVA-specific TCR) effector CD8⁺ T cells, respectively. There was a dose-dependent increase in T cell–mediated killing with both co-culture systems (Fig. 1F). When Pmel T cells were co-cultured with CKI27 pre-treated B78H1 cells, a B16 clone that does not present gp100 peptide, killing was completely abolished. Furthermore, we observed a significant loss of killing upon adding an MHC-I blocking antibody to the OT-1:B16-YFP-OVA co-culture, suggesting that the effect was largely MHC dependent (Fig. 1F). Overall, these results indicated that CKI27 directly increased expression of several immune-related proteins of interest, including MHC expression, on tumor cells, which resulted in improved CD8⁺ T-cell target recognition and thus killing.

Intermittent MEK inhibition relieves suppression of T-cell function

Since T cells rely on MAPK/ERK signaling for proliferation, cytokine production and effector functions (29), we determined the impact of MEK inhibition on T-cell activation and function. Several groups have shown that MEKis suppress T-cell function, thus we investigated whether an intermittent dosing schedule would allow T-cell recovery (18,21). CTV-labelled murine and human T cells were sub-optimally stimulated and treated with CKI27 for either 72–96hr continuously or 16–24hr followed by a 56–72hr washout. Flow cytometry was used to measure CTV dilutions and expression of various activation markers. While T-cell proliferation was significantly and dose-dependently inhibited with continuous treatment, washout conditions alleviated this effect. However, proliferation was not restored to untreated levels (Fig. 2A–B, Supplementary Fig. S4A). Similarly, expression of several co-inhibitory, co-stimulatory, and activation markers on CD8⁺ and CD4⁺ T cells decreased dose-dependently with CKI27. Washout of CKI27 enabled T cells to recover expression of many of these markers to some extent (Fig. 2C, Supplementary Fig. S4B). Since cytokines produced by CD4⁺ T cells can activate CD8⁺ T cells, we analyzed the supernatants from CD5⁺ isolated T cells. Cytokine analysis of the supernatants from CKI27-treated T cells revealed that many pro- and anti-inflammatory cytokines, as well as chemokines, were dose-dependently decreased but rescued with the washout conditions (Fig. 2D, Supplementary Fig. S4C). Continuous CKI27 treatment during priming of OT-1 T cells significantly inhibited their killing of B16-YFP-OVA cells, but washout conditions completely rescued this (Fig. 2E).

Figure 2. Intermittent CKI27 treatment relieves suppressive effects of MEK inhibition on T-cell proliferation, cytokine production, and effector function.

(A-D) CD5+ T cells purified from naïve C57BL/6 spleens were labelled with CellTrace Violet (CTV), sub-optimally stimulated with α-CD3/α-CD28 Dynabeads, and treated with DMSO, continuous CKI27 (72hr), or washout CKI27 (16hr on, 52hr off); n=3. Experiments were repeated 2–3 times. (A) Proliferation as measured by % of CTV^low CD8+ and CD4+ T cells. (B) Proliferation fold change of CD8+ and CD4+ T cells was calculated by normalizing to DMSO. (C) FACS analysis of co-inhibitory, co-stimulatory, and activation markers on CD8+ and CD4+ T cells. Heatmaps represent fold changes of positive percentages of each marker normalized to DMSO. (D) Cytokine analysis of supernatants collected from CD5+ T cells. Heatmap represents fold changes of concentrations (pg/mL) of proteins normalized to DMSO. (E) DMSO, continuous CKI27, or washout CKI27 treated OT-1 primed T cells were co-cultured with SIINFEKL peptide pulsed B16-YFP. Killing was assessed after 48hr using a Celigo Imaging Cytometer; n=5–6. Data are shown as mean±SEM. One-way ANOVA test with Tukey’s correction for multiple comparisons was used. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001)

To compare the effect of daily versus intermittent dosing in vivo, we adapted our schedule from a phase I clinical trial exploring different regimens of CKI27 in human patients with solid tumors. The final recommended regimen was 2.7mg 4on/3off to increase the dose and therapeutic index of CKI27 (14). We chose a 5mg/kg 4on/3off regimen that maintained similar efficacy without additional toxicity compared to the widely used 2mg/kg daily regimen in our LLC tumor–bearing model. We explored the effects of both regimens on the frequency of immune populations 1-, 4-, and 7-days post treatment (dpt) in the spleen, TDLN, and tumor (Supplementary Fig. S5A). Both regimens slightly increased the absolute number of CD45⁺ immune cells, including T cells, in the spleen after 1 dpt, which was then dose-dependently decreased after 4 dpt. Tregs were reduced by ~50%, which was more than CD8⁺ and CD4⁺ effector T cells. After 7 dpt, the daily regimen further decreased immune populations, while the intermittent regimen partially rescued some populations (Supplementary Fig. S5B). Immune populations in the TDLN and tumor were also altered but in a different pattern. The absolute numbers of immune cells in the TDLN remained unchanged with the daily regimen and progressively increased with the intermittent regimen (Supplementary Fig. S5C). Although the frequency of tumor-infiltrating lymphocytes (TILs) slightly increased after 1 dpt, it decreased after 4 and 7 dpt regardless of treatment schedules (Supplementary Fig. S5D). To summarize, while intermittent MEKi treatment only partially rescued T-cell function, it provided additional benefits compared to daily treatment, such as increasing the frequency of T cells in the TDLNs of LLC tumor–bearing mice.

Combination of intermittent MEKi with co-stimulation further rescues T-cell function

Although intermittent treatment relieved some of the inhibitory effects of MEKi on T-cell function, this rescue was incomplete. Several groups have reported that agonist antibodies against co-stimulatory molecules 4–1BB, OX-40, and GITR of the TNFR family restored T-cell function following MEK inhibition (24,25,30). Therefore, we explored whether combining intermittent MEKi with anti-GITR or anti-OX-40 would completely rescue T-cell suppression in vitro in the setting of sub-optimal activation. Both anti-GITR and anti-OX-40 rescued proliferation of CKI27 washout treated CTV-labelled T cells, but only anti-GITR completely restored proliferation to normal levels (Fig. 3A). Similar trends were observed with co-inhibitory, co-stimulatory, and activation marker expression, as well as cytokine production in CD8⁺ and CD4⁺ T cells (Fig. 3B–C, Supplementary Figs. S6–S8). This complete rescue was also observed in T cells from PBMCs treated with washout CKI27 and GITR-L (Supplementary Fig S9). In addition, combining washout CKI27 with either anti-GITR or anti-OX-40 during priming of OT-1 T cells completely rescued killing of B16-YFP-OVA tumor cells (Fig. 3D). Signaling pathways such as MAPK, NF-κB, and PI3K/Akt/mTOR have been shown to be activated downstream of TNFR signaling (22). Several groups have proposed that agonist antibodies targeting TNFRs activate these compensatory pathways to restore the function of MEK-inhibited T cells (24,25). To determine how GITR and OX-40 co-stimulation rescued CKI27-treated T cells, we performed immunoblotting on sub-optimally stimulated CD8⁺ T cells treated with different drug combinations. Combination of CKI27 and anti-GITR upregulated pp38 and pJNK MAPK signaling while CKI27 and anti-OX-40 only upregulated pJNK compared to CKI27 alone (Fig. 3E–F). Additionally, CKI27 and anti-GITR or anti-OX-40 also upregulated pp70S6K, a protein downstream of mTOR signaling, but levels of pp65, a component of NF-κB, were unchanged (Fig. 3E–F). These results suggested that intermittent MEKi combined with anti-GITR can completely rescue MEKi suppression of T-cell function by activating alternative MAPK and PI3K/Akt/mTOR signaling pathways.

Figure 3. Combination with co-stimulatory antibodies targeting GITR and OX-40 further rescues T-cell function by upregulating alternative signaling pathways.

(A-C) CD5⁺ T cells purified from naïve C57BL/6 spleens were labelled with CTV, sub-optimally stimulated with anti-CD3/anti-CD28 Dynabeads, and treated with DMSO, continuous CKI27 (72hr), washout CKI27 (16hr on, 52hr off), isotypes, anti-GITR, and/or anti-OX-40; n=3. Experiments were repeated 2–3 times. (A) Proliferation as measured by % of CTV^low CD8+ and CD4+ T cells. Data are shown as mean±SEM. Two-way ANOVA test with Bonferroni’s correction for multiple comparisons was used. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001) (B) FACS analysis of co-inhibitory, co-stimulatory, and activation markers on CD8⁺ and CD4⁺ T cells. Heatmaps represent fold changes of positive percentages of each marker normalized to DMSO. (C) Cytokine analysis of supernatants collected from CD5+ T cells. Heatmap represents fold changes of concentrations (pg/mL) of proteins normalized to DMSO. (D) DMSO, washout CKI27, isotypes, anti-GITR, and/or anti-OX-40 treated OT-1 primed T cells were co-cultured with unpulsed or SIINFEKL peptide pulsed B16-YFP. Killing was assessed after 48hr using a Celigo Imaging Cytometer; n=4. Data are shown as mean±SEM. One-way ANOVA test with Tukey’s correction for multiple comparisons was used. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001) (E) Sub-optimally anti-CD3/anti-CD28 stimulated CD8⁺ T cells were treated with DMSO, CKI27, isotypes, anti-GITR, and/or anti-OX-40 for 72hr. Lysates were assayed by immunoblotting to determine the level of ERK, MEK, p38, JNK, p70S6k, and p65 phosphorylation. (F) Quantitation of the relative density of bands were normalized to the loading control (vinculin). The ratio was calculated by normalizing phosphorylated protein to total protein.

Intermittent MEKi and GITR co-stimulation improves the anti-tumor effect of ICB

We have previously shown that selumetinib combined with anti-CTLA-4 ICB prolonged survival in LLC tumor–bearing mice (13). Trametinib combined with anti-CTLA-4 has been shown to reduce tumor growth in CT26 tumor–bearing mice as well (18). However, these antitumor effects were marginal, suggesting that MEKis may be partially suppressing the beneficial effects of ICB on T cells. Given that our data suggested that agonist anti-GITR rescued T-cell function to a greater degree than anti-OX-40, we tested the combination of intermittent CKI27 with anti-GITR and anti-CTLA-4 ICB in LLC and CT26 tumor models (Fig. 4A,F). Immunotherapies (anti-GITR or anti-CTLA-4), either alone or combined with each other, did not have any effect on LLC tumor growth or survival, compared to the vehicle. Intermittent CKI27 alone reduced tumor growth and prolonged survival and CKI27 in combination with anti-CTLA-4, but not anti-GITR, slightly enhanced these effects. Notably, the triple combination of CKI27+anti-GITR+anti-CTLA-4 further reduced tumor growth and significantly prolonged survival (Fig. 4B and Supplementary Fig. S10A). The benefit of adding anti-GITR+anti-CTLA-4 to intermittent CKI27 on growth and survival was completely abolished in Rag1^−/− immunodeficient mice, which lack T and B cells, and with CD8 depletion suggesting that the antitumor effect was dependent on T cells (Fig. 4C–D and Supplementary Fig. S10B–C). Finally, when the long-term survivors were re-challenged with LLC on the opposite flank, half of them remained tumor-free, indicating there was establishment of an immune memory response (Fig. 4E and Supplementary Fig. S10D).

Figure 4. The anti-tumor effect of intermittent MEK inhibition, agonist immunotherapy, and checkpoint blockade triple combination is T-cell dependent and leads to a memory response.

(A) LLC mouse schema of treatment timing. (B-E) LLC tumor–bearing mice were treated with vehicle, isotypes, anti-GITR, anti-CTLA-4, 5mg/kg 4on/3off CKI27, and/or anti-CD8 for 4 weeks and survival was monitored over time. (B) Survival of immunocompetent mice. (C) Survival of immunodeficient mice. (D) Survival of CD8-depleted mice. (E) Survival of mice from (B) that were re-challenged. (F) CT26 mouse schema of treatment timing. (G-J) CT26 tumor–bearing mice were treated with vehicle, isotypes, anti-GITR, anti-CTLA-4, 2mg/kg 4on/3off CKI27, and/or anti-CD8 for 4 weeks and survival was monitored over time. (G Survival of immunocompetent mice. (H) Survival of immunodeficient mice. (I) Survival of CD8-depleted mice. (J) Survival of mice from (F) that were re-challenged. Log ranked (Mantel-Cox) test for survival proportions was used. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001). Experiments were repeated 2–3 times.

All treatments were more efficacious in mice bearing CT26 tumors; therefore, the dosage of CKI27 was decreased to 2mg/kg 4on/3off to determine whether there were any improvements with immunotherapy combinations. This is likely due to CT26 being a highly immunogenic tumor and thus more responsive to immunotherapy, compared to the poorly immunogenic LLC (31). All monotherapy arms reduced tumor growth and prolonged survival in CT26 tumor–bearing mice. Double combinations anti-GITR+anti-CTLA-4, CKI27+anti-GITR, and CKI27+anti-CTLA-4 further improved this efficacy (Fig. 4G and Supplementary Fig. S10E). The triple combination had the most striking antitumor effect, which was lost in Rag1^−/− immunodeficient mice and with CD8 depletion (Fig. 4H–I and Supplementary Fig. S10F–G). All long-term survivors that were re-challenged remained tumor free (Fig. 4J and Supplementary Fig. S10H). To summarize, the combination of intermittent MEKi with anti-GITR and anti-CTLA-4 elicited potent antitumor effects and immune memory responses in Kras mutant LLC and CT26 mouse models.

The triple combination enhances activated and effector/memory T cell subsets

We next sought to determine the impact of intermittent CKI27+anti-GITR+anti-CTLA-4 on immune cell populations in the TLDNs and tumors of LLC and CT26 challenged mice after 7 dpt (Fig. 5A). There were consistent trends but also many differences between the models, likely because LLC tumors have low T-cell infiltration while CT26 tumors are highly infiltrated (31). In LLC-challenged mice treated with anti-GITR+anti-CTLA-4, there was visible TDLN enlargement (Supplementary Fig. S11A–B). This was accompanied by changes in the absolute number of CD45⁺ cells, specifically: CD19⁺ B cells and antigen-presenting cells (APCs) including F4/80⁺MHC-II⁺ macrophages and CD11c⁺MHC-II⁺ DCs were slightly decreased with CKI27 but rescued with the addition of anti-GITR+anti-CTLA-4 compared to the control groups (Fig. 5B). The same trend was observed for CD8⁺ and CD4⁺Foxp3^– T effector (Teff) and CD4⁺Foxp3⁺ (Treg) frequencies (Supplementary Fig. S11C–E). In the TDLNs of CT26-challenged mice, these changes were less pronounced. There was a consistent decrease in immune populations with CKI27, but only APCs were increased in the CKI27+anti-GITR+anti-CTLA-4 group (Supplementary Fig. S12A–B). Instead of substantial changes in the frequencies of T cells, we primarily observed alterations in their phenotypes.

Figure 5. The triple combination increases activation, proliferation, and memory subsets of CD8⁺ T cells and CD4⁺ Teffs while destabilizing Tregs in the TDLN and tumor.

(A) Schema of LLC tumor–mice treated with vehicle, CKI27, isotypes, anti-GITR, and/or anti-CTLA-4. TDLNs and tumors from all timepoints were harvested on day 21 (7 days post treatment). (B) Absolute number of immune cell populations in the TDLN; n=4–5. (C) Phenotypes of T cells from the TDLN including representative flow cytometry plots; n=9–10. (D) Tumor weights (mg), numbers of cells/mg, and CD8:Treg ratio of TILs; n=4–5. (E) Phenotypes of TILs; n=4–5. Data are shown as mean±SEM. One-way ANOVA test with Bonferroni’s correction for multiple comparisons was used in all panels. Significance levels are indicated by asterisks (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001). Experiments were repeated 2–3 times.

Immunophenotyping was performed by analyzing expression of several T-cell activation (PD-1, CTLA-4, 4–1BB, CD25, granzyme B) and proliferation (Ki67) markers (32). Inhibitory receptors PD-1 and CTLA-4 have been linked to exhausted T cells, but also increase on T cells upon activation (33,34). Transcription factors T-bet and eomesodermin (Eomes) and cell adhesion molecules CD44 and CD62L play important roles in T-cell function and memory development. T cells that differentiate into effector or effector memory (T_EM) are T-bet⁺Eomes^– or CD44⁺CD62L^– and central memory (T_CM) are T-bet^lowEomes⁺ or CD44⁺CD62L⁺ (35,36). In LLC TDLNs, CD8⁺ and CD4⁺ Teff activation and proliferation markers along with T_EM and T_CM subsets slightly decreased with CKI27 treatment but were completely rescued with the addition of anti-GITR+anti-CTLA-4 (Fig. 5C). The triple combination further enhanced these markers and subsets compared to anti-GITR+anti-CTLA-4 alone. None of these changes were observed in Tregs (Fig. 5C). In CT26 TDLNs, Teff trends were largely the same, except the triple combination only partially rescued the inhibitory effects of CKI27 when compared to anti-GITR+anti-CTLA-4 alone (Supplementary Fig. S12C).

We correspondingly immunophenotyped the TILs and observed more variability in immune subsets and their expression of markers compared to TDLNs. Since CKI27 drastically reduced the sizes of LLC and CT26 tumors, we normalized the TIL numbers to tumor weight for analysis. In LLC tumors (Supplementary Fig. S13), CD45⁺ TILs, including Teffs and B cells, were slightly rescued with the triple combination compared to anti-GITR+anti-CTLA-4 alone. Tregs were slightly decreased and thus the CD8:Treg ratio was increased with the triple combination compared to other groups. APCs and potentially suppressor Ly6C⁺ monocytic and Ly6G⁺ granulocytic myeloid cells (Fig. 5D) were decreased with CKI27 alone and CKI27+anti-GITR+anti-CTLA-4 compared with vehicle. In contrast, CT26 tumors (Supplementary Fig. S14) had increased CD45⁺ TILs including Teffs, Tregs, B cells, APCs, and Ly6C⁺ monocytes with CKI27 alone and the triple combination albeit to a lesser extent, compared to vehicle. Immunotherapy alone did not change the frequencies compared to vehicle. Alterations in the phenotype of LLC and CT26 Teffs were largely the same as those observed in the TDLN. Particularly, Tregs from all treatment groups appeared destabilized based on decreased expression of activation markers including CTLA-4 and CD25 (37) (Fig. 5E). Taken together, these findings suggest that the antitumor effect of intermittent MEKi+anti-GITR+anti-CTLA-4 is due to increased activation, proliferation, and effector/memory differentiation of CD8⁺ and CD4⁺ Teffs in the TDLNs and tumors of LLC- and CT26-challenged mice. Myeloid-derived suppressor cells (MDSCs) and Tregs were inhibited by both MEKi and immunotherapy in LLC and CT26 tumors, potentially further contributing to the antitumor mechanism.

The triple combination alters the gene expression in immune cells

To confirm our findings from flow cytometry analysis, we performed 5’ scRNA-seq and cellular indexing of transcriptomes and epitopes sequencing (CITE-seq) on live CD45⁺ cells isolated from LLC TDLNs of different treatment groups. The treatment conditions were the same as described above (Fig. 5A). 26,080 CD45⁺ cells passed quality control and were clustered based on top differentiated genes. We annotated 5 clusters based on known lineage marker genes listed in Supplementary Table S2, including CD8⁺ T cells, CD4⁺ T cells, Tregs, B cells, and plasma cells (Fig. 6A). There was a large increase in plasma cell counts and decrease in CD4⁺ T cell counts in both anti-GITR+anti-CTLA-4 and CKI27+anti-GITR+anti-CTLA-4 treatment groups (Fig. 6B). T-cell populations were then further subclustered separately into CD8⁺ T cells (n=5,393 cells) and CD4⁺ T cells (n=6,968 cells). CD8⁺ T cells were annotated into subsets according to top differentiated and marker genes commonly described in literature (38,39). The 4 annotated CD8⁺ T-cell subsets were naïve (Dapl1, Ccr7, Lef1, Sell, Tcf7), cytotoxic (Gzma, Gzmk, Eomes, Cxcr3, Il2ra, Il2rb, Ccl5), memory (Stmn1, Birc5, Cenpa, Cks1b, Rrm2, Ccnb1, S100a6, Lgals1), and natural killer T cells (NKT) (Klra7, Klre1, Klrc1, Klrk1, Klrd1) (Fig. 6C and Supplementary Fig. S15A). CD4⁺ T cells were annotated into 5 subsets including: naïve (Lef1, Ccr7, Dapl1, Tcf7), T helper type 1 and T follicular helper (Th1/Tfh)-like (Id2, Cxcr6, Cxcr3, CD200, Tox, Bcla2a1b), Treg (Foxp3, Il2ra, Tnfrsf4, Ctla4, Tnfrsf9, Tnfrsf18, Tigit), early activated (Egr1, Egr2, Cd69, Tagap, Tnf, Id3), and IFN activated (Ifit1, Ifit3, Irf7, Stat1) (Fig. 6D, Supplementary Fig. S15B). Counts of activated and cytotoxic CD8⁺ T-cell, activated and Th1/Tfh-like CD4⁺ T-cell, and NKT-cell clusters were all highest in the triple combination group (Fig. 6E–F and Supplementary Fig. S15C).

Figure 6. The triple combination favorably alters the genetic profile of immune cells in the TDLN.

LLC tumor–bearing mice were treated with vehicle, CKI27, isotypes, anti-GITR, and/or anti-CTLA-4. TLDNs were harvested on day 21 (7 days post treatment). Live CD45⁺ cells were FACS sorted and processed for scRNA-seq. (A) UMAP of all treatment groups showing different clusters and annotations. (B) Quantification of cell counts from (A). (C-D) UMAPs of CD8⁺ and CD4⁺ T cells from all treatment groups showing different clusters and annotations. (E-F) Quantification of cell counts from (C-D). (G) Heatmap showing CD8⁺ and CD4⁺ T-cell gene expression of effector genes. (H) UMAPs of CD8⁺ and CD4⁺ T cells from each treatment group showing specific activation genes. (I-K) GSEA plots of CD8⁺ and CD4⁺ T cells comparing specific groups for different signaling gene sets.

To further examine the differences in T-cell subsets between all the treatment groups, we looked at expression of Teff-related genes derived from a previously published gene signature (40). Total CD8⁺ and CD4⁺ T cells from the anti-GITR+anti-CTLA-4 and CKI27+anti-GITR+anti-CTLA-4 groups had high expression of genes from this signature compared to vehicle or CKI27 (Fig. 6G). Moreover, there were many genes that were highly expressed in the triple combination compared to immunotherapy alone (Fig. 6G). Other genes highly expressed on T cells from the triple combination included those related to pro-inflammatory chemokines and cytokines (Cxcr3, Il2rb, Ifngr1, Stat1), granzymes (Gzma, Gzmm), and memory, which can play important roles in achieving a durable response (Tbx21, Eomes, Il7r). In addition, protein expression of an activation marker that had low gene expression, CD44, also increased upon CKI27+anti-GITR+anti-CTLA-4 treatment, as measured by CITE-seq. Clusters expressing these activation genes and proteins were CD8⁺ memory, cytotoxic, NKT, Tfh/Th1-like, and CD4⁺ activated cells (Fig. 6H and Supplementary Fig. S15A–D). GSEA was then used to evaluate normalized enrichment scores (NES) of several signaling-related gene sets in T cells from different treatment groups. As we expected, the MAPK cascade was significantly less enriched in T cells from the CKI27 group compared to the vehicle (Fig. 6I). Consistent with prior in vitro experiments, the phosphatidylinositol 3 kinase (PI3K) signaling and p38 MAPK cascade gene sets were also weakly enriched with the triple combination compared to CKI27 alone (Fig. 6J–K). These results suggested that intermittent MEKi combined with GITR co-stimulation and CTLA-4 blockade induces the expression of activation and cytotoxicity related genes in T cells from the TDLNs of LLC tumor–bearing mice. This supports the immunophenotypic switch toward an effector response observed by flow cytometry analysis. These observations also suggest that the addition of anti-GITR and anti-CTLA-4 may be an effective strategy to re-activate the antitumor immune response in Kras mutant cancers treated with MEKis.

Discussion

MEKis induce both positive and negative immunomodulatory effects that must be balanced to achieve optimal tumor control. MAPK/ERK inhibition has shown promise in clinical trials when combined with immunotherapy due to its favorable effects on the tumor microenvironment (16,41,42). Consistent with preclinical findings from studies investigating other MEKis, we confirmed that the MEKi CKI27 also dose-dependently increased tumor MHC expression, which was further amplified with the addition of IFNγ to mimic an inflammatory tumor microenvironment. (18,19,24). This resulted in increased T cell–mediated killing that was completely abolished with non-peptide presenting tumor cells and largely diminished by blocking MHC. MEK inhibition of tumor cells also produced extracellular ATP and HMGB1, which could activate DCs and mediate a pro-inflammatory tumor microenvironment (43). We also showed that MEK inhibition modulated expression of checkpoint ligands PD-L1, CD80, and CD86, which can have implications for combination with checkpoint blockade targeting PD-1 or CTLA-4. Oncogenic RAS signaling has been shown to stabilize PD-L1 mRNA and increase protein expression, which was reversible with MEK inhibition (27). Similarly, we observed a dose-dependent decrease of PD-L1 expression in CKI27-treated KRAS mutant tumor cell lines LLC, CT26, HCT-116, and AsPC-1. MEK inhibition has also been shown to increase CD80 and CD86 expression on DCs and tumor cells, aligning with our findings using various tumor cell lines (24,44).

Intermittent dosing regimens have been shown to delay tumor resistance mechanisms often associated with MAPK/ERK inhibitors (12,45). Using this strategy with CKI27, which prevented the relief of negative feedback in KRAS mutant tumor cells more efficiently than other MEKis, potently controlled tumor growth in LLC- and CT26-challenged mice. Several clinical trials have reported that combining BRAFis, MEKis and checkpoint blockade resulted in low response rates and significant hepatotoxicity (46,47). Introducing a break in treatment would allow for a higher dose to be given without additional toxicities and may even positively impact several immune populations (13). Compared to daily dosing, we observed a recovery of immune cell frequencies, such as APCs, in the spleen and an expansion in the TDLN of LLC tumor–bearing mice treated with an intermittent CKI27 regimen. This suggested a higher dose of MEKi potentially increased antigen presentation, and T-cell priming and expansion in the TDLN, which agreed with previous findings (21).

Adding T-cell co-stimulation has been shown to completely rescue the suppressive effects of MEK inhibition on T cells (24,25). Accordingly, we showed that agonist antibodies targeting GITR and OX-40 or GITR engagement further rescued T-cell inhibition when combined with MEKi washout in vitro, although only GITR co-stimulation resulted in full rescue. Similar to previous reports (24,25), phospho-proteins from the p38, JNK, and PI3k/Akt/mTOR pathways were upregulated on CD8⁺ T cells upon addition of GITR or OX-40 co-stimulation to MEKi. However, we did not reliably observe upregulation of the NF-κB pathway, which may be explained by the dynamic modulation of these pathways by co-stimulation (22). In LLC and CT26 tumor–bearing mice, the triple combination of intermittent MEKi, GITR co-stimulation, and CTLA-4 blockade significantly decreased tumor growth and prolonged survival. These antitumor effects were dependent on the adaptive immune response and specifically, CD8⁺ T cells, since they were abrogated in immunodeficient and CD8-depleted mice. Additionally, tumor re-challenged mice that were previously treated with the triple combination remained largely protected, indicated that this combination therapy elicited a long-term immunological memory response.

LLC and CT26 tumors both harbor Kras mutations but have distinct immune profiles (31). Overall, the effects of the therapies on immune cells were more heterogenous in CT26 tumor–bearing mice than in LLC tumor–bearing mice. This could be explained by the higher immunogenicity and increased TILs in CT26 tumors, which variably affected their tumor growth rate and differential response to therapies. The triple combination significantly altered the immune phenotypes of TDLNs from LLC tumor–bearing mice, underscoring the importance of T-cell priming and activation. TDLNs from both LLC- and CT26-challenged mice exhibited a reduction in all immune populations upon intermittent MEK inhibition with CKI27, which was mostly rescued with the addition of anti-GITR+anti-CTLA-4. The triple combination increased the expression of proliferation and activation markers on CD8⁺ and CD4⁺ Teffs and skewed them towards a memory phenotype, while leaving Tregs unaffected. Complementing our flow cytometry analyses, single-cell transcriptome sequencing analysis of TLDNs from LLC-challenged mice treated with the triple combination revealed higher counts of CD8⁺ and CD4⁺ T cells from the cytotoxic, activated, and differentiated/memory clusters. These T-cell clusters also upregulated expression of several effector, activation, and cytotoxicity related genes and had slightly enriched p38 and PI3k/Akt/mTOR signaling pathway signatures.

Further analyses of TILs from LLC-challenged mice revealed a slight increase in the CD8:Treg ratio with the triple combination along with phenotypic changes of Teffs that were consistent with those observed in the TDLNs. Our data on TILs from CT26-challenged mice correlated well with those published by several groups. CKI27 alone increased TILs and activation of Teffs in CT26 tumors as described with other MEKis (18,44); however, the triple combination increased TILs and certain markers to a lesser extent. We hypothesized that this may be because CKI27 increased GITR and CTLA-4 expression on T cells, leading to some depletion by anti-GITR and anti-CTLA-4. Treg depletion in the tumor, which is mediated largely by FcγRIV binding to the anti-GITR IgG2b and anti-CTLA-4 IgG2a Fc portions, has been shown to contribute to the antitumor effect of these immunotherapies. In agreement with our observations, both antibodies do not deplete Tregs in the TDLN, likely due to differences in expression levels (48,49). We observed only a slight depletion of Tregs in LLC and CT26 tumors with the triple combination, however their phenotype suggested that they were significantly destabilized (37). LLC and CT26 tumors are known to contain substantial amounts of MDSCs (31,50). Monocytic and granulocytic MDSCs were significantly decreased in LLC tumors by MEKi, agreeing with previous studies (51). In contrast, only granulocytic MDSCs decreased upon MEKi treatment, while Ly6C⁺ myeloid cells were increased in CT26 tumors. The same trend was observed in a prior work, where Ly6C⁺ myeloid cells in the tumor were classified as intermediary differentiating monocytes that a MEKi could prevent from becoming suppressive tumor-associated macrophages (TAMs) (44). It is important to highlight that although the recent development of direct KRAS^G12C inhibitors (52) can target tumors while sparing other immune cells, the off-target effects of MEKis on immunosuppressive Tregs, MDSCs, and TAMs can be advantageous, as we have shown here.

Together, our study and other work, provide evidence that T-cell co-stimulation can be effectively combined with MEKis to rescue the inhibitory effects on T cells (24,25,30). Specifically, we have shown that combination of a high dose/intermittent treatment regimen using a novel MEKi, CKI27, with GITR co-stimulation can further enhance T-cell activation through upregulating alternative signaling pathways. This led to a more durable tumor growth control and enhanced memory phenotype of T cells when combined with CTLA-4 checkpoint blockade. Finally, these data may also inform the design of future clinical trials with MEKis such that these strategies could reduce toxicities while increasing efficacy.

Synopsis.

Despite having several positive immunomodulatory effects, MEKis also suppress T-cell function. The authors show this suppression can be mitigated by combining an intermittent MEKi regimen with GITR co-stimulation for an improved antitumor effect with CTLA-4 checkpoint blockade.