Inhibition of Granulocytic Myeloid-Derived Suppressor Cells Overcomes Resistance to Immune Checkpoint Inhibition in LKB1-Deficient Non–Small Cell Lung Cancer

The development of immune checkpoint inhibitors (ICI) has transformed the therapeutic landscape of non–small cell lung cancer (NSCLC). Currently, anti–PD-1/PD-L1 treatment is the first-line therapy for patients with advanced-stage NSCLC, either as monotherapy or in combination with chemotherapy (1–3). Although anti–PD-1 therapy has led to improved survival in a subset of patients with NSCLC (2–4), the majority of patients do not respond and many develop resistance (5). Studies reveal that responses to ICIs are associated with preexisting CD8⁺ T cells (6) and high baseline PD-L1 expression in the tumor microenvironment (TME; ref. 7). Tumor mutational burden (TMB; ref 8) and clonal neoantigen burden (9) have also been associated with improved responses to ICIs. In contrast, factors that mediate resistance to PD-1/PD-L1 blockade remain poorly understood.

LKB1 inactivating mutations are found in 20% of patients with NSCLC and one third of KRAS-mutant NSCLCs (10). LKB1 is an upstream activator of the energy-sensing kinase AMPK, which inhibits mTOR signaling (11). Therefore, loss of LKB1 confers a direct proliferative advantage to cancer cells. Studies revealed that loss of LKB1 in NSCLC is associated with a distinct immune profile characterized by diminished T-cell infiltration, low or absent PD-L1 expression, and abundance of neutrophils in the TME (12–14). A recent landmark report demonstrated that inactivating mutations of LKB1 in NSCLC are the major genetic driver of primary resistance to anti–PD-1 therapy (13).

Herein, we report that loss of LKB1 induces heightened production of the ELR⁺ CXC chemokines in in vitro and in vivo lung cancer models. The abundance of the ELR⁺ CXC chemokines in Kras-mutant NSCLCs with LKB1 deficiency is associated with an enrichment of granulocytic myeloid-derived suppressor cells (G-MDSC) locally within the TME and systemically in peripheral blood and spleen of tumor-bearing mice. Depletion of G-MDSCs by anti–Gr-1 or functional blockade of G-MDSCs with ATRA reverses immune suppression and sensitizes LKB1-deficient tumors to PD-1 blockade. ATRA therapy represents a potential therapeutic strategy to overcome resistance to anti–PD-1 immunotherapy in LKB1-mutant NSCLC.

Immortalized human bronchial epithelial cell (HBEC) lines and HBEC-cancer cell pairs were generously provided by John D. Minna (UT Southwestern Medical Center, Dallas, TX; ref. 15) and cultured, as previously described (16). Stable HBEC cell lines containing either a control beta-galactosidase (βgal) or a mutant KRAS^V12 (KRAS) were generated, as previously described (17). NSCLC cell lines H441, H1838, H1793, H1568, H322, A549, and H838, as well as the prostate cancer cell line, MyC-CaP, were purchased from the ATCC and cultured in complete media [RPMI-1640 (Corning) with 10% FBS]. Murine NSCLC lines established from Kras^G12DTp53^−/−Luc (KP), Kras^G12DTp53^+/−Lkb1^−/−Luc (KPL) tumors were previously described (18). Cells were maintained at 37°C in a humidified incubator with 5% CO₂, and used before 5 passages. Cell lines were genotyped by STR profiling (DNA IQ System and Powerplex 1.2 system, Promega) and confirmed to be free of Mycoplasma (MycoAlert, Lonza).

Stable isogenic cell lines were established via transduction with retro- or lentivirus and selected with puromycin (EMD chemicals) for 10 to 14 days, as previously described (19). Short-hairpin RNA (shRNA) lentiviral constructs containing a nonsilencing sequence or a sequence against LKB1 (CATCTACACTCAGGACTTCAC; shLKB1) were purchased from Sigma. Retroviral constructs carrying empty vector (V), wild type LKB1 (WT-LKB1) or kinase dead LKB1 (KD-LKB1) were purchased from Addgene. For transient LKB1 silencing, cells were transfected with siRNA as previously described (19). SiRNAs against LKB1 (siLKB1) or scramble control (sictrl) were pooled siRNAs purchased from Dharmacon GE healthcare.

Luminescent-based cytokine screening was performed with a customized panel of 45 human cytokines/growth factors (Bio-Rad). Fresh medium was added to HBEC cells at 20% confluence and cells were incubated for 72 hours. Supernatants were collected and centrifuged to remove floating cells. Samples were analyzed in triplicate. Results were normalized to 1 × 10⁶ cells from matched samples.

Western Blots were performed, as previously described (19). Anti-LKB1 antibody was purchased from Cell Signaling Technology (Clone 27D10). Anti-GAPDH antibody was from Advanced Immunochemical Inc. (Clone 6C5).

Cells at 20% confluence were incubated for 72 hours in 1 mL of fresh media. CXCL1/CXCL8 concentrations in the supernatant were quantified with DuoSet ELISA (R&D Systems). Murine serum was prepared by resting 0.5 mL of whole blood at room temperature for 30 minutes followed by centrifugation at 2,000 × g for 10 minutes. Bronchoalveolar lavage (BAL) supernatant was harvested after samples were centrifuged at 2,000 × g for 10 minutes. Supernatants of cells transfected with siLKB1 or sictrl were collected 72 hours post-transfection. Bicinchoninic acid assay was performed on lysates from adherent cells to determine the protein concentration. ELISA results were normalized by protein concentration.

Total RNA was isolated using Quick-RNA MiniPrep (Zymo), and cDNA was synthesized with the High-Capacity RNA-to-cDNA Kit (Life Technologies). Expression levels of chemokine genes were examined by qPCR using gene-specific primers (Supplementary Table) and normalized to GAPDH (19).

RNA-seq and genotyping of cancer cell lines were downloaded from Broad Institute Cancer Cell Line Encyclopedia (CCLE. https://portals.broadinstitute.org/ccle, date: 12–24–2019). NSCLC cells were hierarchically classified into high and low clusters based on chemokine expression. Each chemokine was normalized to the geometric mean of its expression in all cells. The occurrence of KRAS and LKB1 double mutations in each cluster was compared using the Fisher exact test. Custom codes are available at https://github.com/Tian-hao/CCLE.

Studies with genetically engineered murine models (GEMM) used Lox-Stop-Lox Kras^G12D;Lkb1^Lox/Lox;Rosa26-Lox-Stop-Lox-Luc (KL), Lox-Stop-Lox Kras^G12D;Tp53^Lox/Lox;Rosa26-Lox-Stop-Lox-Luc (KP), and Lox-Stop-Lox Kras^G12D;Rosa26-Lox-Stop-Lox-Luc (K) mice on an FVB background (20). Lung tumors were induced by intranasal administration of 2.5 × 10⁶ plaque forming units (PFU) of Adeno-Cre (Gene Transfer Vector Core, University of Iowa) as previously described (21). KL and KP mice were euthanized 10 to 12 weeks after tumor induction and K mice 25 to 28 weeks after tumor induction. Lung tumors were dissected and subjected to downstream assays. For early-stage BAL studies, mice were euthanized 4 weeks after administration of 2.5 × 10⁷ PFUs of Adeno-Cre. Mice were housed in pathogen-free facilities at UCLA. All experimental procedures were approved by the UCLA Animal Research Committee.

FVB and NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice were purchased from Charles River Laboratories. Tumor cells were implanted in 7 to 9-week-old mice subcutaneously. To obtain murine blood, tumor-bearing mice were subjected to retro-orbital bleeding on days 9 and 14 and cardiac puncture at euthanasia on day 15. For therapeutic studies, mice bearing approximately 50 mm³ tumors were randomized and treated with 200 μg of anti–PD-1 antibody (BioXcell, Clone RMP1–14), 200 μg of anti–Gr-1 antibody (BioXcell, Clone BE0075) or isotype control via intraperitoneal injections 2 to 3 times per week. In ATRA studies, 200 μg of ATRA (dissolved in 10% DMSO, 10% Tween 80, and 80% PBS) was administered daily, starting on day 5. Bioluminescence images were obtained by IVIS Spectrum imager 10 minutes after intraperitoneal injection of D-luciferin (150 mg/kg). In secondary tumor challenge studies, mice were euthanized when tumor volume reached 1,200 mm³.

Single-cell suspension from murine spleens and tumors were prepared as previously described (18). To obtain circulating leukocytes, murine blood collected in Ethylenediaminetetraacetic acid (EDTA) was treated with red blood cell lysis buffer (BioLegend). Surface staining and intracellular staining for FOXP3 and Ki-67 were performed as previously described (18). IFNγ and TNFα production was evaluated by intracellular staining after in vitro stimulation with Cell Stimulation Cocktail (eBioscience) for 4 hours, using the intracellular fixation and permeabilization buffer set (eBioscience). Data acquisition was performed on Attune NxT cytometer (Thermo Fisher Scientific), and data analyzed by FlowJo software (TreeStar). Reagents and antibodies used are detailed in the Supplementary Table.

Cells were plated in culture media with or without ATRA in 96-well plates (1,000 tumor cells or 3,000 G-MDSCs per well) in triplicates. Proliferation was measured using ATPlite 1step Luminescence Assay Kit (PerkinElmer) every 24 hours up to 120 hours.

Single-cell suspension of 6 tumors from each group (ATRA-treated and untreated) were pooled 10 days after tumor inoculation. Live G-MDSCs (MHCII^loCD11b⁺Ly6C^loLy6G⁺) were sorted with BD FACSAria II (BD Biosciences). Splenic CD3⁺ T cells were isolated from naïve mice using magnetic negative selection (BioLegend). T cells were labeled with CellTrace Violet (Invitrogen) and plated in a 96-well U-bottom plate (1 × 10⁵ cells per well) in complete media. Flow-sorted G-MDSCs from control or ATRA-treated tumor-bearing mice were added at varying proportions to T cells and incubated overnight. The following day, T cells were treated with 1 × 10⁵ αCD3/αCD28 Dynabeads (Thermo Fisher Scientific) and 10 ng/mL murine IL2 (PeproTech). T-cell proliferation was evaluated by flow cytometry 4 days after stimulation. For in vitro treatment, sorted G-MDSCs from control mice were pretreated with 2 μmol/L ATRA for 24 hours. Culture media were replaced with fresh media to remove ATRA prior to coculturing with T cells.

Experiments were performed at least three times, unless otherwise indicated. Results from one representative experiment are shown. Statistical analyses were performed in Prism 9 (GraphPad) unless otherwise noted. Statistical significance was determined using an unpaired, parametric t test with 95% confidence interval. Results are reported as mean ± SEM, unless indicated. Tumor growth curve was analyzed using two-way ANOVA with Tukey after test for time-associated comparison among multiple groups. Statistical significance is reported as the following: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

To understand the effect of LKB1-loss on the production of cytokines in lung cancer, we first used immortalized HBECs that lack other cancer phenotypes and retain the ability to differentiate into normal bronchial epithelium (15). Given that LKB1 mutations frequently co-occur with KRAS mutations in human NSCLC, we stably knocked down LKB1 by shRNA in HBEC3 cells containing either a control βgal or a mutant KRAS vector (Fig. 1A). Screening a panel of commonly dysregulated cytokines and growth factors in human cancer revealed that LKB1 loss in HBEC3 cells increased the production of many of these factors (36/45 in βgal and 35/45 in KRAS). Notably, LKB1 knockdown resulted in upregulation of the ELR⁺ CXC chemokines Gro-α (CXCL1) and IL8 (CXCL8) in both βgal and KRAS HBEC3 cells (Fig. 1B). These results were validated by ELISA (Fig. 1C).

To evaluate whether this upregulation was broadly applicable, we transiently silenced LKB1 via siRNA in four additional parental HBEC lines (H2, H3, H4, and H7; Supplementary Fig. S1A) and found increased CXCL1 and CXCL8 expression at both RNA and protein levels (Fig. 1D). Expression of other ELR⁺ CXC genes, including CXCL2, CXCL3, and CXCL5, was also increased in all HBECs with LKB1 knockdown compared with controls, except for CXCL5, which was not detected in H3 and H4 (Fig. 1E). Transient LKB1 knockdown also resulted in elevated CXCL8 secretion in KRAS-transduced HBEC cell lines (H4-KRAS and H7-KRAS; Fig. 1F). We examined CXCL8 production in two sets of paired HBECs and cancer cells (HCC) derived from the same patient, HBEC4058/HCC4058 and HBEC4087/HCC4087 (Fig. 1G). Cancer cells produced higher amounts of CXCL8 at baseline compared with their HBEC counterparts, and LKB1-loss resulted in significantly increased CXCL8 secretion (Fig. 1G). These data demonstrate that loss of LKB1 induces ELR⁺ CXC chemokine expression in HBECs and cancer cells.

We stably knocked down LKB1 in the KRAS-mutant H441 NSCLC cell line and observed elevated expression of CXCL2, CXCL3, CXCL5, and CXCL8 at the transcriptional level and increased CXCL8 protein production by ELISA (Fig. 2A and Supplementary Fig. S1B). Similarly, transient knockdown of LKB1 in KRAS wild-type NSCLC cell lines, H1838 and H1793, led to a 2- to 9-fold increase in the expression of ELR⁺ CXC chemokines, as revealed by qPCR, except for CXCL8 expression that did not change in H1793 (Fig. 2B and Supplementary Fig. S1C). This observation may be due to the complex transcriptional regulation of CXCL8, which can be governed by promoter methylation (22).

Next, we re-expressed WT-LKB1 or KD-LKB1 in A549 cells, which harbor an intrinsic LKB1 mutation that results in the deletion of the LKB1 protein (Fig. 2C). Re-introduction of WT-LKB1 in A549 cells decreased the expression of CXCL1, CXCL2, CXCL3, CXCL5, and CXCL8 (Fig. 2C). A decrease in CXCL1 and CXCL8 protein production was demonstrated by ELISA (Fig. 2D). In contrast, the expression of ELR⁺ chemokines was not altered in A549 cells that expressed KD-LKB1 except for a statistically significant minor reduction in CXCL8. These findings indicate that the kinase activity of LKB1 is required for the suppression of ELR⁺ CXC chemokines. Similar results were observed with re-introduction of WT-LKB1 and KD-LKB1 in another LKB1-deficient NSCLC cell line H838, except for the expression of CXCL5 that was not altered by LKB1 status (Fig. 2E and Supplementary Fig. S1D). Methylation of the CXCL5 gene has also been reported as a regulatory mechanism (23). To assess whether LKB1 loss results in selective induction of ELR⁺ CXC chemokines, the expression of CXCL9 and CXCL10 was assessed in A549 and H838 cells. CXCL10 expression was not altered by introduction of WT-LKB1 in H838 cells (Supplementary Fig. S1E), whereas the expression of CXCL9 was below detection. CXCL9 and CXCL10 expressions were undetectable in isogenic A549 cell lines. Collectively, these findings establish LKB1 as a negative regulator of ELR⁺ CXC chemokine transcription in NSCLC cell lines, which requires the kinase activity of LKB1.

An unsupervised clustering analysis of chemokine expression in 138 NSCLC cell lines from CCLE (24) revealed that 16 of 18 cell lines with co-occurring KRAS/LKB1 mutations were stratified into Cluster 1 (C1) with high ELR⁺ CXC chemokine expression (P = 0.00394; Fig. 2F). However, LKB1 mutation alone did not lead to a significant C1 distribution of the cell lines (P = 1). This is likely due to the fact that multiple genomic alterations can upregulate the expression of ELR⁺ CXC chemokines in cancer cells (25–29). We observed that cancer cells consistently produce higher amounts of CXCL8 compared with their HBEC counterparts (Fig. 1G). These results suggest that KRAS and LKB1 co-mutation may serve as a genetic biomarker that predicts high expression of ELR⁺ CXC chemokines in human NSCLC.

In vivo expression of ELR⁺ CXC chemokines was examined using GEMMs of LKB1-deficient NSCLC. Increased expression of ELR⁺ CXC chemokines was observed in lung tumors from KL mice compared with KP or K mice (Fig. 3A). Given that ELR⁺ CXC chemokines mediate similar biological functions through common receptors (CXCR1 and CXCR2), an overall score that incorporated the expression of ELR⁺ CXC chemokines was calculated to reflect their cumulative abundance. The mean overall chemokine score in KL tumors was more than 4 log-fold higher compared with that in KP or K tumors (Fig. 3B). No difference was observed between KP and K tumors. We also evaluated the protein level of CXCL1 as a representative ELR⁺ CXC chemokine in the BAL fluid of KL and KP mice and noted a significantly higher level in KL mice (Fig. 3C and Supplementary Fig. S2A). Similarly, gene expression analysis of ELR⁺ CXC chemokines in multiple murine NSCLC cell lines generated from GEMMs (Supplementary Fig. S2B) revealed an increased expression and a higher overall chemokine score in KPL cell lines compared with that in KP cell lines (Fig. 3D).

Given that ELR⁺ CXC chemokines play a critical role in the trafficking of G-MDSCs from bone marrow to inflammatory sites (30, 31), we assessed whether LKB1-loss in GEMM tumors results in an enrichment of G-MDSC in the TME. A significantly higher percentage of G-MDSCs was observed within the immune population of the TME in KL tumors than KP tumors of comparable size (58.5% vs. 7.6%, P < 0.0001; Fig. 3E and Supplementary Fig. S2C). The content of monocytic-MDSCs (M-MDSCs) was less than 1% of the total immune cells in KL and KP tumors (Supplementary Fig. S2D). To further investigate the association between the level of ELR⁺ CXC chemokines and G-MDSC tumor infiltration, three KPL and two KP cell lines that possessed graded levels of ELR⁺ CXC chemokines (Fig. 3F) were used in syngeneic immunocompetent mice. The prevalence of G-MDSCs in the TME of these tumors strongly correlated with the expression levels of ELR⁺ CXC chemokines in the corresponding cell lines (r = 0.92; Fig. 3G and Supplementary Fig. S2E).

We then evaluated the kinetics of peripheral G-MDSC accumulation in mice bearing KPL and KP tumors. A gradual enrichment of G-MDSCs was observed in the blood of mice from days 9 to 14 after tumor inoculation, with a higher accumulation of G-MDSCs in mice bearing KPL tumors compared with the KP tumor-bearing mice (Fig. 3H). Similarly, splenic G-MDSCs were more abundant in tumor-bearing KPL mice compared with KP mice (Fig. 3I). A higher percentage of CXCR2⁺G-MDSCs was observed in the blood and the spleen of KPL tumor-bearing mice compared with that of KP tumor-bearing mice (Supplementary Fig. S2F). A heightened level of circulating CXCL1 protein was also observed in the blood of mice with KPL tumors compared with those with KP tumors (Supplementary Fig. S2G). These data suggest that LKB1 deficiency in murine models of NSCLC leads to a heightened production of ELR⁺ CXC chemokines and is associated with local and systemic enrichment of G-MDSCs.

MDSCs have emerged as potent suppressive immune mediators that facilitate tumorigenesis, metastasis, and resistance to therapy (32). We hypothesized that G-MDSCs mediate resistance to anti–PD-1 therapy in LKB1-deficient murine NSCLC. To test this hypothesis, the capacity of G-MDSC depletion via antibody to enhance the efficacy of anti–PD-1 therapy was evaluated in syngeneic murine models bearing 1940A KPL tumors (designated as KPL-P) that contain abundant G-MDSCs (Fig. 3G). No antitumor efficacy was observed with anti–Gr-1 and anti–PD-1 mono- or combination therapies (Supplementary Fig. S3A). The lack of efficacy in this study could be due to the paucity of tumor neoantigens, given that GEMMs of NSCLC possess low TMB (33, 34). Therefore, 1940A KPL cells with increased TMB (designated as KPL-3M), induced by in vitro exposure to the carcinogen N-methyl-N-nitrosourea (MNU), were used (18). KPL-3M cells had analogous expression of Cxcl1, Cxcl3, Cxcl5, and Cxcl7 chemokines compared with KPL-P cells, and an increase in Cxcl2 expression (Supplementary Fig. S3B). Anti–PD-1 monotherapy in KPL-3M tumor-bearing mice elicited limited efficacy and anti–Gr-1 monotherapy did not affect tumor growth; however, combination therapy with anti–Gr-1 and anti–PD-1 resulted in a synergistic antitumor effect, leading to a complete eradication of 88% of tumors and growth stabilization of remaining tumors (Fig. 4A).

To understand the mechanisms of the antitumor immunity in this model, tumors were immunophenotyped 10 days after treatment initiation when tumor weights were comparable (Supplementary Fig. S3C). Treatment with anti–Gr-1 antibody led to a robust depletion of tumor-infiltrating G-MDSCs (Fig. 4B). To account for the variability introduced by G-MDSC depletion, each immune cell subtype was normalized relative to the number of tumor cells. Anti–Gr-1 treatment did not lead to changes in tumor-associated macrophages (TAM), conventional DCs (cDC), natural killer (NK) cells, FOXP3⁻CD4 or CD8 T cells in the TME (Fig. 4B), but it did enhance the proliferation and function of FOXP3⁻CD4 and CD8 tumor-infiltrating lymphocytes (TIL), as determined by intracellular staining of Ki-67 and TNFα/IFNγ, respectively (Fig. 4C). Anti–Gr-1 monotherapy also enhanced NK proliferation and IFNγ secretion (Fig. 4D). Anti–PD-1 monotherapy also led to an increase in CD8 T cells secreting TNFα and IFNγ (Fig. 4C). Combination therapy with anti–Gr-1 and anti–PD-1 resulted in a significant decrease of G-MDSCs compared with the control, although not as robust as anti–Gr-1 monotherapy (Fig. 4B). A significant increase was observed in TAMs, cDCs, NKs, FOXP3⁻CD4, and CD8 T cells following combination therapy (Fig. 4B). Importantly, combination therapy resulted in the most significant improvement in proliferation and cytokine secretions of both NK and T cells (Fig. 4C and D). This observation is consistent with the robust antitumor immunity observed in response to combination therapy. Upregulation of PD-L1 in tumor cells was also noted following all treatments, with the highest magnitude observed with combination therapy (Fig. 4E). These data suggest that G-MDSC depletion via anti–Gr-1 antibody enhances the proliferation and function of TILs and sensitizes LKB1-deficient tumors to anti–PD-1 therapy.

ATRA was assessed to determine whether a pharmacological agent that interferes with MDSC function could be implemented to overcome resistance to anti–PD-1 in LKB1-deficient NSCLC. ATRA alters MDSC function via down regulation of deleterious reactive oxygen species (ROS; ref. 35). Combination therapy with ATRA and anti–PD-1 resulted in robust antitumor efficacy with the eradication of over 70% of KPL-3M tumors (8/11) and the growth stabilization of remaining tumors (Fig. 5A). ATRA monotherapy led to modest efficacy similar to that of anti–PD-1 monotherapy.

Because ATRA can potentially inhibit tumor proliferation, studies were performed to determine whether the efficacy of ATRA monotherapy in this model is driven by the immune-modulatory effect of ATRA or its direct inhibition of tumor growth. Treatment of KPL-3M cells with ATRA in vitro did not inhibit tumor cell proliferation (Supplementary Fig. S4). Similarly, no differences in tumor growth were observed in KPL-3M tumor-bearing immunodeficient NSG mice treated with either vehicle or ATRA (Fig. 5B). These data suggest that the reduction of tumor growth by ATRA in immunocompetent mice is predominantly immune-mediated.

To assess the immune determinants of response to ATRA and combination therapy, the immune composition of the TME was evaluated 12 days following the initiation of treatment when tumor weights were generally comparable (Supplementary Fig. S5A). Evaluation of the myeloid compartment of the TME (Supplementary Fig. S5B) revealed that ATRA and anti–PD-1 mono- and combination therapies resulted in decreased G-MDSCs with the largest magnitude change observed in combination therapy (Fig. 6A). ATRA monotherapy did not lead to changes in the frequency of M-MDSC, TAM, cDC1, cDC2, or monocyte-derived inflammatory DC (inf DC) compared with vehicle control (Supplementary Fig. S5C). There was a significant increase in M-MDSCs and TAMs upon ATRA and anti–PD-1 combination therapy compared with vehicle control, but no difference in the DC populations (Supplementary Fig. S5C).

Within the lymphoid compartment of the TME, ATRA, and anti–PD-1 mono- and combination therapies led to a significant increase in NK cells and TILs, including CD8, FOXP3⁻CD4, and CD4⁺CD25⁺FOXP3⁺ regulatory T (Treg) cells (Fig. 6B). The increase in TILs and NK cells was most pronounced in the combination group. The activation, proliferation, and function of TILs were also assessed. ATRA monotherapy increased the number of CD44⁺PD-1⁺TIM3⁻FOXP3⁻CD4 and CD44⁺PD-1⁺TIM3⁻CD8 T cells, which represent activated antigen-experienced TILs that are not terminally exhausted (Fig. 6C; ref. 36). Combination therapy of ATRA and anti–PD-1 resulted in a greater increase in these populations compared with ATRA monotherapy. ATRA alone and in combination with anti–PD-1 enhanced the proliferation of CD4 and CD8 TILs, with the greatest increase noted in the combination group (Fig. 6D). ATRA and anti–PD-1 combination therapy also led to a significantly higher percentage of PD-1⁺FOXP3⁻CD4 and PD-1⁺CD8 T cells secreting TNFα and IFNγ (Fig. 6E). Increased proliferation and IFNγ secretion were also noted in tumor-infiltrating NK cells following combination therapy (Fig. 6F). Similarly, the highest magnitude of PD-L1 upregulation by tumor and myeloid cells was observed in the ATRA and anti–PD-1 combination group (Supplementary Fig. S5D and S5E). These data suggest that ATRA therapy increases TILs and enhances their activation and proliferation. The addition of anti–PD-1 to ATRA therapy results in further enhancement of the proliferation and augmentation of the effector function of these TILs.

To assess whether ATRA modulation of G-MDSCs mediated the observed enhancement of T-cell proliferation in the TME, tumor-infiltrating G-MDSCs from KPL-3M tumor-bearing mice treated with ATRA or vehicle were flow-sorted. Coculturing of G-MDSCs with T cells revealed that G-MDSCs from ATRA-treated tumor-bearing mice had reduced T-cell suppressive capacity compared with G-MDSCs from untreated tumor-bearing control mice (Fig. 6G). Similarly, in vitro ATRA treatment of G-MDSCs from control mice diminished the suppression of T-cell proliferation compared with untreated G-MDSCs (Supplementary Fig. S5F). ATRA also exhibited a minor inhibition on GMDSC proliferation in vitro (Supplementary Fig. S5G). These data suggest that ATRA-induced potentiation of anti–PD-1 efficacy is partially mediated through direct ATRA suppression of G-MDSC proliferation and function.

A significant reduction in G-MDSCs and M-MDSCs was observed in the spleen of KPL-3M tumor-bearing mice following ATRA monotherapy and combination therapy, but this was not evident with anti–PD-1 monotherapy (Fig. 7A). No difference was noted in the frequency of macrophages with treatments (Supplementary Fig. S6A). Evaluation of the T-cell subgroups in the spleen revealed no differences in the overall abundance of FOXP3⁻CD4, CD8 or Tregs following any treatment (Supplementary Fig. S6B). An increase in CD44⁺CD62L⁻ effector memory (EM) T cells with a higher percentage of PD-1⁺ cells was observed within the FOXP3⁻CD4 T-cell population, following anti–PD-1 mono- or combination therapies (Fig. 7B). Neither the proliferation nor the cytokine secretion (TNFα and IFNγ) of PD-1⁺EM CD4 cells were altered by therapy (Fig. 7B and Supplementary Fig. S6C). There was no difference in the overall frequency of CD8 EM T-cell populations among treatment groups, but a higher percentage of PD-1⁺ cells was observed within the CD8 EM T-cell population following anti–PD-1 mono- or combination therapies (Fig. 7C). No difference in the proliferation of PD-1⁺ CD8 EM cells was observed following treatments, but ATRA mono- and combination therapies resulted in an increase in cytokine-secreting cells (Fig. 7C and Supplementary Fig. S6D). A higher percentage of CD44⁺CD62L⁺ central memory (CM) CD8 T cells was found in response to ATRA and anti–PD-1 combination therapy. CM CD8 cells demonstrated enhanced proliferation following ATRA and anti–PD-1 mono- or combination therapies (Fig. 7D). A significantly higher level of IFNγ secretion in splenic NK cells was also noted following ATRA mono- or combination therapies (Supplementary Fig. S6E). These data provide compelling evidence that combination therapy with ATRA and anti–PD-1 results in enhanced systemic T-cell responses in KPL-3M tumor-bearing mice.

Next, we assessed whether the antitumor efficacy associated with combination therapy of ATRA and anti–PD-1 resulted in the establishment of systemic tumor-specific T-cell immunity. Mice with complete tumor eradication following therapy with ATRA and anti–PD-1 combination were re-challenged with KPL-3M cells 3 months after the complete antitumor response. Naïve mice were included as controls. Although naïve mice succumbed to tumor progression, the re-challenged mice eradicated 100% of tumors following an initial tumor growth (Fig. 7E). In contrast, a secondary challenge with a distinct cancer cell line from the same FVB background (MyC-CaP) resulted in tumor growth in all the cured and naïve control mice (Fig. 7F). These data indicate that ATRA therapy in combination with anti–PD-1 elicits durable systemic antitumor immunity.

Loss of LKB1 is a primary driver of resistance to anti–PD-1 therapy in KRAS-mutant NSCLC (13). Studies indicate that inactivation of LKB1 in NSCLC is associated with a distinct TME characterized by diminished T-cell infiltration, low or absent PD-L1 expression, and abundance of neutrophils (12–14). Our work extends the previous studies in several key areas. We report that: (i) LKB1 is a negative regulator of ELR⁺ CXC chemokine transcription in HBECs, murine and human NSCLC cells; (ii) heightened levels of ELR⁺ CXC chemokines in murine LKB1-deficient NSCLC models correlate with increased systemic and local G-MDSC enrichment; (iii) depletion of G-MDSCs via anti–Gr-1 antibody or their functional inhibition via ATRA activates TILs and sensitizes LKB1-deficent tumors to PD-1 blockade; (iv) anti–PD-1 combined with ATRA therapy induces durable antitumor T-cell responses and systemic antitumor immunity. These findings enhance our understanding of the critical mediators of immunosuppression in LKB1-deficient NSCLC and provide preclinical evidence for combining PD-1 blockade with therapies that interfere with MDSCs.

Although prior murine and human studies reveal an enrichment of neutrophils in LKB1-deficient NSCLC (12, 14), the molecular mediators of this phenotype are not well understood. Koyama and colleagues (12) identified that loss of LKB1 in murine models of NSCLC results in an increased secretion of cytokines involved in neutrophil production and accumulation, including G-CSF, IL6, and CXCL7. Our findings further establish LKB1 as a negative regulator of the ELR⁺ CXC chemokine family in HBECs as well as murine and human lung cancer cell lines. We found a robust positive correlation between the increased tumor expression of ELR⁺ CXC chemokines and the frequency of G-MDSCs in the TME and the periphery of tumor-bearing mice, implicating ELR⁺ CXC chemokines as potential mediators of G-MDSC/neutrophil accumulation in LKB1-deficient NSCLC. Previous studies have revealed Hippo and Wnt signaling as two oncogenic pathways downstream of LKB1 that regulate ELR⁺ CXC chemokines (37–41). Future mechanistic studies are necessary to delineate the specific molecular mechanisms of LKB1-mediated regulation of ELR⁺ CXC chemokines in NSCLC.

Our current studies support the hypothesis that G-MDSC accumulation mediates resistance to immunotherapy in LKB1-defecient murine NSCLC. This is supported by the finding that G-MDSC depletion via anti–Gr-1 antibody resulted in improved proliferation and cytotoxic function of NK and T cells in the TME and sensitized LKB1-deficient tumors to anti–PD-1 therapy. This hypothesis is further corroborated by clinical studies in patients with NSCLC that implicate G-MDSC/neutrophil and ELR⁺ CXC chemokines as mediators of resistance to ICIs. Kim and colleagues (42) demonstrated that a high frequency of circulating LOX-1⁺ G-MDSCs and low frequency of Treg cells in patients with NSCLC are associated with a decreased progression-free survival following treatment with Nivolumab. Kargl and colleagues (43) reported that neutrophils are negatively correlated with CD4 and CD8 T-cell content in NSCLC TME, and that a high neutrophil to CD8 T-cell ratio is associated with worse outcomes in patients with NSCLC treated with ICIs (44). A recent report by Schalper and colleagues (45) demonstrated that patients with NSCLC with elevated serum levels of CXCL8 have greater accumulation of intratumoral neutrophils and reduced clinical benefits in response to ICIs. Consistent with these findings, there are several early-phase clinical trials in NSCLC evaluating the combination of ICIs with therapies that target G-MDSC/neutrophil trafficking into tumors (46).

We speculated that pharmacological treatment with ATRA would allow for a translatable approach to ameliorate MDSC-mediated immunosuppression and augment host antitumor immune responses. ATRA is a widely used drug that induces the differentiation of immature myelocytic cells and is a standard component of induction therapy for patients with acute promyelocytic leukemia (47). Our studies in the KPL-3M murine model revealed that ATRA synergized with anti–PD-1 and resulted in systemic tumor-specific immunity. Mechanistically, ATRA did not appear to have direct antitumor cytotoxicity, based on the lack of in vivo efficacy in NSG mice and the absence of inhibition on tumor proliferation in vitro. In the immunocompetent models, treatment with ATRA resulted in profound remodeling of the local and systemic immune landscape, including a reduction in splenic and tumor G-MDSCs, and an increase in tumor-infiltrating NK cells and antigen-experienced TILs with an enhanced proliferation capacity. Our studies provide compelling evidence that the efficacy of ATRA treatment in LKB1-deficient NSCLC is partially mediated through ATRA-mediated suppression of G-MDSC proliferation and function. These results add to the previous reports demonstrating that ATRA can promote differentiation of MDSC into macrophages and DCs and decrease the suppressive effects of MDSCs (35, 48, 49).

In summary, we report that loss of LKB1 in NSCLC induces heightened production of ELR⁺ CXC chemokines. The abundance of ELR⁺ CXC chemokines in LKB1-deficient NSCLC murine models correlates with an enrichment of G-MDSCs, which drives resistance to anti–PD-1 therapy. The removal of G-MDSC–mediated immunosuppression via depletion with anti–Gr-1 antibody, or functional inhibition with ATRA, restores antitumor T-cell responses and sensitizes resistant LKB1-deficient NSCLC to anti–PD-1 therapy. These results provide preclinical evidence in support of the clinical investigation of ATRA therapy to augment the efficacy of anti–PD-1 immunotherapy in patients with LKB-1–deficient NSCLC.

S.M. Dubinett reports other support from Early Diagnostics, Johnson & Johnson Lung Cancer Initiative, LungLife AI, Inc., and T-Cure Bioscience, Inc. outside the submitted work. No disclosures were reported by the other authors.

R. Li: Conceptualization, data curation, formal analysis, investigation, writing–original draft. R. Salehi-Rad: Conceptualization, data curation, formal analysis, investigation, writing–original draft, writing–review and editing. W. Crosson: Data curation, writing–review and editing. M. Momcilovic: Data curation, investigation. R.J. Lim: Data curation, investigation. S.L. Ong: Data curation, formal analysis, investigation. Z.L. Huang: Data curation, formal analysis, investigation. T. Zhang: Formal analysis. J. Abascal: Data curation, investigation. C. Dumitras: Data curation, investigation. Z. Jing: Data curation, investigation. S.J. Park: Data curation, formal analysis, investigation. K. Krysan: Supervision, writing–original draft, writing–review and editing. D.B. Shackelford: Supervision. L.M. Tran: Formal analysis, supervision. B. Liu: Supervision, investigation, writing–original draft, writing–review and editing. S.M. Dubinett: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.

The authors thank John Minna for providing HBEC lines and both Lauren Winter and Tamara Silva for administrative support. This research was supported in part by funding from the National Cancer Institute (1U01-CA196408), National Heart Lung and Blood Institute (T32-HL072752), Medical Research Funds from the Department of Veteran Affairs, and NIH/NCATS UL1TR001881. The UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core is supported by the National Institutes of Health (NIH; #CA-16042 and #AI-28697).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.