Immunotherapy has shown limited efficacy in patients with EGFR-mutated lung cancer. Efforts to enhance the immunogenicity of EGFR-mutated lung cancer have been unsuccessful to date. Here, we discover that MET amplification, the most common mechanism of resistance to third-generation EGFR tyrosine kinase inhibitors (TKI), activates tumor cell STING, an emerging determinant of cancer immunogenicity (1). However, STING activation was restrained by ectonucleosidase CD73, which is induced in MET-amplified, EGFR-TKI–resistant cells. Systematic genomic analyses and cell line studies confirmed upregulation of CD73 in MET-amplified and MET-activated lung cancer contexts, which depends on coinduction of FOSL1. Pemetrexed (PEM), which is commonly used following EGFR-TKI treatment failure, was identified as an effective potentiator of STING-dependent TBK1-IRF3-STAT1 signaling in MET-amplified, EGFR-TKI–resistant cells. However, PEM treatment also induced adenosine production, which inhibited T-cell responsiveness. In an allogenic humanized mouse model, CD73 deletion enhanced immunogenicity of MET-amplified, EGFR-TKI–resistant cells, and PEM treatment promoted robust responses regardless of CD73 status. Using a physiologic antigen recognition model, inactivation of CD73 significantly increased antigen-specific CD8+ T-cell immunogenicity following PEM treatment. These data reveal that combined PEM and CD73 inhibition can co-opt tumor cell STING induction in TKI-resistant EGFR-mutated lung cancers and promote immunogenicity.
MET amplification upregulates CD73 to suppress tumor cell STING induction and T-cell responsiveness in TKI-resistant, EGFR-mutated lung cancer, identifying a strategy to enhance immunogenicity and improve treatment.
EGFR-mutated non–small lung cancers (NSCLC) effectively respond to osimertinib and other third-generation EGFR tyrosine kinase inhibitors (EGFR-TKI); however, most patients present at an advanced stage and almost invariably develop acquired drug resistance (2). Approaches to combat EGFR-TKI resistance have centered around targeting downstream or parallel mitogenic signaling pathways such as MEK, MET, and IGF1R, which can compensate for and/or bypass EGFR-TKI inhibition (3). However, this strategy has been limited by the presence of drug-tolerant persister cells and the diversity of genetic and epigenetic mechanisms that can facilitate expansion of resistant clones (4).
MET amplification and/or overexpression occurs in up to 25% of patients with NSCLCs overall (5, 6). It represents one of the most common resistance mechanisms to osimertinib, and along with MET exon 14 mutation represents a distinct targetable subtype of lung adenocarcinoma (7, 8). Several clinical trials examining the efficacy of combination therapy with MET inhibitors for EGFR-mutated NSCLCs are currently underway. In these trials, adverse events were found to be more severe than with EGFR-TKIs alone, although within acceptable limits. Combination therapy with the MET inhibitor savolitinib has advanced into a phase III trial and improves response of osimertinib resistant patients, but median progression-free survival in the phase II study was only 5.5 months, highlighting the limited durable activity of directly targeting this compensatory signaling pathway (9–11).
When T-cell immunity can be successfully engaged against lung cancer, responses are substantially more durable than for targeted therapies (12, 13). However, studies of immune checkpoint blockade (ICB) have revealed that patients with EGFR-mutated lung cancer rarely ever respond to immunotherapy, despite its impressive overall success in other subsets of NSCLC (14). Low tumor mutational burden and PD-L1 expression, along with low rate of CD8+ tumor-infiltrating T cells are considered to be reasons for the failure of response to ICB, but the underlying mechanisms remain unknown (15, 16). On the other hand, the oncogenic activation of MET promotes immune escape by upregulating different transcripts such as CD274 (encoding PD-L1) or SOCS1, among others, involved in immunosuppression (17).
Multiple studies have demonstrated an emerging role for innate immune signaling pathway dysregulation in EGFR-mutated lung cancer, and more specifically adaptive resistance to EGFR TKIs. These finding suggest that understanding and characterizing specific mechanisms of innate immune pathway dysregulation could uncover novel immunotherapeutic opportunities. For example, activation of STAT3 and NFκB signaling is implicated in promoting survival in response to EGFR-TKI treatment (18, 19). In addition, our group previously reported induction of STING, a critical downstream sensor of cytosolic double-stranded DNA (dsDNA) and a key component of the innate immunity, in the EGFR-TKI drug-tolerant persister cell state, together with secretion of specific proinflammatory cytokines (20). An EGFR-TKI–induced antiviral response involving upregulation of type I IFN signaling was also recently described by Gong and colleagues (21), highlighting the potential of such states to prime antitumor immunity. These data suggest that the residual cells that actually mediate acquired resistance to targeted therapy in EGFR-mutated tumors could be ideal targets for harnessing the innate immune response to promote tumor clearance.
However, induction of counterregulatory factors such as CD73, which generates the immunosuppressive metabolite adenosine, may suppress tumor immunogenicity (22), particularly in EGFR-mutated NSCLC (23, 24). Human CD73 is encoded by NT5E, which is located on the long arm (q) of chromosome 6. The canonical pathway leading to extracellular adenosine production involves the degradation of extracellular ATP into AMP and subsequently into adenosine, by the action of membrane-bound ectonucleotidases CD39 and CD73, respectively (25, 26). Extracellular adenosine (eADO) can signal through four different adenosine receptors, designated A1, A2a, A2b, and A3, dampening the immune response by suppressing immune effector cell function (22). More specifically, an increase of eADO, mainly through the adenosine A2A receptor signaling, impairs T-cell cytotoxicity and natural killer cell function and induces the suppression of antigen-presenting cells (27, 28). All of these molecules, among others, contribute to adenosinergic signaling in the tumor microenvironment, and their overexpression has been associated with metastatic disease and worse clinical outcomes in various tumor types, including NSCLC (29–31).
In this study, we sought to characterize STING expression in established EGFR-TKI–resistant lung cancer cells, discovering that its upregulation is linked to MET amplification. However, while examining antigen-specific T-cell activation in this MET-activated cellular context, we discovered a key role for CD73 in restraining STING-driven immunogenicity, prompting us to examine the potential of activating STING and inhibiting CD73 to overcome this barrier.
Materials and Methods
HEK293T, H1650, H441, and H3255 cells were purchased from ATCC. H23, H1975, and H820 cells were originally obtained from the Broad Institute, and H69 and H69M were originally obtained from the laboratory of Dr. Joan Albanell and were authenticated by short tandem repeat genotyping. HCC827, HCC827GR6, HCC827EPR, PC9, HCC2935, HCC2279, HCC4006, DFCI-202, EBC1, H596, and DFCI-633 cells were kindly provided by Dr. P.A. Jänne at Dana-Farber Cancer Institute (DFCI), Boston, MA. Jurkat76 cell line was provided by M.H.M Heemskerk at Leiden University Medical Center, Leiden, the Netherlands. HEK293T cell was cultured in DMEM (Thermo Fisher Scientific, catalog no. 11965-118), 10% FBS (GeminiBio products, catalog no. 100-106), and 1% penicillin/streptomycin (Gibco, #15140-122). H1650, H3255, H1975, H820, HCC827, HCC827GR6, HCC827EPR, PC9, HCC2935, HCC2279, HCC4006, H69, H69M, H441, EBC1, H596, H23, DFCI-202, DFCI-633, and Jurkat76 cell lines were cultured at 37°C in 5% CO2 using RPMI1640 (Thermo Fisher Scientific, catalog no.11875-119), 10% FBS and 1% penicillin/streptomycin. Mycoplasma infection was regularly checked by PCR using the condition media derived from each cell line. The sequences of the primers used for checking Mycoplasma infection are listed in Supplementary Table S1.
Reagents and treatments
The following reagents were used: osimertinib (Selleckchem, catalog no. S7297), savolitinib (Selleckchem, catalog no. S7674), pemetrexed (Selleckchem, catalog no. S5971), docetaxel (Selleckchem, catalog no. S1148), vinorelbine (Selleckchem, catalog no. S4269), etoposide (Selleckchem, catalog no. S1225), Olaparib (Selleckchem, catalog no. S1060), barasertib (Selleckchem, catalog no. S1147), cisplatin (Selleckchem, catalog no. S1166), ruxolitinib (Selleckchem, catalog no. S1378), trametinib (Selleckchem, catalog no. S2673), crizotinib (Selleckchem, catalog no. S1068), AB680 (MedChemExpress, catalog no. HY-125286), adenosine (Sigma-Aldrich, catalog no. A9251), AMP (Sigma-Aldrich, catalog no. A1752), APCP (Sigma-Aldrich, catalog no. M8386), Recombinant Human HGF (R&D Systems, catalog no. 294-HG-005/CF).
Patient samples and IHC staining
Patients with NSCLC with EGFR mutation were identified through the DFCI PROFILE database. Formalin-fixed paraffin-embedded with 4-μm-thick were stained at the Division of Pathology in Brigham & Women's Hospital Pathology Core. Staining for the following antibodies was performed on BOND-III, the fully automated IHC and ISH stainer (Leica Biosystems): STING (Cell Signaling Technology, catalog no. 13647, dilution 1:50), CD73 (Cell Signaling Technology, catalog no. 13160, dilution 1:50), c-MET (Abcam, catalog no. ab243930, dilution 1:100). Poly-HRP IgG reagent from Bond Polymer Refine Detection Kit DC9800 was used to bind rabbit antibody. Expression levels of STING, CD73, and c-MET staining in tumor cells were visually evaluated by a pathologist who was blinded to other data. We used a IHC score system (H-score) that was calculated by multiplying the membranous intensity score in tumor cells (0, absent; 1, weak; 2, moderate; 3, strong) by the percentage of stained cells (0%–100%) to yield a value of 0–300 m and classified into high versus low expression at the median cut-off point (32).
CRISPR/Cas9 system and lentiviral infection
Target sequences for CRISPR interference were designed using the single-guide RNA (sgRNA) designer (http://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design). A nontargeting sgRNA from the Gecko library v2 was used as a scramble sgRNA. sgRNA target sequences are listed in Supplementary Table S1.
A total of 3 × 106 HEK293T cells were plated onto a 60-mm dish and transfected using X-tremeGENE HP DNA Transfection Reagent (Roche, catalog no. 06366236001) with 1 μg of lentivirus-based expression vectors together with 1 μg of pCMV-dR8.91 and 1 μg of pCMV-VSV-G. After 48 hours incubation, the media containing lentivirus particles were collected, passed through a 0.45 μmol/L filter, and concentrated using Lenti-X Concentrator (Clontech, catalog no. 631231). The constructs of TCR-LMP2B were purchased from VectorBuilder company. For selection of virally infected cells, 0.5–2 μg/mL of puromycin (pCRISPR-v2 sgRNAs, plx307-hMET) or 1.5–8 μg/mL of blasticidin (plx304-NanoLuc, plx304-hSTING, plx304-CD73, or plx304-hFRA1) was used 24 hours postinfection.
RNA extraction was performed using RNeasy Mini Kit (Qiagen, catalog no. 74106). RNA samples (1 μg) were reverse transcribed using SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific, catalog no. 1683483). qRT-PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific, catalog no. 4367659). The sequences of the primers used for qRT-PCR are listed in Supplementary Table S1.
Cells were lysed in RIPA lysis and extraction buffer (Thermo Fisher Scientific, catalog no. 89900) containing 1× protease inhibitors (Roche, catalog no. 11-836-145-001) and phosphatase inhibitor (50 μmol/L NaF and 100 μmol/L Na3VO4). Immunoblotting was performed as described previously (33) using antibodies that specifically recognize MET (Cell Signaling Technology, catalog no. 8198), Phospho-MET (Cell Signaling Technology, catalog no. 3077), STING (Cell Signaling Technology, catalog no. 13647), CD73 (Cell Signaling Technology, catalog no. 13160), TBK1 (Cell Signaling Technology, catalog no. 3013), Phospho-TBK1 (Cell Signaling Technology, catalog no. 5483), IRF3 (Cell Signaling Technology, catalog no. 11904), phospho-IRF3 (Cell Signaling Technology, catalog no. 4947), STAT1 (Cell Signaling Technology, catalog no. 9172), phosphor-STAT1 (Cell Signaling Technology, catalog no. 9167), ERK (Cell Signaling Technology, catalog no. 9107), phosphor-ERK (Cell Signaling Technology, catalog no. 4370), AKT (Cell Signaling Technology, catalog no. 9272), phosphor-AKT (Cell Signaling Technology, catalog no. 4060), FRA1 (Abcam, catalog no. 124722), phospho-FRA1 (Cell Signaling Technology, catalog no. 5841), cGAS(Cell Signaling Technology, catalog no. 15102), and β-actin (Cell Signaling Technology, catalog no. 3700). Secondary antibodies were from LI-COR Bioscience: IRDye 680LT Goat anti-Mouse IgG (#926-68020), IRDye 800CW Goat anti-Rabbit IgG (#926-32211). Imaging of blots was performed using LI-COR Odyssey system.
Flow cytometry assay
Cells in culture were examined for surface expression by flow cytometry with BD Biosciences LSRFortessa and the following antibodies: anti-HLA-A,B,C (BioLegend, catalog no. 311436), anti-CD73 (BioLegend, catalog no. 344004), anti-CD39 (R&D Systems, catalog no. FAB4397A), anti-PD-L1 (BioLegend, catalog no. 329718), anti-CD45 (BioLegend, catalog no. 368511) or isotype IgG control antibodies (BioLegend, catalog no. 400268, BioLegend, catalog no. 400114, R&D Systems, catalog no. IC002A, BioLegend, catalog no. 400326, BioLegend, catalog no. 400122). LMP2 Tetramer-SSCSSCPLSK was purchased from MBL. Cells resuspended in 100 μL PBS containing 3% FBS were stained by each antibody. FlowJo v10 was used to perform analysis of flow cytometry raw data.
Cell viability assay
A total of 2,000–5,000 cells were seeded in 96-well plates, allowed to attach overnight, and then incubated with growth media containing drugs as indicated for 96 hours. Values of CellTiter-Glo (CTG) Luminescent Cell Viability assay (Promega) after 96 hours were normalized to vehicle-treated cells. Plates were read on a Tecan Infinite M200 Pro plate reader and analysis was performed using Prism7 (GraphPad Software). All conditions were tested in triplicate. The values represent the average of three technical replicates and a representative experiment from at least two independent experiments (biological replicates).
We treated each cell line with DNA-damaging compounds for 48 hours. Viable cells were replated after treatment and after 24 hours cell culture medium was replaced with medium containing 1% FBS. After 72 hours, 50 μmol/L AMP was added 30 minutes before collecting conditioned medium (CM) and then adenosine quantification was performed according to the instructions in Adenosine Assay Kit (Abcam, catalog no. ab211094). All values were normalized to CM containing 1% FBS.
A total of 2–5 × 105 cells were plated onto a 6-well plate and transfected using X-tremeGENE HP DNA Transfection Reagent (Roche, catalog no. 06366236001) with the indicated amount of poly (dA:dT; Invivogen, catalog no. tlrl-patn).
siRNAs targeting for MET (s8702, catalog no. 4390824) and negative control (catalog no. 4390843) were purchased from Thermo Fisher Scientific. Cells were transfected with 40 nmol/L of the siRNAs using Lipofectamine RNAiMAX Transfection Reagent (catalog no. 13778030) following the manufacturer's instructions, and collected after 48 hours culture.
Human CXCL10 (R&D Systems, catalog no. DIP100), human IFNγ (R&D Systems, catalog no. DIF50C), and human IL2 (R&D Systems, catalog no. D2050) ELISAs were performed according to the manufacturer's instructions. Conditioned media from each cell lines was collected after 24 or 72 hours culture.
Murine and tumor implantation studies
Female NSG (NOD-scid IL2Rgamma null) mice, 3 weeks old were purchased from The Jackson Laboratory. Animals were acclimated for 5 days before initiation of the study. The study was conducted at DFCI with the approval of the Institutional Animal Care and Use Committee in an Association for Assessment and Accreditation of Laboratory Animal Care International accredited vivarium.
Whole body radiation of 2 Gy was administered to the 4-week-old NSG mice using the X-Rad 225Cx Image Guided Biological Irradiator System (Precision X-Ray). Within 2 hours after radiation, mice were implanted intravenously with 35,000 human CD34+ cells from a single donor cord blood (HumanCells Bioscience). Mice were monitored for human immune cell engraftment. At 13 weeks after human CD34+ cell implantation, mice were randomized using Studylog software for hCD45+ cell engraftment and implanted subcutaneously with 5 × 106 HCC827GR6 or HCC827GR6-CD73 knockout (KO) cells with 30% Matrigel (Thermo Fisher Scientific) in the hind-flank (Supplementary Table S2). Tumor growth monitoring was initiated 1 week after tumor cell implantation with n = 8/group. The tumor volume was calculated using the following formula: (mm3) = length × width × width × 0.5. On day 31 after implantation, n = 5 mice/xenograft were euthanized, and tumors collected fresh for FACS analysis and a small piece fixed in formalin for IHC. In the remaining n = 3/xenograft on days 35–39, mice were treated with pemetrexed (purchased from DFCI pharmacy), 100 mg/kg i.p. Once the treatment was completed, tumors were monitored at least twice a week. Data collection and analysis were not performed blind to the conditions of the experiments.
Fresh tumors were mechanically and enzymatically disaggregated in dissociation buffer consisting of RPMI (Life Technologies) +10% FBS (HyClone), 100 U/mL collagenase type IV (Life Technologies), and 50 μg/mL DNase I (Roche). Suspension was incubated at 37°C for 45 minutes and then further mechanically dissociated. Red blood cells were removed from samples using red blood cell lysis buffer (BioLegend). Samples were pelleted and then resuspended in fresh RPMI +10% FBS and strained through a 40 μm filter. Cells were incubated with the Live/Dead Zombie NIR (BioLegend) for 5 minutes the dark at room temperature. Fc receptors were blocked prior to surface antibody staining using 1:1 mix of Human and Mouse TruStain FcX Blocking Reagent (BioLegend). Cells were stained for 15 minutes on ice in the dark and washed 2× with PBS + 2% FBS. Cells were analyzed on a BD LSRFortessa with FACSDiva software (BD Biosciences). Data were analyzed using FlowJo software version 10.5.3. Antibodies were specific for the following markers: CD16(3G8) (BioLegend, catalog no. 302006), CD8(RPA-T8) (Thermo Fisher Scientific, catalog no. BDB560662), CD56 (GDC56) (BioLegend, catalog no. 318348), PD-L1 (29E.2A3) (BioLegend, catalog no. 329708), CD3 (UCHT1) (BioLegend, catalog no. 300424), HLA-DR (G46-6; Thermo Fisher Scientific, catalog no. BDB562804), CD45RO (UCHL1) (BioLegend, catalog no. 304237), CD15 (SSEA-1) (BioLegend, catalog no. 323028), CD19 (HIB19) (BioLegend, catalog no. 302243), CD45 (H130) (BioLegend, catalog no. 304050), CD4(RPA-T4) (BioLegend, catalog no. 300554), CD14(M5E2) (BioLegend, catalog no. 301840).
Statistical significance was assessed using unpaired two-tailed Student t test, one-way ANOVA followed by Tukey post hoc test, or two-way ANOVA followed by Tukey post hoc test. P values less than 0.05 were considered significant. Asterisks used to indicate significance correspond with: *, P < 0.05; **, P < 0.005; ***, P < 0.001. Columns represent means ± SD. In one-way or two-way ANOVA followed by post hoc tests, we showed asterisks only in pairs of our interest. GraphPad Prism7 was used for all statistical analysis.
RNA sequencing for cancer cell lines analysis
We first compared between protein and mRNA levels of CD73/NT5E across cancer cells using quantitative proteomic and RNA sequencing (RNA-seq) data from Cancer Cell Line Encyclopedia (CCLE; refs. 37, 38) from DepMap portal (https://depmap.org/portal). For RNA-seq data, TPM (transcripts per million) values as defined by RSEM and log2 transformed using a pseudocount of 1 to avoid zeros. We also compared the CD73 protein levels to identify top associated CRISPR dependencies by using Chronos scores from DepMap 21Q4 release (39, 40). Reverse-phase protein array (RPPA) data were derived as described previously (34). We performed our analysis on 272 cancer cell lines from solid tumors, which overlapped among the three datasets. We calculated the degree of association between each feature using probabilistic models and Information-Theoretic metrics of association (IC, information coefficient). The statistical significance was estimated using an empirical permutation test (P values) and FDR after performing 1,000 permutations.
The datasets generated during and/or analyzed in this study are available from the corresponding author upon reasonable request.
All patients were consented to an Institutional Review Board (IRB)-approved protocol about specimen collection and correlation with clinical data (IRB#02-180). This study was performed in accordance with the Declaration of Helsinki.
MET-driven EGFR-TKI resistance is associated with induction of tumor cell STING
We used the parental HCC827 lung cancer cell line that harbors an EGFR-activating mutation (deletion in exon19) and two isogenic resistant cell lines: HCC827GR6, which harbors MET amplification, and HCC827EPR, which harbors EGFR T790M mutation, to compare innate immune signaling in these two different contexts (Fig. 1A). STING was strongly upregulated in HCC827GR6 cells relative to HCC827 and HCC827EPR cells (Fig. 1B; Supplementary Fig. S1A). As expected, HCC827GR6 cells were sensitive to the MET-TKI savolitinib in combination with osimertinib (Supplementary Fig. S1B and S1C). Savolitinib and osimertinib treatment also reduced levels of STING protein in HCC827GR6 cells, suggesting the possibility that MET signaling might be responsible for maintaining STING expression in conjunction with EGFR (Fig. 1C).
To assess whether the relationship between elevated MET and tumor cell STING expression might be more broadly generalizable across EGFR-mutated lung cancer cell lines, we next analyzed transcriptomic data from the CCLE (Fig. 1D). Indeed, high STING (also known as TMEM173) and MET mRNA expression were significantly correlated with each other in this dataset (R = 0.6495, P = 0.0223). To validate this finding further, we used IHC to analyze tumor cell–specific levels of STING and MET protein, across a panel of 38 patient-derived EGFR-mutated NSCLC xenograft (PDX) samples, enriched for EGFR-TKI–resistant samples. High-intensity STING staining was found to be robustly associated with high MET protein levels in tumor cells (P = 0.0193, Fisher exact test; Fig. 1E; Supplementary Table S3), reinforcing that STING is upregulated in concert with MET in EGFR-mutated lung cancers.
To assess the causal relationship between MET signaling and regulation of STING expression in EGFR-mutated lung cancers more directly, we next used CRISPR/CAS9 to delete MET specifically in HCC827GR6 or DFCI-202 cells, another model of MET-amplified EGFR-mutated lung cancer cells that upregulated STING (Supplementary Fig. S1D and S1E). MET deletion decreased STING levels (Fig.1F), which was especially apparent following generation of single-cell clones in HCC827GR6 cells. (Fig. 1G; Supplementary Fig. S1F). HCC827GR6 cells transfected with siRNA or different sgRNAs targeting MET also downregulated STING levels in accordance with the degree of MET suppression (Supplementary Fig. S1G and S1H). Conversely, exogenous overexpression of MET in HCC827 in multiple additional EGFR-mutated cell lines (H1975 and HCC4006) induced STING expression (Fig. 1H). Together, these findings demonstrate that MET amplification promotes expression of STING in EGFR-mutated lung cancer cells.
We next explored the downstream impact on cytoplasmic dsDNA sensing in MET-driven EGFR-TKI–resistant lung cancer cells. Transfection of HCC827GR6 cells with the dsDNA mimic poly (dA:dT) induced pTBK1, pIRF3, and pSTAT1, which was impaired following MET or STING deletion (Fig. 1I). In consonance with this observation, MET or STING deletion in HCC827GR6 cells also reduced poly (dA:dT) induced CXCL10 secretion, a potent T-cell chemokine activated downstream of IRF3 and STAT1 (Fig. 1J and K). Pathway activation was not substantially observed in the parental HCC827 cell line following poly (dA:dT) treatment (Supplementary Fig. S1I). We also observed similar results using the downstream cGAS product 2′3′ cGAMP, which preferentially activated STING signaling in HCC827GR6 cells as compared with HCC827cells (Supplementary Fig. S1J and S1K). Thus, activation of STING by oncogenic MET regulates downstream TBK1-IRF3-STAT1 signaling in EGFR-TKI–resistant lung cancer cells.
Impaired T-cell antigen-specific recognition of HCC827GR6 cells despite elevated STING
Tumor cell STING activation is an increasingly recognized determinant of effective antigen presentation because it can improve display of immunogenic peptides to CD8 T cells (35). To assess the impact of elevated STING on antigen-presenting function in MET-driven resistant EGFR-mutated lung cancer, we generated a model of antigen-specific recognition using HCC827 cells. Because HCC827 cells express the HLA A*11 allele, we established a model using a defined epitope (SSCSSCPLSK) from LMP2 of Epstein-Barr Virus, which binds HLA A*11 and for which a specific αβTCR has been defined (36). We introduced this T-cell receptor (TCR) into the J76 Jurkat T-cell line (37) and confirmed that these J76 TCR-LMP2B cells expressed the αβTCR relative to parental J76 cells (Fig. 2A). Next, we used the LMP2 Tetramer-SSCSSCPLSK to confirm that J76 TCR-LMP2B cells indeed express the SSCSSCPLSK/HLA A*11-specific TCR (Fig. 2B). JC76 TCR-LMP2B cells were utilized as effector cells (E), and HCC827 cells as target cells (T), which were loaded with the SSCSSCPLSK peptide. We confirmed that J76 TCR-LMP2B cells produced increasing levels of IFNγ in response to increasing E:T ratios, confirming antigen-specific recognition of HCC827 cells using this assay (Fig. 2C).
We therefore used this model to compare the immunogenicity of HCC827 and HCC827GR6 cells. In contrast to our expectation, TCR-LMP2B cells produced greater levels of IFNγ and IL2 in coculture with HCC827 cells as compared with HCC827GR6 cells (Fig. 2D). These data suggested that, despite elevated STING expression, coinduction of additional factors that suppress immunogenicity might be present in HCC827GR6 cells, which prompted us to further explore the potential mechanisms that could impair immunogenicity.
CD73 coactivation with STING in MET-driven EGFR-TKI–resistant cells
To understand additional factors that could influence tumor immunogenicity in the context of MET amplification, we first integrated RNA-seq data from NSCLC samples in The Cancer Genome Atlas (TCGA) and the CCLE database, and examined the top genes coexpressed with MET in both datasets (Fig. 3A). Twenty-eight genes were identified as candidates, including the immune suppressive factors NT5E (encoding CD73), which showed a strong correlation with MET expression in CCLE NSCLCs (R = 0.5169, P < 0.0001; Fig. 3B). We therefore examined the coexpression of these genes with MET specifically in EGFR-mutated lung adenocarcinoma in the CCLE, which identified a significant correlation between MET and NT5E expression (R = 0.6932, P = 0.0124), but not with CD274 (R = 0.3881, P = 0.2126; Fig. 3C; Supplementary Fig. S2A). PD-L1 cell surface expression was also lower in HCC827GR6 cells, as compared with parental HCC827 and HCC827EPR cells, arguing against PD-L1 as a significant determinant of immune suppression in this setting (Supplementary Fig. S2B). Furthermore, MET was among the top CRISPR genetic codependencies with elevated NT5E mRNA expression and CD73 protein levels in the CCLE database, suggesting an important functional relationship between MET and CD73 (Fig. 3D). Given these findings, we therefore aimed to understand the contribution of MET amplification in the control of CD73 levels.
Indeed, similar to STING, CD73 was strongly upregulated in HCC827GR6 cells (Fig. 3E), and was specifically downregulated by savolitinib treatment, alone or in combination with osimertinib (Fig. 3F). In contrast, we did not observe expression of the upstream ectonucleosidase CD39 in either HCC827 or HCC827GR6 cells (Fig. 3G). We confirmed by IHC across the same panel of 38 EGFR-mutated NSCLC PDX samples that CD73 was robustly associated with MET protein levels in tumor cells (P = 0.0029, Fisher exact test; Fig. 3H; Supplementary Table S4). We therefore also measured CD73 levels on the cell surface and by immunoblot in both HCC827GR6 and DFCI-202 cells following MET CRISPR deletion, and observed downregulation compared with control scramble guides (Figs. 3I and J; Supplementary Fig. S2C and S2D). Furthermore, the same CRISPR-derived in HCC827GR6 single-cell clones, effectively depleted MET, STING, and CD73 (Fig. 3K; Supplementary Fig. S2E), as did MET siRNA treatment (Supplementary Fig. S2F). Taken together, these findings confirmed MET-driven coexpression of CD73 with STING in resistant EGFR-mutated lung cancers.
CD73 is more generally regulated by oncogenic activation of MET signaling in lung cancer cells
We next explored to what extent MET hyperactivation is directly involved in the upregulation of CD73, regardless of EGFR mutation status. We first examined previously described H69/H69M cell lines, a paired model of neuroendocrine to mesenchymal cell state switch induced by chronic hepatocyte growth factor (HGF) exposure, which activates both MET and STING expression in the H69M derivative (38, 39) We found that the levels of CD73 were strongly increased by immunoblot and flow cytometry in H69M cells compared with the parental H69 cells (Fig. 4A and B). We next selected a panel of lung cancer cell lines with varying MET status, including amplification (METamp), exon 14 skipping mutation (METex14), and wild type (Fig. 4C). Both EBC1 cells (METamp) and H441 cells (METamp, KRASmut) exhibited increased pMET and CD73 levels, whereas H596 cells, which harbor an METex14 mutation, did not show elevated levels of CD73 at baseline but also had low levels of p-MET relative to these other cell lines. To elucidate whether the depletion of oncogenic MET reduced CD73 expression, we used CRISPR/CAS9 to generate MET KO in EBC1 cells. Compared with the control, the depletion of MET led to a strong reduction in the levels of total and cell surface CD73 expression (Fig. 4D), as previously observed in HCC827GR6. We also treated EBC1 cells with the MET inhibitor crizotinib, which downregulated p-MET and CD73 levels in parallel (Fig. 4E). The impact of MET inhibition on CD73 suppression was also similar to that of MEK/ERK inhibition by trametinib, which has been linked to regulation of CD73 levels in melanoma (Fig. 4E; ref. 40). Finally, because H596 cells failed to express high levels of CD73 but lacked robust p-MET, we treated with HGF to stimulate MET activity, which resulted in an increase in CD73, whereas the simultaneous addition of crizotinib decreased the levels of p-MET and of CD73 (Fig. 4F). Taken together, these data provide strong additional evidence that oncogenic activation of MET and downstream signaling induces CD73 across lung cancer histologies.
CD73 is regulated by FRA1 in MET-amplified EGFR-TKI–resistant cells and generates adenosine
The top genetic dependencies in cancer cell lines with high levels of CD73, uncovered not only MET, but also the FOS like antigen 1 (FOSL1) as the top hit (Fig. 3C). We also noticed that FOSL1 was among the top coexpressed genes with MET in our TCGA/CCLE expression analysis (Fig. 3A). FRA1 codes for a MEK-ERK–dependent transcription factor. Although it has not been previously implicated in regulation of CD73 in lung cancer, FRA1 is involved in MAPK signaling, suggesting its upregulation could provide a link between MET, MEK/ERK activation, and control of CD73 expression, which has been reported to drive CD73 expression in melanoma. Indeed, analysis of RPPA data in the CCLE revealed induction of FRA1 and MEK/ERK signaling as well as MET in CD73 high lung cancer cell lines (Supplementary Fig. S3A). We also confirmed that levels of protein and RNA for FRA1 were both significantly upregulated in HCC827GR6 compared with HCC827 cells (Fig. 5A). In addition, MET deletion suppressed FRA1 expression in HCC827GR6 as well as DFCI-202 cells, supporting a direct functional relationship between MET activation and induction of FRA1 (Fig. 5B; Supplementary Fig. S3B).
To assess the direct relationship between FRA1 and regulation of CD73 expression in EGFR-mutated lung cancers, we next overexpressed FRA1 in HCC827 cells, which increased the expression of CD73 (Fig. 5C). We also deleted FRA1 using CRISPR/CAS9 in HCC827GR6 or DFCI-202 cells, which decreased CD73 levels (Fig. 5D and E; Supplementary Fig. S3C), confirming that FRA1 plays a key role to induce CD73 downstream of MET. In further consonance with this notion, whereas treatment of HCC827GR6 cells with savolitinib alone modestly diminished CD73 levels, trametinib treatment substantially downregulated CD73 expression (Fig. 5F), similar to what we observed in EBC1 cells (Fig. 4E). Thus, FRA1 and pERK coactivation downstream of MET induces CD73 in EGFR-TKI–resistant cells.
We next generated a protocol to evaluate the production of adenosine for a functional analysis of CD73 activity in this context (Fig. 5G). Adenosine levels in conditioned media were significantly reduced in HCC827GR6 cells following CD73 deletion, as compared with control cells (Fig. 5H). Conversely, production of adenosine was also significantly increased in HCC827 cells engineered to overexpress CD73 (Fig. 5I). Next, we compared levels of adenosine production between HCC827 and HCC827GR6 cells using this assay, and confirmed increased levels in HCC827GR6 cells, in consonance with their higher baseline levels of CD73 expression (Fig. 5J). Furthermore, adenosine production was decreased following MET deletion in HCC827GR6 cells with decreased CD73 (Fig. 5K). Taken together, these data confirm that MET signaling and FRA1 promote functional activation of CD73 in EGFR-mutated lung cancers.
Pemetrexed treatment co-opts cGAS-STING signaling in MET-amplified EGFR-TKI–resistant cells
We considered the possibility that co-opting this state by treating cells with existing lung cancer chemotherapeutic agents to drive STING pathway activation might overcome the immune suppressive consequences of CD73 activation. We first screened CXCL10 as an activation marker for STING, using multiple chemotherapeutic agents, following pulse treatment of HCC827GR6 cells (Supplementary Fig. S3D). Pemetrexed (PEM) and docetaxel (DOC) were the two most effective agents at increasing CXCL10 (Fig. 6A). We also confirmed that PEM treatment substantially activated TBK1-IRF3-STAT1 signaling along with CXCL10 in HCC827GR6 cells, but not in HCC827 and HCC827EPR cells, and that STING deletion suppressed the majority of this effect (Fig. 6B and C; Supplementary Fig. S3E and S3F). We also compared levels of downstream STAT1 activation following PEM versus DOC treatment across the different HCC827 parental and EGFR-TKI–resistant models. Whereas DOC treatment induced STAT1 activation in all three cell lines, suggesting a nonspecific effect, PEM treatment induced STAT1 activation and CXCL10 secretion/mRNA expression only in HCC827GR6 cells (Fig. 6D–F). Furthermore, PEM treatment uniquely induced IFNB mRNA expression in HCC827GR6 cells (Fig. 6G). Finally, we observed that PEM was uniquely capable of inducing 2′-3′ cGAMP accumulation in HCC827GR6 cells (Fig. 6H), and that both PEM-induced STAT1 activation and 2′-3′ cGAMP accumulation was ablated by cGAS KO in HCC827GR6 cells (Fig. 6I and J). Taken together, PEM treatment co-opts endogenous cGAS-STING activation in MET-driven EGFR-TKI–resistant cell lines to promote TBK1-IRF3-STAT1 signaling, potentially enhancing the immunogenicity of this cell state.
PEM-induced immunogenicity is restrained by CD73
We therefore examined the implications of these findings on restoring immunogenicity using an allogeneic stem cell humanized xenograft mouse model as well as the TCR specific model we generated in vitro. Baseline HCC827GR6 CD73 KO xenograft tumor growth was significantly delayed in the humanized model as compared with control tumors (P = 0.0276; Fig. 7A). In contrast, we confirmed that CD73 deletion had no impact on cell proliferation rate in vitro (Supplementary Fig. S4A). We validated that circulating human CD8+ T cells were established in the humanized model and found that the frequency of CD8+ T cells infiltrating into tumors harvested at >1,000 mm3 in size was significantly higher in CD73-deleted HCC827GR6 xenografts (P = 0.0432; Fig. 7B), and redistributed from the tumor periphery into the intratumoral compartment by CD8 IHC (Fig. 7C and D).
Next, we treated the remaining set of large tumors (>1,000 mm3) with a single course of PEM, to assess the impact on tumor response in the setting of high tumor cell STING content (Fig. 7E). We observed dramatic reduction in tumor size in both CD73KO and control xenografted tumors (Fig. 7F). Because of complete tumor regressions, we were unable to quantify T-cell expansion within these tumors; however, reassessment from blood revealed a trend towards greater peripheral expansion of circulating CD8 T cells and dendritic cells in CD73 KO mice (Supplementary Fig. S4B–S4D). In consonance with this impressive impact of PEM treatment, we found that cell surface expression of HLA-A,B,C after treatment with PEM was specifically increased in HCC827GR6 cells, but not in HCC827 and HCC827EPR cells (Fig. 7G and H). We also confirmed that STING depletion specifically decreased baseline HLA-A,B,C expression in HCC827GR6 cells (Supplementary Fig. S5A) and that the increase in HLA-A,B,C associated with PEM treatment in HCC827GR6 cells was suppressed, albeit incompletely, by deleting STING (Supplementary Fig. S5B). These data confirm that PEM treatment is capable of restoring immunogenicity despite activation of CD73, though in the context of allogeneic T-cell response.
Finally, we examined the potential role of CD73 signaling in restraining this immunogenicity in the antigen specific model, given the artificial nature of humanized mouse models and the amplified impact of allogenic reponse. Indeed, we found that PEM treatment further induced CD73 expression in HCC827-GR6 cells and DFCI-202 (Fig 7I; Supplementary Fig. S5C). Consistent with this observation, PEM enhanced basal production of adenosine, which was completely ablated by CD73 KO (Fig. 7J; Supplementary Fig. S5D and S5E). Using the antigen-specific J76 TCR-LMP2B coupled with specific recognition of the HLA A*11-associated SSCSSCPLSK peptide on HCC827GR6 cells, we confirmed that increasing the concentration of adenosine in this model decreased IFNγ production (Supplementary Fig. S5F). In this context, PEM treatment coupled with deletion of CD73 cooperated to promote TCR activation as measured by IFNγ and IL2 production (Fig. 7K). Furthermore, when AB 680 was used as a CD73 inhibitor instead of CRISPR CD73 KO in HCC827GR6 cells, the same combination effect with PEM was confirmed (Supplementary Fig. S5G). These data reveal that PEM treatment and CD73 inhibition together co-operate to enhance activation of specific T cells recognizing model antigens (Fig. 7L).
Multiple strategies for targeting EGFR-TKI bypass pathways, including MET, have been examined, but to date limited durable activity has been observed in clinical practice (41–43). Here we report that MET amplification creates a broader cell state change in EGFR-TKI–resistant NSCLC cells, which is associated with induction of STING and CD73, and could be leveraged for a novel therapeutic strategy. Although immune suppressive at baseline, co-opting STING activation via PEM treatment, while preventing generation of adenosine via CD73 inhibition, promotes an immunogenic cell state that facilitates antitumoral response. Indeed, PEM therapy in a humanized allogeneic mouse model results in robust tumor regressions, and when combined with CD73 KO significantly enhances TCR recognition of a model T-cell antigen.
Adenosine production downstream of the ectonucleotidases CD39 and CD73, creates an immune-suppressive tumor immune microenvironment and contributes to resistance to cancer immunotherapy (22, 27, 44). CD73 has been considered as an emerging immune checkpoint and a potential therapeutic target in the EGFR-mutant setting (23). Combining CD73 inhibition with other immunotherapies is currently under investigation in this context and in NSCLC more generally (NCT03822351, NCT03334617). Although we did not observe significant upregulation of PD-L1 in our models and EGFR-mutated lung cancer fails to respond to anti-PD(L)1 blockade (15, 16, 45), our results suggest that priming innate immune response in combination with CD73 inhibition might be more effective at triggering immunogenicity.
Our data also support the idea that cotargeting STING and CD73 might be more effective in MET-activated lung cancer cells. We previously linked epigenetic regulation of STING expression with HGF-MET activation and mesenchymal cell state (39), consistent with our finding that STING is uniquely upregulated in MET-amplified EGFR-TKI–resistant cells. Furthermore, we discover here that activation of FOSL1 (FRA1) and MEK-ERK signaling in concert with MET simultaneously promotes CD73 expression. This finding is also reminiscent of similar work in melanoma that has implicated MAPK pathway activation and a mesenchymal cells state with CD73 upregulation in that context (40).
Importantly, we discover that PEM, the most frequently used cytotoxic drug in EGFR-mutated lung cancer (46) once they become resistant to targeted therapy, enhances immunogenicity of EGFR-mutated lung cancer by activating the STING pathway. This occurs most prominently in the context of MET amplification, increasing production of type I IFN, CXCL10, and HLA-A,B,C expression, but also generating adenosine. Recently, the phase II COAST clinical trial, evaluating the combination of oleclumab, a CD73 inhibitor, with durvalumab (anti-PDL1) has shown promising results after chemoradiation in locally advanced NSCLC (47). Because radiotherapy is another known activator of the cGAS-STING pathway (48), this early clinical observation may lend further credence to the idea that STING activation and CD73 inhibition is critical in promoting immunogenicity. It would be interesting to explore whether MET expression is linked to enhanced activity in this context. In addition, CD73 inhibition could potentially enhance immunogenicity of novel MET antibody drug conjugates, which have the potential to activate STING and are also being tested in EGFR-mutant MET-amplified NSCLC and in MET-activated NSCLC more broadly (5). Furthermore, MET amplification is also broadly found in cancers other than NSCLC, such as gastroesophageal carcinomas and the effect of combining CD73 inhibition with approaches that activate STING could be examined in these additional subgroups of carcinomas (49, 50).
Finally, multiple CD73 inhibitors in addition to oleclumab are in clinical trial development (e.g., NCT05431270, NCT04797468, NCT05143970, NCT04572152, and NCT03454451), and PEM is a standard chemotherapeutic option in lung adenocarcinoma. Clinical translation of this proof of principle therapeutic combination and determination of activity in MET-driven subsets of lung cancer should therefore be relatively straightforward.
S. Kitajima reports grants from Boehringer Ingelheim outside the submitted work. K. Haratani reports personal fees from AS ONE Corporation, Bristol-Myers Squibb Co. Ltd., Chugai Pharmaceutical Co. Ltd., Ono Pharmaceutical Co. Ltd.; grants and personal fees from AstraZeneca K.K. and MSD K.K. outside the submitted work. S.K. Sundararaman reports personal fees from LEK Consulting outside the submitted work. E.H. Knelson reports other support from Merck and grants from Takeda outside the submitted work. R. Uppaluri reports personal fees from Merck, Inc. outside the submitted work. P.C. Gokhale reports grants from Marengo Therapeutics, Epizyme Inc., Daiichi Sankyo, Foghorn Therapeutics, 28-7 Therapeutics, Moderna, and Kymera Therapeutics outside the submitted work. P.A. Jänne reports grants and personal fees from AstraZeneca during the conduct of the study; grants and personal fees from Boehringer Ingelheim, Eli Lilly, Daiichi Sankyo, Takeda Oncology; personal fees from Pfizer, Roche/Genentech, Chugai Pharmaceuticals, SFJ Pharmaceuticals, Voronoi, Biocartis, Novartis, Sanofi Oncology, Mirati Therapeutics, Transcenta, Silicon Therapeuticsons, Syndax, Nuvalent, Bayer, Esai, Allorion Therapeutics, Accutar Biootech, AbbVie; grants from Revolution Medicines, PUMA, and Astellas outside the submitted work; in addition, P.A. Jänne has a patent for EGFR mutations issued and licensed to LabCorp. D.A. Barbie reports grants from Loxo Oncology at Lilly during the conduct of the study; personal fees from Qiagen/N of One and Exo Therapeutics; grants from BMS, Gilead, Takeda, and Novartis; other support from Xsphera Biosciences outside the submitted work. No disclosures were reported by the other authors.
R. Yoshida: Conceptualization, data curation, software, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Saigi: Conceptualization, data curation, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing. T. Tani: Conceptualization, resources, data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. B.F. Springer: Data curation, investigation, visualization, writing–original draft, writing–review and editing. H. Shibata: Data curation, formal analysis, investigation, visualization, methodology. S. Kitajima: Data curation, investigation, methodology. N.R. Mahadevan: Formal analysis, investigation, methodology. M. Campisi: Investigation, methodology. W. Kim: Software, formal analysis, investigation, visualization, methodology. Y. Kobayashi: Investigation, methodology. T.C. Thai: Software, formal analysis, investigation, visualization, methodology. K. Haratani: Investigation, methodology. Y. Yamamoto: Investigation. S.K. Sundararaman: Investigation, methodology. E.H. Knelson: Resources, investigation. A. Vajdi: Formal analysis, investigation. I. Canadas: Resources, supervision, investigation. R. Uppaluri: Formal analysis, supervision, investigation. C.P. Paweletz: Supervision, investigation, methodology. J.J. Miret: Supervision, investigation. P.H. Lizotte: Investigation, methodology. P.C. Gokhale: Investigation, methodology. P.A. Jänne: Conceptualization, resources, supervision, investigation, methodology, writing–original draft, writing–review and editing. D.A. Barbie: Conceptualization, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing.
This work was supported by NIH R01CA190294 (D.A. Barbie), a DFCI-Lilly Oncology Research Grant (D.A. Barbie), and the Candace Bagby fund for Lung Cancer Research (D.A. Barbie). M. Saigi received a SEOM grant and M. Saigi is currently supported by a Joan Rodes contract from the Instituto de Salud Carlos III (JR20/00015). The authors also thank Andrea Protti and Quang-De Nyugen for help with radiation of humanized mice.
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).