Loss of STING expression is prognostic in non–small cell lung cancer

Abstract Background Stimulator of interferon (IFN) genes (STING) is a protein that promotes type I IFN production essential for T‐cell activation. In this study, we aim to characterize STING expression comprehensively using The Cancer Genome Atlas (TCGA) database, cell lines, and patient tumor samples stained with immunohistochemistry. Methods Two cohorts were evaluated comprising 721 non–small cell lung cancer (NSCLC) patients and 55 NSCLC cell lines for STING and cyclic GMP‐AMP synthase (cGAS) expression using immunohistochemistry. Moreover, an independent cohort of n = 499 patients from the TCGA database was analyzed. Methylation was evaluated on STING and cGAS in five STING‐negative NSCLC cell lines. Results STING RNA expression positively correlates with T cell function and development genes, negatively correlates with cell proliferation and associated with increased survival (5‐year‐overall survival [OS] 47.3% vs. 38.8%, p = 0.033). STING protein expression is significantly higher in adenocarcinoma (AC) and is lost with increasing stages of AC. STING‐positivity is significantly higher in mutant EGFR and KRAS tumors. STING‐positive NSCLC patients identified with immunohistochemistry (H‐score > 50) have increased survival (median OS: 58 vs. 35 months, p = 0.02). Treatment of STING‐negative cell lines with a demethylating agent restores STING expression. Conclusions STING is ubiquitously expressed in NSCLC and associated with T cell function genes, AC histology, EGFR, and KRAS mutations and improved overall survival.

Treatment of STING-negative cell lines with a demethylating agent restores STING expression.
Conclusions: STING is ubiquitously expressed in NSCLC and associated with T cell function genes, AC histology, EGFR, and KRAS mutations and improved overall survival. advanced-stage disease that were previously not anticipated now have been accomplished, and 5-year overall survival (OS) increased to 20% in unselected and up to 40% in PD-L1 high expressing patients. [1][2][3][4] Nevertheless, patients now treated with various anti-PD immunotherapies in the frontline, advanced NSCLC still have poor outcomes, and even with curative resection, 30%-55% of patients with limited-stage disease relapse. 5 Thus, new innovative approaches to select patients for therapy that can cost-effectively be integrated into routine practice that complement current treatments are required. The tumor microenvironment (TME), the interaction of immunotherapy with the balance of the altering immune-and therapy response, is a novel direction to enhance treatment efficacy.

STING (Stimulator of Interferon [IFN]
Genes, tmem173, MITA, and MYPS) is a protein responsible for controlling anticancer immune responses to leaked self-or non-self DNA. 6 STING is a transmembrane component of the endoplasmic reticulum that produces type I IFNs (IFNα/β) essential for activating dendritic cells and thus antigen presentation and T-cell priming. [7][8][9] Agonists of STING that spike IFN production and show potent immune response are currently under investigation in clinical trials and are of particular interest in combination with checkpoint therapies targeting pathways such as PD-L1 or cytotoxic T-lymphocyte-associated protein 4 (CTLA4). 10,11 Importantly, recent studies have shown in animal models that knocking out STING and cGAS expression results in a nonresponse to PD-L1 checkpoint therapy, whereas control mice responded well to PD-L1 checkpoint inhibition. 12 STING and cGAS are thus thought to be essential for the antitumor response of PD-1/PD-L1 checkpoint inhibition.
Recent studies in hepatocellular, 13 gastric 14 and colorectal cancer, 15,16 and in melanoma 17 have shown that STING expression was decreased in tumor, compared with healthy tissues. Additionally, STING is frequently lost during tumor progression, and loss of STING/cGAS correlates with poor survival. One common reported mechanism of STING or cGAS loss in tumors is due to upregulated methylation of their respective promoter regions. 10 The cytoplasmic DNA sensor cGAS (cyclic GMP-AMP synthase, and MB21D1) 18 can detect leaked self or non-self DNA and, in response, will synthesize the cyclic dinucleotide (CDN) cyclic GMP-AMP. 19 cGAMP binds STING specifically, activating STING and causing dramatic conformational changes and translocation of STING from the ER to the perinuclear area, where STING acts as an adapter protein essential for immune signaling following the detection of tumor DNA.
Our study aims to analyze the presence and the expression landscape of STING and cGAS protein according to key clinicopathological parameters, including stage, sex, histological type, mutational status, and survival. Furthermore, to validate our results on the human tissues, we analyzed the methylation of STING and cGAS in NSCLC cell lines using demethylating agents.

| Patient cohorts and tissue microarrays
Tumor tissue samples from a total of 721 patients were included in our study (N = 419 AC and N = 302 SCC, Table S1) from four sources ( Figure 1B)

| Immunohistochemistry (IHC) and scoring
Cell lines, tissues, and TMA sections were stained for immunohistochemistry (IHC) using the Ventana Benchmark XT autostainer, like previously described. 20 STING antibody (Cell Signaling #13647) was diluted at 1:400 for IHC using Signal Stain antibody diluent (Cell Signaling #8112). cGAS antibody (Novus Biologics #NBP1-86761) was diluted at 1:300 with Signal Stain diluent. Scoring of STING and cGAS IHC was based on a percentage of tumor cell marker expression (0%-100%) multiplied by staining intensity (0, 1+, 2+ and 3+) to provide an H-score (range 0-300). H-scores for multiple cores were averaged. In line with previously reported scoring methods, a cutoff below an H-score of 50 was considered negative. cGAS expression at ≥1 was identified as positive based on tumor cell staining. Scoring was carried out by two independent observers with software-assisted (ImageJ) manual cell counting. and specimens were evaluated using an Olympus BX43 brightfield microscope and Olympus DP71 camera and cellSens software.

| Statistical analyses
Associations between clinicopathologic characteristics and STING or cGAS expression were analyzed using the χ 2 test. Receiver operating characteristic (ROC) curves were used to define optimal survival cutoffs for STING mRNA expressions, and survival analysis was   Figure 2F). The latter suggests that STING RNA expression is a favorable prognosticator in AC but not in SCC.

| IHC expression of STING and cGAS in NSCLC cell lines and tissues
We investigated the expression of STING and cGAS using IHC Stage I, 33% Stage II, 31% Stage III, 20% Stage IV, and 35% total; n = 302) ( Figure 4C). Expression of cGAS was higher in AC (94%) than SCC (75%) and showed no correlation with stage ( Figure S3).

| Clinical relevance of STING expression in NSCLC tissues
The "annotated cohort" of 417 cases was analyzed to correlate  Figure 5C) showed a nonsignificant increase in OS. Figure 5D shows survival relative to clinicopathological characteristics according to univariate and multivariate analysis. STING-high expression was prognostic in the univariate analysis and we found conventionally strong prognosticators such as stage, age, and PD-L1 significant in the multivariate analysis too.

| Methylation of STING and cGAS promoters in NSCLC
Since methylation of STING and cGAS promoters has been implicated in the loss of STING or cGAS protein expression, we analyzed TCGA NSCLC methylation data and sought to restore STING or cGAS expression in cell lines using the demethylating agent 5ʹAZADC  The STING pathway might be defective in lung tumors that can alter responses, as reported in other cancers. Others showed that STING/cGAS expression was lost in tumor tissues; however many of these studies involve only a small number of cases. [23][24][25] To our knowledge, STING expression has not been extensively studied in larger patient cohorts with clinicopathological data correlation. TCGA data analysis of STING and cGAS in AC and SCC shows that low expression of STING in adenocarcinoma, but not squamous cell carcinoma, correlates with poor survival. Further TCGA analysis shows F I G U R E 4 STING Expression by IHC in NSCLC tumor samples. Immunohistochemistry for STING expression was performed on a collection of 55 NSCLC cell lines (A) and 721 individual NSCLC cases with stage and histology data provided. (B) An increased frequency of STING loss in the higher tumor stage occurred in AC and SCC subsets (C). SCC (vs. AC) histology was associated with a significantly lower expression level of STING irrespective of tumor stage. STING expression was lost in an average of 34% of AC tumors compared to 65% loss in SCC histology (D). In the annotated cohort containing TMAs with triplicate cases from 414 NSCLC patients, the STING-positive case number was significantly higher in females compared to males (66.9% vs. 51.2%, p = 0.003) (E) STING positivity was significantly higher in EGFR mutant versus wild type (76.2% vs. 50.0%, p = 0.04) and in KRAS mutant versus wild type (74.4% vs. 49.4%, p = 0.001) patients. AC, adenocarcinoma; NSCLC, non-small cell lung cancer; SCC, squamous cell carcinoma; STING, stimulator of interferon genes; TMAs, tissue microarrays STING expression correlates positively with expression genes identified as "T-cell signature genes" in lung cancer, while STING expression negatively correlates with common tumor proliferation genes. 26 Analysis of TCGA Illumina Methylation 450k database shows increased methylation of STING and cGAS genes in AC and SCC.
Activating STING to transform immunologically refractive cold tumors to a hot phenotype is a potential new therapeutic approach. 27,28 Inhibitors to the enzyme poly-ADP ribose polymerase (PARP) are new agents proved experimentally to reduce DNA damage repair, potentially increasing STING activation. 29 Another possibility is the inhibition of DNA damage response (DDR) proteins that were shown to activate the STING/TBK1/IRF3 molecular pathway, followed by a significant increase in chemokine levels  The underlying mechanism for STING and cGAS loss has been attributed to hypermethylation of the promoter regions for both genes. Given that epigenetic modifications are common in NSCLC, it is reasonable to expect methylation to play a role in STING and cGAS expression. TCGA analysis of methylation for both STING and cGAS genes showed much higher methylation in SCC than in AC. The differences in methylation between AC and SCC might explain why STING expression is more remarkable in AC than in SCC. Recent studies combining PD-L1 inhibitors with low-dose demethylating agents such as azacytadine have been shown to improve outcomes. 33 Combinations of STING agonists with lowdose Azacytadine could increase STING activity and sensitize low STING expressers, especially SCC tumors, to STING-targeted therapies.
Moreover, a recent study showed that cisplatin treatment increases the activation of the STING/cGAS pathway and is associated with higher PD-L1 expression in multiple NSCLC preclinical models in both AC and SCC. 25 Small molecule substances, such as cyclic dinucleotides (CDNs) derived from bacteria might directly activate the STING signaling pathway. CDNs increased the infiltration of tumorspecific cytotoxic T-cells and enhanced the therapeutic efficacy of anti-PD-1 and anti-CTLA-4, reprogramming immunosuppressive, M2polarized tumor-associated macrophages to a pro-inflammatory M1macrophages. 34 It was also reported that STING-activating nanovaccines delivering tumor neo-antigens could effectuate intense and persistent antigen-specific T-cell responses, which were followed by vigorous immunotherapeutic efficacy in numerous murine cancer models. 35,36 Of note, our study has possible clinical and therapeutic implications. Others showed in a preclinical ovarian cancer study that survival of mice treated with a combination of STING, carboplatin, agonist, and anti-PD-1 antibody was the longest. 37 In the clinical setting, a number of effective small-molecule STING agonists emerged, already under phase I-III clinical trials, including ADU-S100 (NCT03937141), BMS-986301 (NCT03956680), and DMXAA specifically trialed in NSCLC (NCT00674102) combined with carboplatin and paclitaxel. 38 CDN-based STING agonist, MK-1454 combined F I G U R E 6 Restoration of STING and cGAS expression by demethylation (A) Selected cell lines were either incubated in standard media or media containing 10 µM of demethylating agent 5ʹAZADC. 40 µg cell lysate was analyzed for cGAS and STING expression via western blot. Restoration of STING and cGAS was seen in the majority of cell lines treated. (B) and (C) TCGA Illumina 450k Methylation database was examined for methylation of STING and cGAS promoters in both adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC) data sets. Methylation of promoters was much higher in tumors than normal, with SCC showing higher average methylation than AC. In AC, STING methylation was much more common than cGAS. AC, adenocarcinoma; cGAS, cyclic GMP-AMP synthase; NSCLC, non-small cell lung cancer; STING, stimulator of interferon genes; TCGA, The Cancer Genome Atlas with pembrolizumab is progressing from Phase I to Phase II clinical trial (NCT04220866) in advanced solid tumor indication. 39 However, systemic administration of STING agonists may induce pathological inflammation. This concern is based on the fact that overactivation of STING might be involved in a broad range of autoimmune conditions. 40 Still, a safe, effective, and efficient way of STING-agonism in advanced-stage solid tumors, including NSCLC could strongly potentiate PD-L1 immunotherapies and transform these aggressive malignancies into chronic conditions.
Limitations of this study include that we have no comprehensive data collected on specific treatments in subgroups. Moreover, our "annotated cohort" is from overwhelmingly early-stage NSCLC patients, and we have no data on the potential clinical behavior without STING targeted specific treatments administered in advanced-stage patients.

| CONCLUSIONS
Our sizable clinical patient cohort shows that STING is extensively expressed in NSCLC. High STING tumor expression correlates with improved survival, early-stage disease, and EGFR and KRAS mutations. Our data are further supported by the coincidence of STING and T-cell activation genes, cell line demethylation, and TCGA data.
Furthermore, our study provides multiple aspects of STING expression's translational relevance. Along with other studies on immunotherapy and STING associations, our data serve as a reasonable basis for further exploring the exact clinical role of STING in NSCLC.