Lung cancer ranks the second in incidence and top in mortality among malignancies worldwide (1), of which lung adenocarcinoma is the most common subtype (2). The prognosis of advanced and recurrent lung cancer is usually poor because most standard treatments by cytotoxic anticancer drugs only have limited therapeutic effects. In recent years, with a better understanding of immune response regulation, the immune checkpoint inhibitor (ICI) therapy showed improved survival rates in multiple cancers including non-small-cell lung cancer (NSCLC). The principle of ICIs is to reactivate immune cells by using specific antibodies against inhibitory signaling molecules such as CTLA-4 and PD-1 expressed on tumor and immune cells. Currently, the approved drugs of anti-CTLA-4 (ipilimumab), anti-PD-1 (nivolumab and pembrolizumab), anti-PD-L1 (atezolizumab, avelumab, and durvalumab), and their combinations have performed significant improvements in treating advanced NSCLC patients (3–6). Lung adenocarcinoma accounts for 80% of NSCLC and benefits most from ICIs therapy. However, it was also reported that there were still only partial LUAD patients responsive to ICIs (7)

PD-L1 expression has been widely used as an ICI response predictive marker, but the sensitivity and specificity are not very consistent due to different antibodies and cutoff values used for PD-L1 test (3, 8, 9). Meanwhile, PD-L1 expression cannot accurately reflect the complicated tumor immune microenvironment (10). Recent studies have also reported that tumor mutation burden (TMB) is closely related to the efficacy of ICIs response (11, 12) and can also be used as a predictive marker for the efficacy of ICI treatment. Like PD-L1, the cut-off value of TMB is controversial (13–15). Additionally, TMB alone does not directly produce neoantigen processing by major histocompatibility complex (MHC) class molecules, thus the accuracy of TMB as a predictor for ICI treatment is modest. Neoantigen expressed on tumor cells is one of the main targets for an effective antitumor T-cell response (16), but difficult to be identified. Therefore, identifying novel markers that can efficiently and accurately predict ICI responses is urgent. One promising area for this research is DNA damage repair (DDR). To ensure the integrity of the genome, cells activate DNA damage repair pathways to repair genetic lesions (SNP, Indel, etc.) during the process of DNA replication. DDR consists of eight pathways including miss match repair (MMR), base excision repair (BER), nucleotide excision repair (NER), direct damage repair (DR), homologous dependent recombination (HDR), nonhomologous end joining (NHEJ), fanconi anemia pathway (FA), and translesion DNA synthesis (TLS). Defects in DDR pathways lead to the accumulation of genomic aberrations and an elevated TMB (17–19), thus promoting tumor development (20). Many studies have shown that mutations in DDR pathway genes are associated with ICIs responses (20, 21). Patients who have DDR genomic alterations usually have a better clinical benefit after ICIs therapy (19, 21).

The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm is a computational method that uses gene expression profiles to predict the ICIs response, particularly successful in NSCLC and melanoma (22). TIDE uses a specific set of marker genes to estimate dysfunction of tumor-infiltrating cytotoxic T lymphocyte and exclusion of CTL by an immunosuppressive factor to predict patients’ response to ICIs. Patients with lower TIDE scores have a lower chance of antitumor immune escape, thus having a higher response rate of ICIs treatment (22). The TIDE score exhibited a higher accuracy than PD-L1 expression level and TMB in predicting the overall survival of patients treated with ICIs (20, 23, 24). Some studies also have reported its utility in predicting or evaluating the ICIs efficacy (24–28).

We identified seven significant genes (e.g., DUT, MGMT, POLH, RAD1, RAD17, TYMS, and YWHAG) strongly associated with prognosis from DDR pathways using Cox regression analysis.

Patients with lower expression of DUT, TYMS, and YWHAG but higher expression of MGMT, POLH, RAD1, and RAD17 had a better prognosis. Based on the expressions and weights calculated by Cox regression on these genes, we developed a classifier, GDPLichi (Genes of DDR to Predict LUAD immune checkpoint inhibitors), as the signature to predict the ICIs response. LUAD patients were classified into low- and high-risk groups based on the cutoff of the GDPLichi score. Many features, including PDCD1, CTLA4, PD-L1 expression, TMB, and neoantigen, displayed strong discerning abilities in the survival analysis of these two subgroups. Especially, the high-risk subgroup had a worse prognosis but is presumably more efficacious towards ICIs treatment.

Lung cancer ranks second in incidence and top in mortality among malignancies worldwide (1). Recently, immunotherapy has become an important new therapeutic approach in treating multiple types of cancer with promising results. It has greatly changed the landscape of cancer care. Many studies have shown that mutations in DDR pathway genes are associated with the prognosis of LUAD patients (20, 21), however, using the DDR gene expression profile as a molecular signature to predict the response to ICIs of LUAD patients has not been reported yet. In this study, we constructed a GDPLichi model based on seven DDR genes (DUT, MGMT, POLH, RAD1, RAD17, TYMS, and YWHAG), to classify LUAD into two distinct subgroups: low- and high-risk groups. Thymidylate synthase (TYMS) is a critical target for cancer chemotherapy (45). Tyrosine 3-Monooxygenase/Tryptophan 5-Monooxygenase Activation Protein Gamma (YWHAG) is also known as 14-3-3γ. A recent study reported that knockdown of YWHAG suppresses epithelial-mesenchymal transition (EMT) and reduces the metastatic potential of human NSCLC (46). O-6-Methylguanine-DNA Methyltransferase (MGMT) catalyzes the transfer of methyl groups from O(6)-alkylguanine and other methylated moieties of the DNA to its molecule. A low protein expression of MGMT was found in the bronchial epithelium of patients with lung cancer as compared to healthy controls, suggesting that there is a negative correlation between MGMT expression and lung cancer risk (47). DNA polymerase eta (POLH) is a DNA polymerase belonging to a subset of tumor suppressor proteins required for maintaining genome integrity (48). RAD1 encodes a component of a heterotrimeric cell cycle checkpoint complex, known as the 9-1-1 complex, that is activated to stop cell cycle progression in response to DNA damage or incomplete DNA replication. RAD17 is a cell cycle checkpoint gene required for cell cycle arrest and DNA damage repair in response to DNA damage. This protein recruits the RAD1-RAD9-HUS1 checkpoint protein complex onto chromatin after DNA damage and initiates DNA repair.

Gene Set Enrichment Analysis (GSEA) is a computational method that determines whether a priori defined set of genes shows statistically significant, concordant differences between two biological states. GSEA showed that cellular response to hypoxia, MAPK signaling, noncanonical NF-κB signaling, and cell cycle pathways relating to cell proliferation was significantly enriched in the high-risk group, which might account for a higher malignancy and poorer prognosis of LUAD patients in the high-risk group. Oxygen deprivation (hypoxia) is a feature of solid tumors that promotes genomic instability, enhanced aggressiveness, and metastases and is an important factor in treatment resistance and poor survival (49). The MAPK pathways converge in the amplification of key molecules that sustain cell proliferation, growth, and survival processes (50, 51). Noncanonical NF-κB signaling contains NIK phosphorylates IKK/and helps IKK/to phosphorylate p100. Mutations in various upstream regulators (TRAF2, TRAF3, cIAP1&2, CD40) lead to increased stability of NIK and subsequent activation of the noncanonical NF-kB pathway, and this mechanism of activation appears to be important for different cancer types including DLBC and lung cancer (52). The human cell cycle is a tightly regulated process with checkpoints in place to ensure genomic integrity. Recent studies have shown that CDK4 and CDK6 inhibitors can promote T cell activation (53) and reverse T cell exclusion, thus leading to a better response to ICIs (54). Taken together, this suggests that tumor cells in the high-risk group proliferated faster, leading to increased malignancy.

In addition, there were decreased B cells, CD8⁺ T cells, CD8⁺ central memory T cells, HSC, MDC, and immune scores found in the high-risk group. The CD8 T cell-dependent killing of cancer cells could produce interferon-gamma (IFN-γ) and then activate antitumor immunity (55). Myeloid dendritic cells (MDC) are crucial for the activation of antigen-specific CD8 T Cells. A recent study reported that anti-tumor effects of DCs can be reduced by a low DC count, low antigen presentation efficiency of tumor-infiltrating DCs, and a weak ability of DC to migrate into tumor (56). Many studies reported that B cell infiltration was associated with a favorable prognosis in NSCLC (57–60). Hematopoietic stem cells (HSC) are a very small group of source cells that can self-renew and generate various blood cells and immune cells. Tumor immune infiltrating cells migrate from blood to tumor tissues and play an important role in immune regulation. Lots of studies have shown that tumor immune infiltrating cells are closely related to the efficacy of ICIs and prognosis (61, 62).

Interestingly, we noticed that there were increased TMB, neoantigen, immune checkpoint molecules, and immuno-suppression genes’ expression in the high-risk group. Meanwhile, the expression of MHC II genes that express on antigen-presenting cells only in the low-risk group was significantly lower than in the high-risk group. It has been widely studied that higher TMB, neoantigen, and immune checkpoint molecules are indicators implicated in a better response to ICI treatment (8, 11, 12). Therefore, it is suggested that the high-risk group might be more sensitive to immunotherapies as compared to the low-risk group classified by the GDPLichi model.

We further examined genomic mutations in both the low- and high-risk groups and identified 12 candidates from the first top 25 mutated genes, whose mutation frequency has a significant difference between low- and high-risk groups classified by the GDPLichi model. Most of these genes are associated with TMB (35, 36), which could be used to predict the efficacy of immunotherapy.

The TIDE algorithm is a computational method that uses the expression profile of immune-related genes to predict the ICIs response. It is particularly successful in NSCLC and melanoma (22) and has exhibited a higher accuracy than PD-L1 expression level or TMB alone in predicting overall survival of patients treated with ICIs (13, 19, 20). Further analysis revealed that the TIDE scores in the high-risk group were significantly lower than the low-risk group, suggesting that patient of the high-risk group is more sensitive to response for ICIs. This conclusion was verified in the other external datasets (GSE31210, GSE50081, GSE30219).

In conclusion, we firstly identified two prognostically and clinically relevant subgroups of LUAD using the GDPLichi model which was constructed from seven DDR-risk genes. Patients from the high-risk group showed lower TIDE scores, and are thus more responsive to ICIs. The limit of this research was that it was retrospective, and results should thus be further confirmed by prospective studies.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material .

Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

BH, YHL, and YLi designed the study. YLe and SD collected the data and performed the analysis. FY, TG, XX, YZ, LC, CFQ, NL, XZ, CL, YK, KH, and NW helped to analyze the data. YLe and SD wrote the paper. BH, TG, and XZ revised the paper. All the authors read and approved the final manuscript.

This research was supported by the National Natural Science Foundation of China (grant number 81971470 and 32170913) and Shenzhen Science and Technology Commission Fund (program number JCYJ20190809143803732 and JCYJ20210324120602008) to BH and Chinese Postdoctoral Science Foundation (grant number 2020M683062) to NL.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.