The DDR-related gene signature with cell cycle checkpoint function predicts prognosis, immune activity, and chemoradiotherapy response in lung adenocarcinoma

Cancer is the primary threat to human public health, with an estimated 23.6 million new cancer cases and 10 million cancer deaths by 2019 worldwide [1]. Lung adenocarcinoma (LUAD), the primary pathological type of lung cancer, remains the predominant cause of cancer deaths in most countries [2, 3]. Despite significant breakthroughs achieved in the treatment of LUAD, radiation and drug resistance after treatments are still the main challenges to the survival of patients [4]. Emerging studies suggest that autophagy-related signature [5], immune-related signature [6], and methylation-related markers [7] can be used as prognostic markers to predict the prognosis of LUAD patients, while the heterogeneity of the tumor makes it difficult to evaluate the prognosis of each patient precisely [8]. Therefore, discovering and identifying a distinctive prognostic signature for LUAD to accurately assess patient outcomes and facilitate individual tumor treatment remains critical.

DNA damage can be recognized and repaired by the cell's internal DNA damage repair (DDR) mechanism. Incorrect repair is one of the leading causes of cancer’s occurrence and development [9]. Studies have shown that DDR acts as a barrier to tumorigenesis in the early stages of LUAD. Yet, it can promote malignant growth in tumor cells with defective genomic maintenance mechanisms [10]. In addition, DDR is associated with radiation resistance in LUAD cells. The serine proteinase inhibitor clade E member 2 (SERPINE2), a DDR-related gene, regulates radiation sensitivity. Its high expression positively correlates with poor prognosis in patients with LUAD [11]. ERK5 increases the radiation resistance of LUAD cells by enhancing the DNA damage response, leading to a poor prognosis for patients [12]. Ubiquitin-specific protease 14 (USP14) is a modulator of the double-strand break (DSB) repair pathway that increases radiation resistance in LUAD cells, leading to poor treatment in patients [13]. It is no doubt that DDR-related genes are closely related to LUAD prognosis.

Cell cycle checkpoints, acting as DNA surveillance mechanisms, can prevent the accumulation and propagation of genetic errors during cell division [14]. Activating different cell cycle checkpoints is also considered an essential process in DDR, allowing cells time to repair their damaged DNA before moving to the next cell cycle stage [15]. When DNA double-strand breaks (DSBs), the primary type of DNA damage, the MRE11/NBS1/RAD5 complex activates the ataxia telangiectasia mutated (ATM) /checkpoint kinase 2 (CHK2) pathway to promote the S-phase cell cycle arrest and the p53-associated G1/S-phase checkpoint [16, 17]. Furthermore, the cell cycle checkpoint is closely related to LUAD prognosis. Some studies suggested that genetic variations in the CHEK2 gene may play a key role in predicting the toxicity and prognosis of NSCLC [18]. Higher checkpoint gene PRKCSH expression, which suppresses the activation of the STAT6/p53 pathway, was correlated with a poorer prognosis and more significant infiltration of most immune cell types in patients with lung cancer [19].

As mentioned above, both DDR and cell cycle checkpoint are closely related to the prognosis of LUAD, while the cell cycle checkpoint also has a vital role in DDR. In addition, radiotherapy and chemotherapy are two of the three primary means of cancer treatments, which rely on causing DNA breaks to kill tumor cells. Therefore, the expression of genes related to DDR and cell cycle regulation also directly affect chemoradiotherapy results. A reasonable hypothesis could be generated that genes with both cell cycle checkpoint and DDR functions may have more critical effects on LUAD prognosis.

Consequently, the DDR-related genes with cell cycle checkpoint function, here called DCGs, were selected to construct a prognosis model for LUAD by a systematical method and its effect on immune activity and response to chemoradiotherapy were further explored to reveal the causes behind the poor prognosis.

In recent years, the association between some molecular markers and the prognosis of LUAD has been found in a great number of studies. For example, Koh et al. found that PD-1 overexpression in patients with LUAD resulted in poor overall survival and progression-free survival [29]. Wang et al. indicated that the decreased expression of miR-133a in patients with LUAD was related to poor prognosis [30]. Takamizawa et al. found that the overexpression of let-7 microRNA in postoperative patients with LUAD was associated with a relatively short survival time [31]. However, due to the high heterogeneity of cancer, single-function genes explaining the patient prognosis may be farfetched. Furthermore, some problems hamper the prediction accuracy of these prognostic signatures, such as insufficient clinical samples and a lack of external independent validation.

The cell cycle checkpoints play an essential role in DDR. Moreover, it and DDR are respectively related to LUAD prognosis. Therefore, we sought to select genes with both cell cycle checkpoints and DDR functions in this study, developed a prognosis model in the TCGA-LUAD cohort, and validated it in two independent GEO cohorts. To make the results more scientific and reliable, the TCGA cohort was divided into a training set and a test set according to the 2:1 ratio. The training set was used to construct the model, the test set was used to test the model’s efficacy, and the independent GEO cohorts were finally used to verify the model. The LASSO regression, screening for the key variables, has been widely used to construct prognosis models for different tumors, and the patient classification based on risk score, which is calculated by the LASSO coefficient, has feasible clinical guiding significance [32‐34]. Thus, we combined univariate COX regression and LASSO regression to establish a 4-DCGs prognostic model with high predictive accuracy for LUAD survival, especially in 3-year survival (AUC = 0.711, 0.710). Greater AUC indicates a higher diagnostic value of the test, and AUC greater than 0.7 means high test accuracy. From this, the model we constructed has some clinical application value. In addition, to improve the clinical applicability, we also identified independent prognostic factors combined with the patient's clinical features to establish a prognostic nomogram with good predictive power. For example, a 70-year-old LUAD patient with stage II, T3, and risk score (4-DCGs signature) equal to 6.8 would score a total of 122 points (12 points for age, 20 points for stage, 20 points for T classification, and 70 points for risk score). For this patient, the predicted survival for 1 year, 3 years, and 5 years was 63.0%, 19%, and < 10.0%, respectively.

The 4-DCGs signature consists of PLK1, PLK2, PRKDC, and NBN. PLK1 (Polo-like kinase 1), a member of the polo family of serine/threonine protein kinases, is an essential regulator of cell cycle progression that induces activation of DNA damage checkpoint [35, 36]. The overexpression of PLK1 appeared in various cancers with poor prognosis and survival. In addition, studies have shown that inhibition of PLK1 promotes tumor cell apoptosis in lung cancer [37, 38]. PLK2, also called SNK, regulates the replication of centrosomes during cell division and can be induced by P53 to activate the G2 checkpoints in the DNA damage response [39]. It was reported that PLK2 promotes tumor growth by targeting the FBXW7/ Cyclin E pathway [40, 41]. The PRKDC gene encodes DNA-PK protein kinase, a protein kinase required for cell cycle progression during mitosis and the NHEJ pathway. Moreover, PRKDC can be a drug target for immune checkpoint inhibitors, while the inhibition of DNA-PK can also enhance the chemosensitivity and radiosensitivity of NSCLC. NBN, encoding the Nibrin protein, is a component of the Mre11-Rad50-Nbs (MRN) complex, which can trigger cell cycle checkpoint activation through interaction with ATM proteins, and plays a vital role in the DDR [42]. Increased expression of NBN genes in breast and ovarian cancer cells can induce chemoresistance and poor prognosis, and mutations in NBN can also cause Nijmegen breakage syndrome (NBS), leading to low immune function and abnormal lymphocyte function in patients [43, 44]. To sum up, consistent with our results, high expression of 4-DCGs was closely associated with cancer progression and poor prognosis, and also suggested that these four genes may influence the immune environment and chemoradiation resistance of LUAD patients.

Increasing evidence suggests that tumor development and progression largely depend on the complex microenvironment in which they reside, including the tumor cells and their surrounding immune cells [45]. Therefore, to further explore the potential prognostic mechanism of risk grouping based on the 4-DCGs signature model, we compared the compositions of the 22 tumor-infiltrating immune cells in the two risk groups. The results showed that infiltration levels of naive B cells, activated NK cells, monocytes, and activated dendritic cells were significantly lower in the high-risk group. In contrast, M0 macrophages and resting NK cells were higher in the high-risk group. Naive B cells are a type of lymphocyte, and many studies have reported that its infiltration level correlated with a favorable prognosis in NSCLC [46, 47]. NK cells involved in tumor immunity can be divided into resting and activated subtypes [48]. Generally, the higher the proportion of resting NK cells or the lower the proportion of activated NK cells is, the higher the level of tumor infiltration will be, which favors the formation of the tumor microenvironment [49]. Recruitment of monocytes in the early stages of tumor progression can be found in multiple types of cancer, where monocytes directly kill malignant cells by cytokine-mediated cell death and phagocytosis [50]. An emerging study indicated that the anti-tumor effect of dendritic cells (DCs) could be reduced by the low DCs count inducing the low antigen presentation efficiency of tumor-invasive DCs [51]. Tumor-associated macrophages (TAMs) function as a promoter during tumor progression. TAM consists of several macrophages’ phenotypes, including M0 (inactivated macrophages), M1 (classical activated), and M2 (alternately activated). M2 cells are polarized from M0 macrophages and promote immunosuppression and angiogenesis by producing immunosuppressive factors, interleukin-10 [52, 53]. It follows then that our high-risk group with low anti-tumor immune cells and high pro-tumor immune cells showed a low tumor-suppressive immune microenvironment. In other words, our 4-DCGs signature could somewhat predict the immune activity of LUAD patients.

Furthermore, in our study, the immune checkpoint expression levels were generally elevated in the high-risk group, and immune checkpoints (PDCD1, TIGIT, and CD276) expression were positively correlated with the risk scores. Immune activation can be increased by blocking immune checkpoints. Thus immune checkpoint gene expression is considered an indicator of immunotherapy response in clinical practice. Blocking the immune checkpoint has become a novel approach to eliminating the immunosuppressive microenvironment to enable tumor immunotherapy [54, 55]. Lower immunoactivities’ cell infiltration and higher immune checkpoint expression may explain why a poor prognosis appeared in high-risk LUAD patients. Targeted immune checkpoints, in turn, may be a viable option for immunotherapy of the high-risk group.

Radioresistance in cancer cells remains the main limitation in radiotherapy applications. DNA double-strand breaks are the most lethal damage caused by ionizing radiation and trigger a series of DNA damage responses (DDRs) that help cells recover from radiation damage. These DDRs confer radioresistance to the tumor, bringing a poor prognosis [56]. The 4-DCGs have DDR function, therefore we compared the ssGSEA scores of the DDR pathway, X-ray, and UV response in different risk groups. The results displayed that almost all DDR pathways, X-ray, and UV responses were upregulated in the high-risk group, indicating the high radioresistance may contribute to a poor prognosis in patients of the high-risk group. It also suggested that the 4-DCGs can predict radiotherapy responsiveness in LUAD patients.

Platinum-based chemotherapy is the cornerstone of treatment for LUAD patients. However, many patients relapse because of resistance to tumor-killing drugs, leading to a poor prognosis [57, 58]. Therefore, the distinction of primary-resistant LUAD patient populations can maximize the clinical benefit of these patients [59]. Our functional enrichment results showed that the differential genes in the two risk groups were significantly enriched in pathways with platinum drug resistance, thus speculating that our risk grouping could predict drug sensitivity. Unsurprisingly, sensitivity analysis of common anti-tumor drugs displayed that the high-risk group showed low sensitivity in the common anti-lung adenocarcinoma drugs (Cisplatin, Crizotinib, Nilotinib) and other anti-tumor drugs, indicating that the low drug sensitivity contributes to a poor prognosis in the high-risk group. In other words, our grouping could distinguish drug-sensitive populations to some extent. We found that Erlotinib had high sensitivity in the high-risk group, suggesting that the treatment for the high-risk group could adopt Erlotinib. In addition, we also predicted relevant small molecule drug targets using the drug analysis database, which offers some references for the clinical treatment of risk grouping.

In conclusion, the 4-DCGs signature participates in the DNA damage repair and cell cycle checkpoints regulation, and its higher expression implies the better repair in damaged cells of high-risk group patients after chemoradiotherapy, meaning the higher chemoradioresistence. Therefore, combined with the immune activity analysis, the prognosis model constructed by the 4-DCGs signature could reasonably predict the prognosis of LUAD patients. Our study provides auxiliary guidance for the clinical therapy of LUAD, while more clinical cohorts and experiments are required to validate these results further.

The prognosis model based on the 4-DCGs signature could well predict the survival rate, immune activity, and chemoradiotherapy responsiveness of LUAD patients. This study provides some guidance for the treatment and prognosis evaluation of LUAD.