Head and neck squamous cell carcinoma (HNSCC) is a frequent malignant tumor derived from the mucosal epithelium in the oral cavity, pharynx, and larynx, which presents an incidence rate of approximately 600,000 new cases and 350,000 deaths annually worldwide and an increasing prevalence among older adults (1, 2). In order to establish etiological risk factors related to lethal malignancy of HNSCC, in-depth studies have been conducted to analyze the genetic, epigenetic, and environmental factors that may trigger the occurrence of tumor (3). However, no effective screening strategy for HNSCC has been found so far, and careful physical examination is still the main method for early detection. In addition, most patients with advanced HNSCC present no clinical history of precancerous lesions (3). Therefore, there is an urgent need to identify novel biomarkers that are able to predict the progression of HNSCC precancerous lesions, treatment response, and survival.

Substantial efforts are being made to identify biomarkers for prejudging the prognosis of HNSCC patients. For example, the CXC chemokines have been proven to provide functional prognosis and suitable therapeutic value for HNSCC (4). A research cohort of 383 HNSCC patients revealed that elevated expression of NR2F6 is related to poorer recurrence-free survival, which could be used as a new prognostic indicator for early detection of local recurrences in patients with HNSCC (5). Recently, increasing autophagy- and ferroptosis-related risk signatures have been constructed to prognosticate survival for patients with HNSCC by a set of bioinformatics analyses (6–8). However, accurate and reliable prognostic indicators or models are still insufficient to improve the clinical outcomes of HNSCC patients.

The tumor microenvironment (TME) refers to a complex ecosystem involving persistent complex interactions between cancer cells, immune cells, stromal cells, and extracellular matrix components, which can play a significant role in tumor initiation and metastasis (9, 10). Substantial evidence has shown that a deep understanding of the heterogeneity of the TME within each type of cancer is essential to identify predictive biomarkers of patient outcomes that can routinely be used in the clinic (11, 12). Numerous reviews of published datasets have shown that the aging microenvironment can dramatically affect normal cells of the TME, which may have a huge impact on tumor progression (13–15). As one of the hallmarks of aging, inflammaging can result in degeneration of tissue and destruction of acute inflammation, which is closely correlated with the occurrence and progression of cancer (16, 17). Cellular senescence plays an important role in promoting the aging process and has been recognized as a primary factor that links inflammaging to a variety of age-related malignant tumors (17, 18). Age-related accumulation of senescence-associated secretory phenotype (SASP) cells can promote cancer progression by remaking the primary and metastatic microenvironment over time to a state where malignant cells are more likely to grow. Previous results have shown that cellular senescence is regulated by aging-related genes (AGs), which present important functions in tumor malignancy (19, 20). Given the tremendous changes in the extracellular matrix, secreted factors, and immune system as age increases, there may be potential clinical implications for immunotherapy and targeted therapy in patients with HNSCC. However, there is limited knowledge about the comprehensive correlations of the AGs with the tumor immune microenvironment (TIME) and the prognosis of HNSCC patients.

In this study, a risk model of the four AGs was constructed and was recognized as an independent prognostic index for HNSCC patients. The correlations of the risk model with immune score, immune cell infiltration, and potential immune checkpoints were systematically assessed based on the aging-related signatures. Furthermore, the effect of the four selected AGs on the TIME was explored to reveal their potential functions in tumor progression. Our findings contributed to clarify the regulatory mechanisms of the AGs related to the TIME and implied that these AGs might be functional prognostic biomarkers for predicting clinical outcomes of HNSCC patients and provide new targets for improving the immunotherapy response.

The quality of life and life expectancy for HNSCC patients have been significantly improved by advances in robotic surgery, radiation and chemotherapy, immunotherapy, and molecular characterization of human cancers (3). However, the high mortality rate and outcomes of patients with advanced-stage HNSCC have remained mostly unchanged due to the lack of effective and reliable predictive biomarkers for monitoring HNSCC development (1, 3). Thus, the identification of novel disease-related prognostic biomarkers with adequate performance and clinical convenience in HNSCC is necessary to develop novel therapeutic approaches for improving treatment outcomes.

Studies have highlighted that many of the hallmarks of aging are the same as those of cancer, such as epigenetic changes, altered intracellular communication, changes in proteostasis, mitochondrial dysfunction, and cellular senescence (26). Some of these shared common characteristics, including increased accumulation of genomic damage, telomere attrition, epigenetic alteration, impaired proteostasis, and nutritional perception disorders, may be attributed to tumor progression (27–29). It has been demonstrated that age-dependent changes in inflammation and immune cell infiltration levels played important roles in tumor progression and malignancy (30). Therefore, systematically exploring the functions of AGs in the TIME is necessary to identify the function of the aging microenvironment in HNSCC progression. However, few studies to date have explored the associations of AGs with the TIME and the OS of patients with HNSCC.

In this study, we identified 41 differentially expressed AGs by GSEA based on the training (TCGA) dataset and further constructed a four-AG risk signature using the LASSO Cox regression analysis. It was confirmed that the risk model of the four selected AGs could be recognized as a satisfactory independent prognostic indicator for predicting the clinical outcomes of HNSCC patients. BAK1, DKK1, and MIF acted as risk factors while CDKN2A acted as a protective factor in the risk model. BAK1 played important roles in the mitochondrial apoptosis process, which could alter the permeability of the mitochondrial outer membrane to induce cell apoptosis (31). DKK1 is an antagonist of Wnt signaling, which regulates various cellular and biological processes that play important roles in cell senescence, cell apoptosis, differentiation, and metastasis in various tissues and numerous cancers (32, 33). Previous studies had shown that elevated expression of DKK1 was an independent adverse prognostic indicator of survival in HNSCC (34, 35). MIF has been determined as a pro-inflammatory cytokine that plays multiple roles in inflammation and angiogenesis and was associated with carcinogenesis (36). It had been reported that overexpression of MIF promoted tumor metastasis and was notably associated with poor prognosis of pancreatic ductal adenocarcinoma (37). Kindt etal. (38) demonstrated that the elevated expression of MIF resulted in tumor progression and poorer prognosis of HNSCC. For CDKN2A, it had been shown that methylation of CDKN2A was highly associated with HNSCC carcinogenesis and bad OS, which could be a diagnostic and prognostic indicator for HNSCC (39). In addition, it has been demonstrated that the copy number loss of CDKN2A could be an effective prognostic biomarker to independently predict poor survival in human papilloma virus (HPV)-negative HNSCC (40).

The clinic correlation analysis confirmed that the risk score was highly associated with tumor grade and pathological T stage. Notably, stratification survival analysis revealed that the four-AG risk signature possessed an accurate predictive value for prognosis in multiple HNSCC subtypes sorted by age, gender, tumor grade, pathological TNM stage, pathological T stage, pathological N stage, and therapy. Calibration curves for the 2-, 3-, and 5-year OS nomogram model showed good consistency between nomogram predictions and actual observations. Notably, the 2-, 3-, and 5-year AUC values of ROC curves for the four-AG signature in both the training and testing cohorts were higher than 0.65, which is better than those of previous prognostic signatures of the seven AGs for HNSCC (41). In addition, the total point of clinic efficacy in our nomogram is much higher than those of previously developed and validated models (8, 41). These results suggested that the identified four-AG risk model has a reliable predictive value for prognosis of HNSCC.

The TME in HNSCC harbors a complex interplay between tumor cells and stromal cells, including endothelial cells, cancer-associated fibroblasts (CAFs), and immune cells, which can promote HNSCC development (3). Numerous studies have reported that HNSCCs are highly infiltrated by immune cells, and different immunophenotypes combining several molecular features have been identified as classifiers for HNSCC, which might be useful in predicting response to different therapies, especially checkpoint inhibition (42, 43). Tumor-infiltrating lymphocytes [including T cells, B cells, and natural killer (NK) cells] and myeloid lineage cells (including macrophages, neutrophils, dendritic cells, and myeloid-derived suppressor cells) are the main immune components of the HNSCC TME (3). Previous studies showed that high levels of tumor-infiltrating lymphocytes were usually correlated with better prognosis of HNSCC depending on the balance of cells with antitumor activity vs. those with immunosuppressive activities (44, 45). Effector T cells and NK cells mainly regulated the antitumor immunity in the TME, whereas Tregs, myeloid dendritic cells, and M2 macrophages largely regulated the immune suppression and tumor cell growth. It had been proven that high infiltration levels of Tregs, myeloid dendritic cells, neutrophils, or M2 macrophages were related to advanced-stage HNSCC or bad OS, whereas high infiltration levels of CD8+ effector T cells and NK cells in the TME were related to better outcomes (46). In this study, patients of the high-risk group contained higher infiltration levels of resting NK cells, M0 macrophages, M2 macrophages, and resting mast cells while having lower infiltration levels of memory B cells, CD8+ T cells, follicular helper T cells, Tregs, and activated mast cells than did those of the low-risk group. In addition, negative correlations between the risk score and infiltration levels of the B cells, CD4+ T cells, CD8+ T cells, neutrophils, macrophages, and myeloid dendritic cells were observed based on TIMER database. All these data suggested that the integral index of the immunoscore obtained on the basis of the expression of the four AGs included in our prognostic signature and expression levels of these AGs are significantly correlated with immunosuppressive activity for HNSCC patients. To explore the mechanisms of AGs on the TIME, the correlations between the four selected AGs and immune cell infiltration levels were investigated. It was found that patients of subtypes of high BAK1 and DKK1 expressions contained lower infiltration levels of memory B cells, CD8+ T cells, follicular helper T cells, Tregs, and activated mast cells, suggesting a low immunosuppressive activity for HNSCC patients of the high-expression groups. Based on the TIMER database, HNSCC patients in the subtype of high BAK1 expression included significantly higher infiltration levels of CD8+ T cells and neutrophils while having a lower infiltration level of B cells. HNSCC patients of the high-MIF expression subtype showed a lower infiltration level of B cells, CD4+ T cells, CD8+ T cells, neutrophils, macrophages, and myeloid dendritic cells than did those of the low-MIF expression subtype. Results showed that the selected AGs in our risk model (BAK1, DKK1, and MIF) were negatively related to the infiltration levels of tumor-infiltrating lymphocytes (including T cells, B cells, and macrophages), suggesting a low immunosuppressive activity for HNSCC patients of the high-risk groups.

Furthermore, we illustrated the relationships of immune checkpoints with the risk score and the four selected AGs to predict the response to different therapies, particularly checkpoint inhibition. The immune checkpoint molecule is an inhibitory regulatory molecule in the immune system, which is essential for maintaining self-tolerance, preventing autoimmune response, and minimizing tissue damage by controlling the time and intensity of the immune response (47). Immune checkpoint molecules expressed on immune cells will inhibit the function of immune cells and prevent the body from producing effective antitumor immune responses, which can result in tumor immune escape (48). Thus, the decreased expression of immune checkpoints may prevent tumor immune escape to improve the outcomes of immunotherapy. Our results showed that the expressions of CTLA4, PDCD1, and TIGIT were found to be downregulated in the high-risk subtype; BAK1 showed crucially positive correlations with the expressions of CD274, LAG3, and PDCD1LG2 while having a negative correlation with SIGLEC15 expression; DKK1 showed a positive relationship with the expression of PDCD1LG2 while having a negative relationship with the expression of TIGIT; CDKN2A revealed a positive correlation with PDCD1 expression; and MIF revealed significantly negative correlations with the expressions of CD274, CTLA4, HAVCR2, PDCD1, PDCD1LG2, and TIGIT. Hence, the four selected AGs in our risk model (BAK1, DKK1, CDKN2A, and MIF) performed important functions on immune checkpoint expressions, which might be regarded as potential targets to improve the outcomes of immunotherapy by downregulating the expressions of immune checkpoints.

Moreover, the relationships of the ICB response score (TIDE score) with the risk score and the four selected AGs were evaluated. TIDE uses a set of gene expression markers to evaluate two different tumor immune escape mechanisms, including the dysfunction of tumor-infiltrating cytotoxic T lymphocytes (CTLs) and the rejection of CTLs by immunosuppressive factors (24). The high ICB response score indicates the poor efficacy of the immune checkpoint-blocking therapy (ICB) and the short survival period after receiving ICB treatment. Our results revealed that the ICB response scores were crucially higher in the high-risk subtype, high-DKK1 expression subtype, and high-MIF expression subtype than those in the low subtypes, which suggest poor immunotherapy efficacy for patients with HNSCC in the high-risk groups.

Nevertheless, there are several limitations in this study. First, the risk signature of the four AGs was obtained and validated based on TCGA datasets. The prognostic value of the AG-related signature in HNSCC patients has not been externally validated due to the lack of our own adequate available data. External validation should be performed based on our own clinic data in the future. Second, the relationship between the risk signature and the immune cell infiltration level and the role of the AG-related signature in immunotherapy efficacy were revealed via comprehensive bioinformatics analysis. More in vitro and in vivo experiments are needed to confirm these conclusions. Laboratory experiments will be done in future research to validate the specific roles of the AG-related signature in immunotherapy efficacy for patients with HNSCC.

In conclusion, our study constructed a risk signature of four AGs that could be utilized as an independent prognostic indicator for HNSCC patients and showed important functions in immunosuppression of HNSCC. Meanwhile, the AG-related signature might provide potential targets to improve the outcomes of immunotherapy.

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 .

FC and JL contributed to conception and design of the study. FC, XG, and MX organized the database and performed the statistical analysis. FC wrote the first draft of the article. FY, JW, and CY wrote sections of the article. All authors contributed to article revision and read and approved the submitted version.

This work was supported by grants from the Science and Technology Planning Project of Guangzhou (202102010041) and Health Science and Technology Project of Guangzhou (20211A010014).

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.