Oral squamous cell carcinoma (OSCC) accounts for most head and neck squamous cell carcinomas, and is the eighth most common cancer worldwide (Bray et al., 2018; Huang L. et al., 2021). Tobacco consumption is thought to be a major etiological factor of OSCC (Das et al., 2007). Although the progress in treatment techniques of OSCC has been notable in recent decades, the overall 5-years survival rate and recurrence rate (approximately 50%) remain disappointing (Nör and Gutkind, 2018). Therefore, it is crucial to develop an efficient and personalized therapeutic strategy for patients with OSCC. Recently, extensive efforts have been dedicated to the identification of prognostic biomarkers or signatures for OSCC based on gene expression or DNA methylation (Huang Z. et al., 2021; Huang et al., 2021b; Zou et al., 2021), whereas their specific roles in guiding personalized treatment still need to be explored in depth.

Oxidative stress is characterized by the imbalance between oxidant and antioxidant production, which contributes to an excess of reactive oxygen species (ROS) and can activate proto-oncogenes and inactivate cancer suppressor genes (Gorrini et al., 2013; Biselli-Chicote et al., 2019). ROS have been identified as a potentially critical factor in driving tumorigenesis and cancer progression (Qiu et al., 2020). Patients with OSCC have been shown to present an elevated level of oxidative stress and a compromised capacity of antioxidant defense (Choudhari et al., 2014). Oxidative stress can induce oxidative damage of DNA and protein, enhancing lipid peroxidation and antioxidant defense disorders, which, if unrepaired, can promote the formation and progression of oral cancer (Toyokuni et al., 1995; Choudhari et al., 2014). Additionally, ROS production is involved in the development of oral cancer in chewers of tobacco and areca nuts (Stich and Anders, 1989), and strictly correlates with the clinical stage in patients with advanced cancer (Mantovani et al., 2002). Furthermore, oxidative stress, as an additional metabolic feature, plays a pivotal immunoregulatory role in the tumor microenvironment (TME) (Cubillos-Ruiz et al., 2015; Maj et al., 2017). Previous studies have demonstrated that oxidative stress could not only alter the phenotype and function of myeloid dendritic cells (DCs) in the TME (Cubillos-Ruiz et al., 2015) but also control tumor Treg cell functional behavior and temper the therapeutic efficacy of immune checkpoint therapy (Maj et al., 2017). Importantly, aberrant levels of ROS can profoundly affect the tumor heterogeneity by modifying the DNA structure of cancer cells, frequently leading to chemotherapeutic resistance (de Sá Junior et al., 2017). Nevertheless, the specific roles of oxidative stress genes (OSGs) in the prognosis, immune status, and chemotherapy response of OSCC remain largely unclear.

In this study, we filtered out six prognostic OSGs to construct a predictive signature according to mRNA expression data from The Cancer Genome Atlas (TCGA) database. Then, the prognostic value of the signature and its connection with clinical features were thoroughly explored in TCGA-OSCC cohort and validated in an independent OSCC cohort GSE41613. Additionally, this signature was shown to have close connections with immune status, immunotherapy response, and chemotherapy sensitivity. Overall, our results demonstrate the potential roles of OSGs in the prognosis, immune status, and drug response of OSCC, and provide a reliable tool for predicting the prognosis of patients with OSCC and guiding clinical treatment.

Oxidative stress, caused by ROS accumulation, plays a crucial role in cancer cells throughout the initiation, progression, metastasis, recurrence, and therapy of multiple tumors (Hayes et al., 2020). Accumulating evidence shows that the ROS levels not only correlate with tumor growth but also affect both the TME and the sensitivity of cancer cells to chemotherapeutic agents (Sosa et al., 2013; de Sá Junior et al., 2017). Additionally, oxidative stress-related gene signatures have been identified to be a reliable and efficient tool to predict the prognosis and progression of cancers (Qiu et al., 2020; Wu et al., 2021). Considering the above views, OSGs may represent valuable biomarkers for predicting the prognosis, immune status, and drug sensitivity of cancers and thus help clinicians define individual treatment plans for patients. However, the predictive effect of OSGs in OSCC remains unclear and is yet to be thoroughly investigated. Hence, in this study, we first filtered out 239 aberrantly expressed OSGs in OSCC and explored their potential functional pathways in OSCC. Next, we selected eight prognostic OSGs, according to which an oxidative stress-related signature with six OSGs was conducted. Moreover, the main functions active in different subgroups classified by the risk signature were identified using GSEA analysis. Finally, to further define the specific roles of OSGs in OSCC, we deeply evaluated the correlations between the oxidative stress-related signature and the prognosis, clinical features, immune status, immunotherapy, and drug sensitivity of OSCC.

The oxidative stress-related prediction signature comprised five risk-associated OSGs (HPRT1, ADA, PLAU, VEGFA, and CXCL8) and one protection-associated OSG (CTLA4), which were verified to be prognostic indicators in OSCC via univariate Cox regression and KM survival analyses. Moreover, these five risk OSGs were significantly upregulated in the HRisk group, while CTLA4, as a protection gene, was notably overexpressed in the LRisk group. HPRT1, a salvage enzyme involved in nucleotide production and recycling in cell cycle modulation, has been shown to promote proliferation and metastasis of head and neck squamous cell carcinoma (HNSCC) through direct interaction with STAT3 and has been implicated as a promising prognostic indicator and potential therapeutic target for HNSCC (Wang et al., 2021b). ADA, a housekeeping enzyme crucial in purine metabolism, makes a certain contribution to the regulation of inflammatory reactions and immune status (Bagheri et al., 2019), and its inhibitor has been shown to be obviously associated with reduced tumor size and decreased aggressiveness of cancer cells (Kutryb-Zajac et al., 2018; Bagheri et al., 2019). Additionally, Wang et al. reported that pre-operative serum ADA levels could be a reliable independent prognostic predictor of OSCC (Zhu et al., 2019). Previous studies demonstrated that the upregulated expression of PLAU (uPA), VEGFA, and CXCL8 (IL8) could promote the occurrence and progression of OSCC and might be important prognostic factors for patients with OSCC (Hundsdorfer et al., 2005; Sales et al., 2016; Magnussen et al., 2017; de Matos et al., 2019; Reyimu et al., 2021). The immune checkpoint inhibitory receptor CTLA4 has also been identified as a potential therapeutic target for OSCC (Bundela et al., 2014) and can enhance the therapeutic efficacy of anti-PD-1 immunotherapy in patients with HPV⁺ OSCC (Dorta-Estremera et al., 2019). Combining the results of the current and previous studies, we believe that these six OSGs could serve as reliable prognostic biomarkers and provide potential treatment targets for OSCC.

The prediction signature, composed of six prognostic OSGs with different weighting coefficients, was verified to be an effective prognostic indicator of OSCC according to the training and validation cohorts using univariate Cox regression and KM survival analyses. Additionally, patients with OSCC in the HRisk group had significantly shorter OS than those in the LRisk group. Multivariate Cox regression analysis indicated that the six-gene prediction signature was an independent prognostic predictor of OS of patients with OSCC. Furthermore, the results of ROC curves and AUC values validated the high prediction accuracy of this signature. In terms of the relationship between the risk score and clinical features, patients with tumor stage III–IV or T_3–4 were significantly associated with a higher risk score, which suggested that the risk score increases with the progression of OSCC, showing the poorer prognosis of patients with the higher risk score.

We next performed GSEA analysis in both the training and validation cohorts to define the specific signal pathways involved in the oxidative stress-related signature. The results demonstrated that cell cycle-related pathways such as the cell cycle, spliceosome, and base excision repair were apparently activated in the HRisk group. Moreover, oxidative stress-mediated ROS production exerts a key influence on cell cycle dysregulation by incorporating phosphorylation, ubiquitination, and receptor activation, which can contribute to aberrant cell proliferation and promote tumor progression (Verbon et al., 2012; Ahmad et al., 2020). Therefore, the activated cell cycle-related pathways may relate to the poor prognosis of patients with OSCC in the HRisk group. Meanwhile, autoimmunity-related pathways (i.e., autoimmune thyroid disease and systemic lupus erythematosus) were enriched in the LRisk group, illustrating that the patients with OSCC in the LRisk group might present an active immune state.

Growing evidences suggest that the TME plays a critical role in carcinogenesis, tumor progression, and survival, among which the immune microenvironment serves as a determinative factor (Hinshaw and Shevde, 2019). Thus, we next explored the relationship between the constructed signature and the immune status of OSCC. According to the ESTIMATE algorithm, a higher risk score was notably correlated with a lower immune score based on TCGA-OSCC and GEO-OSCC cohorts, suggesting that the high-risk score indicated an immune-suppressive state of patients with OSCC. Additionally, patients in the HRisk group displayed significantly higher tumor immune evasion scores compared to low-risk patients, and the immune evasion score showed a positive correlation with the risk score on the basis of the TIDE results in TCGA-OSCC cohort, further demonstrating the poor prognosis and immune-suppressive state of patients in the HRisk group. When we estimated the correlation between the prognostic signature and immune functions using ssGSEA, we found that compared to patients in the HRisk group, those in the LRisk group presented higher enrichment scores of cytolytic activity, promoting inflammation, the type II IFN response, and HLA. Of note, the abovementioned immune functions were all positively associated with a favorable prognosis in patients with OSCC according to both the training and validation cohorts via KM survival analysis, which contributes to the beneficial prognosis of low-risk patients. Rooney et al. reported that the cytolytic activity was associated with a modest but significant pan-cancer survival benefit and was connected to counter-regulatory immune responses (Rooney et al., 2015). Consistently, high cytolytic activity in tumor-free tongue tissue conferred improved prognosis in patients with tongue squamous cell carcinoma (Gu et al., 2019). Type II IFN (IFNγ) is thought to play a crucial role in cancer immunosurveillance, with the ability to promote anti-tumor immunity by increasing tumor immunogenicity (Dunn et al., 2006). Naturally, HLA markers are essential for antigen presentation and display a pivotal role in anti-tumor immunity by enhancing immunosurveillance and preventing immune escape. Additionally, high HLA class I expression in OSCC shows a significantly positive connection with a better prognosis (Koike et al., 2020; Anderson et al., 2021). Besides, it is worth emphasizing that various immune cells in the TME participate in anti-tumor immune responses as main components. Hence, we simultaneously calculated the enrichment level of immune cells in different risk subgroups using ssGSEA. The results showed that patients in the LRisk group displayed significantly higher infiltration of DCs, NK cells, mast cells, neutrophils, B cells, helper T cells, and Treg cells, each of which had a positive relationship with favorable prognosis in patients with OSCC. As the most potent antigen-presenting cells, DCs underpin the successful generation of anti-tumor immune responses by initiating and regulating innate and adaptive immune responses in the TME, and thus, targeting DCs is a promising strategy to improve the efficacy of current immunotherapies (Verneau et al., 2020). NK cells, with the potent ability to kill tumor cells, induce remodeling of the oral TME via IFN-γ and TNF-α, as well as prevent tumor growth and metastasis (Jewett et al., 2018). Furthermore, both CD103⁺ DCs and activated NK cells have been shown to have a favorable prognosis in OSCC (Hadler-Olsen and Wirsing, 2019; Xiao et al., 2019). Interestingly, Tregs were significantly connected to an improved prognosis of OSCC in our study, despite being recognized as immunosuppressive cells in numerous cancers. Among patients with OSCC, a high level of infiltrated Tregs has been proven to be notably associated with a lower frequency of lymph node metastasis and prolonged OS (Bron et al., 2013). Moreover, Hanakawa et al. suggested that Tregs in the TME may prevent tumor cell invasion and metastasis by inhibiting inflammatory processes (Hanakawa et al., 2014). In brief, the immune-related results in our study expectedly cohere with those in previous studies and the improved prognosis of patients in the LRisk group may be a result of the active immune state and increasing anti-tumor immune responses in these patients.

In addition to the TME, as the central component of cancers, tumor cells naturally play a vital role in tumorigenesis. Tumor cells present distinct phenotypic states with different functional attributes, among which, cancer stem cells (CSCs) possess the principal properties of self-renewal, clonal tumor initiation capacity, and clonal long-term repopulation potential (Plaks et al., 2015). From this perspective, CSCs can facilitate the initiation and progression of tumors, which may induce drug resistance (Plaks et al., 2015; Malta et al., 2018). Thus, we evaluated the correlation between the prognostic signature and the tumor stemness of OSCC. Based on DNAss and RNAss, patients with OSCC in the HRisk group presented higher tumor stemness than those in the LRisk group, and the risk score showed a positive connection with tumor stemness. These results suggested that this oxidative stress-related signature could predict tumor progression and invasion for patients with OSCC and simultaneously explain the poor prognosis of high-risk patients.

Presently, personalized medicine, concentrating on designing specific therapeutics that are best suited for an individual patient based on genome information, is considered to be the future direction in oncotherapy (Jackson and Chester, 2015). To verify whether the prognostic signature could guide clinicians to develop an effective personalized treatment decision for patients with OSCC, we thoroughly investigated the correlation between the constructed signature and response to immunotherapy or chemotherapy. Immunotherapy, particularly ICI treatment, offers a reliable alternative or adjunctive therapy to conventional therapies for refractory patients with OSCC (Dorta-Estremera et al., 2019). The TIDE score has been identified as a potent biomarker to predict the tumor response to anti-PD1 or anti-CTLA4 (Jiang et al., 2018), while the TMB score has also been validated as an available tool to help oncologists select patients who may benefit from ICIs (Choucair et al., 2020). Encouragingly, the signature in our study was not only closely correlated with both TIDE and TMB scores but was also efficient in predicting the immunotherapy response with high accuracy (AUC = 0.766). Next, we comprehensively evaluated the efficacy of chemotherapy drugs in different risk subgroups and their relationship to the risk score on the basis of three public drug sensitivity databases (CGP, GDSC, and CTRP) (Garnett et al., 2012; Basu et al., 2013; Yang et al., 2013). The IC50 values of eight common chemotherapy drugs for OSCC (Cisplatin, Paclitaxel, Cytarabine, Docetaxel, Doxorubicin, Gemcitabine, Methotrexate, and 5-Fluorouracil), on the basis of the NCCN guidelines Version 2.2021, and one novel anti-cancer drug, Gefitinib, based on previous studies, were calculated and estimated. Consequently, we found that patients with low-risk scores were more sensitive to 5-Fluorouracil, Gemcitabine, and Gefitinib, while those with high-risk scores showed more sensitive responses to paclitaxel and docetaxel, which indicated that the prognostic signature could facilitate personalized chemotherapy decisions. Furthermore, 5-Fluorouracil, Gemcitabine, and Gefitinib generate ROS, which can cause cancer cell death by inducing oxidative damage, whereas CSCs can increase cellular adaptive responses to ROS to result in chemoresistance (Okon et al., 2015; Blondy et al., 2020; Xue et al., 2020). As mentioned above, OSCC patients with high-risk scores presented notably elevated tumor stemness, which may explain their higher resistance to 5-Fluorouracil, Gemcitabine, and Gefitinib. Moreover, paclitaxel and docetaxel are both anti-cancer agents belonging to the taxane family and can inhibit cancer cell proliferation by inducing cell cycle arrest (Ashrafizadeh et al., 2021). On the basis of the GSEA above, cell cycle-related pathways, such as cell cycle, spliceosome, and base excision repair, were apparently activated with increased risk scores, which shed light on the higher sensitivity of high-risk patients to Paclitaxel and Docetaxel. To summarize, considering the strong correlation of the prognostic signature with both immunotherapy and chemotherapy, we present a valid and robust tool that can guide clinicians to make effective personalized treatment decisions for patients with OSCC with different risk levels.

Despite considerable strides in treatment regimens and new drugs for OSCC, the survival rate has been poor and unsatisfactory in recent decades (Bray et al., 2018; Nör and Gutkind, 2018). Thus, the development of novel drugs for OSCC is still necessary. The new use of old drugs is more cost-effective compared to the exploration and development of novel drugs (Cheng et al., 2021). CMap, a public database for predicting small molecule drugs according to transcriptional expression data, has been used to predict and screen potential novel drug candidates for cancers in previous studies (Cheng et al., 2021; Wang M. et al., 2021). Therefore, to uncover novel drug candidates for patients with OSCC with high-risk scores, we evaluated DEGs between the two risk subgroups and finally screened out 10 potential therapeutic agents, including Dextromethorphan and AH-6809, in the CMap database, all of which have been barely explored in OSCC. Dextromethorphan is a safe FDA-approved drug with few undesirable side effects and usually serves as an effective antitussive agent (Silva and Dinis-Oliveira, 2020). Wang et al. found that Dextromethorphan and Metformin, at their pharmacological doses, could synergistically repress nicotine-enhanced cancer-initiating cell properties and halt tumor progression by directly targeting CHRNA7 to inhibit JAK2/STAT3/SOX2 signaling in esophageal squamous cell carcinoma and perhaps other nicotine-sensitive cancer types (Wang et al., 2021a). Additionally, AH-6809 was reported to inhibit the proliferation of non-small cell lung cancer cells by antagonizing prostaglandin receptors (Casibang and Moody, 2002). Hence, the results of CMap analysis may provide some new promising drug candidates for high-risk patients with OSCC and identify a viable direction for the future research on chemotherapy drugs.

In summary, we established a novel prognostic signature with six OSGs. On the basis of both TCGA-OSCC and GSE41613 cohorts, the signature was proven to be an independent prognostic factor with high accuracy, as well as an impactful indicator for predicting the prognosis and immune status of patients with OSCC. Meanwhile, the constructed signature demonstrated that patients with high-risk scores might benefit more from ICI treatment compared to those with low-risk scores, and the risk score presented a close interaction with the TME and chemotherapy sensitivity. Our findings may also provide valuable new insight into the roles of oxidative stress in the prognosis, TME, immune status, immunotherapy response, and chemotherapy sensitivity of patients with OSCC. The oxidative stress-related signature may provide a valid and robust tool that can not only efficiently predict the prognosis and immune status but also guide clinicians to develop effective personalized therapeutic strategies for patients with OSCC.