Abstract

   Background: N7-methylguanosine (m7G) methyltransferases and microRNAs
   (miRNAs) are closely associated with tumor progression. However, the
   role of m7G methyltransferase-related miRNAs as prognostic markers in
   oral squamous cell carcinoma (OSCC) has not been studied. This study
   aimed to explore the m7G methyltransferase-related miRNAs in OSCC,
   establish a prognostic model based on m7G methyltransferase-related
   miRNAs, investigate their correlation with immune cell infiltration,
   and assess their potential prognostic value.

   Methods: Transcriptional and clinical data of patients with OSCC were
   obtained from The Cancer Genome Atlas (TCGA) database. TargetScan and
   miRWalk were used to predict m7G methyltransferase-related miRNAs.
   Subsequently, differentially expressed m7G methyltransferase-related
   miRNAs in TCGA-OSCC were selected. Cox and least absolute shrinkage and
   selection operator (LASSO) regression analyses were used to build an
   m7G methyltransferase-related miRNA risk prognostic model for
   TCGA-OSCC. Patients were stratified into high- and low-risk groups. The
   predictive and diagnostic accuracies of the risk prognostic model were
   further validated using Kaplan-Meier survival analysis, receiver
   operating characteristic (ROC) curve analysis, independent prognosis
   analysis, and nomogram plots. Finally, quantitative real-time
   polymerase chain reaction (qPCR) was used to validate the expression
   levels of m7G methyltransferase-related miRNAs in postoperative cancer
   and adjacent normal tissues from 60 patients with OSCC.

   Results: Through Cox and LASSO regression analysis, six candidate
   miRNAs (hsa-miR-338-3p, hsa-miR-1251-3p, hsa-miR-3129-5p,
   hsa-miR-4633-3p, hsa-miR-216a-3p, and hsa-miR-6503-3p) most relevant to
   the prognosis of patients with OSCC were identified to construct an m7G
   methyltransferase-related miRNA risk prognostic model. In this model,
   the overall survival (OS) of the high-risk group was significantly
   shorter than that of the low-risk group (P < 0.001). The model
   effectively predicted prognosis and served as an independent prognostic
   indicator for patients with OSCC. Compared with the low-risk group, the
   high-risk group exhibited a significantly increased capacity for immune
   cell infiltration (P < 0.05), while the activation and initiation
   abilities of immune cells were decreased. Finally, six m7G
   methyltransferase-related miRNAs were validated in OSCC tissue samples.

   Conclusion: The risk prognostic model based on six m7G
   methyltransferase-related miRNAs can predict the OS rate of patients
   with OSCC and has the potential to guide individualized treatment. This
   prognostic model is closely associated with immune cell infiltration in
   patients with OSCC.

   Keywords: oral squamous cell carcinoma, N7-methylguanosine
   methyltransferase, microRNA, prognosis, immune microenvironment

1. Introduction

   Head and neck squamous cell carcinoma (HNSCC) is the sixth most common
   cancer, with approximately 700,000 diagnosed cases [41]^1^,[42]^2. Oral
   squamous cell carcinoma (OSCC) accounts for approximately 90% of all
   malignant tumors of the head and neck region [43]^1^,[44]^3. OSCC is a
   heterogeneous tumor originating from the inner layer of the oral
   mucosa, with a global incidence, particularly in developing countries.
   Its incidence is higher in male patients than in female patients
   [45]^4. Although primary tumor resection remains the standard treatment
   [46]^5, OSCC's high invasiveness often leads to postoperative
   recurrence and metastasis [47]^6, resulting in a relatively low
   survival rate of 5 years [48]^7. Although targeted therapy and
   immunotherapy have improved outcomes [49]^8^,[50]^9, their application
   is limited by individual variability and issues related to drug
   resistance [51]^10. Therefore, effective biomarkers are needed to
   predict OSCC prognosis, enable accurate assessment of patient
   prognosis, and guide subsequent treatment.

   N7-methylguanosine is a commonly occurring RNA post-transcriptional
   modification formed by methylation of the seventh nitrogen atom of
   guanine at the RNA purine position [52]^11. It plays vital roles in
   different RNA stages such as RNA transcription, processing,
   degradation, and translation. In addition to being presented at 5' end
   and internal positions of mRNA in eukaryotes [53]^12^-[54]^14, m7G
   modification is also widely found in rRNA, tRNA, and miRNA
   [55]^15^,[56]^16. However, unusual m7G modifications are often linked
   to tumor occurrence and progression [57]^17. The m7G regulatory factors
   methyltransferase 1 (METTL1) and WD Repeat Domain 4 (WDR4) participate
   in the regulation of various cancer types, including HNSCC, liver
   cancer, bladder cancer, and lung cancer, by changing the m7G
   modification levels of miRNAs and tRNAs [58]^18^,[59]^19. METTL1 and
   WDR4 are significantly upregulated, increasing m7G modifications in
   tRNAs, which enhances oncogenic mRNA translation, promoting tumor
   progression and poor prognosis. METTL1 knockout alters the link between
   cancer cells and their microenvironment in HNSCC. The ratio of CD4^+ T
   cells to Tregs is significantly reduced, whereas the permeation of
   CD4^+ memory T cells, CD4^+ naïve T cells, and CD8^+ naïve T cells
   increases [60]^19. These findings highlight the crucial roles of METTL1
   and WDR4 in HNSCC.

   MicroRNAs (miRNAs) are single-stranded noncoding RNAs with lengths
   ranging from 19-25 nucleotides. They have vital functions in
   post-transcriptional regulation by binding to mRNAs and long noncoding
   RNAs that influence important biological processes such as cell
   proliferation, differentiation, and apoptosis [61]^20. In addition,
   miRNAs are closely related to immune response and angiogenesis
   [62]^21^,[63]^22. However, studies have confirmed that the
   dysregulation of miRNAs is closely linked to cell metastasis and drug
   resistance in cancer cells. Several miRNAs have been identified as
   possible diagnostic and prognostic biomarkers for OSCC and used to
   construct prognostic models [64]^23^-[65]^26. However, the role of m7G
   methyltransferase-related miRNAs in OSCC remains unclear. Therefore, it
   is essential to further investigate their significance as prognostic
   biomarkers for patients with OSCC and provide insights into novel OSCC
   treatment methods.

   Based on this evaluation, we predicted the upstream miRNAs associated
   with the N7-methyltransferase METTL1 and the WDR4 complex. Using the
   TCGA database, a risk prognosis model for m7G methyltransferase-related
   miRNAs was constructed and validated in our study.

2. Materials and Methods

2.1 Data Collection and Processing

   Transcriptome mRNA sequencing data (240 tumor and 17 normal samples)
   and miRNA sequencing data (245 tumor and 17 normal samples) of patients
   with OSCC were acquired from the TCGA database
   ([66]https://portal.gdc.cancer.gov/) [67]^27. Corresponding clinical
   information, including patient age, sex, tumor grade, TNM stage,
   survival time, and tumor status, was also retrieved. We processed the
   data using Perl software by parsing and extracting mRNA and miRNA
   expression levels, generating respective expression matrices with genes
   or miRNAs as rows and samples as columns. We created a clinical data
   matrix with patients in rows and clinical features in columns. These
   matrices were integrated based on patient identifiers to correlate all
   the data types for each patient. The quality control steps included
   checking for missing values, normalizing expression data, and filtering
   out low-quality features, ensuring that subsequent analyses were based
   on high-quality and well-integrated datasets [68]^27.

2.2 m7G Modification-Related miRNA Screening and Enrichment Analysis

   The R package "limma" was employed for comparison of expression
   intensities of both m7G methyltransferase genes METTL1 and WDR4 in
   TCGA-OSCC tissues and normal tissues. Next, upstream miRNAs targeting
   METTL1 and WDR4 were predicted using the TargetScan database
   ([69]https://www.targetscan.org/) and miRWalk
   ([70]http://mirwalk.umm.uni-heidelberg.de/), respectively. This
   intersection was used to obtain a list of miRNAs related to m7G
   modifications. Subsequently, the "edgeR" R package was utilized to
   analyze the expression intensities of these m7G modification-associated
   miRNAs inside TCGA-OSCC and normal tissues. FDR < 0.05 and
   |log[2]Foldchange| > 1 were applied to identify significantly
   differentially expressed m7G modification-related miRNAs, and a network
   relationship diagram was drawn using the Cytoscape 2.8 tool [71]^28.
   Finally, the FunRich tool was used to enhance the differential
   expression of these m7G modification-associated miRNAs.

2.3 Construction and Evaluation of Prognostic Model with Basis on m7G
Modification-Related miRNAs

   Initially, 245 patients with OSCC were randomly categorized into
   training and testing sets in a ratio of 7:3. The training set comprised
   172 patients for model construction and the testing set included 73
   patients to validate the model's performance. Subsequently, the
   Kaplan-Meier "survival" R package was used for Cox regression analysis
   to assess the prognostic significance of the differentially expressed
   m7G modification-associated miRNAs (P < 0.05). The "glmnet" R package
   was then applied for least absolute shrinkage and selection operator
   (LASSO) to refine and select miRNAs for constructing the prognostic
   model. The risk score calculation formula for each patient was
   described as follows: Risk Score = (0.326 * expression of
   hsa-miR-338-3p) + (1.597 * expression of hsa-miR-1251-3p) + (2.288 *
   expression of hsa-miR-4633-3p) + (0.743 * expression of
   hsa-miR-216a-3p) + (0.348 * expression of hsa-miR-6503-3p) - (0.494 *
   expression of hsa-miR-3129-5p).

   Next, using the median risk scores in the training set as the
   threshold, patients with OSCC in the training, testing, and entire TCGA
   sample sets were stratified into high- or low-risk groups for
   subsequent examination and validation. The Kaplan-Meier curve (log-rank
   analysis) was employed to assess the survival level differences between
   the high- and low-risk groups. A prime component evaluation was
   utilized to evaluate accurate grouping, and model accuracy was assessed
   by calculating the area under the receiver operating characteristic
   (ROC) curve. A concordance index was used to assess the discriminative
   ability of the model. Finally, univariate and multivariate Cox
   proportional hazards analyses were conducted to evaluate the risk
   scores of the model to check the clinical features and self-regulating
   predictive biomarkers in patients with OSCC.

2.4 Nomogram Construction

   To enhance clinical applicability, the "rms" R package was used to
   create a nomogram, integrating model risk scores and patient clinical
   data. This nomogram helped quantify factors affecting the analysis of
   patients with OSCC and predict the survival period of patients. A
   calibration curve was then generated to assess the model's predictive
   accuracy, demonstrating consistency between predicted and actual
   survival times.

2.5 Association Analysis between Prognostic Model and Tumor Immune
Microenvironment

   The R packages "estimate,” "GSEABase," and "GSVA" were employed to
   execute the ESTIMATED algorithm as well as conduct single sample Gene
   Set Enrichment Analysis [72]^29. This allowed the evaluation of 23
   immune cell infiltrations and 13 immune-associated functional scores
   for each TCGA-OSCC model. Subsequently, we assessed immune cell
   penetration and patient immunity within the high- and low-risk groups.
   In addition, expression level difference analyses were performed to
   identify checkpoints for immune-associated genes among the various
   subgroups. Finally, to evaluate the antitumor immune function scores
   for each patient, we used Immunophenotype Element Server and Tracking
   Tumor ([73]http://biocc.hrbmu.edu.cn/TIP/index.jsp) to compare the
   antitumor immunity systems between the two risk groups.

2.6 Prediction of Potential Regulatory Mechanisms and Sensitivity for
Potential Therapeutic Drugs between High- and Low-Risk Subgroups

   Enhancement of gene sets was employed to identify signal passages and
   differentially expressed genes between the high- and low-risk groups.
   This was performed to unveil the potential regulatory mechanisms
   contributing to the prognostic differences between these two risk
   subgroups. Additionally, utilizing R software and the "pRRophetic"
   package, we calculated half- maximal inhibitory concentration (IC[50]),
   commonly used to assess antitumor drug efficacy. Drug sensitivity was
   also detected in the high- and low-risk groups. Lower IC[50] values
   indicated high sensitivity of the drugs for treatment.

2.7 Detection of Relative m7G Expression Levels in Modification-Related
miRNAs within Clinical OSCC Tissues

   Clinical OSCC tissues and paired adjacent normal tissues were acquired
   from 60 patients at the Oral and Maxillofacial Surgery Section of the
   Peking University Shenzhen Hospital. These patients had not undergone
   cancer treatment before surgery. The tissue samples were rapidly
   transferred and stored in liquid nitrogen, which was approved by the
   Ethical Society of Peking University Shenzhen Hospital (grant number:
   2022-117). Informed consent was obtained from all patients or their
   relatives. Total RNAs was purified from OSCC and normal tissues using
   TRIzol reagent (Takara Bio Inc., Kusatsu, Japan). cDNA for miRNA
   examination was generated using the stem-loop method (Accurate Bio
   Inc., Hunan, China) [74]^13. Subsequently, real-time quantitative PCR
   was performed using SYBR^® Green Premix Pro Taq HS qPCR Kit II
   (Accurate Bio Inc, Hunan, China) on a LightCycler® 480 PCR instrument
   (Roche, Indianapolis, IN, USA). The expression of the six m7G
   methyltransferase-associated miRNAs were normalized to U6 and
   calculated using the 2^^-ΔΔCt method in triplicate. The primer
   sequences are listed in [75]Supplementary Table 1.

2.8 Statistical Analysis

   The statistical analysis of public data was performed using R (version
   4.2.1), and experimental data were visualized and statistically
   analyzed using GraphPad Prism 9.5.1. Significance testing for
   differences was conducted using the Wilcoxon test, and correlation
   analysis was performed using the Spearman's rank relationship method,
   Statistical significance was set at p<0.05.

3. Results

3.1 Identification of Differentially Expressed m7G
Methyltransferase-Associated miRNAs in TCGA-OSCC Tissues

   Analysis of TCGA transcriptome data revealed that the RNA levels of m7G
   methyltransferase METTL1 and WDR4 were significantly higher in OSCC
   tissues than in normal tissues (Figure [76]1A, B). METTL1 and WDR4 were
   significantly and positively correlated (r = 0.51, P < 0.001; Figure
   [77]1C). Subsequently, the upstream miRNAs related to the m7G
   methyltransferase METTL1-WDR4 complex were predicted using the
   TargetScan and miRWalk online databases (Figure [78]1D). In total, 346
   miRNAs associated with m7G methylation were identified. Among them, 56
   miRNAs were differentially expressed in TCGA-OSCC tissues (Figure
   [79]1E), with 24 upregulated and 32 downregulated miRNAs (FDR < 0.05,
   |log[2] Fold Change| > 1). Among the 56 miRNAs, 46 targeted METTL1 and
   10 targeted WDR4 (Figure [80]1F).

Figure 1.

   [81]Figure 1
   [82]Open in a new tab

   Expression and regulatory network analysis of m7G methyltransferase
   complex METTL1/WDR4 mRNA and associated miRNAs in TCGA-OSCC. (A, B)
   METTL1 and WDR4 expression levels in the TCGA-OSCC cohort. (C)
   Correlation evaluation of expression between WDR4 and METTL1. (D)
   Prediction scores of miRNAs binding to METTL1 and WDR4 mRNA using
   TargetScan and miRWalk online databases (Venn diagram). (E) Volcano
   plot depicting differential expressed m7G methyltransferase-associated
   miRNAs between normal and tumor tissues in the TCGA-OSCC cohort. (F)
   Co-expression network of miRNA-m7G-associated genes.

3.2 Functional Evaluation for m7G Methyltransferase-Associated miRNAs in
TCGA-OSCC

   Functional enhancement evaluation of m7G methyltransferase-associated
   miRNAs using the FunRich tool revealed enrichment in 89 items related
   to biological processes, 147 items related to molecular functions, 443
   items related to cellular components, and 494 items related to
   biological pathways. In terms of biological processes, these miRNAs
   were significantly enriched in cell communication, signal transduction,
   nucleic acid metabolism, and immune response (Figure [83]2A). In terms
   of cellular components, these miRNAs were distributed in the nucleus,
   cytoplasm, exosomes, and plasma membrane (Figure [84]2B). The molecular
   functions were primarily associated with transcription factor activity,
   serine or threonine kinase activity, cell adhesion molecule activity,
   and T-cell receptor activity (Figure [85]2C). Additionally, m7G
   methyltransferase-associated miRNAs were found to be involved in
   glypican-, IFN-γ-, and IL5-mediated signaling pathways (Figure [86]2D).

Figure 2.

   [87]Figure 2
   [88]Open in a new tab

   Enrichment analysis of 56 m7G methyltransferase-associated miRNAs. (A)
   Biological processes, (B) cellular components, (C) molecular functions,
   and (D) biological pathways of m7G methyltransferase-associated miRNAs.

3.3 Construction and Validation of m7G Methyltransferase-Associated miRNA
Prognostic Model

   Cox and LASSO regression analyses were performed for the 56 m7G
   methyltransferase-associated miRNAs and six representative prognostic
   miRNAs (Figure [89]3A, B). The corresponding Cox coefficients are shown
   in Table [90]1. LASSO analysis was validated through cross-validation
   (Figure [91]3C). Subsequently, 3'UTR targeting analysis revealed
   binding sites in the 3'UTR region of the METTL1 and WDR4 mRNA for these
   six miRNAs (Figure [92]3D). Using these six m7G
   methyltransferase-associated miRNAs, a predictive risk model was
   developed for a training set of 172 samples. Based on the median risk
   score of the training set (0.91896), the 172 TCGA-OSCC cases were
   categorized into low- or high-risk groups. The number of deaths in the
   high-risk group was significantly higher than that in the low-risk
   group (P < 0.001). As the danger score increased, patient deaths also
   increased (Figure [93]4A). Subsequently, model applicability was
   further validated in the test and all-sample sets. Based on the median
   risk score from the training set, patients in the test and all-sample
   sets were categorized into low- or high-risk groups (Figure [94]4B, C).
   According to the Kaplan-Meier survival curve results, the survival time
   of the low-risk group was significantly longer than that of the
   high-risk group.

Figure 3.

   [95]Figure 3
   [96]Open in a new tab

   Establishment of the m7G methyltransferase-associated miRNA prognostic
   model. (A) Univariate Cox regression analysis of six m7G
   methyltransferase-associated miRNAs shown in a forest plot. (B) LASSO
   regression analysis. (C) Cross-validation. (D) Prediction results of
   binding sites at the 3'UTR of METTL1 and WDR4 mRNA for six miRNAs.

Table 1.

   The six candidate miRNAs and their corresponding coefficients involved
   in the model construction.
       miRNAs      Coefficient
   hsa-miR-338-3p  0.326343
   hsa-miR-1251-3p 1.59744
   hsa-miR-4633-3p 2.288034
   hsa-miR-216a-3p 0.742978
   hsa-miR-6503-3p 0.347863
   hsa-miR-3129-5p -0.49423
   [97]Open in a new tab

Figure 4.

   [98]Figure 4
   [99]Open in a new tab

   Prognostic model for m7G methyltransferase-associated miRNAs
   constructed based on TCGA-OSCC patient data. (A) Risk value
   distribution and Kaplan-Meier survival curve for the training set. (B)
   Risk value distribution and Kaplan-Meier survival curve for the testing
   set. (C) Risk value distribution and Kaplan-Meier survival curve for
   the entire TCGA sample.

   The primer component investigation results for the all-sample set
   showed that the transcriptional features of the high- and low-risk
   patient groups were significantly different (Figure [100]5A). This was
   characterized by substantial inter-group differences and good
   intra-group repeatability, indicating distinct transcriptional features
   between the two risk groups and allowing for effective differentiation.
   To evaluate the prognostic model, ROC curves were generated for the
   training set, yielding AUC scores above 0.7 at 1, 3, and 5 years. The
   model demonstrated strong accuracy in predicting survival at these
   intervals (Figure [101]5B). This phenomenon was further confirmed in
   the test and all-sample sets (Figure [102]5C, D). Additionally, the
   concordance index (C-index) of our risk score model was better than
   that of clinical features, such as age, sex, tumor grade, and TNM
   phase, confirming its reliability as a diagnostic tool for OSCC (Figure
   [103]5E).

Figure 5.

   [104]Figure 5
   [105]Open in a new tab

   Accuracy of the constructed m7G methyltransferase-associated miRNA
   prognostic model in predicting survival rates for patients with OSCC.
   (A) Principal component analysis of TCGA whole-sample data. (B)
   Time-reliant ROC curve for the training set. (C) Time-reliant ROC curve
   for the testing set. (D) Time-reliant ROC curve for the complete TCGA
   sample. (E) Concordance index analysis comparing risk scores with
   clinical variables such as age, sex, tumor grade, and TNM stage.

3.4 Establishment and Validation of Nomogram

   According to the univariate Cox analysis results shown in Figure
   [106]6A, the risk score and TNM staging were significantly related to
   patient diagnosis (P < 0.01). Multivariate Cox regression testing
   results further indicated that both the risk score and TNM stage could
   independently serve as prognostic biomarkers for OSCC (P < 0.05),
   offering valuable insights into patient outcomes (Figure [107]6B).

Figure 6.

   [108]Figure 6
   [109]Open in a new tab

   Construction of a nomogram. (A, B) Forest plots showing results of
   univariate and multivariate regression analyses for risk scores. (C)
   Nomogram combining risk values from the prognostic model with patient
   TNM staging. (D) Calibration curve for the nomogram. (E-G) ROC curves
   of the nomogram for predicting survival at 1, 3 and 5 years.

   Based on the independent prognostic analysis results, a nomogram graph
   that included the model's risk score and the patient's TNM stage was
   drawn to increase the application of the experimental model (Figure
   [110]6C). This nomogram aimed to determine the overall survival (OS)
   rate of patients with OSCC at 1, 3 and 5 years. The calibration curve
   showed increased consistency within the estimated time limit and the
   actual observed survival time in patients with OSCC (Figure [111]6D).
   This result indicated that the survival time of patients with OSCC
   could be efficiently predicted using the nomogram. Moreover, the
   nomogram's ROC curves displayed AUC values of approximately 0.7 for 1,
   3, and 5 years (s), outperforming other clinical variables (Figure
   [112]6E-G). This suggests that the nomogram, along with the basis of
   the model's danger values, provides more accurate predictions of OS
   rates in patients with OSCC than predictions based on independent
   clinical variables. This highlights the nomogram's potential for
   precise clinical diagnosis.

3.5 High- and Low-Risk Groups Exhibited Distinct Tumor Immune
Microenvironments

   To expand the immunotherapy options for patients with OSCC owing to
   limited treatment options, we conducted immune-related analyses on a
   prognostic model. Using the ESTIMATE algorithm, we found that the
   immune score in the high-risk group was higher than that in the
   low-risk group (Figure [113]7A), indicating greater immune cell
   penetration in the high-risk group. This implies a potential link
   between the risk score and resistant microenvironment in patients with
   OSCC. Subsequently, the ssGSEA algorithm was used to assess the
   permeation levels for 23 resistant cell varieties and enrichment scores
   for 13 protection-related pathways in patients between the two
   hazardous categories. The results showed that the high-risk group had a
   stronger potential for resistant permeation and higher proportions of
   initiated CD8^+ T cells, type I T helper cells, CD56 natural killer
   cells, and immature dendritic cells than the low-risk group (Figure
   [114]7B). Additionally, cytolytic activity and inflammation promotion
   were significantly upregulated in the high-risk group (Figure [115]7C).
   The expression of multiple immunity-verifying genes was also
   significantly elevated in the high-risk group (Figure [116]7D),
   indicating that the resistant microenvironments of high-risk patients
   may be abnormally active.

Figure 7.

   [117]Figure 7
   [118]Open in a new tab

   Immune infiltration and tumor microenvironment in high- and low-risk
   OSCC groups. (A) Comparison of immune levels between high- and low-risk
   groups. (B) Comparison of infiltration levels for 23 resistant cell
   types within high- and low-risk groups. (C) Comparison of elevated
   values for 13 resistance-associated pathways within high- and low-risk
   groups. (D) Comparison of resistant gene expression levels between
   high- and low-risk groups. (E) Assessment of anti-cancer-immune
   activity in the cancer-immunity cycle for patients with OSCC in the
   high- and low-risk groups.

   Furthermore, the TIP meta-server method was used to evaluate the
   anti-cancer-immune functions in high- and low-risk patients across the
   cancer-immune phase, including cancerous antigen release, antigen
   presentation, T-cell recognition, and cancer cell killing. As shown in
   Figure [119]7E, the high-risk group showed significantly stronger
   immune cell infiltration than the low-risk group (P < 0.05). However,
   regarding immune cell priming and activation, the high-risk group shown
   weaker activity than the low-risk group. These results suggest that the
   promotion of immune cell activation within the high-risk group and the
   improvement of immune cell penetration within the low-risk group might
   contribute to better clinical outcomes in patients with OSCC.

3.6 Pathway Enrichment Analysis and Tumor Drug Sensitivity Analysis between
High- and Low-Risk Groups

   We conducted an enrichment study for gene sets to explore potential
   regulatory processes that differed between high- and low-risk groups.
   Additionally, we performed a tumor drug sensitivity analysis on
   TCGA-OSCC patient data to provide a theoretical foundation for
   personalized treatment. The pathway enrichment results, as illustrated
   in Figure [120]8A, revealed significant upregulation of the oxidative
   phosphorylation pathway in the high-risk group. Furthermore, as
   depicted in Figure [121]8C, a comparison of the generally utilized
   IC[50] scores of cancer therapeutic drugs between the two categories
   showed a markedly lower IC[50] value for the oxidative phosphorylation
   inhibitor phenformin in the high-risk subgroup (P < 0.001). Moreover,
   the patient risk scores were significantly negatively associated with
   phenformin IC[50] values (r = -0.24, P < 0.001). These findings suggest
   that phenformin may be an effective therapeutic drug for patients with
   high-risk OSCC. In the low-risk group, signaling pathways, such as
   those associated with epithelial-mesenchymal transition and
   transforming growth factor-beta, were notably upregulated (Figure
   [122]8B). Additionally, the comparison revealed that the multikinase
   inhibitor AMG-706 (motesanib) presented significantly low IC[50] values
   in low-risk group (P < 0.001). These results indicate that motesanib
   might be more effective in patients with low-risk OSCC. As a
   multikinase inhibitor, motesanib can interfere with various biological
   functions such as angiogenesis, cell growth, and EMT (Figure [123]8D).
   These findings suggest that our m7G methyltransferase-related miRNA
   prognostic model performs well in guiding drug treatment in patients
   with OSCC.

Figure 8.

   [124]Figure 8
   [125]Open in a new tab

   Pathway enrichment investigation and chemotherapy drug sensitivity
   analysis within high- and low-risk groups. (A, B) Pathway enrichment
   analysis results within high- and low-risk groups. (C, D) IC[50] values
   of phenformin and AMG-706 for the two risk groups.

3.7 Expression Level Detection of m7G Methyltransferase-Related miRNAs in
OSCC Tissues

   The expression levels of six prognostic model m7G
   methyltransferase-related miRNAs were assessed using qPCR in
   postoperative OSCC tissues and their corresponding neighboring
   noncancerous tissues. As shown in Figure [126]9A-F, the expression
   levels of hsa-miR-216a-3p, hsa-miR-1251-3p, hsa-miR-3129-5p,
   hsa-miR-4633-3p, and hsa-miR-6503-3p were elevated in OSCC tissues
   relative to those in noncancerous tissues. However, hsa-miR-338-3p
   showed lower expression in OSCC tissues than in noncancerous tissues.
   However, further statistical analysis revealed that only hsa-miR-338-3p
   (P < 0.01), hsa-miR-3129-5p (P < 0.01), hsa-miR-4633-3p (P < 0.05), and
   hsa-miR-6503-3p (P < 0.001) differed significantly between OSCC and
   noncancerous tissues. Considering the comprehensive analysis of TCGA
   data, hsa-miR-338-3p and hsa-miR-3129-5p can be considered more
   reliable miRNAs for the m7G methyltransferase-related miRNA prognosis
   model genes.

Figure 9.

   [127]Figure 9
   [128]Open in a new tab

   Expression levels of six candidate miRNAs associated with m7G
   methyltransferase genes in OSCC tissues. (A-F) Relative expression
   levels of six miRNAs were assessed using qPCR within 60 pairs of OSCC
   tissues and adjacent normal tissues.

4. Discussion

   OSCC is a highly heterogeneous tumor. Different tumor progression and
   prognoses may be observed even in patients with OSCC with similar tumor
   grading and clinical staging. While there are various treatment methods
   such as surgery, chemotherapy, and radiotherapy that can, to some
   extent, improve the survival time of patients with OSCC, the prognosis
   has been unsatisfactory for those in advanced stages and with
   metastatic OSCC. The predictive capabilities of some confirmed
   biomarkers are limited and insufficient for OSCC diagnosis, treatment,
   and patient survival evaluation. Therefore, exploring new prognostic
   markers for the accurate prediction of OSCC patient prognosis and
   providing guidance for personalized treatment to avoid undertreatment
   and overtreatment are crucial. miRNAs are associated with cancer
   development and treatment. Studies have suggested a crucial function of
   m7G methyltransferase-related genes in the incidence and growth of
   cancer, especially the m7G regulatory factors METTL1 and WDR4. It has
   been confirmed that the expression of these two factors is
   significantly upregulated in neck and head squamous carcinoma, leading
   to an increase in the m7G modification level of tRNA and the promotion
   of tumor progression [129]^19^,[130]^30. Subsequently, m7G
   methyltransferase-associated miRNAs have been increasingly used in
   cancer research. Models predicting patient prognosis have been
   successfully constructed using m7G methyltransferase-related miRNAs in
   liver cancer, breast cancer, renal clear cell carcinoma, and lung
   adenocarcinoma, confirming their potential value and role as tumor
   prognostic factors [131]^31^-[132]^35. However, there have been no
   reports on m7G methyltransferase-associated miRNAs in OSCC. Therefore,
   the latent interface of m7G methyltransferase-related miRNAs can be
   used to determine their probable predictive value.

   We used the m7G methyltransferases METTL1 and WDR4 to predict miRNAs
   associated with m7G methylation. We identified 56 differentially
   expressed m7G methylation-related miRNAs in OSCC tissues. These miRNAs
   are primarily involved in the regulation of tumor metabolism and
   immune-related pathways. Abnormal activation of these signaling
   pathways induces OSCC growth, invasion, metastasis, and changes in the
   tumor-resistant microenvironment [133]^36^-[134]^38. We constructed a
   novel OSCC risk prognostic model using six representative m7G
   methylation-related miRNAs (hsa-miR-338-3p, hsa-miR-1251-3p,
   hsa-miR-4633-3p, hsa-miR-216a-3p, hsa-miR-6503-3p, and
   hsa-miR-3129-5p). Among these miRNAs, hsa-miR-338-3p, hsa-miR-1251-3p,
   hsa-miR-4633-3p, hsa-miR-216a-3p, and hsa-miR-6503-3p are considered
   risk factors for OSCC, whereas hsa-miR-3129-5p is considered a
   protective factor. Although these six miRNAs have been reported in
   other cancers, studies on OSCC are limited. Zhao et al. found that
   miR-338-3p is an oncogene in mucoepidermoid carcinoma [135]^39, while
   Abbas et al. discovered that miR-216a-3p may promote oral cancer
   progression by targeting adenylate cyclase 2 (ADCY2) [136]^40.
   Additionally, Martinez et al. found that the upregulation of
   miR-6503-3p is associated with decreased stability of primary cilia in
   thyroid cells, potentially triggering autoimmune thyroid diseases
   [137]^41. Cao et al. found that the upregulation of lncRNA MALAT1
   mediates doxorubicin immunity and hepatocellular carcinoma progression
   through miR-3129-5p [138]^42^,[139]^43. Based on these six m7G
   methylation-related miRNAs, we constructed a prognostic risk model to
   develop new strategies for the prognosis and treatment of OSCC.
   Furthermore, qPCR was used to analyze miRNA expression in OSCC tissues
   and adjacent normal tissues. The results showed major expression
   differences in hsa-miR-338-3p, hsa-miR-3129-5p, hsa-miR-4633-3p, and
   hsa-miR-6503-3p between the OSCC tissues and adjacent normal samples.
   Through integrated analysis of TCGA dataset results, hsa-miR-338-3p and
   hsa-miR-3129-5p were found to be the most reliable miRNAs for
   predicting the m7G methylation-related miRNA model. Specifically,
   hsa-miR-3129-5p, a protective factor against OSCC, was upregulated in
   OSCC tissues, possibly inhibiting OSCC progression by enhancing its
   targeting of WDR4. Conversely, hsa-miR-338-3p, a risk factor for OSCC,
   was significantly downregulated in OSCC tissues, consistent with a
   previous analysis.

   Owing to the high heterogeneity and complex etiological factors of OSCC
   tumors, TNM staging is also challenging for predicting or describing
   the individual risks and prognoses of patients with OSCC at the same
   stage. In comparison with TNM staging, our constructed m7G
   methyltransferase-related miRNA prognostic model exhibited
   self-assessment values for 1, 3, and 5 years prognosis, all exceeding
   0.7. This indicated that the prognostic model built on m7G
   methyltransferase-related miRNAs exhibited excellent presentation for
   predicting the diagnosis of patients with OSCC. Furthermore, the
   concordance index curve demonstrated that our prognostic model was more
   reliable for assessing patient prognosis than TNM staging and other
   clinical indicators. Independent prognostic analysis confirmed the
   effectiveness of the diagnostic model as a self-regulating indicator
   for patients with OSCC.

   The possible regulatory processes underlying the differences between
   the two subgroups were investigated. Notably, we found an upregulated
   oxidative phosphorylation pathway in the high-risk group compared with
   the low-risk group. This suggests that inhibitors of oxidative
   phosphorylation might have had a stronger antitumor effect in the
   high-risk group of patients with OSCC. Additionally, higher enrichment
   of gene sets associated with Parkinson's disease and myocardial
   contraction was found in this group, suggesting that patients with
   these conditions may have higher risk scores. Interestingly, the
   potential associations between these diseases open more possibilities
   for the development of new drugs for the treatment of OSCC. For
   instance, Eizuka et al. discovered that L-3,4-dihydroxyphenylalanine, a
   permitted drug for the treatment of Parkinson's disease, could
   potentially become a new drug for the treatment of OSCC. This is
   because it hinders OSCC development by down-regulating SYT12 expression
   [140]^44^,[141]^45.

   The tumor microenvironment varies among patients, and these differences
   contribute to the proliferation, migration, and invasion capabilities
   of OSCC cells, explaining the significant variations in prognosis among
   patients [142]^46^-[143]^48. High-risk patients exhibited increased
   immune penetration, indicating higher immunity. However, considering
   patient survival curves, this abnormally active immune status did not
   improve patient prognosis. Additionally, most immune-related gene
   expression levels were higher in the high-risk group, suggesting that
   increased inflammation and elevated immune checkpoint activity may
   contribute to immunodeficiency, partially explaining the poor prognosis
   of high-risk patients with OSCC. Moreover, compared to low-risk
   patients, high-risk patients had relatively weaker immune cell
   activation functions, although immune permeation into the tumor was
   significantly stronger. This implies that promoting immune cell
   activation in high-risk patients and improving immune cell penetration
   in low-risk patients may improve the clinical prognosis of patients
   with OSCC [144]^49. Furthermore, we propose that the upregulation of
   immune checkpoint-associated genes in high-risk patients may represent
   potential targets for immunotherapy in this population. Notably,
   TNFRSF14 (HVEM) and its receptor BTLA (CD272) were significantly
   upregulated in high-risk patients. The upregulation of TNFRSF14 in
   tumor cells contributes to the inhibition of antitumor resistance by
   BTLA, leading to disease development and worse diagnostic outcomes
   [145]^50^,[146]^51. Therefore, targeting BTLA-HVEM in high-risk
   patients may be a feasible immunotherapeutic approach.

   Currently, radiotherapy and chemotherapy are the main approaches for
   the postoperative treatment of patients with OSCC. In chemotherapy,
   platinum-based drugs such as 5-fluorouracil, paclitaxel, and
   doxorubicin are the most commonly used drugs. However, owing to
   individual differences among patients and multidrug resistance, the
   application of these drugs in OSCC treatment is limited, often leading
   to chemotherapy failure [147]^52^,[148]^53. Therefore, to guide
   personalized treatment more effectively, we compared the IC[50] values
   of general anticancer drugs between the two subgroups. These results
   indicated that phenformin may be more suitable for the treatment of
   high-risk patients. Phenformin, an oxidative phosphorylation inhibitor,
   disrupts the energy metabolism of tumor cells and inhibits angiogenesis
   within tumors by inhibiting oxidative phosphorylation, effectively
   suppressing OSCC growth [149]^35^,[150]^53^-[151]^55. Conversely,
   motesanib (AMG-706) may be more suitable for treating low-risk patient
   subgroups. It is a potent multi-kinase inhibitor that can overcome
   resistance to doxorubicin and paclitaxel by inhibiting ABCB1 efflux
   action [152]^43^,[153]^56^-[154]^58. Therefore, motesanib is expected
   to serve as an adjuvant chemotherapy drug, enhancing the sensitivity of
   low-risk patients to chemotherapeutic drugs and increasing the
   effectiveness of treatment.

   This study has a few limitations. Firstly, the analyzed data were
   derived from public databases with limited sample sizes. In addition,
   experimental data were lacking. Therefore, future studies should
   explore the functions and processes of m7G methyltransferase-associated
   miRNAs in patients with OSCC. Future studies should also focus on
   elucidating the deep-seated mechanisms by which m7G
   methyltransferase-related miRNAs trigger the activation of immune
   checkpoint molecules to uncover potential targets for specific
   immunotherapy in OSCC.

   In conclusion, our results revealed the diagnostic score of m7G
   methyltransferase-related miRNAs. We also developed an m7G
   methyltransferase-related miRNA prognostic model associated with
   distinct clinical outcomes and immune characteristics in patients with
   OSCC. Our findings lay the foundation for developing new prognostic
   models for OSCC, with important clinical significance for personalized
   treatment strategies in patients with OSCC.

Supplementary Material

   Supplementary table.
   [155]jcav15p6022s1.pdf^ (89.4KB, pdf)

Acknowledgments