Abstract
Background
To identify the prognostic markers of oral squamous cell carcinoma (OSCC), the genetic heterogeneity of the pathological stages was investigated.
Methods
The data of 295 patients with primary OSCC obtained from the Cancer Genome Atlas were studied. The genetic prognostic landscape of the pathological stages was systematically analyzed by Cox regressions, Fisher’s exact tests, and Gene Ontology (GO) enrichment.
Results
Stage 4 patients had a poor prognosis in univariate and multivariate Cox models. Transforming growth factor-beta (TGF-β) pathway alterations were found more frequently in stage 4, whereas alterations in cell cycle pathways were significant in stages 1, 2, and 3. The differentially mutated genes were divided into three groups: risk genes of high stage, hazardless genes, and risk genes of low stage. The risk genes of low stage (RNF112, AKR7L, ZSCAN5C, and ZPBP) were independent prognostic factors with stage 4 and treatment modality in multivariate Cox regressions. Additionally, in genetic interaction analysis, NOMO1 and ZNF333 had a high co-occurrence in high stage, and WIZ had high co-occurrence in low stage. In GO enrichment, the prognostic genes were clustered at the functional term of RNA polymerase II transcription, and ZNF333 had an association with RNA transcription.
Conclusion
The genetic mutation type and ratio of tumor heterogeneity are different for each stage of OSCC, and stratification of OSCC patients with differential therapeutic efficacy is needed. Risk genes of both high and low stages must be identified in patients diagnosed with low-stage OSCC. Mutations in NOMO1, ZNF333, and WIZ should be considered as potential prognostic markers.
Supplementary Information
The online version contains supplementary material available at 10.1007/s12105-022-01516-8.
Keywords: Head and neck cancer, Oral squamous cell carcinoma, DNA, Neoplasm staging, Prognosis, Transforming growth factor beta
Introduction
Head and neck cancer is the seventh most common type of cancer worldwide [1]. Head and neck squamous cell carcinoma (HNSCC) at different anatomic sites has a different etiology, clinical features, genetic profile, and prognosis. Oral squamous cell carcinoma (OSCC) is the most common subtype of HNSCC [2]. Recent advances in next-generation sequencing of HNSCC samples have revealed that HNSCC development is driven by gene mutations [3, 4]. It is thought to develop through a multi-step carcinogenic process, including oncogene activation and loss-of-function mutations of tumor suppressor genes. Another genomic analysis revealed molecular OSCC subgroups that reflect etiological and prognostic correlation [5, 6]. However, genetic prognosis has not been studied by the OSCC stage defined by the American Joint Committee on Cancer (AJCC).
In recent years, accumulating evidence has pointed to the existence of intra-tumor heterogeneity [7]. Many human tumors are histopathologically diverse, containing regions demarcated by various degrees of differentiation, proliferation, vascularity, inflammation, or invasiveness [7]. Therefore, it is expected that cellular heterogeneity would be observed in each OSCC stage, which would affect the prognosis of the disease. Accordingly, the type and ratio of genetic mutation will be different for each stage. These findings will provide clues for the design of clinical trials for targeted therapies and stratification of OSCC patients with differential therapeutic efficacy.
In this study, the contribution of prognostic genetic mutations to the AJCC stage of OSCC at five levels of analysis was determined. First, the prognostic significance and hazard ratio (HR) of the OSCC stages and other major clinical variables were checked. Second, the mutations in genes of 10 well-known oncogenic signaling pathways in a previous study were mapped [8] and compared with the OSCC stages. Additionally, differentially mutated genes were identified in each OSCC stage. Third, the prognostic significance and hazard ratio of each gene were analyzed with major clinical variables to verify the prognostic effects of the significant differentially mutated genes. Fourth, genetic interaction analyses were conducted to examine co-occurring and mutually exclusive features of these prognostically significant genes. Lastly, Gene Ontology (GO) enrichment was performed to discover biological pathways closely associated with the OSCC prognosis. Collectively, our study systematically investigated the genetic prognosis of the pathological stages of OSCC.
Materials and Methods
Patients
The data of patients with OSCC who had underwent surgical resection were downloaded from the Cancer Genome Atlas (TCGA) portal [4]. Patients with tumors in the pharynx, hypopharynx, larynx, and oropharynx were excluded. Among the OSCC patients, cases involving the hard palate, upper gum, lower gum, and retromolar area were included as a tumor site subgroup of the maxilla and mandible. Among the 320 patients initially included in the OSCC clinical dataset, 15 patients were excluded because of a prior malignancy. And three patients who were positive for HPV (human papilloma virus) were excluded. Whole exome somatic mutation data and survival data were available for 295 of the remaining 302 patients in whom the mutation profile was analyzed. To determine the accurate pathological stage defined by AJCC, two pathologists analyzed the data of 295 OSCC patients independently. This study was performed in the Genomic Data Commons (GDC) Data Use Agreement and the study-specific Data Use Certification Agreement available in the database of Genotypes and Phenotypes (dbGaP).
Somatic Mutation Data
Whole exome somatic mutation data of the TCGA cohort released on October 12, 2012 were downloaded from the TCGA data portal [4]. The detected somatic variants of all the 295 patients were downloaded in the Mutation Annotation Format (MAF). Whole exome sequencing was performed using the Illumina Genome Analyzer IIX platform. Detailed information on the sequencing, quality control, raw data processing, and validation procedure is provided elsewhere [4, 9]. After downloading the data, the somatic mutation profile was analyzed for each tumor. Nonsense mutations, frameshift indels, and splice-site mutations were considered as loss-of-function mutations.
Statistical Analysis
Univariate and multivariate Cox regression analyses were performed to estimate the HRs and 95% confidence intervals (95% CIs) for overall survival after adjusting for all the major clinical variables. Regression diagnostics were performed using Schoenfeld and df beta residuals to verify the underlying assumptions of the Cox models. Survival data were analyzed using the Kaplan–Meier method, and the overall survival was compared between the two groups using the log-rank test. Fisher’s exact tests were performed to investigate oncogenic signaling pathways and differentially mutated genes defined by stage groups, A 2 × 2 contingency table of frequencies of mutations was calculated for every gene from the stage groups followed by Fisher’s exact test to identify genes showing significant differences in their mutation frequencies. To examine genetic interaction, pair-wise Fisher’s exact tests were also conducted. For a pair of genes, the pattern of exclusiveness or co-occurrence was estimated by performing a Fisher’s exact test on a 2 × 2 contingency table containing frequencies of mutated and nonmutated samples. All statistical tests were two-sided, and the threshold for statistical significance was set at p < 0.05. All statistical analyses were performed using R software (version 4.1.3).
Functional Enrichment Analysis
GO enrichment analysis was performed to identify biological processes and pathways that were differentially expressed. The Database for Annotation Visualization and Integration Discovery (DAVID) network software (National Institutes of Health, Bethesda, Maryland) contains nearly all major public bioinformatics resources and is designed to facilitate high-throughput gene functional analysis. DAVID enriches the biological information of individual genes and can be used to annotate gene-related biological mechanisms using standardized gene terminology [10].
Visualization
Oncoplots were generated using Complex Heatmap Bioconductor package. Plots generated to display results from analysis modules, such as oncogenic signaling pathways, forest plots, genetic interactions, and genetic network plots, were generated using ggplot2 and basic R plot functions. The major visualization modules were used in maftools R package [11].
Results
Patients and Stage Prognosis
In the present study, 295 patients with primary OSCC of the TCGA were analyzed. The average age of patients was 61.8 years (range, 19–90 years), and there were 197 men and 98 women. The clinicopathological data of patients are presented in Table 1. Of the 295 cases, 137 patients had died, and the median survival time was 48.6 (95% CI, 33.1–85.7) months. The median follow-up period of the patients who were alive was 30 months. Univariate Cox proportional hazard regression was performed on all the major clinical variables. The analysis revealed an increased risk on comparison of stage 4 with stages 1, 2, and 3 (p = 0.02, 0.0005, and 0.04, respectively; Table 2). And there were significant differences in the values of age, surgical margin, extracapsular extension, lymphovascular invasion, and perineural invasion. In addition, multivariate Cox proportional hazard model was adjusted by all the major clinical variables. Stage 4, age, extracapsular extension, and treatment modality had statistically significant results (Table 2). Patients with stages 1, 2, and 3 had 0.17, 0.35, and 0.60 times less odds than those with stage 4 (p = 0.02, 0.0005, 0.04, respectively; Table 2). Therefore, all the 295 patients were classified into two groups of stages 1, 2, and 3 in one group and stage 4 in the other group. In a Kaplan–Meier curve, 10-year disease-free survival was significantly longer in stage 1, 2, and 3 groups than in stage 4 patients (p < 0.0001, Supplementary Fig. S1).
Table 1.
Patients’ clinicopathological data
| AJCC stage | Stage 1 | Stage 2 | Stage 3 | Stage 4 | Total |
|---|---|---|---|---|---|
| Variable | No. (%) | No. (%) | No. (%) | No. (%) | No. (%) |
| No. of patients | 16 (5.4) | 53 (18.0) | 59 (20.0) | 167 (56.6) | 295 |
| Age (yrs.)a | 61 (24–87) | 67 (34–87) | 62 (24–87) | 60 (19–90) | |
| Sex | |||||
| Male | 8 (50.0) | 32 (60.4) | 42 (71.2) | 115 (68.9) | 197 (66.8) |
| Female | 8 (50.0) | 21 (39.6) | 17 (28.8) | 52 (31.1) | 98 (33.2) |
| Race | |||||
| White | 15 (93.8) | 48 (90.6) | 50 (84.7) | 145 (86.8) | 258 (87.5) |
| Nonwhite | 0 | 3 (5.7) | 7 (11.9) | 18 (10.8) | 28 (9.5) |
| Unknown | 1 (6.2) | 2 (3.7) | 2 (3.4) | 4 (2.4) | 9 (3.0) |
| Primary tumor site | |||||
| Tongue | 12 (75.0) | 22 (41.5) | 30 (50.8) | 57 (34.1) | 121 (41.0) |
| Floor of mouth | 1 (6.3) | 8 (15.1) | 9 (15.2) | 40 (24.0) | 58 (19.7) |
| Buccal mucosa | 0 | 4 (7.5) | 8 (13.6) | 10 (6.0) | 22 (7.5) |
| Mandible and maxilla | 0 | 3 (5.7) | 3 (5.1) | 17 (10.2) | 23 (7.8) |
| Lip | 2 (12.5) | 0 | 0 | 1 (0.6) | 3 (1.0) |
| Mouth, unknown | 1 (6.2) | 16 (30.2) | 9 (15.3) | 42 (25.1) | 68 (23.0) |
| Surgical margin | |||||
| Negative | 15 (93.7) | 47 (88.7) | 53 (89.8) | 135 (80.8) | 250 (84.7) |
| Positive | 1 (6.3) | 4 (7.5) | 4 (6.8) | 24 (14.4) | 33 (11.2) |
| Unknown | 0 | 2 (3.8) | 2 (3.4) | 8 (4.8) | 12 (4.1) |
| Extracapsular extension | |||||
| Absence | 11 (68.8) | 33 (62.3) | 41 (69.5) | 78 (46.7) | 163 (55.3) |
| Presence | 0 | 0 | 6 (10.2) | 54 (32.3) | 60 (20.3) |
| Unknown | 5 (31.2) | 20 (37.7) | 12 (20.3) | 35 (21.0) | 72 (24.4) |
| Lymphovascular invasion | |||||
| Absence | 10 (62.5) | 30 (56.6) | 33 (56.0) | 78 (46.7) | 151 (51.2) |
| Presence | 2 (12.5) | 3 (5.7) | 13 (22.0) | 50 (30.0) | 68 (23.0) |
| Unknown | 4(25.0) | 20 (37.7) | 13 (22.0) | 39 (23.3) | 76 (25.8) |
| Perineural invasion | |||||
| Absence | 6 (37.5) | 25 (47.2) | 22 (37.3) | 48 (28.7) | 101 (34.2) |
| Presence | 3 (18.8) | 12 (22.6) | 27 (45.8) | 89 (53.3) | 131 (44.4) |
| Unknown | 7 (43.7) | 16 (30.2) | 10 (16.9) | 30 (18.0) | 63 (21.4) |
| Treatment modality | |||||
| Surgery | 9 (56.3) | 28 (52.8) | 20 (33.9) | 31 (18.6) | 88 (29.8) |
| Surgery + RT | 6 (37.5) | 13 (24.5) | 24 (40.7) | 52 (31.1) | 95 (32.2) |
| Surgery + CCRT | 0 | 4 (7.6) | 9 (15.2) | 69 (41.3) | 82 (27.8) |
| Unknown | 1 (6.2) | 8 (15.1) | 6 (10.2) | 15 (9.0) | 30 (10.2) |
AJCC American Joint Committee on Cancer; RT radiotherapy; CCRT concurrent chemotherapy and radiotherapy
amedian (range) value
Table 2.
Univariate and multivariate Cox regression models of all the major clinicopathological variables in all oral squamous cell carcinoma cases (N = 295)
| Variable | Univariate analysis | Multivariate analysisa | ||||
|---|---|---|---|---|---|---|
| HR | 95%CI | P | HR | 95%CI | P | |
| AJCC stage | ||||||
| Stage 4 vs stage 1 | 0.19 | 0.05–0.75 | 0.02* | 0.17 | 0.04-0.73 | 0.02* |
| Stage 4 vs stage 2 | 0.40 | 0.24–0.67 | 0.0005* | 0.35 | 0.20-0.63 | 0.0005* |
| Stage 4 vs stage 3 | 0.63 | 0.40–0.99 | 0.04* | 0.60 | 0.37-0.98 | 0.04* |
| Age | 1.02 | 1.01–1.03 | 0.01* | 1.02 | 1.01-1.04 | 0.008* |
| Sex | 1.16 | 0.82–1.64 | 0.40 | 0.98 | 0.66-1.45 | 0.92 |
| Race | ||||||
| White vs nonwhite | 1.45 | 0.81–2.57 | 0.21 | 1.52 | 0.82–2.83 | 0.19 |
| White vs unknown | 0.86 | 0.32–2.32 | 0.76 | 0.59 | 0.20–1.73 | 0.34 |
| Tumor site | ||||||
| Tongue vs other sitesb | 1.15 | 0.77–1.72 | 0.48 | 0.90 | 0.58–1.41 | 0.66 |
| Tongue vs unknown | 1.17 | 0.77–1.79 | 0.47 | 0.78 | 0.49–1.26 | 0.32 |
| Surgical margin | ||||||
| Negative vs positive | 1.69 | 1.08–2.66 | 0.02* | 1.55 | 0.95–2.55 | 0.08 |
| Negative vs unknown | 0.86 | 0.32–2.33 | 0.76 | 0.87 | 0.31–2.48 | 0.79 |
| Extracapsular extension | ||||||
| Absence vs presence | 3.10 | 2.07–4.66 | <0.00001* | 2.67 | 1.66–4.31 | 0.00005* |
| Absence vs unknown | 1.32 | 0.87–2.00 | 0.19 | 1.15 | 0.72–1.85 | 0.56 |
| Lymphovascular invasion | ||||||
| Absence vs presence | 1.73 | 1.15–2.61 | 0.009* | 1.37 | 0.87–2.17 | 0.17 |
| Absence vs unknown | 1.13 | 0.74–1.72 | 0.58 | 0.85 | 0.45–1.61 | 0.62 |
| Perineural invasion | ||||||
| Absence vs presence | 1.69 | 1.11–2.55 | 0.01* | 1.32 | 0.84–2.08 | 0.23 |
| Absence vs unknown | 1.33 | 0.80–2.19 | 0.27 | 1.38 | 0.65–2.95 | 0.40 |
| Treatment modality | ||||||
| Surgery vs surgery + RT | 0.66 | 0.41–1.04 | 0.07 | 0.46 | 0.28–0.76 | 0.002* |
| Surgery vs surgery + CCRT | 1.06 | 0.68–1.66 | 0.78 | 0.43 | 0.24–0.75 | 0.003* |
| Surgery vs unknown | 1.66 | 1.00–2.77 | 0.05 | 1.44 | 0.80–2.59 | 0.22 |
AJCC American Joint Committee on Cancer; RT radiotherapy; CCRT, concurrent chemotherapy and radiotherapy; HR hazard ratio; CI confidence interval
*p < 0.05
aThe statistical significance of fitted model of Cox proportional hazard ratio was calculated by Likelihood ratio test. (p = 0.000000003)
bOther tumor sites included floor of mouth, buccal mucosa, mandible, maxilla, and lip
Stage Comparison Analysis
There were two groups comprising 128 patients in stages 1, 2, and 3 and 167 patients in stage 4. According to tumor suppressor and oncogenic genes of 10 well-known signaling pathways [8], the two groups were analyzed by Fisher’s exact tests. Alterations in the cell cycle pathway were more frequently found in the lower stages of OSCC (p = 0.0003; Fig. 1a; Supplementary Table S1). Stage 4 had more alterations in the transforming growth factor-beta (TGF-β) pathway (p = 0.02; Fig. 1b; Supplementary Table S1). Figure 1b shows that mutations of tumor suppressor genes of the TGF-β pathway were found more frequently in stage 4 of OSCC. The statistics of comparing stages 1, 2, and 3 with 4 revealed 71 genes that were differentially mutated (p < 0.01; Fig. 1c). Among them, 27 genes (CNGB1, GDA, VCX3A, FCGR3A, CSTF3, GATA2, GBP5, GPR160, IREB2, RASL12, SCIN, USP6NL, ZNF670, AKAP7, TUBA3D, MAGEA3, FANCI, HOXD3, KRI1, NTRK1, TNFRSF6B, CDK5R2, LGMN, MECOM, INTS9, STRA6, WDR46) were significantly enriched in stage 4, and the other 44 genes were significantly enriched in stages 1, 2, and 3.
Fig. 1.
Oncogenic signaling pathway and stage comparison analysis a Tumor suppressor and oncogenic genes of Cell Cycle pathway were found more frequently in low stage. b High stage had more alterations of tumor suppressor genes of TGF-beta pathway than those of low stage. c Differentially mutated genes compared stage 1, 2, and 3 with stage 4 displayed as a forest plot (n = 128 and 167, respectively; Bar, 95% confidence interval of odds ratio;***p < 0.001; **p < 0.01)
Gene Prognosis Analysis
A total of 476 differentially mutated genes below the p-value of 0.05 were identified in the previous stage comparison analysis. Among the 472 genes, 37 genes had statistically significant (p < 0.05; Table 3) relevance to OSCC prognosis. Among them, 13 genes (DDOST, WIZ, LRRC74A, MTPAP, AP5M1, RAD17, KCNG4, UBASH3B, LRRFIP1, ARHGEF39, GRXCR1, ADAM32, PLEKHA4) were enriched in stages 1, 2, and 3 and HRs were less than 1.0. However, 20 genes (MAP2K5, OTX2, ZFAND5, ADGRG1, SLC35G5, CPNE1, IVL, SLC5A12, NOMO1, VRTN, CDC27, ZNF266, FOXC1, DPH1, INTS9, ANAPC2, FOXK1, ZNF333, HOXD3, SIAH2) were enriched in stage 4 and had HRs more than 1.0. The remaining four genes had HRs more than 1.0 and were enriched in the lower stages. These four risk genes were mutated in stages 2 and 3 at a high rate (RNF112, 71.4%; ZPBP, 85.7%; AKR7L, 88.9%; ZSCAN5C, 88.9%). Oncoplots of the 37 genes were drawn in all 295 OSCC patients as is shown in Fig. 2. The 37 genes were adjusted by multivariate Cox model with all the major clinical variables. The risk genes of lower stages (RNF112, AKR7L, ZSCAN5C, and ZPBP; Table 4) had statistical significance (p = 0.00004, 0.002, 0.0004, and 0.0009, respectively) with stage 4, age, extracapsular extension, and treatment modality. The mutant-type cases of RNF112, AKR7L, ZSCAN5C, and ZPBP genes were 8.11, 4.48, 4.27, and 4.85 times worse prognosis than the wild-type patients, respectively.
Table 3.
Significant genes of stage comparison and gene prognosis analyses in all oral squamous cell carcinoma cases (p < 0.05; N = 295)
| Hugo Symbol | Stage1,2,3 | Stage 4 | Stage Comparea | Gene Prognosisb | Hazard Ratio |
|---|---|---|---|---|---|
| No. | No. | p valuec | p valuec | ||
| Risk genes in high stage | |||||
| MAP2K5 | 0 | 7 | 0.020 | 0.001* | 3.55 |
| OTX2 | 0 | 6 | 0.038 | 0.005* | 3.36 |
| ZFAND5 | 0 | 6 | 0.038 | 0.009* | 2.83 |
| ADGRG1 | 0 | 6 | 0.038 | 0.015 | 2.90 |
| SLC35G5 | 56 | 97 | 0.019 | 0.018 | 1.51 |
| CPNE1 | 2 | 12 | 0.027 | 0.018 | 2.14 |
| IVL | 0 | 6 | 0.038 | 0.021 | 3.06 |
| SLC5A12 | 5 | 17 | 0.046 | 0.024 | 1.84 |
| NOMO1 | 2 | 12 | 0.027 | 0.029 | 2.03 |
| VRTN | 2 | 12 | 0.027 | 0.030 | 1.96 |
| CDC27 | 9 | 25 | 0.042 | 0.032 | 1.70 |
| ZNF266 | 0 | 8 | 0.011 | 0.038 | 2.33 |
| FOXC1 | 3 | 16 | 0.015 | 0.042 | 1.89 |
| DPH1 | 1 | 10 | 0.026 | 0.042 | 2.07 |
| INTS9 | 2 | 14 | 0.010 | 0.042 | 1.83 |
| ANAPC2 | 1 | 10 | 0.026 | 0.043 | 2.16 |
| FOXK1 | 0 | 7 | 0.020 | 0.043 | 2.45 |
| ZNF333 | 2 | 11 | 0.045 | 0.044 | 1.92 |
| HOXD3 | 1 | 12 | 0.008* | 0.045 | 1.97 |
| SIAH2 | 0 | 7 | 0.020 | 0.046 | 2.42 |
| Risk genes in low stage | |||||
| RNF112 | 6 | 1 | 0.045 | 0.004* | 3.49 |
| ZPBP | 6 | 1 | 0.045 | 0.011 | 2.80 |
| AKR7L | 8 | 1 | 0.012 | 0.016 | 2.65 |
| ZSCAN5C | 8 | 2 | 0.023 | 0.023 | 2.24 |
| Hazardless genes in low stage | |||||
| DDOST | 6 | 1 | 0.045 | 0.006* | < 0.01 |
| WIZ | 10 | 2 | 0.006* | 0.018 | 0.22 |
| LRRC74A | 7 | 2 | 0.043 | 0.020 | 0.14 |
| MTPAP | 7 | 1 | 0.023 | 0.021 | 0.13 |
| AP5M1 | 7 | 2 | 0.043 | 0.026 | < 0.01 |
| RAD17 | 11 | 2 | 0.003* | 0.030 | 0.35 |
| KCNG4 | 11 | 0 | < 0.001* | 0.030 | 0.24 |
| UBASH3B | 6 | 0 | 0.006* | 0.032 | 0.15 |
| LRRFIP1 | 11 | 5 | 0.041 | 0.038 | 0.25 |
| ARHGEF39 | 11 | 2 | 0.003* | 0.041 | 0.26 |
| GRXCR1 | 7 | 0 | 0.003* | 0.045 | 0.17 |
| ADAM32 | 7 | 2 | 0.043 | 0.046 | 0.27 |
| PLEKHA4 | 11 | 4 | 0.029 | 0.049 | 0.33 |
aFisher’s exact test
bUnivariate Cox proportional hazard ratio
cp < 0.05
*p < 0.01
Fig. 2.
Oncoplots of the 37 prognostic significant genes in stage 1, 2, 3, and 4 patients (n = 128 and 167, respectively)
Table 4.
Multivariate Cox regression models of prognostic significant genes in oral squamous cell carcinoma (N = 295)
| RNF112a | AKR7La | ZSCAN5Ca | ZPBPa | |||||
|---|---|---|---|---|---|---|---|---|
| Variable | HR | P | HR | P | HR | P | HR | P |
| Gene | ||||||||
| Wild type vs mutant type | 8.11 | 0.00004* | 4.48 | 0.002* | 4.27 | 0.0004* | 4.85 | 0.0009* |
| AJCC stage | ||||||||
| Stage 1 vs 4 | 7.37 | 0.008* | 5.62 | 0.02* | 6.10 | 0.02* | 5.63 | 0.02* |
| Stage 2 vs 4 | 3.27 | 0.0001* | 3.17 | 0.0002* | 3.07 | 0.0002* | 3.06 | 0.0002* |
| Stage 3 vs 4 | 1.78 | 0.02* | 1.83 | 0.02* | 2.04 | 0.008* | 1.88 | 0.02* |
| Age | 1.02 | 0.01* | 1.03 | 0.005* | 1.03 | 0.007* | 1.03 | 0.006* |
| Sex | 1.00 | 0.99 | 1.01 | 0.96 | 0.95 | 0.82 | 0.98 | 0.93 |
| Race | ||||||||
| White vs nonwhite | 1.52 | 0.19 | 1.63 | 0.12 | 1.58 | 0.15 | 1.46 | 0.23 |
| White vs unknown | 0.62 | 0.39 | 0.64 | 0.42 | 0.60 | 0.36 | 0.62 | 0.39 |
| Tumor site | ||||||||
| Tongue vs other sitesb | 0.88 | 0.56 | 0.90 | 0.65 | 0.91 | 0.68 | 0.87 | 0.53 |
| Tongue vs unknown | 0.72 | 0.18 | 0.79 | 0.35 | 0.84 | 0.48 | 0.71 | 0.17 |
| Surgical margin | ||||||||
| Negative vs positive | 1.61 | 0.06 | 1.58 | 0.07 | 1.54 | 0.09 | 1.64 | 0.05 |
| Negative vs unknown | 0.86 | 0.78 | 0.89 | 0.82 | 0.80 | 0.68 | 0.89 | 0.82 |
| Extracapsular extension | ||||||||
| Absence vs presence | 2.75 | 0.00004* | 2.78 | 0.00003* | 2.68 | 0.00006* | 2.75 | 0.00004* |
| Absence vs unknown | 1.26 | 0.35 | 1.18 | 0.49 | 1.08 | 0.75 | 1.20 | 0.45 |
| Lymphovascular invasion | ||||||||
| Absence vs presence | 1.46 | 0.11 | 1.45 | 0.12 | 1.44 | 0.12 | 1.44 | 0.13 |
| Absence vs unknown | 0.91 | 0.76 | 0.94 | 0.85 | 0.85 | 0.61 | 0.93 | 0.82 |
| Perineural invasion | ||||||||
| Absence vs presence | 1.23 | 0.37 | 1.19 | 0.45 | 1.30 | 0.26 | 1.32 | 0.23 |
| Absence vs unknown | 1.29 | 0.51 | 1.25 | 0.57 | 1.45 | 0.33 | 1.34 | 0.44 |
| Treatment | ||||||||
| Surgery vs surgery + RT | 0.42 | 0.0008* | 0.51 | 0.009* | 0.43 | 0.001* | 0.51 | 0.009* |
| Surgery vs surgery + CCRT | 0.40 | 0.002* | 0.47 | 0.01* | 0.40 | 0.002* | 0.46 | 0.01* |
| Surgery vs unknown | 1.33 | 0.35 | 1.59 | 0.13 | 1.35 | 0.32 | 1.58 | 0.14 |
AJCC American Joint Committee on Cancer; RT radiotherapy; CCRT concurrent chemotherapy and radiotherapy; HR hazard ratio
*p < 0.05
aThe statistical significance of fitted model of Cox proportional hazard ratio was calculated by Likelihood ratio test. (p = 0.00000000007, 0.0000000003, 0.0000000001, and 0.0000000002, respectively)
bOther tumor sites included floor of mouth, buccal mucosa, mandible, maxilla, and lip
Genetic Interaction Analysis
Pair-wise Fisher’s exact tests were performed in all 295 cases to examine co-occurring and mutually exclusive features of the 37 prognostic significant genes. By these analyses, 57 co-occurring and one mutually exclusive feature were discovered (p < 0.05; Fig. 3a). In a mutually exclusive manner, ARHGEF39 had HR less than 1.0 and SLC35G5 had HR more than 1.0. Fifty-seven pair-wise co-occurrences arose significantly among all the 37 genes except among four genes (SLC35G5, IVL, and ANAPC2). To clarify the co-occurring features, a genetic interaction network was drawn of these 34 genes (Fig. 3b). According to this genetic network, risk genes of high stage and hazardless genes of low stage were mutated in a separate and aggregated pattern. Moreover, RNF112, a risk gene found in the low stages of OSCC interfered between high- and low-risk mutated genes.
Fig. 3.
Genetic interaction analysis a Co-occurring and mutually exclusive gene pairs of the 37 prognostic significant genes displayed as a triangular matrix (N = 295). b A genetic network of the co-occurring interactions was presented by stages of oral squamous cell carcinoma
Biological Pathways and Drug Interactions
To discover biological pathways closely associated with OSCC prognosis, the online tool DAVID was used to perform GO enrichment analysis of the significant prognostic genes. The genes were clustered according to the functional terms “regulation of transcription from RNA polymerase II promoter” (p = 0.002), “RNA polymerase II core promoter proximal region sequence-specific DNA binding” (p = 0.004), “RNA polymerase II transcription factor activity, sequence-specific DNA binding” (p = 0.001), and “nucleus” (p = 0.027; Table 5). In addition, the genes were checked for known drug–gene interactions and druggable categories compiled from Drug Gene Interaction Database (Supplementary Fig. S2). Of the 16 genes, eight risk genes of high stage (MAP2K5, SLC35G5, ADGRG1, CPNE1, CDC27, ANAPC2, FOXC1, SIAH2, SLC5A12), six hazardless genes of low stage (DDOST, WIZ, RAD17, KCNG4, UBASH3B, ADAM32), and two risk genes of the lower stages (RNF112, AKR7L) had available therapies.
Table 5.
Gene ontology enrichment analysis of the prognostic significant genes
| Category | Term | No | P | Genes | Fold Enrichment |
|---|---|---|---|---|---|
| GOTERM BP_DIRECT | GO:0006357 ~ regulation of transcription from RNA polymerase II promoter | 10 | 0.002* | WIZ, LRRFIP1, ZSCAN5C, FOXC1, FOXK1, OTX2, ZNF333, VRTN, ZNF266, HOXD3 | 3.30 |
| GOTERM CC_DIRECT | GO:0005634 ~ nucleus | 17 | 0.027* | WIZ, UBASH3B, RAD17, LRRFIP1, ZSCAN5C, ZPBP, RNF112, FOXC1, FOXK1, CDC27, CPNE1, OTX2, INTS9, ZNF333, MAP2K5, ZNF266, HOXD3 | 1.62 |
| GOTERM MF_DIRECT | GO:0000978 ~ RNA polymerase II core promoter proximal region sequence-specific DNA binding | 8 | 0.004* |
WIZ, LRRFIP1, ZSCAN5C, FOXC1, FOXK1, OTX2, ZNF333, HOXD3 |
3.68 |
| GOTERM MF_DIRECT | GO:0000981 ~ RNA polymerase II transcription factor activity, sequence-specific DNA binding | 9 | 0.001* |
WIZ, LRRFIP1, ZSCAN5C, FOXC1, FOXK1, OTX2, ZNF333, ZNF266, HOXD3 |
3.90 |
GO gene ontology
*p < 0.05
Discussion
In this study, stage 4 OSCC patients had a higher HR than stages 1, 2, and 3 in univariate and multivariate Cox regressions (Table 2). Thus, we compared oncogenic signaling pathways and differentially mutated genes of stages 1, 2, and 3 with those of stage 4. Moreover, each gene was analyzed by univariate Cox model to verify the prognostic effects of the significant differentially mutated genes. According to the significance and HR, there were three groups (Table 3); risk genes of stage 4, hazardless genes, and risk genes of stages 1, 2, and 3. Multivariate Cox analysis was performed adjusted by all the major clinical variables to confirm independent prognostic markers. As a result, there were four stage-independent risk genes (RNF112, AKR7L, ZSCAN5C, and ZPBP; Table 4) of low stage.
In the stage comparison analysis, stage 4 had more alterations in the genes of the TGF-β pathway than those of the lower stages (Fig. 1b). Mutations of tumor suppressor genes of TGF-β pathway were found more frequently in stage 4. This is consistent with the findings of a previous report showing that TGF-β pathways had a significantly negative impact on disease-free survival [12]. While exposed to defective TGF-β signaling, tumor cells acquire resistance to TGF-β-induced growth inhibition and undergo epithelial–mesenchymal transition [13, 14]. The lower stages of OSCC had more alterations in the cell cycle pathways than the high stage (Fig. 1a). Mutations of the cell cycle pathway inhibit OSCC proliferation by causing cell cycle arrest [15]. Additionally, of differentially mutated genes compared with stage 4, KCNG4 and TKT were significantly mutated in stage 1, 2, and 3 (p < 0.0001; Fig. 1c). KCNG4 (potassium voltage-gated channel modifier subfamily G member 4) is associated with analgesia [16] and multiple sclerosis [17]. TKT had been reported to be involved in the malignant progression of HNSCC [18]; however, there is very limited information on TKT in OSCC.
In the gene prognosis analysis, there were three groups of risk genes; risk genes of stage 4, hazardless genes and risk genes of stages 1, 2, and 3. MAP2K5 and OTX2 were significant risk genes found in stage 4 (p < 0.01) and had high HR (HR = 3.55 and 3.36, respectively; Table 3). MAP2K5 has been reported to be related to the prognosis of breast [19] and gastric [20] cancers. OTX2 is a potent oncogene and a predictive factor of medulloblastoma [21, 22]. On the other hand, the mutation in DDOST was enriched in the lower stages (p = 0.006; HR < 0.01; Table 3). This is in agreement with a report that the mutation in DDOST served as a susceptibility variant in esophageal squamous cell carcinoma [23]. However, the prognostic effects of MAP2K5, OTX2, and DDOST have not been studied in HNSCC. In the results of multivariate Cox proportional hazard model, there were four stage-independent risk genes (RNF112, AKR7L, ZSCAN5C, and ZPBP; Table 4) of low stage. These four risk genes were mutated in stages 2 and 3 at high rates. However, since they have a risk similar to that of stage 4 mutations, the multivariate Cox models showed significant results (Table 4). Ring finger protein 112 (RNF112) is a member of the RING finger protein family and plays an important role in neuronal differentiation [24]. Aldoceto-reducing enzyme (AKR) family has been reported as a prognostic biomarker in HNSCC [25, 26]. Zinc finger and SCAN (ZSCAN) family also has important roles in cancer progression [27]. Further studies on AKR7L and ZSCAN5C are needed in OSCC. ZPBP is associated with polymerization of extracellular proteins that is dysregulated in cancer [28].
In the genetic interaction analysis, co-occurring and mutually exclusive features of the 37 prognostically significant genes were examined. According to the genetic network (Fig. 3), risk gene mutations of high stage and hazardless gene mutations of low stage showed a separate and aggregated pattern. The reason for this is that the type and ratio of gene mutations in tumor heterogeneity [7] are different for each stage of OSCC. Among the risk genes of stage 4, NOMO1 and ZNF333 had the highest co-occurrences with other genes. NOMO and transmembrane protein 147 (TMEM147) bind to different Nicalin domains, which form a complex to antagonize Nodal/TGFβ signaling at a more downstream step [29]. Nodal/TGFβ signaling pathways promote tumorigenesis; however, as the Nodal signaling antagonist, NOMO mutation cannot suppress tumorigenesis [29]. ZNF333 is associated with the regulation of transcription via the RNA polymerase II promoter (Table 5). RNA polymerase II transcription factor activity, sequence-specific DNA binding are biological pathways closely associated with OSCC prognosis. WIZ had the highest co-occurrence with other hazardless genes of stages 1, 2, and 3. WIZ plays an important role during cell replication and genomic stability [30]. Thus, mutations of the WIZ gene can have a better prognosis.
The present study has several limitations from the point of view of bioinformatics. Although we found the genes significantly related to the OSCC prognosis, the mechanism underlying the effects remains unknown. To improve our understanding regarding the prognostically significant genes, we reviewed the role of the genes in different diseases and investigated the mechanisms underlying their effects. In addition, there is a possibility of overfitting by analyzing in one patient dataset. It is recommended to confirm the results obtained using other test datasets. However, the results of the present study can be reliable because they were analyzed using a large group of patients with OSCC.
Conclusion
There were high TGF-β pathway alterations in stage 4, while cell cycle pathways showed significant results in stages 1, 2, and 3. TGF-β mutation caused resistance to TGF-β-induced growth inhibition and led to epithelial–mesenchymal transition [13, 14], resulting in poor prognosis. Cell cycle mutations enabled relatively good prognosis by reducing proliferation [15]. The prognosis of differentially mutated genes in stages 1, 2, and 3 compared to stage 4 identified three groups: risk genes of high stage, hazardless genes, and risk genes of low stage. Among them, risk genes RNF112, AKR7L, ZSCAN5C, and ZPBP have pathogenic mutations independent of the stage. Thus, they need to be investigated carefully when diagnosing and treating patients in stages 1, 2, and 3. Additionally, Nomo1 and ZNF333 had a high co-occurrence in stage 4 and WIZ in stages 1, 2, and 3. Nomo1 is a Nodal/TGFβ antagonist [29], and ZNF333 is associated with RNA polymerase II transcription. WIZ has an important role in cell duplication [30]. We can conclude that the genetic mutation type and ratio of tumor cells heterogeneity are different for each stage of OSCC, and stratification of OSCC patients with differential therapeutic efficacy is needed.
Supplementary Information
Below is the link to the electronic supplementary material.
Authors Contributions
All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by SKK, JIL, and KK. The first draft of the manuscript was written by SKK and all authors commented on the previous versions of the manuscript. All authors read and approved the final manuscript.
Funding
This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education [2020R1A4A1019423 and 2022R1I1A1A01062894].
Data Availability
In this study, publicly archived datasets were investigated and analyzed. The data were downloaded from the Cancer Genome Atlas (TCGA) portal (https://portal.gdc.cancer.gov/).
Code Availability
All statistical analyses were performed using R software (version 4.1.3). The Database for Annotation Visualization and Integration Discovery (DAVID) network software was used to identify biological pathways. The visualization modules were used in maftools R package.
Declarations
Conflicts of interest
None declared.
Ethics Approval
This study was performed in the GDC Data Use Agreement and the study-specific Data Use Certification Agreement available in dbGaP.
Consent to Participate
We agreed not to attempt to reidentify any individual participant in any study represented by GDC data, for any purpose whatsoever. We agreed to have read and understand study-specific Data Use Agreements and to comply with any additional restrictions therein. We agreed to abide by the NIH Genomic Data Sharing Policy (GDS).
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer. 2015;136(5):E359–E386. doi: 10.1002/ijc.29210. [DOI] [PubMed] [Google Scholar]
- 2.Chaturvedi AK, Anderson WF, Lortet-Tieulent J, Curado MP, Ferlay J, Franceschi S, et al. Worldwide trends in incidence rates for oral cavity and oropharyngeal cancers. J Clin Oncol. 2013;31(36):4550–4559. doi: 10.1200/JCO.2013.50.3870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Stransky N, Egloff AM, Tward AD, Kostic AD, Cibulskis K, Sivachenko A, et al. The mutational landscape of head and neck squamous cell carcinoma. Science. 2011;333(6046):1157–1160. doi: 10.1126/science.1208130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cancer Genome Atlas N. Comprehensive genomic characterization of head and neck squamous cell carcinomas. Nature. 2015;517(7536):576–582. doi: 10.1038/nature14129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Su SC, Lin CW, Liu YF, Fan WL, Chen MK, Yu CP, et al. Exome sequencing of oral squamous cell carcinoma reveals molecular subgroups and novel therapeutic opportunities. Theranostics. 2017;7(5):1088–1099. doi: 10.7150/thno.18551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chamorro Petronacci CM, García García A, Padín Iruegas E, Rivas Mundiña B, Lorenzo Pouso AI, Pérez SM. Identification of Prognosis Associated microRNAs in HNSCC Subtypes Based on TCGA Dataset. Medicina. 2020 doi: 10.3390/medicina56100535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646–674. doi: 10.1016/j.cell.2011.02.013. [DOI] [PubMed] [Google Scholar]
- 8.Sanchez-Vega F, Mina M, Armenia J, Chatila WK, Luna A, La KC, et al. Oncogenic signaling pathways in the cancer genome atlas. Cell. 2018;173(2):321–337. doi: 10.1016/j.cell.2018.03.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.India Project Team of the International Cancer Genome C Mutational landscape of gingivo-buccal oral squamous cell carcinoma reveals new recurrently-mutated genes and molecular subgroups. Nat Commun. 2013 doi: 10.1038/ncomms3873. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Jiao X, Sherman BT, da Huang W, Stephens R, Baseler MW, Lane HC, et al. DAVID-WS: a stateful web service to facilitate gene/protein list analysis. Bioinformatics. 2012;28(13):1805–1806. doi: 10.1093/bioinformatics/bts251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28(11):1747–1756. doi: 10.1101/gr.239244.118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Fan WL, Yang LY, Hsieh JC, Lin TC, Lu MJ, Liao CT. Prognostic genetic biomarkers based on oncogenic signaling pathways for outcome prediction in patients with oral cavity squamous cell carcinoma. Cancers. 2021 doi: 10.3390/cancers13112709. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pang X, Tang YL, Liang XH. Transforming growth factor-β signaling in head and neck squamous cell carcinoma: Insights into cellular responses. Oncol Lett. 2018;16(4):4799–4806. doi: 10.3892/ol.2018.9319. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Oshimori N, Oristian D, Fuchs E. TGF-β promotes heterogeneity and drug resistance in squamous cell carcinoma. Cell. 2015;160(5):963–976. doi: 10.1016/j.cell.2015.01.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Li X, Ding R, Han Z, Ma Z, Wang Y. Targeting of cell cycle and let-7a/STAT3 pathway by niclosamide inhibits proliferation, migration and invasion in oral squamous cell carcinoma cells. Biomed Pharmacother. 2017;96:434–442. doi: 10.1016/j.biopha.2017.09.149. [DOI] [PubMed] [Google Scholar]
- 16.Lee MC, Nahorski MS, Hockley JRF, Lu VB, Ison G, Pattison LA, et al. Human Labor Pain Is Influenced by the Voltage-Gated Potassium Channel K(V)6.4 Subunit. Cell Rep. 2020 doi: 10.1016/j.celrep.2020.107941. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Vilariño-Güell C, Zimprich A, Martinelli-Boneschi F, Herculano B, Wang Z, Matesanz F, et al. Exome sequencing in multiple sclerosis families identifies 12 candidate genes and nominates biological pathways for the genesis of disease. PLoS Genet. 2019 doi: 10.1371/journal.pgen.1008180. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Tang YC, Hsiao JR, Jiang SS, Chang JY, Chu PY, Liu KJ, et al. c-MYC-directed NRF2 drives malignant progression of head and neck cancer via glucose-6-phosphate dehydrogenase and transketolase activation. Theranostics. 2021;11(11):5232–5247. doi: 10.7150/thno.53417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wang L, Ji X, Mao C, Yu R. BAY-885, a mitogen-activated protein kinase kinase 5 inhibitor, induces apoptosis by regulating the endoplasmic reticulum stress/Mcl-1/Bim pathway in breast cancer cells. Bioengineered. 2022;13(5):12888–12898. doi: 10.1080/21655979.2022.2078557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wang JR, Shen GN, Luo YH, Piao XJ, Shen M, Liu C, et al. The compound 2-(naphthalene-2-thio)-5,8-dimethoxy-1,4-naphthoquinone induces apoptosis via reactive oxygen species-regulated mitogen-activated protein kinase, protein kinase B, and signal transducer and activator of transcription 3 signaling in human gastric cancer cells. Drug Dev Res. 2018;79(6):295–306. doi: 10.1002/ddr.21442. [DOI] [PubMed] [Google Scholar]
- 21.Zagozewski J, Shahriary GM, Morrison LC, Saulnier O, Stromecki M, Fresnoza A, et al. An OTX2-PAX3 signaling axis regulates Group 3 medulloblastoma cell fate. Nat Commun. 2020;11(1):3627. doi: 10.1038/s41467-020-17357-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Figueira Muoio VM, Uno M, Oba-Shinjo S, da Silva R, Araújo Pereira BJ, Clara C, et al. OTX1 and OTX2 Genes in Medulloblastoma. World Neurosurg. 2019;127:e58–e64. doi: 10.1016/j.wneu.2019.02.013. [DOI] [PubMed] [Google Scholar]
- 23.Donner I, Katainen R, Tanskanen T, Kaasinen E, Aavikko M, Ovaska K, et al. Candidate susceptibility variants for esophageal squamous cell carcinoma. Genes Chromosomes Cancer. 2017;56(6):453–459. doi: 10.1002/gcc.22448. [DOI] [PubMed] [Google Scholar]
- 24.Lin DY, Huang CC, Hsieh YT, Lin HC, Pao PC, Tsou JH, et al. Analysis of the interaction between Zinc finger protein 179 (Znf179) and promyelocytic leukemia zinc finger (Plzf) J Biomed Sci. 2013;20(1):98. doi: 10.1186/1423-0127-20-98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Peraldo-Neia C, Ostano P, Mello-Grand M, Guana F, Gregnanin I, Boschi D, et al. AKR1C3 is a biomarker and druggable target for oropharyngeal tumors. Cell Oncol (Dordr) 2021;44(2):357–372. doi: 10.1007/s13402-020-00571-z. [DOI] [PubMed] [Google Scholar]
- 26.Liu J, Ban H, Liu Y, Ni J. The expression and significance of AKR1B10 in laryngeal squamous cell carcinoma. Sci Rep. 2021;11(1):18228. doi: 10.1038/s41598-021-97648-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Huang M, Chen Y, Han D, Lei Z, Chu X. Role of the zinc finger and SCAN domain-containing transcription factors in cancer. Am J Cancer Res. 2019;9(5):816–836. [PMC free article] [PubMed] [Google Scholar]
- 28.Drees L, Königsmann T, Jaspers MHJ, Pflanz R, Riedel D, Schuh R. Conserved function of the matriptase-prostasin proteolytic cascade during epithelial morphogenesis. PLoS Genet. 2019 doi: 10.1371/journal.pgen.1007882. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sun Y, Li T, Qian X. Biological Role of Nodal Modulator: A Comprehensive Review of the Last Two Decades. DNA Cell Biol. 2022;41(4):336–341. doi: 10.1089/dna.2021.0944. [DOI] [PubMed] [Google Scholar]
- 30.Srivastava M, Chen Z, Zhang H, Tang M, Wang C, Jung SY, et al. Replisome Dynamics and Their Functional Relevance upon DNA Damage through the PCNA Interactome. Cell Rep. 2018;25(13):3869–83.e4. doi: 10.1016/j.celrep.2018.11.099. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
In this study, publicly archived datasets were investigated and analyzed. The data were downloaded from the Cancer Genome Atlas (TCGA) portal (https://portal.gdc.cancer.gov/).
All statistical analyses were performed using R software (version 4.1.3). The Database for Annotation Visualization and Integration Discovery (DAVID) network software was used to identify biological pathways. The visualization modules were used in maftools R package.