Assessment of the Mutation Profile of Tonsillar Squamous Cell Carcinomas Using Targeted Next-Generation Sequencing

Data regarding driver mutation profiles in tonsillar squamous cell carcinomas (TSCCs) remain scarce, limiting the understanding of its pathogenesis and unexpected behavior in the updated staging system. We investigated the incidence of clinically relevant mutations and their contribution in the prognosis of the condition, and their association with human papillomavirus (HPV) infection and adjuvant therapy. We subjected 43 surgically resected TSCC samples to targeted next-generation sequencing, determined their HPV status using polymerase chain reaction, and performed The Cancer Genomic Atlas and Gene Set Enrichment analyses. Thirty-five TSCC samples (81.4%) showed at least one oncogenic/likely oncogenic mutation among twenty-nine cancer-related genes. The top five mutated genes were TP53 (46.5%), PIK3CA (25.6%), PTEN (18.6%), EGFR (16.3%), and SMAD4 (14.0%). The EGFR pathway was the most frequently affected (51.2%), followed by the p53 (48.8%), PI3K (39.5%), and RTK (34.9%) pathways. The gene set enrichment analysis confirmed that the genes involved in signal transduction, such as growth factor receptors and second messengers, EGFR tyrosine kinase inhibitors, and PI3K signaling pathways, were mostly related with TSCCs. TP53 mutation was an independent prognostic factor predicting worse overall survival in the adjuvant therapy group. RTK mutations were related to survival in all patients and in the HPV-positive group, but multivariate analyses showed no significance. In conclusion, oncogenic/likely oncogenic mutations were relatively high in TSCCs, and TP53 and RTK mutations may be candidate predictors for poor prognosis in the adjuvant therapy and HPV-positive groups, respectively, under the updated staging system.


Introduction
Squamous cell carcinomas (SCCs) of the tonsillar region (TSCCs) are the most prevailing oropharyngeal cancers, constituting 15-20% of all oropharyngeal malignancies and one of the subsites with the highest human papillomavirus (HPV) positivity rate [1,2]. Oropharyngeal cancers are associated with HPV infection, alcohol consumption, smoking, environmental pollution, and genetic risk factors [3]; however, awareness of HPV as a causative agent for these cancers is low [4], resulting in a public health burden [5]. Although the 5-year survival in low-stage cases is >90%, it decreases to <20% in advanced tonsil cancers [1,2,6]. Early detection of tonsil cancer is challenging [5]; the peculiar tonsillar anatomical features make the tumor proliferating beneath the surface mucosa unrecognizable, as do the several lymphatic vessel plexuses present around the tonsils. Thus, the clinical course of TSCC is more aggressive, with significant risk for higher stage and earlier dissemination at diagnosis, as compared to that for other head and neck cancers [7,8].
Combined surgery and radiotherapy/chemotherapy is a standard practice for local or regionally advanced tonsil cancers [9,10] because TSCC is relatively chemo/radiosensitive [7,8,11] and HPV is involved in a more favorable prognosis and response to radiochemotherapy, as compared to those observed in case of HPV-unrelated oropharyngeal cancers [2,10]. In this context, the eighth edition of the American Joint Committee on Cancer (AJCC) TNM classification in 2017 specified HPV-related oropharyngeal SCCs as a separate entity and downstaged this group as compared to the HPV-negative group [12]. However, treatment failure can develop unexpectedly in a loco-regionally controlled disease as therapy-refractory recurrence [13]; 13-25% of patients with HPV-positive oropharyngeal SCCs, including TSCC, develop local/distant recurrence and have a course of disease similar to those of HPV-negative oropharyngeal cancers [14], which may imply potential genetic involvement in tumor behavior and therapeutic drug response at the molecular level [15]. The molecular profiles of tonsil cancers and their clinical or pathological relation under the new eighth-edition AJCC staging system are poorly understood. Therefore, identifying molecular markers under the updated AJCC staging system that are diagnostic and predictive of clinical outcomes may be clinically significant so as to recommend further therapeutic options for potentially responsive patients, thereby improving patient prognosis.
Currently, molecular profiling of various cancers using next-generation sequencing (NGS) has been increasingly implemented in the oncology field for diagnostic and therapeutic applications. A broad range of somatic genetic alterations have been illustrated in head and neck SCCs (HNSCCs), where cancer-specific mutations, including small deletions or insertions and single-nucleotide polymorphisms, are known to influence driver genes during tumor development and progression [15][16][17]. However, despite the increasing prevalence and substantial public health impact of tonsil cancers [18], molecular data that may illustrate the unexpected behaviors or characteristic histological findings of HPV-positive and -negative TSCCs are limited. Owing to the emergence of personalized medicine in conjunction with molecular profiling, there is now a crucial need for detailed research on TSCCs.
Hence, we carried out this retrospective study to detect the frequency of pathogenic mutations in tonsil cancers, to find potential targets for treatment through an NGS cancer panel, and tested whether somatic mutation status or any cancer pathway are associated with risk factors or survival.

Patients and Tumor Samples
The study was based on formalin-fixed, paraffin-embedded (FFPE) tissues acquired from 43 patients with TSCC who underwent primary surgical resection with no prior treatment between 1997 and 2018, and whose complete medical records were available at the present institution. After curative surgery, adjuvant therapy was carried out. In total, 6 patients underwent post-operative radiotherapy, while 21 patients underwent radiochemotherapy following surgical resection. The remaining 16 patients underwent surgery alone. Radiation doses spanned from 5040 to 7200 cGy/36 fractions over eight weeks.
Hematoxylin-and-eosin-stained slides of all cases were examined, and diagnosis and histological characterization were performed by two expert pathologists (MJK and SJC) according to the 2017 World Health Organization classification of head and neck tumors [6]. Patients were re-staged according to the 8th edition of the AJCC TNM classification [12]. Clinical information was analyzed based on medical records and radiological results. Heavy smoking and heavy alcohol consumption were defined as >20 packs/year and >14 drinks/week, respectively [12].
The study protocol was permitted by the Institutional Review Board (IRB no. 2019-06-012) and conducted following the relevant guidelines and regulations (Declaration of Helsinki).

DNA Extraction
Tumor parts on the two 10 µm sections from every FFPE tissue block, on glass slides, were manually macrodissected from the unstained tissue sections to enrich for a tumor cell population of >50%. DNA was extracted and purified using the Ion AmpliSeq™ Direct FFPE DNA Kit (Thermo Fisher Scientific, Waltham, MA, USA) and QIAamp DSP DNA FFPE Tissue Kit (QIAGEN, Hilden, Germany), respectively, as previously described [19]. The yield of purified genomic DNA was measured utilizing a Qubit ® 2.0 Fluorometer and Qubit ® dsDNA HS Assay Kit (Thermo Fisher Scientific) prior to library preparation for sequencing.

Immunohistochemistry
Programmed death-ligand 1 (PD-L1) staining of 4 µm thick tissue sections was performed using the US Food and Drug Administration-approved PD-L1 22C3 pharmDx Kit (Dako North America Inc., Carpinteria, CA, USA) on the Dako Autostainer Link 48, as per the manufacturer's instructions. The slides were assigned depending on the percentage of positive cells divided by the number of fields, to calculate the average value for each individual case, defined at 200× magnification. The PD-L1 combined positive score was calculated as the ratio of PD-L1-positive cells (tumor or immune cells) to the total number of tumor cells × 100 and classified as positive (≥1) or negative (<1).

Library Preparation and Sequencing
In this study, an Ion Personal Genome Machine (PGM) Sequencer (Thermo Fisher Scientific) was used as the NGS platform. Library preparation for individual samples was accomplished using the Ion AmpliSeq™ Library Kit 2.0 (Thermo Fisher Scientific) and Ion AmpliSeq™ Cancer HotSpot Panel v2 (Thermo Fisher Scientific), as per the manufacturer's directions [19]. The targeted cancer panel sequences 2790 mutations in 50 oncogenes and tumor suppressor genes with known cancer correlations. Generally, 10 ng of genomic DNA from each sample was applied to prepare barcoded libraries, using Ion Xpress™ Barcode Adapters (Thermo Fisher Scientific). Libraries were pooled to a final concentration of 100 pmol/L using the Ion Library Universal Quantification Kit (Thermo Fisher Scientific), and emulsion polymerase chain reaction was performed utilizing the Ion Torrent™ OneTouch™ 2 System. Every pool was loaded onto an Ion 318v2 Chip (Thermo Fisher Scientific) for single-end sequence analysis with an Ion PGM Sequencer, while applying 500 flows (125 cycles) for 200-base-read sequencing.

Data Analyses, TCGA Data, and Gene Set Enrichment Analysis (GSEA)
Raw sequence reads were mapped against the human reference genome hg19 and cleaned prior to variant calling. Ion Torrent platform-specific pipeline software (Torrent Suite version 5.0.5) was operated throughout the variant calling steps. Annotation was carried out using the VEP and OncoKB™ (an expert-guided precision oncology knowledge base) annotator [21]. To detect confident putative somatic variants, annotated raw variants were filtered using the following criteria: (1) non-synonymous single-nucleotide variant or short insertion or deletion in coding regions; (2) coverage ≥50× and variant allele frequency ≥5%; (3) minor allele frequency <0.1% in gnomAD and 1000 Genomes Project; (4) detected as 'Oncogenic' or 'Likely Oncogenic' by the OncoKB™ annotator. The resulting list of variants was manually reviewed and visually confirmed using the Integrated Genomics Viewer (http://www.broadinstitute.org/igv/; accessed on 15 June 2020). R and MutationMapper (https://www.cbioportal.org/mutation_mapper, accessed on 10 January 2023) were used for visualization.
To compare the data with those of previous cancer genome studies on tonsil cancer and verify hotspot mutations, we downloaded the 'Head and Neck Squamous Cell Carcinoma from the Cancer Genome Atlas (TCGA, PanCancer Atlas)' (TCGA HNSC) (http:// cancergenome.nih.gov/abouttcga (accessed on 5 September 2022)) sequence variant dataset in mutation annotation format (MAF) from cBioPortal (https://www.cbioportal.org/) (accessed on 5 September 2022). We performed OncoKB™ annotation on the TCGA HNSC MAF file and filtering using the same criteria as described above for comparison under the same conditions as our data.
The 29 altered genes identified in this study were subjected to GSEA [22] using the canonical pathways of the Molecular Signatures Database (http://www.gsea-msigdb.org/ gsea/msigdb/human/annotate.jsp; accessed on 10 January 2023), including BIOCARTA, KEGG, PID, REACTOME, and WIKIPATHWAYS, which determine pathway correlations with major biomolecular and cellular procedures [23]. For genes that could be assigned to multiple pathways, we chose the signaling with the highest fraction of mapped genes. Additionally, knowledge of the functional classification of the major signaling molecules in the chosen pathway was used. A result was considered significant if the cut-off of the false discovery rate was ≤ 0.01 and p-value was < 0.05.

Statistical Analyses
Statistical analyses of the data for pairs of groups were conducted using the Fisher's exact test for categorical variables and Wilcoxon rank-sum test for continuous variables. Survival curves were produced using the Kaplan-Meier method, and a log-rank test was performed to calculate the statistical significance between groups. Overall survival (OS) was clarified as the interval from the first day of surgery until death. Disease-free survival (DFS) was described as the interval from the first day of surgery until tumor recurrence. The OS and DFS were analyzed until February 2019. Univariate and multivariate analyses applying the Cox proportional hazard regression model were used to measure the hazard ratios (HRs) and 95% confidence intervals (CIs) for specific variables involved in OS and DFS. The prognostic variables (p < 0.1) observed using univariate analysis by means of the log-rank test were additionally analyzed using multivariate analysis. A two-sided p < 0.05 was considered to indicate a statistically significant difference. Statistical analyses were carried out using Rex software (version 3.3.1, RexSoft Inc., Seoul, Republic of Korea).

Molecular Profiling in Overall TSCCs
The genetic landscape of all the cases is summarized in Figure 1A. On average, there were 405,114 reads (range 128,509-711,217) per case. An average of 96.4% of the reads (range 81.9-99.6%) per case was mapped to the intended targeted areas of the human genome. All regions had an average coverage of 1531× (range 163×-3417×). After the multiple filtering process described in the Materials and Methods section, we identified 120 variant calls of validated mutations for 29 genes in 35 cases (81.4%) as oncogenic/likely oncogenic mutations in the OncoKB™ database, including 98 (81.7%) missense point mutations (single-nucleotide variants) in 32 cases, 13 (10.8%) nonsense mutations in 9 cases, and 9 (7.5%) frameshift deletion mutations in 8 cases (Table S1). A total of 26 of the 43 cases (60.5%) harbored concurrent mutations in two or more genes.  The four mutation types have been distinguished using different colors. The right panel represents the comparison of mutation frequencies among our cases, overall head and neck cancers, and tonsil cancers retrieved from the TCGA data. Locations of TP53 (B) and PIK3CA (C) mutations found in tonsillar squamous cell carcinomas. A circle represents the presence of a single mutation case. Deep-red-and dark-blue/black-lined circles indicate cases in this study with and without HPV infection, respectively; others have been retrieved from the TCGA tonsil cancer and TCGA HNSC data. The line length depends on the number of mutations detected in that codon. The colored boxes are specific functional domains. Above the lollipops, the frequent variants have been annotated as the amino acid change at that particular site. HPV, human papillomavirus; TCGA, The Cancer Genomic Atlas.
The most repeated mutation in the HPV-positive tumors was in TP53 (37.1%, 13/35), followed by those in PIK3CA

Impact of Gene Mutations and Pathway Mutations on Clinicopathological Features
Next, we assessed the influence of important gene mutations and the primarily affected gene pathways in tonsil cancers on clinicopathological features (Table 3). Significantly more patients with TP53 mutations were heavy smokers (p = 0.033), HPV-negative (p = 0.017), node-negative (pN0; p = 0.038), and PD-L1-negative (p = 0.011), as compared to those without. CDKN2A mutations were considerably correlated with a history of heavy alcohol consumption (p = 0.045) and node-negative tumors (p = 0.014) (data not shown). FBXW7 mutations were correlated with a history of heavy smoking (p = 0.048) (data not shown). With the exceptions of the TP53, CDKN2A, and FBXW7 mutations, no other associations were detected between a single somatic mutation and the evaluated clinicopathological variables. We further examined whether specific somatic mutations might be related to the discordant cases between the p16 immunohistochemistry and HPV PCR results. We found 10 discordant cases, consisting of HPV-DNA-detected but p16-negative tumors (n = 8) and HPV-DNA-negative but p16-positive tumors (n = 2). Among the somatic mutations in the 29 genes detected in this study, the SMARCB1 mutation tended to be more frequently detected in the discordant cases than in the concordant ones (p = 0.020).

Impact of Gene Mutations and Pathway Mutations on Clinicopathological Features
Next, we assessed the influence of important gene mutations and the primarily affected gene pathways in tonsil cancers on clinicopathological features (Table 3). Significantly more patients with TP53 mutations were heavy smokers (p = 0.033), HPV-negative (p = 0.017), node-negative (pN0; p = 0.038), and PD-L1-negative (p = 0.011), as compared to those without. CDKN2A mutations were considerably correlated with a history of heavy alcohol consumption (p = 0.045) and node-negative tumors (p = 0.014) (data not shown). FBXW7 mutations were correlated with a history of heavy smoking (p = 0.048) (data not shown). With the exceptions of the TP53, CDKN2A, and FBXW7 mutations, no other associations were detected between a single somatic mutation and the evaluated clinicopathological variables. MT, mutant; WT, wild-type; HPV, human papillomavirus; AJCC, American Joint Committee on Cancer; EGFR, epidermal growth factor receptor; RTK, receptor tyrosine kinase; PD-L1, programmed death-ligand 1. * Statistically significant, at p < 0.05.P53 pathway mutations were associated with male sex (p = 0.048), heavy smoking history (p = 0.015), HPV negativity (p = 0.021), and node-negative tumors (p = 0.046), whereas those in the RTK pathway were strongly associated with advanced stages (p = 0.019). However, there were no meaningful correlations between the PI3K pathway mutations and clinicopathological features.

Prognostic Genetic Factors in Overall Tonsil Cancers
We analyzed the prognostic relationships of genetic alterations and clinicopathological features with OS and DFS in patients with TSCC (Table 4). With respect to OS ( Figure 5A), RTK pathway mutations (p = 0.007), older age (>60 years; p = 0.011), and advanced disease stage (p = 0.003) were associated with shorter patient survival. Factors with p < 0.1 in the univariate analysis were incorporated into the subsequent multivariate analysis. Older age (p = 0.020, HR = 3.565, 95% CI = 1.221-10.412) and advanced stage (p = 0.006, HR = 5.246, 95% CI = 1.596-17.235) were independent predictors of shorter OS in patients with TSCC; however, RTK pathway mutations did not reach statistical significance in the multivariate analysis. Kaplan-Meier survival analyses showed that patients with HPV-positive TSCCs had significantly better OS rates than those with HPV-negative TSCCs (mean 80 months vs. 25 months; p = 0.049); however, the difference was not statistically significant and showed only borderline significance in the univariate analysis carried out using the Cox proportional hazard regression model (p = 0.058). In contrast, patients with either RTK pathway mutations (p = 0.042) or advanced stage disease (p = 0.008) exhibited worse prognostic outcomes in terms of DFS ( Figure 5B). RTK pathway mutations failed to reach statistical significance in the multivariate analysis. Only advanced stage was an independent negative prognostic factor influencing DFS (p = 0.029, HR = 2.856, 95% CI = 1.110-7.345) in the multivariate analyses. tivariate analysis. Kaplan-Meier survival analyses showed that patients with HPV-positive TSCCs had significantly better OS rates than those with HPV-negative TSCCs (mean 80 months vs. 25 months; p = 0.049); however, the difference was not statistically significant and showed only borderline significance in the univariate analysis carried out using the Cox proportional hazard regression model (p = 0.058). Overall (E) and disease-free (F) survival by TP53 mutation status in patients who underwent curative surgery and adjuvant therapy. The TP53 mutation strongly correlates with decreased overall survival in patients following resection and adjuvant therapy. HPV, human papillomavirus; RTK, receptor tyrosine kinase. Overall (E) and disease-free (F) survival by TP53 mutation status in patients who underwent curative surgery and adjuvant therapy. The TP53 mutation strongly correlates with decreased overall survival in patients following resection and adjuvant therapy. HPV, human papillomavirus; RTK, receptor tyrosine kinase.

Prognostic Genetic Factors in Patients Who Received Adjuvant Therapy following Surgery
We next investigated the prognostic potential of genetic alterations in patients with TSCC who underwent adjuvant therapy following surgical resection (Table 6). TP53 mutations (p = 0.011), p53 pathway mutations (p = 0.029), and advanced stage (p = 0.017) were highly correlated with decreased OS in these patients ( Figure 5E). Multivariate analyses revealed that TP53 mutation (p = 0.022, HR = 4.348, 95% CI = 1.242-15.223) and higher stage (p = 0.024, HR = 4.856, 95% CI = 1.228-19.201) were independent prognostic factors predicting a decrease in OS in those who received adjuvant therapy.  With respect to DFS, TP53 mutations showed borderline statistical significance in affecting DFS in both the univariate (p = 0.075) and multivariate (p = 0.059) analyses ( Figure 5F). Advanced stage was the only independent poor prognostic factor for DFS in patients who underwent resection with adjuvant therapy (p = 0.039, HR = 3.075, 95% CI = 1.057-8.944).

Comparisons with TCGA Data
We compared our results with TCGA data for validation purposes. When we exclusively sorted tonsil cancer data from the TCGA database, we found only 38 cases of tonsil cancer, which were fewer than the number of cases in our study. Owing to this small number, we unavoidably had to retrieve 487 cases of overall HNSC from the TCGA database.
Survival analyses of TCGA tonsil cancer data were performed in accordance with the mutation status of TP53, PIK3CA, and RTK pathway genes ( Figure 6). TP53 mutations had worse prognostic consequences for OS (p < 0.0001) and DFS (p = 0.032) in the TCGA tonsil cancer cohort. PIK3CA and RTK pathway gene mutations, on the other hand, were not correlated with OS or DFS in the TCGA tonsil cancer cohort, even in the HPV-positive group.

Identification of Gene Signatures Using GSEA
We then carried out a GSEA with the identified mutations of the 29 genes. The top 20 enriched gene sets are shown in Figure 7. Several malignancy-related gene sets, including head and neck SCCs, glioblastoma, melanoma, endometrial cancer, prostate cancer, and breast cancer, were strongly enriched. The genes involved in signal transduction, such as growth factor receptors and second messengers, EGFR tyrosine kinase inhibitors, and PI3K signaling pathways, were mostly related to the 29 genes observed in TSCCs.

Identification of Gene Signatures Using GSEA
We then carried out a GSEA with the identified mutations of the 29 genes. The top 20 enriched gene sets are shown in Figure 7. Several malignancy-related gene sets, including head and neck SCCs, glioblastoma, melanoma, endometrial cancer, prostate cancer, and breast cancer, were strongly enriched. The genes involved in signal transduction, such as growth factor receptors and second messengers, EGFR tyrosine kinase inhibitors, and PI3K signaling pathways, were mostly related to the 29 genes observed in TSCCs.

Discussion
In our study, somatic mutations were relatively common in tonsil cancers, of which 81.4% had one or more oncogenic/likely oncogenic mutations in twenty-nine cancer-related genes, and 48.6% displayed concurrent mutations in two or more genes. TP53 (46.5%), PIK3CA (25.6%), PTEN (18.6%), EGFR (16.3%), and SMAD4 (14.0%) were the top five mutated genes. The fact that these genes are the most frequently altered genes in HNSCC, including TSCCs, represents a critical therapeutic target in patients with TSCCs [17,[25][26][27]. In this study, the presence of TP53 mutations was highly correlated with reduced OS and DFS in patients who underwent adjuvant treatment following surgical resection, and multivariate analysis disclosed it to be an independent factor associated with inferior OS. As the TP53 gene is widely inactivated for its tumor-suppressive function in a number of cancers, and TP53 mutation has been reported to confer resistance to cancer therapies, its reactivation or restoration may be a highly attractive therapeutic target for treating the disease, resulting in tumor regression [28]. Unfortunately, this trial is generally regarded as difficult and has only been demonstrated in animal and cell line models [29][30][31]. Synergistic effects have been displayed in the TSCC cell line when complementing single PI3K, PARP, and WEE1 inhibitor medications with 10 Gy or combining these inhibitors [31]. In a current Phase I clinical trial, the WEE1 kinase inhibitor adavosertib (AZD1775) has exhibited beneficials effects on TP53 mutants in the context of HNSCC [32].
The pathways involving EGFR signaling (51.2%), p53 signaling (48.8%), PI3K (39.5%), and RTKs (34.9%) were found to be frequently involved, suggesting that they may contribute to the tumorigenesis and progression of tonsil cancer [11,27,[33][34][35][36][37]. These mutations were also identified in the TCGA HNSC data; however, TCGA tonsil cancer data were available for only a small number (n = 38) of patients with tonsil cancer, which is a smaller patient cohort than that in our study, thus reflecting the importance of our research. In our GSEA, the 29 genes detected in TSCCs were confirmed to be linked with the gene signatures in signal transduction by growth factor receptors and second messengers, EGFR tyrosine kinase inhibitors, and PI3K signaling pathways. In the present study,

Discussion
In our study, somatic mutations were relatively common in tonsil cancers, of which 81.4% had one or more oncogenic/likely oncogenic mutations in twenty-nine cancer-related genes, and 48.6% displayed concurrent mutations in two or more genes. TP53 (46.5%), PIK3CA (25.6%), PTEN (18.6%), EGFR (16.3%), and SMAD4 (14.0%) were the top five mutated genes. The fact that these genes are the most frequently altered genes in HNSCC, including TSCCs, represents a critical therapeutic target in patients with TSCCs [17,[25][26][27]. In this study, the presence of TP53 mutations was highly correlated with reduced OS and DFS in patients who underwent adjuvant treatment following surgical resection, and multivariate analysis disclosed it to be an independent factor associated with inferior OS. As the TP53 gene is widely inactivated for its tumor-suppressive function in a number of cancers, and TP53 mutation has been reported to confer resistance to cancer therapies, its reactivation or restoration may be a highly attractive therapeutic target for treating the disease, resulting in tumor regression [28]. Unfortunately, this trial is generally regarded as difficult and has only been demonstrated in animal and cell line models [29][30][31]. Synergistic effects have been displayed in the TSCC cell line when complementing single PI3K, PARP, and WEE1 inhibitor medications with 10 Gy or combining these inhibitors [31]. In a current Phase I clinical trial, the WEE1 kinase inhibitor adavosertib (AZD1775) has exhibited beneficials effects on TP53 mutants in the context of HNSCC [32].
The pathways involving EGFR signaling (51.2%), p53 signaling (48.8%), PI3K (39.5%), and RTKs (34.9%) were found to be frequently involved, suggesting that they may contribute to the tumorigenesis and progression of tonsil cancer [11,27,[33][34][35][36][37]. These mutations were also identified in the TCGA HNSC data; however, TCGA tonsil cancer data were available for only a small number (n = 38) of patients with tonsil cancer, which is a smaller patient cohort than that in our study, thus reflecting the importance of our research. In our GSEA, the 29 genes detected in TSCCs were confirmed to be linked with the gene signatures in signal transduction by growth factor receptors and second messengers, EGFR tyrosine kinase inhibitors, and PI3K signaling pathways. In the present study, RTK pathway mutations, including EGFR, were considerably associated with reduced OS and DFS upon univariate analyses, in the overall TSCCs as well as the HPV-positive subgroup; however, multivariate analyses did not reach statistical significance in this case. Overall, RTK somatic mutations, including EGFR and MET, mediate therapeutic resistance in HNSCC [33]. However, the prognostic and predictive relevance of individual RTK mutations are not clear in HNSCC, including tonsil cancer [33]. As these subsets of patients may potentially benefit from a drug targeting the related pathways, effective screening of molecular-targeted genes may be of great clinical relevance. Targeting EGFR utilizing anti-EGFR monoclonal antibodies or kinase domain inhibitors with concomitant radiation is a treatment choice for patients with HNSCC [38]. Several tyrosine kinase inhibitors, such as gefitinib, erlotinib, and afatinib, have also been used to treat HNSCC [39].
TP53 and PIK3CA mutations were the top two common alterations in tonsil cancer, independent of HPV infection status. We found no relationship between clinicopathological variables and PIK3CA mutation in tonsil cancer. The lack of specific clinicodemographic features concerned with PIK3CA mutation has been suggested in oropharyngeal cancers [27], although some studies have characterized rare patients with decreased exposure to tobacco or HPV-positivity by the clinicopathological features of PIK3CA mutation [16,40]. Previous studies have reported that TP53 mutations are common in HPV-negative tumors, while PIK3CA mutations are frequent in HPV-positive cancers [27,40,41]. Recent NGS-based studies identified TP53 (9.3-25.0% vs. 63.8-85.7%) and PIK3CA (20.1-21.4% vs. 6.4-32.1%) in HPV-positive and -negative oropharyngeal SCCs [42], respectively, indicating that although the total number of TP53 and PIK3CA mutations may differ depending on the HPV status, these may still be relatively prevalent [16,42]. Meanwhile, HPV-positive and -negative tonsil cancers may likely have different preferences for cancer-related pathways and detailed somatic mutations: HPV-positive tumors tend to have higher rates of RB1, HRAS, and FGFR3 mutations. Of note, FGFR3 has been identified as a common variant, particularly in HPV-positive tonsil/base of tongue cancers, where FGFR3 mutations have indicated worse prognosis of patients [27]. Similarly, in our study, RTK mutations, including FGFR3, were associated with low OS and DFS in the HPV-positive group upon univariate analysis. In contrast, HPV-negative tumors showed a higher tendency for TP53, FBXW7, PDGFRA, CDKN2A, NOTCH1, AKT1, and FLT3 mutations in the present study. Reder et al. [42] also demonstrated that HPV-negative oropharyngeal cancers have a significantly higher frequency of mutations in TP53, FAT1, KRAS, NOTCH1, and NRAS, with the reported FBXW7 (25.0%) mutation frequency being consistent to that observed in our study. TP53, CDKN2A, and NOTCH1 have repeatedly been reported to be associated with HPV-negative oral/oropharyngeal SCCs [16,42]. Likewise, the clinicopathological correlation of our study suggested that TP53, CDKN2A, and FBXW7 mutations may be closely linked with HPV-negative smoking/alcohol-consumption-related tumors; TP53 mutation was associated with heavy smoking, HPV-negativity, node-negativity (pN0), and PD-L1-negativity; CDKN2A mutation with heavy alcohol drinking history and node-negative tumors; and FBXW7 mutation with heavy smoking history. Mutations in CDKN2A and TP53 have been described among the earliest genetic events in oral dysplastic lesions [15], which may more likely indicate HPV-negative tumors. Notably, there are no precursor lesions of HPV-associated oropharyngeal SCCs [5,43].
The eighth edition of the AJCC staging system for oropharyngeal cancer has been updated with the incorporation of HPV infection and extranodal extension [12], and now offers better stratification between early or advanced stages and T or N categories for OS than the seventh edition for tonsil cancers [20]. However, the prognostic implication of the various somatic mutations in tonsil cancers has rarely been reported, with little information on staging under the eighth edition of the AJCC guidelines [41,42]. Under this updated staging system, we identified the AJCC stage as the best independent prognostic indicator in tonsil cancers, regardless of HPV status or adjuvant therapy. RTK pathway mutations were likely associated with adverse prognostic impact on OS and DFS in patients with TSCC, although they were not an independent prognostic factor. Nevertheless, RTK pathway gene mutations were more clinically relevant to the advanced stage of tonsil cancers than other mutations in this study. With respect to HPV, HPV infection did not carry an independent prognostic implication for OS and DFS in our study, which may be due to the notable contribution of HPV to the T and N categories of the staging system. The impact of the staging modification might have obscured the effect of HPV because our study analyzed all HPV-positive as well as -negative tumors [20].
TP53 mutations, despite having the highest frequency (46.5%), occurred at only 6.7% of the hotspot codons in our study, indicating that most TP53 mutations are widely distributed across exons 4-8 in non-hotspot positions in tonsil cancers. PIK3CA variants, the second most common (25.6%) in TSCCs, were mainly located in non-hotspot regions, and their hotspot mutations (E542, E545, and H1047) only accounted for one-fifth of PIK3CA variants. This may emphasize the need for caution in the production of sequencing primers and their interpretation when attempting identification using direct sequencing. Earlier studies using direct sequencing and hotspot-targeted primers have reported low rates of TP53 (21-23%) [34,44] and PIK3CA (16-19.5%) mutations in oral/oropharyngeal SCCs [36,40,45]. Instead, a comprehensive NGS panel may be effective in assessing a broad spectrum of genetic alterations in TSCCs. Recent studies on oropharyngeal cancers using NGS platforms have reported high TP53 (33-55.4%) and PIK3CA (26.8%) mutation frequencies, comparable to those observed in our study [41,42,46]. Notably, 58.1% of tonsil cancers harbored candidate target mutations in 14 genes (ALK, ATM, EGFR, ERBB2, FGFR3, FLT3, IDH1, IDH2, KIT, KRAS, MET, PDGFRA, PIK3CA, and SMARCB1) with level 1 evidence in OncoKB™, which could be implicated in targeted therapies in the future. In addition, we found that SMARCB1 mutations were more frequent in the 10 discordant cases (23.3%) between p16 immunohistochemistry and HPV PCR results. Although the significance of discordant cancers is controversial, patients with discordant cases demonstrate a borderline significant trend toward survival differences compared to those with concordant cases, which may be a potential therapeutic strategy for discrepant tonsil cancers with SMARCB1 mutation [47][48][49]. This may also imply that NGS-based gene analyses of tonsil cancer may potentially contribute to providing comprehensive treatment options, including molecularly targeted therapies; however, in the current scenario, NGS testing is not yet routine.
One of the limitations of this study is that we used a limited cancer-related gene panel that lacked information on copy number alterations. Another limitation is the small number of tonsil cancers included in the NGS analyses. Nevertheless, our study improved the understanding of mutation profiles and their clinical significance in poorly elucidated tonsil cancers, with the highest count of HPV and SCC sites among studies on oropharyngeal cancers in the era of precision medicine. Our data showed a similar pattern of mutation frequencies to TCGA tonsil cancer data; the latter was derived from a smaller patient cohort than that considered in our study. Because tonsil cancer data are very scarce even in the TCGA head and neck cancer database, our findings may be valuable for further understanding the genomic mutation profile of cancers of the tonsils, a specialized oropharyngeal organ among head and neck structures.

Conclusions
In conclusion, the present study revealed that oncogenic/likely oncogenic mutations are relatively high in tonsil cancers, and that TP53 and RTK pathway mutations may be candidate predictors for detrimental prognosis in patients receiving adjuvant therapy and for overall tonsil cancer patients/HPV-positive subgroup, respectively, according to the AJCC eighth edition staging system. The findings of this study may pave the way for further prospective investigations and may be clinically relevant for the potential personalized therapeutic options for tonsil cancer.
Supplementary Materials: The following supporting information can be downloaded at https://www. mdpi.com/article/10.3390/biomedicines11030851/s1, Table S1: Full list of pathogenic mutations.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest:
The authors declare no conflict of interest.