Comparison of mutation landscapes of pretreatment versus recurrent squamous cell carcinoma of the oral cavity: The possible mechanism of resistance to standard treatment

Abstract Background A high recurrent rate of oral squamous cell carcinoma (OSCC) is a major concern in head and neck cancer treatment. The study of the genetic mutation landscape in recurrent OSCC may provide information on certain mutations associated with the pathobiology and treatment response of the OSCC. Aim We investigated the mutation landscape of matched pretreatment and recurrent tumors to understand the influence of genetic mutations on the pathobiology and clinical outcomes in OSCC. Methods and Results We sequenced 33 formalin‐fixed paraffin‐embedded (FFPE) recurrent tumors, primary tumors, and primary tumors before recurrence that matched the recurrent tumors collected from Rajavithi Hospital during 2019–2021. We identified recurrent mutations from these samples by the Oncomine Ion Torrent‐based next‐generation sequencing on the 517 cancer‐associated gene panel. From the results, we found that the most commonly mutated gene in the cohort is a histone methyltransferase KMT2D (54.55%), implicating that aberrance in epigenetic regulation may play a role in oral cancer tumorigenesis. Functional protein association network analysis of the gene frequently mutated in the recurrent tumors showed enrichment of genes that regulate the cancer cell cycle, that is, MRE11A, CDKN2A, and CYLD. This finding was confirmed in the primary‐recurring matched pair. We found that recurrent tumors possess a small but recurring group of genes, with presumably the subclonal mutations driving the recurrence of the tumor, suggesting that the recurrent disease originated from a small fraction of the cancer cell that survives standard treatment. These genes were absent in the primary tumor with a good response to the standard treatment. On the other hand, we found an enrichment of DNA repair genes, namely ATR, BRCA1, BRCA2, RAD50, and MUTYH, in nonrecurrent tumors suggesting that the mutations in the DNA repair pathway may at least partially explain the different response to the standard treatment. Conclusions Our pilot study identified pathways of carcinogenesis in oral cancer and specific gene sets that indicate treatment responses and prognoses in this group of patients.

tumors suggesting that the mutations in the DNA repair pathway may at least partially explain the different response to the standard treatment.
Conclusions: Our pilot study identified pathways of carcinogenesis in oral cancer and specific gene sets that indicate treatment responses and prognoses in this group of patients.

K E Y W O R D S
cancer cell cycle, DNA repair, epigenetics, mutation landscape, recurrent oral squamous cell carcinoma

| INTRODUCTION
Oral squamous cell carcinoma (OSCC) is a prevalent type of head and neck cancer. 1,2In 2020, the worldwide incidence of new OSCC cases was 377 713 cases, and the number of deaths from OSCC was 177 757. 3 The major risk factors for OSCC are smoking, alcohol consumption, and human papillomavirus. 4,5While surgical therapy followed by radiotherapy or radio-chemotherapy is the standard treatment for OSCC, 6 recurrence after treatment can occur, ranging from 20% up to 86%. 7,8The recurrence of OSCC is a major problem for treatment and can result in poor survival rates. 9The recurrence rate of OSCC is high, following standard treatment: approximately 40%-60% in advanced stages [10][11][12][13] and 10%-25% in earlier stages. 14,15Recurrence of OSCC is usually difficult to treat due to the limited treatment options available, and these recurrent tumors represent the progeny of resistant cancer cells that successfully evaded standard treatment. 16Moreover, there are currently no predictive biomarkers for the recurrence after standard treatment of OSCC.Therefore, studying the detailed genetic mechanism of recurrence after standard treatment in OSCC is essential to guide future clinical practices.
The tumorigenesis of OSCC is believed to be driven by specific genetic alterations.Therefore, understanding the genetic alteration landscape in OSCC can aid in the development of precision medicine.8][19][20] There are a number of reports on the genetic mutation landscape of OSCC.Previous reports of the snapshot OSCC mutation landscape revealed TP53, CDKN2A, PIK3CA, HRAS, NOTCH1, CHUK, and ELAVLI as frequently mutated genes. 19,21,22However, only a few studies focused on the relative genetic information between primary and recurrent OSCC tumors.
Although ideally, matched primary and recurrent OSCC tumors from the same patients should be compared for the highest quality data, very few such data exist.
To understand the dynamic change of genetic mutations that are influenced by standard treatment, a study with a small number of patients on mutation landscapes in recurrent and metastatic head and neck cancer was performed. 23The result revealed that mutations of C17orf104, ITR3, and DDR2 were specifically found in the recurrent or metastatic tumor but not in the primary.This result is suggesting that the activation of particular genes may facilitate head and neck cancer recurrence.
In this study, we focus on OSCC and aim to identify mutated genes that may contribute to recurrence after standard treatment by comparing gene mutations from FFPE tissue samples of recurrent OSCC before and after standard treatment, a portion of which are paired, matched tumors.We also studied mutations in OSCC with good response to the standard therapy relative to recurrent tumors, hoping to identify mutated genes that may confer sensitivity to standard treatment by studying genes that were correlated with complete response compared with genes that correlated to recurrence.

| Sample collection and inclusion criteria
The retrospective study protocol and archival FFPE samples analyzed were approved by the Rajavithi ethic committee (EC No.64102) and Siriraj Institutional Board (SIRB No. 104/2564[IRB1], COA no.Si 344/2021), whereby the anonymized archival FFPE samples were provided to the researchers without the requirement for patient consent.Naïve oral cancer patients who underwent surgery, followed by standard treatment during 2019-2021 in Rajavithi Hospital, were included in this study.The recurrent group samples included FFPE from primary oral cancer patients who had regular follow-ups in the clinic with evidence of recurrence within 24 months.Of these samples, 7 cases have matching primary and recurrent FFPE blocks.The nonrecurrent control group included FFPE in oral cancer patients who had regular follow-ups in the clinic with no evidence of recurrence for at least 36 months and found 9 cases.The patients who were lost to follow-up or had no pathological report of recurrence were excluded from the trial.

| Next-generation sequencing of OSCC
The genomic DNA was extracted from FFPE by using MagMAX™ Cell-Free Total Nucleic Acid Isolation Kit (ThermoFisher Scientific).
Qubit quantification was performed after DNA extraction.TaqMan™ GUSB gene assay was performed as a proxy determination of the amplifiable FFPE DNA.Deaminated cytosine bases, commonly found in FFPE specimens, were enzymatically removed by treatment with uracil DNA glycosylase (ThermoFisher Scientific).Target sequencing libraries were constructed with Oncomine™ Comprehensive Assay Plus (DNA+RNA) (ThermoFisher Scientific) with a target 517 cancerassociated gene panel.Then, target sequencing was performed by Ion GeneStudio™ S5 System (ThermoFisher Scientific) following the manufacturer's instructions.

| Next-generation sequencing analysis
The sequencing data were uploaded to Ion Reporter software (V.The results then were manually analyzed for only likely pathogenic or pathogenic variants according to ClinVar (V.20201121), with a sequence coverage of at least 100Â.

| Protein-protein interaction networks functional enrichment analysis
The pathway or protein-protein interaction networks that may involve the pathogenesis of OSCC was determined by a search tool for retrieval of interacting genes (STRING) database. 24The lists of the gene identified to be unique to the recurrent, and the non-recurrent tumors were applied to the STRING tool with the active interaction sources set to include the data from experiments, databases, and coexpression (only Homo sapiens specified) to construct the proteinprotein interaction networks.

| Patient characteristics
We performed an investigation in 33

| KMT2D may play role in oral cancer tumorigenesis
For all samples, we opted for targeted deep sequencing of 517 cancer-associated genes, which divided the results into three groups, the primary tumor which did not recur, the primary tumor before recurrence, and recurrent tumors post-standard treatment.The sequencing results were supplied in the Supplementary Information.The resulting mutations were shown in Figure S2 with the TP53 gene as a top mutated gene.The mutations representing the pathogenic and likely pathogenic genes in our cohort were selected as described.The resulting mutation landscape was shown in Figure 1A and Figure S3.Our data revealed that the KMT2D mutation, which is an epigenetic gene, is the most commonly found gene among all three groups.The mutation rate of the KMT2D was 54.55%.A similar finding was also found in The Cancer Genome Atlas (TCGA), and in the study by Nisa and colleagues 25 in 2018, in which the KMT2D mutations were 14.46% and 66.67%, respectively (Figure 1B).Our study confirmed that KMT2D may play a role in oral cancer tumorigenesis but may not be the ideal marker for disease prognosis.

| Mutations in genes regulating cancer cell cycle and differentiation may confer the recurrence of OSCC after completed treatment
To identify gene mutations that may contribute to a recurrence of OSCC, the three groups of OSCC patients, gene mutation frequencies were compared as shown in Figure 2A.Of all the gene mutations detected, a total of six (CDKN2A, CYLD, MER11A, CIC, GRID2, and PDGFRA), were uniquely associated with tumor recurrences.To identify potential pathways of enrichment for these mutated genes, we used STRING to identify gene pathways and the results showed enrichments of pathways involved with cancer mutant cell cycle and differentiation (Figure 2B).To establish how important these mutated genes are (Figure 2A,B), we compared the mutated genes in a group of recurrent OSCC patients, who have paired match tumors before and after treatment as shown in Figure 3A.The results confirmed that the gene identified in Figure 2A,B was also found in the match-paired recurrent tumors after treatment.These results suggest that the mutations in cancer cell-cycle regulation/differentiation genes may play an important major role in the recurrence of OSCC after standard treatment is completed.

|
The recurrent disease may be rare clones of OSCC that survive during treatment Figures 1A and 2A show that the number of mutant genes found in recurrent tumors is less than in the nonrecurrent and primary tumors before relapse.This result may come from a small fraction of OSCC that survive standard treatment; hence, the treatment may allow the survival of the cells containing these mutations.To investigate this hypothesis, we searched and found the genes identified from Figure 2A in the VCF files in both nonrecurrent and primary tumors before relapse with VAF ≤5%.Interestingly, this set of the genes that correlated with recurrence was specifically found in primary tumors before relapse, but never found in the nonrecurrent tumors (56% compared with 0% of cases containing at least one gene mutation associated with recurrent oral cancer in VAF ≤5%) (Figure 3B).This result suggests that recurrent OSCC originated from a small fraction of OSCC that survived initial treatment.

| Mutations in DNA repair genes may confer nonrecurrent cases
Figure 2A showed the result of 24 mutated genes, which were found only in nonrecurrent tumors.To identify the pathway of enrichment for these mutated genes, STRING was used to analyze proteins in the same complex.We found enrichment of DNA repair genes (ATR, BRCA1, BRCA2, RAD50, and MUTYH) as shown in Figure 3A.These mutated genes, especially the mutated BRCA1, ATR, and RAD50 genes, were found in more than one nonrecurrent tumor (22.22%) (Figure 2B).We also identified mutated DNA repair genes in all three groups of our OSCC and found that 55.6% of cases with a gene mutation in the DNA repair pathway can be found in nonrecurrent tumors.
In contrast, only 0% of cases with a gene mutation in the DNA repair pathway can be found in primary tumors recurrent tumors, respectively.This result suggests that mutant DNA repair genes may be a unique characteristic of tumors that do not recur.

| DISCUSSION
To elucidate the mutational profiles of OSCC that may contribute to drug response, we set out to perform next generation sequencing on a cohort containing primary versus recurrent tumors, some of which matched-pair tissue samples.We found that the most common mutation found in our OSCC study was KMT2D, which is also reported to be one of the most common mutations in the TCGA cohort and again in another previous study. 25Mutation of KMT2D was also common in other types of cancer, such as small-cell lung cancer and pheochromocytoma. 26,27The role of KMT2D mutation in tumorigenesis was studied by Maitituoheti and colleagues in 2020 who found that KMT2D-deficient melanoma cells were associated with an epigenetic change in H3K4me1-marked enhancer activtion. 28While a further study of the KMT2D function in the epigenetic and tumorigenesis of OSCC is still required, our finding suggests that KMT2D-mediated epigenetic changes may play an important role in OSCC tumorigenesis.
Our study identified a set of unique genes in recurrent tumors, which have three genes that may contribute to a recurrence factor in our recurrent OSCC.These were CDKN2A, MRE11A, and CYLD.
CDKN2A (p16) gene has been well studied in head and neck cancer including OSCC 18,29,30 and it shows the correlation with lossof-function CDKN2A in the major recurrence of hand-neck cancer. 31,32By analyzing a set of unique genes in recurrent tumors using STRING, we identified the deregulation of the cell cycle pathway, which appears to be closely associated with recurrence in OSCC.This finding was supported by a report on the function of CDKN2A and CYLD.
CDKN2A gene functions lead to cell cycle arrests by inhibiting the function of CDK4 protein. 335][36] Interestingly, data from a mouse model with cdkn2a À/À genotype showed that cdkn2a loss of function is a rate-limiting step in oral cancer formation 37 and targeting the CDK4/CDKN1A/RB1 pathway may be an attractive strategy to treat OSCC. 38CYLD also can regulate cell-cycle progression by inactivating HDAC6 and increasing the levels of acetylated tubulin. 39Loss of function mutation of CYLD has been shown to be associated with cisplatin resistance in oral cancer. 40In the same publication, suppression of CYLD by CYLD-specific siRNA can reverse the cisplatin-resistant phenotype, indicating that CYLD is one of the major proteins that facilitate oral cancer resistance.Even with the limited number of patients in this cohort, our analyses were informative and precise in identifying the major genes known to confer resistance to OSCC.Our study also pointed out that the deregulation of the cell cycle regulatory proteins may be a key event that facilitates OSCC resistance to standard therapy.Therefore, these findings highlight small molecule cell cycle inhibitors as a potential therapeutic target for recurrent OSCC. 41,42terogeneity and clonal selection may also be driving the recurrence in OSCC. 43Our results shown in Figure 3B also suggest that standard treatment induces subclonal selection pressure.This is because, we found a set of unique genes in recurrent tumors preexisting in the primary tumor before the treatment as a subclonal population.Some studies on OSCC show the clonal evolution from subclonal types in primary tumor. 43,44These preexisting genes in recurrent tumors, also detected in the primary tumor, were previously studied in breast cancer. 33The study demonstrated that there are subclonal populations of breast cancer cells that contain a mutation landscape that promotes resistance to chemotherapy. 45e unique finding of DNA repair gene mutations in tumors with a good response to radiotherapy or chemo-radiotherapy was supported in a previous study in a mouse model of engrafted BRCA1deficient tumors.The mouse models showed hypersensitivity to radiation.46 The finding of mutant DNA repair genes may be one factor to indicate one of the prediction methods for patients who may benefit from standard treatment as shown in Figure 4C.In agreement with our result, cancer cells containing DNA repair gene defects were shown to be hypersensitive to platinum-based chemotherapy, such as cisplatin.47 This may explain the typically good responses to DNA damage-induced first-line therapies in this disease.It is also interesting to see, whether PARP inhibitors, such as olaparib and rucaparib, which are currently used in ovarian and breast cancer with DNA repair defects, 48 may be also used in OSCC containing mutations of DNA repair genes.From these data, it appears to be a plausible conclusion that the use of data on DNA repair gene mutations may help to guide treatment plans for primary OSCC as well as the follow-up after the completion of each treatment.On the other hand, the ability to avoid unnecessary treatments with low probability of treatment success may improve patient quality of life.To confirm the findings of this study, we plan to expand our cohort by collecting more samples and applying a new molecular technology to identify the molecular changes in this cancer.
In conclusion, our study found a strong relation between the high rate of KMT2D mutation and OSCC, suggesting role of epigenetics in 5.18) (ThermoFisher Scientific) and analyzed by Oncomine Comprehensive Plus w2.3 DNA-Single sample workflow using human genome assembly GRCh37(hg 19) as a reference for alignment.The abundance of FFPE artifacts was assessed by the software via three parameters, that is, deamination score above 60, the signature pattern , UV damage, or FFPE processing, and abnormally high tumor mutation burden (above 10 muts/Mb).If one of the criteria is met, FFPE artifacts were filtered using a variant allele frequency (VAF) cutoff 10%.The sequencing quality was assessed by the software via four measurements, that is, mapped read >22 million, mean depth >800Â, percent uniformity >80%, and percent on target >85%.Gene annotation was applied by the Oncomine Comprehensive Plus Annotations v1.2.Minor allele frequency was filtered using the Database for Single Nucleotide Polymorphisms (dbsnp v. 154) and 5000Exomes (V.20 161 108).Somatic variant calling was performed by applying the filter chain "Oncomine™ extended 5.18v2," which essentially selected only the somatic the Catalogue Of Somatic Mutations In Cancer variants.

F I G U R E 1
Mutation landscape of oral squamous cell carcinoma (OSCC) patients in this study.(A).Heatmap of top 20 frequently mutated genes.Each row indicates the gene and each column indicates the patient.The bars on the righthand side show numbers of patients containing mutated genes in each row.Different colors presenting types of mutation shown at the bottom of the figure.Colored bars on top of the figure indicate groups of OSCC patients.(B).Percents of mutated KMT2D from three studies.

F I G U R E 2
Sets of unique genes found in each group of oral squamous cell carcinoma (OSCC) sample.(A) Venn diagram shows sets of unique genes found in this study.Different colors in the Venn diagram represent the groups of OSCC sample, nonrecurrent tumor (green), as a recurrent tumor before standard treatment (orange), and recurrent tumor after standard treatment (red).Red box shows the set of unique genes only found in a recurrent tumor after standard treatment and green box shows the set of unique genes only found in a nonrecurrent tumor.(B).Genes uniquely found in recurrent tumors after standard treatment are genes in cell cycle regulation.Lines connected between genes indicate functional or physical interactions between these genes based on search tool for retrieval of interacting genes.F I G U R E 3 Enrichment of cell cycle regulation gene in pairmatched samples.(A).Venn diagram of the pair-matched tumor shows a set of unique genes in cell cycle regulation.Brown color indicates recurrent tumors before standard treatment and red color indicates recurrent tumors after standard treatment.Red box shows the set of unique genes only found in a recurrent tumor after standard treatment.The text with red color shows the unique genes that regulate the cell cycle.(B).Bar graphs of the percent of cases containing at least one cell cycle regulation gene (≥5% variant allele frequency [VAF]).

F I G U R E 4
Mutations DNA repair genes may confer sensitivity to the standard treatment.(A) Enrichment of genes in DNA repair pathway.The set of genes uniquely found in nonrecurrent tumors.Lines connected between genes in the DNA repair pathway indicate functional or physical interactions between these genes based on search tool for retrieval of interacting genes.(B) Bar graphs show the numbers of cases containing mutations in DNA repair genes only found in nonrecurrent tumors.(C) Percents of patients with and without mutations genes in DNA repair pathway.OSCC tumorigenesis.We also found evidence of preexisting subclones in the primary tumor, staying dormant in the pretreatment tumors of our recurrent OSCC group.A set of unique genes that regulate the DNA repair pathway in the tumors will indicate a good outcome after completed treatment.This pilot study may also facilitate us to conduct a new model clinical trial for molecular-guided therapy in the future.