Feline oral squamous cell carcinoma (FOSCC) is the fourth most common cancer, and the most commonly found malignant oral tumor in cats (1), with a 1-year survival rate of < 10% (2). FOSCC arises from the normal squamous epithelium of the oral (1) (gingiva, tongue, and sublingual regions) cavity. FOSCC rarely metastasizes to distant locations; however, the lymph nodes can be affected in 13–31% of cases (3). Early studies indirectly implicated using flea collars, feeding predominantly canned foods in the diet to increase the risk of development of FOSCC up to 5.3-fold (3) as well as environmental tobacco smoke (4). There are currently no broadly effective treatment options for most cats with FOSCC. Surgical excision has shown to be an effective method to treat FOSCC and other oral tumors with cats having 1-year survival rates of over 80% (5, 6). Chemotherapy or radiation are additional alternatives, but most owners don't opt for these treatments due to side effects and for most cats, the median survival time is only 2–6 months (7, 8). FOSCC presents with similar features as HNSCC, such as inflammation, spontaneous formation, heterogeneous cell environment, natural tumor, and host immune system interactions, and thus may present an opportunity to model comparative therapeutics (9, 10).

HNSCC is the sixth most common cancer worldwide, with 550,000 new cases per year, and has a 5-year survival rate of 50% (11, 12). Like FOSCC, if HNSCC is diagnosed in the early stages survival rates are much higher (2, 13, 14). Predicted risk factors for HNSCC include exposure to tobacco smoke, alcohol, and infection with HPV (13–15). Molecular similarities between the two species have been reported but these presumptions are only based on candidate gene studies in FOSCC. Both humans and cats show the perturbed function of p53, altering cell metabolism, preventing cell cycle arrest, and apoptosis (16). Gain of function in p53 has also been observed in HNSCC and is associated with enhanced tumor progression, invasive cell growth, and metastatic potential (17). Overexpression of EGFR is found in 69–100% of FOSCC and 90% of HNSCC cases, driving cycle progression, and facilitating the invasion of oral tissues (10, 18–20).

Naturally occurring companion animal models of cancer are becoming integral to the understanding of tumor evolution and progression (1, 6). The mouse is the standard model, yet those models lack critical factors such as spontaneous tumor formation as a result of acquiring somatic mutations in shared habitats, progressive tumor heterogeneity, and often a cancer-conditioned immune system. Understanding the genomic environment of both cat and human OSCC will help inform future translational studies. With many unknowns in the genetics of FOSCC, we aimed to characterize the mutational and transcriptional profile of FOSCC and contrast this with HNSCC. To accomplish this, FOSCC tumor tissue and matched blood samples were used for whole exome sequencing (WES), and RNA-seq was generated on FOSCC tumor tissue and oral cavity samples from healthy cats. Somatic mutations, their corresponding contribution to the tumor mutation burden (TMB), and differentially expressed genes (DEG) were assessed to compile a preliminary report of FOSCC genetics.

In this first study of the somatic mutations and gene expression variation present in FOSCC, we describe the similarities and differences when comparing to HNSCC, its most closely related cancer type in humans. Although FOSCC is the most common oral tumor in domestic cats, our knowledge of its genetic properties is very limited. The few characterized molecular features of FOSCC include overexpression of EGFR and perturbed p53 expression (19, 37). When examining the commonalities of FOSCC and HNSCC, overexpression of EGFR in 90% of 750 HNSCC tumors sequenced, and TP53 mutated in 41% of cases is also observed (38). We hypothesized that altered genes in FOSCC would be known to be implicated in HSCC, and would present alternate candidates for hypotheses testing of mechanisms of action (35). We find several mutated genes in common (Supplementary Table S4) the strongest candidate being TP53, which is the most mutated in HNSCC and all cancer types (2, 39–41). Given that TP53 is a tumor suppressor gene and variation in TP53 and is a predictive marker for immunotherapy in HNSCC (16) it is reasonable to hypothesize that if feline immune checkpoint drugs were available their use in the rapidly growing FOSCC could be efficacious. In agreement with our findings, p53 mutations have been seen in other FOSCC studies (4, 37).

Immunotherapy is not an option in feline cancer treatment today but may likely be available in the future. Since TMB has been shown to be useful in some human cancer types, e.g., HHNSCC non-small cell lung cancer and melanoma (42–44), we sought to compare FOSCC to HNSCC for this metric despite our small cohort size. Recent studies on HNSCC have found that mutations in TP53 are associated with high TMB and low overall survival rates, and coincidently HNSCC patients with high TMB have higher response rates to immune checkpoint therapy (23, 24). Moreover, HNSCC with TMB values of >5.0 is associated with poor prognosis (45, 46). Two of our FOSCC samples fall into what we suggest is a high TMB value at 8.5 and 6 and if immune checkpoint therapy was available could be evaluated for tumor control. But we recognize this is very speculative at this stage of evaluating FOSCC genetics. More FOSCC sequencing experiments to obtain better estimates of TMB are needed as well as the future availability of immune checkpoint inhibitors to improve outcomes for this very lethal cancer.

Other FOSCC recurrent or single gene mutations of interest we evaluated included KAT2B, ARID1A, MED12L, and HOXB3 mostly due to their involvement in cell growth. KAT2B is a member of the lysine transferases that are responsible for the acetylation of a broad range of proteins that can function as tumor suppressors or oncogenes (47). HNSCC cell lines have shown universal loss of KAT2B (11) and HNSCC tumors show significantly lower KAT2B expression compared to normal tissue (48). One FOSCC sample had a missense variant in KAT2B (Supplementary Table S4) and it was significantly downregulated in our RNA-seq data (Supplementary Table S4) suggesting a candidate driver gene role in FOSCC. A KMT2D mutation was identified in only one sample but has similar epigenetic properties to KAT2B (49). Studies using The Cancer Genome Atlas database (TCGA) have associated mutations occurring in KMT2D, to the open chromatin state, thereby promoting gene expression (50, 51). Mutations in KAT2B and KMT2D may induce epigenetic changes in both HNSCC and FOSCC that could advocate for the treatment with epigenetic drug control of both (47, 51). A missense mutation in ARID1A was found in one sample (Supplementary Table S4), however, this gene did not show altered expression in FOSCC. ARID1A functions as a tumor stemness repressor by disrupting the function of the p53 or PTEN pathways (52, 53). This gene is often deleted in many human cancers (52) yet with no known role in feline cancers.

Several other candidate genes (MED12L, HOXB3, and PXYLP1 with one gene mutation per sample) were found to overlap in HNSCC and FOSCC (Supplementary Table S4). Mediator Complex Subunit 12L (MED12L) works by activating the kinase activity of CDK8 which regulates the growth and division of cells (54). MED12L is differentially overexpressed in many cancers however, the MED12L complex is altered in only 3.05% of HNSCC patients (55, 56). The HOX genes regulate a wide range of cell activity including proliferation and migration thus making their contribution to FOSCC beyond interpretation at this stage. HNSCC studies have shown an overall increased expression of all HOX genes, including HOXB3 (57–59). No overlap in actionable genes mutated in FOSCC was found when searching OncoKB relative to HNSCC. We believe this is due to the low sample size of our cohort which only additional sequencing can address.

Altered gene expression is often used to discover and advance gene candidates for further study of their oncogenic roles. Transcriptional profiling showed that FOSCC is enriched with genes known to be upregulated in Human OSCC (34) suggesting conserved gene regulatory mechanisms. While these findings were predicted given the remarkable clinical, pathological, and genetic parallels between both tumors (9, 37, 60–62) they in no way confirm comparative origins or outcomes. The functional gene and canonical pathway enrichment analyses provided us further insight into the possible activation of several pathways that were seen in FOSCC that included EMT, hypoxia- and inflammation-related pathways (61–65), indicating the complementary value of gene expression analysis in this study. EMT typically involves the expression of transcription factors that can activate this cellular program (i.e., SNAI1, TWIST1, ZEB1, and ZEB2), and is characterized by the upregulation of mesenchymal-related genes (e.g., FN1, VIM, CDH2, and metalloproteinases) and downregulation of epithelial-related genes (i.e., CDH1 and cytokeratins) (65–67). These studies are insightful into the aggressive biological behavior of FOSCC in part due to the activation of the EMT program. Not surprisingly activation of the hypoxia and angiogenesis pathways, as well as several inflammation pathways are predicted to be implicated in oral cancer with rapid tumor growth creating areas of ischemia and necrosis, and the production of reactive oxygen species (62). These events also induce the expression of transcription factors that can stimulate angiogenesis while simultaneously eliciting a local inflammatory response inducing expression of PTGS2 (i.e., COX-2), with signaling via NF-kB, TNF, IL-6, and TGFB, all of which were found to be significantly enriched in FOSCC.

In the context of the recurrent TP53 somatic mutations observed, it was interesting to note that the p53 pathway was not significantly activated across the set of FOSCC tumors compared to healthy controls (GSEA FDR > 0.2). This observation suggests that TP53 mutations in FOSCC represent loss-of-function events and that the mechanisms that would normally result in p53 pathway activation are likely impaired. Consistent with cancer studies in people (68), TP53 and relevant target genes (e.g., CDKN1A) known to be expressed in wild-type cells were found to be downregulated in FOSCC, while genes that would be expected to be upregulated in TP53 mutated cells (e.g., E2F1, MYBL2, and FOXM1) were enriched in FOSCC. Overall, these findings strongly implicate TP53 somatic mutations as driver events in FOSCC tumorigenesis.

Another aspect of FOSCC revealed by RNA-seq was the variation among tumor expression profiles. Indeed, although cluster analysis revealed a distinct molecular phenotype when compared to control samples, we observed differences among tumor samples among individual genes as well as for relevant biological pathway activity (Figure 3, Supplementary Figure S6). This tumor heterogeneity may reflect inter-individual differences (i.e., variations in tumor stage, anatomical location, breed, age, other clinical and environmental factors, etc.), and inherent variations among primary clinical samples of naturally occurring disease, which are comprised of heterogeneous cell types and frequencies and are typically collected in a clinical setting rather than under strict research protocols.

There are several limitations to this small study, with the most important factor being cohort size. Only six samples were available at the time of sequencing, therefore conclusions between potential causal variants in the cat in common with the human were found less frequently. Another limitation of the study is that healthy mucosal tissue from case animals was not available to use as matched RNA-seq controls, due to limitations of the clinical sampling protocol. Tumors also originated from different anatomical locations and were collected and processed at different times, which can contribute to transcriptional variation. Lastly, aside from TP53, the common mutations identified with functional effects have not been studied at length and the implication of variants in these genes are unknown. Further study is in progress in a larger cohort of cats where more significant conclusions may be drawn.

The first exome resource was used to evaluate the somatic mutational landscape of FOSCC and gene expression changes were identified using RNA-seq. We observed several mutations in TP53, consistent with what is seen in HNSCC, and several other genes also overlapped between FOSCC and HNSCC. Our study suggests that similar genes initiate tumorigenesis in both species and perhaps FOSCC may serve as a comparative model of treatment in HNSCC. Similarities in the mechanism of FOSCC and HNSCC shown in the RNA-seq studies, such as genes implicated in inflammation further demonstrate the possible use of the domestic feline as a model for HNSCC.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material.

WW and JB provided samples for analysis. AR completed variant calling and analysis. ZS, OG, and SC assisted in variant calling and analysis. JKG and SP completed sample collection as well as RNA-sequencing analysis. All authors contributed to the article and approved the submitted version.