Novel Molecular Targets for Feline Oral Squamous Cell Carcinoma

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Wachiraphan Supsavhad

Graduate Program in Comparative and Veterinary Medicine

The Ohio State University

2016

Dissertation Committee

Thomas J. Rosol, Advisor

James DeWille

Ramiro Toribio

Gwendolen Lorch

Theodore M. Brasky

Copyrighted by

Wachiraphan Supsavhad

2016

Abstract

Feline oral squamous cell carcinoma (FOSCC) is the most common oral cancer in cats.

This cancer is very aggressive and has the ability to invade into the adjacent bone resulting in poor quality of life and a grave prognosis in affected cats. FOSCC has been shown to be comparable to human OSCC in many aspects. However, the pathogenesis of

FOSCC remains unclear. Similar to human OSCC, FOSCCs respond poorly to the available standard therapy. Therefore, novel molecular targets as well as effective therapies are required for this cancer.

The usefulness of p16, p53, and pRb immunohistochemistry (IHC) for human

OSCC classification has been reported. Low p16 frequently associates with non-viral- associated OSCCs, while high p16 with low p53 and pRb usually occurs in HPV- associated OSCC. In addition, a favorable prognosis was found in patients with HPV- associated OSCC and a grave prognosis was observed in patients with non-viral- associated OSCC. In this study, the IHC pattern of p16, p53, and pRb in FOSCC was investigated. We found that low p16 IHC is common in FOSCCs and high p16 IHC is present in a subset of patients with this cancer. However feline papillomavirus was not detected in these samples. These results indicate that FOSCCs have more than one type of classification and a variation in pathogenesis may occur in FOSCC.

Increased osteoprotegerin (OPG) has been found to reduce tumor size and bone invasion in many human bone-invasive cancers. Low levels of feline OPG (fOPG)

ii mRNA expression were previously observed in a bone-invasive FOSCC cell line

(SCCF2). The role of fOPG in bone-invasive FOSCC in vitro and in vivo was investigated using a novel bone-invasive FOSCC cell line with high fOPG mRNA expression (SCCF2-OPGv.1).We found that overexpression of fOPG reduced tumor growth and inhibited bone invasion in bone-invasive FOSCC in vivo. Thus, fOPG could be effective adjuvant therapy for FOSCC-associated bone lysis.

Telomerase, a protein complex with cell immortalization activity, has been speculated as a novel therapeutic target for human cancers. Telomerase transcriptase catalytic subunit (TERT) and the non-coding RNA template are the main components of this protein. Pre-mRNA alternative splicing occurs in human TERT gene and more than

20 human TERT alternative splice variants were identified. In this study, high levels of telomerase activity were observed in FOSCC cells (SCCF1, SCCF2, and SCCF3) and in

2 of 3 FOSCC tumors. Full-length and 10 novel feline TERT (fTERT) alternative splice variants were identified. Full-length and the deletion of exon 10 (Del-e10) splice variant of fTERT were the two common variants in FOSCCs with high telomerase activity found in this study. These findings suggest that alternative splicing is important for the regulation of telomerase activity and provide critical information to further the understanding of telomerase in feline cancers.

iii

Acknowledgments

I would like to take this opportunity to give special thanks to people who have contributed to my success.

First and foremost, I would like to sincerely express my gratitude to my advisor, Dr.

Thomas J. Rosol who has given me an opportunity to pursue my studies at the Ohio State

University. The wonderful experience, constant guidance and support that I have received during my study prepared me well for more challenges in my future career.

I would like to take this chance to thank to Drs. James DeWille, Ramiro Toribio, and

Gwendolen Lorch, members of my dissertation committee, for their valuable technical advice. My gratitude is extended to Dr. Wessel Dirksen, Rosol lab manager, for significant contributions on day-to-day problem solving related to experimental work, suggestions, and considerable editorial help on this dissertation. In addition, I would like to thank all of the previous and present Rosol lab members (Dr. Eason Hildreth, Dr.

Jessica Simmons, Dr. Said Elshafae, Dr. Bardes Hassan, Dr. Nicole Kohart, Dr. Aylin

Demirer, Lucas Altstadt, James Feller and Taylor Wickware) for their support and for contributing to a pleasant working environment during my study.

I am very thankful for the staff in the Histology/Immunohistochemistry Core Lab,

College of Veterinary Medicine, The Ohio State University for the wonderful histopathology slides. This work could not be done without their help.

I am grateful for the tuition scholarship from the Faculty of Veterinary Medicine at

Kasetsart University, Thailand, during the first two and half years of my Ph.D. study and the fellowship from the Department of Defense (DOD) for during last 1.5 year. Without this support, it would not have been possible to complete my studies successfully.

Last but not least, I cannot forget to thank my family members and my beloved husband for the unconditional love, care, and encouragement. I would never come this far without their constant support.

Vita

1999...... Chinat Pittayakom High School

2005...... D.V.M., Kasetsart University, Thailand

2012...... M.S. Comparative and Veterinary Medicine. The Ohio State University 2012 to 2015 ...... Graduate Student Department of Veterinary Biosciences, The Ohio State University 2015 to present ...... Post Doctoral Fellow Department of Veterinary Biosciences, The Ohio State University

Publications

W. Supsavhad, W.P. Dirksen, C.K. Martin, T.J. Rosol. Animal models of head and neck squamous cell carcinoma. The Veterinary Journal. 210: 7-16. October 2015

B.E. Hildreth, K.M. Hernon, W.P. Dirksen, J. Leong, W. Supsavhad, P.N. Boyaka, T.J. Rosol, R.E. Toribio. Deletion of the nuclear localization sequence and C-terminus of parathyroid hormone-related protein decreases osteogenesis and chondrogenesis but increases adipogenesis and myogenesis in murine bone marrow stromal cells. Journal of Tissue Engineering. 6. October 2015; DOI: 10.1177/2041731415609298 vi

J.K. Simmons, B.E. III Hildreth, W. Supsavhad, S.M. Elshafae, B.B. Hassan, W.P. Dirksen, R.E. Toribio, T.J. Rosol. Animal model of bone metastasis. Veterinary Pathology. 52(5). May 2015

S.M. Elshafae, B.B. Hassan, W. Supsavhad, W.P. Dirksen, R.Y. Camiener, H. Ding, M. F. Tweedle, T. J Rosol. Gastrin-releasing peptide receptor (GRPr) promotes EMT, growth, and invasion in canine prostate cancer. The Prostate. 76(9). March 2016

B.B. Hassan, S.M. Elshafae, W. Supsavhad, J.K. Simmons, W.P. Dirksen, S.M. Sokkar, T.J. Rosol. Feline mammary cancer: novel nude mouse model and molecular characterization of invasion and metastasis genes. Veterinary Pathology. June 2016

Fields of Study

Major Field: Comparative and Veterinary Medicine

vii

Table of Contents

Abstract ...... ii

Acknowledgments...... iv

Vita ...... vi

List of Tables ...... xiii

List of Figures ...... xiv

Chapter 1: Animal Models of Head and Neck Squamous Cell Carcinoma ...... 1

Abstract ...... 1

Introduction ...... 2

Hamster model of HNSCC ...... 5

Rodent models of HNSCC ...... 7

Rat and mouse chemically-induced carcinogenesis model ...... 7

Xenograft transplantation models of HNSCC ...... 9

Transgenic mouse models of HNSCC ...... 11

Domestic animal models of spontaneous HNSCC ...... 13

Dog model of HNSCC ...... 14

viii

Cat model of HNSCC ...... 15

Conclusions ...... 22

Chapter 2: p16, pRb, and p53 in Feline Oral Squamous Cell Carcinoma ...... 33

Abstract ...... 33

Introduction ...... 34

Materials and Methods ...... 36

Sample collection ...... 36

Immunostaining ...... 37

p16 mRNA in 3 FOSCC cell lines...... 38

Papillomavirus L1 major capsid DNA investigation ...... 39

PCR product purification and sequencing ...... 40

Statistical analysis...... 40

Results ...... 41

Relative p16 mRNA quantification in 3 FOSCC cell lines ...... 41

Immunohistochemistry (IHC) ...... 41

Papillomavirus L1 capsid DNA amplification ...... 42

Sequencing...... 43

Discussion ...... 43

Conclusions ...... 48

Chapter 3: The Role of Osteoprotegerin in Bone-invasive Feline Oral Squamous Cell

Carcinoma ...... 61

Abstract ...... 61

Introduction ...... 63

Materials and Methods ...... 65

Cell lines and tissues...... 65

Feline OPG cDNA cloning ...... 66

Stable transfection of fOPG in SCCF2 cells ...... 66

Relative ratio of feline OPG splice variant ...... 67

Quantitative RT-PCR ...... 68

Feline OPG secretion ...... 68

Murine calvaria co-culture ...... 69

Lentiviral luciferase transduction ...... 70

In vivo studies ...... 71

Urine calcium concentration ...... 72

Statistical analysis...... 72

Results ...... 73

Feline OPG cDNA cloning ...... 73

The relative ratio of full-length and splice variant of feline OPG mRNA in SCCF2

cell lines, FOSCC tumors, and normal cat tissues ...... 73

fOPG overexpression in SCCF2 cell line ...... 74

Secreted fOPG protein in SCCF2-OPGv.1...... 74

In vitro bone lysis ...... 75

In vivo tumor growth ...... 75

Faxitron radiography ...... 76

Histopathological findings ...... 76

Urine calcium concentration ...... 77

Discussion ...... 77

Conclusions ...... 80

Chapter 4: Feline Oral Squamous Cell Carcinoma, Telomerase Activity, and Novel

Alternative Splicing Cloning and Expression of TERT ...... 91

Abstract ...... 91

Introduction ...... 93

Materials and Methods ...... 95

Cell lines and sample collection ...... 95

Relative fTERT mRNA expression ...... 96

Protein extraction and telomerase activity quantification ...... 97

fTERT cDNA cloning ...... 98

fTERT alternative splice variant quantification ...... 100

Statistical Analyses ...... 100

Results ...... 101

Relative fTERT mRNA expression ...... 101

Telomerase activity in FOSCC ...... 101

Full-length and alternative splice variants of fTERT ...... 102

The relative proportion of the fTERT splice variants in FOSCC ...... 103

Discussion ...... 103

Conclusions ...... 107

References ...... 115

xii

List of Tables

Table 1. Summary of advantages and disadvantages of animal models for head and neck squamous cell carcinoma (HNSCC) research ...... 24

Table 2. Classification of p16, pRb, and p53 immunohistochemistry staining ...... 49

Table 3. Signalment and immunohistochemistry data from 43 cats with FOSCC ...... 50

Table 4 Immunoreactivity to p16, p53, and pRb in Three FOSCC Xenograft Tumors and

Primary Feline Bowenoid in situ Cutaneous Squamous Carcinoma (FBISC) ...... 51

Table 5. RT-PCR primers for feline OPG, RANKL and GAPDH ...... 81

Table 6. The relative amount of full-length (fOPGv.1) to fOPGv.2 splice variant of feline osteoprotegerin mRNA ...... 81

Table 7. PCR primers for semi-quantification of feline TERT alternative splicing ...... 107

Table 8. PCR cycles used for individual feline oral squamous cell carcinoma (FOSCC) cell lines, tumors, normal cat oral tissues, normal cat testis and plasmid cDNA (positive control)...... 108

Table 9. Full-length and alternative splice variants of feline TERT...... 109

Table 10. Relative proportion of full-length feline TERT and feline TERT alternative splice variants...... 110

xiii

List of Figures

Figure 1. Hamster cheek pouch HNSCC model ...... 26

Figure 2. Histologic progression in the hamster cheek pouch model following DMBA exposure ...... 27

Figure 3. Rat tongue HNSCC model ...... 28

Figure 4. Nude mouse model of invasive feline HNSCC ...... 29

Figure 5. Cat with late-stage HNSCC of the left maxilla with extensive growth and invasion of the maxillary bone (arrowheads). Scale is 5 mm ...... 30

Figure 6. Cross-section of a cat tongue with SCC that originated in the ventral regions with marked invasion of the muscularis (arrowheads). Scale is 5 mm ...... 30

Figure 7. Cat with gingival HNSCC of the anterior mandible ...... 31

Figure 8. Pathogenesis of bone invasion by HNSCC ...... 32

Figure 9. Alignment of human, chimpanzee, rabbit, cat, cow and mouse p16 cDNA sequences...... 52

Figure 10. Relative p16 mRNA expression in 3 feline oral squamous cell carcinoma

(FOSCC) cell lines compared to normal cat gingiva...... 53

Figure 11. p16 immunohistochemistry of FOSCC and normal cat oral tissues ...... 54

Figure 12. pRb and p53 immunohistochemistry of FOSCC and normal cat oral tissues. 56

xiv

Figure 13. Immunohistochemistry of p16, p53, and pRb in spontaneous feline oral squamous cell carcinomas...... 58

Figure 14. The average age of cats with FOSCC...... 59

Figure 15. Feline p16 predicted cDNA sequence for exons 1α and 2...... 60

Figure 16. SCCF2 and SCCF2-OPGv.1 cell morphology in vitro ...... 82

Figure 17. Relative mRNA levels of feline OPG, RANKL and RANKL: OPG ratio in

SCCF2-OPGv.1 and SCCF2-OPGv.2 cells compared to SCCF2 and SCCF2-pcDNA cells

...... 83

Figure 18. Murine calvaria co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells with and without zoledronic acid (Zol) treatment ...... 84

Figure 19. Bone area, bone mineral density of calvaria and calcium concentration of calvaria co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells ...... 85

Figure 20. SCCF2Luc and SCCF2-OPGv.1Luc in nude mice ...... 86

Figure 21. Average SCCF2Luc and SCCF2-OPGv.1Luc tumor volume in nude mice ... 87

Figure 22. Radiographs of maxillary bone from nude mice with SCCF2Luc and SCCF2-

OPGv.1Luc tumors (week 6) ...... 88

Figure 23. Histopathology of SCCF2Luc and SCCF2-OPGv.1Luc xenografts ...... 89

Figure 24. Urine calcium concentration from nude mice with SCCF2Luc and

SCCF2OPGv.1Luc tumors ...... 90

Figure 25. Relative feline TERT (fTERT) mRNA expression ...... 111

Figure 26. Telomerase activity in feline oral squamous cell carcinoma (FOSCC) cell lines and tumors and normal cat oral tissues ...... 112

Figure 27 Diagram of feline TERT (fTERT) gene, mRNA and alternative splice isoforms

...... 113

Figure 28. Full-length and splice variants of feline TERT in feline oral squamous cell carcinoma (FOSCC) cell lines and tumors and normal cat tissues ...... 114

xvi

Chapter 1: Animal Models of Head and Neck Squamous Cell Carcinoma

Abstract

Head and neck squamous cell carcinoma (HNSCC) is the most common oral cancer worldwide. Local bone invasion into the maxilla or mandible and metastasis to regional lymph nodes often result in a poor prognosis, decreased quality of life and shortened survival time for HNSCC patients. Poor response to treatment and clinical outcomes are the major concerns in this aggressive cancer. Multiple animal models have been developed to replicate spontaneous HNSCC and to investigate genetic alterations and novel therapeutic targets. This review provides an overview of HNSCC as well as the traditional animal models used in HNSCC preclinical research. The value and challenges of each in vivo model are discussed. Similarity between HNSCC in humans and cats and the possibility of using spontaneous feline oral squamous cell carcinoma (FOSCC) as a model for HNSCC in translational research are highlighted.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer in

Europe and the United States [1, 2] and is a common malignancy in developing countries including southern Asia [3]. In 2013, approximately 54,000 Americans were diagnosed with HNSCC and 11,500 died from this aggressive disease [4]. The aerodigestive tract, including the gingiva, tongue and larynx, is the most common site for HNSCC [2]. The five-year survival rate of patients with HNSCC is approximately 50% [1, 5]. The prognosis for patients with recurrent disease, bone invasion or metastasis is poor [5, 6].

HNSCC is more common in males than females [3, 7]. HNSCC usually occurs in older patients; however, the incidence of human papilloma virus-related oropharynx SCC in younger populations has been increasing [7]. More than half of the HNSCC patients were diagnosed at the progressive stage of disease (T3 or T4), in which cancers have invaded the adjacent bone or have metastasized to regional lymph nodes [8].

Tobacco use and alcohol consumption are the major risk factors for HNSCC worldwide, followed by high-risk human papillomavirus (HPV) infection [2, 7, 9].

High-risk HPVs, including HPV-16, -18, and -31 were detected in 19% of

HNSCC [7], most frequently in the oropharynx [10]. Fewer genomic alterations were reported in HPV-related HNSCC compared to HPV-negative HNSCC [9]. Even though there is a correlation between high intensity p16 immunostaining and a more favorable prognosis in patients with HPV-induced HNSCC, high p16 is not always indicative of a

favorable prognosis since high p16 expression is sometimes associated with HPV- negative HNSCC [10-13].

Local invasion of the submucosa and bone, in addition to lymph node metastasis, are common histological features of HNSCC [8, 12]. Different types of bone invasion have been found in HNSCC and include erosive, infiltrative and mixed patterns, which have different histological features and different rates of 3-year disease-free incidences

[5]. HNSCC with infiltrative bone invasion has a lower rate of 3-year disease-free status compared to the erosive bone invasion form of HNSCC.

Whole exome sequencing in tumors from 100 patients revealed the molecular background in human HNSCC. Mutation/inactivation of genes that play a role in cellular proliferation, including TP53, Rb, CDKN2A (p16, INK4A), and CCND1, as well as cell survival-regulated genes such as PIK3CA, PTEN, CASP8 and HRAS, was commonly observed. Expression of genes that are associated with adhesion and invasion, such as

TGF-β, SMAD and FAT1, and osteoclast regulation, including PTHrP, OPG, RANKL and

RANK, were also detected in HNSCC that had evidence of bone invasion and/or metastasis [1, 2, 14, 15]. TP53 is the most commonly mutated gene in the progressive stage of HNSCC in developed countries [1, 2], while inactivation of H-ras and amplification of Myc gene are most commonly detected in HNSCC in developing countries [3]. Furthermore, mutation in the TP53 gene was suggested to be frequently involved with the synergistic actions between tobacco smoke and alcohol consumption in the pathogenesis of HNSCC [7].

Even though mutations of EGFR are rare in HNSCC, amplification of this gene is common and correlated with poor prognosis [2]. The overexpression of EGFR was found in 30-90% of HNSCC patients [1, 2, 6, 7]. Interestingly, expression of EGFRvIII, a truncated isoform of EGFR that can be activated without an EGF ligand and causes tyrosine kinase inhibitor (RTK) resistance was reported in 42% of HNSCC [2]. HER-2

(C-erbB-2) expression, which was related to shorter overall survival, was found in 40-

60% of HNSCC [16]. EGFR overexpression was associated with decreased disease-free survival (DFS) and overall survival (OS) in HNSCC patients [6]. Multiple EGFR inhibitors have been developed, including antibodies that prevent ligand-binding and receptor dimerization, such as cetuximab, and small molecule inhibitors that target tyrosine kinase receptor (RTK), such as erotinib, gefitinib, lapatinib, and recently afatinib. However, only cetuximab has been approved by the FDA for HNSCC, which resulted in a minimal to modest response (13% partial regression; 33% stable disease;

38% progressive disease) depending on the study [6, 17, 18]. To date, a highly effective single or combination therapy has not been reported for clinical therapy of HNSCC [17].

In addition to genomic instability, the alteration of non-coding genes, especially microRNA (miRNA/miR) genes, occurs in HNSCC. MiRs regulate expression of many genes at the post-transcriptional level leading to reduced translation. These small non- coding RNAs have an important role in normal biology, cell development, and cancer

[19]. Deregulation of miRs has been found to correlate with the pathogenesis of multiple human cancers including HNSCC [19, 20]. MiR-21 has been suggested to function as an

oncogene in HNSCC because it was commonly up-regulated [21-23]. Up-regulation of miR-7, miR-155, miR-130b, miR-233 and miR-34b, as well as down-regulation of miR-

100, miR-99a, miR-125b and miR-375 were also reported in HNSCC [21, 23]. The applicability of these specific miRNAs as prognostic biomarkers and/or treatment targets of HNSCC is currently being investigated [20].

Various animal models have been developed for studying the pathogenesis, genetic backgrounds and novel therapeutic development in HNSCC. Although no single model can recapitulate all aspects of HNSCC, the information gained from animal models is essential for the advancement of treatment and diagnosis in HNSCC [24, 25].

In this review, the traditional animal models are classified according to their species, techniques, benefits and limitations. In addition, spontaneous feline oral squamous cell carcinoma (FOSCC) is highlighted as an alternative model for HNSCC preclinical investigations for potential application to human medicine. Table 1 shows a summary of the advantages and disadvantage of the animal models.

Hamster model of HNSCC

The hamster model is a time-honored animal model for investigation of HNSCC, particularly for research focusing on carcinogenesis and cancer chemoprevention [26-29].

Carcinogenic agents are applied directly to the mucosa of the buccal cavity of hamsters to induce tumorigenesis. Initially, the polycyclic hydrocarbon, 9,10-dimethyl-1,2-

benzanthracene (DMBA) was directly painted on to the mucosal cheek pouch 3 times/week and tumors developed locally in 16 weeks after induction (Figure 1). The histologic progression of the hamster cheek pouch model following DMBA exposure is demonstrated in figure 2, which closely mimics the progression of disease in humans.

The DMBA hamster tongue model, which uses mechanical trauma of the tongue and local DMBA application, was also developed. The hamster model of HNSCC, regardless of anatomic site, allows research on premalignant lesions and evaluation of chemical agents on carcinogenesis. Investigations on the interactions between host immunity, cancer, and the microenvironment are possible since the hamsters are not immunosuppressed; however; there is evidence to suggest that the cheek pouch is immunoprivileged (e.g., xenograft tumors have been successfully engrafted in the cheek pouch) and may not fully mimic the intact immune system of human patients [8].

Although tumor growth is easily monitored in vivo using this model, since humans do not have a cheek pouch casts some doubt on whether the hamster cheek pouch accurately mimics the pathobiology and mechanisms of spontaneous HNSCC in people [8, 30, 31].

Furthermore, the inconsistent ability of this model to develop metastatic tumors has raised some skepticism regarding the malignancy of these tumors [8].

Increased expression of EGF, TGF-α, H-ras and p53 in HNSCCs has been described in this model [30]. Similarly, increased expression of several receptor tyrosine kinases relevant to human HNSCC has been observed in the cheek pouch model of chemical carcinogenesis, including FGFR2 and FGFR3, and EGFR and ErbB2 [32].

Although increased expression of erbB3 mRNA has been demonstrated in human

HNSCC, no corresponding increase in erbB3 was observed in the carcinogenic process in the hamster cheek pouch. Hamster p16 protein has a biochemical structure similar to human p16 [33]. Approximately 70% of DMBA-induced tumors in the hamster cheek pouch have p16/CDKN2A alterations [34]. Evaluation of p16 levels using immunohistochemistry in this model revealed that loss of p16 is an early event during carcinogenesis, with reduced expression evident in dysplastic lesions, prior to the development of invasive neoplasia. Homozygous deletion of TP53 and CDKN2A was found in oral tumors of the Syrian golden hamster, rat and humans [35].

Rodent models of HNSCC

Murine models are widely used to investigate HNSCC. Models have been developed in both immunocompetent and immunosuppressed rodents.

Rat and mouse chemically-induced carcinogenesis model

The oil soluble agent DMBA failed to induce oral tumors in the rat; therefore, water soluble 4-NQO was applied by either directly painting the palate or tongue or adding it to the drinking water of rats or mice. Oral tumors developed in 2-6 months after addition of

4-NQO into the drinking water and 12-16 weeks following the direct painting of 4-NQO

(Figure 3) [30, 36]. The benefits and drawbacks of this model are similar to the hamster

cheek pouch model; however, addition of 4-NQO to drinking water is also associated with tumor formation in other regions of the gastrointestinal tract and the paws [8].

Carcinogens that stimulate malignant transformation and tumor formation can be used on genetically engineered rats in order to determine the contributions of different mutations to the pathogenesis of oral SCC. For example, this model showed that mutation of the adenomatous polyposis coli (APC) gene in rats increased inflammation and oral tumor formation in rats exposed to 4-NQO in their drinking water [37]. This model also has been used for studies of cancer detection methods. For example, the properties of diffusion reflection using bio-conjugated gold nanorods were used to detect squamous cell carcinoma of the rat tongue, and this technology might be useful for detection of residual disease intraoperatively in people [38]. A common feature of carcinogen-induced tumorigenesis in these models is the lengthy experimental protocol (20 and 37 weeks for the two previous examples) [37, 38]. Unfortunately, the development of invasive carcinoma is rare in this model and the common genetic alterations found in HNSCC are also infrequently observed [30]. Therefore, the chemically-induced carcinogenesis models do not recapitulate human HNSCC pathogenesis [30]. However, they are useful for studying the influence of carcinogens in oral oncogenesis. For example, a similar model in B6C3F1 lacI mice was used to demonstrate the role of tobacco in carcinogenesis of human HNSCC [39, 40].

Xenograft transplantation models of HNSCC

Subcutaneous xenografts

Human and animal HNSCC tissue or cell lines have been inoculated into the subcutis of immunocompromised mice, including athymic nude and severe combined immunodeficiency (SCID) mice. Subcutaneous xenografts have been used to investigate cancer drugs for treatment of HNSCC. A mouse xenograft tumor model was used to evaluate the efficacy of novel targeted dendrimer chemotherapy in reducing cellular toxicity of methotrexate in HNSCC [41]. Subcutaneous inoculation of high ALDH- expressing cells from UM-SCC-104 (the human papillomavirus-16-positive HNSCC cell line) into a non-obese diabetic/SCID (NOD/SCID) mouse revealed the presence of self- renewing and differentiation capabilities of cancer stem cells (CSCs) in this cell line [42].

The advantages of this model include a relatively short interval between tumor inoculation and tumor formation (in contrast to carcinogen induction) and convenient monitoring of tumor size. However, the deficiencies in host immunity, lack of local tissue invasion, and infrequent metastasis are important weaknesses of this model. In addition, the usefulness of this model to recapitulate human HNSCC has been questioned since the response of chemotherapeutic agents in xenograft tumors did not reflect drug responses in

HNSCC patients [8, 9, 30].

Orthotopic transplantation models of HNSCC

Human and animal tissue or cell lines have been injected into various locations of the head to model HNSCC. Locations have included the palate, masseter muscle [43, 44], tongue and the myelohyoid muscle, [8, 30], cavity of an extracted incisor, and the submucosa of the maxillary gingiva [45]. This approach is more appropriate for studies focusing on the pathogenesis or treatment of established and invasive OSCC, rather than tumorigenesis and cancer prevention. An advantage of this model is that tumors are produced in specific anatomic regions that more closely mimic HNSCC as it occurs in humans [46]. Since bone-invasive tumors frequently occur in HNSCC patients, the orthotopic animal models have been recognized as appropriate to recapitulate tumor invasion into oral bone [8, 9, 45, 47]. Transplanted orthotopic OSCC cells grow faster and require lower numbers of cells for tumor development in vivo than transplanted cells in subcutaneous xenograft models [48]. However, the number of orthotopic models for bone-invasive HNSCC is limited. Only a few human HNSCC cell lines, including BHY,

UMSCC, UMSCC11A, UMSCC11B, BICR31, BICR56 and HSC3, were found to invade bone in vivo [47]. Recently, a feline OSCC cell line (SCCF2) was developed that closely mimics bone-invasive OSCC when tumor cells were injected in the submucosa of the maxilla of nude mice (Figure 4) [45]. Bone resorption was associated with PTHrP expression in this feline model. Cancer stem cells (CSCs) and up-regulation of epithelial- to-mesenchymal transformation (EMT) genes were observed in tumors of the tongue of

SCID mice that were injected with a highly aggressive HNSCC cell line as well as

primary human HNSCC tissue samples [9]. An orthotopic model using luciferase- transfected OSCC cell lines was successfully developed for tumor growth and metastasis in vivo [49]. This technique allows researchers to visualize the effect of novel anti- tumorigenic agents on tumor growth, bone invasion and regional metastasis in vivo.

Injection of MDA-1986, a human aggressive OSCC cell line, directly into the tongue submucosa of nude mice developed metastases in 30% of lymph nodes and 20% of lungs

[48]. However, in general, metastases to regional lymph nodes are uncommon and the model lacks an interaction between tumor and host immunity in the immunocompromised rodents [8, 30].

Osteolytic calvaria mouse model of HNSCC

A novel bone invasive human HNSCC cell line (YD-39) originating from an OSCC with mandibular bone invasion was subcutaneously inoculated over the calvaria of nude mice.

Transplanted mice were found to develop tumors with prominent osteolytic bone lesions

[50]. The high rate of adjacent bone invasion and osteolysis developed by this model could be useful for studying pathogenesis of bone invasion and osteolysis in HNSCC.

Transgenic mouse models of HNSCC

Recently, genetically modified mouse models have been developed to investigate the role of specific genes in the pathogenesis of HNSCC. Both conventional transgenic and conditional/inducible genetically modified mouse models have been produced.

Conventional genetically modified models were produced by either inducing (knockin) or deleting (knockout) specific cancer genes. Transgenic mice expressing cyclin-D1 were generated using the Epstein-Barr virus (EBV) ED-L2 promoter in 1997 [9, 51, 52]. The roles of the EGFR, TP53 and CDK-4 genes in the pathogenesis of HNSCC were demonstrated using this model. Transgenic mice with bone-invasive oral cancer were produced in L2-cyclin-D1/p53+/- mice by crossing cyclin-D1 knockin and p53 heterozygous knockout mice (p53+/- mice) [51-54]. N, N-dibutylnitrosamine (DBN) was shown to induce oral carcinogenesis in heterozygous p53 knockout (p53+/-) mice [52,

55], as well as the metastatic phenotype of HNSCC that was observed in DMBA-induced mice heterozygous for the TP53 gene (B6;129S2-Trp53tm1Tyj) [56]. The conventional genetically modified mouse models have certain limitations due to uncontrolled transgene expression and non-oral tumor development [8, 30, 52].

Conditional genetically modified mouse models have used the Cre-LoxP technology and epithelial-specific promoters to generate inducible cancer systems [30,

52]. Cre, a bacteriophage P1 DNA recombinase, is used to catalyze the excision of DNA at loxP sites. Transgenes have been targeted for expression in basal epithelial cells of the oral cavity using specific promoters, such as keratin-5 or-14. The inducible agents, such as progesterone-responsive agent (PR), tetracycline-responsive element (Tet) or estrogen- receptor have been used to regulate transgene expression [30]. The role of the HPV-16

E7 oncogene in tumor initiation in transgenic mice as well as the role of ErbB2 (Her-2) oncogene in epithelial tumor development in k14-rtA/TetRE-ErbB2 transgenic mice was

reported [57, 58]. The inducible activation of K-ras in oral epithelium using conditional transgenic mice revealed the significance of K-ras in tumor initiation [59]. Invasive oral and cutaneous squamous carcinomas developed in some of the progeny of conditionally expressed K-ras mice, which emphasized the potential of K-ras alone to induce squamous cell carcinogenesis [60]. The conditional genetically modified mouse models of HNSCC may not mimic cancer initiation and progression as it occurs in human oral carcinogenesis, which requires the involvement of multiple genes [30, 52]. The microenvironment surrounding the induced tumors in mice is also different from spontaneous human HNSCC, since mutations occur in the tumor cells of spontaneous cancers, while transgenes are carried in both stromal and cancer cells of induced tumors

[61]. The use of genetically modified mice can be limited by cost, model development and complicated breeding protocols [52].

Domestic animal models of spontaneous HNSCC

Induced tumor models of HNSCC fail to recapitulate human HNSCC in many aspects, including the initiation period, genomic instability, tumor heterogeneity, tumor microenvironment, and host immunity-tumor cell interaction [25]. Spontaneous HNSCC in companion animals, especially cats and dogs, can serve as a useful natural model for

HNSCC. HNSCC is particularly common in aged cats, but is infrequent in dogs compared to cats [62].

Dog model of HNSCC

HNSCC is the second most common oral cancer following melanoma in dogs [63].

Canine HNSCC can be categorized into 2 groups, tonsillar and non-tonsillar, according to the different regions of origin and biological behaviors [64, 65]. While canine tonsillar squamous cell carcinomas (SCC) are rare and have aggressive local invasion with metastasis, non-tonsillar SCCs are more common with less invasion and rare metastasis.

[66]. Approximately 80% of dogs with non-tonsillar HNSCC have a favorable prognosis and only 19% develop metastasis [63]. Gingiva is the most common site for non-tonsillar

SCC, while cancers on the tongue occur less frequently [67]. Most dogs with non- tonsillar HNSCC have a good prognosis after surgical or radiation therapy [64]. Cisplatin in combination with piroxicam had anti-tumorigenic effects within accepted toxicity levels in dogs with HNSCC [66, 68]. Even though, there was no significant correlation between clinical stages, histological features, tumor locations, and the survival times in

33 dogs with HNSCC, the poorly differentiated subtype of HNSCC is suspected to have the progressive phenotype that mimics the infiltrative form of human HNSCC [63].

HNSCC without metastasis that occurs in dogs is recognized as a good model for photodynamic therapy (PDT) study. Eight of 11 dogs with HNSCC were cured after treatment with 0.3 mg kg-1 of pyropheophorbide-alpha-hexyl-ether (HPPH, Photochlor) as the photosensitizer in combination with surgery and irradiation [69]. Canine oral papillomaviruses (COPV) were suspected to play a role in canine HNSCC, especially in young dogs [70]. COPV DNAs were detected in 3 out of 29 canine HNSCC using PCR

as well as in situ hybridization (ISH) [70]. However, the evidence that COPV plays a role in the pathogenesis of HNSCC in dogs is incomplete [71]. COPV DNA, but not RNA, was found in 10% (3/33) canine HNSCC samples and it was suggested that COPV does not have a significant role in HNSCC cancer pathogenesis in dogs because the COPV

DNA could be unrelated to the cause of the cancer [72].

Cat model of HNSCC

SCC is the most common oral cancer in cats (Figures. 5 and 6). Ten percent of all tumors diagnosed in cats are oral tumors and 75% of feline oral cancers are SCC [25, 73-79].

Gingiva followed by sublingual and tongue were reported as the most commonly affected sites for feline HNSCC [78]. Cats and humans can develop lingual SCC, which is an unusual location for oral SCC in other species [5, 24, 67]. While SCCs are commonly found at ventral part of tongue in cats (Figure 6), human HNSCC are more frequently found in the lateral border of tongue [67]. Similar to humans, HNSCC frequently occurs in aged cats with no apparent breed or sex predisposition [76]. There is a high degree of biologic similarity between feline and non-viral-associated human HNSCC [25, 80]. Cats are usually presented for veterinary care of HNSCC at a late stage of disease and although available treatment options include surgery, chemotherapy, and radiation therapy, the advanced state of disease in many of these patients makes treatment challenging. The prognosis is grave for most cats and time to death or euthanasia is short

(2-6 months) [77]. Clinical trials in cats with spontaneous HNSCC is a potentially useful approach for translational cancer research and novel therapy development [24, 80, 81].

Risk factors

Cats sharing the same environment with their owners and their meticulous grooming habits may result in exposure to similar environmental chemical risk factors [24, 25, 75,

79, 80, 82]. Environmental tobacco smoke (ETS) exposure, feeding of canned food and papillomavirus infection were suspected as possible factors involved in feline HNSCC development [7, 75, 79, 83]. However, a significant association between ETS exposure and the increased risk of HNSCC occurrence in household cats was not found [83]. An association between p53 overexpression and ETS exposure was reported in feline

HNSCC biopsy samples [75]. Although papillomavirus DNA is frequently detected in feline cutaneous SCC and Bowenoid in situ cutaneous carcinomas [84-87], papillomavirus has not been detected frequently in feline HNSCC [86, 88, 89]. A high intensity of p16 immunostaining (an indicator of papillomavirus infection in human

HNSCC) was found in 7% (2/30) of feline HNSCC samples [90] and 14% (6/43) of feline HNSCC biopsies (Supsavhad et al., unpublished data). However, no papilloma viral capsid DNA was detected in the p16-positive samples. A subset of human HNSCC also has p16 overexpression without papillomavirus infection. The pathogenesis of nonviral-induced p16 up-regulation in both human and feline HNSCC still remains to be elucidated [10, 12].

Tumor biology

Similar histopathological features and biological behavior exist between human and feline HNSCC [24, 25, 77, 80, 91]. Spontaneous feline HNSCCs are heterogeneous tumors with natural tumor and host immune interactions [24]. Inflammation is common in the feline cancers. Metastases of feline HNSCC were found in mandibular lymph nodes in 13-31% of cases [76, 92]. Metastatic rate of HNSCC in this animal is comparable to human HNSCCs, where 15-20% of cases develop metastases [93]. The gingiva is the most common site for bone-invasive feline HNSCC (Figure 7) [78]. Bone invasion, lymph node metastasis and frequent spontaneous recurrence after therapy in feline HNSCC recapitulate the progressive behavior of human HNSCC [75, 76]. There is a high morbidity and a low survival rate in both humans (50% five-year survival rate) and cats (10% 1-year survival rate) with invasive HNSCC [5, 73, 78]. In addition, late-stage cancer with bone invasion at the time of diagnosis was frequently found in cats as well as in humans with HNSCC [8, 78]. The average survival time of feline HNSCC after diagnosis was very short because the majority of the cats were euthanized after diagnosis

[24].

Treatment and clinical outcomes

Multimodality treatment including surgery, radiation therapy and chemotherapy is the common strategy used for both cats and humans with HNSCC. However, there is a poor

response to the routine treatment in cats [77]. Few approved drugs exist for human

HNSCC [17]. Similarly, treatment of feline HNSCC is rarely curative in veterinary clinical practice [94]. Chemotherapeutic drugs routinely used in HNSCC include cisplatin and piroxicam, and novel therapeutic agents such as gefitinib and masitinib have also been investigated in cats [6, 24, 73, 79, 95]. Combination therapy with piroxicam and masitinib significantly inhibited cancer cell proliferation in vitro using feline SCC cells

[79]. Hypoxia in feline HNSCC has been evaluated using Cu-ATSM positron emission tomography/computer tomography in 12 cats and revealed that cat cancers had similar intratumoral hypoxia as that reported in human HNSCC [25].

The effect of an amino-bisphosphonate, zoledronic acid, was examined using the feline orthotopic model of bone-invasive HNSCC [45, 47]. Zoledronic acid is a third generation, nitrogen-containing bisphosphonate that is a potent inhibitor of osteoclastic bone resorption [96]. Zoledronic acid reduced HNSCC tumor growth and bone loss in vivo, but did not prevent tumor cell invasion into the maxilla. Interestingly, zoledronic acid was ineffective at inhibiting osteoclastic bone resorption at the tumor-bone interface, but was more effective on bone surfaces without tumor cell invasion. Zoledronic acid treatment of cats with bone-invasive HNSCC resulted in decreased vascular endothelial growth factor (VEGF) and C-telopeptide (CTx) serum concentration as well as slower tumor growth compared to untreated group [97]. Recently, the feasibility of using 1-2 mg/kg of the bisphosphonate, pamidronic acid, in cats with bone-invasive cancer for 3-4 weeks was reported without short-term toxicity issues [98]. However, the effectiveness of

this drug in tumor growth suppression or bone pain palliation in feline HNSCC has not been reported.

Genetic alterations

Comparative analysis of the cat genome with 6 mammals including humans revealed a high degree of similarity. There is a combined overall homology of 90% between 20,285 putative feline genes and the human genome and 201 feline miRNAs were homologous to human miRNAs [74]. Since the majority of HNSCCs have genetic instability, many of the genes that are altered in the pathogenesis of human HNSCC have also been investigated in cats with oral cancer.

The TP53 tumor suppressor gene has an important role in cell cycle regulation.

Somatic mutation of TP53 was detected in more than 50% of human HNSCCs and many of these mutations led to increased levels of the mutated p53 protein, which was associated with a poor clinical outcome [2, 99]. 17% (7/41) of feline HNSCC had high p53 staining intensity as detected by immunohistochemistry (Supsavhad et al., unpublished data). Inactivation of the CDKN2A gene, which encodes the p16 protein, was found in 7-30% of human HNSCCs [2]. Low intensity p16 immunostaining was found in

48% (108/225) of HNSCCs [100]. In cats, low intensity of p16 immunostaining was variably observed. Only 7% (2/30) of the feline HNSCCs had low intensity of p16 immunostaining as reported by Munday et al.,while low p16 immunostaining was found in 42% (18/43) of the cats in our recent study (Supsavhad et al., unpublished data). The

correlation between p16 expression and clinical outcome has not been investigated in feline HNSCC.

EGFR is a tyrosine kinase receptor and is important in various human cancers including HNSCC. EGFR is a therapeutic target of HNSCC because the majority of these cancers have EGFR overexpression [6]. There is a high degree of homology between feline and human EGFR, especially in the ATP binding pocket [82]. High EGFR expression levels in feline HNSCC were detected in vivo [79]. An inverse correlation between EGFR expression and overall survival time (OST) in humans with HNSCC was reported [6].The correlation between EGFR expression and overall survival time (OST) in cats with HNSCC is not clear. One study reported a trend for an inverse correlation between EGFR and OST, while another reported a trend for a positive correlation between EGFR and OST in feline HNSCC [101, 102]. RNA interference (RNAi) was used to target EGFR in combination with radiotherapy and caused an additive effect on inhibition of cancer cell proliferation in both human and cat HNSCC in vitro [82].

Consequently, several tyrosine kinase inhibitors have been developed [1, 2, 6, 7].

Gefitinib, one of the EGFR inhibitors, can reduce cell migration and proliferation of feline HNSCC in vitro [73, 82]. Resistance to tyrosine kinase inhibitors, especially gefinitib, has been reported in various cancers including HNSCC [73]. Gefinitib resistance was also induced in a feline HNSCC cell line, which was used as an in vitro model of tyrosine kinase inhibitor resistance [73].

Cancer stem cells (CSCs) have self-renewal capability and produce both proliferative and differentiated cells in cancers [9, 73]. EMT (epithelial to mesenchymal transformation) and EMT gene expression are characteristics of invasive epithelial cancer cells [9]. Overexpression of CSC-associated proteins, such as ALDH1 and CD133, and

EMT-related proteins, such as vimentin, Snail and fibronectin, were reported in highly invasive feline HNSCCs as well as an orthotopic mouse model of human HNSCC [9, 73].

Bone invasion in HNSCC is dependent on osteoclast activation and function

(Figure 8). PTHrP and the NFҡB ligands (RANKL/RANK and osteoprotegerin, OPG) are involved in human HNSCC [5, 103]. PTHrP contributes to bone resorption by inducing

RANKL expression in osteoblasts or stromal cells, which binds to its receptor (RANK) on osteoclast precursors or mature osteoclasts to increase osteoclast number and function

[103]. In addition, TGF-β released from bone during osteolysis and bone invasion was reported to not only increase tumor cell proliferation but also increase the production of

PTHrP [103]. In contrast, OPG is a soluble decoy receptor for RANKL and inhibits osteoclast development and activation [5, 15]. The ratio of RANKL to OPG is the major control mechanism for bone resorption in vivo [5, 14]. Similar to humans, high PTHrP expression was found in feline HNSCC and was associated with bone invasion and osteoclastogenesis in vivo and in vitro [45, 78]. Comparative alignment between feline and human PTHrP cDNA revealed a high similarity between the two species [77]. TGF-β was shown to induce PTHrP expression and inhibit OPG expression in bone-invasive feline HNSCC. This emphasizes the usefulness of the feline orthotopic model for

investigations on bone-invasive HNSCC [45]. Treating SCCF2 cells with EGFR ligands including amphiregulin and TGFα resulted in significantly increased PTHrP expression levels in a feline HNSCC bone-invasive cell line (our unpublished data). However, the

EGFR downstream signaling pathway that up-regulates PTHrP still needs to be elucidated.

Cyclooxygenase (COX) is a catalytic enzyme that converts arachidonic acid to prostanoids. Cox-2 expression is associated with both inflammation and carcinogenesis

[95]. High expression of Cox-2 was detected in both human and feline HNSCC [79, 95,

104]. High levels of Cox-2 immunoreactivity were observed in 18% (6/34) of FOSCCs compared to 75% in human HNSCCs [95]. The success of co-treatment with piroxicam

(Cox-2 inhibitor) and masitinib to inhibit cancer cell proliferation as well as decrease the level of Cox-2 expression was reported in both human and feline HNSCC cell lines [79].

Additionally, a related drug, meloxicam, reduced feline HNSCC tumor growth in an orthotopic nude mouse model of feline OSCC [105].

Conclusions

No single animal model is sufficient for investigation of the pathogenesis and treatment of HNSCC. Genetically modified mice are useful for in vivo models to interrogate the role of specific genes and genetic modifications in the pathogenesis of HNSCC.

Traditional carcinogenesis models, such as the hamster cheek pouch model, have benefits

in studying carcinogenesis and treatment prevention strategies. Mouse orthotopic models are well suited for investigations of bone-invasive HNSCC. The development and assessment of various novel therapies have relied on these traditional animal models.

Unfortunately, there is a high rate of failure in predicting clinical responses in human

HNSCC patients using rodent models. In addition, rodent models are not effective at inducing robust lymph node metastasis. There are many pathophysiologic and genetic similarities between HNSCC in both humans and cats. This suggests that spontaneous

HNSCC in cats is a potential valuable model for translational research on novel drug, genetic, nanoparticle, and interventional therapies using clinical trials of client-owned cats in the clinical veterinary setting.

Animal Model Advantages Disadvantages Develop premalignant lesions Time consuming Can be used for Laborious chemoprevention studies Hamster Carcinogenesis Chemical hazards Host-tumor cell interaction is Cheek pouch does not exist in humans conserved Site may be immunocompromised Immunocompetent host

Time consuming Develop premalignant lesions Laborious Murine Carcinogenesis Host-tumor cell interaction is Chemical hazards conserved Metastasis and bone invasion are rare

Metastasis and bone invasion are rare Time saving Lack of host immunity-tumor cell Subcutaneous Multiple cell lines available xenografts interaction from humans, rodents, and cats Tumors formed in ectopic sites

Develop tumors in oral region Lack of host immunity-tumor cell Orthotopic xenografts Develop bone-invasive cancer interaction

Gene expression can be Excessive gene expression Conventional manipulated Non-oral tumors may develop genetically modified Useful for study of genetic Time consuming mice alterations Limited accessibility

Continued

Table 1. Summary of advantages and disadvantages of animal models for head and neck squamous cell carcinoma (HNSCC) research

Table 1 continued

Animal Model Advantages Disadvantages Gene expression can be Excessive gene expression

Conditional manipulated Does not mimic multiple gene

genetically modified Develop oral tumors alterations in spontaneous cancers

mice Useful for study of genetic Time consuming

alterations Limited accessibility

Similar to spontaneous HNSCC

Common in aged cats

Useful for clinical trials of novel Species-specific drug metabolism and therapies Cat Spontaneous solubility issues Bone invasion is common Reagent availability Cancers occur in oral cavity and Clinical trial approval needed pharynx

Metastases to regional lymph

nodes

Uncommon

Useful for clinical trials of novel Species-specific drug metabolism and

Dog Spontaneous therapies solubility issues

Cancers occur in oral cavity Reagent availability

Metastasis and bone invasion are rare

Figure 1. Hamster cheek pouch HNSCC model

Hamster cheek pouch HNSCC model. Hamster was exposed to 0.2% DMBA for 6 weeks in the cheek pouch and sacrificed at 12 weeks. Multiple squamous cell carcinomas (arrowheads) and a papilloma (arrow) were present in the cheek pouch. Reprinted with permission from Warner et al., Chemoprevention of Oral Cancer by Topical Application of Black Raspberries on High At-Risk Mucosa, Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, 2014.

Figure 2. Histologic progression in the hamster cheek pouch model following DMBA exposure

Histologic progression in the hamster cheek pouch model following DMBA exposure. Reprinted with permission from Warner et al., “Chemoprevention of Oral Cancer by Topical Application of Black Raspberries on High At-Risk Mucosa,” Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, 2014.

Figure 3. Rat tongue HNSCC model

4NQO (20 ppm) was added to the drinking water for 14 weeks and the rat was sacrificed at 20 weeks. 4NQO (20 ppm) was added to the drinking water for 14 weeks and the rat was sacrificed at 20 weeks. A large squamous cell carcinoma (4x6 mm) is present on the right side dorsal surface of the tongue with several smaller peripheral lesions. Reprinted with permission from Warner et al., “Chemoprevention of Oral Cancer by Topical Application of Black Raspberries on High At-Risk Mucosa,” Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, 2014.

Figure 4. Nude mouse model of invasive feline HNSCC

Nude mouse model of invasive feline HNSCC. Feline SCCF2-luciferase cells were injected into the maxillary gingiva. The SCCF2 cells form an invasive carcinoma that leads to bone resorption of the maxilla and invasion into the nasal cavity. Note that the incisor is spared. Tumor cell growth can be monitored by in vivo bioluminescent imaging (inset). NB=new bone formation. I=incisor.

Figure 5. Cat with late-stage HNSCC of the left maxilla with extensive growth and invasion of the maxillary bone (arrowheads). Scale is 5 mm

Figure 6. Cross-section of a cat tongue with SCC that originated in the ventral regions with marked invasion of the muscularis (arrowheads). Scale is 5 mm

Figure 7. Cat with gingival HNSCC of the anterior mandible

The squamous cell carcinoma invaded the gingiva and mandibular bone causing bone lysis around the incisor and anterior molar (radiograph top panel; histopathology lower panel).

Figure 8. Pathogenesis of bone invasion by HNSCC

Tumor-derived cytokines, such as parathyroid hormone-related protein (PTHrP), bind to bone stromal cells or osteoblasts to induce the formation of osteoclasts and activate the function of osteoclasts. The common mediator of osteoclast function produced by osteoblasts is RANK ligand, which binds to its receptor, RANK, on osteoclasts. HNSCC is known to produce PTHrP and RANK ligand. Growth factors, such as TGF-β and Ca2+ released from resorbing bone further enhance PTHrP and cytokine production by the cancer cells resulting in a positive feedback loop (‘vicious cycle’) of cytokine production and bone resorption.

Chapter 2: p16, pRb, and p53 in Feline Oral Squamous Cell Carcinoma

Abstract

Feline oral squamous cell carcinoma (FOSCC) is a highly aggressive head and neck cancer in cats with an unknown molecular pathogenesis. In this study, p16, p53, and pRb proteins were detected and quantified by immunohistochemistry in forty-three FOSCC primary tumors and three FOSCC xenografts. p16 mRNA levels were also measured in three FOSCC cell lines (SCCF1, F2, and F3), which were consistent with their p16 immunoreactivity. Feline SCCF1 cells had very high levels of p16 protein and mRNA

(55-fold greater) compared to SCCF2 and F3. A partial feline p16 cDNA sequence was amplified and sequenced. The average age of cats with FOSCC with high p16 immunoreactivity was significantly lower than the average age in the low p16 group.

Eighteen of 43 (42%) FOSCCs had low p16 intensity, while 6/43 (14%) had high p16 immunoreactivity. Feline papillomavirus L1 (major capsid) DNA was not detected in the

SCC cell lines or the FOSCCs with high p16 immunostaining. Five of 6 (83%) of the high p16 FOSCC had low p53, but only 1/6 (17%) had low pRb immunoreactivity. In

summary, the staining pattern of p16, p53, and pRb in FOSCC was different from human head and neck squamous cell carcinoma and feline cutaneous squamous cell carcinoma.

The majority of FOSCCs have low p16 immunostaining intensity, therefore, inactivation of CDKN2A is suspected to play a role in the pathogenesis of FOSCC. A subset of

FOSCCs had increased p16 protein, which supports an alternate pathogenesis of cancer in these cats.

Introduction

Feline oral squamous cell carcinoma (FOSCC) is one of the most common cancers in cats

[94]. Over 70% of FOSCC invade the adjacent mandible or maxilla, resulting in bone lysis and a short survival time (less than 1 year) [78]. Pathogenesis of this cancer is not well-understood and few treatment options are available with limited success.

FOSCC and human head and neck cancer (HNSCC) have a high degree of biological similarity including location of tumors and bone-invasive behavior [25, 80]. In humans,

HNSCC can be classified in two major categories; human papillomavirus (HPV)-induced and non-viral-associated HNSCC [106]. A favorable prognosis as well as good treatment response was commonly reported in patients with the HPV HNSCC, while most patients with non-viral-associated HNSCC have a poor prognosis. Inactivation or mutation of tumor suppressor genes, especially CDKN2A and TP53, were reported in more than 60% of non-viral-associated HNSCC patients [106].

TP53 is a tumor suppressor gene that encodes p53 protein and has an important role in regulating cell cycle and tumor progression. More than 50% of HNSCC patients were reported to have either TP53 mutations or increased p53 protein (usually due to mutations) detected by immunohistochemistry (IHC), which were associated with a poor clinical outcome and invasive tumor progression [99, 106]. High p53 immunostaining was reported in 43% (10/23) of FOSCCs, while 35% (8/23) of FOSCCs did not have high p53 staining [107].

Low p16 protein immunoreactivity was found in 48-83% of human HNSCCs

[100, 108, 109], while high p16 immunoreactivity was found in 37-59% of human

HNSCC [108, 110]. Loss of p16 protein immunoreactivity in human HNSCC correlated with CDKN2A inactivation, especially point mutations, loss of heterozygosity, promoter hypermethylation, or homozygous deletion [106, 109]. In contrast, increased p16 immunoreactivity correlated with high-risk-human papilloma virus (HPV)-induced

HNSCC [111].

Papillomavirus (PV) is a small, double-stranded circular DNA virus. Its genome has non- coding upstream, early, and late regions. The early region consists of 8 genes (E1 to E8).

In high-risk human PVs (HPV), including HPV-16,-18 and -31, E6 and E7 are oncogenic and cause degradation of p53 and pRb during the cell cycle [112]. An association between the presence of feline catus papillomavirus type 2 (FcaPV2) DNA and p16 and pRb immunostaining in preneoplastic and neoplastic cutaneous SCC lesions has been reported in cats [90]. Increased p16 with decreased pRb was suggested to be a marker for

FcaPV in feline cutaneous SCC [90]. However, the presence of papillomavirus and the relationship between p16, p53, and pRb and papillomavirus infection in FOSCC are still unclear. Moreover, FOSCC subtypes have not been classified in this species.

We hypothesized that FOSCC could be classified in a manner similar to HNSCC using the immunohistochemical (IHC) pattern and intensity of p16, pRb, and p53 staining. High p16 with a low pRb and p53 immunostaining profile was expected in PV- related FOSCC, while low p16 and either a high or low p53 IHC pattern were proposed in non-PV FOSCC. The goals of our study were to investigate the 1) pattern of immunoreactivity to p16, pRb, and p53 in FOSCC; 2) determine the presence of papillomavirus DNA in 3 FOSCC cell lines and formalin-fixed, paraffin-embedded

(FFPE) FOSCC samples with high p16 immunostaining intensity; 3) correlate p16 immunoreactivity and mRNA expression in FOSCC cell lines; and 4) sequence a region of feline p16 cDNA.

Materials and Methods

Sample collection

Forty-three FFPE FOSCC specimens originating from the gingiva (n=9), tongue (n=11), sublingual region (n=12), and mandible or maxilla (n=11) were obtained. Three FOSCC cell lines (SCCF1, SCCF2 and SCCF3) previously developed by our laboratory [45, 81] and xenograft tissues from nude mice with SCCF1, 2, and 3 were also used [45]. Since

positive p16 immunostaining has been reported in feline Bowenoid in situ carcinoma

(FBISC) [84], a FBISC was used as a positive control for p16. One normal cat gingiva was used for the qRT-PCR, and two normal cat gingiva FFPE samples were used to compare to the FOSCC IHC results.

Immunostaining

Forty-three FOSCC FFPE, 3 FOSCC xenografts, 1 FBISC and 2 cat gingival tissues were cut into 5 µm sections, deparaffinized, rehydrated, and incubated in antigen retrieval solution (Dako, Carpentaria, CA) at 80°C for 40 min. and cooled to room temperature for

10 min. Hydrogen peroxide (3%) was used to block endogenous peroxidase and serum- free protein block (Dako) was applied. Samples were incubated overnight at 4°C with either 1) mouse anti-human p16 monoclonal antibody [84] (BD Bioscience, San Jose,

CA; clone G175-405, 550834; 1:30) in phosphate-buffered saline (PBS); 2) mouse anti- human non-phosphorylated retinoblastoma (pRb) antibody [84] (BD Bioscience, clone

G3-245, 554136; 1:40) in PBS; or 3) mouse anti-human p53 antibody [84] (BD

Bioscience, clone PAb 240, 554166; 1:100) in PBS. The anti-pRb antibody was specific for nuclear pRb. Anti-mouse monoclonal secondary antibody (Vector Laboratories, BA-

9200, Burlingame, CA) was used at 1:200 in PBS for 30 min. Slides were stained with

Vector ABC Elite complex (Dako) for 30 min., diaminobenzidine tetrahydrochloride

(Dako) for 5 min., and with Mayer’s hematoxylin (Thermo Scientific, Fremont, CA) for 1 min.

The highest intensity pRb and p53 immunostained samples were used as positive controls, while omission of the primary antibody was used as a negative control. Normal tissue surrounding each tumor served as an internal control. IHC staining of p16, pRb, and p53 was classified into one of 3 categories: 1) absent to low; 2) moderate; and 3) high immunoreactivity, according to the average percentage of immunostained cells and the intensity level of the immunostaining in 5 different 400x fields (Table 2).

p16 mRNA in 3 FOSCC cell lines

RNA was purified from SCCF1, SCCF2, and SCCF3 cells and a normal cat gingiva using the Absolutely RNA RT-PCR miniprep kit (Agilent Technologies, Santa Clara, CA).

Normal cat gingiva was collected from a cat euthanized for other reasons, snap frozen in liquid nitrogen, and stored at -80°C.

The quantitative RT-PCR (qRT-PCR) conditions have previously been described [113].

A consensus p16 cDNA sequence (Figure 9) of exons 1α (126 nucleotides) and 2 (307 nucleotides) was obtained by aligning p16 cDNAs from multiple species including human, chimpanzee, rabbit, cow and mouse and comparing to the updated cat genome sequence (Genbank accession NM 000077.4) from the National Center for Biotechnology

Information (NCBI) database. Feline p16 exon 3 was not included in the consensus sequence, as the genomic DNA sequence for exon 3 remains questionable due to

disagreements in the alignment of exon 3 between species. Therefore, sequences of exons

1α and 2 were used for primer design. Primers Fp16m2S (5’-

GCCTGGGTCGGAGCCCGATT-3’) and Fp162AS (5’-TGCAGCACCACCAGCGTGTC-3’) were designed using NCBI Primer-Blast software to specifically detect a 166-nucleotide amplicon spanning the intron between exons 1α and 2 of the p16 gene. The amplicon was partially sequenced and included 95 nucleotides. The amplicon and the forward primer

(total of 115 nucleotides) were aligned with p16 cDNA sequences from several species, which confirmed that this region of the predicted cDNA sequence and the junction of exons 1α and 2 are correct (Figure 9). Feline beta-2 microglobulin (forward primer: 5’-

CTACTTCTGGCGCTGCTCTG-3’ and reverse primer: 5’-CCTGAACCTTTGGAGAATGC-3’) was used as a control housekeeping gene in this experiment.

Papillomavirus L1 major capsid DNA investigation

The presence of papillomavirus L1 (major capsid) DNA was investigated in SCCF1, F2, and F3 cells as well as all FFPE FOSCC samples that had high p16 immunostaining.

FBISC was used as a positive control for FcaPV2 DNA. Liver, spleen, kidney, lymph node, and lung tissues from a normal cat were used as negative controls. DNA was isolated using the DNeasy blood and tissue kit (Qiagen, Valencia, CA) and quantified using a spectrophotometer (NanoVue, 4282 V2.0.0, Fisher Scientific, Pittsburgh, PA).

Four published primer pairs including JMPF/R [89] (specific for L1-FcaPV2),

MY09/11[114] (consensus primers of the L1 region in many PVs), and JMY2F/R [85]

and JMY3F/R [85] (specific primers for L1-PV DNA originally amplified from a feline cutaneous SCC) were used to amplify papillomavirus L1 capsid DNA.

PCR product purification and sequencing

PCR products were separated on 2% agarose gels and purified using a QIA quick gel extraction kit (Qiagen) and sequenced. Sequences were evaluated using the NCBI

BLAST and compared to known sequences in the NCBI database.

Statistical analysis

A one-way ANOVA with a Tukey’s post-hoc analysis for multiple comparisons was used to examine the difference 1) between ages of cats with FOSCCs in each immunohistochemistry category and 2) p16 mRNA expression in SCCF1, F2, and F3 cells and normal cat gingiva. An unpaired t-test with Welch’s correction was used to compare p16 mRNA levels between SCCF3 cells and normal gingiva. For each IHC antibody, the percentage of cases within each scoring criterion and their 95% confidence intervals (CI) were calculated. No overlap between the CIs indicates a statistically significant difference between groups at a significance threshold of =0.05. Therefore, statistical significance was established at p < 0.05 for all comparisons.

Results

Relative p16 mRNA quantification in 3 FOSCC cell lines

SCCF1 had significantly greater p16 mRNA levels compared to SCCF2, SCCF3 and normal gingiva (p <0.0001) (Figure 10).

Immunohistochemistry (IHC)

Three IHC intensity groups were classified according to the percentage of positive cells and staining intensity as defined in Table 2. There was minimal to no staining variation in different regions of tumors for all antigens. Two samples evaluated for p16 expression were excluded from the pRb and p53 IHC due to insufficient tissue. The intensity of immunoreactivity was categorized as absent to low, moderate or high (Figures. 11A-11C,

12A-12C and 12E-12G; Table 3). The p16 immunostaining was present in both the nucleus and cytoplasm of the tumor cells. It was found that 18/43 (42%) of the FOSCC had low p16 IHC expression. High p16 immunostaining was present in 6/43 (14%) of the

FOSCC. This was significantly lower than the number of tumors with moderate p16 immunostaining (19/43, 44%) (Figure 13). Nuclear immunoreactivity was used to determine the presence of pRb. Fourteen of 41 (34%) FOSCCs had low to absent pRb staining intensity. In contrast, high intensity of pRb was found in 5/41 (12%) of samples.

Most FOSCCs had moderate intensity of pRb staining (22/41, 54%; greater incidence, p<0.05, compared to low or high intensity) (Figure 13). p53 IHC staining was also

limited to the nuclei. High p53 staining intensity was present in 7/41 (17%) of the samples, while increased p53 immunostaining was not observed in 34/41 (83%) FOSCCs

(Figure 13).

The histopathological characteristics of FOSCC in all of the different IHC groups were similar. The epithelium of normal cat gingiva and tongue had moderate staining intensity for p16 and a low to moderate intensity for pRb and p53 immunostaining (Figures. 11D,

12D and 12H). SCCF1 xenograft had a high intensity of p16 immunostaining in both the nucleus and cytoplasm of the neoplastic cells (Table 4). In contrast, p16 immunostaining was absent in both SCCF2 and SCCF3 xenografts (Figures. 11E-11G).

The average age of FOSCC cats with high intensity p16 staining (mean=10.6 years, SD =

4.0) was significantly lower than cats in the low intensity group (mean= 14.6 years, SD =

4.3), (p < 0.05) (Figure 14).

Papillomavirus L1 capsid DNA amplification

PV-L1 capsid DNA was not amplified in the SCCF1, F2, or F3 cell lines or FOSCC samples with high p16 immunostaining intensity, but it was amplified in the FBISC positive control using the JMPF/R primers.

Sequencing

A consensus p16 cDNA sequence was obtained by aligning p16 cDNAs from multiple species and comparing them to the updated cat genomic sequence (Genbank accession

NM000077.4) from the NCBI database. Using BLAST software, a predicted feline p16 cDNA sequence was generated (Figure 15). RT-PCR products amplified in the SCCF1 cell line and normal cat gingiva were sequenced and aligned to the predicted feline p16 cDNA, which confirmed the predicted splice site for exons 1α and 2. The RT-PCR sequences had 100% homology with the predicted p16/CDKN2A cDNA of Felis catus.

Although the full-length p16 cDNA could not be amplified due to the difficulty in designing functional RT-PCR primers for the cat and inadequate sequence information for the very short open reading frame of exon 3, a partial p16 cDNA sequence (Genbank accession: Banklt1884793 Seq1 KU508421) was amplified, confirmed, and found useful to measure p16 mRNA in feline cells and tissues (Figure 15).

Discussion

High intensity p16 and low intensity p53 and pRb immunostaining correlated with a subset of high-risk- HPV-induced HNSCC in humans [108]. In contrast, low intensity p16 immunostaining was reported in non-viral-associated HNSCC, which had either mutation or inactivation of CDKN2A [109, 115]. For this reason, p16 IHC serves as a useful tool for HNSCC classification and prognosis in humans.

The number of FOSCC with high intensity p16 staining found in this study (6/43, 14%) was greater than in a previous study (2/30, 7%) [90]. Based on HNSCC in humans, we hypothesized that the majority of FOSCCs would have either low or high p16. However, the results of the present study show that there were significant differences between cats and humans with oral cancer. Most FOSCC in cats had a moderate staining intensity of p16 and smaller subsets had high or low staining intensity. In humans, high intensity p16 staining was associated with high-risk HPV infection and a better prognosis, while loss of p16 immunostaining was associated with CDKN2A inactivation and a worse clinical outcome [99, 106]. However, the connection between p16 immunostaining and prognosis in FOSCC has not been reported, therefore, the clinical outcome of FOSCC with different p16 staining intensity should be investigated further.

Interestingly, cats with FOSCC and abundant p16 protein were significantly younger than cats with FOSCC and low levels of p16. In humans, HPV was suspected to be involved in the pathogenesis of HNSCC in young patients [116, 117]. Therefore, we speculate that the pathogenesis of FOSCC in younger cats with high p16 staining intensity may differ from FOSCC in older cats with low p16. It is also possible that FOSCC with high p16 intensity may be associated with a feline papillomavirus infection that has not been discovered yet. We were unable to amplify papillomavirus L1 capsid DNA from the

SCCF1 cell line and FOSCC FFPE samples that had high intensity of p16 staining. The combined data from two studies showed that papillomavirus capsid DNA was only amplified in 3 of 47 FOSCC samples (p16 expression was not measured in these two

studies) [86, 118]. In another study, only two of thirty FOSCC samples had high p16 immunoreactivity, but FcaPV was not detected in these two specimens [90]. Therefore, papillomavirus infection may only be involved rarely in the pathogenesis of FOSCC.

However, these relatively small sample numbers may be inadequate to draw a definitive conclusion on the role of papillomavirus in FOSCC. All primer pairs used to amplify papillomavirus DNA in these experiments were designed against the L1-regions of

FcaPV2 and the conserved sequences from an alignment between different types of PVs.

These primers may not detect other PVs that possibly exist in the FOSCC samples.

Low p53 and pRb were found in a subset of high risk-HPV-induced HNSCCs that had high p16 [108]. The interaction between the high-risk human papillomavirus E6 protein and p53 causes degradation of p53, while the interaction between E7 protein and pRb results in degradation of pRb and increases free p16 [112, 119]. A lack of pRb correlated with high p16 and FcaPV2 infection in feline cutaneous squamous cell carcinoma [84]. In this study, high p16 immunoreactivity was associated with low p53 in 5 out of 6 samples; however, only 1 of 6 FOSCC with high p16 had low pRb. Instead, FOSCC samples with high p16 were more likely to have moderate pRb (4 out of 6 samples). Therefore, the association between p16, p53, and pRb in FOSCC is unclear. Other proteins may be involved in up-regulation of p16 in FOSCC or the relationship between pRb and p16 in

FOSCC may be different from HNSCC in humans.

We found that forty-two percent of FOSCC had low p16. These figures were greater than a previous study with 2/30 (7%) of cats having low p16 [90]. These differences may be due to regional differences and location of the FOSCC tissues within the oral cavity, as well as larger sample size, which might explain the greater proportion of low p16 that was found in the present study. The most common tumor location in the previous study was gingiva (12/30), while in this present study there was an even proportion between gingiva (9), tongue (11), sublingual (12) and mandible or maxilla (11). In addition,

FOSCC in cats from different countries may have a distinct pathogenesis and protein expression. Loss of p16 immunostaining correlated with inactivation of the CDKN2A in

HNSCC [109]. Hypermethylation of a CpG island in the CDKN2A promoter region and homozygous deletion mutation of p16 has been reported in various cancers, particularly

HNSCC [106, 109, 120, 121]. It is possible that many FOSCCs are associated with mutation or inactivation of the CDKN2A tumor suppressor gene. Unfortunately, there are still gaps in the feline whole genome sequence in the NCBI database and the genomic sequences surrounding the CDKN2A exons remain undefined. Completion of the feline p16 gene sequence will be essential for evaluating inactivation of the CDKN2A in

FOSCC.

We found that p16 mRNA levels in SCCF1, SCCF2, and SCCF3 cell lines were consistent with p16 immunostaining in the xenograft samples. In particular, a high level of p16 mRNA expression was present in the SCCF1 cell line, which also showed a high p16 staining intensity.

Low p16 has been observed more often in the late stages of HNSCC (T3 or T4) [100], which has a less favorable prognosis compared to HPV-induced-HNSCC [99]. p16 protein expression was suggested to possess prognostic utility in squamous cell carcinoma of the nasal planum in cats, since cats with high p16 had a longer survival compared to those with loss of p16 [122]. In addition, transfection of wild-type p16 into

HNSCC cell lines decreased cancer cell proliferation in vitro [121, 123].

Increased p53 is the predictor of TP53 mutation in various cancers including

HNSCC [124]. Seventeen percent (7/41) of FOSCC samples had a high intensity of p53 immunostaining. Therefore, somatic mutation of TP53 with increased expression and decreased activity of p53 may occur in some FOSCCs. Increased p53 immunostaining was not observed in 83% of FOSCCs. Lack of or low p53 immunostaining can be found in cells with TP53 inactivation as well as normal p53 expression. However, a subset of

FOSCCs with low p53 immunostaining might be associated with TP53 inactivation, since

24-53% of HNSCC in humans with low p53 immunostaining were related to TP53 mutations [124, 125]. Twenty-seven percent (11/41) of FOSCC samples in this study had low p53 and p16 staining. Since CDKN2A encodes both the p16 and p14 proteins, inactivation of CDKN2A may also lead to loss of p14 if one of the common exons is inactivated or deleted. p14 stabilizes p53 by preventing the association between MDM2 and p53. Without p14, MDM2 independently binds and targets p53 protein for proteosomal degradation [126]. There is still no evidence that CDKN2A inactivation causes both p16 and p14 loss in HNSCC. However, the loss of both p16 and p14 IHC

was found in 30% (9/30) of non-small cell lung cancers (NSCLC) and 67% (19/28) of pancreatic cancers [127]. Inactivation of CDKN2A is a potential mechanism to indirectly inhibit p53 protein expression. PV-related FOSCC was suspected in 12% (5/41) of samples that had low p53 and high p16 immunostaining intensity. However, papillomavirus capsid DNA was not amplified in these samples. Our data suggests that inactivation of the TP53 or CDKN2A may occur in FOSCC to directly or indirectly silence p53 expression. Inactivation of CDKN2A, as well as TP53, in the pathogenesis of

FOSCC will require further investigation.

Conclusions

The immunohistochemical staining patterns for p16, Rb, and p53 have been described for

FOSCC and provide important information on the potential pathogenesis and future improvements in the diagnosis of FOSCC. The immunohistochemistry patterns of p16, p53, and pRb in FOSCC were different from human HNSCC and feline cutaneous SCC.

The p16 immunostaining intensity was consistent with relative p16 mRNA levels in three

FOSCC cell lines. Partial sequence analysis of p16 DNA was completed and shown to be identical for a FOSCC with high p16 and normal feline gingiva. There is still limited evidence that papillomavirus plays a role in FOSCC, but it should not be excluded as a potential pathogenic mechanism. It is likely that subclassifications of oral SCC in cats

based on tumor suppressor gene expression will guide research investigations and prognosis and treatment of cats with OSCC in the future.

Protein Absent to Low Moderate High p16 <20% of cancer 20-50% of cancer cells >50% of cancer cells have low have low to cells have high cytoplasmic intermediate cytoplasmic and intensity cytoplasmic and nuclear nuclear staining staining intensity pRb and p53 <20% of cancer 20-50% of cancer cells >50% of cancer cells have low to have high nuclear cells have high intermediate staining intensity, or nuclear staining nuclear staining >50% of cancer cells intensity intensity have low to intermediate nuclear staining intensity Table 2. Classification of p16, pRb, and p53 immunohistochemistry staining

Signalment IHC Results No. Breed Sex Age (Year) Location p16 pRb p53 1 DSH FS 14 Gingiva Low Moderate Low 2 DSL MC 20 Gingiva Low Low N/A 3 DSH F 11 Gingiva High High Low 4 Himalayan FS 16 Gingiva Moderate Moderate Low 5 DSH MC 15 Gingiva Moderate Moderate Moderate 6 DSH FS N/A Gingiva Moderate Moderate Low 7 Himalayan MC 12 Gingiva Low High Low 8 DSH MC 11 Gingiva Moderate Moderate High 9 DSH MC 17 Gingiva Moderate Moderate Low 10 DSH MC 13 Tongue Low High High 11 Siamese FS 13 Tongue Low Low Moderate 12 DSH MC 16 Tongue Low Moderate Low 13 DLH MC 11 Tongue Moderate Low Low 14 DSH FS 13 Tongue Low Low Low 15 Exotic SH F 10 Tongue Moderate N/A Low 16 DSH MC 11.5 Tongue Moderate Moderate Low 17 DLH F 10 Tongue Moderate Moderate Low 18 DSH MC 6 Tongue Low Low High 19 DSH FS 15.5 Tongue Low Low High 20 DSH F 10 Tongue Moderate Moderate Low 21 DSH F 12 Sublingual Low Low Moderate 22 Angora FS 7 Sublingual Low Moderate Low 23 DSH MC 4 Sublingual Moderate Low Low 24 DSH M 14+ Sublingual Moderate Low High 25 Persian FS 13.5 Sublingual High Moderate Low 26 DSH MC 12 Sublingual Moderate Moderate Low 27 DSH M 12.5 Sublingual Low Low Low 28 DLS FS 17 Sublingual Low High High 29 DSH FS 19 Sublingual Low Moderate Low 30 DSH FS 14 Sublingual Moderate Moderate Low 31 DLH MC 17 Sublingual Low Moderate Low 32 DSH FS 19 Sublingual Low N/A Low 33 DSH MC 13 Mandi or Maxi High Moderate High 34 DSH MC 12 Mandi or Maxi High Low Low 35 DLH F 6.5 Mandi or Maxi Moderate Moderate Low 36 DSH MC 12 Mandi or Maxi High Moderate Low 37 DSH F 16 Mandi or Maxi Low Low Low 38 DSH F 20 Mandi or Maxi Low Low Low 39 DSH FS 2 Mandi or Maxi High Moderate Low 40 DLH MC 14 Mandi or Maxi Moderate Low Low 41 Himalayan F 11 Mandi or Maxi Moderate Moderate Low 42 DSH MC 13 Mandi or Maxi Moderate Moderate Low 43 DSH FS 9 Mandi or Maxi Moderate High N/A Table 3. Signalment and immunohistochemistry data from 43 cats with FOSCC

DSH: Domestic short hair, DLH: Domestic long hair, F: Female, FS: Spayed female, M: Male, MC: Castrated male, N/A: Not applicable, Mandi or Maxi: Mandible or Maxilla. 50

Sample p16 IHC p53 IHC pRB IHC

SCCF1 High Low Low SCCF2 Low Low Low SCCF3 Low Moderate Low FBISC High Low Low Table 4 Immunoreactivity to p16, p53, and pRb in Three FOSCC Xenograft Tumors and Primary Feline Bowenoid in situ Cutaneous Squamous Carcinoma (FBISC)

Figure 9. Alignment of human, chimpanzee, rabbit, cat, cow and mouse p16 cDNA sequences.

The 115-nucleotide RT-PCR amplicon spanning the junction between exons 1α and 2 of the cat p16 cDNA was aligned with the corresponding regions of p16 from the indicated species. The dot indicates the first nucleotide of exon 2. An asterisk (*) indicates a nucleotide that is identical between all the aligned sequences, a colon (:) indicates a nucleotide of strongly similar properties between aligned sequences and a period (.) indicates a nucleotide of weakly similar properties between aligned sequences. Alignment was performed using ClustalW2 software (European Bioinformatics Institute ftp server).

Figure 10. Relative p16 mRNA expression in 3 feline oral squamous cell carcinoma (FOSCC) cell lines compared to normal cat gingiva.

All qRT-PCR was repeated in triplicate using 3 different passages of each cell line. The relative p16 mRNA expression in SCCF1 (n=3) was significantly greater (55-fold) than in SCCF2 (n=3), SCCF3 (n=3), and normal gingiva (**p ˂ 0.0001). p16 mRNA was significantly reduced in the SCCF3 cells compared to normal gingiva (*p = 0.027).

Figure 11. p16 immunohistochemistry of FOSCC and normal cat oral tissues

Absent to low p16 immunoreactivity. Very low intensity of p16 immunostaining was present in less than 20% of the neoplastic cells (A). Moderate p16 staining. Moderate levels of p16 immunoreactivity were present in the cytoplasm of some of the FOSCC cells. Small numbers of nuclei had a high intensity of p16 immunostaining (B). High 54

intensity p16 staining. Both nuclear and cytoplasmic p16 immunostaining were prominent in the tumor cells (C). Normal gingival epithelium, cat. Immunoreactivity to p16 protein in normal cat oral tissues. There was a moderate intensity of p16 staining in normal gingival epithelium (D). p16 IHC in a mouse xenograft tumor derived from the SCCF1 cell line. High p16 intensity (brown color) was observed in both the nuclei and cytoplasm of SCCF1 tumor cells (E). p16 IHC in a mouse xenograft tumor derived from the SCCF2 cell line. Low to absent p16 intensity was present in SCCF2 cells (F). p16 IHC in a mouse xenograft tumor derived from the SCCF3 cell line. Low to absent p16 staining intensity was observed (G). (DAB/hematoxylin)

Figure 12. pRb and p53 immunohistochemistry of FOSCC and normal cat oral tissues.

Less than 50% of the neoplastic cells had a low intensity of nuclear immunostaining to pRb (A). Microscopic features of moderate pRb intensity in FOSCC. Moderate pRb staining intensity was present in more than 50% of neoplastic nuclei (B). High intensity of pRb in FOSCC. Almost all of the neoplastic nuclei have high intensity immunostaining for pRb protein (C). Normal gingival epithelium, cat. pRb IHC of normal cat oral tissue. Moderate pRb staining intensity was present in epithelial cell nuclei (D). Microscopic features of absent to low intensity of p53 staining in FOSCC. The immunoreactivity to p53 protein was absent from neoplastic nuclei (E). Microscopic features of moderate p53 staining intensity in FOSCC. Moderately intense immunoreactivity to p53 was present in most of the neoplastic nuclei (F). Microscopic features of high intensity p53 immunostaining in FOSCC. Prominent p53 immunostaining (dark brown color) was visible within neoplastic nuclei (G). Normal gingival epithelium, cat. Immunostaining of p53 protein in normal cat oral tissue. Low intensity p53 staining was present (H). (DAB/hematoxylin)

Figure 13. Immunohistochemistry of p16, p53, and pRb in spontaneous feline oral squamous cell carcinomas.

The numbers in parentheses (percentages) indicate the 95% confidence intervals (CIs) of each group. No overlap between CIs indicates statistically significant differences between groups (p ˂ 0.05). a Significantly different from absent to low group. b Significantly different from moderate group. c Significantly different from high group.

Figure 14. The average age of cats with FOSCC.

The average age of cats with FOSCC with high intensity p16 staining group (n=6, mean = 10.6 years, SD = 4.0) was significantly lower than the age of cats in the absent to low p16 immunostaining intensity group (n=18, mean = 14.6 years, SD = 4.3) (*p < 0.05).

There was no significant difference between the average ages of cats with absent to low and moderate p16 (n=18) and between moderate and high p16 staining intensity. The plus sign represents the means, the middle line of the boxes indicates the medians, the boxes

contain 50% of samples, and top and bottom error bars represent maximum and minimum age of cats in each group respectively.

Figure 15. Feline p16 predicted cDNA sequence for exons 1α and 2.

The feline p16 cDNA sequence shown was obtained from the updated cat genome sequence (Genbank accession NM 000077.4) after alignment with known cDNA sequences from other species. The sequenced feline p16 cDNA amplicon (Genbank accession: Banklt1884793 Seq1 KU508421) (115 nucleotides in bold) consists of the downstream end of exon 1α and the beginning of exon 2. The ATG start site of exon 1α is underlined.

Chapter 3: The Role of Osteoprotegerin in Bone-invasive Feline Oral Squamous Cell

Carcinoma

Abstract

Feline oral squamous cell carcinoma (FOSCC) is the most common and aggressive cancer in the oral cavity of cats that often induces bone lysis in the mandible or maxilla.

Osteoprotegerin (OPG) a glycoprotein cytokine receptor, reduces tumor growth and inhibits tumor invasion into the adjacent bone in bone-invasive human cancers. We hypothesized that overexpression of feline OPG (fOPG) would reduce bone loss, invasion and tumor growth in xenograft models with bone-invasive FOSCC tumors. Full-length

(OPGv.1) cDNA and an alternative splice variant (OPGv.2) of fOPG were amplified, cloned and sequenced from feline SCCF3 cells. The fOPGv.1 and fOPGv.2 cDNAs were stably transfected into SCCF2 cells. Relative levels of fOPG and fRANKL mRNA were quantified in SCCF2-OPGv.1, SCCF2-OPGv.2, parental SCCF2, and SCCF2-pcDNA cells using qRT-PCR. fOPG mRNA expression in SCCF2-OPGv.1 and SCCF2-OPGv.2 cells was greater (˃ 106-fold) than in the parental SCCF2 and SCCF2-pcDNA cells. The

ratio of fRANKL and OPG mRNA expression in the parental SCCF2 cells was significantly greater than in SCCF2-OPGv.1 and SCCF2-OPGv.2 cells (˃ 106-fold).

Mouse calvaria co-cultured with the conditioned medium from SCCF2-OPGv.1 or

SCCF2-pcDNA cells were used to investigate the effect of fOPG overexpression on bone resorption in vitro. A similar degree of in vitro bone lysis occurred in mouse calvaria that were co-cultured with the conditioned medium of SCCF2-OPGv.1 and SCCF2-pcDNA cells. Nude mice were injected with luciferase expressing SCCF2Luc or SCCF2-

OPGv.1Luc cell lines in the peri-maxillary gingival submucosa. Mice with peri-maxillary

SCCF2Luc tumors had significantly greater tumor volumes (126 ± 35 mm3) compared to

SCCF2-OPGv.1Luc tumors (28 ± 27 mm3) (p ˂ 0.05). Maxillary bone lysis was found in

3 of 5 mice with SCCF2-Luc tumors but, was not present in mice with SCCF2-

OPGv.1Luc tumors (p = 0.58). Overexpression of fOPGv.1 in SCCF2 cells significantly reduced tumor growth in vivo (p ˂ 0.05). These findings are consistent with the results of exogenous OPG treatment in mice with human OSCC tumors where OPG reduced tumor growth and bone loss. These data indicate that fOPG has potential as a therapeutic candidate for FOSCC-associated bone loss.

Introduction

Feline oral squamous cell carcinoma (FOSCC) is an aggressive cancer with limited treatment options [24, 128]. Complete surgical resection can rarely be accomplished in cats with FOSCC due to the local invasiveness of this cancer. FOSCC of the gingiva and oral cavity has a strong propensity to invade maxillary and mandibular bone, which leads to morbidity and the need for euthanasia [129]. Reduction of tumor size and inhibition of the ability of FOSCC to invade oral bone could improve the quality of life of cats with this invasive cancer and serve as important adjunct therapy with other standard treatments.

Bisphosphonates are inhibitors of osteoclasts and approved for human use to inhibit bone resorption associated with osteolytic bone metastases. The effect of a third generation bisphosphonate, zoledronic acid (Zol), in inhibiting tumor-induced bone invasion was previously investigated in FOSCC [129-131]. Even though Zol decreased bone resorption and tumor growth associated with FOSCC, it was unable to prevent all bone resorption or local FOSCC invasion into bone [129-131].

FOSCC-associated bone loss is part of a ‘vicious’ cycle of tumor growth, when tumor invades bone, the resorbing bone provides factors, such as transforming growth factor-beta to induce cancer cell growth [132]. Therefore, inhibition of bone resorption will reduce both local tumor invasion and growth.

Osteoclastic bone resorption requires RANK ligand (RANKL) production from osteoblasts to induce osteoclast function [133]. The binding of RANKL to its receptor

(RANK) on osteoclasts activates osteoclast differentiation and function [134].

Osteoprotegerin (OPG) is also produced from osteoblasts and regulates bone resorption by serving as a decoy receptor and blocking the binding between RANKL and RANK

[77]. The balance of RANKL, RANK and OPG expressions regulates osteoclastic bone resorption [134]. The local activity of osteoclasts is determined by the RANKL to OPG ratio. The higher the level of RANKL, the greater the amount of bone resorption. In contrast, the higher the level of OPG, the lower the amount of bone resorption

Osteoclastogenesis induced by human oral squamous cell carcinoma (OSCC) cell lines (HSC-2 and BHY) was associated with OPG suppression but not RANKL expression [5]. Similar results were reported in FOSCC since a bone-invasive FOSCC cell line (SCCF2) had lower levels of feline OPG (fOPG) mRNA expression compared to

SCCF1 and SCCF3 cells, which have a low capacity to invade bone [129, 132]. Increased

OPG in human tumor-associated-bone disease has been found to reduce tumor growth and bone invasion in vivo. When exogenous OPG was administrated to mice with human

OSCC, a reduction in tumor size, number of osteoclasts and bone invasion was observed

[135]. Similar results were reported in bone-invasive human prostate and breast cancers.

Tumor growth and bone invasion were decreased in tumors that overexpressed OPG and in tumor-bearing mice after administration of exogenous OPG [136].

We hypothesized that fOPG would inhibit bone lysis and tumor growth in bone- invasive FOSCC. The overexpression of fOPG in bone-invasive FOSCC cells was expected to reduce the ability of tumor cells to invade into the adjacent bone and decrease tumor growth. In this study, we developed a FOSCC cell line with fOPG overexpression and investigated the role of fOPG in FOSCC-associated bone lysis in vitro and in vivo.

Materials and Methods

Cell lines and tissues

Three FOSCC cell lines including SCCF1, SCCF2 and SCCF3 cells were grown in

Dulbecco’s Modified Eagle’s Medium (DMEM) with high glucose medium (Gibco by

Life Technologies, Waltham, MA) supplemented with 10% fetal bovine serum (FBS,

Gibco by Life Technologies), L-glutamine and penicillin/streptomycin at 37 °C with 5%

CO2. Three different FOSCC tumors (tumor 1 originated from mandible, tumor 2 from oral mucosa and tumor 3 from gingiva) were obtained from The Ohio State University

(OSU) College of Veterinary Medicine Biospecimen Repository (Dr Holly Borghese, coordinator, and Emily Kuhn). Normal cat tissues including lip (n=1), gingiva (n=2) and tongue (n=4) were obtained from healthy control research cats. Tissues were immediately snap frozen in liquid nitrogen and kept at -80°C until used.

Feline OPG cDNA cloning

The SCCF3 cell line was used to amplify fOPG cDNA since it had the highest levels of fOPG mRNA expression between the 3 FOSCC cell lines. RT-PCR was performed using fOPG primers (forward: 5’-ATAACGTGATGAGTGCACGGG-3’, reverse: 5’-

GTGCCCTGTGGCCAAATCT-3’). PCR fragments were ligated into the TOPO TA cloning kit vector (Invitrogen by Thermo Fisher Scientific, Waltham, MA) and transformed into chemically competent TOP10 (E.coli delivertive) cells. Individual colonies were selected and grown in culture tubes containing LB broth supplemented with 50 µg/mL of ampicillin overnight. Plasmid DNA was extracted using the HighSpeed Plasmid Mini Kit

(MidSci, Valley Park, MO). A set of restriction enzymes, including EcoRI and PstI, were used to confirm the size and determine the orientation of the inserts. Full-length

(OPGv.1) and a splice variant of fOPG missing exon 2 (OPGv.2) were cloned. The plasmid DNA was submitted for Sanger sequencing to confirm the absence of mutations.

Stable transfection of fOPG in SCCF2 cells

The two fOPG variants were cloned into the expression vector pcDNA3.1 that has the

CMV promoter and neomycin and ampicillin resistance cassettes (ThermoFisher

Scientific). SCCF2 cells were then stably transfected with 1 μg of pcDNA3.1

(+)/OPGv.1, pcDNA3.1 (+)/OPGv.2 or pcDNA3.1-empty vector using 2.5 μl

Lipofectamine™ LTX, 1 μl PLUS™ reagent and 200 μl of antibiotic and serum-free

Opti-MEM® (Invitrogen by Thermo Fisher Scientific). Successfully transfected cells

were selected by treatment with 0.6 mg/ml Geneticin® (Invitrogen by ThermoFisher

Scientific) for 4 weeks.

Relative ratio of feline OPG splice variant

The relative proportion between fOPGv.1 (full-length fOPG) and fOPGv.2 (alternative splice variant of fOPG) mRNA was quantified in all 3 FOSCC cell lines, 3 FOSCC tumors and normal cat oral tissues (n=7) using semi-qRT-PCR. The plasmid cDNA clones of fOPGv.1 and fOPGv.2 were used as positive controls. The fOPGv.1 and

OPGv.2 variants were amplified using fOPG-exon 2 primers (forward: 5’-

AATGAACCGGCTGCTGTACT-3’and reverse: 5’-ACCCATCTGGACATCTCTTGC-3’). To obtain an optimal yield (non-saturated PCR products), appropriate PCR cycles for individual samples (SCCF1=36-37 cycles, SCCF3=35-36 cycles, gingiva=34 cycles, lip=32 cycles, tongue=34 cycles, testis=32-34 cycles, FOSCC tissues =30-33 cycles and control plasmid =22 cycles).were determined [137]. The PCR products of fOPGv.1 and fOPGv.2 variants were separated on a 2% agarose gel and imaged using a Typhoon 9410

(Amersham Biosciences Crop, Piscataway, NJ). The relative proportion of fOPGv.1 and fOPGv.2 in each sample was quantified by gel densitometry using ImageQuant TL v.2 software (GE Healthcare Life Sciences, Marlborough, MA).

Quantitative RT-PCR

RNA samples were isolated from SCCF2-OPGv.1 (passages 2, 3 and 4), SCCF2-OPGv.2

(passages 2, 3 and 4), SCCF2-pcDNA 3.1 (passages 2, 3 and 4) and parental SCCF2

(passages 106, 108 and 116) cells using QuickGene DNA cultured cell HC kit (AutoGen,

Holliston, MA). The relative mRNA expression levels of fOPG and fRANKL were normalized to feline glyceraldehyde 3-phosphate dehydrogenase (fGAPDH) using qRT-

PCR, as described in protocol elsewhere [138]. Primers for fGAPDH, fOPG and fRANKL are listed in Table 5.

Feline OPG secretion

To confirm that FOSCC cells overexpressing fOPG mRNA secreted fOPG, a mouse

Osteoprotegerin/TNFRSF11B Quantikine ELISA Kit (R&D systems, Minneapolis, MN) was used to quantify the fOPG protein in the conditioned medium of parental SCCF2,

SCCF2-pcDNA3.1, SCCF2-OPGv.1, and SCCF2-OPGv.2 cells. Western blot analysis was also performed to measure the secreted fOPG protein from SCCF2-OPGv.1 and

SCCF2-pcDNA. Cells (2.5x105) in 2 ml growth medium were seeded into a 6-well plate.

Growth medium was replaced with serum-free medium 24 hr after incubation at 37°C and 5% CO2. Cells and the serum-free conditioned medium were collected at 72 hr. The tumor cells were harvested with 0.25% trypsin EDTA, washed with phosphate-buffered saline (PBS) and centrifuged at 1,000 rpm. Supernatant was removed and the cell pellets’ were resuspended in 200µL of 1X lysis buffer containing 1% protease inhibitor (Cell

Signaling Technology, Danvers, MA). The serum-free conditioned medium was concentrated using acetone precipitation. Protein levels were quantified using the Bio-

Rad protein assay kit (Bio-Rad, Hercules, CA). Protein from tumor cells and conditioned medium (15 µg) were used to perform Western blots [139]. A mouse monoclonal OPG antibody (E-10) (1:500; sc-390518) obtained from Santa Cruz Biotechnology was used as the primary antibody. Goat anti-mouse IgG-CFL 488 (1:2,000; sc362257) (Santa Cruz

Biotechnology) was used as the secondary antibody. Mouse monoclonal β-Actin (C4)

(1:1,000; sc47778) was used as loading control. An image of membrane was taken using a Typhoon 9410 Imager (Amersham Biosciences Crop). The protein intensities were measured by gel densitometry using ImageQuant TL v.2 software (GE Healthcare Life

Sciences).

Murine calvaria co-culture

BGJb medium (Gibco) supplemented with 10% bovine serum albumin RIA grade (BSA,

Sigma Aldrich Co., St Louis, MO) and 50ng/ml Normocin (InvivoGen, San Diego, CA) was used for the bone culture medium. Ten pairs of calvaria bone specimens from 7-day- old mouse pups were obtained using 3.5 mm biopsy punches of the parietal bone (Miltex,

Rietheim-Weilheim, Germany) after euthanasia. Since Zol has been shown to inhibit bone-lysis in vitro [140], overnight incubation of calvaria with Zol was used as negative control. Five pairs of calvaria were incubated overnight in bone culture medium supplemented with 10nM zoledronic acid (Zol: Novartis, Cambridge, MA) and the other

5 pairs were incubated in bone culture medium supplemented with the same amount of

PBS in 37°C and 5% CO2 [130]. After overnight incubation, all calvaria were rinsed gently in PBS. In both Zol and PBS treatment groups, each pair of calvaria was separately co-cultured with 50% of bone culture medium and 50% of conditioned medium from SCCF2-OPGv.1 or SCCF2-pcDNA cells for 1 week. The conditioned medium was changed every 3 days. Co-cultured medium was collected and frozen at -

20°C. The calcium concentrations in co-culture medium were quantified using

QuantiChromTM calcium assay kit (Bio Assay System, Hayward, CA). Calvarial bones were radiographed using a Faxitron cabinet X-ray system (Hewlett-Packard, Palo Alto,

CA). Total bone area and bone mineral density (BMD) of the mouse calvaria were quantified using OsiriX for Mac software (Pixmeo, Geneva, Switzerland). Calvaria were fixed in 95% ethanol and 5% acetic acid for 10 min, followed by rinsing 3 times with

PBS (Gibco) before staining the active osteoclasts for tartrate resistant acid phosphatase using an Acid Phosphatase, Leukocyte (TRAP) kit (387A, Sigma Aldrich).

Lentiviral luciferase transduction

SCCF2Luc cells were previously developed in our lab [129]. SCCF2-OPGv.1 cells were stably transfected with Luciferase-containing lentivirus (VC 2192 (VSV-G), a kind gift from Dr. Amanda Panfil from Dr. Patrick Green’s lab (The Ohio State University) [138].

The expression of luciferase was confirmed by bioluminescent imaging (photon/sec) of

serial dilutions of SCCF2-OPGv.1Luc cells using the IVIS 100 imaging machine (Caliper life sciences, Waltham, MA) [141].

In vivo studies

Twelve 6-week-old nu/nu male mice were purchased from the Target Validation Shared

Resource (TVSR) at the Ohio State University Comprehensive Cancer Center

(OSUCCC). All experiments were approved by the OSU Institutional Lab Animal Care and Use Committee using criteria based on both the Animal Welfare Act and the Public

Health Services “Guide for the Care and Use of Laboratory Animals”. Nude mice were anesthetized with isoflurane before injecting the left side of their peri-maxillary gingiva submucosa with tumor cells that were diluted in 0.1 ml PBS. Six mice were injected with

5-10x106 SCCF2Luc cells and the other six mice were injected with 5-10x106 SCCF2-

OPGv.1Luc cells. [130]. To measure the calcium concentration in urine, samples were collected daily from all mice by manual expression of the urinary bladder from day one to day 42. Tumor sizes were measured using a digital caliper every 3-4 days. To measure viable tumor cells in vivo, bioluminescent images (BLI) were obtained once a week as previously described [142]. Euthanasia was performed at 42 days after tumor cell injection. The heads with mandibles were removed and radiographed using a Faxitron cabinet X-ray system (Hewlett-Packard, Palo Alto, CA) [130]. Specimens were fixed in

10% neutral-buffered formalin for 48 hr, transferred to 70% ethanol for 24 hr and decalcified in 10% EDTA (pH 7) for 1 week before submitting for histopathological

processing. To investigate the microscopic bone invasion, samples were paraffin embedded, cut and stained with hematoxylin and eosin (H&E) and examined using light microscopy.

Urine calcium concentration

Daily urine samples from individual mice injected with SCCF2Luc or SCCF2- pcDNALuc cells were pooled together (one pool per week). Individual mouse urine calcium concentrations were measured using QuantiChromTM calcium assay kit

(Bioassay Systems, Hayward, CA).

Statistical analysis

Two-tailed student’s t-test was used to analyze the statistical differences between 1) relative mRNA levels of SCCF2 and SCCF2-OPGv.1, SCCF2 and SCCF2-OPGv.2,

SCCF2-pcDNA and SCCF2-OPGv.1, and SCCF2-pcDNA and SCCF2-OPGv.2; 2) calcium concentrations of conditioned medium from SCCF2-pcDNA and SCCF2-

OPGv.1 cells co-cultured with calvarial bone; 3) bone area of calvaria co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells and 4) bone mineral density of calvaria co-cultured with conditioned medium from SCCF2-pcDNA and

SCCF2-OPGv.1. A two-way ANOVA was used to analyze the statistical difference between SCCF2Luc and SCCF2-OPGv.1Luc tumors volume in week 1 to 6 as well as between urine calcium concentrations of mice with SCCF2Luc and SCCF2-OPGv.1Luc

tumors in weeks 1, 3 and 6. The statistical difference between numbers of mice with bone invasion in SCCF2Luc-bearing mice and SCCF2-OPGv.1Luc-bearing mice were determined using Chi-square. Statistical significance was considered to be p ˂ 0.05.

Results

Feline OPG cDNA cloning

A full-length fOPG (OPGv.1) and fOPG alternative splice variant (OPGv.2) were amplified, cloned and sequenced from the SCCF3 cell line. The fOPGv.1 contained 1,441 nucleotides spinning 5 exons (GenBank no. KX610462). The fOPGv.2 alternative splice variant contained 1,071 nucleotides with complete deletion of exon 2 (nt 1-370)

(GenBank no. KX610463). The OPGv.2 splice variant was predicted to generate a truncated fOPG protein using NCBI ORF Finder software.

The relative ratio of full-length and splice variant of feline OPG mRNA in SCCF2 cell lines, FOSCC tumors, and normal cat tissues

The relative proportion of fOPGv.2 splice variant was less than 40% in SCCF1 (n=3) and

SCCF3 (n=3) cell lines. Full-length and fOPGv.2 splice variant were not amplified in

SCCF2 (n=3). The percentage of fOPGv.2 expression varied from 0 to 38% in normal cat oral tissues (Table 6). Interestingly, an alternative splice variant was not found in FOSCC tissues (n=3) (Table 6)

fOPG overexpression in SCCF2 cell line

FOSCC cells overexpressing either variant of fOPG (SCCF2-OPGv.1 and SCCF2-

OPGv.2) were developed for this study. The cytological characteristics of SCCF2-

OPGv.1 and SCCF2-OPGv.2 cells were similar. Both SCCF2-OPGv.1 and SCCF2-

OPGv.2 formed clusters of round to polygonal cells with enlarge nucleoli. Larger tumor cells were observed in these two cells compared to the parental SCCF2 cells (Figure 16).

The relative fOPG mRNA levels in SCCF2-OPGv.1 (8x106-fold) and in SCCF2-OPGv.2

(3x106 -fold) were significantly greater than those in the parental SCCF2 and SCCF2- pcDNA cells (p ˂ 0.05). The relative fRANKL mRNA levels between the parental

SCCF2, SCCF2-pcDNA and the fOPG overexpression cells (SCCF2-OPGv.1 and

SCCF2-OPGv.2) were not different. Thus, the ratio between fRANKL and fOPG mRNA expression was significantly lower in SCCF2-OPGv.1 and SCCF2-OPGv.2 cells than that in the parental SCCF2 and SCCF2-pcDNA (˃106-fold) (p ˂ 0.05) (Figure 17).

Secreted fOPG protein in SCCF2-OPGv.1

A Mouse Osteoprotegerin/TNFRSF11B Quantikine ELISA (R&D systems) and western blots using a mouse OPG monoclonal antibody as the primary antibody were used to measure the secreted OPG protein. However, we were not able to detect fOPG protein in either fOPG overexpressing cells (SCCF2-OPGv.1 and SCCF2-OPGv.2) or in the

parental SCCF2 and SCCF2-pcDNA cells. We do not know if this is because OPG is not secreted or because the antibody against murine OPG does not recognize feline OPG.

In vitro bone lysis

In vitro bone lysis was observed in mouse calvaria that were co-cultured with conditioned medium from SCCF2-pcDNA and from SCCF2-OPGv.1 cells (Figure 18). Unexpectedly, conditioned medium from SCCF2-OPGv.1 cells (mean=13.92 mg/dL, SD = 1.86) that were co-cultured with calvarial bone had significantly greater calcium levels than co- conditioned medium from SCCF2pcDNA cells (mean = 8.39 mg/dL, SD = 0.74) (p ˂

0.05). However, bone mineral density and the remaining bone area of these calvaria were not statistically different (Figure 19).

In vivo tumor growth

In the first two weeks following injection of SCCF2Luc cells into mice, tumors were not visible macroscopically but could be detected by bioluminescent imaging (BLI) in all mice. One of six SCCF2Luc bearing mice did not develop a tumor and was excluded from the results. In 2 of 6 (33%) SCCF2-OPGv.1Luc-bearing mice, BLI signal was detected in the third week but was absent in the fourth week after tumor implantation. No histological evidence of tumor cells was found in these two SCCF2-OPGv.1Luc-bearing mice. From the third week to the end of the experiment, SCCF2Luc bearing mice (n=5)

had greater average tumor volumes (126 mm3 ± 35) than those in (28 mm3 ± 27) of

SCCF2-OPGv.1Luc-bearing mice (n=6) (p ˂ 0.05) (Figures 20 and 21).

Faxitron radiography

Three of five (60%) SCCF2Luc-bearing mice had moderate to severe maxillary bone lysis as determined by radiography. There was no evidence of maxillary bone loss on the radiographs of the SCCF2-OPGv.1Luc-bearing mice (Figure 22). While not statistically significant due to low sample numbers, there was a trend of greater maxillary bone loss in

SCCF2Luc-bearing mice (3/5) compared to SCCF2-OPGv.1Luc-bearing mice (0/4) (p =

0.058)

Histopathological findings

Both SCCF2Luc and SCCF2-OPGv.1Luc xenografts had the characteristics of a well- differentiated squamous cell carcinoma. Multiple islands of keratin pearls surrounded by large polygonal cancer cells with hyperchromatic nuclei were observed. Invasion of tumor cells into maxillary bone and moderate to severe osteoclastic bone lysis were found in three of five SCCF2Luc xenografts. Tumor cells were not found in 2 of 6

SCCF2-OPGv.1Luc xenografts. In the other four SCCF2-OPGv.1Luc xenografts, tumor cells were adjacent to maxillary bone surface but there was no evidence of bone lysis

(Figure 23).

Urine calcium concentration

Urine calcium concentrations in SCCF2Luc (n=6) and SCCF2-OPGv.1Luc-bearing mice

(n=6) in the first, third and sixth weeks of the experiment were compared. In the first and third weeks, SCCF2Luc-bearing mice had significantly greater urine calcium concentrations compared to SCCF2-OPGv.1Luc-bearing mice (Figure 24).

Discussion

Inhibition of RANKL using OPG or soluble RANK receptor has been found to reduce tumor-induced bone lysis in human OSCC-associated bone loss [5, 15]. In FOSCC, fOPG is the only option for RANKL inhibition, since a cat-specific soluble RANK receptor has not been developed. In this study, the therapeutic role of fOPG in bone-invasive FOSCC was investigated using novel fOPG overexpressing FOSCC cell lines.

The RANKL to OPG ratio has been known to serve as an important in vivo indicator of osteoclastic bone resorption [134]. Low OPG mRNA expression was observed in a human OSCC cell lines associated with bone loss (HSC-2 and BHY cells) as well as in the bone-invasive FOSCC cells (SCCF2) [132, 143]. This suggests that OPG levels are critical for bone lysis induced by OSCC

In humans, full-length OPG mRNA has 2,400 nucleotides that span exons 1-5 and two alternative splice variants [144, 145]. Variants 1 and 2 of human OPG had an additional 1,800 and 4,000 nucleotides, respectively, from the insertion of intron 2 to the

standard variant [145]. The splice variants of OPG have been speculated to have a role in human cancer pathogenesis; however, their function remains unknown [144]. In this study, full-length (fOPGv.1) and an alternative splice variant of fOPG (fOPGv.2) were amplified from SCCF3 cells. The fOPGv.2 variant had a complete deletion of the 370 nucleotides in exon 2. The fOPGv.2 alternative splice variant was expressed at high levels in SCCF1, SCCF3 and gingiva (16-38%), but it was expressed at very low levels in the remaining normal cat tissues (0-3%) compared to fOPv.1. Neither fOPGv.1 nor fOPGv.2 variant mRNA expression was found in SCCF2. In addition, none of the three

FOSCC tumors investigated in this study expressed the fOPGv.2 variant. These results suggest that the fOPGv.2 alternative splice variant is uncommon in cats and may not be important for the pathogenesis of FOSCC. However, a larger sample size may need to more accurately identify the incidence. The existence of other potential fOPG alternative splice variants in FOSCC needs to be investigated.

Since mRNA levels do not always correlate with protein expression, it was necessary to verify that OPG was secreted in the overexpressing cell lines. However, fOPG protein was not found in the conditioned medium from SCCF2-OPGv.1, SCCF2-

OPGv.2, parental SCCF2 or SCCF2-pcDNA cells. This could mean that OPG was not secreted by these cells. However, a lack of cross-reactivity between mouse and feline

OPG resulting in an inability to detect fOPG protein is a likely possibility. A commercially available cat-specific OPG ELISA is not recommended by the manufacturer for measuring the secreted fOPG in conditioned medium (BIOTANG,

Lexington, MA). In addition, a cat-specific OPG antibody that can be used for Western blot analysis is not currently available. Western blots with antibodies that reliably detect fOPG or mass spectrometry would greatly facilitate further identification of this important protein.

Significant osteoclastic bone resorption was previously observed in mouse calvaria that were co-cultured with SCCF2 cells [129]. Our study revealed no difference in the degree of bone resorption between mouse calvaria that were co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells. However, calcium concentrations in the conditioned medium from SCCF2-OPGv.1 cells that was co- cultured with calvaria were significantly greater than the co-cultured conditioned medium from SCCF2-pcDNA cells (p ˂ 0.05). These results were unexpected; however, we suspect that FOSCC-induced bone resorption in vitro may not be regulated by the

RANKL/OPG rates.

A previous study reported that exogenous OPG treatment (subcutaneous injection) in mice with bone-invasive human OSCC cells (B88) resulted in a reduction of bone invasion, tumor size and osteoclast number [135]. In the current study, nude mice with SCCF2Luc tumors had greater tumor volume compared to mice with SCCF2-

OPGv.1Luc tumors (p ˂ 0.05). The difference in tumor growth between mice in these two groups could be observed as early as three weeks after tumor xenograft. These observations were consistent with the results found in OPG C4-2 cells (a human OPG

overexpression prostate cancer cell line) where overexpression of OPG resulted in decreased tumor growth in mice with OPG C4-2 tumors [146].

In vivo bioluminescence was absent after week three of tumor formation in two

SCCF2-OPGv.1Luc mice. At six week, tumor cells were absent in the maxilla suggesting that fOPG induced regression of tumors in FOSCC in vivo. However, larger sample sizes are needed to confirm these findings.

Numbers of mice with maxillary bone loss appear to be greater in SCCF2Luc- bearing mice compared to SCCF2-OPGv.1Luc-bearing mice. These findings demonstrated that overexpression of fOPG inhibited the ability of FOSCC cells to invade into the adjacent bone and reduce osteoclastic bone resorption in vivo. However, a larger sample size is required to ensure the inhibitory effect of fOPG on bone loss. In addition, mice with SCCF2Luc tumors had greater urine calcium concentrations than mice with

SCCF2-OPGv.1Luc tumors (p ˂ 0.05) in the first 3 weeks. Based on these results, we speculate that the overexpression of fOPG reduced urine calcium excretion in the early period after tumor cell engraftment due to a local of maxillary bone invasion and secondary bone lysis.

Conclusions

In this study, we developed a novel fOPG FOSCC cell line (SCCF2-OPGv.1) that overexpressed fOPG. Overexpression of fOPG decreased tumor growth and inhibited bone invasion and resorption in vivo. The SCCF2-OPGv.1 cell line is useful tool to

investigate the role of fOPG in bone-invasive FOSCC in vivo. fOPG may be an effective adjuvant treatment for FOSCC-associated bone loss.

Primer Primer pair Product (bp) fOPG forward: 5’-CCAGCTACTGAAGTTATGGAAACA -3’ 118 reverse: 5’-GAGGTTTGTGTGTCCGAGGT-3’ fRANKL forward: 5’-AAGGGGTATGACCCACAATG-3’ 117 reverse: 5’-AAGCCCTCGGACTGTAACAA-3’ fGAPDH forward:5’-GGCGTGAACCACGAGAAGTA-3’ 144 reverse:5’-GATGGCATGGACTGTGGTCA-3’ Table 5. RT-PCR primers for feline OPG, RANKL and GAPDH

Cell lines FOSCC tissues Normal tissue Splice variants SCCF1 SCCF2 SCCF3 1 2 3 Gingiva Lip Tongue (n=3) (n=3) (n=3) (n=1) (n=1) (n=1) (n=2) (n=1) (n=4) OPGv.1 (full-length) 0.66 ND 0.70 1.00 1.00 1.00 0.62 0.97 100 OPGv.2 0.34 ND 0.30 0.00 0.00 0.00 0.38 0.03 0.00 Table 6. The relative amount of full-length (fOPGv.1) to fOPGv.2 splice variant of feline osteoprotegerin mRNA

ND: Not detected

Figure 16. SCCF2 and SCCF2-OPGv.1 cell morphology in vitro

The parental SCCF2 and SCCF2-OPGv.1 cells were adherent, round to polygonal cells with cell to cell adhesion. However, the SCCF2-OPGv.1 were larger and rounder with more cytoplasm compared to the parental SCCF2 cells. (x40)

Figure 17. Relative mRNA levels of feline OPG, RANKL and RANKL: OPG ratio in SCCF2-OPGv.1 and SCCF2-OPGv.2 cells compared to SCCF2 and SCCF2-pcDNA cells

Feline OPG mRNA levels in SCCF2-OPGv.1 and SCCF2-OPGv.2 cells were much higher than in SCCF2 and SCCF2-pcDNA cells (*p ˂ 0.0001) (A). The level of fRANKL mRNA expression was not different between the cells (B). The fRANKL to fOPG mRNA expression ratio was dramatically lower (Y-axis is a log scale) in SCCF2-OPGv.1 and

SCCF2-OPGv.2 cells compared to SCCF2 and SCCF2-pcDNA cells (*p ˂ 0.0001) (C). 83

Figure 18. Murine calvaria co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells with and without zoledronic acid (Zol) treatment

There was a similar degree of in vitro bone loss in murine calvaria that were co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells. Bone resorption was inhibited in the calvaria that were incubated with 10 nM Zol overnight before co-culturing with conditioned medium.

Figure 19. Bone area, bone mineral density of calvaria and calcium concentration of calvaria co-cultured with conditioned medium from SCCF2-pcDNA and SCCF2-OPGv.1 cells

There was no difference in bone area and bone mineral density (BMD) of mouse calvaria, which were co-cultured in the conditioned medium from SCCF2-pcDNA and SCCF2-

OPGv.1 cells (A, B). The conditioned medium of SCCF2-OPG cells co-cultured with calvaria had greater calcium concentration than conditioned medium of SCCF2-pcDNA cells co-cultured with calvaria (*p ˂ 0.05) (C). Means are indicated by the plus signs, medians are indicated by the middle lines of the boxes, the top and bottom error bars represent maximum and minimum bone area (A), BMD of calvaria (B) and calcium concentrations (C) of co-conditioned medium in each group.

Figure 20. SCCF2Luc and SCCF2-OPGv.1Luc in nude mice

SCCF2Luc tumors were larger compared to the SCCF2-OPGv.1Luc tumors (arrow) (bar

= 10 mm).

Figure 21. Average SCCF2Luc and SCCF2-OPGv.1Luc tumor volume in nude mice

SCCF2Luc tumors grew faster and larger (*p ˂ 0.05) compared to SCCF2-OPGv.1Luc tumors in nude mice. Black circles and squares indicate mean of average tumor volume in mice with SCCF2Luc and mice with SCCF2-OPGv.1Luc, respectively. Two side bars indicate standard deviation (SD) of data.

Figure 22. Radiographs of maxillary bone from nude mice with SCCF2Luc and SCCF2- OPGv.1Luc tumors (week 6)

Nude mice with SCCF2Luc tumors had moderate to severe maxillary bone lysis (arrows)

(upper panel). Maxillary bone lysis was not present in nude mice with SCCF2-

OPGv.1Luc tumors (lower panel).

Figure 23. Histopathology of SCCF2Luc and SCCF2-OPGv.1Luc xenografts

Well differentiated SCCF2Luc squamous cell carcinoma (SCC) cells invaded into the maxillary bone (MB) and induce osteoclastic bone resorption (arrows) (A, C). In the

SCCF2-OPGv.1Luc xenografts, the well differentiated SCC tumor cells grew adjacent to the maxillary bone, but did not invade into the bone (B, D). (40X, H&E stain)

Figure 24. Urine calcium concentration from nude mice with SCCF2Luc and SCCF2OPGv.1Luc tumors

There was greater urinary calcium in mice with SCCF2Luc xenografts compared to mice with SCCF2-OPGv.1Luc tumors during weeks 1 and 3 (p ˂ 0.05) only. Mean of calcium concentration in first, third and sixth week is indicated by the plus signs, medians are indicated by the middle lines of the boxes, the top and bottom error bars represent maximum and minimum calcium concentration.

Chapter 4: Feline Oral Squamous Cell Carcinoma, Telomerase Activity, and Novel

Alternative Splicing Cloning and Expression of TERT

Abstract

Telomerase is a complex protein that has major functions in cell replication and immortalization. Telomerase activity and telomerase transcriptase catalytic subunit,

(hTERT) mRNA levels have been investigated in various human cancers for decades.

Studies have demonstrated increased levels of telomerase activity and up-regulation of hTERT mRNA in cancer cells and tumors when compared to normal samples.

Alternative splicing is thought to regulate the immortalizing activity of telomerase.

Evaluation of telomerase activity in feline cancers has been limited due to the lack of a complete feline TERT (fTERT) cDNA. Telomerase activity has been reported in some feline cancers; however, the correlation between telomerase activity and level of fTERT mRNA expression has not been addressed. In this study, 3 feline oral squamous cell carcinoma (FOSCC) cell lines (SCCF1, SCCF2 and SCCF3), 3 FOSCC tumors, normal cat oral tissues and cat testis were used to measure the levels of telomerase activity and

fTERT mRNA expression in FOSCC. fTERT mRNA levels were quantified in all samples using qRT-PCR. Telomerase activity was measured using the quantitative telomerase detection kit (QTD). Full-length fTERT cDNA cloning was performed with mRNA from SCCF2 cells. Dominant fTERT isoforms were investigated in all samples using semi-quantitative RT-PCR. Significantly higher telomerase activity was found in 3

FOSCC cell lines (SCCF1, SCCF2 and SCCF3) and 2 out of 3 FOSCC tumors when compared to normal cat oral tissues (lip, gingiva, and tongue). Full-length fTERT cDNA and 10 novel cat-specific alternative splice variants were successfully amplified, cloned and sequenced. The relative levels of full-length fTERT and the del-e10 splice variant paralleled the level of telomerase activity in each sample. Alternative splicing may have a role in the regulation of feline telomerase activity. To our knowledge, this is the first study that has successfully amplified the complete fTERT cDNA and identified the fTERT alternative splice variants. This information will be useful for understanding the role of telomerase in feline cancers and may also shed light on the role of telomerase in human cancer.

Introduction

Telomerase is a 15-20 kDa ribonucleoprotein enzyme complex that adds a DNA sequence repeat (TTAGGG) to the 3’-end of the DNA strand in telomere regions [147,

148]. This complex consists of 2 major subunits, including the telomerase transcriptase catalytic subunit (TERT) and a 581-nucleotide non-coding RNA template (TERC). These

2 subunits and their associated proteins in the T-loop and G-quadruplex complexes cooperate to prevent shortening of the telomere, which leads to cellular apoptosis and senescence [148-150]. Generation of immortalized primary cell lines by using human

TERT (hTERT) transfection has been reported in human as well as in cat primary cells

[151]. Increased telomerase activity and overexpression of TERT mRNA has been reported in more than 90% of human cancers, including oral squamous cell carcinoma

(OSCC) as well as normal proliferative cells, such as embryogenic cells of ovaries and spermatogonia cells of testis [152, 153]. TERT mRNA was detected in 76% (35/46) of human oral squamous cell carcinoma (OSCC) compared to 0% (0/15) of normal oral samples [153]. Up-regulation of telomerase activity was reported in 46% (21/46) to 86%

(36/42) of OSCC cases and in 50% (4/8) to 67% (8/12) of pre-malignant conditions of the oral mucosa [154, 155].

The biological characteristics of telomerase activity appear to be similar between humans and cats [91, 156]. The average telomere repeat length in humans (5-12 kb) is more similar to cats (4.7-26.3 kb) than mice (50-150 kb) [156-158]. Detection of fTERT

mRNA has been reported in feline mammary cancer cell lines and normal cat tissues

[159].

Telomerase activity has been studied in certain feline tumors and normal tissues, including feline oral squamous cell carcinoma (FOSCC) [156, 159-164]. FOSCC is a very aggressive head and neck cancer in cats that can invade into the adjacent bone and induce bone lysis [76, 129]. The similarity between FOSCC and human OSCC has been reported [165].

Alternative splicing is an important gene regulation mechanism that accounts for a tremendous amount of diversity in protein translation and function [166]. Alternative splicing occurs in approximately 95% of human multi-exon genes [166]. Twelve alternative splice sites (ASPSs) and 22 alternative splice variants of hTERT have been described [167-170]. Alternative splicing was suspected to regulate the pathogenic role of telomerase in human cancers [166]. Even though none of the discovered hTERT splice variants had reverse transcriptase activity, some of them appeared to have a role in cancer regulation [167, 171]. Dominant negative regulatory activity was demonstrated in the hTERT α-deletion isoform [166, 169, 170]. The hTERT β-deletion isoform was also found to have a role in breast cancer cell growth [167]. TERT alternative splicing was reported in non-human species, including dogs [169] and chickens [172]. TERT splice variants, such as hTERT α-deletion, were proposed as a novel therapeutic targets and biomarkers for cancer [166].

In cats, only 7% (237 bp) of the feline-TERT (fTERT) sequence has been amplified from one feline lymphoma cell line [163]. The correlation between the level of telomerase activity and fTERT mRNA expression as well as a complete fTERT cDNA sequence and fTERT alternative splicing information are unknown.

In the present study, the telomerase activity and fTERT mRNA expression in FOSCC cell lines, FOSCC tumors, normal cat oral tissues and testis were investigated. A full-length fTERT cDNA sequence and 10 novel, cat-specific fTERT alternative splice variants were amplified and cloned. The association between relative expression of splice variants and telomerase activity was also examined. The cloning of full-length fTERT, novel fTERT alternative splice variants and the information about telomerase activity reported in this study will provide fundamental knowledge for studying the role of telomerase in feline cancers.

Materials and Methods

Cell lines and sample collection

The FOSCC cell lines (SCCF1, SCCF2 and SCCF3), which have previously been developed in our laboratory [130], were grown in DMEM high-glucose medium (Gibco,

Life Technologies, Waltham, MA) supplemented with 10% fetal bovine serum (FBS,

Gibco), L-glutamine and penicillin/streptomycin and maintained at 37°C with 5% carbon dioxide. Normal cat oral tissues (gingiva (n=3), tongue (n=4) and lip (n=1)) and normal

testis (n=3) were used as negative and positive controls, respectively. Control tissues were obtained from appropriately healthy research cats. The three FOSCC tumors used in this study were collected by Dr. Holly Borghese and Ms. Emily Kuhn (coordinator and assistant, respectively) of the Ohio State Veterinary Biospecimen Repository, which is supported by Clinical and Translational Science Awards of the National Institutes of

Health under award number UL1TR001070. Tissues were immediately frozen in liquid nitrogen and stored at -80°C.

Relative fTERT mRNA expression

Total RNA was isolated from 3 different passages of SCCF1 (passages 63, 64 and 65),

SCCF2 (passages 59, 60 and 61), and SCCF3 (passages 53, 54 and 55) cell lines using

QuickGene RNA cultured cell HCs kit (Autogen; Holliston, MA). RNA was also extracted from normal cat oral tissues (gingiva (n=3), tongue (n=4) and lip (n=1)), normal cat testis (n=3) and 3 frozen FOSCC tumors. Quantitative RT-PCR was performed as previously described [138]. Two set of primers that amplified an PCR product between exons 3 and 5 (fTERT exon 3-5, forward: 5’-GGAGCCAGTTGCAGAGCATA-3’, reverse:

5’-GTGCCGGACCTTCTTGTCTC-3’) and exons 11 and 12 (fTERT exon 11-12, forward:

5’-CGGTGTTCCTGAGTATGGCT-3’, reverse: 5’-GAATGGAGGTCTGGGCGTAA-3’) were used to amplify the fTERT mRNA. Feline ubiquitin (fUBC) primers (forward: 5’-

AAGCCGGGAGTTCCGTTG-3’, reverse: 5’-CACGAAGATCTGCATGGTCAAG-3’) were used to amplify the housekeeping gene. All normal cat gingiva samples were excluded

from qRT-PCR of fTERT due to an unacceptable degree of variability of housekeeping gene mRNA levels.

Protein extraction and telomerase activity quantification

SCCF1, 2 and 3 cells (0.1-1x106) were harvested with 0.25% trypsin EDTA (Gibco by

Life Technologies), washed with phosphate-buffered saline (PBS) followed by centrifugation at 1,000 rpm and supernatant removed. The remaining cell pellets were resuspended in 200µL of 1X lysis buffer (Cell Signaling Technology, Danvers, MA) that contained 1% protease inhibitor (Cell Signaling Technology). FOSCC tumors 50-100 mg

(n=3), normal cat oral tissues (gingiva (n=2), tongue (n=2) and lip (n=1)), and cat testis

(n=3) were homogenized in 200 µL of 1X lysis buffer containing 1% protease inhibitor

(Cell Signaling Technology) using a gentleMACS dissociator (Miltenyi Biotec, San

Diego, CA). Cell suspensions were incubated on ice for 30 min and briefly sonicated 3 times followed by centrifugation at 12,000 rpm at 4°C for 30 min. Protein levels were quantified from recovered supernatants using a Bio-Rad protein assay kit (Bio-Rad,

Hercules, CA).

A commercially available PCR-based Quantitative Telomerase Detection Kit (QTD:

Allied Biotech, Inc, Vallejo, CA ) was used to quantify telomerase activity in SCCF1, 2 and 3 cell lines, 3 frozen FOSCC tumors and normal cat oral tissues (gingiva (n=2), tongue (n=2) and lip (n=1)) in triplicate. To perform the PCR reaction, a mixture of 0.7

µg protein lysate, 11.5 µL DNase-free water (Gibco) and 12.5 µL of 2X QTD premix

containing the telomere template and essential PCR reagents for qRT-PCR was prepared.

The samples with greater telomerase enzyme activity had greater PCR product amplification. A standard curve made from serial dilutions of telomere sequences

(provided in the kit) was used to convert the amount of PCR product (crossing points) to telomerase activity (attomoles).

fTERT cDNA cloning

The fTERT cDNA was cloned in 2 fragments, a 5’ fragment and a 3’ fragment, since exon 1 is almost 80% GC rich and it was difficult to obtain an amplified PCR fragments containing this exon. Furthermore, the sequence for cat TERT exon 1 is incorrect in the

NCBI database (XM_011287361.1), thus it was difficult finding primers to amplify the 5’ fragment. Nested PCR was required to obtain the 5’ fragment. The initial RT-PCR product was amplified from cDNA made from SCCF2 cells using Pfx polymerase

(Invitrogen, Thermo Fisher Scientific, Waltham, MA) and 5’-open reading frame

(5’ORF) primers (forward: 5’-GCCCCTCCCCGACCATG-3’ and reverse: 5’-

CTTCACTTGCGAGGTGAGGT-3’). There was no obvious PCR product of the correct size, therefore, a region of the gel where the correct sized product was expected to be present was excised and PCR was performed on the purified DNA with a nested set of primers

(forward: 5’-ACCATGCCGCGCGCGCCCAG-3’ and reverse: 5’-

GCCTGATCACGTTGGCAATC-3). Platinum Taq DNA polymerase (Invitrogen, Thermo

Fisher Scientific) was added to the final elongation step in order to add an adenine

overhang for TA cloning. For the 3’ fragment, the PCR reaction was performed using the

Platinum Taq DNA polymerase (Invitrogen) and 3’ORF primers (forward: 5’-

GCCAGTTGCAGAGCATAGGA-3’ and reverse: 5’-CTCTCTTCCACAAGGACACGG-3’). PCR fragments were ligated into separate TOPO TA cloning kit vectors (Invitrogen) and transformed into chemically competent Escherichia coli TOP10 (E.coli derivative) cells.

Individual colonies were selected and grown overnight in culture tubes containing LB broth and 50 µg/mL of ampicillin. The plasmid DNA was extracted using the HighSpeed

Plasmid Mini Kit (MidSci, Valley Park, MO). To confirm the size and determine the orientation of the fragment, restriction digests using EcoRI and PstI restriction enzymes were used. The plasmid DNA was submitted for sequencing to check for completeness and the absence of point mutations. This cloning strategy yielded full-length 5’ and 3’ fTERT fragments, as well as 10 novel, cat-specific fTERT alternative splice variants. The full-length 5’ and 3’ fragments were then cut with AclI and NotI (NEB, Ipswich, MA), and ligated with T4 DNA ligase (NEB) to obtain the complete fTERT cDNA. Individual colonies were selected and plasmid DNA was extracted. The complete fTERT cDNA was cloned into the EcoRI site of the expression vector, pcDNA3.1 (Thermo Fisher

Scientific). The complete plasmid was submitted for sequencing. NCBI ORF Finder software was used to predict the transcribed protein of each variant.

fTERT alternative splice variant quantification

Semi-quantitative RT-PCR was performed to investigate fTERT splice variants expression in the 3 FOSCC cell lines, 3 FOSCC tumors, 7 normal cat oral tissues and normal cat testis [137]. Primers were designed to bracket selected exons (Table 7).

Different PCR cycles were used in each individual cDNA sample to obtain optimal amounts (non saturated PCR products) of splice variant PCR products (Table 8). The fTERT splice variants were separated and visualized by 1.5% agarose gel electrophoresis.

Plasmid cDNA clones of the splice variants were use as positive controls. The relative proportion of splice variants in each sample were quantified using gel densitometry

(ImageQuant TL v.2 software: GE Healthcare Life Sciences, Marlborough, MA).

Statistical Analyses

The difference between relative mRNA expression levels of FOSCC cell lines and normal cat oral tissues and the difference between relative mRNA expression of FOSCC tumors and normal cat oral tissues was determined using a unpaired t-test with Welch’s correction. An unpaired t-test with Welch’s correction was also used to determine the difference between telomerase activity of FOSCC cell lines and normal cat oral tissues and between FOSCC tumors and normal cat oral tissues. A P-value less than 0.05 was considered statistically significant. Due to variation between values of mRNA expression and telomerase activity in each sample, Log10 scale was used to present the data.

100

Results

Relative fTERT mRNA expression qRT-PCR using primers to the exon 3-5 region showed an 8-fold greater level of fTERT mRNA in SCCF2 compared to normal cat oral tissues (p = 0.0001). The levels of fTERT mRNA expression in SCCF3 were significantly lower than normal cat oral tissues (15- fold, p = 0.025). Two of three FOSCC tumors (tumors 2 and 3) had greater fTERT mRNA levels (˃ 8-fold) compared to normal cat oral tissues (Figure 25A). qRT-PCR using primers to the exon 11-12 region showed greater levels of fTERT in

SCCF1 (5-fold, p = 0.0014) and SCCF2 (10-fold, p = 0.0035) compared to normal cat oral tissues. Two of three FOSCC tumors also had greater fTERT mRNA (˃ 10-fold) compared to normal cat oral tissues (Figure 25B). In the exon 11-12 region, there was no statistically significant difference in fTERT mRNA levels between SCCF3 and normal cat oral tissues.

Telomerase activity in FOSCC

Telomerase activity in SCCF1, SCCF2 and SCCF3 was significantly greater than normal cat oral tissues (p ˂ 0.0001) (Figure 26). Two of 3 FOSCC tumors also had greater telomerase activity than normal cat oral tissues (p ˂ 0.001). The three cat testicular samples were excluded because protein with telomerase activity could not be successfully isolated from these samples.

101

Full-length and alternative splice variants of fTERT

The SCCF2 cell line, which expresses high telomerase activity and fTERT mRNA, was used to amplify full-length fTERT cDNA. The full-length fTERT containing 3,457 nucleotides spaning16 exons (1-16) and 10 novel fTERT splice variants were amplified, cloned and sequenced (Figure 27).

Five novel, cat-specific fTERT splice variants were amplified in the 5’region of fTERT

(exon 1 to proximal region of exon 5) (Figure 27D). The Del-e2, e3 splice variant had a complete deletion of exon 2 (nt 1-1,288) and exon 3 (nt 1-196). This splice variant was predicted to generate a frame shifted protein. The 4 other splice variants within in the 5’ fragment included various combinations of exon 2 deletion and intron 3 insertion. The

Del-e2, Ins-i3_1 isoform had a deletion of 1,233 nucleotides within exon 2 (nt 1-1,233) with insertion of the complete intron 3 (nt 1-1,326). The last three 5’ splice variants had a complete deletion of exon 2 (nt 1-1,288) with an insertion of 570 nucleotides of intron 3

(nt 230-799; Del-e2, Ins-i3_2), an insertion of 1,097 nucleotides within intron 3 (nt 230-

1,326; Del-e2, Ins-i3_3), and an insertion of 799 nucleotides within intron 3 (nt 1-799;

Del-e2, Ins-i3_4). These 4 splice variants were predicted to generate proteins without telomerase activity (Table 9).

The other 5 novel, cat-specific fTERT splice variants were observed in the 3’region of fTERT (terminal region of exon 3 to exon 16) (Figure 27C). The Del-e10 splice variant had a complete deletion of exon 10 (nt 1-72). The Del-e9, e10 splice variant had a deletion of exon 9 (nt 1-114) and exon 10 (nt 1-72). The Del-e9, e10, e13 splice variant

102

had a complete deletion of exons 9, 10 and 13 (nt 1-62). In the Del-e9, e13 splice variant, exons 9 (nt 1-114) and 13 (nt 1-62) were deleted. The Del-e10, e13, e15 variant had complete deletion of exon 15 (nt 1-138) in addition to exons 9 and 13. The Del-e10 and

Del-e9, e10 splice variants were predicted to generate fTERT proteins with short deletions in the c-terminus, while splice variants that contained the complete exon 13 deletion (Del-e10, e13, Del-e10, e13, e15, and Del-e9, e10, e13 splice isoforms) were predicted to generate frame shifted proteins (Table 9).

The relative proportion of the fTERT splice variants in FOSCC

RT-PCR was performed using a primer pair that spanned exons 8 and 14 in order to quantify the proportion of full-length and four of the novel fTERT splice variants (Del- e10; Del-e9, e10; Del-e13 and Del-e9, e10, e13) in all samples. One gingiva sample was excluded from the results since no amplification was detected. The results are listed in

Table 10. The additional six novel fTERT isoforms were amplified but were not quantified.

Discussion

Up-regulation of telomerase activity has been reported in 95% of all human cancers and up to 86% of human OSCC [154, 155].. Alignment of TERT and the telomerase RNA component (TERC) between various vertebrates including humans and cats revealed that

103

both TERT and TERC are highly conserved between species (80 % identity). In the present study, we investigated the similarity between feline telomerase activity and fTERT mRNA expression. We observed increased telomerase activity in all 3 FOSCC cell lines and two of three FOSCC tumors. These findings are comparable to what occurs in human OSCC.

The absence of telomerase activity has been found in 10-15% of all human cancers [173]. A subset of telomerase negative cancers utilized the alternative lengthening telomere (ALT) to maintain telomere length independently from the activity of telomerase [174]. ALT was detected in approximately 4% (228/6110) of human cancers [173]. Heterogeneity of telomere length and the presence of promyelocytic leukemia (PML) protein nuclear bodies have been used to identify the ALT phenotype

[175]. In this study, telomerase activity was not increased in 1 of the 3 FOSCC tumors.

The ALT mechanism may have been activated in this cancer since a low level of telomerase activity was detected. To confirm ALT in this FOSCC sample, measuring telomere length, detection of DNA circles (C-circles) as well as identification of PML protein will be required.

Two sets of fTERT primers for amplification of different regions of fTERT cDNA were used to amplify total fTERT mRNA in each sample. The levels of full-length fTERT mRNA expression from the two different regions were quantified. We found that fTERT mRNA levels in FOSCC cells and tumors varied between the two regions. Future

104

investigations will be necessary to determine whether there is additional alternative splicing in these two selected coding regions.

A strong correlation between telomerase activity and the level of full-length

TERT mRNA expression has been demonstrated in human cancers [152]. In this study, the total fTERT mRNA level did not consistently correlate to the level of telomerase activity in each sample. Thus, the discrepancy between fTERT mRNA levels and telomerase activity could be associated with alternative splicing of fTERT mRNA (see below).

To date, 22 human-specific alternative splice variants of TERT have been reported [167, 171]. The hTERTα-deletion (deletion in exon 6), the hTERTβ-deletion

(deletion of exons 7 and 8) and the hTERTγ-deletion (deletion of exon 11) are the three major splice variants of TERT in humans [167, 171]. In chickens, 18 alternative splice isoforms (V1-V18) located between exons 2 and 12 have been described [172]. The full- length and the V4 splice variant (deletion of exon 5) were the major TERT variants in chickens. In the present study, full-length and 10 novel, cat-specific alternative splice variants of fTERT were amplified, cloned and sequenced from SCCF2 cells. The alternative splice variants of fTERT that were amplified in this study were completely different from the variants of hTERT. The deletion of exons 6, 7, 8 or 11 that were found in the dominant splice variants of hTERT did not occur in any of the 10 alternative splice variants of fTERT. In addition, there was no counterpart to the splice variants in the 3’ end of fTERT in other species for which fTERT alternative splicing is known. However,

105

the five splice variants in the 5’ end of fTERT had some similarity to the chicken TERT splice variants. The deletion of exon 3 was found in the Del-e2, e3 variant of fTERT as well as in the V3 variant of chicken TERT. The insertion of intron 3 was found in the

Del-e2 and Ins-i3_1, 2, 3 and 4 variants of fTERT and in the V2 variant of chicken

TERT. The complexity of fTERT pre-mRNA alternative splicing and the distinct ASPSs found in this study demonstrated the unique and complex regulation of the TERT gene in cats.

The major variants of hTERT were speculated to have an important role in the pathogenesis of human cancers [168, 171]. In this study, we found that full-length and the

Del-e10 variant, which has a complete deletion of exon 10, were the major fTERT variants in all 3 FOSCC cell lines, 2 of 3 FOSCC tumors and normal cat testis. Full- length fTERT was not the major variant in 1 FOSCC tumor (tumor 3), normal cat gingiva and tongue. Interestingly, full-length and the Del-e10 variant of fTERT were the major variants in the samples with high levels of telomerase activity. This finding suggested that both full-length and the Del-e10 splice variant of fTERT may contribute to the higher telomerase activity in FOSCC. The ability of full-length and Del-e10 as well as other splice variants of fTERT to maintain the telomere length and immortalize telomerase- negative cells in cats is currently under investigation.

106

Conclusions

Increased telomerase activity was found in FOSCC cell lines and two of three FOSCC tumors. However, the level of telomerase activity in FOSCC cell lines and tumors did not consistently correlate with the level of total fTERT mRNA expression. Several variants of fTERT were identified. The disagreement between telomerase activity and the level of fTERT mRNA expression may have been a result of the alternative mRNA splicing. In the samples with high levels of telomerase activity, full-length and the Del-e10 variant of fTERT were the major variants. This observation suggests that full-length and the Del- e10 splice variants play a major role in the up-regulation of telomerase activity. The alternative splice variants identified in this study will be essential tools for studying the role of alternative splicing in telomerase activity in feline cancer.

Target Primer Primer pair Product (bp) exon8-exon14 fTERT4S/7AS forward: 5’-AACCACGTCATCAGGATCGG -3’ 659 reverse: 5’-CGGAGGAAAAAGGAGGGGTTC-3’ exon6-exon11 fTERT16S/15AS forward: 5’-ACGTGATCAGGCCTCAAGAAAA-3’ 682 reverse: 5’-CTCGAGCGAATGGAGGTCTG-3’ exon4-exon11 fTERT3’ORF1S/fTERT15AS forward: 5’- GACCCACTCTGCTGACATCC-3’ 1,036 reverse: 5’-CTCGAGCGAATGGAGGTCTG-3’ Table 7. PCR primers for semi-quantification of feline TERT alternative splicing

107

Samples fTERT4S/7AS SCCF1/1 36 SCCF1/2 36 SCCF1/3 36 SCCF2/1 34 SCCF2/2 34 SCCF2/3 34 SCCF3/1 40 SCCF3/2 39 SCCF3/3 42 FOSCC1 37 FOSCC2 35 FOSCC3 42 Gingiva-1 41 Gingiva-2 42 Gingiva-3 39 Lip 40 Tongue-1 42 Tongue-2 42 Tongue-3 38 Tongue-4 39 Testis-1 33 Testis-2 34 Tesitis-3 33 Control 21 Table 8. PCR cycles used for individual feline oral squamous cell carcinoma (FOSCC) cell lines, tumors, normal cat oral tissues, normal cat testis and plasmid cDNA (positive control).

108

Isoform Alternative splice sites Splicing type Predicted protein GenBank accession no. Full-length - - fTERT KX620456

Del-e10 Exon 10 (nt 1-72) Exonic deletion In-frame deletion KX620462

Del-e10, e13 Exon 10 (nt 1-72) Exonic deletion Frame shift mutation KX620465 Exon 13 (nt 1-62)

Del-e10, e13,e 15 Exon 10 (nt 1-72) Exonic deletion Frame shift mutation KX620466 Exon 13 (nt 1-62) Exon 15 (nt 1-138) Del-e9, e10 Exon 9 (nt 1-114) Exonic deletion In-frame deletion KX620463 Exon 10 (nt 1-72)

Del-e9, e10, e13 Exon 9 (nt 1-114) Exonic deletion Frame shift mutation KX620464 Exon 10 (nt 1-72) Exon 13 (nt 1-62) Del-e2, e3 Exon 2 (nt 1-1,288) Exonic deletion Non-telomerase KX620457 Exon 3 (nt 1-196)

Del-e2, Ins-i3_1 Exon 2 (nt 1-1,233) Exonic deletion Non-telomerase KX620458 Intron 3 (nt 1-1,326) Intronic insertion

Del-e2, Ins-i3_2 Exon 2 (nt 1-1,288) Exonic deletion Non-telomerase KX620459 Intron 3 (nt 230-799) Intronic insertion

Del-e2, Ins-i3_3 Exon 2 (nt 1-1,288) Exonic deletion Non-telomerase KX620460 Intron 3 (nt 230-1,326) Intronic insertion

Del-e2, Ins-i3_3 Exon 2 (nt 1-1,288) Exonic deletion Non-telomerase KX620461 Intron 3 (nt 1-799) Intronic insertion

Table 9. Full-length and alternative splice variants of feline TERT

109

Cell lines FOSCC tissues Normal tissue Splice variants SCCF1 SCCF2 SCCF3 1 2 3 Gingiva Lip Tongue Testis (n=3) (n=3) (n=3) (n=1) (n=1) (n=1) (n=2) (n=1) (n=4) (n=3) Full-length 0.37 0.49 0.50 0.17 0.44 0.41 0.20 0.43 0.19 0.43 Del-e10 0.39 0.24 0.49 0.36 0.55 0.58 0.27 0.15 0.35 0.44 Del-e10, e13 0.15 0.17 0.00 0.27 0.00 0.00 0.27 0.15 0.30 0.08 Del-e9, e10 0.07 0.09 0.00 0.20 0.00 0.00 0.26 0.08 0.12 0.00 Del-e9, e10, e13 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.19 0.04 0.05 Table 10. Relative proportion of full-length feline TERT and feline TERT alternative splice variants

110

Figure 25. Relative feline TERT (fTERT) mRNA expression

SCCF2 (n=3) had greater fTERT (exons 3-5) mRNA compared to normal cat oral tissues

(n=5) (**p < 0.0001) (A). SCCF3 (n=3) had lower fTERT (exons 3-5) compared to normal cat oral tissues (*p = 0.025) (A). Relative fTERT (exons 11-12) mRNA in SCCF1

(n=3) (*p = 0.014) and SCCF2 (n=3) (*p = 0.0035) were greater than normal cat oral tissues (n=5) (B). Two of three feline oral squamous cell carcinoma tumors (tumors 2 and

3) had greater levels of fTERT (exons 3-5 and exons 11-12) mRNA than normal cat oral tissues (A and B). The Y-axis of theses graphs is on log10 scale.

111

Figure 26. Telomerase activity in feline oral squamous cell carcinoma (FOSCC) cell lines and tumors and normal cat oral tissues

All 3 feline oral squamous cell carcinoma cell lines (SCCF1, SCCF2, SCCF3) had significantly greater telomerse activity (˃10-fold) compared to normal cat oral tissues

(gingiva (n=2), lip (n=1) and tongue (n=1) (**p ˂ 0.0001). FOSCC tumor 2 (*p =

0.0006) and tumor 3 (**p ˂ 0.0001) had significantly greater telomerase activity (˃ 10- fold) compared to normal cat oral tissues. Telomerase activity in all samples were measured in triplicate. Two sides bars indicate standard deviation (SD) of data. The Y- axis of this graph is presented in log10 scale.

112

Figure 27 Diagram of feline TERT (fTERT) gene, mRNA and alternative splice isoforms

The fTERT gene has 16 exons (black boxes) and 15 introns (A). The numbers in parentheses indicate the number of nucleotides in each exon. fTERT mRNA contains 16 exons (B). Alternative splice sites were observed in exons 2, 3, 9, 10, 13 and 15 (gray boxes). Five feline TERT alternative splice variants, Del-e10; Del-e10, e13; Del-e10, e13, e15; Del-e9, e10, and Del-e9, e10, e13 isoforms were amplified and sequenced in the 3’open reading frame (ORF) of feline TERT cDNA (C). Five feline TERT alternative

113

splice variants, Del-e2, e3; Del-e2, Ins-i3_1; Del-e2, Ins-i3_2; Del-e2, Ins-i3_3 and, Del- e2, Ins-i3_4 were amplified, cloned and sequenced in the 5’ORF of fTERT cDNA (D).

Figure 28. Full-length and splice variants of feline TERT in feline oral squamous cell carcinoma (FOSCC) cell lines and tumors and normal cat tissues

The fTERT 4S/7AS primers were used to amplify full-length and alternative splice variants of fTERT between exons 8 and 14. The full-length and 4 cat-specific alternative splice variants, including Del-e10; Del-e10, e13; Del-e9, e10 and Del-e9, e10, e13 were amplified in FOSCC cell lines and tumors (upper panel) and in normal cat tissues (lower panel). The positive controls for each variant and the markers of DNA size were indicated on the right and the left sides of both panels, respectively.

114

References

1. Kim, S., Animal models of cancer in the head and neck region. Clin Exp Otorhinolaryngol, 2009. 2(2): p. 55-60.

2. Rothenberg, S.M. and L.W. Ellisen, The molecular pathogenesis of head and neck squamous cell carcinoma. Journal of Clinical Investigation, 2012. 122(6): p. 1951-7.

3. Joshi, P., et al., Head and neck cancers in developing countries. Rambam Maimonides Med J, 2014. 5(2): p. e0009.

4. Siegel, R., D. Naishadham, and A. Jemal, Cancer statistics, 2013. CA Cancer J Clin, 2013. 63(1): p. 11-30.

5. Jimi, E., et al., The RANKL/RANK system as a therapeutic target for bone invasion by oral squamous cell carcinoma (Review). Int J Oncol, 2013. 42(3): p. 803-9.

6. Ferrarotto, R. and K.A. Gold, Afatinib in the treatment of head and neck squamous cell carcinoma. Expert Opinion on Investigational Drugs, 2014. 23(1): p. 135- 43.

7. Zygogianni, A.G., et al., Oral squamous cell cancer: early detection and the role of alcohol and smoking. Head Neck Oncol, 2011. 3: p. 2.

8. Smith, L.P. and G.R. Thomas, Animal models for the study of squamous cell carcinoma of the upper aerodigestive tract: a historical perspective with review of their utility and limitations. Part A. Chemically-induced de novo cancer, syngeneic animal 115

models of HNSCC, animal models of transplanted xenogeneic human tumors. Int J Cancer, 2006. 118(9): p. 2111-22.

9. Masood, R., et al., A novel orthotopic mouse model of head and neck cancer and lymph node metastasis. Oncogenesis, 2013. 2: p. e68.

10. Bishop, J.A., et al., Detection of transcriptionally active high-risk HPV in patients with head and neck squamous cell carcinoma as visualized by a novel E6/E7 mRNA in situ hybridization method. American Journal of Surgical Pathology, 2012. 36(12): p. 1874-82.

11. Geisler, S.A., et al., p16 and p53 Protein expression as prognostic indicators of survival and disease recurrence from head and neck cancer. Clinical Cancer Research, 2002. 8(11): p. 3445-53.

12. Ukpo, O.C., et al., High-risk human papillomavirus E6/E7 mRNA detection by a novel in situ hybridization assay strongly correlates with p16 expression and patient outcomes in oropharyngeal squamous cell carcinoma. American Journal of Surgical Pathology 2011. 35(9): p. 1343-50.

13. Shinohara, S., et al., Prognostic impact of p16 and p53 expression in oropharyngeal squamous cell carcinomas. Japanese Journal of Clinical Oncology, 2014. 44(3): p. 232-40.

14. Boyce, B.F. and L. Xing, Biology of RANK, RANKL, and osteoprotegerin. Arthritis Research & Therapy, 2007. 9 Suppl 1: p. S1.

15. Valero, C., et al., Osteoprotegerin and bone mass in squamous cell head and neck cancer patients. Calcif Tissue Int, 2006. 78(6): p. 343-7.

16. Parise Junior, O., et al., Prognostic impact of p53, c-erbB-2 and epidermal growth factor receptor on head and neck carcinoma. Sao Paulo Medical Journal, 2004. 122(6): p. 264-8. 116

17. Young, M.R. and T.A. Day, Immune regulatory activity of vitamin d3 in head and neck cancer. Cancers (Basel), 2013. 5(3): p. 1072-85.

18. Vermorken, J.B., et al., Open-label, uncontrolled, multicenter phase II study to evaluate the efficacy and toxicity of cetuximab as a single agent in patients with recurrent and/or metastatic squamous cell carcinoma of the head and neck who failed to respond to platinum-based therapy. Journal of Clinical Oncology, 2007. 25(16): p. 2171-7.

19. Courthod, G., et al., The role of microRNA in head and neck cancer: current knowledge and perspectives. Molecules, 2014. 19(5): p. 5704-16.

20. Sethi, N., et al., MicroRNAs and head and neck cancer: reviewing the first decade of research. European Journal of Cancer, 2014. 50(15): p. 2619-35.

21. Chen, D., et al., MicroRNA Deregulations in Head and Neck Squamous Cell Carcinomas. Journal of Oral & Maxillofacial Research, 2013. 4(1): p. e2.

22. Kimura, S., et al., Expression of microRNAs in squamous cell carcinoma of human head and neck and the esophagus: miR-205 and miR-21 are specific markers for HNSCC and ESCC. Oncol Rep, 2010. 23(6): p. 1625-33.

23. Ramdas, L., et al., miRNA expression profiles in head and neck squamous cell carcinoma and adjacent normal tissue. Head Neck, 2009. 31(5): p. 642-54.

24. Wypij, J.M., A naturally occurring feline model of head and neck squamous cell carcinoma. Patholog Res Int, 2013. 2013: p. 502197.

25. Ballegeer, E.A., et al., Evaluation of hypoxia in a feline model of head and neck cancer using (6)(4)Cu-ATSM positron emission tomography/computed tomography. BMC Cancer, 2013. 13: p. 218.

117

26. Casto, B.C., et al., Chemoprevention of oral cancer by lyophilized strawberries. Anticancer Research, 2013. 33(11): p. 4757-66.

27. Casto, B.C., et al., Chemoprevention of oral cancer by black raspberries. Anticancer Research 2002. 22(6C): p. 4005-15.

28. Schwartz, J.L., D. Sloane, and G. Shklar, Prevention and inhibition of oral cancer in the hamster buccal pouch model associated with carotenoid immune enhancement. Tumour Biology, 1989. 10(6): p. 297-309.

29. Warner, B.M., et al., Chemoprevention of oral cancer by topical application of black raspberries on high at-risk mucosa. Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology, 2014. 118(6): p. 674-83.

30. Lu, S.L., H. Herrington, and X.J. Wang, Mouse models for human head and neck squamous cell carcinomas. Head Neck, 2006. 28(10): p. 945-54.

31. Salley, J.J., Experimental carcinogenesis in the cheek pouch of the Syrian hamster. J Dent Res, 1954. 33(2): p. 253-62.

32. Vairaktaris, E., et al., The hamster model of sequential oral oncogenesis. Oral Oncology, 2008. 44(4): p. 315-24.

33. Li, J., et al., Expression and characterization of Syrian golden hamster p16, a homologue of human tumor suppressor p16 INK4A. Biochem Biophys Res Commun, 2003. 304(2): p. 241-7.

34. Li, J., et al., Tumor suppressor p16(INK4A)/Cdkn2a alterations in 7, 12- dimethylbenz(a)anthracene (DMBA)-induced hamster cheek pouch tumors. Mol Carcinog, 2008. 47(10): p. 733-8.

118

35. Muscarella, P., et al., Identification and sequencing of the Syrian Golden hamster (Mesocricetus auratus) p16(INK4a) and p15(INK4b) cDNAs and their homozygous gene deletion in cheek pouch and pancreatic tumor cells. Gene, 2001. 278(1-2): p. 235-43.

36. Schoop, R.A., M.H. Noteborn, and R.J. Baatenburg de Jong, A mouse model for oral squamous cell carcinoma. J Mol Histol, 2009. 40(3): p. 177-81.

37. Tanaka, T., et al., Apc-Mutant Kyoto Apc Delta (KAD) Rats Are Susceptible to 4- NQO-Induced Tongue Carcinogenesis. Cancers (Basel), 2014. 6(3): p. 1522-39.

38. Fixler, D., et al., Diffusion Reflection: A Novel Method for Detection of Oral Cancer. Journal of Dental Research, 2014. 93(6): p. 602-606.

39. Mallery, S.R., et al., Clinical and biochemical studies support smokeless tobacco's carcinogenic potential in the human oral cavity. Cancer Prev Res (Phila), 2014. 7(1): p. 23-32.

40. Guttenplan, J.B., et al., Mutagenesis and carcinogenesis induced by dibenzo[a,l]pyrene in the mouse oral cavity: a potential new model for oral cancer. International Journal of Cancer, 2012. 130(12): p. 2783-90.

41. Ward, B.B., et al., Targeted dendrimer chemotherapy in an animal model for head and neck squamous cell carcinoma. J Oral Maxillofac Surg, 2011. 69(9): p. 2452-9.

42. Tang, A.L., et al., UM-SCC-104: a new human papillomavirus-16-positive cancer stem cell-containing head and neck squamous cell carcinoma cell line. Head Neck, 2012. 34(10): p. 1480-91.

43. Nomura, T., et al., Establishment of a murine model of bone invasion by oral squamous cell carcinoma. Oral Oncol, 2007. 43(3): p. 257-62.

44. Aubry, K., et al., Characterization of a new rat model of head and neck squamous cell carcinoma. In Vivo, 2008. 22(4): p. 403-8. 119

45. Martin, C.K., et al., Characterization of bone resorption in novel in vitro and in vivo models of oral squamous cell carcinoma. Oral Oncology, 2012. 48(6): p. 491-9.

46. Lin, C.J., et al., Head and neck squamous cell carcinoma cell lines: established models and rationale for selection. Head Neck, 2007. 29(2): p. 163-88.

47. Martin, C.K., et al., Zoledronic acid reduces bone loss and tumor growth in an orthotopic xenograft model of osteolytic oral squamous cell carcinoma. Cancer Research, 2010. 70(21): p. 8607-16.

48. Myers, J.N., et al., An orthotopic nude mouse model of oral tongue squamous cell carcinoma. Clin Cancer Res, 2002. 8(1): p. 293-8.

49. Sano, D. and J.N. Myers, Xenograft models of head and neck cancers. Head Neck Oncol, 2009. 1: p. 32.

50. Hwang, Y.S., et al., An orthotopic and osteolytic model with a newly established oral squamous cell carcinoma cell line. Arch Oral Biol, 2013. 58(2): p. 218-25.

51. Nakagawa, H., et al., The targeting of the cyclin D1 oncogene by an Epstein-Barr virus promoter in transgenic mice causes dysplasia in the tongue, esophagus and forestomach. Oncogene, 1997. 14(10): p. 1185-90.

52. Thomas, G.R. and L.P. Smith, Animal models for the study of squamous cell carcinoma of the upper aerodigestive tract: a historical perspective with review of their utility and limitations part B. Transgenic mouse models. Int J Cancer, 2006. 118(10): p. 2379-80.

53. Mueller, A., et al., A transgenic mouse model with cyclin D1 overexpression results in cell cycle, epidermal growth factor receptor, and p53 abnormalities. Cancer Res, 1997. 57(24): p. 5542-9.

120

54. Opitz, O.G., et al., A mouse model of oral-esophageal carcinogenesis. Onkologie, 2005. 28(1): p. 44-8.

55. Nishikawa, T., et al., High susceptibility of p53 knockout mice to esophageal and urinary bladder carcinogenesis induced by N, N-dibutylnitrosamine. Cancer Lett, 2003. 194(1): p. 45-54.

56. Ku, T.K., et al., Loss of p53 expression correlates with metastatic phenotype and transcriptional profile in a new mouse model of head and neck cancer. Mol Cancer Res, 2007. 5(4): p. 351-62.

57. Herber, R., et al., Squamous epithelial hyperplasia and carcinoma in mice transgenic for the human papillomavirus type 16 E7 oncogene. J Virol, 1996. 70(3): p. 1873-81.

58. Xie, W., et al., Conditional expression of the ErbB2 oncogene elicits reversible hyperplasia in stratified epithelia and up-regulation of TGFalpha expression in transgenic mice. Oncogene, 1999. 18(24): p. 3593-607.

59. Caulin, C., et al., Inducible activation of oncogenic K-ras results in tumor formation in the oral cavity. Cancer Reserach, 2004. 64(15): p. 5054-8.

60. Vitale-Cross, L., et al., Conditional expression of K-ras in an epithelial compartment that includes the stem cells is sufficient to promote squamous cell carcinogenesis. Cancer Res, 2004. 64(24): p. 8804-7.

61. Gilbert, M.R., Lim, C.W., Kim, S., Head and neck cancer, in Experiment Metastasis: Modelling and Analysis, M. Anastasia, Editor. 2013, Spinger science: online. p. 7-16.

62. Nemec, A., et al., Oral papillary squamous cell carcinoma in twelve dogs. J Comp Pathol, 2014. 150(2-3): p. 155-61.

121

63. Fulton, A.J., et al., Risk factors associated with survival in dogs with nontonsillar oral squamous cell carcinoma 31 cases (1990-2010). Journal of The American Veterinary Medical Association, 2013. 243(5): p. 696-702.

64. Grier, C.K. and M.N. Mayer, Radiation therapy of canine nontonsillar squamous cell carcinoma. Canadian Veterinary Journal, 2007. 48(11): p. 1189-91.

65. Nemec, A., et al., Histological subtypes of oral non-tonsillar squamous cell carcinoma in dogs. Journal of Comparative Pathology, 2012. 147(2-3): p. 111-20.

66. de Vos, J.P., et al., Piroxicam and carboplatin as a combination treatment of canine oral non-tonsillar squamous cell carcinoma: a pilot study and a literature review of a canine model of human head and neck squamous cell carcinoma. Veterinary and Comparative Oncology, 2005. 3(1): p. 16-24.

67. Gardner, D.G., Spontaneous squamous cell carcinomas of the oral region in domestic animals: a review and consideration of their relevance to human research. Oral Diseases, 1996. 2(2): p. 148-54.

68. Boria, P.A., et al., Evaluation of cisplatin combined with piroxicam for the treatment of oral malignant melanoma and oral squamous cell carcinoma in dogs. Journal of the American Veterinary Medical Association, 2004. 224(3): p. 388-94.

69. McCaw, D.L., et al., Treatment of canine oral squamous cell carcinomas with photodynamic therapy. British Journal of Cancer, 2000. 82(7): p. 1297-9.

70. Teifke, J.P., C.V. Lohr, and H. Shirasawa, Detection of canine oral papillomavirus-DNA in canine oral squamous cell carcinomas and p53 overexpressing skin papillomas of the dog using the polymerase chain reaction and non-radioactive in situ hybridization. Veterinary Microbiology, 1998. 60(2-4): p. 119-30.

122

71. Munday, J.S., et al., Multiple oral carcinomas associated with a novel papillomavirus in a dog. Journal of Veterinary Diagnostic Investigation, 2015. 27(2): p. 221-5.

72. Porcellato, I., et al., A retrospective investigation on canine papillomavirus 1 (CPV1) in oral oncogenesis reveals dogs are not a suitable animal model for high-risk HPV-induced oral cancer. PLoS One, 2014. 9(11): p. e112833.

73. Pang, L.Y., et al., Identification of tumour initiating cells in feline head and neck squamous cell carcinoma and evidence for gefitinib induced epithelial to mesenchymal transition. Vet J, 2012. 193(1): p. 46-52.

74. Pontius, J.U., et al., Initial sequence and comparative analysis of the cat genome. Genome Res, 2007. 17(11): p. 1675-89.

75. Snyder, L.A., et al., p53 expression and environmental tobacco smoke exposure in feline oral squamous cell carcinoma. Vet Pathol, 2004. 41(3): p. 209-14.

76. Soltero-Rivera, M.M., et al., Prevalence of regional and distant metastasis in cats with advanced oral squamous cell carcinoma: 49 cases (2005-2011). J Feline Med Surg, 2014. 16(2): p. 164-9.

77. Tannehill-Gregg, S.H., A.L. Levine, and T.J. Rosol, Feline head and neck squamous cell carcinoma: a natural model for the human disease and development of a mouse model. Vet Comp Oncol, 2006. 4(2): p. 84-97.

78. Martin, C.K., et al., Bone-invasive oral squamous cell carcinoma in cats: pathology and expression of parathyroid hormone-related protein. Veterinary Pathology, 2011. 48(1): p. 302-12.

79. Rathore, K., M. Alexander, and M. Cekanova, Piroxicam inhibits Masitinib- induced cyclooxygenase 2 expression in oral squamous cell carcinoma cells in vitro. Translational Research, 2014. 164(2): p. 158-68.

123

80. MacEwen, E.G., Spontaneous tumors in dogs and cats: models for the study of cancer biology and treatment. Cancer Metastasis Rev, 1990. 9(2): p. 125-36.

81. Tannehill-Gregg, S., E. Kergosien, and T.J. Rosol, Feline head and neck squamous cell carcinoma cell line: characterization, production of parathyroid hormone- related protein, and regulation by transforming growth factor-beta. In Vitro Cellular & Developmental Biology-Animal 2001. 37(10): p. 676-83.

82. Bergkvist, G.T., et al., Studies on the inhibition of feline EGFR in squamous cell carcinoma: enhancement of radiosensitivity and rescue of resistance to small molecule inhibitors. Cancer Biology & Therapy, 2011. 11(11): p. 927-37.

83. Bertone, E.R., L.A. Snyder, and A.S. Moore, Environmental and lifestyle risk factors for oral squamous cell carcinoma in domestic cats. Journal of Veterinary Internal Medicine, 2003. 17(4): p. 557-62.

84. Munday, J.S. and D. Aberdein, Loss of retinoblastoma protein, but not p53, is associated with the presence of papillomaviral DNA in feline viral plaques, Bowenoid in situ carcinomas, and squamous cell carcinomas. Veterinary Pathology, 2012. 49(3): p. 538-45.

85. Munday, J.S., I. Gibson, and A.F. French, Papillomaviral DNA and increased p16CDKN2A protein are frequently present within feline cutaneous squamous cell carcinomas in ultraviolet-protected skin. Veterinary Dermatology, 2011. 22(4): p. 360-6.

86. Munday, J.S., et al., Detection of papillomaviral DNA sequences in a feline oral squamous cell carcinoma. Research in Veterinary Science, 2009. 86(2): p. 359-61.

87. Munday, J.S., et al., Detection of papillomaviral sequences in feline Bowenoid in situ carcinoma using consensus primers. Veterinary Dermatology, 2007. 18(4): p. 241-5.

124

88. Munday, J.S. and M. Kiupel, Papillomavirus-associated cutaneous neoplasia in mammals. Veterinary Pathology, 2010. 47(2): p. 254-64.

89. Munday, J.S., et al., Amplification of papillomaviral DNA sequences from a high proportion of feline cutaneous in situ and invasive squamous cell carcinomas using a nested polymerase chain reaction. Veterinary Dermatology, 2008. 19(5): p. 259-63.

90. Munday, J.S., C.G. Knight, and A.F. French, Evaluation of feline oral squamous cell carcinomas for p16CDKN2A protein immunoreactivity and the presence of papillomaviral DNA. Research in Veterinary Science, 2011. 90(2): p. 280-3.

91. Pang, L.Y. and D.J. Argyle, Using naturally occurring tumours in dogs and cats to study telomerase and cancer stem cell biology. Biochim Biophys Acta, 2009. 1792(4): p. 380-91.

92. Postorino-Reeves, N.C., J.M. Turrel, and S.J. Withrow, Oral squamous cell carcinoma in the cat. Journal of the American Animal Hospital Association, 1993. 29(5): p. 438-441.

93. Genden, E.M., et al., Neck disease and distant metastases. Oral Oncology, 2003. 39(3): p. 207-12.

94. Hayes, A.M., et al., Survival of 54 cats with oral squamous cell carcinoma in United Kingdom general practice. Journal of Small Animal Practice, 2007. 48(7): p. 394- 9.

95. DiBernardi, L., et al., Study of feline oral squamous cell carcinoma: potential target for cyclooxygenase inhibitor treatment. Prostaglandins, Leukotrienes, and Essential Fatty Acids, 2007. 76(4): p. 245-50.

96. Goldsby, R.E., et al., Feasibility and dose discovery analysis of zoledronic acid with concurrent chemotherapy in the treatment of newly diagnosed metastatic osteosarcoma: a report from the Children's Oncology Group. European Journal of Cancer, 2013. 49(10): p. 2384-91. 125

97. Wypij, J.M., et al., In vivo and in vitro efficacy of zoledronate for treating oral squamous cell carcinoma in cats. Journal of Veterinary Internal Medicine 2008. 22(1): p. 158-63.

98. Wypij, J.M. and D.A. Heller, Pamidronate disodium for palliative therapy of feline bone-invasive tumors. Vet Med Int, 2014. 2014: p. 675172.

99. Ganci, F., et al., Molecular genetics and biology of head and neck squamous cell carcinoma: Implications for diagnosis, prognosis and treatment, in Head and Neck Cancer, M. Agulnik, Editor. 2012, InTech Europe: Rijeka, Croatia. p. 73-122.

100. Yuen, P.W., et al., Clinicopathological significance of p16 gene expression in the surgical treatment of head and neck squamous cell carcinomas. Journal of Clinical Pathology, 2002. 55(1): p. 58-60.

101. Yoshikawa, H., et al., Immunohistochemical characterization of feline oral squamous cell carcinoma. Am J Vet Res, 2012. 73(11): p. 1801-6.

102. Bergkvist, G.T., et al., Expression of epidermal growth factor receptor (EGFR) and Ki67 in feline oral squamous cell carcinomas (FOSCC). Veterinary and Comparative Oncology 2011. 9(2): p. 106-17.

103. Takayama, Y., et al., Parathyroid-related protein plays a critical role in bone invasion by oral squamous cell carcinoma. Int J Oncol, 2010. 36(6): p. 1387-94.

104. Hayes, A., et al., COX-1 and COX-2 expression in feline oral squamous cell carcinoma. Journal of Comparative Pathology, 2006. 135(2-3): p. 93-9.

105. Martin, C.K., et al., Combined zoledronic acid and meloxicam reduced bone loss and tumour growth in an orthotopic mouse model of bone-invasive oral squamous cell carcinoma. Vet Comp Oncol, 2013: p. epub May 8, 2013. 126

106. Leemans, C.R., B.J. Braakhuis, and R.H. Brakenhoff, The molecular biology of head and neck cancer. Nature Reviews Cancer, 2011. 11(1): p. 9-22.

107. Snyder, L.A., et al., p53 expression and environmental tobacco smoke exposure in feline oral squamous cell carcinoma. Veterinary Pathology, 2004. 41(3): p. 209-14.

108. Langendijk, J.A. and A. Psyrri, The prognostic significance of p16 overexpression in oropharyngeal squamous cell carcinoma: implications for treatment strategies and future clinical studies. Annnals of Oncology, 2010. 21(10): p. 1931-4.

109. Reed, A.L., et al., High frequency of p16 (CDKN2/MTS-1/INK4A) inactivation in head and neck squamous cell carcinoma. Cancer Research, 1996. 56(16): p. 3630-3.

110. Chen, Y.J., et al., High p16 expression predicts a positive response to chemoradiotherapy in stage IVa/b head and neck squamous cell carcinoma. Asian Pacific Journal of Cancer Prevention, 2011. 12(3): p. 649-55.

111. Singhi, A.D. and W.H. Westra, Comparison of human papillomavirus in situ hybridization and p16 immunohistochemistry in the detection of human papillomavirus- associated head and neck cancer based on a prospective clinical experience. Cancer, 2010. 116(9): p. 2166-73.

112. Lesnikova, I., et al., Rapid, sensitive, type specific PCR detection of the E7 region of human papillomavirus type 16 and 18 from paraffin embedded sections of cervical carcinoma. Infectious Agents and Cancer, 2010. 5: p. 2.

113. Simmons, J.K., et al., Canine prostate cancer cell line (Probasco) produces osteoblastic metastases in vivo. The Prostate, 2014. 74(13): p. 1251-65.

114. Chan, S.Y., et al., Analysis of genomic sequences of 95 papillomavirus types: uniting typing, phylogeny, and taxonomy. Journal of Virology, 1995. 69(5): p. 3074-83.

127

115. Cunningham, L.L., Jr., et al., Overexpression of p16INK4 is a reliable marker of human papillomavirus-induced oral high-grade squamous dysplasia. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, and Endodontics, 2006. 102(1): p. 77-81.

116. Verschuur, H.P., et al., A matched control study of treatment outcome in young patients with squamous cell carcinoma of the head and neck. Laryngoscope, 1999. 109(2 Pt 1): p. 249-58.

117. Shiboski, C.H., B.L. Schmidt, and R.C. Jordan, Tongue and tonsil carcinoma: increasing trends in the U.S. population ages 20-44 years. Cancer, 2005. 103(9): p. 1843- 9.

118. O'Neill, S.H., et al., Detection of human papillomavirus DNA in feline premalignant and invasive squamous cell carcinoma. Veterinary Dermatology, 2011. 22(1): p. 68-74.

119. Lie, A.K. and G. Kristensen, Human papillomavirus E6/E7 mRNA testing as a predictive marker for cervical carcinoma. Expert Review of Molecular Diagnostics 2008. 8(4): p. 405-15.

120. Merlo, A., et al., 5' CpG island methylation is associated with transcriptional silencing of the tumour suppressor p16/CDKN2/MTS1 in human cancers. Nature Medicine, 1995. 1(7): p. 686-92.

121. Liggett, W.H., Jr. and D. Sidransky, Role of the p16 tumor suppressor gene in cancer. Journal of Clinical Oncology, 1998. 16(3): p. 1197-206.

122. Munday, J.S., et al., The presence of p16 CDKN2A protein immunostaining within feline nasal planum squamous cell carcinomas is associated with an increased survival time and the presence of papillomaviral DNA. Veterinary Pathology, 2013. 50(2): p. 269- 73.

128

123. Jin, X., et al., Cell cycle arrest and inhibition of tumor cell proliferation by the p16INK4 gene mediated by an adenovirus vector. Cancer Research 1995. 55(15): p. 3250-3.

124. Saunders, M.E., et al., Patterns of p53 gene mutations in head and neck cancer: full-length gene sequencing and results of primary radiotherapy. Clinical Cancer Research, 1999. 5(9): p. 2455-63.

125. Peltonen, J.K., et al., p53 in head and neck cancer: functional consequences and environmental implications of TP53 mutations. Head Neck Oncology, 2010. 2: p. 36.

126. Brown, C.J., et al., Awakening guardian angels: drugging the p53 pathway. Nature Reviews Cancer, 2009. 9(12): p. 862-73.

127. Geradts, J., R.E. Wilentz, and H. Roberts, Immunohistochemical [corrected] detection of the alternate INK4a-encoded tumor suppressor protein p14(ARF) in archival human cancers and cell lines using commercial antibodies: correlation with p16(INK4a) expression. Modern Pathology, 2001. 14(11): p. 1162-8.

128. Bilgic, O., et al., Feline Oral Squamous Cell Carcinoma: Clinical Manifestations and Literature Review. J Vet Dent, 2015. 32(1): p. 30-40.

129. Martin, C.K., et al., Bone-invasive oral squamous cell carcinoma in cats: pathology and expression of parathyroid hormone-related protein. Vet Pathol, 2011. 48(1): p. 302-12.

130. Martin, C.K., et al., Zoledronic acid reduces bone loss and tumor growth in an orthotopic xenograft model of osteolytic oral squamous cell carcinoma. Cancer Res, 2010. 70(21): p. 8607-16.

131. Wypij, J.M., et al., In vivo and in vitro efficacy of zoledronate for treating oral squamous cell carcinoma in cats. J Vet Intern Med, 2008. 22(1): p. 158-63.

129

132. Martin, C.K., et al., Characterization of bone resorption in novel in vitro and in vivo models of oral squamous cell carcinoma. Oral Oncol, 2012. 48(6): p. 491-9.

133. Rosol, T.J., et al., Animal models of bone metastasis. Cancer, 2003. 97(3 Suppl): p. 748-57.

134. Boyce, B.F., et al., Regulation of bone remodeling and emerging breakthrough drugs for osteoporosis and osteolytic bone metastases. Kidney Int Suppl, 2003(85): p. S2-5.

135. Shin, M., et al., The inhibition of RANKL/RANK signaling by osteoprotegerin suppresses bone invasion by oral squamous cell carcinoma cells. Carcinogenesis, 2011. 32(11): p. 1634-40.

136. Holen, I. and C.M. Shipman, Role of osteoprotegerin (OPG) in cancer. Clin Sci (Lond), 2006. 110(3): p. 279-91.

137. Bracco, L., et al., Methods and platforms for the quantification of splice variants' expression. Prog Mol Subcell Biol, 2006. 44: p. 1-25.

138. Simmons, J.K., et al., Canine prostate cancer cell line (Probasco) produces osteoblastic metastases in vivo. Prostate, 2014. 74(13): p. 1251-65.

139. Gilda, J.E., et al., Western Blotting Inaccuracies with Unverified Antibodies: Need for a Western Blotting Minimal Reporting Standard (WBMRS). PLoS One, 2015. 10(8): p. e0135392.

140. Thudi, N.K., et al., Zoledronic acid decreased osteolysis but not bone metastasis in a nude mouse model of canine prostate cancer with mixed bone lesions. Prostate, 2008. 68(10): p. 1116-25.

130

141. Jenkins, D.E., et al., Bioluminescent human breast cancer cell lines that permit rapid and sensitive in vivo detection of mammary tumors and multiple metastases in immune deficient mice. Breast Cancer Res, 2005. 7(4): p. R444-54.

142. Tannehill-Gregg, S., E. Kergosien, and T.J. Rosol, Feline head and neck squamous cell carcinoma cell line: characterization, production of parathyroid hormone- related protein, and regulation by transforming growth factor-beta. In Vitro Cell Dev Biol Anim, 2001. 37(10): p. 676-83.

143. Tada, T., et al., Oral squamous cell carcinoma cells induce osteoclast differentiation by suppression of osteoprotegerin expression in osteoblasts. Int J Cancer, 2005. 116(2): p. 253-62.

144. Waterman, E.A., et al., The antibody MAB8051 directed against osteoprotegerin detects carbonic anhydrase II: implications for association studies with human cancers. Int J Cancer, 2007. 121(9): p. 1958-66.

145. Morinaga, T., et al., Cloning and characterization of the gene encoding human osteoprotegerin/osteoclastogenesis-inhibitory factor. Eur J Biochem, 1998. 254(3): p. 685-91.

146. Corey, E., et al., Osteoprotegerin in prostate cancer bone metastasis. Cancer Res, 2005. 65(5): p. 1710-8.

147. Mergny, J.L., et al., Natural and pharmacological regulation of telomerase. Nucleic Acids Res, 2002. 30(4): p. 839-65.

148. Harley, C.B., et al., Telomerase, cell immortality, and cancer. Cold Spring Harb Symp Quant Biol, 1994. 59: p. 307-15.

149. Keith, W.N., T.R. Jeffry Evans, and R.M. Glasspool, Telomerase and cancer: time to move from a promising target to a clinical reality. J Pathol, 2001. 195(4): p. 404- 14.

131

150. Chen, J.L., M.A. Blasco, and C.W. Greider, Secondary structure of vertebrate telomerase RNA. Cell, 2000. 100(5): p. 503-14.

151. Olyslaegers, D.A., et al., Generation and characterization of feline arterial and venous endothelial cell lines for the study of the vascular endothelium. BMC Vet Res, 2013. 9: p. 170.

152. Ruden, M. and N. Puri, Novel anticancer therapeutics targeting telomerase. Cancer Treat Rev, 2013. 39(5): p. 444-56.

153. Lee, B.K., et al., Diagnostic and prognostic relevance of expression of human telomerase subunits in oral cancer. Int J Oncol, 2001. 19(5): p. 1063-8.

154. Ries, J.C., et al., Correlation of telomerase activity, clinical prognosis and therapy in oral carcinogenesis. Anticancer Res, 2001. 21(2A): p. 1057-63.

155. Chang, L.Y., et al., Telomerase activity and in situ telomerase RNA expression in oral carcinogenesis. J Oral Pathol Med, 1999. 28(9): p. 389-96.

156. McKevitt, T.P., et al., A cohort study of telomere and telomerase biology in cats. Am J Vet Res, 2003. 64(12): p. 1496-9.

157. Gomes, N.M., et al., Comparative biology of mammalian telomeres: hypotheses on ancestral states and the roles of telomeres in longevity determination. Aging Cell, 2011. 10(5): p. 761-8.

158. Calado, R.T. and B. Dumitriu, Telomere dynamics in mice and humans. Semin Hematol, 2013. 50(2): p. 165-74.

159. Uyama, R., et al., Establishment and characterization of eight feline mammary adenocarcinoma cell lines. J Vet Med Sci, 2005. 67(12): p. 1273-6.

132

160. Cadile, C.D., et al., Telomerase activity as a marker for malignancy in feline tissues. Am J Vet Res, 2001. 62(10): p. 1578-81.

161. Fusaro, L., et al., Quantitative analysis of telomerase in feline mammary tissues. J Vet Diagn Invest, 2009. 21(3): p. 369-73.

162. Mandrioli, L., et al., Immunohistochemical expression of h-telomerase reverse transcriptase in canine and feline meningiomas. J Vet Sci, 2007. 8(2): p. 111-5.

163. Yazawa, M., et al., Molecular cloning of the feline telomerase reverse transcriptase (TERT) gene and its expression in cell lines and normal tissues. J Vet Med Sci, 2003. 65(5): p. 573-7.

164. Yoshikawa, H., et al., Predicting clinical outcome in feline oral squamous cell carcinoma: tumour initiating cells, telomeres and telomerase. Vet Comp Oncol, 2014.

165. Supsavhad, W., et al., Animal models of head and neck squamous cell carcinoma. Vet J, 2016. 210: p. 7-16.

166. Liu, S. and C. Cheng, Alternative RNA splicing and cancer. Wiley Interdiscip Rev RNA, 2013. 4(5): p. 547-66.

167. Wong, M.S., W.E. Wright, and J.W. Shay, Alternative splicing regulation of telomerase: a new paradigm? Trends Genet, 2014. 30(10): p. 430-8.

168. Wang, Y., et al., Differentiating alternative splice variant patterns of human telomerase reverse transcriptase in thyroid neoplasms. Thyroid, 2008. 18(10): p. 1055- 63.

169. Angelopoulou, K., et al., Canis familiaris telomerase reverse transcriptase undergoes alternative splicing. Mamm Genome, 2008. 19(9): p. 647-53.

133

170. Saeboe-Larssen, S., E. Fossberg, and G. Gaudernack, Characterization of novel alternative splicing sites in human telomerase reverse transcriptase (hTERT): analysis of expression and mutual correlation in mRNA isoforms from normal and tumour tissues. BMC Mol Biol, 2006. 7: p. 26.

171. Liu, Y., et al., Quantification of alternative splicing variants of human telomerase reverse transcriptase and correlations with telomerase activity in lung cancer. PLoS One, 2012. 7(6): p. e38868.

172. Chang, H. and M.E. Delany, Complicated RNA splicing of chicken telomerase reverse transcriptase revealed by profiling cells both positive and negative for telomerase activity. Gene, 2006. 379: p. 33-9.

173. Heaphy, C.M., et al., Prevalence of the alternative lengthening of telomeres telomere maintenance mechanism in human cancer subtypes. Am J Pathol, 2011. 179(4): p. 1608-15.

174. Dilley, R.L. and R.A. Greenberg, ALTernative Telomere Maintenance and Cancer. Trends Cancer, 2015. 1(2): p. 145-156.

175. Henson, J.D., et al., Alternative lengthening of telomeres in mammalian cells. Oncogene, 2002. 21(4): p. 598-610.

134