Mechanisms and Treatment of Bone Resorption in Models of 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

Chelsea Kathleen Martin

Graduate Program in Veterinary Biosciences

The Ohio State University

2010

Dissertation Committee:

Thomas Rosol, Advisor

Steven Weisbrode

Christopher Weghorst

Cheryl London

Chelsea Kathleen Martin

2010

Abstract

Bone invasion negatively effects prognosis for patients with oral squamous cell

carcinoma (OSCC). Despite the clinical significance, mechanisms responsible for OSCC

invasion into bone are not fully understood. There is limited availability of experimental models which recapitulate the bone invasive behavior of OSCC in which to study disease pathogenesis and treatment. The purpose of this body of work was to develop and characterize models of bone invasive OSCC, to investigate the factors responsible for osteoclastic bone resorption, and to evaluate treatment strategies designed to inhibit bone resorption and tumor growth.

The first objective was to characterize a naturally occurring model of bone invasive

OSCC in domestic cats, and to determine if expression of parathyroid hormone-related protein (PTHrP) is associated with osteolytic behavior. Bone invasive feline OSCC was characterized by osteoclastic bone resorption and PTHrP expression and nuclear localization was increased in osteolytic tumors; supporting a relationship between PTHrP expression and bone-invasive behavior.

The second objective was to investigate the mechanisms responsible for OSCC- associated bone resorption in relevant in vitro and in vivo models of bone invasive

OSCC. In bone invasive OSCC cells, PTHrP expression was associated with ability to

stimulate in vitro and in vivo bone resorption, in addition to stimulation of receptor

activator of nuclear factor κB ligand (RANKL) expression and inhibition of

osteoprotegerin (OPG) in murine preosteoblasts. In vitro bone culture released factors

which stimulated tumor cell proliferation and PTHrP expression in feline and human

OSCC cells, suggesting the existence of a vicious cycle between OSCC growth and

osteoclastic bone resorption.

The final two objectives were to evaluate the effectiveness of targeting osteoclastic

bone resorption in a novel, bioluminescent orthotopic model of maxillary invasive

OSCC. Zoledronic acid (ZOL, an aminobisphosphonate osteoclast inhibitor) reduced

tumor growth and bone resorption associated with xenograft growth. OSCC cells

expressed cyclooxygenase-2 (COX-2), and COX-2 expression was stimulated by factors

in bone conditioned medium. A preferential cyxlooxygenase-2 inhibitor (meloxicam) was

more effective than ZOL at reducing xenograft growth but did not have a significant

effect on bone resorption. The combination of meloxicam and ZOL was well tolerated

but did not stimulate additional tumor suppression or inhibition of bone loss compared to

meloxicam or ZOL monotherapy.

These studies support the role of OSCC-expression of PTHrP and COX-2 in the pathogenesis of bone invasive OSCC, and support the presence of a vicious cycle between OSCC growth and bone resorption. Additionally, valuable experimental models of maxillary invasive OSCC were developed for the advancement of OSCC mechanistic and preclinical research.

iii

Dedication

Dedicated to Luke and Sam

Acknowledgments

I am very grateful to my advisor, Dr. Thomas Rosol, for his guidance and support throughout my graduate research training, and to my graduate committee members; Dr.

Steven Weisbrode, Dr. Christopher Weghorst and Dr. Cheryl London, for their helpful input and advice. Additionally, I’m thankful to all the members of the veterinary anatomic pathology training program, in particular Dr. Rosol and Dr. Weisbrode, for providing me the opportunity to participate in the residency program.

Successful completion of my dissertation research would not have been possible without the help of Dr. Wessel Dirksen, who committed a vast amount of time and effort to assist me, and provided a wealth of technical support and advice. Thank you, Wessel!

Additionally, I benefitted personally and professionally from interaction and friendship with all of the past and current members of the Rosol Laboratory. Thank you to Dr.

Sherry Shu, Ms. Jillian Werbeck, Dr. Nandu Thudi, Dr. Lisa Lanigan, Dr. Tobie Wolfe,

Dr. Ramiro Toribio, Dr. Smitha Pilla, Dr. Prasad Nadella, Dr. John Leong, Dr. Eason

Hildreth, Dr. Wendy Lorch, Dr. Jessica Simmons and Ms. Stephanie Flansburg-Cruz for their valuable assistance and support. My research is built upon a strong foundation of feline oral squamous cell carcinoma investigation performed by Dr. Sarah Tannehill-

Gregg. I am also grateful to the undergraduate and veterinary students who assisted me

with these projects, including Ms. Liza Nusinovich, Ms. Kristin Heller and Mr. Dan

Yanik.

I relied on many people in the Department, College and University during my training;

Ms. Angelia Morris for slide digitization and Mr. Alan Flechtner and Ms. Anne Saulsbery

in the Histology/Immunohistochemistry Core for slide preparation, Mr. Tim Vojt for

assistance with figures, Dr. Päivi Rajala-Schultz for assistance with statistics, Ms.

Michelle Carlton (Department of Radiology, Wright Center of Innovation) for microCT

technical support, Dr. Matthew Allen and Ms. Sue Ringler for IVIS support, Dr. Mamuro

Yamaguchi for transmission electron microscopy, Ms. Jadwiga Labanowska at the

Molecular Cytogenetics Shared Resource for karyotyping, and all of the ULAR staff in

the Sisson Hall and Animal Medical Center vivaria for facilitating my in vivo studies. My research experience has benefitted greatly from the willingness of all these people to assist me.

Tumor specimens critical to completing my research objectives were generously provided by IDEXX laboratories, Veterinary Diagnostics Ltd., Clinilab Inc.,

HistoTechniques LTD.; Dr. Paul Stromberg, Dr. Weisbrode, and Dr. Rosol in the

Department of Veterinary Biosciences; Dr. Deborah Prescott at MedVet Medical Center for Pets; Dr. Robert Knapp at Knapp Veterinary Hospital Inc.; Dr. Meredith Weller at

The Cat Doctor; and Dr. Cynthia White at Town and Country Animal Clinic.

My research and training were supported by the Morris Animal Foundation (grant

number D03FE-029); and the National Cancer Institute (award number F32CA130458).

Many heartfelt thanks to Mom, Dad, Grammy, Tara, Tex and Tuesday for their love and support. It has been hard being away from home, but our visits and phone calls always lifted my spirits and gave me strength. My sincerest gratitude to Marn and Larr;

Jen, John, Nic, Eli and Alex; and Lisa for their tremendous support over the past 6 years.

I am especially grateful to Marn, Larr and Jen for their wonderful care of Sam.

Above all, thank you Luke and Sam. I could not have done it without you.

vii

Vita

December 14th, 1973...... Born – Vancouver, Canada

1999...... D.V.M., University of Prince Edward Island

1999-2004 ...... Associate Veterinarian Newaygo Veterinary Services, Michigan

2004 to present ...... Postdoctoral Fellow Department of Veterinary Biosciences, The Ohio State University

2006 to present...... Veterinary Anatomic Pathology Resident Department of Veterinary Biosciences, The Ohio State University

Publications

F.S.B. Kibenge, B. Qian, J.R. Cleghorn and C.K. Martin. Infectious bursal disease virus polyprotein processing does not involve cellular proteases. Archives of Virology 142: 2401-2419, 1997.

F.S.B. Kibenge, S.K. Whyte, K.L. Hammell, D. Rainnie, M.T. Kibenge and C. K. Martin. A dual infection of infectious salmon anaemia (ISA) virus and a togavirus-like virus in ISA of Atlantic salmon Salmo salar in New Brunswick, Canada. Diseases of Aquatic Organisms 42: 11-15, 2000.

N.K.Thudi, C.K.Martin, M.V.P. Nadella, S.A. Fernandez, J.L.Werbeck, J.J. Pinzone and T.J.Rosol. Zoledronic acid decreased osteolysis but not bone metastasis in a nude mouse model of canine prostate cancer with mixed bone lesions. Prostate 68:1116-25, 2008.

viii

S.T. Shu, C.K. Martin, N.K. Thudi, W.P. Dirksen, and T.J. Rosol. Osteolytic bone resorption in adult T-cell leukemia/lymphoma. Leuk Lymphoma 51:702-14, 2010.

C.K. Martin, J.L. Werbeck, N.K. Thudi, L.G. Lanigan, T.D. Wolfe, R.E. Toribio and T.J. Rosol. Zoledronic acid reduces bone loss and tumor growth in an orthotopic xenograft model of osteolytic oral squamous cell carcinoma. Cancer Research, in press, 2010.

C.K. Martin, S.H. Tannehill-Gregg, T.D. Wolfe and T.J. Rosol. Pathology and expression of parathyroid hormone related protein in bone invasive feline oral squamous cell carcinoma. Vet. Pathol., in press, 2010.

N.K. Thudi, C.K. Martin, S. Murahari, S.S. Shu, L.G. Lanigan, J.L. Werbeck, E.T. Keller, L.K. McCauley, J.J. Pinzone and T.J. Rosol. Dickkopf-1 (DKK-1) stimulated prostate cancer growth and metastasis and inhibited bone formation in osteoblastic bone metastasis. The Prostate, in press, 2010.

Fields of Study

Major Field: Veterinary Biological Science

Table of Contents

Abstract...... ii

Dedication...... iv

Acknowledgments...... v

Vita...... viii

Table of Contents...... x

List of Tables ...... xiii

List of Figures...... xiv

Chapter 1 : Bone Invasive Oral Squamous Cell Carcinoma in Cats: Pathology and expression of Parathyroid Hormone-Related Protein ...... 1

Abstract ...... 1

Introduction ...... 2

Materials and Methods...... 3

Results ...... 9

Discussion ...... 13

Chapter 2 : Characterization of Bone Resorption in Novel In Vitro and In Vivo Models of

Oral Squamous Cell Carcinoma...... 33

Abstract ...... 33

Introduction ...... 34

Materials and Methods...... 37

Results ...... 48

Discussion ...... 53

Chapter 3 : Zoledronic Acid Reduced Bone Loss and Tumor Growth in an Orthotopic

Xenograft Model of Osteolytic Oral Squamous Cell Carcinoma ...... 78

Abstract ...... 78

Introduction ...... 79

Materials and Methods...... 81

Results ...... 86

Discussion ...... 90

Chapter 4 : Combined Zoledronic Acid and Meloxicam Reduced Bone Loss and Tumor

Growth in an Orthotopic Mouse Model of Bone Invasive Oral Squamous Cell carcinoma

...... 105

Abstract ...... 105

Introduction ...... 106

Materials and Methods...... 108 xi

Results ...... 116

Discussion ...... 123

Bibliography ...... 141

xii

List of Tables

Table 1.1. Comparison of breed, hair length, gender and neuter status between control

cats and cats with OSCC...... 22

Table 2.1: Behavior of OSCC xenografts in nude mice...... 61

xiii

List of Figures

Figure 1.1: Feline OSCC population demographics...... 24

Figure 1.2: Histopathology of bone invasive OSCC ...... 26

Figure 1.3: PTHrP immunohistochemistry positive and negative controls ...... 28

Figure 1.4: PTHrP expression in feline OSCC ...... 29

Figure 1.5: PTHrP expression in normal feline oral epithelium...... 31

Figure 1.6: PTHrP-positive OSCC cells and nuclei were more frequent in biopsies with

evidence of osteolysis ...... 32

Figure 2.1: SCCF2 and SCCF2 microscopic morphology ...... 62

Figure 2.2: SCCF2 and SCCF3 ultrastructure ...... 64

Figure 2.3: SCCF2 and SCCF3 karyotype...... 65

Figure 2.4: Feline and human OSCC cells induced in vitro bone resorption and

osteoclastogenesis...... 67

Figure 2.5: In vivo bioluminescent imaging revealed progressive SCCF2Luc xenograft

growth and pulmonary metastasis...... 68

Figure 2.6: SCCF2Luc and UMSCC12Luc were associated with the greatest degree of bone resorption in vivo...... 70

xiv

Figure 2.7: Bone-invasive OSCC cells expressed more PTHrP and less OPG compared to

minimally bone-invasive OSCC cells...... 72

Figure 2.8: Bone invasive OSCC cells induce RANKL expression and suppress OPG expression in murine preosteoblasts ...... 74

Figure 2.9: Bone conditioned medium stimulated proliferation of OSCC cells and

increased OSCC-expression of PTHrP ...... 76

Figure 3.1: ZOL treatment reduced SCCF2Luc xenograft growth...... 96

Figure 3.2: ZOL inhibited SCCF2Luc xenograft-associated osteolysis ...... 97

Figure 3.3: ZOL prevented loss of bone surface area and bone volume in SCC2Luc

xenografts...... 99

Figure 3.4: ZOL reduced SCCF2Luc-associated osteolysis and induced vacuolar

degeneration of osteoclasts but did not reduce invasion...... 100

Figure 3.5: ZOL-mediated reduction in bone loss was characterized by reduced loss of

pre-existing bone and reduced number of bone-resorbing osteoclasts at the tumor-bone

interface...... 102

Figure 3.6: PTHrP was increased in mice bearing SCCF2Luc xenografts but was not

associated with hypercalcemia...... 104

Figure 4.1: OSCC cells express COX-1 and COX-2, and SCCF2 growth inhibition did

not occur at physiologically relevant doses of COX inhibitors ...... 130

Figure 4.2 Meloxicam reduced tumor growth ...... 132

Figure 4.3: Zoledronic acid reduced bone loss and was associated with osteoclast

vacuolar degeneration ...... 133

Figure 4.4: Zoledronic acid reduced loss of bone volume...... 135

Figure 4.5: Zoledronic acid reduced loss of bone area ...... 137

Figure 4.6: Zoledronic acid reduced osteoclast number and increased osteoclast size and number of nuclei ...... 139

xvi

Chapter 1 : Bone Invasive Oral Squamous Cell Carcinoma in Cats: Pathology and expression of Parathyroid Hormone-Related Protein

Abstract

Feline oral squamous cell carcinoma (OSCC) is the most common oral tumor in cats.

There is no effective treatment and the average duration of survival after diagnosis is only

2 months. Feline OSCC is frequently associated with osteolysis; however, the

mechanisms responsible are unknown. The objective of this study was to characterize the

epidemiology and pathology of bone-invasive OSCC in cats and to determine the

expression of select bone resorption agonists. Four-hundred and fifty-one cases of feline

OSCC were evaluated. There was no gender or breed predisposition; however, there were

more intact cats in the OSCC group compared to controls. Gingiva was the most common

site, followed by the sublingual region and tongue. Cats with lingual OSCC were younger

(mean of 11.9 years) compared to cats with gingival OSCC (mean of 13.6 years). In

addition to osteolysis, there was also periosteal new bone formation, osseous metaplasia

of tumor stroma, and direct apposition of OSCC to fragments of bone suggestive of bone-

binding behavior.

Eighty-two cases were selected for immunohistochemical detection of parathyroid hormone-related protein (PTHrP). Specimens with osteolysis had increased PTHrP expression and nuclear localization compared to OSCC without osteolysis. Thirty-eight

biopsies of OSCC with osteolysis were evaluated for expression of tumor necrosis factor-

α (TNFα). Only 4 biopsies had TNFα expression in a small proportion of tumor cells.

Increased tumor-expression of PTHrP, and increased localization of PTHrP to the

nucleus, were associated with osteolysis and may play an important role in bone resorption and tumor invasion in cats with OSCC.

Introduction

Oral squamous cell carcinoma (OSCC) is the most commonly diagnosed tumor of the oral cavity in cats (1, 2). It is extremely aggressive and characterized by destruction of oral tissues and bone invasion, and is often associated with tooth loss, ulceration and secondary bacterial infections (3). Although early surgical removal may be curative, most cats have invasive disease at the time of diagnosis and either requires extensive resections that are associated with morbidity (4, 5) or the tumors are so extensive that they are no longer surgical candidates (3). There is still no effective treatment for feline OSCC and average duration of survival after diagnosis rarely exceeds 2 months (6).

The purpose of this study was to describe the epidemiology and pathology of cats with

OSCC with emphasis on characterizing OSCC-associated bone resorption, and to

determine the expression of various candidate regulators of osteoclastic activity. We

evaluated 451 surgical biopsies in which we characterized the patient signalment and

history, clinical signs, size of lesion, histopathology and type of bone interaction.

Factors that can play an important role in bone resorption include parathyroid hormone related protein (PTHrP), interleukin 1α (IL-1α), tumor necrosis factors alpha (TNFα), receptor activator of nuclear factor κB ligand (RANKL), osteoprotegerin (OPG) and

transforming growth factor-beta 1 (TGF-β1). RANKL is a direct stimulator of osteoclast

activation normally expressed by osteoblasts, but is also expressed by a number of bone

invasive tumors including feline (7) and human OSCC (8).

PTHrP, IL-1α and TNFα stimulate osteoblast-expression of RANKL, and are expressed

in human OSCC (9). PTHrP expression has also been demonstrated in a feline OSCC

cell line (10). OPG is a soluble receptor for RANKL and inhibits activation of

osteoclasts. OPG is expressed at low levels in human OSCC (8). TGF-β1 is a growth

factor that has both stimulatory and inhibitory effects on osteoclastic bone resorption, and

has been shown to be expressed by human OSCC epithelial cells (11) and stroma (12).

Materials and Methods

Biopsy material and patient signalment

Four hundred and fifty-one (451) hematoxylin and eosin (HE)-stained tissue sections

from surgical biopsies of feline OSCC were acquired from Veterinary Diagnostics, Ltd.

(Columbus, Ohio), IDEXX laboratories (Worthington, Ohio) and Clinilab, Inc.

(Valpariaso, Indiana). Biopsies were submitted between 1998 and 2006, from 282

veterinary practices and practitioners from 11 states. Three hundred and nine (309) cases

(68%) were from Ohio. Tumors were assigned to the following anatomic locations based

on information provided: tongue, sublingual (includes ventral tongue, frenulum, mouth

floor), gingiva (includes tumors associated with teeth, mandible, maxilla and jaw), hard

palate, soft palate, tonsils, pharynx, larynx, buccal mucosa, lip and ‘not reported’.

Clinical history including symptoms, size of tumor, and evidence of bone resorption were

summarized.

The breed, hair length (long, short or undetermined based on reported breed), gender

distribution and reproductive status (intact or neutered) of 389 cats (≥ 8 years of age)

with OSCC were compared to 391 cats (≥ 8 years of age) that had tissue submitted to

IDEXX laboratories in 2005, regardless of diagnosis. Limiting the cases to this age group

was performed to minimize the effect of age on neuter status, particularly in the control

group which included a greater proportion of immature cats compared to the OSCC

group.

The HE-stained tissue sections were evaluated microscopically to confirm the presence

of OSCC, and to determine the presence of bone in the biopsy material. Specimens were

evaluated for evidence of new bone formation (high osteocyte density and woven

collagen pattern), osteoclastic bone resorption (eroded bone surfaces with or without

osteoclasts), and evidence of OSCC adhesion to bone (OSCC in direct contact with

bone).

Immunohistochemistry for the detection of bone resorption mediators in feline tissues

Immunohistochemical protocols were designed for the detection of several candidate

factors known to regulate osteoclastic bone resorption including PTHrP, TNFα, IL-1α,

RANKL, OPG and TGF-β1. The primary antibodies were polyclonal rabbit anti-human

PTHrP (34-53, AB-2, Calbiochem, La Jolla, CA), polyclonal goat anti-human TNFα

(R&D Systems, Minneapolis, MN), polyclonal goat anti-human IL-1α (R&D Systems), monoclonal mouse anti-human RANKL (clone 70525, R&D Systems), polyclonal goat anti-human OPG (R&D Systems) and polyclonal rabbit anti-human TGF-β1 (V) (Santa

Cruz Biotechnology, Inc., Santa Cruz, CA). The secondary antibodies were biotin-labeled goat anti-rabbit IgG (Zymed, San Francisco, CA), biotin-labeled horse anti-goat IgG

(Vector Laboratories, Burlingame, CA) and biotin-labeled horse anti-mouse IgG (Vector

Laboratories).

Feline skin served as positive control tissue for PTHrP and TGF-β1, mouse skin and feline ovary for TGF-β1, feline mammary carcinoma for TNFα, feline bone and lymph node for RANKL and OPG, and feline lymph node for IL-1α. A range of primary antibody dilutions was evaluated, ranging from 0.2 µg/mL to 5 µg/mL for all antibodies except PTHrP (0.2 µg/mL to 2 µg/mL), TGF-β1 (0.05 µg/mL to 4 µg/mL) and RANKL

(1 µg/mL to 25 µg/mL). Blocking peptides for PTHrP (Calbiochem), TNFα (R&D

Systems) and TGF-β1 (Santa Cruz Biotechnology) were pre-incubated with the primary antibodies (10:1 peptide to antibody) to determine the specificity of the primary antibody.

Blocking peptides for RANKL, OPG and IL-1α were not used because these antibodies did not work on feline control tissues and further optimization was not pursued.

Two forms of antigen retrieval were evaluated; target retrieval solution (Dako,

Carpinteria, CA) at 95°C for 20 minutes followed by room temperature cooling for 20

minutes, and digestion with 0.02 units/ml recombinant α2,3-neuraminidase (Calbiochem)

for 60 minutes at room temperature. Specificity of secondary antibodies was evaluated by

omitting the primary antibody from the protocol. Protocols that yielded signals on

positive control tissue, and those that demonstrated specificity of the primary and

secondary antibodies, were selected for evaluating feline OSCC specimens (PTHrP and

TNFα). Positive and negative controls were run with all OSCC slides.

Pancytokeratin (AE1/AE3) immunohistochemistry was performed on select tissues

with evidence of OSCC adhesion to bone to confirm the direct apposition of tumor cells

with the bone matrix, and was performed in the Histology and Immunohistochemistry

Laboratory in the Department of Veterinary Biosciences.

Immunohistochemistry for PTHrP and TNFα in feline OSCC biopsies

Forty-four (44) feline OSCC tumors with microscopic evidence of osteolysis and 38

feline OSCC tumors without microscopic evidence or reported clinical history of

osteolysis were selected. Tissues were routinely processed, sectioned and mounted on

glass slides. Sections were dewaxed and rehydrated through stepwise washes of xylene

and descending concentrations of ethanol followed by water.

For PTHrP immunohistochemistry, antigen retrieval was achieved by neuraminidase digestion as described above. Endogenous peroxidases were quenched with a ready-to- use peroxidase blocking reagent (Dako) for 30 minutes at room temperature. Non-

specific binding was blocked with serum-free, ready-to-use protein block (Dako) for 20

minutes at room temperature. Protein block was replaced by PTHrP primary antibody diluted to 1µg/ml in serum-free protein block. Tissues were incubated with primary antibody in a humidified chamber at 4 °C for 16 hours. After washing in phosphate- buffered saline (PBS), tissue sections were incubated with biotin-labeled goat anti-rabbit

IgG. The avidin–biotin–peroxidase method (Vector Laboratories) was then applied according to the manufacturer's instructions. Color development was achieved with

DAKO diaminobenzidine (DAB) chromogen system. Feline skin served as a positive control. For negative control purposes, the primary antibody was preadsorbed (4° C for

16 hrs) with PTHrP blocking peptide before being applied to the tissues.

TNFα immunohistochemistry was performed as for PTHrP except antigen retrieval was achieved with heated target retrieval solution as described above, the primary antibody

(or blocked antibody for negative controls) was applied for 40 hours in a 4°C humidified chamber, the secondary antibody was biotin-labeled horse anti-goat IgG, and the positive

control tissue was feline mammary carcinoma. Tissue sections were not counterstained.

Slides were imaged using Nomarski bright field microscopy and black and white

photography.

To determine if PTHrP was expressed differentially in the oral cavity by anatomic site,

immunohistochemistry was performed on normal feline gingiva and lingual tissue that

was collected from 12 cats during postmortem examination. All tissues were fixed in

10% buffered formalin, paraffin embedded, and sectioned at 5 µm prior to

immunohistochemical evaluation.

Grading PTHrP immunohistochemistry

Specimens that were evaluated using immunohistochemistry for the detection of PTHrP

were scored based on a subjective assessment of overall signal intensity in OSCC cells

(absent, light, moderate or heavy), proportion of PTHrP-positive OSCC cells (regardless of cellular localization; 0-24%, 25-49%, 50-74% and 75-100%), and the proportion of

OSCC cells with PTHrP positive nuclei (0-24%, 25-49%, 50-74% and 75-100%). The

distribution of PTHrP intensity, cellular staining and nuclear staining scores were

compared between specimens with and without bone resorption.

Only a small number of tumors with osteolysis demonstrated TNFα positivity (4 of 38),

and in only a small proportion of cells (less than 5%). Comparison to tumors without osteolysis was not pursued because evaluation of this bone resorption agonist in tumors with osteolysis was predominantly negative.

Statistical analysis

Age and tumor site were displayed as histograms. Average age by tumor site was

displayed as means ± standard error. Age of cats with lingual OSCC was compared to age

of cats with gingival, sublingual or ‘other’ sites of OSCC using Student’s t-test when data

was normally distributed (determined with the Shapiro Wilk test) or Wilcoxon rank sum

(Mann-Whitney) test when not normally distributed. The P value of 0.05 divided by the

number of comparisons (3) was accepted as significant (P value of 0.017). Categorical

data (breed, hair length, gender, neuter status and PTHrP scores) were analyzed using

Fishers exact test. All analyses were performed with STATA intercooled 10 software

(Cary, NC).

Results

Epidemiology of feline OSCC

Age was reported for 416 cats. The average age of cats with OSCC was age 13 years

(range 1.5 to 22) but was not normally distributed (Shapiro Wilk, P=0.00017) (figure

1.1A). Gingiva was the most common location (51%, figure 1.1B). Forty percent (40%) of gingival tumors were specified as mandibular and 37% as maxillary. The locations of the remaining 23% of gingival tumors were not provided. Twenty-three percent (23%) of

feline OSCC biopsies were sublingual and 11% were lingual. Tumors from buccal

mucosa, hard palate, soft palate, larynx, pharynx, tonsil, and lip each represented ≤ 2% of

the total. Anatomic location was not provided for the remaining 5%.

The average age of cats with lingual OSCC was 11.9 years old, younger than cats with

gingival OSCC (13.6 years; P=0.0004 2-tailed t-test) and cats with OSCC from other

sites (13.5 years; combined hard and soft palate, buccal, tonsil, larynx, pharynx, lip and

unspecified location; P=0.0094 2-tailed t-test). Although the average age of cats with

lingual OSCC was younger than cats with sublingual OSCC (13.0 years), the difference

9 was not statistically significant when adjusted for multiple comparisons (Wilcoxon rank sum P= 0.028, figure 1.1C).

There was no significant difference in breed, hair length or gender distribution between cats in the control group or in the OSCC group (table 1.1); however, there were more intact cats in the OSCC group (21%) compared to the control group (15%) (P=0.015, fisher’s exact test). Ulceration was the most common symptom reported (36%), followed by loose teeth (17%) and facial swelling (13%) (n=230). Average duration of symptoms was 5.2 weeks (range 0.1 to 36 weeks, n=72). Average tumor size was 2.4 cm diameter

(range 1 to 8 cm, n=32). Clinical evidence of bone involvement (radiography, tooth loss, clinical observation) was reported in 95 cases.

Pathology and characterization of bone involvement

Thirty-seven percent (37%) of biopsies contained bone. Of bone-containing specimens,

87% had evidence of bone resorption, 46% had evidence of new bone formation, and

49% demonstrated direct contact of OSCC cells with eroded bone surfaces suggestive of

OSCC adhesion to bone. Identification of new bone was based on high osteocyte density and woven collagen pattern, in contrast to the low osteocyte density and lamellar collagen pattern characteristic of mature cortical bone. In some cases, new bone was organized in a radiating pattern of trabeculae extending from the surface of mature bone consistent with reactive, periosteal new bone formation (Figures 1.2A and B). In other specimens, immature bone existed as irregular spicules within dense bands of fibroplasia between

nests and islands of OSCC, interpreted as tumor-induced metaplastic bone formation

(Figures 1.2C and D).

Bone resorption was characterized by eroded bone with irregular, scalloped surfaces

(Figure 1.2E), and occasionally with large multinucleated osteoclasts in resorption pits on the bone surface (Figure 1.2F). In almost all specimens with new bone formation, osteoclastic resorption of the immature bone was observed. OSCC cells occasionally surrounded fragments of eroded bone (Figure 1.2G) and were in direct contact with the bone surface with no intervening stromal cells (Figure 1.2H).

Immunohistochemistry optimization

RANKL, OPG and IL-1α were not detectable in positive control tissues and evaluation of feline OSCC specimens with these antibodies was not performed. The TGF-β1

antibody produced strong signal in feline skin, feline ovary and mouse ovary, but

specificity could not be demonstrated because preadsorption of the antibody with TGF-β1

blocking peptide failed to block the signal, therefore TGF-β1 was not evaluated in the feline OSCC specimens. The TNFα antibody produced strong cytoplasmic signal in feline mammary carcinoma (data not shown), and the PTHrP antibody produced moderately intense cytoplasmic signal in feline epidermal cells (Figure 1.3). Both TNFα

and PTHrP positive reactions were successfully blocked with their respective blocking

peptides.

Expression of TNFα in feline OSCC specimens with bone resorption

The optimized TNFα protocol was used to evaluate 32 cases of feline OSCC with evidence of osteolysis. Twenty-eight (28) cases (88%) of feline OSCC with osteolysis were negative for TNFα expression, the remaining 4 cases demonstrated positivity for

TNFα in only 1 to 5 % of tumor cells (data not shown). Evaluation of additional biopsies with bone resorption, and biopsies without bone resorption, were not performed.

Expression of PTHrP in feline OSCC specimens with and without bone resorption

Overall, PTHrP was expressed in most OSCC cells in a diffuse, finely granular cytoplasmic pattern; however, a subset of PTHrP-expressing cells also demonstrated nuclear localization of PTHrP (Figure 1.3C). Most biopsies without osteolysis (Figures

1.4A and B) had light to moderate cytoplasmic staining for PTHrP. Most biopsies with osteolysis demonstrated moderate to heavy PTHrP staining that frequently localized to the nucleus in addition to the cytoplasm (Figures 1.4C and D). Fibroblasts in tumor stroma often demonstrated light to moderate staining; and bone lining cells, osteoblasts and osteocytes in bone-containing biopsies were also PTHrP-positive. PTHrP staining in postmortem samples of feline gingiva and tongue (Figures 1.5A and B) was light to moderate intensity in epithelial cells.

The distribution of and intensity of PTHrP localization was scored and compared between tumors with and without evidence of osteolysis. Heavy PTHrP staining intensity was more common in feline OSCC with osteolysis compared to OSCC without osteolysis

(P=0.004, Fisher’s exact test, data not shown). PTHrP-positive tumor cells were more

common in biopsies with osteolysis compared to OSCC without osteolysis (Fisher’s

exact test, P=0.001, Figure 1.6). Eighty percent of OSCC biopsies with osteolysis had

75% -100% PTHrP-positive cells, compared to only 39% of the OSCC specimens without evidence of osteolysis. Positive nuclei were more common in biopsies of feline

OSCC with osteolysis, compared to OSCC biopsies without osteolysis (Fisher’s exact

test, P<0.0001, Figure 22). Forty-one percent (41%) of OSCC specimens with bone

resorption had nuclear localization of PTHrP in 75-100% of OSCC cells, compared to

only 8% of specimens without bone resorption. There was no statistical difference in

epithelial cell expression of PTHrP between normal feline gingiva and tongue

(postmortem samples) in terms of intensity, percent positive epithelial cells, and percent

positive nuclei (Fisher’s exact test, data not shown).

Discussion

Epidemiology

The population of cats with OSCC in this study was similar in age (mean age 13 years,

range 1.5 to 22 years) and gender distribution (53.7% female and 46.3% male) to a cohort

of 227 cats with OSCC reported in 1989 by Stebbins et al. (mean age 12.5 years, range 3

to 21, 55.5% female and 44.5% male) (1, 2). Although females appear to be slightly overrepresented in these studies, there was no statistical difference between cats with

OSCC and the control cats in this study. OSCC most often affects older humans; however, there is a clear gender association with a male to female ratio of 2: 0.1 (13).

Most human cases occur in men over 45 years-old who use tobacco and/or consume

alcohol (14). While cats do not share the same risk factors as most human patients with

OSCC (although some cats are exposed to environmental tobacco smoke in their home)

(15), they may have similarities to a small subgroup of human patients with OSCC who

are unique in that they do not smoke or consume alcohol, they tend to be female, have

tumors of the oral cavity (as opposed to the oropharynx, a common site among people

who smoke and consume alcohol), and are older than patients who smoke and consume

alcohol. Risk factors for the development of OSCC in this subgroup of human patients

are unknown (16).

There was no significant difference in breed, gender or hair length between control cats

and cats with OSCC. These findings are similar to what Bertone et al. reported in 2003,

when they compared patient signalment and history between 36 cats with OSCC and 112

cats with renal disease (15). Similarly, no breed predisposition was detected in the study

performed by Stebbins et al. Interestingly, our study demonstrated that there were more

intact cats (regardless of gender) in the OSCC group than in the control group (21%

compared to 15%). The reason why intact cats are overrepresented in the OSCC group is

unknown, but may involve the influence of hormones such as estrogen or testosterone on

the development and progression of OSCC, or the role of other factors associated with

neuter status such as living indoors or outdoors, type of diet, and exposure to infectious

agents and environmental carcinogens.

Cats with lingual OSCC were significantly younger than cats with OSCC of the gingiva

and of other sites excluding the sublingual area; however the reason for this is unknown.

Lingual OSCC may be more rapidly progressive than OSCC of other sites in the oral cavity, or the clinical signs are more rapidly detectable in lingual OSCC; however, this would mean that cats with gingival OSCC were living an average of 1.5 years longer than cats with lingual OSCC; which is highly unlikely given the aggressive and rapidly progressive nature of this tumor. It is more likely that lingual OSCC actually developed at a younger age, and is suggestive that the pathogenesis of lingual OSCC differs from that of other locations.

We hypothesize that lingual epithelium differs in susceptibility to the development of

OSCC compared to other sites in the oral cavity. In humans with OSCC, variation in the expression of polymorphic xenobiotic metabolizing enzymes in oral mucosa, such as those of the cytochrome P450 family, has been implicated as a reason for increased individual susceptibility to oral cancer as a result of exposure to carcinogens in tobacco smoke (13). Site specific expression of xenobiotic metabolizing enzymes within the oral cavity has not been reported in cats or humans; however, a study by Robinson et al. demonstrated that carboxylesterase activity in the oral cavity of mice varied depending on location, with an apparent correlation between lesion distribution following exposure to a high concentrations of vinyl acetate (a carcinogen) and high carboxylesterase activity in different areas of the oral cavity (17). Bertone et al. demonstrated a 2-fold increased risk for the development of OSCC in cats exposed to environmental tobacco smoke (15).

Perhaps the expression of polymorphic and site-specific xenobiotic metabolizing

enzymes in the lingual mucosa makes some cats more susceptible to carcinogenesis and

lead to development OSCC of the tongue at an earlier age. Cats may be exposed to

carcinogens in environmental tobacco smoke or other sources as they are deposited on

the hair coat and introduced to the tongue through grooming (18).

Pathology

The most common location of OSCC was gingiva (including tumors associated with

teeth, mandible, maxilla and jaw), followed by the sublingual region (including ventral

tongue, frenulum, mouth floor) and tongue. These findings differ from reports stating that the sublingual area (6) and tongue (1) are the most common sites. This discrepancy may

be due to differences in how tumor sites were classified; for example, Dorn et al. did not

have a site designation for sublingual tumors but did separate gingival tumors from those

of the dental alveolus. We did not designate dental alveolus as a site separate from

gingiva, and we did not separate maxillary tumors from mandibular tumors.

Bone invasion was most commonly found in gingival OSCC, as expected given the

close proximity of the gingiva to underlying bone. Bone invasion was characterized by

osteoclastic bone resorption, new bone formation, and desmoplasia. A proportion of

biopsies that contained bone demonstrated islands of OSCC surrounding and in direct

apposition with the surface of eroded bone, suggesting that OSCC cells may be capable

of adhering to bone. To our knowledge, OSCC adherence to bone has not been reported;

however, human breast and prostate cancer cells have been shown to adhere to bone

matrix in vitro (19). Carcinoma cells adhering to bone matrix was mediated by the

interaction of integrin receptors on the tumor cells with collagen 1 (20) and bone

sialoprotein (19). TGF-β1, a growth factor stored in bone matrix and released in an active

form during bone resorption, has been shown to stimulate prostate carcinoma cell

adhesion (21).

Although certain carcinoma cells appear to adhere to bone, it is unlikely that

carcinomas directly resorb bone (22, 23). The significance of OSCC-bone binding is not

known; however, exposed bone surfaces may serve as a scaffold upon which OSCC can

migrate, much like perineural invasion. OSCC adhesion to bone may be particularly

significant if it promotes tumor cell migration and proliferation as has been demonstrated

with human breast (19) and prostate cancer cells (20), and warrants further investigation

in order to confirm the adhesion of OSCC cells to bone matrix.

Immature bone in feline OSCC occurred as either interwoven, radiating trabeculae of periosteal new bone, or as irregular spicules of metaplastic bone arising within tracts of

tumor-associated desmoplasia adjacent to islands and cords of tumor cells. The induction

of periosteal new bone formation is an expected event, resulting from activation of the

periosteum during expansile and invasive growth of the tumor. In contrast, the formation

of immature bone within areas of tumor-induced desmoplasia is likely the result of OSCC

cells inducing the expression of bone formation agonists, such as bone morphogenetic

proteins (BMPs) and isoforms of TGF-β1 (24) in the adjacent stroma in conjunction with

metaplastic transition of stromal cells from a fibroblastic to osteoblastic phenotype.

Expression of TNFα and PTHrP

A role for tumor-expressed TNFα in the pathogenesis of OSCC-induced bone

resorption was not supported in this study, as only a few biopsies with evidence of bone resorption had detectable TNFα, (4 of 32 biopsies) and in only a small number of cells.

These findings are in contrast to the work of Shibahara et al. who demonstrated that

TNFα expression was increased in cases of bone invasive human OSCC (9); however, a

more recent study by Van Cann et al. revealed that TNFα was not differentially expressed

in human OSCC with and without invasion into the mandibular medullary canal (25).

PTHrP was detected in feline OSCC with and without evidence of bone resorption

using an antibody directed towards midregion (34-53) PTHrP. AntiPTHrP antibodies

directed towards this epitope have been used to detect PTHrP in a feline OSCC cell line

(10) and in various canine tissues (epidermis, myoepithelial cells of dermal apocrine

glands, mammary gland, anal sac epithelium) (26). Staining intensity of PTHrP

expression in OSCC cells varied from absent to heavy, and was also detected in stromal

cells surrounding islands of OSCC, in osteoblasts lining trabeculae of new bone,

osteocytes, and in overlying oral epithelium. This staining pattern is not unexpected, as

PTHrP is known to be ubiquitous in expression (27), and PTHrP has been previously

demonstrated to be expressed by canine dermal fibroblasts (28), canine prostate stromal

cells (28); mouse (29) and canine osteoblasts(26), and in murine preosteoblasts and

osteocytes (30).

Immunohistochemistry has been previously used to detect PTHrP in feline neoplasms

including lung carcinomas, a thyroid carcinoma and lymphoma (31). Additionally, an

immunoradiomimetic assay (IRMA) for human PTHrP was used to detect elevated

plasma levels of PTHrP in cats with humoral hypercalcemia of malignancy (31). It is

unknown if the increased PTHrP expression observed in feline OSCC with osteolysis is

the cause or result of bone resorption. It is likely that both relationships are occurring.

PTHrP is known to stimulate osteoclastic bone resorption by increasing the expression of RANKL in osteoblasts, which in turn leads to the maturation and activation of osteoclasts resulting in bone resorption. Bone resorption leads to the liberation of

activated growth factors from the bone matrix, such as TGF-β1, in addition to the release

of calcium. Both calcium and TGF-β1 can stimulate increased production of PTHrP (32).

In fact, TGF-β1 has been shown to increase PTHrP expression in a feline OSCC cell line,

SCCF1 (10).

The prognostic significance of increased PTHrP is controversial. On one hand, PTHrP

expression in human colorectal cancer and gastric cancer correlated with depth of

invasion and metastasis (33). In contrast, PTHrP expression in human non-small cell lung

carcinomas was associated with a survival advantage among female patients, and

increased PTHrP expression has been associated with longer survival in patients with

ductal mammary carcinoma (33). The effect of PTHrP expression on prognosis in

veterinary patients has not been studied.

PTHrP is capable of stimulating osteoclastic bone resorption by increasing osteoblast

expression of RANKL and subsequent recruitment and activation of osteoclasts, and

stimulates bone formation by promoting the maturation and survival of matrix-producing

osteoblasts (34). PTHrP expression in murine osteoblasts has been shown to be necessary

for bone formation (29), and expression of PTHrP in human OSCC cells has been shown to induce the in vitro formation of osteoclasts by stimulating RANKL expression in murine osteoblasts (35). It is likely that tumor-derived PTHrP had a stimulatory role in both formation of new bone and osteoclastic bone resorption observed in the feline

OSCC biopsies.

OSCC biopsies with evidence of bone resorption demonstrated an increased percentage of cells with nuclear localization of PTHrP. The significance of nuclear localization of

PTHrP in feline OSCC with osteolysis is unknown, but it may support tumor cell

proliferation, survival, adhesion and invasion as has been shown for human breast cancer

cell lines (36). PTHrP is known to localize to the nucleus and nucleolus during the G1

phase of the cell cycle, and has been shown to inhibit apoptosis of chondrocytes and

tumor cells (32).

Feline OSCC was most commonly located in the gingiva, and frequently demonstrated

bone involvement which was characterized by osteoclastic bone resorption, formation of

new bone (periosteal new bone and metaplastic bone), and direct apposition of OSCC

cells with bone suggesting bone adhesion. OSCC-expression of TNFα was not important

to the pathogenesis of bone resorption since only a small proportion of biopsies

demonstrated TNFα expression in OSCC cells. In contrast, PTHrP expression was

commonly expressed in OSCC biopsies with and without evidence of osteolysis. The

proportion of PTHrP-positive cells and nuclei, in addition to overall staining intensity,

was increased in tumors with evidence of osteolysis. These results support a relationship

20 between PTHrP expression in feline OSCC cells and osteoclastic bone resorption; however, further investigation is necessary to determine the significance of PTHrP expression and nuclear localization in feline OSCC.

Control OSCC % % Breedb DSH 68 69 DLH 17 19 DMH 3 4 Siamese 3 1 Himalayan 2 3 Persian 2 3 Mixed 2 1 Otherc 3 0 Hair lengthd Short 73 70 Long 25 29 Unspecified 2 1 Gender Female 7 12 Spayed Female 47 42 Male 8 10 Castrated Male 38 36 Neuter Statuse Intact 15 21* Neutered 85 78 Table 1.1. Comparison of breed, hair length, gender and neuter status between control cats and cats with OSCCa aControl group consisted of 391 cats, OSCC group consisted of 389 cats. Only cats at least 8 years-old were included. bBreed abbreviations include DSH (domestic short hair), DMH (domestic medium hair) and DLH (domestic long hair)

cOther breeds that represented less than 1% of both groups included Main Coon,

Abyssinian, Rag doll, Balinese, Egyptian Mau, Munchkin, Bengal, Cornish Rex, Manx,

Angora and Burmese.

dHair length was designated based on breed. ‘Short Hair’ = DSH, Siamese, Abyssinian,

Balinese, Egyptian Mau, Munchkin, Bengal, Cornish rex, Manx and Burmese. ‘Long

Hair’ = DMH, DLH, Himalayan, Persian, Main Coon, Rag doll and Angora.

‘Unspecified’ were mixed breed.

eNeuter status: There were more intact cats in the OSCC group compared to the control

group (*P=0.01).

Figure 1.1: Feline OSCC population demographics

A. Age distribution of cats with OSCC. The average age of cats diagnosed with OSCC was 13 years (n=541), with a range of 1.5 to 22 years. Age was not normally distributed

(skewed towards older cats). B. Anatomic location of feline OSCC. The most common location of OSCC was gingiva (including tumors associated with teeth, mandible, maxilla and jaw), followed by sublingual (includes ventral tongue, frenulum, mouth floor) and tongue. C. Cats with lingual OSCC were younger than cats with OSCC of the gingiva and other sites. The average age of cats with lingual OSCC was 11.9 years old, younger than cats with gingival OSCC (13.6 years; *P=0.0004) and cats with OSCC from other sites

(13.5 years; combined hard and soft palate, buccal, tonsil, larynx, pharynx, lip and unspecified; *P=0.0094). Although the average age of cats with lingual OSCC was younger than cats with sublingual SCC (13.0 years), the difference was not statistically when adjusted for multiple comparisons (Wilcoxon rank sum P= 0.028).

A B

C D

E F

G H

Figure 1.2: Histopathology of bone invasive OSCC

A. Gingival OSCC, low magnification revealing small islands of OSCC (arrows) with radiating trabeculae of periosteal new bone (PNB) (HE). B. Higher magnification of A

revealing OSCC (black arrows) surrounding small fragments of bone adjacent to lamellar

cortical bone (CB) and periosteal new bone (PNB)(HE). C. Gingival OSCC, Low magnification revealing islands of OSCC (arrows) deep to the gingival surface (G) with desmoplasia and irregular trabeculae of immature metaplastic bone (MB) (HE). D.

Higher magnification of C demonstrating metaplastic bone formation (MB) within the tumor stroma adjacent to OSCC (black arrow) (HE). E. Gingival OSCC, islands of OSCC

(arrows) infiltrating between tooth root dentin (D) and irregularly surfaced (eroded) alveolar bone (AB) (HE). F. OSCC (black arrows) adjacent to immature bone (B) with numerous multinucleated osteoclasts in resorption pits (white arrows) (HE). G. OSCC cells (arrows) surround small fragments of previously eroded bone (B) (HE). H. Gingival

OSCC, cytokeratin (AE1/AE3) -positive OSCC cells (arrows) surrounding previously eroded bone (B) (DAB chromogen and hematoxylin counterstain).

A B

C D

Figure 1.3: PTHrP immunohistochemistry positive and negative controls

A, PTHrP positive control (cat skin). PTHrP was detected in the cytoplasm of epidermal

keratinocytes (white arrow) and in dermal fibroblasts (arrowheads), but not in epithelial

cell nuclei (black arrows)(DAB, no counterstain). B, PTHrP negative control.

Preincubation of the PTHrP primary antibody with PTHrP blocking peptide eliminated staining in the positive control tissue (white arrow = epidermis, black arrows = nuclei)(DAB, no counterstain). C, Gingival OSCC. Moderate PTHrP staining was detected in the cytoplasm (white arrows) and nucleus (black arrows) of OSCC cells.

Fibrous stroma (desmoplasia) indicated by white asterisk. D, Preincubation of the PTHrP primary antibody with PTHrP blocking peptide eliminated staining (white arrows =

OSCC cells, black arrows = nuclei, white asterisk = desmoplasia)(DAB, no counterstain).

A B

C D

Figure 1.4: PTHrP expression in feline OSCC

A, Lingual OSCC without osteolysis. Light PTHrP staining (cytoplasmic) in a low proportion of OSCC cells (white arrows) without nuclear localization (black arrows)

(DAB, no counterstain). B, Lingual OSCC without osteolysis. Moderate cytoplasmic

PTHrP staining (white arrows) in a high proportion of OSCC cells without nuclear localization (black arrows)(DAB, no counterstain). C, Gingival OSCC with osteolysis.

Moderate PTHrP staining of a high proportion of OSCC cells with nuclear localization

(black arrows) in OSCC cells surrounding a fragment of previously eroded bone (DAB,

29 no counterstain). D. Gingival OSCC with osteolysis. Heavy PTHrP staining of a high proportion of OSCC cells demonstrating nuclear localization (black arrows, bone is out of field of view) (DAB, no counterstain).

Figure 1.5: PTHrP expression in normal feline oral epithelium

A, Normal gingiva. Moderate PTHrP staining of gingival suprabasalar epithelial cells without nuclear localization (DAB, no counterstain). B, Normal tongue. Moderate PTHrP staining of lingual suprabasalar epithelial cells without nuclear localization (DAB, no counterstain).

Figure 1.6: PTHrP-positive OSCC cells and nuclei were more frequent in biopsies with evidence of osteolysis

Each bar represents 100% of biopsies in each group (with and without evidence of osteolysis). Each bar is subdivided to demonstrate the proportion of biopsies with 0-24%,

25-49%, 50-74% and 75-100% PTHrP-positive cells or PTHrP-positive nuclei, as indicated. PTHrP positive tumor cells were more common in biopsies with osteolysis compared to OSCC without osteolysis (*P=0.001). PTHrP positive tumor nuclei were more common in biopsies with osteolysis compared to OSCC without osteolysis

(*P<0.0001).

Chapter 2 : Characterization of Bone Resorption in Novel In Vitro and In Vivo Models of Oral Squamous Cell Carcinoma

Abstract

Head and neck cancer is the 8th most common human cancer in the world, the majority of which are oral squamous cell carcinoma (OSCC). Bone invasion occurs frequently and is associated with poor prognosis and reduced survival. Advances in the understanding and treatment of bone invasive OSCC has been complicated by the limited availability of model systems that recapitulate the osteolytic behavior of OSCC. Similar to humans,

OSCC is the most commonly diagnosed oral tumor of cats. OSCC in both species is associated with osteoclastic bone resorption; however, the mechanisms responsible are not fully understood. The objective of this study was to investigate the significance of tumor-derived parathyroid hormone-related protein (PTHrP) in novel models of bone invasive OSCC utilizing cell lines derived from spontaneous human and feline OSCC tumors.

In vitro expression of PTHrP mRNA and protein in human and feline OSCC cell lines was associated with the ability to induce in vitro and in vivo osteoclastogenesis and bone

33 resorption. Bone invasive OSCC cells stimulated the expression of receptor activator of nuclear factor κB ligand (RANKL, an agonist of osteoclastogenesis) in murine preosteoblasts and suppressed osteoprotegerin expression (OPG, an inhibitor of osteoclastic bone resorption). Culture of human and feline OSCC cells in bone- conditioned medium increased PTHrP secretion and increased tumor cell proliferation.

Human and feline OSCC expressing the highest levels of PTHrP were associated with in vitro bone resorption and osteoclastogenesis and in addition to marked osteoclastic bone resorption and bone invasion when injected adjacent to the maxilla in nude mice.

Additionally, transfection of OSCC cell lines with luciferase permitted sensitive determination of tumor growth using in vivo bioluminescent imaging.

In conclusion, OSCC-induced bone resorption was associated with tumor cell secretion of PTHrP and with increased preosteoblast-expression of RANKL and suppressed preosteoblast expression of OPG. Bone-conditioned medium increased OSCC proliferation and secretion of PTHrP. These preclinical models of OSCC recapitulate the bone invasive phenotype characteristic of the disease in both humans and cats, and will be useful to future studies of bone invasive OSCC.

Introduction

Oral and oropharyngeal cancer is the 8th most common cancer world-wide (37). In

2009, The American Cancer Society estimated that 35,720 people in the U.S. would be

diagnosed with cancer of the oral cavity and pharynx and 7,600 people would die (38).

Approximately 90% of oral and oropharyngeal tumors are squamous cell carcinoma

(OSCC) (39-41). There has been minimal improvement in the 5-year disease specific survival for OSCC, which is currently 61% for all stages combined (42). Development of

successful therapies depends on the development of OSCC animal models which

faithfully recapitulate complicated tumor–host interactions including those related to

immune response, angiogenesis, invasion and metastasis (43).

OSCC frequently invades bone and is associated with osteoclastic bone resorption (44,

45). The bone invasive behavior of OSCC contributes to the clinical morbidity of the patient and is associated with poorer prognosis (46-51). Despite the frequency and clinical impact of bone invasion in OSCC, the mechanisms responsible for the induction of osteoclastic bone resorption and bone invasion remain poorly understood.

Numerous animal models are available for the study of head and neck cancer; however, many are designed to study the early stages of carcinogenesis and involve exposing tissues of the oral cavity of hamsters, mice and rats to carcinogenic agents such as dimethylbenzanthracine (DMBA) and 4-nitroquinolone oxide (4NQO), or involve injection of primary or established OSCC cell lines subcutaneously in syngeneic or immunocompromised rodent models resulting in noninvasive tumor growth (43, 52).

There are few preclinical, in vivo models which recapitulate the bone invasive behavior of OSCC in which to evaluate therapeutic agents.

In vivo OSCC studies are dominated by 2 osteolytic OSCC cell lines; BHY, derived from a human gingival OSCC (53), and SCCVII, derived from a C(3)H/HeN murine

mouth floor OSCC (54). The objective of this study was to develop a relevant in vitro and

in vivo model of OSCC-associated bone resorption utilizing cell lines derived from primary OSCC tumors from humans and domestic cats. As in humans, OSCC is the most commonly diagnosed tumor of the oral cavity in cat (1, 2), and has a highly invasive,

osteolytic phenotype (3, 55) with similarities in clinical progression and pathology

compared to human OSCC (18).

We hypothesized that OSCC invasion into bone is facilitated by tumor-derived

parathyroid hormone-related protein (PTHrP) and bone-derived factors such as

transforming growth factor beta (TGF-β1). A panel of novel and established OSCC cell

lines was evaluated for the ability to stimulate osteoclastic bone resorption in vitro and in

vivo, to increase preosteoblast-expression of receptor activator of nuclear factor κB ligand

(RANKL, an activator of osteoclastogenesis) and to inhibit preosteoblast-expression of

osteoprotegerin (OPG, the soluble receptor of RANKL and inhibitor of

osteoclastogenesis). Cell lines derived from primary OSCC tumors in cats in addition to

human OSCC cell lines were evaluated to determine if the mechanism of bone invasion

in OSCC was similar between the two diverse species, and to better characterize bone

invasive feline OSCC as a spontaneous model of the human disease.

Materials and Methods

Cell culture reagents and established cell lines

Cell culture medium and related materials (William’s E, high glucose DMEM, BGJb +

Glutamax, αMEM + Glutamax, fetal bovine serum (FBS), L-glutamine, 0.25% trypsin-

EDTA, 0.05% trypsin-EDTA, and phosphate buffered saline) were purchased from

Invitrogen (Carlsbad, CA). Cell culture antimicrobials (Primocin and Normocin) were purchased from InvivoGen (San Diego, CA). Cells were grown in BD Falcon cell culture flasks or multiwell plates as described (BD Biosciences, San Jose, CA). Bovine serum albumin (BSA) was purchased from Sigma-Aldrich (St. Louis, MO). Serum free medium was prepared as serum-containing medium except FBS was replaced with 0.1% BSA.

Human cell lines (A253, salivary squamous cell carcinoma; SCC25, lingual squamous cell carcinoma; NHDF, dermal fibroblasts) and murine preosteoblast cells (MC3T3) were purchased from ATCC (Manassas, VA). The human laryngeal squamous cell carcinoma cell line, UMSCC12, was kindly provided by Dr. Thomas Carey at the University of

Michigan. The feline laryngeal squamous cell carcinoma cell line, SCCF1, was previously derived and characterized by Tannehill-Gregg et al. (10).

Cell line derivation

The SCCF2 cell line was derived from a bone-invasive gingival squamous cell

carcinoma of a 7-year-old male castrated domestic shorthaired cat and the SCCF3 cell

line was derived from a lingual squamous cell carcinoma of a 12-year-old male castrated

domestic medium-haired cat using methods previously described (10). Breifly, tumor

tissue was harvested immediately following euthanasia and placed in 4°C Willams E medium supplemented with 100 µg/ml Normocin™, 10% FBS and 2mM L-glutamine.

Tissue was minced and digested for 2 hours in 1000 units/ml collagenase type 4

(Worthington Biochemical, Lakewood, NJ) in serum-free Williams E medium on a rocker platform at 37°C. The digested tissue was seeded into cell culture flasks and maintained in complete medium with Primocin™ (InvivoGen). Fibroblasts were preferentially removed from primary cultures with by sequential trypsinization in 0.05%

trypsin-EDTA. Cells were switched to growth medium consisting of DMEM high

glucose medium supplemented with 10% FBS, 2mM L-Glutamine, and Normocin™.

Cultured cells were periodically tested for mycoplasma using a PCR-based assay

(VenorGeM®; Sigma-Aldrich, St. Louis, MO). Phase contrast microscopy and digital

photography was used to image colonies of adherent cells at passage 25. Tumor-

associated fibroblasts (TAF) from a primary feline gingival OSCC unrelated to SCCF2

and SCCF3 were isolated and cultured using the methods described above.

Cytokeratin immunocytochemistry

SCCF2 and SCCF3 cells were grown on sterile glass coverslips, rinsed in PBS, fixed in ice-cold 100% methanol and air dried. Coverslips were rehydrated in 3 washes of PBS and were submitted for pancytokeratin (AE1 / AE3, DAKO, Carpinteria, CA) immunocytochemistry at the Histology and Immunohistochemistry Core in the

Department of Veterinary Biosciences at The Ohio State University. The coverslips were routinely mounted on glass slides and imaged using light microscopy and digital photography.

Growth curves

Eighty-thousand (8x104) SCCF2 and SCCF3 cells were seeded into T-25 flasks in

complete growth medium. The cells from three flasks per cell line were harvested and

counted every 2 to 3 days until the cultures reached confluence. The doubling time of

SCCF2 and SCCF3 cell during the logarithmic phase of growth were determined using

Minitab version 14.1 statistical software (Minitab, State College, PA). Doubling time was

determined by the average doubling time of three separate experiments.

Electron microscopy

SCCF2 and SCCF3 cells were grown to dense confluence on glass strips. The cells

were rinsed in PBS and fixed in 3% glutararaldehyde, 0.2 M sodium phosphate (pH 7.2)

and 1mM NaN3 for 5 minutes on ice followed by 3 hours at room temperature under

vaccuum. The cells were rinsed in PBS, and incubated in 1% OsO4 at 4°C for 4 days.

OsO4 was removed with several washes of 10 mM sodium phosphate buffer, and the cells

dehydraded through an ethanol gradient ending in 100% prolylene oxide followed by

embedding in EPON resin. Ultrathin sections were prepared using an ultramicrotome,

stained with 1% uranyl acetate and lead citrate, and visualized using a Philips EM 300

electron microscope.

Karyotype

SCCF2 (passage 3) and SCCF3 (passage 6) cells were seeded into T-25 flasks.

Subconfluent cultures were submitted to the Molecular Cytogenetics Shared Resource at

The Ohio State University where metaphase chromosome preparations were prepared and evaluated for numeric and structural chromosomal abnormalities. The equivalent of human cytogenetic nomenclature was used to describe feline karyotypes (56) and the band nomenclature was based on published ideograms (57).

In vitro bone resorption and osteoclastogenesis

Calvarial bone was harvested from 5 to 8-day-old euthanized mouse pups and divided

into two bone disks using a 4 mm diameter dermal biopsy punch. Calvarial disks were

cultured overnight in BGJb + Glutamax medium supplemented with 0.1% BSA and

Primocin. Forty-eight hours prior to co-culture, 1.25x104 cells (NHDF, UMSCC12,

A253, SCC25, SCCF1, SCCF2 and SCCF3) were plated per well of 24-well plates.

Medium was changed to high glucose DMEM supplemented with 0.1% BSA and one

bone disk was added per well. Medium was replaced every 48 hours for 7 to 14 days as

indicated. Calvaria were fixed in neutral-buffered formalin followed by 70% ethanol and

stained for tartrate-resistant acid phosphatase (TRAP) activity using a commercially

available kit (Sigma-Aldrich, St. Louis, MO) as previously described (58). TRAP-stained

calvaria were mounted on glass slides and coverslipped for digital photography. Bone area for each disk was calculated (Image-Pro plus histomorphometry software, Bathesda,

MD) and compared between each cell line.

Bone marrow mononuclear cells (BMMC) were cultured in medium conditioned by

OSCC cells and evaluated for TRAP activity. Marrow cells were flushed from the femurs, tibia and humeri of 4 to 6-week-old male mice using αMEM medium +

Glutamax, supplemented with 10% FBS and normocin. Mononuclear cells were seeded at 2.5x105 cells per well of a 24-well plate. The cells were cultured in αMEM growth medium supplemented with 30 ng/ml recombinant murine RANKL (Peprotech Inc.,

Rocky Hills, NJ) and 5 ng/ml recombinant murine MCSF (R&D Systems, Minneapolis,

MN) for 72 hours. Medium was replaced with 1:1 mixture of αMEM growth medium and conditioned medium or unconditioned medium. Conditioned medium was prepared by culturing OSCC cells in DMEM growth medium and harvesting every 48 hours. Medium was stored at -30C and filter sterilized prior to use. Four wells of BMMCs were cultured per cell line. Medium was changed 3 days latera and cells were fixed 6 days later (10% neutral-buffered formalin followed by 70% ethanol). TRAP staining was performed using a commercially available kit as previously described (59).

Transfection with luciferase-yellow fluorescent protein

A plasmid containing a 2183-base pair luciferase-yellow fluorescent protein (YFP)

fusion construct driven by a CMV promoter and containing neomycin and ampicillin

resistance cassettes (pCDNA3.1(+).yLuc-YFP) was kindly provided by Dr. Christopher

Contag (Stanford University, Stanford, CA, USA). SCCF2, SCCF3 and UMSCC12 cells were transfected with 1 µg pCDNA3.1(+).yLuc-YFP using 2.5 µl Lipofectamine™ LTX,

1 µl PLUS™ Reagent and 200 µl of antibiotic and serum-free Opti-MEM® (Invitrogen),

according to the manufacturer’s instructions. Cells were selected for antibiotic resistance

with 0.6 – 0.8 mg/ml Geneticin® (Invitrogen). Luciferase expression in surviving cells

was determined using the IVIS 100 system in growth medium supplemented with 142.5

µg/ml D-Luciferin (Caliper Life Sciences, Hopkinton, MA). Colonies with the highest

bioluminescence were selected using sterile cloning disks (Research Products

International, Mt. Prospect, IL) soaked in 0.25% trypsin-EDTA and expanded. Three to

six clones with the highest bioluminescence were selected and pooled. The resulting cell

lines were named SCCF2Luc, SCCF3Luc and UMSCC12Luc. Human OSCC cell lines

A253 and SCC25 were not permissive to luciferase-transfection. Luciferase expression

was periodically checked by imaging adherent cells or by imaging serial dilutions using

the IVIS 100 as described above. The SCCF1Luc cells were previously transfected by

Tannehill-Gregg et al. and were referred to as SCCF110YFP-Luc cells (18).

Orthotopic nude mouse model of bone invasive OSCC

All animals were 6-week-old male nu/nu mice (NCI, Frederick, MD). Animal care

procedures were approved by the 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”. Forty mice were assigned to 1 of 4

groups; UMSCC12Luc, SCCF1Luc, SCCF2Luc and SCCF3Luc. Each cell line was

injected into 10 mice, which were divided into two groups of 5 mice. One group received

cells suspended in a 1:1 mixture of PBS and Matrigel (BD Matrigel™ , high concentration, phenol red free, Franklin Lakes, NJ) , and the other received cells suspended in PBS alone. Mice received 5x105 or 1x106 cells suspended in 0.1ml of

vehicle. Xenograft progression did not differ in animals receiving 1x106 cells, so they

were combined with mice receiving 5x105 cells for purposes of summarizing the data.

The xenografts were permitted to grow until the tumors reached 1cm in diameter, unless

early removal criteria were reached (weight loss of 15% or tumor ulceration). Mice had

bioluminescent imaging every 1 to 2 weeks. Mice were euthanized by cervical

dislocation during isoflurane anesthesia immediately following acquisition of a

bioluminescent image. Tissues were fixed in 10% neutral-buffered formalin until

histologic processing and evaluation.

Bioluminescent imaging (BLI)

In vivo BLI was performed with the IVIS 100 system as previously described (18).

Results were analyzed using LivingImage® software, version 2.2 (Caliper Life Sciences).

Mice were injected intraperitoneally with 4.3 mg D-luciferin in sterile PBS, and imaged

while under isofluorane anesthesia. A circular region of interest (ROI) surrounding the

xenograft was used to determine the radiance (photons/second) for each xenograft. Peak

ROI radiance was used for comparisons. Ex vivo imaging of submandibular and cervical lymph nodes, lungs, liver, spleen and kidneys was performed on all mice immediately following BLI and euthanasia.

Faxitron radiography and histopathology

The mandible was removed from each skull and the degree of bone loss was evaluated

qualitatively using a Faxitron cabinet X-ray system (Hewlett-Packard, McMinnville, OR)

at 45 kVp for 3.5 min. The skulls were decalcified in 10% EDTA pH 7.4 at 4°C for 7 to

14 days. The skulls were divided into 3mm thick sections, paraffin-embedded, sectioned

at 5µm and stained with hematoxylin and eosin (HE) followed by microscopic evaluation.

Xenografts were evaluated histologically for evidence of bone invasion. Enzymatic

histochemistry for TRAP-positive osteoclasts was performed on tissues from

representative mice as previously described (60).

Real-time RT-PCR

OSCC cells (UMSCC12, A253, SCC25, SCCF1, SCCF2 and SCCF3) were evaluated

using real-time reverse transcriptase polymerase chain reaction for RT-PCR expression of

PTHrP, colony stimulating factor-1 (CSF-1), RANKL and OPG. Murine preosteoblasts

were evaluated for expression of RANKL and OPG. RNA extraction, reverse 44 transcription, PCR, and real-time PCR were performed as previously described (59). The feline primers for real-time RT-PCR included β2-microglobulin (B2M) (forward:

CTACTTCTGGCGCTGCTCG and reverse: CCTGAACCTTTGGAGAATGC), PTHrP

(forward: GCGCTCGGTGGAGGAACTCG and reverse:

AGCAGGCTTGGAGTTGGGGGA), RANKL (forward:

AAGGGGTATGACCCACAATG and reverse: AAGCCCTCGGACTGTAACAA), osteoprotegerin (forward: CCAGCTACTGAAGTTATGGAAACA and reverse:

GAGGTTTGTGTGTCCGAGGT), and CSF-1 (forward:

GGCGGAGCCATCGAGAGCCT and reverse: ACGCCCCCGTGATGTCTCGT ).

OSCC gene expression and effect of OSCC conditioned medium on MC3T3 expression of RANKL and OPG

OSCC and TAF cells were seeded at 1x105 cells per well in six-well plates and grown for 48 hours in growth medium followed by overnight serum starvation and 6-hour stimulation with serum-containing growth medium. RNA extraction and real-time RT-

PCR was performed as described above and expressed as relative gene expression compared to the lowest expressing cell line. OSCC conditioned medium was prepared by culturing UMSCC12, SCCF2 and SCCF3 cells in serum free medium and harvesting conditioned medium every 48 hours. MC3T3 cells were seeded in 6-well plates at a density of 1x105 cells per well in serum-containing α-MEM growth medium and cultured for 48 hours, followed by serum-free medium for 24 hours. Medium was changed to 50%

serum-free α-MEM medium with 0.1% BSA, 100 µg/ml normal chicken IgY, and 50%

OSCC-conditioned medium or unconditioned medium and incubated for 3 hours.

Alternatively, MC3T3 cells were cultured in 10nM human PTHrP (1-86, Abcam,

Cambridge, MA). RNA isolation and real-time RT-PCR was performed using methods

and primer sequences previously described (59) to detect the expression of murine

RANKL and OPG in MC3T3 cells.

Effect of bone-conditioned medium on OSCC proliferation

Calvarial disks were harvested from 3 to 5-day-old mouse pups as described above, and

cultured in serum-containing high glucose DMEM at a density of 2 disks/ml media.

Medium was collected and changed every 48 hours. Fibroblasts (NHDF and feline

fibroblasts isolated from SCCF2 and SCCF3 primary tumors) were grown in growth

media, and media was harvested every 48 hours. Human OSCC cells were exposed to

NHDF-conditioned medium and feline OSCC cells were exposed to feline fibroblast

conditioned medium. 1x103 NHDF or OSCC cells were seeded per well in 96-well plates

in growth medium. After 24 hours, the medium was changed to 50% growth medium /

50% conditioned medium (bone or fibroblast). The conditioned medium was changed at

48 hours. A commercial MTT assay (Promega, Madison, WI) was performed according

to the manufacturer’s instructions after 96 hours of culture in conditioned media in order

to determine the effect of bone-conditioned medium on OSCC proliferation.

Effect of bone-conditioned medium on OSCC expression of PTHrP

OSCC cell lines were cultured in serum-free, bone-conditioned medium, MC3T3-

conditioned medium, or 12 ng/ml recombinant human TGF-β1 (R&D). Bone-conditioned

medium was prepared by culturing 1 calvarium per 2 ml of serum-free medium (high

glucose DMEM). Medium was harvested every 96 hours for 8 days. MC3T3-conditioned

medium was prepared from confluent cultures grown in serum free high-glucose DMEM

every 96 hours for 8 days. OSCC cells were seeded in 6-well plates at a density of 1x105

cells per plate in growth medium for 48 hours followed by serum-free medium for 24

hours. Medium was replaced with 50% unconditioned serum-free medium and 50%

conditioned serum-free medium for 3 hours (mRNA expression) or for 24 hours (PTHrP

secretion). PTHrP mRNA expression was measured using real-time RT-PCR and the

concentration of PTHrP secreted into the culture medium was determined with a PTHrP

immunoradiometric assay (Diagnostic Systems Laboratories Inc., Webster, TX, USA)

according to the manufacturer’s instructions.

Statistical analysis

Results are displayed as dot plots or bar graphs with means and standard error. Data

were analyzed using Student’s t-test or ANOVA and Bonferroni’s post hoc test.

Normalized gene expression data (∆CT) was analyzed for statistical significance using

ANOVA and Bonferroni’s post hoc test, and graphically represented by showing relative

expression compared to the cell line with the lowest expression. Data with P values less

than 0.05 were considered statistically significant. All statistical comparisons were

performed with STATA Intercooled 10 software (Cary, NC).

Results

Cell line derivation and morphologic characterization

The SCCF2 primary tumor was composed of a well differentiated squamous cell

carcinoma with evidence of bone resorption (figure 2.1A). The SCCF3 primary tumor

was a well differentiated lingual squamous cell carcinoma which had infiltrated and

expanded the entire thickness of the tongue (figure 2.1B). Both tumors were composed of

islands and cords of neoplastic epithelium demonstrating squamous differentiation with

the formation of keratin pearls (most pronounced in the SCCF3 tumor) and the induction

of dense fibrous stroma (desmoplasia). Culture of minced tumor from both tissue donors

resulted in the growth of adherent round to polygonal cells that maintained close cell-to-

cell contact and formed colonies composed of cells in a cobblestone arrangement typical

of epithelial cells (figures 2.1C and D). All of the SCCF2 and SCCF3 cells were

pancytokeratin positive (figures 2.1E and F). Transmission electron microscopy revealed

that both cell lines had intercellular junctions (desmosomes) and tonofilaments (typical of

epithelial cells), in addition to irregular nuclear envelopes (typical of neoplastic cells)

(figure 2.2). Degenerative changes including swollen mitochondria and numerous 48

phagolysosomes, attributed to stress associated with high density in vitro culture. The

average doubling time of SCCF2 cells (27 ± 3 hours) was slightly shorter than SCCF3

cells (30 hours ± 3 hours)(data not shown).

SCCF2 and SCCF3 cells were aneuploid with multiple marker chromosomes (figure

2.3). There were no common structural abnormalities between the two cell lines. SCCF2

cells (figure 2.3A) were 4n with chromosomal additions to B1 (short arm), D4 and F1.

No normal D4 chromosomes were observed and no Y chromosomes were observed.

Approximately one half of the SCCF3 cells (figure 2.3B) were diploid and the remaining

cells were tetraploid; however, all had the same chromosomal abnormalities which

included additions to B1 (long arm), C2, F2 and deletion of part of D1. No normal E3

chromosomes were observed in SCCF3 cells.

UMSCC12 and SCCF2 cells induced in vitro bone resorption and osteoclastogenesis

OSCC co-culture with murine bone was performed to determine which OSCC cell lines induced bone resorption in vitro. Co-culture of UMSCC12 and SCCF2 cells with calvaria resulted in significantly reduced bone area compared to calvarial bone co-cultured with

NHDF cells (figure 2.4A). Culture of murine BMMC’s in a 1:1 ratio of osteoclastogenesis medium and conditioned medium from UMSCC12, SCC25 and

SCCF2 cells induced the formation of numerous large, multinucleated TRAP-positive osteoclasts (figure 2.4B). Few osteoclasts were formed in the presence of SCCF3 and

NHDF medium. Similarly few osteoclasts were formed in TAF-conditioned medium

(data not shown).

Luciferase-expressing UMSCC12 and SCCF2 cells induced osteoclastic bone resorption

and invaded maxillary bone in vivo

BLI was used to evaluate SCCF1Luc, SCCF2Luc, SCCF3Luc and UMSCC12Luc xenograft growth with and without Matrigel, and to detect regional and distant metastasis ex vivo (figure 2.5A, B, C and table 2.1). Histopathologic examination was used to confirm the presence of metastasis (figure 2.5D and table 2.1), and to confirm that bioluminescence corresponded with the presence of OSCC xenografts (figure 2.6).

The four cell lines differed in incidence of sustained xenograft growth, rate of tumor progression, bone-invasive behavior and metastasis (table 2.1). In the absence of

Matrigel, the only cell line that progressive xenograft growth in all mice was SCCF3Luc cells. Mice bearing SCCF3Luc xenografts met removal criteria (tumor size and / or weight loss) most rapidly (19-28 days), compared to the other cell lines. In the absence of

Matrigel, xenografts developed from SCCF2Luc cells in 4 of 5 mice, from

UMSCC12Luc cells in 3 of 5 mice, and from SCCF1Luc cells in 2 of 5 mice. Mice bearing UMSCC12Luc xenografts demonstrated the slowest rate of tumor progression

(83-90 days). Interestingly, addition of Matrigel increased the incidence of xenograft

formation to 5 of 5 mice for UMSCC12Luc and SCCF1Luc cells, but reduced the

incidence of sustained xenograft growth SCCF2Luc injected mice. Addition of Matrigel

50 increased the rate of tumor progression by the greatest degree in SCCF1Luc-bearing mice followed by UMSCC12Luc-bearing mice.

The greatest degree of osteoclastic bone resorption and maxillary invasion was observed in SCCF2Luc-bearing mice, followed by UMSCC12Luc bearing mice (table 2.1 and figure 2.6A-D). Mild to moderate bone resorption was observed in mice injected with

SCCF3Luc cells, and minimal to no bone resorption was observed in mice injected with

SCCF1Luc cells. Periosteal new bone was observed in SCCF2Luc and SCCF3Luc- bearing mice, which was observed at the periphery of the tumor-bone interface (figure

2.A and B). Bone invasion was associated with the presence of numerous TRAP-positive osteoclasts in resorption pits on the maxillary surfaces adjacent to the xenograft (figure

2.6D).

OSCC expression of osteoclastogenic factors and effect of OSCC conditioned medium on

MC3T3 expression of RANKL and OPG

The UMSCC12 and SCCF2 cells expressed significantly greater PTHrP mRNA (5-fold and 50-fold) compared to A253 and SCC25 or SCCF3 and TAF cells respectively (figure

2.7A). The PTHrP mRNA levels corresponded with the ability of UMSCC12 and SCCF2 cells to induce osteoclastic bone resorption. Interestingly, UMSCC12 expressed the highest levels of CSF-1 mRNA (colony stimulating factor-1, synonymous with macrophage colony stimulating factor, a cytokine capable of stimulating osteoclast formation) compared to either A253 or SCC25 suggesting that CSF-1 may have contributed to the stimulation of osteoclast formation. In contrast, bone-invasive SCCF2 51

cells expressed the lowest levels of CSF-1 mRNA compared to the other feline OSCC cell lines (figure 2.7B). RANKL expression was relatively low in UMSCC12 and SCCF2 cells (figure 2.7C) compared to A253 cells and TAF cells, suggesting that the ability to stimulate osteoclast formation and bone resorption does not rely solely on OSCC- expression of RANKL. TAF cells expressed the highest levels of CSF-1 and RANKL mRNA compared to the feline OSCC cell lines, indicating that the tumor stroma may have a significant role in stimulating osteoclast formation in bone invasive OSCC.

Interestingly, cells associated with less bone resorption (A253 and SCCF3), expressed the highest levels of OPG (figure 2.7D).

RANKL mRNA expression was not detectable in unstimulated MC3T3 cells, but was induced by 10µM PTHrP 1-86, SCCF2-CM and UMSCC12-CM (figure 2.8A). SCCF3- conditioned medium did not induce MC3T3-expression of RANKL (figure 2.8B), corresponding to the low degree of bone resorption and osteoclastogenesis observed with this cell line.

Unstimulated MC3T3 cells expressed OPG, which was reduced by the addition of

PTHrP 1-86. SCCF3 conditioned medium stimulated the expression of OPG, in contrast to UMSCC12 conditioned medium which inhibited OPG expression. There was no effect of SCCF2-conditioned medium on MC3T3-expression of OPG compared to untreated

MC3T3 cells. Taken together, exposure of MC3T3 cells to SCCF2 or UMSCC12 conditioned medium increased RANKL expression relative to OPG expression (a pro- osteoclastogenesis expression profile).

Effect of bone-conditioned medium on OSCC proliferation and expression of PTHrP

OSCC cells cultured in bone conditioned medium proliferated more rapidly compared to unconditioned medium (UMSCC12, figure 2.9A), or both unconditioned medium and fibroblast conditioned medium (A253, SCCF1, SCCF2 and SCCF3). A similar response was not observed from NHDF cells. Bone-conditioned medium stimulated OSCC- expression of PTHrP mRNA (figure 2.9B). TGF-β1 stimulated SCCF2-expression of

PTHrP mRNA; however, TGF-β1 did not significantly stimulate PTHrP expression in the other OSCC cell lines. There was no effect of MC3T3-conditioned medium on

UMSCC12 or SCCF2-expression of PTHrP (figure 2.9B). Bone-conditioned medium stimulated PTHrP secretion into culture medium in all OSCC cell lines evaluated

(UMSCC12, SCC25, SCCF1, SCCF2 and SCCF3).

Discussion

We have developed a novel and useful in vitro and in vivo model of OSCC-associated osteoclastic bone resorption utilizing cell lines derived from human and feline cancers.

Modeling bone invasive OSCC in vitro using bone-co-culture, and in vivo using an orthotopic model of maxillary invasion, has not been previously reported. There are few orthotopic models of head and neck cancer that demonstrate the invasive behavior typical of natural disease, including the induction of osteoclastic bone resorption and subsequent

53 invasion into the maxilla or mandible. In vivo studies of bone invasive OSCC include two osteolytic OSCC cell lines; BHY, derived from a human gingival OSCC (53), and

SCCVII, derived from a C(3)H/HeN murine mouth floor OSCC (54). Both BHY and

SCCVII cells are typically injected percutaneously into the masseter muscle to achieve mandibular invasion. Other cell lines reported to have bone invasive activity in vivo include UMSCC1 injected percutaneously ventral to the mandible (61), UMSCC11A,

UMSCC11B, BICR31 and BICR56 injected intraorally into the floor of the mouth (62,

63), and HSC3 injected percutaneously adjacent to the parietal bone (35).

Our objective was to develop a novel model of maxillary-invasive OSCC which reliably stimulates osteoclastic bone invasion and mimics disease progression in natural disease, and to determine the role of PTHrP and bone-derived factors in the pathogenesis of osteoclastic bone resorption. In addition to utilizing human cell lines, we developed and evaluated three feline OSCC cell lines. These experiments confirmed the importance of PTHrP using cells from two species. Characterization of feline OSCC will promote the use of naturally occurring OSCC in cats as a spontaneous model of the human disease.

The prognosis of OSCC in cats is poor with few treatment options. Prior to this study, the only feline OSCC cell line available was SCCF1, which demonstrated limited ability to induce bone resorption (18).

Human OSCC karyotypes are frequently complex with numerous numeric and structural abnormalities (64, 65). Both SCCF2 and SCCF3 cells had complex karyotypes

(3 or more numeric and/or structural changes) with numerous marker chromosomes.

SCCF2 and SCCF3 cell lines had few cytogenetic abnormalities in common, which were

54 limited to loss of E3 (SCCF3 had no normal E3 chromosomes, SCCF2 was missing one

E3 chromosome), and each had loss of one F1 chromosome. The E3 chromosome has homology to human chromosomes 7 and 16 (neither of which are commonly lost in human oral SCC) (64), and the F1 chromosome has homology to human chromosome 4

(lost in 19 of 106 human OSCC) (64). Early alterations in the progression of oral cancer in humans include loss of genetic material at chromosomes 9p21 (p16), followed by alteration of 17p13 (P53), and as a later event, gain of chromosomal material at 11q13

(cyclin D) (66). Interestingly, no normal D4 chromosomes were observed in the SCCF2 cells (homologue of human chromosome 9 and site of p16) (67, 68). Neither cell line had abnormalities of E1 (homologue of human chromosome 17 and locus of TP53) (67, 68) or gains in D1 (homologue of human chromosome 11q and locus for cyclin D) (67, 68).

The karyotype of the SCCF1 cells was previously reported, which was characterized by aneuploidy with numerous marker chromosomes and missing chromosomes (specific structural changes were not detected) (10).

The results suggest that feline OSCC demonstrate a variety of chromosomal abnormalities and indicate these tumors are genetically unstable and capable of acquiring and propagating significant chromosomal rearrangements while escaping cell death.

Further study of the integrity and function of p16, p53 and cyclin D in the feline cell lines is needed.

Co-culture of OSCC cells or conditioned medium with bone marrow cells or osteoclast precursors has been previously used to study OSCC-associated osteoclastogenesis (69-

71). This study also utilized murine calvarial bone in co-cultures with OSCC cells in

order to observe physical evidence of bone resorption in the presence of a more complete

bone environment including osteocytes, bone lining cells, osteoblasts, osteoclasts, fibrous

tissue and endothelial cells. Murine bone co-cultures have been utilized to study other

forms of osteolytic neoplasia such as Adult T-cell Leukemia / Lymphoma (59) and prostate cancer (72). The combination of both assays demonstrated the ability of OSCC

cells of human and feline origin to stimulate osteoclast formation and bone resorption.

Matrigel was previously shown to improve tumor engraftment in nude mice injected

with SCCF1Luc cells adjacent to the mandible, adjacent to the calvarium, or into the

medullary canal of the tibia (18). The SCCF1Luc xenografts rarely invaded bone, consistent with what was found in the experiments presented here. Addition of Matrigel to the cell suspension vehicle proved beneficial in supporting xenograft growth in

SCCF1Luc and UMSCC12Luc cells. Mice bearing SCCF3Luc xenografts had a high rate of tumor engraftment and tumor growth and there was no benefit of adding Matrigel.

Matrigel appeared to improve tumor growth in the early stages of xenograft formation in

SCCF2Luc-bearing mice, but surprisingly, the tumors regressed in all but 1 mouse.

The invasive phenotype in vivo was similar to the ability of the cells to induce bone resorption in vitro (UMSCC12 and SCCF2 cells induced the greatest degree of bone resorption in vitro and in vivo). The relatively short duration of SCCF3Luc xenograft growth resulted from the rapidity by which mice reached termination criteria for tumor size. It is possible that the short period of time that the SCCF3Luc xenografts were permitted to grow contributed to the low degree of bone invasion; however, the low degree of in vivo bone resorption is consistent with the low degree of in vitro bone

resorption and osteoclastogenesis associated with this cell line. The ability of SCCF2Luc

cells to induce osteoclastic bone resorption and invade maxillary bone without Matrigel

make this model ideal for future studies of bone invasive OSCC.

Human and feline OSCC cell lines demonstrated variable ability to induce in vitro

bone resorption and osteoclast formation. The human cell line UMSCC12 and the feline

cell line SCCF2 were associated with the greatest degree of bone resorption and

osteoclast formation in vitro and the greatest degree of bone resorption in vivo.

Interestingly, these cell lines also expressed the greatest levels of PTHrP.

The human and feline OSCC cell lines evaluated in this study all expressed PTHrP.

This is consistent with previous reports showing that PTHrP is commonly expressed in human OSCC tissue (35, 69, 73-75) and cell lines (69), in addition to feline OSCC tissue

(55) and SCCF1 cells (10). OSCC-expression of PTHrP has been shown to influence in

vitro tumor cell proliferation, migration and invasiveness (76), and to participate in the

regulation of osteoclastogenesis by increasing osteoblast expression of RANKL (35, 69,

77).

Studies investigating the mechanisms by which OSCC-derived PTHrP stimulate

osteoclastogenesis are somewhat contradictory. Tada et al. reported that bone-invasive

BHY cells expressed low levels of PTHrP, and that the ability of OSCC cells to induce osteoclastogenesis was correlated with expression of TNFα and was associated with a

suppression of OPG rather than stimulation of RANKL in murine osteoblasts (70).

Furthermore, treatment with a TNFR1 inhibitor reduced osteoclastogenesis associated

with BHY co-culture (70). More recently, BHY cells have been shown to express PTHrP

and to stimulate RANKL expression in MC3T3 cells (69, 77), although there was no

effect on MC3T3 expression of OPG (77). Kayamori et al. reported that a PTHrP

neutralizing antibody partially inhibited the BHY-stimulation of RANKL expression in

ST-2 cells (rat osteoblasts) and osteoclast formation in vitro (35). BHY conditioned

medium did not alter ST-2 expression of OPG (35). In a recent report, knockdown of

PTHrP in murine SCCVII cells inhibited osteoclastic bone resorption of the mandible in

mice (54).

We demonstrated that OSCC cell lines proliferated more rapidly and had increased expression and secretion of PTHrP when cultured in bone-conditioned medium. The factors responsible these effects are unknown; however, bone is a reservoir of numerous

growth factors with the ability to promote tumor progression and include transforming

growth factor β (TGF-β1), fibroblast growth factor (FGF), insulin-like growth factors

(IGFs) I and II, platelet derived growth factor (PDGF), and bone morphogenic proteins

(BMP) (78). In fact, TGF-β1 is known to stimulate PTHrP expression in a variety of tumor cells and is theorized to function in a vicious cycle of tumor growth and bone invasion in skeletal metastasis of human breast cancer (79). We have observed that bone- conditioned medium contains latent TGF-β1 using a commercially available TGF-β1

ELISA (unpublished data). TGF-β1 stimulated PTHrP expression in SCCF2 cells. In all cell lines, the TGF-β1 response was less than what was achieved with bone-conditioned medium, suggesting that TGF-β1 in bone-conditioned medium is not solely responsible for the stimulatory effect on PTHrP. Calcium released from resorbing bone has been proposed to play a role in stimulating tumor-expression of PTHrP in bone-metastatic

breast cancer (80), and resorbing bone has been shown to release calcium into culture

medium (59). It is possible that bone-derived calcium in the conditioned medium

contributed to the stimulation of PTHrP in the OSCC cell lines.

The ability of TGF-β1 to stimulate SCCF2-expression of PTHrP is in contrast to the

findings of Goda et al., who reported that TGF-β1 treatment stimulated RANKL

expression in an OSCC (HSC-2) cell line, however there was no effect on expression of

PTHrP (81). Differences in experimental design may have contributed to this

discrepancy, since they used less TGF-β1 (5ng/ml) for a longer duration (24 hours)

compared to what was used in the experiment presented here. We determined that maximal stimulation of PTHrP expression in SCCF2 cells was achieved with 12 ng/ml

TGF-β1 for 3 hours. Our findings are in agreement with those of Takayama et al., who

demonstrated that PTHrP expression in SCCVII cells was stimulated by TGF-β1 (54).

The role of bone-derived TGF-β1 in stimulating OSCC proliferation is unknown. TGF-

β1 normally inhibits epithelial cell proliferation, however; abnormalities of TGF-β1 receptor signaling have been frequently documented in human OSCC and include alterations that inhibit the normal functionality of the TGF-β type II receptor (TGFbRII) and downstream signaling molecules Smad2 and Smad4, resulting in a loss of TGF-β1 mediated growth inhibition (52). Reduced responsiveness to the inhibitory effects of

TGF-β1 are associated with tumor progression and include loss of epithelial cell adhesion, remodeling of extracellular matrix, enhanced angiogenesis, increased cell proliferation, decreased differentiation and increased transformation capacity of the tumor cells (52, 82). The SCCF2 cells demonstrated that they are capable of responding

59 to TGF-β1 treatment by increasing PTHrP expression, and therefore appear to have TGF-

β receptor functionality. Further study is required in order to characterize the downstream events of TGF-β1 receptor signaling in the SCCF2 cells, and to determine if they are indeed insensitive to the antiproliferative effects of TGF-β1.

We have developed a novel model of OSCC-associated osteoclastic bone resorption in vitro using co-culture of OSCC cells with mouse calvarial bone and in vivo using an orthotopic, bioluminescent model of maxillary invasive OSCC. We observed that the ability of OSCC cells to stimulate osteoclastic bone resorption was correlated with

PTHrP expression, and that OSCC-derived factors stimulated RANKL and inhibited

OPG in murine preosteoblasts. Bone conditioned medium and TGF-β1 stimulated OSCC- expression of PTHrP, supporting the hypothesis that OSCC invasion into bone was facilitated by a vicious cycle of tumor-derived PTHrP and bone-derived factors. These preclinical models of OSCC recapitulate the bone invasive phenotype characteristic of the disease in both humans and cats, and will be useful to future studies of bone invasive

OSCC.

Progressive Maxillary osteolysis Metastasis Cell Line Vehicle N Growth Days (range) and Invasion BLI / Histo UMSCC12Luc PBS 5 3 86 (83‐90) Moderate to marked 2 / 2 Lymph node PBS + Matrigel 5 5 74 (49‐90) Moderate to marked 1 / 0 Lymph node SCCF1Luc PBS 5 2 70 (49‐90) None PBS + Matrigel 5 5 50 (35‐64) Minimal to none 1 / 0 Lymph node SCCF2Luc PBS 5 4 49 (46‐57) Marked 2 / 2 Lungs PBS + Matrigel 5 1 47 Marked SCCF3Luc PBS 5 5 25 (19‐28) Mild PBS + Matrigel 5 5 21 (14‐26) Mild to moderate

Table 2.1: Behavior of OSCC xenografts in nude mice.

Figure 2.1: SCCF2 and SCCF2 microscopic morphology

SCCF2 and SCCF3 cell lines were derived from spontaneous feline OSCC cancers.

SCCF2 was derived from a bone invasive maxillary OSCC (A). OSCC cells formed thin cords occasionally surrounding foci of keratin and sloughed tumor cells, within dense fibrous stroma and adjacent to a fragment of partially resorbed, necrotic, lamellar bone

(HE). SCCF3 was derived from an invasive lingual OSCC (B). The tumor was composed

of large cystic spaces lined by neoplastic squamous epithelium and contained large

quantities of keratin and necrotic tumor cells. SCCF2 (C) and SCCF3 (D) cells both grew

in vitro as colonies of adherent cells with typical round to polygonal cell morphology.

SCCF2 (E) and SCCF3 (F) cells grown on glass slides were cytokeratin positive

(Pancytokeratin immunocytochemistry, DAB and hematoxylin counterstain).

SCCF2 SCCF3 AB

C D

Figure 2.2: SCCF2 and SCCF3 ultrastructure

SCCF2 and SCCF3 cells were grown on glass and were examined using transmission electron microscopy. Both cell types demonstrated desmosomes (A and B), and cytoplasmic linear structural elements interpreted as tonofilaments (C and D). Cells of both types frequently had irregular nuclear envelopes, swollen mitochondria, and membrane bound cytoplasmic structures with heterogenous, irregular contents

(phagolysosomes). Scale bars: A, 0.5 µM; B, 2.5 µM; C, 1 µM; D, 2.5 µM.

Figure 2.3: SCCF2 and SCCF3 karyotype

Both SCCF2 and SCCF3 cells were aneuploid with numerous marker chromosomes.

SCCF2 cells (A) were 4n with chromosomal additions to B1 (short arm), D4 and F1. No 65

normal D4 chromosomes were observed and no Y chromosomes were observed. SCCF2

karyotype: 70-80<4n>, XX, -Y, -Y, -A3, -A3, add(B1)(p12)x2, -C1, -D4, -D4, add(D4)(q14)x2, -E2, -E3, -E3, -F1, add(F1)(p10)x2,-F2, +mar1x2, +mar2x2, +mar3,

+mar4x4, +mar5x2 [cp20]. Approximately one half of the SCCF3 cells (B) were diploid

and the remaining cells were tetraploid; however, all had the same chromosomal

abnormalities which included additions to B1 (long arm), C2, F2 and deletion of part of

D1. No normal E3 chromosomes were observed in SCCF3 cells. SCCF3 karyotype: 36,

XY, -A1, -A2, add(B1)(q14), -B3, -B4, add(C2)(q24), del(D1)(q22), -D2, -E3, -E3, -F1, -

F2, add(F2)(p10), +mar1, +mar2, +mar3, +mar4, +mar5, +mar6, +mar7 [20] /72, idemx2[10].

Figure 2.4: Feline and human OSCC cells induced in vitro bone resorption and

osteoclastogenesis

Murine calvarial bone discs (4 mm diameter) were co-cultured with OSCC cells for 7

(black circles) or 14 (white circles) days (A). Calvaria were fixed, stained for TRAP

activity, photographed and evaluated for changes in bone area using histomorphometry.

SCCF2 (feline) and UMSCC12 (human) cells stimulated the most bone resorption.

*P<0.05, compared to NHDF (7 days only, ANOVA and Bonferroni’s post hoc test).

SCC25 was not included in the statistical comparison because of an insufficient number

of bone disks. Murine BMSCs were cultured in conditioned medium from human and

feline OSCC cells. Each panel is a representative image from one of 4 co-cultures.

UMSCC12 cells stimulated larger and more numerous osteoclasts compared to SCC25

and NHDF cells (human fibroblasts), and SCCF2 stimulated larger and more numerous

osteoclasts compared to SCCF3 cells. 67

Figure 2.5: In vivo bioluminescent imaging revealed progressive SCCF2Luc xenograft

growth and pulmonary metastasis

In vivo bioluminescent imaging revealed progressively increased bioluminescent signal

over time (A). The addition of Matrigel (white circles) initially supported increased

tumor growth, but tumor growth was better maintained when cells were injected without

Matrigel (black circles). Panel B shows increasing bioluminescent signal in a mouse bearing an SCCF2Luc xenograft without Matrigel. Ex vivo imaging of the viscera was

68 used to detect pulmonary metastasis (C). Multifocal bioluminescent signal in the lungs corresponded to microscopic pulmonary metastasis of OSCC (D). Bar = 100µM. Inset, higher magnification to show detail of metastatic OSCC.

C Figure 2.6: SCCF2Luc and UMSCC12Luc were associated with the greatest degree of bone resorption in vivo

Representative mice from each group are presented (SCCF1Luc and UMSCC12Luc with

Matrigel, SCCF2Luc and SCCF3Luc without Matrigel). A: Minimal to no evidence of bone resorption was radiographically evident in SCCF1 and SCCF3-tumor-bearing mice.

Moderate to marked bone resorption (white arrow heads) was observed in the SCCF2Luc and UMSCC12Luc-bearing mice. White arrows indicate reactive bone formation. B and

C: SCCF1Luc xenografts were rarely associated with bone resorption despite close proximity of tumor to bone. SCCF2Luc xenografts were characterized by marked bone loss and invasion with the formation of numerous large, TRAP-positive osteoclasts (D).

Less bone invasion was observed in the SCCF3Luc xenografts; however, osteoclastic bone resorption was occasionally observed (D). UMSCC12Luc frequently invaded bone

(B and C) and stimulated osteoclastic bone resorption (D). Scale bars: A, 1 mm; B, 1 mm; C and D, 50 µm.

A * *

B * *

C * *

D * *

Figure 2.7: Bone-invasive OSCC cells expressed more PTHrP and less OPG compared to minimally bone-invasive OSCC cells

UMSCC12 and SCCF2 cells stimulated a high degree of osteoclastic bone resorption in vitro and in vivo. A. UMSCC12 cells expressed the most PTHrP mRNA compared to

A253 and SCC25 cells, and SCCF2 cells expressed the most PTHrP compared to SCCF1 and SCCF3 cells. UMSCC12 and TAF cells (feline fibroblasts) expressed the most CSF-

1. C. A253 and TAF cells expressed the most RANKL. D. A253 and SCCF3 cells (both were associated with low degrees of bone resorption) expressed the greatest amount of

OPG mRNA. (*P<0.05, ANOVA and Bonferroni post hoc test).

A 4

1 mRANKL / mB2M / mB2M mRANKL (relative to PTHrP 1-86) PTHrP to (relative 0 Control PTHrP 1-86 SCCF2 CM SCCF3 CM UMSCC12 CM

4 B * 3

mOPG / mOPG mB2M 1 * (relative to control) to (relative * 0 Control PTHrP 1-86 SCCF2 CM SCCF3 CM UMSCC12 CM

Figure 2.8: Bone invasive OSCC cells induce RANKL expression and suppress OPG expression in murine preosteoblasts

A. RANKL expression was not detected in unstimulated MC3T3 cells, but was induced by 10 µM PTHrP 1-86. SCCF2-conditioned medium and UMSCC12-conditioned medium induced RANKL mRNA expression in MC3T3 cells. SCCF3-conditioned medium did not induce MC3T3 expression. Statistical comparison was not performed because there was no expression in control and SCCF3-CM-treated cells. B. OPG expression was inhibited by 10 µM PTHrP 1-86 and UMSCC12-conditioned medium

(*P<0.05, ANOVA and Bonferroni post hoc test). SCCF3-conditioned medium stimulated MC3T3-expression of OPG (*P<0.05, ANOVA and Bonferroni post hoc test).

SCCF2-CM had no effect of OPG expression compared to untreated cells.

* *

* * * *

Figure 2.9: Bone conditioned medium stimulated proliferation of OSCC cells and increased OSCC-expression of PTHrP

A. An MTT assay was used to demonstrate that bone-conditioned medium stimulated

OSCC cell proliferation (*P<0.05, compared to unconditioned medium or fibroblast conditioned medium, $P<0.05 compared to unconditioned medium, ANOVA and

Bonferroni’s post-hoc-test). Bone or fibroblast-conditioned medium did not stimulate proliferation of human dermal fibroblasts (NHDF). B. Bone-conditioned medium stimulated OSCC-expression of PTHrP mRNA compared to unconditioned medium

(*P<0.05, ANOVA and Bonferroni’s post-hoc-test). There was no effect of MC3T3- conditioned medium on OSCC-expression of PTHrP. TGF-β1 (12 ng/ml) stimulated

PTHrP expression in SCCF2 cells (*P<0.05, ANOVA and Bonferroni’s post-hoc-test) C.

Bone-conditioned medium stimulated OSCC-secretion of PTHrP into culture medium compared to control medium (*P<0.05, 2-tailed t-test).

Chapter 3 : Zoledronic Acid Reduced Bone Loss and Tumor Growth in an Orthotopic Xenograft Model of Osteolytic Oral Squamous Cell Carcinoma

Abstract

Squamous cell carcinoma is the most common form of oral cancer. Destruction and invasion of mandibular and maxillary bone frequently occurs and contributes to morbidity and mortality. We hypothesized that the bisphosphonate drug, zoledronic acid

(ZOL), would inhibit tumor-induced osteolysis and reduce tumor growth and invasion in a murine xenograft model of bone-invasive oral squamous cell carcinoma (OSCC) derived from an osteolytic feline OSCC. Luciferase-expressing OSCC cells (SCCF2Luc) were injected into the perimaxillary subgingiva of nude mice which were then treated with 100 µg/kg ZOL or vehicle. ZOL treatment reduced tumor growth and prevented loss of bone volume and surface area, but had no effect on tumor invasion. Effects on bone were associated with reduced osteolysis and increased periosteal new bone formation.

ZOL-mediated inhibition of tumor-induced osteolysis was characterized by reduced numbers of tartrate-resistant acid phosphatase-positive osteoclasts at the tumor-bone interface, where it was associated with osteoclast vacuolar degeneration. The ratio of eroded to total bone surface was not affected by treatment, arguing that ZOL-mediated 78

inhibition of osteolysis was independent of effects on osteoclast activation or initiation of

bone resorption. In summary, our results establish that ZOL can reduce OSCC-induced

osteolysis and may be valuable as an adjuvant therapy in OSCC to preserve mandibular

and maxillary bone volume and function.

Introduction

Oral and oropharyngeal cancer is the 8th most common cancer world-wide (37).

Approximately 90% of oral and oropharyngeal tumors are squamous cell carcinoma

(OSCC) (39, 41). Despite advances in cancer treatment, the 5-year disease specific

survival for OSCC has shown only moderate improvement over the past 30 years, and is

currently 61% for all stages combined (42). OSCC frequently invades bone (44) and is

characterized by osteoclastic bone resorption (45). Bone invasion contributes to the

clinical morbidity of OSCC patients and is associated with poorer prognosis (46, 48).

Despite the frequency and clinical impact of bone invasion in OSCC, the mechanisms

responsible for bone resorption and invasion are poorly understood.

We hypothesize that OSCC invasion into bone is facilitated by a vicious cycle of tumor growth and bone resorption, and inhibition of the vicious cycle by targeting osteoclastic bone resorption with Zoledronic acid, a third generation bisphosphonate, will reduce bone loss and tumor growth. The vicious cycle theory was described in bone metastases of human breast cancer as a relationship between tumor-derived parathyroid hormone-

79 related protein (PTHrP) and bone-derived transforming growth factor beta 1 (TGF-β1)

(79). OSCC expresses several factors with the potential to stimulate osteolysis including

PTHrP (9, 35, 69), tumor necrosis factor-alpha (TNF-α) (70), interleukin-6 and interleukin-11 (9). Regardless of the number and type of bone resorption agonists expressed by OSCC cells, the physical destruction of bone is accomplished by bone- resorbing osteoclasts.

Osteoclastic bone resorption is inhibited by bisphosphonates. Nitrogen-containing bisphosphonates, such as zoledronic acid (ZOL), interfere with the mevalonate (MVA) pathway by inhibiting farnesyl pyrophosphate synthase (83). Inhibition of the MVA pathway leads to reduced prenylation of small guanosine-triphosphate (GTP)-binding proteins responsible for osteoclast function and survival (84) and leads to apoptosis (85).

We hypothesize that inhibition of osteoclastic bone resorption with ZOL will not only reduce bone loss but will also inhibit OSCC xenograft growth and invasion by antagonizing the vicious cycle of osteolysis and tumor growth.

We utilized a novel, bone invasive, bioluminescent orthotopic xenograft nude mouse model of OSCC using PTHrP-expressing cells derived from a bone invasive feline

OSCC. The inhibitory activity of ZOL on OSCC growth and tumor-associated osteolysis was investigated using a combination of in vivo bioluminescent imaging (BLI), Faxitron radiography, quantitative micro-computed tomography (microCT) and detailed maxillary histomorphometry. We found that ZOL may be an effective adjuvant treatment for preventing bone resorption associated with OSCC.

Materials and Methods

Cells and reagents

SCCF2 cells were derived in the Rosol laboratory in 2005 from a bone-invasive

gingival squamous cell carcinoma of a 7-year old male castrated cat using methods

previously described and were stably transfected with a plasmid containing a luciferase-

yellow fluorescent protein (YFP) fusion construct (86). SCCF2 cells were determined to

be of feline origin based on karyotype, and to be of epithelial origin based on pancytokeratin immunocytochemistry and the presence of desmosomes detected using electron microscopy. Zoledronic acid (Zometa; Novartis, East Hanover, NJ) was purchased from the James Cancer Center at The Ohio State University. ZOL vehicle was prepared according to the package insert and consisted of 44 mg/ml mannitol and 4.8 mg/ml sodium citrate in sterile water.

Animals and treatments

All animals were 6-week-old male nu/nu mice (NCI, Frederick, MD). Animal care

procedures were approved by the 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”. Sixty mice were assigned to 1 of 4

groups; nontumor-bearing vehicle-treated mice, tumor-bearing vehicle-treated mice,

nontumor-bearing ZOL-treated mice and tumor-bearing ZOL-treated mice.

General anesthesia of mice was induced and maintained with inhaled isoflurane. One million (1x106) SCCF2Luc cells in 0.1ml phosphate buffered saline (PBS; Invitrogen)

were injected through the maxillary gingiva into the perimaxillary gingival submucosa using a 1 ml syringe and 26-gauge needle (BD). Non-tumor-bearing mice were similarly

anesthetized and injected with cell-free PBS. Treatment was initiated 7 days following

injection of SCCF2Luc cells, and consisted of twice weekly subcutaneous injections of

ZOL at a dose of 100 µg/kg. Non-treated mice received twice weekly injections of

vehicle at the same volume. Two mice met early removal criteria and were removed from

the study (one ZOL-treated, tumor-bearing mouse and one vehicle-treated, non-tumor-

bearing mouse). Mice were euthanized by cervical dislocation during isoflurane anesthesia on the 28th day of treatment, and blood was collected by cardiac-puncture for

determination of plasma PTHrP and total calcium concentrations. Tissues were fixed in

4°C 10% neutral buffered formalin for 48 hours, followed by storage in 70% ethanol at

4°C until histologic processing and evaluation.

Bioluminescent imaging

In vivo BLI was performed with the IVISTM 100 system as previously described (18).

Results were analyzed using LivingImage® software, version 2.2 (Caliper Life Sciences).

Mice were injected intraperitoneally with 4.3 mg D-Luciferin in sterile PBS, and imaged

while under isoflurane anesthesia. A circular region of interest (ROI) surrounding the

xenograft was used to determine the radiance (photons/second) for each xenograft. Peak

ROI radiance was used for comparisons. ROI radiance was normalized by dividing radiance at day 7, 21 and 28 by the radiance at day 1.

Faxitron radiography and micro-computed tomography

The mandible was removed from each skull and the degree of bone loss was evaluated qualitatively using a Faxitron cabinet X-ray system (Hewlett-Packard, McMinnville, OR) at 45 kVp for 3.5 min. Bone loss was evaluated quantitatively using microCT (Siemens

Inveon Preclinical CT scanner and Inveon Research Workplace 3-Dimensional Image

Software, Siemens AG, Munich, Germany). Images were acquired in 300 exposures over

360 degrees, at 40 KVp, 500MA, 490 millisecond exposure, Bin 2, and a medium-high system magnification with a pixel width of 20.3 µm. Image data was reconstructed using

Cobra software (Exxim, Pleasanton CA) and analyzed using 3D analysis software

(Inveon Research Workplace 3-Dimensional Image Software, Siemens).

An ROI consisting of 100 consecutive image slices of the rostral skull (total thickness of 2.03 mm) was selected. The rostral margin of the ROI was defined by the rostral commissure of the palatine fossae and extended caudally. The ROI included the region of xenograft growth and incorporated pre-maxillary and maxillary bone. Intensity thresholds

(Hounsfield units) for extracting bone and teeth from surrounding soft tissue were determined by visually assessing the quality of segmentation for representative mice and corresponded to voxel intensity equal to or greater than 1000 Hounsfield units.

Segmentation thresholds were kept constant for all mice. ROI bone surface area and bone

volume were determined with the analysis software and compared between treatment

groups.

Histopathology, TRAP histochemistry and histomorphometry

The skulls were decalcified in 10% EDTA pH 7.4 at 4°C for 14 days. Rostral skull with xenograft tumor was paraffin embedded, sectioned at 5µm and stained with hematoxylin and eosin (HE) followed by microscopic evaluation. Invasiveness was determined by visually identifying tumor cells at the level of the periodontal ligament of the maxillary incisor beneath nasal respiratory epithelium. Slides were scanned using the Aperio

ScanScope slide scanner (Aperio, Vista, CA). Two-dimensional tumor area was determined using morphometry tools in Imagescope software (Aperio). The degree of maxillary bone loss was determined by comparing maxillary bone area between the nontumor-bearing side and the tumor-bearing side for vehicle and ZOL-treated mice.

Maxillary bone was classified as either pre-existing (mature) bone or tumor-induced new bone (immature) based on collagen pattern (woven or lamellar), osteocyte density, and anatomic location.

Enzymatic histochemistry for tartrate-resistant acid phosphatase (TRAP, Sigma-

Aldrich) was performed on tissue sections as previously described (60). Bone histomorphometry was completed with the Imagescope software. The average percent perimeter of eroded bone, number of active osteoclasts, and average osteoclast length were compared between treated and untreated tumor-bearing mice. Only sections with a

minimum of 3 mm of direct tumor-bone interface were included (13 vehicle-treated mice

and 12 ZOL-treated mice).

PTHrP and Calcium concentration assays

Mouse plasma concentrations of PTHrP were measured using a commercially available

two-site immunoradiometric PTHrP (1-86) assay (Diagnostic Systems Laboratories Inc.,

Webster, TX, USA). Plasma calcium concentration was measured using a commercially

available calcium assay kit (Bioassay Systems, Hayward, CA).

Statistical analysis

Results were displayed as means ± standard error. Normality was determined with the

Shapiro Wilk test. Normally distributed, continuous data were analyzed using Student’s t-

test and multiple group comparisons were made by one way ANOVA followed by

Bonferroni post hoc test. In the event the data were not normally distributed, a Wilcoxon

rank sum (Mann-Whitney) test was performed. Data with P values less than 0.05 were

considered statistically significant. When multiple comparisons were made within a data

set (maxillary histomorphometry), the P value of 0.05 divided by the number of

comparisons (7) was accepted as significant (P value of 0.007). Categorical data

(presence of invasion) was analyzed using Fishers exact test. All analyses were performed with STATA intercooled 10 software (Cary, NC).

Results

ZOL treatment reduced tumor growth

Bioluminescent signal was detectable in all mice at all time points, with no initial difference between vehicle and ZOL-treated groups. Tumor formation became visibly

evident as mild facial swelling at 7 to 14 days after injection of tumor cells. Gross tumor

formation occurred in all mice, which was located at the junction of maxillary and

premaxillary bone.

At the 28th day of treatment, ZOL treatment reduced tumor radiance by 43% (P=0.0097,

2-tailed t test, figure 3.1). Histomorphometric evaluation of tumor sections revealed that

ZOL treatment reduced tumor area (total area minus area of central necrosis) by 14%

(P=0.075 2-tailed t-test, data not shown). The greater difference in tumor growth detected

with BLI was attributed to the higher specificity of BLI for viable tumors cells with

exclusion of necrotic areas and stromal elements, in addition to normalization to initial

radiance measurements which corrected for variation in the number of injected tumor

cells.

Zoledronic acid reduced loss of bone surface area and volume

Faxitron radiography revealed reduced bone loss and increased periosteal new bone

formation in the region of xenograft growth in the ZOL-treated mice compared to

vehicle-treated mice (figure 3.2A). Reconstructed, 3-dimensional skull images (figure

3.2B) revealed that ZOL-treatment reduced, but did not eliminate loss of bone at the

86 tumor-bone interface, and was accompanied by extensive bone remodeling with new bone formation. Quantitative determination of bone volume and surface area was performed on a 2mm thick maxillary ROI (figure 3.2C). Xenograft growth in vehicle- treated mice reduced bone surface area by 14% (P<0.0001, ANOVA) and bone volume by 19% (P<0.0001, ANOVA) (figure 3.3). There was no statistically significant loss of bone surface area and volume in ZOL-treated, tumor-bearing mice compared to nontumor-bearing mice. ZOL treatment increased bone surface area and bone volume in tumor-bearing mice by 23% and 39% respectively (P<0.0001, ANOVA) compared to untreated tumor-bearing mice.

Zoledronic acid reduced microscopic evidence of maxillary bone loss and induced osteoclast degeneration but did not reduce invasion

SCCF2Luc xenografts were composed of islands and trabeculae of malignant epithelial cells demonstrating variable degrees of squamous differentiation and keratinization with regions of central necrosis. There was no difference in xenograft cellular morphology between vehicle and ZOL-treated mice. SCCF2Luc xenografts were poorly vascularized, however; vessels appeared slightly more numerous and larger at the tumor-bone interface and were associated with larger aggregates of tumor cells. This pattern was observed in mice of both vehicle and ZOL-treated groups.

SCCF2Luc xenografts were associated with marked osteolysis of premaxillary and maxillary bone (figure 3.4A and B). Zoledronic acid inhibited osteoclast formation and bone loss in vivo (figure 3.4A-D). Despite reduced bone loss, ZOL had no effect on

tumor invasion (figure 3.4C and D). Tumor invasion around the maxillary incisor was

observed in 13 of 15 vehicle-treated mice and 12 of 14 ZOL-treated mice. Tumor invasion into the nasal passage was observed in 12 of 15 vehicle-treated mice and in 10

of 14 ZOL-treated mice.

Nine (9) of 14 tumor-bearing, ZOL-treated mice had foci of bone necrosis within

xenografts, and 6 of these mice also demonstrated necrosis of the maxillary incisor pulp

cavity. Bone necrosis was often located in regions of tumor necrosis and only occurred in

mice with pronounced tumor invasion. Bone and tooth necrosis was not observed on the

non-tumor-bearing side of the maxilla, and was not observed in ZOL-treated, non-tumor-

bearing mice. No vehicle-treated mice had evidence of tooth necrosis, although 1 of 15

vehicle-treated tumor-bearing mice had evidence of bone necrosis.

TRAP-positive osteoclasts were observed at the tumor-bone interface in treated and

untreated mice; however, osteoclasts in ZOL-treated mice (figure 3.4D) had variable

degrees of cytoplasmic vacuolar degeneration (numerous, variably sized, circular, clear

cytoplasmic vacuoles) compared to those in vehicle-treated mice (figure 3.4C).

Zoledronic acid reduced loss of maxillary bone area and resulted in fewer osteoclasts

Maxillary bone area was determined using Imagescope software (figure 3.5A). There was no difference in total bone area on the non-tumor-bearing side of vehicle and ZOL- treated mice. Xenograft growth was associated with a 48% reduction in total bone area

(pre-existing bone and new bone combined) compared to the non-tumor-bearing side

(P=0.0002, Wilcoxon rank-sum), whereas ZOL treatment prevented loss of total bone

area. Xenograft growth was associated with a 64% reduction in pre-existing bone

(P=0.0003, Wilcoxon rank-sum) in vehicle treated mice, compared to a 33% reduction in pre-existing bone in ZOL-treated mice (P=0.0002 2-tailed t-test). ZOL-treated mice retained 2-fold more total bone compared to vehicle-treated mice (P=0.0004, Wilcoxon rank-sum), and 2-fold more pre-existing bone compared to vehicle treated mice (P=0.005,

Wilcoxon rank-sum). P values < 0.007 were considered statistically significant (adjusted

for multiple comparisons). The preservation of total bone in ZOL-treated mice, as

detected by microCT and histomorphometry, was attributed to reduced loss of pre-

existing bone and increased new bone formation.

Activated osteoclasts at the xenograft-bone interface were detected using TRAP

histochemistry. There was no statistically significant difference in the ratio of eroded to

total bone surface between ZOL and vehicle-treated mice (figure 3.5B); however, ZOL

reduced the number of osteoclasts per millimeter of bone surface by 52% (P=0.0003 2-

tailed t-test). The average osteoclast length was increased in ZOL-treated mice by 39%

(P=0.0001, Wilcoxon rank-sum).

Plasma PTHrP was increased in tumor-bearing mice

SCCF2Luc xenograft growth resulted in elevated plasma PTHrP in vivo (figure 3.6).

Tumor-bearing mice (n=12) had a 2.7-fold increase (P=0.037, ANOVA) in plasma

PTHrP (0.65 pM vs. 2.42 pM) compared to non-tumor-bearing control mice (n=5). ZOL treatment resulted in a non-statistically significant, 20% reduction in plasma PTHrP in

89 tumor-bearing mice. Elevated plasma PTHrP in tumor-bearing mice was not associated with hypercalcemia (figure 3.6).

Discussion

There are few preclinical, in vivo models that recapitulate the bone invasive behavior of

OSCC. In vivo OSCC studies have been reported using 2 osteolytic OSCC cell lines;

BHY, derived from a human gingival OSCC (87), and SCCVII, derived from a

C(3)H/HeN mouse OSCC from ventral mouth (88, 89). Other cell lines demonstrating variable bone invasion in vivo include UMSCC1 (61), UMSCC11A, UMSCC11B (62),

BICR31, BICR56 (63), and HSC3 (35). We have designed a novel, bone invasive, bioluminescent orthotopic xenograft nude mouse model of OSCC using SCCF2Luc cells derived from a feline OSCC. SCCF2Luc cells reliably develop into invasive xenografts without requiring co-injection of fibroblasts (required for BHY cells) (87) or Matrigel

(required for UMSCC11A cells) (62). Tumor progression in the SCCF2Luc model are typical of spontaneous bone invasive OSCC, demonstrating rapid tumor growth with bone invasion characterized by osteoclastic bone resorption.

ZOL treatment reduced tumor growth and bone loss in the SCCF2Luc xenograft model, and was associated with a reduced number of activated osteoclasts at the tumor-bone interface. These results are similar to those reported by Cui et al (2005), who demonstrated that a third generation bisphosphonate (YM529, Minodronic acid) reduced

osteoclast number and tumor growth in the SCCVII syngeneic mouse model of bone

invasive OSCC (88). In contrast to our study, Cui et al. injected SCCVII cells

percutaneously into the masseter muscle adjacent to the mandible rather than

orthotopically into the oral submucosa, and they did not quantify the effects of YM529

treatment on tumor-induced bone loss. The addition of complete histomorphometric

characterization of bone loss in our study not only corroborated the subjective and

objective findings of the radiographic and microCT examination, but demonstrated the

effect of ZOL-treatment on pre-existing bone independent of periosteal new bone

formation.

There are several possible mechanisms for reduced osteoclastic bone resorption in the

ZOL-treated mice. ZOL inhibits osteoclastic bone resorption by inducing osteoclast

apoptosis through impaired prenylation of small GTPases as a result of farnesyl

transferase inhibition (90), accumulation of unprenylated small GTPases in their active

state with inappropriate activation of downstream signaling pathways, and the induction

of an ATP analogue which induces direct apoptosis (91). Additionally, ZOL has also been shown to increase the in vitro expression of the osteoclast inhibitor, osteoprotegerin

(OPG), in human osteoblasts (92). Our study revealed that osteoclasts in ZOL-treated

mice continued to form and maintained their ability to resorb bone, but existed in reduced

numbers and demonstrated evidence of vacuolar degeneration. The mechanism of the

vacuolar degeneration is not known; however, a study by Coxon et al. (2003) demonstrated that targeted inhibition of the small GTPase Rab resulted in reduced osteoclast function accompanied by dome-shaped cell morphology and large intracellular

91 vacuoles which was attributed to disruption of Rab-dependent intracellular membrane trafficking in osteoclasts (93). In our study, accumulation of numerous cytoplasmic vacuoles may have contributed to the increased osteoclast diameter (length) observed in the ZOL-treated mice.

ZOL significantly reduced, but did not completely inhibit, bone resorption in this model of bone-invasive OSCC. Interestingly; tumor-bearing mice continued to lose pre-existing bone during ZOL therapy, albeit at a reduced level, and demonstrated increased periosteal new bone formation. This may be explained by differential affinity of ZOL for different bone surfaces. ZOL is reported to be incorporated more readily into bone as it is being formed by osteoblasts, and into bone exposed by osteoclast activity, compared to quiescent bone surfaces (90). Periosteal new bone formation would be relatively spared from bone resorption because of the higher levels of ZOL expected to be incorporated into the matrix as it is formed (90). ZOL treatment did not totally eliminate the presence of osteoclasts; therefore, it appears that osteoclasts in the tumor-bone microenvironment are at least partially protected from the pro-apoptotic effects of ZOL. Interestingly, a recent in vitro study revealed that a human OSCC cell line was capable of reducing osteoclast apoptosis through downregulation of the pro-apoptotic factor Bim in osteoclasts (71).

ZOL has been shown to inhibit tumor growth in other models through direct antineoplastic mechanisms which include inhibition of cancer cell adhesion, invasion, viability and angiogenesis (94, 95). For example, the antiangiogenic activity of bisphosphonates is associated with reduced circulating levels of vascular endothelial

growth factor (VEGF) in humans with metastatic bone disease (96), and in cats with

OSCC (97). There was no histologic evidence of an antiangiogenic effect ZOL treatment in this study. Although pro-apoptotic activity of ZOL in tumor cells has been demonstrated in human cell lines (98, 99), microscopic evaluation of ZOL-treated

SCCF2Luc tumors did not reveal evidence of increased necrosis or apoptosis. The complex pharmacokinetics of zoledronic acid in the bone microenvironment pose significant challenges to in vitro studies attempting to model the exposure of tumor cells to clinically relevant concentrations of ZOL and make comparisons of in vitro and in vivo cytotoxicity difficult.

It is likely that ZOL-mediated inhibition of osteoclastic bone resorption contributed to reduced xenograft growth in this model. Culture of SCCF2 cells in bone-conditioned medium stimulated PTHrP secretion in vitro, and SCCF2-conditioned medium induced murine preosteoblast expression of RANKL (86). ZOL demonstrated and in vivo

inhibition of SCCF2-associated bone resorption. Interestingly, ZOL treatment reduced

the average plasma PTHrP concentration in tumor-bearing mice by 20%, although

statistical significance was not reached. Taken together, the data supports the role of

tumor-derived PTHrP in a vicious cycle of OSCC growth and osteoclastic bone

resorption. Bone-derived factors that may have stimulated SCCF2-expression of PTHrP

include transforming growth factor β1 (TGF-β1), fibroblast growth factor (FGF), insulin-

like growth factors (IGFs) I and II, platelet derived growth factor (PDGF), and bone

morphogenic proteins (BMP) (78). By reducing osteoclastic bone resorption, ZOL has

93 the potential to reduce the liberation of stored growth factors making a less favorable environment for the growth of tumor cells (92).

OSCC-expression of PTHrP influences in vitro tumor cell proliferation, migration and invasiveness (76), and participates in the regulation of osteoclastogenesis by increasing stromal cell expression of RANKL (35). A small subset of OSCC patients experienced hypercalcemia with increased serum PTHrP concentration (73). Plasma PTHrP concentrations were increased in mice bearing SCCF2Luc xenografts. Although there was no evidence of a humoral role of PTHrP in this model (lack of hypercalcemia), it is likely that tumor-derived PTHrP in the xenograft-bone microenvironment functioned in a paracrine manner to stimulate osteoclastic bone resorption by increasing RANKL expression in osteoblasts.

Osteonecrosis of the jaw (ONJ) is occasionally observed in cancer patients receiving aminobisphosphonate therapy (estimated incidence of 5% to 10%) (100) and is characterized by exposure of mandibular or maxillary bone which fails to heal over a period of six to eight weeks (101) In this study, ZOL was well tolerated in both the tumor-bearing and non-tumor-bearing mice and there was no evidence of ONJ. The presence of tumor-associated necrotic bone within xenografts of ZOL-treated mice was attributed to the retention of bone undergoing necrosis as a result of tumor infiltration.

Bone necrosis would be expected in the vehicle treated mice as well, however the robust osteoclastic response in vehicle-treated mice would have remove necrotic bone as it developed. The cause of tooth necrosis in ZOL-treated mice was not evident, but appeared as an extension of bone necrosis and occurred in mice that demonstrated tumor

infiltration near the base of the tooth root. It is possible that OSCC patients with existing bone disease of the mandible or maxilla may be at increased risk for the development of

ONJ; however, the potential benefits of protecting bone volume and function in cases of non-resectable, bone invasive cancer may outweigh the risks in some patients.

ZOL inhibited OSCC-induced osteolysis and reduced tumor growth in this xenograft mouse model; however, the presence of invasion was unchanged. The results of this experiment suggest that ZOL would be valuable as an adjunct therapy with the purpose of maintaining mandibular and maxillary bone volume and function, particularly if combined with antineoplastic agents capable of reducing tumor invasion. Further studies are required to characterize the factors released from bone which stimulate OSCC growth and PTHrP secretion, and to determine if ZOL may provide an additive or synergistic antitumor effect when combined with other therapeutic modalities for the treatment of bone invasive OSCC.

Figure 3.1: ZOL treatment reduced SCCF2Luc xenograft growth

Mice were imaged at day 1,7,21 and 28 of treatment. On day 28, the relative bioluminescence was reduced 43% in ZOL-treated mice compared to vehicle treated mice (*P=0.0097).

Figure 3.2: ZOL inhibited SCCF2Luc xenograft-associated osteolysis

A, Faxitron radiography of maxilla from 3 representative mice from each treatment group

(xenograft margins = arrowheads, osteolysis = arrows, new bone formation = asterisks).

ZOL reduced osteolysis and increased periosteal new bone formation compared to vehicle-treated mice. B, Three-dimensional renderings of microCT data. Normal premaxillary bone (black arrow) and maxillary bone (white arrow) are indicated on the

Sham + Vehicle mouse. SCCF2Luc xenografts in vehicle-treated mice were associated with marked loss of lateral pre-maxillary and maxillary bone revealing the root of the maxillary incisor (white arrows). ZOL reduced osteolysis of maxillary and premaxillary bone (minimal exposure of incisor root, black circle) and increased periosteal new bone 97 formation. C, A 2mm thick region of interest (ROI) was selected (grey band) allowing bone surface area and volume to be calculated and compared between treatment groups.

Figure 3.3: ZOL prevented loss of bone surface area and bone volume in SCC2Luc

xenografts

SCCF2Luc xenograft growth resulted in a significant reduction in bone surface area

(14%, *P<0.0001) and bone volume (19%, *P<0.0001) compared to non-tumor-bearing mice. ZOL prevented loss of bone surface area and bone volume. ZOL treatment increased bone surface area and bone volume in tumor-bearing mice by 23% and 39% respectively (P<0.0001) compared to untreated tumor-bearing mice.

Figure 3.4: ZOL reduced SCCF2Luc-associated osteolysis and induced vacuolar degeneration of osteoclasts but did not reduce invasion

Photomicrographs of H&E-stained tissue sections. A, SCCF2Luc xenografts were associated with marked osteolysis of premaxillary and maxillary bone (white asterisks)

100 along the infiltrating tumor (black arrows, bar = 500µm). Higher magnification from

Panel A (box), illustrating islands of squamous carcinoma cells (arrows) infiltrating the remaining bone (white asterisk, bar = 100µm). B, Treatment with ZOL reduced resorption of premaxillary and maxillary bone (black asterisks, Bar = 500µm). Despite reduced bone resorption, higher magnification (box) revealed islands of squamous carcinoma cells infiltrating the bone and the periodontal ligament (black arrows, bar =

100µm). TRAP-positive osteoclasts were observed at the tumor-bone interface of bone vehicle (C) and ZOL (D) treated mice. Osteoclasts were multinucleated (small black arrows), and in ZOL-treated mice (D) were characterized by vacuolar degeneration (large black arrows). Islands of FOSCC are indicated by ‘T’. (Bar = 50µm)

101

Figure 3.5: ZOL-mediated reduction in bone loss was characterized by reduced loss of

pre-existing bone and reduced number of bone-resorbing osteoclasts at the tumor-bone

interface

A. SCCF2Luc xenografts were associated with a 48% reduction in total bone area (pre-

existing bone and new bone combined) compared to the non-tumor-bearing side

(*P=0.0002). ZOL prevented loss of total bone area. Vehicle-treated mice lost 64% of

pre-existing bone area (*P=0.0003) compared to 33% in ZOL-treated mice (*P=0.0002).

ZOL-treated mice retained 2-fold more total bone compared to vehicle-treated mice

(*P=0.0004, Wilcoxon rank-sum), and 2-fold more pre-existing bone compared to vehicle treated mice (*P=0.005). B. Histomorphometric evaluation of TRAP-stained 102 sections revealed no difference in the percentage of eroded bone between ZOL- and vehicle-treated mice; however, ZOL reduced the number of osteoclasts per millimeter by

52% (*P=0.0003). The average osteoclast length was increased in the ZOL-treated mice by 39% (*P=0.0001).

103

Figure 3.6: PTHrP was increased in mice bearing SCCF2Luc xenografts but was not associated with hypercalcemia

A. Plasma PTHrP was increased in tumor-bearing mice (2.42 pM) compared to control mice (0.65pM, *P=0.037); however, ZOL treatment did not significantly reduce plasma

PTHrP concentrations (NS). Xenograft growth and ZOL-treatment had no effect on plasma calcium compared to control mice. B. SCCF2 cells secreted PTHrP into the culture medium (N=3). Culture of SCCF2 cells in bone-conditioned medium increased

PTHrP secretion 2.6-fold. *P<0.05, Student’s t-test.

104

Chapter 4 : Combined Zoledronic Acid and Meloxicam Reduced Bone Loss and Tumor Growth in an Orthotopic Mouse Model of Bone Invasive Oral Squamous Cell carcinoma

Abstract

Squamous cell carcinoma is the most common form of oral cancer. Destruction and invasion of mandibular and maxillary bone occurs frequently and contributes to morbidity and mortality. We hypothesized that combining a preferential cyclooxygenase-

2 inhibitor, meloxicam, with a bisphosphonate drug, zoledronic acid (ZOL), would inhibit tumor-induced osteolysis and tumor growth and invasion in a murine xenograft model of bone-invasive oral squamous cell carcinoma (OSCC) derived from an osteolytic feline OSCC. Luciferase-expressing OSCC cells (SCCF2Luc) were injected into the perimaxillary gingival submucosa of nude mice, which were treated with 100 µg/kg ZOL twice weekly, 0.3 mg/kg meloxicam daily, combined ZOL and meloxicam, or vehicle.

Meloxicam alone and meloxicam + ZOL reduced tumor growth. ZOL alone and meloxicam + ZOL inhibited osteoclastic bone resorption and maxillary bone loss.

Inhibition of tumor-induced osteolysis in ZOL and ZOL + meloxicam treated mice was characterized by reduced numbers of tartrate-resistant acid phosphatase-positive osteoclasts at the tumor-bone interface. ZOL also induced vacuolar degeneration. There 105 was a trend for reduced loss of pre-existing bone, reduced eroded surfaces, and larger osteoclasts in meloxicam-treated mice. The combination of meloxicam and ZOL was well tolerated but did not stimulate additional tumor suppression or inhibition of bone loss compared to meloxicam- or ZOL-monotherapy. The combination of meloxicam and

ZOL may be valuable as adjuvant therapy for OSCC.

Introduction

Bone invasion associated with oral squamous cell carcinoma (OSCC) frequently occurs and negatively impacts prognosis (46-51). The objective of this study was to evaluate the effectiveness of targeting the interaction between tumor growth and bone destruction by combining a bone resorption inhibitor (zoledronic acid) with a cyclooxygenase inhibitor

(meloxicam).

OSCC produces cytokines that can serve as agonists of osteoclastic bone resorption including prostaglandin E2 (PGE2) (102, 103), parathyroid hormone related-protein

(PTHrP), tumor necrosis factor-alpha (TNF-α), interleukin-6 and interleukin-11 (9, 87,

88). High levels of PGE2 in OSCC are due to up-regulation of COX-2 (104, 105).

Tumor-associated bone resorption releases growth factors such as TGF-β1 (79), which have been shown stimulate tumor cell secretion of PGE2 through increased expression of

COX-2 (106). Up-regulation of COX-2 causes increased synthesis of prostaglandins and is known to increase cell proliferation (107), promote angiogenesis (108), inhibit immune

106

surveillance (109), inhibit apoptosis (110) and enhance tumor invasiveness (111). PGE2 stimulates osteoclastogenesis through up-regulation of receptor activator of nuclear factor

κB ligand (RANKL) expression in osteoblasts and bone marrow stromal cells (112-114),

in addition to inhibition of OPG secretion and upregulation of receptor activator of

nuclear factor κB (RANK) expression in osteoclasts (115).

Bisphosphonates prevent bone loss by antagonizing the mevalonate pathway resulting

in osteoclast inhibition (116). Zoledronic acid (ZOL, a third generation

aminobisphosphonate) reduced tumor growth and bone invasion in a model of bone

invasive OSCC utilizing a bone invasive feline OSCC cell line (117). The ability of COX

inhibitors to reduce tumor growth has been documented in rodent models of human

OSCC (118-120), lung adenocarcinoma (121), bladder transitional cell carcinoma (122),

mammary carcinoma (123), and carcinogenesis of tongue SCC in rats (124). There are no

published reports of a bisphosphonate being used in any form of combination therapy for

bone invasive OSCC. The hypothesis of this study was that the combination of

zoledronic acid with a COX-2 inhibitor (meloxicam) will reduce bone resorption and

tumor growth in an orthotopic model of bone invasive OSCC in nude mice.

We have designed a bone invasive, bioluminescent orthotopic xenograft nude mouse

model of OSCC using cells derived from a bone invasive feline OSCC. As in humans,

OSCC is the most commonly diagnosed tumor of the oral cavity in cats (1, 2), and has a

highly invasive, osteolytic phenotype (55) with similarities in clinical progression and

pathology compared to human OSCC (55).

107

We used real-time reverse transcriptase polymerase chain reaction (real-time RT-PCR)

to measure expression of COX-1 and COX-2 in a panel of feline and human OSCC cell

lines, and determined that culture of SCCF2 cells (feline bone invasive OSCC) in bone

conditioned medium stimulated SCCF2-expression of COX-2. The effect of ZOL and

meloxicam therapy on OSCC growth and tumor-associated osteolysis was investigated

using bioluminescent imaging, Faxitron radiography, micro-computed tomography and

maxillary histomorphometry. We found that combination therapy of ZOL and meloxicam

reduced tumor growth, invasion, and bone loss and may be an effective adjuvant

treatment for bone invasive OSCC.

Materials and Methods

Cells and reagents

Feline OSCC cell lines (SCCF1, laryngeal SCC; SCCF2, gingival SCC and SCCF3,

lingual SCC) and feline OSCC tumor-associated fibroblasts (TAF) were derived as previously described (86). Human cell lines (A253, salivary squamous cell carcinoma;

SCC25, lingual squamous cell carcinoma) and murine preosteoblast cells (MC3T3) were purchased from ATCC (Manassas, VA). The human laryngeal squamous cell carcinoma cell line, UMSCC12, was kindly provided by Dr. Thomas Carey at the University of

Michigan. Cells were maintained in growth medium consisting of high glucose DMEM supplemented with 10% fetal bovine serum (FBS), 2mM L-Glutamine (Invitrogen,

108

Calisbad, CA), and Normocin (Invivogen, San Diego, CA). Zoledronic acid (Zometa;

Novartis, East Hanover, NJ) was purchased from the James Cancer Center at The Ohio

State University. Meloxicam (Metacam; Boehringer Ingelheim, Ridgefield, CT) was

purchased from the Veterinary Medical Center at The Ohio State University. SCCF2 cells

were previously transfected with a luciferase-yellow fluorescent protein (YFP) fusion

construct (pCDNA3.1(+).yLuc-YFP, kindly provided by Christopher Contag (Stanford

University, Stanford, CA), as previously described (86).

Real-time RT-PCR

OSCC cells (UMSCC12, A253, SCCC25, SCCF1, SCCF2 and SCCF3) were evaluated

using real-time reverse transcriptase polymerase chain reaction for expression of COX-1

and COX-2 mRNA relative to β-2 microglobulin mRNA. RNA extraction, reverse

transcription, PCR, and real-time RT-PCR were performed as previously described (59).

The feline primers for real-time RT-PCR included β2-microglobulin (B2M) (forward:

CTACTTCTGGCGCTGCTCTG and reverse: CCTGAACCTTTGGAGAATGC), COX-

1 (forward: CTGGGGTGATGAGCAACTCT and reverse:

GGAAGTAACCGCTCAACTGC) and COX-2 (forward:

AGGACTGGGCCATGGGGTGG and reverse: CTGGCCCACAGCAAACCGCA). The

human primers included COX-1 (forward: GAGCAGCTTTTCCAGACGAC and reverse:

GCAGGAAATAGCCACTCAGC) and COX-2 (forward:

GCAGTTGTTCCAGACAAGCA

and reverse: GCCACTCAAGTGTTGCACAT).

109

OSCC cells were seeded at 1x105 cells per well in six-well plates and grown for 48 hours in growth medium followed by overnight serum starvation and 6-hour stimulation with serum-containing growth medium. Gene expression was expressed as relative expression compared to the cell line with lowest expression.

To determine if bone-conditioned medium contained factors that stimulate COX-2 expression, SCCF2 cells were cultured in serum-free, MC3T3-conditioned medium

(murine preosteoblasts), or murine bone-conditioned medium. Bone-conditioned medium was prepared by culturing 1 calvarium per 2 ml of serum-free medium (high glucose

DMEM). Medium was harvested every 96 hours for 8 days. MC3T3-conditioned medium was prepared from confluent cultures grown in serum free high-glucose DMEM every 96 hours for 8 days. Conditioned medium was filter sterilized and stored at -30°C until use.

Unconditioned medium was prepared as conditioned medium except there were no cells or bone in the flasks. SCCF2 cells were seeded in 6-well plates at a density of 1x105 cells per plate in growth medium for 48 hours followed by serum-free medium for 24 hours.

Medium was replaced with 50% unconditioned serum-free medium and 50% conditioned serum-free medium for 3 hours. RNA extraction, reverse transcriptase PCR and real-time

PCR was performed as described above.

Effect of bone-conditioned medium on OSCC production of PGE2

OSCC cell lines were cultured in serum-free bone-conditioned medium or unconditioned medium. Bone-conditioned medium was prepared as described above.

110

SCCF1, SCCF2, SCCF3, UMSCC12 and SCC25 cells were seeded in 6-well plates at a density of 1x105 cells per plate in growth medium for 48 hours followed by serum-free medium for 24 hours. Medium was replaced with 50% unconditioned serum-free medium and 50% conditioned serum-free medium for 24 hours. The concentration of PGE2 in the conditioned medium was measured using a commercially available PGE2 ELISA

(Parameter PGE2 kit, R&D Systems) according to the manufacturer’s instructions. The lower limit of PGE2 detection was 82.5 pg/ml.

Effect of cyclooxygenase inhibition on OSCC proliferation

SCCF2 cells were seeded into 96-well culture plates at 3x103 cells / well and cultured overnight in growth medium. Medium was replaced with growth medium containing various concentrations (1 to 1000 µM) of indomethacin (Sigma-Aldrich, St. Louis, MO), piroxicam (Sigma-Aldrich) or meloxicam (Metacam, Boehringer Ingelheim) for 72 hours and a commercially available MTT assay was performed according to the manufacturer’s instructions (Promega, Madison, WI). The drug concentration which resulted in a 50% reduction in viability (IC50) was calculated from graphs of viability (%) plotted against concentration of drug.

Animals and treatments

All animals were 6-week-old male nu/nu mice (NCI, Frederick, MD). Mice were housed in micro-isolator cages and provided with food pellets and water ad libitum.

Animal care procedures were approved by the Institutional Lab Animal Care and Use 111

Committee using criteria based on both the Animal Welfare Act and the Public Health

Services “Guide for the Care and Use of Laboratory Animals”. Mice were systematically randomized to 1 of 6 groups; nontumor-bearing vehicle-treated mice, nontumor-bearing

ZOL-treated mice, tumor-bearing vehicle-treated mice, tumor-bearing meloxicam-treated mice, tumor-bearing ZOL-treated mice, and tumor-bearing combination-treated mice.

One mouse died of undetermined cause within 24 hours of injection, 2 mice were removed because of significant leakage of tumor suspension, 3 mice were removed because of low initial signal (ROI radiance < 1x107), and one mouse was removed because it was a high BLI outlier at day 0 (Grubb’s test). In order to evaluate the effect of treatment on OSCC-associated bone resorption, all mice were microscopically evaluated at the end of the study to confirm contact between the xenograft and the maxilla. Mice with xenografts demonstrating minimal-to-no contact with the maxilla were removed from the study and not included in the analyses.

The number of mice included in each group was 5 nontumor-bearing, untreated mice; 5 nontumor-bearing, ZOL-treated mice; 11 vehicle-treated, tumor-bearing mice; 11 meloxicam-treated, tumor-bearing mice; 12 ZOL-treated, tumor-bearing mice; and 10 combination-treated, tumor-bearing mice. The xenograft of 1 mouse completely regressed (meloxicam monotherapy group), and was not removed from analysis because there was evidence of prior contact with bone (resolving bone resorption and surface remodeling) and statistical analysis (Grubb’s test) did not identify the mouse as a bioluminescent outlier. One meloxicam-treated mouse met early removal criteria for

112

weight loss at day 21, but had tumor growth and bone invasion similar to other mice in the group and was kept in the analysis.

Mice were anesthetized with isoflurane and injected with 1x106 SCCF2Luc cells

suspended in 0.1 ml of sterile PBS and injected below the maxillary gingiva adjacent to

the maxillary incisors as previously described (117). Nontumor-bearing mice were

similarly anesthetized and injected with cell-free PBS. Treatment was initiated 3 days

following injection of SCCF2Luc cells, and consisted of twice weekly subcutaneous

injections of ZOL at 100 µg/kg diluted in 0.9% sodium chloride (saline). Meloxicam was

diluted in saline and administered by intraperitoneal (ip) injection at a dose of 0.3 mg/kg

once daily. Vehicle-treated mice received daily ip injections of saline at the same volume

as what was administered to treated mice. Mice were weighed weekly. Bioluminescent

images were collected at the onset of treatment and following 14 and 21 days of

treatment. Mice were euthanized by carbon dioxide inhalation and cervical dislocation

following 25 days of treatment (28 days of xenograft growth). Tissues were fixed in 4°C

10% buffered formalin for 48 hours, followed by storage in 4°C 70% ethanol until

routine histologic processing and evaluation.

Bioluminescent imaging

Mice were injected ip with 4.3 mg D-Luciferin (Caliper Life Sciences, Hopkinton, MA)

dissolved in sterile PBS, and imaged while under isoflurane anesthesia. In vivo

bioluminescent imaging was performed using an IVIS 100 system (Caliper Life Sciences)

and analyzed using LivingImage® software, version 2.2 (Xenogen Corporation) as

113

previously described (18, 117). Region of interest (ROI) radiance values were normalized

by dividing radiance values at day 28 by the radiance at the onset of treatment and are

expressed as the fold-change.

Faxitron radiography and micro-computed tomography

Five random mice from each tumor-bearing group, and all 5 mice in each of the

nontumor-bearing groups, were selected for Faxitron radiography and microcomputed

tomography. The mandible was removed from each skull and the degree of maxillary and

premaxillary bone loss was evaluated qualitatively using a Faxitron cabinet X-ray system

(Hewlett-Packard, McMinnville, OR) as previously described (117). Bone loss was

measured for 5 randomly selected mice from each tumor-bearing group and all 5 mice

from each nontumor-bearing mice, using micro-computed tomography (microCT)

(Siemens Inveon Preclinical CT scanner and Inveon Research Workplace 3-Dimensional

Image Software, Siemens AG, Munich, Germany).

Images were acquired in 400 exposures over 360 degrees, at 80 KVp, 500MA, 175

millisecond exposure, Bin 4 and a pixel width of 38.8 µm. Image data were reconstructed

using Cobra software (Exxim, Pleasanton CA) and analyzed using 3D analysis software

(Inveon Research Workplace 3-Dimensional Image Software, Siemens). A 2 mm thick

ROI was selected that extended caudally from the rostral commissure of the palatine

fossae and included the region of xenograft growth. Intensity thresholds for extracting

bone and teeth from surrounding soft tissue were determined by visually assessing the quality of segmentation and were kept constant for all mice. ROI bone volume was

114

determined with the analysis software and compared between treatment groups. Higher

resolution acquisitions were taken for figures.

Histopathology, TRAP histochemistry and histomorphometry

The skulls were decalcified in 10% EDTA pH 7.4 at 4°C for 14 days. The rostral skull was divided into 3mm thick frontal sections, paraffin embedded, sectioned and stained with hematoxylin and eosin (HE) followed by microscopic evaluation. The degree of invasiveness was determined by visually identifying tumor cells at the level of the periodontal ligament of the maxillary incisor, and within the nasal cavity (tumor cells observed immediately beneath nasal respiratory epithelium). HE-stained slides were scanned using the Aperio ScanScope slide scanner (Aperio, Vista CA). The degree of maxillary bone loss was measured by expressing bone area on the tumor-bearing side as a percentage of bone area on the nontumor-bearing side. Maxillary bone was classified as

either pre-existing (mature) bone or new bone (immature) based on collagen pattern

(woven or lamellar), osteocyte density, and anatomic location.

Enzymatic histochemistry for tartrate-resistant acid phosphatase (TRAP, Sigma-

Aldrich) was completed as previously described (60). Bone histomorphometry was

performed with Imagescope software (Aperio). The average percentage of eroded bone,

number of activated osteoclasts, osteoclast area and number of nuclei per osteoclast were

determined for the lateral aspect of the maxillary bone at the invasive tumor and

compared between treatment groups.

115

Statistical analysis

Results are displayed as means ± standard error. Normalized gene expression data

(∆CT) was analyzed for statistical significance using ANOVA and Bonferroni’s post hoc test, and graphically represented by showing relative expression compared to the cell line

with the lowest expression. Data from the in vivo experiment was analyzed by

comparing each treatment group to the control group using Student’s t-test. In the event

that data was not normally distributed, a Wilcoxon rank sum (Mann-Whitney) test was

performed. The in vivo data was evaluated by comparing each treatment group to the

vehicle group using three separate tests; therefore, a standard P value of 0.05 divided by 3

was considered significant (adjusted for multiple comparisons, P value of 0.017).

Categorical data (presence of invasion) was analyzed using Fishers exact test. All

comparisons were performed with STATA intercooled 10 (Cary, NC). Outliers were

detected using Grubb’s test (free online software at GraphPad.com).

Results

OSCC expression of COX-1 and COX-2

In order to determine if feline OSCC cell lines expressed COX-1 and COX-2, semi-

quantitative real-time RT-PCR was performed on a panel of feline and human OSCC cell

lines. It has been reported that SCCF2 cells and UMSCC12 cells induce the greatest

116

degree of osteoclastic bone resorption (86) compared to SCCF1, SCCF3 cells and A253

cells. COX-1 expression was detected at the mRNA level in all OSCC cell lines (figure

4.1A), but was not associated with the osteolytic phenotype (SCCF2 and UMSCC12

expressed the lowest levels of COX-1). Interestingly, TAF cells (feline OSCC tumor-

associated fibroblasts) expressed the highest amount of COX-1. COX-2 mRNA was

detectable in all OSCC cell lines (figure 4.1B), but was not associated with osteolytic

activity (A253 and SCCF3 expressed the most COX-2, but did not stimulate the most bone resorption).

In order to determine if COX-2 could be stimulated by factors in bone-conditioned

medium, SCCF2 cells were cultured for 24 hours in various types of conditioned

medium; unconditioned (control medium), MC3T3 (murine preosteoblasts) or murine

bone-conditioned medium. Bone-conditioned medium stimulated COX-2 expression in

SCCF2 cells but MC3T3 cell did not (figure 4.1C). COX activity was determined by

measuring prostaglandin E2 in culture medium of OSCC cells treated with unconditioned

medium or bone conditioned medium. In unconditioned medium; UMSCC12, SCC25 and

SCCF3 produced detectable levels of PGE2 after 24 hours of culture (153 pg/ml, 474

pg/ml and 200 pg/ml respectively, data not shown), however SCCF1 and SCCF2 did not

(A253 and TAF production of PGE2 was not evaluated). Bone-conditioned medium

alone contained 3840 pg/ml PGE2. There was no detectable stimulation of PGE2

secretion in OSCC cells treated with bone-conditioned medium.

117

Effect of COX inhibition on SCCF2 viability

In order to determine if COX inhibition would reduce SCCF2 viability, SCCF2 cells were cultured in various concentrations of two non-preferential COX inhibitors

(indomethacin and piroxicam) and a preferential COX-2 inhibitor (meloxicam) for 72 hours and an MTT assay was performed (figure 4.D). The IC50 of indomethacin for

SCCF2 cells was not reached (>1000 µM). The IC50 of piroxicam was 940 µM and the

IC50 of meloxicam was 270 µM. Although meloxicam had the lowest IC50 of all three

COX inhibitors, the effective concentration was well above a physiologically relevant dose (3 µM in cats, 2.5 µM to 7 µM in humans).

Effect of therapy on tumor growth and invasion

At the 21st day of treatment, meloxicam reduced tumor bioluminescence by 32% compared to vehicle-treated mice (P=0.0143, 2-tailed t test, figure 4.2), and combination therapy reduced bioluminescence by 36% (P=0.0134, 2-tailed t test, figure 2b). ZOL monotherapy resulted in a non-statistically significant 17% reduction in bioluminescence

(P=0.1509, 2-tailed t test, figure 4.2).

SCCF2Luc xenografts were composed of islands and cords of malignant epithelial cells demonstrating variable degrees of squamous differentiation and keratinization with regions of central necrosis. Maxillary bone loss was most evident in vehicle- or meloxicam-treated mice compared to ZOL- and combination-treated mice (figures 4.3A and 3B). There was no difference in the incidence of tumor invasion into sulcus of the maxillary incisor between treatment groups (figure 4.3B), which was observed in 90% of 118 vehicle-treated mice, 91% of meloxicam-treated mice, 92% of ZOL-treated mice, and

80% of combination-treated mice. ZOL, alone or combined with meloxicam, caused a non-statistically significant reduction in the incidence of tumor invasion into the nasal cavity; which was observed in 64% of vehicle-treated mice, 73% of meloxicam-treated mice, 33% of ZOL-treated mice and 40% of combination-treated mice (P=0.192, Pearson chi square).

Xenograft invasion into bone was associated with numerous TRAP-positive osteoclasts in resorption pits at the tumor-bone interface (figure 4.3C). Cytoplasmic vacuolation of some osteoclasts was observed in ZOL-treated mice (alone and in combination with meloxicam).

Tumor-bearing mice occasionally demonstrated mild to moderate bone necrosis at the invasive front of the tumor. Bone necrosis was observed in 5 of 10 ZOL + meloxicam treated-mice, 5 of 12 ZOL-only treated mice, 1 of 10 meloxicam-only treated-mice, and 1 of 11 vehicle-treated mice. Bone necrosis was only observed in xenografts with significant bone resorption. Bone and tooth necrosis was not observed on the non-tumor- bearing side of the maxilla, and was not observed in ZOL-treated, non-tumor-bearing mice. Two (2) combination treated mice and 1 ZOL-only treated mouse had evidence of tooth necrosis in addition to bone necrosis, and only occurred in mice with pronounced tumor invasion.

119

Zoledronic acid decreased loss of bone volume

In order to demonstrate the effect of treatment on maxillary bone loss, Faxitron radiography and microcomputed tomography (microCT) was performed. There was a qualitative reduction in bone loss and increased periosteal new bone formation in the region of xenograft growth in the ZOL-treated mice (monotherapy or combined with meloxicam) compared to vehicle-treated mice (figure 4.4A and B). MicroCT of 5 randomly selected mice from each treatment group was performed to quantify changes in bone volume in a 2mm thick region of interest adjacent to the xenograft. Meloxicam monotherapy had no effect on bone volume compared to vehicle-treated mice; however,

ZOL therapy (alone and combined with meloxicam), was associated with significantly increased maxillary bone volume compared to vehicle-treated mice (figure 4.4C).

Combination treatment did not result in an additional increase in bone volume compared to ZOL-monotherapy. Additionally, ZOL-treatment resulted in increased bone volume compared to vehicle-treated, non-tumor-bearing mice (dashed line) and ZOL-treated, nontumor-bearing mice (solid line). The increased bone volume was attributed to new bone formation.

Effect of zoledronic acid and meloxicam on bone area

Maxillary bone area on the nontumor-bearing side and tumor-bearing side was measured using Imagescope software. Total bone (pre-existing bone and new bone combined) on the tumor-bearing side was expressed as a percentage of bone area on the nontumor-bearing side (figure 5.5A). Xenograft growth in vehicle-treated mice was 120

associated with a 47% reduction in total bone area compared to the non-tumor-bearing

side (dashed line). There was no effect of meloxicam treatment on bone area (loss of

46%). ZOL monotherapy and ZOL-meloxicam combination therapy reduced loss of total

bone compared to vehicle-treated mice (P<0.00001 for both comparisons, 2-tailed t-test).

Mice treated with ZOL or combination therapy had increased total bone compared to the

nontumor-bearing side (15% and 27% more total bone, respectively).

Xenograft growth was associated with a 65% reduction in pre-existing bone compared

to the nontumor-bearing side (dashed line, figure 4.5B). Meloxicam treatment resulted in

a non-statistically significant reduction in pre-existing bone loss (56% compared to 65%

in vehicle-treated mice). Mice treated with ZOL monotherapy retained greater pre-

existing bone compared to vehicle-treated mice (P=0.0003, Wilcoxon rank-sum,

corresponding to a loss of 4.7%). Loss of pre-existing bone was inhibited by ZOL +

meloxicam (P=0.0002, Wilcoxon rank-sum). To determine how much of the bone on the

tumor-bearing side was composed of immature, reactive periosteal bone, the area of new bone was measured and expressed as a percentage of the total bone on the tumor-bearing side (figure 4.5C). New bone represented the lowest percentage of total bone in the ZOL

and ZOL + meloxicam-treated mice (P=0.0013 and P=0.0054, 2-tailed t-tests) which was

attributed the high retention of pre-existing bone. P values < 0.0167 were considered

statistically significant (adjusted for multiple comparisons).

121

Zoledronic acid reduced the number of osteoclasts and increased osteoclast size and

number of nuclei

Unstained tissue sections were evaluated for the presence of activated osteoclasts at the

xenograft-bone interface using TRAP histochemistry. Osteoclasts appeared as

multinucleated TRAP-positive cells in howships lacunae on the bone surface. ZOL

monotherapy and ZOL-meloxicam combination therapy significantly reduced osteoclast

number (P<0.0001 for both comparisons, 2-tailed t-test, figure 4.6A). There was no effect

of meloxicam monotherapy on the number of osteoclasts per millimeter of bone surface.

Addition of meloxicam to ZOL did not reduce osteoclast number compared to ZOL

monotherapy.

The largest osteoclasts (figure 4.6B) were observed in combination-treated mice

(P=0.0018, 2-tailed t-test, unequal variances), followed by ZOL-monotherapy-treated

mice (P=0.0001, 2-tailed t-test, equal variances). There was a trend for meloxicam

monotherapy to result in larger osteoclasts compared to vehicle-treated mice, but there

was no statistical difference (P=0.0594, 2-tailed t-test). The effect of treatment on the

number of nuclei per osteoclast (figure 4.6C) was similar to the effect on osteoclast size.

ZOL-monotherapy and ZOL-meloxicam-combination therapy resulted in the highest numbers of osteoclast nuclei (P=0.004 and P=0.007 respectively, 2-tailed t-tests).

There was no statistically significant difference in the ratio of eroded to total bone surface between each of the treatment groups and the vehicle-treated group (figure 4.6D), despite an apparent trend for reduced eroded surfaces in ZOL-treated mice and ZOL +

122

meloxicam-treated mice (P=0.0218 for combination-treated mice, 2-tailed t-test, greater than the adjusted P value of 0.0167).

Discussion

Patients with bone invasive OSCC have a worse prognosis compared to patients

without evidence of bone invasion (46-51). The purpose of this study was to determine if

the inhibitory effect of ZOL on OSCC tumor growth and bone invasion could be

enhanced by the addition of a preferential COX-2 inhibitor, Meloxicam. We previously

reported that ZOL monotherapy reduced tumor growth and osteoclastic bone resorption

in an orthotopic mouse model of maxillary invasive OSCC (117). COX inhibitors have

been shown to reduce tumor growth in rodent models of OSCC (118-120). The purpose

of this study was to determine if the inhibitory effects of ZOL on tumor growth and bone

loss could be improved by the addition of a preferential COX-2 inhibitor, Meloxicam.

COX-1 and COX-2 mRNA was detectable in all human and feline OSCC cell lines;

however, there was no apparent association with osteolytic activity. For example, COX-1

and COX-2 expression was greatest in SCCF3 cells, but SCCF3 cells were not associated

with a strong osteoclastic response (86). The OSCC cell lines that were associated with

the most in vitro bone resorption and osteoclast formation (UMSCC12 and SCCF2) had

relatively low COX-1 and COX-2 expression. These findings suggest that COX

expression in OSCC cell lines is not important in the pathogenesis of bone invasion;

123

however, SCCF2 cells did have increased COX-2 expression when stimulated with bone-

conditioned medium. It is possible that OSCC cells express more COX-2, and produce more PGE2, when in close proximity with resorbing bone.

Feline tumor-associated fibroblasts expressed much more COX-1 mRNA compared to feline OSCC cells, suggesting that the tumor stroma may be a significant source of PGE2 which could be available to stimulate bone resorption regardless of PGE2 production from the tumor cells themselves. We have observed that murine calvarial bone secretes

PGE2 into conditioned medium (data not shown), suggesting that bone itself may serve as a source of PGE2 in the OSCC-bone microenvironment.

SCCF2 cells expressed more COX-2 when exposed to bone-conditioned medium. In gastrointestinal tumors, increased COX-2 expression has been attributed to transcriptional and post-translational regulation rather than copy number gain of the COX-2 gene (125).

Transcriptional factors that have been associated with increased COX-2 expression include nuclear factor kappaB (NFκB), cAMP response element-binding protein (CREB), nuclear factor of activated T-cells (NFAT), activator protein-1 (AP-1), peroxisome proliferator-activated receptor (PPAR) and hypoxia-inducible factor 1, alpha subunit

(HIF1A) (125).

SCCF2 proliferation was not inhibited by physiologically relevant doses of indomethacin, piroxicam or meloxicam. Although the lowest IC50 was achieved with meloxicam, there was no effect at physiologically relevant doses. In contrast, meloxicam treatment reduced the rate of SCCF2Luc xenograft growth in vivo, alone or in

124

combination with zoledronic acid. There was no significant additional tumor suppression

when meloxicam was combined with zoledronic acid.

The mechanism by which meloxicam inhibited tumor growth in this model is unknown.

PGE2 has been shown to be increased in a variety of tumor tissues, and has been shown

to stimulate tumor cell proliferation while inhibiting apoptosis, in addition to promoting

angiogeneses, tumor invasiveness, escape from immune surveillance, and resistance to

chemotherapeutic drugs (125). The antiproliferative and pro-apoptotic effects of

cyclooxygenase inhibitors have been associated with both COX-dependant and COX-

independent mechanisms (126).

COX-2 inhibitors have demonstrated antiproliferative effects on OSCC in vitro, albeit

at concentrations higher than needed to inhibit PGE2 production (126). Meloxicam was

shown to inhibit proliferation of a human OSCC cell line (KB) at a relatively high

concentration (IC50 of 386 µM, higher than IC50 concentrations of celecoxib, NS-198

and nimulside) (126). A similarly high IC50 of meloxicam was observed in SCCF2 cells.

While there are no published reports of the effects of meloxicam on clinical or preclinical in vivo OSCC studies, Meloxicam has been shown to induce tumor cell apoptosis in human patients with esophageal cancer (127) and reduce tumor growth in animals models of other forms of cancer including osteosarcoma (128), hepatocellular carcinoma

(129)and ovarian carcinoma (130).

Tumor cell expression of COX-2 and subsequent production of PGE2 is not the only target of a systemically administered COX-2 inhibitor. Meloxicam administration may have reduced PGE2 production in the tumor stroma or in the resorbing bone (COX

125

expression was demonstrated in feline OSCC-associated fibroblasts and ex vivo murine bone cultures were shown to release PGE2 into conditioned medium).

Regardless of the source, PGE2 activation of prostaglandin E (EP) receptors is capable of transactivating epidermal growth factor receptor (EGFR), potentially leading to increased angiogenesis, tumor invasion, and reduced apoptosis (125). EGFR is known to be overexpressed in human OSCC (131, 132), and EGFR expression has been demonstrated in spontaneous feline OSCC tumors (133) and in the three feline OSCC cell lines used in this study (unpublished data).

As expected, ZOL therapy reduced bone loss in vivo that was characterized by increased bone volume and reduced numbers of osteoclasts. ZOL is known to interfere with the mevalonate (MVA) pathway by inhibiting farnesyl pyrophosphate synthase (83,

116, 134) leading to reduced prenylation of small guanosine-triphosphate (GTP)-binding proteins required for osteoclast function and survival (84, 85).

ZOL-treatment increased osteoclast size and number of nuclei in this study, consistent with previously reported data (117, 135). Interestingly, the presence of giant, hypernucleated osteoclasts has been observed in women on long term bisphosphonate therapy for the treatment of osteoporosis, but increased numbers of osteoclasts were also observed (136). The mechanism of hypernucleation is not known; however, Weinstein et al. speculated that inhibition of bone resorption would have caused reduced local calcium concentration and reduced apoptotic signals in the osteoclast, leading to increased duration of survival and more time for fusion of osteoclasts with mononuclear progenitors (136). Since we observed reduced osteoclast number in the ZOL-treated mice

126

with OSCC xenografts, it is unlikely that reduced osteoclast apoptosis and increased

osteoclast lifespan was the reason for increased osteoclast size and number of nuclei

observed in this study.

We previously reported that ZOL reduced, but did not eliminate, loss of pre-existing

bone in a bone invasive model of OSCC (117). In contrast, this study showed that ZOL

almost completely inhibited loss of pre-existing bone. The improved effectiveness of

ZOL in this study may be due to the fact that treatment was initiated earlier in the course of xenograft growth (3 days following tumor cell injection compared to 7 days in the previous study). Additionally, the total duration of xenograft growth in this study was 7 days shorter compared to the previous study. The combination of earlier treatment and shorter duration of tumor growth may be the reason for the increased effectiveness of

ZOL against tumor-associated bone loss.

Osteonecrosis of the jaw (ONJ) occurs in a low percentage of cancer patients treated

with bisphosphonates (estimated incidence of 5% to 10%) (100). ZOL (alone and in

combination with meloxicam) was well tolerated in the mice and there was no evidence

of ONJ. Small amounts of tumor-associated necrotic bone within the xenografts of ZOL- treated mice was attributed to the retention of bone undergoing necrosis as a result of tumor infiltration, since the antiresorptive effects of ZOL would inhibit removal of necrotic bone. Bone necrosis was rarely observed in non ZOL-treated animals (1 vehicle- treated mouse and 1 meloxicam-treated mouse). Tooth necrosis appeared as an extension of bone necrosis and was observed in few mice (3 of 22 mice). Bone necrosis was associated with invasive behavior, and is suggestive that OSCC patients with existing

127

bone disease may be at increased risk for the development of ONJ. The potential benefits

of reducing bone loss and invasion in cases of non-resectable, bone invasive cancer likely outweigh the risks.

The data suggested that Meloxicam mildly increased retention of pre-existing bone, reduced eroded surfaces and increased osteoclast size. This trend was observed when meloxicam monotherapy was compared to untreated mice and when ZOL + meloxicam combination therapy was compared to ZOL monotherapy, suggesting that COX-2 inhibition may have mild inhibitory effect on osteoclastic bone resorption in OSCC.

Hiraga et al. showed that bone-derived TGF-β1 promoted osteoclastogenesis and bone resorption by inducing COX-2-expression and PGE2 production in MDA-MB-231 cells

(human breast cancer), which up-regulated RANKL expression in osteoblasts (106).

Similarly, cultured bone released factors which stimulated COX-2 and PTHrP expression in SCCF2 cells, and SCCF2 cells stimulated RANKL expression in murine preosteoblasts

(86). In the MDA-MB-321 model, COX-2 inhibitors significantly suppressed bone metastases with decreased osteoclast number and increased apoptosis in MDA-MB-231

cells. These findings led to the conclusion that breast cancer-derived COX-2 plays a critical role in the promotion of osteolytic bone metastases by cooperating with PTHrP to stimulate osteoclastic bone resorption (106).

Meloxicam was more effective than ZOL at reducing xenograft growth but did not have

a significant effect on bone resorption. The combination of meloxicam and ZOL was well

tolerated but did not stimulate additional tumor suppression or inhibition of bone loss

128 compared to meloxicam- or ZOL-monotherapy. Although a synergistic effect of tumor progression was not observed, the results indicate that meloxicam and zoledronic acid would be of benefit in the management of oral squamous cell carcinoma.

129

A * *

* # B #

2.2 C 9 * D 28

1.87 mRNA control)

mRNA

/ 6

1.6 Control) B2M

B2M / 5 to /

1.4

4 untreated

1.2 mRNA

mRNA

2 (relative 3 ‐ 1 COX2 2 COX NS

0.8(relative 1 0.6 0 ControlControl MC3T3 50% CM MC3T3‐CM Bone CM 50% F2+Bone Bone‐CM CM

Figure 4.1: OSCC cells express COX-1 and COX-2, and SCCF2 growth inhibition did not occur at physiologically relevant doses of COX inhibitors

SCCF2 and UMSCC12 cells have been previously shown to stimulate osteoclastic bone resorption. COX expression was measured using real-time RT-PCR and mRNA levels

130

were expressed relative to the lowest-expressing cell line. A. COX-1 mRNA was detected

in all OSCC cell lines, and was lowest in the osteolytic SCCF2 and UMSCC12 cells.

TAF cells (feline OSCC tumor-associated fibroblasts) expressed the greatest amount of

COX-1. B. COX-2 mRNA was detectable in all OSCC cell lines, but was highest in A253

and SCCF3 cells (associated with the lowest levels of osteoclast activity). (*highest expression compared to all other cell lines, P<0.05, ANOVA and Bonferroni post hoc

test; # SCCF2 and SCCF3 had higher expression than the other cell lines, P<0.05

ANOVA and Bonferroni post hoc test). C. SCCF2 cells were cultured for 24 hours in unconditioned medium (control), MC3T3 conditioned medium or murine bone conditioned medium. Bone-conditioned medium stimulated COX-2 expression in SCCF2 cells (*P<0.05 ANOVA and Bonferroni post hoc test), but MC3T3-conditioned medium did not (NS = not significant compared to control). D. SCCF2 cells were cultured in indomethacin and piroxicam (non-preferential COX inhibitors) and meloxicam (a preferential COX-2 inhibitor) for 72 hours and an MTT assay was performed. The IC50 was lowest for meloxicam (270 µM). The physiologically relevant dose of meloxicam is

3 µM in cats, 2.5 µM to 7 µM in humans.

131

Figure 4.2 Meloxicam reduced tumor growth

On day 21 of treatment, meloxicam reduced tumor radiance by 32% compared to the

vehicle-treated mice, and combination therapy reduced tumor radiance by 36%.

*P<0.0167. ZOL monotherapy resulted in a non-statistically significant 17% reduction in tumor radiance.

132

Vehicle Meloxicam Zoledronic acid Combined Histopathology

TRAP

Figure 4.3: Zoledronic acid reduced bone loss and was associated with osteoclast vacuolar degeneration

A. Maxillary bone loss was most common in vehicle- and meloxicam-treated mice. B.

Tumor invasion around the maxillary incisor occurred with similar frequency between

133 treatment groups. C. Bone invasion was associated with TRAP-positive osteoclasts in resorption pits at the tumor-bone interface, and vacuolar degeneration of osteoclasts was observed in ZOL-treated mice (alone and in combination with meloxicam).

134

Vehicle Meloxicam Zoledronic acid Combined MicroCT Faxitron

28 26 * *

) 24 3 22 (mm 20 18 Volume

Bone 14 12 10 Vehicle Meloxicam Zoledronic Acid Combined

Figure 4.4: Zoledronic acid reduced loss of bone volume

A. Faxitron radiography of a representative mouse from each group. There was reduced maxillary bone loss and increased new bone formation in the ZOL-treated mice 135

(monotherapy or combined with meloxicam) compared to vehicle-treated mice and

meloxicam-treated mice. B. Reconstructed microCT images of a representative mouse

from each group showing reduced maxillary bone loss and increased new bone formation

in the ZOL- treated mice (monotherapy or combined with meloxicam) compared to

vehicle-treated mice or meloxicam-treated mice. C. Quantitative microCT analysis of

bone volume was performed on five randomly selected mice from each treatment group.

Meloxicam monotherapy had no effect on bone volume compared to vehicle-treated

mice; however, ZOL therapy (alone and combined with meloxicam), was associated with

significantly increased maxillary bone volume compared to vehicle-treated mice

(*P<0.0167). ZOL-treatment resulted in increased bone volume compared to vehicle- treated, non-tumor-bearing mice (dashed line) and ZOL-treated, non-tumor-bearing mice

(solid line). Increased bone volume was attributed to new bone formation.

136

Figure 4.5: Zoledronic acid reduced loss of bone area

137

Maxillary bone area on the non-tumor-bearing side and tumor-bearing side was measured using Imagescope software. A. Xenograft growth in vehicle-treated mice was associated with a 47% reduction in total bone area compared to the non-tumor-bearing side (dashed line). There was no effect of meloxicam treatment on bone area (loss of 46%). ZOL monotherapy and ZOL-meloxicam combination therapy eliminated loss of total bone compared to vehicle-treated mice (*P<0.0167). Mice treated with ZOL or combination therapy had increased total bone compared to the nontumor-bearing side (15% and 27% more total bone, respectively). B. Xenograft growth was associated with a 65% reduction in pre-existing bone compared to the nontumor-bearing side (dashed line). Meloxicam treatment resulted in a non-statistically significant reduction in pre-existing bone loss.

Mice treated with ZOL monotherapy retained greater pre-existing bone compared to vehicle-treated mice (*P<0.0167). ZOL-meloxicam also retained more pre-existing bone compared to vehicle-treated mice (*P<0.0167) and completely inhibited loss of pre- existing bone area. C. New bone area was determined and expressed as a percentage of the total bone on the tumor-bearing side. New bone represented the lowest percentage of total bone in the ZOL and ZOL-meloxicam combination treated mice (*P<0.0167) which was attributed the high retention of pre-existing bone.

138

14 800 A B 700 12 *

) 600 10 2 * (µm

mm 500

/ 8 area 400 6 * * 300 Osteoclasts

4 Osteoclast 200

2 100

0 0 Vehicle Meloxicam Zoledronic acid Combined Vehicle Meloxicam Zoledronic acid Combined C 3.0 D 95 * 2.5 * 90 (%) 2.0 85 osteoclast

Surface 1.5 80 nuclei

of 1.0 75 Resorbed

Number 0.5 70

0.0 65 Vehicle Meloxicam Zoledronic acid Combined Vehicle Meloxicam Zoledronic acid Combined

Figure 4.6: Zoledronic acid reduced osteoclast number and increased osteoclast size and number of nuclei

A. ZOL monotherapy and ZOL-meloxicam combination therapy significantly reduced osteoclast number (*P<0.0167) compared to vehicle-treated mice, but there was no effect of meloxicam monotherapy on the number of osteoclasts. B. The largest osteoclasts were observed in combination-treated mice (*P<0.0167), followed by ZOL-treated mice

(*P<0.0167). There was a trend for meloxicam monotherapy to result in larger osteoclasts compared to vehicle-treated mice, but there was no statistical difference (P>0.0167). C.

ZOL-monotherapy and ZOL-meloxicam-combination therapy resulted in the greatest

139 number of osteoclast nuclei per cell (*P<0.0167). D. There was a trend for ZOL + meloxicam-treated mice to have the lowest percentage of eroded bone surface

(P>0.0167).

140

Bibliography

1. Dorn CR, Priester WA. Epidemiologic analysis of oral and pharyngeal cancer in dogs, cats, horses, and cattle. J Am Vet Med Assoc 1976;169(11):1202-6.

2. Stebbins K, E., Morse C, C., Goldschmidt MH. Feline oral neoplasia: A ten-year survey. Vet Pathol 1989;26(2):121-8.

3. Morris J, Dobson J. Head and neck. In: Morris J, Dobson J, editors. Small Animal Oncology. Alder Press Ltd, Oxford: Blackwell Science Ltd.; 2001. p. 94-124.

4. Hutson C, A., Willauer C, C., Walder E, J., Stone J, L., Klein MK. Treatment of mandibular squamous cell carcinoma in cats by use of mandibulectomy and radiotherapy: Seven cases (1987-1989). J Am Vet Med Assoc 1992;201(5):777-81.

5. Northrup NC, Selting KA, Rassnick KM, et al. Outcomes of cats with oral tumors treated with mandibulectomy: 42 cases. J Am Anim Hosp Assoc 2006;42(5):350-60.

6. Hayes AM, Adams VJ, Scase TJ, Murphy S. Survival of 54 cats with oral squamous cell carcinoma in united kingdom general practice. J Small Anim Pract 2007;48(7):394-9.

7. Barger AM, Fan TM, de Lorimier LP, Sprandel IT, O'Dell-Anderson K. Expression of receptor activator of nuclear factor kappa-B ligand (RANKL) in neoplasms of dogs and cats. J Vet Intern Med 2007;21(1):133-40.

8. Chuang FH, Hsue SS, Wu CW, Chen YK. Immunohistochemical expression of RANKL, RANK, and OPG in human oral squamous cell carcinoma. J Oral Pathol Med 2009;38(10):753-8.

9. Shibahara T, Nomura T, Cui NH, Noma H. A study of osteoclast-related cytokines in mandibular invasion by squamous cell carcinoma. Int J Oral Maxillofac Surg 2005;34(7):789-93.

10. Tannehill-Gregg S, Kergosien E, Rosol TJ. 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):676-83.

11. Pries R, Wollenberg B. Cytokines in head and neck cancer. Cytokine Growth Factor Rev 2006;17(3):141-6. 141

12. Rosenthal E, McCrory A, Talbert M, Young G, Murphy-Ullrich J, Gladson C. Elevated expression of TGF-beta1 in head and neck cancer-associated fibroblasts. Mol Carcinog 2004;40(2):116-21.

13. Walker DM, Boey G, McDonald LA. The pathology of oral cancer. Pathology 2003;35(5):376-83.

14. Scully C, Bagan J. Oral squamous cell carcinoma: Overview of current understanding of aetiopathogenesis and clinical implications. Oral Dis 2009;15(6):388-99.

15. Bertone ER, Snyder LA, Moore AS. Environmental and lifestyle risk factors for oral squamous cell carcinoma in domestic cats. J Vet Intern Med 2003;17(4):557-62.

16. Farshadpour F, Hordijk GJ, Koole R, Slootweg PJ. Non-smoking and non-drinking patients with head and neck squamous cell carcinoma: A distinct population. Oral Dis 2007;13(2):239-43.

17. Robinson DA, Bogdanffy MS, Reed CJ. Histochemical localisation of carboxylesterase activity in rat and mouse oral cavity mucosa. Toxicology 2002;180(3):209-20.

18. Tannehill-Gregg SH, Levine AL, Rosol TJ. Feline head and neck squamous cell carcinoma: A natural model for the human disease and development of a mouse model. Veterinary and Comparative Oncology 2006;4(2):84-97.

19. Sung V, Stubbs JT,3rd, Fisher L, Aaron AD, Thompson EW. Bone sialoprotein supports breast cancer cell adhesion proliferation and migration through differential usage of the alpha(v)beta3 and alpha(v)beta5 integrins. J Cell Physiol 1998;176(3):482- 94.

20. Kiefer JA, Farach-Carson MC. Type I collagen-mediated proliferation of PC3 prostate carcinoma cell line: Implications for enhanced growth in the bone microenvironment. Matrix Biol 2001;20(7):429-37.

21. Kostenuik PJ, Sanchez-Sweatman O, Orr FW, Singh G. Bone cell matrix promotes the adhesion of human prostatic carcinoma cells via the alpha 2 beta 1 integrin. Clin Exp Metastasis 1996;14(1):19-26.

22. Goltzman D. Osteolysis and cancer. J Clin Invest 2001;107(10):1219-20.

23. Tomita A, Kasaoka T, Inui T, et al. Human breast adenocarcinoma (MDA-231) and human lung squamous cell carcinoma (hara) do not have the ability to cause bone resorption by themselves during the establishment of bone metastasis. Clin Exp Metastasis 2008;25(4):437-44. 142

24. Heliotis M, Ripamonti U, Ferretti C, Kerawala C, Mantalaris A, Tsiridis E. The basic science of bone induction. Br J Oral Maxillofac Surg 2009;47(7):511-4.

25. Van Cann EM, Slootweg PJ, de Wilde PC, et al. The prediction of mandibular invasion by squamous cell carcinomas with the expression of osteoclast-related cytokines in biopsy specimens. Int J Oral Maxillofac Surg 2009;38(3):279-84.

26. Okada H, Nishijima Y, Yoshino T, Grone A, Capen CC, Rosol TJ. Immunohistochemical localization of parathyroid hormone-related protein in canine mammary tumors. Vet Pathol 1997;34(4):356-9.

27. Clemens TL, Cormier S, Eichinger A, et al. Parathyroid hormone-related protein and its receptors: Nuclear functions and roles in the renal and cardiovascular systems, the placental trophoblasts and the pancreatic islets. Br J Pharmacol 2001;134(6):1113-36.

28. Blomme EA, Werkmeister JR, Zhou H, Kartsogiannis V, Capen CC, Rosol TJ. Parathyroid hormone-related protein expression and secretion in a skin organotypic culture system. Endocrine 1998;8(2):143-51.

29. Miao D, He B, Jiang Y, et al. Osteoblast-derived PTHrP is a potent endogenous bone anabolic agent that modifies the therapeutic efficacy of administered PTH 1-34. J Clin Invest 2005;115(9):2402-11.

30. Okazaki K, Jingushi S, Ikenoue T, Urabe K, Sakai H, Iwamoto Y. Expression of parathyroid hormone-related peptide and insulin-like growth factor I during rat fracture healing. J Orthop Res 2003;21(3):511-20.

31. Bolliger AP, Graham PA, Richard V, Rosol TJ, Nachreiner RF, Refsal KR. Detection of parathyroid hormone-related protein in cats with humoral hypercalcemia of malignancy. Vet Clin Pathol 2002;31(1):3-8.

32. Liao J, McCauley LK. Skeletal metastasis: Established and emerging roles of parathyroid hormone related protein (PTHrP). Cancer Metastasis Rev 2006.

33. Nishihara M, Kanematsu T, Taguchi T, Razzaque MS. PTHrP and tumorigenesis: Is there a role in prognosis? Ann N Y Acad Sci 2007;1117:385-92.

34. Martin TJ. Osteoblast-derived PTHrP is a physiological regulator of bone formation. J Clin Invest 2005;115(9):2322-4.

35. Kayamori K, Sakamoto K, Nakashima T, et al. Roles of interleukin-6 and parathyroid hormone-related peptide in osteoclast formation associated with oral cancers: Significance of interleukin-6 synthesized by stromal cells in response to cancer cells. Am J Pathol 2010;176(2):968-80. 143

36. Bhatia V, Saini MK, Falzon M. Nuclear PTHrP targeting regulates PTHrP secretion and enhances LoVo cell growth and survival. Regul Pept 2009;158(1-3):149-55.

37. Shibuya K, Mathers CD, Boschi-Pinto C, Lopez AD, Murray CJ. Global and regional estimates of cancer mortality and incidence by site: II. results for the global burden of disease 2000. BMC Cancer 2002;2:37.

38. American Cancer Society. Cancer facts & figures 2009. Atlanta: American Cancer Society; 2009.

39. Funk G, F., Karnell LH, Robinson R, A., Zhen W, K., Trask D, K., Hoffman HT. Presentation, treatment, and outcome of oral cavity cancer: A national cancer data base report. Head Neck 2002;24(2):165-80.

40. Silverman S, Jr., editor. Oral cancer. 4th ed. Hamilton, Ontario, Canada: BC Decker Inc.; 1998.

41. Silverman S,Jr. Demographics and occurrence of oral and pharyngeal cancers. the outcomes, the trends, the challenge. J Am Dent Assoc 2001;132 Suppl:7S-11S.

42. SEER cancer statistics review, 1975-2006 [homepage on the Internet]. Bethesda, MD: National Cancer Institute. 2009.

43. Smith LP, Thomas GR. 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 models of HNSCC, animal models of transplanted xenogeneic human tumors. Int J Cancer 2006;118(9):2111-22.

44. Kimura Y, Sumi M, Sumi T, Ariji Y, Ariji E, Nakamura T. Deep extension from carcinoma arising from the gingiva: CT and MR imaging features. AJNR Am J Neuroradiol 2002;23(3):468-72.

45. Ito M, Izumi N, Cheng J, et al. Jaw bone remodeling at the invasion front of gingival squamous cell carcinomas. J Oral Pathol Med 2003;32(1):10-7.

46. Patel RS, Dirven R, Clark JR, Swinson BD, Gao K, O'Brien CJ. The prognostic impact of extent of bone invasion and extent of bone resection in oral carcinoma. Laryngoscope 2008;118(5):780-5.

47. Munoz Guerra MF, Naval Gias L, Campo FR, Perez JS. Marginal and segmental mandibulectomy in patients with oral cancer: A statistical analysis of 106 cases. J Oral Maxillofac Surg 2003;61(11):1289-96.

144

48. Shaw RJ, Brown JS, Woolgar JA, Lowe D, Rogers SN, Vaughan ED. The influence of the pattern of mandibular invasion on recurrence and survival in oral squamous cell carcinoma. Head Neck 2004;26(10):861-9.

49. Ogura I, Kurabayashi T, Amagasa T, Okada N, Sasaki T. Mandibular bone invasion by gingival carcinoma on dental CT images as an indicator of cervical lymph node metastasis. Dentomaxillofac Radiol 2002;31(6):339-43.

50. Ogura I, Kurabayashi T, Sasaki T, Amagasa T, Okada N, Kaneda T. Maxillary bone invasion by gingival carcinoma as an indicator of cervical metastasis. Dentomaxillofac Radiol 2003;32(5):291-4.

51. Wong RJ, Keel SB, Glynn RJ, Varvares MA. Histological pattern of mandibular invasion by oral squamous cell carcinoma. Laryngoscope 2000;110(1):65-72.

52. Lu SL, Herrington H, Wang XJ. Mouse models for human head and neck squamous cell carcinomas. Head Neck 2006;28(10):945-54.

53. Erdem NF, Carlson ER, Gerard DA. Characterization of gene expression profiles of 3 different human oral squamous cell carcinoma cell lines with different invasion and metastatic capacities. J Oral Maxillofac Surg 2008;66(5):918-27.

54. Takayama Y, Mori T, Nomura T, Shibahara T, Sakamoto M. Parathyroid-related protein plays a critical role in bone invasion by oral squamous cell carcinoma. Int J Oncol 2010;36(6):1387-94.

55. Martin CK, Tannehill-Gregg SH, Wolfe TD, Rosol TJ. Bone invasive oral squamous cell carcinoma in cats: Pathology and expression of parathyroid hormone-related protein. Vet Pathol In Press.

56. ISCN 2005: An international system for human cytogenetic nomenclature. Shaffer LG and Tommerup N, editors. Basel, Switzerland: S. Karger; 2005.

57. Cho KW, Youn HY, Watari T, Tsujimoto H, Hasegawa A, Satoh H. A proposed nomenclature of the domestic cat karyotype. Cytogenet Cell Genet 1997;79(1-2):71-8.

58. Meghji S, Morrison MS, Henderson B, Arnett TR. pH dependence of bone resorption: Mouse calvarial osteoclasts are activated by acidosis. Am J Physiol Endocrinol Metab 2001;280(1):E112-9.

59. Shu ST, Martin CK, Thudi NK, Dirksen WP, Rosol TJ. Osteolytic bone resorption in adult T-cell leukemia/lymphoma. Leuk Lymphoma 2010;51(4):702-14.

145

60. Thudi NK, Martin CK, Nadella MV, 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):1116-25.

61. Simon C, Nemechek AJ, Boyd D, et al. An orthotopic floor-of-mouth cancer model allows quantification of tumor invasion. Laryngoscope 1998;108(11 Pt 1):1686-91.

62. Henson B, Li F, Coatney DD, et al. An orthotopic floor-of-mouth model for locoregional growth and spread of human squamous cell carcinoma. J Oral Pathol Med 2007;36(6):363-70.

63. Prime SS, Eveson JW, Stone AM, et al. Metastatic dissemination of human malignant oral keratinocyte cell lines following orthotopic transplantation reflects response to TGF- beta 1. J Pathol 2004;203(4):927-32.

64. Jin C, Jin Y, Wennerberg J, Annertz K, Enoksson J, Mertens F. Cytogenetic abnormalities in 106 oral squamous cell carcinomas. Cancer Genet Cytogenet 2006;164(1):44-53.

65. Martin CL, Reshmi SC, Ried T, et al. Chromosomal imbalances in oral squamous cell carcinoma: Examination of 31 cell lines and review of the literature. Oral Oncol 2008;44(4):369-82.

66. Lingen MW, Kumar V. Head and neck. In: Kumar V, Abbas AK, Fausto N, editors. Robbins and Cotran Pathologic Basis of Disease. 7th ed. Philadelphia, Pennsylvania: Elsevier Saunders; 2005. p. 773.

67. Murphy WJ, Davis B, David VA, et al. A 1.5-mb-resolution radiation hybrid map of the cat genome and comparative analysis with the canine and human genomes. Genomics 2007;89(2):189-96.

68. Cat genome project: Genetic maps of the domestic cat [homepage on the Internet]. National Cancer Institiute, Laboratory of Genomic Diversity. 1997 October, 1997.

69. Deyama Y, Tei K, Yoshimura Y, et al. Oral squamous cell carcinomas stimulate osteoclast differentiation. Oncol Rep 2008;20(3):663-8.

70. Tada T, Jimi E, Okamoto M, Ozeki S, Okabe K. Oral squamous cell carcinoma cells induce osteoclast differentiation by suppression of osteoprotegerin expression in osteoblasts. Int J Cancer 2005;116(2):253-62.

71. Tada T, Shin M, Fukushima H, et al. Oral squamous cell carcinoma cells modulate osteoclast function by RANKL-dependent and -independent mechanisms. Cancer Lett 2009;274(1):126-31. 146

72. Nordstrand A, Nilsson J, Tieva A, Wikstrom P, Lerner UH, Widmark A. Establishment and validation of an in vitro co-culture model to study the interactions between bone and prostate cancer cells. Clin Exp Metastasis 2009;26(8):945-53.

73. Fujikawa M, Takata Y, Okamura K, et al. Hypercalcemia associated with parathyroid hormone-related protein at the terminal stage of uncomplicated squamous cell carcinoma in the head and neck region. Head Neck 2002;24(1):56-62.

74. Kornberg LJ, Villaret D, Popp M, et al. Gene expression profiling in squamous cell carcinoma of the oral cavity shows abnormalities in several signaling pathways. Laryngoscope 2005;115(4):690-8.

75. Tsuchimochi M, Kameta A, Sue M, Katagiri M. Immunohistochemical localization of parathyroid hormone-related protein (PTHrP) and serum PTHrP in normocalcemic patients with oral squamous cell carcinoma. Odontology 2005;93(1):61-71.

76. Yamada T, Tsuda M, Ohba Y, Kawaguchi H, Totsuka Y, Shindoh M. PTHrP promotes malignancy of human oral cancer cell downstream of the EGFR signaling. Biochem Biophys Res Commun 2008;368(3):575-81.

77. Ishikuro M, Sakamoto K, Kayamori K, et al. Significance of the fibrous stroma in bone invasion by human gingival squamous cell carcinomas. Bone 2008;43(3):621-7.

78. Roodman GD. Mechanisms of bone metastasis. N Engl J Med 2004;350(16):1655-64.

79. Chirgwin J, M., Guise TA. Molecular mechanisms of tumor-bone interactions in osteolytic metastases. Crit Rev Eukaryot Gene Expr 2000;10(2):159-78.

80. Yoneda T, Hiraga T. Crosstalk between cancer cells and bone microenvironment in bone metastasis. Biochem Biophys Res Commun 2005;328(3):679-87.

81. Goda T, Shimo T, Yoshihama Y, et al. Bone destruction by invading oral squamous carcinoma cells mediated by the transforming growth factor-beta signalling pathway. Anticancer Res 2010;30(7):2615-23.

82. Deshpande AM, Wong DT. Molecular mechanisms of head and neck cancer. Expert Review of Anticancer Therapy 2008;8(5):799-809. Available from: http://dx.doi.org/10.1586/14737140.8.5.799.

83. Kavanagh KL, Guo K, Dunford JE, et al. The molecular mechanism of nitrogen- containing bisphosphonates as antiosteoporosis drugs. Proc Natl Acad Sci U S A 2006;103(20):7829-34.

84. Hall A. Rho GTPases and the actin cytoskeleton. Science 1998;279(5350):509-14. 147

85. Luckman SP, Hughes DE, Coxon FP, Graham R, Russell G, Rogers MJ. Nitrogen- containing bisphosphonates inhibit the mevalonate pathway and prevent post- translational prenylation of GTP-binding proteins, including ras. J Bone Miner Res 1998;13(4):581-9.

86. Martin CK. Mechanisms and treatment of bone resorption in models of oral squamous cell carcinoma. 2010:Chapter 2.

87. Okamoto M, Hiura K, Ohe G, et al. Mechanism for bone invasion of oral cancer cells mediated by interleukin-6 in vitro and in vivo. Cancer 2000;89(9):1966-75.

88. Cui N, Nomura T, Noma H, et al. Effect of YM529 on a model of mandibular invasion by oral squamous cell carcinoma in mice. Clin Cancer Res 2005;11(7):2713-9.

89. Nomura T, Shibahara T, Katakura A, Matsubara S, Takano N. Establishment of a murine model of bone invasion by oral squamous cell carcinoma. Oral Oncol 2007;43(3):257-62.

90. Kimmel DB. Mechanism of action, pharmacokinetic and pharmacodynamic profile, and clinical applications of nitrogen-containing bisphosphonates. J Dent Res 2007;86(11):1022-33.

91. Monkkonen H, Auriola S, Lehenkari P, et al. A new endogenous ATP analog (ApppI) inhibits the mitochondrial adenine nucleotide translocase (ANT) and is responsible for the apoptosis induced by nitrogen-containing bisphosphonates. Br J Pharmacol 2006;147(4):437-45.

92. Yuasa T, Kimura S, Ashihara E, Habuchi T, Maekawa T. Zoledronic acid - a multiplicity of anti-cancer action. Curr Med Chem 2007;14(20):2126-35.

93. Coxon FP, Rogers MJ. The role of prenylated small GTP-binding proteins in the regulation of osteoclast function. Calcif Tissue Int 2003;72(1):80-4.

94. Dass CR, Choong PF. Zoledronic acid inhibits osteosarcoma growth in an orthotopic model. Mol Cancer Ther 2007;6(12 Pt 1):3263-70.

95. Guise TA. Antitumor effects of bisphosphonates: Promising preclinical evidence. Cancer Treat Rev 2008;34 Suppl 1:S19-24.

96. Lipton A. Emerging role of bisphosphonates in the clinic--antitumor activity and prevention of metastasis to bone. Cancer Treat Rev 2008;34 Suppl 1:S25-30.

148

97. Wypij JM, Fan TM, Fredrickson RL, Barger AM, de Lorimier LP, Charney SC. In vivo and in vitro efficacy of zoledronate for treating oral squamous cell carcinoma in cats. J Vet Intern Med 2008;22(1):158-63.

98. Peng H, Sohara Y, Moats RA, et al. The activity of zoledronic acid on neuroblastoma bone metastasis involves inhibition of osteoclasts and tumor cell survival and proliferation. Cancer Res 2007;67(19):9346-55.

99. Raikkonen J, Crockett JC, Rogers MJ, Monkkonen H, Auriola S, Monkkonen J. Zoledronic acid induces formation of a pro-apoptotic ATP analogue and isopentenyl pyrophosphate in osteoclasts in vivo and in MCF-7 cells in vitro. Br J Pharmacol 2009;157(3):427-35.

100. Edwards BJ, Gounder M, McKoy JM, et al. Pharmacovigilance and reporting oversight in US FDA fast-track process: Bisphosphonates and osteonecrosis of the jaw. Lancet Oncol 2008;9(12):1166-72.

101. Reid IR. Osteonecrosis of the jaw: Who gets it, and why? Bone 2009;44(1):4-10.

102. Karmali RA, Wustrow T, Thaler HT, Strong EW. Prostaglandins in carcinomas of the head and neck. Cancer Lett 1984;22(3):333-6.

103. Jung TT, Berlinger NT, Juhn SK. Prostaglandins in squamous cell carcinoma of the head and neck: A preliminary study. Laryngoscope 1985;95(3):307-12.

104. Chan G, Boyle JO, Yang EK, et al. Cyclooxygenase-2 expression is up-regulated in squamous cell carcinoma of the head and neck. Cancer Res 1999;59(5):991-4.

105. Hayes A, Scase T, Miller J, Murphy S, Sparkes A, Adams V. COX-1 and COX-2 expression in feline oral squamous cell carcinoma. J Comp Pathol 2006;135(2-3):93-9.

106. Hiraga T, Myoui A, Choi ME, Yoshikawa H, Yoneda T. Stimulation of cyclooxygenase-2 expression by bone-derived transforming growth factor-beta enhances bone metastases in breast cancer. Cancer Res 2006;66(4):2067-73.

107. Sheng H, Shao J, Morrow JD, Beauchamp RD, DuBois RN. Modulation of apoptosis and bcl-2 expression by prostaglandin E2 in human colon cancer cells. Cancer Res 1998;58(2):362-6.

108. Tsujii M, Kawano S, Tsuji S, Sawaoka H, Hori M, DuBois RN. Cyclooxygenase regulates angiogenesis induced by colon cancer cells. Cell 1998;93(5):705-16.

109. Huang M, Stolina M, Sharma S, et al. Non-small cell lung cancer cyclooxygenase-2- dependent regulation of cytokine balance in lymphocytes and macrophages: Up- 149

regulation of interleukin 10 and down-regulation of interleukin 12 production. Cancer Res 1998;58(6):1208-16.

110. Tsujii M, DuBois RN. Alterations in cellular adhesion and apoptosis in epithelial cells overexpressing prostaglandin endoperoxide synthase 2. Cell 1995;83(3):493-501.

111. Tsujii M, Kawano S, DuBois RN. Cyclooxygenase-2 expression in human colon cancer cells increases metastatic potential. Proc Natl Acad Sci U S A 1997;94(7):3336- 40.

112. Ono K, Akatsu T, Murakami T, et al. Involvement of cyclo-oxygenase-2 in osteoclast formation and bone destruction in bone metastasis of mammary carcinoma cell lines. J Bone Miner Res 2002;17(5):774-81.

113. Akatsu T, Takahashi N, Debari K, et al. Prostaglandins promote osteoclastlike cell formation by a mechanism involving cyclic adenosine 3',5'-monophosphate in mouse bone marrow cell cultures. Journal of bone and mineral research : the official journal of the American Society for Bone and Mineral Research 1989;4(1):29-35.

114. Yasuda H, Shima N, Nakagawa N, et al. Osteoclast differentiation factor is a ligand for osteoprotegerin/osteoclastogenesis-inhibitory factor and is identical to TRANCE/RANKL. Proc Natl Acad Sci U S A 1998;95(7):3597-602.

115. Liu XH, Kirschenbaum A, Yao S, Levine AC. Interactive effect of interleukin-6 and prostaglandin E2 on osteoclastogenesis via the OPG/RANKL/RANK system. Ann N Y Acad Sci 2006;1068:225-33.

116. Green JR, Clezardin P. Mechanisms of bisphosphonate effects on osteoclasts, tumor cell growth, and metastasis. Am J Clin Oncol 2002;25(6 Suppl 1):S3-9.

117. Martin CK, Werbeck JL, Thudi NK, et al. Zoledronic acid reduces bone loss and tumor growth in an orthotopic xenograft model of osteolytic oral squamous cell carcinoma. Cancer Res In Press.

118. Wang Z, Fuentes CF, Shapshay SM. Antiangiogenic and chemopreventive activities of celecoxib in oral carcinoma cell. Laryngoscope 2002;112(5):839-43.

119. Li N, Sood S, Wang S, et al. Overexpression of 5-lipoxygenase and cyclooxygenase 2 in hamster and human oral cancer and chemopreventive effects of zileuton and celecoxib. Clin Cancer Res 2005;11(5):2089-96.

120. Roh JL, Sung MW, Kim KH. Suppression of accelerated tumor growth in surgical wounds by celecoxib and indomethacin. Head Neck 2005;27(4):326-32.

150

121. Sievers EM, Bart RD, Backhus LM, et al. Evaluation of cyclooxygenase-2 inhibition in an orthotopic murine model of lung cancer for dose-dependent effect. J Thorac Cardiovasc Surg 2005;129(6):1242-9.

122. Smakman N, Schaap N, Snijckers CM, Borel Rinkes IH, Kranenburg O. NS-398, a selective cyclooxygenase-2 inhibitor, reduces experimental bladder carcinoma outgrowth by inhibiting tumor cell proliferation. Urology 2005;66(2):434-40.

123. Yoshinaka R, Shibata MA, Morimoto J, Tanigawa N, Otsuki Y. COX-2 inhibitor celecoxib suppresses tumor growth and lung metastasis of a murine mammary cancer. Anticancer Res 2006;26(6B):4245-54.

124. Tanaka T, Nishikawa A, Mori Y, Morishita Y, Mori H. Inhibitory effects of non- steroidal anti-inflammatory drugs, piroxicam and indomethacin on 4-nitroquinoline 1- oxide-induced tongue carcinogenesis in male ACI/N rats. Cancer Lett 1989;48(3):177-82.

125. Wu WK, Sung JJ, Lee CW, Yu J, Cho CH. Cyclooxygenase-2 in tumorigenesis of gastrointestinal cancers: An update on the molecular mechanisms. Cancer Lett 2010;295(1):7-16.

126. Ko SH, Choi GJ, Lee JH, Han YA, Lim SJ, Kim SH. Differential effects of selective cyclooxygenase-2 inhibitors in inhibiting proliferation and induction of apoptosis in oral squamous cell carcinoma. Oncol Rep 2008;19(2):425-33.

127. Liu JF, Zhang SW, Jamieson GG, et al. The effects of a COX-2 inhibitor meloxicam on squamous cell carcinoma of the esophagus in vivo. Int J Cancer 2008;122(7):1639-44.

128. Naruse T, Nishida Y, Hosono K, Ishiguro N. Meloxicam inhibits osteosarcoma growth, invasiveness and metastasis by COX-2-dependent and independent routes. Carcinogenesis 2006;27(3):584-92.

129. Jiang X, Li H, Qiao H, Jiang H, Xu R, Sun X. Combining kallistatin gene therapy and meloxicam to treat hepatocellular carcinoma in mice. Cancer Sci 2009;100(11):2226- 33.

130. Xin B, Yokoyama Y, Shigeto T, Mizunuma H. Anti-tumor effect of non-steroidal anti-inflammatory drugs on human ovarian cancers. Pathol Oncol Res 2007;13(4):365-9.

131. Laimer K, Spizzo G, Gastl G, et al. High EGFR expression predicts poor prognosis in patients with squamous cell carcinoma of the oral cavity and oropharynx: A TMA- based immunohistochemical analysis. Oral Oncol 2006.

151

132. Martin-Ezquerra G, Salgado R, Toll A, et al. Multiple genetic copy number alterations in oral squamous cell carcinoma: Study of MYC, TP53, CCDN1, EGFR and ERBB2 status in primary and metastatic tumors. Br J Dermatol 2010.

133. Looper JS, Malarkey DE, Ruslander D, Proulx D, Thrall DE. Epidermal growth factor receptor expression in feline oral squamous cell carcinomas. Veterinary and Comparative Oncology 2006;4:33-40.

134. Dunford JE, Thompson K, Coxon FP, et al. Structure-activity relationships for inhibition of farnesyl diphosphate synthase in vitro and inhibition of bone resorption in vivo by nitrogen-containing bisphosphonates. J Pharmacol Exp Ther 2001;296(2):235-42.

135. Bi Y, Gao Y, Ehirchiou D, et al. Bisphosphonates cause osteonecrosis of the jaw- like disease in mice. Am J Pathol 2010;177(1):280-90.

136. Weinstein RS, Roberson PK, Manolagas SC. Giant osteoclast formation and long- term oral bisphosphonate therapy. N Engl J Med 2009;360(1):53-62.

152