ABSTRACT ANGSTADT, ANDREA Y. Evaluation of the Genomic

ABSTRACT

ANGSTADT, ANDREA Y. Evaluation of the Genomic Aberrations in Canine Osteosarcoma and Their Resemblance to the Human Counterpart. (Under the direction of Dr. Matthew Breen).

In the last decade the domestic dog has emerged as an ideal biomedical model of complex genetic diseases such as cancers. Cancer in the dog occurs spontaneously and several studies have concluded that human and canine cancers have similar characteristics such as presentation of disease, rate of metastases, genetic dysregulation, and survival rates. Furthermore, in the genomic era the dog genome was found more homologous in sequence conservation to humans than mice, making it a valuable model organism for genetic study in addition to pathophysiological analysis. Osteosarcoma (OS), the most commonly diagnosed malignant bone tumor in humans and dogs, is one such cancer that would benefit from comparative genomic analysis. In humans, OS is a rare cancer diagnosed in fewer than 1,000 people per year in the USA, while in the domestic dog population the annual number of new cases is estimated to far exceed 10,000. This high rate of disease occurrence in dogs provides a unique opportunity to study the genomic imbalances in canine OS and their translational value to human OS as a means to identify important alterations involved in disease etiology. OS in humans is characterized by extremely complex karyotypes which contain both structural changes (translocations and/or rearrangements) and DNA copy number changes. Metaphase and array comparative genomic hybridization (aCGH) has assisted in uncovering the genetic imbalances that are associated with human OS phenotype. In dog OS, previous low-resolution (10-20Mb) aCGH analysis identified a wide range of recurrent copy number aberrations (CNAs), indicative of a similar level of genomic instability to human OS. To further interpret chaotic OS karyotypes a genome-wide approach was taken to identify, characterize, and directly compare genomic instability in canine and human OS. For identification of genome-wide CNAs 123 cases of canine OS were profiled by 1Mb-resolution aCGH, 23 of the 123 cases were subsequently profiled by ~27kb-resolution aCGH and 15 cases of human OS were profiled by ~100kb-resolution aCGH. Subsequent fluorescence in-situ hybridization (FISH) analysis was used to confirm aCGH data, quantify numerical imbalances, and visualize structural abnormalities in a subset of dog OS cases. Characterization of the affect that CNA has on the expression of select cancer associated genes revealed that imbalance and transcriptional dysregulation in canine OS also paralleled human OS. Specifically, changes in RUNX2, TUSC3, and PTEN expression levels correlated with genomic copy number status in dog OS. This analysis showcased RUNX2 as an ‘OS associated gene’ and TUSC3 as a tumor suppressor gene involved in canine OS. In addition, direct comparison of genomic imbalance in human and dog OS using high resolution oligonucleotide aCGH indicated that the ‘OS associated genes’ RUNX2, CDKN2A/CDKN2B, MYC, RB1, and PTEN resided in orthologous microaberration regions (<500kb) with similar CNA patterns supporting that these genes are key genetic players driving OS progression. Similarities in genome-wide CNA patterns in OS between orthologous regions of the human and dog genome were also found suggesting that characterization of genes in these regions may identify additional alterations important for OS manifestation. Ultimately, this large scale screening of genomic imbalance in canine OS reiterates the value of the dog as a biomedical model of human OS while pinpointing key genes dysregulated in the disease in dogs. Upon further investigation, the genes with parallel CNA frequencies in human and dog OS may serve as possible targets of novel genetic therapeutics that once developed and tried in dogs could be translational to human patients.

Evaluation of the Genomic Aberrations in Canine Osteosarcoma and Their Resemblance to the Human Counterpart

by Andrea Y. Angstadt

A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Functional Genomics

Raleigh, North Carolina

2010

APPROVED BY:

______Dr. Matthew Breen Dr. Dahlia M. Nielsen Committee Chair

______Dr. David E. Malarkey Dr. Marlene L. Hauck DEDICATION

To my parents, Roy and Ann Young, for your love, guidance, and support

To my husband, Nathan, for joining me in this journey

ii BIOGRAPHY

Andrea developed her love for biology at an early age through hiking trips, raising farm animals, and grade school science fair projects. After graduating high school she went to Penn State University and completed a degree in animal bioscience with a minor in biology. During her time at Penn State she worked in a laboratory studying the genetic affect of external and internal factors on cacao plant growth and response as a means to improve plant productivity. She also completed an undergraduate research project identifying potential mutations in the leptin receptor gene associated with the diabetic and obese phenotype in the Butterball strain of mice. These projects increased her interest in genetics and its’ role in veterinary and human diseases. Therefore, she decided to continue her research training in the Functional Genomics Doctorate program at North Carolina State

University.

iii ACKNOWLEDGMENTS

I would like to thank my committee members, Dr. David Malarkey, Dr. Dahlia

Nielsen, and Dr. Marlene Hauck for providing support, discussions, and aid in my research progress. Thanks to Dr. Dahlia Nielsen and Dr. Alison Motsinger-Reif for helping me learn how to manage large datasets and conduct appropriate statistical analyses and to Dr. Eric

Stone for your excellent teaching of bioinformatics. Dr. Ted Emigh, it was a pleasure to TA for your genetic class and I appreciate the aid you gave in developing my teaching skills. To the Breen Lab through the years; Tessa Breen, Benoit Hedan, Shannon Becker, Rachael

Thomas, Christina Williams, Eric Seiser, Pei-Chen Tsai, Kate Kelley, Katie Kennedy, and

Kristen Maloney, thanks for making the work environment so much fun. I cherish all our discussions and will miss you all! To my father, my brother Hugh, and Erin Parker thanks so much for your editing skills and advice. Lastly, a special thanks to Dr. Matthew Breen for your direction and guidance through my graduate years and future endeavors.

iv TABLE OF CONTENTS

LIST OF TABLES...... viii

LIST OF FIGURES ...... x

LIST OF ABBREVIATIONS ...... xiii

Chapter I: Literature Review ...... 1

Overview...... 2 Dog as a suitable model of human disease ...... 3 The pathology of osteosarcoma ...... 9 Molecular and genetic dysregulation of osteosarcoma...... 13 Genomic Studies ...... 17 Thesis Outline ...... 31 References...... 32

Chapter II: Characterization of canine osteosarcoma by array comparative genomic hybridization and qRT-PCR: Signatures of genomic imbalance in canine osteosarcoma parallels the human counterpart ...... 49

Abstract...... 50 Introduction...... 51 Materials and Methods...... 53 Tissue Specimens...... 53 Array Comparative Genomic Hybridization (aCGH)...... 55 Fluorescence in-situ Hybridization (FISH)...... 56 Quantitative RT-PCR...... 57 Statistical Analysis...... 58 Results...... 59 Clinical Assessment...... 59 Abundant Genomic Instability in Canine OS ...... 60 FISH Validation of aCGH Data...... 65

v Breed and Morphological Subtype Specific Associated DNA Copy Number Aberrations ...... 68 qRT-PCR...... 70 Discussion...... 71 Acknowledgements...... 78 References...... 90

Chapter III: A genome-wide approach to comparative oncology: High-resolution oligonucleotide aCGH of canine and human OS pinpoints communal microaberrations ...... 101

Abstract...... 102 Introduction...... 103 Materials and Methods...... 107 Tissue Specimens...... 107 Array Comparative Genomic Hybridization (aCGH)...... 109 Array Comparative Genomic Hybridization Analysis...... 110 Identification of Orthologous Canine and Human Regions...... 111 Results...... 112 Refining Regions of Genomic Aberration in Canine OS...... 112 Extensive Genomic Imbalance in Human OS ...... 117 Comparative Regions of Genomic Imbalance in Canine and Human OS ...... 121 Discussion...... 125 References...... 131

Chapter IV: Conclusion ...... 141

Concluding remarks...... 142 Future evaluation of the complexity in canine OS...... 147 References...... 151

Appendices...... 158

vi Appendix I: Tackling the characterization of canine chromosomal breakpoints with an integrated in-situ/in-silco approach: The canine PAR and PAB...... 159

Abstract...... 160

Introduction...... 160

Materials and Methods...... 162 Clone Selection ...... 162 Cytogenetic Evaluation...... 162 Computational Analysis of Overlapping BAC Clones ...... 164 Results...... 164 Cytogenetic interpretation of the canine PAR ...... 164 Computational Analysis of the PAB...... 164 Comparison of Canine and Human PAR-Specific Genes...... 165 Discussion...... 167 References...... 168

Appendix II: Visualization of copy number changes in OS by FISH...... 170

vii LIST OF TABLES

Chapter I:

Table 1: Tumor incidence rate (all sites) estimated in pet dogs by country and tumor characteristics...... 7

Table 2: Previously defined molecular factors dysregulated in canine OS ...... 14

Table 3: Summary of the cytogenetic abnormalities associated with human OS...... 23

Table 4: Summary of highly recurrent genomic aberrations (≥30%) in 38 patients of canine OS indentified by low-resolution (10-20Mb) aCGH...... 29

Table 5: Summary of copy number imbalances for known cancer-associated genes in 38 patients of canine OS identified by low-resolution (10-20Mb) aCGH...... 29

Chapter II:

Table 1: Summary of breed and morphological subtype of 123 canine OS tumor samples analyzed for DNA copy number aberrations by genome integrated 1Mb-resolution aCGH ..54

Table 2: Primer sequence of genes analyzed by qRT-PCR analysis of canine OS RNA...... 58

Table 3: The top twenty autosomal regions that exhibited the highest percentages of copy number gain in our canine OS cases...... 64

Table 4: The top twenty autosomal regions that exhibited the highest percentages of copy number loss in our canine OS cases...... 65

Supplementary Table 1: Detailed list of clinical information concerning 123 OS tumor samples analyzed for DNA copy number aberrations by genome integrated 1Mb-resolution aCGH ...... 84

Supplementary Table 2: Genomic position and BAC ID for clones selected using the UCSC genome browser (http://www.genome.ucsc.edu/) ...... 88

Supplementary Table 3: Raw Ct values of non-neoplastic bone and 11 canine OS samples for reference gene c12orf43...... 89

viii Chapter III:

Table 1: Clinical summary of 23 canine OS tumor samples analyzed for DNA copy number aberrations (CNAs) on Agilent’s canine180,000 feature G3 SurePrint oligonucleotide aCGH platform...... 108

Table 2: Clinical summary of 15 human OS tumor samples analyzed for DNA copy number aberrations on Agilent’s 60,000 feature G3 SurePrint oligonucleotide aCGH platform...... 109

Table 3: Frequency of autosomal CNAs (gain and loss) identified by the ADM-2 algorithm with a threshold of 6 and log2 ratio cutoff of 0.2 in 23 cases of canine OS...... 116

Table 4: Frequency of CNAs (gain and loss) identified by the ADM-2 algorithm with a threshold of 6 and log2 ratio cutoff of 0.5 in 15 cases of human OS...... 120

ix LIST OF FIGURES

Chapter I:

Figure 1: Dog cancer incidence rates for females and males estimated from tumor specimens received by the ATR from local veterinarians in Genoa, Italy across the study period (1985- 2002) for all cancers and selected cancer sites by calendar period...... 9

Figure 2: SKY analysis of human OS...... 20

Figure 3: FISH analysis of dog OS as directed by aCGH...... 30

Chapter II:

Figure 1: Frequency of copy number aberration in 123 canine OS samples...... 62

Figure 2: Bivariate fit of CNA percent loss/gain for 123 canine OS cases by the physical size of the aberrations in Mb...... 63

Figure 3: Whole genome 1Mb aCGH profile of a 12 year old female Golden Retriever diagnosed with osteoblastic OS located in the left proximal humerus, co-hybridized with DNA derived from a blood sample from the same patient ...... 67

Figure 4: Further characterization of copy number aberration occurrences across 1,066 regions of aberration commonality in the 38 dog autosomes ...... 69

Figure 5: Quantitative RT-PCR on 26 canine OS patients for six genes located in regions that had a high occurrence of CNAs...... 71

Supplementary Figure 1: Kaplan-Meier survival curve comparing different treatment protocols for canine OS patients...... 79

x Supplementary Figure 4: Graphical representation of the PCA performed to evaluate global differences in genome-wide DNA copy number aberrations evident in our five breed groups ...... 82

Supplementary Figure 5: Graphical representation of the PCA performed to evaluate global differences in genome-wide DNA copy number aberrations evident in our three morphological subtypes ...... 83

Chapter III:

Figure 1: Array CGH profiles of OS12, an 8 year-old male Rottweiler with osteoblastic OS ...... 113

Figure 2: Frequency of copy number aberrations in 23 canine OS samples ...... 115

Figure 3: Box plots of frequency of autosomal CNAs (gain and loss) in 23 canine OS cases (array call resolution ~27kb) identified by the ADM-2 algorithm with a threshold of 6 and log2 ratio cutoff of 0.2...... 117

Figure 4: Array CGH profile (100kb resolution view) of HOS5, a 12-year old female with OS in the left distal femur...... 118

Figure 5: Frequency of copy number aberrations in 15 human OS samples...... 119

Figure 6: Box plots of frequencies of CNAs (gain and loss) in 15 human OS cases (array call resolution ~100kb) identified by the ADM-2 algorithm with a threshold of 6 and log2 ratio cutoff of 0.5...... 121

Figure 8: Chromosome profiles of CNA frequencies of 23 canine OS cases (~27kb resolution) and 15 human OS cases (~100kb resolution) ...... 124

Appendix I:

Figure 1: Schematic representation of the location of the primary BAC clones on CFA Xp and CFA Y, used to cytogenetically identify the PAB ...... 163

xi Figure 2: The sequence within BAC clones 172L08 and 397M10, canFam(chrX:6580407- 6630000), searched against 6.2 million unassembled shotgun sequences read from a male poodle genome using BLASTn...... 165

Figure 3: Position of PAR genes in 1Mb from the p-arm telomere and their comparative status in human (NCBI Build 36.1), horse (equCab1), dog (canFam2), and mouse (NCBI Build 37) according to the UCSC genome browser (http://www.genome.ucsc.edu/)...... 166

Appendix II:

Figure 1: Whole genome aCGH profile of an 8 year old male Great Pyrenees diagnosed with fibroblastic osteosarcoma located in the right proximal tibia co-hybridized with DNA derived from a blood sample from the same patient...... 171

Figure 2: Whole genome 1Mb aCGH profile of a 12 year old female Golden Retriever diagnosed with osteoblastic OS located in the left proximal humerus, co-hybridized with DNA derived from a blood sample from the same patient ...... 172

Figure 3: Interphase multicolor FISH analysis of three zinc fixed paraffin embedded canine OS tissues using BAC clone probe sets encompassing three genes (TSC2, MYC, and TUSC3) ...... 173

xii LIST OF ABBREVIATIONS

OS-osteosarcoma

BAC-bacterial artificial chromosome

CGH- comparative genome hybridization aCGH-array comparative genome hybridization

FISH-fluorescence in-situ hybridization

SKY- spectral karyotyping mBAND-multicolor banding

CNA-copy number aberration

CFA-Canis Familiaris

HSA-Homo sapiens

BLAST-basic local alignment and search tool

PAR- pseudoautosomal region

PAB- pseudoautosomal breakpoint

RB1-retinoblastoma 1

PTEN -phosphatase and tensin homolog

MYC-v-myc myelocytomatosis viral oncogene homolog (avian)

RUNX2- runt-related transcription factor 2

TUSC3- tumor suppressor candidate 3

CDKN2A/CDKN2B-cyclin-dependent kinase inhibitor 2A/2B

xiii

Chapter I

Literature Review

1 Overview

Cancer is a complex genetic disease, divided into over 100 subtypes. It presents with a wide range of symptoms that often result in extensive pain and suffering for an individual. Even in today’s society with the vast advances in health care it is still the second leading cause of death in humans, only topped by heart disease (Jemal et al. 2009). As cancer studies in humans are usually difficult because of ethical boundaries, selection of an appropriate model organism is extremely important. In the last decade the domestic dog (Canis familiaris, CFA) has emerged as an ideal model organism for the study of several human diseases including cancer because of strong anatomical, pathological, physiological, and genetic similarities (Paoloni and Khanna 2008).

One such cancer is osteosarcoma (OS), the most common primary malignant bone tumor in humans and dogs. Previous studies have identified similar characteristics between human and canine OS in the presentation of disease pathology, rate of metastases, genetic dysregulation, and survival rates (Brodey and Riser 1969; Berg et al. 1995; Bergman et al. 1996; van Leeuwen et al. 1997; Mendoza et al. 1998; Ferracini et al. 2000; Levine and Fleischli 2000; Kirpensteijn et al. 2002a; Bailey et al. 2003; Boston et al. 2006; Boston et al. 2007; Kirpensteijn et al. 2008; De Maria et al. 2009; Fieten et al. 2009; Fossey et al. 2009; McCleese et al. 2009). However, these studies were not able to identify genome-wide aberration similarities shared between canine and human OS because of the lack of genomic resources and tools. The genomic era saw an increase in the number of genomic tools for humans (Lander et al. 2001) as well as genomic sequencing of non-traditional model organisms, including the dog in 2005 (Lindblad-Toh et al. 2005). These newly developed genomic tools provided the ability to directly compare regions of the canine genome to the human genome and characterize parallels in genomic abnormalities in diseases like OS. Understanding the genetic mechanisms important in canine OS not only provides insight into cancer regulation and novel areas of therapeutics in dogs, but also in humans.

2 In this thesis I took a genome-wide approach to identify, characterize, and directly compare genomic instability in canine and human OS. I sought to demonstrate the value of the dog as a model organism through the identification of parallels in cytogenetic abnormalities in osteosarcoma (OS) while further evaluating and interpreting the disease aberrations in canines. I hypothesized that an array comparative genome hybridization (aCGH) study of the genomic imbalance at a higher resolution would reveal that the disease in the dog is as chaotic in nature as it is in the human counterpart. Evidence supporting this hypothesis is found in the previous low resolution aCGH (10-20Mb) study of canine OS, which concluded that many genomic regions were subject to copy number changes (Thomas et al. 2009). Thomas et al. (2009) also found that the pattern of copy number changes differed between dog breeds for specific cancer related genes within these regions. I believed this pattern would continue in the high-resolution analysis, suggesting that genomic imbalances may be result of an individual’s genetic framework. In order to properly interpret OS abnormalities I originally had planned to combine 1Mb-resolution bacterial artificial chromosome (BAC) aCGH, fluorescence in-situ hybridization (FISH), and computational analysis to determine precise locations of chromosome breakpoint region within aberrant OS karyotypes. I demonstrated the success of this approach in Appendix I by characterizing the naturally occurring breakpoint present on all canine X chromosomes - the pseudoautosomal breakpoint (PAB) (Young et al. 2008). Due to technological advancements I was subsequently able to enhance sections of this methodology by using oligonucleotide aCGH evaluation of both dog and human OS. Overall, this thesis demonstrates the translational potential of the dog as a model organism for human OS while further characterizing the chaotic genomic imbalance present in canine OS, enhancing our understanding of OS genetic abnormalities in man and his ‘best friend.’

Dog as a suitable biomedical model of human disease

One of the most important goals of biomedical research is to understand the genetic basis of human diseases. As many challenges, both ethical and physical, are present in

3 human-based studies the use of an appropriate model organism for the study of human diseases is important. In the last decade the domestic dog (Canis familiaris) has emerged as an ideal model organism for the study of several human diseases. According to the 2009/2010 National Pet Owners Survey conducted by the American Pet Products Manufacturers’ Association (APPMA) nearly half of US households (45.6%) are dog owners with the total number of pet dogs in the US estimated at approximately 77.5 million (http://www.americanpetproducts.org/press_industrytrends.asp). These dogs are susceptible to over 450 spontaneously arising diseases and approximately 360 of these are analogous to human diseases (Parker and Ostrander 2005; Wayne and Ostrander 2007; Shearin and Ostrander 2010). In addition, a subset of these 77.5 million dogs is divided into over 350 distinct breeds worldwide (Shearin and Ostrander 2010) with 167 breeds recognized in the USA by the American Kennel Club (AKC, http://www.akc.org/).

Distinct dog breeds provide the most compelling reason to use the dog as a model for spontaneously occurring human diseases because of the unique population structure formed during the domestication and development of the modern day purebred dogs. Each breed is defined by specific behavioral and physical characteristics that were driven to exceptionally high frequency by population bottlenecks and strong artificial selection (Karlsson and Lindblad-Toh 2008). The process of each breed arising from a limited number of founders and frequent use of popular sires has caused genetic diversity in dogs to be largely between breeds with limited locus and disease heterogeneity within breeds (Ostrander and Kruglyak 2000; Lindblad-Toh et al. 2005; Shearin and Ostrander 2010). In addition, since most breeds are fewer than 200 years old they have long stretches of linkage disequilibrium (LD) and long haplotype blocks, reducing the overall number of markers and individuals compared with humans needed to investigate genetic variants associated with morphological and disease variation (Sutter and Ostrander 2004; Lindblad-Toh et al. 2005). In turn, this makes genetic studies in dogs theoretically simpler and more straightforward (Shearin and Ostrander 2010) as the limited locus and disease heterogeneity in each closed population of purebred dogs offer several of the statistical advantages of studies performed in

4 geographically isolated human populations, such as those carried out in Finland or Iceland (Ostrander and Kruglyak 2000; Shearin and Ostrander 2010), as both environmental and genetic factors are somewhat controlled. Previous genome-wide analysis strategies have exploited the breed structure of dogs to investigate the genetic basis of specific traits in dogs. One study found an IGF-1 single nucleotide polymorphisms (SNP) haplotype suggestive of causing differences in body size in dogs (Sutter et al. 2007; Jones et al. 2008) and more recently 915 dogs from 80 domestic dog breeds were genotyped to discover 51 regions of the dog genome associated with phenotypic variation including average breed body size and external body dimensions (Boyko et al. 2010).

Domestic dog breeds have fascinated geneticists since the early 1900s when the first inherited characteristics and disorders such as coat color, hairlessness, barking, and hemophilia were published in the Journal of Heredity (Little 1914; Wright 1917; Whitney 1929; Hutt et al. 1948). As technology advanced, incorporation of cytogenetic techniques were introduced to the understanding of the genetics behind dog phenotypes (Karlsson and Lindblad-Toh 2008). In 1997 the first genetic linkage map of the dog genome (Mellersh et al. 1997), containing 150 microsatellite markers was published. In 1999 the gene associated with canine narcolepsy was identified and intensified interest in the dog as a model for human disease (Lin et al. 1999). The next few years saw the development of chromosome- specific single locus FISH probes for dogs (Breen et al. 2001; Thomas et al. 2001a; Thomas et al. 2001b; Breen et al. 2004), a 1Mb-resolution radiation hybrid map of the canine genome (Guyon et al. 2003), and survey sequencing and comparative analysis of the dog genome (Kirkness et al. 2003). In 2005, the Canine Genome Sequencing Project (Lindblad-Toh et al. 2005) further enhanced canine research by initiating the development of state-of-the-art tools from which to map traits in the dog. The availability of an assortment of genetic tools are useful in assisting researchers in defining phenotypes, selecting sample sets, identifying trait loci through linkage, association or selection mapping, identifying gene function and genomic variants, and testing likely variants for function (Karlsson and Lindblad-Toh 2008).

5 These tools also allow for the characterization of the genomic dysregulation that leads to disease phenotype in dogs (Lindblad-Toh et al. 2005; Karlsson and Lindblad-Toh 2008).

Although the dog had long been used as a model system for drug discovery and development research because of its similarities to human anatomy and physiology (Khanna et al. 2006), until the genomic age few had compared genetic disease similarities between dogs and humans. Analysis of the dog genome demonstrated that it was more homologous in sequence conservation to humans than mice, the foremost model for genetic studies in mammals since the beginning of the twentieth century (Kirkness et al. 2003; Lindblad-Toh et al. 2005). The mouse has been one of the most widely used model organisms and does boast an impressive availability of experimental resources that have been useful when testing the effect of genetic manipulation on single genes or interactions of a subset of genes (Paigen 1995; Bucan and Abel 2002). Yet in contrast to other vertebrates such as dogs, the mouse is not very useful when attempting to identify putative mutations in human diseases that are polygenic in nature. Also diseases in mouse are often induced while diseases in dogs occur spontaneously over the course of their lifetime and include several diseases also seen in humans such as cancers, diabetes, heart disease, eye diseases, epilepsy, deafness and even psychiatric disease such as obsessive compulsive disorder (Overall 2000; Khanna et al. 2006; Gershwin 2007; Withrow and Vail 2007).

In particular, the unique genetic architecture and the availability of genomic tools make dogs an ideal genetic model for the study of human cancers. As an artifact of the strong selection that breeders have imposed over the years to produce populations of dogs with common morphological and behavioral traits (Shearin and Ostrander 2010) an enrichment of cancer associated alleles has arisen within different breeds or closely related breeds. Dogs are susceptible to a variety of cancers that share strong pathophysiological similarities to human cancers such as non-Hodgkin lymphoma, osteosarcoma, melanoma, prostate carcinoma, lung carcinoma, mammary carcinoma, head and neck carcinoma, and soft-tissue sarcoma (Khanna et al. 2006). Although it is difficult to accurately quantify the

6 incidence rate of canine cancer, studies have shown that tumors occur at a frequency paralleling human neoplasms (Cadieu and Ostrander 2007; Breen and Modiano 2008). Studies conducted by European and North American veterinary cancer registries have sought to estimate the occurrence rate of spontaneous tumors in pet dogs (Table 1), but these studies should not be considered a complete representation of the occurrence of canine cancer as they have variable reference population. Although the Animal Tumor Registry (ATR) of Genoa, Italy did recently provide an accurate assessment of the occurrence rate of select canine cancers (Figure 1). The incidence rates for all canine cancers separated out for gender and for selected site-specific cancers by calendar period are shown in Figure 1 (A and B). The ATR found that all cancer incidences were three times higher in female dogs than in male dogs, but this difference can be attributed to the high rate of mammary cancer in females. Although there is variation in the incidence of canine cancer we can infer from these studies that in terms of epidemiology cancer is a common disease in dogs much like that seen in human populations.

Table 1: Tumor incidence rates (all sites) estimated in pet dogs by country and tumor characteristics (Merlo et al. 2008). aSize of the estimated canine population at risk bCrude rate defined as the number of cases per 100,000 dogs/year cAll dogs insured with a single UK pet insurance company dClaims for veterinary treatment identified as being related to neoplasia

Number of dogs Population estimated to be Incidence Country Years location at riska Tumors rateb 1997- UK insured UK 1998 dogsc 130,684 Cancersd 747.9 All neoplasia-related claimsd 1948 1963- California, USA 1966 Alameda County 1031 Cancers 381 Nonmalignant tumors 1130 Veterinary Ontario, Canada 1999 clinics 63,500 Cancers 850 All tumors 3970 1985- Italy 1994 Genoa County 127,600 Cancers 310 Nonmalignant tumors 760 Italy 2001 Local health unit 9812 Cancers 958.4

Along with the similar cancer morbidities, the ‘essence’ of the problem of human cancers is also captured by canine cancer by the characterization of tumor growth over long periods of time in the setting of an intact immune system and inter-individual and intra- tumoral heterogeneity. Similarly, the development of recurrent or resistant disease and metastasis to relevant distant sites such as the lungs and/or liver is also seen in dogs (Khanna et al. 2006). In addition, the manner in which dogs respond to tumor treatment is comparable to humans (Cadieu and Ostrander 2007) thus creating an opportunity to use dogs for timely assessment of new cancer therapies as the course of cancer progression is compressed in dogs (Khanna et al. 2006). Characterization of canine cancers at the genomic level and identification of specific causative mutations are the groundwork on which we can begin to develop new drugs. Inspecting individual canine cancers from multiple levels of genetic regulation such as the studies presented in chapter II and chapter III of this thesis provide more building blocks in which the collective genomics community can use to develop superior therapeutics for human and canine cancer.

Figure 1: Dog cancer incidence rates for (A) females and (B) males estimated from tumor specimens received by the ATR from local veterinarians in Genoa, Italy across the study period (1985-2002) for all cancers and selected cancer sites by calendar period. Vertical bands represent the 95% confidence intervals of the estimated incidence rates. (Adapted from Merlo et al. 2008)

The pathology of osteosarcoma

Osteosarcoma (OS) is a malignant mesenchymal neoplasm of primitive bone cells which produce an extracellular matrix of osteoid. This presence of tumor osteoid is the basis for histological diagnosis and differentiates OS from other sarcomas of the bone. In dogs, OS accounts for approximately 85% of all primary bone tumors, making it the most common primary bone tumor in dogs (Withrow and Vail 2007). The annual number of new cases of canine OS in the US is estimated to far exceed 10,000 (Withrow et al. 1991; Fossey et al. 2009). Even though fewer than 1,000 people every year in the US are diagnosed with OS it is still the most common primary bone tumor in humans (Mirabello et al. 2009b; Paoloni et al. 2009). Canine OS is almost exclusively observed in the large and giant breeds as shown by a

9 review of 1462 cases, which concluded that dogs weighing > 40 kg accounted for 29% of all cases whereas only 5% of the reported cases weighed < 15 kg (Withrow and Vail 2007). Other studies provide further support for the claim that large dog breeds such as Scottish Deerhounds, Saint Bernards, Great Danes, Irish Setters, Doberman pinschers, Rottweilers, German Shepherds, and Golden Retrievers, are more predisposed to develop OS (Brodey and Riser 1969; Brodey and Abt 1976; Straw et al. 1991; Ru et al. 1998; McNeill et al. 2007). Canine OS also tends to affect middle-aged to older dogs, with a median age of seven years, though studies have reported a bimodal distribution with a second small peak in young dogs aged between 18 and 24 months (Spodnick et al. 1992; Boston et al. 2006; Mueller et al. 2007; Withrow and Vail 2007). The majority of human OS cases occur in adolescence but there is a well described second peak of incidence in the elderly (seventh and eighth decades) (Mirabello et al. 2009a; Mirabello et al. 2009b). The occurrence of OS in large breeds as well as within pedigrees suggest an hereditary basis for the formation of canine OS (Withrow and Vail 2007). Previous publications reported male dogs to be more affected by the disease than females (up to 1.5:1) (Brodey and Riser 1969; Brodey and Abt 1976; Misdorp and Hart 1979; Spodnick et al. 1992) yet evaluation of 1775 cases of canine OS treated at Colorado State University between 1978 and 2005 found the male to female ratio to be equal (Withrow and Vail 2007).

Most cases of canine OS (75%) originate in the appendicular skeleton (humerus, femur, radius, tibia, and ulna) in the metaphysis of long bones (Liptak et al. 2004; Selvarajah and Kirpensteijn 2010) with the remainder occurring in the axial skeleton (flat bones of the skull, ribs, vertebrae, sternum, and pelvis) (Hammer et al. 1995; Dickerson et al. 2001). The front limbs are affected twice as often as rear limbs, with the distal radius (35%) and proximal humerus (18%) being the most common locations (Knecht and Priester 1978). A few cases of primary tumors arising at extra-skeletal sites have been described (Kuntz et al. 1998; Langenbach et al. 1998) even though OS is classified as a ‘sarcoma of the bone.’ Extra-skeletal OS is a rare neoplasm in older dogs (median age 11 yrs) and the survival prognosis for these dogs is poor (Patnaik 1990; Kuntz et al. 1998; Langenbach et al. 1998).

10 In human patients the disease most often arises in the metaphysis of long bones such as the distal femur, proximal tibia, and proximal humerus (Mirabello et al. 2009a; Mirabello et al. 2009b), which is similar to the physiological locations of canine OS.

In humans as well as dogs OS has an aggressive and invasive nature leading to local skeletal destruction as seen in radiographic evidence of both osteoproductive and osteolytic lesions. It is also highly metastatic, predominantly spreading to the lungs with a lower spread frequency to distant bones, regional lymph nodes (Hillers et al. 2005), and other soft tissues (Peremans et al. 2003; Gorman et al. 2006). Several histological subclassifications exist for OS that are based on the type and amount of matrix and characteristics of cancer cells (i.e. osteoblastic, fibroblastic, telangiectatic, chondroblastic and mixed forms) (Kirpensteijn et al. 2002a; Tang et al. 2008). Examination of the tumor’s histopathological features is the current diagnostic standard and there can be considerable variation in the histological appearance both between and within individual neoplasms making it important to obtain a histological analysis of the tumor following definitive excision. Several factors, such as alkaline phosphatase (ALP) levels, histological grade, and microvascular density, have also been identified as prognostic indicators in dogs, with the most important factor being the detection of metastases already present at the time of diagnosis (Mueller et al. 2007). Histologically, metastatic lesions usually appear identical to the primary tumor but they frequently exhibit a greater degree of necrosis (Selvarajah and Kirpensteijn 2010). A histopathological grading system has been applied to canine OS and it was concluded that tumors classified as grade III had a significantly poorer prognosis than grade I and II neoplasms (Kirpensteijn et al. 2002a). However, this grading system was not significantly prognostic for tumors from non-appendicular sites.

Although advancements in disease management for dogs have occurred over the years, treatment of OS has seen little advancement. Dogs that receive palliative care alone have a median survival time of 1-3 months, while those receiving surgery (usually amputation of the affected limb) with no chemotherapy have a median survival time of 1-6

11 months (Brodey and Abt 1976; Spodnick et al. 1992; Kirpensteijn et al. 2002a; Kirpensteijn et al. 2002b; Withrow and Vail 2007) and patients receiving a combination of amputation and chemotherapy treatment have the best prognosis; 50% 1-year and 20% 2-year survival rates (Mueller et al. 2007; Withrow and Vail 2007). ‘Limb-sparing’ is used only to selectively remove tumors located in the distal radius, ulna, and tibia (Straw and Withrow 1996; Boston et al. 2007) and even with removal of the primary tumor and chemotherapy treatment 90% still succumb to metastatic disease (Selvarajah and Kirpensteijn 2010). In human OS the introduction of new chemotherapy regimens in the 1980’s, which included treatment both before and after definitive surgical resection, helped improve the five-year survival rate to approximately 70%, but little improvement to this rate has been made in the last decade (Mirabello et al. 2009b)

Adjuvant therapy at specialized veterinary practices usually consists of multi-modal chemotherapy regimes, treatment with bisphosphonates or immune modulators and palliative radiation (Selvarajah and Kirpensteijn 2010). As stated earlier, combined amputation and chemotherapy contribute significantly to improving survival time and the chemotherapy agents usually consist of cisplatin, carboplatin, and doxorubicin. These agents have been used at varying dosages and treatment intervals (Berg et al. 1995; Bergman et al. 1996; Chun et al. 2000; Bailey et al. 2003), yet to date no differences in survival were seen when these treatments were compared with pre- or post-operative chemotherapy (Berg et al. 1997). In addition, no differences in disease-free-interval (DFI) have been reported for dogs treated using single- or multi-agent chemotherapeutic regimes. Once metastasis has been detected chemotherapy is often ineffective at improving survival beyond the short term (Brodey and Abt 1976; Kirpensteijn et al. 2002a; Kirpensteijn et al. 2002b). One study found that prolonged use of chemotherapy, specifically cisplatin, is often not an option as the harsh side-effects counteract any clinical benefit the canine patient may receive from the treatment (Barabas et al. 2008) and to date studies have not concluded if the growth of metastases is restricted with aggressive chemotherapy. There are a small number of canine OS cases that do not develop metastatic disease once the primary tumor has been removed without

12 chemotherapy treatment (Selvarajah et al. 2009), suggesting genetic composition of both patient and tumor play a role in prognostic outcome.

Molecular and genetic dysregulation of OS

Recent studies have suggested that specific genetic pathways may play an important role in the pathogenesis of osteosarcoma. These studies focused on characterizing known candidate genes (Table 2) implicated in the pathogenesis or progression of OS in dogs. Alterations of the tumor suppressor genes RB1 (retinoblastoma 1) (Araki et al. 1991) and TP53 (tumor protein p53) (Toguchida et al. 1992; McIntyre et al. 1994) are known to be involved in the pathogenesis of human OS. Past research suggest the role of RB1 gene abnormalities may not be as important in canine OS as analysis of 21 canine OS cases found no deletions or gross alterations of the gene (Mendoza et al. 1998). Another study reported that canine OS cell lines contained mutations that indirectly inactivated the three RB family members, RB1, p107, and p130, simultaneously (Levine and Fleischli 2000). Mutations of TP53 have been found in several canine OS studies (van Leeuwen et al. 1997; Johnson et al. 1998; Levine and Fleischli 2000). More recently it was suggested that these mutations predict patient survival (Kirpensteijn et al. 2008) as dogs with mutated TP53 had a significantly shorter survival time that correlated with elevated serum alkaline phosphatase concentrations and tumor histological grade. In addition, Zhang et al (Zhang et al. 2009) concluded that the canine TP53 family of proteins performs similar biological activities in OS as in human proteins. It remains unclear whether the TP53 mutation occurs in the metastases as well as in the primary tumor (van Leeuwen et al. 1997; Kirpensteijn et al. 2008) as other mutations in the tumor could contribute to tumor spread. Yet, mutations of TP53 may be an independent prognostic indicator and predictor of a more malignant phenotype of neoplasm which could be used in predicting patient outcome (Selvarajah and Kirpensteijn 2010). Another tumor suppressor gene, PTEN (phosphatase and tensin homolog) has been shown to be commonly mutated in the majority of canine cell lines and tumors evaluated. This mutation resulted in subsequent down-regulation of protein expression but no prognostic significance for this gene has been reported (Levine et al. 2002).

Table 2: Previously defined molecular factors dysregulated in canine OS. Gene/Protein Descriptive Name Comments Publications Not mutated, but cell lines contained mutations that Mendoza et al. 1998; RB1 retinoblastoma 1 indirectly inactivated genes Levine and Fleischli 2000 Mutated and/or van Leeuwen et al. 1997; overexpressed, protein Johnson et al. 1998; perform similar biological Levine and Fleischli activites to those seen in 2000; Kirpensteijn et al. TP53 tumor protein p53 human OS 2008; Zhang et al. 2009 Mutated or downregulated in cell lines and tumors, no phosphatase and prognostic significance PTEN tensin homolog reported to date Levine et al. 2002 v-erb-b2 erythroblastic leukemia viral oncogene homolog Overexpressed in several 2/epidermal growth canine OS lines and primary erbb-2/HER-2 factor receptor 2 tumors Flint et al. 2004 insulin-like growth role in tumor cell invasion factor-1/insulin-like and growth, previous clinical growth factor-1 trial supression did not MacEwen et al. 2004; IGF-1/IGF-1R receptor improve outcome Khanna et al. 2002 Mesenchymal- epithelial Expressed in lung metastases transition/hepatocyte and expression in primary Ferracini et al. 2000; growth factor tumor thought to lead to Fieten et al. 2009; De MET/HGFR receptor metastases Maria et al. 2009 platelet-derived growth factor beta polypeptide (simian Low level overexpression in sarcoma viral (v-sis) canine OS cell lines and Kochevar et al. 1990; sis/PDGF oncogene homolog) primary tumors Levine et al. 2002 Low level overexpression in canine OS cell lines and v-myc primary tumors, recent myelocytomatosis research saw no significant viral oncogene changes in expression Kochevar et al. 1990; MYC homolog (avian) relative to osteoblasts Levine et al. 2002 Upregulated in canine OS but undergos alternative Telomerase reverse splicing, need more study Kow et al. 2008; TERT transcriptase gene into therapuetic value Angelopoulou et al. 2008

14 Table 2: Continued Gene/Protein Descriptive Name Comments Publications member of the ezrin-radixin- moesin (ERM) High expression associated Ezrin protein family with metastatic phenotype Khanna et al. 2004 High concentrations in three cell lines and 30 primary tumor tissues, may be responsible for local disease progression and mestastasis but recent trial of inhibition Loukopoulos et al. 2004; matrix did not improve patient Lana et al. 2000; Moore MMPs metalloproteinases outcome et al. 2007 Upregulated in some canine OS, prognostic value, inhibition in cell line caused Mullins et al. 2004; COX-2 Cyclooxygenase-2 loss of cellular growth Wolfesberger et al. 2006

Along with tumor suppressor genes several proto-oncogenes have been identified as playing a role in producing the canine OS phenotype. V-erb-b2 erythroblastic leukemia viral oncogene homolog 2 (erbB-2) which encodes for epidermal growth factor receptor 2 (HER- 2) is thought to be important in tumor transformation and it was found overexpressed in 86% (6/7) of canine OS cell lines and 40% (4/10) of primary tumor samples (Flint et al. 2004). In human OS the role of erbB-2 is controversial as in some studies over-expression has been found and it was related to poor prognosis (Gorlick et al. 1999; Zhou et al. 2003) while other studies found no over-expression of HER-2/neu in the tumors (Maitra et al. 2001; Anninga et al. 2004). The sis/PDGFB oncogene was found over-expressed in a subset of four canine OS cell lines and nine primary tumors (Kochevar et al. 1990; Levine 2002) suggesting a possible autocrine growth factor loop involved in the pathogenesis. Within these studies slightly elevated levels of expression of MYC in primary tumors and cell lines was also indentified (Kochevar et al. 1990; Levine 2002) and this phenomenon has also been found in human OS (Gamberi et al. 1998; de Nigris et al. 2007). Although a recent study that assessed the relative expression of 16 genes including MYC as potential biomarkers of human OS

15 oncogenesis and chemotherapy response showed no significant changes in MYC expression in tumors relative to normal osteoblasts suggesting that MYC is not dysregulated at a transcriptional level in human OS (Sadikovic et al. 2010). Mesenchymal-epithelial transition (MET) is another proto-oncogene and encodes a protein known as c-Met or hepatocyte growth factor receptor (HGFR). MET can activate oncogenic pathways and participate in angiogenesis and metastases in neoplasias. Expression of MET has been found in canine OS lung metastases (Ferracini et al. 2000) and expression of MET in primary tumors was found to predict metatasis via lymphatics (Fieten et al. 2009). MET mRNA is expressed in both human and canine OS cell lines and small molecular inhibitors of MET have been shown in vitro to impair the invasive properties of canine OS cells suggesting the gene as an avenue of study for future treatment options (De Maria et al. 2009).

There are a few genes linked to OS that are not specifically classified as tumor suppressor genes or oncogenes. One such gene, the telomerase reverse transcriptase gene (TERT), is responsible for catalyzing the addition of telomeric sequences onto the 3’-ends of chromosomes de novo and is involved in cell proliferation and tumorigenesis in canine OS. The majority of canine OS patients are telomerase positive suggesting that it may be a valuable target for canine OS therapy (Kow et al. 2008). A recent study, however, concluded that dog TERT undergoes alternative splicing and the different isoforms may have implications in gene regulation. Thus, more studies would need to be conducted on the regulation of telomerase activity in canine normal and OS cells in order to properly understand the role of this gene in the disease (Angelopoulou et al. 2008).

In addition to gene based studies of canine OS, several groups have analyzed protein dysregulation in the disease. The matrix metalloproteinases (MMPs) are zinc-dependent enzymes that may be partially responsible for local tumor progression and metastasis. The MMPs have been implicated as biomarkers of shorter disease free interval (DFI) in human OS (Uchibori et al. 2006) as well as a predictor of survival following neo-adjuvant chemotherapy (Foukas et al. 2002). In canine OS, MMP 2 and MMP 9 were found in high concentrations in three canine OS cell lines (Loukopoulos et al. 2004) and 30 primary canine

16 OS cases had greater MMP expression in tumor than stromal cells, suggesting the proteins may be partially responsible for local disease progression and metastasis (Lana et al. 2000). Yet, a recent study that inhibited MMP-2 expression using BAY 12-9566, a non-peptide biphenyl MMP inhibitor, in combination with doxorubicin chemotherapy in 303 dogs with OS concluded that the inhibition of MMP-2 did not improve survival rates (Moore et al. 2007). The insulin-like growth factor-1 (IGF-1) and its receptor IGF-1R have been shown to play a significant role in OS cell growth and invasion of canine and human OS cell lines (MacEwen et al. 2004). Yet, a previous randomized blind, placebo-controlled preclinical study in dogs with OS concluded that suppression of IGF levels did not sufficiently improve patient outcome (Khanna et al. 2002). Cyclooxygenase (COX-2), an enzyme involved in apoptosis, is upregulated in canine appendicular OS and increased expression was correlated with a significant decrease in survival time for patients relative to tumors that had minimal expression of the enzyme (Mullins et al. 2004). Analysis of the affect of COX-2 inhibition by meloxicam, a non-steroidal anti-inflammatory drug, on the canine OS cell line D-17 found that meloxicam induced a time- and dose-dependent inhibition of cellular growth (Wolfesberger et al. 2006), yet more studies are needed to conclude its potential for treatment of osteosarcoma in dogs. High expression of another protein, ezrin, a member of the ezrin- radixin-moesin (ERM) family, is associated with early development of metastases in canine OS patients. In pediatric OS, patterns of high expression are linked to poor clinical outcomes (Khanna et al. 2004).

Genomic Studies

Several analyses have noted that canine OS is highly comparable to its human counterpart with respect to disease histology, high metastatic rate, poor long-term survival, and the dysregulation pattern of known tumor-associated genes and/or proteins. Yet, until recently the comparability of genomic alterations in canine OS to human OS had not been explored. The completion of high quality annotated assemblies for both the human and canine genomes (Lander et al. 2001; Kirkness et al. 2003; Lindblad-Toh et al. 2005) has

17 made it easier to assess changes in the dog and make direct genomic comparisons between the two.

In human OS, extensive research has identified the cytogenetic and molecular abnormalities, presenting a thorough knowledge basis from which to compare aberrations to canine OS. In contrast to other human sarcomas conventional OS is not associated with a specific recurrent translocation or any other chromosomal rearrangement. In fact, it is characterized by an array of sequential and well-orchestrated genetic changes that involve numerous tumor-suppressor genes and oncogenes (Sandberg and Bridge 2003). Cytogenetically, human OS shows markedly abnormal karyotypes that contain complex structural changes (translocations and/or rearrangements) and DNA copy number changes (Bayani et al. 2007). Evaluation of human OS karyotypes using a combination of comparative genomic hybridization (CGH), spectral karyotyping (SKY), and multicolor banding (mBAND) has identified extensive and frequent genome reorganization, including a high degree of aneuploidy, gene amplification, and multiple unbalanced chromosomal rearrangements (Bridge et al. 1997; Zielenska et al. 2001; Ozaki et al. 2002; Bayani et al. 2003; Ozaki et al. 2003; Lim et al. 2004; Man et al. 2004; Lim et al. 2005; dos Santos Aguiar et al. 2007; Selvarajah et al. 2008; Sadikovic et al. 2009). The ploidy number in human OS ranges from haploidy to near-hexaploid and fewer than 5% of the human OS tumors that contain copy number aberrations (CNAs) had only single karyotype changes (Sandberg and Bridge 2003). An example of SKY characterization of the chaotic karyotypes in human OS is shown in figure 2 (Bayani et al. 2003). This study showed that most karotypes were pseudotriploid for all chromosomes, or had loss or gain of several chromosomes. The overall findings revealed a high frequency of centrometric rearrangements and recurrent aberrations of human 8q23-q24, 17p11-p13, and 20q (Bayani et al. 2003) (Figure 2A,C).

In addition to SKY analysis, CGH (both metaphase and array based) methods have detected extensive numerical imbalance in human OS. At a limited resolution metaphase

18 CGH identified numerous regions of copy number changes (Lau et al. 2004; Atiye et al. 2005; Ohata et al. 2006; dos Santos Aguiar et al. 2007), but development of array CGH technology over the last decade has allowed researchers to ‘zoom in’ on CNAs by enhancing the genomic resolution and ability to identify subtle imbalances such as microdeletions (Squire et al. 2003; Selvarajah et al. 2007). Looking across publications seeking to characterize the genomic instability in human OS, overlapping regions of copy number changes could be identified (Table 3). The regions with large overlap are located within human chromosome bands 6p22-p21, 8q24 and 17p13.2-p11.2 and are consistently involved

in high copy number gain (log2 ratio of tumor:reference ≥ 1.0) or amplification (log2 ratio of tumor:reference ≥ 2.0) (Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Batanian et al. 2002; Bayani et al. 2003; Man et al. 2004; dos Santos Aguiar et al. 2007; Selvarajah et al. 2008; Sadikovic et al. 2009; Yang et al. 2010). These large regions of recurrent genomic amplification are believed to play an important role in tumor progression (Sadikovic et al. 2009).

Figure 2: SKY analysis of human OS. (A-C) All aberrations, the inverted DAPI and RGB (red-green-blue) images of the fluorescently labeled chromosomes are illustrated. (D) Only SKY-classified colors are illustrated. (A) Aberrations involving chromosome 8 in four primary OS tumors and three cell lines. (B) Evolution of new rearrangements of der(4)t(4;7)(p16;?p15) in primary tumor OS10. (C) Aberrations involving chromosome 20 in eight primary OS tumors and two cell lines. (D) Complete karyotypes from SKY analysis of primary tumor 0S9 (Bayani et al. 2003).

In the last few years, some investigators have sought to identify the effect of cytogenetic abnormalities in human OS on tumor progression through evaluation of both copy number and expression changes. In turn they have showcased specific genes that they believe to be important in OS because of dysregulation at both the numerical and transcriptional level. One study used a region specific array for human chromosome 6p12-21 with 108 overlapping bacterial artificial chromosomes (BACs) and P1 artificial clones (PACs), covering a 28.8Mb region at 0.26 Mb intervals (Lu et al. 2008). This array was able to refine the 6p amplicon to 7.9 Mb in 26 primary OS tumors and identify target genes (MAPK14, MAPK13, CDKN1A, PIM1, MDGA1, BTB9, DNAH8, CCND3, PTK7, CDC5L, and RUNX2). Through qRT-PCR analysis of these genes and protein screening by western

20 blotting and immunohistochemistry indicated that CDC5L, a cell cycle regulator important

for G2-M transition, was a likely candidate oncogene for the 6p12-21 amplicon in human OS (Lu et al. 2008). A similar study profiled 36 primary tumors and 20 cell lines with a 1Mb resolution genomic microarray that consisted of 4,549 BAC and PAC clones as well as the minimal tiling-path between 1q12 and the beginning of 1q25 (Kresse et al. 2009). This study revealed that a small region in 3q13.31 (2.1Mb) was frequently deleted (56%). This region was validated by FISH and contained the LSAMP (limbic system-associated membrane protein) gene which has been reported as a candidate tumor suppressor gene in other human cancer types (Kresse et al. 2009). In addition qRT-PCR analysis found that expression of LSAMP was downregulated and that lower expression was associated with poor survival (Kresse et al. 2009).

Over the last three years researchers have begun to use oligonucleotide based aCGH technology to identify and characterize genetic abnormalities in human OS. Using the Human Genome CGH 44K microarray (Agilent Technologies) to analyze ten tumor specimens Maire et al. were able to identify CNAs spanning the entire human genome at a median resolution of 75kb (Maire et al. 2009). They investigated the expression level of RECQL4, a gene that is mutated in patients with Rothmund-Thomson syndrome that develop OS and also maps to the cytoband 8q24 (often rearranged or gained in copy number) (Maire et al. 2009). There was no evidence to indicate that disruption of 8q24 in OS led to the elevated expression of RECQL4, but FISH and aCGH analysis of the tumor DNA showed that expression levels were strongly copy number-dependent (Maire et al. 2009). The same Human Genome CGH 44K array (Agilent Technologies) was used to evaluate the role of the WWOX gene in 10 human OS cases in combination with immunohistochemical staining of 55 formalin-fixed and paraffin embedded tissues (Yang et al. 2010). Loss of WWOX is associated with several human cancers and the protein was recently reported to play a critical role in bone metabolism by suppressing RUNX2 transactivation in osteoblasts resulting in inhibition of normal differentiation. Deletion of WWOX was found in 30% (3/10) of cases and the protein was undetected in 62% (34/55) of the paraffin samples (Yang et al. 2010).

Sadikovic et al. were the first to demonstrate combined analysis of multiple levels of dysregulation in human OS by using novel cytogenetic and gene regulation molecular tools (Sadikovic et al. 2009). They used an integrative approach for genome-wide high resolution profiling of genetic, epigenetic, and gene expression to study the combined-effects of genetic and epigenetic mechanisms on cancer-related gene networks (Sadikovic et al. 2009). This approach was then applied to a panel of primary human OS tumors in order to characterize the correlation between global changes in copy number, DNA methylation, and gene expression. They reported strong associations between copy number aberration, over expression, and hypomethylation leading to disease progression (Sadikovic et al. 2009). To date this integrative approach has not been completed on canine OS, but some studies have looked at specific aspects of whole genome genetic and epigenetic dysregulation. One study using gene expression profiling on a canine specific cDNA microarray identified 51 transcripts to be differently expressed in canine OS with common upregulation of these genes in patients with a short (<6 months) survival time. These over-expressed genes were associated cell proliferation, drug resistance or metastasis, and were involved in known tumor deregulated pathways in human OS including Wnt signaling, integrin signaling, and Chemokine/cytokine signaling (Selvarajah et al. 2009). A separate study used parallel oligonucleotide array platforms to make a direct comparison of canine (15 primary tumors and two cell lines) and pediatric osteosarcoma (15 primary tumors and three cell lines) expression profiles (Paoloni et al. 2009). Shared orthologous regions between species were identified and normalized, and it was concluded that OS expression signatures between the canine and human samples were indistinguishable by hierarchical clustering (Paoloni et al. 2009). This study strongly supported the similarities between human and canine OS and the value in taking a comparative oncology approach to improve our understanding of cancer biology.

22 Table 3: Summary of the cytogenetic abnormalities associated with human OS.

Human Copy Chromosomal Number regions Aberration Publications Start End Zielenska et al. 2001; Batanian et al. 2002; Man et al. 2004; Atiye et al. 2005; dos Santos Aguiar et al. 2007; Yang et al. 1p36.32 gain 2010 2,300,001 5,400,000 Ozaki et al. 2002; Batanian et al. 2002; Atiye et al. 2005; Ohata et al. 2006; Kresse et al. 2009; Sadikovic et al. 2009; Yang et al. 1q21.1-q21.3 gain 2010 142,600,001 155,000,000 Ozaki et al. 2002; Batanian et al. 2002; Atiye et al. 2005; Ohata et al. 2006; Kresse et al. 2009;Yang 1q21.3-q22 gain et al. 2010 150,300,001 156,500,000 Zielenska et al. 2001; Man et al. 1q25.1 loss 2004 172,900,001 176,000,000 1q43-44 loss Ohata et al. 2006 236,600,001 249,250,621 2p gain dos Santos Aguiar et al. 2007 1 90,500,000 Zielenska et al. 2001; Ozaki et 2q34-qter loss al. 2002, Lau et al. 2004 209,000,001 243,199,373 3q gain dos Santos Aguiar et al. 2007 1 87,900,000 3p14.1 loss Lau et al. 2004; Man et al. 2004 63,700,001 69,800,000 Batanian et al. 2002; Kresse et 3q13.31 loss al. 2009 113,500,001 117,300,000

23 Table 3: Continued Human Copy Chromosomal Number regions Aberration Publications Start End 4p15.1 loss Man et al. 2004 27,700,001 35,800,000 4p14 gain Ohata et al. 2006 35,800,001 41,200,000 4p12-p13 gain Ohata et al. 2006 41,200,001 48,200,000 Ozaki et al. 2002; Batanian et al. 4q13-q21 gain 2002; Ohata et al. 2006 59,500,001 88,000,000 4q22-q23 gain Ohata et al. 2006 88,000,001 101,100,000 4q24-q26 gain Ohata et al. 2006 101,100,001 120,800,000 Ozaki et al. 2002; Zielenska et al. 2001; Lau et al. 2004; dos Santos 5p14-p15.2 gain Aguiar et al. 2007 15,000,000 28,900,000 Ohata et al. 2006; dos Santos 5q11.2 gain Aguiar et al. 2007 50,700,001 58,900,000 5q12.3-q13.2 loss Yang et al. 2010 63,200,001 73,300,000 5q14.3-q22.2 loss Yang et al. 2010 82,800,001 113,100,000 5q32-q33.1 loss Ozaki et al. 2002 144,500,001 152,700,000 5q34 loss Ohata et al. 2006 159,900,001 168,500,000 Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Batanian et al. 2002; Bayani et al. 2003; Squire et al. 2003; Man et al. 2004; dos Santos Aguiar et al. 2007; Selvarajah et al. 2008; Yang 6p22-p21 gain et al. 2010 15,200,001 46,200,000 Ozaki et al. 2002; Man et al. 2004; Lau et al. 2004; dos Santos Aguiar 6p21-p12 gain et al. 2007; Lu et al. 2008 30,400,001 57,000,000 Atiye et al. 2005; Sadikovic et al. 6p12.3-p21.1 gain 2009 40,500,001 51,800,000

6q12 loss Man et al. 2004 63,400,001 70,000,000

24 Table 3: Continued Human Copy Chromosomal Number regions Aberration Publications Start End 6q15-q16 loss Ohata et al. 2006 88,000,001 105,500,000

6q16.1 loss Ozaki et al. 2002; Kresse et al. 2009 93,100,001 99,500,000

6q16.3 loss Ozaki et al. 2002; Man et al. 2004 100,600,001 105,500,000 6q21 loss Ohata et al. 2006 105,500,001 114,600,000 6q22 loss Ohata et al. 2006 114,600,001 130,300,000

7p21 gain Ozaki et al. 2002 7,300,001 20,900,000 8p23-p22 loss Ohata et al. 2006 1 19,000,000 Ozaki et al. 2002; Ohata et al. 2006; 8p21-p21.3 loss Sadikovic et al. 2009 19,000,001 23,300,000

8q11 gain Ohata et al. 2006 45,600,001 55,500,000

8q12-q13 gain Ohata et al. 2006 55,500,001 73,900,000 Ozaki et al. 2002; Batanian et al. 2002; Man et al. 2004; Lau et al. 2004; Ohata et al. 2006; Kresse et 8q22.1 gain al. 2009 93,300,001 99,000,000

Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Batanian et al. 2002; Bayani et al. 2003; Squire et al. 2003; Man et al. 2004; Lau et al. 2004; Atiye et al. 2005; Ohata et al. 2006; Selvarajah 8q24 gain et al. 2008; Maire et al. 2009 117,700,001 146,364,022 Ozaki et al. 2002; Ohata et al. 2006; 9p21 loss Lau et al. 2004 1 47,300,000

25 Table 3: Continued Human Copy Chromosomal Number regions Aberration Publications Start End

Ozaki et al. 2002; Lau et al. 10p loss 2004; Ohata et al. 2006 1 38,000,000 Zielenska et al. 2001; Ozaki et al. 2002; Ohata et al. 2006; 10q22.1-q22.2 loss Kresse et al. 2009 70,600,001 77,700,000

10q25-q26 loss Ohata et al. 2006 105,800,001 135,534,747

11q13 gain Lau et al. 2004 63,400,001 77,100,000

11q23 gain Lau et al. 2004 110,400,001 121,200,000 12p gain Ozaki et al. 2002 1 33,300,000 Man et al. 2004; Lau et al. 2004; Atiye et al. 2005; Lu, et 12q13-q15 gain al. 2008; Kresse et al. 2009 46,400,001 71,500,000 Zielenska et al. 2001; Man et al. 2004; Lau et al. 2004; Ohata 13q12.2 loss et al. 2006 27,800,001 28,900,000 Ozaki et al. 2002; Ohata et al. 13q14.2 loss 2006; Kresse et al. 2009 47,300,001 50,900,000 Man et al. 2004; Lau et al. 2004; Kresse et al. 2009, Ohata et al. 2006; Ohnstad et al. 2009; 13q21.1 loss Yang et al. 2010 55,300,001 59,600,000 14q11.2 loss dos Santos Aguiar et al. 2007 19,100,001 24,600,000

Lau et al. 2004; Atiye et al. 14q32 loss 2005; Ohata et al. 2006 89,800,001 107,349,540 Lau et al. 2004; dos Santos 15q loss Aguiar et al. 2007 1 15,800,000 Zielenska et al. 2001; Man et 16p13 gain al. 2004 1 16,800,000

26 Table 3: Continued Human Copy Chromosomal Number regions Aberration Publications Start End Ozaki et al. 2002, Batanian et al. 2002; Lau et al. 2004; Ohata et al. 16p12-p13.1 loss 2006; dos Santos Aguiar et al. 2007 10,500,001 28,100,000 Ozaki et al. 2002; Lau et al. 2004; 16q23-q24 loss Ohata et al. 2006 74,100,001 90,354,753 17p13 loss Ohata et al. 2006 1 10,700,000 Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Ozaki et al. 2002; Batanian et al. 2002; Bayani et al. 2003; Squire et al. 2003; Man et al. 2004; Lau et al. 2004; Atiye et al. 2005; Selvarajah et al. 2008; Lu et al. 17p13.2-p11.2 gain 2008 3,300,001 22,200,000

17q25 gain Atiye et al. 2005 70,900,001 81,195,210

18q12-q22 loss Ohata et al. 2006 25,000,001 73,100,000

18q22 gain Lau et al. 2004 61,600,001 73,100,000 18q23 loss Ohata et al. 2006 73,100,001 78,077,248

Zielenska et al. 2001; Ozaki et al. 19p13.11-p13.2 gain 2002; Yang et al. 2010 6,900,001 20,000,000

20p11.2 gain Lau et al. 2004 17,900,001 25,600,000

20q gain Ozaki et al. 2002; Lau et al. 2004 29,400,001 63,025,520 Ozaki et al. 2002, Batanian et al. 21q gain 2002; Atiye et al. 2005 14,300,001 48,129,895

21q11.2-q21 loss dos Santos Aguiar et al. 2007 14,300,001 31,500,000

22q11-q13 gain Atiye et al. 2005 14,700,001 51,304,566

22q13 loss Ohata et al. 2006 37,600,001 51,304,566

27 While many expression-based studies of canine OS have been completed, analysis of genomic CNAs in canine OS has only recently been done. Low-resolution (10-20Mb) aCGH analysis of 38 cases of canine OS consisting of two breeds (Rottweilers and Golden Retrievers) identified a high degree of genomic instability that resulted in extensive cytogenetic disorganization, as is the case in human OS (Thomas et al. 2009). An example of this genomic instability is shown in figure 3, FISH analysis of 10 BAC clones on metaphase and interphase nuclei from a case of osteoblastic OS in a female Rottweiler (Thomas et al. 2009). The metaphase spreads contained numerous bi-armed metacentric chromosomes which are likely the result of fusion events between acrocentric chromosomes. This figure also shows the extent of the copy number aberrations concluded from the aCGH analysis as MYC has upwards of 11 copies and PTEN is a homozygous deletion. Summarization of the aCGH data regions in 38 cases identified at least 30% copy number gain or loss (table 4). In addition to defining regions of recurrent genomic imbalances Thomas et al. cataloged the copy number imbalances in 11 known oncogenes and tumor suppressor genes (table 5). Of further interest, a statistically significant difference characterizing the deletion of the WTI gene was found between Golden Retrievers and Rottweilers. None of the nine Golden Retriever’s contained deletion of this gene in their array profile while WT1 deletion occurred in 48% (14/29) of Rottweilers (Thomas et al. 2009). This study was the first to demonstrate the extent of cytogenetic imbalance in canine OS, greatly supporting the importance of more accurate interpretation of the extent and comparability of genomic aberrations in canine OS to human OS.

28 Table 4: Summary of highly recurrent genomic aberrations (≥30%) in 38 patients of canine OS indentified by low-resolution (10-20Mb) aCGH. (Adapted from Thomas et al. 2009) Canine Chromosomal Position Copy Number Region (Mb) Aberration 1q36 116.7 gain 13q13 28.3 gain 13q14 29.4 gain 13q21.1 36.1 gain 13q21.3 50.1 gain 16q12prox 13.6 loss 16q21mid 30.6 loss 18q21-22.1 21.2 gain 18q22.3 45.7 loss 26q24-q25 28.1 loss 26q25 39.9 loss 29q12 13 loss 31q15.3 37.6 gain 35q14-q15 12.5 loss

Table 5: Summary of copy number imbalances for known cancer-associated genes in 38 patients of canine OS identified by low-resolution (10-20Mb) aCGH. (Adapted from Thomas et al. 2009) Canine Chromosomal Position Copy Number Region (Mb) Aberration Gene 5q21 35.5 15% gain/20% loss TP53 7q23 70.0 10% gain/8% gain YES1 11q14-q15 35.0 10% gain/10% loss CDKN2A 13q13 28.3 40% gain/5% loss MYC 13q21.3 50.1 30% gain/1% loss KIT 15q24.3 54.3 15% gain/10% loss CDK4 18q21disr-22.1prox 21.2 40% gain/20% loss HRAS 18q22.3 45.7 5% gain/40% loss WT1 22q11.2 6.0 1% gain/30% loss RB1 26q24disr-q25 39.9 0% gain/42% loss PTEN 31q11-q12 6.5 5% gain/2% loss MDM2

Figure 3: FISH analysis of dog OS as directed by aCGH. (a) & (b) Multicolor FISH analysis of 10 BAC clones on metaphase (left) and interphase nuclei (right) from a case of osteoblastic OS in a female Rottweiler. (c) Presence of an isochromosome of CFA 5 as shown by resulting probe hybridization pattern. (d) Compilation of copy number data, based on FISH analysis demonstrated in (a) and (b). (Thomas et al. 2009).

30 Thesis Outline

In the following chapters of this thesis I enhanced the previous study of genomic imbalance in canine OS (Thomas et al. 2009) by conducting aCGH analysis; at 1Mb- resolution on 123 cases of canine OS and at ~27kb resolution for 23 of the 123 cases. Specific dog breeds, Greyhounds, Golden Retrievers, Rottweilers, and Great Pyrenees, were recruited to identify if any regions of genomic instability were associated with genetic background. Subsequent metaphase and interphase fluorescence in-situ hybridization (FISH) analysis were performed on cells from select canine OS patients to visualize and calculate the extent of copy number gain and loss. My aCGH analysis found that previously defined cancer related genes resided in genomic regions of high aberration frequency on CFA 6, CFA 9, CFA 12, CFA 13, CFA 16, and CFA 26. I then determined the effect of copy number aberration (CNA) on the transcriptional regulation of these genes. Lastly, 15 cases of human OS were profiled at ~100kb resolution to identify orthologous regions of CNAs between dogs and humans. In summary, this thesis (1) characterizes the genomic imbalance in canine OS, (2) concludes that ortholgous regions between humans and canines have similar patterns of CNAs, and (3) identifies cancer-associated genes in regions of high aberration frequency warranting future evaluation in canine and human OS. The parallels that I found in the following chapters between genetic aberrations in canine and human OS help to identify recurrent cross-species abnormalities that could be driving cancer progression and development and suggest that new preventions and therapies developed in the dog may be translational to human patients.

31 References

Angelopoulou, K., M. Zavlaris, N. Papaioannou and I. Vlemmas (2008). "Canis familiaris telomerase reverse transcriptase undergoes alternative splicing." Mamm Genome 19(9): 647-53.

Anninga, J. K., M. J. van de Vijver, A. M. Cleton-Jansen, P. M. Kristel, A. H. Taminiau, M. Nooij, R. M. Egeler and P. C. Hogendoorn (2004). "Overexpression of the HER-2 oncogene does not play a role in high-grade osteosarcomas." Eur J Cancer 40(7): 963- 70.

Araki, N., A. Uchida, T. Kimura, H. Yoshikawa, Y. Aoki, T. Ueda, S. Takai, T. Miki and K. Ono (1991). "Involvement of the retinoblastoma gene in primary osteosarcomas and other bone and soft-tissue tumors." Clin Orthop Relat Res(270): 271-7.

Atiye, J., M. Wolf, S. Kaur, O. Monni, T. Bohling, A. Kivioja, E. Tas, M. Serra, M. Tarkkanen and S. Knuutila (2005). "Gene amplifications in osteosarcoma-CGH microarray analysis." Genes Chromosomes Cancer 42(2): 158-63.

Bailey, D., H. Erb, L. Williams, D. Ruslander and M. Hauck (2003). "Carboplatin and doxorubicin combination chemotherapy for the treatment of appendicular osteosarcoma in the dog." J Vet Intern Med 17(2): 199-205.

Barabas, K., R. Milner, D. Lurie and C. Adin (2008). "Cisplatin: a review of toxicities and therapeutic applications." Vet Comp Oncol 6(1): 1-18.

Batanian, J. R., L. R. Cavalli, N. M. Aldosari, E. Ma, C. Sotelo-Avila, M. B. Ramos, J. D. Rone, C. M. Thorpe and B. R. Haddad (2002). "Evaluation of paediatric osteosarcomas by classic cytogenetic and CGH analyses." Mol Pathol 55(6): 389-93.

Bayani, J., S. Selvarajah, G. Maire, B. Vukovic, K. Al-Romaih, M. Zielenska and J. A. Squire (2007). "Genomic mechanisms and measurement of structural and numerical instability in cancer cells." Semin Cancer Biol 17(1): 5-18.

Bayani, J., M. Zielenska, A. Pandita, K. Al-Romaih, J. Karaskova, K. Harrison, J. A. Bridge, P. Sorensen, P. Thorner and J. A. Squire (2003). "Spectral karyotyping identifies recurrent complex rearrangements of chromosomes 8, 17, and 20 in osteosarcomas." Genes Chromosomes Cancer 36(1): 7-16.

32 Berg, J., M. C. Gebhardt and W. M. Rand (1997). "Effect of timing of postoperative chemotherapy on survival of dogs with osteosarcoma." Cancer 79(7): 1343-50.

Berg, J., M. J. Weinstein, D. S. Springfield and W. M. Rand (1995). "Results of surgery and doxorubicin chemotherapy in dogs with osteosarcoma." J Am Vet Med Assoc 206(10): 1555-60.

Bergman, P. J., E. G. MacEwen, I. D. Kurzman, C. J. Henry, A. S. Hammer, D. W. Knapp, A. Hale, S. A. Kruth, M. K. Klein, J. Klausner, A. M. Norris, D. McCaw, R. C. Straw and S. J. Withrow (1996). "Amputation and carboplatin for treatment of dogs with osteosarcoma: 48 cases (1991 to 1993)." J Vet Intern Med 10(2): 76-81.

Boston, S. E., F. Duerr, N. Bacon, S. Larue, E. J. Ehrhart and S. Withrow (2007). "Intraoperative radiation for limb sparing of the distal aspect of the radius without transcarpal plating in five dogs." Vet Surg 36(4): 314-23.

Boston, S. E., N. P. Ehrhart, W. S. Dernell, M. Lafferty and S. J. Withrow (2006). "Evaluation of survival time in dogs with stage III osteosarcoma that undergo treatment: 90 cases (1985-2004)." J Am Vet Med Assoc 228(12): 1905-8.

Boyko, A. R., P. Quignon, L. Li, J. J. Schoenebeck, J. D. Degenhardt, K. E. Lohmueller, K. Zhao, A. Brisbin, H. G. Parker, B. M. Vonholdt, M. Cargill, A. Auton, A. Reynolds, A. G. Elkahloun, M. Castelhano, D. S. Mosher, N. B. Sutter, G. S. Johnson, J. Novembre, M. J. Hubisz, A. Siepel, R. K. Wayne, C. D. Bustamante and E. A. Ostrander (2010). "A simple genetic architecture underlies morphological variation in dogs." PLoS Biol 8(8).

Breen, M., C. Hitte, T. D. Lorentzen, R. Thomas, E. Cadieu, L. Sabacan, A. Scott, G. Evanno, H. G. Parker, E. F. Kirkness, R. Hudson, R. Guyon, G. G. Mahairas, B. Gelfenbeyn, C. M. Fraser, C. Andre, F. Galibert and E. A. Ostrander (2004). "An integrated 4249 marker FISH/RH map of the canine genome." BMC Genomics 5(1): 65.

Breen, M., S. Jouquand, C. Renier, C. S. Mellersh, C. Hitte, N. G. Holmes, A. Cheron, N. Suter, F. Vignaux, A. E. Bristow, C. Priat, E. McCann, C. Andre, S. Boundy, P. Gitsham, R. Thomas, W. L. Bridge, H. F. Spriggs, E. J. Ryder, A. Curson, J. Sampson, E. A. Ostrander, M. M. Binns and F. Galibert (2001). "Chromosome- specific single-locus FISH probes allow anchorage of an 1800-marker integrated

33 radiation-hybrid/linkage map of the domestic dog genome to all chromosomes." Genome Res 11(10): 1784-95.

Breen, M. and J. F. Modiano (2008). "Evolutionarily conserved cytogenetic changes in hematological malignancies of dogs and humans--man and his best friend share more than companionship." Chromosome Res 16(1): 145-54.

Bridge, J. A., M. Nelson, E. McComb, M. H. McGuire, H. Rosenthal, G. Vergara, G. E. Maale, S. Spanier and J. R. Neff (1997). "Cytogenetic findings in 73 osteosarcoma specimens and a review of the literature." Cancer Genet Cytogenet 95(1): 74-87.

Brodey, R. S. and D. A. Abt (1976). "Results of surgical treatment in 65 dogs with osteosarcoma." J Am Vet Med Assoc 168(11): 1032-5.

Brodey, R. S. and W. H. Riser (1969). "Canine osteosarcoma. A clinicopathologic study of 194 cases." Clin Orthop Relat Res 62: 54-64.

Bucan, M. and T. Abel (2002). "The mouse: genetics meets behaviour." Nat Rev Genet 3(2): 114-23.

Cadieu, E. and E. A. Ostrander (2007). "Canine genetics offers new mechanisms for the study of human cancer." Cancer Epidemiol Biomarkers Prev 16(11): 2181-3.

Chun, R., I. D. Kurzman, C. G. Couto, J. Klausner, C. Henry and E. G. MacEwen (2000). "Cisplatin and doxorubicin combination chemotherapy for the treatment of canine osteosarcoma: a pilot study." J Vet Intern Med 14(5): 495-8.

de Nigris, F., C. Botti, R. Rossiello, E. Crimi, V. Sica and C. Napoli (2007). "Cooperation between Myc and YY1 provides novel silencing transcriptional targets of alpha3beta1-integrin in tumour cells." Oncogene 26(3): 382-94.

34 Dickerson, M. E., R. L. Page, T. A. LaDue, M. L. Hauck, D. E. Thrall, M. E. Stebbins and G. S. Price (2001). "Retrospective analysis of axial skeleton osteosarcoma in 22 large- breed dogs." J Vet Intern Med 15(2): 120-4. dos Santos Aguiar, S., L. de Jesus Girotto Zambaldi, A. M. dos Santos, W. Pinto, Jr. and S. R. Brandalise (2007). "Comparative genomic hybridization analysis of abnormalities in chromosome 21 in childhood osteosarcoma." Cancer Genet Cytogenet 175(1): 35- 40.

Ferracini, R., P. Angelini, E. Cagliero, A. Linari, M. Martano, J. Wunder and P. Buracco (2000). "MET oncogene aberrant expression in canine osteosarcoma." J Orthop Res 18(2): 253-6.

Fieten, H., B. Spee, J. Ijzer, M. J. Kik, L. C. Penning and J. Kirpensteijn (2009). "Expression of hepatocyte growth factor and the proto-oncogenic receptor c-Met in canine osteosarcoma." Vet Pathol 46(5): 869-77.

Flint, A. F., L. U'Ren, M. E. Legare, S. J. Withrow, W. Dernell and W. H. Hanneman (2004). "Overexpression of the erbB-2 proto-oncogene in canine osteosarcoma cell lines and tumors." Vet Pathol 41(3): 291-6.

Forus, A., D. O. Weghuis, D. Smeets, O. Fodstad, O. Myklebost and A. Geurts van Kessel (1995). "Comparative genomic hybridization analysis of human sarcomas: II. Identification of novel amplicons at 6p and 17p in osteosarcomas." Genes Chromosomes Cancer 14(1): 15-21.

Fossey, S. L., A. T. Liao, J. K. McCleese, M. D. Bear, J. Lin, P. K. Li, W. C. Kisseberth and C. A. London (2009). "Characterization of STAT3 activation and expression in canine and human osteosarcoma." BMC Cancer 9(1): 81.

Foukas, A. F., N. S. Deshmukh, R. J. Grimer, D. C. Mangham, E. G. Mangos and S. Taylor (2002). "Stage-IIB osteosarcomas around the knee. A study of MMP-9 in surviving tumour cells." J Bone Joint Surg Br 84(5): 706-11.

Gamberi, G., M. S. Benassi, T. Bohling, P. Ragazzini, L. Molendini, M. R. Sollazzo, F. Pompetti, M. Merli, G. Magagnoli, A. Balladelli and P. Picci (1998). "C-myc and c- fos in human osteosarcoma: prognostic value of mRNA and protein expression." Oncology 55(6): 556-63.

35 Gershwin, L. J. (2007). "Veterinary autoimmunity: autoimmune diseases in domestic animals." Ann N Y Acad Sci 1109: 109-16.

Gorlick, R., A. G. Huvos, G. Heller, A. Aledo, G. P. Beardsley, J. H. Healey and P. A. Meyers (1999). "Expression of HER2/erbB-2 correlates with survival in osteosarcoma." J Clin Oncol 17(9): 2781-8.

Gorman, E., A. M. Barger, J. M. Wypij and M. E. Pinkerton (2006). "Cutaneous metastasis of primary appendicular osteosarcoma in a dog." Vet Clin Pathol 35(3): 358-61.

Guyon, R., T. D. Lorentzen, C. Hitte, L. Kim, E. Cadieu, H. G. Parker, P. Quignon, J. K. Lowe, C. Renier, B. Gelfenbeyn, F. Vignaux, H. B. DeFrance, S. Gloux, G. G. Mahairas, C. Andre, F. Galibert and E. A. Ostrander (2003). "A 1-Mb resolution radiation hybrid map of the canine genome." Proc Natl Acad Sci U S A 100(9): 5296- 301.

Hammer, A. S., F. R. Weeren, S. E. Weisbrode and S. L. Padgett (1995). "Prognostic factors in dogs with osteosarcomas of the flat or irregular bones." J Am Anim Hosp Assoc 31(4): 321-6.

Hillers, K. R., W. S. Dernell, M. H. Lafferty, S. J. Withrow and S. E. Lana (2005). "Incidence and prognostic importance of lymph node metastases in dogs with appendicular osteosarcoma: 228 cases (1986-2003)." J Am Vet Med Assoc 226(8): 1364-7.

Hutt, F. B., C. G. Rickard and R. A. Field (1948). "Sex-linked hemophilia in dogs." J. Heredity 39: 3-9.

Jemal, A., R. Siegel, E. Ward, Y. Hao, J. Xu and M. J. Thun (2009). "Cancer statistics, 2009." CA Cancer J Clin 59(4): 225-49.

Johnson, A. S., C. G. Couto and C. M. Weghorst (1998). "Mutation of the p53 tumor suppressor gene in spontaneously occurring osteosarcomas of the dog." Carcinogenesis 19(1): 213-7.

Jones, P., K. Chase, A. Martin, P. Davern, E. A. Ostrander and K. G. Lark (2008). "Single- nucleotide-polymorphism-based association mapping of dog stereotypes." Genetics 179(2): 1033-44.

36 Karlsson, E. K. and K. Lindblad-Toh (2008). "Leader of the pack: gene mapping in dogs and other model organisms." Nat Rev Genet 9(9): 713-25.

Khanna, C., K. Lindblad-Toh, D. Vail, C. London, P. Bergman, L. Barber, M. Breen, B. Kitchell, E. McNeil, J. F. Modiano, S. Niemi, K. E. Comstock, E. Ostrander, S. Westmoreland and S. Withrow (2006). "The dog as a cancer model." Nat Biotechnol 24(9): 1065-6.

Khanna, C., J. Prehn, D. Hayden, R. D. Cassaday, J. Caylor, S. Jacob, S. M. Bose, S. H. Hong, S. M. Hewitt and L. J. Helman (2002). "A randomized controlled trial of octreotide pamoate long-acting release and carboplatin versus carboplatin alone in dogs with naturally occurring osteosarcoma: evaluation of insulin-like growth factor suppression and chemotherapy." Clin Cancer Res 8(7): 2406-12.

Khanna, C., X. Wan, S. Bose, R. Cassaday, O. Olomu, A. Mendoza, C. Yeung, R. Gorlick, S. M. Hewitt and L. J. Helman (2004). "The membrane-cytoskeleton linker ezrin is necessary for osteosarcoma metastasis." Nat Med 10(2): 182-6.

Kirkness, E. F., V. Bafna, A. L. Halpern, S. Levy, K. Remington, D. B. Rusch, A. L. Delcher, M. Pop, W. Wang, C. M. Fraser and J. C. Venter (2003). "The dog genome: survey sequencing and comparative analysis." Science 301(5641): 1898-903.

Kirpensteijn, J., M. Kik, G. R. Rutteman and E. Teske (2002a). "Prognostic significance of a new histologic grading system for canine osteosarcoma." Vet Pathol 39(2): 240-6.

Kirpensteijn, J., M. Kik, E. Teske and G. R. Rutteman (2008). "TP53 gene mutations in canine osteosarcoma." Vet Surg 37(5): 454-60.

Kirpensteijn, J., E. Teske, M. Kik, T. Klenner and G. R. Rutteman (2002b). "Lobaplatin as an adjuvant chemotherapy to surgery in canine appendicular osteosarcoma: a phase II evaluation." Anticancer Res 22(5): 2765-70.

Knecht, C. D. and W. A. Priester (1978). "Musculoskeletal tumors in dogs." J Am Vet Med Assoc 172(1): 72-4.

Kochevar, D. T., J. Kochevar and L. Garrett (1990). "Low level amplification of c-sis and c- myc in a spontaneous osteosarcoma model." Cancer Lett 53(2-3): 213-22.

37 Kow, K., D. H. Thamm, J. Terry, K. Grunerud, S. M. Bailey, S. J. Withrow and S. E. Lana (2008). "Impact of telomerase status on canine osteosarcoma patients." J Vet Intern Med 22(6): 1366-72.

Kresse, S. H., H. O. Ohnstad, E. B. Paulsen, B. Bjerkehagen, K. Szuhai, M. Serra, K. L. Schaefer, O. Myklebost and L. A. Meza-Zepeda (2009). "LSAMP, a novel candidate tumor suppressor gene in human osteosarcomas, identified by array comparative genomic hybridization." Genes Chromosomes Cancer 48(8): 679-93.

Kuntz, C. A., W. S. Dernell, B. E. Powers and S. Withrow (1998). "Extraskeletal osteosarcomas in dogs: 14 cases." J Am Anim Hosp Assoc 34(1): 26-30.

Lana, S. E., G. K. Ogilvie, R. A. Hansen, B. E. Powers, W. S. Dernell and S. J. Withrow (2000). "Identification of matrix metalloproteinases in canine neoplastic tissue." Am J Vet Res 61(2): 111-4.

Lander, E. S., L. M. Linton, B. Birren, C. Nusbaum, M. C. Zody, J. Baldwin, K. Devon, K. Dewar, M. Doyle, W. FitzHugh, R. Funke, D. Gage, K. Harris, A. Heaford, J. Howland, L. Kann, J. Lehoczky, R. LeVine, P. McEwan, K. McKernan, J. Meldrim, J. P. Mesirov, C. Miranda, W. Morris, J. Naylor, C. Raymond, M. Rosetti, R. Santos, A. Sheridan, C. Sougnez, N. Stange-Thomann, N. Stojanovic, A. Subramanian, D. Wyman, J. Rogers, J. Sulston, R. Ainscough, S. Beck, D. Bentley, J. Burton, C. Clee, N. Carter, A. Coulson, R. Deadman, P. Deloukas, A. Dunham, I. Dunham, R. Durbin, L. French, D. Grafham, S. Gregory, T. Hubbard, S. Humphray, A. Hunt, M. Jones, C. Lloyd, A. McMurray, L. Matthews, S. Mercer, S. Milne, J. C. Mullikin, A. Mungall, R. Plumb, M. Ross, R. Shownkeen, S. Sims, R. H. Waterston, R. K. Wilson, L. W. Hillier, J. D. McPherson, M. A. Marra, E. R. Mardis, L. A. Fulton, A. T. Chinwalla, K. H. Pepin, W. R. Gish, S. L. Chissoe, M. C. Wendl, K. D. Delehaunty, T. L. Miner, A. Delehaunty, J. B. Kramer, L. L. Cook, R. S. Fulton, D. L. Johnson, P. J. Minx, S. W. Clifton, T. Hawkins, E. Branscomb, P. Predki, P. Richardson, S. Wenning, T. Slezak, N. Doggett, J. F. Cheng, A. Olsen, S. Lucas, C. Elkin, E. Uberbacher, M. Frazier, R. A. Gibbs, D. M. Muzny, S. E. Scherer, J. B. Bouck, E. J. Sodergren, K. C. Worley, C. M. Rives, J. H. Gorrell, M. L. Metzker, S. L. Naylor, R. S. Kucherlapati, D. L. Nelson, G. M. Weinstock, Y. Sakaki, A. Fujiyama, M. Hattori, T. Yada, A. Toyoda, T. Itoh, C. Kawagoe, H. Watanabe, Y. Totoki, T. Taylor, J. Weissenbach, R. Heilig, W. Saurin, F. Artiguenave, P. Brottier, T. Bruls, E. Pelletier, C. Robert, P. Wincker, D. R. Smith, L. Doucette-Stamm, M. Rubenfield, K. Weinstock, H. M. Lee, J. Dubois, A. Rosenthal, M. Platzer, G. Nyakatura, S. Taudien, A. Rump, H. Yang, J. Yu, J. Wang, G. Huang, J. Gu, L. Hood, L. Rowen, A. Madan, S. Qin, R. W. Davis, N. A. Federspiel, A. P. Abola, M. J. Proctor, R. M. Myers, J. Schmutz, M. Dickson, J. Grimwood, D. R. Cox, M. V. Olson, R. Kaul, N. Shimizu, K. Kawasaki, S.

38 Minoshima, G. A. Evans, M. Athanasiou, R. Schultz, B. A. Roe, F. Chen, H. Pan, J. Ramser, H. Lehrach, R. Reinhardt, W. R. McCombie, M. de la Bastide, N. Dedhia, H. Blocker, K. Hornischer, G. Nordsiek, R. Agarwala, L. Aravind, J. A. Bailey, A. Bateman, S. Batzoglou, E. Birney, P. Bork, D. G. Brown, C. B. Burge, L. Cerutti, H. C. Chen, D. Church, M. Clamp, R. R. Copley, T. Doerks, S. R. Eddy, E. E. Eichler, T. S. Furey, J. Galagan, J. G. Gilbert, C. Harmon, Y. Hayashizaki, D. Haussler, H. Hermjakob, K. Hokamp, W. Jang, L. S. Johnson, T. A. Jones, S. Kasif, A. Kaspryzk, S. Kennedy, W. J. Kent, P. Kitts, E. V. Koonin, I. Korf, D. Kulp, D. Lancet, T. M. Lowe, A. McLysaght, T. Mikkelsen, J. V. Moran, N. Mulder, V. J. Pollara, C. P. Ponting, G. Schuler, J. Schultz, G. Slater, A. F. Smit, E. Stupka, J. Szustakowski, D. Thierry-Mieg, J. Thierry-Mieg, L. Wagner, J. Wallis, R. Wheeler, A. Williams, Y. I. Wolf, K. H. Wolfe, S. P. Yang, R. F. Yeh, F. Collins, M. S. Guyer, J. Peterson, A. Felsenfeld, K. A. Wetterstrand, A. Patrinos, M. J. Morgan, P. de Jong, J. J. Catanese, K. Osoegawa, H. Shizuya, S. Choi and Y. J. Chen (2001). "Initial sequencing and analysis of the human genome." Nature 409(6822): 860-921.

Langenbach, A., M. A. Anderson, D. M. Dambach, K. U. Sorenmo and F. D. Shofer (1998). "Extraskeletal osteosarcomas in dogs: a retrospective study of 169 cases (1986- 1996)." J Am Anim Hosp Assoc 34(2): 113-20.

Lau, C. C., C. P. Harris, X. Y. Lu, L. Perlaky, S. Gogineni, M. Chintagumpala, J. Hicks, M. E. Johnson, N. A. Davino, A. G. Huvos, P. A. Meyers, J. H. Healy, R. Gorlick and P. H. Rao (2004). "Frequent amplification and rearrangement of chromosomal bands 6p12-p21 and 17p11.2 in osteosarcoma." Genes Chromosomes Cancer 39(1): 11-21.

Levine, R. A. (2002). "Overexpression of the sis oncogene in a canine osteosarcoma cell line." Vet Pathol 39(3): 411-2.

Levine, R. A. and M. A. Fleischli (2000). "Inactivation of p53 and retinoblastoma family pathways in canine osteosarcoma cell lines." Vet Pathol 37(1): 54-61.

Levine, R. A., T. Forest and C. Smith (2002). "Tumor suppressor PTEN is mutated in canine osteosarcoma cell lines and tumors." Vet Pathol 39(3): 372-8.

Lim, G., J. Karaskova, B. Beheshti, B. Vukovic, J. Bayani, S. Selvarajah, S. K. Watson, W. L. Lam, M. Zielenska and J. A. Squire (2005). "An integrated mBAND and submegabase resolution tiling set (SMRT) CGH array analysis of focal amplification, microdeletions, and ladder structures consistent with breakage-fusion-bridge cycle events in osteosarcoma." Genes Chromosomes Cancer 42(4): 392-403.

39 Lim, G., J. Karaskova, B. Vukovic, J. Bayani, B. Beheshti, M. Bernardini, J. A. Squire and M. Zielenska (2004). "Combined spectral karyotyping, multicolor banding, and microarray comparative genomic hybridization analysis provides a detailed characterization of complex structural chromosomal rearrangements associated with gene amplification in the osteosarcoma cell line MG-63." Cancer Genet Cytogenet 153(2): 158-64.

Lin, L., J. Faraco, R. Li, H. Kadotani, W. Rogers, X. Lin, X. Qiu, P. J. de Jong, S. Nishino and E. Mignot (1999). "The sleep disorder canine narcolepsy is caused by a mutation in the hypocretin (orexin) receptor 2 gene." Cell 98(3): 365-76.

Lindblad-Toh, K., C. M. Wade, T. S. Mikkelsen, E. K. Karlsson, D. B. Jaffe, M. Kamal, M. Clamp, J. L. Chang, E. J. Kulbokas, 3rd, M. C. Zody, E. Mauceli, X. Xie, M. Breen, R. K. Wayne, E. A. Ostrander, C. P. Ponting, F. Galibert, D. R. Smith, P. J. DeJong, E. Kirkness, P. Alvarez, T. Biagi, W. Brockman, J. Butler, C. W. Chin, A. Cook, J. Cuff, M. J. Daly, D. DeCaprio, S. Gnerre, M. Grabherr, M. Kellis, M. Kleber, C. Bardeleben, L. Goodstadt, A. Heger, C. Hitte, L. Kim, K. P. Koepfli, H. G. Parker, J. P. Pollinger, S. M. Searle, N. B. Sutter, R. Thomas, C. Webber, J. Baldwin, A. Abebe, A. Abouelleil, L. Aftuck, M. Ait-Zahra, T. Aldredge, N. Allen, P. An, S. Anderson, C. Antoine, H. Arachchi, A. Aslam, L. Ayotte, P. Bachantsang, A. Barry, T. Bayul, M. Benamara, A. Berlin, D. Bessette, B. Blitshteyn, T. Bloom, J. Blye, L. Boguslavskiy, C. Bonnet, B. Boukhgalter, A. Brown, P. Cahill, N. Calixte, J. Camarata, Y. Cheshatsang, J. Chu, M. Citroen, A. Collymore, P. Cooke, T. Dawoe, R. Daza, K. Decktor, S. DeGray, N. Dhargay, K. Dooley, P. Dorje, K. Dorjee, L. Dorris, N. Duffey, A. Dupes, O. Egbiremolen, R. Elong, J. Falk, A. Farina, S. Faro, D. Ferguson, P. Ferreira, S. Fisher, M. FitzGerald, K. Foley, C. Foley, A. Franke, D. Friedrich, D. Gage, M. Garber, G. Gearin, G. Giannoukos, T. Goode, A. Goyette, J. Graham, E. Grandbois, K. Gyaltsen, N. Hafez, D. Hagopian, B. Hagos, J. Hall, C. Healy, R. Hegarty, T. Honan, A. Horn, N. Houde, L. Hughes, L. Hunnicutt, M. Husby, B. Jester, C. Jones, A. Kamat, B. Kanga, C. Kells, D. Khazanovich, A. C. Kieu, P. Kisner, M. Kumar, K. Lance, T. Landers, M. Lara, W. Lee, J. P. Leger, N. Lennon, L. Leuper, S. LeVine, J. Liu, X. Liu, Y. Lokyitsang, T. Lokyitsang, A. Lui, J. Macdonald, J. Major, R. Marabella, K. Maru, C. Matthews, S. McDonough, T. Mehta, J. Meldrim, A. Melnikov, L. Meneus, A. Mihalev, T. Mihova, K. Miller, R. Mittelman, V. Mlenga, L. Mulrain, G. Munson, A. Navidi, J. Naylor, T. Nguyen, N. Nguyen, C. Nguyen, R. Nicol, N. Norbu, C. Norbu, N. Novod, T. Nyima, P. Olandt, B. O'Neill, K. O'Neill, S. Osman, L. Oyono, C. Patti, D. Perrin, P. Phunkhang, F. Pierre, M. Priest, A. Rachupka, S. Raghuraman, R. Rameau, V. Ray, C. Raymond, F. Rege, C. Rise, J. Rogers, P. Rogov, J. Sahalie, S. Settipalli, T. Sharpe, T. Shea, M. Sheehan, N. Sherpa, J. Shi, D. Shih, J. Sloan, C. Smith, T. Sparrow, J. Stalker, N. Stange-Thomann, S. Stavropoulos, C. Stone, S. Stone, S. Sykes, P. Tchuinga, P. Tenzing, S. Tesfaye, D. Thoulutsang, Y. Thoulutsang, K. Topham, I. Topping, T.

40 Tsamla, H. Vassiliev, V. Venkataraman, A. Vo, T. Wangchuk, T. Wangdi, M. Weiand, J. Wilkinson, A. Wilson, S. Yadav, S. Yang, X. Yang, G. Young, Q. Yu, J. Zainoun, L. Zembek, A. Zimmer and E. S. Lander (2005). "Genome sequence, comparative analysis and haplotype structure of the domestic dog." Nature 438(7069): 803-19.

Liptak, J. M., W. S. Dernell, R. C. Straw, S. A. Rizzo, M. H. Lafferty and S. J. Withrow (2004). "Proximal radial and distal humeral osteosarcoma in 12 dogs." J Am Anim Hosp Assoc 40(6): 461-7.

Little, C. C. (1914). "Coat color in pointer dogs." J Heredity 5: 244-248.

Loukopoulos, P., T. O'Brien, M. Ghoddusi, B. A. Mungall and W. F. Robinson (2004). "Characterisation of three novel canine osteosarcoma cell lines producing high levels of matrix metalloproteinases." Res Vet Sci 77(2): 131-41.

Lu, X. Y., Y. Lu, Y. J. Zhao, K. Jaeweon, J. Kang, L. Xiao-Nan, G. Ge, R. Meyer, L. Perlaky, J. Hicks, M. Chintagumpala, W. W. Cai, M. Ladanyi, R. Gorlick, C. C. Lau, D. Pati, M. Sheldon and P. H. Rao (2008). "Cell cycle regulator gene CDC5L, a potential target for 6p12-p21 amplicon in osteosarcoma." Mol Cancer Res 6(6): 937- 46.

MacEwen, E. G., J. Pastor, J. Kutzke, R. Tsan, I. D. Kurzman, D. H. Thamm, M. Wilson and R. Radinsky (2004). "IGF-1 receptor contributes to the malignant phenotype in human and canine osteosarcoma." J Cell Biochem 92(1): 77-91.

Maire, G., M. Yoshimoto, S. Chilton-MacNeill, P. S. Thorner, M. Zielenska and J. A. Squire (2009). "Recurrent RECQL4 imbalance and increased gene expression levels are associated with structural chromosomal instability in sporadic osteosarcoma." Neoplasia 11(3): 260-8, 3p following 268.

Maitra, A., D. Wanzer, A. G. Weinberg and R. Ashfaq (2001). "Amplification of the HER- 2/neu oncogene is uncommon in pediatric osteosarcomas." Cancer 92(3): 677-83.

Man, T. K., X. Y. Lu, K. Jaeweon, L. Perlaky, C. P. Harris, S. Shah, M. Ladanyi, R. Gorlick, C. C. Lau and P. H. Rao (2004). "Genome-wide array comparative genomic hybridization analysis reveals distinct amplifications in osteosarcoma." BMC Cancer 4: 45.

41 McCleese, J. K., M. D. Bear, S. L. Fossey, R. M. Mihalek, K. P. Foley, W. Ying, J. Barsoum and C. A. London (2009). "The novel HSP90 inhibitor STA-1474 exhibits biologic activity against osteosarcoma cell lines." Int J Cancer 125(12): 2792-801.

McIntyre, J. F., B. Smith-Sorensen, S. H. Friend, J. Kassell, A. L. Borresen, Y. X. Yan, C. Russo, J. Sato, N. Barbier, J. Miser and et al. (1994). "Germline mutations of the p53 tumor suppressor gene in children with osteosarcoma." J Clin Oncol 12(5): 925-30.

McNeill, C. J., B. Overley, F. S. Shofer, M. S. Kent, C. A. Clifford, M. Samluk, S. Haney, T. J. Van Winkle and K. U. Sorenmo (2007). "Characterization of the biological behaviour of appendicular osteosarcoma in Rottweilers and a comparison with other breeds: a review of 258 dogs*." Vet Comp Oncol 5(2): 90-8.

Mellersh, C. S., A. A. Langston, G. M. Acland, M. A. Fleming, K. Ray, N. A. Wiegand, L. V. Francisco, M. Gibbs, G. D. Aguirre and E. A. Ostrander (1997). "A linkage map of the canine genome." Genomics 46(3): 326-36.

Mendoza, S., T. Konishi, W. S. Dernell, S. J. Withrow and C. W. Miller (1998). "Status of the p53, Rb and MDM2 genes in canine osteosarcoma." Anticancer Res 18(6A): 4449-53.

Merlo, D. F., L. Rossi, C. Pellegrino, M. Ceppi, U. Cardellino, C. Capurro, A. Ratto, P. L. Sambucco, V. Sestito, G. Tanara and V. Bocchini (2008). "Cancer incidence in pet dogs: findings of the Animal Tumor Registry of Genoa, Italy." J Vet Intern Med 22(4): 976-84.

Mirabello, L., R. J. Troisi and S. A. Savage (2009a). "International osteosarcoma incidence patterns in children and adolescents, middle ages and elderly persons." Int J Cancer 125(1): 229-34.

Mirabello, L., R. J. Troisi and S. A. Savage (2009b). "Osteosarcoma incidence and survival rates from 1973 to 2004: data from the Surveillance, Epidemiology, and End Results Program." Cancer 115(7): 1531-43.

Misdorp, W. and A. A. Hart (1979). "Some prognostic and epidemiologic factors in canine osteosarcoma." J Natl Cancer Inst 62(3): 537-45.

42 Moore, A. S., W. S. Dernell, G. K. Ogilvie, O. Kristal, R. Elmslie, B. Kitchell, S. Susaneck, R. Rosenthal, M. K. Klein, J. Obradovich, A. Legendre, T. Haddad, K. Hahn, B. E. Powers and D. Warren (2007). "Doxorubicin and BAY 12-9566 for the treatment of osteosarcoma in dogs: a randomized, double-blind, placebo-controlled study." J Vet Intern Med 21(4): 783-90.

Mueller, F., B. Fuchs and B. Kaser-Hotz (2007). "Comparative biology of human and canine osteosarcoma." Anticancer Res 27(1A): 155-64.

Mullins, M. N., S. E. Lana, W. S. Dernell, G. K. Ogilvie, S. J. Withrow and E. J. Ehrhart (2004). "Cyclooxygenase-2 expression in canine appendicular osteosarcomas." J Vet Intern Med 18(6): 859-65.

Ohata, N., S. Ito, A. Yoshida, T. Kunisada, K. Numoto, Y. Jitsumori, H. Kanzaki, T. Ozaki, K. Shimizu and M. Ouchida (2006). "Highly frequent allelic loss of chromosome 6q16-23 in osteosarcoma: involvement of cyclin C in osteosarcoma." Int J Mol Med 18(6): 1153-8.

Ostrander, E. A. and L. Kruglyak (2000). "Unleashing the canine genome." Genome Res 10(9): 1271-4.

Overall, K. L. (2000). "Natural animal models of human psychiatric conditions: assessment of mechanism and validity." Prog Neuropsychopharmacol Biol Psychiatry 24(5): 727- 76.

Ozaki, T., T. Neumann, D. Wai, K. L. Schafer, F. van Valen, N. Lindner, C. Scheel, W. Bocker, W. Winkelmann, B. Dockhorn-Dworniczak, J. Horst and C. Poremba (2003). "Chromosomal alterations in osteosarcoma cell lines revealed by comparative genomic hybridization and multicolor karyotyping." Cancer Genet Cytogenet 140(2): 145-52.

Ozaki, T., K. L. Schaefer, D. Wai, H. Buerger, S. Flege, N. Lindner, M. Kevric, R. Diallo, A. Bankfalvi, C. Brinkschmidt, H. Juergens, W. Winkelmann, B. Dockhorn- Dworniczak, S. S. Bielack and C. Poremba (2002). "Genetic imbalances revealed by comparative genomic hybridization in osteosarcomas." Int J Cancer 102(4): 355-65.

Paigen, K. (1995). "A miracle enough: the power of mice." Nat Med 1(3): 215-20.

43 Paoloni, M., S. Davis, S. Lana, S. Withrow, L. Sangiorgi, P. Picci, S. Hewitt, T. Triche, P. Meltzer and C. Khanna (2009). "Canine tumor cross-species genomics uncovers targets linked to osteosarcoma progression." BMC Genomics 10: 625.

Paoloni, M. and C. Khanna (2008). "Translation of new cancer treatments from pet dogs to humans." Nat Rev Cancer 8(2): 147-56.

Parker, H. G. and E. A. Ostrander (2005). "Canine genomics and genetics: running with the pack." PLoS Genet 1(5): e58.

Patnaik, A. K. (1990). "Canine extraskeletal osteosarcoma and chondrosarcoma: a clinicopathologic study of 14 cases." Vet Pathol 27(1): 46-55.

Peremans, K., A. Otte, F. Verschooten, H. Van Bree and R. Dierckx (2003). "Soft tissue metastasis of an osteosarcoma of the humerus in a four-legged patient." Eur J Nucl Med Mol Imaging 30(1): 188.

Ru, G., B. Terracini and L. T. Glickman (1998). "Host related risk factors for canine osteosarcoma." Vet J 156(1): 31-9.

Sadikovic, B., P. Thorner, S. Chilton-Macneill, J. W. Martin, N. K. Cervigne, J. Squire and M. Zielenska (2010). "Expression analysis of genes associated with human osteosarcoma tumors shows correlation of RUNX2 overexpression with poor response to chemotherapy." BMC Cancer 10: 202.

Sadikovic, B., M. Yoshimoto, S. Chilton-MacNeill, P. Thorner, J. A. Squire and M. Zielenska (2009). "Identification of interactive networks of gene expression associated with osteosarcoma oncogenesis by integrated molecular profiling." Hum Mol Genet 18(11): 1962-75.

Sandberg, A. A. and J. A. Bridge (2003). "Updates on the cytogenetics and molecular genetics of bone and soft tissue tumors: osteosarcoma and related tumors." Cancer Genet Cytogenet 145(1): 1-30.

Selvarajah, G. T. and J. Kirpensteijn (2010). "Prognostic and predictive biomarkers of canine osteosarcoma." Vet J.

44 Selvarajah, G. T., J. Kirpensteijn, M. E. van Wolferen, N. A. Rao, H. Fieten and J. A. Mol (2009). "Gene expression profiling of canine osteosarcoma reveals genes associated with short and long survival times." Mol Cancer 8: 72.

Selvarajah, S., M. Yoshimoto, O. Ludkovski, P. C. Park, J. Bayani, P. Thorner, G. Maire, J. A. Squire and M. Zielenska (2008). "Genomic signatures of chromosomal instability and osteosarcoma progression detected by high resolution array CGH and interphase FISH." Cytogenet Genome Res 122(1): 5-15.

Selvarajah, S., M. Yoshimoto, G. Maire, J. Paderova, J. Bayani, J. A. Squire and M. Zielenska (2007). "Identification of cryptic microaberrations in osteosarcoma by high-definition oligonucleotide array comparative genomic hybridization." Cancer Genet Cytogenet 179(1): 52-61.

Shearin, A. L. and E. A. Ostrander (2010). "Leading the way: canine models of genomics and disease." Dis Model Mech 3(1-2): 27-34.

Simons, A., I. M. Janssen, R. F. Suijkerbuijk, R. P. Veth, M. Pruszczynski, C. A. Hulsbergen-van de Kaa, S. du Manoir and A. Geurts van Kessel (1997). "Isolation of osteosarcoma-associated amplified DNA sequences using representational difference analysis." Genes Chromosomes Cancer 20(2): 196-200.

Spodnick, G. J., J. Berg, W. M. Rand, S. H. Schelling, G. Couto, H. J. Harvey, R. A. Henderson, G. MacEwen, N. Mauldin, D. L. McCaw and et al. (1992). "Prognosis for dogs with appendicular osteosarcoma treated by amputation alone: 162 cases (1978- 1988)." J Am Vet Med Assoc 200(7): 995-9.

Squire, J. A., J. Pei, P. Marrano, B. Beheshti, J. Bayani, G. Lim, L. Moldovan and M. Zielenska (2003). "High-resolution mapping of amplifications and deletions in pediatric osteosarcoma by use of CGH analysis of cDNA microarrays." Genes Chromosomes Cancer 38(3): 215-25.

Straw, R. C. and S. J. Withrow (1996). "Limb-sparing surgery versus amputation for dogs with bone tumors." Vet Clin North Am Small Anim Pract 26(1): 135-43.

Straw, R. C., S. J. Withrow, S. L. Richter, B. E. Powers, M. K. Klein, N. C. Postorino, S. M. LaRue, G. K. Ogilvie, D. M. Vail, W. B. Morrison and et al. (1991). "Amputation and cisplatin for treatment of canine osteosarcoma." J Vet Intern Med 5(4): 205-10.

45 Sutter, N. B., C. D. Bustamante, K. Chase, M. M. Gray, K. Zhao, L. Zhu, B. Padhukasahasram, E. Karlins, S. Davis, P. G. Jones, P. Quignon, G. S. Johnson, H. G. Parker, N. Fretwell, D. S. Mosher, D. F. Lawler, E. Satyaraj, M. Nordborg, K. G. Lark, R. K. Wayne and E. A. Ostrander (2007). "A single IGF1 allele is a major determinant of small size in dogs." Science 316(5821): 112-5.

Sutter, N. B. and E. A. Ostrander (2004). "Dog star rising: the canine genetic system." Nat Rev Genet 5(12): 900-10.

Tang, N., W. X. Song, J. Luo, R. C. Haydon and T. C. He (2008). "Osteosarcoma development and stem cell differentiation." Clin Orthop Relat Res 466(9): 2114-30.

Tarkkanen, M., T. Bohling, G. Gamberi, P. Ragazzini, M. S. Benassi, A. Kivioja, P. Kallio, I. Elomaa, P. Picci and S. Knuutila (1998). "Comparative genomic hybridization of low-grade central osteosarcoma." Mod Pathol 11(5): 421-6.

Tarkkanen, M., I. Elomaa, C. Blomqvist, A. H. Kivioja, P. Kellokumpu-Lehtinen, T. Bohling, J. Valle and S. Knuutila (1999). "DNA sequence copy number increase at 8q: a potential new prognostic marker in high-grade osteosarcoma." Int J Cancer 84(2): 114-21.

Tarkkanen, M., R. Karhu, A. Kallioniemi, I. Elomaa, A. H. Kivioja, J. Nevalainen, T. Bohling, E. Karaharju, E. Hyytinen, S. Knuutila and et al. (1995). "Gains and losses of DNA sequences in osteosarcomas by comparative genomic hybridization." Cancer Res 55(6): 1334-8.

Thomas, R., M. Breen and M. M. Binns (2001a). "Chromosome assignment of six dog genes by FISH, and correlation with dog-human Zoo-FISH data." Anim Genet 32(3): 148- 51.

Thomas, R., M. Breen, P. Deloukas, N. G. Holmes and M. M. Binns (2001b). "An integrated cytogenetic, radiation-hybrid, and comparative map of dog chromosome 5." Mamm Genome 12(5): 371-5.

Thomas, R., H. J. Wang, P. C. Tsai, C. F. Langford, S. P. Fosmire, C. M. Jubala, D. M. Getzy, G. R. Cutter, J. F. Modiano and M. Breen (2009). "Influence of genetic background on tumor karyotypes: evidence for breed-associated cytogenetic aberrations in canine appendicular osteosarcoma." Chromosome Res 17(3): 365-77.

46 Toguchida, J., T. Yamaguchi, S. H. Dayton, R. L. Beauchamp, G. E. Herrera, K. Ishizaki, T. Yamamuro, P. A. Meyers, J. B. Little, M. S. Sasaki and et al. (1992). "Prevalence and spectrum of germline mutations of the p53 gene among patients with sarcoma." N Engl J Med 326(20): 1301-8.

Uchibori, M., Y. Nishida, T. Nagasaka, Y. Yamada, K. Nakanishi and N. Ishiguro (2006). "Increased expression of membrane-type matrix metalloproteinase-1 is correlated with poor prognosis in patients with osteosarcoma." Int J Oncol 28(1): 33-42. van Leeuwen, I. S., C. J. Cornelisse, W. Misdorp, S. A. Goedegebuure, J. Kirpensteijn and G. R. Rutteman (1997). "P53 gene mutations in osteosarcomas in the dog." Cancer Lett 111(1-2): 173-8.

Wayne, R. K. and E. A. Ostrander (2007). "Lessons learned from the dog genome." Trends Genet 23(11): 557-67.

Whitney, L. F. (1929). "Heredity of the trail barking propensity in dogs." J Heredity 20: 561- 562.

Withrow, S. J., B. E. Powers, R. C. Straw and R. M. Wilkins (1991). "Comparative aspects of osteosarcoma. Dog versus man." Clin Orthop Relat Res(270): 159-68.

Withrow, S. J. and D. Vail, Eds. (2007). Small Animal Clinical Oncology. Philadelphia, WB Sanders.

Wolfesberger, B., I. Walter, C. Hoelzl, J. G. Thalhammer and M. Egerbacher (2006). "Antineoplastic effect of the cyclooxygenase inhibitor meloxicam on canine osteosarcoma cells." Res Vet Sci 80(3): 308-16.

Wright, S. (1917). "The hairless dog." J Heredity 8: 519-520.

Yang, J., D. Cogdell, D. Yang, L. Hu, H. Li, H. Zheng, X. Du, Y. Pang, J. Trent, K. Chen and W. Zhang (2010). "Deletion of the WWOX gene and frequent loss of its protein expression in human osteosarcoma." Cancer Lett 291(1): 31-8.

Young, A. C., E. F. Kirkness and M. Breen (2008). "Tackling the characterization of canine chromosomal breakpoints with an integrated in-situ/in-silico approach: the canine PAR and PAB." Chromosome Res 16(8): 1193-202.

47 Zhang, J., X. Chen, M. S. Kent and C. O. Rodriguez (2009). "Establishment of a dog model for the p53 family pathway and identification of a novel isoform of p21 cyclin- dependent kinase inhibitor." Mol Cancer Res 7(1): 67-78.

Zhou, H., R. L. Randall, A. R. Brothman, T. Maxwell, C. M. Coffin and R. E. Goldsby (2003). "Her-2/neu expression in osteosarcoma increases risk of lung metastasis and can be associated with gene amplification." J Pediatr Hematol Oncol 25(1): 27-32.

Zielenska, M., J. Bayani, A. Pandita, S. Toledo, P. Marrano, J. Andrade, A. Petrilli, P. Thorner, P. Sorensen and J. A. Squire (2001). "Comparative genomic hybridization analysis identifies gains of 1p35 approximately p36 and chromosome 19 in osteosarcoma." Cancer Genet Cytogenet 130(1): 14-21.

Chapter II

Characterization of canine osteosarcoma by array comparative genomic hybridization and qRT-PCR: Signatures of genomic imbalance in canine osteosarcoma parallels the human counterpart

Andrea Y. Angstadta, Alison Motsinger-Reifb,c,d, Rachael Thomasa,b, William C. Kisseberthe,f, C. Guillermo Coutoe,f, Dawn L. Duvalg, Dahlia M. Nielsenc,h, Jaime F. Modianoi,j, and Matthew Breena,b,k

aDepartment of Molecular Biomedical Sciences, North Carolina State University, Raleigh, NC, USA bCenter for Comparative Medicine and Translational Research, North Carolina State University, Raleigh, NC, USA cBioinformatics Research Center, North Carolina State University, Raleigh, NC, USA dDepartment of Statistics, North Carolina State University, Raleigh, NC, USA eDepartment of Veterinary Clinical Sciences, College of Veterinary Medicine, The Ohio State University, Columbus, OH, USA fComprehensive Cancer Center and Solove Research Institute, The Ohio State University, Columbus, OH, USA gDepartment of Clinical Sciences, Animal Cancer Center, Colorado State University, Fort Collins, CO, USA hDepartment of Genetics, North Carolina State University, Raleigh, NC, USA iMasonic Cancer Center, University of Minnesota, Minneapolis, MN, USA j Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, St. Paul, MN, USA kCancer Genetics Program, UNC Lineberger Comprehensive Cancer Center, Chapel Hill, NC, USA

49 Abstract:

Osteosarcoma (OS) is the most commonly diagnosed malignant bone tumor in humans and dogs, characterized in humans by extremely complex karyotypes with a high frequency of genomic imbalance. Evaluation of genomic signatures in human OS using array-based comparative genomic hybridization (aCGH) has assisted in uncovering genetic mechanisms that result in disease phenotype. Previous low-resolution (10-20Mb) aCGH analysis of canine OS identified a wide range of recurrent copy number aberrations, indicating extensive genomic instability. To further interpret and compare chaotic OS karyotypes, we profiled 123 canine OS tumors by 1Mb-resolution aCGH and concluded that high frequency aberrations in canine and human OS are orthologous. Expression analysis of select cancer related genes in regions of genomic imbalance revealed that imbalance and transcriptional dysregulation in canine OS parallel human OS. Of particular interest, changes in RUNX2, TUSC3, and PTEN expression levels correlated with genomic copy number status, showcasing RUNX2 as an OS associated gene and TUSC3 as a tumor suppressor candidate. Large scale screening of genomic imbalance in canine OS further validates the use of the dog as a suitable model for human cancers, supporting the idea that dysregulation discovered in canine cancers will provide an avenue for complementary study in human counterparts.

50 Introduction: Comparison of the high quality annotated assemblies for the human and canine genomes (Lander et al. 2001; Kirkness et al. 2003; Lindblad-Toh et al. 2005) revealed a high degree of similarity. The human and canine autosomal genomes are packaged into 22 and 38 chromosomes, respectively. However, comparative genomics has allowed us now to consider these two species in the context of 200+ shared syntenic regions (Derrien et al. 2007). By comparing cytogenetic aberrations in human and canine cancers, this comparative approach already has been shown to provide an effective means to reduce the size of shared aberrations to much smaller sub-regions and thus accelerate gene discovery (Thomas et al. 2009a). In addition, it has been demonstrated that several cytogenetic abnormalities present in naturally occurring canine cancers are evolutionarily conserved with those in the corresponding human cancers, suggesting that they share an ancestrally retained pathogenetic basis for cancer (Breen and Modiano 2008; Thomas et al. 2009b). Genetic factors, combined with the spontaneous incidence of a variety of canine tumors sharing remarkable pathophysiological similarities with human cancers, now broadly support the role of the dog as a highly suitable biomedical model for cancer research (Khanna et al. 2006).

Osteosarcoma (OS) is the most commonly diagnosed primary malignant tumor of the bone in humans and dogs (Withrow and Vail 2007; Mirabello et al. 2009a; Mirabello et al. 2009b). In the USA, human OS is diagnosed in fewer than 1,000 people per year (Mirabello et al. 2009b; Paoloni et al. 2009), while in the domestic dog population the annual number of new cases is estimated to far exceed 10,000 (Withrow et al. 1991; Fossey et al. 2009). The high rate of disease occurrence in dogs provides a unique opportunity to evaluate the extent of genomic imbalance in canine OS and its comparability to the human disease. At the cytogenetic level, human OS is characterized by markedly abnormal karyotypes containing both complex structural changes (translocations and/or rearrangements) and DNA copy number changes (Bayani et al. 2007). Molecular cytogenetic evaluation of human OS karyotypes using a combination of comparative genomic hybridization (CGH), spectral karyotyping (SKY) and multicolor banding (mBAND) has identified extensive and frequent

51 genome reorganization, including a high degree of aneuploidy, gene amplification, and multiple unbalanced chromosomal rearrangements (Bridge et al. 1997; Bayani et al. 2003; Ozaki et al. 2003; Lim et al. 2004; Man et al. 2004; Lim et al. 2005). In human OS the DNA contained within human chromosome bands 6p22-p21, 8q24 and 17p12-p11.2 has been found consistently to be involved in high copy number gain (log2 ratio of tumor:reference ≥

1.0) or amplification (log2 ratio of tumor:reference ≥ 2.0) (Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Bayani et al. 2003; Squire et al. 2003; Selvarajah et al. 2008). These large regions of recurrent genomic amplification are believed to play an important role in tumor progression (Lu et al. 2008; Sadikovic et al. 2009). Recent studies in human OS have reported strong associations between copy number aberration, over expression and hypomethylation leading to disease progression (Sadikovic et al. 2009).

Until recently cytogenetic characterization of canine OS had not been performed. Preliminary low-resolution (10-20Mb) aCGH analysis of 38 cases of canine OS identified a high degree of genomic instability resulting in extensive cytogenetic disorganization, consistent with observations of human OS (Thomas et al. 2009b). The high degree of aneuploidy and structural instability present in both human and canine OS suggest that the disease may involve alterations of the common signal transduction pathways. Genomic disorganization in both species has been reported to involve the dysregulation of well known tumor suppressor genes (TP53, RB1, PTEN, CDKN2A, CDKN2B) and oncogenes (MYC and MET) (Feugeas et al. 1996; Levine and Fleischli 2000; Tsuchiya et al. 2000; Gokgoz et al. 2001; Levine et al. 2002; Overholtzer et al. 2003; Kirpensteijn et al. 2008; De Maria et al. 2009; Zhang et al. 2009). Several previous studies have sought to define the roles of specific tumor suppressor genes and oncogenes in primary canine OS as well as canine OS cell lines, using expression and proteomic analysis (Levine and Fleischli 2000; Levine et al. 2002; Kirpensteijn et al. 2008; Selvarajah et al. 2008; Fossey et al. 2009; Zhang et al. 2009). The scope of these studies did not allow the evaluation of dysregulation of canine OS from a genome-wide view. In this current study we investigated further the chaotic karyotypes in

52 canine OS by profiling 123 cases with genome-integrated 1 Mb-resolution aCGH, and subsequent florescence in situ hybridization (FISH) analysis. The focus for sample collection was four breeds, each of which has a high incidence of OS; Golden Retriever, Great Pyrenees, retired racing Greyhound, and Rottweiler. Samples from other breeds were collectively described as a ‘non-target’ breed group. We identified recurrent high frequency DNA copy number aberrations (CNAs) within our sample set orthologous to regions of genomic imbalance in human OS and observed that genomic instability is associated with morphological subtypes (osteoblastic, chondroblastic, and fibroblastic) and breed. To evaluate the effect of genomic CNA on gene expression, qRT-PCR analysis of a subset of genes was performed on 26 canine OS samples.

Our genome-wide study of canine OS represents the largest canine cancer patient cohort evaluated by aCGH to date and serves to advance our knowledge of the genomic abnormalities present in the canine disease. In spite of the complexity of the karyotypes we identified several key similarities shared between human and canine OS that further reinforce the dog as a valuable spontaneous biomedical model for human OS. In addition, our characterization of CNAs in canine OS provides valuable insight into the genetic mechanisms responsible for disease phenotype that may result in identification of prognostic markers and therapeutic targets for dogs and humans.

Materials and Methods:

Tissue Specimens Canine OS tissues were collected from cases of spontaneously occurring canine OS either by institutional or community veterinary practices in the United States between November 1999 and July 2009. A total of 123 canine OS tumor specimens (Table 1 and Supplementary Table 1) were obtained during routine clinical evaluation under approved protocols for informed consent by owners and prior to the initiation of chemotherapy or radiotherapy. Tissue specimens were processed as described previously (Thomas et al. 2005). Briefly, tumor tissue was surgically removed under sterile conditions as a part of a diagnostic

53 biopsy procedure and grossly normal tissue was excised from the specimen. Tumor tissue was then 1) fixed in formalin and submitted for histological evaluation of hematoxylin and eosin (H&E)-stained sections, using the criteria of Kirpensteijn et al (Kirpensteijn et al. 2002), 2) either processed immediately for DNA/RNA extraction or snap-frozen in liquid nitrogen for subsequent extraction, and 3) used to initiate primary cell cultures in RPMI-1640 medium supplemented with 10% fetal bovine serum, 2mM L-Glutamine and 100μg/ml Primocin.

Table 1: Summary of breed and morphological subtype of 123 canine OS tumor samples analyzed for DNA copy number aberrations by genome integrated 1Mb-resolution aCGH. A detailed list of all cases is provided in supplementary table 1. OS subtype Total Breed Cases Chondroblastic Fibroblastic Osteoblastic Undetermined American Bulldog 1 1 Bernese Mountain Dog 1 1 Borzoi 1 1 Boxer 1 1 Briard 1 1 Collie 1 1 Doberman 4 1 3 Flat Coat Retriever 2 2 German Shepherd 1 1 Golden Retriever 22 3 1 13 5 Great Dane 2 1 1 Great Pyrenees 13 3 2 4 4 Greyhound 25 1 1 16 7 Irish Setter 2 2 Labrador Retriever 3 1 2 Mastiff 2 1 1 Mix 4 1 3 Rhodesian Ridgeback 2 1 1 Rottweiler 34 4 6 21 3 Saint Bernard 1 1 Total 123 12 13 67 31

54 Array Comparative Genomic Hybridization (aCGH) A genomic microarray comprising cytogenetically-mapped bacterial artificial chromosome (BAC) clones was used for array CGH analysis. Clones used for this array were distributed at approximately 1 Mb intervals throughout each dog autosome and the X chromosome (all from the CHORI-82 dog BAC library (http://bacpac.chori.org, BACPAC Resources, Children’s Hospital Oakland Research Institute, Oakland, CA) and was supplemented with additional clones representing cancer related genes (Thomas et al. 2008; Thomas et al. 2009a). Reference DNA samples used in this study were derived from peripheral blood samples, either from patient’s own blood or from equimolar pools of DNA from ≥10 cancer free, healthy, sex-mismatched dogs of the same breed. Conventional phenol:chloroform extraction was used to isolate high molecular weight DNA from both tumor and reference samples. Tumor (test) and reference DNA samples were differentially labeled with Cyanine-3-dCTP and Cyanine-5-dCTP respectively and then combined in the presence of dog Cot1 DNA as competitor. The probe mixture was denatured and applied to the microarray for 40 hours at 37ºC as described previously (Thomas et al. 2007). Following post hybridization stringency washes the arrays were scanned at 10μm resolution (Perkin Elmer ScanArray Express). The fluorescence intensity at each genomic locus was quantified

and used to calculate the log2 ratio of test:reference DNA for each locus on the array.

Threshold limits for CNAs were set at log2 ratio (test:reference) values equivalent to 1.15:1 (copy number gain) and 0.85:1 (copy number loss), and the data analyzed by the aCGH Smooth algorithm (Jong et al. 2003). Briefly, the aCGH Smooth algorithm uses a clustering- based approach to identify homogenous groups of probes that define regions of gain or loss.

The thresholds selected were used to discretize the log2 ratio data from all 123 canine OS aCGH data sets and the ratios converted to ordinal data (loss, normal, gain), which was then evaluated for disease association. The percent of CNAs was calculated for each of the autosomal loci on the array. All 123 cases were analyzed for inflection points in regions of chromosomal gain, loss, or normal copy number across cases, thus collapsing the data into larger regions of contiguous copy number status. To make direct comparisons of the genomic imbalance shared between human and dog, orthologous regions of human

55 chromosomal segments defined previously to have a high rate of aneuploidy in OS were identified in the canine genome using the UCSC Genome Browser (http://genome.ucsc.edu/) and Autograph tool (Derrien et al. 2007).

Fluorescence in-situ Hybridization (FISH) Where fresh OS tumor tissue was available, primary cell cultures were initiated using the same biopsy specimen that was used for DNA isolation. Metaphase chromosome preparations and interphase nuclei were produced either directly from the primary tumor tissue or from low passage (n≤3) primary cultures using conventional techniques of colcemid arrest, hypotonic treatment, and methanol-glacial acetic acid fixation (Breen et al. 1999). Twenty-five micron sections were obtained from zinc-fixed and paraffin-embedded OS tissue and deparaffinized with xylene prior to immersion in 100% ethanol. Samples were then immersed in pepsin at 37ºC for 20 min and left at 4ºC for 48 hrs followed by resuspension in 1xPBS and methanol-glacial acetic acid fixation. Cell suspensions were dropped onto clean glass slides and air-dried. Multicolor FISH analysis was performed to evaluate the distribution of selected copy number changes indicated by the aCGH data. BAC clones were selected from within chromosomal regions displaying a range of both normal and aberrant copy number changes, differentially labeled and used as probes as described previously (Breen et al. 2004). Overlapping BAC clones for FISH were selected using the UCSC genome browser (http://genome.ucsc.edu/) for candidate gene-specific regions exhibiting high genomic imbalance and/or possible cancer-related function, including cell proliferation, transformation, and/or apoptosis (Supplementary Table 2). Metaphase preparations from clinically healthy dogs were used first to confirm that all BAC probes in the FISH set exhibited the expected copy number in non-neoplastic cells (n=2). Images from a minimum of 30 cells were used to evaluate the copy number status of each probe. Due to the increased depth of field of fixed nuclei, images were acquired using incremental capture within the SmartCapture 3 program (Digital Scientific Ltd, Cambridge, UK) to accurately identify and quantify signals in different focal planes.

56 Quantitative RT-PCR

A subset of the sample population (n=26) was selected for extraction of total RNA. These were processed using the Qiagen RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions, following grinding of the tumor tissue with a mortar and pestle in the presence of liquid nitrogen. The quality and quantity of RNA was assessed using spectrophotometry (Nanodrop 1000, Nanodrop Technologies) and a RNA 6000 Nano Labchip on the Agilent 2100 Bioanalyzer (Agilent). Samples with an RNA integrity number (RIN) > 6 were included in the study. Total RNA was treated for residual genomic DNA (TURBO DNA-free Kit, Ambion) and 1μg was reverse transcribed into cDNA (Quantitect Reverse Transcription Kit, Qiagen). Quantitative real-time PCR (qRT-PCR) analysis was then performed using the QuantiFast SYBR Green Kit (Qiagen) and an iCycler (BioRad) to evaluate the transcriptional status of genes selected based on aCGH data of these loci. Primers were designed using primer blast (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) to ensure each sequence was specific for the gene of interest within the dog genome (Table 2). The reaction efficiency for all primers was calculated, using serial dilutions of cDNA isolated from non-neoplastic dog femurs and only primer pairs with efficiencies ranging from 95%-105% were used in subsequent analysis. Relative quantification was performed as described elsewhere (Pfaffl 2001) using c12orf43 as the reference gene, selected on the basis of having both normal copy number in >97% of canine OS sample evaluated by aCGH, and a stable expression pattern in qRT-PCR evaluation of 11 canine osteosarcoma samples (Supplementary Table 3) and 10 non-neoplastic samples (Tsai et al, in preparation). We expressed the relative mRNA levels in the tumors as -ΔΔCT where ΔCT is difference in the threshold PCR cycle (Ct) value of mRNA of our gene of interest and the corresponding control (c12orf43) in each reaction. Fold change was then calculated using the primer efficiencies and -ΔΔCT.

57 Table 2: Primer sequence of genes analyzed by qRT-PCR analysis of canine OS RNA.

Accession Number Gene Location Sense Primer (5'-3') Antisense Primer (5'-3') Chr6q22:41.90- XM_859874.1 TSC2 41.94Mb TCGTCGGACATCAACAACAT CCGCAGAGTCCGTGTTAGAT Chr9q11.2:3.29 XM_540336.2 RHOC 4-3.329Mb CATCGACAGCCCCGACAGCC GCACGGGCTCCTGCTTCATCT Chr12q13:16.7 XM_532158.2 RUNX2 3-16.84Mb TCACTCCACCACCCCGCTGT TGAAGCACCTGCCTGGCTCT Chr13q13:28.2 NM_001003246.1 MYC 3-28.24Mb TCGCCTATTTGGGAAGACAC AAGCTGACGTTGAGAGGCAT Chr16q23:41.8 AGTGCCATGGTCCAAATCACAT XM_844352.1 TUSC3 2-42.04Mb GCCTAGTGGGATTAGGCCTGGTGG CTTC Chr26q25:40.9 NM_001003192.1 PTEN 2-40.98Mb ACTTTGAGTTCCCTCAGCCA AGGTTTCCTCTGGTCCTGGT Chr26q21:19.8 XM_849238.1 C12orf43 5-19.86Mb GCCTGGGGCTTGGAGCAGTG TGGGCTCGGAATTCGGGGGT

Statistical analysis:

Clinical data for all 123 cases in this study were compiled. A total of 99 cases had data on survival time, defined as the time period from the date of diagnosis until death (including death by euthanasia). Patients still alive at time of analysis, or deceased from causes other than their OS, were censored. The Kaplan-Meier method was used to generate survival curves and tests of nonparametric proportional hazards were used to assess significance between the groups in question for the different tests.

Principal components analysis (PCA) (Pearson 1901) was used to characterize global patterns of copy number variation across the entire genome. By performing an eigenvalue decomposition of a data matrix after centering the data for each attribute around its mean, PCA transforms a number of possible correlated variables into a smaller number of uncorrelated variables. These principal components account for maximal amounts of variation in the data (Pearson 1901). PCA was used in the current study to evaluate genome- wide differences in aberration frequency between breeds and morphological subtypes (Pritchard et al. 2000; Reich and Goldstein 2001; Hoggart et al. 2003).

Association analyses were performed with Fisher’s exact tests to correlate aberration frequencies with the five breed groups in the OS dataset; Golden retriever, Great Pyrenees,

58 Greyhound, Rottweiler, and the non-target breed group. Fisher’s exact tests also were used to associate aberration frequencies within the three cellular morphological subtypes of OS: osteoblastic, chondroblastic and fibroblastic. To control family-wise error rates and correct for the multiple comparisons performed, a Bonferroni correction was performed to derive empirically p-value cut-offs of significance that correspond to a family-wise error rate of 0.05 (Abdi 2007). The correlation between aberration region size and percentages of CNAs in our case set was tested by linear regression. A student’s t-test and/or non-parametric Kruskal-Wallis test was performed to evaluate the differences between breed groups and morphological subtypes within PCA components and when evaluating differences between groups in our qRT-PCR data. Statistical analyses of the data were performed in JMP Genomics v4 and SAS 9.1.3 (SAS Institute).

Results:

Clinical assessment

A total of 123 cases of canine OS with diverse pathological diagnoses (Table 1 and Supplementary Table 1) were profiled within this study. The majority of cases for which we had morphological information (n=93) presented with the osteoblastic (73%) morphological cell subtype (Table 1), while the remainder comprised approximately equal proportions of fibroblastic (14%) and chondroblastic (13%) OS. Of the dogs in this study with characterized tumor location (n=73), 63% and 37% had OS in the front and rear legs, respectively. The most common anatomical location for the tumors was in the appendicular skeleton (radius 35%, humerus 27%, tibia 25%, femur 12%, ulna 0.03%) with one patient having the disease in the maxilla (Supp. Table 1). Though a total of 19 breeds were represented in our study population, it is important to highlight that these cases were not selected randomly. We specifically targeted a collection of just four breeds totaling 94 (76%) of the 123 cases; retired racing Greyhound (n=25), Great Pyrenees (n=13), Golden Retriever (n=22), and Rottweiler (n=34) (Table 1). The remaining 29 cases (including mixed-breed dogs) were selected randomly and combined into a fifth, ‘non-target’ breed group for statistical analyses. The clinical characteristics of our cohort, however, are very much in

59 accordance with a typical random selection of dog breeds and so notwithstanding the intentional bias in the breeds selected, there was no apparent overall bias in the frequency of tumor histology, gender, age at diagnosis etc.

Survival time, measured in weeks from the date of diagnosis to death, and detailed treatment information (Supp. Table 1) were available for 99 and 79 dogs respectively, in the study. These 79 cases were grouped into three categories: those that received amputation and chemotherapy (n=56); amputation alone (n=15); and palliative care (n=8). All clinical covariates were tested for statistically significant relevance to survival time. The breed (p=0.7982), gender (p=0.5747), tumor location (p=0.0767), and morphological subtype (p=0.2969) were not shown to have a statistically significant effect on survival time. As expected, treatment type received by each dog did reveal a statistically significant effect on survival time, demonstrated by a nonparametric proportional hazards test p=0.0004 (Supplementary Figure 1). These data indicated that more aggressive treatment corresponded to longer survival time. Also, in subsequent iterations of survival analysis with array data, type of treatment was always significantly correlated with survival time.

Abundant genomic instability in canine OS

Cases in our canine OS cohort typically presented with chaotic aCGH profiles, comprising whole chromosome aneuploidy, single locus CNAs and structural rearrangements, all of which have been shown previously both in canine and human OS (Bayani et al. 2003; Bayani et al. 2007; Selvarajah et al. 2008; Thomas et al. 2009b). The extent of the genomic imbalances is represented in Figure 1, which denotes the percentages

of gains and losses at ~1Mb intervals across all 38 canine autosomes based on log2 ratios before and after data processing by the aCGH Smooth algorithm (Jong et al. 2003) (Figure 1A,B). Prior to the data being processed by the aCGH Smooth algorithm (Figure 1A) all loci exhibited some frequency of copy number gain and/or loss, reiterating the chaotic nature of the disease. Application of the aCGH smooth algorithm exposed the most prominent

60 genomic regions containing CNAs within our canine OS dataset. Analysis of the 123 cases for inflection points in regions of chromosomal gain, loss, or normal copy number across cases collapsed the data into 1,066 larger regions of recurrent contiguous copy number status (Figure 1C). The large number of shared regional aberrations indicates the broad positional range of gain/loss in each OS patient. The distribution shift across the autosomes shows the degree of disease complexity per chromosome and the variance of genomic aberrations between patients (Figure 1B,C). Subsequent data analysis was based on comparison between aberration frequencies within these 1,066 regions of the genome. Setting the level of recurrence at ≥10% aberration, regions of the genome were more likely to experience copy number gain (15%) than copy number loss (10%). Within the sample population seven chromosomes contained regions that had >10% copy number loss, while 23 chromosomes had regions with a frequency of copy number gain >10%. Chromosome 23 (CFA 23) appeared to be the least aberrant chromosome with little variation between patients. As seen in Figure 1C, 46 loci (spaced at ~1Mb intervals) on CFA 23 could be segmented into just nine aberration regions, with no regional aberration frequency >4%. The majority of regions experiencing a >10% aberration frequency of gain or loss in canine OS were <5Mb in size, while regions with < 5% aberration frequency were generally larger in size (Figure 2). Linear regression analysis concluded this pattern of high aberration frequency and smaller aberration size to be statistically significant for regions experiencing both gain (p=1.549e-8) and/or loss (p=0.00987).

Figure 1: Frequency of copy number aberration in 123 canine OS samples. The 38 dog autosomes are listed on the x-axis. The y-axis shows the percentage of cases with copy number gain (red dots) and copy number loss (green dots), for each ~1 Mb interval. (A) Graphical representation of the aberration percentages across 123 canine OS patients based on the raw log2 ratio values for chromosomal loci. (B) Data in A after being processed with the aCGH Smooth algorithm (Jong et al. 2003). (C) Graphical representation of aberration percentages in B, now for 1,066 regions of collapsed data, based on aberration commonality. Regions of CNAs were collapsed according to inflection points of copy number gain, loss, or normality across 123 canine OS patients which caused the chromosomal size and distribution differences between A, B (spaced at 1 Mb intervals) and C (which is spaced according to the number of recurrent aberration regions present within each chromosome).

Identification of recurrent genomic aberrations in the 1,066 regions allowed further characterization of CNAs and were in agreement with our previous findings (Thomas et al. 2009b). The highest frequency (>20%) of genomic copy number increase in the canine cohort was seen for regions on five chromosomes; CFA 13q11-22.2 (cen-66Mb) (% gain

62 >10% for all loci), CFA 9q11.1-q11.2 (cen-6.0Mb), CFA 17q11.1-11.2 (cen-5.4Mb), CFA 31q15.3 (38-42Mb), CFA 34q11 (14.23-14.41Mb) (Table 3, Figure 1B,C). Several of these regions are well annotated in the canine genome assembly and contain a variety of genes (Table 3), some of which have been shown previously to be involved in human and/or canine cancers, including OS (Kim et al. 1994; Shay 1997; Nasir et al. 2001; Mueller et al. 2007; Kow et al. 2008; Tang et al. 2008; Eckerle et al. 2009; Narumiya et al. 2009). When comparing canine and human OS, it is interesting to note that the orthologous regions: HSA 6p21.1 (40.6-45.2 Mb)/CFA 12q13 (12.21-16.73Mb), containing RUNX2; and 8q24 (117.7- 147.6Mb)/ CFA13q12.3-q13 (19.1Mb-41.3Mb), containing MYC both displayed >15% gain in our canine OS cohort (Figure 1C).

Figure 2: Bivariate fit of CNA percent loss/gain for 123 canine OS cases by the physical size of the aberrations in Mb.

Genomic regions exhibiting the highest frequency of copy number loss involved only three chromosomes; CFA 26q22-q23 (39.66-41.72Mb); CFA 16q11-25.2 (cen-62Mb) (% loss >10% for all loci); and CFA 18q22.3 (32.16-32.37Mb) (Table 4; Figure 1B,C). The tumor suppressor gene PTEN (40.8-40.9Mb) is located within a BAC clone on our canine 1Mb array and was part of the region of loss on CFA 26q22-q23. Loss of this gene has been noted previously in both canine and human OS (Mueller et al. 2007). As an entire

63 chromosome, CFA 16 experienced the highest frequency of loss with all 33 aberrant regions containing a loss frequency >10%. Several genes on CFA 16, including MTUS1 and TUSC3, are possible tumor suppressor gene candidates (Table 4).

Table 3: The top twenty autosomal regions that exhibited the highest percentages of copy number gain in our canine OS cases. Genes are those annotated and present within the corresponding regions in the dog (canFam2) genome, extracted using the UCSC genome browser (http://www.genome.ucsc.edu/). Start End % Chromosome (Mb) (Mb) Gain Genes 13q13 26.92 27.11 31.93 13q13 23.95 24.15 31.25 ACP1, FAM150B 17q11.1 3.20 3.40 30.63 13q13 26.02 26.22 30.00 SQLE, NSMCE2 MAFA, RHOJ, RAC3, RHOC, MAFG, SIRT7, P4HB, 9q11.1 3.27 4.75 29.27 PDE6G, ACTG1, BAIAP2, RPTOR, CARD14 KRTAP, TAF9B, UBE2G2, ITGB2, ADARB1, SLC19A1, 31q15.3 40.69 42.07 29.27 COL18A1, PCBP3, COL6A1, MCM3AP, PCNT 13q13 25.12 25.31 27.97 TNFRSF11B, MAL2, CTGF, ENPP2, RTN3, DEPDC6, 13q13 20.58 22.90 27.64 TAF2, SNTB1, BAI1, JRK, PSCA, SLURP1, CYP11B2, GLI4, RHPN1, 13q21.1 39.70 40.85 27.64 MAFA, NAPRT1, DGAT1 13q21.1 38.34 38.55 26.32 PTK2 9q11.2 5.88 6.04 25.64 BIRC5, TMC8, TMC6, 13q12.3 18.43 18.61 25.42 34q11 14.23 14.41 25.42 SLC6A3, TERT 13q13 28.09 28.27 25.20 MYC 13q13 29.35 29.53 24.56 31q15.3 39.66 39.88 24.56 PDXK, HSF2BP 13q11 3.53 6.17 24.39 STK3, NPM1, VPS13B, COX6C, POLR2K 17q11.1-11.2 5.25 5.45 24.35 31q15.3 38.70 38.91 23.93 C2CD2, ZNF295, UMODL1, ABCG1 13q21.1 34.87 35.06 23.73 SNORA32

64 Table 4: The top twenty autosomal regions that exhibited the highest percentages of copy number loss in our canine OS cases. Genes are those annotated and present within the corresponding regions of the dog (canFam2) genome, extracted using the UCSC genome browser (http://www.genome.ucsc.edu/) Start End % Chromosome (Mb) (Mb) Loss Genes 26q23 41.51 41.72 31.30 LIPF, LIPK, LIPN, LIPM, ANKRD22 26q23 40.84 41.08 29.75 ATAD1, PTEN 26q23 40.58 40.74 25.00 MINPP1 26q22 39.66 39.85 23.33 PRKG1 16q21 29.32 29.52 18.92 ADAM32, ADAM9 RBPMS, SMS, DCTN6, DUSP4, PPP1R3B, MFHAS1, 16q22 36.75 39.18 18.70 GOT2 FGF20, FGF22, ZDHHC2, CNOT7, MTMR7, PDGFRL, 16q23 43.09 45.46 18.70 MTUS1, FGL1 16q14 19.52 19.70 18.33 16q12 9.90 10.10 18.26 TRYX3 BRAF, NDUFB2, RAB19, TBXAS1, HIPK2, CHRM2, 16q12-13 11.10 15.51 17.89 SHROOM3 16q14 21.01 21.97 17.89 PAXIP1, HTR5A, INSIG1, EN2, RBM33, SHH 16q23 41.36 42.27 17.89 SGCZ, TUSC3, MAGT1 16q23 46.31 46.50 17.65 16q12 5.56 5.74 17.50 CNTNAP2 16q12 9.05 9.24 17.50 TAS2R16 16q12 6.33 6.50 17.36 CNTNAP2 16q12 7.64 7.82 17.36 16q22 40.22 40.42 17.36 SGCZ 16q24 49.87 50.06 17.21 WWC2, DNAJC19 18q22.3 32.16 32.37 17.17

FISH validation of aCGH data

To validate the aCGH data, interphase and metaphase FISH analyses were conducted. As an example, Figure 3 demonstrates the typical extent of numerical (and structural) aberrations present in the cases evaluated in this study. Additional metaphase FISH analysis is found in Appendix II. In addition to cells from primary cultures, 18 zinc-fixed paraffin embedded canine OS cases (highlighted in Supplementary Table 1) were evaluated by interphase FISH. Multiple BAC clone probe sets were used (Supplementary Table 2) to

65 visualize copy number status of selected genes (TSC2, RHOC, RUNX2, MYC, TUCS3, PTEN). The FISH analysis further confirmed our aCGH analysis and the high degree of tumor heterogeneity in canine OS (Supplementary Figure 2 and 3, Appendix II).

Figure 3:(A) Whole genome 1Mb aCGH profile of a 12 year old female Golden Retriever diagnosed with osteoblastic OS located in the left proximal humerus, co-hybridized with DNA derived from a blood sample from the same patient. The data profile is plotted as the log2 tumor DNA: reference DNA ratio after median block normalization and background subtraction of the replicate spots for each locus on the array. Points colored in grey represent loci that were indicated as having normal copy number (n=2), while loci colored red and green represent those that had an increase or decrease, respectively, in copy number. Selected BAC clones for FISH are colored to represent the fluorochrome with which each one was labeled and arrows point to their corresponding position on the aCGH profile. The clone 326P22 (orange signal) was used as a control for the FISH analysis since it was not indicated as having aberrant copy number by aCGH (B) Interphase and metaphase FISH analysis of the OS case used in A. i) An interphase nucleus from the patient, representative of one of the 40% of cells that exhibited normal copy number for all three BAC clones (n=2). ii) An abnormal interphase nucleus from the patient presenting with eight copies of two CFA 13 BACs, 287D11 (red signal) and 326P15 (green signal), and two copies of a control BAC 326P22 (orange signal). iii) Metaphase spread from the patient demonstrating structural aberrations of CFA 13 (inset) along with three copies of 287D11 and 326P15. (C) Compilation of copy number data, based on FISH analysis. The y-axis demonstrates the percentages of nuclei for a certain copy number status for each clone and the x-axis represents the BAC clone address, chromosome, position, and log2 ratio value from the corresponding aCGH analysis.

67 Breed and morphological subtype specific associated DNA copy number aberrations

The high level of genomic imbalance present within the canine OS patients attests to the arduous/challenging nature of characterizing recurrent aberrations and variations between different dog breeds and morphological subtypes. Figure 4 illustrates the variation in aberration frequencies of the 1,066 regions of recurrent aberration between the five major breeds groups (Figure 4A) and three morphological OS subtypes (Figure 4B). Principal components analysis (PCA) (Pearson 1901) was performed to evaluate differences in genome-wide DNA copy number aberrations evident in our five breed groups and three morphological subtypes. Using the entire sample set for the breed PCA, the results indicated four significant components that define the data. The resulting eigenvalues for the first four components were tested for association with breed using the nonparametric Kruskal-Wallis test. These results indicated a statistically significant difference in data distribution between breed status and the third component (Supplementary Figure 4) (p=0.0022). A Fisher’s exact test, however, revealed no significant differences (after applying a multiple testing correction) between aCGH-defined regional aberrations and breed groups, when comparing results between all five breed groups together and individually. It is likely that this is a matter of limited sample size, the power of the analysis being reduced by the small numbers for each individual breed. The Fisher’s exact test indicated one region, CFA 15q12 (10.64- 10.84Mb), as trending toward displaying a different aberration frequency between the five groups (uncorrected p< 0.003) (Figure 4A). Forty four percent (11/25 cases) of Greyhounds had a high frequency of DNA copy number gain in this region of CFA 15, while in the other breed groups the frequency of gain was greatly reduced; Great Pyrenees (0%; 0/13 cases), Golden Retriever (4.5%; 1/22 cases), Rottweiler (17.6%; 6/34 cases), and non target/other breeds (6.9%; 2/29 cases).

Figure 4: Further characterization of copy number aberration occurrences across 1,066 regions of aberration commonality in the 38 dog autosomes (x-axis). Percentages of copy number gain/loss are shown on the y-axis (A) Percentage CNA in different dog breeds represented within our 123 case dataset. The ‘non-target’ breed group is a combination of 29 dogs representing 15 breeds, each with fewer than five cases (Table 1). (B) Percentage of CNA present in the 93 (of 123) cases within the study that presented with information about morphological subtype; fibroblastic, chondroblastic, and osteoblastic.

PCA was conducted on the 93 samples where OS morphological subtype information was available. A scree plot of results indicated four significant components also that define the global data. The eigenvalues generated for the first four components were tested for association with morphological subtype using the nonparametric Kruskal-Wallis test. No one component was found to have a statistically significant difference in any of the morphological subtypes. Results of a student’s t-test found a statistically significant difference between eigenvalues defined by the PCA in component 2 for the osteoblastic and

69 chondroblastic subtypes (Bonferroni corrected p<0.05) (Supplementary Figure 5). The Fisher’s exact test comparing the aCGH aberration results for patients with osteoblastic or chondroblastic OS also found a significant difference (Bonferroni corrected p< 0.05) on two consecutive regions of CFA 5q11-q12 (3.14-7.02Mb and 8.01-9.26Mb). Of the 12 chondroblastic OS patients, 4 (33%) demonstrated copy number gain in both these two regions, while none of the 67 osteoblastic patients displayed copy number gain of these regions (Figure 4B).

qRT-PCR

To assess the correspondence of DNA copy number changes present in canine OS with an increase or decrease in expression we analyzed six candidate genes (TSC2, RHOC, RUNX2, MYC, TUCS3, PTEN). We compared expression levels to those present in RNA isolated from a non-neoplastic dog femur from a 1-2 year old pit bull mix. Evaluation of 26 canine OS cases indicated that four of the six genes showed expression changes (>2 fold change) that were in agreement with the corresponding genomic DNA CNA revealed by aCGH (Figure 5). The BACs on our 1Mb array that contaned MYC and TSC2 both had copy number gain (25.2% and 16.3%) when profiling 123 cases of canine OS, but we did not see any changes in expression levels of these two genes. qRT-PCR data indicated that RHOC and RUNX2 were over-expressed and PTEN and TUSC3 were under-expressed in our canine OS patients. Although RHOC was over-expressed it did not show a significant association between mean fold change and CNA when using a non-parametric Kruskal-Wallis test (p=0.4715). Loss of PTEN expression is consistent with previous research (Levine et al. 2002) on canine OS and unlike RHOC we did find a statistically significant association between the mean fold change for cases that exhibited either a normal or loss of copy number using a non-parametric Kruskal-Wallis test (p=0.0158). In addition, RUNX2 (p=0.0077) and TUSC3 (p=0.0449) displayed a statistically significant association between mean fold change and CNA suggesting that CNAs for these particular genes does affect the expression level. Using a Cox Proportional Hazard test expression changes in these patients were not

70 identified as having a significant effect on their survival time. This is because treatment still remains the driving factor affecting patient outcome (p=0.0016) (Supplementary Figure 1).

Figure 5: Quantitative RT-PCR on 26 canine OS patients for six genes located in regions that had a high occurrence of CNAs. The results are displayed as fold change in expression relative to non-neoplastic dog femur RNA normalized by the expression of c12orf43. The color of the marker refers to the array CGH call (gain, loss, or normal) of each OS case.

Discussion

In human OS the eminently complex karyotypes and rarity of the disease have confounded the ability to define the genetic changes that ultimately lead to the disease phenotype. Several recent reviews have sought to summarize current knowledge concerning

71 the abundance of genomic changes present in human OS (Papachristou and Papavassiliou 2007; Clark et al. 2008; Tang et al. 2008) and have noted the genomic instability present in the disease. Likewise, our own recent evaluation of 14 human OS tumors with high-density oligonucleotide CGH arrays confirmed the cytogenetically chaotic nature of the disease (Thayanithy et al. submitted). In the domestic dog OS is not a rare disease, with reported annual cases being at least ten fold higher than in humans, yet very few studies have profiled the disease for its genomic instability. In previous preliminary studies we reported that dogs (n=38) with OS have a broad range of numerical and structural cytogenetic aberrations, leading to dosage imbalances of known oncogenes and tumor suppressor genes (Thomas et al. 2009b). In the present study we profiled 123 canine OS cases with various clinical backgrounds using a custom 1Mb-resolution aCGH platform comprising genome integrated BAC clones. We observed a high degree of tumor heterogeneity resembling that present in human OS, found regional genomic aberrations that are associated with specific breeds and morphological subtypes, and identified new target genes involved in canine OS. In addition we analyzed several genes in regions of genomic instability by qRT-PCR to determine if alteration to gene dosage corresponded with transcriptional dysregulation. Thus, we have concluded that orthologous genomic regions of humans and dogs have some of the same types of CNAs in OS as well as similar expression patterns.

When analyzing canine OS for genomic signatures, regions experiencing a high frequency of copy number aberration in our cohort were more likely to be smaller in length (Mb) than regions with a low aberration frequency (Figure 2). This phenomenon supports our conclusion that canine OS presents with high genomic imbalance and with copy changes more likely to involve sub-chromosomal regions than entire chromosomes. DNA copy number gain was much more prevalent than DNA copy number loss, involving >3 fold the number of chromosomes (23 ‘vs’ 7), a feature that is very similar to the distribution of CNAs in human OS (Man et al. 2004; Selvarajah et al. 2008). CFA 13 contained the most regions with the highest frequency of copy number gain (>20% for all regions) in our dataset, indicative of a whole chromosome copy number increase in some cases (Figure 1, Table 3).

72 In human OS, high-level amplifications have been seen in 8q21-24 (Man et al. 2004; Selvarajah et al. 2008; Sadikovic et al. 2009), which is orthologous to the centromeric half of CFA 13q12.3-q13, providing appeal for further study. An example of the extent of CNA in canine OS is shown by our hybridization of two BAC clones within this CFA 13 region onto cells derived from the OS of a 12-year female Golden Retriever (Figure 3). We found more than eight copies of both clones, along with structural rearrangements of CFA 13 in this case (Figure 3B), demonstrating that copy number aberrations are associated with structural alterations in the OS genome. Numerous genes map within this shared region on CFA13 (Table 3), several of which (including ENPP2, MYC, PSCA, RHOJ, RAC3, RHOC, CARD14, BIRC5, TMC6, and TMC8) already have been associated with known cancer phenotypes, cell growth and intracellular signal transduction pathways (Kurima et al. 2003; Courts et al. 2007; Mueller et al. 2007; Tang et al. 2008; Eckerle et al. 2009; Jonkers and Moolenaar 2009; Narumiya et al. 2009). Seven to 12 percent of human OS have amplifications of the MYC oncogene (Tang et al. 2008) and copy number gains of MYC have been confirmed previously in canine OS (Thomas et al. 2009b). FISH analysis of MYC on cases in our dataset also demonstrated copy number gains of the oncogene (Supplementary Figure 2) and our expression analysis of MYC concluded that although the gene is subject to recurrent gene dosage increase in our canine OS cohort it does not appear to produce a corresponding increase in expression levels (Figure 5). Previously, over-expression of MYC has been found to play a role in human OS (Gamberi et al. 1998; de Nigris et al. 2007), yet very recently Sadikovic et al showed that no significant changes in expression of MYC occurred in tumors in comparison with normal osteoblasts (Sadikovic et al. 2010). Our findings in canine OS support this recent observation. Further investigation into effects of myc protein expression would be necessary to understand the role of copy number gain of MYC in canine OS.

The orthologous region to human chromosome 6p21.1 (40.6-45.2 Mb), a key region of copy number gain in human OS (Selvarajah et al. 2008; Sadikovic et al. 2009), is CFA12q13 (12.21-16.73Mb). While this region was not one of the top twenty regions that experienced the highest percentages of copy number gain (Table 3), it did have >15%

73 frequency of gain across the sample population. Within this region is the RUNX2 gene, a member of the RUNX gene family of differentiation mediators expressed at different stages of osteoblast development (Ito 2004). It has been hypothesized that the amplification and over-expression of RUNX2 in primary human OS tumors could disrupt G2/M cell cycle checkpoints, and downstream OS-specific changes, such as genomic polyploidization and failure of bone differentiation (Sadikovic et al. 2009). When studying the functions of RUNX2 in immortalized mouse calvarial derived MC3T3-E1 osteoblasts and rat and human

OS cell lines, a high expression level of RUNX2 in late G1 and mitosis was reported (San Martin et al. 2009). These data suggested that the cell cycle control of RUNX2 gene expression is impaired in OS and this dysregulation may lead to the pathogenesis of OS (San Martin et al. 2009). We conducted qRT-PCR to identify if copy number gain of RUNX2 is associated with over-expression of the gene in canine OS. We found a statistically significant difference between cases with normal or increase in copy number and change in expression. This same pattern of increase in copy number leading to an increase in expression has been found in human OS (Sadikovic et al. 2009) making this gene valuable to study of OS in both species. Additionally, recent evaluation of 16 genes as potential biomarkers of human OS oncogenesis and chemotherapy response concluded that a significant increase in RUNX2 expression corresponded with poor patient response to chemotherapy relative to the good responders (Sadikovic et al. 2010). Thus, our conclusions concerning RUNX2 suggest that it is an OS-associated gene, important for the disease progression and of interest for further characterization in canine OS. To our knowledge this is the first indication of the possible role of RUNX2 in canine OS.

Regions experiencing highly recurrent copy number loss reside largely on two chromosomes, CFA 26 and CFA 16 (Table 4). In human OS the PTEN tumor suppressor gene has a high frequency of copy number loss, often a homozygous deletion resulting in complete loss of PTEN expression (Freeman et al. 2008). Our current findings (Figure 1, Table 4, Supplementary Figure 3) support two previous studies: our own, which identified frequent deletion of PTEN (Thomas et al. 2009b); and one in which PTEN was found

74 mutated or downregulated in a high percentage of canine OS cell lines and tumors (Levine et al. 2002). Our qRT-PCR data revealed PTEN as under-expressed in several cases (Figure 5) with the expressional change associated with CNA. Thus our data suggest that in combination with other factors, a change in gene dosage may have an affect on the transcriptional regulation of PTEN in canine OS. Another region of high copy number loss in our dataset contained the candidate tumor suppressor gene, TUSC3. To our knowledge, loss of TUSC3 has not been described previously either in human or canine OS. Yet, high resolution aCGH of human breast cancer cell lines and primary tumors concluded that the copy number loss of TUSC3 resulted in decreased expression in cell lines, and absent or reduced expression in 31% of primary breast tumors (Cooke et al. 2008). Our expression analysis indicated that TUSC3 was under-expressed and we found a statistically significant difference between mean RNA fold change and deletion indicating that, as with PTEN, copy number loss in combination with other factors may be causal in reducing gene expression. As far as we are aware this study has been the first to consider TUSC3 as a gene involved in canine OS. In addition, TUSC3 might be of interest for study in human OS due to the high degree of similarity we found concerning disease copy number aberrations in dogs and humans.

In combination with defining a broad range of genomic CNAs and the effect that CNAs may elicit on transcriptional regulation of genes, we attempted to identify global differences in genome-wide DNA copy number aberrations between the five breed groups and three morphological subtypes represented in our sample set. PCA identified a statistically significant difference in global data distribution between breed status and the third component (Supplementary Figure 2) (p=0.0022). Along with the PCA a Fisher’s exact test was performed to identify specific CNAs that are different between the five breed groups studied. Although no significant differences were found (after a multiple testing correction), the region CFA 15q12 (10.64-10.84Mb) did have a low p-value suggesting that different aberration frequencies are present in these five groups; Greyhounds had a high frequency of gain in this region compared to the other breed groups (Figure 4A). Low power, due to a

75 limited sample size for each breed group, is a potential explanation for these results and further studies are needed to fully elucidate the possible significance of this region. A search of this region did not reveal any gene of particular interest at this time.

PCA also was conducted on the 93 samples for which morphological subtype information was available and a Student’s t-test revealed a statistically significant difference between the osteoblastic and chondroblastic subtypes in component 2 (Bonferroni corrected p<0.05) (Supplementary Figure 3). In contrast with the breed analysis, a Fisher’s exact test comparing genomic CNAs between osteoblastic and chondroblastic tumors identified two statistically significant regions of CFA 5 within the region 5q11-q12 (3.14-7.02Mb and 8.01- 9.26Mb). A difference in biological behavior between the morphological subtypes of OS is not well established. Interestingly, of the 12 chondroblastic OS patients 33% had a copy number gain in both of these two regions, while none of the 67 osteoblastic patients displayed copy number gains in these regions (Figure 4B). Although this was a small sample size and the majority of canine OS patients had osteoblastic OS, investigation of this region could serve to distinguish one subtype from another. Several genes are located within these two regions and further characterization of their possible effect on the tumor phenotype could lead to the development of more tailored therapies based on the histological profile of the tumor.

In profiling 123 cases of canine OS by 1Mb aCGH, we concluded that the high occurrence of genetic alterations characteristic of human OS is a striking feature of canine OS also. We identified several new candidate genes exhibiting high frequency of CNAs, including gain of RUNX2 and loss of TUSC3. The majority of these genes are involved in dysregulation of signal transduction pathways in a wide range of human cancers and previously have been defined as candidate tumor suppressor genes and oncogenes. Our study supports previous research concerning canine OS (Levine et al. 2002; Thomas et al. 2009b) and highlights TUSC3 as a possible tumor suppressor gene, since copy number loss appeared to be involved in the transcriptional dysregulation of the gene. We were able to categorize

76 specific CNAs into different canine OS morphological subtypes and reveal possible breed- associated CNAs (within the context of a biased population), suggesting that disease pathogenesis may differ depending on an individual’s genetic background. Further evaluation of the regions of CNAs, by molecular and immunohistochemical approaches may identify specific differences in the genetic pathways. We also were able to show that the canine orthologous regions to HSA 6p21.1 and HSA 8q24 presented with the same pattern of CNAs in OS of the dog, suggesting an evolutionarily conserved genetic basis. Of specific interest is the gene RUNX2, which resides within the orthologous regions defined by HSA 6p21.1 and CFA 12q13. This gene exhibited an increase in expression along with a gain in copy number in canine OS. Several recent studies have focused on investigating the dysregulation of RUNX2 in human OS (Pereira et al. 2009; Sadikovic et al. 2009; San Martin et al. 2009; Shapovalov et al. 2009; Sadikovic et al. 2010) and the comparability of our data suggests that a good avenue for further investigation would be to explore the dysregulation of RUNX2 in canine OS.

The discoveries reported here provide additional insight into the chaotic genome organization of a cancer that plagues more than 10,000 dogs per year in the United States. This study highlights the remarkably high extent to which cytogenetic abnormalities in OS are conserved between both species, supporting the role of the dog as a highly valid comparative model of human OS. Analysis of selected gene expression patterns in human and canine OS indicated similarities at the transcriptional level also, further supporting the functional significance of such aberrations. Overall, these data reinforce the impact that genomic analysis of canine OS may have on pinpointing key genes in human OS, which warrant subsequent examination.

77 Acknowledgements We thank the CCMTR at NCSU College of Veterinary Medicine; The Ohio State University College of Veterinary Medicine Biospecimen Repository; Susan Fosmire, Cristan Jubala, Katherine Gavin, and Mitzi Lewellen in the Modiano lab; and Irene Mok and Sue Lana at Colorado State University Animal Cancer Center College of Veterinary Medicine for their aid in sample collection. Additionally, we thank Milcah Scott for organization of clinical information concerning samples from UMN and technical assistance in RNA extraction and Kristin Maloney for technical assistance with paraffin FISH. We also thank Gary Cutter and Laurence Hunter for statistical advice. This work was supported in part by funds from the American Kennel Club Canine Health Foundation awarded to MB and JM. AA is a recipient of an NCSU Functional Genomic Fellowship.

Supplementary Figure 1: Kaplan-Meier survival curve comparing different treatment protocols for canine OS patients. Proportional hazards test revealed statistically significance in survival time (p=0.0004) depending on the type of treatment received.

Supplementary Figure 2: Interphase multicolor FISH analysis of three zinc fixed paraffin embedded canine OS tissues using BAC clone probe sets encompassing three genes (TSC2, MYC, and TUSC3) (Supplementary Table 2). TSC2 is represented by the orange signal, MYC by the green signal, and TUSC3 by the pink signal. In each graph the x-axis represents the gene and the y-axis demonstrates the percentages of nuclei for the stated copy number of each clone. A, B, and C show FISH analysis of three cases; A=10 yr old male Rottweiler with osteoblastic OS, B=12 yr old female Great Pyrenees with chondroblastic OS, C=6 yr old female retired racing Greyhound with OS. For each case data are shown as: i) interphase nucleus of a cell from each case presenting with a normal copy number (n=2) of each of the three genes being assessed; ii, iii) abnormal interphase nucleus from each patient demonstrating the presence of an aberrant copy number (n≠2) for one or more of TSC2 MYC and TUSC3. iv) Compilation of copy number data of these three genes for each case, based on FISH analysis.

Supplementary Figure 3: Interphase multicolor FISH analysis of three zinc fixed paraffin embedded canine OS tissues using BAC clone probe sets encompassing three genes (RUNX2, RHOC and PTEN) (Supplementary Table 2). RUNX2 is represented by the orange signal, RHOC by the green signal, and PTEN by the pink signal. In each graph the x-axis represents the gene and the y-axis demonstrates the percentages of nuclei for the stated copy number of each clone. A, B, and C show FISH analysis of three cases; A=12 yr old female Great Pyrenees with chondroblastic OS, B=6 yr old female Golden Retriever with osteoblastic OS, C=6 yr old female retired racing Greyhound with OS. For each case data are shown as: i) interphase nucleus of a cell from each case presenting with a normal copy number (n=2) of each of the three genes being assessed; ii, iii) abnormal interphase nucleus from each patient demonstrating the presence of an aberrant copy number (n≠2) for one or more of RUNX2, RHOC and PTEN. iv) Compilation of copy number data of these three genes for each case, based on FISH analysis.

Supplementary Figure 4: Graphical representation of the PCA performed to evaluate global differences in genome-wide DNA copy number aberrations evident in our five breed groups. (A). Placement of each dog within the first three components identified by PCA to explain global differences in data distribution. The color of the circle responds to the breed group shown in B. The blue arrows correlate to the spatial regions between components and help to visually see the differences in breed data placement. (B) Results of a Kruskal-Wallis test for component association with breed. Component 3, which accounted for 5.0% of global data variation, was found to have a statistically significant difference in global data distribution between breed stasis (p= 0.0022).

Supplementary Figure 5: Graphical representation of the PCA performed to evaluate global differences in genome-wide DNA copy number aberrations evident in our three morphological subtypes. (A). Placement of each dog within the first three components identified by PCA to explain global differences in data distribution. The color of the circle responds to the morphological subtype shown in B. The blue arrows correlate to the spatial regions between components and help to visually see the differences in morphological subtype data placement. (B) Results of a Kruskal-Wallis and Student’s t test for component association with morphological subtype. Component 2, which accounted for 5.0% of global data variation, was found to have a statistically significant difference in global data distribution between the osteoblastic and chrondroblastic subtype based on a Student’s t test (p= 0.0043).

83 Supplementary Table 1: Detailed list of clinical information concerning 123 OS tumor samples analyzed for DNA copy number aberrations by genome integrated 1Mb-resolution aCGH. Highlighted cases are those that were evaluated by FISH analysis. Censor morphological Anatomical Age Survival (1=yes, Breed Sex subtype Location Treatment onset (weeks) 0=no) Borzoi F Osteoblastic tibia 11.9 0.00 0 Rottweiler M Osteoblastic Palliative 3.7 0.00 0 Rottweiler MC Fibroblastic None 9.9 0.00 0 Rottweiler MC Osteoblastic None 0.00 0 Great Dane Fibroblastic None 8.5 0.14 0 Labrador Retriever MC Radius Amp_Chemo 12 0.29 0 Mastiff FS Palliative 12.3 0.29 0 Golden Retriever F Osteoblastic Amputation 5.02 0.43 0 Rottweiler MC Osteoblastic Amputation 9.1 1.00 0 Hemipelvectom Rottweiler MI Chondroblastic y 9 1.14 0 Rottweiler M Osteoblastic Radius Palliative 9.91 1.14 0 Rhodesian Ridgeback M Osteoblastic Radius Palliative 2 1.57 0 Rottweiler MI Chondroblastic Amputation 9.5 1.71 0 Doberman MC Osteoblastic tibia Amp_Chemo 5 4.14 0 Golden Retriever FS Chondroblastic Radius Amp_Chemo 6 4.43 1 Saint Bernard FS Osteoblastic Femur Amputation 5.8 5.14 0 Greyhound MC Amp_Chemo 5.71 0 Golden Retriever FS Osteoblastic tibia Palliative 2.67 6.57 0 Greyhound MC Chondroblastic Radius Amp_Chemo 7.86 0 Rottweiler M Osteoblastic Humerus Amp_Chemo 11 8.57 0 Great Dane MC Chondroblastic tibia Amp_Chemo 9 9.71 1 Rottweiler MC Amp_Chemo 9.86 0 Golden Retriever MI Radius Palliative 8 10.14 0 Rottweiler F Osteoblastic Humerus Amp_Chemo 6.8 10.14 0 Greyhound MC Osteoblastic Amp_Chemo 11.00 0 Greyhound FS Osteoblastic Humerus Amp_Chemo 5 11.14 0 Golden Retriever M Chondroblastic Humerus Amp_Chemo 9 11.57 0 Lab FS Amp_Chemo 13.57 0 Greyhound MC Osteoblastic Humerus Amp_Chemo 6 15.43 0 Greyhound MC Femur Amp_Chemo 5 15.86 0

84 Supplementary Table 1: Continued Censor morphological Anatomical Age Survival (1=yes, Breed Sex subtype Location Treatment onset (weeks) 0=no) Rottweiler FS Osteoblastic Humerus Amputation 8 15.86 0 Rottweiler FS Fibroblastic Amp_Chemo 7.2 17.57 0 American Bulldog FS Osteoblastic Amp_Chemo 6.9 18.14 0 Irish Setter M Osteoblastic Palliative 8.5 18.71 0 Labrador Retriever FS Osteoblastic Humerus Amp_Chemo 10 18.86 0 Golden Retriever F Osteoblastic Radius Amp_Chemo 1.5 19.14 0 Greyhound MC Osteoblastic Humerus Amp_Chemo 12 19.71 0 Rottweiler MC Radius Amp_Chemo 9 21.00 1 Mastiff F Osteoblastic Amputation 6 21.57 0 Greyhound FS Humerus Amp_Chemo 6 21.71 0 Great Pyrenees FS Chondroblastic Radius Amputation 12 22.00 0 Rottweiler FS Fibroblastic tibia Amp_Chemo 7 22.00 0 Irish Setter FS Osteoblastic Amputation 11 22.43 0 Greyhound FS Osteoblastic Radius Amp_Chemo 14 23.43 0 Greyhound FS Osteoblastic Humerus Amp_Chemo 10 23.71 0 Great Pyrenees MC Osteoblastic Radius Amp_Chemo 8 24.29 0 Golden Retriever MC Humerus Amp_Chemo 10 25.43 0 Rottweiler FS tibia 8.92 26.14 0 Greyhound MC Osteoblastic tibia Amp_Chemo 5 27.14 0 Doberman MC Osteoblastic Femur Amputation 4 27.71 0 Golden Retriever FS Osteoblastic tibia Palliative 3 28.43 0 Greyhound FS Osteoblastic Humerus Amp_Chemo 11 28.57 0 Golden Retriever M Osteoblastic Amputation 7 31.71 0 Golden Retriever FS Osteoblastic Amputation 9.8 33.29 0 Golden Retriever FS Amp_Chemo 12 34.29 0 Greyhound MC Humerus Amp_Chemo 9 35.86 0 Rottweiler MI Chondroblastic tibia Amputation 7 38.86 1 Great Pyrenees MC Radius Amp_Chemo 3 39.29 0 Rottweiler MI Osteoblastic Humerus Amputation 10 42.14 0 Rottweiler MC Osteoblastic tibia Amputation 5 43.29 0

85 Supplementary Table 1: Continued Censor morphological Anatomical Age Survival (1=yes, Breed Sex subtype Location Treatment onset (weeks) 0=no) Great Pyrenees FS Osteoblastic Radius Amp_Chemo 9 44.86 0 Great Pyrenees M Amp_Chemo 9.7 45.00 0 Golden Retriever MC Osteoblastic Amp_Chemo 8.11 47.86 0 Great Pyrenees FS Fibroblastic Amp_Chemo 8.7 47.86 0 Rottweiler FS Osteoblastic Radius Amp_Chemo 7 48.43 0 Mix FS Amp_Chemo 49.71 0 Mix MC Amp_Chemo 50.86 0 Greyhound MC Osteoblastic tibia Amp_Chemo 10 52.57 0 Great Pyrenees MC Radius Amp_Chemo 2 54.43 0 Rottweiler FS Osteoblastic Amp_Chemo 12.08 58.86 0 Greyhound MC Osteoblastic tibia Amp_Chemo 6 59.00 0 Great Pyrenees MC Osteoblastic Radius Chemo 8 60.14 1 Greyhound MC Osteoblastic tibia Amp_Chemo 11 62.14 1 Great Pyrenees FS Chondroblastic Ulna Chemo 7 64.14 0 Greyhound MC Amp_Chemo 8 65.00 0 Rottweiler MC Osteoblastic Radius Amp_Chemo 8 66.57 0 Doberman MC Fibroblastic tibia Amp_Chemo 67.71 0 Greyhound MC Osteoblastic Radius Amp_Chemo 6 68.00 0 Golden Retriever FS Osteoblastic Radius Amp_Chemo 6 70.86 0 Golden Retriever MC Osteoblastic Radius Amputation 8 71.71 0 Great Pyrenees MC Chondroblastic Radius Amp_Chemo 5 73.14 0 Great Pyrenees M Fibroblastic tibia Amp_Chemo 8 75.57 Greyhound MC Osteoblastic Humerus Amp_Chemo 8 77.00 0 Rottweiler MC Osteoblastic Radius Amputation 9 80.57 1 Great Pyrenees MC Osteoblastic tibia Amp_Chemo 10 83.86 0 Greyhound FS Osteoblastic Femur Amp_Chemo 7 86.43 0 Mix MC Osteoblastic Humerus Amp_Chemo 10 89.14 0 German Shephard MC Osteoblastic Amputation 8 89.57 0 Mix MC Amp_Chemo 99.14 0 Rottweiler F Osteoblastic Radius 4 105.86 0

86 Supplementary Table 1: Continued Censor morphological Anatomical Age Survival (1=yes, Breed Sex subtype Location Treatment onset (weeks) 0=no) Golden Retriever FS Amp_Chemo 126.43 0 Greyhound MC Osteoblastic tibia Amp_Chemo 10 142.43 1 Greyhound MC Osteoblastic Radius Amp_Chemo 10 144.14 0 Greyhound MC Osteoblastic Femur Amp_Chemo 9 144.29 1 Greyhound FS Femur Amp_Chemo 8 144.57 0 Golden Retriever FS Chondroblastic Humerus Amp_Chemo 9.4 169.29 0 Golden Chemo_Local Retriever FS Osteoblastic Maxilla Resection 6 207.43 0 Rottweiler M Fibroblastic Amp_Chemo 8.9 278.14 0 Rottweiler F Osteoblastic Amp_Chemo 6 305.14 1 Golden Retriever FS Fibroblastic Ulna 11.5 Great Pyrenees F Amp_Chemo 7 Bernese Mountain Dog Chemo 8 Collie MC tibia Amputation 7 Briard M Femur 12 Rottweiler F Osteoblastic Palliative 10 0 Boxer FS Fibroblastic 9 0 Golden Retriever MC Osteoblastic 7 0 Flat Coat Retriever M 9.5 Flat Coat Retriever F 9 Rottweiler M Osteoblastic 9 Rottweiler M Osteoblastic Humerus Amputation 7.8 Golden Retriever M Rottweiler F Osteoblastic Amputation 9 Rottweiler F Osteoblastic 9 Doberman MC Osteoblastic Radius Amputation 2.5 Rottweiler M Osteoblastic Amputation 2 Rhodesian Ridgeback MC Amputation 1 Golden Retriever FS Osteoblastic Humerus Amputation 12 Rottweiler FS Fibroblastic 9

87 Supplementary Table 1: Continued morphological Anatomical Age Survival Censor Breed Sex subtype Location Treatment onset (weeks) (1=yes, 0=no) Golden Retriever FS Osteoblastic 6.1 Greyhound MC Fibroblastic Amputation 8.9 Rottweiler F Chondroblastic 5 Rottweiler F Fibroblastic Femur Amputation

Supplementary Table 2: Genomic position and BAC ID for clones selected using the UCSC genome browser (http://www.genome.ucsc.edu/). These clones encompass six genes residing within regions of high CNA occurrence in our canine OS cohort. BAC Start position End position ID Chromosome (bp) (bp) Gene 520K05 6 41684612 41929698 TSC2 390J08 6 41842425 42061613 TSC2 109P13 6 41994977 42256794 TSC2 53D15 9 3128434 3367455 RHOC 7K11 9 3283815 3479840 RHOC 522D16 9 3435308 3626893 RHOC 208D17 12 16698011 16877413 RUNX2 65E02 12 16797093 16967703 RUNX2 523M05 13 27948711 28179583 MYC 335M01 13 28085282 28265714 MYC 322J03 13 28259484 28465596 MYC 53N12 16 41682593 41850298 TUSC3 60N08 16 41837470 42052144 TUSC3 314G09 26 40729598 40911630 PTEN 521G14 26 40844018 41078623 PTEN 54G12 26 41030754 41232429 PTEN

88 Supplementary Table 3: Raw Ct values of non-neoplastic bone and 11 canine OS samples for reference gene c12orf43. Reference Sample Dog Case Gene CT 1 CT 2 CT 3 AVE Standard Dev non-neoplastic bone c12orf43 29.60 29.20 29.40 29.40 0.20 1 c12orf43 28.00 28.30 28.30 28.20 0.17 2 c12orf43 29.10 28.70 28.80 28.87 0.21 3 c12orf43 28.60 28.80 28.70 28.70 0.10 4 c12orf43 28.40 28.90 28.90 28.73 0.29 5 c12orf43 29.60 29.60 29.50 29.57 0.06 6 c12orf43 28.80 29.30 28.90 29.00 0.26 7 c12orf43 28.60 28.50 28.80 28.63 0.15 8 c12orf43 29.10 29.40 29.10 29.20 0.17 9 c12orf43 28.70 28.40 28.40 28.50 0.17 10 c12orf43 28.30 28.30 28.60 28.40 0.17 11 c12orf43 29.80 29.50 30.10 29.80 0.30

89 References

Abdi, H. (2007). Bonferroni and Šidák corrections for multiple comparisons. Encyclopedia of Measurement and Statistics. N. J. Salkind. Thousand Oaks, CA, Sage.

Breen, M., C. F. Langford, N. P. Carter, N. G. Holmes, H. F. Dickens, R. Thomas, N. Suter, E. J. Ryder, M. Pope and M. M. Binns (1999). "FISH mapping and identification of canine chromosomes." J Hered 90(1): 27-30.

Clark, J. C., C. R. Dass and P. F. Choong (2008). "A review of clinical and molecular prognostic factors in osteosarcoma." J Cancer Res Clin Oncol 134(3): 281-97.

Cooke, S. L., J. C. Pole, S. F. Chin, I. O. Ellis, C. Caldas and P. A. Edwards (2008). "High- resolution array CGH clarifies events occurring on 8p in carcinogenesis." BMC Cancer 8: 288.

90 Courts, C., M. Montesinos-Rongen, J. I. Martin-Subero, A. Brunn, D. Siemer, R. Zuhlke- Jenisch, H. Pels, A. Jurgens, U. Schlegel, I. G. Schmidt-Wolf, C. Schaller, G. Reifenberger, M. Sabel, U. Warnecke-Eberz, O. D. Wiestler, R. Kuppers, R. Siebert and M. Deckert (2007). "Transcriptional profiling of the nuclear factor-kappaB pathway identifies a subgroup of primary lymphoma of the central nervous system with low BCL10 expression." J Neuropathol Exp Neurol 66(3): 230-7.

De Maria, R., S. Miretti, S. Iussich, M. Olivero, E. Morello, A. Bertotti, J. G. Christensen, B. Biolatti, R. A. Levine, P. Buracco and M. F. Di Renzo (2009). "met oncogene activation qualifies spontaneous canine osteosarcoma as a suitable pre-clinical model of human osteosarcoma." J Pathol 218(3): 399-408. de Nigris, F., C. Botti, R. Rossiello, E. Crimi, V. Sica and C. Napoli (2007). "Cooperation between Myc and YY1 provides novel silencing transcriptional targets of alpha3beta1-integrin in tumour cells." Oncogene 26(3): 382-94.

Derrien, T., C. Andre, F. Galibert and C. Hitte (2007). "AutoGRAPH: an interactive web server for automating and visualizing comparative genome maps." Bioinformatics 23(4): 498-9.

Eckerle, I., D. Muth, J. Batzler, K. O. Henrich, W. Lutz, M. Fischer, O. Witt, M. Schwab and F. Westermann (2009). "Regulation of BIRC5 and its isoform BIRC5-2B in neuroblastoma." Cancer Lett.

Feugeas, O., N. Guriec, A. Babin-Boilletot, L. Marcellin, P. Simon, S. Babin, A. Thyss, P. Hofman, P. Terrier, C. Kalifa, M. Brunat-Mentigny, L. M. Patricot and F. Oberling (1996). "Loss of heterozygosity of the RB gene is a poor prognostic factor in patients with osteosarcoma." J Clin Oncol 14(2): 467-72.

91 Freeman, S. S., S. W. Allen, R. Ganti, J. Wu, J. Ma, X. Su, G. Neale, J. S. Dome, N. C. Daw and J. D. Khoury (2008). "Copy number gains in EGFR and copy number losses in PTEN are common events in osteosarcoma tumors." Cancer 113(6): 1453-61.

Gokgoz, N., J. S. Wunder, S. Mousses, S. Eskandarian, R. S. Bell and I. L. Andrulis (2001). "Comparison of p53 mutations in patients with localized osteosarcoma and metastatic osteosarcoma." Cancer 92(8): 2181-9.

Hoggart, C. J., E. J. Parra, M. D. Shriver, C. Bonilla, R. A. Kittles, D. G. Clayton and P. M. McKeigue (2003). "Control of confounding of genetic associations in stratified populations." Am J Hum Genet 72(6): 1492-1504.

Ito, Y. (2004). "Oncogenic potential of the RUNX gene family: 'overview'." Oncogene 23(24): 4198-208.

Jong, K., E. Marchiori, A. van der Vaart, B. Ylstra, M. Weiss and G. Meijer (2003). "Chromosomal breakpoint detection in human cancer." Applications of Evolutionary Computing 2611: 54-65.

Jonkers, J. and W. H. Moolenaar (2009). "Mammary tumorigenesis through LPA receptor signaling." Cancer Cell 15(6): 457-9.

Kim, N. W., M. A. Piatyszek, K. R. Prowse, C. B. Harley, M. D. West, P. L. Ho, G. M. Coviello, W. E. Wright, S. L. Weinrich and J. W. Shay (1994). "Specific association of human telomerase activity with immortal cells and cancer." Science 266(5193): 2011-5.

92 Kirkness, E. F., V. Bafna, A. L. Halpern, S. Levy, K. Remington, D. B. Rusch, A. L. Delcher, M. Pop, W. Wang, C. M. Fraser and J. C. Venter (2003). "The dog genome: survey sequencing and comparative analysis." Science 301(5641): 1898-903.

Kirpensteijn, J., M. Kik, G. R. Rutteman and E. Teske (2002). "Prognostic significance of a new histologic grading system for canine osteosarcoma." Vet Pathol 39(2): 240-6.

Kirpensteijn, J., M. Kik, E. Teske and G. R. Rutteman (2008). "TP53 gene mutations in canine osteosarcoma." Vet Surg 37(5): 454-60.

Kow, K., D. H. Thamm, J. Terry, K. Grunerud, S. M. Bailey, S. J. Withrow and S. E. Lana (2008). "Impact of telomerase status on canine osteosarcoma patients." J Vet Intern Med 22(6): 1366-72.

Kurima, K., Y. Yang, K. Sorber and A. J. Griffith (2003). "Characterization of the transmembrane channel-like (TMC) gene family: functional clues from hearing loss and epidermodysplasia verruciformis." Genomics 82(3): 300-8.

93 N. A. Federspiel, A. P. Abola, M. J. Proctor, R. M. Myers, J. Schmutz, M. Dickson, J. Grimwood, D. R. Cox, M. V. Olson, R. Kaul, N. Shimizu, K. Kawasaki, S. Minoshima, G. A. Evans, M. Athanasiou, R. Schultz, B. A. Roe, F. Chen, H. Pan, J. Ramser, H. Lehrach, R. Reinhardt, W. R. McCombie, M. de la Bastide, N. Dedhia, H. Blocker, K. Hornischer, G. Nordsiek, R. Agarwala, L. Aravind, J. A. Bailey, A. Bateman, S. Batzoglou, E. Birney, P. Bork, D. G. Brown, C. B. Burge, L. Cerutti, H. C. Chen, D. Church, M. Clamp, R. R. Copley, T. Doerks, S. R. Eddy, E. E. Eichler, T. S. Furey, J. Galagan, J. G. Gilbert, C. Harmon, Y. Hayashizaki, D. Haussler, H. Hermjakob, K. Hokamp, W. Jang, L. S. Johnson, T. A. Jones, S. Kasif, A. Kaspryzk, S. Kennedy, W. J. Kent, P. Kitts, E. V. Koonin, I. Korf, D. Kulp, D. Lancet, T. M. Lowe, A. McLysaght, T. Mikkelsen, J. V. Moran, N. Mulder, V. J. Pollara, C. P. Ponting, G. Schuler, J. Schultz, G. Slater, A. F. Smit, E. Stupka, J. Szustakowski, D. Thierry-Mieg, J. Thierry-Mieg, L. Wagner, J. Wallis, R. Wheeler, A. Williams, Y. I. Wolf, K. H. Wolfe, S. P. Yang, R. F. Yeh, F. Collins, M. S. Guyer, J. Peterson, A. Felsenfeld, K. A. Wetterstrand, A. Patrinos, M. J. Morgan, P. de Jong, J. J. Catanese, K. Osoegawa, H. Shizuya, S. Choi and Y. J. Chen (2001). "Initial sequencing and analysis of the human genome." Nature 409(6822): 860-921.

Levine, R. A. and M. A. Fleischli (2000). "Inactivation of p53 and retinoblastoma family pathways in canine osteosarcoma cell lines." Vet Pathol 37(1): 54-61.

Levine, R. A., T. Forest and C. Smith (2002). "Tumor suppressor PTEN is mutated in canine osteosarcoma cell lines and tumors." Vet Pathol 39(3): 372-8.

Lim, G., J. Karaskova, B. Vukovic, J. Bayani, B. Beheshti, M. Bernardini, J. A. Squire and M. Zielenska (2004). "Combined spectral karyotyping, multicolor banding, and microarray comparative genomic hybridization analysis provides a detailed characterization of complex structural chromosomal rearrangements associated with gene amplification in the osteosarcoma cell line MG-63." Cancer Genet Cytogenet 153(2): 158-64.

94 Cuff, M. J. Daly, D. DeCaprio, S. Gnerre, M. Grabherr, M. Kellis, M. Kleber, C. Bardeleben, L. Goodstadt, A. Heger, C. Hitte, L. Kim, K. P. Koepfli, H. G. Parker, J. P. Pollinger, S. M. Searle, N. B. Sutter, R. Thomas, C. Webber, J. Baldwin, A. Abebe, A. Abouelleil, L. Aftuck, M. Ait-Zahra, T. Aldredge, N. Allen, P. An, S. Anderson, C. Antoine, H. Arachchi, A. Aslam, L. Ayotte, P. Bachantsang, A. Barry, T. Bayul, M. Benamara, A. Berlin, D. Bessette, B. Blitshteyn, T. Bloom, J. Blye, L. Boguslavskiy, C. Bonnet, B. Boukhgalter, A. Brown, P. Cahill, N. Calixte, J. Camarata, Y. Cheshatsang, J. Chu, M. Citroen, A. Collymore, P. Cooke, T. Dawoe, R. Daza, K. Decktor, S. DeGray, N. Dhargay, K. Dooley, P. Dorje, K. Dorjee, L. Dorris, N. Duffey, A. Dupes, O. Egbiremolen, R. Elong, J. Falk, A. Farina, S. Faro, D. Ferguson, P. Ferreira, S. Fisher, M. FitzGerald, K. Foley, C. Foley, A. Franke, D. Friedrich, D. Gage, M. Garber, G. Gearin, G. Giannoukos, T. Goode, A. Goyette, J. Graham, E. Grandbois, K. Gyaltsen, N. Hafez, D. Hagopian, B. Hagos, J. Hall, C. Healy, R. Hegarty, T. Honan, A. Horn, N. Houde, L. Hughes, L. Hunnicutt, M. Husby, B. Jester, C. Jones, A. Kamat, B. Kanga, C. Kells, D. Khazanovich, A. C. Kieu, P. Kisner, M. Kumar, K. Lance, T. Landers, M. Lara, W. Lee, J. P. Leger, N. Lennon, L. Leuper, S. LeVine, J. Liu, X. Liu, Y. Lokyitsang, T. Lokyitsang, A. Lui, J. Macdonald, J. Major, R. Marabella, K. Maru, C. Matthews, S. McDonough, T. Mehta, J. Meldrim, A. Melnikov, L. Meneus, A. Mihalev, T. Mihova, K. Miller, R. Mittelman, V. Mlenga, L. Mulrain, G. Munson, A. Navidi, J. Naylor, T. Nguyen, N. Nguyen, C. Nguyen, R. Nicol, N. Norbu, C. Norbu, N. Novod, T. Nyima, P. Olandt, B. O'Neill, K. O'Neill, S. Osman, L. Oyono, C. Patti, D. Perrin, P. Phunkhang, F. Pierre, M. Priest, A. Rachupka, S. Raghuraman, R. Rameau, V. Ray, C. Raymond, F. Rege, C. Rise, J. Rogers, P. Rogov, J. Sahalie, S. Settipalli, T. Sharpe, T. Shea, M. Sheehan, N. Sherpa, J. Shi, D. Shih, J. Sloan, C. Smith, T. Sparrow, J. Stalker, N. Stange-Thomann, S. Stavropoulos, C. Stone, S. Stone, S. Sykes, P. Tchuinga, P. Tenzing, S. Tesfaye, D. Thoulutsang, Y. Thoulutsang, K. Topham, I. Topping, T. Tsamla, H. Vassiliev, V. Venkataraman, A. Vo, T. Wangchuk, T. Wangdi, M. Weiand, J. Wilkinson, A. Wilson, S. Yadav, S. Yang, X. Yang, G. Young, Q. Yu, J. Zainoun, L. Zembek, A. Zimmer and E. S. Lander (2005). "Genome sequence, comparative analysis and haplotype structure of the domestic dog." Nature 438(7069): 803-19.

Man, T. K., X. Y. Lu, K. Jaeweon, L. Perlaky, C. P. Harris, S. Shah, M. Ladanyi, R. Gorlick, C. C. Lau and P. H. Rao (2004). "Genome-wide array comparative genomic

95 hybridization analysis reveals distinct amplifications in osteosarcoma." BMC Cancer 4: 45.

Mirabello, L., R. J. Troisi and S. A. Savage (2009a). "International osteosarcoma incidence patterns in children and adolescents, middle ages and elderly persons." Int J Cancer 125(1): 229-34.

Mueller, F., B. Fuchs and B. Kaser-Hotz (2007). "Comparative biology of human and canine osteosarcoma." Anticancer Res 27(1A): 155-64.

Narumiya, S., M. Tanji and T. Ishizaki (2009). "Rho signaling, ROCK and mDia1, in transformation, metastasis and invasion." Cancer Metastasis Rev 28(1-2): 65-76.

Nasir, L., P. Devlin, T. McKevitt, G. Rutteman and D. J. Argyle (2001). "Telomere lengths and telomerase activity in dog tissues: a potential model system to study human telomere and telomerase biology." Neoplasia 3(4): 351-9.

Overholtzer, M., P. H. Rao, R. Favis, X. Y. Lu, M. B. Elowitz, F. Barany, M. Ladanyi, R. Gorlick and A. J. Levine (2003). "The presence of p53 mutations in human osteosarcomas correlates with high levels of genomic instability." Proc Natl Acad Sci U S A 100(20): 11547-52.

Paoloni, M., S. Davis, S. Lana, S. Withrow, L. Sangiorgi, P. Picci, S. Hewitt, T. Triche, P. Meltzer and C. Khanna (2009). "Canine tumor cross-species genomics uncovers targets linked to osteosarcoma progression." BMC Genomics 10: 625.

96 Papachristou, D. J. and A. G. Papavassiliou (2007). "Osteosarcoma and chondrosarcoma: new signaling pathways as targets for novel therapeutic interventions." Int J Biochem Cell Biol 39(5): 857-62.

Pearson, K. (1901). "On Lines and Planes of Closest Fit to Systems of Points in Space." Philosophical Magazine 2 2(6): 559-572.

Pereira, B. P., Y. Zhou, A. Gupta, D. T. Leong, K. Z. Aung, L. Ling, R. W. Pho, M. Galindo, M. Salto-Tellez, G. S. Stein, S. M. Cool, A. J. van Wijnen and S. S. Nathan (2009). "Runx2, p53, and pRB status as diagnostic parameters for deregulation of osteoblast growth and differentiation in a new pre-chemotherapeutic osteosarcoma cell line (OS1)." J Cell Physiol 221(3): 778-88.

Pfaffl, M. W. (2001). "A new mathematical model for relative quantification in real-time RT- PCR." Nucleic Acids Res 29(9): e45.

Pritchard, J. K., M. Stephens, N. A. Rosenberg and P. Donnelly (2000). "Association mapping in structured populations." Am J Hum Genet 67(1): 170-81.

Reich, D. E. and D. B. Goldstein (2001). "Detecting association in a case-control study while correcting for population stratification." Genet Epidemiol 20(1): 4-16.

San Martin, I. A., N. Varela, M. Gaete, K. Villegas, M. Osorio, J. C. Tapia, M. Antonelli, E. E. Mancilla, B. P. Pereira, S. S. Nathan, J. B. Lian, J. L. Stein, G. S. Stein, A. J. van Wijnen and M. Galindo (2009). "Impaired cell cycle regulation of the osteoblast- related heterodimeric transcription factor Runx2-Cbfbeta in osteosarcoma cells." J Cell Physiol 221(3): 560-71.

97 Selvarajah, S., M. Yoshimoto, O. Ludkovski, P. C. Park, J. Bayani, P. Thorner, G. Maire, J. A. Squire and M. Zielenska (2008). "Genomic signatures of chromosomal instability and osteosarcoma progression detected by high resolution array CGH and interphase FISH." Cytogenet Genome Res 122(1): 5-15.

Shapovalov, Y., D. Benavidez, D. Zuch and R. A. Eliseev (2009). "proteasome inhibition with bortezomib suppresses growth and induces apoptosis in osteosarcoma." Int J Cancer.

Shay, J. W. (1997). "Telomerase in human development and cancer." J Cell Physiol 173(2): 266-70.

Tang, N., W. X. Song, J. Luo, R. C. Haydon and T. C. He (2008). "Osteosarcoma development and stem cell differentiation." Clin Orthop Relat Res 466(9): 2114-30.

98 Thayanithy, V., A. L. Sarver, R. V. Kartha, C. Park, M. Scott, A. Y. Angstadt, M. Breen, C. J. Steer, J. F. Modiano and S. Subramanian (submitted). "Perturbation in the 14q32 miRNAs-cMYC-miR-17~92 gene network contributes to osteosarcoma and is associated with survival outcome." Cancer Cell.

Thomas, R., S. E. Duke, S. K. Bloom, T. E. Breen, A. C. Young, E. Feiste, E. L. Seiser, P. C. Tsai, C. F. Langford, P. Ellis, E. K. Karlsson, K. Lindblad-Toh and M. Breen (2007). "A cytogenetically characterized, genome-anchored 10-Mb BAC set and CGH array for the domestic dog." J Hered 98(5): 474-84.

Thomas, R., S. E. Duke, E. K. Karlsson, A. Evans, P. Ellis, K. Lindblad-Toh, C. F. Langford and M. Breen (2008). "A genome assembly-integrated dog 1 Mb BAC microarray: a cytogenetic resource for canine cancer studies and comparative genomic analysis." Cytogenet Genome Res 122(2): 110-21.

Thomas, R., S. E. Duke, H. J. Wang, T. E. Breen, R. J. Higgins, K. E. Linder, P. Ellis, C. F. Langford, P. J. Dickinson, N. J. Olby and M. Breen (2009a). "'Putting our heads together': insights into genomic conservation between human and canine intracranial tumors." J Neurooncol 94(3): 333-49.

Thomas, R., A. Scott, C. F. Langford, S. P. Fosmire, C. M. Jubala, T. D. Lorentzen, C. Hitte, E. K. Karlsson, E. Kirkness, E. A. Ostrander, F. Galibert, K. Lindblad-Toh, J. F. Modiano and M. Breen (2005). "Construction of a 2-Mb resolution BAC microarray for CGH analysis of canine tumors." Genome Res 15(12): 1831-7.

Thomas, R., H. J. Wang, P. C. Tsai, C. F. Langford, S. P. Fosmire, C. M. Jubala, D. M. Getzy, G. R. Cutter, J. F. Modiano and M. Breen (2009b). "Influence of genetic background on tumor karyotypes: evidence for breed-associated cytogenetic aberrations in canine appendicular osteosarcoma." Chromosome Res 17(3): 365-77.

Tsuchiya, T., K. Sekine, S. Hinohara, T. Namiki, T. Nobori and Y. Kaneko (2000). "Analysis of the p16INK4, p14ARF, p15, TP53, and MDM2 genes and their prognostic implications in osteosarcoma and Ewing sarcoma." Cancer Genet Cytogenet 120(2): 91-8.

Withrow, S. J., B. E. Powers, R. C. Straw and R. M. Wilkins (1991). "Comparative aspects of osteosarcoma. Dog versus man." Clin Orthop Relat Res(270): 159-68.

99 Withrow, S. J. and D. Vail, Eds. (2007). Small Animal Clinical Oncology. Philadelphia, WB Sanders.

Zhang, J., X. Chen, M. S. Kent and C. O. Rodriguez (2009). "Establishment of a dog model for the p53 family pathway and identification of a novel isoform of p21 cyclin- dependent kinase inhibitor." Mol Cancer Res 7(1): 67-78.

100

Chapter III

A genome-wide approach to comparative oncology: High-resolution oligonucleotide aCGH of canine and human OS pinpoints communal microaberrations

Andrea Y. Angstadta, Venugopal Thayanithyb, Subbaya Subramanianb,c, Jaime F. Modianoc,d, and Matthew Breena,g,h

aDepartment of Molecular Biomedical Sciences, North Carolina State University, Raleigh, NC, USA bDepartment of Laboratory Medicine and Pathology, University of Minnesota, Minneapolis, MN, USA cMasonic Cancer Center, University of Minnesota, Minneapolis, MN, USA d Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, St. Paul, MN, USA gCenter for Comparative Medicine and Translational Research, North Carolina State University, Raleigh, NC, USA hCancer Genetics Program, UNC Lineberger Comprehensive Cancer Center, Chapel Hill, NC, USA

101 Abstract:

Cancer is a genomic disease and therefore it is challenging to separate key genetic alterations driving disease progression from passenger alterations not directly involved in disease etiology. Advancements in molecular tools in the genomic era combined with cross- species evaluation of cancers have helped to address these difficulties, by identifying similarities in genomic abnormalities. Spontaneous cancers in dogs are ideal for cross- species genome-wide evaluation because of the strong anatomical and physiological similarities to human malignancies. In this comparative oncology study we directly compared genomic instability in naturally occurring osteosarcoma (OS), the most common malignant tumor of the bone in humans and dogs. We profiled 23 cases of canine OS and 15 cases of human OS by high resolution oligonucleotide array comparative genomic hybridization (aCGH). Consequently, we identified microaberrations (<500kb) in human and dog OS demonstrating the analogous level of cytogenetic complexity in both species. We also found parallels in genome-wide CNA patterns in OS between orthologous regions of the human and dog genome. Within cross-species microaberration regions were the ‘OS associated genes,’ MYC, CDKN2A/CDKN2B, RUNX2, PTEN, and RB1, suggesting that they are driver alterations in OS. Foremost, this study supports further comparative studies of dog and human OS while overcoming previous challenges in interpreting genomic instability. Identification of parallel genome-wide aberrations in dog and human OS provides avenues for further investigation that could pinpoint novel targets of genetic therapies that once developed and tried in dogs, may be translatable to human patients.

102

Introduction:

Comparative oncology, described as the discipline that integrates the study of naturally occurring cancers in animals into studies of human cancer biology and therapy (Paoloni and Khanna 2008), has become increasingly popular in recent years and is most often referred to as the study of cancers in companion animals such as cats, dogs, horses, ferrets, and other small animals (Klein et al. 1986; Nasir and Reid 1999; Antinoff and Hahn 2004; Seltenhammer et al. 2004; Hershey et al. 2005; Khanna et al. 2006; Withrow and Vail 2007). Specifically, the domestic dog is a key model in comparative oncology because of its long history in biomedical research, strong anatomical and physiological similarities to humans, and the fact that a vast number of pet dogs per year are diagnosed with and treated for cancer (estimated to be over 1 million per year in the US) (Khanna et al. 2006; Paoloni and Khanna 2008). Furthermore, unlike laboratory induced mice models of human cancers dogs acquire cancer spontaneously while living in the same environments as people and analysis of the dog genome demonstrated that it was more homologous in sequence conservation to humans than mice (Kirkness et al. 2003; Lindblad-Toh et al. 2005). As a means to answer questions about cancer therapy, investigation of cancer in dogs is not novel but the recent advancement in genome annotation and molecular tools in both dogs (Breen et al. 1999; Breen et al. 2001; Kirkness et al. 2003; Breen et al. 2004; Lindblad-Toh et al. 2005; Thomas et al. 2005; Thomas et al. 2008) and humans (Lander et al. 2001; Zielenska et al. 2001; Squire et al. 2003; Selvarajah et al. 2008; Sadikovic et al. 2009) has enhanced our ability to characterize genomic abnormalities in complex diseases. Recently launched cooperative initiatives such as the National Cancer Institute’s (NCI’s) Comparative Oncology Trials Consortium (COTC; http://ccr.cancer.gov/resources/cop/COTC.asp) and the Canine Comparative Oncology and Genomics Consortium (CCOGC; http://www.ccogc.net), now provide the infrastructure and resources needed to integrate data from genomic studies in naturally occurring cancer models into the development of new cancer therapeutics (Gordon et al. 2009).

103

A predominant goal of the COTC and CCOGC is to identify key genetic alterations in cancers that can be deemed as drivers of the disease progression. Therefore, it can be hypothesized that cross-species evaluation of genomic aberrations in cancers will pinpoint causative genetic alterations that can be separated from abnormalities which are only a result of cancer instability. Osteosarcoma (OS) is a highly metastatic cancer of the bone in humans and dogs that would benefit from genome-wide comparative study as strong pathophysiological similarities exist between humans and dogs (Khanna et al. 2006). Fewer than 1,000 people per year in the US are diagnosed with OS while in the domestic dog population the annual number of new cases is estimated to far exceed 10,000 (Withrow et al. 1991; Fossey et al. 2009). The introduction of new chemotherapy regimens in the 1980’s, which included treatment both before and after definitive surgical resection, improved the five-year survival rate to approximately 70%, but little improvement to this rate has been made over the last decade (Mirabello et al. 2009). Advancements in health care have also provided a similar benefit to dogs with OS as patients receiving a combination of amputation and chemotherapy treatment have the best prognosis; 50% 1-year and 20% 2-year survival rates (Mueller et al. 2007; Withrow and Vail 2007). These similarities combined with the high occurrence rate and the rapid time course of OS progression in dogs when compared to humans, presents the opportunity to assess novel therapeutic agents in dogs in a quicker manner than is possible in many human trials (Paoloni and Khanna 2008). As such, comparative genomic OS studies completed in dogs are valuable to the development of new human cancer drugs, devices, and imaging techniques (Gordon et al. 2009).

In addition to the physical parallels in human and canine OS previous studies have identified analogous molecular abnormalities (Johnson et al. 1998; Gokgoz et al. 2001; Fossey et al. 2009; Paoloni et al. 2009). Human OS is characterized by a high frequency of genomic instability, highly heterogeneous karyotypes (both intra- and inter-tumor) and gross changes in gene expression (Batanian et al. 2002; Sandberg and Bridge 2003; Squire et al. 2003; Lim et al. 2004; Lim et al. 2005; Selvarajah et al. 2007; Sadikovic et al. 2010), making

104 it difficult to identify key genomic players in the cancer. The same pattern of genomic instability has been found in canine OS (Paoloni et al. 2009; Selvarajah et al. 2009; Thomas et al. 2009) and a few of the candidate genes implicated in the pathogenesis and progression of OS in children have also been characterized in the canine disease. Notable examples of these genes are PTEN (phosphatase and tensin homolog), RB1 (retinoblastoma), ezrin (villin- 2), c-met (meschymal-epithelial transition factor), erbB-2 (v-erb-b2 erythroblastic leukemaia viral oncogene homolog 2), and TP53 (tumor protein 53) (Mendoza et al. 1998; Ferracini et al. 2000; Levine and Fleischli 2000; Levine et al. 2002; Flint et al. 2004; Kirpensteijn et al. 2008). Some studies have conducted comparative analysis and treatment of previously identified gene abnormalities in human and canine OS (Fossey et al. 2009; McCleese et al. 2009; Wittenburg et al. 2010). One example is the study that concluded that constitutive activation of STAT3 (a key participant in tumor cell survival, proliferation, and metastasis) was present in both canine and human OS (Fossey et al. 2009). In human and canine cell lines downregulation of STAT3 resulted in decreased cell proliferation and viability, inducing caspase-3/7 mediated apoptosis in treated cells, suggesting that STAT3 activation contributes to the survival and proliferation of human and canine OS cells (Fossey et al. 2009). This method was again used in two separate studies that tested the biologic activity of a novel heat shock protein 90 (HSP90) inhibitor, STA-1474, (McCleese et al. 2009) and the in vitro and in vivo effects of the HDAC inhibitor valproic acid (VPA) on doxorubicin sensitivity in both human and canine OS (Wittenburg et al. 2010). Human and canine OS cell lines, normal canine osteoblasts, and mice bearing canine OS xenografts were treated with STA-1474 and evaluated for effects on proliferation, apoptosis, and known HSP90 proteins. STA-1474 treatment promoted loss of cell viability, inhibition of cell proliferation, induction of apoptosis in OS cell lines, and exhibited selectivity for OS cells versus normal canine osteoblasts, supporting HSP90 as a therapeutic target for OS (McCleese et al. 2009). Pre- incubation with VPA followed by doxorubicin treatment in canine and human OS cells and a canine OS xenograft model also resulted in significant growth inhibition and potentiation of apoptosis (Wittenburg et al. 2010). These studies highlight the potential of how comparative

105 studies between humans and dogs can characterize specific gene and protein dysregulation involved in OS and identify new cancer treatment targets for both dogs and humans. As OS is hallmarked by high genomic instability in humans and dogs (Sadikovic et al. 2009; Thomas et al. 2009), singular gene research is limited in its ability to accurately identify all genetic alterations that lead to the cancer’s development and progression and therefore should be tested for therapeutic potential. Recent advancements in molecular technology have helped overcome these boundaries by increasing our ability to conduct genome-wide studies. Due to the sequencing of various genomes and the availability of online databases we can now evaluate cross-species data in order to characterize similar genetic abnormalities between numerous species, including humans and dogs. To date very few investigators have applied this concept, but one group demonstrated the analytical power of this method by using parallel human and canine oligonucleotide array platforms (Paoloni et al. 2009) to identify OS gene expression signatures for primary tumors and normal tissues from both dogs and pediatric patients. Strong similarities between human and canine OS were identified and hierarchical cluster analysis of 265 orthologous transcripts could not distinguish the cancers based on species alone. Cross-species target mining identified two genes, IL-8 (interleukin-8) and SLC1A3 (solute carrier family 1, member 3) that were identified as uniformly expressed in canine OS but not in pediatric patients. Analysis of an independent population of pediatric OS patients concluded that expression of these genes was associated with decreased survival demonstrating the importance of a comparative oncology approach for improvement of our understanding of cancer biology and therapies (Paoloni et al. 2009).

Here we demonstrate the power of genomic comparative oncology to identify conserved copy number aberrations (CNAs) in humans and dogs by high resolution oligonucleotide array comparative genome hybridization (aCGH) analysis of 23 cases of canine OS and 15 cases of human OS. We identified recurrent micro-regions of aberration (microaberrations, <500kb) in both human and canine cases. Further characterization of these microaberrations, which included regions of high copy number gain, amplification, and

106 suggestive homologous loss, identified specific genes subject to numerical imbalances. Comparison of patterns of CNAs in orthologous regions between human and canine OS cases found several recurrent aberrations suggestive of driving development of the cancer. Consequently, the extensive similarities that we define in genomic CNAs between canine and human OS cases reveal potential targets for genetic therapies in dogs that may be translational to human patients.

Materials and Methods:

Tissue specimens

Canine OS tissues were collected from cases of spontaneously occurring canine OS either by institutional or community veterinary practices in the United States between November 1999 and July 2009. A total of 23 canine OS tumor specimens (Table 1) were obtained during routine clinical evaluation under approved protocols for informed consent by owners and prior to the initiation of chemotherapy or radiotherapy. Tissue specimens were processed as described previously (Thomas et al. 2005). Briefly, tumor tissue was surgically removed under sterile conditions as a part of a diagnostic biopsy procedure and grossly normal tissue was excised from the specimen. Tumor tissue was then 1) fixed in formalin and submitted for histological evaluation of hematoxylin and eosin (H&E)-stained sections, using the criteria of Kirpensteijn et al (Kirpensteijn et al. 2002) and 2) either processed immediately for DNA extraction or snap-frozen in liquid nitrogen for subsequent extraction. Human OS DNA samples and clinical information were acquired from the tissue procurement facility at Masonic Cancer Center, University of Minnesota (Table 2). The quality and quantity of DNA was assessed using spectrophotometry (Nanodrop 1000, Nanodrop Technologies) and by visualizing on a 2% agarose gel.

107 Table 1: Clinical summary of 23 canine OS tumor samples analyzed for DNA copy number aberrations on Agilent’s canine 180,000 feature G3 SurePrint oligonucleotide aCGH platform. Age Morphological Tumor Case Breed Sex onset subtype Location OS1 Rottweiler M 10 Osteoblastic Humerus OS2 Rottweiler M 9 Osteoblastic Radius Golden OS3 Retriever F 12 Non-productive Humerus Golden OS4 Retriever M 10 Non-productive Humerus OS5 Great Pyrenees F 12 Chondroblastic Radius OS6 Greyhound F 7 Osteoblastic Femur OS7 Great Pyrenees M 8 Osteoblastic Radius OS8 Great Pyrenees M 5 Chondroblastic Radius OS9 Great Pyrenees M 3 Poorly differentiated Radius OS10 Greyhound F 6 Humerus Golden OS11 Retriever M 8 Non-productive Radius / Ulna OS12 Rottweiler M 8 Osteoblastic Radius OS13 Rottweiler M 7 Chondroblastic Tibia OS14 Greyhound M 5 Non-productive Femur Golden OS15 Retriever M 8 Osteoblastic Radius OS16 Rottweiler M 5 Osteoblastic Tibia Golden OS17 Retriever F 6 Chondroblastic Radius Golden OS18 Retriever F 6 Osteoblastic Maxilla OS19 Great Dane M 9 Chondroblastic Tibia OS20 Great Pyrenees M 8 Fibroblastic Tibia OS21 Rottweiler M 3.7 Osteoblastic American OS22 Bulldog F 6.9 Osteoblastic Golden OS23 Retriever F 2.67 Osteoblastic Tibia

108 Table 2: Clinical summary of 15 human OS tumor samples analyzed for DNA copy number aberrations on Agilent’s 60,000 feature G3 SurePrint oligonucleotide aCGH platform. Case Age Gender Tumor site Mediastinum hOS1 16 F (bronchus) hOS2 10 M right distal femur hOS3 40 M tibia hOS4 16 M femur hOS5 12 F left distal femur hOS6 19 M left lung hOS7 19 M left lung hOS8 14 M left distal femur hOS9 17 M left distal femur hOS10 22 M left lung, lower lobe hOS11 31 F left lung, upper lobe hOS12 42 M right forearm hOS13 31 F right lung, upper lobe hOS14 24 M hOS15 20 M right middle lobe lung

Array comparative genome hybridization

A total of 23 canine OS DNA samples, which had already been previously hybridized onto a 1Mb-resolution BAC array (Angstadt et al. submitted), were hybridized onto Agilent Canine CGH 4x180K microarrays, each field comprising 180,000 unique 60-mer oligonucleotide probes at ~13.5kb spacing throughout the canine genome (ID# 25522, Agilent Technologies). The Bioprime Labeling Kit (Invitrogen) was used to label 500ng of tumor and reference DNA which consisted of a genomic DNA pool generated from an equimolar mix of a series of non-neoplastic female or male breed matched blood samples and then combined in the presence of 48ul (2μM) canine Cot1 DNA. The sex-mismatched probe mixture was then precipitated at -80ºC for 45 min and resuspended in 44µl HPLC water, 11µl Agilent 10X blocking agent, and 55µl Agilent HI-RPM hybridization buffer. This mixture was denatured at 95ºC for 5 min, and incubated at 37ºC for 30 min before application to the array. Hybridization was carried out for 40 hrs at 65ºC using an Agilent microarray hybridization chamber in a rotating oven. DNA from 15 cases of human OS were hybridized onto the Agilent Human CGH 8x60K microarrays which contain 60,000 unique

109 60-mer oligonucleotide probes at ~50 kb spacing throughout the human genome (ID# 21924,

Agilent Technologies). Reference DNA consisted of a genomic DNA pool generated from either an equimolar mix of a series of non-neoplastic female or male blood samples (Promega). 500ng of labeled tumor and reference probe were then combined in the presence of 2µl (1mg/ml) human Cot1 DNA (Invitrogen). The sex-mismatched probe mixture was then precipitated at -80ºC for 45 min and resuspended in 22µl HPLC water, 5.5ul Agilent 10X blocking agent, and 27.5µl Agilent HI-RPM hybridization buffer. This mixture was denatured at 95ºC for 5 min, and incubated at 37ºC for 30 min before application to the array. Hybridization was carried out for 24 hrs at 65ºC using an Agilent microarray hybridization chamber in a rotating oven. Both canine and human arrays were washed according to manufacturer’s recommendations; air dried, and scanned at 3μm resolution using an Agilent G2565CA DNA Microarray Scanner with SureScan High resolution technology (Agilent Technologies).

Array comparative genome hybridization analysis

The Agilent canine 180K and human 60K arrays were assessed for data quality by the ‘Quality Metrics’ report produced in Agilent’s Feature extraction software (v10.5)( Agilent Technologies). The data report from Feature Extraction was then imported into Agilent Genome Workbench Software (v5.0) (Agilent Technologies). In order to produce aberration calls and assess overall array quality, an estimate of the noise was obtained by calculating the spread of the ratio differences between consecutive probes (DLRspread) along all chromosomes. The ADM-2 algorithm was then used to scan for chromosomal intervals containing at least three probes (~27kb resolution for canine arrays and ~100kb resolution for human arrays), for which the mean interval ratio was significantly different from zero and with ratios outside the threshold standard deviations. The threshold value for the ADM-2 algorithm was set to 6.0 to reduce inherent sample noise without loss of true consistent intervals along each chromosome. To detect the extent of the CNAs we set the mean

aberration log2 ratio thresholds after ADM-2 algorithm processing equivalent to or exceeding ~1:1.5 ratio (0.5) for canine and human arrays (Selvarajah et al. 2008). As previously

110 reported in aCGH analysis of human OS, gains were classified as high copy number gains if

log2 ratio of tumor:reference ≥1.0 and as amplifications where the log2 ratio of tumor:reference ≥ 2.0 (Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Bayani et al. 2003; Squire et al. 2003; Selvarajah et al. 2008). We considered regions suggestive of

homozygous loss to have a log2 ratio of tumor:reference ≤ -2.0 for both human and canine cases. A sliding window of 50kb for canine arrays and 100kb for human arrays was set for visualization purposes (sliding window does not affect ADM-2 algorithm calculation). Further details concerning the aberration analysis principles are available at the Agilent website (www.agilent.com). We considered microaberrations to be <500kb as a consensus on size has not been defined but previous studies using lower resolution technology called these regions <750kb (Selvarajah et al. 2007; Selvarajah et al. 2008; Sadikovic et al. 2009). A non-parametric Kruskal-Wallis test was performed to evaluate the differences between aberration size and percentages of total aberrations in our canine and human aCGH analysis. Statistical analyses of the data were performed in JMP Genomics v4 and SAS 9.1.3 (SAS Institute).

Identification of orthologous canine and human regions

To make direct comparisons of the genomic imbalance shared between human and dog, orthologous regions of human chromosomal segments defined previously shown to have a high rate of aneuploidy in OS were identified in the canine genome using the UCSC Genome Browser (http://genome.ucsc.edu/) and Autograph tool (Derrien et al. 2007). Specifically, Autograph identified conserved segments and breakpoints regions between canine and human chromosomes, constructing synteny maps. These synteny maps were used to identify orthologous regions of the canine genome to regions on human chromosomes. In addition, Autograph was used to indentify combinations of human chromosomal regions that were orthologous to entire canine chromosomes.

111 Results:

Refining regions of genomic aberration in canine OS

Twenty three cases of canine (CFA) OS previously assessed by 1Mb-resolution BAC aCGH (Angstadt et al. submitted) were reanalyzed for genomic imbalances on a canine 180,000 feature oligonucleotide aCGH platform. To quantify the total number of imbalances in each case we applied the ADM-2 algorithm (Agilent Genome Workbench Software (v5.0)) with a threshold of 6.0 and a log2 ratio threshold of 0.5. As was expected from our 1Mb array data we found the same high level of genomic instability in high resolution aCGH analysis of canine OS, but now were able to identify numerous additional CNAs due to the >74 fold increase in resolution. The increased detection rate of oligo arrays is shown by the comparison of the BAC and oligo aCGH profiles of OS12, an 8 year-old male Rottweiler with osteoblastic OS (Figure 1). Previous analysis of CNAs at 1Mb-resolution (Figure 1A) revealed several regions of genomic imbalance in the case but analysis of the same tumor at ~27kb resolution (Figure 1B) enabled us to identify microaberrations (<500kb) of gain and loss that exist between larger regions of imbalance. The oligo-array analysis identified numerous CNAs (gain=red bars; loss=green bars) not indicated at 1Mb-resolution between larger regions of gain or loss in CFA 13 and CFA 16 (Figure 1B). The oligo-array analysis

also further characterized a region of gain (Figure 1B) on CFA 12q12-q14 (log2 ratio 1.2312),

indicating a high copy number gain of the RUNX2 locus (CFA 12q13:16.73-16.85Mb, log2 ratio 1.491).

112

Figure 1: Array CGH profiles of OS12, an 8 year-old male Rottweiler with osteoblastic OS. The log2 ratio value is shown on the y-axis and the x-axis represents the chromosome and position of the 38 dog autosomes. (A) 1Mb-resolution BAC aCGH profile showing gained (red) and lost (green) loci. Loci displayed in grey has log2 ratios indicative of a copy number of n=2. (B) Agilent’s 180k oligonucleotide aCGH profile (~50kb resolution view) of the same case, demonstrating the considerably higher number of datapoints. Cytoband views of CNAs for C) CFA 12, D) 13, and E) 16. Red and green bars above and below the aberrant regions indicate the extent of gain and loss, respectively.

The penetrance of each locus on the oligo-array was calculated after identification of genomic regions of copy number gain or loss by the ADM-2 algorithm in each of the 23 cases. Profiles of genomic imbalance frequencies of our cohort concluded by both the BAC array after data processing by the aCGH Smooth algorithm and oligo-array after data processing by the ADM-2 algorithm are represented in Figure 2. The plots show very similar

113 profiles of regional copy number gain and loss across the genome but as expected the oligo- array found several microaberration regions not identified by the BAC array. One of these regions is within CFA 2q24.1 at 36.4-36.8Mb, which contained intervals of high gain frequencies (>40%) and lies between two clones on the BAC array, 326N08 and 186J05 situated at CFA 2q24.1:35.9-36.1Mb and 36.8-37.0Mb. Analysis of common aberration regions identified by the oligo-array across all canine cases revealed a total of 6348 regions of copy number gain and a total of 6076 regions of loss with 26.6% (1688/6348) of gain regions and 13% (792/6076) of loss regions involved in >30% of cases. CFA 3q32:65.33- 65.34Mb, CFA 6q22:41.91-41.92Mb, CFA 9q11.2:5.24-5.36Mb, and CFA 13q21.1:39- 39.16Mb all had the highest gain frequency (73.6%) in our canine cohort and CFA 11q16:44.26– 44.30Mb had the highest loss frequency of 69.57%.

114

Figure 2: Frequency of copy number aberrations in 23 canine OS samples. The 38 dog autosomes are listed on the x-axis. The y-axis shows the percentage of cases with copy number gain (red) and copy number loss (green), for (A) each ~1 Mb interval, based on log2 ratios after data processing by the aCGH Smooth algorithm (Jong et al. 2003) and (B) each ~27kb interval, based on log2 ratios calls by the ADM-2 algorithm (threshold of 6.0 and log2 ratio threshold of 0.5).

All 23 canine OS cases had regions of high copy number gain (log2 ratio≥1.0), distributed across all 38 autosomes, while 19 of these cases had regions of amplification (log2 ratio≥2.0). Two regions of amplification, CFA 1q23:59.59-59.62Mb and CFA 37q17:33.69- 33.71Mb, and two regions of high copy number gain, CFA 6q14:17.49-17.51Mb and CFA 10q11:4.11-4.15Mb, were found in 26% (6/23) of the cases. In addition the region of high copy number gain, CFA 3q32:65.33-65.35Mb, was found in 34.8% (8/23) of the cases.

Regions suggestive of homozygous loss (log2 ratio≤-2.0) were found in 19 dogs and were distributed across 28 autosomes. Three of these regions; CFA 20q17:56.26-56.28Mb, CFA

115 24q23:35.84-35.85Mb, CFA 26q24:30.24-30.29Mb, were identical in 13% (3/23) of the cases. Although only a select number of homozygous losses were identical in different cases, overlap was found between CFA 26q25:40.92-40.98Mb and CFA 11q16:44.25-44.26Mb. Dividing the copy number gains and losses for each canine case into categories based on aberration size (Table 3 and Figure 3) most cases have a higher frequency of regions experiencing gains <500kb in size than regions >1Mb in size (Kruskal-Wallis non-parametric test p=1.27e-7). The lowest percentage of copy number losses for all 23 cases involved regions >500kb and <1Mb (Figure 3) (Kruskal-Wallis non-parametric test p=0.0094).

Table 3: Frequency of autosomal CNAs (gain and loss) separated by aberration size and identified by the ADM-2 algorithm with a threshold of 6.0 and log2 ratio cutoff of 0.5 in 23 cases of canine OS. 100kb<%CNAs %CNAs≤100kb <500kb 500kb<%CNAs<1Mb %CNA>1Mb Total Case Aberrations GAIN LOSS GAIN LOSS GAIN LOSS GAIN LOSS OS1 574 48.78 1.57 28.75 1.74 7.84 0.87 9.23 1.22 OS2 366 50.27 1.09 23.77 2.19 8.74 0.27 10.38 3.28 OS3 76 14.47 11.84 15.79 13.16 7.89 3.95 14.47 18.42 OS4 150 45.33 6.67 24.67 2.67 2.67 1.33 8.67 8.00 OS5 925 53.73 3.03 21.19 5.08 3.78 1.08 7.46 4.65 OS6 306 51.31 3.59 20.59 2.61 6.54 0.98 7.84 6.54 OS7 232 5.60 56.47 1.72 21.55 0.43 5.17 1.29 7.76 OS8 79 2.53 26.58 12.66 16.46 3.80 5.06 10.13 22.78 OS9 119 16.81 10.92 8.40 17.65 5.88 4.20 7.56 28.57 OS10 75 36.00 14.67 22.67 10.67 0.00 4.00 5.33 6.67 OS11 199 31.16 7.04 18.59 12.56 3.52 2.51 9.55 15.08 OS12 135 25.93 8.89 14.81 5.93 13.33 0.74 15.56 14.81 OS13 490 0.41 27.35 3.27 31.02 1.43 8.78 4.90 22.86 OS14 1412 50.35 0.64 30.38 0.71 6.16 0.57 10.48 0.71 OS15 44 9.09 45.45 2.27 36.36 0.00 2.27 0.00 4.55 OS16 736 61.14 1.49 22.15 1.77 5.84 1.09 4.48 2.04 OS17 276 37.32 2.17 23.19 2.90 6.52 1.09 17.39 9.42 OS18 795 60.38 1.01 25.03 0.75 3.52 0.13 8.43 0.75 OS19 598 1.00 46.32 1.17 29.26 0.84 8.70 1.51 11.20 OS20 851 54.17 0.47 27.50 0.82 5.29 0.35 10.22 1.18 OS21 746 45.04 1.07 25.74 0.40 8.98 0.00 18.63 0.13 OS22 607 56.51 0.82 28.17 1.65 5.77 0.33 6.10 0.66 OS23 73 17.81 10.96 8.22 10.96 2.74 10.96 8.22 30.14

116

Figure 3: Box plots of frequency of autosomal CNAs (gain and loss) in 23 canine OS cases (array call resolution ~27kb) identified by the ADM-2 algorithm with a threshold of 6.0 and log2 ratio cutoff of 0.5. They are separated by aberration size and only cover the 38 canine autosomes as sex-mismatching was performed for each array.

Extensive genomic imbalance in human OS

A total of 15 human OS (hOS) (Table 2) cases were analyzed using a 60,000 feature oligonucleotide aCGH platform (~100kb analysis resolution). To quantify the total number of imbalances in each case we applied the ADM-2 algorithm (Agilent Genome Workbench

Software (v5.0)) with a threshold of 6.0 and a log2 ratio threshold of 0.5. Most of the human OS aCGH profiles were chaotic in nature with many aberrations. This is demonstrated by the genomic plot of hOS5, a 12-year old female with OS in the left distal femur (Figure 4). This patient had gain of human chromosome bands 6p22-p21, 8q24, and 17p12-p11.2. The case

also had regional amplification of HSA 6p21.1-p12.3:41.63- 46.44Mb (log2 ratio 2.04)

117 containing RUNX2 (chr6p21.1:45.3-45.52Mb) which we found gained in 46% of our human cases as well as 34.8% of our canine OS cases including the Rottweiler in Figure 1.

Figure 4: Array CGH profile (100kb resolution view) of A) hOS5, a 12-year old female with OS in the left distal femur. The log2 ratio value is shown on the y-axis and the x-axis represents the chromosome and position of the 22 human autosomes. B) HSA 6, C) HSA 8, and D) HSA 17 are highlighted to demonstrate regions of copy number aberrations in the case previously found in several human OS patients. Bars above and below the aberration regions indicate gain (red) and loss (green), respectively.

Aberration calls from each case were used to calculate the frequency of copy number gains and losses for each chromosomal locus (Figure 5). In general, numerous recurrent regions of gains and losses were found. Analysis of overlapping regions of copy number gains across all cases revealed a total of 1014 regions with 12% (122/1014) of these regions

118 involved in >30% of cases. The highest frequency of gain, 60%, was found in HSA 6p21.1:43.09-43.53Mb and HSA 8q24.21:128.816-128.821Mb. A smaller number of overlapping regions, 705, were lost and only 3% (24/705) of these regions had a loss frequency >30%. The highest frequency of loss (46.7%) was found on HSA 3q13.32:117.89- 117.94Mb and HSA 3q13.32:117.94-118.1Mb.

Figure 5: Frequency of copy number aberrations in 15 human OS samples. The 22 human autosomes are listed on the x-axis. The y-axis shows the percentage of cases with copy number gain (red) and copy number loss (green), for each ~100kb interval, based on log2 ratios calls by the ADM-2 algorithm with a threshold of 6.0 and log2 ratio threshold of 0.5.

Ten cases of human OS had amplification regions (log2 ratio≥2.0) spread across 13

autosomes and all 15 cases had regions of high copy number gain (log2 ratio≥1.0) spread across all 22 autosomes. There was positional overlap between regions of amplification containing HSA 6p21.1 in the 10 cases with the minimum size being 2.5 Mb. Four identical regions of high copy number gain were found in 13% (2/15) of our human cohort: HSA 8q24.3:145.25-145.30Mb, HSA 12q15:67.47-68.37Mb, HSA 16q23.1:73.81-73.84Mb, and HSA 2q11.2:96.89- 96.891Mb. Regions suggestive of homozygous loss were found in 12 patients spread across seven chromosomes. Similar to high copy number gains only 13% (2/15) contained identical regions, HSA 7q35:146.22-146.52Mb and HSA 8p11.23- p11.22:39.38-39.50Mb. The frequency of copy number gains and losses for each case based

119 on aberration size (Table 4 and Figure 6) is more variable then what we observed in canine OS. The lowest percentage of aberrations for most cases involved regions >500kb and <1Mb (Figure 6) and the most common aberration size differed for each case. A non-parametric Kruskal-Wallis test concluded that patients were more likely to have copy number gains (p=0.0056) or losses (p=0.0096) that were either >1Mb or <100kb than the other size categories.

Table 4: Frequency of CNAs (gain and loss) separated by aberration size and identified by the ADM-2 algorithm with a threshold of 6.0 and log2 ratio cutoff of 0.5 in 15 cases of human OS. The total aberrations in each case only cover the 22 human autosomes as sex- mismatching was performed for each array. 100kb<%CNAs< %CNAs≤100kb 500kb 500kb<%CNAs<1Mb %CNA>1Mb Total Case Aberrations GAIN LOSS GAIN LOSS GAIN LOSS GAIN LOSS hOS1 105 8.57 5.71 25.71 0.00 22.86 0.00 36.19 0.95 hOS2 7 14.29 28.57 0.00 14.29 14.29 0.00 28.57 0.00 hOS3 16 62.50 6.25 0.00 0.00 6.25 0.00 25.00 0.00 hOS4 49 61.22 4.08 16.33 0.00 6.12 0.00 12.24 0.00 hOS5 142 1.41 2.82 10.56 2.82 14.08 2.11 59.15 7.04 hOS6 22 9.09 13.64 22.73 4.55 0.00 9.09 40.91 0.00 hOS7 113 1.77 46.02 4.42 19.47 4.42 3.54 9.73 10.62 hOS8 187 16.58 21.39 12.83 17.65 4.81 2.67 3.74 20.32 hOS9 155 8.39 18.71 17.42 9.03 5.16 3.23 25.16 12.90 hOS10 72 13.89 2.78 9.72 4.17 9.72 2.78 48.61 8.33 hOS11 29 58.62 3.45 10.34 0.00 3.45 0.00 24.14 0.00 hOS12 44 25.00 18.18 27.27 4.55 6.82 4.55 11.36 2.27 hOS13 81 3.70 6.17 8.64 3.70 14.81 2.47 50.62 9.88 hOS14 79 3.80 18.99 6.33 20.25 3.80 5.06 12.66 29.11 hOS15 74 22.97 4.05 10.81 9.46 8.11 2.70 16.22 25.68

120

Figure 6: Box plots of frequencies of CNAs (gain and loss) in 15 human OS cases (array call resolution ~100kb) identified by the ADM-2 algorithm with a threshold of 6.0 and log2 ratio cutoff of 0.5. They are separated by aberration size.

Comparative regions of genomic imbalance in canine and human OS

The Autograph tool (Derrien et al. 2007) was used to identify orthologous regions in the canine and human genome as a means to directly compare regions of genomic imbalance in canine OS against previously defined regions of cytogenetic abnormalities in human OS (Zielenska et al. 2001; Bayani et al. 2003; Sandberg and Bridge 2003; Selvarajah et al. 2008; Sadikovic et al. 2009). The gain and loss frequencies in identified human and canine orthologous regions were then compared between our human and canine OS datasets. We found that among our canine OS samples the canine orthologous regions of HSA 1p36.32 (CFA 5q32), HSA 1q21.2-q22 (CFA 17q23-q24, 7q17, 38q15.2), HSA Chr6p22-p21 (CFA 2q11-q21.1), and HSA 8q24 (CFA 13q12.3-q21.2), showed similar CNA patterns to both originally published human OS aCGH data and our own aCGH evaluation of human OS (Figure 7). All of these comparative regions were mostly gained in copy number in our human and canine OS cases.

121

Figure 7: Canine orthologous regions of the human genome that demonstrate similar aberration frequencies in 15 cases of human OS arrayed at ~100kb resolution and 23 canine OS cases arrayed at ~27kb resolution. The x-axis represents the chromosomal cytoband while the y-axis represents the percentage of gain (red) and loss (green) per each locus on the array. (A) HSA 1p36.32 and 1q21.1-q22; orthologous to CFA 5q32 and 17q23-q24, 7q17, 38q15.2. (B) HSA 6p22-p21; orthologous to CFA 12q11-q21.1 (C) HSA 8q24; orthologous to CFA 13q12.3-q21.2.

122 We then sought to determine if chromosomal CNA patterns in our canine OS cohort showed similar patterns in our human OS aCGH analysis if orthologous human regions were directly mapped to entire canine chromosomes. We found that four canine chromosomes (CFA 8, 9, 22, and 24) showed similar frequencies of copy number gains and losses when compared to the aberration frequencies of mapped human chromosomal regions (Figure 8). CFA 8, 22, and 24 are orthologous to single human chromosomes with CFA 22 only consisting of part of HSA 13. The penetrance plots of copy number changes were extremely similar as regions on CFA 8 and HSA 14 were gained more than lost, regions on CFA 22 and HSA 13 were lost more than gained, and the orthologous regions CFA 24q21-q25 and HSA 20q11.21-q13.33, were both gained at a frequency of 50-70%. CFA 9 is orthologous to two separate human chromosomes. Combining and ordering HSA 17 and HSA 9 sequence orthologous to CFA 9 showed comparable aberration patterns. Specifically, an increase in copy number gain at CFA 9q25 (HSA 9q34.3) (Figure 8) was prominent in the human and canine cases.

123

Figure 8: Chromosome profiles of CNA frequencies of 23 canine OS cases (~27kb resolution) and 15 human OS cases (~100kb resolution). The x-axis is the percentage of gain (red) and loss (green) while the y-axis is the chromosomal position and cytobands. Mapped human chromosomal regions appear to the right of canine (A) CFA 8, (B) CFA 9, (C) CFA 22, (D) CFA 24 with a bar on the right side of the human chromosome representing descending human sequence and a bar on the left side representing ascending human sequence. A line through the human chromosome frequency profile notes the start of sequence from a separate chromosome.

124 Discussion:

Over the last 30-40 years investigators have evaluated spontaneous occurrence of cancer in dogs to make important contributions to the understanding and practice of human oncology in fields such as basic tumor biology and immunology (Gordon et al. 2009). Yet, only in the last decade has comparative oncology been used as a tool to aid in the identification of genome-wide abnormalities in human diseases such as cancer. This is largely due to the advancement of molecular technology and genomic sequencing, which allows for direct comparison of species at a genome-wide level. Throughout cancer progression genomic abnormalities accumulate making it harder to pinpoint the genetic changes which originally contributed to the disease manifestation (Tang et al. 2010). Therefore, cross-species genomic evaluation of cancers enables us to identify the genomic instability which is driving disease development and progression. One such disease that would benefit from cross-species genomic evaluation is osteosarcoma (OS), a malignant cancer of the bone that contains several physiological similarities in disease presentation and progression between humans and dogs. Previous studies have demonstrated comparable expression signatures of OS tumors between dog and human cases (Mendoza et al. 1998; Paoloni et al. 2009; Zhang et al. 2009). Because of these studies and previous cytogenetic evaluation of canine OS that indicated that the tumors contained a high rate of genomic imbalance (Thomas et al. 2009; Angstadt et al. submitted), dog OS is an ideal candidate for further comparative study aimed at directly comparing genome-wide CNAs in humans and dogs with OS. Identification of analogous regions in OS subject to genomic imbalance in both the human and canine genome could highlight cancer associated abnormalities and thus areas of interest for the development of translational cancer treatment drugs.

OS in both humans and dogs presents with chaotic karyotypes and large regions of genomic instability (Bayani et al. 2003; Selvarajah et al. 2007; Selvarajah et al. 2008; Thomas et al. 2009). We found genomic instability in our oligo aCGH analysis of both human (~100kb resolution) and canine (~27kb resolution) OS (Figure 2 and 4). Using high resolution oligo-array analysis allowed us to identify microaberrations (<500kb) which were

125 previously undetectable with a 1Mb-resolution BAC array (Angstadt et al. submitted) in 23 cases of canine OS. Included in these aberrations was a chromosomal region of high frequency gain (>40%), located on CFA 2q24.1:36.4-36.8Mb, which contains the RNA editing gene, adenosine deaminase RNA specific B2 mRNA (ADARB2). Similar to previous results (Angstadt et al. submitted) the canine OS oligo aCGH analysis found that microaberrations were more likely to occur than whole chromosomal aneuploidy with copy number gains mainly <500kb in size. Unlike what was defined in dogs, humans OS aCGH showed average gain aberration sizes to be >1Mb. The size of copy number losses was variable for both human and canines. However, the differences in array analysis resolution (~100kb for human and ~27kb for canine) may be the reason for the variation in aberration size between humans and dogs. As our human array resolution was only approximately ≤100kb that may explain why dogs had more copy number gains ≤100kb and therefore direct comparison of aberration size cannot be drawn from this analysis. Yet, the sheer number of microaberrations in both human and dog OS demonstrates the complex nature of the disease in relation to genomic instability.

After identification of the most frequent regional copy number gains and losses in our dog and human sample sets several similarities in CNA patterns were found for singular genes within these regions. Because of the high similarity of instability in humans and dogs it can be assumed that these genes are directly contributing to the cancer phenotype. The highest gain frequencies (73.6%) among our canine cohort were located at CFA 3q32:65.33- 65.34Mb, CFA 6q22:41.91-41.92Mb, CFA 9q11.2:5.24-5.36Mb, and CFA 13q21.1:39- 39.16Mb. Currently no gene annotation exists for regions within CFA 3q32, CFA 9q11, and CFA13q21.1 but the human orthologous region to CFA 13q21.1 (8q24.3) also did have a high frequency of gain (46.7%) in our human OS cases suggesting that this genomic region is important in OS. The orthologous region to CFA 6q22 (HSA 16p13.3), which contains the tumor suppressor gene tuberous sclerosis 2 isoform 4 (TSC2), was not a region of high gain (only 6%), indicating that CFA 6q22 may be a canine OS specific aberration. Even though TSC2 had a high copy number gain frequency shown by both the oligo aCGH and previous

126 1Mb-resolution BAC aCGH and FISH analysis, evaluation of 26 canine OS cases by qRT- PCR did not show over-expression of the gene (Angstadt et al. submitted). Mutations in TSC2 are linked to the tuberous sclerosis complex (TSC) in humans and renal tumors are associated with TSC (Inoki and Guan 2009). This implies that because altered levels of TSC2 expression in dog OS cases are not associated with a gain of gene dosage, a mutation in the gene maybe causing dysregulation. The level of TSC2 protein expression would also need to be analyzed to understand the effect of increase gene dosage.

The highest percentage of copy number loss (69.57%) in our canine cases was in CFA 11q16:44.26-44.30Mb which is orthologous to HSA 9p21.3:21.94-21.98Mb and was lost in 13.3% of our human cohort. The genes CDKN2A/CDKN2B (CFA 11q16:44.25- 44.26Mb) are located in this region. CDKN2A/CDKN2B along with PTEN (chr26q25:40.92- 40.98Mb) was also located in overlapping regions of homozygous loss in our canine cohort with overall copy number loss frequencies of 65.2%. The degree of PTEN loss in dogs was previously shown by FISH analysis (Angstadt et al. submitted). In our human dataset, these genes were lost at a frequency of 13.3% for PTEN (Chr10q23.2:89.63-89.72Mb) and 20% for the entire extent of CDKN2A/CDKN2B (Chr9p21.3:21.97-22Mb), indicating that the loss of these genes is important in OS. Prior research in both human and canine OS concluded that PTEN had a high frequency of copy number loss (Freeman et al. 2008; Thomas et al. 2009) often as a homozygous deletion resulting in complete loss of PTEN expression. The role of CDKN2A/CDKN2B in OS has not been evaluated in dogs but it was inactivated by deletions and mutations in human OS (Tsuchiya et al. 2000; Mohseny et al. 2009). The fact that both PTEN and CDKN2A/CDKN2B are deleted at high frequencies in both species implies that they are driver alterations in OS.

Oligo aCGH analysis of human OS mirrored previous aCGH studies with the highest gain frequencies (60%) found in HSA 6p21.1:43.09-43.53Mb and HSA 8q24.21:128.816- 128.821Mb (Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Bayani et al. 2003; Squire et al. 2003;

127 Selvarajah et al. 2008). The HSA 6p21.1 region, orthologous to CFA 12q13:14.48-14.9Mb (copy number gain 52.2%), codes for several genes, PTK7, SRF, CUL9, C6orf108, SLC22A7, CRIP3, ABCC10, DLK2, TJAP1, YIPF3, XP05, and the regions HSA 8q24.21, orthologous to CFA 13q13:28.23-28.24 (copy number gain 56.5%), codes for MYC. Over-expression of MYC has been found to play a role in human OS (Gamberi et al. 1998; de Nigris et al. 2007), yet very recently Sadikovic et al. showed that no significant changes in expression of MYC occurred in tumors in comparison with normal osteoblasts (Sadikovic et al. 2010). Similarly, copy number gains of MYC have been shown previously in canine OS (Thomas et al. 2009; Angstadt et al. submitted) although no change in MYC expression was found in either normal or gain of copy number tumors (Angstadt et al. submitted). Additional studies of MYC aimed at identifying mutations and protein expression changes would be beneficial to interpret the role of MYC in OS. Chromosomal regions HSA 3q13.32:117.89-117.94Mb and HSA 3q13.32:117.94-118.1Mb contained the highest frequency of loss (46.7%) in our analysis, but currently no gene annotation exits for these regions. Further sequencing and annotation within these regions could identify genes involved in OS.

In addition to identifying similar instable genes in human and dog OS we found genome-wide parallel CNA patterns between orthologous regions in our canine and human OS cohort. Canine orthologous regions of HSA 1p36.32 (CFA 5q32), HSA 1q21.2-q22 (CFA 17q23-q24, 7q17, 38q15.2), HSA Chr6p22-p21 (CFA 2q11-q21.1), and HSA 8q24 (CFA 13q12.3-q21.2), all had similar frequencies of CNA to both previously reported human OS aCGH data and our own aCGH evaluation of human OS (Figure 5). These observations are to the first to indicate that in addition to anatomical, physiological, and singular gene abnormalities similarities, at a genome-wide level human and dog OS share orthologous regions of recurrent CNA. Several cancer associated genes reside within these regions of copy number gain, including PRDM16, TNFRSF14, TP73, MUC1, CDC5L, RUNX2, and MYC. Expression analysis in human OS has shown that CDC5L (Lu et al. 2008) and RUNX2 (Sadikovic et al. 2009; Sadikovic et al. 2010), a member of the RUNX gene family of differentiation mediators expressed at different stages of osteoblast development (Ito 2004),

128 are over-expressed in the cancer. RUNX2 was also previously found over-expressed in canine OS (Angstadt et al. submitted). In addition, RUNX2 has been found over-expressed in human OS patients that showed a poor response to chemotherapy relative to good responders (Sadikovic et al. 2010). Additional characterization of RUNX2 in canine OS would aid in uncovering its role in disease progression and may identify a treatment methodology to arrest the dysregulation of the gene.

To further demonstrate the parallels in CNA frequencies orthologous human regions were directly mapped to whole canine chromosomes (CFA 8, 9, 22, and 24). Similar frequencies of copy number gains were found on CFA 8 orthologous to HSA 14, and CFA 25q21-25 orthologous to 20q11.21-q13.33. Although CFA 9 mapped to two human chromosomes a common pattern of copy number gains existed between the orthologous regions, particularly with the increased gain frequencies appearing on the orthologous regions to CFA 9q25, HSA 9q34.3. A common copy pattern of copy number loss was found in the region of HSA 13 orthologous to CFA 22 (Figure 6). The tumor suppressor gene RB1 is located in the region of loss on CFA 22 (6.01-6.1Mb) and HSA 13 (48.88-48.94Mb). RB1 alterations are frequent in human OS and it is estimated that 60% of human patients have RB- associated abnormalities (Feugeas et al. 1996). Recent expression analysis of genes associated with human OS found loss of RB1 expression in tumors relative to normal osteoblasts (Sadikovic et al. 2010). As RB1 deletion has not been extensively characterized in canine OS further study could indicate a similar trend in dogs.

OS is a highly metastatic disease in dogs and humans subject to extensive genomic instability, producing it difficult to differentiate between genetic abnormalities that are directly responsible for disease manifestation and those that are only passenger abnormalities, having no effect on disease etiology. Using high resolution oligonucleotide aCGH to identify recurrent CNAs in OS in combination with cross-species evaluation of 23 cases of canine OS and 15 cases of human OS addressed these difficulties. We identified microaberrations in the disease, large parallels in regional CNAs in OS between humans and

129 dogs, and pinpointed ‘OS associated genes’ subject to imbalance. The occurrence of microaberrations in both human and canine OS demonstrates the complexity of the disease and challenges involved in identification of key genetic players. The combination of high resolution aCGH technology and cross-species genome synteny maps allowed us to detect genomic regions of orthology in the human and canine genome subject to similar patterns of imbalance. In addition we showed that MYC, CDKN2A/CDKN2B, PTEN, RUNX2, and RB1 are subject to similar CNAs in human and dogs and thus are ‘OS associated genes’. Additional investigation into the genome-wide CNA parallels between human and canine orthologous regions that we identified will expand the list of OS associated genes, helping to pinpoint all key genetic players in the disease. This analysis supports further comparative studies of dog and human OS while tackling some of the challenges involved in identification of driver genetic abnormalities leading to OS development and progression. Subsequently, these genetic abnormalities could then serve as targets of novel genetic therapeutics that once developed and tried in dogs may be translatable to human patients.

130 References

Angstadt, A. Y., A. Motsinger-Reif, R. Thomas, W. C. Kisseberth, C. G. Couto, D. L. Duval, D. M. Nielsen, J. F. Modiano and M. Breen (submitted). "Characterization of canine osteosarcoma by array comparative genomic hybridization and qRT-PCR: Signatures of genomic imbalance in canine osteosarcoma parallels the human counterpart." Cancer Genet Cytogenet.

Antinoff, N. and K. Hahn (2004). "Ferret oncology: diseases, diagnostics, and therapeutics." Vet Clin North Am Exot Anim Pract 7(3): 579-625, vi.

131 de Nigris, F., C. Botti, R. Rossiello, E. Crimi, V. Sica and C. Napoli (2007). "Cooperation between Myc and YY1 provides novel silencing transcriptional targets of alpha3beta1-integrin in tumour cells." Oncogene 26(3): 382-94.

Derrien, T., C. Andre, F. Galibert and C. Hitte (2007). "AutoGRAPH: an interactive web server for automating and visualizing comparative genome maps." Bioinformatics 23(4): 498-9.

Ferracini, R., P. Angelini, E. Cagliero, A. Linari, M. Martano, J. Wunder and P. Buracco (2000). "MET oncogene aberrant expression in canine osteosarcoma." J Orthop Res 18(2): 253-6.

Freeman, S. S., S. W. Allen, R. Ganti, J. Wu, J. Ma, X. Su, G. Neale, J. S. Dome, N. C. Daw and J. D. Khoury (2008). "Copy number gains in EGFR and copy number losses in PTEN are common events in osteosarcoma tumors." Cancer 113(6): 1453-61.

132 Gokgoz, N., J. S. Wunder, S. Mousses, S. Eskandarian, R. S. Bell and I. L. Andrulis (2001). "Comparison of p53 mutations in patients with localized osteosarcoma and metastatic osteosarcoma." Cancer 92(8): 2181-9.

Gordon, I., M. Paoloni, C. Mazcko and C. Khanna (2009). "The Comparative Oncology Trials Consortium: using spontaneously occurring cancers in dogs to inform the cancer drug development pathway." PLoS Med 6(10): e1000161.

Hershey, A. E., R. R. Dubielzig, M. L. Padilla and S. C. Helfand (2005). "Aberrant p53 expression in feline vaccine-associated sarcomas and correlation with prognosis." Vet Pathol 42(6): 805-11.

Inoki, K. and K. L. Guan (2009). "Tuberous sclerosis complex, implication from a rare genetic disease to common cancer treatment." Hum Mol Genet 18(R1): R94-100.

Ito, Y. (2004). "Oncogenic potential of the RUNX gene family: 'overview'." Oncogene 23(24): 4198-208.

Johnson, A. S., C. G. Couto and C. M. Weghorst (1998). "Mutation of the p53 tumor suppressor gene in spontaneously occurring osteosarcomas of the dog." Carcinogenesis 19(1): 213-7.

Jong, K., E. Marchiori, A. van der Vaart, B. Ylstra, M. Weiss and G. Meijer (2003). "Chromosomal breakpoint detection in human cancer." Applications of Evolutionary Computing 2611: 54-65.

Kirpensteijn, J., M. Kik, G. R. Rutteman and E. Teske (2002). "Prognostic significance of a new histologic grading system for canine osteosarcoma." Vet Pathol 39(2): 240-6.

133 Kirpensteijn, J., M. Kik, E. Teske and G. R. Rutteman (2008). "TP53 gene mutations in canine osteosarcoma." Vet Surg 37(5): 454-60.

Klein, W. R., G. E. Bras, W. Misdorp, P. A. Steerenberg, W. H. de Jong, R. H. Tiesjema, A. W. Kersjes and E. J. Ruitenberg (1986). "Equine sarcoid: BCG immunotherapy compared to cryosurgery in a prospective randomised clinical trial." Cancer Immunol Immunother 21(2): 133-40.

134 Ponting, G. Schuler, J. Schultz, G. Slater, A. F. Smit, E. Stupka, J. Szustakowski, D. Thierry-Mieg, J. Thierry-Mieg, L. Wagner, J. Wallis, R. Wheeler, A. Williams, Y. I. Wolf, K. H. Wolfe, S. P. Yang, R. F. Yeh, F. Collins, M. S. Guyer, J. Peterson, A. Felsenfeld, K. A. Wetterstrand, A. Patrinos, M. J. Morgan, P. de Jong, J. J. Catanese, K. Osoegawa, H. Shizuya, S. Choi and Y. J. Chen (2001). "Initial sequencing and analysis of the human genome." Nature 409(6822): 860-921.

Levine, R. A. and M. A. Fleischli (2000). "Inactivation of p53 and retinoblastoma family pathways in canine osteosarcoma cell lines." Vet Pathol 37(1): 54-61.

Levine, R. A., T. Forest and C. Smith (2002). "Tumor suppressor PTEN is mutated in canine osteosarcoma cell lines and tumors." Vet Pathol 39(3): 372-8.

135 Friedrich, D. Gage, M. Garber, G. Gearin, G. Giannoukos, T. Goode, A. Goyette, J. Graham, E. Grandbois, K. Gyaltsen, N. Hafez, D. Hagopian, B. Hagos, J. Hall, C. Healy, R. Hegarty, T. Honan, A. Horn, N. Houde, L. Hughes, L. Hunnicutt, M. Husby, B. Jester, C. Jones, A. Kamat, B. Kanga, C. Kells, D. Khazanovich, A. C. Kieu, P. Kisner, M. Kumar, K. Lance, T. Landers, M. Lara, W. Lee, J. P. Leger, N. Lennon, L. Leuper, S. LeVine, J. Liu, X. Liu, Y. Lokyitsang, T. Lokyitsang, A. Lui, J. Macdonald, J. Major, R. Marabella, K. Maru, C. Matthews, S. McDonough, T. Mehta, J. Meldrim, A. Melnikov, L. Meneus, A. Mihalev, T. Mihova, K. Miller, R. Mittelman, V. Mlenga, L. Mulrain, G. Munson, A. Navidi, J. Naylor, T. Nguyen, N. Nguyen, C. Nguyen, R. Nicol, N. Norbu, C. Norbu, N. Novod, T. Nyima, P. Olandt, B. O'Neill, K. O'Neill, S. Osman, L. Oyono, C. Patti, D. Perrin, P. Phunkhang, F. Pierre, M. Priest, A. Rachupka, S. Raghuraman, R. Rameau, V. Ray, C. Raymond, F. Rege, C. Rise, J. Rogers, P. Rogov, J. Sahalie, S. Settipalli, T. Sharpe, T. Shea, M. Sheehan, N. Sherpa, J. Shi, D. Shih, J. Sloan, C. Smith, T. Sparrow, J. Stalker, N. Stange-Thomann, S. Stavropoulos, C. Stone, S. Stone, S. Sykes, P. Tchuinga, P. Tenzing, S. Tesfaye, D. Thoulutsang, Y. Thoulutsang, K. Topham, I. Topping, T. Tsamla, H. Vassiliev, V. Venkataraman, A. Vo, T. Wangchuk, T. Wangdi, M. Weiand, J. Wilkinson, A. Wilson, S. Yadav, S. Yang, X. Yang, G. Young, Q. Yu, J. Zainoun, L. Zembek, A. Zimmer and E. S. Lander (2005). "Genome sequence, comparative analysis and haplotype structure of the domestic dog." Nature 438(7069): 803-19.

McCleese, J. K., M. D. Bear, S. L. Fossey, R. M. Mihalek, K. P. Foley, W. Ying, J. Barsoum and C. A. London (2009). "The novel HSP90 inhibitor STA-1474 exhibits biologic activity against osteosarcoma cell lines." Int J Cancer 125(12): 2792-801.

Mendoza, S., T. Konishi, W. S. Dernell, S. J. Withrow and C. W. Miller (1998). "Status of the p53, Rb and MDM2 genes in canine osteosarcoma." Anticancer Res 18(6A): 4449-53.

Mirabello, L., R. J. Troisi and S. A. Savage (2009). "Osteosarcoma incidence and survival rates from 1973 to 2004: data from the Surveillance, Epidemiology, and End Results Program." Cancer 115(7): 1531-43.

136 Mohseny, A. B., K. Szuhai, S. Romeo, E. P. Buddingh, I. Briaire-de Bruijn, D. de Jong, M. van Pel, A. M. Cleton-Jansen and P. C. Hogendoorn (2009). "Osteosarcoma originates from mesenchymal stem cells in consequence of aneuploidization and genomic loss of Cdkn2." J Pathol 219(3): 294-305.

Mueller, F., B. Fuchs and B. Kaser-Hotz (2007). "Comparative biology of human and canine osteosarcoma." Anticancer Res 27(1A): 155-64.

Nasir, L. and S. W. Reid (1999). "Bovine papillomaviral gene expression in equine sarcoid tumours." Virus Res 61(2): 171-5.

Paoloni, M. and C. Khanna (2008). "Translation of new cancer treatments from pet dogs to humans." Nat Rev Cancer 8(2): 147-56.

Sandberg, A. A. and J. A. Bridge (2003). "Updates on the cytogenetics and molecular genetics of bone and soft tissue tumors: osteosarcoma and related tumors." Cancer Genet Cytogenet 145(1): 1-30.

Seltenhammer, M. H., E. Heere-Ress, S. Brandt, T. Druml, B. Jansen, H. Pehamberger and G. W. Niebauer (2004). "Comparative histopathology of grey-horse-melanoma and human malignant melanoma." Pigment Cell Res 17(6): 674-81.

137 Selvarajah, G. T., J. Kirpensteijn, M. E. van Wolferen, N. A. Rao, H. Fieten and J. A. Mol (2009). "Gene expression profiling of canine osteosarcoma reveals genes associated with short and long survival times." Mol Cancer 8: 72.

Tang, J., S. Le, L. Sun, X. Yan, M. Zhang, J. Macleod, B. Leroy, N. Northrup, A. Ellis, T. J. Yeatman, Y. Liang, M. E. Zwick and S. Zhao (2010). "Copy number abnormalities in sporadic canine colorectal cancers." Genome Res 20(3): 341-50.

Tarkkanen, M., R. Karhu, A. Kallioniemi, I. Elomaa, A. H. Kivioja, J. Nevalainen, T. Bohling, E. Karaharju, E. Hyytinen, S. Knuutila and et al. (1995). "Gains and losses

138 of DNA sequences in osteosarcomas by comparative genomic hybridization." Cancer Res 55(6): 1334-8.

Withrow, S. J., B. E. Powers, R. C. Straw and R. M. Wilkins (1991). "Comparative aspects of osteosarcoma. Dog versus man." Clin Orthop Relat Res(270): 159-68.

Withrow, S. J. and D. Vail, Eds. (2007). Small Animal Clinical Oncology. Philadelphia, WB Sanders.

Wittenburg, L. A., L. Bisson, B. J. Rose, C. Korch and D. H. Thamm (2010). "The histone deacetylase inhibitor valproic acid sensitizes human and canine osteosarcoma to doxorubicin." Cancer Chemother Pharmacol.

139 Zielenska, M., J. Bayani, A. Pandita, S. Toledo, P. Marrano, J. Andrade, A. Petrilli, P. Thorner, P. Sorensen and J. A. Squire (2001). "Comparative genomic hybridization analysis identifies gains of 1p35 approximately p36 and chromosome 19 in osteosarcoma." Cancer Genet Cytogenet 130(1): 14-21.

140

Chapter IV

Conclusions

141 Concluding Remarks

According to the American Cancer Society (ACS, http://www.cancer.org) the estimated number of new cancer cases for 2010, excluding basal and squamous cell skin cancers and in-situ carcinomas, is roughly 1.5 million people. Although advances in health care from the 1970s to 2000s have increased the 5-year survival rate for all cancers from 50% to 66% (ACS, http://www.cancer.org), we are still struggling to arrest the disease completely. Since human-based studies of cancer are faced with difficulties associated with ethical dilemmas, an appropriate model organism is necessary to increase our understanding of disease pathogenesis and progression. Since the beginning of the 12th century, the foremost model organism for genetic study of complex human diseases such as cancers has been the mouse (Kirkness et al. 2003; Lindblad-Toh et al. 2005). The mouse does boast an impressive array of experimental resources (Paigen 1995; Bucan and Abel 2002) for genetic studies, but it is really only ideal for studying a previously identified putative mutation and is less useful for finding mutations in the disease. In the last decade, one model organism, the domestic dog, has begun to fill this gap between mouse-based studies and human diseases, primarily due to advances in the availability of genetic experimental resources (Breen et al. 1999; Breen et al. 2001; Kirkness et al. 2003; Breen et al. 2004; Lindblad-Toh et al. 2005). Unlike induced cancers in mice, dogs spontaneously develop cancers which, therefore, share many pathophysiological similarities with human cancers. In addition, comparison of the high quality annotated assemblies for both human and canine genomes (Lander et al. 2001; Kirkness et al. 2003; Lindblad-Toh et al. 2005) have revealed a high degree of similarity suggesting that genetic factors subject to dysregulation in human and canine cancers may also be similar.

Within this thesis I addressed the similarities that exist between genetic dysregulation in one such cancer, osteosarcoma (OS), the most common primary malignant bone tumor in humans and dogs (Withrow and Vail 2007; Mirabello et al. 2009a; Mirabello et al. 2009b). In the USA, fewer than 1,000 people per year (Mirabello et al. 2009b; Paoloni et al. 2009) are diagnosed with OS, whereas OS affects the domestic dog at a rate of over 10,000 new

142 cases per year in the USA (Withrow et al. 1991; Fossey et al. 2009). Thus, because of the sheer numbers of canine patients diagnosed with OS, I could evaluate a larger sample set, increasing disease understanding in dogs while providing translational information for human OS. Through a genome-wide approach, I (1) identified the cytogenetic abnormalities in 123 cases of canine OS, (2) characterized the frequency of these genomic instabilities and their effect on transcriptional dysregulation of select genes, and (3) directly compared genomic imbalance in canine and human OS.

Cytogenetically, human OS is known to have markedly abnormal karyotypes that contain complex structural changes (translocations and/or rearrangements) and DNA copy number changes (Bayani et al. 2007). Contrary to other human sarcomas, conventional OS is not associated with a specific recurrent translocation or other chromosomal rearrangement. In fact, it is characterized by an array of sequential and well-orchestrated genetic changes that involve numerous tumor-suppressor genes and oncogenes (Sandberg and Bridge 2003). Low-resolution (10-20Mb) array comparative genomic hybridization (aCGH) analysis of 38 cases of canine OS indicated that a high degree of genomic instability, consistent with observations of human OS, was present in canine OS (Thomas et al. 2009). To enhance this analysis I conducted both 1Mb-resolution and ~27kb-resolution aCGH analyses of canine OS. I found a similar level of cytogenetic chaos previously identified in canine and human OS demonstrated by the fact that all canine autosomes had some level of imbalance in one or several of the 123 dogs. This imbalance in dog OS involved whole chromosomal aneuploidy, single-locus copy number aberrations (CNAs), and structural rearrangements (Chapters II and III). In addition, analysis of CNAs at ~27kb identified microaberrations (<500kb) of gain and loss that existed between larger regions of imbalance. These microabberations were more likely to be <100kb for copy number gain and >500kb and <1Mb for copy number loss, reiterating the cytogenetic complexity of canine OS (Chapter III). Fluorescence in-situ hybridization (FISH) validation of a subset of canine OS cases visually confirmed CNAs found in the aCGH analysis and demonstrated the typical extent of

143 numerical (and structural) aberrations present in the cases evaluated (Chapter II and Appendix II).

The modern domestic dog encompasses over 350 physically diverse breeds defined by specific behavioral and physical characteristics that have been driven to exceptionally high frequency by population bottlenecks and strong artificial selection (Sutter and Ostrander 2004; Karlsson and Lindblad-Toh 2008; Shearin and Ostrander 2010). Thus, I hypothesized that different breeds would be subject to unique patterns of CNAs, the result of the individual genetic framework of each breed. Previous studies of genomic instability in canine OS supported this hypothesis since they found that a pattern of copy number changes differed between dog breeds for specific cancer-associated genes (Thomas et al. 2009). A total of 19 breeds with focus on four main breeds, Greyhound, Great Pyrenees, Golden Retriever, and Rottweiler, were recruited to address this hypothesis. I did not find a significant difference between aCGH-defined regional aberrations and breed groups (Chapter II). This result is possibly because of small sample sizes, <35 cases were available for each of the individual breed groups, which limited the statistical power of the analyses. In addition to identifying if a difference was present between aCGH-defined regional aberrations and breed groups, I also analyzed the differences in aCGH data between three morphological cell subtypes, osteoblastic, chondroblastic, and fibroblastic. In contrast with the breed analysis, a statistically significant difference was found between genomic CNAs in osteoblastic and chondroblastic tumors (Chapter II). A difference in biological behavior between the morphological subtypes of OS was not well established, yet I showed that it may be possible to distinguish morphological cell subtypes by specific CNAs.

Several previous studies have sought to define the roles of specific tumor suppressor genes and oncogenes in primary canine OS as well as canine OS cell lines, using expression and proteomic analysis (Levine and Fleischli 2000; Levine et al. 2002; Kirpensteijn et al. 2008; Selvarajah et al. 2008; Fossey et al. 2009; Zhang et al. 2009). The scope of these studies did not analyze the effect that numerical imbalance has on the transcriptional

144 regulation of these genes. I selected six cancer-associated genes (TSC2, RHOC, RUNX2, MYC, TUCS3, PTEN) in genomic regions I had identified as experiencing high aberration frequency (Chapters II and III) to identify if copy number change was one factor involved in transcriptional dysregulation of these genes. The extent of the copy number changes of these genes identified by aCGH analysis was confirmed by FISH analysis (Chapter II), reinforcing the value of analyzing them for expression changes. Evaluation of 26 of the 123 cases of canine OS by qRT-PCR for these select genes concluded that CNAs were a factor involved in the over-expression of the osteoblast developmental gene RUNX2 and under-expression of the tumor suppressor genes, TUSC3 and PTEN. Since a statistically significant difference between copy number changes and expression was not found for all genes, it appeared that imbalances were only associated with transcriptional dysregulation for select genes. Yet, to date this was the first analysis to showcase the involvement of RUNX2 and TUSC3 in canine OS (Chapter II). TSC2 and MYC were also found in regions of high gain frequencies when the analysis resolution was increased by oligonucletide aCGH analysis (1-2Mb to ~27kb) (Chapter III), but since I did not find over-expression of these genes (Chapter II) their role in dog OS pathology remains unclear. Mutations in TSC2 are linked to the tuberous sclerosis complex (TSC) in humans and renal tumors are associated with TSC (Inoki and Guan 2009). This implies that because altered levels of TSC2 expression in dog OS cases are not associated with a gain of gene dosage, a mutation in the gene maybe causing dysregulation. Over-expression of MYC has been shown in human OS (Gamberi et al. 1998; de Nigris et al. 2007), yet very recently Sadikovic et al. showed that no significant changes in expression of MYC occurred in tumors in comparison with normal osteoblasts (Sadikovic et al. 2010). In addition, data generated in Chapter III was included in a collaboration study (Thayanithy et al. submitted) which found that restoration of downregulated microRNAs in human OS at the HSA chr14q32 locus reduced the steady-state levels of MYC and induced apoptosis in the human OS cell line, SaoS2. Their combined analysis of miRNA, mRNA, and DNA CNAs in both human and canine OS support a model where the deregulation of a proapoptotic network involving 14q32 miRNAs, MYC, and miR-17-92 plays a key role in OS (Thayanithy et al. submitted). Therefore, further characterization of TSC2 and MYC aimed

145 at identifying their role in deregulated cellular pathways would be beneficial to interpret their role in dog OS.

A common theme throughout this thesis has been one of comparative oncology, a discipline that often refers to the integration of study of naturally occurring cancers in animals into studies of human cancer biology and therapy (Paoloni and Khanna 2008). Due to the advances in molecular technology we can now conduct genomic cross-species data mining in a comparative oncology setting in order to characterize similar genetic abnormalities between human and canine diseases (Paoloni et al. 2009). To date only one study had conducted genomic cross-species data mining on OS using parallel oligonucleotide array platforms (Paoloni et al. 2009) to identify OS expression signatures for primary tumors and normal tissues from both dogs and humans. They found strong similarities between human and canine OS, and hierarchical cluster analysis of 265 orthologous transcripts could not distinguish the cancers based on species alone (Paoloni et al. 2009). This study implies that numerical imbalances in OS may also be similar between humans and canines. In fact, this hypothesis was addressed in both Chapter II and Chapter III of this thesis. 1Mb- resolution aCGH profiling of 123 canine OS cases (Chapter II) indicated that the high occurrence of genetic alterations characteristic of human OS was also a striking feature of canine OS. Subsequently, higher resolution oligonucleotide aCGH analyses (Chapter III) of OS at ~27kb on 23 of 123 dog cases and ~100kb on 15 human cases allowed for direct comparisons of genomic instability in orthologous regions of the human and canine genome. Matching previous analyses of genomic imbalance in human OS (Forus et al. 1995; Tarkkanen et al. 1995; Simons et al. 1997; Tarkkanen et al. 1998; Tarkkanen et al. 1999; Zielenska et al. 2001; Bayani et al. 2003; Squire et al. 2003; Selvarajah et al. 2008), most of the human OS aCGH profiles were chaotic in nature with many aberrations. The same instability I saw at 1Mb-resolution for canine OS was also seen at ~27kb-resolution, and my ability to find microaberrations was greatly enhanced by the oligo-array analysis. Directly comparing regions of high aberration frequency in my human OS population to my canine OS population concluded that the canine orthologous regions of HSA 1p36.32 (CFA 5q32),

146 HSA 1q21.2-q22 (CFA 17q23-q24, 7q17, 38q15.2), HSA Chr6p22-p21 (CFA 2q11-q21.1), and HSA 8q24 (CFA 13q12.3-q21.2), had parallel frequencies of CNA to both originally published human OS aCGH data and my own aCGH evaluation of human OS (Chapter III). In addition, when the human orthologous regions to CFA 8 (HSA 14), CFA 9 (HSA 17 and HSA 9), CFA 22 (HSA 13), and CFA 24 (HSA 20) were mapped to their respective canine chromosome, an analogous pattern of CNA was found (Chapter III). Thus, higher resolution oligonucleotide analysis of both canine and human OS demonstrated that large parallels exist in regional CNAs. Characterization of these parallels allowed for the pinpointing of the ‘OS associated genes,’ MYC, CDKN2A/CDKN2B, PTEN, RUNX2, and RB1, which were subject to imbalance in both humans and dogs (Chapter III). My ability to identify cross-species OS associated genes supports additional characterization of parallel genome-wide CNAs in orthologous human and dog regions as a means to identify new candidate genes responsible for OS manifestation.

Foremost, the discoveries in this thesis increase the understanding of cytogenetic abnormalities in canine OS which impacts a large number of pet dogs each year in the USA. Secondarily, my results support the dog as a translational biomedical model for human OS as shown by the extensive parallels in disease genomic imbalance. Lastly, the genes that I identified as ‘OS associated genes’ because they had similar aberration patterns in humans and dogs may serve as possible targets of novel genetic therapeutics. Due to the naturally shorter lifespan of dogs these novel therapeutics could be developed and tried in less time in dogs with OS, serving to advance treatment options for both humans and dogs.

Future evaluation of the complexity in Canine OS

The work presented in this thesis provides a solid foundation upon which several future genomic and gene-based studies could be conducted to analyze abnormalities in canine OS and their comparability to human OS. The genomic instabilities that I have identified and characterized here are the building blocks for enhanced understanding of the biological

147 mechanisms in OS. For example, the significant associations found between the CNA pattern and expression changes of RUNX2 and TUSC3 need to be further evaluated. RUNX2 is known to be over-expressed and to show gain in copy number in human OS (Sadikovic et al. 2009; Sadikovic et al. 2010), but these are only recent observations. Considering the fact that I found this gene to have a similar aberration pattern in human and dog OS, further analysis of this gene and its protein regulation in dogs would aid both species. TUSC3 is a novel candidate tumor suppressor gene and since this gene has not been evaluated in either canine or human OS to date, it would merit investigation. Additional characterization of both genes at both the transcriptional and translational level may prove prognostic significance and/or identify new therapeutic targets.

Along with gene-based studies of OS, my thesis supports the need for future genome- wide analysis of canine OS. At the beginning of this study one goal was to identify if any CNAs were associated with specific dog breeds affected by OS and if particular CNAs were of prognostic significance. I was unable to conclude that any specific CNAs were associated with a particular dog breed, but I did find a significant difference in CNA frequencies between different cell morphological subtypes, suggesting that an individual’s genetic framework does have an effect on aberrations present in dog OS. One reason behind my inability to associate any CNAs with specific breeds may be causal of the increase in array analysis resolution from previous research at 10-20Mb (Thomas et al. 2009), that indicated that a breed’s genetic background influenced tumor karyotypes for specific tumor associated genes, to 1Mb and ~27kb resolution. As dog OS is cytogenetically complex an increase in aCGH analysis resolution identified more aberration regions and breakpoints. This complexity caused difficulty in determining significant differences in CNAs between dog breeds because we were performing 1066 statistical tests with the 1Mb-resolution array data, one test for each genomic region subject to imbalance (Bonferroni adjusted p-value <0.05, Chapter II). Thus the differences between dog breeds in aberration frequencies in my 1Mb- resolution array analysis (assuming 30 cases per breed group) would need to be 60% or better to have an 80% power to detect a difference. In addition, in order to detect differences in

148 aberration frequency between breeds of 30% and 40% I would need ~105 and ~66 cases equally represented for each breed (Power and Sample Size Program v3.0.17, Vanderbilt University, http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/PowerSampleSize?CGISESSID=14c47 115b869aea72315e87dd308ae59). As we only arrayed 23 cases of dog OS on the ~27Kb- resolution oligo-array we again did not have the appropriate number of cases for each breed to detect a difference in CNA frequencies. Therefore, an increase in the number of individuals representing each breed would add more statistical power to define if associations between specific CNAs and a breed’s genetic framework exist or if the chaotic genomic imbalance in OS is only an artifact of the disease itself.

Integration of genome-wide CNA analyses with genomic analyses at the transcriptional level would also increase our understanding of disease pathogenesis. Few studies (Paoloni et al. 2009; Selvarajah et al. 2009) have conducted genome-wide expression profiling of canine OS, and the combination of these data with the genomic instability that I have found would provide a powerful source upon which to define the key genetic players in canine OS. Ideally, evaluation of genome-wide CNAs and expression changes in the same dog OS cases would provide the most accurate data. A recent study of human OS demonstrated an integrated molecular profiling approach combining analysis of genome- wide copy number changes, expression changes, and methylation patterns (Sadikovic et al. 2009) for five cases of human OS. Although they only studied five cases, Sadikovic et al. (2009) were able to identify interactive networks of biological regulation associated with human OS oncogenesis. As molecular technologies continue to advance in future years for both dogs and humans, limitations of this type of analysis that currently exist will be less prominent. The study of multiple areas of molecular regulation in the same dog patient will help to separate driver genetic abnormalities from passenger abnormalities that do not have an effect on OS etiology and are only a consequence of the genomic instability in the cancer.

149 In today’s society companion animals are a large part of American families. Therefore, the desire to provide better health care to our pet ‘family members’ has increased, causing the need for more thorough research of genetic aberrations in complex animal diseases like OS. The combination of my research on dog OS with past, present, and future research should shed more light on abnormalities in dogs while having comparative value to human OS patients. The discoveries in this research put us one step closer to identifying the biological mechanisms behind both human and dog OS and to finding ways to treat the disease, easing patient pain and suffering.

150 References

Bucan, M. and T. Abel (2002). "The mouse: genetics meets behaviour." Nat Rev Genet 3(2): 114-23. de Nigris, F., C. Botti, R. Rossiello, E. Crimi, V. Sica and C. Napoli (2007). "Cooperation between Myc and YY1 provides novel silencing transcriptional targets of alpha3beta1-integrin in tumour cells." Oncogene 26(3): 382-94.

Forus, A., D. O. Weghuis, D. Smeets, O. Fodstad, O. Myklebost and A. Geurts van Kessel (1995). "Comparative genomic hybridization analysis of human sarcomas: II.

151 Identification of novel amplicons at 6p and 17p in osteosarcomas." Genes Chromosomes Cancer 14(1): 15-21.

Inoki, K. and K. L. Guan (2009). "Tuberous sclerosis complex, implication from a rare genetic disease to common cancer treatment." Hum Mol Genet 18(R1): R94-100.

Karlsson, E. K. and K. Lindblad-Toh (2008). "Leader of the pack: gene mapping in dogs and other model organisms." Nat Rev Genet 9(9): 713-25.

Kirpensteijn, J., M. Kik, E. Teske and G. R. Rutteman (2008). "TP53 gene mutations in canine osteosarcoma." Vet Surg 37(5): 454-60.

152 Slezak, N. Doggett, J. F. Cheng, A. Olsen, S. Lucas, C. Elkin, E. Uberbacher, M. Frazier, R. A. Gibbs, D. M. Muzny, S. E. Scherer, J. B. Bouck, E. J. Sodergren, K. C. Worley, C. M. Rives, J. H. Gorrell, M. L. Metzker, S. L. Naylor, R. S. Kucherlapati, D. L. Nelson, G. M. Weinstock, Y. Sakaki, A. Fujiyama, M. Hattori, T. Yada, A. Toyoda, T. Itoh, C. Kawagoe, H. Watanabe, Y. Totoki, T. Taylor, J. Weissenbach, R. Heilig, W. Saurin, F. Artiguenave, P. Brottier, T. Bruls, E. Pelletier, C. Robert, P. Wincker, D. R. Smith, L. Doucette-Stamm, M. Rubenfield, K. Weinstock, H. M. Lee, J. Dubois, A. Rosenthal, M. Platzer, G. Nyakatura, S. Taudien, A. Rump, H. Yang, J. Yu, J. Wang, G. Huang, J. Gu, L. Hood, L. Rowen, A. Madan, S. Qin, R. W. Davis, N. A. Federspiel, A. P. Abola, M. J. Proctor, R. M. Myers, J. Schmutz, M. Dickson, J. Grimwood, D. R. Cox, M. V. Olson, R. Kaul, N. Shimizu, K. Kawasaki, S. Minoshima, G. A. Evans, M. Athanasiou, R. Schultz, B. A. Roe, F. Chen, H. Pan, J. Ramser, H. Lehrach, R. Reinhardt, W. R. McCombie, M. de la Bastide, N. Dedhia, H. Blocker, K. Hornischer, G. Nordsiek, R. Agarwala, L. Aravind, J. A. Bailey, A. Bateman, S. Batzoglou, E. Birney, P. Bork, D. G. Brown, C. B. Burge, L. Cerutti, H. C. Chen, D. Church, M. Clamp, R. R. Copley, T. Doerks, S. R. Eddy, E. E. Eichler, T. S. Furey, J. Galagan, J. G. Gilbert, C. Harmon, Y. Hayashizaki, D. Haussler, H. Hermjakob, K. Hokamp, W. Jang, L. S. Johnson, T. A. Jones, S. Kasif, A. Kaspryzk, S. Kennedy, W. J. Kent, P. Kitts, E. V. Koonin, I. Korf, D. Kulp, D. Lancet, T. M. Lowe, A. McLysaght, T. Mikkelsen, J. V. Moran, N. Mulder, V. J. Pollara, C. P. Ponting, G. Schuler, J. Schultz, G. Slater, A. F. Smit, E. Stupka, J. Szustakowski, D. Thierry-Mieg, J. Thierry-Mieg, L. Wagner, J. Wallis, R. Wheeler, A. Williams, Y. I. Wolf, K. H. Wolfe, S. P. Yang, R. F. Yeh, F. Collins, M. S. Guyer, J. Peterson, A. Felsenfeld, K. A. Wetterstrand, A. Patrinos, M. J. Morgan, P. de Jong, J. J. Catanese, K. Osoegawa, H. Shizuya, S. Choi and Y. J. Chen (2001). "Initial sequencing and analysis of the human genome." Nature 409(6822): 860-921.

Levine, R. A. and M. A. Fleischli (2000). "Inactivation of p53 and retinoblastoma family pathways in canine osteosarcoma cell lines." Vet Pathol 37(1): 54-61.

Levine, R. A., T. Forest and C. Smith (2002). "Tumor suppressor PTEN is mutated in canine osteosarcoma cell lines and tumors." Vet Pathol 39(3): 372-8.

153 T. Bayul, M. Benamara, A. Berlin, D. Bessette, B. Blitshteyn, T. Bloom, J. Blye, L. Boguslavskiy, C. Bonnet, B. Boukhgalter, A. Brown, P. Cahill, N. Calixte, J. Camarata, Y. Cheshatsang, J. Chu, M. Citroen, A. Collymore, P. Cooke, T. Dawoe, R. Daza, K. Decktor, S. DeGray, N. Dhargay, K. Dooley, P. Dorje, K. Dorjee, L. Dorris, N. Duffey, A. Dupes, O. Egbiremolen, R. Elong, J. Falk, A. Farina, S. Faro, D. Ferguson, P. Ferreira, S. Fisher, M. FitzGerald, K. Foley, C. Foley, A. Franke, D. Friedrich, D. Gage, M. Garber, G. Gearin, G. Giannoukos, T. Goode, A. Goyette, J. Graham, E. Grandbois, K. Gyaltsen, N. Hafez, D. Hagopian, B. Hagos, J. Hall, C. Healy, R. Hegarty, T. Honan, A. Horn, N. Houde, L. Hughes, L. Hunnicutt, M. Husby, B. Jester, C. Jones, A. Kamat, B. Kanga, C. Kells, D. Khazanovich, A. C. Kieu, P. Kisner, M. Kumar, K. Lance, T. Landers, M. Lara, W. Lee, J. P. Leger, N. Lennon, L. Leuper, S. LeVine, J. Liu, X. Liu, Y. Lokyitsang, T. Lokyitsang, A. Lui, J. Macdonald, J. Major, R. Marabella, K. Maru, C. Matthews, S. McDonough, T. Mehta, J. Meldrim, A. Melnikov, L. Meneus, A. Mihalev, T. Mihova, K. Miller, R. Mittelman, V. Mlenga, L. Mulrain, G. Munson, A. Navidi, J. Naylor, T. Nguyen, N. Nguyen, C. Nguyen, R. Nicol, N. Norbu, C. Norbu, N. Novod, T. Nyima, P. Olandt, B. O'Neill, K. O'Neill, S. Osman, L. Oyono, C. Patti, D. Perrin, P. Phunkhang, F. Pierre, M. Priest, A. Rachupka, S. Raghuraman, R. Rameau, V. Ray, C. Raymond, F. Rege, C. Rise, J. Rogers, P. Rogov, J. Sahalie, S. Settipalli, T. Sharpe, T. Shea, M. Sheehan, N. Sherpa, J. Shi, D. Shih, J. Sloan, C. Smith, T. Sparrow, J. Stalker, N. Stange-Thomann, S. Stavropoulos, C. Stone, S. Stone, S. Sykes, P. Tchuinga, P. Tenzing, S. Tesfaye, D. Thoulutsang, Y. Thoulutsang, K. Topham, I. Topping, T. Tsamla, H. Vassiliev, V. Venkataraman, A. Vo, T. Wangchuk, T. Wangdi, M. Weiand, J. Wilkinson, A. Wilson, S. Yadav, S. Yang, X. Yang, G. Young, Q. Yu, J. Zainoun, L. Zembek, A. Zimmer and E. S. Lander (2005). "Genome sequence, comparative analysis and haplotype structure of the domestic dog." Nature 438(7069): 803-19.

Mirabello, L., R. J. Troisi and S. A. Savage (2009a). "International osteosarcoma incidence patterns in children and adolescents, middle ages and elderly persons." Int J Cancer 125(1): 229-34.

Paigen, K. (1995). "A miracle enough: the power of mice." Nat Med 1(3): 215-20.

154 Paoloni, M. and C. Khanna (2008). "Translation of new cancer treatments from pet dogs to humans." Nat Rev Cancer 8(2): 147-56.

Sandberg, A. A. and J. A. Bridge (2003). "Updates on the cytogenetics and molecular genetics of bone and soft tissue tumors: osteosarcoma and related tumors." Cancer Genet Cytogenet 145(1): 1-30.

Selvarajah, G. T., J. Kirpensteijn, M. E. van Wolferen, N. A. Rao, H. Fieten and J. A. Mol (2009). "Gene expression profiling of canine osteosarcoma reveals genes associated with short and long survival times." Mol Cancer 8: 72.

Shearin, A. L. and E. A. Ostrander (2010). "Leading the way: canine models of genomics and disease." Dis Model Mech 3(1-2): 27-34.

155 Sutter, N. B. and E. A. Ostrander (2004). "Dog star rising: the canine genetic system." Nat Rev Genet 5(12): 900-10.

Thayanithy, V., A. L. Sarver, R. V. Kartha, C. Park, M. Scott, A. Y. Angstadt, M. Breen, C. J. Steer, J. F. Modiano and S. Subramanian (submitted). "Perturbation in the 14q32 miRNAs-cMYC-miR-17~92 gene network contributes to osteosarcoma and is associated with survival outcome." Cancer Cell.

Withrow, S. J., B. E. Powers, R. C. Straw and R. M. Wilkins (1991). "Comparative aspects of osteosarcoma. Dog versus man." Clin Orthop Relat Res(270): 159-68.

Withrow, S. J. and D. Vail, Eds. (2007). Small Animal Clinical Oncology. Philadelphia, WB Sanders.

156 Zielenska, M., J. Bayani, A. Pandita, S. Toledo, P. Marrano, J. Andrade, A. Petrilli, P. Thorner, P. Sorensen and J. A. Squire (2001). "Comparative genomic hybridization analysis identifies gains of 1p35 approximately p36 and chromosome 19 in osteosarcoma." Cancer Genet Cytogenet 130(1): 14-21.

157 APPENDICES

158 Appendix I

Tackling the characterization of canine chromosomal breakpoints with an integrated in-situ/in-silco approach: The canine PAR and PAB

Young, Andrea C1, Kirkness, Ewen F2 and Breen, Matthew1,3

1Dept. of Molecular Biomedical Sciences, College of Veterinary Medicine, North Carolina State University, Raleigh, NC 27606

2The J. Craig Venter Institute, 9712 Medical Center Drive, Rockville, MD, USA 20850

3Center for Comparative Medicine and Translational Research, NCSU, Raleigh, NC, USA

Accepted for publication in Chromosome Research (2008) Vol 16(8): 1193-1202.

159

160

161

162

163

164

165

166

167

168

169

Appendix II

Visualization of copy number changes in canine OS by FISH

170

Figure 1: (A) Whole genome 1Mb aCGH profile of an 8 year old male Great Pyrenees diagnosed with fibroblastic osteosarcoma located in the right proximal tibia co-hybridized with DNA derived from a blood sample from the same patient (OS20). The data profile is plotted as the log2 tumor DNA: reference DNA ratio after median block normalization and background subtraction of the replicate spots for each locus on the array. Points colored in grey represent loci that are a normal copy number (n=2). In contrast, loci that are gained (n > 2) in the patient appear in red and loci that are lost (n=1 and n=0) appear in green and blue, respectively. Selected BAC clones for FISH are colored to represent the fluorochrome with which each one was labeled and arrows point to their corresponding position on the aCGH profile. The clone 330D16 (red signal) was used as a control for the reaction since it was not indicated as aberrant by aCGH (n=2). (B) Interphase FISH analysis of the OS case used in A. i-v) An interphase nucleus representative of abnormal copy number for BAC clones 40GO9 (n=13), 328M18 (n=10), 335J10 (n=1), and 126B16 (n=6). The control BAC clone, 330D16 had normal copy number (n=2). vi) Composite image of all probes in aberrant interphase nucleus. (C) Metaphase FISH analysis of the OS case used in A. i) metaphase spread from the patient demonstrating structural aberration of BAC clones. ii) larger view of the tandem duplication of 40G09 and 328M18. (D) Compilation of copy number data, based on FISH analysis. The y-axis demonstrates the percentages of nuclei for a certain copy number status for each clone and the x-axis represents the BAC clone address, chromosome, position, and log2 ratio value from the corresponding aCGH analysis.

171

Figure 2:(A) Whole genome 1Mb aCGH profile of a 12 year old female Golden Retriever diagnosed with osteoblastic OS located in the left proximal humerus, co-hybridized with DNA derived from a blood sample from the same patient. The data profile is plotted as the log2 tumor DNA: reference DNA ratio after median block normalization and background subtraction of the replicate spots for each locus on the array. Points colored in grey represent loci that were indicated as having normal copy number (n=2), while loci colored red and green represent those that had an increase or decrease, respectively, in copy number. Selected BAC clones for FISH are colored to represent the fluorochrome with which each one was labeled and arrows point to their corresponding position on the aCGH profile. (B) i, ii) Metaphase FISH analysis of the OS case used in A demonstrating structural aberrations of 98B16 (KIT oncogene) on CFA 13 (inset) along with two copies of 323C15 (CFA 16 lost by aCGH but adjacent to clones with n=2) (C) Interphase FISH analysis of the OS case used in A. i) An interphase nucleus from the patient that exhibited normal copy number for both BAC clones (n=2) ii, iii) Abnormal interphase nucleus from the patient presenting with two copies of 323C15 and nine and seven copies of KIT. iv) Compilation of copy number data, based on FISH analysis. The y-axis demonstrates the percentages of nuclei for a certain copy number status for each clone and the x-axis represents the BAC clone address, chromosome, position, and log2 ratio value from the corresponding aCGH analysis.

172

Figure 3: Interphase multicolor FISH analysis of three zinc fixed paraffin embedded canine OS tissues using BAC clone probe sets encompassing three genes (TSC2, MYC, and TUSC3). TSC2 is represented by the orange signal, MYC by the green signal, and TUSC3 by the pink signal. In each graph the x-axis represents the gene and the y-axis demonstrates the percentages of nuclei for the stated copy number of each clone. A, B, and C show FISH analysis of three cases; A=5 yr old male Great Pyrenees with chondroblastic OS (OS8), B=8 yr old male Golden Retriever with OS (OS11), C=8 yr old male Rottweiler with osteoblastic OS (OS12). For each case data are shown as: i) interphase nucleus of a cell from each case presenting with a normal copy number (n=2) of each of the three genes being assessed; ii, iii) abnormal interphase nucleus from each patient demonstrating the presence of an aberrant copy number (n≠2) for one or more of TSC2, MYC, and TUSC3. iv) Compilation of copy number data of these three genes for each case, based on FISH analysis.

173