UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The molecular characterization of the progression of oral squamous cell carcinoma Towle, Rebecca 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_november_towle_rebecca.pdf [ 6.47MB ]
Metadata
JSON: 24-1.0313412.json
JSON-LD: 24-1.0313412-ld.json
RDF/XML (Pretty): 24-1.0313412-rdf.xml
RDF/JSON: 24-1.0313412-rdf.json
Turtle: 24-1.0313412-turtle.txt
N-Triples: 24-1.0313412-rdf-ntriples.txt
Original Record: 24-1.0313412-source.json
Full Text
24-1.0313412-fulltext.txt
Citation
24-1.0313412.ris

Full Text

  THE MOLECULAR CHARACTERIZATION OF THE PROGRESSION OF ORAL SQUAMOUS CELL CARCINOMA by  Rebecca Towle  B.Sc., The University of British Columbia, 2012  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Interdisciplinary Oncology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2016  © Rebecca Towle, 2016    - ii - Abstract Oral squamous cell carcinoma (OSCC) is the most common subtype of head and neck cancer and has a relatively low five year survival rate of ~50%. One of the reasons for this high mortality rate is that patients are generally diagnosed at late stages.  OSCC develops through a typical histological progression and although lesions in the oral cavity are visible at the premalignant stage, it is not possible to predict which lesions will progress based on histology alone.  In-depth analysis of genome-wide molecular alterations may identify novel genes or pathways that can be used as biomarkers or therapeutic targets in order to improve survival rates of this disease.   In this thesis, I perform DNA methylation, gene expression and miRNA profiling on a panel of patient tissue samples, each with a paired adjacent normal, dysplasia and either a carcinoma in situ or squamous cell carcinoma, taken from a single contiguous disease field within a patient’s oral cavity.  My hypotheses are that the epigenetic landscape of OSCC becomes progressively more deregulated throughout the different histological stages and that the most frequently altered molecular events identified at the dysplasia stage may be crucial for premalignant disease development and progression. A high level of deregulation in both methylation and miRNA patterns as the disease progresses is observed, and a number of highly frequent molecular events are identified.  Several of these molecular events are then functionally validated to assess the ability to contribute to tumorigenesis in oral premalignant lesions. Taken together, this thesis provides one of the most comprehensive epigenetic analyses of paired normal, dysplasia and CIS/SCC biopsies with regards to DNA methylation and miRNA profiling.  In addition to providing a deeper insight into the molecular mechanisms at play within the premalignant lesions, we also validate the ability of these mechanisms to directly contribute to tumorigenesis.       - iii - Preface The research conducted in this thesis has ethical approval through the University of British Columbia Research Ethics Board (Certificate number: H10-01694).  Chapters 3, 4, 6 and 7 were co-authored for publication.  The full author lists are as follows:  A version of Chapter 3 has been published as: Towle, R.,  Truong, D., Hogg, K., Robinson, W. P., Poh, C. F., & Garnis, C. (2013). Global analysis of DNA methylation changes during progression of oral cancer. Oral Oncology, 49(11), 1033-1042.  I am first author of this work.  I wrote the manuscript, conducted experiments, performed data analysis and interpreted the results.  A version of Chapter 4 has been published as: Towle, R.* , Gorenchtein, M*., Dickman, C., Zhu, Y., Poh, C. F. & Garnis, C (2014). Dysregulation of microRNAs across oral squamous cell carcinoma fields in non-smokers. JBR Journal of Interdisciplinary Medicine and Dental Science.  I am the co-first author of this work and wrote the manuscript, conducted data analysis and interpreted the results.  A version of Chapter 6 has been accepted for publication: Towle, R.,  Truong, D. & Garnis, C.(2016). Epigenetic mediated silencing of EYA4 contributes to tumorigenesis in oral dysplastic cells. Genes, Chromosomes and Cancer, 55(7), 568-576. I am first author of this work and wrote the manuscript, conducted experiments, performed data analysis and interpreted the results.  A version of Chapter 7 has been published as: Towle, R*,I am first author of this work and wrote the manuscript, helped conduct experiments and data analysis, and interpreted the results  Tsui, I. F., Zhu, Y., MacLellan, S., Poh, C. F., & Garnis, C. (2014). Recurring DNA copy number gain at chromosome 9p13 plays a role in the activation of multiple candidate oncogenes in progressing oral premalignant lesions. Cancer Medicine, 3(5), 1170-1184.   - iv - Table of Contents Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables ..................................................................................................................................x List of Figures ............................................................................................................................... xi List of Abbreviations ................................................................................................................. xiii Acknowledgements .................................................................................................................... xiv Dedication .....................................................................................................................................xv Chapter 1: Introduction ................................................................................................................1 1.1 Introduction to Oral Cancer ............................................................................................ 1 1.2 Oral Cancer Etiological Factors ...................................................................................... 3 1.3 Oral Cancer Progression ................................................................................................. 4 1.4 Field Cancerization ......................................................................................................... 7 1.5 Molecular Pathology of Oral Cancer .............................................................................. 8 1.6 Thesis Theme and Rationale ........................................................................................... 9 1.7 Research Question, Objectives and Hypotheses ........................................................... 10 1.8 Specific Aims and Thesis Outline................................................................................. 11 Chapter 2: Sample Collection and Analysis ..............................................................................13 2.1 Sample Accrual ............................................................................................................. 13 2.2 Nucleic Acid Extraction ................................................................................................ 15 2.3 DNA Methylation Microarray ...................................................................................... 15 2.4 Gene Expression Microarray ........................................................................................ 16   - v - 2.5 miRNA Analysis ........................................................................................................... 17 Chapter 3: DNA Methylation throughout the Progression of Oral Squamous Cell Carcinoma ....................................................................................................................................18 3.1 Introduction ................................................................................................................... 18 3.2 Methods......................................................................................................................... 20 3.2.1 Sample Cohort .......................................................................................................... 20 3.2.2 Sample Preparation and DNA Methylation Profiling Experiments .......................... 20 3.2.3 Gene Expression Profiling ........................................................................................ 20 3.2.4 Statistical Analyses ................................................................................................... 23 3.2.5 Bisulfite Pyrosequencing .......................................................................................... 23 3.2.6 TCGA Illumina 450K Microarray Data Analysis ..................................................... 24 3.3 Results and Discussion ................................................................................................. 24 3.3.1 Global DNA Methylation Landscape of Oral Tissues .............................................. 25 3.3.2 DNA Methylation Changes Throughout Oral Cancer Progression .......................... 26 3.3.3 Status of Previously Implicated Candidate Genes .................................................... 32 3.3.4 Frequently Altered Gene Candidates in Dysplasia ................................................... 33 3.3.5 Frequently Altered Gene Candidates in CIS/OSCC ................................................. 36 3.3.6 Epigenetically Driven Pathway Disruption in Oral CIS/OSCC ................................ 37 3.3.7 Analysis of Candidate Methylation Events in TCGA Datasets ................................ 43 3.4 Conclusions ................................................................................................................... 44 Chapter 4: Deregulation of miRNAs throughout the Progression of OSCC in Never-smokers .........................................................................................................................................46 4.1 Introduction ................................................................................................................... 46   - vi - 4.2 Materials and Methods .................................................................................................. 47 4.2.1 Sample Accrual ......................................................................................................... 47 4.2.2 RNA Isolation and miRNA Expression Quantification ............................................ 48 4.2.3 Quantitative RT-PCR of Candidate Genes ............................................................... 48 4.2.4 Development of Tissue Microarrays (TMAs) for Verification and Validation of Candidate miRNA Deregulation ........................................................................................... 49 4.2.5 in situ Hybridization (ISH) ....................................................................................... 49 4.3 Results ........................................................................................................................... 50 4.3.1 miRNA Expression in Oral Cancer Progression ....................................................... 50 4.3.2 miRNA Dysregulation Throughout Oral Cancer Progression .................................. 51 4.3.3 Candidate miRNAs Contributing to Oral Dysplasias ............................................... 51 4.3.4 miRNAs Dysregulated in More Advanced Disease.................................................. 56 4.3.5 in situ Hybridization Analysis .................................................................................. 58 4.3.6 Validation of Highly Frequent Deregulated miRNAs in the Tumor Stage in the TCGA Data Set of Oral Cavity Tumors ............................................................................... 58 4.4 Discussion ..................................................................................................................... 61 Chapter 5: Integrative Analysis of miRNA, Methylation and Gene Expression in Five Patients ..........................................................................................................................................63 5.1 Introduction ................................................................................................................... 63 5.2 Integrative Analysis of miRNA, Methylation and Expression Data Sets ..................... 65 5.2.1 Study Design ............................................................................................................. 65 5.2.2 Statistical Analysis .................................................................................................... 67 5.3 Results ........................................................................................................................... 67   - vii - 5.3.1 Highly Inversely Correlated Events .......................................................................... 67 5.3.2 Correlated Hypermethylated Genes are More Prevalent at the Dysplasia Stage ...... 68 5.3.3 Frequently Deregulated Genes among 5 Patient Samples ........................................ 68 5.3.4 Gene Ontology Analysis in Dysplasia and Tumor ................................................... 70 5.3.5 Pathway Analysis Shows Different Mechanisms of Perturbation in Patients .......... 73 5.4 Discussion ..................................................................................................................... 78 Chapter 6: Methylation Mediated Silencing of EYA4 in Oral Dysplasia ...............................80 6.1 Introduction ................................................................................................................... 80 6.2 Materials and Methods .................................................................................................. 81 6.2.1 Integrative Analysis of DNA Methylation and Gene Expression Microarray Data . 81 6.2.2 TCGA Data Analysis ................................................................................................ 82 6.2.3 Cell Lines .................................................................................................................. 82 6.2.4 qPCR ......................................................................................................................... 83 6.2.5 Lentiviral Transduction ............................................................................................. 83 6.3 MTT Proliferation Assay .............................................................................................. 83 6.3.1 Colony Formation Assay .......................................................................................... 84 6.3.2 Annexin V Apoptosis Assay ..................................................................................... 84 6.3.3 g-H2AX Analysis...................................................................................................... 84 6.3.4 Comet Assays............................................................................................................ 85 6.4 Results ........................................................................................................................... 85 6.4.1 EYA4 Hypermethylation and Gene Silencing is Observed in Dysplasia Samples of Oral Cancer Patients ............................................................................................................. 85 6.4.2 EYA4 Methylation is Correlated with Reduced EYA4 expression ........................... 88   - viii - 6.4.3 EYA4 is Silenced in Oral Dysplasia Cell Lines ........................................................ 89 6.4.4 Over-expression of EYA4 Impacts Cellular Proliferation, Apoptosis and DNA Damage Repair in Oral Dysplasia Cell Lines ....................................................................... 89 6.5 Discussion ..................................................................................................................... 95 Chapter 7: Focal Amplification of 9p13 in Oral Premalignant Lesions .................................98 7.1 Introduction ................................................................................................................... 98 7.2 Materials and Methods .................................................................................................. 99 7.2.1 Whole Genome Characterization of DNA Copy Number Alterations in OPLs ....... 99 7.2.2 Gene Expression Profiling Analysis ......................................................................... 99 7.2.3 Accrual of Fresh Oral Tumor Tissues for DNA Copy Number and Gene Expression Analyses ................................................................................................................................ 99 7.2.4 Statistical Analysis of Genomic Profiles ................................................................ 100 7.2.5 Validation of Gene Candidates in Tissue Microarrays (TMA) .............................. 100 7.2.6 Cell Culture and Reagents ...................................................................................... 101 7.2.7 shRNA Lentiviral Vector Knock-down .................................................................. 101 7.2.8 Plasmid Construction and Viral Transduction ........................................................ 102 7.2.9 Real-time Polymerase Chain Reaction (PCR) of mRNA Expression .................... 102 7.2.10 Western Blotting ................................................................................................. 103 7.2.11 MTT Cell Viability Assay .................................................................................. 104 7.2.12 Soft Agar Colony Formation Assay .................................................................... 104 7.3 Results ......................................................................................................................... 105 7.3.1 Identification of Recurring DNA Gain at 9p13.3 in Progressing OPLs ................. 105   - ix - 7.3.2 Inhibition of Gene Candidates Within the 9p13.3 Amplicon Diminishes Oral Cancer Phenotypes .......................................................................................................................... 107 7.3.3 Over-expression of Candidate Genes Enhances Cell Proliferation and Anchorage-Independent Growth............................................................................................................ 109 7.3.4 9p13 Amplification Within Multiple Biopsies From a Single Patient .................... 113 7.3.5 Frequency of 9p13 Gain in Various Cancers .......................................................... 114 7.4 Discussion ................................................................................................................... 114 7.5 Conclusion .................................................................................................................. 120 Chapter 8: Discussion and Conclusions ...................................................................................121 8.1 Summary of Findings .................................................................................................. 121 8.1.1 Epigenetic Analysis of the Different Histological Stages of OSCC ....................... 121 8.1.2 Assessment of Frequent Molecular Events as Being Implicated in Progression .... 122 8.2 Conclusions Regarding Hypotheses ........................................................................... 123 8.3 Strengths and Limitations ........................................................................................... 124 8.4 Overall Significance of this Work .............................................................................. 126 8.5 Future Directions ........................................................................................................ 128 Bibliography ...............................................................................................................................130 Appendix A ............................................................................................................................. 144 A.1 Chapter 4 Validation of miRNA Expression in Individual qPCR Assays .............. 144 A.2 Chapter 7 Tissue Microarray Patient Demographic Information ........................... 144   - x - List of Tables Table 1.1 Summary of TMN Staging in Oral Cancer ..................................................................... 2 Table 2.1 Patient Demographic Table .......................................................................................... 14 Table 3.1 Demographics of Methylation Patients......................................................................... 22 Table 3.2 Differentially Methylated Probes.................................................................................. 30 Table 3.3 Methylation Deregulation of Known Gene Candidates ................................................ 33 Table 3.4 Identified Candidate Genes in Dysplasia ...................................................................... 34 Table 3.5 Frequency of WNT Pathway Candidate Genes ............................................................ 40 Table 3.6 Frequency of MAPK Pathway Candidate Genes .......................................................... 40 Table 3.7 Validation of Frequently Methylated Genes in TCGA Methylation Data.................... 43 Table 4.1 Patient Demographic Information ................................................................................ 48 Table 4.2 Highly Frequent Deregulated miRNAs ........................................................................ 54 Table 4.3 Validation of Frequently Downregulated miRNAs in the TCGA Cohort .................... 59 Table 4.4 Validation of Frequently Upregulated miRNAs in the TCGA Cohort ......................... 60 Table 5.1 Data Analysis Grouping Scheme .................................................................................. 64 Table 5.2 Genes Enriched in “Pathways in Cancer Signaling” in Dysplasia Samples ................. 75 Table 5.3 Genes Enriched in Focal Adhesion Signaling in CIS/SCC Samples ............................ 76 Table 6.1 Overview of EYA4 methylation and expression patterns in BCCRC Samples ........... 87 Table 6.2 GSEA Analysis Results for Signaling Pathways Enriched in EYA4 Low Samples .... 92    - xi - List of Figures Figure 1.1 Oral Cancer Progression from Normal Epithlia to Squamous Cell Carcinoma ............ 5 Figure 3.1 Obtaining multiple biopsies from an oral cancer field ................................................ 21 Figure 3.2 Mean proportion of low or unmethylated and high methylated CpGs ........................ 27 Figure 3.3 Methylation patterns of a sampls set of 30 biopsies using hiarchical clustering ......... 28 Figure 3.4 Methylation patterns in oral dysplasia and CIS/SCC .................................................. 29 Figure 3.5 General trends in DNA methylation patterns during progression ............................... 31 Figure 3.6 Signaling network of candidate genes in WNT and MAPK pathways ....................... 38 Figure 3.7 Correlation between % methylation using microarray and pyrosequencing ............... 41 Figure 3.8 DNA methylation status of MAPK314, TNF and RASGFR1 in validation cohort .... 42 Figure 4.1 Clinical characterization and corresponding histology within an oral cancer field .... 52 Figure 4.2 Number of miRNAs Deregulated Throughout Progression ........................................ 53 Figure 4.3 Trends in Deregulated miRNA .................................................................................... 57 Figure 5.1 Flow Chart of Approach to Integrative Analysis in 5 Patient Samples ....................... 66 Figure 5.2 Proportion of Significant Genes in Each Patient  ........................................................ 69 Figure 5.3 Number of Genes Deregulated in at least 2/5 Patients ................................................ 71 Figure 5.4 Representative Categories of Biological Process Gene Ontology Terms ................... 72 Figure 5.5 Genes Deregulated in Dysplasia Samples in "Pathways in Cancer" ........................... 74 Figure 5.6 Genes Deregulated in CIS/SCC Samples in "Focal Adhesion" Pahway ..................... 77 Figure 6.1 Mean Beta Value for EYA4 Promoter Region in BCCRC Samples ........................... 88 Figure 6.2 Analysis of TCGA Oral Cavity Patient Samples for EYA4 Methylation ................... 91 Figure 6.3 Proliferation and Apotposis Results for POE9n-tert and DOK ................................... 94 Figure 6.4 Assessment of DNA Damage Repair .......................................................................... 95   - xii - Figure 7.1 Detection of 9p13 gain using FISH Analysis ............................................................ 107 Figure 7.2 Alignment of Genomic Alteration Data of Five Low-Grade OPLs .......................... 109 Figure 7.3 Knockdown of Candidate Genes in Tumor Cell Line ............................................... 111 Figure 7.4 Overexpression of Candidate Genes in Tumor Cell Line ......................................... 112 Figure 7.5 Overexpression of Candidate Genes in Dysplasia Cell Line ..................................... 113 Figure 7.6 DNA Gain at 9p13 and Corresponding Increase in mRNA  of Genes in One Patient .............................................................................................................. 116       - xiii - List of Abbreviations CGH comparative genomic hybridization CIS carcinoma in situ COOLs Canadian Optically Guided Approach for Oral Lesions Surgical Trial CpG cytosine guanine dinucleotide DCTN3 dynactin 3 DNA deoxyribonucleic acid EYA4 EYA transcriptional co-activator and phosphatase 4 FV fluorescence visualization GO gene ontology GSEA gene set enrichment analysis HPV human papillomavirus ISH in situ hybridization KEGG Kyoto Encyclopedia of Genes and Genomes LNA locked nucleic acid LOH loss of heterozygosity MAPK mitogen-activated protein kinase miRNA/miR microRNA MTT 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide OPL oral premalignant lesion OSCC oral squamous cell carcinoma qRT-PCR quantitative real time PCR RNA ribonucleic acid STOML2 stomatin like-2 TCGA The Cancer Genome Atlas TMA tissue microarray TRC The RNAi Consortium VCP valosin containing protein WHO World Health Organization WNT wingless-related integration site BCCRC British Columbia Cancer Research Centre ANOVA analysis of variance PCR  polymerase chain reaction         - xiv - Acknowledgements I would like to express my sincerest gratitude to all those who have supported me during my PhD. Without their support, this thesis would not have been possible.  I will be forever grateful for my supervisor, Dr. Cathie Garnis, for giving me the opportunity to work with her, and for her mentorship and support.  I could not have asked for a better supervisor. I would also like to thank my supervisory committee members, Dr. Catherine Poh, Dr. Wendy Robinson and Dr. Will Lockwood for their invaluable scientific critique and guidance throughout my PhD.   I would like to acknowledge the funding bodies that supported the work in this thesis: the Canadian Institute of Health Research, the Pacific Otolaryngology Foundation, and  the Rotary Hearing Foundation.  I would also like to acknowledge the following scholarship support: the Interdisciplinary Oncology Program Entrance Award and the Canadian Cancer Society.            I would also like to thank members of the Cathie Garnis lab, both past and present, who have contributed to and provided helpful discussion and insight to the work in this thesis.  I would also like to thank my colleagues in the Integrative Oncology Department at the British Columbia Cancer Research Centre who have provided advice and support.         - xv - Dedication To my family   - 1 - Chapter 1: Introduction 1.1 Introduction to Oral Cancer Oral squamous cell carcinoma (OSCC) is the most common form of head and neck cancers, and accounts for the deaths of tens of thousands of people worldwide each year. In 2012, ~300,000 people were diagnosed with oral cancer world-wide 1.  Although there have been advances in detection, treatment and screening for this disease, the survival rate has remained stagnant at ~50% over five years for the past several decades2.   Late stage diagnosis is one of the biggest factors contributing to the low survival rate of OSCC.  The stage at which the disease is diagnosed at is the biggest indicator of prognosis, as patients diagnosed at Stage I  have an 80% survival rate over five years, whereas the survival rate at Stage IV is a dismal 20% 3.  Table 1.1 has a summary of the different TMN stages of oral cancer.   As with many other types of cancer, treatment at the earliest stages, and particularly at the premalignant stage, would greatly increase the survival rate for this disease4.   The most common anatomical sites where OSCC can develop are the tongue, floor of mouth, maxillary/mandibular gingiva, buccal mucosa and hard palate with the majority of cases originating from the tongue2. Given the ease of access to the oral cavity, lesions are readily identifiable at early stages; however aggressive intervention is rarely performed as it is impossible to identify patients at risk for progression based on the current gold standard, the histology.  Doctors generally take a “watch-and-wait” strategy for these lesions - meaning the time for early intervention is lost for patients that will ultimately progress. Treatment strategies for OSCC patients largely depend on the extent of the disease, with surgical resection being the key component in both early and late stages of disease. Radiation and/or chemotherapy is used as adjuvant therapy if the tumor is particularly large, surgical margins are not     - 2 - Table 1.1Stage  Summary of TMN Staging in Oral Cancer TNM Classification Description I T1N0M0 Tumor 2cm or less; No nodal or distant metastasis II T2N0M0 Tumor >2cm but <4cm No nodal or distant metastasis III T3N0M0 T1,2 OR 3N1M0 Tumor >4cm or metastasis in a single ipsilateral lymph node (<3cm in dimension) No distant metastasis IV Any T4 Any N2 or N3 Any M1 Tumor invades adjacent structure; or Metastasis in lymph node >3cm  or Distant metastasis  clean or if there is evidence of invasion to the lymph nodes or other sites.   Surgical intervention can have a high level of morbidity, especially at later stages of disease, with major impacts on mastication, deglutition and speech.  Reconstructive surgery may be required after the initial removal of the tumor.     - 3 - 1.2 Oral Cancer Etiological Factors Oral cancer is a carcinogen driven disease and the major risk factors include tobacco and alcohol use, betel nut chewing and HPV infection.  The incidence of OSCC increases with age, with the majority of patients diagnosed >50 years of age5.  Traditionally, oral cancer patients have been older males who are both drinkers and smokers6.  However, there has been an increasing incidence of younger (<40 years old), never-smoker, non-drinker patients being diagnosed with tongue squamous cell carcinoma, particular in females without apparent risk factors3,7,8.   Smokers account for 25% and drinkers for 7-19% of cases world-wide, and are the predominant cause of OSCC in North America.  Other suggested risk factors for OSCC include poor diet, viral and fungal infections, and poor dental hygiene9-11.  However, more in depth analysis of these factors are required in order to fully confirm the role these risks have on OSCC development.  Cigarette smoke contains a high number of known carcinogens that can contribute to the development of oral cancer including nitrosamines, polycyclic aromatic hydrocarbons, and acetaldehyde12. These carcinogens can cause DNA damage which in turn can activate oncogenic or silence tumor suppressive pathways.  The specific mechanism for the role of alcohol in tumorigenesis is less clear, as pure ethanol itself is not carcinogenic.  There are several mechanisms which may be responsible for the increased risk due to alcohol consumption.  Firstly, acetaldehyde, a carcinogenic molecule, is formed during the metabolism of alcohol13.  The bulk of ethanol metabolism occurs within the liver however, high levels of acetaldehyde can be found in saliva after alcohol consumption, potentially due to ethanol metabolizing microbes in the oral cavity14.  Alcohol also causes inflammation and can act as a solvent (allowing other carcinogens such as those from tobacco to enter epithelia cells easily), potentially contributing to tumorigenesis via these mechanisms.  Although either smoking or excessive alcohol use alone is sufficient to increase the risk of developing oral cancer, the combination of these two risk factors appears to work synergistically for   - 4 - a much larger increase in risk, with almost a six fold increase in risk in patients who are both smokers and drinkers15,16.   Oral cancer is a global problem, and the major risk factors for a given demographic vary worldwide.  In North America, the largest risk factors are smoking and alcohol use as discussed above.  Conversely, in South Asian countries, the major risk factor for developing oral cancer (which is highly prevalent in these regions) is chronic betel quid nut use.   Despite being traditionally associated as a carcinogen driven disease, there is an increasing incidence world wide of young (<40 years of age) patients diagnosed with OSCC17,18.  These patients appear to lack etiological factors that could have caused their disease.  They are generally non-smokers, non-drinkers and exhibit good oral hygiene.  Oropharyngeal cancers have also seen a similar increase in incidence of diagnosis in younger non-smoker, non-drinking patients19.  This incidence has been attributed to increasing incidence of HPV infection, which account for 60 to 70% of OPC cases5.  Interestingly,  HPV related OSCC cases are rarer and only account for 5-10%20 of cases in the oral cavity, indicating the virus may not be responsible for this increase in oral cavity patients, however this remains controversial21.     1.3 Oral Cancer Progression OSCC, as with most types of cancer, is a disease driven by the accumulation of genetic and epigenetic abnormalities, and develops through a series of histological stages from hyperplasia to various degrees of dysplasia, to carcinoma in situ (CIS), and finally to invasive disease (Figure 1.1).  Oral lesions are classified clinically based on appearance as most commonly a leukoplakia or an erythroplakia.      - 5 -    Figure 1.1  Oral cancer progression from normal epithelia to squamous cell carcinoma                - 6 - Leukoplakia is characterized by the World Health Organization (WHO) as “a predominately white lesion of the oral mucosa that cannot be characterized as any other definable lesion”22.  These lesions are considered to be at an increased risk of progression compared to normal epithelia, but have only an approximately 1-2% risk of progression23.  Many leukoplakias are diagnosed as hyperplasia or dysplasia.  Erythroplakia is defined as “any lesion of the oral mucosa that presents as bright red velvety plaques which cannot be characterized clinically or pathologically as any other recognizable condition”24.  These lesions are at a much higher risk for progression than leukoplakia as histologically they are generally diagnosed as severe dysplasia or carcinoma in situ.  Transformation rates of erthryoplakia are much higher than leukoplakia and range from 14-50%25 The current gold standard for determining the risk of progression for a lesion is to biopsy and assess tissue histology, as a patient diagnosed with a mild dysplasia is less likely to progress compared with patient with a severe dysplasia or CIS.  The criteria for histological diagnosis for the various stages has been set by the WHO26.  However, diagnoses can be somewhat subjective and consequently, exact diagnoses can vary among individual pathologists.   The earliest histological stages a lesion may be diagnosed is hyperplasia.  At this stage, there are no abnormal cytological characteristics, but the tissue exhibits an abnormal increase in the proliferation of cells27.  Hyperplasia may arise in tissue that is chronically exposed to carcinogens such as alcohol or tobacco, or due to inflammation from chronic injury such as through chewing, improperly fitted dentures or broken teeth. Dysplasias are diagnosed based on cytological and architectural changes in the epithelia27. Dysplasias can be categorized as mild, moderate or severe and these grades fall on a spectrum rather than distinct categories. Due to this, categorization of dysplasias can be subjective and to the pathologist’s discretion. Generally, mild dysplasia exhibit minimal cytological changes and   - 7 - architectural disturbance is limited to the lower third of epithelia.  Moderate dysplasias show cytological changes and architectural disturbance is limited to the lower two thirds of epithelia.  Finally, severe dysplasia exhibits a high level of cytological aberrations, as well as architectural disturbances surpassing the last third of epithelia.  Accordingly, the more severe the dysplasia at diagnosis, the more likely the disease will progress. Patients diagnosed with a moderate or severe dysplasia can have almost twice the risk of progression than those with mild dysplasias27.  A lesion is diagnosed as CIS when architectural disturbance is apparent among the entire thickness of the epithelia and if there is a high degree of cellular atypia28. The major difference between a CIS and an invasive tumor is that the lesion has yet to break through the basement membrane of the epithelia.  At this stage, malignant transformation is apparent, however, invasion to surrounding tissue is not present, hence “cancer in one place”.  Once the lesion has broken through this membrane layer, the lesion is potentially invasive and consequently the histological diagnosis is an SCC.     1.4 Field Cancerization Cancers that develop as a result of exposure to carcinogens often exhibit a phenomenon known as field cancerization, in which histopathologically abnormal tissue extends beyond a neoplastic lesion’s visible boundaries29. The recurrent exposure to these carcinogens leads to genetic and epigenetic aberrations within a single cell or across a number of cells, and expansion of these clones leads to an accumulation of these aberrations and results in areas of differing histological characteristics within a single contiguous field. This concept was first proposed  in 1953 by Slaughter et al. based on the observation that benign epithelia surrounding malignant tissue appeared abnormal in oral cancer tissue29.  This phenomenon has also been observed to occur in a number of cancer types including lung, esophagus, breast, bladder and colorectum30,31.   - 8 -  Field cancerization was observed prior to the advent of molecular profiling, but current definitions of this concept now include a molecular component that illustrates the expanse of field cancerization beyond that which is histologically apparent in the tissue30.  It has been shown that even surrounding histologically normal tissue can exhibit early molecular aberrations32. The field effect in oral cancer is largely responsible for the high rates of recurrence and the development of multiple primary tumors.  A tumor is highly likely to recur if the extent of the malignant field is not adequately excised, leaving cells that can progress to invasive disease once again. Oral cancer also has one of the highest rates of multiple primary tumors of all cancers types33, as they can arise from independent abnormal fields that coalesce30.  1.5 Molecular Pathology of Oral Cancer OSCC is initiated and progresses through an accumulation of genetic and epigenetic aberrations. Advanced histological stages of oral cancer exhibit a high level of genomic instability and deregulation of molecular mechanisms that keep processes such as maintenance of DNA, methylation patterns and miRNA processing in check.  This molecular instability results in a high level of molecular abnormalities including the deletion or amplification of chromosomal regions, point mutations, promoter-methylation mediated silencing and miRNA deregulation34,35.  With such a high amount of molecular instability, it is difficult to differentiate between genes that become deregulated as a consequence of these factors from those that are actively driving tumorigenesis.  A number of driver genes have been identified in late stages of oral cancer.  Commonly reported oncogenes in OSCC often play a role in growth promotion, invasion or through evasion of apoptosis and cell cycle checkpoints. Commonly reported oncogenes include EGFR, KRAS, c-myc, PIK3CA, CCND1 and TGF-α35-37.  Tumor suppressor genes are also reported as being silenced, including p53, CDKN2A, E-cadherin, MGMT, DAPK1, PTEN, SMAD4 and RUNX335,36,38,39.    - 9 - Chromosomal alterations found at the invasive stages of OSCC include loss of 11q, 4q and in chromosome 8  40,41. Major signaling pathways that have been found to be deregulated in OSCC include the WNT signaling pathway, TGFbeta, and PI3k-PTEN-AKT42-44.  These pathways play a major role in cell proliferation, survival and differentiation. Studies assessing the molecular events in premalignant tissue are rarer as these samples are difficult to profile due the small size of the lesions.  However, several highly frequent early events in oral cancer development have been identified.  One of the most frequently reported early events in OSCC premalignancy is copy number loss at 9p2145. This was first identified by loss of heterozygosity at 9p and further documented as being lost in the majority of premalignant or early carcinogenic lesions46,47. Other regions exhibiting chromosomal loss at early stages of OSCC include regions on 3p and 17p35,48.  This genomic region contains CDKN2A, the gene encoding p16 and p14, which are cyclin-dependent kinase inhibitors and tumor suppressor genes for OSCC.  Loss of p53 also occurs early in tumorigenesis and was found in 35% of the histologically normal tissue surrounding dysplastic lesions32.   Deregulation of miRNAs also can contribute to tumorigenesis in OSCC by either an over-expressed miRNA that silenced a tumor suppressor gene or the decreased expression of a miRNA no longer suppressing the translation of an oncogenic mRNA. One of the most frequently reported deregulated miRNA is mir-2149.  This miRNA is often over-expressed in oral cancer and has been found to play a role in inhibiting apoptosis.  1.6 Thesis Theme and Rationale The overall theme for this thesis centers on characterizing the molecular aberrations that evolve throughout the progression of OSCC. Comparatively little is known about the premalignant stages of this disease, especially on a genome-wide level. Clinical intervention for cancer is well   - 10 - known to be more effective at earlier stages, and it is currently impossible to differentiate between lesions that will progress further from one that will not based on histology alone.  Thus, in depth analysis of the epigenetic background in the early stages of disease will provide a great insight into the molecular mechanisms defining the premalignant versus malignant stages.  The significance of these mechanisms with regards to progression can then be determined through long-term follow-up of patients with dysplasias who have not yet progressed into invasive disease.  Furthermore, assessing a variety of molecular mechanisms from the same set of tissue samples will aid in giving a clearer understanding of the genes and signal transduction pathways that are implicated in this disease.  This could provide new strategies therapeutically for targeting genes or pathways to increase survival from this disease.    1.7 Research Question, Objectives and Hypotheses The overarching research question for the work described in this thesis is: How does the epigenetic landscape of oral dysplasia change throughout the progression to OSCC?The hypotheses for this work are:   My overall objectives are to characterize the epigenetic events in invasive lesions and their paired premalignant precursors, and to identify novel genes or pathways that play an active role in the earliest stages of tumorigenesis in OSCC. 1. The epigenetic landscape of OSCC will become sequentially more deregulated as the disease progresses through the different histological stages 2.  The most frequently altered molecular events identified at the dysplasia stage may be crucial for premalignant disease development and progression    - 11 - 1.8 Specific Aims and Thesis Outline Chapter 2 is focused on the patient tissue sample set analyzed in Chapters 3-6 and the methodology that provides the groundwork for the bulk of the experiments and analysis within this thesis.  This chapter describes sample collection and processing, and provides a brief summary of the technologies used to analyze the tissue.   Aim 1: Profile sets of patient matched tissue of differing stages of progression for DNA methylation, gene expression and/or miRNA profiles Chapter 3-5 focus on set of 14 patient samples, each with paired sets of biopsies from differing histologies.  DNA methylation, miRNA profiling and mRNA profiling are performed on subsets of these samples. Chapter 3 focuses on the DNA methylation landscape on a global level, identifying trends in the different histological stages and commenting on frequently hypermethylated/hypomethylated genes.  Chapter 4 provides an analysis on the miRNA profiles of a similar set of samples as in Chapter 3, but taking in consideration the increasing incidence of non-smoking patients with OSCC, this sample set is comprised entirely of non-smoking patients.  Overall trends in the patterns of miRNA expression are discussed, and highly frequent miRNA events are identified. Aim 2: Assessment of the epigenetic events throughout the progression of OSCC The previous Aim focuses on DNA methylation and miRNA as separate entities in each respective chapter, and the goal of Chapter 4 is to take the data from the previous chapters and provide an integrative analysis for patients that have a complete set of data from all three platforms (DNA methylation, gene expression, miRNA).  Highly correlated genes are identified in paired DNA methylation/gene expression and miRNA/miRNA target expression using the data from the previous Aim 3. Integration of genome-wide data to gain in depth information in patient samples   - 12 - chapters and highly frequent events in the integrative samples are assessed by pathway and gene ontology (GO) analysis. Chapter 6 takes a deeper look at the candidate tumor suppressor gene, EYA4, one of the most frequently hypermethylated/silenced genes within the dysplasia stage identified in Chapter 3.  Assessment of this gene in The Cancer Genome Atlas (TCGA) publicly available oral cancer data set is performed and a functional analysis of the role of EYA4 in dysplastic cells is accomplished through in vitro tests in dysplasia oral cell lines.  Chapter 6 provides an analysis of the miRNA profiles throughout the different stages of OSCC, assessing frequently up or down-regulated miRNA. The previous Aims focus on specific genetic interactions and how the contribute to tumorigenesis, the goal of further our understanding of OSCC progression and development.  However, clinical utility of this information in the context of progression necessitates assessing if these alterations can themselves be biomarkers of progression.  In order to do this, oral premalignant lesions without a disease field must be assessed.  Through long-term follow-up, those lesions that will ultimately progress will be identified and molecular events associated with those lesions may be the ones crucial for progression. Aim 4: To identify the significance of recurrent molecular events at the dysplasia stage of disease Through previous work by the Garnis lab a recurrent DNA copy number gain of ~2.49 Mbp at chromosome 9p13 in oral premalignant lesions that later progressed to invasive lesions was identified.  In this aim, an assessment of the significance of this amplification is performed by identifying candidate oncogenes within this region. In vitro silencing and activation experiments are performed for each of these genes in oral cancer cell lines found that each gene is independently capable of up-regulating proliferation and anchorage-independent growth.    - 13 - Chapter 2: Sample Collection and Analysis 2.1 Sample Accrual The samples analyzed in Chapters 3-6 were collected as part of the Canadian Optically- Guided Approach for Oral Lesions (COOLs) Surgical Trial50.  This Phase III pan-Canadian clinical trial is investigating surgical removal of oral lesions with surgical margins determined with the aid of a fluorescence visualization (FV) device.  The goal of this trial is to reduce the rate of recurrence, as recurrent and secondary primary tumors can be caused by insufficient surgical margins. The increased margins allows for sampling of different histopathological stages of disease from within a single contiguous disease field.   The FV device is able to indicate regions of disease proximal to clinically evident oral tumors that are not apparent through standard white light51-53.   A blue/violet light (400-460 nm) is used to scan the oral mucosa. Normal tissues will re-emit this light as pale green, whereas abnormal tissues do not exhibit this auto-fluorescence and instead appear as dark patches. The entire area of abnormal tissue is excised from the patient’s oral cavity and punch biopsies are performed on various areas of the tissue.  The histology of these biopsies is reviewed by an oral pathologist in order to determine the stage of the lesion For each patient, histologically different biopsies, including adjacent normal tissue, dysplasia and either a CIS or squamous cell carcinoma were collected at the same time from a single, contiguous field in a patient’s mouth (See Figure 3.1).    All samples were collected under informed, written consent and collected as approved by the University of British Columbia - British Columbia Cancer Agency Research Ethics Board (ID#: H10-01694). The ability to obtain multiple biopsies from a single patient and to use each patient as their own control facilitates much more stringent analyses, since differences in miRNA expression, DNA methylation and gene expression arise due to tissue and timing variability, as well as underlying genetics and extrinsic factors such as smoking and diet, are eliminated.   - 14 - The patient cohort included 11 non-smokers, 2 former smokers and 1 current smoker.  The ages of the patients ranged from 31 to 82.  Detailed demographic information is listed in Table 2.1. Ten patients were profiled for DNA methylation, thirteen for gene expression and nine for miRNA.  Five patient samples had data profiling for all three molecular platforms. We were unable to perform all three platforms for each biopsy due an insufficient amount of material for some tissue samples. A detailed list of patients and which were profiled on each platform is also provided in Table 2.1. Table 2.1Sample ID  Patient Demographic Table Diagnosis Age Sex Site Smoking Habit Data 1878 SCC 48 M Left Tongue NS Me; E; Mi 1953 CIS 53 F Left Tongue NS Me; E; Mi 3002 SCC 40 F Left Lateral Tongue NS Me; E; Mi 5266 SCC 54 M Right Tongue NS Me; E; Mi 7057 SCC 52 F Right Lateral Tongue NS Me; E; Mi 1858 SCC 65 M Right Tongue FS Me; E 1896 CIS 61 M Right Tongue S Me; E 1979 CIS 31 F Right Front of Mouth NS Me; E 1999 CIS 50 M Right Tongue NS Me; E 7059 SCC 68 F Left Lateral Tongue FS Me; E 2000 SCC 58 F Right Tongue NS Mi; E 2076 SCC 82 F Left Tongue NS Mi; E 7071 CIS 71 M Left Ventral Tongue NS Mi; E 7058 SCC 43 F Right Tongue NS Mi Me: Illumina27K Methylation, E: Agilent Expression, Mi: Exiqon qRTPCR miRNA Panels   - 15 -  2.2 Nucleic Acid Extraction Fresh frozen tissue samples underwent pathological review to confirm pathology stages and were then microdissected to obtain high tissue purity by an oral cancer pathologist to ensure tissue specimens had >85% purity.   RNA and/or DNA were then extracted from the samples.  RNA extractions were performed using the TRIZol reagent and isopropanol precipitation as per the manufacturer’s instructions (Life Technologies).  Residual DNA was removed using the DNAse-1 kit (Ambion).  DNA was extracted using the standard phenol/chloroform procedures and ethanol precipitation.  Residual RNA was removed using RNAse-A treatment as per manufacturer’s instructions (Qiagen).  Sample quality was assessed by visualization using gel electrophoresis and quantified using a spectrometer. RNA samples were only profiled if they appeared to have minimal degradation (i.e. if the 28S rRNA and 18S rRNA bands remained above approximately 200 base pairs).  2.3 DNA Methylation Microarray Sodium bisulfite conversion of genomic DNA was performed for all accrued DNA samples using the EZ DNA Methylation Kit following manufacturer’s protocol (Zymo). Infinium HumanMethylation27 BeadChip arrays  – which evaluate the methylation status of 27,578 CpG dinucleotides (corresponding to 14,473 unique genes) – were performed as fee for service by the Centre for Applied Genomics (www.tcag.ca) using 1 μg of bisulfite-converted DNA from each sample. Methylation levels derived from resultant data were determined using GenomeStudio software (Illumina). We present the methylation status of each CpG site as a β-value, which is the ratio of the methylated probe intensity versus the overall intensity (sum of methylated and unmethylated probe intensities plus a constant α, where α = 100)54. β-values with a detection P-value   - 16 - > 0.05 were eliminated from analysis55. There was a range in the number of  probes where signal was not above background (failed probes) throughout the samples with a minimum of 6 rejected probes to a maximum of ~5000.  To account for the variability in the number of probes failed probes in each biopsy, we compared the proportion of altered probes to the total number of successful probes for each sample.  Changes in CpG methylation status for a given lesion were calculated at each probe by subtracting the β-value from the paired normal tissue from the β-value of the lesion (thus yielding a Δβ-value). A CpG was considered hypermethylated if Δβ ≥ 0.15 or hypomethylated if Δβ ≤ -0.15.  There is no general consensus on what Δβ constitutes a biologically change in DNA methylation as this is dependent on the data being analyzed56, and generally Δβ of 0.1-0.2 are commonly used in the literature as cut-offs for differential methylation.  We felt 0.15 would be appropriate in our dataset as this generally was in the top 90-95th  2.4 Gene Expression Microarray Total RNA was analyzed using the Agilent Whole Human Genome Microarray 4x44K.    The microarrays were scanned using the Axon GenePix 4000B scanner.  Array data was normalized using GenePix Pro 6.1.  Normalization processing was done by dividing the background subtracted array intensity value by the median array intensity of each microarray57.  The median array intensity was determined using the background-subtracted intensity value for all spots, except for the control type spots.  A gene was considered to be over-expressed if its expression increased by at least two fold compared to the adjacent normal samples.  Similarly, a gene was deemed under-expressed if the expression decreased by at least two fold from the adjacent normal sample.    - 17 - 2.5 miRNA Analysis For global miRNA expression profiling, 40 ng total RNA from each tissue sample was reverse transcribed using the miRCURY LNA Universal RT miRNA PCR, Polyadenylation and cDNA synthesis kit (Exiqon). cDNA was then analyzed by quantitative real-time PCR (qRT-PCR) on the miRNA Ready-to-Use PCR, Human panel I and panel II with the miRCURY LNA Universal RT miRNA PCR, SYBR Green master mix according to the manufacturer’s protocol (Exiqon). Briefly, a 10 µL mix comprised of 0.005 ng/µL cDNA, 2x SYBR Green master mix (Exiqon) and 50x ROX dye (Invitrogen) was distributed into individual wells across two separate 384-well panels. Collectively, these panels contain 742 human miRNAs. All assays were quantified on the ViiA™ 7 Real-Time PCR System (Applied Biosystems). Amplification occurred under the following conditions: 95°C for 10 min, followed by 40 cycles of 95°C for 10 s and 60°C for 60 s. At the end of the PCR cycles, melting curve analyses were performed. Only assays with distinct melting curves and CT ≤ 35 were included in analyses. All qRT-PCR data were normalized using global mean and analyzed using the comparative Ct method using the non-malignant sample as the comparator the. Data were then filtered based on a minimum 2-fold expression change relative to the paired non-malignant sample. Candidate miRNAs were selected based on high frequencies of dysregulation and their highest (for over-expressers) and lowest (for under-expressers) average fold change.   - 18 - Chapter 3: DNA Methylation throughout the Progression of Oral Squamous Cell Carcinoma  3.1  Introduction DNA methylation is an epigenetic mechanism that has been observed to become highly deregulated in cancer, and associated with  the silencing of tumor suppressor genes and genomic instability58.  This epigenetic modification occurs when a methyl group is covalently added to the 5’ carbon of a cytosine.  Within the human genome, this occurs almost exclusively in cytosines adjacent to guanines (referred to as CpG dinucleotides).  The function of DNA methylation in humans is generally to maintain the repression of non-expressed genes or non-coding regions of the genome (such as LINE-1 elements or retrotransposons).  Methylation can also occur within gene bodies and is generally positively correlated with gene expression, however the function of this mechanism has yet to be elucidated59,60.   DNA methylation is well documented as becoming deregulated in a variety of cancer types including oral cancer 58,61.   This deregulation generally occurs as hypermethylation (a CpG becoming methylated) or hypomethylation (a CpG losing a methyl group).  Hypermethylation generally occurs at site specific regions of the genome, often within the promoter region of a gene.  This has been correlated with the inactivation of tumor suppressor genes. Hypomethylation occurs on a global level within the genome, commonly occurring within regions such as LINE-1 or retrotransposons62,63.  The loss of methylation can allow for the activation of these elements, subsequently contributing to genomic instability64.  Furthermore, the loss of methylation at the promoter region of gene can result in the activation of expression of that gene and cause in the expression of oncogenes.  As the type of methylation microarray used to analyze the patient samples within this chapter is biased towards CpGs mapping to the promoter regions of genes, the focus of this chapter will be on promoter related DNA methylation.     - 19 - Very few studies have analyzed methylation changes in oral dysplasia – and those that have assessed such changes in these oral precancerous lesions (OPLs) have only evaluated one or a small number of genes previously implicated in invasive OSCC 65-68.While epigenetic changes are understood to be drivers of epithelial cancer progression and high resolution methods now exist for assessing such changes 69, to date there have been no reports evaluating genome-wide DNA methylation changes for oral dysplasias. Further, there have only been a limited number of genome-wide methylation studies investigating head and neck cancers in general70-73.   To our knowledge, this is the first report on the genome-wide methylation patterns in paired OSCC lesions of differing histological stages.  Analysis of aberrantly methylated genes in tumorigenesis is advantageous not only to further our knowledge of the overall disease process, but also as methylation shows great promise as a biomarker due to its overall stability and ability to be detected in saliva74,75. Promoter methylation has been used to identify several tumor suppressor genes in oral cancer, such as p16, DAPK1and MGMT 66,76,77. However, a reliable biomarker for progression has yet to be identified for oral cancer.  The objective of the following chapter is to characterize the methylation landscape throughout the stages of oral squamous cell carcinoma progression.  I predict that there will be an increasing amount of hypermethylated and hypomethylated genes as the disease progresses, and distinct DNA methylation profiles between the adjacent normal and tumor tissue.  Further, I predict that novel genes that are frequently hypermethylated or hypomethylated in our dataset will be identified.  These genes may provide new insights into the molecular mechanisms that characterize the stages of this disease.  To test these assumptions, the CpG island methylation status at tens of thousands of sites in the genomes of dysplasias, CIS/OSCCs, and healthy adjacent normal oral tissue will be evaluated. The analyses of these different tissue states involved patient-paired samples; all pathology states (dysplastic, CIS/OSCC, and normal/non-dysplastic) were collected from each   - 20 - patient contributing samples to our study. Internally-paired comparisons from these cases have identified stage-specific DNA methylation changes, including recurrent methylation alterations at specific genes and recurrent perturbations of established cell signaling pathways. Herein novel insights into the biology of early oral tumorigenesis are reported.  A version of Chapter 3 has been published in Oral Oncology with the following co-authors: D Truong, K Hogg, W.P. Robinson, CF Poh and C. Garnis.  3.2 Methods 3.2.1 Sample Cohort Table 3.1 outlines demographic information for the patients contributing samples for this study. Figure 3.1 illustrates biopsy selection for a representative case enrolled in this study. Each fresh-frozen biopsy samples was manually micro-dissected by an expert oral pathologist and DNA was extracted using a standard phenol: chloroform extraction protocol. A further sample set of CIS/OSCC samples from an independent cohort of 11 patients (Table 3.1) were analyzed by bisulfite pyrosequencing for candidate genes.   3.2.2 Sample Preparation and DNA Methylation Profiling Experiments DNA samples were profiled for DNA methylation using the Illumina 27K microarray as described in Chapter 2.4.   3.2.3  Gene Expression Profiling RNA samples were profiled for mRNA expression using the Agilent 4x44k microarray as described in Chapter 2.5.     - 21 -   Figure 3.1  Obtaining multiple biopsies from an oral cancer field. A) White light visualization of the oral cancer field. White circles represent the areas where punch biopsies were obtained. B) Fluorescence visualization (FV) of the disease field represented in A. C-E) Photomicrographs of hematoxylin and eosin stained slides of varying histology observed across the FV-defined disease field: C) adjacent normal (N), D) dysplasia, and E) CIS.       - 22 -   Table 3.1  Patient Demographics Discovery Set Patients  Total patients 10 Mean age 52 Age range 31-68 Number of males 5 Number of females 5 Former Smokers 2 Current Smokers 1 Non-Smokers 7 Cancer Site  Tongue 9 Floor of Mouth 1 Diagnosis  Oral Squamous Cell Carcinoma (OSCC) 9 Carcinoma in situ (CIS) 1   Validation Set Patients Total patients 11 Mean age 59 Age range 45-82 Number of males 5 Number of females 6 Former Smokers 2 Non-Smokers 9 Cancer Site  Tongue 10 Gingiva 1 Diagnosis  Oral Squamous Cell Carcinoma (OSCC) 11     - 23 - 3.2.4 Statistical Analyses Microarray data for all 30 evaluated DNA samples were assessed using average linkage hierarchical clustering based on the Pearson un-centered correlation coefficient (Genesis software, Graz University of Technology)78. Data from genes located on the X and Y chromosomes were eliminated from analyses. Only CpGs successfully detected in all 30 samples were included.  A non-parametric Wilcoxon signed rank test was performed to identify CpG sites that had significantly different DNA methylation profiles between biopsies. The list of significant CpGs associated with gene promoters was further refined by comparing the Δβ in the group. The proportion of probes that were unmethylated and methylated was calculated working within the total number of probes that were successfully detected for each case. The number of methylated probes in the three sample types was analyzed using a repeated measures ANOVA with a post hoc Tukey test. Differences between hypermethylation and hypomethylation events were determined by calculating the number of probes that were hypermethylated (Δβ > 0.15) and hypomethylated (Δβ < -0.15) and comparing to the total number of successful probes for each case. This data was then analyzed using a t-test and variability measured using coefficient of variation and F-test of equality of variances.   3.2.5 Bisulfite Pyrosequencing Candidate genes were validated using bisulfite pyrosequencing. For dysplasia samples, pyrosequencing was performed to assay the same CpG sites that were frequently aberrantly methylated in six of the patients in our original cohort. CpG sites, associated with candidate genes selected from the CIS/OSCC group, were assessed for DNA methylation status in eleven additional patients with paired adjacent normal and OSCC samples. Forward, reverse, and sequencing primers were designed using Pyro Q-CpG software (Qiagen Ltd;Primer sequences and conditions published online79). The PCR reactions were prepared with ~30ng bisulfite converted DNA, 1x PCR buffer   - 24 - (Qiagen Ltd), 0.2 mM dNTPs (Invitrogen), 0.4uM forward and reverse primers (Alpha DNA) and 0.5 U DNA polymerase (HotStarTaq, Qiagen Ltd.). Pyrosequencing was performed using the Pyromark Q96 MD Pyrosequencer (Qiagen Ltd.) as per manufacturer’s instructions and using 0.3 µM of sequencing primer. For the candidate genes, where possible, more than one CpG site in addition to the site of interest from the microarray was included in the assay (CpG sites; EYA4: cg07327468, KCNAB3: cg14918082, MAP3K14: cg15639951 TNF: cg11484872, RASGFR1: cg16154416). Within assays, DNA methylation at CpG sites were significantly correlated (Spearman’s Rho; MAP3K14: R = 0.90, TNF: R = 0.98, and RASGFR1: R = 0.96) and values were therefore averaged across the region.  3.2.6 TCGA Illumina 450K Microarray Data Analysis The Cancer Genome Atlas (TCGA) head and neck squamous cell carcinoma dataset was mined for data for patients with paired normal and tumors arising from the oral cavity (using data from samples collected from the tongue, floor of mouth, buccal mucosa, oral cavity and hard palate, n=32). DNA methylation (from the Illumina HumanMethylation450K microarray) was processed and normalized using the R package minfi80.  Differences were assessed using a Student’s t-test on the paired samples.   3.3 Results and Discussion Several groups have argued that epigenetic abnormalities play a key role in the earliest stages of cancer initiation81-83. To identify the epigenetic events that are “drivers” of disease progression as opposed to “bystanders” (alterations that do not drive cancer pathway disruptions), it is imperative to identify the molecular event or mechanism, such as aberrant DNA methylation, that may be causal to gene expression changes 84. In this report, we investigated genome-wide DNA methylation   - 25 - alterations in a unique panel of paired tissues that represent adjacent normal, dysplasia, and CIS/OSCC tissues. Our use of a fluorescence visualization (FV)-scope to identify disease field margins, which can be undetectable by standard white light, often results in histopathologically-heterogeneous excised tissues (Figure 1). Obtaining multiple biopsies from these excised tissues allows us to acquire samples of varied histopathological grade. Although it has been demonstrated that histologically normal tissue adjacent to an OSCC lesion often also exhibit molecular alterations, the advantage of using paired samples obtained from the same individual at the same time allows us to investigate how molecular events change throughout disease progression while controlling for potential confounding factors associated with using unpaired samples 85,86. This is particularly relevant for DNA methylation studies since DNA methylation patterns are known to be tissue specific and can be altered by extrinsic factors such as smoking77,87-89   3.3.1 Global DNA Methylation Landscape of Oral Tissues To characterize promoter methylation patterns during oral cancer progression, we profiled 30 micro-dissected biopsies from ten different patients; from each patient, a dysplastic lesion, CIS or OSCC, and a paired adjacent normal tissue specimen were profiled using methylation microarrays.  The Illumina HumanMethylation27K microarray is biased towards gene promoter regions. We noted that for all histological grades the majority of probes were unmethylated or low-methylated (<30% methylation). A three-way ANOVA with post hoc Tukey test was used to determine statistically significant differences among the three histological groups with respect to the proportion of probes that were low-methylated versus high-methylated (Figure 3.2). There was a statistically significant increase in the number of highly methylated CpGs (P = 0.004) in the CIS/OSCC group compared to the adjacent normal, but there were no significant differences between the normal and dysplasia or dysplasia and CIS/OSCC groups.   - 26 - To identify similarities and differences among the DNA methylation profiles, we performed an unsupervised hierarchical clustering analysis of the entire normalized data set for CpGs that were detected in all samples (Figure 3.3). Two main clusters emerged from this analysis, one including all normal biopsies and the other including all CIS/OSCC cases. Dysplastic biopsies were split across these clusters; seven dysplasias grouped within the normal cluster and three within the CIS/OSCC cluster, indicating a great deal of heterogeneity among the dysplastic DNA methylation profiles. However, when dysplastic grade was factored into analysis a clearer progressive trend emerged; mild dysplasias (the lowest grade of dysplasia) clustered with non-malignant tissues. The four moderate dysplasias and two severe dysplasias were split across the clusters, reflecting their transitional state.  3.3.2 DNA Methylation Changes Throughout Oral Cancer Progression   To identify epigenetic alterations playing a role in early oral tumorigenesis, we compared genome-wide DNA methylation profiling results for 1) oral dysplastic biopsies and 2) CIS/OSCC biopsies to results for paired normal biopsies. Signed rank analysis was performed to identify statistically significant differences between compared groups. Table 3.2 summarizes these results.   We observed that both hyper- and hypomethylation events were more prevalent in CIS/OSCC (with 1933 and 2006 altered CpGs, respectively) relative to paired dysplasias (761 and 93, respectively) (Table 3.2). We also noted that while hypermethylation was more prevalent in dysplastic tissues relative to hypomethylation, hypermethylation and hypomethylation occurred in CIS/OSCC at approximately the same frequency (Figure 3.4). The variability of the DNA methylation events (both hyper- and hypo-) were not statistically different between dysplasia and CIS/OSCC samples (P > 0.4 for both hyper- and hypomethylation events). Between group comparisons revealed that the CIS/OSCC group showed the most significant between group differences regarding both hyper-and hypomethylation. With respect to the dysplasia    - 27 -      Figure 3.2  Mean proportion of low or unmethylated (β-value =<0.3) and high-methylated (beta value 0.3-1) CpGs in normal, dysplasia, and CIS/ OSCC biopsies. Error bars indicate standard error across ten samples and circles demonstrate the minimum and maximum of proportions for each level of methylation.         - 28 -   Figure 3.3  Methylation patterns of a sample set of 30 biopsies that were analyzed using hierarchical clustering of results from 21395 CpG sites. This sample set was comprised of results from paired normal, dysplasia, and CIS/OSCC samples from ten patients. Sample names are listed across the top. Red indicates highly methylated and green indicates low methylated. N: Adjacent Normal. Mild Dys: Mild dysplasia. Mod Dys: Moderate dysplasia. Sev Dys: Severe dysplasia.    - 29 -   Figure 3.4  Methylation patterns in oral dysplasia and CIS/OSCC. Mean proportion of probes that have a Δβ of >0.15 (hypermethylated) and <-0.15 (hypomethylated) compared to the total number of CpGs for each sample in A) dysplasia and B) CIS/OSCC samples. Similarly, results are shown for comparisons at CpGs that were C) hypermethylated or D) hypomethylated in oral dysplasia and CIS/OSCC. The central line inside the box represents the mean, the box represents the standard error, the whiskers represent minimum and maximum values observed (n = 10).          - 30 - biopsies, the most prevalent alterations observed were hypermethylation events that were maintained in CIS/OSCC, followed by hypomethylation events that were maintained in CIS/OSCC (Figure 3.5). Of the total hypermethylation events detected in the dysplasia biopsies, a small proportion (2.09%), showed a further increase in DNA methylation, while 8.33% showed a decrease in DNA methylation in the paired CIS/OSCC.  Of note, cases with a higher amount of hypermethylation also exhibited a high degree of hypomethylation, suggesting the potential for broader dysregulation of DNA methylation machinery.  Table 3.2Dysplasia  Differentially-Methylated Probes CpGs Genes After sign rank analysis   4822 4048 After further cut-offs*   Total 854 695 Hypermethylated 761 605 Hypomethylated 93 90 *Present in at least 40% of cases; Δβ of 0.15 CIS/OSCC CpGs Genes After sign rank analysis   10232 7396 After further cut-offs*  Total 3939 3061 Hypermethylated 1933 1455 Hypomethylated 2006 1606    *Present in at least 40% of cases; Δβ of 0.15      - 31 -       Figure 3.5 General trends in DNA methylation patterns during the transition from normal to dysplasia and dysplasia to CIS/OSCC.  Each bar represents mean proportion of CpGs that exhibited each particular change across 10 cases and error bars present standard error.            - 32 - 3.3.3 Status of Previously Implicated Candidate Genes Several genes have been  reported to be aberrantly methylated in OSCC 90. Many of these genes have also been identified as having altered DNA methylation states in our sample set (Table 3.3); four of the six previously implicated candidate genes were observed to be aberrantly methylated in 10/10 (100%) of CIS/OSCC cases in our data set. DAPK1, RUNX3, and secreted frizzled-related protein 4 (SFRP4) were hypermethylated in 10/10 (100%) of the CIS/OSCC biopsies and ≥6/10 (60%) in the corresponding dysplasia biopsies. CDKN2A and MGMT were detected as hypermethylated at slightly lower frequencies in both CIS/OSCC and dysplasias (7/10 and 5/10, respectively). Matched mRNA expression data revealed concurrent decreases in expression in multiple cases that also exhibited hypermethylation of many of these genes. For example, six patients exhibited a decrease in DAPK1 expression (with two exhibiting deregulation in the dysplasia stage, three in the CIS/OSCC, and one in both stages). Five patients also exhibited SFRP4 down-regulation (with one exhibiting this change at the dysplasia stage and four showing it in CIS/OSCC biopsies). Interestingly, CDKN2A was down-regulated in two patients with matched hypermethylation (one in the dysplasia and the other in the CIS/OSCC biopsy); RUNX1 was down-regulated in a single patient where hypermethylation was observed (a CIS/OSCC biopsy); and MGMT was not found to be down-regulated in any patients exhibiting hypermethylation of this gene. There are multiple potential explanations for why hypermethylation of these genes was not associated with a concurrent decrease in gene expression, including the possibility that methylated regions may actually harbor promoters/enhancers for other genes governing oral cancer processes or that this localized methylation is governing malignant phenotypes via a different mechanism.       - 33 - Table 3.3 Methylation Deregulation of Known Gene Candidates  Frequency in Dysplasia N=10 Frequency in CIS/OSCC N=10 DAPK1 7/10 10/10 RUNX3 6/10 10/10 SFRP4 6/10 10/10 CDKN2A 5/10 7/10 MGMT 6/10 5/10 MPO 6/10 10/10   3.3.4 Frequently Altered Gene Candidates in Dysplasia To identify potential causal events in oral cancer initiation and progression, we identified highly frequent DNA methylation changes in both the dysplasia and CIS/OSCC biopsies.  A change was defined as a Δβ of at least 0.15. In the dysplasia samples, no change of this magnitude was observed for all samples, but 6 genes showed changes in 8 of the 10 samples (Table 3.4). Three of the six frequently hypermethylated genes (TRHDE, ZNF454, and KCNAB3) have never previously been linked to any cancer type. EYA4, ZNF671, and HOXA9, on the other hand, have all been shown to be deregulated through aberrant DNA methylation in various cancers91-95. When assessing the DNA methylation status of the six most frequently hypermethylated genes detected in dysplasia in the paired CIS/OSCC, EYA4, HOXA9, and TRHDE exhibited additional increased methylation in the paired CIS/OSCC in three cases and consistent methylation compared to the paired dysplasia in four cases, and a decrease in methylation in one case.   - 34 - Table 3.4 Identified Candidate Genes in Dysplasia Hypermethylated Gene Name Symbol Description Frequency   Dysplasia CIS/ OSCC  Potassium voltage-gated channel, shaker-related subfamily beta member 3 KCNAB3 Function unknown 8/10 9/10 Eyes absent 4 EYA4 Transcriptional co-activator involved in recruiting DNA repair machinery 8/10 10/10 Homeobox A9 HOXA9 Transcription factor involved in establishing cell identity during development 8/10 10/10 Thyrotropin-releasing hormone degrading ectoenzyme TRHDE An ectoenzyme that degrades thyrotropin-releasing hormone (TRH) 8/10 9/10 Zinc finger protein 454 ZNF454 Function unknown 8/10 10/10 Zinc finger protein 671 ZNF671 Function unknown 8/10 10/10  Hypomethylated Gene Name  Symbol  Description  Frequency  Dysplasia CIS/ OSCC  Potassium channel, subfamily K, member 18 KCNK18 Potassium channel protein 7/10 9/10 Complement component 1, q subcomponent, B chain C1QB Subunit of complement component 1 6/10 10/10 Caspase 8 CASP8 Thiol protease involved in apoptosis 6/10 9/10 Complement factor H-related 2 CFHR2 a serum protein that is related to complement protein H 6/10 8/10 Transient receptor potential cation channel M2 TRPM2 Calcium permeable cation channel on plasma membrane 6/10 10/10 Hepatocyte nuclear factor 4, alpha HNF4A Transcription factor essential for gastrointestinal tract development 6/10 7/10 Chromosome 12 open reading frame 36 C12orf36 Function unknown 6/10 10/10 Myeloperoxidase MPO Microbicidal enzyme 6/10 10/10   - 35 -  ZNF671 and ZNF454 showed increased DNA methylation in two CIS/OSCC cases, decrease in one case, and no additional changes in five cases. KCNAB3 showed an increase in one case and a decrease in one case and no change in six cases.  For all cases where increased DNA methylation of these six candidate genes was not detected in the dysplasia, an increase was observed in the paired CIS/OSCC sample compared to the paired normal.  Data from dysplasia cases also exhibited recurrently hypomethylated genes, though at lower frequency than the hypermethylated candidates. One gene, KCNK18 was observed to be hypomethylated in 7/10 (70%) of cases, and seven additional genes were hypomethylated in at least 6/10 (60%) of cases (C1QB, CASP8, MPO, CFHR2, TRPM2, HNF4A and C12orf36). Of the hypomethylated candidate genes identified in the dysplasia biopsies, five (KCNK18, MPO, C1QB, C12orf36 and CFHR2) have never been associated with cancer. The other three (TRPM2, CASP8, and HNF4A) have been previously linked to other cancers96-103. The majority of paired CIS/OSCC cases showed no additional changes in DNA methylation compared to the dysplasia. However, KCNK18 and C1QB showed further decreased methylation in three of the paired CIS/OSCC. Most of the dysplasia cases where no change in methylation was detected compared to normal showed a decrease in methylation in the paired CIS/OSCC.  Promoter methylation changes at KCNAB3 and EYA4 were validated by bisulfite pyrosequencing in six out of the ten patients originally profiled (as these were the only cases with sufficient DNA remaining for this analysis). For this comparison, only paired specimens from normal and dysplastic tissues were analyzed. Pyrosequencing data confirmed trends exhibited in these CpG sites and were highly correlative to our DNA methylation array data (R = 0.92 for EYA4 and R = 0.94 for KCNAB3; Figure 3.7).    - 36 - Gene expression data revealed that EYA4 expression was decreased by at least two-fold in six of eight patients that exhibited hypermethylation of this gene.  KCNAB3 was down-regulated in three out of the eight patients with aberrant methylation.   The high frequency of promoter methylation of candidate genes, the persistence of this alteration status into paired CIS/OSCC cases in many instances, and the observation of concurrent gene expression changes associated with methylation status in many cases suggest an early role for many of these genes in oral tumorigenesis.  3.3.5 Frequently Altered Gene Candidates in CIS/OSCC A greater number of high frequency methylation changes were observed in profiles from CIS/OSCC biopsies (relative to results from dysplasia biopsies). A total of 93 genes were hypermethylated and 139 hypomethylated in 10/10 (100%) of the CIS/OSCC biopsies (Supplemental Table 2). Interestingly, none of the previously implicated genes deregulated by DNA methylation in head and neck cancers were observed at this high frequency, reinforcing the need for unbiased genome-wide DNA methylation analyses. Among the 93 genes we detected as frequently hypermethylated in CIS/OSCC tissues, 63.4% have been reported as deregulated in various types of epithelial cancer, 15.3% have been implicated in HNSCC and 66% have been observed to have aberrant methylation in epithelial cancers. Among the 139 hypomethylated genes, 34.5% have been previously reported to be deregulated in epithelial cancers, 14.6% have been implicated in HNSCC, and 12.5% have been reported to be deregulated specifically by aberrant DNA methylation in epithelial cancers. The discrepancy between the high proportions of previously identified hypermethylated genes compared to the low number of hypomethylated genes is likely due to the tendency of studies to focus on promoter hypermethylation as opposed to hypomethylation.   - 37 -  3.3.6 Epigenetically Driven Pathway Disruption in Oral CIS/OSCC To identify genes in known biological pathways and/or processes that may be deregulated due to altered DNA methylation, we used ClueGo, a Cytoscape plug-in that integrates Gene Ontology as well as KEGG pathways 104. This software, in addition to identifying enrichment of genes involved in inflammation and metabolism pathways, also highlighted an enrichment of genes showing aberrant DNA methylation in cancer related pathways in general and more specifically the WNT and MAPK pathways (Figure 3.6).  The function of the canonical WNT pathway is to regulate the transcription of β-catenin, which regulates transcription factors involved in cancer promoting activities such as cell proliferation and survival 105. The WNT pathway has been previously implicated in oral cancer106,107. Our results show that the majority of genes deregulated by DNA methylation work to activate the WNT pathway. Antagonists of this pathway – including various SFRPs, SOX17, and WIF1 – have previously been reported as silenced by promoter methylation in a variety of cancers, including OSCC 108-110, with the most frequently reported hypermethylation events in OSCC previously reported as occurring at SFRP4 and SFRP5 in 78% and 59% of cases, respectively 109. We observed similar patterns of hypermethylation of the SFRPs but at slightly different frequencies. We also detected a much higher frequency of hypermethylation for WIF1 than previous studies. In addition to WNT pathway members previously reported, we also detected highly frequent aberrant methylation in several additional members of the WNT pathway (Table 3.5). Several of these candidates also showed frequent deregulation in paired dysplasia samples. Aberrant methylation of the MAPK pathway members has been reported in several cancer types but has yet to be reported in oral cancer111-113. The MAPK pathway is involved in translating   - 38 -        Figure 3.6  Signaling network of candidates genes in Wnt and MAPK pathways that are frequently deregulated by aberrant DNA methylation in CIS/OSCC. Genes with deregulated DNA methylation in WNT (hexagons), MAPK (diamonds) and other pathways (ovals) in cancer present in at least 8/10 (80%) CIS/OSCC samples.  Genes belonging to the represented signaling cascades that did not directly interact with candidate genes defined by our analyses were not represented to facilitate clarity. Overlapping shapes indicate genes are present in more than one pathway.  Hypomethylation is denoted by grey shapes and hypermethylation by black shapes.  Dotted arrows indicate indirect relationships and solid arrows indicate direct relationships. Gene interactions are shown as activating (arrows) or inhibiting (blocked arrows).      - 39 -  extracellular signals to cellular machinery to control cell processes, including growth, proliferation,  differentiation and apoptosis 114. As with the WNT pathway, this signal cascade is highly implicated in tumorigenesis. We have found several genes within this pathway to exhibit frequent aberrant DNA methylation in oral cancer and dysplasia (Table 3.6). Several pathway members of the classical MAPK pathway are hypermethylated which would result in deactivation of the pathway; however, the JNK and p38 MAP kinase pathway appears to be activated by the aberrant DNA methylation detected in our sample set.  Matched gene expression profiles revealed that TNF, MAP3K14, and RASGFR1 are also deregulated at the mRNA level in multiple patients. Six out of the nine patients with TNF hypomethylation also showed a ≥2-fold increase in expression. RASGFR1 was detected as down-regulated in four of nine patients with hypermethylation within the promoter of this gene. Similarly, three of eight patients with hypermethylation at MAP3K14 exhibited a subsequent decrease in expression.   Three of the genes in the MAPK pathway found to exhibit recurring DNA methylation alteration – TNF, MAP3K14, and RASGRF1 – were validated in an independent cohort of paired adjacent normal and OSCC samples using bisulfite pyrosequencing (n = 11, Figure 3.7). The MAP3K14 promoter was found to be significantly hypermethylated (P = 0.0029), with 7/11 SCCs exhibiting at least a 21% change in methylation compared to adjacent normal tissues (Figure 3.8A). TNF also showed a significant trend towards hypomethylation (P = 0.0068). Five out of eleven  patients showed a decrease of methylation of at least 17% (Figure 3.8B). Although RASGRF1 did not show significant hypermethylation (P = 0.12), hypermethylation at this CpG island was exhibited in 4/11 patients (Figure 3.8C).     - 40 - Table 3.5 Frequency of WNT pathway candidate genes   Hypermethylated   Dysplasia N=10 CIS/ OSCC N=10 CCND1 6/10 8/10 NFATC2 3/10 8/10 PRKCB1 6/10 9/10 SFRP1 1/10 5/10 SFRP2 5/10 8/10 SFRP4 6/10 10/10 SOX17 7/10 10/10 WIF1 4/10 8/10 WNT5B 2/10 8/10  Hypomethylated   Dysplasia N=10 CIS/ OSCC N=10 PLCB2 1/10 8/10 WNT8A 1/10 9/10 SFRP5 0/10 1/10   Table 3.6 Frequency of MAP Kinase pathway candidate genes   Hypermethylated   Dysplasia N=10 CIS/ OSCC N=10 CACNA1A 5/10 9/10 FGF3 4/10 8/10 MAP3K14 3/10 8/10 NFATC2 3/10 8/10 PRKCB1 6/10 9/10 PTPRR 5/10 8/10 RASGRF1 5/10 9/10  Hypomethylated   Dysplasia N=10 CIS/ OSCC N=10 CACNG3 3/10 9/10 FGF4 3/10 8/10 FGF6 3/10 8/10 TNF 3/10 9/10    - 41 -  Figure 3.7 Correlation between % methylation determined using the microarray (X) and bisulfite pyrosequencing (Y).  Adjacent normal and dysplasia DNA from six patients from the original discovery cohort were assessed using bisulfite pyrosequencing in EYA4 (n-10) and KCNAB3 (n=12).  Correlation coefficient determined by Spearman’s Rho test.      - 42 -  Figure 3.8 DNA methylation status of MAP3K14 (A), TNF (B) and RASGFR1 (C) in a validation cohort of 11 patients profiled by bisulfite pyrosequencing. Each line indicates the change in percentage DNA methylation from adjacent normal (NM) to OSCC.  P-values were determined by Wilcoxon signed rank test.     - 43 - Deregulation of several genes within a given pathway, or linked pathways, has striking implications in a time where there is significant emphasis on developing targeted therapies: it will make little sense to target a single gene that can be functionally bypassed by deregulating other factors in the same signaling cascade.  3.3.7 Analysis of Candidate Methylation Events in TCGA Datasets In order to determine if the CpGs we identified as displaying frequent aberrant methylation were similarly deregulated in independent cohorts of oral cancer data, we assessed the methylation status of the genes listed in Table 3.4-6 in the oral cavity samples from the TCGA Head and Neck Squamous Cell Carcinoma data set.  We identified a high level of concordance in the level of methylation in the in the original analysis cohort (referred to as the British Columbia Cancer Research Centre (BCCRC) cohort here on) compared to the TCGA cohort (Table 3.7).  Of all the genes identified as differentially methylated in the BCCRC cohort, all but one (SFRP1) had a significant decrease in methylation.  Taken together, the similarity in change in methylation patterns of these genes in these two data sets gives further evidence that these genes are highly frequent events in oral cancer patients.    Table 3.7 Validation of Differentially Methylated CpGs in  TCGA Data Hypermethylated  BCCRC Cohort  TCGA Cohort        Probe ID Gene Frequency n=10 TTEST Frequency  n=32 TTEST cg02919422 SOX17 1 3.01E-05 0.96875 5.03E-16 cg01381846 HOXA9 1 1.53E-07 0.9375 5.86E-16 cg03355526 ZNF454 1 2.93E-06 0.84375 5.51E-15 cg08261094 SFRP4 1 3.18E-06 0.875 1.04E-11 cg19831575 FGF4 0.9 3.48E-06 0.875 6.86E-14    - 44 - Hypermethylated  BCCRC Cohort  TCGA Cohort        Probe ID Gene Frequency n=10 TTEST Frequency  n=32 TTEST cg19427610 WIF1 0.8 2.27E-04 0.75 1.34E-10 cg23207990 SFRP2 0.8 1.88E-04 0.8125 2.25E-10 cg15337897 FGF3 0.8 1.64E-04 0.75 6.21E-10 cg24176563 EYA4 0.8 8.62E-05 0.78125 1.01E-09 cg19246110 ZNF671 0.8 2.08E-04 0.8125 1.11E-09 cg15639951 MAP3K14 0.8 8.95E-05 0.59375 1.88E-09 cg11086066 NFATC2 0.8 1.37E-04 0.75 4.11E-07 cg24505122 WNT5B 0.8 4.13E-03 0.375 1.48E-04       Hypomethylated  BCCRC Cohort  TCGA Cohort         Probe ID Gene Frequency n=10 TTEST Frequency n=32 TTEST cg23834593 HNF4A 1 8.06E-08 0.8125 2.54E-13 cg09421562 MPO 1 3.27E-06 0.875 2.68E-13 cg08191854 TRPM2 1 6.63E-06 0.84375 5.09E-13 cg03941108 C1QB 1 3.53E-06 0.8125 7.99E-13 cg07637239 KCNK18 0.9 1.48E-05 0.84375 1.20E-14 cg08603768 WNT8A 0.9 4.47E-07 0.71875 2.85E-10 cg11484872 TNF 0.9 4.28E-05 0.65625 9.08E-09 cg26799474 CASP8 0.8 2.23E-04 0.96875 2.32E-19 cg09551916 CFHR2 0.8 5.86E-05 0.78125 1.54E-11 cg06166767 SFRP1 0.8 2.83E-05 0.125 1.55E-01 cg04721098 CACNG3 0.7 1.79E-04 0.78125 3.10E-12 cg10207745 C12orf36 0.7 3.17E-04 0.84375 4.26E-12 cg02240622 PLCB2 0.5 1.21E-04 0.3125 8.61E-06 cg01731341 FGF6 0.4 3.34E-04 0.375 5.13E-06  3.4 Conclusions DNA methylation is an important mechanism for regulating gene expression and it is often disrupted in cancer. This first comprehensive epigenetic profiling of oral cancers and precancers shows distinct CpG methylation changes at different disease stages.  Our rare sample set consists of multiple biopsies of various histological grade obtained from the same individual at the same time, which facilitates the robust identification of DNA methylation changes during oral cancer progression by reducing the impact of potential confounding factors like differences in smoking   - 45 - habits, diet, age, etc. This rigorous analytical approach, when applied to a larger sample collection, may facilitate robust delineation of novel biomarkers and candidates for new targeted therapies, critical tools for improving approaches to disease management – and oral cancer outcomes.    - 46 - Chapter 4: Deregulation of miRNAs throughout the Progression of OSCC in Never-smokers  4.1  Introduction MicroRNAs (miRNAs) are a class of short ~22 nucleotide-long, non-coding, single stranded RNAs that control gene expression of up to a third of the human genome 115. After undergoing a series of processing events and being incorporated into the RNA induced silencing complex, miRNAs bind to 3’ untranslated regions on target mRNAs. This interaction leads to either mRNA degradation or translational repression. Since many miRNA targets are oncogenes and tumor suppressors, miRNA signaling is involved in numerous cellular processes that contribute to tumor development, including cell differentiation, proliferation, and apoptosis. In the last decade, aberrant miRNA expression has been implicated in the development of multiple tumor types116,117.  The sample cohort this chapter focuses on is comprised of non-smoking patients.  Tobacco use (particularly smoking) is identified as a major etiological factor for oral cancers118, however OSCC also arise in individuals without smoking history. Some of these cases have been attributed to the presence of the human papillomavirus (HPV) while the causative factors in others remain unknown119,120.   Emerging evidence suggests that OSCCs in non-smoking individuals may have distinct disease characteristics. OSCC patients without history of tobacco or alcohol use exhibit a female predominance, higher average age, and a preponderance of tumor presentation at the mandibular alveolar ridge and maxilla at diagnosis121-125. Regarding molecular alterations, a lower proliferation index (based on Ki-67 expression) has been noted in tumor-adjacent epithelia of non-smoking head and neck cancer patients relative to smoker patients (with the majority of samples originating in the oral cavity)126. Expression of p53 in tumor-adjacent mucosa of non-smoking and non-drinking OSCC   - 47 - patients is lower than in comparable tissues from patients with tobacco and alcohol use histories124,127. In addition, p53 mutations are less prevalent in non-smoking OSCC patients relative to smoking patients and, when p53 mutations are present in non-smoking OSCC patients, they are confined to cytosine phosphate guanosine (CpG) dinucleotide “hot spots”128-130. Other studies show that non-smoking individuals typically harbor fewer tumors with a loss of heterozygosity (LOH) at the 3p, 4q, and 11q13 chromosomal loci130. Additionally, OSCCs from non-smoker subjects have shown a higher frequency of amplification of chromosomal locus 3q26-27 and gain of chromosome 1p when compared to the smoking OSCC patient population131. While the exact causes for the different clinic-pathological behavior of OSCC malignancies in non-smokers remains enigmatic, these collective findings suggest a unique mechanism of pathogenesis.  To date, there have been few reports regarding the role of miRNAs in oral premalignant lesions (OPLs)132,133. Here, we report results of an analysis of miRNA dysregulation in OPLs and OSCCs, with analyzed specimens derived from a rare cohort of internally-controlled oral tissues collected from contiguous disease fields defined by fluorescence visualization (FV) device. We describe trends in altered miRNA behaviors during disease progression and identify specific miRNAs recurrently dysregulated at different stages of oral premalignancy/ malignancy in non-smoking individuals.  A version of this chapter has been published in  the Journal of Interdisciplinary Medicine and Dental Science with the following co-authors: M. Gorenchtein, C Dickman, Y Zhu, CF Poh and C Garnis.  4.2 Materials and Methods 4.2.1 Sample Accrual   Samples were collected as described in Chapter 2.1.  A table listing patient demographics can be found in Table 4.1.  HPV status was determined using in situ hybridization (ISH).   - 48 - Table 4.1 Patient Demographic Information. Patient ID Gender Smoking Status HPV Status Lesion Site Age Diagnosis 1878 M NS Negative Tongue 48 SCC 1953 F NS Negative Tongue 53 CIS 2000 F NS Negative Tongue 58 SCC 2076 F NS Negative Tongue 82 SCC 3002 F NS Negative Tongue 41 SCC 5266 M NS Negative Tongue 55 SCC 7057 F NS Negative Tongue 53 SCC 7071 M NS Negative Tongue 71 CIS 7058 F NS Negative Tongue 43 SCC  4.2.2 RNA Isolation and miRNA Expression Quantification RNA was isolated as per Chapter 2.2.  RNA samples were profiled for mRNA expression using the Agilent 4x 44 K microarrays as described in Chapter 2.3.2  4.2.3 Quantitative RT-PCR of Candidate Genes  The expression of selected miRNA candidates (miR-224, miR-135a, miR-143, miR-223, miR-155,  and miR-605) was re-examined, using individual miRNA PCR primer sets (Exiqon) and following the same protocol as was used for the 384-well panels described above.  These candidates were determined based on the frequency of their expression within our sample set and due to their potential role in tumorigenesis based on a review of the literature of these miRNAs in other cancer types.   Due to their stable expression levels across the sample set during the above global miRNA expression analysis, miR-103 and miR-23b were chosen as reference genes for data normalization. All assays were run in triplicate on the MicroAmp Fast Optical 96-Well Reaction Plates (Applied Biosystems) and no enzyme controls were included.    - 49 - 4.2.4 Development of Tissue Microarrays (TMAs) for Verification and Validation of Candidate miRNA Deregulation Formalin fixed paraffin embedded (FFPE) specimens corresponding to each of the 27 fresh frozen patient tissue samples were used for assembly of TMAs to verify the expression of candidate miRNAs. Briefly, 2 mm cores were obtained in duplicate from each paraffin biopsy and distributed amongst 2 recipient TMA blocks using a specific arraying device (Manual Tissue Arrayer MTA-1, BeecherInstruments, Inc. WI, USA). An independent validation set, a TMA was constructed consisting of premalignant and malignant archival patient tissues from the British Columbia Oral Biopsy Service. Individual 1 mm tissue cores from 26 primary dysplasias were deposited into a single TMA block. A second TMA block was built from 31 independent OSCC tumors, each sampled in 2 replicates (0.6-mm-diameter tissue cores).  Multiple 6-μm sections were cut from each TMA block with a microtome (Leica RM2235, Leica Biosystems, Ashbourne, Ireland) and used for in situ hybridization analysis.   4.2.5 in situ Hybridization (ISH)  Digoxigenin (DIG)-labeled locked nucleic acid (LNA) modified probes targeting miR-155, positive control (U6 snRNA), and negative control (scrambled-miRNA) were obtained from manufacturer (Exiqon, Woburn, MA). The entire procedure was performed according to manufacturer protocols, with minor adjustments. Briefly, 6 µm thick formalin-fixed paraffin embedded tissue sections were deparaffinized with xylene, rehydrated in graded ethanol and treated with 300µL Proteinase-K (15 min for slides stained with U6 probe; 20 min for slides stained with miRNA and scrambled probes) at 37°C. Tissue sections were then dehydrated through successive ethanol washes and hybridized with miRNA-specific (100 nM), U6 (1 nM) and scrambled (40 nM) probes for 1 h. Hybridization temperatures were 48°C for the miRNA-specific probes and 54°C for   - 50 - the U6 and scrambled probes. Following stringent washes in SSC buffers, the sections were incubated with alkaline phosphatase-conjugated anti-DIG (Roche) at dilutions 1:500 for miRNA probes and 1:800 for U6 and scrambled probes. This was performed at room temperature for 1 h. ISH signal was then detected by incubation with 4-nitro-blue tetrazolium (NBT) and 5-brom-4-chloro-3′-Indolylphosphate (BCIP) substrate (Roche) at 30°C for 2 h. Nuclear fast red (Vector Laboratories) was used as a counterstain and Eukitt mounting medium (VWR) for cover glass mounting. Results were analyzed the next day with an Olympus MVX10 microscope equipped with a charge-coupled device camera and CellP software (Olympus). The results were examined and scored for cytoplasmic staining by a pathologist. The dominant staining intensity was scored as: 0 = negative; 1 =positive.   4.3 Results 4.3.1 miRNA Expression in Oral Cancer Progression  We describe miRNA expression results from a rare cohort of non-smoker oral cancer patients (n = 9), where analyzed tissues were obtained from dysplasia, CIS/OSCC, and paired adjacent normal biopsies collected from within a single contiguous field of diseased oral tissue (Figure 4.1). As only a subset of miRNAs are expressed in a given tissue type or disease state 134,135, Figure 4.2 summarizes the number of detected miRNAs in each histopathological group (normal, dysplasia, and CIS/OSCC). A repeated measure ANOVA with a post hoc Tukey test was performed to determine statistical differences between the three groups. The total number of detected miRNAs was not statistically different between the normal and the dysplasia groups, however statistically significant differences were observed between the normal group versus the CIS/OSCC group and between the dysplasia versus the CIS/OSCC group (P = 0.0009, for both comparisons). Figure 4.2B also shows the number of miRNAs in each group  for which expression was solely detected within that single group.    - 51 - 4.3.2 miRNA Dysregulation Throughout Oral Cancer Progression  To identify miRNAs potentially playing a role in oral tumorigenesis, we undertook paired analyses of miRNA expression data from each of the dysplasia and CIS/OSCC cases versus miRNA expression results obtained from associated normal biopsies. A two-fold change cut-off was used to identify differences between compared samples within a given patient. Table 4.2 summarizes these results. Two hundred and sixty miRNAs were identified as dysregulated in both dysplasia and CIS/OSCC groups. Fifty-four miRNAs were found to have altered expression in the dysplasia group alone while eighty-seven miRNAs were found to be expressed only in the CIS/OSCC group. An average across all samples revealed more miRNA dysregulation in the CIS/OSCC group relative to the dysplasia group (with decreased candidate miRNA expression relative to normal cases the more common direction of alteration in both group comparisons). Figure 4.3 shows the pattern of miRNA deregulation in each patient sample set.  4.3.3 Candidate miRNAs Contributing to Oral Dysplasias  Those miRNAs recurrently altered across a disease sample set are more likely to represent causal alterations driving oral cancer progression. In the dysplasia group, the most frequently occurring miRNA alteration events were down-regulation of miR-886-5p, miR-375 and miR-143-3p relative to paired normal tissues (observed in 5/9 (56%) of cases), followed by down-regulation  of eleven other miRNAs (all observed in 4 out of 9 (44%) patients). Highly recurring down-regulated miRNAs from the dysplasia group showed variable trends in the paired CIS/OSCC cases (relative to normal), with miR-886-5p remaining the same, miR-375 increasing, and miR-143-3p decreasing in frequency.  The most frequently up-regulated miRNAs seen in the dysplasia group included miR-142-3p, miR-146a-3p, miR-150-5p, miR-182-3p, miR-187-3p, miR-224-5p miR-26b-3p, miR-577 and miR-501-5p (observed in 4/9 (44%) of cases).  Three of these recurrently up-   - 52 -  Figure 4.1 Clinical characterization and corresponding histology within an oral cancer field of case # 7071.  A) White light visualization. B)Fluorescent visualization of lesions.  White circles represent areas biopsied.  C-E) photomicrographs of hematoxylin and eosin stained slides of varying histology: C) adjacent normal, D) dysplasia and  E) carcinoma in situ.  F-H: Staining of mir-155 in F) adjacent normal, G) dysplasia and  H) carcinoma in situ . Digoxigenin (DIG)-labeled locked nucleic acid (LNA) modified probes targeting miR-155, positive control (U6 snRNA), and negative control (scrambled-miRNA) were obtained from manufacturer (Exiqon). The entire procedure was performed according to manufacturer protocols, with minor adjustments. Results were examined and scored for cytoplasmic staining by the pathologist (CFP). The dominant staining intensity was scored as: 0 = negative; 1 = positive. I) Fold change as determined by qRT-PCR of mir-155 in dysplasia and CIS compared to adjacent normal.   - 53 -   Figure 4.2 A)Venn diagram showing patterns of miRNA detection throughout each histopathological stage across nine samples.  Each number represents miRNA detected in at least one case in adjacent normal, dysplasia and/or CIS/OSCC.  B) Mean number of miRNA detected in normal, dysplasia, and tumor biopsies.  Error bars indicate standard error across 9 samples and circles demonstrate the minimum and maximum of number detected for each biopsy group.  A B   - 54 - Table 4.2 Highly Frequent Deregulated miRNAs. miRNA  Up in Dysplasia Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-142-3p 4 7  2.7 hsa-miR-146a-5p 4 7  2.7 hsa-miR-150-5p 4 6  5.5 hsa-miR-501-5p 4 4  3.1 hsa-miR-182-3p 4 2  3.5 hsa-miR-187-3p 4 2  2.8 hsa-miR-224-5p 4 2  2.3 hsa-miR-577 4 1  6.7 hsa-miR-26b-3p 4 0  2.6 miRNA  Down in Dysplasia Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-375 5 8  0.26 hsa-miR-886-5p 5 5  0.33 hsa-miR-143-3p 5 0  0.33 hsa-miR-135a-5p 4 6  0.13 hsa-miR-204-5p 4 5  0.35 hsa-miR-195-5p 4 3  0.35 hsa-miR-451a 4 3  0.25 hsa-miR-886-3p 4 3  0.32 hsa-miR-1-3p 4 2  0.19 hsa-miR-133a-3p 4 2  0.28 hsa-miR-133b 4 2  0.27 hsa-miR-31-3p 4 1  0.34 hsa-miR-424-5p 4 1  0.32 hsa-miR-126-3p 4 0  0.4 miRNA  Up in CIS/OSCC Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-21-5p 2 8  9.4 hsa-miR-424-5p 2 8  6.7 hsa-miR-142-3p 4 7  4.5 hsa-miR-146a-5p 4 7  4.1 hsa-miR-155-5p 3 7  5.1 hsa-miR-223-3p 2 7  7.1 hsa-miR-31-3p 1 7  16.3 hsa-miR-150-5p 4 6  3.3 hsa-miR-132-3p 2 6  3.4 hsa-miR-31-5p 2 6  22.9 hsa-miR-133a-3p 1 6  35   - 55 - miRNA  Up in CIS/OSCC Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-133b 1 6  34.3 hsa-miR-146b-3p 3 5  3.4 hsa-miR-146b-5p 3 5  6.5 hsa-miR-142-5p 2 5  4.9 hsa-miR-1-3p 1 5  64.1 hsa-miR-503-5p 1 5  6.1 hsa-miR-181b-5p 0 5  3.1 hsa-miR-501-5p 4 4  2.9 hsa-let-7i-3p 3 4  2.9 hsa-miR-455-3p 2 4  5.9 hsa-miR-132-5p 1 4  3.3 hsa-miR-136-5p 1 4  4.8 hsa-miR-421 1 4  2.7 hsa-miR-505-3p 1 4  2.8 hsa-miR-671-3p 1 4  2.7 hsa-miR-21-3p 0 4  5.9 hsa-miR-22-3p 0 4  2.8 hsa-miR-34b-5p 0 4  4.5 miRNA  Down in CIS/OSCC Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-375 5 8  0.05 hsa-miR-605-5p 1 8  0.23 hsa-miR-1260a 2 7  0.18 hsa-miR-135a-5p 4 6  0.14 hsa-miR-203a-3p 1 6  0.27 hsa-miR-886-5p 5 5  0.28 hsa-miR-204-5p 4 5  0.18 hsa-let-7c-5p 1 5  0.36 hsa-miR-99a-5p 1 5  0.2 hsa-miR-200b-3p 0 5  0.28 hsa-miR-224-3p 2 4  0.31 hsa-miR-125b-2-3p 1 4  0.33 hsa-miR-200a-3p 1 4  0.3 hsa-miR-26a-5p 1 4  0.41 hsa-miR-26a-1-3p 1 4  0.4 hsa-miR-99a-3p 1 4  0.33 hsa-miR-1238-3p 0 4  0.3 hsa-miR-149-5p 0 4  0.21 hsa-miR-200b-5p 0 4  0.29   - 56 - miRNA  Down in CIS/OSCC Freq in Dysplasia Freq in CIS/OSCC Mean Fold Change1 hsa-miR-23b-3p 0 4  0.45 hsa-miR-643 0 4   0.18  (1across cases with >2x increase) regulated miRNAs, mir-142-3p, miR-146a-3p and miR-150-5p, increased in frequency in the paired CIS/OSCC cases.The other miRNAs became less frequent in the paired CIS/OSCC biopsies, with the exception of miR501-5p, which had no change in frequency.    4.3.4 miRNAs Dysregulated in More Advanced Disease  In the CIS/OSCC group, the most frequently up- and down-regulated miRNAs were observed in 8/9 (89%) of cases (through comparison against individual paired normal cases). Highly frequent up-regulation of miR-21-5p and miR-424-5p and highly frequent down-regulation of miR-375 and  miR-605-5p were observed as critical in the advanced disease stage as these miRNAs were observed to be minimally altered in the dysplasia cases (≤20% of cases).  MiR-375 expression was an exception, as it was detected as down-regulated in 80% of CIS/OSCC cases and 50% of dysplasia cases. MiRNAs up-regulated in CIS/OSCC cases relative to paired normal tissues – including miR-146a-5p, miR-142-3p, miR-155-5p, and miR-223-3p – were found at high frequencies as well (7/9 CIS/OSCC cases (78%). These miRNAs were also more likely to be observed as concurrently up-regulated in paired dysplasia cases.   A subset of miRNAs (miR-224-3p, miR-135a-5p, miR-143-3p, miR-223, miR-155-5p, and miR-605-5p) was selected for validation (based on criteria described in the Methods section). Expression patterns for this subset were independently validated by individual qRT-PCR expression analysis (Appendix A.1). Replicate experiments revealed expression behaviors consistent with initial findings for all candidates except for miR-224-3p which was subsequently removed from further analysis.     - 57 -  Figure 4.3  Trends in deregulated miRNA.  Number of miRNA upregulated (ddCT > 2) or downregulated (ddCT<0.5) for dysplasia and tumor relative to adjacent normal for all nine cases profiled.                 - 58 - 4.3.5 in situ Hybridization Analysis    MiR-155-5p was identified as frequently up-regulated in both dysplasia and CIS/OSCC tissues. It also exhibited the highest average fold-changes in expression among candidate miRNAs in both dysplasia (4.14-fold change) and CIS/OSCC (4.77-fold change) groups relative to normal tissues (Table 4.2). With several recent reports presenting miR-155 as an oncogene candidate in multiple cancer types, we selected this candidate for further evaluation in a larger sample set136-145.  Two TMAs were constructed for verification and validation purposes. The first TMA consisted of biopsies from the nine patients used for the above qRT-PCR analysis. This TMA was used to confirm a correlation of miR-155 expression between ISH staining and qRT-PCR data. ISH staining for miR-155-5p correlated with the RT-PCR data for all samples represented on the TMA. Figure 1 is a representative example of this correlation.  The next TMA consisted of an independent panel of 29 tissue cores representing both CIS and dysplasia (n = 18) and OSCC (n = 11) cases. Staining of this TMA revealed a high level of expression of miR-155-5p, with elevated expression observed for 64% (7/11) of OSCC cores and 94% (17/18) of CIS and dysplasia cores. The samples on the TMA contained smokers and non smokers as well as HPV positive and negative cases.  MiR-155-5p expression did not correlate with HPV infection (present in 28% of the cores). It was observed to be highly expressed in TMA from both non-smokers (14 out of 19, 74%) and smokers (10 out of 10, 100%).  4.3.6 Validation of Highly Frequent Deregulated miRNAs in the Tumor Stage in the TCGA Data Set of Oral Cavity Tumors  We analyzed the Cancer Genome Atlas miRNA-Seq data for paired normal-tumor patient samples from tissue originating from the oral cavity.  When comparing the fold change of miRNAs we identified as deregulated in our sample set, we found a fairly low agreement between the two data   - 59 - sets (Table 4.3). Of the 24 miRNAs we identified as being recurrently down-regulated in our sample set, 25%  had a significant decrease in the TCGA set (as determined by Student’s t-test).  There was a higher proportion of up-regulated miRNAs validating in TCGA dataset.  Of the 30 miRNAs we found to be recurrently up-regulated, approximately 37% had a significant increase in expression in the tumor samples. Table 4.3 Validation of Frequently Down-regulated miRNAs in the TCGA Cohort miRNA Accession Number BCCRC Down Freq TCGA Down Freq TCGA Status TCGA pValue miR-99a-3p MIMAT0004511 0.556 0.897 Down 3.80E-09 miR-204-5p MIMAT0000265 0.667 0.862 Down 5.53E-09 miR-26a-1-3p MIMAT0004499 0.444 0.586 Down 7.84E-08 let-7c-5p MIMAT0000064 0.556 0.828 Down 1.39E-07 miR-125b-2-3p MIMAT0004603 0.556 0.690 Down 1.49E-06 miR-195-5p MIMAT0000461 0.333 0.655 Down 1.61E-05 miR-224-5p MIMAT0000281 0.333 0.000 Up 1.87E-07 miR-451a MIMAT0001631 0.333 0.000 Up 6.41E-06 miR-577 MIMAT0003242 0.444 0.000 Up 1.28E-03 miR-224-3p MIMAT0009198 0.444 0.069 Up 1.11E-02 miR-99a-5p MIMAT0000097 0.556 0.034 Up 4.12E-05 miR-1260a MIMAT0005911 0.778 0.000 No Change 1.22E-02 miR-23b-3p MIMAT0000418 0.444 0.310 No Change 5.83E-02 miR-26a-5p MIMAT0000082 0.444 0.310 No Change 5.83E-02 miR-1238-3p MIMAT0005593 0.667 0.207 No Change 1.25E-01 miR-203a-3p MIMAT0000264 0.667 0.379 No Change 1.32E-01 miR-135a-5p MIMAT0000428 0.667 0.000 No Change 1.70E-01 miR-200b-5p MIMAT0004571 0.444 0.172 No Change 2.20E-01 miR-200b-3p MIMAT0000318 0.556 0.241 No Change 2.98E-01 miR-149-5p MIMAT0000450 0.444 0.241 No Change 4.42E-01 miR-375 MIMAT0000728 0.889 0.069 No Change 4.44E-01 miR-200a-3p MIMAT0000682 0.444 0.138 No Change 5.43E-01 miR-643 MIMAT0003313 0.556 0.138 No Change 7.93E-01 miR-605-5p MIMAT0003273 0.889 0.000 No Change 1.00E+00      - 60 - Table 4.4 Validation of Frequently Up-regulated miRNAs in the TCGA Cohort miRNA Accession Number BCCRC Up Freq TCGA Up Freq TCGA Status TCGA pValue miR-21-5p MIMAT0000076 0.889 0.759 Up 9.89E-11 miR-501-5p MIMAT0002872 0.556 0.897 Up 5.10E-08 miR-181b-5p MIMAT0000257 0.556 0.483 Up 2.94E-07 miR-424-5p MIMAT0001341 0.889 0.414 Up 4.19E-06 miR-31-3p MIMAT0004504 0.778 0.828 Up 8.56E-05 miR-505-3p MIMAT0002876 0.444 0.655 Up 1.45E-04 miR-21-3p MIMAT0004494 0.778 0.483 Up 3.64E-04 miR-503-5p MIMAT0002874 0.889 0.241 Up 3.75E-04 miR-34b-5p MIMAT0000685 0.667 0.724 Up 7.87E-04 miR-187-3p MIMAT0000262 0.667 0.414 Up 1.15E-02 miR-132-3p MIMAT0000426 0.778 0.207 Up 2.85E-02 miR-421 MIMAT0003339 0.556 0.034 Down 4.97E-04 miR-136-5p MIMAT0000448 0.444 0.000 Down 1.62E-03 miR-1-3p MIMAT0000416 0.667 0.138 Down 9.05E-03 miR-133a-3p MIMAT0000427 0.778 0.138 Down 1.63E-02 miR-455-3p MIMAT0004784 0.444 0.103 Down 1.81E-02 miR-133b MIMAT0000770 0.778 0.103 Down 2.91E-02 miR-22-3p MIMAT0000077 0.444 0.207 No Change 5.95E-02 miR-146a-5p MIMAT0000449 0.778 0.207 No Change 1.23E-01 miR-146b-5p MIMAT0002809 0.667 0.379 No Change 2.77E-01 miR-223-3p MIMAT0000280 0.778 0.414 No Change 3.48E-01 let-7i-3p MIMAT0004585 0.444 0.138 No Change 4.66E-01 miR-132-5p MIMAT0004594 0.667 0.103 No Change 5.42E-01 miR-142-3p MIMAT0000434 0.778 0.345 No Change 5.58E-01 miR-146b-3p MIMAT0004766 0.556 0.310 No Change 6.12E-01 miR-150-5p MIMAT0000451 0.667 0.172 No Change 6.39E-01 miR-31-5p MIMAT0000089 0.667 0.276 No Change 6.77E-01 miR-155-5p MIMAT0000646 0.778 0.310 No Change 6.82E-01 miR-142-5p MIMAT0000433 0.778 0.241 No Change 9.72E-01 miR-671-3p MIMAT0004819 0.444 0.000 No Change 9.82E-01      - 61 - 4.4 Discussion  The detection of altered disease fields reveals therapeutic implications. We report miRNA expression changes specific to different histological disease stages (Figure 4.2). In light of widespread efforts to create molecularly-targeted therapies, the reality that different disease-causing molecular alterations may exist across a disease field depending on histological stage shows that a given targeted therapy may not be effective across that entire field – thus setting the stage for post-treatment recurrence. Targeted therapies, therefore, would ideally be selected based on their capacity to be effective against as many cells in a field of diseased as tissue as possible. Based on our analyses, the most attractive miRNA candidates for downstream analysis as novel therapeutic targets would be those that exhibited altered expression in both dysplasia and CIS/OSCC groups relative to paired normal tissues.  MiR-155-5p represents such a candidate. As noted above, it has previously been described as having oncogenic function in several cancer types, including recent work in invasive oral cancer136-145. In the context of oral malignancies, miR-155-5p is understood to down-regulate the tumor suppressor CDC73. It is also  up-regulated by the TGFβ/ Smad pathway, which is activated in oral cancers 146,147. Based on our stringent patient-paired analyses, we detected elevated miR-155-5p expression in both dysplasia and CIS/OSCC groups (Figure 1 F-I). Using in situ hybridization of miR-155-5p, our TMA work corroborated these findings and confirmed them in an independent sample set.   Given emerging evidence from multiple cancer types that molecular subgroups of disease exist based on smoking status, we initially restricted our analyses to tissues from non-smokers to minimize the possible impact of sample heterogeneity. As outcomes in head and neck cancers generally  differ by smoking status.  Restricting analysis to a given subgroup is imperative.  Smokers and non-smokers with oral cancer may require different management and treatment strategies in   - 62 - order to see improved outcomes 148,149. When expression of our miR-155-5p candidate was evaluated in a TMA comprised of tissues from both smokers and non-smokers, it was found to be up-regulated in both groups and no correlation was observed between miR-155-5p expression and smoking status. Neither was a correlation noted between miR-155-5p expression and HPV status (a different etiological factor that could play a critical role in development of OSCC in non-smokers). A larger sample set is needed to confirm that no difference in miR-155-5p expression exists between groups based on these etiological factors (smoking, HPV). Also, it may be that other key miRNA candidates do exhibit different expression across these subgroups. Etiological factors should be an a priori consideration when designing a study to examine molecular drivers of disease.  Our work brings us to multiple key findings. First, novel imaging technologies capable of clearly delineating fields of cancerous tissue can be leveraged to facilitate stringent investigations into the molecular basis of cancer progression. Second, molecular heterogeneity across these fields of diseased tissue must be accounted for as we select targets for novel anticancer therapies, lest significant time and resources be invested in targeting candidates that are not causal to disease across an entire disease field. Third, miR-155-5p may represent an early and sustained driver of oral tumorigenesis in non-smokers – and perhaps in smokers as well. With human trials of anti-miRNA drugs like miraversin already underway, we will soon see such drugs developed for an oncology context150,151. Based on our results, we believe the impacts of field cancerization on miRNA expression should significantly inform these efforts.        - 63 - Chapter 5: Integrative Analysis of miRNA, Methylation and Gene Expression in Five Patients  5.1 Introduction In the previous chapters, we profiled a total of fourteen patient sample sets of adjacent normal, dysplasia and squamous cell carcinoma samples on DNA methylation microarrays (n=10), gene expression microarrays (n=13) and for 722 miRNAs using qRTPCR (n=9). These chapters focus on these modes of deregulation (DNA methylation and miRNA profiles) in cancer as separate entities.  The goal of this chapter is to integrate the data from the previous chapters and provide a detailed multidimensional analysis for the patient samples that have been profiled on all three data platforms (n=5).   For clarity, Table 5.1 describes the breakdown of the different groupings of patients that underwent analysis within this chapter.  Unfortunately, not all patients were profiled on all platforms due to insufficient tissue for DNA extraction (n=4) or input required for mRNA microarray profiling (n=1).  Table 5.1 shows an overview of the patient case numbers involved in the three sets of analyses: the MethExp (integrative analysis of methylation and expression data), MiRExp (integrative analysis of miRNA and expression data) and Integrated (analysis of patient samples with miRNA, methylation and expression data) cohort. To approach the integrative analysis for these data sets, I first performed separate correlation analyses on the MethExp and MiRExp cohorts in order to identify epigenetic events that are significantly inversely correlated with gene expression level.  I then assessed each patient from the Integrated cohort individually and identified a subset of genes that have a change in gene expression that is correlated with an inverse change in either miRNA status or   - 64 - methylation status.  From this narrowed list of potential candidate genes, I performed pathway and gene ontology analyses in order to identify overall trends within these patient samples.    Table 5.1  Data Analysis Grouping Scheme Cohort Name Platforms Analyzed Patients MethExp • Illumina 27K Microarray (14, 475 Entrez Genes)  • Agilent Gene Expression 4x44K (19,596 Entrez Genes) 1878 1953 3002 5266 7057 1858 1896 1979 1999 7059 MiRExp • Exiqon (722 miRNAs) Agilent Gene Expression   • Agilent Gene Expression 4x44K (19,596 Entrez Genes) 1878 1953 3002 5266 7057 2000 2076 7071 Integrated • Illumina 27K Microarray (14, 475 Entrez Genes)  • Agilent Gene Expression 4x44K (19,596 Entrez Genes)  • Exiqon (722 miRNAs) Agilent Gene Expression 1878 1953 3002 5266 7057 (Note: Patient 7058 was omitted due to lack of expression data)      - 65 - 5.2 Integrative Analysis of miRNA, Methylation and Expression Data Sets 5.2.1 Study Design A flow chart indicating the approach for the integrative analysis contained within this chapter is presented in Figure 5.1.   Confirmed targets of the miRNA found on the qPCR panel were extracted from the MirTarBase Release 6.1 list of miRNA and gene interactions that are supported by strong evidence (Reporter assay or Western blot).  This list was used as the basis for integrating the miRNA and mRNA data.  The fold change values for the all detectable miRNA in our dataset were correlated with the fold change values for the detectable mRNA microarray data.  The dysplasia and tumor sample sets were each analyzed separately and thus a list of significantly inversely correlated genes was generated for each stage. A paired Spearman correlation with 1000 permutations was performed in R.  After analysis, the significantly inversely correlated (p<0.05) miRNA and mRNA target pairs were identified in the five integrative patient data sets. The delta beta values for the CpGs on the Illumina microarray were correlated with the fold change values for the detectable mRNA microarray data as per annotation in the Illumina 27K microarray annotation for the CpGs that correspond with mRNA promoter regions.  Like the MiRExp cohort, a separate list of significantly inversely correlated genes was generated for the dysplasia and tumor data. A paired Spearman correlation with 1000 permutations was performed in R.  After analysis, the significantly inversely correlated (p<0.05) DNA methylation and mRNA target pairs were identified in the five integrative patient data sets. The lists of significantly inversely correlated genes for the MiRExp and MethExp cohorts for both dysplasia and tumor stages were then used to identify significant events found in  dysplasia and tumor sample from each patient in the Integrated cohort.     - 66 -  Figure 5.1 Flow chart of approach to integrative analysis in 5 patient samples          - 67 - 5.2.2 Statistical Analysis  Pathway and gene ontology enrichment analysis was performed using the ToppFun function from the ToppGene suite142.  Gene ontology: Biological Processes terms were categorized into broad terms using the ReviGO program152.  5.3 Results 5.3.1 Highly Inversely Correlated Events After performing a Spearman correlation analysis between in the MethExp cohort, we identified 335 and 964 genes as having a significant inverse correlation with methylation status in dysplasia and CIS/SCC samples, respectively.  Of the 335 highly correlated genes with dysplasia, 16 have both hypermethylated and gene silencing in >2/5 patients and two have both hypomethylation and increased gene expression in 2/5 patients.  Similarly, for the CIS/SCC samples, of the 964 inversely correlated genes, nine had both hypermethylated and gene silencing in >2/5 patients and 32 had both hypomethylation and increased gene expression in 2/5 patients. As for the MiRExp data analysis cohort, 96 and 83 mRNA targets were found to be inversely correlated with the expression of their respective miRNAs in dysplasia and CIS/SCC samples, respectively. In the dysplasia samples, three of these genes were found to have up-regulated miRNAs corresponding to down-regulated mRNA targets in 2/5 cases, and 33 were found to have down-regulated miRNAs corresponding to up-regulated mRNA targets in 2/5 cases.  In the CIS/SCC samples, six miRNAs were up-regulated and 36 were down-regulated.    - 68 - 5.3.2 Correlated Hypermethylated Genes are More Prevalent at the Dysplasia Stage In order to provide a thorough assessment of the epigenetic events of the patients with data from all three platforms, I took the list of highly inversely correlated genes and identified the genes in each patient sample that a) had a change in gene expression >2 fold, and b) had either a change in miRNA expression of >2 or a change of beta value >0.15.    Figure 5.2 A-B shows the proportion of methylation events in each of the five patients the dysplasia and CIS/SCC stage.  The overall trend appears to be more hypermethylated genes in the dysplasia stage with hypomethylation becoming more predominant in the more advanced lesions.  This is unsurprising, as this corresponds with the general trends of methylation events discussed in Chapter 3, indicating the global methylation events are reflective of the trends seen when integrating the data with gene expression.  The trends for miRNA events remain less clear.  In both dysplasia and CIS/SCC stages there appears to be more down-regulated events corresponding with an increase in the expression of predicted targeted mRNAs.       5.3.3 Frequently Deregulated Genes among 5 Patient Samples  In addition to assessing each patient sample for an overview of the epigenetic changes taking place in both dysplasia and CIS/SCC samples, I also looked across each sample to further our understanding in the overlapping trends among these patients.  Specifically, the goal was to see if the genes deregulated in this sample set were generally targeted by only one mechanism across all five samples, or if both methylation and miRNAs could be responsible for changes in gene expression in different patients.  However, the vast majority of genes were only targeted by one type of deregulation mechanisms at both dysplasia and CIS/SCC stages     - 69 -  Figure 5.2 Proportion of significant genes in each patient displaying deregulation of methylation or miRNA deregulation in a) Dysplasia and B)CIS/SCC (Figure 5.3 A-B). There were only two genes that had an overlap of methylation and miRNA   A B   - 70 - deregulation among the five patients, RUNX2 (silenced in the dysplasia stage) and ATG12 (silenced in the CIS/SCC stage).    5.3.4 Gene Ontology Analysis in Dysplasia and Tumor In order to view greater trends developing in our patient data, I analyzed the Biological Processes Gene Ontology (GO) terms that were highly prevalent in our data set.  For each patient, I performed an enrichment analysis using the ToppGene Suite for Biological Processes GO terms.  After, obtaining a list of significant terms for each of the five patient samples, the top 25 terms that had the highest significance among the patient samples in each stage of disease were inputted into a program called ReviGO to categorize the list of terms into broad categories in order to see overall trends in each stage.   Interestingly, there was a stark contrast in the epigenetic mechanisms at play in the comparison of these two disease stages (Figure 5.4).  At the dysplasia stage, the majority of GO  terms fell under the category “Positive Regulation of Signaling”.  The second largest category was “Localization of Cells”.  However, in the more advanced lesions, the category “Regulation of Cell Proliferation” became vastly more enriched (increasing from 12% of terms in dysplasia to 52% in CIS/SCC).  The second largest category was regulation of apoptosis, which was also found in the dysplasia stage although this increase was more modest from dysplasia (12% of terms) to the CIS/SCC stage (20% of terms).      - 71 -  Figure 5.3 Number of deregulated genes in at least 2/5 patients correlated with methylation, miRNAs or both in A) dysplasia and B) CIS/SCC.    - 72 -   Figure 5.4. Representative categories of Biological Process Gene Ontology terms recurrently enriched among Integrated cohort data analysis              - 73 - 5.3.5 Pathway Analysis Shows Different Mechanisms of Perturbation in Patients  In addition to identifying enrichment to GO terms, I wanted to assess any molecular pathways that may be perturbed in the data set.  In the dysplasia stage of disease, a significant enrichment for a number of general pathways, including the KEGG pathways “Pathways in Cancer” and “Hepatitis B” was found in 3/5 of the dysplasia samples.  Each patient was assessed for any genes that are listed as being involved in “Pathways in Cancer”, of which each patient had at least one gene (Figure 5.5; Table 5.2).  There appeared to be a spectrum of mechanisms where deregulation in major processes occurs. One patient exhibits a high number (9) of genes found within this pathway, all of which correlated with miRNAs, whereas another has a single gene deregulated within this pathway (which is correlated via methylation).    Within the CIS/SCC stage of disease, pathway enrichment analysis identified more specific pathways as significant.  One in particular was “Focal Adhesion”, which has been shown  to be involved in metastasis in tumorigenesis153.  As with the dysplasia genes, in the Focal Adhesion pathway there was also a spectrum of deregulation throughout the different mechanisms of deregulation and different areas of the pathway impacted in different patients (Figure 5.6, Table 5.3). Interestingly, BCL2 was appeared to be deregulated by miRNAs in two patients (Patients 7057 and 5266) and via methylation in another (Patient 3002).  Both patients with miRNA correlations had down-regulation of BCL2 (a tumor suppressor gene), but surprisingly Patient 3002 exhibits hypomethylation and up-regulation of BCL2.  Another interesting aspect is the overlap of genes deregulated among the patient samples.  Two patients had deregulation of both MET and IGF1R, however in the opposite directions.  One patient (1878) had up-regulation of this MET and down-regulation of IGF1R whereas the other (5266)   - 74 -   Figure 5.5  Genes deregulated in dysplasia samples in KEGG “Pathways in Cancer”.  Pathway interactions taken from  KEGG pathway hsa05200 “Pathways in Cancer”. Genes not deregulated have been omitted for clarity.           - 75 - Table 5.2 Genes Enriched in “Pathways in Cancer Signaling” in Dysplasia Samples   Mode of Deregulation  Gene  Expression Status  Name of CpG Probe or miRNA Patient 5266           miR  MYC  Up  hsa-miR-210-3p   miR  FOXO1  Up  hsa-miR-210-3p,  hsa-miR-218-5p,  hsa-miR-29a-3p   miR  MET  Up  hsa-miR-133a-3p hsa-miR-18b-5p hsa-miR-195-5p hsa-miR-200a-3p hsa-miR-34c-5p   miR  RB1  Up  hsa-miR-27a-3p hsa-miR-30e-5p hsa-miR-335-5p hsa-miR-34a-5p   miR  ETS1  Up  hsa-miR-17-5p   miR  BCL2  Up  hsa-miR-1-3p   miR  MMP9  Down  hsa-miR-34c-5p   miR  PTEN  Up  hsa-miR-140-5p hsa-miR-142-3p hsa-miR-195-5p hsa-miR-199a-5p   miR  CDK6  Up  hsa-miR-138-5p hsa-miR-141-3p hsa-miR-142-3p  Patient 3002           miR  MYC  Up  hsa-miR-376a-5p   miR  BCL2  Down  hsa-miR-137   miR  MMP9  Up  hsa-miR-34c-5p Patient 1858           CpG  FN1  Down  cg10692870   CpG  CSF3R  Up  cg07285167   CpG  SHH  Down  cg00577464   CpG  PRKCB1  Down  cg05436658   miR  E2F1  Up  hsa-miR-223-3p   miR  TP53  Down  hsa-miR-200c-3p   CpG  STAT3  Down  cg20716209   - 76 -  Mode of Deregulation  Gene  Expression Status  Name of CpG Probe or miRNA Patient 7057          CpG  RASSF5 Down cg02589695   CpG  CCND1 Down cg16794682          Patient 1953          CpG  CASP8 Up cg25095814  Table 5.3 Genes Enriched in Focal Adhesion Signaling in CIS/SCC Samples   Mode of Deregulation  Gene  Direction of Deregulation  miR or CpG Name  Patient 1878       MiR  FN1 Down hsa-miR-1-3p  MiR  SHC1 Down hsa-miR-27b-3p  MiR  MET Down hsa-miR-1-3p  MiR  THBS2 Down hsa-miR-27b-3p  MiR  IGF1R Up hsa-miR-140-5p  Meth BRAF Up cg02903525  MiR  MAP2K1 Down hsa-miR-34a-5p Patient 5266       MiR  BCL2 Down cg00983520  MiR JUN Up hsa-miR-139-5p  MiR CDC42 Up hsa-miR-195-5p  MiR  MET Up hsa-miR-562 Patient 7057       MiR  IGF1R Down hsa-miR-140-5p  Meth AKT3 Up cg13062935  MiR  MET Up hsa-miR-137  Meth BCL2  Up cg04189838 cg18390025   MiR ITGB1 Down hsa-miR-134-5p  MiR  PXN Up hsa-miR-137 Patient 3002      MiR  BCL2L10  Down cg18270343    - 77 -  Figure 5.6  Genes deregulated in SCC/CIS samples in KEGG “Focal Adhesion”.  Pathway interactions taken from KEGG pathway hsa04510 “Focal Adhesion”. Genes not deregulated have been omitted for clarity.    had the opposite (MET up and IGF1R down).  IGF1R in both samples were correlated with expression of hsa-miR-140-5p, however MET was correlated with different miRNAs (hsa-miR-1-3p in Patient 1878 and hsa-miR-137 in Patient 7057).       - 78 - 5.4 Discussion The goal of this chapter was to provide an in depth integrative analysis for the patient samples covered in this thesis that have been assessed for three levels of molecular analysis – miRNA expression, methylation patterns and gene expression profiles. Given the advent of high throughput genomic and epigenomic profiling and the lower input requirements now needed for microarray or sequencing analysis, there is an increasing amount of data that can be gathered from a single tissue sample.  Large scale analyses and databases of this type of information, such as through TCGA, are now providing us with a wealth of information regarding tissue specimens.  This wealth of information provides us not only with a high level of detail per platform, but allows us to integrate the data from each platform and find even more information regarding the tumor as a whole system.  This type of approach to analysis is beneficial for a number of reasons. One, it provides a means for eliminating genes that are deregulated as a consequence of genomic instability or other molecular processes gone awry in cancer instead of those that are playing active roles in tumorigenesis.  For example, in assessing the gene expression data alone from the five patient samples analyzed through integrative analysis in this chapter, we identify thousands of differentially expressed genes.  Assessing this data solely on this level does not allow for us to apply meaning to what is truly happening within these patient samples, as it is impossible to identify a gene that’s deregulation is crucial for cancer processes such as proliferation, progression or invasion, from one that has become deregulated as a bystander in this process.  By integrating with miRNA or methylation data, we are able to find genes that tie a mechanism to its deregulation.  In our data set, this allowed us to narrow our list of thousands of   - 79 - differentially expressed genes to a smaller, more meaningful list of candidate genes.  Of course aberrant methylation or deregulation of miRNA machinery can cause inconsequential genes to have become deregulated via these mechanisms as well; however, this assessment can be made by investigating the gene itself in in vitro and in vivo models and assessing functional impacts on tumorigenic processes.     In addition to providing a means for identifying more imperative events in tumorigenesis on a single gene basis, integrative analysis is essential for identifying trends in the system as a whole.  Due to our small sample size of 5 patients in this chapter, it was unlikely to find highly significant trends in specific pathways; however, through this analysis we are still able to appreciate the myriad of directions and mechanisms that generic pathways, such as the “Pathways in Cancer” can be deregulated among different patient samples.  By integrating significant data on a number of pathways, it enhances the ability to identify deregulated pathways or processes that may be hit once or twice per platform, but when taken each level into consideration can be identified as being hit more times.  This can be found on a small scale in this analysis, but other groups have assessed this premise in other types of cancer with much larger sample sets and have illustrated how this approach to analysis can create a much richer analysis of the mechanisms at play in tumorigenesis.  Although this analysis was performed on a small sample set, it still provided a number of intriguing results regarding the tumorigenesis process within these patient samples, indicating the importance of an integrative approach to genomic/epigenomic analysis.   - 80 - Chapter 6: Methylation Mediated Silencing of EYA4 in Oral Dysplasia  6.1 Introduction  DNA methylation within the promoter region of a gene has been linked with gene silencing. Within a normal cell, the function of this mechanism is to maintain the repression of genes that must remain unexpressed in that cell type at that time.  Proteins called DNA methyltransferases are responsible for maintaining the methylation patterns during subsequent replications of the genome and the cell59,154.  However, deregulation of these processes is highly prevalent in tumorigenesis.  Genes that are hypermethylated in tumors have been correlated with gene silencing, and many tumor suppressor genes have been found to exhibit promoter methylation mediated silencing155,156.   This is thought to occur by either physically inhibiting the binding of proteins essential for transcription, or by recruiting proteins that have transcription repressive properties.   A number of genes have been observed to be frequently methylated in oral cancer including: p16, MGMT, RUNX2 and DAPK86.  Hypermethylation of genes has been observed even in histologically normal tissue within the boundaries of an abnormal field, indicating that deregulation of methylation is an early event in oral tumorigenesis86.  The bulk of genes identified as being methylated in OPLs are genes already identified as tumor suppressor genes in the advanced stages of disease.    In the previous chapter we assessed DNA methylation throughout the different stages of OSCC.  In addition to profiling the methylation landscape of these lesions and assessing how methylation patterns change, we also wanted to identify novel tumor suppressor genes with gene expression controlled by DNA methylation.  We identified EYA4 as being the most recurrently hypermethylated in our dysplasia sample set, indicating a potential for this gene to be contributing to tumorigenesis in the earliest stages of this disease.   - 81 - EYA4, a protein that acts both as a transcriptional co-activator and as a protein phosphatase, has previously been shown to contribute to tumorigenesis via playing a role in proliferation, apoptosis and DNA damage repair 86,157,158.  This gene has been identified as a potential tumor suppressor gene in other cancer types in a variety of cancer types 159-162.This gene has yet to be implicated in oral cancer progression.   Due to the prevalence of EYA4 hypermethylation in our premalignant samples, in this chapter we investigate the role of EYA4 expression in tumorigenesis. We assess the level of methylation and correlation with gene expression in an expanded cohort of oral cancer tumor specimens from TCGA Head and Neck SCC sample set.  We also assess the impact of EYA4 expression on dysplasia cell lines by over-expressing EYA4 in two cell lines that have no detectable EYA4 expression.  The impact of this change on proliferation, anchorage independent growth, apoptosis and DNA damage repair is assessed.  A version of this chapter has been published in Genes, Chromosomes and Cancer with the following co-authors: D Truong and C Garnis.  6.2 Materials and Methods 6.2.1 Integrative Analysis of DNA Methylation and Gene Expression Microarray Data DNA methylation and mRNA microarray array data was obtained from ten patients with a paired adjacent normal, dysplasia and, either a CIS or OSCC biopsy as previously described 79. This study received ethical approval by the Research Ethical Board at the University of British Columbia (ID#: H10-01694) and all samples were collected with informed consent.  DNA methylation levels were profiled using the Illumina HumanMethylation27K microarray and gene expression was assessed using the Agilent 4x 44 K Human Gene Expression microarrays. A CpG was considered to be differentially methylated if the delta beta value (Δβ) was >0.15 compared to adjacent normal tissue. Similarly, a gene was considered to be differentially expressed if the change in expression was   - 82 - >2-fold compared to paired adjacent normal samples. An ANOVA test was used to determine if there was a significant difference between the methylation of EYA4 throughout the stages of oral cancer.   6.2.2 TCGA Data Analysis The Cancer Genome Atlas (TCGA) head and neck squamous cell carcinoma dataset from TCGA Research Network (http://cancergenome.nih.gov) was mined for data from patients with tumors arising from the oral cavity (using data from samples collected from the tongue, floor of mouth, buccal mucosa, oral cavity and hard palate).   DNA methylation (from the Illumina HumanMethylation450K microarray) and gene expression (RNA-Seq) data were analyzed. All data analysis was performed using GraphPad. The significance of differences between EYA4 CpG island methylation in paired normal and tumor samples was assessed using a paired t-test. Correlation analyses between RNA-Seq data and methylation microarray were performed on the tumor samples by Spearman correlation.  Gene set enrichment analysis (GSEA) was performed on TCGA oral tumor RNA-Seq data using the signal-to-noise metric and permutation tests using samples within the highest and lowest 25th percentile of EYA4 expression (n=146). This identified enrichment in curated gene set databases (including KEGG, Biocarta, Reactome, and GO datasets).   6.2.3 Cell Lines Two oral dysplasia cell lines, POE9n-tert and DOK, were used in this study: DOK from Sigma and POE9n-tert from Harvard Skin Disease Research Center. DOK was grown in DMEM media supplemented with 10% FBS and 5ug/ml hydrocortisone. POE9n-tert was grown using KSFM media with 25μg/ml bovine pituitary extract and 0.2ng/ml EGF. The 293T cells were used to produce   - 83 - lentivirus and were grown in DMEM media supplemented with 10% FBS. The 293T cells were a generous gift from Dr. Aly Karsan.   6.2.4 qPCR  RNA was extracted using the TRIzol protocol as per manufacturer’s instructions (Life Technologies). Any residual DNA was removed using a DNA-Free kit (Ambion) according to manufacturer’s instructions. RNA was converted to cDNA using the Applied Biosystems High Capacity Reverse Transcription kit. EYA4 Taqman gene expression assay was used with the Taqman gene expression master mix. qPCR reactions were performed on the Applied Biosystems Viaa7 machine. 18s was used as an endogenous control. The delta-delta-Ct method was used to determine fold change compared to a control.  6.2.5 Lentiviral Transduction Stable over-expression of EYA4 was performed using the plenti6.3 vector system (Life Technologies), as per manufacturer recommendations. EYA4-plenti6.3 vector was a generous gift from Dr. Wan Lam.  Lentivirus was produced in 293T cells using the Virapower lentiviral transformation kit per manufacturer directions (Life Technologies). After collection, viral supernatant was filter sterilized and stored at -80°C. Over-expression of EYA4 was induced by infecting the cells with viral supernatant and selected using blasticidin (1.5ug/ml) for two weeks  6.3 MTT Proliferation Assay The level of cell proliferation was measured over 5 days using a MTT assay. Two thousand cells were seeded into 96-well plates one day before the start of the experiment. 0.5 mg/ml MTT was added and incubated on the cells for 2-8 hours (dependant on cell line). At the end of incubation 20%   - 84 - SDS was added to lyse the cells. Optical densities were measured on the following day using an EMax plate reader (Molecular Devices). A Student’s t-test was performed on the data from the final day in order to determine if the results were significant.  6.3.1 Colony Formation Assay Colony formation assays were prepared by plating a bottom layer of 0.6% agarose in cell culture media and upper layer of 0.33% agarose in cell culture media. Each tested cell line had ~2000 cells plated in triplicate. Cells were allowed to grow for four weeks and colonies consisting of ≥15 cells were counted independently by two people.   6.3.2 Annexin V Apoptosis Assay Apoptosis was induced in the cell lines by exposing to cisplatin (20uM for 72 hours in POE9n-tert) or by serum starvation for 48 hours (DOK). Apoptosis was measured by using the BD Biosciences Annexin V Kit (BD Biosciences) as per manufacturer instructions. Flow cytometry analysis was performed using a FACscalibur flow cytometer (BD Biosciences) and data was analyzed using FlowJo software. A Student’s t-test was performed in order to determine the significance in change in apoptosis.    6.3.3 g-H2AX Analysis Cell lines were irradiated at 5 Gy and collected at various time points (0hr, 1hr, 2hr, 4hr, 6hr, and 8hr). The cells were collected and stained for Alexa-Fluor 647 anti-H2AX kit (BD Biosciences). Flow cytometry analysis was performed using a FACscalibur flow cytometer (BD Biosciences). Propidium iodide was used in order to gate the cells on G0/G1 phase. The change in intensity of g-  - 85 - H2AX was determined by comparing against unirradiated cells. Data was analyzed using FlowJo software.  6.3.4 Comet Assays DNA damage repair was assessed via comet assays. Briefly, cell lines were irradiated at 10 Gy, aliquoted out and incubated at 37°C until the appropriate time point (0 min, 7 min, 15 min, and 120 min). After the time point was reached, the cells were immediately put on ice and kept there until mounted on the slide. A total of 5000 cells were mixed with 1% low melting point agarose (Life Technologies), placed on glass slides and put in lysis buffer (2.5M NaCl, 0.1M EDTA, 10mM Tris, 1% Triton X-100, pH 10) overnight. The slides were then placed in a 0.3M NaOH + 1mM EDTA buffer, incubated at 4°C for 40 minutes and then run at 1.3V/cm for 30 minutes. After electrophoresis, the slides were placed in a neutralization buffer (0.4M Tris pH 7.5) for 10 minutes and then rinsed with distilled water for 10 minutes. The slides were fixed by submerging in 70% ETOH and then 100% ETOH for 15 minutes. The comets were stained with SybrGold (LifeTechnologies; 1 in 10,000 dilution) and visualized under 200x magnification using Olympus BX61 (Olympus America, Inc., Melville, NY) and ImagePro Plus 5.1 (Media Cybernetics, Silver Spring, MD). Comet scoring and analysis was done in ImageJ using the OpenComet plug-in and using Olive Moment as the measurement 163,164.   6.4 Results 6.4.1 EYA4 Hypermethylation and Gene Silencing is Observed in Dysplasia Samples of Oral Cancer Patients  Our previous analysis of genome-wide DNA methylation throughout paired dysplasia and CIS/OSCC samples identified EYA4 as exhibiting the highest frequency of promoter   - 86 - hypermethylation in dysplasia samples. This trend continued in the patient matched CIS or OSCC samples, with EYA4 methylation occurring in ≥1 measured CpG within the promoter region from all tumor samples (Figure 1). Interestingly, there were three patients (7059, 7057, and 1878) with particularly high DNA methylation within the EYA4 promoter in the dysplasia stage. These samples also had the highest overall level of DNA methylation out of all dysplasia samples79.  We also performed an unsupervised hierarchical clustering analysis of all biopsies and found only these three dysplasias clustered with the CIS/OSCC samples while the remaining dysplasia biopsies clustered with lesion-adjacent normal biopsies. This indicates that the level of EYA4 methylation may be indicative of the overall methylation landscape of the lesion. After integrating the methylation data with paired gene expression data from the same patient samples, we found that the majority of dysplasia samples with EYA4 hypermethylation had a decrease in gene expression by at least 2 fold (Table 6.1). In the CIS/OSCC stage, half of the samples with EYA4 hypermethylation had a ≥2-fold change in gene expression.  Table 6.1 Overview of EYA4 methylation and expression patterns in BCCRC Samples    Dysplasia (n=10)  CIS/SCC (n=10) Hypermethylateda 8 10 Under-expressedb 7 5 Hypermethylated & Under-expressed within same patient sample 6 5 aDelta beta value of >0.15 compared to adjacent normal tissue in at least one CpG within promoter region of EYA4 bFold change of >2 fold compared to adjacent normal tissue in at least one EYA4 probe      - 87 -       Figure 6.1 Mean beta value for CpGs found within CpG island in promoter region of EYA4 across paired patient samples (n=10).  NM: Adjacent normal; DYS: Dysplasia; CIS/OSCC: carcinoma in situ/oral squamous cell carcinoma.  Significant determined by one way ANOVA with post-hoc Tukey HSD (Honestly Significant Difference).  *: p<0.05; *** = p< 0.001.            - 88 - 6.4.2 EYA4 Methylation is Correlated with Reduced EYA4 expression To determine if the observed hypermethylation and gene silencing of EYA4 was present in a larger cohort of oral cancer samples, we mined publicly available TCGA data set for oral cavity specific samples with both DNA methylation microarray data and RNA-Seq data. A significant increase in methylation of EYA4 in tumor tissue compared to paired normal was observed (n=32, Figure 6.2A) and a delta beta value of >0.15 in 24/32 (75%) of tissue samples compared to paired normal tissue was observed. Additionally, we found a highly significant correlation between increased methylation of EYA4 CpG island methylation and decreased gene expression (p<0.0001) (Figure 2B). Neither EYA4 hypermethylation nor low EYA4 expression was correlated with survival, smoking, or HPV status in the TCGA cohort. We also compared the status of EYA4 in oropharyngeal TCGA samples. The level of EYA4 expression was similar in both oral (n=238) and oropharyngeal (n=167) tumors, however the level of methylation in oral tumors was significantly higher (p = 1.28E-06), indicating that EYA4 expression may be controlled by an alternative mechanism in oropharyngeal cancer. The sum of the data presented thus far support the characterization of EYA4 as a tumor suppressor gene in OSCC that may play a role in early tumorigenesis.   As EYA4 is a transcriptional co-activator, we performed a gene set enrichment analysis to identify pathways enriched in TCGA samples with the lowest EYA4 expression (<25th percentile of all oral tumor data sets) compared to the highest (>25th percentile). As expected, we found an enrichment of pathways involving DNA damage repair (including DNA repair, nucleotide excision and repair and DNA damage response), indicating a deregulation of these processes in patients with low EYA4 expression.  In addition to the expected pathways we also observed that in the tumor samples with low EYA4 expression, there was a striking enrichment for Reactome pathways involved in RNA processing (Table 6.2). These molecular processes have not previously been linked to EYA4 signaling.    - 89 -  6.4.3 EYA4 is Silenced in Oral Dysplasia Cell Lines To determine if EYA4 is involved in the early stages of oral cancer, we assessed its status in two oral dysplasia cell lines, POE9n-tert and DOK 160,165. POE9n-tert is a severe oral dysplasia cell line that is immortalized through induced expression of hTERT. DOK is a mild oral dysplasia cell line that was immortalized without induced expression of transgenes. qPCR analysis of both cell lines revealed no detectable expression of EYA4. Methylation-specific PCR DNA from on the cell lines indicated that DOK exhibits methylation at the EYA4 promoter region of interest, but POE9n-tert was unmethylated. Treatment of 5’azacytidine on DOK was not sufficient to induce expression of EYA4. As both cell lines have EYA4 silenced, we over-expressed EYA4 in both cell lines using lentiviral vectors. We also attempted to knockdown EYA4 in two distinct normal oral keratinocyte cell lines (OKF4tert1 and OKF6tert1) that have expression of EYA4.  We found that loss of this gene resulted in cell death in these lines.  6.4.4 Over-expression of EYA4 Impacts Cellular Proliferation, Apoptosis and DNA Damage Repair in Oral Dysplasia Cell Lines An MTT assay was performed to assess the impact of EYA4 expression on the proliferation of the two oral dysplastic cell lines. Both DOK and POE9n-tert cell lines with EYA4 over-expression had a significant decrease in cellular proliferation (Figure 3 A, B). Although over-expression of EYA4 was found to significantly inhibit proliferation in both cell lines, it was not sufficient to induce colony formation in soft agar (data not shown).     - 90 -  Figure 6.2 Analysis of TCGA Oral Cavity Patient Samples.  A) Mean methylation (β-value) for CpGs found within CpG island in promoter region of EYA4 across paired TCGA normal and tumor samples (n=32).  Statistical significance determined using the student’s t-test. B)  Correlation between TCGA DNA methylation and RNA-Seq log2 data for oral cavity tumors (n=238).  Statistical significant determined using the Spearman correlation analysis (r = -0.3368).          - 91 - Table 6.2 Gene set enrichment analysis results for signaling pathways enriched in TCGA oral cancer with low EYA4 expression   Name Database NES p-val FDR q-val METABOLISM OF RNA Reactome 1.92 0.00E+00 4.60E-02 ACTIVATION OF THE MRNA UPON BINDING OF THE CAP BINDING COMPLEX AND EIFS AND SUBSEQUENT BINDING TO 43S Reactome 1.88 2.06E-03 5.86E-02 METABOLISM OF MRNA Reactome 1.88 2.07E-03 4.00E-02 INFLUENZA LIFE CYCLE Reactome 1.88 2.07E-03 3.27E-02 RNA POL I PROMOTER OPENING Reactome 1.87 2.02E-03 2.95E-02 3 UTR MEDIATED TRANSLATIONAL REGULATION Reactome 1.84 2.05E-03 3.86E-02 NONSENSE MEDIATED DECAY ENHANCED BY THE EXON JUNCTION COMPLEX Reactome 1.84 0.00E+00 3.34E-02 FORMATION OF THE TERNARY COMPLEX AND SUBSEQUENTLY THE 43S COMPLEX Reactome 1.84 2.05E-03 2.96E-02 RNA POL I TRANSCRIPTION Reactome 1.83 2.07E-03 3.11E-02 TRNA AMINOACYLATION Reactome 1.83 2.06E-03 2.82E-02 DEPOSITION OF NEW CENPA CONTAINING NUCLEOSOMES AT THE CENTROMERE Reactome 1.82 0.00E+00 3.04E-02 CONDENSED NUCLEAR CHROMOSOME Gene Ontology 1.79 1.97E-03 4.62E-02 S PHASE Reactome 1.78 0.00E+00 4.78E-02    - 92 -  Name Database NES p-val FDR q-val METABOLISM OF NON CODING RNA Reactome 1.77 2.03E-03 5.14E-02  Deregulation of EYA4 was reported to have an impact on apoptosis in other types of cancer cells. We assessed the impact of EYA4 over-expression on the dysplasia cell lines using an Annexin V assay. POE9n-tert-EYA4+ significantly elevated the level of apoptosis compared to the empty control vector cells (POE9n-Ctrl), however this effect was not observed in the other dysplasia cell line (Figure 3 C,D). There was no significant difference in the number of DOK EYA4+ cells undergoing early apoptosis compared to the control cells, but there were significantly less cells in late apoptosis/necrosis. This discrepancy could be due the specific genomic backgrounds of these two lines. DOK cells in general were resistant to apoptosis (after inductions using both cisplatin and serum-starvation) as evidenced by the higher level of live cells. These results indicate that EYA4 can contribute to apoptosis in dysplastic oral cells in some instances.  EYA4, in addition to being a transcriptional co-activator, also exhibits phosphatase activity. Specifically, it has been found to dephosphorylate Tyr-142 on H2AX, a histone variant that is recruited after DNA damage occurs. This dephosphorylation event results in the phosphorylation of Ser-139 on H2AX and triggers the recruitment of DNA damage response proteins. To test the impact I further analyzed the role of EYA4 in cancer via publicly available data and functional analyses in two oral dysplasia cell lines.      - 93 -      Figure 6.3  Proliferation and apoptosis assay results for POE9n-tert and DOK cell lines.  MTT proliferation assay results for A) POE9n-tert and B) DOK over five days.  Statistical significance determined by t-test on the values for the last day of experiment.  Annexin V apoptosis assay for POE9n-tert (C) and DOK (D).  Experiments performed in triplicate. Statistical significance determined by t-test. NS = Not significant.     - 94 -  Figure 6.4 Assessment of DNA damage repair.  (A, B) gH2AX levels after 5 Gy irradiation in A) POE9n-tert and B) DOK over 8 hours. (C, D) Comet assay results for POE9n-tert (C) and DOK (D) to measure DNA damage after 10 Gy irradiation.  Statistical significance is denoted with an asterisk.  (E, F) Relative DNA damage repair compared to level of DNA damage occurring at 0 min after 10 Gy irradiation.     - 95 - 6.5 Discussion  The key to improving survival rates for OSCC will partially lie in the ability to differentiate patients at risk for progression at the premalignant stage. To accomplish this, a deeper understanding of the molecular events that drive this disease is paramount. We previously identified EYA4 as one of the most frequently hypermethylated and silenced genes in oral dysplasia patient samples. Herein, we   EYA4 has been implicated in a variety of other types of cancer, as both a potential tumor suppressor and an oncogene. It has proposed to play an oncogenic role in peripheral nerve sheath tumors 166 and to be a potential oncogene in breast cancer due to the effect of over-expression of EYA1, EYA2 and EYA3 on breast cancer cell lines 167.  EYA4 has predominately been reported as hypermethylated and/or silenced in a variety of cancer types, including lung, colorectal, esophageal and pancreatic cancers91,149,159,161,168. Furthermore, several groups report that loss of EYA4 can occur at the premalignant stage of some types of cancer, including lung carcinoma in situ, ulcerated colitis-associated dysplasia, and Barrett’s esophagus – indicating a potential for its role in the early stages of cancer progression91,157,159,161.   The proteins of the EYA family appear to play a complex role in development and disease, and were the first example of a transcription factor that also exhibits phosphatase activity 169-171 . The specific functions that EYA4 has reported to impact include DNA damage repair, proliferation, apoptosis, and differentiation 158,159,162,167. EYA4 has been implicated in DNA damage repair through its interaction with H2AX, a histone variant of H2A, which enhances the recruitment of DNA damage repair factors to the site of double stranded DNA breaks.  EYAs dephosphorylate Tyr-142 on H2AX, which subsequently promotes and maintains the phosphorylation of Ser-139 on H2AX (becoming g-H2AX) 172.   The role EYA4 plays in apoptosis is more complex as results are conflicting: some have reported it as an inducer of apoptosis through its role as a transcriptional co-  - 96 - activator and others as having an anti-apoptotic role through its role as a tyrosine phosphatase on H2AX 158,159,162. Over-expression of EYA4 in dysplasia cell lines increases the level of g-H2AX as well as the efficiency of DNA damage repair after irradiation (as measured by comet assays).  This suggests that silencing of EYA4 may contribute to OSCC progression through its contribution to the diminished ability of the cell to repair double stranded DNA breaks.   Additionally, over-expression of EYA4 also caused a significant decrease in proliferation in both cell lines indicating this gene can impact proliferation in cell lines.  We found a significant increase in apoptosis in one dysplasia cell line, indicating EYA4’s role in apoptosis may not only be specific to a certain disease type, but also specific to each individual tumor. Although there was no obvious reason for the difference in apoptosis between these two cell lines, it was observed that DOK was more resistant to apoptosis in general. Neither serum starvation nor cisplatin treatment was sufficient to induce the levels of apoptosis seen in the more responsive dysplasia cell line, POE9n-tert. Given EYA4’s role in DNA damage response and apoptosis, its inactivation during tumorigenesis may contribute to inducing genomic instability through inefficient DNA damage repair and lack of apoptosis of these damaged cells. It is well documented that genomic instability increases as lesions develop through the premalignant stage to the invasive stage, and loss of EYA4 may play a role in this process in OPLs.  There was there was no significant impact of EYA4 on proliferation, apoptosis, and DNA damage repair in the tumor cell line CAL-27.  This indicates that the silencing of EYA4 must occur early in tumorigenesis.  Given the high level of genomic instability and molecular aberrations occurring in tumor cell lines, these cells already possesses mechanisms to subvert apoptosis, increase   - 97 - proliferation and response to DNA damage.  Thus, it is unsurprising that the change of one gene would not have a major impact in these factors.   We attempted to restore EYA4 expression in the DOK cell line by demethylating the genome of the cell line using 5’azacytidine.  However, we found this to have no effect on EYA4 expression.  This may be due to the fact that removal of  the methyl groups is not sufficient to induce the expression of EYA4 alone and other transcription factors or other proteins may have been required to induce expression.   A gene set enrichment analysis of the oral cavity TCGA data comparing tumors with EYA4 expression in the top and bottom 25th percentile revealed a strong enrichment for Reactome pathways involved in transcription and translation processes. The link between loss of EYA4 expression and these processes is not clear. This could be due to the role of EYA4 as a transcriptional co-activator or through its phosphorylation activity on H2AX. However, up-regulation in these pathways has been shown to contribute to tumorigenesis and thus may be another mechanism for EYA4’s involvement in tumorigenesis173-175. In conclusion, the OSCC squamous cell carcinoma, and appears to play a role in tumorigenesis via the effects of the loss of EYA4 on the DNA damage repair process and potentially in apoptosis. To determine if the inactivation of EYA4 contributes to the progression of OPLs into invasive tumors, long-term prospective studies must be done on patients that have not yet reached the invasive stage of disease. Further studies should also be performed to further investigate the connection between EYA4 silencing and the enrichment of RNA related pathways.    - 98 - Chapter 7: Focal Amplification of 9p13 in Oral Premalignant Lesions 7.1 Introduction Genomic instability, a hallmark of most epithelial tumors, can drive activation of oncogenes and silencing of tumor suppressor genes via changes in DNA dosage176,177. While many critical DNA amplifications and deletions have been uncovered for invasive tumor tissues, few genomic studies of earlier precancerous lesions have been undertaken. This represents a missed opportunity to identify the molecular origins of disease, as causal genetic alterations are more likely to be discovered through analysis of premalignant tissues. To date, few molecular analyses of OPLs have identified genetic markers to help determine the risk of progression from pre-malignancy to invasive cancer28,178-184.  Loss of chromosome 9p has been reported as both a recurring event and an alteration predictive of increased risk for disease progression in OPLs 28,179,185,186.  However, these studies have typically relied on loss of heterozygosity (LOH) analysis of only a small number of microsatellite markers on chromosome 9p. Furthermore, these studies often excluded analyses of lower-grade OPLs (mild/moderate dysplasia), which is necessary to glean insights into disease initiating molecular changes. This following chapter describes a DNA gain at 9p13 that recurs in low-grade OPLs that subsequently progress to invasive oral cancer, yet is absent in non-progressing low-grade OPLs. Significantly, this alteration is more frequent in progressing low-grade OPLs than canonical loss of chromosome 9p. Also, we provide functional data indicating that this DNA gain spans three potential oncogenes in oral carcinogenesis. This finding suggests that a single, focal DNA copy number gain may be deregulating multiple genes in concert in order to drive progression to invasive oral cancer.  A version of this chapter has been published in Cancer Medicine with the following co-authors: IF Tsui, Y Zhu, S MacLellan, CF Poh and C Garnis.   - 99 -  7.2 Materials and Methods 7.2.1 Whole Genome Characterization of DNA Copy Number Alterations in OPLs This study evaluated 64 OPLs (43 high-grade dysplasias, 7 low-grade dysplasias that later progressed to invasive disease, and 14 low-grade dysplasias that did not progress) for DNA copy number alteration status on chromosome 9p. Previously described whole genome tiling-path array comparative genomic hybridization (CGH) data were used to perform this analysis182. All profiles have been deposited to NCBI Gene Expression Omnibus (GEO), series accession number GSE9193.   7.2.2 Gene Expression Profiling Analysis Expression of the 68 genes mapped within the recurring 9p13.3 DNA gain was evaluated using previously published data with the gene expression profiles of ten tumor samples and their paired adjacent normal tissue (GEO Accession number GSE46802) 79.   7.2.3 Accrual of Fresh Oral Tumor Tissues for DNA Copy Number and Gene Expression Analyses Four fresh frozen tissues were obtained immediately following surgical resection in the operating room53. Collected tissues were micro-dissected based on pathologist guidance. Extracted DNA was hybridized to a whole genome tiling-path CGH microarray 187,188 and corresponding extracted RNA was analyzed by Agilent Whole Human Genome Microarray 4 X 44K (Agilent Technologies, Mississauga, ON). Sample labeling, hybridization, and scanning of the tiling-path CGH arrays was performed as previously described182,188. Labeling and hybridization experiments for the Agilent 4 X 44k array – which interrogates > 41,000 unique human transcripts – were performed according to manufacturer protocols. These gene expression arrays were scanned using Axon   - 100 - GenePix 4000B and 4200A scanners.  Median normalization was performed on Agilent Whole Human Genome Microarray as previously described189. All data are publicly available on GEO (GSE46802).  7.2.4 Statistical Analysis of Genomic Profiles A three-step normalization procedure was used to remove systematic biases on tiling-path CGH arrays as previously described190,191. SeeGH software was used to display log2 signal intensity ratios in relation to genomic locations in the hg17 assembly (NCBI Build 35)190.  Data points with standard deviation >0.075 and signal to noise ratio <3 in either channel were filtered from downstream analyses.    7.2.5 Validation of Gene Candidates in Tissue Microarrays (TMA) An independent validation set of samples consisting of premalignant archival patient tissues from the British Columbia Oral Biopsy Service was used to construct a tissue microarray. Thirty-seven mild and moderate dysplasia cases with known progression status and 26 severe dysplasia and CIS specimens were used for assembly of the TMA (all cases were formalin fixed and paraffin embedded [FFPE]). Patient demographic information is available in Appendix A.2192.  Briefly, one 1-mm core from a represented area was obtained from each paraffin specimen and distributed on recipient TMA blocks using a specific arraying device (Manual Tissue Arrayer MTA-1, Beecher Instruments, Inc. WI, USA). A single 5-μm section was then cut from the TMA block and used for in situ hybridization analysis as previously described 193. One set of 3-colored probes (Vysis, Downers Grove, IL) was performed on the 5-µm tissue sections according to manufacturer instructions, which included CEP9 probe (centromere, SpectrumGreen), 9p13 probe (SpectrumAqua), and 9p21 probe (SpectrumOrange). Signals were captured and imaged using Olympus BX61 and ImagePro Plus 5.1.   - 101 - At least 150 non-overlapping intact nuclei were scored. Samples with >90% nuclei showing signals were considered informative. Sample signals were scored and classified as deletions if >50% of nuclei showed ≤1 signal and as gained or amplified if >10% of nuclei showing ≥4 signals 194.   7.2.6 Cell Culture and Reagents Six oral cancer cell lines (SCC-4, SCC-9, SCC-15, SCC-25, A253, and Cal27) were purchased from the American Type Culture Collection (ATCC). The oral dysplasia cell line POE9n-tert was purchased from the Harvard Skin Disease Research Center Cell Culture Core. DNA was analyzed using whole genome tiling-path CGH arrays and RNA was analyzed by Agilent Whole Human Genome Microarray 4 X 44K. SCC-9 and Cal-27 tongue cell lines were chosen for knock-down and over-expression experiments due to their genomic and expression profiles189. They were maintained according to distributor recommendations. POE9n-tert, a premalignant oral mucosal keratinocyte cell line, was chosen for over-expression experiments due to its genomic and expression profile and was grown at 37°C using keratinocyte serum-free medium supplemented with L-Glutamine, bovine pituitary extract, and epidermal growth factor (Life Technologies, Life Technologies, Gaithersburg, MD). The 293T cells were a generous gift from Dr. Aly Karsan and were cultured in Dulbecco's Modified Eagle's medium with 10% fetal bovine serum at 37°C.  7.2.7 shRNA Lentiviral Vector Knock-down Human pLKO.1 lentiviral shRNA target gene sets were selected from the RNAi consortium (TRC) and were purchased from Open Biosystems (Huntsville, AL). For each of the genes, five shRNAs constructs were tested for knockdown efficiency and the two that showed the best knockdown were selected for further experiments to minimize off-target effects. To produce lentivirus, 293T cells were transfected with pLKO.1 plasmid construct coding a shRNA targeted for   - 102 - each gene candidate with the packaging plasmids VSVG and d8.91 using TransIT-LT1 transfection reagent (Mirus, Mississauga, ON). pLKO.1 empty vector was also transfected into 293T cells to serve as a control. Viral supernatant was collected over two days post transfection, filter sterilized (0.45 µm), and stored in -80ºC.  After transfecting SCC-9 cells with each lentivirus for 24 hours, cells were selected with 2 µg/ml puromycin over three days. All non-transfected cells were dead within three days of selection, while stably transfected SCC-9 cells were effectively cultured in growth media containing 2 µg/ml puromycin.   7.2.8 Plasmid Construction and Viral Transduction Full-length human cDNA expression vectors were purchased from Open Biosystems for each gene candidate, including pCMV-SPORT6 for VCP (BC110913), pOTB7 for STOML2 (BC002442), and pOTB7 for DCTN3 (BC000319). Coding sequences were PCR-amplified and cloned into pLenti4/V5-DEST Gateway® Vector as per the manufacturer protocols (Invitrogen, Carlsbad, CA). Lentivirus supernatant was produced and collected as described above. Twenty-four hours following infection with lentivirus Cal-27 and POE9n-tert cells were selected using 10 µg/ml and 0.5 µg/ml Zeocin, respectively.   7.2.9 Real-time Polymerase Chain Reaction (PCR) of mRNA Expression Total RNA from stably-selected cell lines was extracted using TRIzol (Invitrogen, Carlsbad, CA) and treated with DNA-free DNase Treatment & Removal Reagents (Ambion, Austin, TX). High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA) was used to convert total RNA to cDNA. Real-time PCR using Taqman Universal PCR master mix was performed to analyze RNA expression levels with Applied Biosystems Standard Real-Time PCR   - 103 - systems. Taqman gene expression assays of DCTN3 (Hs00989657_m1), STOML2 (Hs00203730_m1), VCP (Hs00997650_m1), and 18s rRNA (Hs99999901_s1) were purchased from Applied Biosystems. Triplicate reactions were performed for each sample and standard error was calculated using ABI software. Relative expression values using the average of cycle thresholds of target genes and 18s rRNA were calculated using the 2-ΔΔCt method. A value of 1 was assigned to the control empty vector sample.   7.2.10 Western Blotting Cell lysates were harvested in Radioimmunoprecipitation assay buffer (RIPA) [150 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.5% Na-deoxycholate, 1 mM EDTA] with 10mM phosphatase cocktail I and cocktail II (Sigma-Aldrich, St. Louis, MO) and 1:100 of protease inhibitor (Invitrogen, Carlsbad, CA). The protein concentrations were determined using the Bicinchoninic Acid Protein assay kit (Thermo Scientific, Waltham, MA). A total of 20 µg of protein was separated by NuPAGE 4-12% Bis-Tris Gels (Invitrogen, Carlsbad, CA) and transferred to polyvinylidene difluoride membranes (Millipore). Membrane blocking was performed in 5% w/v nonfat dry milk, 1x TBS and 0.1% Tween-20 at room temperature with gentle shaking for one hour for polyclonal anti-DCTN3 primary antibody (Sigma-Aldrich, St. Louis, MO). Membrane blocking was performed in 5% BSA, 1x TBS, and 0.1% Tween-20 at 4˚C with gentle shaking overnight as recommended by the manufacturer before incubation with monoclonal anti-VCP (7F3) (Cell Signaling, Technology, Inc., Danvers, MA), and polyclonal anti-STOML2 (H-180) (Santa Cruz Biotechnology Inc, Santa Cruz, CA). After blocking, the membranes were incubated with the appropriate primary antibody anti-DCTN (1:1000) at 4˚C with gentle shaking overnight, anti-VCP (1:1000) at 4˚C with gentle shaking for 3.5 hours and anti-STOML2 (1:1000) at 4˚C with gentle shaking for 1 hour. After washing, the membranes were incubated with peroxide-conjugated anti-mouse secondary antibody (GE Health   - 104 - care, Buckinghamshire, UK) and HRP-linked anti-rabbit secondary antibody (Cell Signaling Technology, Inc., Danvers, MA) at (1:2000) at room temperature for 45 minutes. Anti-β-actin antibody (Cell Signaling Technology, Inc., Danvers, MA) (1:1000) was used as loading control. Proteins were detected with the enhanced Amersham ECL Western Blotting detection kit (GE Healthcare, Buckinghamshire, UK).  7.2.11 MTT Cell Viability Assay Stably-selected cells with confirmed knockdown or over-expression efficiency were plated in five 96-well plates at a density of 1000-2000 cells per well. The number of viable cells was followed for five days. Colorimetric thiazolyl blue tetrazolium bromide (Sigma-Aldrich, Canada) was added to each well (final concentration 0.5mg/ml). For each plate, the cells were repeatedly plated in six wells and 570 nm absorbance was measured (with reference to 650 nm) using an EMax plate reader (Molecular Devices, Sunnyvale, CA). The mean of the absorbance was plotted against time and standard error of the mean was plotted as error bars. Statistical analysis was carried out using Student’s t-test on day 5 and p < 0.05 was used as a cut-off for statistical significance.   7.2.12 Soft Agar Colony Formation Assay Bottom layer agarose was made to 0.5% in a 12-well plate. The top layer was made with 2000 stably-infected cells in 0.37% agarose using low-melting point agarose. The number of colonies per plate was counted for both the infected cell line and the empty vector control. All experiments were performed in triplicate wells with two biological replicates. Colonies, which consisted of approximately 15 cells, were counted by two independent observers after growing the colonies for four weeks.     - 105 - 7.3 Results 7.3.1 Identification of Recurring DNA Gain at 9p13.3 in Progressing OPLs Our group previously reported that a high degree of global genomic imbalance is associated with low-grade (mild and moderate dysplasia) OPLs that subsequently progressed to invasive cancer (relative to non-progressing low-grade OPLs)178. Within these data, genomic imbalance at chromosome 9p was found to be the most frequently occurring event in progressing low-grade OPLs, with incidence of ~80%. Low-grade OPLs that did not subsequently progress to invasive disease did not harbor genetic alteration at chromosome 9p13. Interestingly, genetic gain at 9p13 occurs more frequently in early OPLs than other canonical early genomic alterations for low-grade OPLs such as loss of chromosome 9p21 or chromosome 3p (78% vs. 56% and 22%, respectively). Using a tissue microarray consisting of an independent validation set of 37 low-grade OPLs (including 23 cases that were known to later progress), we validated our findings by FISH analysis for alteration at 9p21 and 9p13 (Fig. 7.1A-B). Among all low-grade lesions (progressing and non-progressing), 11 cases showed a gain of 9p13 with 6 having normal DNA copy number at 9p21 and 1 exhibiting 9p21 loss. Among low-grade OPLs, 12 of 23 progressing low-grade dysplasias showed abnormal DNA copy number involving at least one analyzed locus while only one of 14 non-progressing OPLs exhibited abnormal DNA copy number at either of these two loci (p = 0.01). Four of 12 (33%) progressing cases showed a gain of all three analyzed loci, suggesting a whole chromosome arm gain. Among progressing cases exhibiting DNA copy number changes by FISH, 10 of 12 cases showed gain of 9p13, while two cases showed deletion of 9p21. Analysis of demographic data for all low-grade OPLs did not show 9p13 gain correlating to age, sex or smoking status.  We also examined the frequency of 9p13 and 9p21 alterations in a panel of 26 high grade OPLs (severe dysplasias and CIS cases) and found only 2 cases with gain of 9p13 and deletion of 9p21 and 3 cases showing no change of 9p13 but deletion of 9p21.      - 106 -       Figure 7.1 A) Detection of 9p13 gain using FISH analysis. A representative tissue microarray of a patient with 9p13 (blue) amplification and normal copy number for 9p21 (red) and Cep-9 (green), shown at 60x magnification. B) 9p13, 9p21 and Cep-9 FISH results of tissue microarrays for 14 non-progressing low-grade lesion, 23 progressing low-grade lesion and 26 high grade lesions. N: normal copy number, +: gain, ++: amplification, -: deletion.                   - 107 - Mapped more specifically, the recurrent high-level DNA copy number gain in progressing OPLs occurs at 9p13.3 (bp 33,751,040 – 36,241,407 [hg19]), which is 2.49 Mbp in size and spans 58 unique RefSeq genes (Fig. 7.2).  We reviewed publicly available gene expression data derived from ten oral cancer tumors to hone this gene list and identified possible oncogenes “driving” the emergence of this recurrent gain79. Candidate genes were selected based on 1) increased expression in tumor samples (relative to normal comparators), 2) previous implication as oncogenes in other cancer types, and 3) plausible positive impacts on tumorigenicity when over-expressed. Seven of 58 genes mapping to 9p13.3 in this dataset were found to be ≥2-fold over-expressed in oral tumors in ≥6 of 10 patients. Of these, three genes have been previously implicated as being over-expressed in malignant processes: Valosin-containing Protein (VCP), Stomatin-like Protein 2 (STOML2), and Dynactin 3 (DCTN3)195-201. We selected these candidates for further study.   7.3.2 Inhibition of Gene Candidates Within the 9p13.3 Amplicon Diminishes Oral Cancer Phenotypes The functional significance of VCP, STOML2, and DCTN3 in oral cancer phenotypes was evaluated by shRNA-mediated knockdown experiments using oral cancer cell line SCC-9. SCC-9 cells were chosen due to the existence of high-level amplification of the 9p13 region (bp 34,085,033-35,428,177 to bp 35,432,063-36,539,166), which corresponds to the 9p13 DNA gain we observed in clinical samples189. To minimize off-target effects, the knockdown efficiencies of five shRNAs for each gene were tested. The two shRNAs demonstrating the most effective knockdown were used for further experiments. The reduced expressions of candidate genes were confirmed at both transcript and protein levels (Fig. 7.3). In summary, our results indicated that knockdown of each of the three candidate genes resulted in reduced cell proliferation rates, with inhibition of VCP and DCTN3    - 108 -       Figure 7.2 Alignment of genomic alteration data of five low-grade oral premalignant lesions (OPLs) reveals copy number gain at chromosome 9p13. Log2 signal intensity ratios of each competitive hybridization with normal reference DNA are plotted. Each black dot represents the log2 signal intensity ratio of an individual clone mapping to the array. Sample identifiers are listed at the top of genomic plots. Vertical lines represent log2 signal intensity ratios of +1, +0.5, 0, -0.5, and -1. Horizontally highlighted regions indicate the location of the CDKN2A tumor suppressor gene and the minimal altered region of the 9p13 amplicon. All known RefSeq genes within the minimal region of alteration are indicated on the right. The sizes of the three candidate genes are not drawn to scale        - 109 - causing the most dramatic reductions in SCC-9 cell growth (Fig. 3 A-D). Additionally, independent knockdown of each candidate gene in SCC-9 cells also decreased anchorage-independent growth in soft agar (Fig. 7.3D).   We also attempted to knockdown these candidate genes in the cell line DOK, an oral dysplasia cell line that also possesses a DNA gain at 9p13. However, in multiple experiments, the loss of the candidate gene resulted in the death of cells before any additional experiments could be performed. The control-vector cells remained viable.  7.3.3  Over-expression of Candidate Genes Enhances Cell Proliferation and Anchorage-Independent Growth In evaluating the ability of each gene candidate to promote growth in oral cancer, we focused on cell line Cal-27, an oral cancer cell line that is DNA copy number neutral at chromosome 9p13. We created Cal-27 sub-lines stably over-expressing each gene candidate using lentiviral constructs. Over-expression of each gene candidate in the stable clone was verified by qRT-PCR and Western blotting (Fig. 7.4). Over-expression of VCP – measured as a 35-fold mRNA increase relative to control (empty vector) infected Cal-27 cells – resulted in the most significant increase in cell proliferation (p=5.7x10-5; Fig. 7.4 A-D). VCP-expressing cells also showed a dramatic increase in the number of colonies formed in soft agar (Fig.7.4D). Increased colony formation was also observed in STOML2- and DCTN3-over-expressing cells, indicating that these candidate genes have a role in enhancing anchorage-independent growth as well.  We also assessed the ability of the candidate genes to affect proliferation in premalignant cells (Figure 7.5). We over-expressed each candidate gene in POE9n-tert, an oral dysplasia cell line that does not possess the 9p13 DNA copy number gain. As with Cal-27 tumor cell lines, each candidate gene caused a significant increase in the proliferation of the candidate gene expressing cell    - 110 -  Figure 7.3 Candidate gene silencing contributes to decreased cell proliferation in SCC-9 cells (an oral cancer cell line harboring the recurrent 9p13 amplicon). Panel (A) presents VCP, (B) DCTN3, (C) STOML2. Each panel contains 3 experiments performed for each gene, presenting images of, knockdown efficiencies (qRT-PCR) for five different shRNAs against the candidate gene, protein expression of two best knockdowns (Western blots), and MTT proliferation results over a five-day period of the two confirmed knockdowns. (D)  shRNA stable knockdown of each candidate gene in SCC-9 cells decreases colony formation in soft agar relative to control SCC-9 cells. The two most effective shRNAs for each candidate gene are shown. Triplicate experiments were performed for each line. The mean numbers of colonies and standard deviations are plotted.            - 111 -     Figure 7.4  Stable over-expression of STOML2 (A), VCP (B), and DCTN3 (C) increases cell growth in Cal-27 cells (an oral cancer cell line not exhibiting the 9p13 amplicon). MTT proliferation results over a five-day period are shown. Each panel for each gene depicts 1) an MTT cell proliferation plot for both control and test lines, 2) qRT-PCR results demonstrating increased expression of a given candidate gene in test line, and 3) Western blots confirming concordant protein over-expression of test line. (D) Stable over-expression of candidate gene in Cal-27 cells increases colony growth in soft agar relative to control Cal-27 cells. Experiments were performed in triplicate. Mean number of colonies and standard deviations are plotted.      - 112 -  Figure 7.5 Stable over-expression of STOML2 (A), VCP (B), and DCTN3 (C) increases cell growth in the dysplasia cells line, POE9n-TERT (an oral dysplasia cell line with no 9p13 amplicon). MTT proliferation results over a five-day period are shown. Each panel for each gene depicts 1) an MTT cell proliferation plot for both control and test lines, 2) qRT-PCR results demonstrating increased expression of a given candidate gene in test line, and 3) Western blots confirming increased amount of protein of test line.     - 113 - lines compared to the empty vector control cell line (VCP: p=1.34 x10-4, DCTN3: p=1.16x10-4, STOML2: p=2.14x10-5; Fig. 5).  7.3.4 9p13 Amplification Within Multiple Biopsies From a Single Patient  Oral tumors have been known to exhibit a high degree of molecular heterogeneity30,202,203. Genomic profiling and the detection of shared DNA alteration boundaries have been used to delineate clonal relationships and identify stage-specific genetic alterations for a variety of cancer types202,204-206. To evaluate whether 9p13 DNA dosage changes lead to over-expression of our three candidate oncogenes during oral cancer progression, we undertook molecular analyses of matched DNA and RNA taken from multiple lesions within an oral cancer field from a single oral cancer patient. This field of diseased oral tissue was defined using a fluorescence visualization (FV) device and histopathological review. DNA and RNA samples were extracted from micro-dissected tissues representing normal, mild dysplasia, CIS, and OSCC tissues within the defined area (Fig. 7.6 A-F)51-53. We detected the 9p13 DNA copy number increase in the CIS and OSCC samples, with both samples exhibiting the same DNA alteration boundary (suggesting a shared clonal origin) (Fig. 7.6.G). This same DNA alteration event was not detected for normal or dysplasia tissues isolated from the same disease field. We noted that no genetic alterations were present in the whole genome profiles of the normal and mild dysplasias – and that genomic instability at 9p13 was first observed in the CIS sample in this oral cancer field. Expression levels of the three 9p13.3 candidate genes were evaluated in RNA samples isolated from each biopsy from the disease field (Fig. 7.6H). Relative to matched normal tissue, DCTN3 was the only candidate showing ≥2-fold over-expression in all stages of disease. VCP and STOML2 were transcriptionally up-regulated only in the CIS lesion and invasive tumor, but did not show a change in those samples without the 9p13 copy number increase. VCP,   - 114 - DCTN3, and STOML2 all showed expression increases concomitant with the emergence of DNA copy number increases at 9p13.3.  7.3.5 Frequency of 9p13 Gain in Various Cancers To determine the significance of 9p13 DNA gain in other cancer cell types, we evaluated 217 cell lines derived from several cancer types207. We found 36 (16.6%) of the cell lines carried regions of genomic gain spanning part of chromosome 9p13, while four (1.8%) harbored high-level DNA amplification of this region. Those four lines with high-level amplification included: a ductal breast carcinoma line (BT-474), a tongue squamous cell carcinoma line (SCC-9), a melanoma line (WM-115), and an osteosarcoma line (MG-63). Out of these four lines, only the amplicon in SCC-9 spanned all three oral cancer candidate genes defined by our analyses of oral cancer tissues (VCP, STOML2, and DCTN3). These results suggest that 9p13 gain is not only to oral tumorigenesis but may also play a role in malignancy in other cancer types.   7.4 Discussion Loss of chromosome 9p has been reported as an early and frequent event in oral tumorigenesis, with selection for the loss of tumor suppressor CDKN2A (at 9p21.3) believed to be the driving force behind the emergence of this alteration46,65. Loss of chromosome 9p has primarily been detected by microsatellite markers for LOH analysis, with reports suggesting 9p LOH frequencies ranging from 28% to 82% in oral precancers. Here, an examination of progressing low-grade OPLs reveals that DNA copy number gain at 9p13 is detected at a frequency of 62.5%, whereas deletions on 9p21-24 are only found in 25% of cases. This discrepancy could be explained by 1) differences in analytical techniques (with LOH analysis not typically performed at the 9p13 locus); 2) differences   - 115 -  Figure 7.6 DNA gain at 9p13 and corresponding increased mRNA expressions in gene candidates detected in biopsies from a single patient. Clinical, pathological, and molecular characterization of different tissues obtained from an oral cancer disease field in a single patient. (A) to (D) are photomicrographs of hematoxylin and eosin stained slides of varying histology: (A) normal, (B) mild dysplasia, (C) carcinoma in situ (CIS), and (D) invasive squamous cell carcinoma (SCC) (Original magnification, 200 X). (E) White light image of the oral cancer at right side of tongue. Location of biopsies and its corresponding histology (A-D) are indicated. (F) Fluorescence visualization (FV) image of the same lesion, with a broad area of dark brown change of FV loss. (G) Alignment of genomic alteration data at chromosome 9p indicates 9p13 amplification in the CIS and SCC cells obtained from this single patient. The red vertical lines indicate log2 signal intensity ratios of +1 and +0.5 and the green lines indicate ratios of -1 and -0.5, Each black dot represents the log2 signal intensity ratio of an individual clone mapping to the array. (H) Increased fold-change mRNA expressions (> two-fold) are found for STOML2, VCP, and DCTN3 as determined by gene expression microarray. No standard error is indicated for STOML2 as only a single probe maps to this gene.    - 116 - in sample cohort make-up; and 3) the fact that earlier reports often involved LOH analysis of only severe dysplasia and CIS lesions when delineating molecular events in OPLs (i.e. low-grade OPLs were not included in analyses or analyzed separately). When we evaluate only high-grade OPLs within our data, we also detect a higher frequency of loss on chromosome 9p21, providing insight into a possible sequence of molecular events occurring during disease progression (data publicly available at GSE9193)178. Additionally, our tissue microarray experiments show 43.5% of progressing low-grade OPLs possess a DNA copy number gain at 9p13, whereas this is much less frequent in the high-grade OPL cohort (7.6%).  A comparison of the available patient demographic data (such as age, sex, smoking history and site of cancer) indicate our patient population is similar to those assessed in previous studies analyzing 9p21 deletion in OPLs28,179.  DNA copy number gain at chromosome 9p13 has been reported as a driver for other epithelial cancer types. The frequency of gain in this region varies widely depending on cancer type, from ~8-75%196,208,209. A study analyzing DNA copy number data across 11 types of cancer – including head and neck cancers – found increased DNA dosage at 9p13.3 to be significantly recurrent210. When only considering the DNA copy number changes in head and neck cancer in this same sample set, gains at chromosome 9p13.3 were found to be one of the most significantly recurring regions of DNA gain210. To the best of our knowledge, only one other group has fine-mapped recurring amplifications in this region to identify oncogene candidates. Kamradt et al. reported a minimal region of alteration of ~1.7 Mb in prostate cancers that spanned two candidate genes – IL11-RA and DCTN3 196. Recently, 9p13 DNA gain has been reported for oral tumors211,212. Pickering et al. found that 26% of OSCC tumors possessed a 9p amplification and suggested an oncogene may be found within this area212. Furthermore, analysis of the cohort of head and neck tumors from TCGA Research Network (http://cancergenome.nih.gov) using cBioPortal for Cancer Genomics and the Integrative Genomics Viewer (IGV) indicate that 25% of oral cavity specific   - 117 - tumors have a copy number gain at 9p13.3 (based on analysis of data from tissue collected from the oral cavity, tongue, floor of mouth, buccal mucosa, and hard palate) 213,214. These findings agree with our results insofar as gain at 9p13 occurs at a lower frequency in later stage/invasive disease as compared to earlier stage OPLs that progress.  To identify oncogene candidates within the 9p13 region, we leveraged existing whole genome alteration data from OPLs and oral tumors and global gene expression data from oral tumors. Then we further honed this list by evaluating only those candidates previously implicated in malignant processes. VCP, STOML2, and DCTN3 were identified by this approach. VCP (p97) is known to activate NF-κB signaling, drive cell proliferation and anti-apoptotic messages, and has been described as up-regulated in multiple cancer types195. STOML2 has been described as a positive regulator of cancer cell growth; its increased expression is associated with esophageal precancers and its expression has been reported as increasing with invasive disease stages in laryngeal squamous cell carcinomas197,199. High expression of STOML2 is also associated with poorer survival in gastric adenocarcinoma, and metastasis and poor survival in lung cancer201,215. DCTN3 (also referred to as DCTN22 or p22) is a subunit of dynactin, which is a protein complex involved in a number of cell processes such as spindle formation, cytokinesis, chromosome movement, and nuclear positioning216. DCTN3 has been associated with progression and metastasis formation in breast cancer217. Each of these genes represented a robust oncogene candidate for downstream functional analysis. Significantly, we found that all three oncogene candidates contributed to malignant processes in oral cancer cells. Independent shRNA-mediated knockdown of each candidate in SCC-9, an OSCC cell line harboring a 9p13 amplicon, resulted in reduced proliferative abilities and reduced capacity for anchorage-independent growth (Fig. 3). Further, lentiviral-mediated over-expression of VCP, DCTN3, and STOML2 in Cal-27 and POE9n-tert, OSCC and dysplasia cell lines with neutral DNA copy number at 9p13, enhanced both proliferation and anchorage independent growth (Fig. 4). These   - 118 - functional studies, plus reports detailing the oncogenic potential of each of the three candidate genes potentially indicate that parallel activation of VCP, STOML2, and DCTN3 may be a critical event in governing OPL phenotypes. Whereas DNA amplification is typically thought to arise by selection for over-expression of a single oncogene176,177 our data indicate that multiple oncogenes may be driving emergence of the 9p13 amplicon in early OPLs. This concept of parallel oncogene activation has previously been reported in ovarian, colorectal, and breast cancers218-220. These experiments give a strong indication that these genes can affect tumorigenicity of tumor cells in vitro, however, in order to gain a more in depth understanding of the effects of the gene in vivo, further investigations using animal models are warranted.  To further elucidate the impact of 9p13 gain on gene expression, we analyzed the DNA dosage and gene expression status of our three putative oncogenes in a series of different staged OPLs captured from within a single disease field (as defined by fluorescence visualization)51-53 (Fig. 6). Normal, mild dysplasia, CIS, and invasive oral carcinoma tissues were studied. In this case, the first sign of genomic instability at 9p occurred within the CIS lesion. The shared breakpoint at the 9p13 gain in both the CIS and OSCC biopsy suggest a shared clonal relationship between these two lesions. We assumed shared clonality between all four biopsies, however as oral cancer potentially arises due to field cancerization, it is a possibility that all of these lesions are not in fact clonally related. The normal and dysplasia oral tissues did not harbor any DNA copy number changes therefore breakpoint mapping was not feasible.  DNA dosage-mediated increases in RNA expression was apparent for all candidate genes.  Furthermore, our results suggest that activation of candidate oncogenes is likely to be governed by additional events rather than a single DNA copy number gain event at 9p13: we detected DCTN3 over-expression in dysplasia tissue in the absence of increased gene dosage. However, all three genes were over-expressed by the invasive disease stage and exhibited DNA copy number gain-mediated expression increases. These results align well with the   - 119 - other clinical and functional data we reported, further supporting the conclusion that that these genes play a role in oral cancer phenotypes. This is also supported by trends of 9p13 gain and increased gene expression of our candidate genes in the oral cavity specific samples from TCGA data. Gain of the 9p13 region was present in 27% of TCGA cases and over-expression of at least one of our candidate genes was present in 29% of cases. For tumors with 9p13 gain (n = 47) that had both gene expression and DNA copy number data available, 70% had an increase in gene expression of at least one of our candidate genes; however, 12% of cases without 9p13 gain showed over-expression of at least one candidates, further supporting the critical role of DNA dosage changes in driving over-expression of our candidate oncogenes (while also suggesting the existence of alternative mechanisms for deregulation of these genes in a smaller subset of cases).   Based on results from the single case we have reported and publicly available oral cancer data, it appears that in addition to DNA copy number changes, other factors may regulate expression of our candidate genes during oral tumorigenesis. Activation of oncogenes by a variety of mechanisms has previously been demonstrated in other types of cancer and reinforces the importance of analysis of tumors on multiple platforms221. The observations that 9p13 gain as well as gene over-expression CIS stage but not in the matched dysplastic lesion in our single patient case indicates that there are multiple molecular pathways that can initiate oral tumorigenesis. Further, while the 9p13 gain event is one of the most highly frequent events within the progressing LGOPLs, it may also contribute to tumorigenesis at later stages as well.  The decrease in frequency of the 9p13 amplicon as disease progresses indicates that this could be an initial step in disease initiation that is later not required or that mechanisms of gene regulation other than DNA amplification are involved in maintaining gene over-expression. This decrease in frequency with increasing oral lesion severity may be explained by the increased genomic instability that is known to occur with disease progression. Tumor progression is a dynamic   - 120 - process, and thus early critical events for progression may lose selection pressure in favor of other molecular aberrations that confer a greater growth advantage222 . Known mechanisms of genome instability such as bridge fusion breakage can utilize a DNA amplification event to drive further fragility of the surrounding chromosome region, thus creating greater genome instability and driving possible loss of the original amplified region223. This increased instability may be masking early genomic alterations that are important for disease initiation224 . Although there is a decrease in frequency of the 9p13 gain with disease progression, there is still a considerable proportion of tumors (24-27%) exhibiting this alteration and/ or gene expression increases at invasive stages (as is evident from TCGA data and previous studies done on OSCC)212.   7.5 Conclusion We report a recurrent region of DNA gain at 9p13 that is frequently detected in low-grade OPLs that subsequently progressed to invasive disease. Within this region, VCP, STOML2 and DCTN3 were identified as candidate oncogenes and we demonstrate that each is able to regulate oral cancer phenotypes. Activation of these putative oncogenes is frequently mediated by this single DNA copy number gain event at chromosome 9p13. Analysis of multiple biopsies from a single oral cancer field suggests that additional levels of molecular deregulation might govern activation of these genes, indicating that their contribution to oral cancer progression is more complex than a simple additive effect. Further analysis of this 9p13 alteration event in oral cancer initiation and progression is warranted.      - 121 - Chapter 8: Discussion and Conclusions 8.1 Summary of Findings OSCC has a dismal survival rate of ~50% over five years, partially due to the disease generally being diagnosed at late stages.  This disease develops through a typical histological progression, and premalignant lesions can be readily identified through examination of the oral cavity. However, since the majority of early lesions do not progress into invasive disease, more aggressive intervention is rarely performed as it is impossible to differentiate patients at risk for progression based on the current gold standard histology.  There are currently no biomarkers or mechanisms for identifying these patients at risk for progression and comparatively little is known regarding the dysplasia stage of disease on a molecular level.  Thus, an in-depth analysis of premalignant lesions may provide insights into genes and molecular pathways crucial for the progression process or molecular alterations that may be used strategically as biomarkers to identify patients at risk for progression.  The work in this thesis provides epigenetic profiling of patient matched adjacent normal, dysplasia and CIS/SCC DNA methylation, miRNA expression and mRNA profiles.  Furthermore, we also investigate frequently deregulated genes that may be vital to progression at early stages of disease.   8.1.1 Epigenetic Analysis of the Different Histological Stages of OSCC The goal of this thesis was to characterize the epigenetic landscape of the earliest stages of this disease and to identify genes potentially contributing to the tumorigenesis of the lesions at these early stages. Chapters 3-5 focus on the epigenetic analysis of patient matched adjacent normal, dysplasia and CIS/SCC from within the same disease field. Within these chapters, we assess the DNA methylation, gene expression and miRNA profiles of these different histological stages.  By assessing   - 122 - these types of samples, we are able to assess the change in these epigenetic mechanisms as the disease progresses.   In Chapter 3 we provide the first genome-wide DNA methylation analysis of a sample set of this nature.  We found that hypermethylation is prevalent at the dysplasia stage and is more or less maintained in the progression through tumorigenesis.  Further, we find hypomethylation to increase after the dysplasia stage.  We also identify a vast number of CpGs that exhibit aberrant DNA methylation in every CIS/SCC sample analyzed and find a high level of concordance of the prevalence of the methylation of these CpGs in TCGA oral cancer dataset.   Examination of the microRNA profiles throughout the different stages of oral cancer had only been performed on small subsets of miRNAs.  In Chapter 4 we assess the most miRNAs from a sample set like this, and the first to focus specifically on non-smokers.  We find distinct differences in the number of miRNAs detectable at the tumor stage compared to normal and dysplasia.  We also identify a number of frequently deregulated miRNAs.     Finally, an integrative analysis of the data discussed in the above chapters is performed in order to provide a detailed overview of the epigenetic mechanisms at play within these tissue biopsies.  The data from five patient samples that have an overlap between all three platforms is assessed for common highly correlated events.  Although this analysis lacks the sample numbers to find highly significant events and pathways implicated among these patient samples, a variety of different mechanisms where processes such as “Pathways in Cancer” are deregulated in the dysplasia and “Focal Adhesion within the tumor genome are identified.   8.1.2 Assessment of Frequent Molecular Events as Being Implicated in Progression In addition to providing in-depth analyses of the epigenetic mechanisms of the earliest stages of disease of oral cancer, my goal was also to assess molecular aberrations frequently occurring at the   - 123 - premalignant stage of disease.  This was accomplished via functional assessments in vitro using oral squamous cell carcinoma cell lines and oral dysplasia cell lines.  In Chapter 6, a further investigation into the validity of EYA4, the most frequently hypermethylated dysplasia gene identified in Chapter 3, as being a potential tumor suppressor gene is performed.  We find evidence to substantiate this hypothesis, as over-expression of EYA4 contributes to decreasing proliferation, increasing apoptotic repair and increasing DNA damage repair capabilities.   Chapters 3-6 focus on dysplastic samples with paired tumors within an abnormal field.  However, to truly link a potential molecular event to impacting progression, genes must be assessed when a patient has yet to progress.  The study in Chapter 7 provides an example for this type of analyses.  In a study prior to this thesis, a recurrent amplification at a region in 9p13 identified in oral progressing tumors was further investigated by assessing the validity of candidate genes within this region.  Three genes, VCP, DCTN3 and STOML2, are identified as possible oncogenes and all three have an impact on oral tumorigenesis in both dysplasia and tumor cell lines.    8.2 Conclusions Regarding Hypotheses The hypotheses in this thesis were 1) the epigenetic landscape of OSCC becomes progressively deregulated throughout the different histological stages and that 2) the most frequently molecular events identified at the dysplasia stage may be crucial for premalignant disease development and progression. In order to assess the validity of Hypothesis 1, we characterized the DNA methylation, mRNA and miRNA profiles of patient-matched adjacent normal, dysplasia and CIS/SCC from within a single disease field.  We found distinct DNA methylation patterns between normal and advanced lesions indicating a change in methylation patterns through progression.  We had initially anticipated seeing increasing levels of both hypermethylation and hypomethylation within the progressive tissue   - 124 - biopsies.  Although we did see an increase in the level of both hyper and hypomethylation in the transition from normal to dysplasia, we only see a change in the level of hypomethylation in the change from dysplasia to tumor.  We also found the number of miRNA detected increases as the disease progresses and see the deregulation of a number of miRNA throughout the different stages of oral cancer. Collectively, we find evidence to support Hypothesis 1 within these results. We assessed the validity of frequently deregulated molecular events on impacting tumorigenicity in order to address Hypothesis 2.  Specifically, we assess the ability of the hypermethylation of EYA4 as a tumor suppressive event in Chapter 6 and the amplification of a region on chromosome 9p13 to increase the expression of potential oncogenic genes in Chapter 7.  In both chapters, we find evidence that deregulation of these genes have the ability to impact tumorigenecity.  Over-expression of EYA4 in dysplastic lines with no detectable EYA4 expression results in a marked decrease in proliferation, DNA damage repair and apoptosis.  Additionally, both over-expression and knockdown experiments in candidate oncogenes within the 9p13 region result in impacts on proliferation and anchorage independent growth in the direction that indicate these genes are potential oncogenes in early oral cancer.  These results give support to Hypothesis 2, that frequently deregulated genes, be it via methylation or copy number aberrations can be identified through high level genomic analysis and can contribute to tumorigenesis.  Taken together, the results obtained in this thesis give evidence to support that there is a transition of epigenetic events through the progression of this disease, and that frequently deregulated genes can contribute to tumorigenesis, highlighting the clinical potential for these type of analysis.    8.3 Strengths and Limitations The approach to this study has a number of strengths.  First, a major strength of this work is the opportunity to assess patient matched lesions of differing histology within a single disease field.    - 125 - This allows for internally matched controls to identify to molecular aberrations as compared to a patient’s own normal tissue.  This limits confounding variables in our sample set, such as age, smoking status, sex or ethnicity that studies using pooled normal samples inevitably possess.  In addition to having paired normal tissue we also have dysplastic tissue, and thus are able to see the transformation of the molecular aberrations throughout the progression of disease. This type of patient cohort is very rare, and only a few studies have utilized a similar approach132. A second strength is the amount of information we are able to acquire regarding these patient samples.  In total, the methylation status of >27,000 CpGs mapping to ~14,000 genes, the expression 41,000 transcripts corresponding to ~19,000 genes and the expression of 722 miRNAs is profiled.  Although this data does not reach the caliber that whole transcriptome sequencing or profiling with the newer Illumina methylation arrays (which profiles ~485,000 CpGs), the data in this thesis nevertheless represents the most detailed interrogation of a sample set containing patient matched normal, dysplasia and advanced lesions in oral cancer studies to our knowledge.    Finally, the use of dysplasia cell lines in addition to tumor cell lines in our in vitro functional assessments also strengthens this work.  This is not only important as the focus of this thesis is on dysplastic tissue, but also these cell lines display a lower degree of genomic instability and molecular aberrations that cancer cell lines often display.  This allows for the identification of functional changes in the cell lines that would not be apparent if they were lost within the genomic background of the cancer cell lines.    Despite these strengths discussed above, this work does have several limitations that must be discussed.  First of all, in Chapter 3, patient samples from both smokers/former smokers (n=3) and non-smokers (n=7) are included.  As it has been demonstrated that the overall methylation profiles of smokers and non-smokers have subtle differences, inclusion of these patients may confound the results.   However, we do note that we did not identify any obvious differences between smokers or   - 126 - non-smokers in Chapter 3, although with the sample size involved in this study we would be unlikely to identify anything statistically significant.  For example, our hierarchical clustering presented in Figure 3.3 the smoker patients do not cluster separately from the non-smokers.  Thus, we likely do not have the sample number required to identify the differences in these populations and are detecting methylation events that are common among both populations of oral cancer patients. However, upon reflection on the implication of including smokers in this analysis, only non-smokers are included in the analysis in Chapter 4. This analysis was also somewhat limited due to the small sample size of patients we investigated.  Only a total of 14 patient sample sets were profiled, ten of which are analyzed with methylation/gene expression and nine of which have miRNA analysis (five samples with data from all platforms).  We were also unable to profile all patients on all molecular platforms insufficient RNA or DNA input. These samples are rare and it is difficult to acquire dysplasia due to the small size of the lesion.  However, although this sample size is small, a number of  highly frequent events in were identified (including events happening between 70-100% of all cases).    8.4 Overall Significance of this Work A major goal of this research is to bring to light more information about the progression of oral cancer, and more specifically, to the dysplasia stage of this disease.  A core part of the goals of this thesis is to present the analysis of highly detailed data regarding the DNA methylation and miRNA landscape of this disease.   The work presented in Chapter 3 is the first to perform genome-wide DNA methylation analysis of patient matched biopsies of differing histological stages.  First, we establish that methylation landscapes change throughout the different stages of disease, with adjacent normal tissue and advanced lesions showing distinct methylation patterns.  Through this analysis, we identify an   - 127 - increase of methylation levels as an event occurring at the premalignant stage of disease and see hypomethylation becoming more prevalent as the disease progresses. These results are of importance on a clinical level as it provides clinicians with more information regarding the epigenetic landscape of OSCC, which is relevant in the context of utilizing epigenetic drugs such as 5’azacytidine as an approach to combat this disease within patients. Our analysis of miRNA changes throughout the different stages of oral cancer in non-smokers is also the first analysis of a sample set of patients from this particular subgroup assessing a large number of miRNAs.  We illustrate that miRNA change throughout the different stages, as we see a significant increase in the number of miRNAs detected as the disease progresses.  In addition to identifying general trends in miRNA deregulation in this sample set, we also comment on the status of a number of genes and provide context for the importance of miR-155-5p up-regulation in the premalignant stage of disease. We also provide further investigation in novel candidate genes that play a role in early tumorigenesis.  We identify EYA4 as a novel tumor suppressor that is recurrently hypermethylated and silenced in oral dysplasia and OSCC, which could potentially have merit as a biomarker for progression in the early stages of oral cancer.   In addition to this we also identify the significance of a recurring amplification predominately within progressing oral premalignant lesions in Chapter 7.  Not only is this one of the few potential biomarkers of progression, but we observe that this change occurs prior to the loss of chromosome regions 9p and 3p - the earliest molecular changes observed in molecular changes prior to this research.  This amplification of 9p13 presents clinical utility as a potential biomarker for progression.  However, further long-term follow up studies are required to fully understand the significance of this finding clinically.      - 128 - 8.5 Future Directions Analysis of the DNA methylation and miRNA profiles highlighted a number of genes which merit further investigation, of which went beyond the scope of this thesis.  Specifically, within the DNA methylation cohort we identified a number of potentially interesting genes. The development of more in-depth DNA methylation microarray by Illumina means that further profiling of these lesions can be done to a much higher scale, with ~485,000 CpGs profiled compared to the ~27,000 done in this thesis.  These data would provide insight into regions of CpG abnormalities beyond those that are found in the promoter region, which is what has been our focus in this research.   The identification of frequent molecular events in early tumorigenesis of oral cancer also warrants further investigation. Further analysis of EYA4 in the context of progression may give insight in a possibility as utility as a biomarker.  The next step in the investigation of this gene in the context of progression is to identify if this is a molecular event that occurs in all dysplasias, or if it is unique to progressing oral premalignant lesions.  Thus, an investigation in patients with oral dysplasias who have not yet progressed into invasive disease should be done to assess if EYA4 is unique to these types of dysplasias.  This could have a clinical utility as it would allow for a biomarker at the early stages to identify patients at risk for progression.   Further investigation into the candidate oncogenes identified in Chapter 7 is also warranted.  As these genes are found within the same frequent amplification on 9p13, it would be prudent to identify the effect of the activation of a combination of these genes in model systems.  The results of this may play a synergistic effect with one another, and may provide more insights of the molecular mechanisms involved in the genes within this amplification.  Furthermore, given the frequency of the amplification and subsequent over-expression of these candidate genes within our tumor sample set, it may be warranted to further investigate the ability to target these genes therapeutically.  Currently   - 129 - there are no available compounds that specifically target STOML2 or DCTN3, however there has been an increased interest in VCP in the context of cancer due to it increasingly identified as over-expressed in other types of cancer and due to its role as an ATPase molecule225-227. One group has identified several compounds that can be used to target VCP, confirming it is a druggable  target228.      - 130 - Bibliography 1 Torre, L. A. et al. Global cancer statistics, 2012. CA: a cancer journal for clinicians 65, 87-108 (2015). 2 Warnakulasuriya, S. Global epidemiology of oral and oropharyngeal cancer. Oral oncology 45, 309-316 (2009). 3 Ries, L. A. et al. SEER cancer statistics review, 1975-2003.  (2006). 4 Lippman, S. M. & Hong, W. K. Molecular markers of the risk of oral cancer. New England Journal of Medicine 344, 1323-1326 (2001). 5 Chaturvedi, A. K. et al. Worldwide trends in incidence rates for oral cavity and oropharyngeal cancers. Journal of Clinical Oncology 31, 4550-4559 (2013). 6 Ram, H. et al. Oral cancer: risk factors and molecular pathogenesis. Journal of maxillofacial and oral surgery 10, 132-137 (2011). 7 Kademani, D. in Mayo Clinic Proceedings.  878-887 (Elsevier). 8 Patel, S. C. et al. Increasing incidence of oral tongue squamous cell carcinoma in young white women, age 18 to 44 years. Journal of Clinical Oncology 29, 1488-1494 (2011). 9 Helen‐Ng, L. C. et al. Dietary pattern and oral cancer risk–a factor analysis study. Community dentistry and oral epidemiology 40, 560-566 (2012). 10 Maier, H., Zöller, J., Herrmann, A., Kreiss, M. & Heller, W.-D. Dental status and oral hygiene in patients with head and neck cancer. Otolaryngology--Head and Neck Surgery 108, 655-661 (1993). 11 Alnuaimi, A. D., Wiesenfeld, D., O’Brien-Simpson, N. M., Reynolds, E. C. & McCullough, M. J. Oral Candida colonization in oral cancer patients and its relationship with traditional risk factors of oral cancer: A matched case-control study. Oral oncology 51, 139-145 (2015). 12 Hecht, S. S. Tobacco carcinogens, their biomarkers and tobacco-induced cancer. Nature Reviews Cancer 3, 733-744 (2003). 13 Boffetta, P. & Hashibe, M. Alcohol and cancer. The lancet oncology 7, 149-156 (2006). 14 Homann, N. et al. Increased salivary acetaldehyde levels in heavy drinkers and smokers: a microbiological approach to oral cavity cancer. Carcinogenesis 21, 663-668 (2000). 15 Blot, W. J. et al. Smoking and drinking in relation to oral and pharyngeal cancer. Cancer research 48, 3282-3287 (1988). 16 Rothman, K. & Keller, A. The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx. Journal of chronic diseases 25, 711-716 (1972). 17 Müller, S., Pan, Y., Li, R. & Chi, A. C. Changing trends in oral squamous cell carcinoma with particular reference to young patients: 1971–2006. The Emory University experience. Head and neck pathology 2, 60-66 (2008). 18 Shiboski, C. H., Schmidt, B. L. & Jordan, R. C. Tongue and tonsil carcinoma: increasing trends in the U.S. population ages 20-44 years. Cancer 103, 1843-1849, doi:10.1002/cncr.20998 (2005). 19 Myers, J. N., Elkins, T., Roberts, D. & Byers, R. M. Squamous cell carcinoma of the tongue in young adults: increasing incidence and factors that predict treatment outcomes. Otolaryngology--Head and Neck Surgery 122, 44-51 (2000). 20 Lingen, M. W. et al. Low etiologic fraction for high-risk human papillomavirus in oral cavity squamous cell carcinomas. Oral oncology 49, 1-8 (2013).   - 131 - 21 Salem, A. Dismissing links between HPV and aggressive tongue cancer in young patients. Annals of oncology, mdp380 (2009). 22 Axéll, T., Pindborg, J., Smith, C. & Waal, I. v. d. Oral white lesions with special reference to precancerous and tobacco‐related lesions: conclusions of an international symposium held in Uppsala, Sweden, May 18–21 1994. Journal of oral pathology & medicine 25, 49-54 (1996). 23 Leemans, C. R., Braakhuis, B. J. & Brakenhoff, R. H. The molecular biology of head and neck cancer. Nature Reviews Cancer 11, 9-22 (2011). 24 Kramer, I. R., Lucas, R. B., Pindborg, J. J. & Sobin, L. H. Definition of leukoplakia and related lesions: an aid to studies on oral precancer. Oral surgery, oral medicine, and oral pathology 46, 518-539 (1978). 25 Bouquot, J. E., Speight, P. M. & Farthing, P. M. Epithelial dysplasia of the oral mucosa—Diagnostic problems and prognostic features. Current Diagnostic Pathology 12, 11-21 (2006). 26 Banes, L., Eveson, J. W., Reichart, P. & Sidransky, D. World Health Organization classification of tumors. Pathology and genetics of head and neck tumors. World Health Organization, New York, p130-131 (2005). 27 Warnakulasuriya, S., Reibel, J., Bouquot, J. & Dabelsteen, E. Oral epithelial dysplasia classification systems: predictive value, utility, weaknesses and scope for improvement. Journal of oral pathology & medicine 37, 127-133 (2008). 28 Mao, L. et al. Frequent microsatellite alterations at chromosomes 9p21 and 3p14 in oral premalignant lesions and their value in cancer risk assessment. Nat Med 2, 682-685 (1996). 29 Slaughter, D. P., Southwick, H. W. & Smejkal, W. “Field cancerization” in oral stratified squamous epithelium. Clinical implications of multicentric origin. Cancer 6, 963-968 (1953). 30 Braakhuis, B. J., Tabor, M. P., Kummer, J. A., Leemans, C. R. & Brakenhoff, R. H. A genetic explanation of Slaughter's concept of field cancerization: evidence and clinical implications. Cancer Res 63, 1727-1730 (2003). 31 Ogden, G., Cowpe, J. & Green, M. Evidence of field change in oral cancer. British Journal of Oral and Maxillofacial Surgery 28, 390-392 (1990). 32 Kerdpon, D., Rich, A. & Reade, P. Expression of p53 in oral mucosal hyperplasia, dysplasia and squamous cell carcinoma. Oral diseases 3, 86-92 (1997). 33 Day, G. L. & Blot, W. J. Second primary tumors in patients with oral cancer. Cancer 70, 14-19 (1992). 34 Pickering, C. R. et al. Integrative genomic characterization of oral squamous cell carcinoma identifies frequent somatic drivers. Cancer discovery 3, 770-781 (2013). 35 Pérez-Sayáns, M. et al. Genetic and molecular alterations associated with oral squamous cell cancer (Review). Oncol Rep 22, 1277-1282 (2009). 36 Won, H. S. et al. Difference in expression of EGFR, pAkt, and PTEN between oropharyngeal and oral cavity squamous cell carcinoma. Oral oncology 48, 985-990 (2012). 37 Bruckman, K. C., Schönleben, F., Qiu, W., Woo, V. L. & Su, G. H. Mutational analyses of the BRAF, KRAS, and PIK3CA genes in oral squamous cell carcinoma. Oral Surgery,   - 132 - Oral Medicine, Oral Pathology, Oral Radiology, and Endodontology 110, 632-637 (2010). 38 Šupić, G., Kozomara, R., Branković-Magić, M., Jović, N. & Magić, Z. Gene hypermethylation in tumor tissue of advanced oral squamous cell carcinoma patients. Oral oncology 45, 1051-1057 (2009). 39 Sartor, M. et al. Role of p16/MTS1, cyclin D1 and RB in primary oral cancer and oral cancer cell lines. British journal of cancer 80, 79 (1999). 40 Huang, Q. et al. Genetic differences detected by comparative genomic hybridization in head and neck squamous cell carcinomas from different tumor sites: construction of oncogenetic trees for tumor progression. Genes, Chromosomes and Cancer 34, 224-233 (2002). 41 Chen, Y. J. et al. Genome‐wide profiling of oral squamous cell carcinoma. J Pathol 204, 326-332 (2004). 42 Cohen, Y. et al. Mutational analysis of PTEN/PIK3CA/AKT pathway in oral squamous cell carcinoma. Oral oncology 47, 946-950 (2011). 43 Pannone, G. et al. WNT pathway in oral cancer: epigenetic inactivation of WNT-inhibitors. Oncol Rep 24, 1035-1041 (2010). 44 Iamaroon, A., Pattamapun, K. & Piboonniyom, S.-o. Aberrant expression of Smad4, a TGF-. BETA. signaling molecule, in oral squamous cell carcinoma. Journal of oral science 48, 105-109 (2006). 45 van der Riet, P. et al. Frequent loss of chromosome 9p21-22 early in head and neck cancer progression. Cancer research 54, 1156-1158 (1994). 46 El-Naggar, A. K. et al. Sequential loss of heterozygosity at microsatellite motifs in preinvasive and invasive head and neck squamous carcinoma. Cancer research 55, 2656-2659 (1995). 47 Papadimitrakopoulou, V. et al. Frequent inactivation of p 16 INK 4 a in oral premalignant lesions. Oncogene 14, 1799-1803 (1997). 48 Tsui, I. F., Rosin, M. P., Zhang, L., Ng, R. T. & Lam, W. L. Multiple aberrations of chromosome 3p detected in oral premalignant lesions. Cancer Prevention Research 1, 424-429 (2008). 49 Li, J. et al. MiR-21 indicates poor prognosis in tongue squamous cell carcinomas as an apoptosis inhibitor. Clinical cancer research 15, 3998-4008 (2009). 50 Poh, C. F. et al. Canadian Optically-guided approach for Oral Lesions Surgical (COOLS) trial: study protocol for a randomized controlled trial. BMC cancer 11, 462 (2011). 51 Lane, P. M. et al. Simple device for the direct visualization of oral-cavity tissue fluorescence. Journal of biomedical optics 11, 024006, doi:10.1117/1.2193157 (2006). 52 Poh, C. F. et al. Direct fluorescence visualization of clinically occult high-risk oral premalignant disease using a simple hand-held device. Head Neck 29, 71-76, doi:10.1002/hed.20468 (2007). 53 Poh, C. F. et al. Fluorescence visualization detection of field alterations in tumor margins of oral cancer patients. Clinical cancer research : an official journal of the American Association for Cancer Research 12, 6716-6722, doi:10.1158/1078-0432.CCR-06-1317 (2006). 54 Bibikova, M. et al. Genome-wide DNA methylation profiling using Infinium(R) assay. Epigenomics 1, 177-200, doi:10.2217/epi.09.14 (2009).   - 133 - 55 Illumina. GenomeStudio® Methylation Module v1.8 User Guide.  (2010). 56 Du, P. et al. Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC bioinformatics 11, 587 (2010). 57 Tsui, I. F. & Garnis, C. Integrative molecular characterization of head and neck cancer cell model genomes. Head & neck 32, 1143-1160 (2010). 58 Das, P. M. & Singal, R. DNA methylation and cancer. Journal of Clinical Oncology 22, 4632-4642 (2004). 59 Jones, P. A. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nature Reviews Genetics 13, 484-492 (2012). 60 Jjingo, D., Conley, A. B., Soojin, V. Y., Lunyak, V. V. & Jordan, I. K. On the presence and role of human gene-body DNA methylation. Oncotarget 3, 462-474 (2012). 61 Laird, P. W. & Jaenisch, R. The role of DNA methylation in cancer genetics and epigenetics. Annual review of genetics 30, 441-464 (1996). 62 Daskalos, A. et al. Hypomethylation of retrotransposable elements correlates with genomic instability in non‐small cell lung cancer. International Journal of Cancer 124, 81-87 (2009). 63 Kitkumthorn, N. & Mutirangura, A. Long interspersed nuclear element-1 hypomethylation in cancer: biology and clinical applications. Clinical epigenetics 2, 315-330 (2011). 64 Richards, K. L. et al. Genome-wide hypomethylation in head and neck cancer is more pronounced in HPV-negative tumors and is associated with genomic instability. PLoS One 4, e4941 (2009). 65 Kresty, L. A. et al. Alterations of p16INK4a and p14ARF in patients with severe oral epithelial dysplasia. Cancer research 62, 5295-5300 (2002). 66 Cao, J. et al. Methylation of p16 CpG island associated with malignant progression of oral epithelial dysplasia: a prospective cohort study. Clinical cancer research 15, 5178-5183 (2009). 67 Gasco, M. et al. Epigenetic inactivation of 14-3-3 σ in oral carcinoma association with p16INK4a silencing and human papillomavirus negativity. Cancer research 62, 2072-2076 (2002). 68 Long, N. K. et al. Hypermethylation of the RECK gene predicts poor prognosis in oral squamous cell carcinomas. Oral oncology 44, 1052-1058 (2008). 69 Bibikova, M. & Fan, J. B. Genome‐wide DNA methylation profiling. Wiley Interdisciplinary Reviews: Systems Biology and Medicine 2, 210-223 (2010). 70 Network, C. G. A. Comprehensive genomic characterization of head and neck squamous cell carcinomas. Nature 517, 576-582 (2015). 71 Poage, G. M. et al. Identification of an epigenetic profile classifier that is associated with survival in head and neck cancer. Cancer research 72, 2728-2737 (2012). 72 Zhang, S. et al. Genome-wide analysis of DNA methylation in tongue squamous cell carcinoma. Oncol Rep 29, 1819-1826 (2013). 73 Jithesh, P. et al. The epigenetic landscape of oral squamous cell carcinoma. British journal of cancer 108, 370-379 (2013). 74 Righini, C. A. et al. Tumor-specific methylation in saliva: a promising biomarker for early detection of head and neck cancer recurrence. Clinical cancer research 13, 1179-1185 (2007).   - 134 - 75 Duffy, M. et al. Methylated genes as new cancer biomarkers. European Journal of Cancer 45, 335-346 (2009). 76 Viswanathan, M., Tsuchida, N. & Shanmugam, G. Promoter hypermethylation profile of tumor‐associated genes p16, p15, hMLH1, MGMT and E‐cadherin in oral squamous cell carcinoma. International Journal of Cancer 105, 41-46 (2003). 77 Kulkarni, V. & Saranath, D. Concurrent hypermethylation of multiple regulatory genes in chewing tobacco associated oral squamous cell carcinomas and adjacent normal tissues. Oral oncology 40, 145-153 (2004). 78 Sturn, A., Quackenbush, J. & Trajanoski, Z. Genesis: cluster analysis of microarray data. Bioinformatics 18, 207-208 (2002). 79 Towle, R. et al. Global analysis of DNA methylation changes during progression of oral cancer. Oral oncology 49, 1033-1042 (2013). 80 Aryee, M. J. et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics, btu049 (2014). 81 Kanai, Y. Genome-wide DNA methylation profiles in precancerous conditions and cancers. Cancer science 101, 36-45, doi:10.1111/j.1349-7006.2009.01383.x (2010). 82 Nephew, K. P. & Huang, T. H. Epigenetic gene silencing in cancer initiation and progression. Cancer letters 190, 125-133 (2003). 83 Feinberg, A. P., Ohlsson, R. & Henikoff, S. The epigenetic progenitor origin of human cancer. Nature reviews. Genetics 7, 21-33, doi:10.1038/nrg1748 (2006). 84 Vucic, E. A. et al. Translating cancer ‘omics’ to improved outcomes. Genome Research 22, 188-195 (2012). 85 Sinha, P. et al. Significance of promoter hypermethylation of p16 gene for margin assessment in carcinoma tongue. Head & neck 31, 1423-1430 (2009). 86 Kato, K. et al. Aberrant promoter hypermethylation of p16 and MGMT genes in oral squamous cell carcinomas and the surrounding normal mucosa. Journal of cancer research and clinical oncology 132, 735-743 (2006). 87 Talikka, M. et al. Genomic impact of cigarette smoke, with application to three smoking-related diseases. Critical reviews in toxicology 42, 877-889 (2012). 88 Toyooka, S. et al. Mutational and epigenetic evidence for independent pathways for lung adenocarcinomas arising in smokers and never smokers. Cancer research 66, 1371-1375 (2006). 89 von Zeidler, S. V., Miracca, E., Nagai, M. & Birman, E. Hypermethylation of the p16 gene in normal oral mucosa of smokers. International journal of molecular medicine 14, 807 (2004). 90 Towle, R. & Garnis, C. Methylation-mediated molecular dysregulation in clinical oral malignancy. Journal of oncology 2012, 170172, doi:10.1155/2012/170172 (2012). 91 Zou, H. et al. Frequent methylation of eyes absent 4 gene in Barrett's esophagus and esophageal adenocarcinoma. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology 14, 830-834, doi:10.1158/1055-9965.EPI-04-0506 (2005). 92 Selamat, S. A. et al. DNA methylation changes in atypical adenomatous hyperplasia, adenocarcinoma in situ, and lung adenocarcinoma. PLoS One 6, e21443, doi:10.1371/journal.pone.0021443 (2011).   - 135 - 93 Arai, E. et al. Single-CpG-resolution methylome analysis identifies clinicopathologically aggressive CpG island methylator phenotype clear cell renal cell carcinomas. Carcinogenesis 33, 1487-1493, doi:10.1093/carcin/bgs177 (2012). 94 Lind, G. E. et al. Novel epigenetically deregulated genes in testicular cancer include homeobox genes and SCGB3A1 (HIN-1). J Pathol 210, 441-449, doi:10.1002/path.2064 (2006). 95 Gilbert, P. M. et al. HOXA9 regulates BRCA1 expression to modulate human breast tumor phenotype. The Journal of clinical investigation 120, 1535-1550, doi:10.1172/JCI39534 (2010). 96 Ishii, M. et al. Facilitation of H2O2-induced A172 human glioblastoma cell death by insertion of oxidative stress-sensitive TRPM2 channels. Anticancer research 27, 3987-3992 (2007). 97 Zeng, X. et al. Novel role for the transient receptor potential channel TRPM2 in prostate cancer cell proliferation. Prostate cancer and prostatic diseases 13, 195-201, doi:10.1038/pcan.2009.55 (2010). 98 Rennie, P. S. et al. Relationship between variant forms of estrogen receptor RNA and an apoptosis-related RNA, TRPM-2, with survival in patients with breast cancer. Cancer 72, 3648-3654 (1993). 99 Fulda, S. Caspase-8 in cancer biology and therapy. Cancer letters 281, 128-133, doi:10.1016/j.canlet.2008.11.023 (2009). 100 Hacker, S. et al. Histone deacetylase inhibitors cooperate with IFN-gamma to restore caspase-8 expression and overcome TRAIL resistance in cancers with silencing of caspase-8. Oncogene 28, 3097-3110, doi:10.1038/onc.2009.161 (2009). 101 Tanaka, T. et al. Dysregulated expression of P1 and P2 promoter-driven hepatocyte nuclear factor-4alpha in the pathogenesis of human cancer. J Pathol 208, 662-672, doi:10.1002/path.1928 (2006). 102 Lazarevich, N. L. et al. Progression of HCC in mice is associated with a downregulation in the expression of hepatocyte nuclear factors. Hepatology 39, 1038-1047, doi:10.1002/hep.20155 (2004). 103 Sel, S., Ebert, T., Ryffel, G. U. & Drewes, T. Human renal cell carcinogenesis is accompanied by a coordinate loss of the tissue specific transcription factors HNF4 alpha and HNF1 alpha. Cancer letters 101, 205-210 (1996). 104 Bindea, G. et al. ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics 25, 1091-1093, doi:10.1093/bioinformatics/btp101 (2009). 105 Reya, T. & Clevers, H. Wnt signalling in stem cells and cancer. Nature 434, 843-850, doi:10.1038/nature03319 (2005). 106 Noguti, J. et al. The role of canonical WNT signaling pathway in oral carcinogenesis: a comprehensive review. Anticancer research 32, 873-878 (2012). 107 Andrade Filho, P. A. et al. Insights from studies with oral cleft genes suggest associations between WNT-pathway genes and risk of oral cancer. Journal of dental research 90, 740-746, doi:10.1177/0022034511401622 (2011). 108 Sogabe, Y. et al. Epigenetic inactivation of SFRP genes in oral squamous cell carcinoma. International journal of oncology 32, 1253-1261 (2008).   - 136 - 109 Pannone, G. et al. WNT pathway in oral cancer: epigenetic inactivation of WNT-inhibitors. Oncol Rep 24, 1035-1041 (2010). 110 Ying, Y. & Tao, Q. Epigenetic disruption of the WNT/beta-catenin signaling pathway in human cancers. Epigenetics : official journal of the DNA Methylation Society 4, 307-312 (2009). 111 Kondo, T. et al. Epigenetically controlled fibroblast growth factor receptor 2 signaling imposes on the RAS/BRAF/mitogen-activated protein kinase pathway to modulate thyroid cancer progression. Cancer Res 67, 5461-5470, doi:10.1158/0008-5472.CAN-06-4477 (2007). 112 Velasco, A. et al. Promoter hypermethylation and expression of sprouty 2 in endometrial carcinoma. Human pathology 42, 185-193, doi:10.1016/j.humpath.2010.08.001 (2011). 113 Nagasaka, T. et al. Colorectal cancer with mutation in BRAF, KRAS, and wild-type with respect to both oncogenes showing different patterns of DNA methylation. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 22, 4584-4594, doi:10.1200/JCO.2004.02.154 (2004). 114 Dhillon, A. S., Hagan, S., Rath, O. & Kolch, W. MAP kinase signalling pathways in cancer. Oncogene 26, 3279-3290, doi:10.1038/sj.onc.1210421 (2007). 115 Lewis, B. P., Burge, C. B. & Bartel, D. P. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 120, 15-20 (2005). 116 Osman, A. MicroRNAs in health and disease--basic science and clinical applications. Clinical laboratory 58, 393 (2012). 117 Cho, W. C. S. MicroRNAs as therapeutic targets and their potential applications in cancer therapy. Expert Opinion on Therapeutic Targets 16, 747-759 (2012). 118 Marshall, J. R. et al. Smoking, alcohol, dentition and diet in the epidemiology of oral cancer. European Journal of Cancer Part B: Oral Oncology 28, 9-15 (1992). 119 Llewellyn, C., Johnson, N. & Warnakulasuriya, K. Risk factors for squamous cell carcinoma of the oral cavity in young people—a comprehensive literature review. Oral oncology 37, 401-418 (2001). 120 Poling, J. et al. Human papillomavirus (HPV) status of non-tobacco related squamous cell carcinomas of the lateral tongue. Oral oncology 50, 306-310 (2014). 121 Kruse, A. L., Bredell, M. & Gratz, K. W. Oral squamous cell carcinoma in non-smoking and non-drinking patients. Head Neck Oncol 2, 24, doi:10.1186/1758-3284-2-24 1758-3284-2-24 [pii] (2010). 122 Constantinides, M. S., Rothstein, S. G. & Persky, M. S. Squamous cell carcinoma in older patients without risk factors. Otolaryngol Head Neck Surg 106, 275-277 (1992). 123 Farshadpour, F., Hordijk, G. J., Koole, R. & Slootweg, P. J. Non-smoking and non-drinking patients with head and neck squamous cell carcinoma: a distinct population. Oral Dis 13, 239-243, doi:ODI1274 [pii] 10.1111/j.1601-0825.2006.01274.x (2007). 124 Farshadpour, F., Hordijk, G. J., Koole, R. & Slootweg, P. J. Head and neck squamous cell carcinoma in non-smoking and non-drinking patients with multiple tumors: etiologic significance of p53 and Ki-67 in non-tumorous epithelium. J Oral Pathol Med 37, 549-554, doi:10.1111/j.1600-0714.2008.00657.x JOP657 [pii] (2008).   - 137 - 125 Wiseman, S. M. et al. Squamous cell carcinoma of the head and neck in nonsmokers and nondrinkers: an analysis of clinicopathologic characteristics and treatment outcomes. Ann Surg Oncol 10, 551-557 (2003). 126 van Oijen, M. G., Gilsing, M., Rijksen, G., Hordijk, G. & Slootweg, P. Increased number of proliferating cells in oral epithelium from smokers and ex-smokers. Oral oncology 34, 297-303 (1998). 127 van Oijen, M. G., van de Craats, J. G. & Slootweg, P. J. p53 overexpression in oral mucosa in relation to smoking. J Pathol 187, 469-474, doi:10.1002/(SICI)1096-9896(199903)187:4<469::AID-PATH262>3.0.CO;2-3 [pii] 10.1002/(SICI)1096-9896(199903)187:4<469::AID-PATH262>3.0.CO;2-3 (1999). 128 Brennan, J. A. et al. Association between cigarette smoking and mutation of the p53 gene in squamous-cell carcinoma of the head and neck. N Engl J Med 332, 712-717, doi:10.1056/NEJM199503163321104 (1995). 129 Sorensen, D. M. et al. Absence of p53 mutations in squamous carcinomas of the tongue in nonsmoking and nondrinking patients younger than 40 years. Arch Otolaryngol Head Neck Surg 123, 503-506 (1997). 130 Koch, W. M., Lango, M., Sewell, D., Zahurak, M. & Sidransky, D. Head and neck cancer in nonsmokers: a distinct clinical and molecular entity. Laryngoscope 109, 1544-1551, doi:10.1097/00005537-199910000-00002 (1999). 131 Singh, B. et al. Chromosomal aberrations in patients with head and neck squamous cell carcinoma do not vary based on severity of tobacco/alcohol exposure. BMC Genet 3, 22 (2002). 132 Santhi, W. et al. Oncogenic microRNAs as biomarkers of oral tumorigenesis and minimal residual disease. Oral oncology 49, 567-575 (2013). 133 Cervigne, N. K. et al. Identification of a microRNA signature associated with progression of leukoplakia to oral carcinoma. Human molecular genetics 18, 4818-4829 (2009). 134 Volinia, S. et al. A microRNA expression signature of human solid tumors defines cancer gene targets. Proceedings of the National Academy of Sciences of the United States of America 103, 2257-2261 (2006). 135 Houbaviy, H. B., Murray, M. F. & Sharp, P. A. Embryonic stem cell-specific MicroRNAs. Developmental cell 5, 351-358 (2003). 136 Chou, J. & Werb, Z. MicroRNAs Play a Big Role in Regulating Ovarian Cancer–Associated Fibroblasts and the Tumor Microenvironment. Cancer discovery 2, 1078-1080 (2012). 137 Gironella, M. et al. Tumor protein 53-induced nuclear protein 1 expression is repressed by miR-155, and its restoration inhibits pancreatic tumor development. Proceedings of the National Academy of Sciences of the United States of America 104, 16170-16175 (2007). 138 Caponi, S. et al. The good, the bad and the ugly: a tale of miR-101, miR-21 and miR-155 in pancreatic intraductal papillary mucinous neoplasms. Annals of oncology (2012). 139 Kong, W. et al. Upregulation of miRNA-155 promotes tumour angiogenesis by targeting VHL and is associated with poor prognosis and triple-negative breast cancer. Oncogene (2013). 140 Rather, M. I., Nagashri, M. N., Swamy, S. S., Gopinath, K. S. & Kumar, A. Oncogenic microRNA-155 down-regulates tumor suppressor CDC73 and promotes oral squamous   - 138 - cell carcinoma cell proliferation: implications for cancer therapeutics. Journal of Biological Chemistry 288, 608-618 (2013). 141 Eis, P. S. et al. Accumulation of miR-155 and BIC RNA in human B cell lymphomas. Proceedings of the National Academy of Sciences of the United States of America 102, 3627-3632 (2005). 142 Chen, H. et al. MicroRNA deregulation and pathway alterations in nasopharyngeal carcinoma. British journal of cancer 100, 1002-1011 (2009). 143 Chang, S. S. et al. MicroRNA alterations in head and neck squamous cell carcinoma. International Journal of Cancer 123, 2791-2797 (2008). 144 Ramdas, L. et al. miRNA expression profiles in head and neck squamous cell carcinoma and adjacent normal tissue. Head & neck 31, 642-654 (2009). 145 Kong, W. et al. MicroRNA-155 regulates cell survival, growth, and chemosensitivity by targeting FOXO3a in breast cancer. Journal of Biological Chemistry 285, 17869-17879 (2010). 146 Kong, W. et al. MicroRNA-155 is regulated by the transforming growth factor β/Smad pathway and contributes to epithelial cell plasticity by targeting RhoA. Molecular and cellular biology 28, 6773-6784 (2008). 147 Wong, D. T. W. TGF-α and oral carcinogenesis. European Journal of Cancer. Part B: Oral Oncology 29, 3-7 (1993). 148 Chen, A. M. et al. Head and neck cancer among lifelong never-smokers and ever-smokers: matched-pair analysis of outcomes after radiation therapy. American Journal of Clinical Oncology 34, 270 (2011). 149 Agarwal, J. P. et al. Prognostic factors in oropharyngeal cancer--analysis of 627 cases receiving definitive radiotherapy. Acta Oncol 48, 1026-1033, doi:10.1080/02841860902845839 (2009). 150 Stenvang, J., Petri, A., Lindow, M., Obad, S. & Kauppinen, S. Inhibition of microRNA function by antimiR oligonucleotides. Silence 3, 1 (2012). 151 Lindow, M. & Kauppinen, S. Discovering the first microRNA-targeted drug. J Cell Biol 199, 407-412 (2012). 152 Supek, F., Bošnjak, M., Škunca, N. & Šmuc, T. REVIGO summarizes and visualizes long lists of gene ontology terms. PLoS One 6, e21800 (2011). 153 Kornberg, L. J. Focal adhesion kinase and its potential involvement in tumor invasion and metastasis. Head & neck 20, 745-752 (1998). 154 Rhee, I. et al. DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature 416, 552-556 (2002). 155 Baylin, S. B. & Herman, J. G. DNA hypermethylation in tumorigenesis: epigenetics joins genetics. Trends in Genetics 16, 168-174 (2000). 156 Herman, J. G. in Seminars in cancer biology.  359-367 (Elsevier). 157 Kisiel, J. B., Garrity-Park, M. M., Taylor, W. R., Smyrk, T. C. & Ahlquist, D. A. Methylated Eyes Absent 4 (EYA4) gene promotor in non-neoplastic mucosa of ulcerative colitis patients with colorectal cancer: evidence for a field effect. Inflammatory bowel diseases 19, 2079-2083 (2013). 158 Cook, P. J. et al. Tyrosine dephosphorylation of H2AX modulates apoptosis and survival decisions. Nature 458, 591-596 (2009).   - 139 - 159 Wilson, I. et al. EYA4 is inactivated biallelically at a high frequency in sporadic lung cancer and is associated with familial lung cancer risk. Oncogene (2013). 160 Chang, S. E., Foster, S., Betts, D. & Marnock, W. E. DOK, a cell line established from human dysplastic oral mucosa, shows a partially transformed non‐malignant phenotype. International Journal of Cancer 52, 896-902 (1992). 161 Osborn, N. K. et al. Aberrant Methylation of the Eyes Absent 4 Gene in Ulcerative Colitis–Associated Dysplasia. Clinical Gastroenterology and Hepatology 4, 212-218 (2006). 162 Clark, S. W., Fee, B. E. & Cleveland, J. L. Misexpression of the eyes absent family triggers the apoptotic program. Journal of Biological Chemistry 277, 3560-3567 (2002). 163 Gyori, B. M., Venkatachalam, G., Thiagarajan, P., Hsu, D. & Clement, M.-V. OpenComet: An automated tool for comet assay image analysis. Redox biology 2, 457-465 (2014). 164 Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to ImageJ: 25 years of image analysis. Nature methods 9, 671-675 (2012). 165 Dickson, M. A. et al. Human keratinocytes that express hTERT and also bypass a p16INK4a-enforced mechanism that limits life span become immortal yet retain normal growth and differentiation characteristics. Molecular and cellular biology 20, 1436-1447 (2000). 166 Miller, S. J. et al. Inhibition of Eyes Absent Homolog 4 expression induces malignant peripheral nerve sheath tumor necrosis. Oncogene 29, 368-379 (2010). 167 Pandey, R. N. et al. The Eyes Absent phosphatase-transactivator proteins promote proliferation, transformation, migration, and invasion of tumor cells. Oncogene 29, 3715-3722 (2010). 168 Kim, J. J. & Grady, W. M. Epigenomic analysis of aberrantly methylated genes in colorectal cancer identifies genes commonly affected by epigenetic alterations. Annals of surgical oncology 18, 2338-2347 (2011). 169 Rayapureddi, J. P. et al. Eyes absent represents a class of protein tyrosine phosphatases. Nature 426, 295-298 (2003). 170 Tootle, T. L. et al. The transcription factor Eyes absent is a protein tyrosine phosphatase. Nature 426, 299-302 (2003). 171 Li, X. et al. Eya protein phosphatase activity regulates Six1–Dach–Eya transcriptional effects in mammalian organogenesis. Nature 426, 247-254 (2003). 172 Paull, T. T. et al. A critical role for histone H2AX in recruitment of repair factors to nuclear foci after DNA damage. Current Biology 10, 886-895 (2000). 173 White, R. J. RNA polymerases I and III, growth control and cancer. Nature reviews. Molecular cell biology 6, 69-78, doi:10.1038/nrm1551 (2005). 174 Zhao, J., Yuan, X., Frodin, M. & Grummt, I. ERK-dependent phosphorylation of the transcription initiation factor TIF-IA is required for RNA polymerase I transcription and cell growth. Molecular cell 11, 405-413 (2003). 175 Blagden, S. P. & Willis, A. E. The biological and therapeutic relevance of mRNA translation in cancer. Nature reviews. Clinical oncology 8, 280-291, doi:10.1038/nrclinonc.2011.16 (2011). 176 Albertson, D. G. Gene amplification in cancer. Trends in genetics : TIG 22, 447-455, doi:10.1016/j.tig.2006.06.007 (2006).   - 140 - 177 Lockwood, W. W. et al. DNA amplification is a ubiquitous mechanism of oncogene activation in lung and other cancers. Oncogene (2008). 178 Garnis, C. et al. Genomic imbalances in precancerous tissues signal oral cancer risk. Molecular cancer 8, 50, doi:10.1186/1476-4598-8-50 (2009). 179 Rosin, M. P. et al. Use of allelic loss to predict malignant risk for low-grade oral epithelial dysplasia. Clinical cancer research : an official journal of the American Association for Cancer Research 6, 357-362 (2000). 180 Lee, J. S. et al. Detection of chromosomal polysomy in oral leukoplakia, a premalignant lesion. J Natl Cancer Inst 85, 1951-1954 (1993). 181 Tsui, I. F., Rosin, M. P., Zhang, L., Ng, R. T. & Lam, W. L. Multiple aberrations of chromosome 3p detected in oral premalignant lesions. Cancer Prev Res (Phila Pa) 1, 424-429, doi:1/6/424 [pii] 10.1158/1940-6207.CAPR-08-0123 [doi] (2008). 182 Tsui, I. F. et al. Multiple pathways in the FGF signaling network are frequently deregulated by gene amplification in oral dysplasias. Int J Cancer 125, 2219-2228, doi:10.1002/ijc.24611 (2009). 183 Cervigne, N. K. et al. Recurrent genomic alterations in sequential progressive leukoplakia and oral cancer: drivers of oral tumorigenesis? Human molecular genetics, ddt657 (2014). 184 Poh, C. et al. Unique FISH patterns associated with cancer progression of oral dysplasia. Journal of dental research 91, 52-57 (2012). 185 el-Naggar, A. K. et al. Sequential loss of heterozygosity at microsatellite motifs in preinvasive and invasive head and neck squamous carcinoma. Cancer Res 55, 2656-2659 (1995). 186 Papadimitrakopoulou, V. et al. Frequent inactivation of p16INK4a in oral premalignant lesions. Oncogene 14, 1799-1803, doi:10.1038/sj.onc.1201010 (1997). 187 Ishkanian, A. S. et al. A tiling resolution DNA microarray with complete coverage of the human genome. Nature genetics 36, 299-303, doi:10.1038/ng1307 (2004). 188 Watson, S. K., DeLeeuw, R. J., Horsman, D. E., Squire, J. A. & Lam, W. L. Cytogenetically balanced translocations are associated with focal copy number alterations. Hum Genet. 120, 795-805 (2007). 189 Tsui, I. F. & Garnis, C. Integrative molecular characterization of head and neck cancer cell model genomes. Head Neck 32, 1143-1160, doi:10.1002/hed.21311 (2010). 190 Chi, B. et al. MD-SeeGH: a platform for integrative analysis of multi-dimensional genomic data. BMC Bioinformatics 9, 243, doi:10.1186/1471-2105-9-243 (2008). 191 Khojasteh, M., Lam, W. L., Ward, R. K. & MacAulay, C. A stepwise framework for the normalization of array CGH data. BMC Bioinformatics 6, 274 (2005). 192 Towle, R. et al. Recurring DNA copy number gain at chromosome 9p13 plays a role in the activation of multiple candidate oncogenes in progressing oral premalignant lesions. Cancer medicine 3, 1170-1184 (2014). 193 Poh, C. et al. Unique FISH patterns associated with cancer progression of oral dysplasia. Journal of dental research 91, 52-57 (2012). 194 Cappuzzo, F. et al. Increased HER2 gene copy number is associated with response to gefitinib therapy in epidermal growth factor receptor–positive non–small-cell lung cancer patients. Journal of Clinical Oncology 23, 5007-5018 (2005).   - 141 - 195 Wang, Q., Song, C. & Li, C. C. Molecular perspectives on p97-VCP: progress in understanding its structure and diverse biological functions. J Struct Biol 146, 44-57, doi:10.1016/j.jsb.2003.11.014 (2004). 196 Kamradt, J. et al. Detection of novel amplicons in prostate cancer by comprehensive genomic profiling of prostate cancer cell lines using oligonucleotide-based arrayCGH. PLoS One 2, e769, doi:10.1371/journal.pone.0000769 (2007). 197 Cao, W. F. et al. Prognostic significance of stomatin-like protein 2 overexpression in laryngeal squamous cell carcinoma: clinical, histologic, and immunohistochemistry analyses with tissue microarray. Human pathology 38, 747-752, doi:10.1016/j.humpath.2006.11.004 (2007). 198 Yamamoto, S. et al. Expression level of valosin-containing protein (VCP) as a prognostic marker for gingival squamous cell carcinoma. Ann Oncol 15, 1432-1438, doi:10.1093/annonc/mdh354 (2004). 199 Zhang, L. et al. Stomatin-like protein 2 is overexpressed in cancer and involved in regulating cell growth and cell adhesion in human esophageal squamous cell carcinoma. Clinical cancer research : an official journal of the American Association for Cancer Research 12, 1639-1646, doi:10.1158/1078-0432.ccr-05-1858 (2006). 200 Fessart, D., Marza, E., Taouji, S., Delom, F. & Chevet, E. P97/CDC-48: proteostasis control in tumor cell biology. Cancer letters (2013). 201 Chang, D. et al. SLP-2 overexpression is associated with tumour distant metastasis and poor prognosis in pulmonary squamous cell carcinoma. Biomarkers 15, 104-110 (2010). 202 Tsui, I. F., Garnis, C. & Poh, C. F. A dynamic oral cancer field: unraveling the underlying biology and its clinical implication. Am J Surg Pathol 33, 1732-1738, doi:10.1097/PAS.0b013e3181b669c2 (2009). 203 van Houten, V. M. et al. Molecular diagnosis of surgical margins and local recurrence in head and neck cancer patients: a prospective study. Clinical cancer research : an official journal of the American Association for Cancer Research 10, 3614-3620, doi:10.1158/1078-0432.ccr-03-0631 (2004). 204 Buys, T. P., Aviel-Ronen, S., Waddell, T. K., Lam, W. L. & Tsao, M. S. Defining genomic alteration boundaries for a combined small cell and non-small cell lung carcinoma. J Thorac Oncol 4, 227-239, doi:10.1097/JTO.0b013e3181952678 (2009). 205 Navin, N. et al. Inferring tumor progression from genomic heterogeneity. Genome Res 20, 68-80, doi:10.1101/gr.099622.109 (2010). 206 Ruiz, C. et al. Advancing a clinically relevant perspective of the clonal nature of cancer. Proc Natl Acad Sci U S A, doi:10.1073/pnas.1104009108 (2011). 207 Chari, R. et al. SIGMA2: a system for the integrative genomic multi-dimensional analysis of cancer genomes, epigenomes, and transcriptomes. BMC Bioinformatics 9, 422, doi:10.1186/1471-2105-9-422 (2008). 208 Guled, M. et al. Array comparative genomic hybridization analysis of olfactory neuroblastoma. Mod Pathol 21, 770-778, doi:10.1038/modpathol.2008.57 (2008). 209 Muleris, M., Almeida, A., Gerbault-Seureau, M., Malfoy, B. & Dutrillaux, B. Detection of DNA amplification in 17 primary breast carcinomas with homogeneously staining regions by a modified comparative genomic hybridization technique. Genes Chromosomes Cancer 10, 160-170 (1994).   - 142 - 210 Zack, T. I. et al. Pan-cancer patterns of somatic copy number alteration. Nature genetics 45, 1134-1140 (2013). 211 Lai, M. T. et al. Talin-1 overexpression defines high risk for aggressive oral squamous cell carcinoma and promotes cancer metastasis. J Pathol 224, 367-376, doi:10.1002/path.2867 (2011). 212 Pickering, C. R. et al. Integrative genomic characterization of oral squamous cell carcinoma identifies frequent somatic drivers. Cancer discovery (2013). 213 Thorvaldsdóttir, H., Robinson, J. T. & Mesirov, J. P. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Briefings in bioinformatics 14, 178-192 (2013). 214 Gao, J. et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Science signaling 6, pl1 (2013). 215 Li, X.-H. et al. Increased expression of stomatin-like protein 2 (STOML2) predicts decreased survival in gastric adenocarcinoma: a retrospective study. Medical Oncology 31, 1-7 (2014). 216 Karki, S., LaMonte, B. & Holzbaur, E. L. Characterization of the p22 subunit of dynactin reveals the localization of cytoplasmic dynein and dynactin to the midbody of dividing cells. J Cell Biol 142, 1023-1034 (1998). 217 Abba, M. C. et al. Breast cancer molecular signatures as determined by SAGE: correlation with lymph node status. Mol Cancer Res 5, 881-890, doi:10.1158/1541-7786.mcr-07-0055 (2007). 218 Brown, L. A. et al. Amplification of 11q13 in ovarian carcinoma. Genes Chromosomes Cancer 47, 481-489, doi:10.1002/gcc.20549 (2008). 219 Carvalho, B. et al. Multiple putative oncogenes at the chromosome 20q amplicon contribute to colorectal adenoma to carcinoma progression. Gut 58, 79-89, doi:10.1136/gut.2007.143065 (2009). 220 Wu, J. et al. Identification and functional analysis of 9p24 amplified genes in human breast cancer. Oncogene, doi:10.1038/onc.2011.227 (2011). 221 Chari, R. et al. Integrating the multiple dimensions of genomic and epigenomic landscapes of cancer. Cancer and Metastasis Reviews 29, 73-93 (2010). 222 Cahill, D. P., Kinzler, K. W., Vogelstein, B. & Lengauer, C. Genetic instability and darwinian selection in tumours. Trends in cell biology 9, M57-M60 (1999). 223 Gisselsson, D. et al. Chromosomal breakage-fusion-bridge events cause genetic intratumor heterogeneity. Proceedings of the National Academy of Sciences 97, 5357-5362 (2000). 224 Garnis, C. et al. Chromosome 5p aberrations are early events in lung cancer: implication of glial cell line-derived neurotrophic factor in disease progression. Oncogene 24, 4806-4812 (2005). 225 Tsujimoto, Y. et al. Elevated expression of valosin-containing protein (p97) is associated with poor prognosis of prostate cancer. Clinical cancer research 10, 3007-3012 (2004). 226 Yamamoto, S. et al. Expression level of valosin-containing protein (p97) is associated with prognosis of esophageal carcinoma. Clinical cancer research 10, 5558-5565 (2004). 227 Meyer, M. F. et al. Valosin-containing protein (VCP/p97)-expression correlates with prognosis of HPV-negative oropharyngeal squamous cell carcinoma (OSCC). PLoS One 9, e114170 (2014).   - 143 - 228 Magnaghi, P. et al. Covalent and allosteric inhibitors of the ATPase VCP/p97 induce cancer cell death. Nature chemical biology 9, 548-556 (2013).                                              - 144 - Appendix A A.1 Chapter 4 Validation of miRNA Expression in Individual qPCR Assays  Dysplasia    Patient Sample miRNA Full Panel Fold Change Individual Assay Fold Change 2000 miR 135 0.0312 0.0041 2000 miR 143 0.3268 0.2639 2000 miR 224 2.3251 1.8846 2076 miR 135a 0.2783 0.113 2076 miR 143 0.3765 0.1726 2076 miR 224 2.4281 0.4511 Tumor     Patient Sample miRNA Full Panel Fold Change Individual Assay Fold Change 7071 miR-605 0.0638 0.5094 2000 miR 605 0.2602 0.2521 2000 miR 223 4.0465 6.7491 2076 miR 223 4.1724 7.0273 7071 miR-155 6.6449 12.9842 2000 miR 155 10.6123 12.0069  A.2  Chapter 7 Tissue Microarray Patient Demographic Information Patient # Site Dx Age Sex Smoking Status 1 Floor of Mouth D1, D2 41 F S 2 Floor of Mouth D2, VH 49 F S 3 R Ventral Tongue D3 71 F S 4 R Lateral Tongue CIS, D2 48 M NA 5 Floor of Mouth D2 64 M NA 6 L Ventral Tongue D3 58 M FS 7 R Tongue D2 45 M NA 8 L Palate D2 64 F S 9 L Palate D1 48 M S 10 R Palate D2 48 M S 11 L Ventral Tongue D1-D2 70 F FS   - 145 -  Patient #  Site  Dx  Age  Sex  Smoking Status 12 R Lower LIP D2 29 M S 13 L Retromolar Ridge D3 60 M FS 14 FOM D2-D3 42 M S 15 L. Lingal Mand/Front of Mouth D2 44 F NS 16 Anterior Ventral. Tongue D1, D2 59 M FS 17 L. Lateral Tongue D3 52 F NS 18 R. Lateral Tongue D1, D2 52 F NS 19 R. Floor of Mouth D2 63 M S 20 R. Lateral Tongue D2 62 F FS 21 R. Ventral. Tongue D2 46 M S 22 R. Anterior Pillar D2 69 M FS 23 L. Lateral Tongue D1 50 F S 24 L. Soft Palate D1 69 M NS 25 L. Lateral Tongue D2 65 M NS 26 R.Posterior Lateral and Ventral Surface of Tongue D2 54 M NS 27 L. Lateral Tongue D2 62 M S 28 Junction R. Ventral Tongue and Floor of Mouth D1, D2 61 M S 29 L. Lateral Tongue D2 47 M NS 30 L. Anterior Ventral Tongue D2 63 F FS 31 R. Retromolar Region D1, D2 69 F S 32 L. Ventral. Tongue D2 42 M NS 33 L. Mid. Ventral. Tongue D2 58 M S 34 R. Ventral. Tongue D2 78 M NS 35 R. Ventral. Tongue D1 78 M NS 36 L Lateral Tongue D3 49 M NS 37 Right Anterior Pillar CIS 72 M FS 38 R Lat Toongue D3 59 M S 39 R Ventral Tongue CIS 71 F S 40 R Anterior Lateral Tongue D3 24 M NS 41 R Posteror Lateral Tongue D3 68 M FS 42 36/37 Region CIS 59 F S 43 L Ventral Tongue D2 69 F FS 44 R Tongue CIS 62 F NS 45 L Ventral Tongue CIS 69 M FS 46 R Lateral Tongue D3 45 F S 47 L Lateral Tongue D3 55 F NS 48 L Floor of Mouth CIS 54 M S 49 R Anterior VentralTongue CIS 52 M S   - 146 - 50 L Anterior Ventral Tongue CIS 61 M FS Patient # Site Dx Age Sex Smoking Status       51 L Anterior Lateral Tongue D3 51 M S 52 L Lateral Tongue SCC 55 M S 53 1. L and R Cheek 2. Floor of Mouth CIS 50 M S 54 L Floor of Mouth CIS 55 F FS 55 L VentralRAL Tongue CIS 54 M S 56 Posteriorr AND Mid Left Lateral Tongue D3 53 F NS 57 R Lateral Tongue D3 57 M NS 58 L Lateral Tongue CIS 45 M FS 59 Anterior Floor of Mouth CIS 58 M NS 60 R Ventral Tongue CIS 53 M S 61 Lingual of Mand. Alveolar CIS 62 M NA 62 Gingiva CIS 69 M FS 63 Tongue D2 75 M NA  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0313412/manifest

Comment

Related Items