Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Bioinformatics approach to investigate genetic differences underlying breast tumours with specific outcomes… Chun, Hye-Jung 2010

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2010_fall_chun_hyejung.pdf [ 25.42MB ]
Metadata
JSON: 24-1.0071196.json
JSON-LD: 24-1.0071196-ld.json
RDF/XML (Pretty): 24-1.0071196-rdf.xml
RDF/JSON: 24-1.0071196-rdf.json
Turtle: 24-1.0071196-turtle.txt
N-Triples: 24-1.0071196-rdf-ntriples.txt
Original Record: 24-1.0071196-source.json
Full Text
24-1.0071196-fulltext.txt
Citation
24-1.0071196.ris

Full Text

BIOINFORMATICS APPROACH TO INVESTIGATE GENETIC DIFFERENCES UNDERLYING BREAST TUMOURS WITH SPECIFIC OUTCOMES OF ADOPTIVE T-CELL THERAPY USING A MOUSE MODEL by Hye-Jung Chun B.Sc., The University of British Columbia, 2004 B.Sc., The University of British Columbia, 2000  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in The Faculty of Graduate Studies (Bioinformatics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2010 © Hye-Jung Chun, 2010  i  ABSTRACT The immune system plays a critical role in cancer prevention and development. The stimulation of natural immune reaction in a cancer patient by adoptive T-cell therapy has shown success in treating metastatic melanomas and renal cell carcinomas. However, the use of adoptive T-cell therapy remains limited due to unpredictable outcomes and low response rates. In particular, adoptive T-cell therapy for breast cancer has not been realized, despite of the presence of immunogenic antigens such as over-expressed HER2, present in 20-40% of breast tumours. Using a unique transgenic mouse model, the global profiles of gene expression, miRNA abundance and single nucleotide variants (SNVs) were investigated to identify the molecular difference of murine mammary tumours with isogenic background, which exhibited complete regression (CR), partial regression (PR) or progressive disease (PD) outcome of adoptive T-cell therapy. The bioinformatics analyses were further carried out to identify uniquely activated pathways, prognostic gene expression signatures, the effect of post-transcriptional gene regulation and mutated genes unique to tumours with specific outcome. The largest differences in gene expression, miRNA and SNV profiles were repeatedly observed between the regressing (CR, PR) and non-regressing (PD) tumours, supporting the attribution of molecular differences to the immunotherapy outcome. In particular, the gene expression signatures derived from genes in immune-related pathways were experimentally validated to be strong prognostic markers for predicting the CR outcome. Comparison with the human breast cancer subtypes further revealed similarities of the non-regressing tumours with the basal subtype, and the regressing tumours with the  ii  HER2 subtype. The difference in miRNA profiles between CR and PR tumours suggested potential translational activities unique to PR, which was nearly identical to CR at the transcriptome level. The findings from this study show that tumour-derivied factors that either promote or suppress the immune system are responsible for the varying outcome of immunotherapy, and that the molecular characteristics can be further applied for the development of clinical prognostic tools, cancer vaccines and drug targets to enhance the efficacy of adoptive T-cell therapy.  iii  TABLE OF CONTENTS ABSTRACT....................................................................................................................... ii TABLE OF CONTENTS ................................................................................................ iv LIST OF TABLES ........................................................................................................... vi LIST OF FIGURES ........................................................................................................ vii ACKNOWLEDGEMENTS .......................................................................................... viii DEDICATION .................................................................................................................. x CO-AUTHORSHIP STATEMENT ............................................................................... xi 1 INTRODUCTION ........................................................................................................ 1 1.1 BIBLIOGRAPHY ...............................................................................................................8  2 GENE EXPRESSION PROFILES ASSOCIATED WITH THE OUTCOME OF ADOPTIVE T-CELL THERAPY IN A MOUSE MODEL OF BREAST CANCER ........................................................................................................................................... 18 2.1 INTRODUCTION.............................................................................................................18 2.2 MATERIAL AND METHODS........................................................................................24 2.2.1 Mouse tumour samples and microarray experiments ..................................................24 2.2.2 Microarray data analysis ..............................................................................................25 2.2.3 Experimental validation of trichostatin A in vivo ........................................................27 2.2.4 Outcome-specific differential expression, classifier identification and pathway enrichment analyses ...............................................................................................................28 2.2.5 Validation of outcome-specific DE genes ...................................................................29 2.2.6 Comparison with molecular profiles from publicly available human breast cancer subtypes .................................................................................................................................30 2.3 RESULTS...........................................................................................................................32 2.3.1 Gene expression profiles of CR, PR and PD tumours .................................................32 2.3.2 Discovery of tricostatin A as a potential therapeutic drug using the Connectivity Map database..................................................................................................................................35 2.3.3 Pathway enrichment analysis of outcome-specific genes............................................39 2.3.4 Validation of the predictive gene expression signature ...............................................47 2.3.5 Comparison with human breast cancer subtypes .........................................................59 2.4 DISCUSSION ....................................................................................................................67 2.5 BIBLIOGRAPHY .............................................................................................................76  3 CHARACTERIZATION OF microRNA PROFILES AND POTENTIAL POSTTRANSCRIPTIONAL REGULATIONS IN GENE EXPRESSION OF BREAST TUMOURS WITH VARYING OUTCOMES AFTER ADOPTIVE T-CELL THERAPY ....................................................................................................................... 86 3.1 INTRODUCTION.............................................................................................................86 3.2 MATERIAL AND METHODS........................................................................................99 3.2.1 Murine tumour samples and small RNA library preparation ......................................99 3.2.2 Massively parallel sequencing of small RNAs by ABI SOLiDTM .............................101 3.2.3 Genome mapping and annotation of small RNA sequences......................................101 3.2.4 Detection of differentially abundant miRNAs...........................................................103  iv  3.2.5 Gene expression changes of predicted target genes of differentially abundant miRNAs .............................................................................................................................................104 3.3 RESULTS.........................................................................................................................106 3.3.1 Sequencing small RNA using SOLiD sequencing platform......................................106 3.3.2 Annotation of small RNAs ........................................................................................118 3.3.3 Differential abundance of miRNAs in tumours with different adoptive immunotherapy outcomes....................................................................................................122 3.3.4 Possible effect of differentially abundant microRNA on the expression of predicted target genes ..........................................................................................................................126 3.4 DISCUSSION ..................................................................................................................133 3.5 BIBLIOGRAPHY ..........................................................................................................141  4 MUTATION PROFILE OF TUMOURS WITH DIFFERENT OUTCOMES AFTER ADOPTIVE T-CELL THERAPY BY HIGH THROUGHPUT TRANSCRIPTOME SEQUENCING ......................................................................... 150 4.1 INTRODUCTION...........................................................................................................150 4.2 MATERIAL AND METHODS......................................................................................154 4.2.1 Mouse breast tumour model ......................................................................................154 4.2.2 RNA-seq library construction and sequencing by ABI SOLiDTM technology ..........155 4.2.3 Genome mapping and novel SNV discovery.............................................................157 4.2.4 Identifying the chromosomal segment that originated from the FVB mouse strain..159 4.2.5 Pathway enrichment and functional enrichment analysis..........................................160 4.3 RESULTS.........................................................................................................................160 4.3.1 Sequencing and mapping of short RNA sequences generated by SOLiD .................160 4.3.2 Discovery of an FVB chromosomal segment in the C57BL/6J genome ...................163 4.3.3 Putative mutation discovery pipeline.........................................................................167 4.3.4 Validation of selected putative mutations by PCR amplicon sequencing .................173 4.4 DISCUSSION ..................................................................................................................175 4.5 BIBLIOGRAPHY ...........................................................................................................180  5 CONCLUSION ......................................................................................................... 184 5.1 BIBLIOGRAPHY ...........................................................................................................193  v  LIST OF TABLES Table 2.1 List of qPCR-validated genes .......................................................................... 49! Table 3.1 Mapping result of miRNA libraries using the MAQ aligner ......................... 107! Table 3.2 Mapping result of miRNA libraries using the SHRiMP aligner.................... 108! Table 3.3 Mapping rate of miRNA sequnces using SHRiMP after trimming the last 10 base pairs................................................................................................................. 117! Table 3.4 Annotation of various RNA classes and genomic elements .......................... 120! Table 3.5 Differentially abundant miRNAs between tumours with different outcome of adoptive T-cell therapy ........................................................................................... 125! Table 4.1 The mapping rate of sequences derived from the mRNA transcriptomes in the RNA-seq experiment using SOLiD ........................................................................ 162! Table 4.2 Top ranking pathways enriched for mutated genes ....................................... 171! Table 4.3 Top ranking biological annotation significantly associated with mutated genes ................................................................................................................................. 172! Table 4.4 List of genes and putative mutations attempted for experimental validation using PCR ............................................................................................................... 174!  vi  LIST OF FIGURES Figure 2.1 The clustering of different NOP tumours....................................................... 33! Figure 2.2 Progression of tumours treated with TSA before adoptive T-cell therapy..... 37! Figure 2.3 Intratumoural infiltration of T-cells after TSA treatment............................... 38! Figure 2.4 Identification of ‘outcome-specific’ differentially expressed genes .............. 39! Figure 2.5 Heatmaps of the outcome-specific gene expression signatures ..................... 42! Figure 2.6 Gene expression level of selected genes for validation using qPCR.............. 49! Figure 2.7 The process of merging and analysis of gene expression profiles from human breast cancer subtypes and mouse NOP tumours to identify the CBC gene set....... 61! Figure 2.8 Comparison of the mouse tumour expression profile with human breast cancer subtypes..................................................................................................................... 63! Figure 2.9 Significance clustering result of 82 CBC gene sets over the merged data from the mouse and human breast tumours....................................................................... 66! Figure 3.1 Phred-like quality of colours called at each base position of 35bp-long sequence in small RNA libraries sequenced by SOLiD ......................................... 109! Figure 3.2 Distribution of log10-transformed SHRiMP statistics ................................. 111! Figure 3.3 Percentages of different colours called at each base across sequences from randomly selected libraries ..................................................................................... 114! Figure 3.4 Distribution of non-coding small RNA classes across all libraries sequenced using SOLiD ........................................................................................................... 119! Figure 3.5 Differentially abundant miRNAs from multiple outcome class comparisons in an area-proportional Venn diagram ........................................................................ 124! Figure 3.6 Cumulative distribution functions and histograms of log2-fold changes of predicted target genes and randomly selected genes from 1,000 iterations............ 129! Figure 4.1 Tumour SNVs that matched SNPs from the Broad mouse Hapmap data generated from 94 laboratory mouse strains using the Affymetrix SNP arrays ..... 164! Figure 4.2 Filtering process to discover putative non-synonymous mutations from tumours subjected to RNA-seq ............................................................................... 168! Figure 4.3 Comparisons of genes that carried the putative non-synonymous mutations detected in tumours from four cell lines sequenced................................................ 169!  vii  ACKNOWLEDGEMENTS My thesis research would not have been possible without Dr. Steven Jones, my supervisor, who gave me the opportunity to embark on what I can only describe as having been a wonderful experience as a graduate student at the Genome Sciences Centre. During the three years of my studies, he continued to provide me with the intellectual support and guidance for my research. His scientific insight often kept me away from following the ‘rabbit holes’ and his thought-provoking questions helped me think more critically during many meetings and discussions. I am most grateful for his wonderful supervision. I also offer my sincere gratitude to Dr. Brad Nelson for his prompt feedback, penetrating questions and attention to details, which helped me see the connections between findings of my research. I also thank him for his precious time taken to fly over from Victoria to attend several meetings in person, and attend numerous teleconference calls. His deep knowledge in immunology also helped me immensely put my bioinformatics analysis results into the immunological perspective. I especially thank my committee member, Dr. Isabella Tai for her stimulating questions during the meetings and for her enthusiasm to help even two days before she gave birth to her daughter. I owe special thanks to Michele Martin, my experimental immunologist liaison, whose insight and knowledge in immunology were vital for my research. Her enthusiasm in this research was critical for the fruition of this research project.  viii  Finally, my heartfelt gratitude goes to Ms. Louise Clarke and Sharon Ruschkowski for all her administrative support with smiles, and to all my colleagues at the Genome Sciences Centre who provided emotional support, friendship as well as countless scientific advice and constructive criticism that immensely helped my research be productive.  ix  DEDICATION  I dedicate this work !  To my remarkable family into whom I had the miraculous fortune to be born my parents, James Gun-Gill & Susanna Ok-Ja Chun, who always back me up with their unfailing love and support, my sis, Helena Hye-Won, for her wit and incredible insight that help me see. !  To my aunt Yoon-Hee, uncle Gwang-Soo, Jong-Sung, and my friend Louisa, who are braving their battles against cancer, who kept me going forward. !  And most of all to J.C. Without Whom I can do nothing.  x  CO-AUTHORSHIP STATEMENT The chapter 2 of this thesis was co-written by myself, Michele L. Martin, Allen Delaney, Brad H. Nelson and Steven J.M. Jones. The overall design of this research project was formulated by Brad H. Nelson, Allen Delaney and Steven J.M. Jones. The preparation of tumour materials, the experimental validation and analyses of the validation result were performed by Michele L. Martin at Trev & Joyce Deeley Research Centre, BCCA in Victoria. The protocols used for mouse experiments conducted by Michele L. Martin were approved by the University of Victoria research ethics board (Protocol number #2008-036, “Analysis of the immune response to cancer in murine models” and #2009-001, “Breeding wild type and genetically modified mice for tumour biology and immunology studies”). The processing of RNA and microarray experiments was carried out in the Centre for Translational and Applied Genomics at the British Columbia Cancer Agency. I carried out all bioinformatics analyses described in chapter 2 and wrote the manuscript. The library construction and the SOLiD sequencing of microRNAs and transcriptomes of murine mammary tumours were performed by Yongjun Zhao, Kevin Ma, Thomas Zeng, Angela Tam and Martin Hirst at the Genome Sciences Centre. I carried out all bioinformatics analysis described in chapters 3 and 4 and wrote the manuscripts.  xi  1 INTRODUCTION The immune system has long been recognized to play an important role in cancer prevention and progression. The first observation of tumour regression caused by immune system was made by William Coley 117 years ago when he discovered the shrinkage in tumour size followed by an acute infection by Streptococcus pyogenes [1]. Many lines of evidence following this first observation further supported the anti-tumour activities of the immune system and led to the concept of ‘immunosurveillance’, which described the protective role of the immune system in promoting tumour suppression and regression [2-7]. Indeed, many studies involving mice and humans showed that hosts with defective immune system were presented with tumours spontaneously rising at a significantly higher frequency compared to the immunocompetent hosts [8-15]. Conversely, transplanted tumours in immunocomptent mice were frequently rejected [9, 16]. Also, the presence of tumour-infiltrating CD4+ helper and CD8+ cytotoxic T lymphocytes (CTLs) was repeatedly shown to strongly associate with good prognosis in many cancer types such as colon [17-20], cervix [21], ovary [22, 23], skin [24-26], lung [27, 28], breast [29-32], and lymphomas [33]. Thus, immunological approaches to treat cancer have long been investigated. Currently, two major clinical strategies have been developed: 1) cancer vaccines that stimulate the host immune system via active immunization against cancer-associated antigens; 2) adoptive T-cell therapy which augments the anti-tumour response through transfusion of a large number of ex vivo expanded tumour-reactive CD4+ helper T-cells and CD8+ CTLs [34-39].  1  Cancer vaccines can be designed to target proteins that are exclusively expressed in cancerous tissues such as NY-BR-1 [40] or mutated in cancer e.g. p53 [41]. Vaccines can be used to prevent cancer via prophylactic vaccination, or to treat cancer via therapeutic vaccination. Hundreds of cancer vaccines have been evaluated in Phase I, II or III clinical trials to date [42]. Also, adoptive T-cell therapy has increasingly shown promise in treating renal cell carcinoma and metastatic melanoma, in which the objective response rate (50% reduction in total tumour size) was observed in approximately half of the patients subjected to adoptive T-cell therapy [43, 44]. A successful immune response to tumour begins with the recognition of immunogenic epitopes presented on the cell surface by major histocompatibility complex (MHC) molecules. After antigen recognition, the activated CTLs migrate and infiltrate the tumour epithelium and engage in cytolytic granule- or ligand-mediated apoptosis of cancer cells. Another unique consequence of anti-cancer immune response is the generation of lasting immunity against cancer through the formation of memory T- and B-cells [2, 11, 37, 45, 46]. One of the key components for effective anti-tumour immune response is the presence of tumour antigens that are sufficiently immunogenic and tumour-specific. Tumour antigens include tissue-differentiation proteins, e.g. Melan-A/MART-1, tyrosinase, gp100/Pmel17 [8, 36, 46, 47]; mutational aberrations, e.g. mutated p53, Ras, cyclindependent kinase 4 (CDK4)-R24C, BCR-abl [8, 36, 46, 48, 49]; over-expressed proteins, e.g. HER2/neu [8, 47]; proteins of viral origin, e.g. HPV-16 E7 [8, 46-48]; or aberrantly expressed proteins such as NY-BR-1 and cancer-testis antigens [8, 48, 50]. More than a  2  thousand tumour antigens have already been described [51], representing a large repertoire of potential immunotherapy targets. Despite the potential advantages of immunotherapy and the discovery of many potential cancer-specific antigens, the clinical application of cancer immunotherapy is currently minimal, due to unpredictable outcomes and low response rates achieved to date [2, 37]. For example, cancer vaccines have achieved an overall objective response rate of only 2.6% [37]. In the case of breast cancer treatment, the use of adoptive T-cell therapy or cancer vaccines has not been currently realized despite the immunogenic HER2 antigen found in 20-40% of the breast cancer patients [48, 52-56]. Several clinical trials of a HER2/neu vaccine demonstrated the increased immune response against tumours with HER2 over-expression, demonstrating the immunogenicity of HER2 antigen [114-117]. The intricate interactions between the humoral and cellular components of the immune system, and between the immune system and cancer cells contribute to the overall antitumour immune response. Coupled with the heterogeneity of individual tumours and interactive microenvironment, the outcome of immunotherapy is largely unpredictable. For example, the lack of avid immune response against cancer can be caused by an insufficient number of activated CTLs, the presence of immunosuppressive regulatory Tcells or deficiencies in the production and secretion of cytokines, which are all critical for T-cell activation and recruitment [37, 45, 46, 57]. The tug-of-war between the tumorsuppressive immune system and the immune-evading cancer cells, described as ‘immunoediting’, can often result in selection and growth of tumours with weakly immunogenic antigens [3, 8, 58]. Tumours have evolved other mechanisms to evade the immune system, including the secretion of immunosuppressive cytokines such as  3  transforming growth factor-! (TGF!) or interleukin-10 (IL-10) [37, 59], and stromamodulating proteins that transform the surrounding microenvironment to encourage tumour growth such as TGF!, platelet-derived growth factor (PDGF) and basic fibroblast growth facgtor (bFGF) [60, 61]. Hence, the molecular characteristics of individual tumours can largely dictate the overall outcome of the adoptive T-cell therapy. In the current era of genomics, the molecular characteristics of cells can be globally assessed using various high-throughput technologies such as microarrays and massively parallel sequencing of whole genome, transcriptome and small non-coding RNAs. In particular, transcriptome profiling by sequencing or microarrays allows global assessment of the absolute or relative abundance of mRNA transcripts with an assumption that the mRNA transcript abundance directly correlates with the gene expression level. Hence, ‘gene expression’ will be used interchangeably with the ‘mRNA transcript abundance’ throughout this study. Global transcriptome profiling by microarrays has long been used to identify molecular characteristics of tumours including prognostic gene expression signatures [62-64], mechanistic signatures of pathways unique to tumour progression [65-67], predictive signatures for the outcome of various anti-cancer treatments [68-71], and molecular classifier signatures that distinguish molecular subtypes of human breast cancer (i.e. luminal-A/B, HER2-over-expressing and basal-like subtypes) [72-74]. Notably, the molecular subtypes of human breast cancer enabled further distinctions between histological subtypes of breast cancer. For example, the medullary breast cancer was found to be an estrogen receptor (ER)-negative basal-like subtype [75]. Similarly, the majority of inflammatory breast cancers with very poor prognosis belong to the ER- and  4  progesterone receptor (PR)-negative subclass, which includes the basal-like and HER2 subtypes [76]. The direct attribution of the genetic features of tumours to disease outcome was demonstrated further by the application of MammaPrint, which was recently approved by the FDA to be used as the first clinical prognostic microarray for breast cancer prognosis based on the expression levels of 70 genes [77]. In addition to transcriptome analyses, recent studies have presented increasing evidence of the profound physiological effects of post-transcriptional gene regulation by microRNAs (miRNAs). MiRNAs were shown to affect a wide range of cellular processes including development [78-83], cellular differentiation [84-86], metabolic pathways [87] and disease [84, 88-91]. The immune system is also under miRNA control. Several miRNAs, such as miR-150, miR-181a, miR-17-92, have been found to influence T- and B-cell differentiation [92]. Others include the activation of innate and adaptive immune response (miR-155) [93, 94]; T- and B-cell sensitivity (miR-155, miR181a) [92, 95] and the acute inflammatory response after the recognition of pathogens (miR-16, miR-146, miR-155, miR-223) [92, 93, 96]. Interestingly, recent studies discovered that miRNAs could be secreted to the extracellular environment and cause physiological changes in neighbouring tumour cells [97, 98]. These findings support the possibility of a novel role of miRNA as an intercellular communicator through heterotypic signaling. Furthermore, miRNA signatures have been shown to clearly differentiate between normal and cancerous breast tissues [99], and between different breast cancer subtypes in humans [100]. Prognostic miRNA signatures have also been discovered in several other cancers [101-105].  5  The advent of massively parallel sequencing (also known as ‘next-generation sequencing’) technology such as 454 Life Sciences, Illumina/Solexa and SOLiD, enabled the molecular characterization of tumours at the resolution of single nucleotides. These ultra high-throughput sequencing experiments allow the cost-effective discovery of novel mutations and the unbiased assessment of the absolute abundance of mRNA transcripts and small RNA molecules. Using these technologies, many studies recently identified cancer-specific mutations in various types of cancer including acute myeloid leukaemia (AML) [106], follicular and diffuse large B-cell lymphomas [107], breast cancer[108], ovarian cancer, [109], and lung cancer [110]. In this study, the molecular characteristics of tumours from a unique transgenic mouse model were investigated to better understand the factors derived from tumours that affect the immune response in the context of adoptive T-cell therapy. To investigate the complex interactions between tumours and adoptively transferred T-cells in vivo, the mouse model was developed by Dr. Brad Nelson and his colleagues [111-113]. These mice are transgenic for the Her2/neu oncogene where the immunogenic epitopes derived from chicken ovalbumin were fused at the C-terminus. Transgenic Her2/neu was overexpressed under the control of the MMTV promoter. Also, the tumour suppressing Tp53 gene was under the control of WAP promoter, and had a dominant negative mutation, giving rise to spontaneously arising mammary tumours that were recognized by transgenic Her2/neu epitope-specific CD4+ T-cells and CD8+ CTLs with transgenic Tcell receptors, which were transferred from syngeneic donor mice. Remarkably, tumours from this mouse model showed a complete spectrum of response to the adoptively transferred T-cells, exhibiting complete regression (CR), partial regression (PR), stable  6  disease (SE) or progressive disease (PD) outcome. Cell lines derived from these tumours were subcutaneously injected into the mammary fat pads of mice, and subsequently grown tumours repeatedly showed the expected outcome after adoptive T-cell therapy, providing a unique experimental system to investigate tumour immunogenicity [111113]. In this study, tumours derived from cell lines of CR, PR and PD response [111-113] were investigated by global assessments of gene expression, miRNA abundance and somatic non-synonymous mutations. Using various bioinformatics methodologies, these profiles were used to characterize the molecular differences between tumours with different outcomes and identify tumour-specific factors that promote or inhibit successful T-cell recruitment and infiltration into the tumour epithelium. The findings from this research provide potential clinical targets to enhance the efficacy of adoptive T-cell therapy, and also generate valuable resources to further test the current hypothesis that tumour-specific factors affect the overall immune response and that the molecular characteristics of tumours can predict the outcome of immunotherapy.  7  1.1 BIBLIOGRAPHY 1. Coley WB: The treatment of malignant tumors by repeated inoculations of erysipelas: with a report of ten original cases. Am J Med Sci 1893, 105:487-511. 2. Janeway C: Immunobiology : the immune system in health and disease: 6th ed. New York: Garland Science; 2005. 3. Dunn GP, Bruce AT, Ikeda H, Old LJ, Schreiber RD: Cancer immunoediting: from immunosurveillance to tumor escape. Nat Immunol 2002, 3(11):991-998. 4. Burnet FM: Immunological surveillance in neoplasia. Transplant Rev 1971, 7:3-25. 5. Burnet FM: The concept of immunological surveillance. Prog Exp Tumor Res 1970, 13:1-27. 6. BURNET M: Immunological Factors in the Process of Carcinogenesis. Br Med Bull 1964, 20:154-158. 7. BURNET M: Cancer; a biological approach. I. The processes of control. Br Med J 1957, 1(5022):779-786. 8. Dunn GP, Old LJ, Schreiber RD: The three Es of cancer immunoediting. Annu Rev Immunol 2004, 22:329-360. 9. Kaplan DH, Shankaran V, Dighe AS, Stockert E, Aguet M, Old LJ, Schreiber RD: Demonstration of an interferon gamma-dependent tumor surveillance system in immunocompetent mice. Proc Natl Acad Sci U S A 1998, 95(13):7556-7561. 10. Shankaran V, Ikeda H, Bruce AT, White JM, Swanson PE, Old LJ, Schreiber RD: IFNgamma and lymphocytes prevent primary tumour development and shape tumour immunogenicity. Nature 2001, 410(6832):1107-1111. 11. Smyth MJ, Thia KY, Street SE, MacGregor D, Godfrey DI, Trapani JA: Perforinmediated cytotoxicity is critical for surveillance of spontaneous lymphoma. J Exp Med 2000, 192(5):755-760. 12. van den Broek ME, Kagi D, Ossendorp F, Toes R, Vamvakas S, Lutz WK, Melief CJ, Zinkernagel RM, Hengartner H: Decreased tumor surveillance in perforin-deficient mice. J Exp Med 1996, 184(5):1781-1790. 13. Gatti RA, Good RA: Occurrence of malignancy in immunodeficiency diseases. A literature review. Cancer 1971, 28(1):89-98. 14. McClain KL: Immunodeficiency states and related malignancies. Cancer Treat Res 1997, 92:39-61. 15. Birkeland SA, Storm HH, Lamm LU, Barlow L, Blohme I, Forsberg B, Eklund B, Fjeldborg O, Friedberg M, Frodin L: Cancer risk after renal transplantation in the Nordic countries, 1964-1986. Int J Cancer 1995, 60(2):183-189. 16. Dighe AS, Richards E, Old LJ, Schreiber RD: Enhanced in vivo growth and resistance to rejection of tumor cells expressing dominant negative IFN gamma receptors. Immunity 1994, 1(6):447-456.  8  17. Morris M, Platell C, Iacopetta B: Tumor-infiltrating lymphocytes and perforation in colon cancer predict positive response to 5-fluorouracil chemotherapy. Clin Cancer Res 2008, 14(5):1413-1417. 18. Pages F, Berger A, Camus M, Sanchez-Cabo F, Costes A, Molidor R, Mlecnik B, Kirilovsky A, Nilsson M, Damotte D, Meatchi T, Bruneval P, Cugnenc PH, Trajanoski Z, Fridman WH, Galon J: Effector memory T cells, early metastasis, and survival in colorectal cancer. N Engl J Med 2005, 353(25):2654-2666. 19. Dalerba P, Maccalli C, Casati C, Castelli C, Parmiani G: Immunology and immunotherapy of colorectal cancer. Crit Rev Oncol Hematol 2003, 46(1):33-57. 20. Galon J, Costes A, Sanchez-Cabo F, Kirilovsky A, Mlecnik B, Lagorce-Pages C, Tosolini M, Camus M, Berger A, Wind P, Zinzindohoue F, Bruneval P, Cugnenc PH, Trajanoski Z, Fridman WH, Pages F: Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science 2006, 313(5795):1960-1964. 21. Tartour E, Gey A, Sastre-Garau X, Lombard Surin I, Mosseri V, Fridman WH: Prognostic value of intratumoral interferon gamma messenger RNA expression in invasive cervical carcinomas. J Natl Cancer Inst 1998, 90(4):287-294. 22. Leffers N, Gooden MJ, de Jong RA, Hoogeboom BN, ten Hoor KA, Hollema H, Boezen HM, van der Zee AG, Daemen T, Nijman HW: Prognostic significance of tumor-infiltrating T-lymphocytes in primary and metastatic lesions of advanced stage ovarian cancer. Cancer Immunol Immunother 2009, 58(3):449-459. 23. Zhang L, Conejo-Garcia JR, Katsaros D, Gimotty PA, Massobrio M, Regnani G, Makrigiannakis A, Gray H, Schlienger K, Liebman MN, Rubin SC, Coukos G: Intratumoral T cells, recurrence, and survival in epithelial ovarian cancer. N Engl J Med 2003, 348(3):203-213. 24. Clemente CG, Mihm MC,Jr, Bufalino R, Zurrida S, Collini P, Cascinelli N: Prognostic value of tumor infiltrating lymphocytes in the vertical growth phase of primary cutaneous melanoma. Cancer 1996, 77(7):1303-1310. 25. Inoue T, Yoneda K, Manabe M, Demitsu T: Spontaneous regression of merkel cell carcinoma: a comparative study of TUNEL index and tumor-infiltrating lymphocytes between spontaneous regression and non-regression group. J Dermatol Sci 2000, 24(3):203-211. 26. Halliday GM, Patel A, Hunt MJ, Tefany FJ, Barnetson RS: Spontaneous regression of human melanoma/nonmelanoma skin cancer: association with infiltrating CD4+ T cells. World J Surg 1995, 19(3):352-358. 27. Dieu-Nosjean MC, Antoine M, Danel C, Heudes D, Wislez M, Poulot V, Rabbe N, Laurans L, Tartour E, de Chaisemartin L, Lebecque S, Fridman WH, Cadranel J: Longterm survival for patients with non-small-cell lung cancer with intratumoral lymphoid structures. J Clin Oncol 2008, 26(27):4410-4417. 28. Kerr KM, Johnson SK, King G, Kennedy MM, Weir J, Jeffrey R: Partial regression in primary carcinoma of the lung: does it occur? Histopathology 1998, 33(1):55-63.  9  29. Lee AH, Gillett CE, Ryder K, Fentiman IS, Miles DW, Millis RR: Different patterns of inflammation and prognosis in invasive carcinoma of the breast. Histopathology 2006, 48(6):692-701. 30. Marques LA, Franco EL, Torloni H, Brentani MM, da Silva-Neto JB, Brentani RR: Independent prognostic value of laminin receptor expression in breast cancer survival. Cancer Res 1990, 50(5):1479-1483. 31. Nixon AJ, Neuberg D, Hayes DF, Gelman R, Connolly JL, Schnitt S, Abner A, Recht A, Vicini F, Harris JR: Relationship of patient age to pathologic features of the tumor and prognosis for patients with stage I or II breast cancer. J Clin Oncol 1994, 12(5):888-894. 32. Rilke F, Colnaghi MI, Cascinelli N, Andreola S, Baldini MT, Bufalino R, Della Porta G, Menard S, Pierotti MA, Testori A: Prognostic significance of HER-2/neu expression in breast cancer and its relationship to other prognostic factors. Int J Cancer 1991, 49(1):44-49. 33. Rooney CM, Smith CA, Ng CY, Loftin SK, Sixbey JW, Gan Y, Srivastava DK, Bowman LC, Krance RA, Brenner MK, Heslop HE: Infusion of cytotoxic T cells for the prevention and treatment of Epstein-Barr virus-induced lymphoma in allogeneic transplant recipients. Blood 1998, 92(5):1549-1555. 34. Gattinoni L, Powell DJ,Jr, Rosenberg SA, Restifo NP: Adoptive immunotherapy for cancer: building on success. Nat Rev Immunol 2006, 6(5):383-393. 35. Dudley ME, Rosenberg SA: Adoptive-cell-transfer therapy for the treatment of patients with cancer. Nat Rev Cancer 2003, 3(9):666-675. 36. Greten TF, Jaffee EM: Cancer vaccines. Curr Protoc Hum Genet 2001, Chapter 13:Unit 13.8. 37. Rosenberg SA, Yang JC, Restifo NP: Cancer immunotherapy: moving beyond current vaccines. Nat Med 2004, 10(9):909-915. 38. Rosenberg SA, Restifo NP, Yang JC, Morgan RA, Dudley ME: Adoptive cell transfer: a clinical path to effective cancer immunotherapy . Nat Rev Cancer 2008, 8(4):299-308. 39. Pardoll DM: Spinning molecular immunology into successful immunotherapy. Nat Rev Immunol 2002, 2(4):227-238. 40. Jager D, Stockert E, Gure AO, Scanlan MJ, Karbach J, Jager E, Knuth A, Old LJ, Chen YT: Identification of a tissue-specific putative transcription factor in breast tissue by serological screening of a breast cancer library. Cancer Res 2001, 61(5):2055-2061. 41. Runnebaum IB, Nagarajan M, Bowman M, Soto D, Sukumar S: Mutations in p53 as potential molecular markers for human breast cancer. Proc Natl Acad Sci U S A 1991, 88(23):10657-10661. 42. Ribas A, Butterfield LH, Glaspy JA, Economou JS: Current developments in cancer vaccines and cellular immunotherapy. J Clin Oncol 2003, 21(12):2415-2432.  10  43. Dudley ME, Wunderlich JR, Robbins PF, Yang JC, Hwu P, Schwartzentruber DJ, Topalian SL, Sherry R, Restifo NP, Hubicki AM, Robinson MR, Raffeld M, Duray P, Seipp CA, Rogers-Freezer L, Morton KE, Mavroukakis SA, White DE, Rosenberg SA: Cancer regression and autoimmunity in patients after clonal repopulation with antitumor lymphocytes. Science 2002, 298(5594):850-854. 44. Dudley ME, Wunderlich JR, Yang JC, Sherry RM, Topalian SL, Restifo NP, Royal RE, Kammula U, White DE, Mavroukakis SA, Rogers LJ, Gracia GJ, Jones SA, Mangiameli DP, Pelletier MM, Gea-Banacloche J, Robinson MR, Berman DM, Filie AC, Abati A, Rosenberg SA: Adoptive cell transfer therapy following non-myeloablative but lymphodepleting chemotherapy for the treatment of patients with refractory metastatic melanoma. J Clin Oncol 2005, 23(10):2346-2357. 45. Begley J, Ribas A: Targeted therapies to improve tumor immunotherapy. Clin Cancer Res 2008, 14(14):4385-4391. 46. Boon T, Cerottini JC, den Eynde B, V., van der Bruggen P, Pel A, V.: Tumor antigens recognized by T lymphocytes. Ann Rev Immun 1994, 12:337-365. 47. Marincola FM, Jaffee EM, Hicklin DJ, Ferrone S: Escape of human solid tumors from T-cell recognition: molecular mechanisms and functional significance. Adv Immunol 2000, 74:181-273. 48. Scanlan MJ, Jager D: Challenges to the development of antigen-specific breast cancer vaccines. Breast Cancer Res 2001, 3(2):95-98. 49. Wolfel T, Hauer M, Schneider J, Serrano M, Wolfel C, Klehmann-Hieb E, De Plaen E, Hankeln T, Meyer zum Buschenfelde KH, Beach D: A p16INK4a-insensitive CDK4 mutant targeted by cytolytic T lymphocytes in a human melanoma. Science 1995, 269(5228):1281-1284. 50. Scanlan MJ, Gure AO, Jungbluth AA, Old LJ, Chen YT: Cancer/testis antigens: an expanding family of targets for cancer immunotherapy. Immunol Rev 2002, 188:2232. 51. Cancer Immunome Database [http://ludwig-sun5.unil.ch/CancerImmunomeDB/] 52. Ioannides CG, Ioannides MG, O'Brian CA: T-cell recognition of oncogene products: a new strategy for immunotherapy. Mol Carcinog 1992, 6(2):77-82. 53. Ioannides C, G., Fisk B, Fan D, Biddison W, E., Wharton JT, O'brian C, A.: Cytotoxic T cells isolated from ovarian malignant ascites recognize a peptide derived from the HER-2/neu proto-oncogene. Cell Immun 1993, 151(1):225-234. 54. Disis ML, Calenoff E, McLaughlin G, Murphy AE, Chen W, Groner B, Jeschke M, Lydon N, McGlynn E, Livingston RB: Existent T-cell and antibody immunity to HER-2/neu protein in patients with breast cancer. Cancer Res 1994, 54(1):16-20. 55. Peoples GE, Smith RC, Linehan DC, Yoshino I, Goedegebuure PS, Eberlein TJ: Shared T cell epitopes in epithelial tumors. Cell Immunol 1995, 164(2):279-286. 56. Fisk B, Blevins TL, Wharton JT, Ioannides CG: Identification of an immunodominant peptide of HER-2/neu protooncogene recognized by ovarian tumor-specific cytotoxic T lymphocyte lines. J Exp Med 1995, 181(6):2109-2117. 11  57. Schroder K, Hertzog PJ, Ravasi T, Hume DA: Interferon-gamma: an overview of signals, mechanisms and functions. J Leukoc Biol 2004, 75(2):163-189. 58. Dunn GP, Koebel CM, Schreiber RD: Interferons, immunity and cancer immunoediting. Nat Rev Immunol 2006, 6(11):836-848. 59. Parish CR: Cancer immunotherapy: the past, the present and the future. Immunol Cell Biol 2003, 81(2):106-113. 60. Mueller MM, Fusenig NE: Friends or foes - bipolar effects of the tumour stroma in cancer. Nat Rev Cancer 2004, 4(11):839-849. 61. Liotta LA, Kohn EC: The microenvironment of the tumour-host interface. Nature 2001, 411(6835):375-379. 62. Teschendorff AE, Miremadi A, Pinder SE, Ellis IO, Caldas C: An immune response gene expression module identifies a good prognosis subtype in estrogen receptor negative breast cancer. Genome Biol 2007, 8(8):R157. 63. Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, Yang F, Talantov D, Timmermans M, Meijer-van Gelder ME, Yu J, Jatkoe T, Berns EM, Atkins D, Foekens JA: Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005, 365(9460):671-679. 64. Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H, Chen H, Omeroglu G, Meterissian S, Omeroglu A, Hallett M, Park M: Stromal gene expression predicts clinical outcome in breast cancer. Nat Med 2008, 14(5):518-527. 65. Ponzo MG, Lesurf R, Petkiewicz S, O'Malley FP, Pinnaduwage D, Andrulis IL, Bull SB, Chughtai N, Zuo D, Souleimanova M, Germain D, Omeroglu A, Cardiff RD, Hallett M, Park M: Met induces mammary tumors with diverse histologies and is associated with poor outcome and human basal breast cancer. Proc Natl Acad Sci U S A 2009, 106(31):12903-12908. 66. Huang S, Chen Y, Podsypanina K, Li Y: Comparison of expression profiles of metastatic versus primary mammary tumors in MMTV-Wnt-1 and MMTV-Neu transgenic mice. Neoplasia 2008, 10(2):118-124. 67. Worschech A, Kmieciak M, Knutson KL, Bear HD, Szalay AA, Wang E, Marincola FM, Manjili MH: Signatures associated with rejection or recurrence in HER-2/neupositive mammary tumors. Cancer Res 2008, 68(7):2436-2446. 68. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH: Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002, 415(6871):530-536. 69. Rouzier R, Perou CM, Symmans WF, Ibrahim N, Cristofanilli M, Anderson K, Hess KR, Stec J, Ayers M, Wagner P, Morandi P, Fan C, Rabiul I, Ross JS, Hortobagyi GN, Pusztai L: Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res 2005, 11(16):5678-5685.  12  70. Schauer M, Janssen KP, Rimkus C, Raggi M, Feith M, Friess H, Theisen J: Microarray-based response prediction in esophageal adenocarcinoma. Clin Cancer Res 2010, 16(1):330-337. 71. Farmer P, Bonnefoi H, Anderle P, Cameron D, Wirapati P, Becette V, Andre S, Piccart M, Campone M, Brain E, Macgrogan G, Petit T, Jassem J, Bibeau F, Blot E, Bogaerts J, Aguet M, Bergh J, Iggo R, Delorenzi M: A stroma-related gene signature predicts resistance to neoadjuvant chemotherapy in breast cancer. Nat Med 2009, 15(1):68-74. 72. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO, Botstein D: Molecular portraits of human breast tumours. Nature 2000, 406(6797):747-752. 73. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, Eisen MB, van de Rijn M, Jeffrey SS, Thorsen T, Quist H, Matese JC, Brown PO, Botstein D, Eystein Lonning P, Borresen-Dale AL: Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 2001, 98(19):10869-10874. 74. Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, Deng S, Johnsen H, Pesich R, Geisler S, Demeter J, Perou CM, Lonning PE, Brown PO, Borresen-Dale AL, Botstein D: Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A 2003, 100(14):8418-8423. 75. Bertucci F, Finetti P, Cervera N, Charafe-Jauffret E, Mamessier E, Adelaide J, Debono S, Houvenaeghel G, Maraninchi D, Viens P, Charpin C, Jacquemier J, Birnbaum D: Gene expression profiling shows medullary breast cancer is a subgroup of basal breast cancers. Cancer Res 2006, 66(9):4636-4644. 76. Nguyen DM, Sam K, Tsimelzon A, Li X, Wong H, Mohsin S, Clark GM, Hilsenbeck SG, Elledge RM, Allred DC, O'Connell P, Chang JC: Molecular heterogeneity of inflammatory breast cancer: a hyperproliferative phenotype. Clin Cancer Res 2006, 12(17):5047-5054. 77. Glas AM, Floore A, Delahaye LJ, Witteveen AT, Pover RC, Bakx N, Lahti-Domenici JS, Bruinsma TJ, Warmoes MO, Bernards R, Wessels LF, Van't Veer LJ: Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006, 7:278. 78. Brennecke J, Hipfner DR, Stark A, Russell RB, Cohen SM: bantam encodes a developmentally regulated microRNA that controls cell proliferation and regulates the proapoptotic gene hid in Drosophila. Cell 2003, 113(1):25-36. 79. Johnston RJ, Hobert O: A microRNA controlling left/right neuronal asymmetry in Caenorhabditis elegans. Nature 2003, 426(6968):845-849. 80. Moss EG, Lee RC, Ambros V: The cold shock domain protein LIN-28 controls developmental timing in C. elegans and is regulated by the lin-4 RNA. Cell 1997, 88(5):637-646.  13  81. Reinhart BJ, Slack FJ, Basson M, Pasquinelli AE, Bettinger JC, Rougvie AE, Horvitz HR, Ruvkun G: The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature 2000, 403(6772):901-906. 82. Lee RC, Feinbaum RL, Ambros V: The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 1993, 75(5):843854. 83. Du T, Zamore PD: microPrimer: the biogenesis and function of microRNA. Development 2005, 132(21):4645-4652. 84. Esquela-Kerscher A, Slack FJ: Oncomirs - microRNAs with a role in cancer. Nat Rev Cancer 2006, 6(4):259-269. 85. Schickel R, Boyerinas B, Park SM, Peter ME: MicroRNAs: key players in the immune system, differentiation, tumorigenesis and cell death. Oncogene 2008, 27(45):5959-5974. 86. Chen CZ, Li L, Lodish HF, Bartel DP: MicroRNAs modulate hematopoietic lineage differentiation. Science 2004, 303(5654):83-86. 87. Wilfred BR, Wang WX, Nelson PT: Energizing miRNA research: a review of the role of miRNAs in lipid metabolism, with a prediction that miR-103/107 regulates human metabolic pathways. Mol Genet Metab 2007, 91(3):209-217. 88. Blenkiron C, Miska EA: miRNAs in cancer: approaches, aetiology, diagnostics and therapy. Hum Mol Genet 2007, 16 Spec No 1:R106-13. 89. Espinosa CES, Slack FJ: The role of microRNAs in cancer. Yale J Biol Med 2006, 79:131-140. 90. Ikeda S, Kong SW, Lu J, Bisping E, Zhang H, Allen PD, Golub TR, Pieske B, Pu WT: Altered microRNA expression in human heart disease. Physiol Genomics 2007, 31(3):367-373. 91. Wang WX, Rajeev BW, Stromberg AJ, Ren N, Tang G, Huang Q, Rigoutsos I, Nelson PT: The expression of microRNA miR-107 decreases early in Alzheimer's disease and may accelerate disease progression through regulation of beta-site amyloid precursor protein-cleaving enzyme 1. J Neurosci 2008, 28(5):1213-1223. 92. Lindsay MA: microRNAs and the immune response. Trends Immunol 2008, 29(7):343-351. 93. Baltimore D, Boldin MP, O'Connell RM, Rao DS, Taganov KD: MicroRNAs: new regulators of immune cell development and function. Nat Immunol 2008, 9(8):839845. 94. Lodish HF, Zhou B, Liu G, Chen CZ: Micromanagement of the immune system by microRNAs. Nat Rev Immunol 2008, 8(2):120-130. 95. Li QJ, Chau J, Ebert PJ, Sylvester G, Min H, Liu G, Braich R, Manoharan M, Soutschek J, Skare P, Klein LO, Davis MM, Chen CZ: miR-181a is an intrinsic modulator of T cell sensitivity and selection. Cell 2007, 129(1):147-161.  14  96. Jing Q, Huang S, Guth S, Zarubin T, Motoyama A, Chen J, Di Padova F, Lin SC, Gram H, Han J: Involvement of microRNA in AU-rich element-mediated mRNA instability. Cell 2005, 120(5):623-634. 97. Yu Z, Willmarth NE, Zhou J, Katiyar S, Wang M, Liu Y, McCue PA, Quong AA, Lisanti MP, Pestell RG: microRNA 17/20 inhibits cellular invasion and tumor metastasis in breast cancer by heterotypic signaling. Proc Natl Acad Sci U S A 2010, 107(18):8231-8236. 98. Valadi H, Ekstrom K, Bossios A, Sjostrand M, Lee JJ, Lotvall JO: Exosomemediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat Cell Biol 2007, 9(6):654-659. 99. Iorio MV, Ferracin M, Liu CG, Veronese A, Spizzo R, Sabbioni S, Magri E, Pedriali M, Fabbri M, Campiglio M, Menard S, Palazzo JP, Rosenberg A, Musiani P, Volinia S, Nenci I, Calin GA, Querzoli P, Negrini M, Croce CM: MicroRNA gene expression deregulation in human breast cancer. Cancer Res 2005, 65(16):7065-7070. 100. Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning MJ, BarbosaMorais NL, Teschendorff AE, Green AR, Ellis IO, Tavare S, Caldas C, Miska EA: MicroRNA expression profiling of human breast cancer identifies new markers of tumor subtype. Genome Biol 2007, 8(10):R214. 101. Calin GA, Ferracin M, Cimmino A, Di Leva G, Shimizu M, Wojcik SE, Iorio MV, Visone R, Sever NI, Fabbri M, Iuliano R, Palumbo T, Pichiorri F, Roldo C, Garzon R, Sevignani C, Rassenti L, Alder H, Volinia S, Liu CG, Kipps TJ, Negrini M, Croce CM: A MicroRNA signature associated with prognosis and progression in chronic lymphocytic leukemia. N Engl J Med 2005, 353(17):1793-1801. 102. Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M, Stephens RM, Okamoto A, Yokota J, Tanaka T, Calin GA, Liu CG, Croce CM, Harris CC: Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell 2006, 9(3):189-198. 103. Takamizawa J, Konishi H, Yanagisawa K, Tomida S, Osada H, Endoh H, Harano T, Yatabe Y, Nagino M, Nimura Y, Mitsudomi T, Takahashi T: Reduced expression of the let-7 microRNAs in human lung cancers in association with shortened postoperative survival. Cancer Res 2004, 64(11):3753-3756. 104. Garzon R, Fabbri M, Cimmino A, Calin GA, Croce CM: MicroRNA expression and function in cancer. Trends Mol Med 2006, 12(12):580-587. 105. Schetter AJ, Leung SY, Sohn JJ, Zanetti KA, Bowman ED, Yanaihara N, Yuen ST, Chan TL, Kwong DL, Au GK, Liu CG, Calin GA, Croce CM, Harris CC: MicroRNA expression profiles associated with prognosis and therapeutic outcome in colon adenocarcinoma. JAMA 2008, 299(4):425-436. 106. Ley TJ, Mardis ER, Ding L, Fulton B, McLellan MD, Chen K, Dooling D, DunfordShore BH, McGrath S, Hickenbotham M, Cook L, Abbott R, Larson DE, Koboldt DC, Pohl C, Smith S, Hawkins A, Abbott S, Locke D, Hillier LW, Miner T, Fulton L, Magrini V, Wylie T, Glasscock J, Conyers J, Sander N, Shi X, Osborne JR, Minx P, Gordon D, Chinwalla A, Zhao Y, Ries RE, Payton JE, Westervelt P, Tomasson MH, Watson M,  15  Baty J, Ivanovich J, Heath S, Shannon WD, Nagarajan R, Walter MJ, Link DC, Graubert TA, DiPersio JF, Wilson RK: DNA sequencing of a cytogenetically normal acute myeloid leukaemia genome. Nature 2008, 456(7218):66-72. 107. Morin RD, Johnson NA, Severson TM, Mungall AJ, An J, Goya R, Paul JE, Boyle M, Woolcock BW, Kuchenbauer F, Yap D, Humphries RK, Griffith OL, Shah S, Zhu H, Kimbara M, Shashkin P, Charlot JF, Tcherpakov M, Corbett R, Tam A, Varhol R, Smailus D, Moksa M, Zhao Y, Delaney A, Qian H, Birol I, Schein J, Moore R, Holt R, Horsman DE, Connors JM, Jones S, Aparicio S, Hirst M, Gascoyne RD, Marra MA: Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat Genet 2010, 42(2):181-185. 108. Shah SP, Morin RD, Khattra J, Prentice L, Pugh T, Burleigh A, Delaney A, Gelmon K, Guliany R, Senz J, Steidl C, Holt RA, Jones S, Sun M, Leung G, Moore R, Severson T, Taylor GA, Teschendorff AE, Tse K, Turashvili G, Varhol R, Warren RL, Watson P, Zhao Y, Caldas C, Huntsman D, Hirst M, Marra MA, Aparicio S: Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature 2009, 461(7265):809-813. 109. Shah SP, Kobel M, Senz J, Morin RD, Clarke BA, Wiegand KC, Leung G, Zayed A, Mehl E, Kalloger SE, Sun M, Giuliany R, Yorida E, Jones S, Varhol R, Swenerton KD, Miller D, Clement PB, Crane C, Madore J, Provencher D, Leung P, DeFazio A, Khattra J, Turashvili G, Zhao Y, Zeng T, Glover JN, Vanderhyden B, Zhao C, Parkinson CA, Jimenez-Linan M, Bowtell DD, Mes-Masson AM, Brenton JD, Aparicio SA, Boyd N, Hirst M, Gilks CB, Marra M, Huntsman DG: Mutation of FOXL2 in granulosa-cell tumors of the ovary. N Engl J Med 2009, 360(26):2719-2729. 110. Ding L, Getz G, Wheeler DA, Mardis ER, McLellan MD, Cibulskis K, Sougnez C, Greulich H, Muzny DM, Morgan MB, Fulton L, Fulton RS, Zhang Q, Wendl MC, Lawrence MS, Larson DE, Chen K, Dooling DJ, Sabo A, Hawes AC, Shen H, Jhangiani SN, Lewis LR, Hall O, Zhu Y, Mathew T, Ren Y, Yao J, Scherer SE, Clerc K, Metcalf GA, Ng B, Milosavljevic A, Gonzalez-Garay ML, Osborne JR, Meyer R, Shi X, Tang Y, Koboldt DC, Lin L, Abbott R, Miner TL, Pohl C, Fewell G, Haipek C, Schmidt H, Dunford-Shore BH, Kraja A, Crosby SD, Sawyer CS, Vickery T, Sander S, Robinson J, Winckler W, Baldwin J, Chirieac LR, Dutt A, Fennell T, Hanna M, Johnson BE, Onofrio RC, Thomas RK, Tonon G, Weir BA, Zhao X, Ziaugra L, Zody MC, Giordano T, Orringer MB, Roth JA, Spitz MR, Wistuba II, Ozenberger B, Good PJ, Chang AC, Beer DG, Watson MA, Ladanyi M, Broderick S, Yoshizawa A, Travis WD, Pao W, Province MA, Weinstock GM, Varmus HE, Gabriel SB, Lander ES, Gibbs RA, Meyerson M, Wilson RK: Somatic mutations affect key pathways in lung adenocarcinoma. Nature 2008, 455(7216):1069-1075. 111. Wall EM, Milne K, Martin ML, Watson PH, Theiss P, Nelson BH: Spontaneous mammary tumors differ widely in their inherent sensitivity to adoptively transferred T cells. Cancer Res 2007, 67(13):6442-6450. 112. Yang T, Martin ML, Nielsen JS, Milne K, Wall EM, Lin W, Watson PH, Nelson BH: Mammary tumors with diverse immunological phenotypes show differing sensitivity to adoptively transferred CD8+ T cells lacking the Cbl-b gene. Cancer Immunol Immunother 2009, 58(11):1865-1875. 16  113. Martin ML, Wall EM, Sandwith E, Girardin A, Milne K, Watson PH, Nelson BH: Density of tumour stroma is correlated to outcome after adoptive transfer of CD4(+) and CD8 (+) T cells in a murine mammary carcinoma model. Breast Cancer Res Treat 2009, 58(11):1865-1875. 114. Hueman MT, Dehqanzada ZA, Novak TE, Gurney JM, Woll MM, Ryan GB, Storrer CE, Fisher C, McLeod DG, Ioannides CG, Ponniah S, Peoples GE: Phase I clinical trial of a HER-2/neu peptide (E75) vaccine for the prevention of prostate-specific antigen recurrence in high-risk prostate cancer patients. Clin Cancer Res 2005, 11(20):74707479. 115. Knutson KL, Schiffman K, Disis ML: Immunization with a HER-2/neu helper peptide vaccine generates HER-2/neu CD8 T-cell immunity in cancer patients. J Clin Invest 2001, 107(4):477-484. 116. Peoples GE, Smith RC, Linehan DC, Yoshino I, Goedegebuure PS, Eberlein TJ: Shared T cell epitopes in epithelial tumors. Cell Immunol 1995, 164(2):279-286. 117. Peoples GE, Holmes JP, Hueman MT, Mittendorf EA, Amin A, Khoo S, Dehqanzada ZA, Gurney JM, Woll MM, Ryan GB, Storrer CE, Craig D, Ioannides CG, Ponniah S: Combined clinical trial results of a HER2/neu (E75) vaccine for the prevention of recurrence in high-risk breast cancer patients: U.S. Military Cancer Institute Clinical Trials Group Study I-01 and I-02. Clin Cancer Res 2008, 14(3):797803.  17  2 GENE EXPRESSION PROFILES ASSOCIATED WITH THE OUTCOME OF ADOPTIVE T-CELL THERAPY IN A MOUSE MODEL OF BREAST CANCER 1  2.1 INTRODUCTION The role of the immune system in cancer prevention and progression has tantalized researchers since the first observation of tumour regression following an acute bacterial infection made by William Coley more than a century ago [1]. Since then, compelling evidence has accumulated supporting a role for the immune system in promoting tumour suppression and regression. This includes an increased frequency of tumourigenesis in immunodeficient or immunocompromised mice and humans [2-9], rejection of transplanted tumours in immunocompetent hosts [3, 10] and strong correlations between patient survival and the presence of tumour-infiltrating CD4+ helper and CD8+ cytotoxic T lymphocytes (CTL) in cancers of the colon [11-14], cervix [15], ovary [16, 17], skin [18-20], lung [21, 22], breast [23-26], and lymphoid system [27]. A successful immune response to cancer requires the effective uptake and presentation of tumour antigens by competent antigen presentation cells (APCs) such as dendritic cells and macrophages, followed by antigen-specific recognition and activation of CD4+ helper T-cells and CD8+ CTLs within the lymph node. Activated CTLs must then migrate to the tumour site travelling through the tumour stroma, infiltrate tumour epithelium and induce direct killing of cancer cells by releasing cytolytic granules 1  A version of this chapter will be submitted for publication. Gene expression profiles associated with outcome of adoptive T-cell therapy in a mouse model of breast cancer. Chun, H.J.E., Martin, M.L., Hsu, J., Ou, D., Delaney, A., Nelson, B.H., Jones, S.J.M. 18  containing perforin, or through Fas-ligand-mediated apoptosis [5, 28-31]. The antitumour immune response not only results in cancer cell death, but can also establish lasting immunity to cancer via the formation of memory B and T lymphocytes. To accomplish a successful anti-tumour immune response, there are many factors that influence the immune system. For example, there are important receptors on APCs such as pattern recognition receptors, which must recognize the antigenic epitopes before presenting them to the CD8+ CTLs. Critical chemokines and cytokines such as granulocyte-macrophage-colony stimulating factor (GM-CSF), interferon gamma (IFN"), interleukin 2 (IL-2) and interleukin 18 (IL-18) must be synthesized and secreted locally and systemically by T-cells and a variety of immune cells to support the recruitment, activation and trafficking of T-cells into the tumour site [29-39]. Deficiencies in the production or secretion of these or other cytokines at any of these stages may result in an ineffective response to the tumour, allowing tumours to evade immune response [2, 37, 38, 40-42]. The intra-tumour infiltration of T-cells that pass through the immunosuppressive tumour microenvironment is also critical for a successful antitumour immune response [43]. Increasing evidence supports adoptive T-cell therapy as a promising approach of treatment for several cancer types. Adoptive T-cell therapy utilizes an infusion of ex vivo expanded, antigen-specific autologous CTLs to augment and optimize the anti-tumour immune response in a patient. Studies have shown tumour regression following adoptive T cell therapy for metastatic renal cell carcinoma [44] and metastatic melanoma [45, 46] in which the objective response rates observed in patients were 35% and 51% respectively. The clinical application of adoptive T cell therapy for breast cancer has not  19  been realized thus far, despite of immunogenic HER2/neu antigen, which is found to be amplified in 20-40% of breast cancer patients [47-52]. Several factors might contribute to unpredictable treatment outcomes and low success rates of immunotherapy. These include a lack of helper factors for CD8+ CTLs from CD4+ helper T lymphocytes [53] and an insufficient number of CD8+ CTLs at the tumour site due to problems with T lymphocyte trafficking and infiltration [54]. The tumour microenvironment can contain molecules that hinder or block T cell migration. For example, stromal cells with altered expression of adhesion molecules such as ICAM1 and VCAM1, or junction proteins including ZO-1 and Claudin-1 and -4, which affect polarization and cell-to-cell communication, can create an unsupportive microenvironment for T-cells to effectively augment the anti-cancer response [55, 56]. Tumours have also evolved a number of well-described local and systemic strategies to evade the immune response by inhibiting the activation of immune cells and blocking the subsequent intratumoural infiltration by CTLs. Some of the known immunosuppressive mechanisms employed by tumours are the secretion of transforming growth factor-! (TGF!) or interleukin-10 (IL-10), and the recruitment of suppressive cellular constituents such as FOXP3+ regulatory T-cells and myloid derived suppressor cells [31, 57-59]. Indeed, the conditioning of a host by depleting regulatory T cells or other immunosuppressive elements (known as ‘lymphodepletion’) before immunotherapy was shown to result in tumour regression in advanced melanoma [45, 46, 60, 61]. The molecular characteristics of individual tumours can be used to differentiate various histopathology subtypes and molecular classifications in human cancers, providing indispensable information for clinical decisions of anti-cancer therapeutic regimens,  20  disease prognosis and prediction of therapeutic outcomes [62]. Numerous studies of global gene expression patterns revealed gene sets that act as distinct signatures indicative of tumour biology and behavior. These include prognostic signatures [63-65], mechanistic signatures of pathways unique to tumour progression [66-68], predictive signatures for the outcome of various anti-cancer treatments [69-72], and molecular classifier signatures that distinguish four subtypes of human breast cancer (luminal-A/B, HER2-overexpressing, basal-like) [73-75] and act as potential markers [76]. The pathways enriched for a particular gene signature can reflect the disease progression based on tumour biology. The basal-like subtype (also known as ‘triple negative’ because of estrogen receptor (ER) negative, progesterone receptor (PR) negative and HER2 over-expression negative status) has the worst prognosis in all four subtypes in human breast cancer. Several gene expression studies have found that the gene signatures of the basal-like subtype are enriched for several hallmark pathways of tumourigenesis including cell cycle, cell proliferation and differentiation, angiogenesis and signal transduction [69, 77, 78]. The transcriptomic differences in tumours are also evident in different histopathological subtypes. The pronounced differences in disease outcome observed between ER-negative and -positive tumours is reflected by the distinct differences in gene expression profiles between the ER-positive luminal A/B subtypes and the ER-negative HER2-overexpressing and the basal-like subtypes [69, 74, 75, 7779] which are predominantly found in grade I and III tumours respectively [62, 80]. Many factors that either promote or impair the T-cell response at the level of the tumour, coupled with the molecular heterogeneity among individual tumours lead to mixed clinical responses to adoptive T-cell therapy, where some tumours regress while others  21  progress even within the same host [43]. This observation supports the concept that local immune suppression may be largely a function of the tumour and the surrounding microenvironment. Identifying the immune-modulating features unique to each tumour will be crucial in understanding the intricate interplay between tumours and adoptively transferred T-cells. Furthermore, characterizing the molecular signatures unique to regressing tumours will not only enhance the current understanding of immunogenicity of tumour, but also provide an unprecedented opportunity to predict the immune response of tumours in the clinical setting, allowing the use of adoptive T-cell therapy as an adjuvant therapy or an alternative treatment strategies for breast cancer or potentially other solid tumours. To better understand the complexities of tumour interacting with adoptively transferred T-cells in vivo, a unique transgenic mouse model of breast cancer was developed by Nelson and his colleagues [43]. In brief, they created a C57BL/6 mouse line that was transgenic for the Her2/neu oncogene, where CD8+ (OT-I) and CD4+ (OT-II) T-cell epitopes from the model antigen ovalbumin (OVA) were fused at the C-terminus. The over-expression of transgenic Her2/neuOT-I/OT-II under the control of MMTV promoter, combined with the effect of dominant negative mutation in Tp53 tumour suppressor gene resulted in spontaneously arising mammary tumours expressing Her2/neu tagged with the OVA epitopes. These tumours were recognized by OVA-specific T-cell receptor transgenic CD8+ and CD4+ T cells from syngeneic donor mice. Remarkably, spontaneous tumours arising in this model showed a complete spectrum of responses to adoptively transferred CD8+ and CD4+ T cells. Approximately one third of the tumours underwent complete regression (CR) in response to adoptively transferred CD8+ and  22  CD4+ T-cells, whereas the rest showed either partial regression (PR) (comprising 25% of tumours), stable disease (SD) (15% of tumours) or progressive disease (PD) (24% of tumours) [43, 81]. Furthermore, Nelson and colleagues successfully generated tumour cell lines from these tumours that demonstrate reproducible responses of CR, PR, SD or PD to adoptively transferred T-cells after implanting them in the mammary fat pads of host mice [43, 81, 82]. Thus, these tumour cell lines and the mouse model provide a unique experimental system to identify histopathological, cellular and molecular factors that dictate the outcome of adoptive T-cell therapy. In this study, tumour cell lines representing the CR, PR and PD outcomes in the mouse model were subjected to gene expression profiling experiments using Affymetrix microarrays. This revealed several immune-related pathways that were present uniquely in tumours of the CR class. Several genes from these pathways show promise as predictive markers for outcome following adoptive T-cell therapy. Finally, by comparing our gene expression profiles to publicly available human breast cancer datasets, tumours in specific outcome groups were found to resemble distinct molecular subtypes of human breast cancer. The findings from this study support the hypothesis that genetic features associated with the tumour can be used to predict the outcome of adoptive T-cell transfer in a mouse model of breast cancer. Importantly, the bioinformatics approach to select a subset of predictive genes and to compare the expression profile from mouse genes with the human data may allow the possibility of translating the findings from this study to clinical applications to predict the outcome of immunotherapy in breast cancer patients.  23  2.2 MATERIAL AND METHODS 2.2.1 Mouse tumour samples and microarray experiments The mouse tumour samples used in this study originated from cell lines generated from a transgenic mouse tumour model described in the previous study by Wall and colleagues [43]. These cell lines were derived from tumours that spontaneously arose from the transgenic mouse model that over-expresses Her2/neu OT-I/OT-II under the control of the MMTV promoter. The tumours generated from these cell lines consistently showed distinctive and reproducible responses to adoptive T-cell transfer therapy, namely: complete regression (CR) (no detectable tumour); partial response (PR) (>50% reduction from the original tumour size); stable disease (SD) (<50% reduction or <25% increase in tumour size); and progressive disease (PD) (>25% increase in tumour size). The naming convention of the NOP mouse tumour cell lines with associated outcome in superscript will be used throughout this chapter. In total, tumours from six NOP tumour cell lines were subjected to mRNA expression profiling using microarrays; NOP21CR, NOP13PR, NOP23PR, NOP6PD, NOP17PD and NOP18PD. Affymetrix (Santa Clara, USA) mouse exon arrays (MoEx-1_0-st-v1) were used with biological duplicates, which is an independent and separate tumour mass that originated from the same cell line (the tumours can be from the same mouse host). Approximately 1x106 cells from a specific cell line were subcutaneously injected under the mammary fat pads of host transgenic mice. When the tumour reached to approximately 50 mm2 in size, tumours were excised from the mouse and immediately snap-frozen in liquid nitrogen.  24  Frozen tumours were subsequently transported to the Centre for Translational and Applied Genomics (CTAG) at the British Columbia Caner Agency to extract and process RNA, and to perform microarray experiments following the protocol from the GeneChip® Whole Transcript Sense Target Labeling Assay (Affymetrix, USA). In brief, 1-2 µg of total RNA was extracted and purified. The poly-adenylated RNA molecules were captured, and subsequently underwent further purification processes to filter mRNAs and concentrate poly-adenylated RNAs in a sample. Purified poly-adenylated RNAs were reverse-transcribed to cDNAs and amplified. The 5’-to-3’ sense strands of cDNA duplexes were isolated and chemically sheared to shorter fragments whose 3’ terminals were chemically bonded with fluorescence labels. The fluorescence-labeled single-stranded cDNA fragments were then hybridized to probes on a microarray, which was subsequently scanned by a laser to produce the fluorescence intensity, which is proportional to the absolute amount of cDNA fragments that hybridized to a probe representing a specific mRNA in mouse transcriptome. The scanned image of a microarray was stored in an image file known as a ‘CEL’ file, which was provided by CTAG as the final product of the microarray experiment.  2.2.2 Microarray data analysis Raw probe intensity values were generated from microarray images in CEL files and analyzed further using the Expression Console software provided by Affymetrix Inc. The raw intensity values were normalized by the sketch-quantile normalization method and then summarized using the median polished Robust Multichip Average (RMA) method [83]. All probe values were analyzed at the gene level. Also, the intensity values of the 25  ‘core’ probes that were designed from the Refseq full-length mRNA sequences were considered for further analyses, because the core probes were designed with the highest confidence in annotation. Microarrays of biological duplicates were considered as independent datasets, and these were treated equally as other microarrays of tumours generated from different cell lines with the same outcome. Notably, the microarrays were prepared in two separate batches on different dates. The systemic bias, also known as the ‘batch effect’, which arose from this non-biological variation was corrected using R script, Combat.R, which was shown to work optimally for a microarray experiment with a small sample size (N < 25) [84]. Genes with differential expression were detected using the Significance Analysis of Microarrays (SAM) methodology [85], which was implemented in the R Bioconductor package, siggenes. As a measure of statistical significance of differential gene expression in tumours from different outcome groups, the false discovery rate (FDR) of 0.10 was set as the significance threshold, except when CR and PR tumour groups were compared. Due to a lack of genes that were significantly differentially express at the FDR 0.10, the FDR threshold of 0.15 was used to define the differentially expressed gene set between CR and PR. To optimize the clustering analysis, probe intensity values were transformed into ratiolike values by normalizing to the median probe values across all arrays. The hierarchical clustering analysis was performed on log2-transformed ratio-like probe values using Java Cluster (version 3.0) [86], which was adapted from the original Cluster software written by Michael Eisen [87]. Clustering of genes and samples was performed based on the iterative distance calculation between two hypothetical clusters by taking all pair-wise  26  distances between data points and subsequently cluster those with the maximum distance. The Pearson correlation coefficient was used to construct the similarity distance matrix. The clustering result was visualized with the TreeView software [86]. From the clustering analysis, differentially expressed genes in CR tumours relative to PD tumours were ranked by q-values according to the SAM analysis. The Connectivity Map tool allowed for a maximum of 100 genes as an input. Thus the top 100 differentially expressed genes were subjected to drug discovery analysis using the Connectivity Map database [88] to find small molecules or drugs that modulate the gene expression patterns that best emulate the expression profiles observed in CR tumours.  2.2.3 Experimental validation of trichostatin A in vivo The experimental validation of trichostatin A (TSA) in vivo was carried out by Michele Martin at the Trev & Joyce Deeley Research Centre, BCCA. In brief, transgenic NOP mice were injected with NOP18PD cells in the mammary fat pads. In the treated mouse group, the mice (Nmouse = 3; Ntumour = 6; Nreplicate = 9) were treated with 0.5mg/kg TSA intraperitoneally once every 24 hours for seven days. After 24 hours following the last treatment, mice were given a regular dose of adoptively transferred T-cells (15x106 CD8+ (OT-I) and 15x106 CD4+ (OT-II) T-cells) intravenously. Blood samples were taken 1, 4 and 7 days following adoptive T-cell therapy. There were three control groups: (1) a group given DMSO (vehicle) every 24 hours instead of TSA (Nmouse = 1; Ntumour = 2; Nreplicate = 2); (2) a group treated with adoptive T-cell therapy alone (Nmouse = 1; Ntumour = 1; Nreplicate = 2) and (3) a group treated with TSA alone (Nmouse = 1; Ntumour = 2; Nreplicate = 3). To assess T-cell infiltration, immune cell scoring was performed using a 27  tissue microarray subjected to immunohistochemistry with antibodies to CD3 and a 5x5 Chaulkey 25-point grid. A tumour stroma and tumour epithelium were evaluated separately for infiltration of positively stained cells. Statistical analysis and the figures generation were performed by Michele Martin using Graph Pad software.  2.2.4 Outcome-specific differential expression, classifier identification and pathway enrichment analyses We determined the genes whose expression profiles were differential exclusively in a single outcome group compared to the others. These ‘outcome-specific’ genes were identified by taking the overlap of differentially expressed genes from two different kinds of comparisons: (1) a comparison of two combined outcome groups against a specific outcome group e.g. for CR-specific genes, CR vs (PR and PD); (2) a series of pair-wise comparisons to exclude differentially expressed genes from a specific outcome group that were also differentially expressed in the other outcome groups e.g. for CR-specific genes, (CR vs PD) – (PR vs PD). Differentially expressed genes between CR and PD tumours (N = 1,242) were analyzed to identify those that can classify different outcomes of adoptive T-cell transfer therapy. The ‘classifier’ analysis was carried out by using the Predictive Analysis of Microarray (PAM) method [89] based on a supervised clustering of nearest shrunken centroids, which was implemented in the pamr function in the R bioconductor package. The threshold for the centroid shrinkage, i.e. the delta value, was chosen to represent the FDR rate of 0.00.  28  Pathway enrichment analysis was done by mapping the differentially expressed genes to well-known canonical pathways using Ingenuity Pathways Analysis software (Ingenuity# Systems, Redwood city, CA, USA) [90]. P-values generated from the Ingenuity software were calculated using modified Fisher’s exact test and then corrected for multiple hypotheses testing using the Benjamini-Hochberg method. Additional pathway analyses were also done with the on-line bioinformatics tool, DAVID [91], which collates pathway information of BioCarta and KEGG databases [92, 93]. The pvalues from DAVID were also corrected for multiple hypothesis testing using the Benjamini-Hochberg method.  2.2.5 Validation of outcome-specific DE genes Genes were ranked by the fold change difference in gene expression between CR and PD tumours. Genes found as classifiers according to PAM analysis were given precedence in ranking. Genes were also considered for their involvement in immune-related pathways that were significantly enriched for outcome-specific genes (corrected p-value < 0.05), the likelihood for therapeutic potential and potential amenability to in vitro and in vivo RNA interference according to the literature. A total of 33 genes were selected for empirical validation using real-time quantitative-PCR (RT-qPCR). RT-qPCR-aided validation was carried out by Michele Martin at the Trev & Joyce Deeley Research Centre, BCCA. In brief, untreated tumours were harvested and immediately snap-frozen in liquid nitrogen. Whole tumour tissue was homogenized from a frozen state, and total RNA was extracted and purified. Subsequently, the purified  29  RNA was reverse transcribed to cDNA and amplified by RT-qPCR. The relative abundance of each mRNA was normalized against the expression level of !-actin.  2.2.6 Comparison with the molecular profiles from human breast cancer subtypes The CEL files of 82 whole-tissue breast tumours from breast cancer patients used in the study by Rouzier and colleagues [70] were obtained from Dr. Lajos Pusztai through a personal correspondence. The authors used HG-U133A Affymetrix 3’ expression array to profile gene expression. Similar to what was done for the mouse data, the raw probe intensity values extracted from the CEL files were normalized using the sketch-quantile normalization method and then summarized using the RMA method by the use of the Expression Console software from Affymetrix. The systematic bias arising from separate batches on different microarray platforms for human and mouse breast tumours was corrected using the distance weighted discrimination (DWD) method, which was shown to work optimally for microarray experiments with a larger sample size (N > 25) [94]. To combine the genes from mouse and human dataset to compare probe intensities across different microarrays, the Ensembl database was used to pair up the annotated orthologous genes between mice and humans. The mouse-human merged gene set was filtered for a one-to-one orthologous relationship between mouse and human genes in the Ensemble database using a web-based cross-databases query program, Biomart [95]. The orthologous pairs of mouse and human genes were further filtered to obtain genes targeted by probes annotated from both the human HG-U133A and mouse exon microarrays. In the case of multiple probes annotating a single gene, the median values 30  were taken to represent the overall gene expression. Genes with ambiguous biological annotation, e.g. a single probe annotated to target multiple genes, were removed from further analyses. The cross-species intrinsic breast cancer gene set between mice and humans (N = 106) was obtained from Jason Herschkowitz via a personal correspondence [96]. The cross-species intrinsic breast cancer genes that did not have a one-to-one orthologous relationship according to the Ensembl database or genes that were not targeted by probes on either human or mouse microarrays were also excluded from further analyses. A total of 82 genes passed the filtering process and were clustered across 94 human and mouse merged microarray datasets. The unsupervised hierarchical clustering on the merged dataset was performed using the Java Cluster (version 3.0) program. The data transformation of the probe intensity values for clustering was carried out by the same process as the mouse microarray data analysis. The parameters used for the clustering algorithm were the same as the clustering analysis of the mouse microarray data. The pvclust function [97] from the R packages was used to generate bootstrap samplings from the human-mouse merged dataset and calculated the approximately unbiased (AU) probability values of the bootstrap re-sampling result to assess the statistical significance of the hierarchical clustering result.  31  2.3 RESULTS ! 2.3.1 Gene expression profiles of CR, PR and PD tumours The NOP tumour lines used in this study have been previously described to show reproducible responses to adoptively transferred CD8+ (OT-I) and CD4+ (OT-II) T-cells [43, 81, 82]. The nomenclature of the tumour cell line consisting of the NOP cell line name with the superscript text to indicate the corresponding response class will be used throughout this chapter. Six cell lines were subjected to gene expression profiling on the Affymetrix platform; NOP21CR, NOP13PR, NOP23PR, NOP6PD, NOP17PD and NOP18PD. Rather than using cultured tumour cells, fresh tissue explants from untreated NOP tumours were used in this study to ensure that gene expression profiles closely reflected the environment encountered by adoptively transferred T-cells in vivo. Tumour cells were implanted subcutaneously in the mammary fat pads of female mice and harvested when they reached a size of approximately 50 mm2, which represents the typical size used previously in adoptive T-cell therapy experiments [43, 81, 82]. RNA was extracted from tumour tissue and subjected to gene expression profiling using the Affymetrix mouse exon array platform. Gene expression profiles were compared amongst three outcome groups: CR, PR and PD. A two-class unpaired modified t-test for three combinations of pair-wise comparisons among CR, PR and PD was performed using the Significance Analysis Microarray (SAM) tool [85] to identify differentially expressed genes. The pair-wise comparison of CR and PD tumours identified 1,242 differentially expressed genes, whereas the pair-wise comparison of PR and PD tumours identified 1,466 differentially expressed genes at a false discovery rate (FDR) of 0.10. Only 6 genes were differentially 32  expressed between CR and PR tumours (Car12, Myo5C, Scg2, Bbox1, Plb1, 3830417A13Rik), suggesting high similarities between these two tumour groups at the transcriptome level. The hierarchical clustering of the normalized data from all twelve arrays showed the high transcriptional similarity between CR and PR tumours (Figure 2.1a). Similarly, the high similarity between the CR and PR groups was also observed from the unsupervised hierarchical clustering analysis that compared differentially expressed genes in all three outcome groups. The hierarchical clustering of normalized data A total of 5,149 genes were discovered to differentially express in CR, PR and PD using the modified ANOVA test from the multi-class unpaired SAM analysis at the FDR of 0.10. A subsequent unsupervised hierarchical clustering analysis of these genes showed a clustering of CR and PR tumours together, leaving PD tumours as a separate group (Figure 2.1b). Figure 2.1 The clustering of different NOP tumours The dendrogram was generated from hierarchical clustering of gene expression profiles from the cell lines NOP21CR, NOP13PR, NOP23PR, NOP6PD, NOP17PD and NOP18PD. This revealed the expected close relationships between biological replicates, clear distinctions among the CR, PR and PD outcome groups, and greater similarity among regressing tumours (CR and PR) compared to non-regressing tumours (PD): (a) Clustering of the normalized data; (b) clustering of differentially expressed genes among CR, PR and PD tumours (a)  33  (b)  The high degree of transcriptional similarities between CR and PR was consistent with the fact that both tumour types were responsive to adoptive T-cell therapy. However, the CR and PR outcomes differ in the extent and duration of regression. This difference might be attributable to the six genes that were differentially expressed between these groups. These genes were Car12, Myo5C, Scg2, Bbox1, Plb1 and 3830417A13Rik, which exhibited a substantial difference in the relative fold change of gene expression between CR and PR tumours: 10.7, 5.2, -5.0, -10.7, -14.6 and -19.2 respectively (a negative fold change indicates the under-expression of PR compared to CR; in this chapter, the over- or under-expression of a gene indicates a relative increase or decrease of a fluorescence intensity level of a probe respectively targeting the gene in one outcome group, which directly correlates the abundance of mRNA transcript of the gene, compared to that from another outcome group).  34  2.3.2 Discovery of tricostatin A as a potential therapeutic drug using the Connectivity Map database To investigate possible therapeutic applications using the gene expression sigatures, a web-based database search tool, the Connectivity Map [88], was used to find small molecules or known drugs that induce gene expression changes similar to those observed in CR tumours. The Connectivity Map searches a database of approximately 7,000 gene expression profiles from Affymetrix HU-133A microarrays from normal human skin tissues and various human cancer cell lines treated with 1,309 compounds including drugs and small molecules. The gene expression signature used to search the Connectivity Map consisted of the top 100 genes that most differentiated between CR and PD tumours, of which 80 genes were relatively over-expressed and 20 genes were relatively under-expressed in CR compared to PD. The Connectivity Map compared this gene expression signature with those in the database and calculated the enrichment score using the gene set enrichment analysis approach [98, 99], which increases the enrichment score for a similar gene expression pattern between two gene expression signatures e.g. over-expression of a gene in both signatures, or decreases the score for an opposite gene expression pattern e.g. over-expression in one signature and under-expression in the other. The expression profile identified from the Connectivity Map that had the most positive enrichment score (score = 0.311) and the most significant p-value (p-val << 0.01) was from the MCF7 human breast cancer cell line treated with trichostatin A (TSA). The positive enrichment score represents the drug that modulates the gene expression in a tumour such that it assimilates the gene expression patterns observed in  35  CR. This suggested that the administration of TSA had a potential to convert PD to CR outcome after adoptive T-cell therapy. Michele Martin at the Trev & Joyce Deeley Research Centre, BCCA, carried out the experimental validation of TSA for modulating the PD response to the CR in vivo. TSA was administered to mice every 24 hours for 7 days, which carried the NOP18PD tumour that were grown under the mammary fat pads followed by adoptive transfer of T-cells. There were three kinds of control mouse groups that were either treated with DMSO or nothing at all prior to adoptive T-cell therapy, or did not receive adoptive T-cell therapy after TSA-treatment. The progression of tumours measured by the tumour area, the proportion of circulating donor T-cells and the extent of T-cell infiltration in tumour epithelium and tumour stroma were measured from the control groups and from the TSAtreated group. The experimental validation revealed that there was no significant change in the outcome of PD tumours that were conditioned with TSA before adoptive T-cell therapy. There was no apparent decrease in the tumour size in the TSA-treated tumours compared to the control groups (Figure 2.2a). Similarly, there was no increase in the number of donor Tcells that circulated bloodstreams in the recipient mouse that was treated with TSA (Figure 2.2b, 2.2c).  36  Figure 2.2 Progression of tumours treated with TSA before adoptive T-cell therapy Progression of tumours treated with TSA prior to receiving adoptively transferred Tcells: (a) Tumour areas measured from the control group and the TSA-treated group; measurement of circulating CD8+ CTLs (b) and CD4+ helper T-cells (c) that were adoptively transferred to a recipient mouse bearing NOP18PD tumours. (a)  (b)  37  (c)  The extent of T-cell infiltration in the tumour epithelium and the tumour stroma also did not differ in TSA-treated tumours compared to non-treated tumours (Figure 2.3).  Figure 2.3 Intratumoural infiltration of T-cells after TSA treatment T-cell infiltration into NOP18PD stroma and epithelium were measured 7 days postadoptive T-cell therapy following TSA (or vector) treatment.  38  2.3.3 Pathway enrichment analysis of outcome-specific genes To identify biological pathways that might regulate the response of tumours to adoptive T-cell therapy, tumours from the most extreme outcome classes, i.e. CR and PD, were subjected to the pathway analysis. Genes that were differentially expressed specifically in a single outcome group compared to the other two groups (FDR = 0.10) were identified in this analysis. These ‘outcome-specific’ genes were identified by taking the overlap of differentially expressed genes from a ‘lenient two-state’ pair-wise comparison and the combined outcome group comparisons (Figure 2.4b). Initially, all three pair-wise combinations were tried to identify outcome-specific genes, however, this approach resulted in an insufficient number of genes to have statistical power in the pathway enrichment analysis (Figure 2.4a). To increase the number of genes for the enrichment analysis, the pair-wise comparison between CR and PR was performed at FDR of 0.15 instead of 0.10 (N = 31).  Figure 2.4 Identification of ‘outcome-specific’ differentially expressed genes Identification of ‘outcome-specific’ differentially expressed genes. The differentially expressed genes were compared in a combination of pair-wise comparisons and the combined outcome group comparisons: (a) Strict comparisons of all three pair-wise comparisons; (b) lenient comparisons using two pair-wise comparisons (blue circles). To compensate for the loss of stringency of detecting the outcome-specific genes, the genes that were also differentially expressed when two combined outcome groups were compared to a single outcome group, (yellow circles) were identified to be outcomespecific genes.  39  (a)  (b)  40  To investigate the pathways involved in a specific outcome, pathway enrichment analysis was carried out using a larger dataset of genes identified from the method of taking the overlap of genes found in both the combined outcome groups and two-state pair-wise comparisons. This approach is basically a subtraction of differentially expressed genes in the other two outcome groups from those in a pair-wise comparison involving a specific outcome group. Since only one out of the two possible pair-wise comparisons involving the other two outcome groups were used (considering all two pair-wise comparisons was the strict comparison described in Figure 2.4a), this ‘lenient’ method would identify more genes as being outcome-specifically differentially expressed than the strict original comparison method. Being less stringent, this approach would invariably have more  41  false positives. However, it would also allow more genes for the pathway analysis, which had subtle changes in gene expression that could collectively influence a single pathway. Using the lenient approach, 229 CR-specific, 42 PR-specific and 880 PDspecific genes were identified. The significant difference in gene expression levels amongst the outcome groups were visualized by the heatmaps after the unsupervised hierarchical clustering of outcome-specific genes (Figure 2.5). Figure 2.5 Heatmaps of the outcome-specific gene expression signatures The outcome-specific gene signatures were compared by the unsupervised hierarchical clustering of outcome-specific DE genes based on complete linkage using Pearson correlation coefficient as similarity distance. (a) Heatmap of 42 PR-specific gene signature; (b) heatmap of 229 CR-specific gene signature; (c) heatmap of 889 PD-specific gene signature (a)  42  (b)  43  (c)  44  To identify whether differential gene expression could be associated with specific pathways, the outcome-specific DE genes were mapped to known pathways using Ingenuity Pathways Analysis (Ingenuity# Systems, Redwood city, CA, USA)[90]. The majority of the outcome-specific pathways were found to be immune-related pathways. Pathways enriched in CR-specific DE genes include: the complement pathway (p-value = 1.06e-08); the pattern recognition receptor pathway (p-value = 1.83e-05); TREM1 signaling pathway (p-value = 3.06e-04); Liver X receptor/Retinoid X receptor (LXR/RXR) activation pathway (p-value = 6.92e-04); cytosolic pattern recognition receptor IRF activation pathway (p-value = 2.48e-03); fibrogenesis pathway (p-value = 0.012); and the dendritic cell maturation pathway (p-value = 0.012). Pathways enriched in PD-specific DE genes include: the pattern recognition receptor pathway (p-value = 4.85e-05); interleukin-3 (Il-3) signaling pathway (p-value = 4.08e04); fibrogenesis pathway (1.82e-03); granulocyte macrophage-colony stimulating factor (GM-CSF) signaling pathway (p-value = 1.82e-03); leukocyte extravasation signaling pathway (p-value = 1.82e-03); Interferon signaling pathway (p-value = 1.82e-03); Fc "receptor class II B (Fc"RIIB) signaling pathway (p-value = 1.82e-03); aminosugars metabolism pathway (p-value = 2.55e-03); cytosolic pattern recognition receptor IRF activation pathway (p-value = 2.86e-03); and the complement system pathway (p-value = 2.86e-03). To test the robustness of the findings from the Ingenuity software, another web-based bioinformatics software, DAVID [91], was used to identify significant pathways enriched for outcome-specific genes. DAVID performs pathway enrichment analysis from multiple pathway databases including KEGG [93] and BioCarta [92]. The most 45  significantly enriched pathway found by DAVID match the pathways identified by the Ingenuity software. The complement pathway was significantly enriched for CR-specific genes in KEGG (p-value (corrected for multiple hypothesis testing using the BenjaminiHochberg method) = 0.0054). The complement pathway was also the top ranked pathway in BioCarta, however the p-value was not significant after multiple hypothesis testing correction (p-value = 0.409). Interestingly, we also found the extra cellular matrix receptor interaction pathway was the most significant pathway enriched for PD-specific genes in KEGG (corrected p-value = 0.00702). Overall, the pathway enrichment analysis using multiple pathway databases confirmed the findings that immune-associated pathways were significantly enriched for CR-specific genes. Notably, some of the top ranking immune-related pathways from the pathway enrichment analysis were enriched for either CR-specific or PD-specific genes. These pathways were especially enriched for genes that were over-expressed in CR relative to PD tumours. The pathways enriched for CR-specific genes included the complement pathway (corrected p-value = 4.29e-09 from the Ingenuity pathway analysis), pattern recognition pathway (corrected p-value = 7.66e-06), TREM1 signaling pathway (corrected p-value = 1.57e-04) and LXR/RXR signaling pathways (corrected p-value = 3.59e-04). The PD-specific genes were also tested for the pathway analysis, with the expectation that PD-specific genes would be differentially under-expressed relative to CR tumours in the same pathways. Intriguingly, different pathways such as the Il-3 signaling pathway (corrected p-value = 4.08E-04), Fc"RIIB signaling pathway (corrected p-value = 4.85E-05), GM-CSF signaling pathway (corrected p-value = 1.82E-03) and the leukocyte  46  extravasation pathway (corrected p-value = 1.82E-03) were found to be enriched for PDspecific genes that were under-expressed relative to CR tumours.  2.3.4 Validation of the predictive gene expression signature To experimentally validate the outcome-specific genes associated with immune-related pathways, 33 genes were selected for validation by quantitative PCR (qPCR) from the list of outcome-specific genes that mapped to top ranking significant pathways from the Ingenuity Pathway Analysis, KEGG and BioCarta databases. The genes were ranked by fold change differences between CR and PD tumours from the microarray experiments, and then further selected (by Michele Martin and Dr. Brad Nelson at the Trev & Joyce Deeley Research Centre, BCCA) based on the likelihood of therapeutic potential, availability of antibodies for the gene products to allow validation at the protein level, and potential amenability to in vitro and in vivo RNA interference. The genes identified to be significant classifiers according to the PAM analysis with the threshold value at 3.0 (FDR = 0.00) were also given precedence in ranking. The experimental validation using qPCR was done by Michele Martin at the Trev & Joyce Deeley Research Centre, BCCA. In addition to the six tumours analyzed by Affymetrix (NOP21CR, NOP13PR, NOP23PR, NOP6PD, NOP17PD and NOP18PD; the “training set”), the expression profiles of the selected genes were assessed from four additional NOP tumours exibiting the PR and PD outcomes (NOP12PR, NOP9PD, NOP14PD and NOP16PD; the “validation set’). Using qPCR, relative gene expression differences in all 10 tumours were assessed and normalized to the gene expression from NOP21CR tumours as a reference, and then compared to differential gene expression 47  measured by Affymetrix analysis for each of the 33 selected genes. The relative gene expression was averaged across the outcome groups for all PR and PD tumours. The consistency in gene expression patterns between those measured by Affymetrix and by qPCR in the PD tumours compared to NOP21CR was used as an initial screen to identify tested genes worthy of further scrutiny for validity. For example, using this approach, if the gene expression pattern from microarray analyses was found to be differentially overexpressed in NOP21CR tumours compared to PD tumours, and the same pattern was not confirmed using qPCR, the gene was not considered valid. As a further benchmark, a fold difference of 1.5 times in magnitude (either over-expressed or under-expressed) in the gene expression determined by qPCR compared to that in NOP21CR tumour was set as an arbitrary cut-off to consider the gene measurably differentially expressed. With these two screening criteria, 14 genes showed expression patterns that continued to correlate with outcome (Table 2.1, Figure 2.6).  48  Figure 2.6 Gene expression level of selected genes for validation using qPCR The expression levels of genes selected for validation (N = 33) were determined using qPCR (expression levels are relative to that in NOP21CR tumour). The numbers on the xaxis on the left graphs indicates the NOP cell lines. The black bars indicate the gene expression from the CR tumours; the grey bars from the PR tumours and the white bars from the PD tumours. The graphs on the right show the average gene expression across the three outcome groups. (a) Genes were successfully validated; (b) genes failed to validate. Figures were generated by Michele Martin. Note: -On the x-axis of the graph on the left, ‘*’ below the NOP cell line number indicates those subjected to the Affymetrix microarray experiments.  (a)  49  50  51  (b)  52  53  54  55  56  Table 2.1 List of qPCR-validated genes Validated genes (N = 14) whose expression correlated with the outcome of adoptive Tcell therapy. The table was modified from the original generated by Michele Martin and Dr. Brad Nelson. Gene Name  Pathway(s) involved  Fold difference: microarrays (CR:PD)  Fold difference: qPCR (CR:PD)  Gene Information  Reln*  Extracellular Matrix Receptor  5.72  8.40  Extracellular glycoprotein regulating neuronal migration during brain development. Silencing implicated in gastric & pancreatic cancer.  Il18*  LXR/RXR, TREM1  2.76  3.57  Induces IFN! secretion in T cells & NK cells. Implicated in inflammatory responses.  Ppp3ca*  GMCSF, IL3  2.31  3.44  Calcineurin; calcium-dependent serine/threonine protein phosphatase. Activates NFATc, upregulates IL-2 expression.  Ifih1  Pattern Recognition  4.42  1.88  MDA5; component of innate immune response (viral RNA sensor). Ligation triggers IFN!mediated apoptosis.  Irf7  Pattern Recognition  6.47  1.79  Controls expression of IFN"/# genes as well as RANTES; induces cellular senescence.  Abca1  LXR/RXR  2.81  1.75  Lipid transporter expressed in epithelial cells, Multi-drug resistance gene.  Blnk  FcyRIIB  2.46  1.69  B cell linker protein, tumor suppressor, involved in humoral immune response. Wide range of tissue expression.  Jak1  IL3  1.9  1.68  Cytokine signal transduction, altered expression implicated in invasive breast cancer.  Cldn1  Leukocyte Extravasation  3.72  1.66  Tight junction protein, metastasis suppressor, epithelial cell expression.  57  Gene Name  Pathway(s) involved  Fold difference: microarrays (CR:PD)  Fold difference: qPCR (CR:PD)  Gene Information  Cd55  Complement  3.86  1.64  Complement regulating factor. If secreted can inhibit T cell function.  Pak1  IL3  2.69  1.56  p21-activated kinase; regulates apoptosis.  TLR7  Pattern Recognition, TREM1  4.89  1.50  Induces NK-mediated anti-tumor response.  Shc1*  GMCSF, FcyRIIB, IL3  -2.24  -2.737  Signal transduction protein; propagation of mitogenic signals. Wide tissue expression.  Cldn4*  Leukocyte Extravasation  -4.67  -13.087  Tight junction protein,wide tissue expression.  Note: -The fold difference from microarrays and qPCR is relative between CR and PD with CR as a reference. For example, the fold difference of 5.72 in Reln means that the expression of Reln was 5.72 times more in CR than in PD. Conversely, the fold difference of -13.1 in Cldn4 indicates that Cldn4 was 13.1 times less in CR compared to PD. -‘*’ indicates a gene with the strong prognostic power.  58  From the validated 14 genes, five genes (Reln, Il18, Ppp3ca, Shc1 and Cldn4) in particular showed the greatest differential expression between CR and PD tumours (Table 2.1). Some genes had patterns of differential expression across the outcome groups that demonstrated diminished differential expression in PD, i.e. highest in CR, intermediate in PR, lowest in PD, as seen in the expression of Abca1, Blnk, Il18, Ppp3ca and Pak1 (Figure 2.5). Conversely, the opposite pattern of gene expression was observed for Reln. Remarkably, these five genes with the strongest differential expression were found to be a strong predictive signature for outcome of adoptive T-cell therapy. In an ongoing project with Dr. Brad Nelson, the signatures from the gene expression patterns of these five genes were used to successfully predict new CR tumour cell lines, NOP20CR and NOP37CR, which was subsequently validated in vivo to reproducibly show the CR outcome after the adoptive T-cell therapy (data not shown). Gene expression levels of the five genes were assessed using qPCR across five new NOP tumours. From the five NOP tumours, NOP20 and NOP37 tumours had the relative gene expression patterns identical to those of CR and later exhibited the complete regression outcome after adoptive T-cell therapy.  2.3.5 Comparison with human breast cancer subtypes Thus far, the findings from this study showed specific gene expression differences between mouse mammary tumours that responded to adoptive T-cell therapy and those that did not. Clinically, there are human breast cancer subtypes known to have a better or worse prognosis than others [62, 74, 78, 100]. The transcriptional similarities between the mouse data and the data from specific human breast cancer subtypes would imply the 59  potential application of the findings in this study to human breast cancer patients. To determine whether the different outcome groups in mouse tumours resembled a specific human breast cancer subtype, the mouse tumour expression profiles were compared with human tumour expression profiles of pre-chemotherapy whole breast tumour tissues of 82 patients from Rouzier and colleagues [70]. A ‘cross-species breast cancer intrinsic gene set (CBC)’ consisting of 82 genes was identified following a series of filtering processes (Figure 2.7), which represented a subset of highly conserved genes between mouse and human that were good classifiers of breast cancer subtypes and showed the least variation within the same breast cancer subtype, but greatest variation between different breast cancer subtypes. The selection of this subset of genes with a more robust gene expression pattern reduced the false discovery due to technical and biological noise.  60  Figure 2.7 The process of merging and analysis of gene expression profiles from human breast cancer subtypes and mouse NOP tumours to identify the CBC gene set.  To construct the CBC dataset, genes contained on both the Affymetrix mouse exon arrays and the HU-U133A arrays used by Rouzier and colleagues were filtered initially to identify the common gene sets targeted by both platforms of microarrays that were in a ‘one-to-one’ orthologous relationship between mouse and human according to the Ensembl database. To identify the cross-species breast cancer intrinsic gene sets within the filtered gene sets, the gene sets were overlapped with the 106 mouse-human breast cancer intrinsic gene sets from the study by Herschkowitz and colleagues [96], and identified 82 genes as the mouse-human (or cross-species) breast cancer intrinsic gene set  61  that were probed by both mouse exon arrays and human HG-U133A arrays from Rouzier and colleagues. To detect transcriptional similarities between the mouse and human tumours, an unsupervised hierarchical clustering analysis over the merged data of human and mouse microarrays was subsequently performed using the CBC subset of 82 genes (Figure 2.8a).  62  Figure 2.8 Comparison of the mouse tumour expression profile with human breast cancer subtypes Comparison of the mouse tumour expression profile with the human breast cancer subtype expression profile. Unsupervised clustering analysis was performed using 82 CBC genes over the merged data from 82 human microarrays and 12 mouse microarrays. Clustering was based on the complete linkage using Pearson correlation coefficient as a similarity distance. (a) A complete heatmap of CBC gene set clustered (green represents relative under-expression, and red represents over-expression); (b) a colour-coded dendrogram indicating mouse samples and human breast cancer subtypes determined from estrogen/progesterone receptor, HER2 over-expression status from the study by Rouzier and colleagues. (a)  63  (b)  64  The clustering result repeatedly demonstrated systematic transcriptional differences between the response (CR and PR) and non-response (PD) outcome groups (Figure 2.8b). All NOPCR and NOPPR tumours clustered together with the human breast cancer samples whose molecular status mostly belonged to HER2-overexpressing subtype, where 92% of the sample was ER- and PR-negative, and 67% was HER2-over-expressing positive. Two out of three NOPPD cell lines clustered with the human breast cancer samples in which 100% of them had ER-negative status; 76% had PR-negative status; and 88% had HER2-over-expressing negative status, resembling the basal-like human breast cancer subtype with the ‘triple-negative’ molecular status, i.e. ER-negative, PR-negative, HER2negative. Intriguingly, NOP18PD tumours clustered with the human breast cancer samples that mostly resembled the molecular status of the luminal-A subtype i.e. ERpositive (85% of the samples) and PR-positive (77%) and HER2-positive (58%). It is notable that all of the mouse mammary tumours in this study were selected specifically in part for their over-expression of Her2/neu, but interestingly, only the responsive (CR and PR) tumour groups showed transcriptional similarities to the HER2-over-expressing human breast cancer subtype whereas the majority of the non-responsive (PD) tumours were similar to the basal subtype. To calculate the significance of the clustering result, a total of 5,000 bootstrap samples were generated using the random sampling method and repeated hierarchical clustering to obtain dendrograms generated from bootstrap samples (Figure 2.9). The frequency of forming the same clusters during 5,000 iterations was calculated and the probability of repeatedly obtaining a specific cluster from multiple iterations was calculated (the approximately unbiased p-value).  65  Figure 2.9 Significance clustering result of 82 CBC gene sets over the merged data from the mouse and human breast tumours The significance of the clustering analysis was calculated by multi-scale bootstrap resampling after 5,000 iterations was represented by the approximately unbiased p-value in red and the bootstrap probability in green.  66  The clustering of PD tumours with the basal-like subtype, and the clustering of the response groups (CR and PR) with the HER2 over-expressing subtype were both found to be significant with a p-value of 0.03 (Figure 2.9).  2.4 DISCUSSION The gene expression profiles were distinctive among CR, PR and PD with higher similarities between the regressing tumours (CR and PR) than the non-regressing (PR) tumours. CR and PR tumours were virtually indistinguishable at the transcriptomic level, with only six genes that were differentially expressed. Although all six genes showed a large difference in gene expression levels between CR and PR, there were no obvious immune-related functions associated to these genes, prompting a hypothesis that the mechanisms responsible for the different regression outcomes of CR and PR tumours might be controlled at the post-transcriptional level. Interestingly, the PD tumour class showed a highly distinct transcriptional profile, which provided support for the hypothesis that the difference in outcomes between the regressing tumours and the nonregressing tumours could be attributed to marked differences in transcriptional profiles. Although speculative, it is possible that the initial regression followed by progression in PR tumours is mechanistically different from the regression in CR tumours and the cancer progression outcome observed in PD tumours. An interesting similarity was observed between the gene signatures comprising the top 100 most differentially expressed genes in CR tumours and those from TSA-treated  67  human breast cancer cell lines from the analysis using the Connectivity Map. TSA is a powerful histone deacetylase inhibitor (HDACi) that has been shown to induce cell cycle arrest, cell differentiation and apoptotic cell death [101-104]. It is also known to affect stromal function by suppressing the accumulation of fibrous connective tissues [101, 105]. Furthermore, recent discoveries heralded the potential application of HDACi as a novel cancer drugs [102, 103, 106] for various tissues or cancer cell lines from breast [107], Hodgkin lymphoma [108], Jurkat lymphoma [109] and colon [109, 110]. Interestingly, several studies reported the immuno-modulating activities of TSA [111, 112], and showed improved immune response to vaccination [113]. In particular, Setiadi and colleagues observed an increase in antigen processing and presentation by tumour cells treated with TSA, which effectively enhanced the overall tumour immunogenicity [123]. Similarly, pre-conditioning of a host with a different HDACi, LAQ824, resulted in sensitization of metastatic melanoma to adoptively transferred T cells in mice [114]. The relevance of TSA activities in immune and stromal functions reported from the literature prompted the experimental validation of TSA as an immune-modulating agent to condition the host before adoptive T-cell therapy. The immuno-modulating effect of TSA on PD tumours did not validate in this study. The dosage and the duration of TSA administration might have been insufficient to induce the TSA-induced sensitivity of tumours to the immune system. Although the effect of TSA on PD tumours did not validate in this study, the compelling evidence from the literature that strongly supports the potential of HDACi as an adjuvant agent shows promise of TSA to increase the overall efficacy of the adoptive T-cell therapy. An increased dosage  68  of TSA and/or a continued administration of TSA throughout the course of adoptive Tcell therapy should be considered for future studies. The identification of outcome-specific genes was accomplished through a combination of ‘lenient’ pair-wise comparisons and the combined outcome group comparisons. Given such high similarities between CR and PR, the analysis could have used CR and PR combined together and then compared against PD to compare the regressing versus the non-regressing tumours. However, the quantitative traits of the PR outcome were not consistent across each PR tumour lines e.g. the extent and duration of regression. Thus, CR and PD tumours, which were consistently the most different and definitive in outcome, were analyzed to investigate the potential pathways involved. The results from the pathway analysis using the outcome-specific genes indicated that the pro-inflammatory pathways were more active in the CR tumours, whereas these pathways were suppressed in the PD tumours, which was suggested by the enrichment of CRspecific over-expressed genes and PD-specific under-expressed genes. Thus, the expression of genes associated with the immune response could be used to differentiate outcomes following adoptive immunotherapy using a pathway enrichment analysis approach. In particular, the activation of the complement pathway observed in CR tumours was reported to modulate the anti-tumour immune response. The complement system is reported to be a ‘double-edged sword’ for controlling tumour growth where it is regarded as being critical for tumour regression by increasing complement receptor-enhanced antibody-dependent cellular cytotoxicity [115]. Conversely, it is reported to enhance tumour growth by promoting pro-inflammatory properties favoring tumour growth and  69  recruitment of myeloid-derived repressor cells [116]. In this study, both activating and inhibitory genes were over-expressed in the CR outcome tumour. The activating genes such as C1q, Masp-2, Cfb, C3a receptor, and C6 were over-expressed exclusively in CR tumours. At the same time, Serping-1/C1 inhibitor and Cd-55/Daf were over-expressed in CR tumours. The effect of the complement pathway in tumour immunogenicity seems to depend on the balance between collective actions of pro-complement genes versus anti-complement genes. The complement pathway is also linked to the Fc! receptor pathway in anti-cancer immune response [117] and is shown to affect the monoclonal antibody-mediated immunotherapy against lymphoma [118, 119] and breast carcinoma [119]. The active complement pathway in CR tumours and the relatively less active Fc! receptor pathway in PD tumours observed from this study supports the importance of the close interaction between these two pathways, and gene signatures from these pathways can potentially be used as predictive markers for outcome of adoptive T-cell therapy. Notably in CR tumours, fibrogenesis and extracellular matrix receptor pathways were enriched in the over-expressed CR-specific gene set, suggesting a higher stromal content in CR tumours. This observation was supported by the findings from Martin and colleagues [81], in which a higher tumour stroma to tumour epithelium ratio was associated with the CR outcome after the adoptive T cell therapy. In the same study, the authors also found that a collagen-rich histological appearance was associated with the tumours compared to PD tumours. The active stroma-related pathways found in CR tumours reiterate the important role of tumour microenvironment in immune modulation, suggesting the possibility of stroma-related genes as potential predictive markers for adoptive immunotherapy.  70  The GM-CSF signaling pathway was significantly enriched for PD-specific genes. Apart from Shc1, the only over-expressed PD-specific compared to CR tumours, the rest of nine genes mapped to this pathway were under-expressed in PD. These genes include Camk2d, Pik3c, Ppp3ca and Stat1. Especially, Ppp3ca is an important catalytic subunit of calcineurin, a Ca2+/calmodulin-dependent protein phosphatase, which activates transcription of GM-CSF [120]. The CSF receptor, Csf2rb, was also found to be underexpressed in PD tumours. GM-CSF is a potent cytokine that plays an important role in the maturation of antigen presenting cells (APCs), which subsequently activate T lymphocytes. One of the ways to increase tumour immunogenicity is to stimulate the antigen presentation by accumulating antigen presenting cells via potent cytokines such as GM-CSF [34]. Indeed, Dranoff and colleagues showed an increased anti-tumour response of the host immune system against tumours transduced with recombinant retroviruses encoding GM-CSF [33] (reviewed in [32, 121] ). The relatively inactive GM-CSF signaling pathway in PD tumours may explain the delayed, reduced and transient activation of T-cells observed in PD tumours as previously described by Martin and colleagues [81]. Finally, the enrichment of PD-specific genes that were underexpressed relative to the CR tumours in the leukocyte extravasation pathway allows for a hypothesis that the suppression of activities in this pathway results in the lack of intratumoural infiltration of T-cells observed in the PD tumours reported by Martin and colleagues [81]. The bioinformatics approach to identify genes associated with a specific outcome and pathways allowed the selection of 33 genes for experimental validation using qPCR. Out of 33 genes, 14 genes were validated to show a correlation between gene expression level  71  and the degree of response to adoptively transferred T-cells. In particular, five genes were experimentally validated to show the most difference in gene expression levels between CR and PD. Interestingly, one of the five genes, Il-18, is a potent inducer of cytolytic T cell responses via the interferon-gamma (IFN-!) pathway, and CR tumour showed the highest expression level of this gene. Indeed, the gene expression level of Il18 showed incremental reduction as the outcome changes from CR to PD, which may indicate that the CR tumours have an increased sensitivity to T-cells relative to PR and PD tumours, contributing to the observed clinical outcomes following the adoptive T-cell therapy. Conversely, Claudin-4 (Cldn4) showed the opposite pattern of gene expression with the lowest level in CR, intermediate in PR and the highest in PD tumours. Cldn4 is an important component of epithelial tight junctions, and increased expression of Cldn4 may result in alterations in membrane functions and structure between tumour epithelial cells and thus impact upon the mechanism of T-cell migration and infiltration into the tumour epithelium. In the studies published by Martin and colleagues [81] and Wall and colleagues, [43], the authors described defects in T-cell infiltration associated with PR and PD tumours, and one could speculate that the increased expression of Cldn4 in PR and PD tumours might be contributing to this observed defect. One of the most remarkable results from this study is the accurate prediction of the new CR cell lines, NOP20CR and NOP37CR based on the gene signatures of five validated genes. The expression level of the five genes from the tumour derived from the NOP20 and NOP37 cell line matched with the signature of CR from this study. Indeed, upon receiving the adoptively transferred T-cells in vivo, the NOP20 and NOP37 tumours exhibited the CR outcome. This shows the strong potential of the predictive signatures  72  consisting of these five genes as predictive markers for adoptive T-cell therapy. Of course, more stringent validation over a larger sample size will be required to test the robustness of this predictive gene signature. Nonetheless, this finding provides an exciting resource for the future development of predictive biomarkers for immunotherapy. The comparison between the expression profiles of CR, PR and PD tumours from mice and those from human breast cancer subtypes revealed intriguing similarities between the PD tumours and the basal-like subtype in human breast cancer. The basal-like subtype is known to have the poorest prognosis among the four subtypes of human breast cancer [62, 74, 77, 78, 100]. The transcriptional similarities between the mouse breast tumours and specific human breast subtypes also reflected similarities in tumour biology. For example, the basal-like and HER2 over-expressing subtypes are known to have a higher proportion (40-80%) of TP53 mutations compared to the luminal subtype in human [74, 122]. Similarly, the NOP tumours contained mutations in Trp53 gene [43]. The clustering of the basal and the HER2 human subtypes to the murine NOP tumours suggested the transcriptional similarities between tumours with mutated p53 in both human and mice. The findings from this study indicated a relatively more active complement pathway in CR and PR tumours suggested by the enrichment of over-expressed genes in CR, which clustered with the ER-negative HER2 over-expressing subtype of human breast cancer. Interestingly, Terschendorff and colleagues described a similar finding in their study where a subclass of ER-negative breast cancer with good prognosis had a significant association with active complement and immune response pathways [63]. Thus, the  73  result from this study and the literature both support the role of active immune response pathways for good prognosis in both mouse and human breast cancers. Furthermore, the importance of the complement immune system in modulating the tumour sensitivity to an immune system suggested by the literature is again supported by this study. The striking differences found between transcriptional profiles of the response (CR and PR) and the non-response (PD) outcome groups reiterated the innate heterogeneity in individual tumours even when the tumours originated from syngeneic murine hosts, all genetically engineered to over-express Her2/neu protein. The clustering of the murine breast cancer groups with specific subtypes of human breast cancer suggests that the mouse model can closely emulate distinct human breast cancer behaviors. Importantly, within the HER2/neu over-expressing subtype in human breast cancer, the literature describes varying clinical outcomes. The immune response-related pathways enriched for outcome-specific genes may play a role in this observation. Notably, the mouse mammary tumours with the poorest response to adoptive T-cell therapy showed the highest similarity in gene expression profile to the human breast cancer subtype with the poorest prognosis, suggesting a potential opportunity for targeted therapy directed towards identified genes of interest described in the murine mammary model. The transcriptional similarities observed between the murine NOP tumours and the human breast cancer subtypes could possibly imply that specific human breast cancer subtypes could be responsive to adoptive T-cell therapy. Thus, the efficacy of the current therapeutic regime for breast cancer could improve by a targeted approach for adjuvant immunotherapy. Finally, the findings from this genomic study are consistent with prior immunological experiments in this model [43, 81], which revealed that outcomes of  74  adoptive T-cell therapy are highly dependent on local tumour-specific factors rather than systemic immune parameters. Moreover, the results from this study suggest several candidate pathways for immune modulation as a means to improve the efficacy of adoptive immunotherapy.  75  2.5 BIBLIOGRAPHY  1. Coley WB: The treatment of malignant tumors by repeated inoculations of erysipelas: with a report of ten original cases. Am J Med Sci 1893, 105:487-511. 2. Dunn GP, Old LJ, Schreiber RD: The three Es of cancer immunoediting. Annu Rev Immunol 2004, 22:329-360. 3. Kaplan DH, Shankaran V, Dighe AS, Stockert E, Aguet M, Old LJ, Schreiber RD: Demonstration of an interferon gamma-dependent tumor surveillance system in immunocompetent mice. Proc Natl Acad Sci U S A 1998, 95(13):7556-7561. 4. Shankaran V, Ikeda H, Bruce AT, White JM, Swanson PE, Old LJ, Schreiber RD: IFNgamma and lymphocytes prevent primary tumour development and shape tumour immunogenicity. Nature 2001, 410(6832):1107-1111. 5. Smyth MJ, Thia KY, Street SE, MacGregor D, Godfrey DI, Trapani JA: Perforinmediated cytotoxicity is critical for surveillance of spontaneous lymphoma. J Exp Med 2000, 192(5):755-760. 6. van den Broek ME, Kagi D, Ossendorp F, Toes R, Vamvakas S, Lutz WK, Melief CJ, Zinkernagel RM, Hengartner H: Decreased tumor surveillance in perforin-deficient mice. J Exp Med 1996, 184(5):1781-1790. 7. Gatti RA, Good RA: Occurrence of malignancy in immunodeficiency diseases. A literature review. Cancer 1971, 28(1):89-98. 8. McClain KL: Immunodeficiency states and related malignancies. Cancer Treat Res 1997, 92:39-61. 9. Birkeland SA, Storm HH, Lamm LU, Barlow L, Blohme I, Forsberg B, Eklund B, Fjeldborg O, Friedberg M, Frodin L: Cancer risk after renal transplantation in the Nordic countries, 1964-1986. Int J Cancer 1995, 60(2):183-189. 10. Dighe AS, Richards E, Old LJ, Schreiber RD: Enhanced in vivo growth and resistance to rejection of tumor cells expressing dominant negative IFN gamma receptors. Immunity 1994, 1(6):447-456. 11. Morris M, Platell C, Iacopetta B: Tumor-infiltrating lymphocytes and perforation in colon cancer predict positive response to 5-fluorouracil chemotherapy. Clin Cancer Res 2008, 14(5):1413-1417. 12. Pages F, Berger A, Camus M, Sanchez-Cabo F, Costes A, Molidor R, Mlecnik B, Kirilovsky A, Nilsson M, Damotte D, Meatchi T, Bruneval P, Cugnenc PH, Trajanoski Z, Fridman WH, Galon J: Effector memory T cells, early metastasis, and survival in colorectal cancer. N Engl J Med 2005, 353(25):2654-2666. 13. Dalerba P, Maccalli C, Casati C, Castelli C, Parmiani G: Immunology and immunotherapy of colorectal cancer. Crit Rev Oncol Hematol 2003, 46(1):33-57.  76  14. Galon J, Costes A, Sanchez-Cabo F, Kirilovsky A, Mlecnik B, Lagorce-Pages C, Tosolini M, Camus M, Berger A, Wind P, Zinzindohoue F, Bruneval P, Cugnenc PH, Trajanoski Z, Fridman WH, Pages F: Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science 2006, 313(5795):1960-1964. 15. Tartour E, Gey A, Sastre-Garau X, Lombard Surin I, Mosseri V, Fridman WH: Prognostic value of intratumoral interferon gamma messenger RNA expression in invasive cervical carcinomas. J Natl Cancer Inst 1998, 90(4):287-294. 16. Leffers N, Gooden MJ, de Jong RA, Hoogeboom BN, ten Hoor KA, Hollema H, Boezen HM, van der Zee AG, Daemen T, Nijman HW: Prognostic significance of tumor-infiltrating T-lymphocytes in primary and metastatic lesions of advanced stage ovarian cancer. Cancer Immunol Immunother 2009, 58(3):449-459. 17. Zhang L, Conejo-Garcia JR, Katsaros D, Gimotty PA, Massobrio M, Regnani G, Makrigiannakis A, Gray H, Schlienger K, Liebman MN, Rubin SC, Coukos G: Intratumoral T cells, recurrence, and survival in epithelial ovarian cancer. N Engl J Med 2003, 348(3):203-213. 18. Clemente CG, Mihm MC,Jr, Bufalino R, Zurrida S, Collini P, Cascinelli N: Prognostic value of tumor infiltrating lymphocytes in the vertical growth phase of primary cutaneous melanoma. Cancer 1996, 77(7):1303-1310. 19. Inoue T, Yoneda K, Manabe M, Demitsu T: Spontaneous regression of merkel cell carcinoma: a comparative study of TUNEL index and tumor-infiltrating lymphocytes between spontaneous regression and non-regression group. J Dermatol Sci 2000, 24(3):203-211. 20. Halliday GM, Patel A, Hunt MJ, Tefany FJ, Barnetson RS: Spontaneous regression of human melanoma/nonmelanoma skin cancer: association with infiltrating CD4+ T cells. World J Surg 1995, 19(3):352-358. 21. Dieu-Nosjean MC, Antoine M, Danel C, Heudes D, Wislez M, Poulot V, Rabbe N, Laurans L, Tartour E, de Chaisemartin L, Lebecque S, Fridman WH, Cadranel J: Longterm survival for patients with non-small-cell lung cancer with intratumoral lymphoid structures. J Clin Oncol 2008, 26(27):4410-4417. 22. Kerr KM, Johnson SK, King G, Kennedy MM, Weir J, Jeffrey R: Partial regression in primary carcinoma of the lung: does it occur? Histopathology 1998, 33(1):55-63. 23. Lee AH, Gillett CE, Ryder K, Fentiman IS, Miles DW, Millis RR: Different patterns of inflammation and prognosis in invasive carcinoma of the breast. Histopathology 2006, 48(6):692-701. 24. Marques LA, Franco EL, Torloni H, Brentani MM, da Silva-Neto JB, Brentani RR: Independent prognostic value of laminin receptor expression in breast cancer survival. Cancer Res 1990, 50(5):1479-1483. 25. Nixon AJ, Neuberg D, Hayes DF, Gelman R, Connolly JL, Schnitt S, Abner A, Recht A, Vicini F, Harris JR: Relationship of patient age to pathologic features of the tumor and prognosis for patients with stage I or II breast cancer. J Clin Oncol 1994, 12(5):888-894. 77  26. Rilke F, Colnaghi MI, Cascinelli N, Andreola S, Baldini MT, Bufalino R, Della Porta G, Menard S, Pierotti MA, Testori A: Prognostic significance of HER-2/neu expression in breast cancer and its relationship to other prognostic factors. Int J Cancer 1991, 49(1):44-49. 27. Rooney CM, Smith CA, Ng CY, Loftin SK, Sixbey JW, Gan Y, Srivastava DK, Bowman LC, Krance RA, Brenner MK, Heslop HE: Infusion of cytotoxic T cells for the prevention and treatment of Epstein-Barr virus-induced lymphoma in allogeneic transplant recipients. Blood 1998, 92(5):1549-1555. 28. Janeway C: Immunobiology : the immune system in health and disease: 6th ed. New York: Garland Science; 2005. 29. Begley J, Ribas A: Targeted therapies to improve tumor immunotherapy. Clin Cancer Res 2008, 14(14):4385-4391. 30. Boon T, Cerottini JC, den Eynde B, V., van der Bruggen P, Pel A, V.: Tumor antigens recognized by T lymphocytes. Ann Rev Immun 1994, 12:337-365. 31. Rosenberg SA, Yang JC, Restifo NP: Cancer immunotherapy: moving beyond current vaccines. Nat Med 2004, 10(9):909-915. 32. Smyth MJ, Godfrey DI, Trapani JA: A fresh look at tumor immunosurveillance and immunotherapy. Nat Immunol 2001, 2(4):293-299. 33. Dranoff G, Jaffee E, Lazenby A, Golumbek P, Levitsky H, Brose K, Jackson V, Hamada H, Pardoll D, Mulligan RC: Vaccination with irradiated tumor cells engineered to secrete murine granulocyte-macrophage colony-stimulating factor stimulates potent, specific, and long-lasting anti-tumor immunity. Proc Natl Acad Sci U S A 1993, 90(8):3539-3543. 34. Blattman JN, Greenberg PD: Cancer immunotherapy: a treatment for the masses. Science 2004, 305(5681):200-205. 35. Chambers SK, Kacinski BM, Ivins CM, Carcangiu ML: Overexpression of epithelial macrophage colony-stimulating factor (CSF-1) and CSF-1 receptor: a poor prognostic factor in epithelial ovarian cancer, contrasted with a protective effect of stromal CSF-1. Clin Cancer Res 1997, 3(6):999-1007. 36. Homey B, Muller A, Zlotnik A: Chemokines: agents for the immunotherapy of cancer? Nat Rev Immunol 2002, 2(3):175-184. 37. Pure E, Allison JP, Schreiber RD: Breaking down the barriers to cancer immunotherapy. Nat Immunol 2005, 6(12):1207-1210. 38. Blankenstein T, Rowley DA, Schreiber H: Cytokines and cancer: experimental systems. Curr Opin Immunol 1991, 3(5):694-698. 39. Schroder K, Hertzog PJ, Ravasi T, Hume DA: Interferon-gamma: an overview of signals, mechanisms and functions. J Leukoc Biol 2004, 75(2):163-189. 40. Dunn GP, Koebel CM, Schreiber RD: Interferons, immunity and cancer immunoediting. Nat Rev Immunol 2006, 6(11):836-848.  78  41. Ikeda H, Old LJ, Schreiber RD: The roles of IFN gamma in protection against tumor development and cancer immunoediting. Cytokine Growth Factor Rev 2002, 13(2):95-109. 42. Dunn GP, Bruce AT, Ikeda H, Old LJ, Schreiber RD: Cancer immunoediting: from immunosurveillance to tumor escape. Nat Immunol 2002, 3(11):991-998. 43. Wall EM, Milne K, Martin ML, Watson PH, Theiss P, Nelson BH: Spontaneous mammary tumors differ widely in their inherent sensitivity to adoptively transferred T cells. Cancer Res 2007, 67(13):6442-6450. 44. Figlin RA, Pierce WC, Kaboo R, Tso CL, Moldawer N, Gitlitz B, deKernion J, Belldegrun A: Treatment of metastatic renal cell carcinoma with nephrectomy, interleukin-2 and cytokine-primed or CD8(+) selected tumor infiltrating lymphocytes from primary tumor. J Urol 1997, 158(3 Pt 1):740-745. 45. Dudley ME, Wunderlich JR, Robbins PF, Yang JC, Hwu P, Schwartzentruber DJ, Topalian SL, Sherry R, Restifo NP, Hubicki AM, Robinson MR, Raffeld M, Duray P, Seipp CA, Rogers-Freezer L, Morton KE, Mavroukakis SA, White DE, Rosenberg SA: Cancer regression and autoimmunity in patients after clonal repopulation with antitumor lymphocytes. Science 2002, 298(5594):850-854. 46. Dudley ME, Wunderlich JR, Yang JC, Sherry RM, Topalian SL, Restifo NP, Royal RE, Kammula U, White DE, Mavroukakis SA, Rogers LJ, Gracia GJ, Jones SA, Mangiameli DP, Pelletier MM, Gea-Banacloche J, Robinson MR, Berman DM, Filie AC, Abati A, Rosenberg SA: Adoptive cell transfer therapy following non-myeloablative but lymphodepleting chemotherapy for the treatment of patients with refractory metastatic melanoma. J Clin Oncol 2005, 23(10):2346-2357. 47. Ioannides CG, Ioannides MG, O'Brian CA: T-cell recognition of oncogene products: a new strategy for immunotherapy. Mol Carcinog 1992, 6(2):77-82. 48. Ioannides C, G., Fisk B, Fan D, Biddison W, E., Wharton JT, O'brian C, A.: Cytotoxic T cells isolated from ovarian malignant ascites recognize a peptide derived from the HER-2/neu proto-oncogene. Cell Immun 1993, 151(1):225-234. 49. Disis ML, Calenoff E, McLaughlin G, Murphy AE, Chen W, Groner B, Jeschke M, Lydon N, McGlynn E, Livingston RB: Existent T-cell and antibody immunity to HER-2/neu protein in patients with breast cancer. Cancer Res 1994, 54(1):16-20. 50. Scanlan MJ, Jager D: Challenges to the development of antigen-specific breast cancer vaccines. Breast Cancer Res 2001, 3(2):95-98. 51. Peoples GE, Smith RC, Linehan DC, Yoshino I, Goedegebuure PS, Eberlein TJ: Shared T cell epitopes in epithelial tumors. Cell Immunol 1995, 164(2):279-286. 52. Fisk B, Blevins TL, Wharton JT, Ioannides CG: Identification of an immunodominant peptide of HER-2/neu protooncogene recognized by ovarian tumor-specific cytotoxic T lymphocyte lines. J Exp Med 1995, 181(6):2109-2117. 53. Wang RF, Wang X, Atwood AC, Topalian SL, Rosenberg SA: Cloning genes encoding MHC class II-restricted antigens: mutated CDC27 as a tumor antigen. Science 1999, 284(5418):1351-1354.  79  54. Rosenberg SA: Progress in human tumour immunology and immunotherapy. Nature 2001, 411(6835):380-384. 55. Ando-Akatsuka Y, Yonemura S, Itoh M, Furuse M, Tsukita S: Differential behavior of E-cadherin and occludin in their colocalization with ZO-1 during the establishment of epithelial cell polarity. J Cell Physiol 1999, 179(2):115-125. 56. Bazzoni G, Dejana E: Endothelial cell-to-cell junctions: molecular organization and role in vascular homeostasis. Physiol Rev 2004, 84(3):869-901. 57. Hori S, Nomura T, Sakaguchi S: Control of regulatory T cell development by the transcription factor Foxp3. Science 2003, 299(5609):1057-1061. 58. Parish CR: Cancer immunotherapy: the past, the present and the future. Immunol Cell Biol 2003, 81(2):106-113. 59. Yang T, Wall EM, Milne K, Theiss P, Watson P, Nelson BH: CD8+ T cells induce complete regression of advanced ovarian cancers by an interleukin (IL)-2/IL-15 dependent mechanism. Clin Cancer Res 2007, 13(23):7172-7180. 60. Hurwitz AA, Yu TF, Leach DR, Allison JP: CTLA-4 blockade synergizes with tumor-derived granulocyte-macrophage colony-stimulating factor for treatment of an experimental mammary carcinoma. Proc Natl Acad Sci U S A 1998, 95(17):1006710071. 61. van Elsas A, Hurwitz AA, Allison JP: Combination immunotherapy of B16 melanoma using anti-cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) and granulocyte/macrophage colony-stimulating factor (GM-CSF)-producing vaccines induces rejection of subcutaneous and metastatic tumors accompanied by autoimmune depigmentation. J Exp Med 1999, 190(3):355-366. 62. Sims AH, Howell A, Howell SJ, Clarke RB: Origins of breast cancer subtypes and therapeutic implications. Nat Clin Pract Oncol 2007, 4(9):516-525. 63. Teschendorff AE, Miremadi A, Pinder SE, Ellis IO, Caldas C: An immune response gene expression module identifies a good prognosis subtype in estrogen receptor negative breast cancer. Genome Biol 2007, 8(8):R157. 64. Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, Yang F, Talantov D, Timmermans M, Meijer-van Gelder ME, Yu J, Jatkoe T, Berns EM, Atkins D, Foekens JA: Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005, 365(9460):671-679. 65. Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H, Chen H, Omeroglu G, Meterissian S, Omeroglu A, Hallett M, Park M: Stromal gene expression predicts clinical outcome in breast cancer. Nat Med 2008, 14(5):518-527. 66. Ponzo MG, Lesurf R, Petkiewicz S, O'Malley FP, Pinnaduwage D, Andrulis IL, Bull SB, Chughtai N, Zuo D, Souleimanova M, Germain D, Omeroglu A, Cardiff RD, Hallett M, Park M: Met induces mammary tumors with diverse histologies and is associated with poor outcome and human basal breast cancer. Proc Natl Acad Sci U S A 2009, 106(31):12903-12908.  80  67. Huang S, Chen Y, Podsypanina K, Li Y: Comparison of expression profiles of metastatic versus primary mammary tumors in MMTV-Wnt-1 and MMTV-Neu transgenic mice. Neoplasia 2008, 10(2):118-124. 68. Worschech A, Kmieciak M, Knutson KL, Bear HD, Szalay AA, Wang E, Marincola FM, Manjili MH: Signatures associated with rejection or recurrence in HER-2/neupositive mammary tumors. Cancer Res 2008, 68(7):2436-2446. 69. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH: Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002, 415(6871):530-536. 70. Rouzier R, Perou CM, Symmans WF, Ibrahim N, Cristofanilli M, Anderson K, Hess KR, Stec J, Ayers M, Wagner P, Morandi P, Fan C, Rabiul I, Ross JS, Hortobagyi GN, Pusztai L: Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res 2005, 11(16):5678-5685. 71. Schauer M, Janssen KP, Rimkus C, Raggi M, Feith M, Friess H, Theisen J: Microarray-based response prediction in esophageal adenocarcinoma. Clin Cancer Res 2010, 16(1):330-337. 72. Farmer P, Bonnefoi H, Anderle P, Cameron D, Wirapati P, Becette V, Andre S, Piccart M, Campone M, Brain E, Macgrogan G, Petit T, Jassem J, Bibeau F, Blot E, Bogaerts J, Aguet M, Bergh J, Iggo R, Delorenzi M: A stroma-related gene signature predicts resistance to neoadjuvant chemotherapy in breast cancer. Nat Med 2009, 15(1):68-74. 73. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO, Botstein D: Molecular portraits of human breast tumours. Nature 2000, 406(6797):747-752. 74. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, Eisen MB, van de Rijn M, Jeffrey SS, Thorsen T, Quist H, Matese JC, Brown PO, Botstein D, Eystein Lonning P, Borresen-Dale AL: Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 2001, 98(19):10869-10874. 75. Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, Deng S, Johnsen H, Pesich R, Geisler S, Demeter J, Perou CM, Lonning PE, Brown PO, Borresen-Dale AL, Botstein D: Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A 2003, 100(14):8418-8423. 76. Charafe-Jauffret E, Ginestier C, Monville F, Finetti P, Adelaide J, Cervera N, Fekairi S, Xerri L, Jacquemier J, Birnbaum D, Bertucci F: Gene expression profiling of breast cell lines identifies potential new basal markers. Oncogene 2006, 25(15):2273-2284. 77. Sotiriou C, Neo SY, McShane LM, Korn EL, Long PM, Jazaeri A, Martiat P, Fox SB, Harris AL, Liu ET: Breast cancer classification and prognosis based on gene expression profiles from a population-based study. Proc Natl Acad Sci U S A 2003, 100(18):10393-10398.  81  78. Sorlie T, Wang Y, Xiao C, Johnsen H, Naume B, Samaha RR, Borresen-Dale AL: Distinct molecular mechanisms underlying clinically relevant subtypes of breast cancer: gene expression analyses across three different platforms. BMC Genomics 2006, 7:127. 79. Hu Z, Fan C, Oh DS, Marron JS, He X, Qaqish BF, Livasy C, Carey LA, Reynolds E, Dressler L, Nobel A, Parker J, Ewend MG, Sawyer LR, Wu J, Liu Y, Nanda R, Tretiakova M, Ruiz Orrico A, Dreher D, Palazzo JP, Perreard L, Nelson E, Mone M, Hansen H, Mullins M, Quackenbush JF, Ellis MJ, Olopade OI, Bernard PS, Perou CM: The molecular portraits of breast tumors are conserved across microarray platforms. BMC Genomics 2006, 7:96. 80. Roylance R, Gorman P, Harris W, Liebmann R, Barnes D, Hanby A, Sheer D: Comparative genomic hybridization of breast tumors stratified by histological grade reveals new insights into the biological progression of breast cancer. Cancer Res 1999, 59(7):1433-1436. 81. Martin ML, Wall EM, Sandwith E, Girardin A, Milne K, Watson PH, Nelson BH: Density of tumour stroma is correlated to outcome after adoptive transfer of CD4(+) and CD8 (+) T cells in a murine mammary carcinoma model. Breast Cancer Res Treat 2009, . 82. Yang T, Martin ML, Nielsen JS, Milne K, Wall EM, Lin W, Watson PH, Nelson BH: Mammary tumors with diverse immunological phenotypes show differing sensitivity to adoptively transferred CD8+ T cells lacking the Cbl-b gene. Cancer Immunol Immunother 2009, 58(11):1865-1875. 83. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP: Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 2003, 31(4):e15. 84. Johnson WE, Li C, Rabinovic A: Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 2007, 8(1):118-127. 85. Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001, 98(9):5116-5121. 86. Jave Cluster 3.0 and Treeview software [http://bonsai.ims.utokyo.ac.jp/~mdehoon/software/cluster/software.htm] 87. Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998, 95(25):1486314868. 88. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, Lerner J, Brunet JP, Subramanian A, Ross KN, Reich M, Hieronymus H, Wei G, Armstrong SA, Haggarty SJ, Clemons PA, Wei R, Carr SA, Lander ES, Golub TR: The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 2006, 313(5795):1929-1935. 89. Tibshirani R, Hastie T, Narasimhan B, Chu G: Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci U S A 2002, 99(10):6567-6572.  82  90. Ingenuity Systems [www.ingenuity.com] 91. Dennis G,Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA: DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 2003, 4(5):P3. 92. BioCarta [http://www.biocarta.com/] 93. Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M: KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res 1999, 27(1):29-34. 94. Benito M, Parker J, Du Q, Wu J, Xiang D, Perou CM, Marron JS: Adjustment of systematic microarray data biases. Bioinformatics 2004, 20(1):105-114. 95. Smedley D, Haider S, Ballester B, Holland R, London D, Thorisson G, Kasprzyk A: BioMart--biological queries made easy. BMC Genomics 2009, 10:22. 96. Herschkowitz JI, Simin K, Weigman VJ, Mikaelian I, Usary J, Hu Z, Rasmussen KE, Jones LP, Assefnia S, Chandrasekharan S, Backlund MG, Yin Y, Khramtsov AI, Bastein R, Quackenbush J, Glazer RI, Brown PH, Green JE, Kopelovich L, Furth PA, Palazzo JP, Olopade OI, Bernard PS, Churchill GA, Van Dyke T, Perou CM: Identification of conserved gene expression features between murine mammary carcinoma models and human breast tumors. Genome Biol 2007, 8(5):R76. 97. Suzuki R, Shimodaira H: Pvclust: an R package for assessing the uncertainty in hierarchical clustering. Bioinformatics 2006, 22(12):1540-1542. 98. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP: Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005, 102(43):15545-15550. 99. Tian L, Greenberg SA, Kong SW, Altschuler J, Kohane IS, Park PJ: Discovering statistically significant pathways in expression profiling studies. Proc Natl Acad Sci U S A 2005, 102(38):13544-13549. 100. Brenton JD, Carey LA, Ahmed AA, Caldas C: Molecular classification and molecular forecasting of breast cancer: ready for clinical application? J Clin Oncol 2005, 23(29):7350-7360. 101. Huber LC, Distler JH, Moritz F, Hemmatazad H, Hauser T, Michel BA, Gay RE, Matucci-Cerinic M, Gay S, Distler O, Jungel A: Trichostatin A prevents the accumulation of extracellular matrix in a mouse model of bleomycin-induced skin fibrosis. Arthritis Rheum 2007, 56(8):2755-2764. 102. Marks PA, Rifkind RA, Richon VM, Breslow R, Miller T, Kelly WK: Histone deacetylases and cancer: causes and therapies. Nat Rev Cancer 2001, 1:194-202. 103. Marks PA, Richon VM, Breslow R, Rifkind RA: Histone deacetylase inhibitors as new cancer drugs. Curr Opin Oncol 2001, 13(6):477-483. 104. Marks PA, Richon VM, Rifkind RA: Histone deacetylase inhibitors: inducers of differentiation or apoptosis of transformed cells. J Natl Cancer Inst 2000, 92(15):1210-1216.  83  105. Rombouts K, Niki T, Wielant A, Hellemans K, Geerts A: Trichostatin A, lead compound for development of antifibrogenic drugs. Acta Gastroenterol Belg 2001, 64(3):239-246. 106. Maio M, Coral S, Fratta E, Altomonte M, Sigalotti L: Epigenetic targets for immune intervention in human malignancies. Oncogene 2003, 22(42):6484-6488. 107. Vigushin DM, Ali S, Pace PE, Mirsaidi N, Ito K, Adcock I, Coombes RC: Trichostatin A is a histone deacetylase inhibitor with potent antitumor activity against breast cancer in vivo. Clin Cancer Res 2001, 7(4):971-976. 108. Buglio D, Georgakis GV, Hanabuchi S, Arima K, Khaskhely NM, Liu YJ, Younes A: Vorinostat inhibits STAT6-mediated TH2 cytokine and TARC production and induces cell death in Hodgkin lymphoma cell lines. Blood 2008, 112(4):1424-1433. 109. Medina V, Edmonds B, Young GP, James R, Appleton S, Zalewski PD: Induction of caspase-3 protease activity and apoptosis by butyrate and trichostatin A (inhibitors of histone deacetylase): dependence on protein synthesis and synergy with a mitochondrial/cytochrome c-dependent pathway. Cancer Res 1997, 57(17):3697-3707. 110. McBain JA, Eastman A, Nobel CS, Mueller GC: Apoptotic death in adenocarcinoma cell lines induced by butyrate and other histone deacetylase inhibitors. Biochem Pharmacol 1997, 53(9):1357-1368. 111. Zampetaki A, Zeng L, Xiao Q, Margariti A, Hu Y, Xu Q: Lacking cytokine production in ES cells and ES-cell-derived vascular cells stimulated by TNF-alpha is rescued by HDAC inhibitor trichostatin A. Am J Physiol Cell Physiol 2007, 293(4):C1226-38. 112. Takahashi I, Miyaji H, Yoshida T, Sato S, Mizukami T: Selective inhibition of IL-2 gene expression by trichostatin A, a potent inhibitor of mammalian histone deacetylase. J Antibiot (Tokyo) 1996, 49(5):453-457. 113. Vanniasinkam T, Ertl H, Tang Q: Trichostatin-A enhances adaptive immune responses to DNA vaccination. J Clin Virol 2006, 36(4):292-297. 114. Vo DD, Prins RM, Begley JL, Donahue TR, Morris LF, Bruhn KW, de la Rocha P, Yang MY, Mok S, Garban HJ, Craft N, Economou JS, Marincola FM, Wang E, Ribas A: Enhanced antitumor activity induced by adoptive T-cell transfer and adjunctive use of the histone deacetylase inhibitor LAQ824. Cancer Res 2009, 69(22):8693-8699. 115. Gelderman KA, Tomlinson S, Ross GD, Gorter A: Complement function in mAbmediated cancer immunotherapy. Trends Immunol 2004, 25(3):158-164. 116. Markiewski MM, DeAngelis RA, Benencia F, Ricklin-Lichtsteiner SK, Koutoulaki A, Gerard C, Coukos G, Lambris JD: Modulation of the antitumor immune response by complement. Nat Immunol 2008, 9(11):1225-1235. 117. Ravetch JV, Clynes RA: Divergent roles for Fc receptors and complement in vivo. Annu Rev Immunol 1998, 16:421-432.  84  118. Cartron G, Dacheux L, Salles G, Solal-Celigny P, Bardos P, Colombat P, Watier H: Therapeutic activity of humanized anti-CD20 monoclonal antibody and polymorphism in IgG Fc receptor FcgammaRIIIa gene. Blood 2002, 99(3):754-758. 119. Clynes RA, Towers TL, Presta LG, Ravetch JV: Inhibitory Fc receptors modulate in vivo cytoxicity against tumor targets. Nat Med 2000, 6(4):443-446. 120. Tsuboi A, Muramatsu M, Tsutsumi A, Arai K, Arai N: Calcineurin activates transcription from the GM-CSF promoter in synergy with either protein kinase C or NF-kappa B/AP-1 in T cells. Biochem Biophys Res Commun 1994, 199(2):10641072. 121. Greten TF, Jaffee EM: Cancer vaccines. Curr Protoc Hum Genet 2001, Chapter 13:Unit 13.8. 122. Carey LA, Perou CM, Livasy CA, Dressler LG, Cowan D, Conway K, Karaca G, Troester MA, Tse CK, Edmiston S, Deming SL, Geradts J, Cheang MC, Nielsen TO, Moorman PG, Earp HS, Millikan RC: Race, breast cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA 2006, 295(21):2492-2502. 123. Setiadi AF, Omilusik K, David MD, Seipp RP, Hartikainen J, Gopaul R, Choi KB, Jefferies WA: Epigenetic enhancement of antigen processing and presentation promotes immune recognition of tumors. Cancer Res 2008, 68(23):9601-9607.  85  3 CHARACTERIZATION OF microRNA PROFILES AND POTENTIAL POST-TRANSCRIPTIONAL REGULATIONS IN GENE EXPRESSION OF BREAST TUMOURS WITH VARYING OUTCOMES AFTER ADOPTIVE T-CELL THERAPY  3.1 INTRODUCTION MicroRNAs (miRNAs) are non-coding RNAs of 18-25 nucleotides in length that regulate gene expression at the post-transcriptional level. Since the first discovery of lin-4 in Caenorhabditis elegans in 1993 [1], miRNAs have been found in every organism investigated to date, including mammals, plants, worms, insects and viruses, with surprisingly high evolutionary conservation across these organisms, which suggest the functional importance of miRNAs [1-3]. Indeed, miRNAs are found to be involved in many vital processes such as growth, development, homeostasis, cellular differentiation and apoptosis [3-7]. MiRNAs originate from long endogenous single stranded RNAs of several kilobases in length, known as primary miRNA transcripts (pri-miRNAs), which are transcribed from a miRNA gene by RNA Polymerase II (Pol II) in the nucleus. The majority of miRNA genes are located within introns of other genes that encode proteins or other functional RNAs. Some miRNA genes are found in isolated genomic regions devoid of genes, or within exons or 3’ untranslated regions (UTRs) of protein-coding genes [3, 6]. The primiRNA transcript has a 5’ 7-methylguanosine cap and 3’ poly-adenylated tail, and local stem-loop structures generated from imperfect base pairing. The stem-loop structures are recognized by a nuclear enzyme complex that includes RNAse III enzymes, Drosha and 86  Pasha (also known as DGCR8), and processed into precursor RNAs (pre-miRNA) of 60110 nucleotides in length that fold into imperfect stem-loop structures. The pre-miRNAs are subsequently transported to the cytoplasm where they are cleaved by another RNAse III enzyme, Dicer, into 18-25 nucleotide-long transient double-stranded RNA fragments. One of the two strands will become the active mature miRNA when it is processed and selectively incorporated into the miRNA-associated multi-protein RNA-induced silencing complex (miRISC), which includes the Argonaute proteins. Finally, the miRISC complex regulates target gene expression by binding to UTRs of the target mRNA transcripts through imperfect complementarity to the mature miRNA incorporated in the complex [3, 6-8]. This process is different from gene silencing by small interfering RNA (siRNAs) through RNA-mediated interference (RNAi). This process occurs mostly in plants where RISC induces the cleavage of the target mRNA transcript after binding with the perfect or nearly perfect complementarity (as reviewed in [3, 9] ). However, miRNAmediated gene silencing does not always involve the cleavage of target mRNA transcripts. Except in only one case of hsa-mir-369-3 discovered so far [10], the binding of miRNA::miRISC complex to its target ultimately leads to the down-regulation of protein expression caused by translational repression via blocking of translation initiation and/or ribosome association, mRNA cleavage mediated by RNA-induced silencing complex (RISC) or mRNA decay [11, 12]. However, the exact mechanism of miRNAmediated protein down-regulation remains unknown. It is estimated that approximately one third of protein-coding genes in human genome have conserved miRNA target sequences [13], suggesting that these genes are under miRNA control. It is currently estimated that there are 255 to 1,000 miRNAs in humans  87  [13, 14]. This comprises of 1-4% of all human genes, making miRNAs one of the largest classes of gene expression regulators [15, 16]. One goal of miRNA research is to identify target genes of miRNAs and their effect on overall physiological processes. The current miRNA target prediction programs such as TargetScan[17], PicTar[18], and miRANDA[19], emphasize the strong sequence homology from comparative genomics analyses to guide the target prediction found between the 5’ end of mature miRNA sequences and the 3’ UTRs of protein-coding transcripts. Especially, the high sequence conservation observed in the first 7-8 nucleotides in mature miRNAs, also known as the ‘miRNA seed’ or the ‘miRNA nucleus’, which are also used to group miRNAs into families, suggests that the conserved motifs found at the 5’ end of miRNAs are implicated with biological significance. Indeed, studies have found that the 5’ end of miRNAs is important for target recognition [6], and the stability and proper loading of miRNA into the miRISC complex [20, 21]. Experimentally verified miRNA targets indicate that the 5’ end of miRNAs have less wobble base pairing of guanine and uracil [22-24] and have a higher sequence complementarity to the target sites compared to the 3’ end [19, 25-27] with occasional mismatches [26, 28]. Another factor that influences miRNA target binding and subsequent translational repression is the free energy of binding between the first 8 nucleotides in the 5’ end of miRNA, suggested by Doench and Sharp [29]. Hence, the current prediction algorithms unanimously focus on the sequence complementarity between the first 7-8 nucleotides of the 5’ end of miRNA and the 3’ UTRs of mRNA transcripts, which are highly conserved across different species. For example, in TargetScan, the comparative genomics analysis is performed for human, mouse, rat, fugu  88  genomes; in miRANDA, the zebrafish genome is added; in PicTar, human, mouse, dog, chicken, fugu genomes are compared. Once the orthologous sequences in the 3’ UTR are identified, the alignments from miRNA seeds are extended allowing various rules reflecting the biological particularities pertaining to RNA binding such as G:U wobble base pairing in TargetScan, the position-specific constraints in miRANDA and the competitive binding of miRNAs represented in a probabilistic model in PicTar. After the candidate targets in 3’ UTRs are found, the folding free energy is calculated to estimate the likelihood of interaction between a miRNA and its target sequence. However, there are more complexities involved with target recognition and binding than just the imperfect base-pairing between the 5’ ends of miRNAs and the 3’ UTRs of mRNA transcripts, which imposes limitations in the current prediction algorithms to accurately identify potential target genes of miRNAs. It is estimated that with the current prediction algorithms, about one third of predicted miRNA targets in human, mouse and rat are false positives [30]. Several reviews [3, 31, 32] have pointed out that limiting the search space to 3’ UTR results in an incomplete target gene pool as the increasing evidence shows the importance of the rest of miRNA sequences in binding of target genes. The 5’ UTRs [33, 34] as well as the open reading frames (ORF) of the mRNA transcripts [13, 35-40] were found to be also important in guiding the target binding of miRNA. Vella and colleagues showed a decrease in the down-regulation of target gene expression when they introduced functional mismatch mutations at the 3’ end of the miRNA [27] and that both 5’ and 3’ ends of the miRNA were required to down-regulate the target gene [41]. In the case of let-7, the entire sequence of the mature miRNA was found to be highly conserved across different phyla from worm to humans, suggesting the  89  functional importance of other sequences in the mature miRNA [42]. The binding site in miRNA is not restricted to the conserved regions either. Farh and colleagues found that non-conserved sites also mediate repression [35]. Further evidence suggests that the efficient binding of miRNA to its target genes is dependent on the cooperation of multiple miRNA seeds [19, 35, 43], the spatial distribution of miRNA seeds [44], positions and sizes of loops and nucleotide bulges in miRNA::mRNA (target) complex [30], and the secondary structure of target mRNA transcripts [45], all of which are not incorporated in the current prediction algorithms. Nonetheless, miRNA-mediated translational repression is inarguably shown by the reduced protein production in cells observed in the experimental validation of miRNA target genes, albeit the degree of protein reduction is found to be mostly subtle [38, 46, 47]. Large-scale proteomic analyses that directly measured the overall protein expression of miRNA target genes showed the small reduction with less than 30% in target proteins [38, 47]. Using quantitative-mass-spectrometry, Baek and colleagues noted that substantial translational repression by a miRNA was occasionally observed for surprisingly few inferred target genes, and that such dramatic repression is rarely detected due to the subtle changes in overall protein levels. It is likely that the real potency of miRNA-mediated regulation comes not from substantial reduction in a handful of key proteins, but the collective effect of the subtle decreases in the protein concentrations expressed from multiple genes with diverse biological functions [12, 38, 47]. Lewis and colleagues [17] estimated that a single miRNA targets approximately 200 genes. Similarly, a study using microarray methodology to detect the change in gene expression also found that the amplification of  90  a single miRNA results in down-regulation of approximately 100 genes [36]. The largescale proteomic study by Selbach and colleagues showed the protein output reduction over an average of 648 genes from a single miRNA change. Notably, these predicted target genes substantially vary in biological functions, including transcription factors, secreted factors, receptors and transporters [18, 19, 30, 36, 48-50]. Thus, the profound physiological effects from miRNA-mediated gene regulation likely come from the precise orchestration of subtle adjustments of the final protein output [12, 38]. Consequently, a slight perturbation in miRNA expression can result in severe abnormalities in an extensive range of cellular functions observed in many studies. MiRNAs affect a wide range of cellular processes including development [1, 22, 25, 26, 28, 51], cellular differentiation [3, 7, 52], metabolic pathways [53] and disease [3, 5, 6, 54, 55]. The effect of miRNAs is not only pronounced at the cellular level, but also at the tissue and organismal levels. The comparative analysis of primate, rodents and dog genomes by Bentwich and colleagues revealed [15] primate-specific miRNA clusters that were expressed specifically in placenta, and also had high sequence similarities to the mir-371, 372, 373 family, which was exclusively expressed in human embryonic stem cells [56]. Disruption of miRNA activities from loss-of-function mutations in Dicer or Argonaute proteins results in extensive abnormalities in various developmental processes in organisms across different phyla (reviewed in [51, 57]), such as abnormal embryogenesis and defective stem cell maintenance in Arabidopsis [58-60]; delayed germ-line stem-cell division in Drosophila [61]; germ-line defects and abnormality in early development in C. elegans [62-64]; abnormal embryonic morphogenesis in zebrafish [65, 66]; and  91  defects in stem-cell differention in mice [67]. Several miRNAs were experimentally verified to cause abnormalities in developmental stages. For example, mutations in the active binding site of lsy-6 disrupt the proper left/right asymmetrical expression patterns of chemoreceptors in the nervous system in C. elegans [26]. Similarly, the ablation of lin-4 and let-7 activities results in abnormal developmental timing in C. elegans [22, 28]. One of the developmental processes largely impacted by miRNA-mediated gene regulation is the immune system. Several miRNAs (miR-181a, miR-221/222, miR-223) are identified to be key regulators in pluripotent hematopoietic stem cells, which differentiate into various types of blood cells [8, 68]. Others such as miR-150, miR-181a, miR-17-92 influence T- and B-cell differentiation [68]; the activation of innate and adoptive immune response (miR-155) [69, 70]; T- and B-cell sensitivity (miR-155, miR181a) [68, 71] or the acute inflammatory response after the recognition of pathogens (miR-16, miR-146, miR-155, miR-223) [68, 69, 72]. Several miRNAs (miR-122, miR196, miR-296, miR-351, miR-431, miR-448) were also found to directly exert an anti-viral effect against hepatitis C virus where the introduction of the anti-sense strands of these miRNAs resulted in a substantial reduction of viral replication [73, 74]. Conversely, miRNAs can also mediate immune evasion as shown by Stern-Ginossar and colleagues, where the expression of a miRNA from human cytomegalovirus (hcmv-miR-UL112) down-regulated the expression of a host gene, MICB, which was crucial for the activation of natural killer cells [75]. MiRNAs also play a significant role in cellular proliferation, apoptosis and differentiation, which are the hallmark pathways deregulated in cancer [3, 6-8, 76-78]. It is currently understood that the role of miRNA in tumourigenesis is to cause aberrant  92  gene expression of target genes, observed from the abnormal level of miRNAs present in cancer cells compared to normal cells [79]. More than 50% of known miRNA genes are located inside or near the chromosomal regions known as fragile sites, which are common breakpoints or minimal regions of loss of heterozygosity or amplification associated with cancers [8, 80]. This observation initially suggested the role of miRNAs in cancer pathogenesis in human [8, 80]. Mir125b-1, for example, is located in the fragile site on chromosome 11q24, known to be frequently deleted in breast, lung, ovarian and cervical cancers [80]. In congruence, a study by Iorio and colleagues found that the same miRNA was one of the most differentially expressed miRNAs in breast cancer [81]. Other studies found that amplification of mir-125b resulted in increased cell proliferation in differentiated cell lines and leukemia cell lines [82, 83], supporting the implication of this miRNA in cancer pathogenesis. Numerous miRNAs were experimentally verified to have deregulation in cancer shown by changes in miRNA gene expression level, which suggested their roles as tumour suppressors or oncogenes. The let-7 miRNA family was significantly underexpressed in 44% of lung cancer patients [77, 84, 85]. A study later revealed that let-7 was a potent tumour suppressor that repressed the expression of RAS protein in human [86], which is one of the most well-known oncogenes involved in cell signaling pathways disrupted in cancer [87]. Conversely, miRNAs can act as oncogenes by repressing tumour suppressor genes or genes that regulate tumour oncogenes [6]. The miRNA, mir155, is linked with increased protein expression of the MYC oncogene, a transcription factor that can induce both cell proliferation and apoptosis and is often mutated or amplified in many types of human cancers [6, 88]. An increased level of mir-155 is  93  found in human B-cell lymphomas [6, 89], especially Burkitt’s lymphoma [90], primary mediastinal, diffuse large B cell and Hodgkin’s lymphomas [91]. A transgenic mouse strain that over-expressed the miR-155 gene rapidly developed B-cell malignancy preceded by the increased proliferation of precursor B-cells [92]. The miRNA cluster miR-17-92, found in a genomic locus often amplified in several types of cancer including B-cell lymphomas, breast, lung, colon and thyroid cancers [6] also affects MYC protein expression, which resulted in the acceleration of tumour growth in a B-cell lymphoma model in mice [93]. In the wake of the genomics era, the miRNA abundance could be globally profiled using various high- throughput technologies such as microarrays, bead-based flow cytometric profiling and cDNA sequencing. The systemic pattern of miRNA expression revealed powerful biomarkers for cancer classification and prognosis, and novel therapeutic targets. As reviewed by Calin and Croce, virtually all types of tumours investigated to date have shown significant differences in miRNA expression profiles compared to normal cells of the same tissue type [79]. For example, the signatures from miRNA expression generated from microarrays of 10 normal tissues and 76 breast cancer tissues showed a clear separation between the two sample types by hierarchical clustering analysis [81]. Using bead-based flow cytometric methodology, Lu and colleagues found that miRNAs were systematically under-expressed in various types of human cancer compared to normal cells regardless of the originating tissues [94]. The authors also found that the overall increase in miRNA levels correlated with normal cellular differentiation, whereas the loss of miRNAs was associated with less differentiated tumours [94]. They also showed that a subset of differentially expressed miRNAs was a  94  powerful classifier that distinguished tumours that were histopathologically poorly differentiated, thus making the miRNA profile a useful tool to clarify diagnostic uncertainty [94]. Their miRNA-based classifier was much more accurate in classifying tumours with uncertain cellular origins with fewer number of miRNA genes compared to the mRNA-based classifier found using the same dataset [94]. The authors also suggested that the increased accuracy in classifying from miRNA signatures, which are derived from much more limited biological information than mRNA signatures (i.e. microarrays that probed 217 miRNAs compared to approximately 16,000 mRNAs), argues for the ultimate importance of miRNAs in normal cellular development. In another study using microarrays by Blenkiron and colleagues, miRNAs signatures were able to distinguish between the basal and the luminal subtypes of human breast cancer [95]. The potential usage of miRNAs as a diagnostic tool is further suggested by the reproducible miRNA levels found in serum [96]. Interestingly, distinctive expression patterns of miRNAs in serum and plasma were identified from healthy individuals compared to patients with different diseases including lung, colorectal cancer and diabetes [96]. The use of serum miRNA profiles hold promise for a non-invasive diagnostic tool for these diseases, which is readily applicable in clinical settings. The prognostic power of miRNA signatures in cancer has also been gaining more recognition from recent discoveries. From analyzing microarray data generated from 94 chronic lymphocytic leukemia samples, Calin and colleagues found 13 miRNAs whose expression levels were significantly correlated with disease aggressiveness [97]. Strikingly, the authors found that the expression level of a single miRNA dictated  95  outcome. Specifically a low level of let-7 strongly correlated with reduced survival in lung cancer patients [77, 84, 85]. Similarly in a colon adenocarcinoma study, overexpression of miR-21 was strongly correlated with poor prognosis [98]. Recently, a novel role of miRNA was discovered as an intercellular communicator through a heterotypically secreted signal [99, 100]. Valadi and colleagues found that RNAs in exosomes, small membrane vesicles of endocytic origin that are secreted to the extracellular environment, are transferable to other cells in mice and human [100]. The RNAs include miRNAs and mRNAs. The transferred mRNAs were found to be functional and were translated into proteins [100], which also supported the potential of functional miRNAs in exosomes as a novel cell-to-cell communicator. These secreted miRNAs are presumed to enter other cells to initiate their translationally repressive activities. The discovery of a hexanucleotide terminal motif in miR-29b that directed nuclear localization [101] opened the further possibility of extracellular miRNAs to be imported into the nucleus and regulate transcription or splicing of target transcripts [101]. In a tantalizing study by Yu and colleagues, the conditioned medium from a human breast cancer cell line that over-expressed miR-17/20, a miRNA that suppressed tumour growth by repressing cyclin D1 gene expression, resulted in a significant reduction in the migration and invasion of aggressive human breast cancer cells cultured in it. Even more surprising was that miR-17/20 in the medium altered the cellular microenvironment that resulted in reduced cell migration and invasion, implicating the role of miRNAs in heterotypic signaling [99]. The potential to use miRNAs as therapeutic agents in cancer patients is currently an active area of research. The idea of miRNA therapeutics is to exogenously introduce  96  corrective synthetic miRNAs in the form of short-interfering RNA (siRNA)-like double stranded RNA (dsRNA) in the case of a deletion or a loss-of-function mutation in miRNA, or inhibit the function of over-expressed miRNAs by introducing synthetic single-strand RNA molecules that are complementary to pre- or mature miRNAs [43, 79, 102, 103]. The development of chemical modification to nucleotides to increase the binding affinity to RNA molecules is already established, such as the addition of 2’-Omethyl group (OMe), 2’-O-Methoxyethyl group (MOE) or chemically modified cholesterol to the 2’-hydroxyl group of the nucleotide [102]. The introduction of OMeoligonucleotides resulted in a potent and irreversible inhibition of the specific target let-7 in C. elegans, Drosophila and human cervical cancer HeLa cell line [104]. Cholesterolconjugated synthetic RNA molecules known as ‘antagomirs’ were also successful in selectively inhibiting miRNA activity by intravenous injection in mice [105]. Anti-sense miRNA-based therapy using OMe/MOE-oligonucleotides was carried out in a clinical trial for the treatment of solid tumours including lung, colorectal and gynecologic cancers, which further highlighted the potential of miRNA as a therapeutic agent [102, 106, 107]. The previously described mouse model of breast cancer [108] that showed varying outcomes after adoptively transferred T-cell therapy provides an unprecedented opportunity to study the intricate, relationships between miRNAs, cancer and the immune system. The transgenic mouse model that over-expressed Her2/neu tagged with antigenic epitopes developed spontaneously arising mammary tumours which exhibited a complete range of outcome after adoptive T-cell therapy: complete regression (CR); partial regression (PR); stable disease (SD) and progressive disease (PD) [108]. When cell lines  97  derived from these tumours were implanted in the mammary fat pads of host mice, the resulting tumours demonstrated reproducible responses of CR, PR, SD and PD, providing a unique experimental system to investigate the immune response against tumours following adoptive T-cell therapy [108]. As described in the previous microarray study, thousands of genes were found to show different transcriptional activities that led to potential pathways responsible for differential responses to immunotherapy. However, the fine adjustments by miRNAs at the last stage of gene expression i.e. protein translation, which simultaneously affect multiple genes, can further elucidate the mechanisms that underlie the varied immunotherapy response at a much more refined level. For example, virtually identical transcriptome landscapes between CR and PR tumours, manifested by few genes (N = 6) with differential gene expressions, may have very different proteomic landscapes that are ultimately responsible for the clear differences in immunotherapy outcomes. MiRNAmediated post-transcriptional regulation could be the key modifier that results in such differences. Furthermore, approximately 10% of the genes were differentially expressed among CR, PR and PD from the previously described microarray study, which represent one dimension of the complex genomic orchestration that affect the immune response. Given the extensive translational regulation of the target mRNA transcripts by miRNAs described above, approximately one third of the rest of the genes that did not show differential expression could potentially be under miRNA control, which can elucidate a broader spectrum of mechanisms behind tumour immunogenicity. In the present study, small RNA fragments isolated from CR, PR and PD tumours were sequenced using the latest massively parallel sequencing technology known as SOLiD  98  (Applied Biosystems, USA). Compared to the probe-based methodology, the highthroughput sequencing of miRNAs allowed unbiased profiling of miRNAs, not limited to a priori knowledge of known miRNA sequences, enabling the discovery of novel miRNAs and other small RNAs in the samples. Deep sequencing also allowed the profiling of rarely expressed miRNAs. Furthermore, the sequence counts can directly translate into the absolute abundance of a particular miRNA at a finer resolution of detection. Another benefit of the sequencing-based profiling of miRNAs (not explored in this study) is the identification of possible RNA editing or mutations in the sequence that may affect the function of the miRNA. Several miRNAs with significantly different expression profiles were identified amongst CR, PR and PD tumours. The gene expression levels of predicted miRNA target genes based on the microarray data did not show significant difference, which could be due to the subtle translational repression and/or the lack of accuracy in predicting the correct target genes.  3.2 MATERIAL AND METHODS 3.2.1 Murine tumour samples and small RNA library preparation Tumour samples were generated as described in the previous microarray study. Tumours of approximately 50 mm2 were grown from subcutaneously implanted tumour cell lines in mammary fat pads of transgenic mice described by Wall et al. [108] and then snap frozen immediately upon harvest for the miRNA sequencing experiment. For miRNA profiling, tumours derived from NOP21CR, NOP23PR, and NOP18PD cell lines were used,  99  which also had been studied in the previous microarray experiment. Additionally, a tumour generated from the NOP12PR cell line, which was not part of the microarray experiment, was included for miRNA sequencing. The samples were prepared for small RNA sequencing using the SOLiD sequencing technology (Applied Biosystems, USA). 10µg of total RNA extracted from NOP21CR, NOP12PR, NOP23PR, and NOP18PD tumours was treated with DNAseI to produce high quality RNA, indicated by the RNA integrity (RI) number above 7. Using the FlashPAGE Fractionator (Ambion, USA), RNAs were enriched for single-stranded small RNAfragments before the SOLiD sequencing libraries were constructed. The SOLiD sequencing libraries were constructed following the small RNA expression kit protocol provided by Applied Biosystems, USA. In brief, purified small RNAs were hybridized and ligated to a mix of adaptor sequences from Applied Biosystems, and then reversed transcribed into cDNA. cDNA templates were amplified by 15 cycles of PCR, and then size selected at 105-150bp after the polyacrylamide gel electrophoresis. The library construction protocol used for NOP21CR and NOP23PR tumours was modified from the original protocol that was used to construct libraries for NOP12PR and NOP18PD tumours. In the modified protocol, the samples were left to hybridize to adaptor sequences overnight compared to two hours in the original protocol. An additional step was added to purify the initial cDNA template before and after the PCR amplification using the Qiagen Minelute PCR Purification Kit (Quiagen, USA) and Invitrogen Purelink Micro PCR Purification Kit (Invitrogen, USA). Also, an in-gel PCR amplification on the cDNA template was done after the PAGE size selection in the new libraries as opposed to raw cDNA templates cleaned from the gel in the original protocol.  100  3.2.2 Massively parallel sequencing of small RNAs by ABI SOLiDTM The cDNA library was amplified by emulsion PCR in which individual oil droplets that were suspended in liquid contained micro-reactors, which contained the cDNA template, PCR reaction enzymes, beads and primers. The concentration of the beads relative to the cDNA template was prepared such that an individual micro-PCR reactor contained a single template to clonally amplify the cDNA template in a massively parallel fashion. The beads enriched for successful template amplification were purified, then deposited onto a glass slide to perform sequencing by ligation of primer probes labeled with one of four fluorescent dyes that represented a specific pair of two adjacent nucleotides. The sequencing was done in 7 ligation cycles with probes interrogating from five different starting bases in a template, resulting in amplicon sequences of 35bp in length. The fluorescent signals were detected during each cycle and converted to corresponding numbers ranging from 0 to 3 in the final sequence files known as colour space fasta files.  3.2.3 Genome mapping and annotation of small RNA sequences The colour space sequences were mapped to the genome of C57BL/6J Mus musculus strain (UCSC version mm9/NCBI version 37) using the alignment programs Maq [109] and Short Read Mapping Package (SHRiMP, version 1.3.1) [110]. Maq searches for ungapped matches in the initial seed of 24bp in length, allowing up to two mismatches in the alignment. Sequences that exceeded the threshold of the maximum mismatch score were calculated for the probability of random alignments to the genome and given a  101  Phred-like mapping score, for which values greater than 10 indicated a statistically significant match to a unique location in a genome. SHRiMP performs vectorized SmithWaterman Algorithm to rapidly identify locations where sequences are likely to map, and then performs the thorough Smith-Waterman local alignment for more accurate final alignments [110]. SHRiMP consists of two programs, rmapper and probcalc, which perform the sequence alignment to the genome and then compute the probability of an alignment created by chance respectively. The default parameters were used for rmapper and probcalc programs except in the rmapper program to increase the sensitivity of mapping short reads (35bp including adapter sequences) to the genome. At least two seeds need to be matched to the window of a region of 53bp in length to be considered for the Smith-Waterman alignment. The mode (the ‘-M’ parameter) was changed for the fast processing speed for the sequence of 35bp in length; the seed matches per window (n) were set at ‘2’; and the seed window length (-w) was set at 150% of a sequence length. A sequence would be aligned to the genome if it matched at least two predetermined seeds that vary in lengths and in the position and number of mismatches, within a search window of 53bp. The Smith-Waterman alignment hit threshold was set with the default value of 68% in the sequence identity. For the probcalc program, the total number of bases in autosomes (chromosome 1 to 19), sex chromosomes and the mitochondrial chromosome were taken as the total length of genome, which was a required parameter for calculating the alignment probability. Custom-made perl scripts were used to parallelize the sequence alignment processes on the computing clusters. Once sequences were mapped to the genome, those that mapped to unique genomic locations were considered for further analyses.  102  The annotation of known mouse miRNAs was obtained from the latest miRBase database at the time of analysis (version 14, GFF version 2/2009-09-14) [111]. The locations of different types of the RNA classes (lincRNA, mRNA, snoRNA, rRNA, scRNA, snRNA, srpRNA, tRNA), genomic repeats (long/short interspersed nuclear elements, satellite repeats, simple repeats, low complexity sequences, rolling circle repeats, other RNA/DNA repeat elements), immunoglobulin genes and pseudogenes were obtained from the MySQL databases of the latest Ensembl (version 57) and the UCSC genome annotation (version mm9) using custom Perl scripts and MySQL commands. Two genomic locations that share at least one base were regarded as overlapped regions. For sequences that mapped to a region annotated with multiple RNA classes and/or repeat elements, the length of the overlap and the precedence of biological annotation were taken into account in the following order: miRNA > tRNA > snRNA > scRNA > snoRNA > srpRNA > rRNA > repeat Element > immunoglobulin > pseudogene > mRNA > lincRNA. The miRNA annotation had the most precedence since it had the most biological relevance to this study. The rest of the RNA classes were ordered by the average length of annotated sequences, favouring a shorter sequence length over a longer sequence due to more specific biological annotation for short sequences and the possibility of long sequences containing overlapping genes.  3.2.4 Detection of differentially abundant miRNAs The sequence counts were normalized to the total number of sequences that uniquely mapped to a genomic location annotated as known mouse miRNAs as it was done similarly by Creighton and colleagues [112]. It is assumed that the sequence count 103  directly correlates with the abundance of particular mouse miRNAs. A sequence count less than 10 was considered noise, and the value was substituted with an infinitesimally small value (i.e. 1e-50) to represent 0 to avoid problems in subsequent calculations. The normalized sequence counts were log2-transformed to stabilize the variance, and then compared between two tumour groups with different immunotherapy outcomes using the two-sample, two-tailed Student’s t-test. P-values were corrected for multiple hypothesis testing against 571 known mouse miRNAs using the Bonferroni method. The fold change difference was also calculated from taking the median of sequence counts for each miRNA after Laplace’s correction of adding a small constant (i.e. 30) was applied to avoid divisions by 0 as well as to prefer the fold changes supported by larger sequence counts to reflect biological significance. MiRNAs with corrected p-value less than 0.05 and with at least 1.5 fold changes in sequence counts were identified as being differentially abundant between two samples. Area-proportional Venn diagrams were created using the BioVenn application [113] to compare differentially abundant miRNAs from multiple comparisons.  3.2.5 Gene expression changes of predicted target genes of differentially abundant miRNAs The list of target genes of known miRNAs, which were predicted by the TargetScan program [17], was obtained from the latest online TargetScan mouse database (release 5.1) [114]. Using custom Perl scripts, predicted target genes of a differentially abundant miRNA were parsed from the list. The gene expression data of the predicted target genes were obtained by taking the fluorescence intensity values of the transcript cluster probe 104  sets for predicted target genes on the previously described Affymetrix mouse exon arrays. The gene expression levels of predicted miRNA target genes were calculated by taking the median of probe intensity values analyzed at the gene level by the Expression Console program provided by Affymetrix. The fold change of gene expression was calculated by taking the ratio of the reverse-transformed gene expression values between two tumour groups. The cumulative density distribution plots for predicted miRNA target gene expression and control gene expression were generated using log2transformed fold changes of predicted target gene expression using a custom R script and the ecdf function in R. The control gene set was generated by taking an equal number of randomly chosen genes probed on the Affymetrix mouse exon microarrays after 1000 iterations of sampling without replacement. The Kolmogorov-Smirnov test was performed to calculate the statistical significance (p-value < 0.05) of the difference in the cumulative density distribution plots of the gene expression distributions between the predicted miRNA target gene set and the control gene set using the ks.test function in R. The distribution of log2-transformed fold changes were plotted in histograms using a customary R script and the hist function in R. To estimate the percentage of significant events occurring randomly in multiple comparisons between the predicted target gene set and the randomly selected gene set, each of the 1000 randomly chosen gene sets was compared against each other (Ncomparison =499,500 from “1000 chooses 2”) and tested for statistical significance using the Kolmogorov-Smirnov test.  105  3.3 RESULTS 3.3.1 Sequencing small RNA using SOLiD sequencing platform The sequencing of small RNA fragments using the latest ‘sequencing-by-di-baseligation’ technology by SOLiD yielded 33,590,237 unfiltered sequences on average, each of 35bp in length. A nucleotide is represented by the overlap of two numbers each of which represents a particular colour corresponding to a series of two adjacent nucleotides. The libraries constructed using the modified protocol (see Material and Methods), which was sequenced using different SOLiD sequencing machines provided later in the project, nearly doubled the previous sequencing throughput, with an average number of sequences of 44,190,577 compared to 22,026,229 sequences generated from the older libraries (Table 3.1 and 3.2). Notably, sequences generated from the older libraries did not have any ‘no-call’ base from which a fluorescence signal could not be detected.  106  Table 3.1 Mapping result of miRNA libraries using the MAQ aligner Sequences generated by SOLiD were mapped to the mouse genome using the Maq aligner. The red coloured cells indicated libraries that produced ambiguous signal detections, which resulted in no alignments to the genome using the MAQ software.  107  Table 3.2 Mapping result of miRNA libraries using the SHRiMP aligner Sequences generated from the SOLiD were mapped to the mouse genome using the SHRiMP aligner. The red coloured cells indicated libraries that produced ambiguous signal detections, which resulted in no alignments to the genome.  108  However, when the quality of color calls was compared across libraries created under different protocols, sequences from libraries constructed from the modified protocol seem to have a higher and more consistent quality across the entire length of a sequence (Figure 3.1). In general, the older libraries seem to have rapid decrease of quality after 25th position in the sequence. The quality of a colour call was represented as Phred-like scores determined based on the relative strength of a particular fluorescence signal. The Phred-like scores indicate the probability of an erroneous sequence call represented in a modified log10 scale. For example, a Phred score of 20 indicates the error rate of 0.01.  Figure 3.1 Phred-like quality of colours called at each base position of 35bp-long sequence in small RNA libraries sequenced by SOLiD Blue lines represent libraries constructed using the original protocol, sequenced from the older machines. Red lines represent libraries constructed using the modified protocol, sequenced from the latest machines.  109  Alignment of the colour-space sequences to the mouse genome using Maq resulted in an unusually low mapping rate averaging at 3.13% (Table 3.1). Again, the bias from different library construction methods and sequencing on different machines may be present since the sequences from the older libraries had on average 4.5% higher mapping rate than those constructed using the modified protocol. The reads were also aligned to the mouse genome using SHRiMP, which resulted in a slightly higher mapping rate although it was still very low, averaging at 7.3% (Table 3.2). The pchance represents the probability of an alignment occurring by chance whereas the pgenome represents the probability that an alignment occurred due to evolutionary events characteristic to the genome such as the rate of mutations including point mutations and insertion or deletion mutations (‘indels’). The normodds, i.e. normalized odds, is a probability odds ratio of pgenome and pchance to represent the statistical significance of an alignment occurring between the particular sequence and the genomic sequence compared to other possible alignments. A credible alignment would result in generating a low pchance (close to 0) and a high pgnome value (close to 1), thus generating a high normodds value (close to 1). To find the statistical threshold to differentiate between the sequences that uniquely mapped to the genome and those that mapped ambiguously, the distributions of pchance, pgenome and normodd values were compared from the alignment result of SOLiD sequences from four randomly selected libraries (Nsequence = 128,975,375) that mapped to a single locus versus multiple loci (Figure 3.2).  110  Figure 3.2 Distribution of log10-transformed SHRiMP statistics (a) pgenome; (b) pchance; (c) and normodds of sequences from randomly selected libraries that mapped uniquely to genome (uniq_*) and those that mapped to multiple locations in the genome (multi_*). The red horizontal line indicates the probability at 0.05 (log10-transformed). The blue horizontal lines in (c) indicate 0.6 and 0.9 (log10transformed), which represent a statistically significant alignment to the genome. (a)  111  (b)  112  (c)  The pchance and pgenome values did not seem to be good measures to differentiate uniquely matching sequences versus promiscuously matching sequences. However, normodds values of all uniquely matching sequences were 1, whereas the median normodds for promiscuously matching sequences ranged between 0.074 and 0.11. From the personal correspondence with the author of the program, Michael Brudno, who reaffirmed that the normodds value has to be at least 0.8 to remove sequences that aligned to the repetitive regions, the sequences that uniquely mapped to the genome with the normodds values of 1 were used for further analyses. Since the lengths of mature miRNA sequences range from 18bp to 23bp, the rest of the sequence generated using the SOLiD sequencing technology belonged to the adapter  113  sequence ligated at the 5’ end of a cDNA template. To estimate the starting position of the adapter sequence, the colour calls from each base of SOLiD sequences, which uniquely mapped to a known miRNAs were compared. In general, one would expect to see the identical colour call at each base consistently throughout the entire length of the sequence. However, the inconsistency of colour calls increases after approximately 25th base in some libraries, which reflects the rapid decrease of sequence quality as well as the varying starting positions of the adapter sequences (Figure 3.3).  Figure 3.3 Percentages of different colours called at each base across sequences from randomly selected libraries The sequences were mapped to a known mouse microRNA using SHRiMP. a) Reads from the library NOP12PR-MM0430 that mapped to mmu-mir-24-2 (N=1043); b) Reads from the library NOP21CR-MM0493 that mapped to mmu-mir-181a-1 (N=1291); c) Reads from the library NOP23PR –MM0502 that mapped to mmu-mir-125a (N=30,382). The first base, the letter T, is the last base of the ligation adapter sequence. Colours represent unique colour calls: Black = ‘T’; red = ‘3’; yellow = ‘2’; green = ‘1’; blue = ‘0’; orange = ‘no call’. (a)  114  (b)  (c)  115  The presence of the adapter sequence would cause many mismatches when the sequences aligned to the genome, which might explain the unusually low mapping rate of SOLiD sequences using Maq, which did not allow more than 3 mismatches in the alignment. On the other hand, in the alignment by the SHRiMP programs, the effect of the adapter sequence on the alignment result did not seem to be problematic shown by many sequences that mapped uniquely to the genome. These alignments were statistically significant with the perfect matches up to 23bp followed by all mismatches to the end of the sequence, which indicates the presence of the adapter sequence. To estimate the extent of the problem caused by the presence of the adapter sequence, the last 10 bases of the sequences were removed before they were aligned to the mouse genome using Maq and SHRiMP. Although the re-alignment result by Maq showed a considerable improvement in the mapping rate from 7.7% to 25.3%, this is still regarded as an unusually low mapping rate, compared to ~70-80% mapping rate observed in the massively parallel sequencing of small RNA fragments using Solexa/Illumina sequencing technology. Morever, the percentage of sequences that align to a unique location in the genome decreased from 47.7% to 25.3% after the trimming of the last 10bp. Using SHRiMP aligner, the percentage of uniquely mapped sequences increased from 68.5% to 73.7% (Table 3.3). However, the mapping rate did not increase significantly, resulting in approximately 3% higher mapping rate after the last 10bp were trimmed. Overall, the removal of the last 10 bp did not improve the mapping rate significantly, indicating that the presence of the adapter sequence did not cause a significant problem in the alignment by SHRiMP.  116  Table 3.3 Mapping rate of miRNA sequnces using SHRiMP after trimming the last 10 base pairs  117  3.3.2 Annotation of small RNAs The annotation of various non-coding RNAs, protein-coding RNAs, genomic repeats and other genomic elements was performed using the miRBase, Ensembl and UCSC genome database. The genome annotation included non-coding small RNAs (miRNA, tRNA, rRNA, snRNA, snoRNA and srpRNA); protein-coding RNA; RNA and DNA repeats and other genomic elements such as immunoglobulin genes and pseudogenes. The RNA and DNA repeat elements included short/long interspersed nuclear elements (SINE/LINE), long terminal elements (LTR), transposons, DNA/RNA repeat elements, micro-satellites, rolling circle repeats and low complexity repeats. The sequences were annotated as a particular genomic element if the position of the sequenced aligned to the genome overlapped with the positions of the annotated genome. In the case of multiple annotations where a particular genomic location or multiple genomic locations that overlapped were annotated with multiple genomic elements, the biological annotation and the length of the overlap were taken as weighting factors for the sequence annotation. The most abundant class was the miRNAs, represented by 37.1% of the uniquely aligned sequences (NmiRNA_sequence = 10,220,756). The second most abundant class was the mRNAs (18.1%; NmRNA_sequence = 4,997,578). About 17% of the sequences mapped to repeat elements. A relatively small proportion of sequences mapped to other small noncoding RNA classes such as scRNA, snoRNA and tRNA (Figure 3.4). 17% of the sequences (Nsequences = 4,674,850) mapped uniquely to genomic loci that had no known function, suggesting that these sequences may represent novel miRNA sequences.  118  Figure 3.4 Distribution of non-coding small RNA classes across all libraries sequenced using SOLiD  There was a dramatic difference in the distributions of sequences annotated into different RNA classes between those that were generated from the older libraries and those that were generated from the libraries constructed under the modified protocol (Table 3.4).  119  Table 3.4 Annotation of various RNA classes and genomic elements Library  miRNA tRNA snRNA scRNA snoRNA srpRNA rRNA repeatElements immunoglobulin pseudogene mRNA lincRNA Run_quadrant Machine (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%)  Unknown (%)  MM0430- tra00000_(3) NOP12_PR tra17386_(3)  SOLID-1 5.31  1.04  0.24  0.69  7.72  0.01  7.48  26.37  0.00  0.15  25.31  0.27  25.40  SOLID-2 5.12  1.09  0.24  0.69  7.35  0.01  6.92  27.91  0.00  0.14  24.93  0.27  25.34  MM0431- tra18058_(2) NOP18_PD tra18206_(3)  SOLID-1 7.00  1.26  0.21  0.80  4.95  0.01  10.16  30.82  0.00  0.10  21.11  0.24  23.34  SOLID-2 6.92  1.16  0.20  0.75  4.99  0.01  10.54  30.41  0.01  0.10  20.96  0.24  23.71  MM0434- tra00000_(4) NOP12_PR tra17386_(4)  SOLID-1 7.46  1.36  0.18  0.42  5.65  0.01  8.81  28.62  0.00  0.11  20.15  0.22  27.00  SOLID-2 6.98  1.35  0.17  0.39  5.15  0.01  8.62  30.76  0.00  0.10  20.02  0.22  26.23  SOLID-1 3.61  1.38  0.25  1.04  7.29  0.01  10.65  27.07  0.00  0.12  24.73  0.28  23.57  MM0457- tra18058_(4) NOP18_PD tra18206_(4)  SOLID-1 4.05  1.41  0.25  1.04  7.17  0.01  10.49  27.24  0.00  0.12  24.54  0.28  23.40  SOLID-2 2.90  1.47  0.24  0.89  6.64  0.01  9.35  29.80  0.00  0.11  23.66  0.29  24.65  tra19544_1 MM0493tra19544_2 NOP21_CR tra19544_3  SOLID-3 36.62 0.52  0.13  2.05  4.16  0.01  0.49  15.39  0.00  0.09  23.09  0.21  17.24  SOLID-3 37.03 0.50  0.13  2.03  4.15  0.01  0.50  14.97  0.00  0.09  23.16  0.20  17.23  SOLID-3 37.79 0.50  0.12  2.02  4.13  0.00  0.48  14.70  0.00  0.09  22.98  0.20  16.99  tra19545_1 MM0494tra19545_2 NOP21_CR tra19545_3  SOLID-3 46.90 0.87  0.09  1.45  3.10  0.01  0.55  15.19  0.00  0.07  16.69  0.14  14.93  SOLID-3 46.52 0.79  0.09  1.46  3.09  0.01  0.58  14.85  0.00  0.06  17.01  0.15  15.40  SOLID-3 46.35 0.79  0.09  1.51  3.11  0.01  0.58  14.64  0.00  0.06  17.18  0.15  15.55  tra19614_1 MM0501tra19614_2 NOP23_PR tra19614_3  SOLID-4 63.30 0.38  0.20  3.40  2.06  0.01  1.45  8.76  0.00  0.05  10.05  0.10  10.22  SOLID-4 64.60 0.36  0.19  3.28  1.95  0.01  1.43  8.45  0.00  0.05  9.68  0.10  9.90  SOLID-4 59.50 0.42  0.22  3.72  2.23  0.02  1.62  9.65  0.00  0.06  11.09  0.12  11.36  tra19615_1 MM0502tra19615_2 NOP23_PR tra19615_3  SOLID-4 67.71 0.76  0.10  2.43  1.39  0.01  2.21  9.51  0.00  0.04  7.78  0.08  7.97  SOLID-4 67.40 0.77  0.10  2.41  1.37  0.01  2.30  9.55  0.00  0.04  7.91  0.08  8.04  SOLID-4 67.74 0.77  0.11  2.50  1.39  0.01  2.32  9.44  0.00  0.04  7.68  0.08  7.92  MM0435NOP18_PD tra18058_(3)  120  Notably, the largest discrepancy in the percentage of sequences annotated between the old libraries and the new ones was in the miRNA annotation. On average, only 5.5% of the uniquely mapped sequences from the old libraries were annotated as miRNAs, whereas 53% of sequences from the libraries constructed under the modified protocol were annotated as miRNAs, which is in agreement with the result from a similar miRNA profiling study by Morin and colleagues using the Solexa/Illumina sequencing methodology [115], where 52% of small RNA sequences was annotated as miRNAs. Conversely, there were more sequences generated from the older libraries that were annotated as other RNA classes compared to those generated from the libraries constructed under the modified protocol. For example, sequences from the older libraries were consistently annotated as mRNA, rRNA and genomic repeats with higher percentages (22.8%; 9.2%; 28.8% respectively) compared to sequences from the libraries constructed under the modified protocol (14.5%; 1.2%; 12.1% respectively). The overall quality of the older libraries seems to be poorer than the libraries that were constructed later. The libraries of all PD tumours and some of the PR tumours were constructed using the original protocol. For analyses concerning the CR and PR tumours, the sequences generated from the old PR libraries were discarded to minimize library-related bias. Sequence data generated from libaries of NOP21CR and NOP23PR tumours, i.e. MM0493, MM0494, MM0501, MM0502, and of NOP18PD tumour, i.e. MM0431, MM0435, MM0457, were considered for further analyses. The data generated from the NOP12PR library were removed.  121  3.3.3 Differential abundance of miRNAs in tumours with different adoptive immunotherapy outcomes Sequences that mapped unambiguously to the mouse genome were aligned to 600 mature miRNA sequences, which represented 571 unique mouse miRNAs genes obtained from the latest miRBase [116]. The number of sequences that aligned to a particular miRNA was assumed to be directly proportional to the absolute abundance of the miRNA in tumour cells. Hence, sequence counts were compared across libraries to detect miRNAs that were differentially expressed in different tumour groups. In this study, there is no way to distinguish between the difference in the transcriptional activity of miRNA genes or the difference in the rate of mature miRNA degradation in cells, therefore the differential abundance and the differential expression of miRNAs will be used interchangeably throughout this chapter. A sequence count of less than 10 was considered as noise and thus regarded as 0. The fold change represented the ratio of sequence counts between two libraries after noise was removed. Again, only sequences that mapped uniquely to a single miRNA were considered further. Sequences from all tumour libraries collectively represented 260 miRNA genes, comprising 45.5% of known miRNA genes. To normalize against the heavy bias in the sequencing throughput generated from different kinds of libraries, the raw sequence counts were normalized to the total number of unambiguously mapped sequences in each library. The Student’s t-test was performed on the log2-transformed normalized sequence counts. The p-values were corrected for multiple testing of 571 miRNA genes using the Bonferroni correction method. Due to the severe difference in the proportion of miRNA sequenced between the older libraries and the libraries  122  constructed under the modified protocol, only CR and PR tumour classes were compared and sequences from the older libraries were excluded. For PD tumour-related analyses, a similar exclusion was not possible because all PD tumour libraries were constructed using the original protocol. The miRNAs were identified as being differentially abundant between two tumour groups if the Bonferroni-corrected p-value was less than 0.05 and the fold change was greater than 1.5. There were 172 miRNAs with a fold change greater than 1.5 in the comparison between PD and CR tumours; 89 miRNAs in PR and CR comparison in which the old PR libraries were removed (129 miRNAs if all PR libraries were included); and 146 miRNAs in the PD and PR comparison. The Student’s t-test revealed 386 miRNAs that showed a significant difference in abundance between PD and CR tumours with the corrected pvalue less than 0.05; 9 miRNAs between PR and CR tumours when the old PR libraries were removed (4 miRNAs if all libraries were accounted for), and 5 miRNAs between PD and PR tumour groups. MiRNAs with a fold change greater than 1.5 and a corrected p-value less than 0.05 were identified as differentially abundant miRNAs. In total, 26 miRNA genes were found to be differentially abundant between PD and CR tumours, 5 miRNAs between PD and PR tumours, and 8 miRNAs between PR and CR tumours (4 miRNAs between PR and CR tumours including the older libraries) (Figure 3.5). These comparisons were done after removing the old libraries i.e. NOP12PR.  123  Figure 3.5 Differentially abundant miRNAs from multiple outcome class comparisons in an area-proportional Venn diagram  Interestingly, almost all differentially expressed miRNAs were under-expressed in the PD tumour group compared to other tumour groups. Also, many differentially expressed miRNAs were present at a much higher level in PR compared to PD or CR tumours. Mmu-mir-541 was found to be differentially abundant in all tumour classes with the highest abundance in PR compared to CR and PD tumours (PR fold change relative to CR and PD tumours = 3.4; the fold change of CR relative to PD = 1.5). Three miRNA genes, mmu-mir-204, mmu-mir-125a and mmu-mir-1-1, were differentially expressed in the CR tumours exclusively. These miRNAs were differentially abundant between CR and PD tumour groups, and between CR and PR groups, but not between the PD and PR tumour groups (Table 3.5). 124  Table 3.5 Differentially abundant miRNAs between tumours with different outcome of adoptive T-cell therapy The fold change (FC) is the ratio of the median of the sequence counts between two groups. The p-values were calculated using Student’s t-test and were corrected for multiple hypothesis testing of 571 known mouse miRNAs using the Bonferroni correction. a) Differentially abundant miRNAs between PD and CR groups. FC is calculated relative to CR; b) Differentially abundant miRNAs between PD and PR. FC is calculated relative to PR; c) Differentially abundant miRNAs between PR and CR after removing the sequence counts from the older libraries. FC is calculated relative to CR; d) Differentially abundant miRNAs between PR and CR. FC is calculated relative to CR. miRNA genes that were differentially abundant specifically in CR tumours are coloured in purple, specifically in PD tumours in green, and in all three outcome classes in blue. (a)  (b)  (c)  125  (d)  The miRNA mmu-mir-204 consistently showed under-expression in CR relative to other tumour groups with a fold change of -10.57 (corrected p-value = 3.76e-07) relative to PR, and -1.57 relative to PD (corrected p-value = 8.73e-06). Conversely, mmu-mir-1-1 was consistently higher in abundance in CR compared to PR (fold difference = 5.97; corrected p-value = 1.11e-16) and PD (fold difference = 5.97; corrected p-value = 1.81e-14). The abundance of mmu-mir-125a showed a mixed pattern where it was more abundant in CR tumours compared to PD (fold difference = 3.98; corrected p-value = 0.00025), but less abundant compared to PR (fold difference = -1.74; corrected p-value = 0.036). In PD tumours, there were two miRNA genes, mmu-mir-10b and mmu-mir-99a, which were under-expressed in the PD tumours compared to both CR and PR tumour groups.  3.3.4 Possible effect of differentially abundant microRNA on the expression of predicted target genes MicroRNAs are known to negatively regulate the expression of their target genes by imperfect complementary binding at the 3’ UTR sites. To investigate the effect of differentially abundant miRNAs on the expression level of their target genes, the gene expression profile of miRNA target genes from the previous Affymetrix microarray data were assessed for significant gene expression changes amongst tumour groups with  126  different immunotherapy outcomes. The online TargetScan database [114] contains a list of predicted target genes of known or predicted miRNAs in mouse using the TargetScan program, which is based on the combination of thermodynamics-based modeling of RNA:RNA duplex interactions and comparative sequence analysis across multiple genomes [17]. The latest TargetScan database (Release 5.1; created on April 2009) contained 7,719 mouse genes that are predicted to be the targets of 153 miRNA clusters, which consist of 241 unique miRNA genes. A single miRNA has hundreds of target genes predicted by TargetScan. Notably, multiple miRNAs were predicted to target the same gene, possibly due to the high sequence similarity in mature miRNA transcripts as well as the biological functions. The TargetScan database version 5.1, which was the latest at the time of the analysis, did not contain predicted target genes for every miRNAs in the miRBase database. The miRNAs that did not have predicted target genes may be those whose target binding sites fall beyond the canonical target regions of conserved 3’ UTRs of mRNA transcripts, reflecting a limitation of current prediction algorithms. Out of 26 differentially abundant miRNAs identified from the CR and PD tumour comparison, 16 miRNAs were predicted to target 3,296 unique genes. From 8 differentially abundant miRNAs between CR and PD tumours, 6 miRNAs were predicted to target 1,535 unique genes. From 5 differentially abundant miRNAs between PR and PD tumours, 3 miRNAs were predicted to target 228 unique genes. The fold changes in the predicted target gene expression were calculated using the previously described microarray dataset. In brief, the median of normalized probe fluorescence intensities represented the gene expression level of a tumour group from twelve Affymetrix microarrays of 3 PD tumours derived from NOP6PD, NOP17PD and  127  NOP18PD cell lines, 2 PR tumours derived from NOP13PR and NOP23PR cell lines, and 1 CR tumour from NOP21CR cell line. A ratio of the median values between two tumour groups was calculated for the fold change. To investigate the effect of the differential abundance of miRNAs on the expression level of their target genes, the distributions of the median log2-transformed fold changes of the predicted target genes in different tumour groups were plotted as cumulative distribution functions (CDFs). A control group was created to represent the neutral case of genes under no repressive control by miRNAs. The expression levels of randomly selected genes from the microarray dataset with an equal number of genes as the predicted target gene set of a particular miRNA were calculated similarly as described above to represent the expression levels of the control group. The calculation of the gene expression of the randomly selected control group was repeated for 1000 iterations using the method of sampling without replacement to increase the robustness of the gene expression values from the control gene set. To estimate the statistical difference in gene expression of the miRNA target genes and the genes under no miRNA control (i.e. p-value < 0.05), the distributions of log2-transformed fold changes of targeted gene set and each of the 1000 control gene sets were compared using the Kolmogorov-Smirnov test (Figure 3.6).  128  Figure 3.6 Cumulative distribution functions and histograms of log2-fold changes of predicted target genes and randomly selected genes from 1,000 iterations The distributions of predicted target genes of differentially abundant miRNAs and 1000 sets of randomly chosen ‘control’ genes were compared using the Kolmogorov-Smirnov test (Bonferroni corrected p-value < 0.05). The frequency of significantly different distibutions is shown in percentage. The percentage of false positive events is indicated as ‘(random)’. CDFs and histograms show the comparisons of expression levels of the control genes and the predicted target genes of differentially abundant miRNAs between (a) CR and PD tumours; b) between CR and PR tumours (the result after the old libraries were removed); and c) between PR and PD tumours. (a)  129  130  (b)  131  (c)  The comparison of the CDFs showed no evidence that supported the gene expression change of the target genes of a differentially abundant miRNA compared to the randomly selected genes. The distribution of log-transformed fold changes in the expression level of genes predicted to be under the control of the most differentially abundant miRNA i.e. mmu-mir-204 (10.57 times more abundant in CR compared to PR) did not differ from the fold changes in expression levels of randomly selected genes. Overall, no evidence of repression in the transcriptional activities of the predicted target genes could be found from this analysis. The CDFs of the fold changes of a target gene set and each of the 1000 control gene sets were compared for statistical differences using Kolmogorov-Smirnov test (Figure 3.6). To estimate the frequency of significant results randomly arising from repeatedly performing such test, each gene set from the 1000 control gene set was compared against each other, resulting in the pair-wise testing of 499,500 (i.e. ‘1000 chooses 2’) randomly  132  generated CDFs for statistical significance using the Kolmogorov-Smirnov test with the threshold p-value set at a maximum of 0.05. On average, 3.8% of the pair-wise comparisons of random distributions resulted as being significant, which represented the frequency of false positives. Interestingly, the frequency of significantly different CDFs from the predicted target genes of the most differentially abundant miRNA, mmu-mir204, was markedly different (25%) compared to the frequency of the expected false positives (3.43%). With the exception of mmu-mir-204, there was no noticeable difference between the frequencies of expected false positives and significantly different CDFs from comparing miRNA target genes and randomly selected genes, demonstrating that the abundance of miRNA did not correlate with the changes in target gene expression levels detected by microarrays.  3.4 DISCUSSION The sequencing of small RNAs in murine breast tumours using the SOLiD technology yielded a tremendous amount of data of more than 33 million sequences generated. However, the mapping of sequences to the mouse genome was heavily influenced by colour-space alignment tools used, the program parameter values and the presence of the adapter sequence at the 3’ end of a sequence. Moreover, regardless of which alignment tool was used, the proportion of the reads mapped to the genome was less than 10%. This unusually low mapping rate can be explained by the quality of the library construction, suggested by the discrepancy in the percentage of sequences annotated into different RNA classes in the earlier libraries versus the libraries constructed later with more improvements in the protocol. Also, the development of the colour-space sequence 133  alignment tools is still at its infancy, which prompts for more improvements to be made in the underlying algorithms for finding better alignment sites. However, the biggest reasons for the low mapping rate observed in this study can be the inadequate testing of different parameter settings of the alignment tool, and the lack of post-sequencing filters for sequences before aligning them to the genome and the consideration of sequences that mapped to more than one genomic location. The sensitivity of the alignment in the SHRiMP program is heavily dictated by program parameters such as the percentage of sequence similarity i.e. number of mismatches, and the position of the mismatches in the spaced seeds from which an initial alignment extends. Rather than forcing the alignment to have perfect matches throughout the sequence until the adapter sequence is found at the 3’ end, a better parameter could have been 1 or 2 mismatches allowed in the middle of a sequence after trimming the last 10-12 bases of the sequence. This would have taken into the account the abiotic factor such as the diminishing trend in sequence qualities across the length of a sequence, and the biological phenomena of miRNA processing such as RNA editing. Also, the natural redundancy of miRNA sequences was not considered during the alignment process. There are miRNA families with high sequence similarities between their members, especially among their mature miRNA transcripts. Assuming that the majority of miRNA transcripts present in cells at the time of sequencing are mature miRNAs in the cytoplasm, many sequences derived from different miRNA genes with high sequence similarity would have mapped to multiple locations in the genome. To estimate the extent of the loss of sequences that ambiguously mapped to the genome by enforcing the unique mapping, the mature miRNA sequences used in this study were aligned to the  134  genome using SHRiMP. Approximately 17% of the mature miRNA sequences mapped ambiguously to multiple places in the genome (data not shown), implying that, although this loss would have not affected the overall mapping rate, the absolute number of sequences used for the analyses could have increased by 17%. However, the encouraging conclusion from these observations is that the sequences selected for further analyses were bona fide miRNA sequences with no ambiguity in their presence in the tumours investigated in this study. Despite the problems imposed by the low mapping rate, more than 3 million sequences were available to identify the differentially expressed miRNAs allowing the analysis at much finer scale than other profiling methodologies such as microarrays. While the majority of sequences was annotated as known miRNAs and other non-coding RNAs, 17% of sequences were mapped to a region with no biological annotation, suggesting that these sequences can be the source of novel miRNAs and the discovery of rarely expressed miRNAs that are uniquely present in breast cancer cells with different tumour immunogenicity. In total, sequences that represent 260 unique miRNA genes (nearly half of all known genes) were identified, which is similar to the finding from a similar miRNA profiling study by Morin and colleagues where they identified 334 miRNA genes [115]. Out of 260 miRNAs expressed collectively in the tumours with different immunotherapy outcomes of CR, PR and PD, 32 of them were differentially expressed amongst the three tumour groups. As repeatedly observed in the histological characteristics as well as transcriptional profiles based on microarray data, the CR and PD tumours have the most difference between their miRNA abundance. It is also interesting to observe that the  135  majority of miRNAs in PD tumours are under-expressed compared to CR. It is possible that this may be due to the poor library quality observed commonly in libraries constructed following the old protocol. However, there is an interesting biological phenomenon associated with systematic under-expression of miRNAs and cancer, which may be manifested by the under-expression of miRNAs in PD tumours. The systemic down-regulation of miRNAs was associated with more deregulated cell states and less differentiated tumours arising from various tissue types in human [94]. The PD group represents tumours with the poorest prognosis in the breast cancer model in this study, which may be the reflection of a more deregulated genomic state characterized by the pronounced under-expression of the miRNAs profiled in this study. Also notable is the observation that almost all miRNAs identified as differentially expressed in PR (with the exception of one) compared to PD or CR are over-expressed, which implies different translational activities unique to PR tumours that were almost indistinguishable from CR at the transcriptional level shown by the previous microarray analyses. Mmu-mir-214, a miRNA commonly found to be differentially expressed in all three tumour groups, has no known biological function found to date. However, the largest fold change observed between CR and PR (10.5 fold over-expression in PR) poses a speculation about its possible role in distinguishing between the CR and PR outcomes after adoptive immunotherapy. There are several miRNAs identified in this study as being differentially expressed that have been known to be involved in various cancers and immune system development. MiR-15a and miR-16-1, miRNAs both under-expressed in PD are well-known tumour suppressor miRNAs. Both miR-16-1 and miR-15a are found to be frequently deleted or  136  down-regulated in B cell chronic lymphocytic leukemia (B-CLL), which is the most common form of leukemia in the Western hemisphere [3, 7, 68, 117]. Interestingly, the expression of miR-15a and miR-16-1 induced apoptosis in a leukemia cell line, suggesting the therapeutic promise of these miRNAs in cancer treatment [3]. Similarly, miR-29, which is under-expressed in PD relative to CR, is also found at a reduced level in B-CLL and non-small cell lung carcinoma (NSCLC) in which miR-29 is involved with apoptotic processes [118]. Enforced expression of miR-29 was also shown to reduce the rate of tumour growth in in vitro and in vivo experiments [7]. The expression level of MiR-10b, which is under-expressed in PD compared to CR and PR (fold change of 4.35 and 2.9 respectively), is inversely correlated with prognosis in breast cancer [7]. Since the higher level of miR-10b expression implies worse prognosis, this finding confounds the prediction from the previously described microarray analyses, in which PD tumours are likely to resemble the basal-like subtype with the worst prognosis in human breast cancer. MiR-181b-2, over-expressed in PR, has no known biological function known to date. However, its close sister, miR-181a, is a potent regulator of T- and B-cell growth and development, T-cell selection and sensitivity to antigens [7, 68, 71]. miR-98, found to be heavily under-expressed in PD tumours (fold change of 7.6), is a synonym for let-7 miRNA. Well-known and one of the first to be discovered, let-7/miR-98 is normally involved with cellular differentiation and growth [7]. In cancer, it acts as a tumour suppressor by negatively regulating the powerful RAS oncogene [3, 86]. Thus, it is found to be down-regulated in various cancerous tissues including lung, colon and ovary, and shown to be a strong prognostic marker correlated with poor prognosis [77, 84-86, 119, 120]. Similarly in breast cancer, reduced expression of let-7/miR-98 was correlated  137  with poor prognosis [81], which is in congruence with what is predicted with the PD tumour model. There were several miRNAs that showed massive fold change difference in the order of hundreds but did not have significant p-values due to large variance in sequence counts from different libraries. The Audic-Claverie testing [121], which takes the raw sequence counts instead of normalized values, may have allowed to select more differentially expressed miRNAs with large fold changes. The effect of repressive regulations of differentially abundant miRNAs on their target genes was not observed in this study using microarray data. One of the reasons for this observation could come from the presence of inaccurately predicted target genes from the TargetScan database, obscuring the statistical testing based on frequency of significant CDF comparisons used in this study. Amongst the current miRNA target gene prediction programs, the TargetScan was shown to have the highest level of accuracy compared to PITA, PicTar and miRanda [38]. However, in the same study, the authors found that two-thirds of their predicted target genes showed no response to the loss of the miRNA in question [38], which suggested that there was a substantial number of false positives predicted by current miRNA target identifying algorithms. A method to improve the false positive rate would be to filter out from the analysis predicted genes with a less confidence in prediction (PCT in the TargetScan list that indicates the score of confidence) or with a longer miRNA seed length, which would require more stringent prediction of target regions. For overall algorithmic improvement as mentioned previously, focusing on the conserved 3’ UTRs of mRNA transcripts, and other factors such as secondary folding structures should be considered for more accurate target predictions.  138  Another reason for the lack of gene expression changes observed from this study could be the lack of power to detect subtle changes in mRNA transcript abundance by the use of microarrays. There is growing evidence that gene suppression exerted by miRNAs results in a surprisingly subtle difference at the protein level. Several large-scale proteomics experiments have shown that, while a single miRNA target range is vast, potentially affecting hundreds of proteins at a time, the translationally repressive effect by miRNAs is rather subtle [12, 38, 47], such that the reduction of the protein level detected was on average as mild as 1.3 in fold change difference [38] and rarely exceeded four-fold [47]. Assuming that mRNA transcript abundance directly correlates with protein levels, the fold change difference of 1.3 might not be detected by the hybridization-based method of interrogating the transcript abundance such as microarrays used in this study. The comparison of transcriptome shotgun sequencing data described in chapter 4 of this study could be an ideal future study to detect subtle differences at the transcriptional level. Conversely, the result of this analysis could also reflect the mechanism of miRNAmediated gene suppression. The lack of transcriptional changes detected in target genes using microarray data support the hypothesis that mRNA-mediated gene suppression may be achieved mainly through translation inhibition e.g. blocking translation initiation or ribosome association, rather than through the degradation of mRNA transcripts. This suggests that the true validation of the effect of miRNA on target genes requires the measurement of the protein level of target genes using experimental methodologies such as Western blots or quantitative-mass-spectrometry-based SILAC (stable isotope labeling with amino acids in cell culture) used in studies by Selbach and colleagues [47] and Baek  139  and colleagues [38]. The use of ‘antagomirs’ [105] to knock-down miRNAs, or to increase the effect of miRNAs through transfection would be necessary tools for further experimental validations of miRNAs.  140  3.5 BIBLIOGRAPHY 1. Lee RC, Feinbaum RL, Ambros V: The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 1993, 75(5):843-854. 2. Bartel DP: MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004, 116(2):281-297. 3. Esquela-Kerscher A, Slack FJ: Oncomirs - microRNAs with a role in cancer. Nat Rev Cancer 2006, 6(4):259-269. 4. Caldas C, Brenton JD: Sizing up miRNAs as cancer genes. Nat Med 2005, 11(7):712714. 5. Blenkiron C, Miska EA: miRNAs in cancer: approaches, aetiology, diagnostics and therapy. Hum Mol Genet 2007, 16 Spec No 1:R106-13. 6. Espinosa CES, Slack FJ: The role of microRNAs in cancer. Yale J Biol Med 2006, 79:131-140. 7. Schickel R, Boyerinas B, Park SM, Peter ME: MicroRNAs: key players in the immune system, differentiation, tumourigenesis and cell death. Oncogene 2008, 27(45):5959-5974. 8. Kanellopoulou C, Monticelli S: A role for microRNAs in the development of the immune system and in the pathogenesis of cancer. Semin Cancer Biol 2008, 18(2):7988. 9. Jackson AL, Bartz SR, Schelter J, Kobayashi SV, Burchard J, Mao M, Li B, Cavet G, Linsley PS: Expression profiling reveals off-target gene regulation by RNAi. Nat Biotechnol 2003, 21(6):635-637. 10. Vasudevan S, Tong Y, Steitz JA: Switching from repression to activation: microRNAs can up-regulate translation. Science 2007, 318(5858):1931-1934. 11. Kim VN: MicroRNA biogenesis: coordinated cropping and dicing. Nat Rev Mol Cell Biol 2005, 6(5):376-385. 12. Xiao C, Rajewsky K: MicroRNA control in the immune system: basic principles. Cell 2009, 136(1):26-36. 13. Lewis BP, Burge CB, Bartel DP: Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 2005, 120(1):15-20. 14. Lim LP, Glasner ME, Yekta S, Burge CB, Bartel DP: Vertebrate microRNA genes. Science 2003, 299(5612):1540. 15. Bentwich I, Avniel A, Karov Y, Aharonov R, Gilad S, Barad O, Barzilai A, Einat P, Einav U, Meiri E, Sharon E, Spector Y, Bentwich Z: Identification of hundreds of conserved and nonconserved human microRNAs. Nat Genet 2005, 37(7):766-770.  141  16. Berezikov E, Guryev V, van de Belt J, Wienholds E, Plasterk RH, Cuppen E: Phylogenetic shadowing and computational identification of human microRNA genes. Cell 2005, 120(1):21-24. 17. Lewis BP, Shih IH, Jones-Rhoades MW, Bartel DP, Burge CB: Prediction of mammalian microRNA targets. Cell 2003, 115(7):787-798. 18. Krek A, Grun D, Poy MN, Wolf R, Rosenberg L, Epstein EJ, MacMenamin P, da Piedade I, Gunsalus KC, Stoffel M, Rajewsky N: Combinatorial microRNA target predictions. Nat Genet 2005, 37(5):495-500. 19. John B, Enright AJ, Aravin A, Tuschl T, Sander C, Marks DS: Human MicroRNA targets. PLoS Biol 2004, 2(11):e363. 20. Khvorova A, Reynolds A, Jayasena SD: Functional siRNAs and miRNAs exhibit strand bias. Cell 2003, 115(2):209-216. 21. Schwarz DS, Hutvagner G, Du T, Xu Z, Aronin N, Zamore PD: Asymmetry in the assembly of the RNAi enzyme complex. Cell 2003, 115(2):199-208. 22. Reinhart BJ, Slack FJ, Basson M, Pasquinelli AE, Bettinger JC, Rougvie AE, Horvitz HR, Ruvkun G: The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature 2000, 403(6772):901-906. 23. Gerlach W, Giegerich R: GUUGle: a utility for fast exact matching under RNA complementary rules including G-U base pairing. Bioinformatics 2006, 22(6):762764. 24. Varani G, McClain WH: The G x U wobble base pair. A fundamental building block of RNA structure crucial to RNA function in diverse biological systems. EMBO Rep 2000, 1(1):18-23. 25. Brennecke J, Hipfner DR, Stark A, Russell RB, Cohen SM: bantam encodes a developmentally regulated microRNA that controls cell proliferation and regulates the proapoptotic gene hid in Drosophila. Cell 2003, 113(1):25-36. 26. Johnston RJ, Hobert O: A microRNA controlling left/right neuronal asymmetry in Caenorhabditis elegans. Nature 2003, 426(6968):845-849. 27. Vella MC, Choi EY, Lin SY, Reinert K, Slack FJ: The C. elegans microRNA let-7 binds to imperfect let-7 complementary sites from the lin-41 3'UTR. Genes Dev 2004, 18(2):132-137. 28. Moss EG, Lee RC, Ambros V: The cold shock domain protein LIN-28 controls developmental timing in C. elegans and is regulated by the lin-4 RNA. Cell 1997, 88(5):637-646. 29. Doench JG, Sharp PA: Specificity of microRNA target selection in translational repression. Genes Dev 2004, 18(5):504-511. 30. Kiriakidou M, Nelson PT, Kouranov A, Fitziev P, Bouyioukos C, Mourelatos Z, Hatzigeorgiou A: A combined computational-experimental approach predicts human microRNA targets. Genes Dev 2004, 18(10):1165-1178.  142  31. Bartel DP: MicroRNAs: target recognition and regulatory functions. Cell 2009, 136(2):215-233. 32. Filipowicz W, Bhattacharyya SN, Sonenberg N: Mechanisms of posttranscriptional regulation by microRNAs: are the answers in sight? Nat Rev Genet 2008, 9(2):102-114. 33. Kloosterman WP, Wienholds E, Ketting RF, Plasterk RH: Substrate requirements for let-7 function in the developing zebrafish embryo. Nucleic Acids Res 2004, 32(21):6284-6291. 34. Lytle JR, Yario TA, Steitz JA: Target mRNAs are repressed as efficiently by microRNA-binding sites in the 5' UTR as in the 3' UTR. Proc Natl Acad Sci U S A 2007, 104(23):9667-9672. 35. Farh KK, Grimson A, Jan C, Lewis BP, Johnston WK, Lim LP, Burge CB, Bartel DP: The widespread impact of mammalian MicroRNAs on mRNA repression and evolution. Science 2005, 310(5755):1817-1821. 36. Lim LP, Lau NC, Garrett-Engele P, Grimson A, Schelter JM, Castle J, Bartel DP, Linsley PS, Johnson JM: Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs. Nature 2005, 433(7027):769-773. 37. Grimson A, Farh KK, Johnston WK, Garrett-Engele P, Lim LP, Bartel DP: MicroRNA targeting specificity in mammals: determinants beyond seed pairing. Mol Cell 2007, 27(1):91-105. 38. Baek D, Villen J, Shin C, Camargo FD, Gygi SP, Bartel DP: The impact of microRNAs on protein output. Nature 2008, 455(7209):64-71. 39. Stark A, Lin MF, Kheradpour P, Pedersen JS, Parts L, Carlson JW, Crosby MA, Rasmussen MD, Roy S, Deoras AN, Ruby JG, Brennecke J, Harvard FlyBase curators, Berkeley Drosophila Genome Project, Hodges E, Hinrichs AS, Caspi A, Paten B, Park SW, Han MV, Maeder ML, Polansky BJ, Robson BE, Aerts S, van Helden J, Hassan B, Gilbert DG, Eastman DA, Rice M, Weir M, Hahn MW, Park Y, Dewey CN, Pachter L, Kent WJ, Haussler D, Lai EC, Bartel DP, Hannon GJ, Kaufman TC, Eisen MB, Clark AG, Smith D, Celniker SE, Gelbart WM, Kellis M: Discovery of functional elements in 12 Drosophila genomes using evolutionary signatures. Nature 2007, 450(7167):219232. 40. Easow G, Teleman AA, Cohen SM: Isolation of microRNA targets by miRNP immunopurification. RNA 2007, 13(8):1198-1204. 41. Vella MC, Reinert K, Slack FJ: Architecture of a validated microRNA::target interaction. Chem Biol 2004, 11(12):1619-1623. 42. Pasquinelli AE, Reinhart BJ, Slack F, Martindale MQ, Kuroda MI, Maller B, Hayward DC, Ball EE, Degnan B, Muller P, Spring J, Srinivasan A, Fishman M, Finnerty J, Corbo J, Levine M, Leahy P, Davidson E, Ruvkun G: Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA. Nature 2000, 408(6808):86-89.  143  43. Doench JG, Petersen CP, Sharp PA: siRNAs can function as miRNAs. Genes Dev 2003, 17(4):438-442. 44. Saetrom P, Heale BS, Snove O,Jr, Aagaard L, Alluin J, Rossi JJ: Distance constraints between microRNA target sites dictate efficacy and cooperativity. Nucleic Acids Res 2007, 35(7):2333-2342. 45. Long D, Lee R, Williams P, Chan CY, Ambros V, Ding Y: Potent effect of target structure on microRNA function. Nat Struct Mol Biol 2007, 14(4):287-294. 46. Bagga S, Bracht J, Hunter S, Massirer K, Holtz J, Eachus R, Pasquinelli AE: Regulation by let-7 and lin-4 miRNAs results in target mRNA degradation. Cell 2005, 122(4):553-563. 47. Selbach M, Schwanhausser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N: Widespread changes in protein synthesis induced by microRNAs. Nature 2008, 455(7209):58-63. 48. O'Donnell KA, Wentzel EA, Zeller KI, Dang CV, Mendell JT: c-Myc-regulated microRNAs modulate E2F1 expression. Nature 2005, 435(7043):839-843. 49. Grosshans H, Johnson T, Reinert KL, Gerstein M, Slack FJ: The temporal patterning microRNA let-7 regulates several transcription factors at the larval to adult transition in C. elegans. Dev Cell 2005, 8(3):321-330. 50. Rajewsky N: microRNA target predictions in animals. Nat Genet 2006, 38 Suppl:S8-13. 51. Du T, Zamore PD: microPrimer: the biogenesis and function of microRNA. Development 2005, 132(21):4645-4652. 52. Chen CZ, Li L, Lodish HF, Bartel DP: MicroRNAs modulate hematopoietic lineage differentiation. Science 2004, 303(5654):83-86. 53. Wilfred BR, Wang WX, Nelson PT: Energizing miRNA research: a review of the role of miRNAs in lipid metabolism, with a prediction that miR-103/107 regulates human metabolic pathways. Mol Genet Metab 2007, 91(3):209-217. 54. Ikeda S, Kong SW, Lu J, Bisping E, Zhang H, Allen PD, Golub TR, Pieske B, Pu WT: Altered microRNA expression in human heart disease. Physiol Genomics 2007, 31(3):367-373. 55. Wang WX, Rajeev BW, Stromberg AJ, Ren N, Tang G, Huang Q, Rigoutsos I, Nelson PT: The expression of microRNA miR-107 decreases early in Alzheimer's disease and may accelerate disease progression through regulation of beta-site amyloid precursor protein-cleaving enzyme 1. J Neurosci 2008, 28(5):1213-1223. 56. Suh MR, Lee Y, Kim JY, Kim SK, Moon SH, Lee JY, Cha KY, Chung HM, Yoon HS, Moon SY, Kim VN, Kim KS: Human embryonic stem cells express a unique set of microRNAs. Dev Biol 2004, 270(2):488-498. 57. He L, Hannon GJ: MicroRNAs: small RNAs with a big role in gene regulation. Nat Rev Genet 2004, 5(7):522-531.  144  58. Park W, Li J, Song R, Messing J, Chen X: CARPEL FACTORY, a Dicer homolog, and HEN1, a novel protein, act in microRNA metabolism in Arabidopsis thaliana. Curr Biol 2002, 12(17):1484-1495. 59. Bohmert K, Camus I, Bellini C, Bouchez D, Caboche M, Benning C: AGO1 defines a novel locus of Arabidopsis controlling leaf development. EMBO J 1998, 17(1):170180. 60. Moussian B, Schoof H, Haecker A, Jurgens G, Laux T: Role of the ZWILLE gene in the regulation of central shoot meristem cell fate during Arabidopsis embryogenesis. EMBO J 1998, 17(6):1799-1809. 61. Hatfield SD, Shcherbata HR, Fischer KA, Nakahara K, Carthew RW, Ruohola-Baker H: Stem cell division is regulated by the microRNA pathway. Nature 2005, 435(7044):974-978. 62. Cox DN, Chao A, Baker J, Chang L, Qiao D, Lin H: A novel class of evolutionarily conserved genes defined by piwi are essential for stem cell self-renewal. Genes Dev 1998, 12(23):3715-3727. 63. Grishok A, Pasquinelli AE, Conte D, Li N, Parrish S, Ha I, Baillie DL, Fire A, Ruvkun G, Mello CC: Genes and mechanisms related to RNA interference regulate expression of the small temporal RNAs that control C. elegans developmental timing. Cell 2001, 106(1):23-34. 64. Knight SW, Bass BL: A role for the RNase III enzyme DCR-1 in RNA interference and germ line development in Caenorhabditis elegans. Science 2001, 293(5538):2269-2271. 65. Giraldez AJ, Cinalli RM, Glasner ME, Enright AJ, Thomson JM, Baskerville S, Hammond SM, Bartel DP, Schier AF: MicroRNAs regulate brain morphogenesis in zebrafish. Science 2005, 308(5723):833-838. 66. Wienholds E, Koudijs MJ, van Eeden FJ, Cuppen E, Plasterk RH: The microRNAproducing enzyme Dicer1 is essential for zebrafish development. Nat Genet 2003, 35(3):217-218. 67. Bernstein E, Kim SY, Carmell MA, Murchison EP, Alcorn H, Li MZ, Mills AA, Elledge SJ, Anderson KV, Hannon GJ: Dicer is essential for mouse development. Nat Genet 2003, 35(3):215-217. 68. Lindsay MA: microRNAs and the immune response. Trends Immunol 2008, 29(7):343-351. 69. Baltimore D, Boldin MP, O'Connell RM, Rao DS, Taganov KD: MicroRNAs: new regulators of immune cell development and function. Nat Immunol 2008, 9(8):839845. 70. Lodish HF, Zhou B, Liu G, Chen CZ: Micromanagement of the immune system by microRNAs. Nat Rev Immunol 2008, 8(2):120-130. 71. Li QJ, Chau J, Ebert PJ, Sylvester G, Min H, Liu G, Braich R, Manoharan M, Soutschek J, Skare P, Klein LO, Davis MM, Chen CZ: miR-181a is an intrinsic modulator of T cell sensitivity and selection. Cell 2007, 129(1):147-161. 145  72. Jing Q, Huang S, Guth S, Zarubin T, Motoyama A, Chen J, Di Padova F, Lin SC, Gram H, Han J: Involvement of microRNA in AU-rich element-mediated mRNA instability. Cell 2005, 120(5):623-634. 73. Pedersen IM, Cheng G, Wieland S, Volinia S, Croce CM, Chisari FV, David M: Interferon modulation of cellular microRNAs as an antiviral mechanism. Nature 2007, 449(7164):919-922. 74. Jopling CL, Yi M, Lancaster AM, Lemon SM, Sarnow P: Modulation of hepatitis C virus RNA abundance by a liver-specific MicroRNA. Science 2005, 309(5740):15771581. 75. Stern-Ginossar N, Elefant N, Zimmermann A, Wolf DG, Saleh N, Biton M, Horwitz E, Prokocimer Z, Prichard M, Hahn G, Goldman-Wohl D, Greenfield C, Yagel S, Hengel H, Altuvia Y, Margalit H, Mandelboim O: Host immune system gene targeting by a viral miRNA. Science 2007, 317(5836):376-381. 76. Sood P, Krek A, Zavolan M, Macino G, Rajewsky N: Cell-type-specific signatures of microRNAs on target mRNA expression. Proc Natl Acad Sci U S A 2006, 103(8):2746-2751. 77. Garzon R, Fabbri M, Cimmino A, Calin GA, Croce CM: MicroRNA expression and function in cancer. Trends Mol Med 2006, 12(12):580-587. 78. Stahlhut Espinosa CE, Slack FJ: The role of microRNAs in cancer. Yale J Biol Med 2006, 79(3-4):131-140. 79. Calin GA, Croce CM: MicroRNA signatures in human cancers. Nat Rev Cancer 2006, 6(11):857-866. 80. Calin GA, Sevignani C, Dumitru CD, Hyslop T, Noch E, Yendamuri S, Shimizu M, Rattan S, Bullrich F, Negrini M, Croce CM: Human microRNA genes are frequently located at fragile sites and genomic regions involved in cancers. Proc Natl Acad Sci U S A 2004, 101(9):2999-3004. 81. Iorio MV, Ferracin M, Liu CG, Veronese A, Spizzo R, Sabbioni S, Magri E, Pedriali M, Fabbri M, Campiglio M, Menard S, Palazzo JP, Rosenberg A, Musiani P, Volinia S, Nenci I, Calin GA, Querzoli P, Negrini M, Croce CM: MicroRNA gene expression deregulation in human breast cancer. Cancer Res 2005, 65(16):7065-7070. 82. Lee YS, Kim HK, Chung S, Kim KS, Dutta A: Depletion of human micro-RNA miR-125b reveals that it is critical for the proliferation of differentiated cells but not for the down-regulation of putative targets during differentiation. J Biol Chem 2005, 280(17):16635-16641. 83. Gefen N, Binder V, Zaliova M, Linka Y, Morrow M, Novosel A, Edry L, Hertzberg L, Shomron N, Williams O, Trka J, Borkhardt A, Izraeli S: Hsa-mir-125b-2 is highly expressed in childhood ETV6/RUNX1 (TEL/AML1) leukemias and confers survival advantage to growth inhibitory signals independent of p53. Leukemia 2010, 24(1):8996. 84. Takamizawa J, Konishi H, Yanagisawa K, Tomida S, Osada H, Endoh H, Harano T, Yatabe Y, Nagino M, Nimura Y, Mitsudomi T, Takahashi T: Reduced expression of the  146  let-7 microRNAs in human lung cancers in association with shortened postoperative survival. Cancer Res 2004, 64(11):3753-3756. 85. Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M, Stephens RM, Okamoto A, Yokota J, Tanaka T, Calin GA, Liu CG, Croce CM, Harris CC: Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell 2006, 9(3):189-198. 86. Johnson SM, Grosshans H, Shingara J, Byrom M, Jarvis R, Cheng A, Labourier E, Reinert KL, Brown D, Slack FJ: RAS is regulated by the let-7 microRNA family. Cell 2005, 120(5):635-647. 87. Malumbres M, Barbacid M: RAS oncogenes: the first 30 years. Nat Rev Cancer 2003, 3(6):459-465. 88. Grandori C, Cowley SM, James LP, Eisenman RN: The Myc/Max/Mad network and the transcriptional control of cell behavior. Annu Rev Cell Dev Biol 2000, 16:653-699. 89. Eis PS, Tam W, Sun L, Chadburn A, Li Z, Gomez MF, Lund E, Dahlberg JE: Accumulation of miR-155 and BIC RNA in human B cell lymphomas. Proc Natl Acad Sci U S A 2005, 102(10):3627-3632. 90. Metzler M, Wilda M, Busch K, Viehmann S, Borkhardt A: High expression of precursor microRNA-155/BIC RNA in children with Burkitt lymphoma. Genes Chromosomes Cancer 2004, 39(2):167-169. 91. Kluiver J, Poppema S, de Jong D, Blokzijl T, Harms G, Jacobs S, Kroesen BJ, van den Berg A: BIC and miR-155 are highly expressed in Hodgkin, primary mediastinal and diffuse large B cell lymphomas. J Pathol 2005, 207(2):243-249. 92. Costinean S, Zanesi N, Pekarsky Y, Tili E, Volinia S, Heerema N, Croce CM: Pre-B cell proliferation and lymphoblastic leukemia/high-grade lymphoma in E(mu)miR155 transgenic mice. Proc Natl Acad Sci U S A 2006, 103(18):7024-7029. 93. He L, Thomson JM, Hemann MT, Hernando-Monge E, Mu D, Goodson S, Powers S, Cordon-Cardo C, Lowe SW, Hannon GJ, Hammond SM: A microRNA polycistron as a potential human oncogene. Nature 2005, 435(7043):828-833. 94. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, Sweet-Cordero A, Ebert BL, Mak RH, Ferrando AA, Downing JR, Jacks T, Horvitz HR, Golub TR: MicroRNA expression profiles classify human cancers. Nature 2005, 435(7043):834838. 95. Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning MJ, BarbosaMorais NL, Teschendorff AE, Green AR, Ellis IO, Tavare S, Caldas C, Miska EA: MicroRNA expression profiling of human breast cancer identifies new markers of tumour subtype. Genome Biol 2007, 8(10):R214. 96. Chen X, Ba Y, Ma L, Cai X, Yin Y, Wang K, Guo J, Zhang Y, Chen J, Guo X, Li Q, Li X, Wang W, Zhang Y, Wang J, Jiang X, Xiang Y, Xu C, Zheng P, Zhang J, Li R, Zhang H, Shang X, Gong T, Ning G, Wang J, Zen K, Zhang J, Zhang CY:  147  Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res 2008, 18(10):997-1006. 97. Calin GA, Ferracin M, Cimmino A, Di Leva G, Shimizu M, Wojcik SE, Iorio MV, Visone R, Sever NI, Fabbri M, Iuliano R, Palumbo T, Pichiorri F, Roldo C, Garzon R, Sevignani C, Rassenti L, Alder H, Volinia S, Liu CG, Kipps TJ, Negrini M, Croce CM: A MicroRNA signature associated with prognosis and progression in chronic lymphocytic leukemia. N Engl J Med 2005, 353(17):1793-1801. 98. Schetter AJ, Leung SY, Sohn JJ, Zanetti KA, Bowman ED, Yanaihara N, Yuen ST, Chan TL, Kwong DL, Au GK, Liu CG, Calin GA, Croce CM, Harris CC: MicroRNA expression profiles associated with prognosis and therapeutic outcome in colon adenocarcinoma. JAMA 2008, 299(4):425-436. 99. Yu Z, Willmarth NE, Zhou J, Katiyar S, Wang M, Liu Y, McCue PA, Quong AA, Lisanti MP, Pestell RG: microRNA 17/20 inhibits cellular invasion and tumour metastasis in breast cancer by heterotypic signaling. Proc Natl Acad Sci U S A 2010, 107(18):8231-8236. 100. Valadi H, Ekstrom K, Bossios A, Sjostrand M, Lee JJ, Lotvall JO: Exosomemediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat Cell Biol 2007, 9(6):654-659. 101. Hwang HW, Wentzel EA, Mendell JT: A hexanucleotide element directs microRNA nuclear import. Science 2007, 315(5808):97-100. 102. Weiler J, Hunziker J, Hall J: Anti-miRNA oligonucleotides (AMOs): ammunition to target miRNAs implicated in human disease? Gene Ther 2006, 13(6):496-502. 103. Bumcrot D, Manoharan M, Koteliansky V, Sah DW: RNAi therapeutics: a potential new class of pharmaceutical drugs. Nat Chem Biol 2006, 2(12):711-719. 104. Hutvagner G, Simard MJ, Mello CC, Zamore PD: Sequence-specific inhibition of small RNA function. PLoS Biol 2004, 2(4):E98. 105. Krutzfeldt J, Rajewsky N, Braich R, Rajeev KG, Tuschl T, Manoharan M, Stoffel M: Silencing of microRNAs in vivo with 'antagomirs'. Nature 2005, 438(7068):685689. 106. Mani S, Goel S, Nesterova M, Martin RM, Grindel JM, Rothenberg ML, Zhang R, Tortora G, Cho-Chung YS: Clinical studies in patients with solid tumours using a second-generation antisense oligonucleotide (GEM 231) targeted against protein kinase A type I. Ann N Y Acad Sci 2003, 1002:252-262. 107. Gleave ME, Monia BP: Antisense therapy for cancer. Nat Rev Cancer 2005, 5(6):468-479. 108. Wall EM, Milne K, Martin ML, Watson PH, Theiss P, Nelson BH: Spontaneous mammary tumours differ widely in their inherent sensitivity to adoptively transferred T cells. Cancer Res 2007, 67(13):6442-6450. 109. Li H, Ruan J, Durbin R: Mapping short DNA sequencing reads and calling variants using mapping quality scores. Genome Res 2008, 18(11):1851-1858.  148  110. Rumble SM, Lacroute P, Dalca AV, Fiume M, Sidow A, Brudno M: SHRiMP: accurate mapping of short color-space reads. PLoS Comput Biol 2009, 5(5):e1000386. 111. Griffiths-Jones S, Saini HK, van Dongen S, Enright AJ: miRBase: tools for microRNA genomics. Nucleic Acids Res 2008, 36(Database issue):D154-8. 112. Creighton CJ, Reid JG, Gunaratne PH: Expression profiling of microRNAs by deep sequencing. Brief Bioinform 2009, 10(5):490-497. 113. Hulsen T, de Vlieg J, Alkema W: BioVenn - a web application for the comparison and visualization of biological lists using area-proportional Venn diagrams. BMC Genomics 2008, 9:488. 114. [http://www.targetscan.org/mmu_50/] 115. Morin RD, O'Connor MD, Griffith M, Kuchenbauer F, Delaney A, Prabhu AL, Zhao Y, McDonald H, Zeng T, Hirst M, Eaves CJ, Marra MA: Application of massively parallel sequencing to microRNA profiling and discovery in human embryonic stem cells. Genome Res 2008, 18(4):610-621. 116. [http://www.mirbase.org/] 117. Calin GA, Dumitru CD, Shimizu M, Bichi R, Zupo S, Noch E, Aldler H, Rattan S, Keating M, Rai K, Rassenti L, Kipps T, Negrini M, Bullrich F, Croce CM: Frequent deletions and down-regulation of micro- RNA genes miR15 and miR16 at 13q14 in chronic lymphocytic leukemia. Proc Natl Acad Sci U S A 2002, 99(24):15524-15529. 118. Pekarsky Y, Santanam U, Cimmino A, Palamarchuk A, Efanov A, Maximov V, Volinia S, Alder H, Liu CG, Rassenti L, Calin GA, Hagan JP, Kipps T, Croce CM: Tcl1 expression in chronic lymphocytic leukemia is regulated by miR-29 and miR-181. Cancer Res 2006, 66(24):11590-11593. 119. Akao Y, Nakagawa Y, Naoe T: let-7 microRNA functions as a potential growth suppressor in human colon cancer cells. Biol Pharm Bull 2006, 29(5):903-906. 120. Shell S, Park SM, Radjabi AR, Schickel R, Kistner EO, Jewell DA, Feig C, Lengyel E, Peter ME: Let-7 expression defines two differentiation stages of cancer. Proc Natl Acad Sci U S A 2007, 104(27):11400-11405. 121. Audic S, Claverie JM: The significance of digital gene expression profiles. Genome Res 1997, 7(10):986-995.  149  4 MUTATION PROFILE OF TUMOURS WITH DIFFERENT OUTCOMES AFTER ADOPTIVE T-CELL THERAPY BY HIGH THROUGHPUT TRANSCRIPTOME SEQUENCING  4.1 INTRODUCTION Permanent DNA changes in a genome, known as mutations, are an evolutionary cauldron, providing a resource for new phenotypes and population diversity. The consequences vary from seemingly benign polymorphisms that occur at low frequency in a population, providing genetic diversity, to point mutations that disrupt the function of genes crucial for the normal development such as mutations in the TP53 oncogene [1]. Protein-changing mutations, i.e. non-synonymous mutations, can confer an advantage or disadvantage to the fitness of individual cells depending on how the changed phenotype affects normal cellular processes. An interesting example is mutation occurring in cancer cells, which results in uncontrolled cellular proliferation. In theory, the same mutation could also be disadvantageous by making cancer cells vulnerable to detection by the immune system. Cancer has long been known as a disease of genetic disarray, characterized by genetic mutations and structural changes in the genome. The advent of massively parallel second-generation short-read sequencing technology such as 454 Life Sciences, Illumina/Solexa and SOLiD has enabled unbiased cost-effective characterization of mutations at the global level. Using this sequencing methodology, many studies have profiled cancer-specific mutations in various cancerous tissues such as acute myeloid leukaemia (AML) [2], follicular and diffuse large B-cell lymphomas [3], breast [4],  150  ovary, [5], and lung [6]. The global mutation profiling analyses were also able to putatively differentiate between passenger and driver mutations in cancer [4, 7], identify pathways uniquely affected in cancer [6] and determine pathogenic mutations that have a strong diagnostic power [5]. Such compelling discoveries in cancer biology using massively parallel sequencing technology have propelled the ongoing large international collaborative effort in The Cancer Genome Atlas (TCGA) to profile every mutation in more than 20 different cancer types to better understand tumour biology at the molecular level [8]. Cancer-specific somatic non-synonymous mutations can provide valuable information about genes ultimately responsible for tumourigenesis, cancer biology and progression. Interestingly, genome-wide mutation profiling studies have shown surprisingly varied numbers of cancer-specific mutations. Using the Illumina/Solexa sequencing technology, Ley and colleagues compared the global mutation profiles between AML and matching normal skin samples [2]. They identified 181 single nucleotide variants (SNVs) that differed from the reference genome and validated ten of them using PCR amplicon sequencing by the conventional Sanger method. A similar methodology used by Shah and colleagues identified 6,410 putative somatic non-synonymous mutations in granulose-cell tumours in ovary and found 15 to have mutations that resulted in significant amino acid dissimilarities, which were proceeded for further validations [5]. Interestingly, large-scale mutation profiling studies using other methods such as targeted re-sequencing and single nucleotide polymorphism (SNP) arrays have revealed hundreds to thousands of cancer-specific somatic non-synonymous mutations. From a mere set of 623 genes profiled by targeted re-sequencing of PCR-amplicons from tumour DNA, Ding  151  and colleagues identified 12 genes that were mutated at a significantly higher frequency in lung adenocarcinoma [6]. Using similar sequencing methodology, Jones and colleagues identified 827 somatic non-synonymous mutations in 304MB of surveyed DNA sequence from colorectal cancers [9]. In another similar study by Wood and colleagues, approximately one hundred mutations were identified from 1,718 genes surveyed in colorectal and breast cancers [7]. Another mutation survey of multiple cancer genomes using Affymetrix SNP arrays identified 921 somatic non-synonymous mutations [10]. One of the therapeutic applications of global mutation profiling is to further the development of cancer vaccines against cancer-specific proteins. The immune system detects cells with abnormal protein sequences such as through infection or mutation by recognizing short protein fragments i.e. antigenic epitopes that are presented on the cell surface by major histocompatibility complex (MHC) molecules. T cells recognize these abnormal peptide-MHC complexes through their antigen receptors. This ultimately leads to perforin- or Fas-ligand-induced apoptosis in the target cell initiated by the CD8+ cytotoxic T-cell [11]. The immune system can recognize mutated and abnormally expressed proteins, both of which are hallmark characteristics of cancerous cells. Thus the use of the immune system to prevent or treat cancer has long been investigated. In breast cancer, there are several potential antigens identified for vaccine development, which can: (1) differentiate between cancerous and normal cells by exclusive expression in cancer e.g. NY-BR-1 [12]; (2) be abnormally over-expressed in the cancer cells, such as HER2/neu [13]; and (3) be mutated in cancer e.g. p53 [14]. Not only could the vaccine be more cost-effective than the adoptive immunotherapy in a clinical setting, it  152  could also be used for cancer prevention through prophylactic vaccination, as well as cancer treatment by therapeutic vaccination. Hundreds of cancer vaccines have been evaluated in Phase I, II or III clinical trials [15]. However, the current efficacy of cancer vaccine is low, exemplified by a meta-analysis of vaccine studies in which merely 2.6% of the patients showed objective responses [16]. Certainly, there is plenty of potential to improve the success rate of cancer vaccines. Drugs are being investigated, which pre-condition the host to improve the effect of immunization. The use of a COX2-inhibitor to decrease tumour-induced immune suppression [17], or anti-CD25 antibodies to deplete immunosuppressive regulatory T cells [18] are such examples. Several histone deacetylase inhibitors such as trichostatin A and LAQ824, have been shown to sensitize the host to adoptive immunotherapy and cancer vaccines [19-22]. Other researchers have suggested the direct targeting of “professional” antigen-presenting cells (APC) i.e. dendritic cells (DC) to improve antigen presentation and increase the avidity of immune response [23-25]. An improved success rate was indeed observed in mice when tumour DNA vaccination was combined with adoptive T-cell transfer and host conditioning by irradiation [26]. As one might expect, one of the areas for the biggest improvement in cancer vaccine may be the identification of immunogenic antigens unique to cancer. A recent study showed successful immunization against breast cancer in a mouse model using a vaccine developed to target !-lactalbumin, a breast-specific protein that is normally expressed exclusively during lactation, but is abnormally over-expressed in the majority of human breast cancer patients [27]. Global cancer-specific mutation characterization by sequencing will  153  provide a valuable resource for discovering more and potentially better antigens for novel cancer vaccines.  4.2 MATERIAL AND METHODS 4.2.1 Mouse breast tumour model Tumour samples were generated similarly as described in the previous microarray and microRNA studies. In brief, tumours were grown from subcutaneously implanted tumour cells in the mammary fat pads of transgenic C57BL/6J mice that over-express immunogenic Her2/neu, described in the study by Wall and colleagues [28]. The current C57BL/6J transgenic mouse model carries two transgenes that have been separately created in different mouse strains (C57BL/6J and FVB/NJ) and then subsequently crossed to create the desired genotype of mutant Trp53 and transgenic Her2/neu. One of the mouse strains was a C57BL/6J mouse that carried the transgenic rat Her2/neu oncogene, over-expressed under the control of the MMTV promoter [29]. The rat Her2/neu was tagged with two different epitopes from chicken ovalbumin at the carboxyl-terminus, each of which was recognized by either CD4+ or CD8+ Tlymphocytes. This line was originally crossed with transgenic FVB/NJ mouse strain that carried the mouse Trp53 gene with a substitution mutation (R172H) whose expression was controlled by the whey acidic protein promoter [29]. To establish this genotype in the C57BL/6J mouse background, the hybrids were backcrossed with the C57BL/6J strain for approximately 10 generations while being selected for the transgenic Her2/neu and mutated Trp53 transgenes.  154  The tumours from the cell lines that were established from the spontaneous mammary tumours in the transgenic mouse model consistently and reproducibly showed distinctive outcomes from the adoptive T-cell transfer immunotherapy, namely complete regression (CR); partial regression (PR) or progressive disease (PD) [28]. Four cell lines were chosen for mutation profiling via sequencing: NOP21CR, NOP12PR, NOP23PR and NOP18PD, which were also subjected to the miRNA profiling experiment described in chapter 3. All cell lines except for NOP12PR were subjected to the microarray experiment described previously. When tumours reached approximately 50 mm2 in diameter, they were harvested and then immediately snap-frozen in the liquid nitrogen to be processed for whole transcriptome profiling in the shotgun sequencing manner known as RNA-seq using SOLiDTM (Supported Oligonucleotide Ligation and Detection) (Applied Biosystems, USA), which is a sequencing technology based on ligation of fluorescent di-nucleotides [30].  4.2.2 RNA-seq library construction and sequencing by ABI SOLiDTM technology mRNA transcripts in tumour samples were processed to profile mutations using SOLiD sequencing technology. Approximately 10µg of total RNA extracted from whole tumours from NOP21CR, NOP12PR, NOP23PR and NOP18PD cell lines was treated with DNaseI for high-quality RNA extraction, indicated by RNA integrity (RI) number greater than 7. The fraction of RNA with poly-adenylated tails was purified using the MACS mRNA Isolation Kit (Miltenyi Biotec, Germany). Double-stranded cDNAs were  155  synthesized from the purified poly-adenylated RNA fragments using Superscript DoubleStranded cDNA Synthesis Kit (Invitrogene, USA). Random hexamer primers from Invitrogen were added at the 3’ end of the cDNA sequence at the concentration of 5µM. The cDNAs were purified and re-suspended in 40µL of nuclease-free TE buffer. The RNA-seq libraries were constructed using the SOLiD standard fragment library construction protocol provided by Applied Biosystems. In brief, the purified cDNAs were sheared by sonication into small fragments of 100 to 110bp in length on average. To allow subsequent ligation to adapter sequences, the 5’ and 3’ ends of sheared cDNAs were ‘polished’ by converting the damaged or protruding ends to 5’-phosphorylated and blunt-ended DNA using End Polishing Enzyme 1 and 2 provided by Applied Biosystems. The polished cDNAs were purified, and then the ends of the cDNAs were ligated with the P1 and P2 adapters provided in the SOLiD Fragment Library Oligos Kit (Applied Biosystems, USA). The ligated cDNAs were size-selected for 150 to 200bp in length and then PCR amplified. The amplified cDNAs were subsequently gel purified using 6% precast TBE PAGE gels, instead of using the PureLink PCR Purification Kit provided by the Applied Biosystems as suggested in the SOLiD RNA-seq library protocol. Also, the Qubit (Invitrogen, USA) was used to quantify the amplified cDNA to determine the concentration of library for SOLiD emulsion PCR instead of quantitative PCR specified in the SOLiD library construction protocol. For the whole transcriptome shotgun sequencing of cDNA libraries, emulsion PCR and subsequent sequencing by ligation using the SOLiD technology were done in the same process as for the cDNA libraries created for miRNA profiling described in the previous  156  miRNA study (see Material and Methods in Chapter 3). As in miRNA sequencing, the length of sequence in colour space was 35bp.  4.2.3 Genome mapping and novel SNV discovery The MAQ alignment tool [31] was used to map SOLiD sequences in colour space to the mouse transcriptome. The reference mouse transcriptome consisted of the genomic sequence of the C57BL/6J strain of Mus musculus (UCSC version mm9/NCBI version 37) and the artificial exon-to-exon junction sequences. The artificial sequences were created by concatenating 50bp of sequence flanking the exon junctions (total 100bp) in known transcripts from the mouse Ensemble genome annotation (version 57) [32], the databases from RefSeq [33], UCSC genome annotation [34], NCBI Aceview [35] and Genscan [36]. MAQ converts the reference mouse transcriptome sequences into colour space sequences to make the alignments, then it searches for an ungapped alignment seed in the first 28bp of the sequence with the lowest number of mismatches in all possible alignments in the genome. A maximum of two mismatches was allowed in the seed. No sequence quality filtering was applied to the original dataset. However, a single quadrant of a run, tra19064_2, from the library MM0470 derived from NOP18PD tumour showed an unusually high number of no-call bases (represented by ‘.’ in the colour space) that originated from a poor fluorescence signal resulting in ambiguous colour calls. MAQ (or any other colour space aligners currently available) does not align sequences with no calls, thus the percentage of sequences mapped to the mouse transcriptome was substantially lower than other sequence runs. The high frequency of poor fluorescence 157  signal was assumed to reflect a poorer quality in overall sequencing. Therefore, the sequences derived from the run, MM0470-tra19064_2, were excluded from further analyses. To detect potential mutations, the SNP discovery pipeline developed by a bioinformatics team at the Genome Sciences Centre (GSC) was used. First, the pipeline searched for potential nucleotide variants, i.e. ungapped mismatches in alignments to the genome, which occurred in the exonic regions of known and predicted genes. Then, the nucleotide variants in the exons were compared to the NCBI mouse SNP database, dbSNP (build 130; NTotal_mouse_SNP = 14,715,422), to identify novel variants that were potential mutations. To ensure that the nucleotide variants identified were not technical artifacts, the mapping quality from MAQ, the measure of confidence of the alignment, needed to be greater than 10. A mapping quality greater than 10 was also determined empirically to indicate an unambiguous mapping to the genome where a sequence can be mapped to a unique genomic location. The sequences that supported a particular single nucleotide variant (SNV) must also be of high sequencing quality (phred-like quality > 20) at the SNV position. Furthermore, at least one third of the total sequences aligned with high sequencing quality, also with the minimum of three sequences, was needed to support the SNV. Finally, SNVs that caused a putative amino acid change in final protein products were identified. Potentially problematic SNVs were also identified by the pipeline and excluded from further analyses, namely the ‘non-ref’ and ‘junction-only’ SNVs. The ‘non-ref’ SNVs refer to those that do not have any supporting base in the aligned sequences that agreed with the reference genome. The ‘junction-only’ SNVs refer to those in which all  158  supporting sequences were aligned to only the artificial junction sequences, and none to the reference genome.  4.2.4 Identifying the chromosomal segment that originated from the FVB mouse strain Mouse SNP data were obtained from the inbred laboratory mouse haplotype map (hapmap) generated by the Broad Institute of Harvard and MIT [37]. The mouse hapmap data contained the genotypes probed by the Affymetrix mouse SNP microarrays and the confidence scores associated with the genotypes. In the mouse hapmap data, a total of 132,286 SNPs were queried from 94 laboratory mouse strains. The probes on the SNP microarrays were designed from the mouse genome assembly of NCBI build 33 (UCSC mm9). SNPs with confidence score less than 0.3 (0 being the best confidence score) were considered for further analyses. From the 132,286 SNPs from the FVB/NJ mouse strain in the Broad dataset, 115,994 high confidence SNPs were identified to pass the confidence score threshold. The FVB SNPs from the Broad hapmap data were compared with SNVs identified in the mouse tumours using a custom Perl script. Those that matched the locations of the FVB SNPs were mapped to the mouse chromosomes. A custom Perl script was used to visualize the mapped FVB SNPs in mouse chromosomes. Chromosome 9 was found to show consistent clusters of FVB SNPs at one of its chromosomal arms across all four tumour samples, and the most downstream position before which no FVB SNPs was found, was determined (63,801,391st base of the chromosome 9) from all four tumours sequenced. The SNVs identified in the tumours beyond this position on the chromosome 9 were discarded from further analyses because 159  they were likely to be nucleotide variants from the FVB mouse strain, which had no relevance to tumour biology investigated in this study. The probability of finding FVB SNPs in tumour samples by chance was calculated using the phyper function in R.  4.2.5 Pathway enrichment and functional enrichment analysis The genes that were mutated exclusively in a single outcome group, CR, PR or PD, were subjected to pathway enrichment analysis through the use of the Ingenuity Pathway Analysis (Ingenuity® Systems, www.ingenuity.com). The enrichment analysis is done by calculating the probability of finding a particular number of mutated genes from the tumour samples in a particular pathway (or with a particular biological function) given the total number of genes involved in the pathway annotated in the Ingenuity database using the Fisher’s exact test. The p-values were adjusted for the multiple hypothesis testing by Benjamini-Hochberg method [38].  4.3 RESULTS 4.3.1 Sequencing and mapping of short RNA sequences generated by SOLiD Mutation profiles were assessed from tumours that exhibit consistent and reproducible outcomes of CR, PR and PD. Whole tumour tissues that originated from cell lines representing NOP21CR, NOP12PR, NOP23PR and NOP18PD were subjected to RNA-seq using the SOLiD technology, in which mRNAs were captured, reverse-transcribed to  160  cDNA and then sequenced in a series of ligation cycles that resulted in colour space sequences of 35bp in length, with each colour-space base representing the transition of two specific adjacent nucleotides. Using the MAQ alignment program, the sequences were aligned to the mouse transcriptome represented by the mouse genomic sequences and the artificial junction sequences consisting of exon-to-exon junctions in all known transcripts in mouse. A total of 653,572,234 sequences of 35bp in length were generated from sequencing four tumour samples, resulting in 62.6X coverage of the transcriptome i.e. the sum of bases sequenced divided by the total length of known transcripts. On average, 53.74% of sequences were mapped to the mouse transcriptome; 77.4% of which were mapped to unique locations on average whereas 19.9% and 2.7% were mapped to multiple loci or resulted in poor alignments respectively (Table 4.1).  161  Table 4.1 The mapping rate of sequences derived from the mRNA transcriptomes in the RNA-seq experiment using SOLiD The sequences were aligned to the mouse genome using the Maq aligner.  ! !  !  162  ! !  In the SOLiD platform, up to four different samples can be loaded on a single glass slide on which the sequencing performs. The tra and the pla numbers refer to a glass slide, followed by a quadrant number. A ‘unique match’ represents unambiguous mapping in which a sequence is mapped to a single location in the transcriptome. Conversely, ‘multiple match’ represents ambiguous mapping to multiple locus. A ‘poor match’ represents the alignment with the low confidence that suspects random alignments to the transcriptome. The percentage of mapped sequences were calculated against the total number of sequences whereas other percentages such as ‘% unique match’ and ‘% poor match’, were calculated against the total number of sequences mapped to the mouse transcriptome. A ‘no call sequence’ represents a sequence that contains ambiguous colour calls, which could not be processed by the MAQ alignment algorithm. The sequences from NOP18PD-MM0470/tra19064_2 (coloured in red) were removed from the further analyses due to poor sequencing qualities resulting from too many no calls.  The average transcript coverage, i.e. the sum of sequences aligned under each base over the total length of transcripts, was 45.2X across all four tumour libraries (42.32X, 40.66X, 51.57X, 46.11X for NOP21CR, NOP23PR, NOP12PR, NOP18PD libraries respectively). Similarly, the average exon coverage, i.e. the sum of sequences aligned under each base over the total length of exons, was 55.3X (47.65X, 49.43X, 66.57X, 57.40X for NOP21CR, NOP23PR, NOP12PR, NOP18PD libraries respectively). Interestingly, an average of 37.1% of mouse transcripts (28.4% of exons) were not detected, thus likely to be rarely expressed in the sequenced tumours, which was indicated by 0X coverage for these transcripts. The sequences that were mapped to unique genomic loci were considered for further analysis. The in-house SNP pipeline was used to identify high-quality SNVs that were likely to be putative non-synonymous mutations in the tumour genome. The ungapped ‘mismatches’ in the alignment between the uniquely mapped sequences to the mouse transcriptome were considered as putative SNVs. Various filtering processes were applied to the putative mutations described in detail in the Material and Methods, such as the sequence quality of each base, the number of high-quality sequences supporting the SNVs, and the detection of possible alignment artefacts. After the sequence quality filtering was applied, a total of 7,807 SNVs was identified from the four tumour samples, of which 4,077 were non-synonymous SNVs.  4.3.2 Discovery of an FVB chromosomal segment in the C75BL/6J genome During the construction of the transgenic mouse model, the FVB mouse strain that carried the mutated Trp53 gene was crossed with the C57BL/6J mouse strain that carried 163  the transgenic rat Her2/neu gene fused with immunogenic epitopes. Although the C57BL/FVB hybrid strains were back-crossed with the C57BL/6J strain multiple times to achieve the nearly pure C57BL/6J genetic background, the possibility remained that the remnant FVB chromosomal segments could still be present in the back-crossed C57BL/6J mouse genome. Any nucleotide change identified within the genomic segments originating from the FVB strain would be unlikely to be relevant to our study and would contribute false positive mutations. To identify FVB chromosomal segments, SNVs identified from the tumours were compared with known SNPs from the FVB mouse strain from the mouse hapmap data from the Broad institute [37]. Hypergeometric testing showed that the SNVs identified from the tumours were significantly enriched for FVB SNPs (p-value << 0.01). In the NOP21CR tumour, 427 SNVs were identical to those found in the mouse hapmap data, of which 23 belonged to the FVB strain (p-value = 1.8e-13). For NOP12PR and NOP23PR tumours, 27 out of 288 SNVs and 9 out of 301 SNVs were from the FVB strain respectively (p-value = 1.0e-10; 0.00017 respectively). In NOP18PD, 129 out of 3898 SNVs were from the FVB strain (p-value = 5.8e-41). The SNVs from the four tumour samples that were identical to those from the Broad mouse hapmap data were visually inspected by mapping to the chromosomes to identify chromosomal segment(s) that possibly originated from the FVB strain (Figure 4.1).  Figure 4.1 Tumour SNVs that matched SNPs from the Broad mouse Hapmap data generated from 94 laboratory mouse strains using the Affymetrix SNP arrays The approximate location of a SNP from the FVB/NJ strain is coloured in pink whereas the blue line indicates a SNP commonly found in one or more non-FVB mouse strains. Each line consists of one pixel in thickness which represents a genomic region of 355kb in length according to the scale drawn here: (a) SNPs from NOP21CR tumour; (b) SNPs from NOP23PR tumour ; (c) SNPs from NOP12PR tumour ; (d) SNPs from NOP18PD tumour.  164  (a)  (b)  165  (c)  (d)  166  Indeed, there was a cluster of FVB SNPs found in the second half of chromosome 9, which was consistently observed across all four tumour samples. In NOP18PD tumour, the cluster of FVB SNPs was visually less apparent than the CR and PR tumours because the number of SNPs detected in PD tumour was 5 to 10 times higher than the other tumours. Nonetheless, the largest frequency of FVB SNPs were found in chromosome 9 (N = 22) in the NOP18PD tumour compared to other chromosomes. SNPs from other mouse strains were also found to be concentrated in the second half of the chromosome 9 where most of the FVB SNPs were found, especially in NOP21CR, NOP12PR and NOP23PR tumours. Even if the region is highly polymorphic across different mouse strains and SNPs are not specifically from the FVB mouse, the SNVs from this region are likely to be identified as false positive mutations. Therefore, the region beyond the most downstream of where the FVB SNPs were found was suspected as the remnant chromosomal segment originated from the FVB strain. Thus, the SNVs found in these loci were excluded from further analyses.  4.3.3 Putative mutation discovery pipeline High-quality non-FVB SNVs were filtered further to identify potential mutations. Novel SNVs that had not been previously identified as SNP, implied that the nucleotide difference was unique to our tumour model and likely to be a somatically accrued mutation. Hence, the novel SNVs that caused amino acid changes in the regions annotated as exons in known transcripts were selected as potential non-synonymous mutations (Figure 4.2).  167  Figure 4.2 Filtering process to discover putative non-synonymous mutations from tumours subjected to RNA-seq Transcriptomes from four tumours with CR, PR (N=2), PD outcomes after the adoptive immunotherapy were profiled to discover putative mutations. The numbers indicate the individual nucleotide variants detected. Multiple variants can be found in a single gene. The total number of genes that carry putative mutations is noted in brackets in the last step.  The PD tumour initially had the most number of SNVs detected (N = 4,871). The numbers of SNVs found in two PR tumours, NOP12 and NOP23, were substantially different initially. However, the final number of putative non-synonymous mutations did not differ much amongst the tumours. After the various filtering steps, a total of 110 SNVs, possible as putative non-synonymous mutations, from 97 unique genes was identified in CR; 85 SNVs in 74 genes in PRNOP23; 79 SNVs in 61 genes PRNOP12 and 167 SNVs in 142 genes PD tumours. Genes with the identified SNVs were compared across the four tumour samples (Figure 4.3). 168  Figure 4.3 Comparisons of genes that carried the putative non-synonymous mutations detected in tumours from four cell lines sequenced.  Trp53 and Her2 were included in the seven genes (Pisd, D030013I16Rik, Ckap5, ENSMUSG00000058140, ENSMUGS00000074482, Trp53, Erbb2/Her2/Neu) that were found to be commonly mutated in all four tumour samples. These two genes were artificially mutated during the construction of the transgenic mouse model. To accelerate the rate of spontaneous tumourigenesis, Trp53 had a substitution mutation, in which the 172nd amino acid, arginine (nucleotide codon = CGC/CGU(T)), was changed to histidine (nucleotide codon = CAC/CAU(T)). Indeed, the substitution mutation of CAC in the Trp53 transcripts, which differed from CGC in the mouse reference genome, were verified in all four tumours sequenced, validating the accuracy of sequencing by the SOLiD technology and of the subsequent bioinformatics approach to identify putative non-synonymous mutations. The mouse model also contained the fusion gene of  169  Her2/neu from rat and two epitopes that originated from chicken ovalbumin. The BLASTP alignment between mouse Her2/neu and rat Her2/neu proteins revealed the protein sequence identity of 95% with a difference of 56 amino acids. At the nucleotide level, the BLASTN alignment showed 92% in sequence identity with a difference of 374 nucleotides. The SNP discovery pipeline would identify the nucleotide difference between the mouse and rat Her2/neu as putative mutations. The analysis identified 15 mutations in Her2/neu from tumour samples. 13 of them were verified as the normal nucleotide difference between rat and mouse Her2/neu genes, 9 of which were commonly found in all four tumours samples whereas the rest were found in NOP12PR and NOP18PD tumours. The other two mutations, each of which was also found in NOP12PR or NOP18PD tumours, were novel nucleotide changes in rat and mouse Her2/neu genes. The mutation found in Her2/neu from NOP12PR occurred at the 1229th amino acid, and truncated the Her2/neu protein by altering serine to a stop. According to the Uniprot database [39], the region containing the mutation belongs to the receptor tyrosine-protein kinase chain in the cytoplasmic domain in Her2/neu where specific tyrosine residues become phosphorylated by receptor tyrosine kinases, which trigger powerful intracellular signal transduction pathways for cell proliferation. The mutation found in NOP18PD changes the normal glutamic acid to lysine at the 211st amino acid in Her2/neu. According to the Uniprot annotation [39], the region that contains this mutation belongs to the extracellular domain, Domain-I, next to Domain-II where the dimerization occurs. The functional importance of this domain is unknown. Genes that were commonly mutated in the response outcome groups of tumours, i.e. CR and PR, were Herpud2, Arhgef3, Fam160b2, Hp1bp3, Plk1, Afap1l1 and Pmch.  170  The majority of the mutations identified were unique to the tumours, with few shared amongst four tumours. The genes that carried CR-specific and PD-specific mutations were subjected to pathway enrichment and functional enrichment analyses to detect the functional relevance of mutated genes. The top five pathways (Table 4.2) and biological functions (Table 3) ranked by p-values were identified using the Ingenuity Pathway Analysis tool. However, after the multiple hypothesis testing correction was applied to the p-values, none of the pathways were found to be enriched for mutated genes found by the RNA-seq experiment (Table 2).  Table 4.2 Top ranking pathways enriched for mutated genes The top ranking pathways were enriched for genes mutated exclusively in (a) NOP21CR tumour and (b) exclusively in NOP18PD using the Ingenuity pathway analysis tool. The p-values were corrected for multiple hypothesis testing. (a)  171  (b)  A small number of genes that carried mutations found exclusively in the NOP21CR tumour were significantly associated with specific biological functions (multiplehypothesis testing corrected p-value < 0.05) including post-transcriptional modification of RNA, intracellular signaling and cancer-related functions (Table 4.3).  Table 4.3 Top ranking biological annotation significantly associated with mutated genes The top ranking biological annotation that significantly associated with genes mutated exclusively in (a) NOP21CR tumour and exclusively in (b) NOP18PD were identified using the Ingenuity pathway analysis tool. The p-values were corrected for multiple hypothesis testing. (a)  172  (b)  4.3.4 Validation of selected putative mutations by PCR amplicon sequencing Validation of the putative mutations was independently carried out by Spencer Martin at the Trev & Joyce Deeley Research Centre, BCCA [40]. Briefly, 18 putative mutations found exclusively within the NOP21CR tumour were selected for PCR-assisted validation, based on their strong potential as cancer epitopes predicted by the NetMHC program [4143] and the Immune Epitope Database and Analysis Resource (IEDB) (http://www.immuneepitope.org) (Table 4.4). PCR primers were designed from the genomic DNA sequences flanking 100bp upstream and downstream of the SNVs. PCR amplicons of the genomic DNAs that contained the putative SNVs were sequenced using conventional capillary-based Sanger sequencing methodology. To verify a somatic mutation, normal tissues from the tail tips of the transgenic NOP mouse and the wild type C57BL/6J mouse were additionally sequenced to compare with the SNVs detected in the NOP21CR tumour.  173  Table 4.4 List of genes and putative mutations attempted for experimental validation using PCR The codons are shown in the 5’ to 3’ orientation with respect to the orientation of the gene. The supporting base is shown as it appears on the sense (‘positive’ or ‘Watson’) strand of the genome.  174  In summary, 17 out of 18 SNVs were not validated by PCR-amplicon sequencing using Sanger sequencing. The PCR-amplicon sequencing revealed that the majority were false positives where the tumours had the same nucleotide in the NOP21CR tumour as the C57BL/6J mouse reference genome at the site of mutation detected by the RNA-seq experiment. The SNV from only one gene, Adss, was validated. However, unlike the bioinformatics analyses which predicted the SNVs to be homozygous (23 sequences with ‘A’ versus 0 sequence with ‘G’ at the mutation loci), the locus was heterozygous with ‘A’ and ‘G’ in the NOP21CR genome. Furthermore, the sequences derived from the normal tissues also revealed that the same locus was homozygous for ‘G’ and ‘A’ in the tail tips of the mouse model and the wild type C57BL/6J respectively, indicating that the putative ‘mutation’ discovered at this locus was most likely a novel SNP. Therefore, none of the 18 putative mutations identified from the analyses were validated to be true somatic mutations.  4.4 DISCUSSION The average mapping rate of 54% of sequences generated from the transcriptomes of four tumour samples, i.e. NOP21CR, NOP12PR, NOP23PR, NOP18PD, was lower than what was normally observed from other RNA-seq experiments using Illumina/Solexa technology (approximately 75%, from a personal correspondence with the bioinformatics team at the GSC). Considering that these libraries were among the earliest ones constructed using the SOLiD RNA-seq protocols, the suboptimal library quality was expected, which  175  reflected the lower mapping rate observed here. More importantly, MAQ may not be the best alignment tool for this dataset because the accuracy of the MAQ algorithm in aligning sequences in colour space has not been rigorously tested by independent studies. Nonetheless, the total number of sequences available for mutation analyses was nearly 350 million sequences of 35bp in length, from a single run of each sample, which was more than enough to cover the diploid human genome twice (Ntotal_bases_sequenced= 12,226,305,630). The comparison between known SNPs from the FVB mouse strain with the putative mutation identified through a bionformatics analysis revealed that the second half of the chromosome 9 in the mouse model genome might have originated from the FVB mouse strain, which was initially crossed with the C57BL/6J strain to achieve the desirable genotype. The presence of the FVB remnant chromosomal segment can be due to the possible incorporation of the mutated Trp53 gene to this region, which confers the selection advantage to be kept throughout the history of the mouse model construction. The total number of SNPs detected in the tumour samples was surprisingly large, ranging from 158 (NOP23PR) to 2,166 (NOP18PD) in the coding regions. The tumours were derived from highly inbred isogenic mice. Also, the di-base encoding in the SOLiD technology renders each base to be interrogated twice, establishing an innate error checking process during sequencing. Thus, the number of false positives expected to be seen using the SOLiD technology was less, even more so for the SNPs detected. Certainly, a high rate of false positive discovery from the analysis is the most likely reason for the unexpectedly many SNPs identified. However, a finding from a recent study suggests that this observation may be normal for highly inbred mice. A study by  176  Watkins-Chow and colleagues who investigated genomic copy number variants (CNVs) within the C57BL/6J inbred mouse strain revealed that a substantially large proportion (64%) of C57BL/6J mice were heterozygous for the CNVs [44]. This work demonstrated that individual mice within a highly inbred population might not be completely isogenic, which may explain the large number of SNPs observed from the CR, PR and PD tumours sequenced in this study. The SNV identification analysis revealed that the PD tumour was once again most varied from the CR and PR tumours, as seen from the gene expression and miRNA profiling experiments in the previously described studies. The PD tumour was presented with the most number of SNVs compared to the CR and PR tumours. The most number of SNVs in the PD tumour can be a possible indication of the lack of chromosomal stability due to non-functional DNA repair mechanisms compared to tumours that showed response to the adoptive T-cell transfer immunotherapy i.e. CR and PR tumours. Interestingly, the overall number of putative mutations did not differ very much amongst the tumour classes, ranging from 79 (NOP12PR) to 167 (PD). Even the CR tumour had a relatively comparable number of putative mutations (N = 110) to the PD tumour. A highly speculative theory can be derived from this observation that the PD tumour may have a higher mutation rate as a result of a relatively more disarrayed cellular state compared to other tumours, and that mutations responsible for cancer-specific phenotypes or for the PD outcome of adoptive immunotherapy are rare and most likely to be commonly found across NOP tumours. The comparison of genes that carried putative mutations showed that surprisingly, the majority of mutated genes were unique to individual tumours with very few mutated  177  genes commonly found across all four tumour samples. Even in two PR tumour samples, only one third of the genes with putative mutations were shared between the two tumours, implicating that there was much heterogeneity among individual tumours despite their similar phenotype and a supposedly nearly identical original genetic background. Although the known mutation in the Trp53 gene and the differences between rat and mouse Her2/neu genes were verified among the putative mutations discovered, only one of the SNVs in the 18 selected genes validated as non-synonymous mutations. No SNVs were validated as somatic non-synonymous mutations. A low validation rate was seen in a similar study by Ley and colleagues [2] where 10 out of 181 putative mutations were validated as true somatic non-synonymous mutations by PCR amplicon sequencing. However, the majority of false positives are likely to result from the bioinformatics analyses used to identify the putative mutations. The number of false positives could be reduced by the use of better alignment and SNP discovery algorithms that are more optimized for color space sequences such as SHRiMP [45] and rna2map (Applied Biosystems) and for detecting SNVs in a transcriptome where the dynamic range of the depth of sequencing is heavily dependent on the abundance of transcripts that highly varies from gene to gene especially in cancer [46]. A novel SNP discovery tool, SNVMix based on a probabilistic binomial mixture model [46], was shown to successfully identify somatic non-synonymous mutations from cancer transcriptomes in recent studies [4, 5] and could be a good candidate SNP discovery pipeline for future studies.  178  No pathway was found to be significantly enriched for genes identified to have putative mutations. The functional enrichment analysis also resulted in insignificant findings. This may not be a surprising result since it is extremely unlikely for mutations to occur in a targeted manner, affecting a group of genes involved in a common pathway. Rather, a mutation in a single key gene can have more impact on overall cellular activities than multiple mutated genes that are in the periphery of a pathway network and do not interact with other proteins. Thus, the enrichment analysis based on a small number of genes is likely to produce a statistically insignificant result. The false positives in the dataset can also confound the enrichment analysis result, shown by the failure of putative mutations to experimentally validate by PCR amplicon sequencing. Although the mutation profiles in this study did not present significant findings, the attempt taken to profile global mutations through a high-throughput RNA-seq methodology followed by bioinformatics analyses to identify putative mutations was a scientifically advanced and valid approach which should be pursued further with more optimized and rigorously tested tools for more accurate alignment and SNV discovery. The sequencing of the normal host genome would also increase the accuracy of somatic mutation discovery. The dataset generated from this experiment can serve as a resource for future mutation profiling or gene expression profiling studies. Furthermore, tumours with the isogenic background that present with different responses to adoptive immunotherapy still provides an unprecedented opportunity to clearly characterize the molecular mechanisms underlying the immune evasion by cancer cells and provide a novel resource for possible cancer vaccines.  179  4.5 BIBLIOGRAPHY 1. Whibley C, Pharoah PD, Hollstein M: P53 Polymorphisms: Cancer Implications. Nat Rev Cancer 2009, 9(2):95-107. 2. Ley TJ, Mardis ER, Ding L, Fulton B, McLellan MD, Chen K, Dooling D, DunfordShore BH, McGrath S, Hickenbotham M, Cook L, Abbott R, Larson DE, Koboldt DC, Pohl C, Smith S, Hawkins A, Abbott S, Locke D, Hillier LW, Miner T, Fulton L, Magrini V, Wylie T, Glasscock J, Conyers J, Sander N, Shi X, Osborne JR, Minx P, Gordon D, Chinwalla A, Zhao Y, Ries RE, Payton JE, Westervelt P, Tomasson MH, Watson M, Baty J, Ivanovich J, Heath S, Shannon WD, Nagarajan R, Walter MJ, Link DC, Graubert TA, DiPersio JF, Wilson RK: DNA sequencing of a cytogenetically normal acute myeloid leukaemia genome. Nature 2008, 456(7218):66-72. 3. Morin RD, Johnson NA, Severson TM, Mungall AJ, An J, Goya R, Paul JE, Boyle M, Woolcock BW, Kuchenbauer F, Yap D, Humphries RK, Griffith OL, Shah S, Zhu H, Kimbara M, Shashkin P, Charlot JF, Tcherpakov M, Corbett R, Tam A, Varhol R, Smailus D, Moksa M, Zhao Y, Delaney A, Qian H, Birol I, Schein J, Moore R, Holt R, Horsman DE, Connors JM, Jones S, Aparicio S, Hirst M, Gascoyne RD, Marra MA: Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat Genet 2010, 42(2):181-185. 4. Shah SP, Morin RD, Khattra J, Prentice L, Pugh T, Burleigh A, Delaney A, Gelmon K, Guliany R, Senz J, Steidl C, Holt RA, Jones S, Sun M, Leung G, Moore R, Severson T, Taylor GA, Teschendorff AE, Tse K, Turashvili G, Varhol R, Warren RL, Watson P, Zhao Y, Caldas C, Huntsman D, Hirst M, Marra MA, Aparicio S: Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature 2009, 461(7265):809-813. 5. Shah SP, Kobel M, Senz J, Morin RD, Clarke BA, Wiegand KC, Leung G, Zayed A, Mehl E, Kalloger SE, Sun M, Giuliany R, Yorida E, Jones S, Varhol R, Swenerton KD, Miller D, Clement PB, Crane C, Madore J, Provencher D, Leung P, DeFazio A, Khattra J, Turashvili G, Zhao Y, Zeng T, Glover JN, Vanderhyden B, Zhao C, Parkinson CA, Jimenez-Linan M, Bowtell DD, Mes-Masson AM, Brenton JD, Aparicio SA, Boyd N, Hirst M, Gilks CB, Marra M, Huntsman DG: Mutation of FOXL2 in granulosa-cell tumors of the ovary. N Engl J Med 2009, 360(26):2719-2729. 6. Ding L, Getz G, Wheeler DA, Mardis ER, McLellan MD, Cibulskis K, Sougnez C, Greulich H, Muzny DM, Morgan MB, Fulton L, Fulton RS, Zhang Q, Wendl MC, Lawrence MS, Larson DE, Chen K, Dooling DJ, Sabo A, Hawes AC, Shen H, Jhangiani SN, Lewis LR, Hall O, Zhu Y, Mathew T, Ren Y, Yao J, Scherer SE, Clerc K, Metcalf GA, Ng B, Milosavljevic A, Gonzalez-Garay ML, Osborne JR, Meyer R, Shi X, Tang Y, Koboldt DC, Lin L, Abbott R, Miner TL, Pohl C, Fewell G, Haipek C, Schmidt H, Dunford-Shore BH, Kraja A, Crosby SD, Sawyer CS, Vickery T, Sander S, Robinson J, Winckler W, Baldwin J, Chirieac LR, Dutt A, Fennell T, Hanna M, Johnson BE, Onofrio RC, Thomas RK, Tonon G, Weir BA, Zhao X, Ziaugra L, Zody MC, Giordano T, Orringer MB, Roth JA, Spitz MR, Wistuba II, Ozenberger B, Good PJ, Chang AC, Beer DG, Watson MA, Ladanyi M, Broderick S, Yoshizawa A, Travis WD, Pao W, Province  180  MA, Weinstock GM, Varmus HE, Gabriel SB, Lander ES, Gibbs RA, Meyerson M, Wilson RK: Somatic mutations affect key pathways in lung adenocarcinoma. Nature 2008, 455(7216):1069-1075. 7. Wood LD, Parsons DW, Jones S, Lin J, Sjoblom T, Leary RJ, Shen D, Boca SM, Barber T, Ptak J, Silliman N, Szabo S, Dezso Z, Ustyanksky V, Nikolskaya T, Nikolsky Y, Karchin R, Wilson PA, Kaminker JS, Zhang Z, Croshaw R, Willis J, Dawson D, Shipitsin M, Willson JK, Sukumar S, Polyak K, Park BH, Pethiyagoda CL, Pant PV, Ballinger DG, Sparks AB, Hartigan J, Smith DR, Suh E, Papadopoulos N, Buckhaults P, Markowitz SD, Parmigiani G, Kinzler KW, Velculescu VE, Vogelstein B: The genomic landscapes of human breast and colorectal cancers. Science 2007, 318(5853):11081113. 8. [http://cancergenome.nih.gov/index.asp] 9. Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, Antal T, Traulsen A, Nowak MA, Siegel C, Velculescu VE, Kinzler KW, Vogelstein B, Willis J, Markowitz SD: Comparative lesion sequencing provides insights into tumor evolution. Proc Natl Acad Sci U S A 2008, 105(11):4283-4288. 10. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G, Davies H, Teague J, Butler A, Stevens C, Edkins S, O'Meara S, Vastrik I, Schmidt EE, Avis T, Barthorpe S, Bhamra G, Buck G, Choudhury B, Clements J, Cole J, Dicks E, Forbes S, Gray K, Halliday K, Harrison R, Hills K, Hinton J, Jenkinson A, Jones D, Menzies A, Mironenko T, Perry J, Raine K, Richardson D, Shepherd R, Small A, Tofts C, Varian J, Webb T, West S, Widaa S, Yates A, Cahill DP, Louis DN, Goldstraw P, Nicholson AG, Brasseur F, Looijenga L, Weber BL, Chiew YE, DeFazio A, Greaves MF, Green AR, Campbell P, Birney E, Easton DF, Chenevix-Trench G, Tan MH, Khoo SK, Teh BT, Yuen ST, Leung SY, Wooster R, Futreal PA, Stratton MR: Patterns of somatic mutation in human cancer genomes. Nature 2007, 446(7132):153-158. 11. Janeway C: Immunobiology : the immune system in health and disease: 6th ed. New York: Garland Science; 2005. 12. Jager D, Stockert E, Gure AO, Scanlan MJ, Karbach J, Jager E, Knuth A, Old LJ, Chen YT: Identification of a tissue-specific putative transcription factor in breast tissue by serological screening of a breast cancer library. Cancer Res 2001, 61(5):2055-2061. 13. Disis ML, Calenoff E, McLaughlin G, Murphy AE, Chen W, Groner B, Jeschke M, Lydon N, McGlynn E, Livingston RB: Existent T-cell and antibody immunity to HER-2/neu protein in patients with breast cancer. Cancer Res 1994, 54(1):16-20. 14. Runnebaum IB, Nagarajan M, Bowman M, Soto D, Sukumar S: Mutations in p53 as potential molecular markers for human breast cancer. Proc Natl Acad Sci U S A 1991, 88(23):10657-10661. 15. Ribas A, Butterfield LH, Glaspy JA, Economou JS: Current developments in cancer vaccines and cellular immunotherapy. J Clin Oncol 2003, 21(12):2415-2432. 16. Rosenberg SA, Yang JC, Restifo NP: Cancer immunotherapy: moving beyond current vaccines. Nat Med 2004, 10(9):909-915.  181  17. Sharma S, Stolina M, Yang SC, Baratelli F, Lin JF, Atianzar K, Luo J, Zhu L, Lin Y, Huang M, Dohadwala M, Batra RK, Dubinett SM: Tumor cyclooxygenase 2-dependent suppression of dendritic cell function. Clin Cancer Res 2003, 9(3):961-968. 18. Attia P, Maker AV, Haworth LR, Rogers-Freezer L, Rosenberg SA: Inability of a fusion protein of IL-2 and diphtheria toxin (Denileukin Diftitox, DAB389IL-2, ONTAK) to eliminate regulatory T lymphocytes in patients with melanoma. J Immunother 2005, 28(6):582-592. 19. Insinga A, Minucci S, Pelicci PG: Mechanisms of selective anticancer action of histone deacetylase inhibitors. Cell Cycle 2005, 4(6):741-743. 20. Vanniasinkam T, Ertl H, Tang Q: Trichostatin-A enhances adaptive immune responses to DNA vaccination. J Clin Virol 2006, 36(4):292-297. 21. Vo DD, Prins RM, Begley JL, Donahue TR, Morris LF, Bruhn KW, de la Rocha P, Yang MY, Mok S, Garban HJ, Craft N, Economou JS, Marincola FM, Wang E, Ribas A: Enhanced antitumor activity induced by adoptive T-cell transfer and adjunctive use of the histone deacetylase inhibitor LAQ824. Cancer Res 2009, 69(22):8693-8699. 22. Setiadi AF, Omilusik K, David MD, Seipp RP, Hartikainen J, Gopaul R, Choi KB, Jefferies WA: Epigenetic enhancement of antigen processing and presentation promotes immune recognition of tumors. Cancer Res 2008, 68(23):9601-9607. 23. Mule JJ: Dendritic cells: at the clinical crossroads. J Clin Invest 2000, 105(6):707708. 24. Murphy GP, Tjoa BA, Simmons SJ, Jarisch J, Bowes VA, Ragde H, Rogers M, Elgamal A, Kenny GM, Cobb OE, Ireton RC, Troychak MJ, Salgaller ML, Boynton AL: Infusion of dendritic cells pulsed with HLA-A2-specific prostate-specific membrane antigen peptides: a phase II prostate cancer vaccine trial involving patients with hormone-refractory metastatic disease. Prostate 1999, 38(1):73-78. 25. Nestle FO, Alijagic S, Gilliet M, Sun Y, Grabbe S, Dummer R, Burg G, Schadendorf D: Vaccination of melanoma patients with peptide- or tumor lysate-pulsed dendritic cells. Nat Med 1998, 4(3):328-332. 26. Jorritsma A, Bins AD, Schumacher TN, Haanen JB: Skewing the T-cell repertoire by combined DNA vaccination, host conditioning, and adoptive transfer. Cancer Res 2008, 68(7):2455-2462. 27. Jaini R, Kesaraju P, Johnson JM, Altuntas CZ, Jane-Wit D, Tuohy VK: An autoimmune-mediated strategy for prophylactic breast cancer vaccination. Nat Med 2010, . 28. Wall EM, Milne K, Martin ML, Watson PH, Theiss P, Nelson BH: Spontaneous mammary tumors differ widely in their inherent sensitivity to adoptively transferred T cells. Cancer Res 2007, 67(13):6442-6450. 29. [http://jaxmice.jax.org/strain/007962.html] 30. Shendure J, Porreca GJ, Reppas NB, Lin X, McCutcheon JP, Rosenbaum AM, Wang MD, Zhang K, Mitra RD, Church GM: Accurate multiplex polony sequencing of an evolved bacterial genome. Science 2005, 309(5741):1728-1732. 182  31. Li H, Ruan J, Durbin R: Mapping short DNA sequencing reads and calling variants using mapping quality scores. Genome Res 2008, 18(11):1851-1858. 32. [http://uswest.ensembl.org/Mus_musculus/Info/Index] 33. [http://www.ncbi.nlm.nih.gov/refseq/] 34. [http://hgdownload.cse.ucsc.edu/downloads.html#mouse] 35. [http://www.ncbi.nlm.nih.gov/IEB/Research/Acembly/index.html?mouse] 36. [http://genes.mit.edu/GENSCAN.html] 37. [http://www.broadinstitute.org/mouse/hapmap/] 38. Benjamini Y, Hochberg Y: Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc 1995, 57(1):289-300. 39. [http://www.uniprot.org/] 40. Martin S: Development of a Vaccine to Target Tumor Specific Mutations Identified by Next Generation Whole Transcriptome Shotgun Sequencing. BSc, Honours. The University of Victoria, 41. Buus S, Lauemoller SL, Worning P, Kesmir C, Frimurer T, Corbet S, Fomsgaard A, Hilden J, Holm A, Brunak S: Sensitive quantitative predictions of peptide-MHC binding by a 'Query by Committee' artificial neural network approach. Tissue Antigens 2003, 62(5):378-384. 42. Nielsen M, Lundegaard C, Worning P, Lauemoller SL, Lamberth K, Buus S, Brunak S, Lund O: Reliable prediction of T-cell epitopes using neural networks with novel sequence representations. Protein Sci 2003, 12(5):1007-1017. 43. Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M: NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8-11. Nucleic Acids Res 2008, 36(Web Server issue):W509-12. 44. Watkins-Chow DE, Pavan WJ: Genomic copy number and expression variation within the C57BL/6J inbred mouse strain. Genome Res 2008, 18(1):60-66. 45. Rumble SM, Lacroute P, Dalca AV, Fiume M, Sidow A, Brudno M: SHRiMP: accurate mapping of short color-space reads. PLoS Comput Biol 2009, 5(5):e1000386. 46. Goya R, Sun MG, Morin RD, Leung G, Ha G, Wiegand KC, Senz J, Crisan A, Marra MA, Hirst M, Huntsman D, Murphy KP, Aparicio S, Shah SP: SNVMix: predicting single nucleotide variants from next-generation sequencing of tumors. Bioinformatics 2010, 26(6):730-736.  183  5 CONCLUSION In this study, the molecular characteristics of murine tumours with different outcomes after adoptive T-cell therapy [1-3] were investigated using bioinformatics approaches to characterize the molecular differences and identify tumour-specific factors that impact the immune response. The genome-wide gene expression patterns, miRNA abundance and somatic mutations were profiled using microarrays and SOLiD sequencing technology. Distinct molecular profiles were identified among CR, PR and PD tumours with significantly higher similarities observed within the regressing tumours (CR and PR) compared to the non-regressing (PD) tumours. This observation was made throughout this study at the levels of mRNA transcripts, non-coding regulatory miRNAs and mutations. The most number of differentially expressed genes and miRNAs were identified in the comparison between PD and regressing (CR, PR) tumours. PD tumours also presented with the most putative mutations compared to the other tumour groups. Such marked differences in gene expression patterns and the SNV frequency in PD tumours could reflect the degree of genomic deregulation, which could ultimately result in the worst PD outcome of adoptive T-cell therapy. CR and PR tumours were virtually indistinguishable at the transcritomic level with only six genes differentially expressed according to the microarray analyses. None of the six genes were associated with immune-related functions. This prompted a hypothesis that mechanisms responsible for the regression outcomes of CR and PR tumours could be different and that these mechanisms were most likely to be affected at the post-  184  transcriptional level. The miRNA profiling of CR and PR tumours indeed showed different patterns, in which almost all of differentially abundant miRNAs in the PR tumours (11 out of 12) were found to be in higher abundance in PR compared to CR and PD tumours, suggesting translational activities unique to PR tumours. Although highly speculative, the distinct molecular profiles of CR, PR and PD also suggested that PR tumour was not a ‘milder’ version of PD tumour nor a ‘stronger’ version of CR tumour, thus leading to another hypothesis that genetic features associated with each tumour group could be used to predict specific outcome of adoptive T-cell therapy in the mouse model of breast cancer. To determine the predictive gene expression signature, outcome-specific genes were first identified, which were differentially expressed exclusively in a specific outcome group. To focus on identifying the most definitive factors for successful immune response to adoptively transferred T-cells, only CR- and PD-specific genes were subjected to the pathway enrichment analysis. The results of the analyses based on multiple pathway databases including Ingenuity [4], BioCarta [5] and KEGG [6] revealed that immunerelated pathways were more active in CR tumours, whereas these pathways were suppressed in PD tumours. Interestingly, both immunologically activating (e.g. C1q, Masp-2, Cfb) and inhibiting genes (e.g. Serping-1/C1 inhibitor, Cd-55/Daf) were overexpressed in the CR tumours. Also pro-inflammatory pathways were found to be more active in CR tumours compared to PD tumours. In particular, the complement pathway was found to be more active in CR tumours relative to PD tumours. The complement pathway is considered a ‘double-edged sword’ in cancer, because it can reduce tumour growth by increasing complement receptor-enhanced antibody-dependent cellular  185  cytotoxicity [7], or promote tumour growth by inducing pro-inflammatory changes that recruit immunosupressive cells [8]. The complement pathway is also linked to the Fc! receptor pathway, which was less active in PD tumours. This pathway has been shown to promote anti-cancer immune response against lymphoma [9, 10] and breast cancer [10]. This result reiterated the importance of close interactions between immunosuppressive and immune-promoting components in determining the overall immune response. Stroma-related pathways were also found to be more active in CR tumours, which provided further support for the recent findings reported by Martin and colleagues [3], in which the authors found a higher stroma-to-tumour-epithelium ratio in CR tumours compared to PD tumours. This histological feature was also found to be significantly associated with the CR outcome after adoptive T cell therapy, suggesting that the overexpression of stroma-related genes in CR tumours could be potential predictive markers for adoptive T-cell therapy. The comparison of gene expression profiles of human breast cancer subtypes revealed intriguing transcriptomic similarities between PD tumours and the basal-like subtype, whereas the regressing (CR and PR) tumours resembled the HER2-over-expressing subtype of human breast cancer. The clustering of murine mammary tumours with specific human breast cancer subtypes suggested that the molecular features of the murine tumours could closely emulate those of human breast cancer, facilitating the development of immunotherapy targeting a specific breast cancer subtype such as the HER2-over-expressing subtype. Interestingly, a study by Terschendorff and colleagues identified a subset of HER2 and basal-like subtypes of breast cancer with active complement and immune response pathways strongly associated with good prognosis  186  [11]. Thus, the result of this analysis showing the significant similarities between CR tumours and HER2 subtype support an important role of the complement pathway in successful immune responses. Moreover, the existing hypothesis, in which the presence of active immune pathways associates with good prognosis, was further supported in both human and murine cancers. One of the remarkable findings from this research is the discovery of predictive gene signatures selected using the bioinformatics approach to identify classifier genes and outcome-specific genes enriched for pathways involved in the tumour microenvironment and immune response. In total, 14 out of 33 genes were experimentally validated by qPCR to show a correlation between the gene expression level and the degree of response to adoptively transferred T-cells (the qPCR validation was carried out by Michele Martin at the Trev & Joyce Deeley Research Centre, BCCA). The five genes that showed the most difference in gene expression levels were subsequently tested for prognostic power by comparing the gene expression profile of these genes in novel NOP tumours, NOP20 and NOP37, whose outcome were previously unknown. The NOP20 and NOP37 tumours matched the gene expression signature of NOP21CR tumours in the microarray experiment, and indeed showed CR outcome after adoptive T-cell therapy. Despite of a low sample size of CR tumours (N = 1), the successful validation of the gene expression signature of the CR outcome suggests the strong predictive power of the gene expression signature identified in this study. Although the predictive gene signature of the five genes was experimentally validated, the bioinformatics analysis used to identify the prognostic classifier genes could have been more sophisticated, instead of resorting to the PAM analysis result without further  187  testing of the robustness of classification. The selection of the initial 33 genes was heavily dependent on prior knowledge of tumour immunogenicity, using the classifier gene information as a weighing factor in ranking. A widely used method of estimating the robustness of the classification and subsequently identifying the optimal number of classifier genes is cross validation of the resulting gene set using the ‘leave-one-out’ method [12]. More robust testing of the predictive analysis could result in identifying novel classifier genes, which could improve the power of the predictive gene signature. The addition of more NOP tumours with CR, PR and PD outcome to the microarray experiment would help identify more comprehensive and robust predictive gene expression signatures. The validation of the predictive gene signature across a larger panel of NOP tumours would also be necessary for further validation of the prognostic gene signature. An interesting resemblance of gene expression signatures between the top 100 genes that were most differentially expressed in CR relative to PD tumours and human breast cancer cell line treated with trichostatin A (TSA) was found using the Connectivity Map database [13]. TSA is a powerful histone deacetylase inhibitor (HDACi), which belongs a novel class of anti-cancer drugs shown to reduce tumour growth of various cancers including breast cancer [14]. In particular, the use of TSA was shown to increase the immunogenicity of tumour cells by enhancing antigen processing and presentation [15]. Furthermore, the adjunctive use of another type of HDACi, LAQ824, has been shown to enhance the immune response of metastatic melanoma to adoptively transferred T cells in mice [16]. In the experimental validation carried out by Michele Martin, the preconditioning treatment with TSA did not improve the sensitivity of PD tumours against  188  adoptively transferred T cells in vivo. Although the effect of TSA did not validate in the murine breast cancer in this study, the bioinformatics approach to use the gene expression profiles and identify the potential drug that could induce gene expression changes similar to those in CR tumours suggested that there is a therapeutic potential to enhance the efficacy of adoptive T-cell therapy, supported by independent findings from published literature. As a future research, the validation could be tried again with an increased dosage of TSA in the host-conditioning regime or a combination of TSA and adoptive Tcell therapy. The unusually low mapping rate observed especially in the miRNA sequence data generated using the SOLiD technology was a significant weakness in detecting the differential abundance of miRNA amongst CR, PR and PD tumours. Inconsistencies in the sequencing throughput, the mapping rate and the small RNA class distributions across libraries constructed using different protocols could have resulted in poor library quality, contributing to the low mapping rate. However, the exploration of different parameters of the SHRiMP alignment tool to identify optimal settings and the processing of sequences to remove adapter sequence could have been carried out more comprehensively. Also, filtering the sequences that mapped uniquely to a single locus could have resulted in exclusion of legitimate miRNA sequences. For future research, sequences that mapped to multiple loci in the genome should be considered to reflect the high sequence similarities observed amongst mature miRNAs in the same miRNA families. The miRNA profiles nonetheless revealed several miRNAs present with significantly different abundance amongst CR, PR and PD tumours that were found to be involved in  189  modulating the immune response or correlated with prognosis of various cancer types. To experimentally confirm the differential abundance of miRNA determined by sequencing, Northern blot or qPCR analysis on some of the most differentially abundant miRNAs could be carried out for the future research. As done with the microarray data, the unsupervised hierarchical clustering could be carried out using the normalized sequence count of miRNAs to identify the miRNA clusters with similar abundance profiles, which could reflect the functional relevance. Similarly, PAM analysis of the miRNA profile could also allow for a detection of classifier miRNA signatures. From similar analyses using miRNA microarray data, researchers were able to identify miRNA signatures that distinguished between different human breast cancer subtypes [17], and between normal and cancerous breast tissues [18]. The lack of changes in the expression level of genes targeted by differentially abundant miRNAs observed in this study could be explained by the subtle effect of miRNAmediated gene suppression, as shown by the small decrease in the protein level described in the studies by Baek and colleagues [19] and Selbach and colleagues [20]. This observation would also support the mechanism of the suppressive gene expression control by miRNAs through blocking translational activities rather than degrading mRNA transcripts. However, the possibility of falsely predicted target genes confounding the overall result still remained. A way to improve this would be to obtain the predicted gene set from more than one prediction algorithm (i.e. TargetScan) and filter genes that were commonly predicted to be a miRNA target from multiple prediction algorithms. Target genes predicted with less statistical confidence could also be removed from further analyses.  190  From sequencing the transcriptomes from CR, PR and PD tumours using the SOLiD technology, 441 novel SNVs that caused amino-acid changes were identified, representing putative non-synonymous mutations. These mutations were subsequently subjected to epitope prediction analysis by Spencer Martin and 18 were selected for experimental validation by PCR amplicon sequencing. From 18 putative mutations attempted for validation, none were found to be true somatic mutations. The false positives were most likely to have come from the lack of accuracy in the alignment to the reference genome and the SNP discovery algorithm. In particular, the alignment tool chosen for this study was MAQ [21], which might have not been the optimal tool of choice for aligning sequences in colour-space since it was originally developed to align short sequences in letter-space such as those generated using the Solexa/Illumina sequencing technology. Other alignment algorithms optimized for colour-space sequences such as SHRiMP [22] and rna2map (Applied Biosystems) should be considered for future research. Moreover, other SNP discovery tool such as SNVMix [23], which incorporates a probabilistic binomial mixture model, could be explored for more accurate detection of SNVs. Overall, the findings from this research revealed several molecular features of tumours that exhibited a complete range of degree in immunogenicity after adoptive T-cell therapy. The mouse model used in this research was a unique experimental system, providing an unprecedented opportunity to investigate the intricate relationships between tumours and the immune system, and amongst different constituents of the immune system. The transcriptome sequencing of murine tumours also provided a unique opportunity to survey mutations in the tumours with syngeneic genomic background,  191  differing only in their immune response. The findings from this research supported the hypothesis that the molecular characteristics of tumours could be used to predict the outcome of adoptive T-cell therapy and that the immune-modulating factors instigated from tumours indeed affected the immunotherapy outcome. Furthermore, the identification of a potential adjuvant drug for immunotherapy, the predictive gene expression signature and transcriptional similarities of specific human breast cancer subtypes supported the possibility of translating these results to clinical applications, bringing adoptive T-cell therapy closer to reality for breast cancer patients.  192  5.1 BIBLIOGRAPHY 1. Wall EM, Milne K, Martin ML, Watson PH, Theiss P, Nelson BH: Spontaneous mammary tumors differ widely in their inherent sensitivity to adoptively transferred T cells. Cancer Res 2007, 67(13):6442-6450. 2. Yang T, Martin ML, Nielsen JS, Milne K, Wall EM, Lin W, Watson PH, Nelson BH: Mammary tumors with diverse immunological phenotypes show differing sensitivity to adoptively transferred CD8+ T cells lacking the Cbl-b gene. Cancer Immunol Immunother 2009, 58(11):1865-1875. 3. Martin ML, Wall EM, Sandwith E, Girardin A, Milne K, Watson PH, Nelson BH: Density of tumour stroma is correlated to outcome after adoptive transfer of CD4(+) and CD8 (+) T cells in a murine mammary carcinoma model. Breast Cancer Res Treat 2009, . 4. Ingenuity Systems [www.ingenuity.com] 5. BioCarta [http://www.biocarta.com/] 6. Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M: KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res 1999, 27(1):29-34. 7. Gelderman KA, Tomlinson S, Ross GD, Gorter A: Complement function in mAbmediated cancer immunotherapy. Trends Immunol 2004, 25(3):158-164. 8. Markiewski MM, DeAngelis RA, Benencia F, Ricklin-Lichtsteiner SK, Koutoulaki A, Gerard C, Coukos G, Lambris JD: Modulation of the antitumor immune response by complement. Nat Immunol 2008, 9(11):1225-1235. 9. Cartron G, Dacheux L, Salles G, Solal-Celigny P, Bardos P, Colombat P, Watier H: Therapeutic activity of humanized anti-CD20 monoclonal antibody and polymorphism in IgG Fc receptor FcgammaRIIIa gene. Blood 2002, 99(3):754-758. 10. Clynes RA, Towers TL, Presta LG, Ravetch JV: Inhibitory Fc receptors modulate in vivo cytoxicity against tumor targets. Nat Med 2000, 6(4):443-446. 11. Teschendorff AE, Miremadi A, Pinder SE, Ellis IO, Caldas C: An immune response gene expression module identifies a good prognosis subtype in estrogen receptor negative breast cancer. Genome Biol 2007, 8(8):R157. 12. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH: Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002, 415(6871):530-536. 13. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, Lerner J, Brunet JP, Subramanian A, Ross KN, Reich M, Hieronymus H, Wei G, Armstrong SA, Haggarty SJ, Clemons PA, Wei R, Carr SA, Lander ES, Golub TR: The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 2006, 313(5795):1929-1935.  193  14. Vigushin DM, Ali S, Pace PE, Mirsaidi N, Ito K, Adcock I, Coombes RC: Trichostatin A is a histone deacetylase inhibitor with potent antitumor activity against breast cancer in vivo. Clin Cancer Res 2001, 7(4):971-976. 15. Setiadi AF, Omilusik K, David MD, Seipp RP, Hartikainen J, Gopaul R, Choi KB, Jefferies WA: Epigenetic enhancement of antigen processing and presentation promotes immune recognition of tumors. Cancer Res 2008, 68(23):9601-9607. 16. Vo DD, Prins RM, Begley JL, Donahue TR, Morris LF, Bruhn KW, de la Rocha P, Yang MY, Mok S, Garban HJ, Craft N, Economou JS, Marincola FM, Wang E, Ribas A: Enhanced antitumor activity induced by adoptive T-cell transfer and adjunctive use of the histone deacetylase inhibitor LAQ824. Cancer Res 2009, 69(22):8693-8699. 17. Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning MJ, BarbosaMorais NL, Teschendorff AE, Green AR, Ellis IO, Tavare S, Caldas C, Miska EA: MicroRNA expression profiling of human breast cancer identifies new markers of tumor subtype. Genome Biol 2007, 8(10):R214. 18. Iorio MV, Ferracin M, Liu CG, Veronese A, Spizzo R, Sabbioni S, Magri E, Pedriali M, Fabbri M, Campiglio M, Menard S, Palazzo JP, Rosenberg A, Musiani P, Volinia S, Nenci I, Calin GA, Querzoli P, Negrini M, Croce CM: MicroRNA gene expression deregulation in human breast cancer. Cancer Res 2005, 65(16):7065-7070. 19. Baek D, Villen J, Shin C, Camargo FD, Gygi SP, Bartel DP: The impact of microRNAs on protein output. Nature 2008, 455(7209):64-71. 20. Selbach M, Schwanhausser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N: Widespread changes in protein synthesis induced by microRNAs. Nature 2008, 455(7209):58-63. 21. Li H, Ruan J, Durbin R: Mapping short DNA sequencing reads and calling variants using mapping quality scores. Genome Res 2008, 18(11):1851-1858. 22. Rumble SM, Lacroute P, Dalca AV, Fiume M, Sidow A, Brudno M: SHRiMP: accurate mapping of short color-space reads. PLoS Comput Biol 2009, 5(5):e1000386. 23. Goya R, Sun MG, Morin RD, Leung G, Ha G, Wiegand KC, Senz J, Crisan A, Marra MA, Hirst M, Huntsman D, Murphy KP, Aparicio S, Shah SP: SNVMix: predicting single nucleotide variants from next-generation sequencing of tumors. Bioinformatics 2010, 26(6):730-736.  194  

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}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            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:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0071196/manifest

Comment

Related Items