CHARACTERIZATION OF EPITHELIOID SARCOMA USING MASSIVELY PARALLEL DNA AND RNA SEQUENCING AND IN VITRO MODELS by Farzad Jamshidi B.Sc. The University of British Columbia, 2009 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Interdisciplinary Oncology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) June 2015 © Farzad Jamshidi, 2015 ii Abstract Epithelioid sarcoma is a soft tissue tumor with an unusual predilection for the distal extremities in young adults. Despite wide-margin resections the 10-year survival is in the range of 50%. The biology of epithelioid sarcoma remains incompletely understood, but one key feature is the loss of SMARCB1. We use whole genome sequencing of five cases of epithelioid sarcoma matched to normal germline DNA, looking for mutations other than SMARCB1. These index cases are supplemented with three additional tumors and three cell lines that undergo whole transcriptome sequencing and are analyzed for somatic point mutations, copy number changes, translocations, and expression patterns. Unlike the situation in other SMARCB1 inactivated tumors, we find a complex genome with a relatively high mutational burden. However, aberrations of SMARCB1 remain the only consistent mutation. Some cases do not show biallelic DNA-level inactivation of this gene which leads us to examine other possible second-hit silencing mechanisms. With the significance attributed to SMARCB1 loss in the genomic landscape of epithelioid sarcoma, we explore two approaches, namely EZH2 and SMARCA4 inhibitions, to specifically exploit this abnormality and suggest novel therapeutic options. iii Preface Contents of section 1.6, Figure 1-3, and Figure 2-3a are from publications on which I am a co-author with appropriate permissions acquired(Jamshidi, Nielsen, & Huntsman, 2015; Jamshidi et al., 2014). Nalan Gogkoz, Mount Sinai Hostpial, Toronto, and I extracted the nucleic acids used for the discovery cohort. I performed the other extractions. I also performed the nucleic acid quality checks and quantification before submission to the Canada’s Michael Smith Genome Sciences Centre where the library preparation and sequencing was completed. Leah Prentice from Dr. David Huntsman’s laboratory helped with the submission and tracking of samples. I performed the validation experiments on Ion Torrent sequencing with supervision of Dr. Stephen Yip. Immunohistochemistry was done at Genetic Pathology Evaluation Centre (GPEC) Vancouver, BC by Christine Chow. The scoring for CDKN2A, SMARCA4, SMARCA2, and ARID1A were done by Dr. Dongxia Gao. The scoring from the FISH on the tissue microarray(TMA) was done by Amy Lum. The TMA sections were kindly supplied by Dr. Alexander Lazar, MD Anderson Cancer Centre. The bioinformatics analysis were based on the programs outlined in Figure 1-3 and were ran by Karey Shumansky at Dr. Sohrab Shah’s laboratory under the supervision of Dr. Ali Bashashati. Figures 2-3B, 2-4, 2-6 were generated by Karey Shumansky. I performed all the remaining experiments, image creation and write up. The work was supported by the Terry Fox Frontiers in Medicine Forme Fruste grant led by Dr. David Huntsman and Dr. Torsten Nielsen. This research was approved by the institutional ethics review board under Research Ethics Board #H08-01411. All inter-institutional material transfer was done with appropriate transfer agreements. iv Table of contents Abstract ............................................................................................................................... ii Preface ................................................................................................................................ iii Table of contents ............................................................................................................... iv List of tables..................................................................................................................... viii List of figures ..................................................................................................................... ix List of abbreviations .......................................................................................................... x Acknowledgements ........................................................................................................... xi Dedication ......................................................................................................................... xii Chapter 1: Introduction .................................................................................... 1 1.1 Aims ................................................................................................................................ 1 1.2 Soft tissue sarcomas ...................................................................................................... 1 1.3 Establishment of a new entity ...................................................................................... 2 1.3.1 Further characterization of epithelioid sarcoma ......................................... 6 1.3.2 Proximal type epithelioid sarcoma .............................................................. 10 1.3.3 Epidemiology and current therapies ........................................................... 14 1.4 SMARCB1 and epithelioid sarcoma .......................................................................... 17 1.4.1 SMARCB1: a tumor suppressor gene ......................................................... 18 1.4.2 SMARCB1 inactivated tumors .................................................................... 19 1.4.3 SMARCB1: a core SWI/SNF member ........................................................ 24 1.4.4 SWI/SNF and cancer .................................................................................... 30 1.5 Other deregulated pathways in epithelioid sarcoma ................................................ 33 v 1.6 Next generation sequencing and study of rare tumors ............................................. 36 1.7 Thesis objective and chapter overview ...................................................................... 38 Chapter 2: Genomic Landscape of Epithelioid Sarcoma ............................ 45 2.1 Introduction .................................................................................................................. 45 2.2 Results ........................................................................................................................... 46 2.2.1 Description of the discovery cohort ............................................................. 46 2.2.2 Copy number changes on tumor samples ................................................... 46 2.2.3 CDKN2A immunohistochemistry in epithelioid sarcoma ......................... 51 2.2.4 Somatic point mutations ............................................................................... 51 2.2.5 Fusions in epithelioid sarcoma ..................................................................... 53 2.2.6 Expression analysis ....................................................................................... 55 2.3 Discussion...................................................................................................................... 58 2.4 Methods ......................................................................................................................... 60 2.4.1 Sample acquisition and nucleic acid extractions ........................................ 60 2.4.2 Deep sequencing ............................................................................................ 61 2.4.3 Bioinformatics analysis ................................................................................. 62 2.4.4 Fluorescent in situ hybridization (FISH) .................................................... 63 2.4.5 PCR and Sanger sequencing ........................................................................ 64 2.4.6 Immunohistochemistry ................................................................................. 64 Chapter 3: SMARCB1 Inactivation in Epithelioid Sarcoma ...................... 75 3.1 Introduction .................................................................................................................. 75 3.2 Results ........................................................................................................................... 76 3.2.1 Sanger sequencing and FISH on tumors .................................................... 76 vi 3.2.2 Cell line model: intact SMARCB1 allele in HSES ...................................... 76 3.2.3 Methylation status ......................................................................................... 78 3.2.4 MicroRNA silencing of SMARCB1 ............................................................. 79 3.3 Discussion...................................................................................................................... 81 3.4 Methods ......................................................................................................................... 83 3.4.1 Multiplex Ligation-dependent Probe Amplification (MLPA) .................. 83 3.4.2 Methylation specific PCR ............................................................................. 83 3.4.3 Cell culture, drug treatment and Western blots ........................................ 83 3.4.4 qPCR .............................................................................................................. 84 3.4.5 Viral infection with sponge miRNA inhibitors ........................................... 84 Chapter 4: In Vitro Studies and Epigenetic Modifying Compounds .......... 89 4.1 Introduction .................................................................................................................. 89 4.2 Results ........................................................................................................................... 90 4.2.1 Inducible induction of SMARCB1 .............................................................. 90 4.2.2 Induction of SMARCB1 and EZH2 inhibition ........................................... 91 4.2.3 EZH2i on epithelioid sarcoma ..................................................................... 92 4.2.4 The residual SWI/SNF complex in epithelioid sarcoma ............................ 93 4.2.5 Synthetic lethality in the residual complex ................................................. 94 4.3 Discussion...................................................................................................................... 95 4.4 Methods ......................................................................................................................... 97 4.4.1 Vector cloning................................................................................................ 97 4.4.2 Viral transduction of the inducible system ................................................. 97 4.4.3 Immunocytochemistry .................................................................................. 98 vii 4.4.4 MTS assay and Incucyte ............................................................................... 99 4.4.5 siRNA knockdown and immunoprecipitation ............................................ 99 4.4.6 qPCR ............................................................................................................ 100 Chapter 5: Conclusions and Future Directions .......................................... 106 5.1 Summary of findings.................................................................................................. 106 5.2 Insights from the genomic landscape of epithelioid sarcoma ................................ 106 5.3 Lineage of epithelioid sarcoma ................................................................................. 108 5.4 SMARCB1 loss in epithelioid sarcoma .................................................................... 108 5.5 Epigenetic modifiers as rational therapies for epithelioid sarcoma ...................... 109 5.6 Residual SWI/SNF complex in epithelioid sarcoma ............................................... 110 5.7 Future direction ......................................................................................................... 111 References ....................................................................................................... 113 Appendices ...................................................................................................... 127 Appendix A: Ion Torrent parameter for VAESBJ ....................................................... 127 Appendix B: Mutations called independently in multiple platforms .......................... 128 Appendix C: Non-open reading frame fusions .............................................................. 132 Appendix D: Cluster outlines for the differential expression ...................................... 135 Appendix E: Top differentially expressed genes from cluster analysis ...................... 136 Appendix F: HDAC inhibitor romidepsin in epithelioid sarcoma .............................. 137 viii List of tables Table 1-1: Immunohistochemistry (IHC) findings in epithelioid sarcoma .................. 43 Table 1-2: Summary of karyotypic studies of epithelioid sarcoma ............................... 44 Table 2-1: Sample list ........................................................................................................ 71 Table 2-2: Genes with evidence of deletion in discovery cases ...................................... 72 Table 2-3: Fusions predicted via deFuse .......................................................................... 73 Table 2-4: FPKM values of epithelioid sarcoma lines compared to ENCODE lines ... 74 Table 3-1: FISH results on epithelioid sarcoma TMA ................................................... 88 ix List of figures Figure 1-1: Histology of epithelioid sarcoma................................................................... 40 Figure 1-2: SWI/SNF mutations in cancer ...................................................................... 41 Figure 1-3: Outline of study design .................................................................................. 42 Figure 2-1: Co-occurrence of NRAS hotspot and SMARCB1 loss in CRINET ........... 65 Figure 2-2: Virtual karyotype of epithelioid sarcoma .................................................... 66 Figure 2-3: Structural landscape of epithelioid sarcoma ............................................... 67 Figure 2-4: Mutation rates of epithelioid sarcoma rates vs other tumors .................... 68 Figure 2-5: Summary of mutations .................................................................................. 69 Figure 2-6: Epithelioid sarcoma expression profiles ...................................................... 70 Figure 3-1: Evaluation for SMARCB1 mutations in epithelioid sarcoma cell lines ..... 85 Figure 3-2: Transcriptional levels of SMARCB1 in epithelioid sarcoma lines ............. 86 Figure 3-3: Epigenetic mechanisms of SMARCB1 silencing in HSES .......................... 87 Figure 4-1: Validations of doxycycline inducible cell lines .......................................... 101 Figure 4-2: EZH2 inhibition and SMARCB1 induction on targets of SMARCB1 ..... 102 Figure 4-3: EZH1/2 inhibition in epithelioid sarcoma ................................................. 103 Figure 4-4: Immunohistochemistry of SWI/SNF in epithelioid sarcoma ................... 104 Figure 4-5: Residual SWI/SNF and potential for synthetic lethality .......................... 105 x List of abbreviations ALL: Acute lymphocytic leukemia CLL: Chronic lymphocytic leukemia DLCBL: Diffuse large B-cell lymphoma EpS: Epithelioid sarcoma FISH: Fluorescent in situ hybridization FFPE: Formalin fixed paraffin-embedded H&E: Hematoxylin and Eosin IHC: Immunohistochemistry Indel: Small Insertion/Deletion PCR: Polymerase Chain Reaction qPCR: Quantitative Polymerase Chain Reaction SCC: Squamous Cell Carcinoma SNP: Single Nucleotide Polymorphism SNV: Single Nucleotide Variant STS: Soft Tissue Sarcoma TMA: Tissue Microarray WB: Western Blot WGSS: Whole Genome Shotgun Sequencing WTSS: Whole Transcriptome Shotgun Sequencing xi Acknowledgements I would like to thank Dr. Torsten Nielsen for his supervision and support. Drs. David Huntsman, Stephen Yip, Michael Underhill, and Poul Sorensen where members of my supervisory committee and I am thankful for their help and input. I would also like to thank all the members of the Nielsen laboratory and the Huntsman laboratory for their guidance and help. Ali Bashashati’s supervision, and Karey Shumansky’s contributions on the bioinformatics analyses were critical for this work and much appreciated. Lastly, I would like to thank Drs. Irene Andrulis and Brendan Dickson from Mt. Sinai Hospital, Toronto, Drs. Dina Lev and Alexander Lazar from MD Anderson Cancer Centre, and Dr. Robin Jones from Seattle Cancer Care Alliance for providing specimen to the discovery and validation cohorts. xii Dedication To my father, Khodamorad Jamshidi, who has been my role model in life and to my mother, Parvin Akhtarkhavary, whose support has been my greatest fortune. To the patients suffering from epithelioid sarcoma, whose selflessness in supporting research and donating samples fuels the hope of curing others befalling the same disease. 1 Chapter 1: Introduction 1.1 Aims The original intended purpose of this research described in this thesis was to use next generation sequencing technologies to study the genomic landscape of epithelioid sarcoma, a relatively understudied soft tissue sarcoma. The results from next-generation sequencing were analyzed to learn about the pathobiology of this tumor and to try to identify driver mutations with the ultimate goal of identifying potential therapeutic targets. 1.2 Soft tissue sarcomas Soft tissue is defined as extra-skeletal non-epithelial tissue excluding the reticuloendothelial system and parenchymal supportive connective tissue such as glia(Weiss, Goldblum, & Enzinger, 2001). Therefore it includes muscle, fat, fibrous tissue, vasculature and by convention the peripheral nervous system. The inclusion of peripheral nervous system is because tumors arising from these sites present as soft tissue masses and create similar diagnostic and therapeutic challenges. Tumors arising from soft tissues can generally be categorized as benign or malignant. Benign variants have limited growth potential, do not metastasize and do not pose major clinical concerns. In the soft tissue, benign tumors are much more common than malignant variants(Weiss et al., 2001). When needed, conservative therapies (such are marginal surgical excision) are sufficient to manage these tumors. However, malignant soft tissue tumors, also known as soft tissue sarcomas are a challenge therapeutically and diagnostically. They can be locally aggressive, recur and metastasize after conventional therapies. 2 One of the special challenges for managing sarcomas is their diversity. There are at least 80 subtypes of STS; about half of behave simply as locally aggressive neoplasms but the rest metastasize at some point in their disease course(Brennan, Antonescu, Moraco, & Singer, 2014). Compared to benign soft tissue tumors, soft tissue sarcomas show less histologic differentiation yet whenever there is evidence of a line of mesenchymal differentiation, the tumors are often named accordingly. Hence, for instance, angiosarcomas represent tumors with vascular differentiation while liposarcomas represent tumors with features of fatty tissue. Although common in animals, soft tissue sarcomas are rare in humans and comprise 1% of all cancers. Despite their relative rarity, a significant number of individuals are still affected. In 2001, the American Cancer Society estimated 8700 new cases and 4400 deaths resulting from soft tissue sarcomas(Greenlee, Hill-Harmon, Murray, & Thun, 2001). Anatomically, the majority of STS occur in the lower limbs, followed by the retroperitoneum/abdomen, the upper limbs and the trunk (Brennan et al., 2014). In the upper extremities, the most common soft tissue sarcomas are epithelioid sarcoma, synovial cell sarcoma and malignant fibrous histiocytoma(Murray, 2004), the latter now termed undifferentiated pleomorphic sarcoma. Epithelioid sarcoma is the most common sarcoma of the hand and wrist(Halling, Wollan, Pritchard, Vlasak, & Nascimento, 1996; Ross, Lewis, Woodruff, & Brennan, 1997), and it has an usual but well-established predilection to occur in the upper extremities(Gustafson & Arner, 1999). 1.3 Establishment of a new entity In a seminal publication in 1970, Franz M. Enzinger described the common features among a collection of 62 soft tissue tumors that were previously generally 3 classified as granulomatous conditions, synovial sarcomas or ulcerating squamous cell carcinomas. Their features included an epithelial-like appearance of many or all tumor cells, the nodular pattern in histology, evidence of necrosis and the tendency to occur along tendons and fascial structures(Enzinger, 1970). He recognized that these pathologies likely represented a single entity and since the large eosinophilic polygonal cells were the most striking histologic feature, he named this new sarcoma “epithelioid sarcoma”. This collection of 62 cases occurred most commonly in the upper extremities, particularly in the fingers and hands, followed by lower extremity. There was a 4:1 male predilection, and the median age at occurrence was 23 with 75% of cases occurring between the ages of 10 and 34. Most commonly the tumors had originally been diagnosed as synovial sarcoma when thought of as malignant (15 cases) or as granulomatous inflammation when cancer was not suspected (16 cases). Tumor masses were either in the dermis, the hypodermis, or the deeper fascia on tendon sheaths. The dermal masses presented as lumps that ulcerated within a few months after being first noticed and thus were commonly mistakenly diagnosed as “infected warts” or “indurated ulcers”. Ulcerations were particularly common over the tibia, patella or the elbow. The hypodermal masses presented as wart-like lumps with slow steady growth. The tumors of the deep fascia were generally larger yet less defined in borders and presented as a single or as a series of nodules. In terms of size, the tumors varied from 2mm to 5cm in diameter. The masses themselves did not cause severe pain unless nearby nerve bundles were impinged upon. Only 10% of patients could recall potential injury at the tumor site within 5-years prior to the first appearance of the mass while 8% specifically denied any injury. Therefore there was no obvious relationship between trauma and tumor 4 occurrence, although a later study would point out this possibility without strong support(Chase & Enzinger, 1985). Perhaps the most unique histologic feature of Enzinger’s epithelioid sarcoma was the presence of “granuloma-like” nodules that at times had necrosis in the center. These nodules were variable in appearances with larger nodules having more prominent central necrosis. Primary tumors had singular nodules while multi-nodular cases were associated with recurrences. In dermal cases, lack of dyskeratotic cells (abnormal keratinizing cells below the stratum granulosum) or prominent acanthosis (hyperplasia of stratum basale and stratum spinosum) helped distinguish epithelioid sarcoma from squamous cell carcinoma of the skin. Cellular heterogeneity was minimal yet populations of spindled cells were found particularly on the edges the epithelioid cell clusters. Recurrences and metastatic tumors tended to have a much more prominent epithelioid cell components whereas the spindled cells were more apparent in early primary cases(Enzinger, 1970). There was never a sharp separation between the spindled cell population and the epithelioid cells, a distinguishing feature from biphasic synovial sarcoma. Furthermore, mucin containing pseudoglandular formations sometimes found in biphasic synovial sarcomas were not seen in any of the epithelioid sarcoma cases. Enzinger also noted that the significant overall eosinophilic appearance seen in epitehioid sarcoma was not solely due to the prominent cytoplasms of the tumor cells but was partly caused by matrix deposition through desmoplasia (a reactive growth of fibrous and connective tissue). Calcification was present in the necrotic portion of granulomatous areas in 13% of cases, and osteoid/bone formation was present in 8% of cases. There was evidence of immune reaction in the majority of cases with siderophages (iron containing 5 macrophages) and lymphocytes infiltrating the tumors. Interestingly, at autopsy, nodular growth into the pleural and peritoneal surfaces, the wall of large vessels and the esophagus was observed in some of metastatic cases. Ultrastructurally, electron microscopy revealed nuclei with scattered chromatin and 1 to 2 nucleoli. Condensed chromatin were seen in small clumps or bands around the periphery of nuclei. Cytoplasm was abundant with moderate amounts of mitochondria and rough endoplasmic reticula. Large amounts of intermediate filaments were observed that were not organized, and their amount was seemingly inversely proportional to that of endoplasmic reticulum. Vacuoles and lysosomes were rare. Filopodia, i.e. sharp extensions beyond the lamellipodia were seen that created intercellular interdigitations. Collagen and ground substance filled the extracellular space(Enzinger, 1970). Therefore in summary, Enzinger recognized a new cancer that presented as small slow growing nodules with high metastatic potential typically occurring in the upper extremities of young adults. These sarcomas had an unusual granulomatous histology with central necrosis. The tumor cells had prominent eosinophilia and polygonal shapes and were “epithelioid”; however, transition to spindled cells was seen in primary cases. The pseudo-granulomas, so called because the tumor cells rather than macrophages made up the granuloma, caused lymphocytic immune reaction in most cases. Compared to inflammatory granulomas, the epithelioid sarcoma pseudogranulomas had less mature cells with a more significant mitotic activity and could be distinguished. Clinically, these tumors tended to metastasize to the lymph nodes, lungs, the pleurae or the skin. Epithelioid sarcomas were previously misdiagnosed as a ulcerating squamous cell carcinoma particularly in superficial cases, synovial sarcoma in more deeply rooted 6 cases, or granulomatous inflammatory conditions in cases where malignancy was not suspected. 1.3.1 Subsequent studies characterizing epithelioid sarcoma After Enzinger’s description of this new tumor, several case reports and reviews in the 1970s and early 1980s reported on additional cases(C. Fisher, 1988; Haas, Palmer, Weinberg, & Beckwith, 1981; Mills, Fechner, Bruns, Bruns, & O'Hara, 1981). With these independent reports and a large study of 241 cases from 19 different countries by Chase and Enzinger(Chase & Enzinger, 1985), the initial findings were confirmed yet slightly modified. It became apparent that male predilection while still there was less prominent. Indeed the male to female ratio has been established to be about 1.5-2:1 (Chase & Enzinger, 1985; Jawad, Extein, Min, & Scully, 2009; Weiss et al., 2001). Furthermore, no geographic or racial predispositions where identified(Chase & Enzinger, 1985). The most common sites were found to be the distal upper extremity (58%), distal lower extremity (15%), proximal lower extremity (12%), proximal upper extremity (10%), trunk (3%) and the head and neck (1%), although case reports of rarer sites such as the vulva(Gallup, Abell, & Morley, 1976; Hall, Grimes, & Goplerud, 1980) and the penis(Iossifides, Ayala, & Johnson, 1979) were also published. The recurrence rate was 77% and the metastatic rate 45%(Chase & Enzinger, 1985). The first site of metastasis was most commonly to the lymph nodes(48%) and the lungs (25%). There are only a handful of “lymphogenous” sarcomas that have a clear propensity for lymph node metastasis: these include rhabdomyosarcoma, epithelioid sarcoma and clear cell sarcoma (Zagars et al., 2003). Such sarcomas have been shown to have a significantly worse 7 prognosis compared to other soft tissue sarcomas when evaluated for 5 and 15-year metastatic control and disease-free survival(Zagars et al., 2003). One of the interesting findings was evidence of metaplasia to various lineages. Areas of bone formation were found in 10% and cartilaginous differentiation was found in an additional case (Chase & Enzinger, 1985) suggesting multipotent capacity of the tumor cells. Moreover, cases with prominent intracytoplasmic vacuoles were recognized, which gave the tumor cells a signet ring appearance(Chase & Enzinger, 1985; Mills et al., 1981). Giant cells were present in 5% of cases but were clearly distinguishable from the bizarre giant cells of malignant fibrous histiocytoma / undifferentiated pleomophic sacomas (Chase & Enzinger, 1985). Another interesting observation was that the peak of occurrence in females was later than that of men (30 yrs vs. 26 yrs), and in general women had better outcome. A higher rate of metastasis in cases with more prominent central necrosis (88% vs. 34%) was reported(Chase & Enzinger, 1985). However, recurrence had an inverse relationship with central necrosis (94% recurrence in cases with little central necrosis vs. 64%). While perplexing, it has been suggested that this relationship could be caused because a more aggressive treatment course is taken when central necrosis is observed. A limitation of all these studies were that they were retrospective referral series subject to bias. The rate of metastasis is about 50% and indeed follow up for this tumor should extend decades as patients died up to 19 years after the initial procedure as a result of late metastases. Furthermore, one of the big challenges is the recognition of the malignancy of this condition given the initial indolent presentation. The average time of symptoms before surgical procedure in a series of 202 cases with available detailed information on 8 clinical presentation was 2 years, with a range from 1 week to 25 years. More recent studies suggest an average time of 1 year of symptoms before diagnosis of epithelioid sarcoma and this improvement likely has to do with improved awareness and recognition of this tumor (Chbani et al., 2009). Immunohistochemically, epithelioid sarcoma has been established to be positive for vimentin, an intermediate filament usually expressed in mesenchymal tissues, as well as epithelial membrane antigen and several cytokeratins associated with epithelial tissues(Chase & Enzinger, 1985; Chbani et al., 2009; Weiss et al., 2001). Although peculiar, this multi-lineage staining is not unique to epithelioid sarcoma: positive staining for both cytokeratins and vimentin is also seen in malignant rhabdoid tumors, pleomorphic adenoma, adenoid cystic carcinoma of the salivary gland, synovial sarcoma, as well as renal, lung and some undifferentiated carcinomas(C. Fisher, 1988). Due to lack of dysplastic processes in the skin in the cases where epithelioid sarcoma occurs in the dermis, and because many tumors arise in deeper tissue along fascia and tendons, epithelioid sarcoma is thought to be a genuine sarcoma, thus primarily a mesenhymal tumor(Chase & Enzinger, 1985). Epithelioid sarcoma is negative for melanoma/neuroectodermal markers (e.g. S100) as well as for endothelial markers such as CD31 and von Willebrand factor (Weiss et al., 2001). Despite insights from immunohistochemistry, the cell of origin is still inconspicuous. Fibrocytic, histiocytic and synovial differentiations were described(E. R. Fisher & Horvat, 1972; Machinami, Kikuchi, & Matsushita, 1982; Soule & Enriquez, 1972) but the tumor histology remains too ambiguous for a clear differentiation pattern to be recognized. Immunohistochemistry findings are generally uniform across tumor cells. One of the most distinctive features is 9 the loss of the INI-1 protein encoded by SMARCB1(Hornick, Dal Cin, & Fletcher, 2009) which is discussed further below. Loss of this protein has been described in a handful but growing number of tumors and was first described in atypical teratoid/rhabdoid tumors of the brain(Biegel et al., 1999) and malignant rhabdoid tumors of the kidney(Versteege et al., 1998). A summary of immunohistochemistry findings in epithelioid sarcoma can be found in Table 1-1. Ultra-structurally, the picture for epithelioid sarcoma is not as consistent as histological or immunohistochemical findings and there have been many additional observations since Enzinger’s 1970 paper on epithelioid sarcoma. One of the most extensive electron microscopy studies of epithelioid sarcoma was reported by Cyril Fisher in 1988. Fisher looked in detail at 7 cases (4 cases on the forearm, 2 on the foot and 1 on the penis) and reported variability in phenotypes. He described 4 main types of epithelioid sarcoma tumor cells suggesting a spectrum of differentiation: (a) cells with few organelles or filaments in a pale staining background of polyribosomes that lacked junctional complexes (these were not always found but were the majority in one case), (b) cells with variable amounts of endoplasmic reticula and Golgi complexes that appeared fibrohistiocytic yet had intercellular desmosome-like junctions, (c) cells with often abundant concentrations of cytoplasmic intermediate filaments (most likely vimentin) that at times pushed the nucleus aside and were more common, and (d) cells with intermediate filaments but also possessing aggregates of tonofilaments (likely cytokertains). Other observations were lipid droplets and presence of microvilli in some cells and general lack of external lamina(C. Fisher, 1988). The spindled epithelioid sarcoma cells were similar to but could be distinguished from fibroblasts because of 10 abundant intermediate filaments and evidence of intercellular junctions. Others have reported similar findings to Fisher’s (b), (c) and (d) cell types (Bloustein, Silverberg, & Waddell, 1976; L. Guillou, Wadden, Coindre, Krausz, & Fletcher, 1997; Ishida, Oka, Matsushita, & Machinami, 1992; Mills et al., 1981), but also reported presence of a population of round compact tumor cells with indented nuclei(Ishida et al., 1992) as well as the existence of intermingled dark and light cells on electron microscopy(Bloustein et al., 1976). Phenotypes similar to cell types (c) and (d) have also been described in malignant rhabdoid tumors of the kidney(Haas et al., 1981). Astonishingly, 17 years before epithelioid sarcoma and malignant rhabdoid tumor would be linked directly through the common loss of SMARCB1(Modena et al., 2005), Fisher described the resemblance of these tumors based on positivity for both vimentin and keratin, their ultra-structural similarities as well as their common most likely mesodermal derivation(C. Fisher, 1988). 1.3.2 Proximal type epithelioid sarcoma An advancement in the classification of epithelioid sarcoma came when Guillou and colleagues realized that the axial and proximal cases of epithelioid sarcoma (i.e. not arising in distal extremities) not only tended to have worse prognoses as previously suggested, but that their histology was different enough for a new sub-classification to be established(L. Guillou et al., 1997). These tumors tended to lack psuedogranulomatous formations, consistently displayed more prominent atypia in the tumor cells, and had rhabdoid features. Out of the 16 cases in their study, the male:female ratio was 1.6:1 and the median age of occurrence was 35.5 with variation from 20-40 years of age. The 11 tumors were located in the pelvis and perineal region, the pubic area, the vulva, the penis, deep soft tissue of the hip, the skin of the neck, the scalp and the proximal forearm. Hasegawa et al. independently validated these findings: in contradistinction to the “classical” distal upper extremity lesions that had psedugranulomatous histology, the newly recognized subgroup of epithelioid sarcoma generally lacked pseudogranulomas (even though necrosis could still be a prominent feature) and had more cellular atypia with rhabdoid features and tended to occur in more proximal/axial locations. They had worse prognosis, were larger in primary occurrence (7.8cm) (Hasegawa et al., 2001) and occurred in somewhat older adults (mid 30s in proximal type epithelioid sarcoma vs. mid 20s of classical epithelioid sarcoma)(L. Guillou et al., 1997; Hasegawa et al., 2001). Electron microscopy of proximal epithelioid sarcoma reveals features similar to Fisher’s cell types (c) and (d) and thus also to malignant rhabdoid tumor (L. Guillou et al., 1997; Hall et al., 1980). Again at times, well developed desmosomal cellular junctions were observed hinting towards epithelial differentiation(L. Guillou et al., 1997). It is important to note that several “proximal” epithelioid sarcoma were included in the Chase and Enzinger’s 1985 study of 241 cases(Chase & Enzinger, 1985) and earlier case reports (Gallup et al., 1976; Hall et al., 1980), and were not considered as a separate entity. In fact upon review of prior literature, Guillou et al. identified 55 additional cases of proximal-type epithelioid sarcoma that were previously diagnosed as epithelioid sarcoma or extra-renal malignant rhabdoid tumors initially. However, because the proximal cases tend to metastasize earlier and seemed more aggressive, Guillou et al. emphasized the value of distinguishing this new subclass of epithelioid sarcoma as a prognostic indicator(L. Guillou et al., 1997). The proximal-type classification is a 12 histologic description rather than a clinical one: even cases in the hand could be “proximal” (L. Guillou et al., 1997). It is suggested that histologically proximal type epithelioid sarcoma is an intermediate between extra-renal malignant rhabdoid tumor and classic epithelioid sarcoma(L. Guillou et al., 1997). In fact, differentiating proximal epithelioid sarcoma from extra-renal malignant rhabdoid tumor can be very difficult and controversial(Chase, 1990). The need to differentiate between these two tumor types again comes down to the worse prognosis of extra-renal malignant rhabdoid tumor even when compared to proximal epithelioid sarcoma(Chase, 1990; Perrone, Swanson, Twiggs, Ulbright, & Dehner, 1989). Yet some argue that outside the kidney, the term rhabdoid tumor is a phenotypic description and not a specific entity and thus proximal epithelioid sarcoma is effectively the same tumor that others have labeled as extra-renal malignant rhabdoid tumor(D. R. Guillou, 1999). Despite this, there are differences that could be pointed out. There is higher percentage of ERG and CD34 positivity in proximal epithelioid sarcoma cases; these proteins rarely stain in malignant rhabdoid tumors(Kohashi, Yamada, et al., 2014). Furthermore, proximal epithelioid sarcoma has weaker staining for SALL4 and glypican-3 (Kohashi, Yamada, et al., 2014; Yoshida et al., 2014). SALL4 immunohistochemistry strongly stains extra-renal malignant rhabdoid tumor, weakly stains proximal epithelioid sarcoma and does not stain classic epithelioid sarcoma(Kohashi, Yamada, et al., 2014). SALL4, a stem-cell associated transcription factor that is only expressed in germ cells in adults, is a marker for nonteratomatous germ cell tumors(Miettinen et al., 2014). On the other hand both ERG and CD34 are vascular markers and their positivity in some epithelioid sarcoma and lack of expression in 13 malignant rhabdoid tumors hints towards perhaps a more vascular endothelial differentiation tendency in epithelioid sarcoma. Positive staining for FLI1 (an homologous transcription factor to ERG) in most epithelioid sarcoma cases is also supportive of this partial vascular differentiation(Stockman et al., 2014). Generally speaking, proximal type epithelioid sarcoma occurs in adults and has a predilection for genitalia whereas extra renal malignant rhabdoid tumor is primarily a tumor of children and can occur in a variety of sites. Smooth muscle actin (SMA) is another possibly differentially expressed protein. Most cases of atypical teratoid/rhabdoid tumor(Rorke, Packer, & Biegel, 1996) and malignant rhabdoid tumor(Kato et al., 2003) show positive SMA staining whereas the majority of both conventional and proximal epithelioid sarcomas are negative for SMA(L. Guillou et al., 1997; Hasegawa et al., 2001). Although some histologic variants of epithelioid sarcoma have been described, generally speaking epithelioid sarcoma, particularly in its classic form, has a relatively homogenous histology and clinical presentation. The existence of such variants could point towards a common progenitor tumor initiating population with a divergent yet inhibited differentiation fate (evidenced by the multiple phenotypes seen in Fisher’s EM studies) or a diverse set of tumor initiating cells that converge in phenotype through a dedifferentiation mechanism (possibly loss of SMARCB1). Given that almost all epithelioid sarcomas have lost SMARCB1 and that there are a diverse set of tumors in other body parts with SMARCB1 loss, which show some similarities with each other and with epithelioid sarcoma, the second possibility has strong support. However, multiple lineages and teratomatous histologies are also described in a variety of SMARCB1 14 negative tumors, particularly atypical teratoid/rhabdoid tumor. Therefore, perhaps there is validity to both scenarios. 1.3.3 Epidemiology and current therapies Because of the rarity of epithelioid sarcoma, the majority of the literature has inherently been from small scale studies often from single referral institutions. Thus most of what has been discussed so far is observational and could fail rigorous statistical scrutiny. This includes Chase and Enzinger’s study of 241 cases(Chase & Enzinger, 1985), which generally had a univariate look at factors affecting patient outcome. Fortunately however, a key study by Jawad et al., used data collected by the Surveillance, Epidemiology, and End Results (SEER) initiative of the National Cancer Institute of United States of America, looking at a collection of 441 epithelioid sarcoma cases (Jawad et al., 2009). The SEER program was initiated with the National Cancer Act of 1971 and is regarded as the standard of quality by cancer registries globally(Howlader, Noone, Yu, & Cronin, 2012). This study by Jawad et al. is the largest study on epithelioid sarcoma to date with perhaps the greatest statistical rigorousness. It should be noted that Jawad et al. included but did not separate variants of epithelioid sarcoma (such as proximal type) in their study. Jawad et al. reported the annual incidence of epithelioid sarcoma to be 0.041 per 100,000 persons. They also found the disease specific survival to be 68% at 5 years and 61% at 10 years(Jawad et al., 2009). Therefore, for conceptualization purposes, one could estimate that 15 individuals are diagnosed with epithelioid sarcoma in Canada annually, a third of whom would die from the disease within 5 years. By the same logic, globally about 3000 individuals are diagnosed with epithelioid sarcoma every year and 1000 of 15 them will succumb to the disease within 5 years if they get first world medical care. Furthermore, the mainstay of therapy is surgical resection with amputations being a common result, particularly in regions which lack sarcoma subspecialty care centres. Therefore, although rare, epithelioid sarcoma certainly has a burden on society that cannot be ignored. Surprisingly, Jawad et al. found that when adjusted for age, the incidence of epithelioid sarcoma is highest in older populations peaking at 75+ year of age. This is contradictory to almost all the literature on epithelioid sarcoma, which categorizes this sarcoma as a disease of young adults(Jawad et al., 2009). However, almost no other study does an age-adjusted analysis of incidence. Similar to other reports, the SEER data set had the largest number of patients with epithelioid sarcoma in the 17-30 years of age group (~24%). However, when normalized to the 2000 US Standard population distribution, the 75+ age group had the highest incidence despite constituting only 10% of all epithelioid sarcoma patients. Using Cox-regression, and therefore attempting to tease out individual contributing factors in a multivariable scenario, the significant prognostic indicators were age, stage, lymph node involvement, and surgical resection. Younger individuals did better, distant spread of disease and lymph node metastasis meant worse outcome, and surgical resection led to better results. Importantly, axial/proximal location, on its own, did not indicate worse outcomes in a statistically significant manner. The previously reported observation of worse outcomes in proximally located epithelioid sarcoma is likely a result of difficulty and hence reduced number of surgical resections. Moreover, epithelioid sarcoma was confirmed to be more common in 16 males (57%), however, gender had no prognostic indications when other variables were taken into account. Wide surgical resection and amputation are treatments of choice when dealing with epithelioid sarcoma despite the resulting morbidities. This is particularly because of the high recurrence rate (77%) with marginal resection(Chase & Enzinger, 1985). In terms of chemotherapeutics, there is no agreed upon standard of care. However, adjuvant doxorubicin and ifosfamide have been used and are not effective(Wolf, Flum, Tanas, Rubin, & Mann, 2008). Adjuvant radiotherapy has also been used but without proven benefits in achieving local control or enhancing survival(Wolf et al., 2008). Therefore, treatment of epithelioid sarcoma is in need of improvement if the 10-year mortality of 50% is to be reduced. In fact, in 2009, Jawad et al. showed no improved survivorship since 1986. There are two main streams where better treatment can be achieved: earlier more accurate diagnostics that would ultimately lead to improved surgical outcomes, and improved medical treatments in the form of effective drugs or other modalities. With greater knowledge about this disease, the first approach can be expected to improve over time. However, the indolent presentation of epithelioid sarcoma, and the low likelihood of early detection particularly when the primary cases are small hypodermal masses (a common scenario in epithelioid sarcoma), means that the greatest potential for progress is likely to be made if we can develop medical therapies that can control metastatic disease. Hence understanding the pathogenesis and biology of epithelioid sarcoma, which is the only way of rationally developing therapies for this tumor, is what we have focused on. To date, one of the most important and consistent biological findings in epithelioid sarcoma is the loss of the tumor suppressor SMARCB1. Understanding the mechanism of 17 its loss and its role in the causation of epithelioid sarcoma could help us in understanding this pathology better. 1.4 SMARCB1 and epithelioid sarcoma Many karyotypic reports, usually single cases, have been reported on epithelioid saroma. A summary of such studies can be found in Table 1-2. From these, the most prominent region undergoing some sort of inactivating change (in the form of fusion or loss) was in 22q11. In 2005, Modena et al. paid particular attention to 22q11. Using multiple fluorescent in situ hybridization (FISH) probes in short term cell cultures of two proximal epithelioid sarcoma cases with prior evidence of 22q11 breakpoints, and additional work with comparative genomic hybridization (CGH) arrays, they narrowed the region of fusion on chromosome 22 to 150 kilo basepairs (kbp)(Modena et al., 2005). Interestingly, a gene previously suggested to be a tumor suppressor in atypical teratoid/rhabdoid tumors and malignant rhabdoid tumors, namely SMARCB1 was found in this region. Expanding their cohort to 11 cases, they found loss of SMARCB1 protein (also known as INI1 or BAF47 or hSNF5) expression in 6 cases (5/6 proximal and 1/5 classic epithelioid sarcoma) using IHC. This breakthrough in epithelioid sarcoma biology needed validation and in 2009, Hornick et al. reported one of largest epithelioid sarcoma immunohistochemistry studies ever published, with 136 cases of epithlioid sarcoma examined for SMARCB1 protein expression(Hornick et al., 2009). They found that 91% and 95% of classic and proximal epithelioid sarcomas respectively had complete loss of SMARCB1. This was a diagnostic breakthrough as now a unifying theme bound all epithelioid sarcomas together and 18 perhaps the 5-9% of cases staining for SMARCB1 were actually different entities with some histologic similarities to epithelioid sarcoma. 1.4.1 SMARCB1: a tumor suppressor gene Before the association of SMARCB1 with epithelioid sarcoma, the first link between SMARCB1 and tumorigenesis and a direct link between the SWI/SNF complex and cancer was established in 1998 in cell lines derived from malignant rhabdoid tumors of kidney and soft tissue(Versteege et al., 1998). These cell lines (including G401 and MON which are still commonly used today) were examined for loss of DNA in the 22q11.2 region. Previous cytogenetic studies of malignant rhabdoid tumors had indicated this region to be deleted commonly, hinting at the potential presence of a tumor suppressor. Versteege et al. used polymerase chain reaction (PCR) amplification of segments across 22q11.2 and were able to narrow the commonly lost region of the cell lines to the markers A006E25 (or SMARCB1), SGC32593, MMP11 and GCT10. They focused on SMARCB1 because DNA from one of the lines that had maintained the aforementioned markers (cell line KD) showed altered migration upon southern blotting against SMARCB1 when treated with EcoRI restriction enzyme. This suggested a small deletion in the SMARCB1 locus and thus allowed further narrowing of the search for the tumor suppressor gene in 22q11.2. Soon afterwards, Biegel et al., also identified somatic and germ-line mutations in SMARCB1 in clinical samples of atypical teratoid/rhabdoid tumor (18 cases) and renal/extra renal malignant rhabdoid tumors (7 and 4 cases respectively)(Biegel et al., 1999). About half of the 29 cases examined had homozygous deletions of SMARCB1 and the other half had inactivating point mutations. Four children had germ-line mutations of whom one was diagnosed with atypical teratoid/rhabdoid 19 tumor of brain and the other three with renal malignant rhabdoid tumor. Almost all cases had a very quiet karyotype with a diploid genome; however, interestingly one of the older children, who was 6 years old with an atypical teratoid/rhabdoid tumor, had a complex genome with 61 chromosomes and multiple aberrations(Biegel et al., 1999). Mouse models support a strong tumor suppressor role for SMARCB1. Homozygous Smarcb1 deletions are embryonically lethal, whereas Smarcb1+/- mice spontaneously develop soft tissue tumors (mainly in the head and neck) with prominent rhaboid cells that stain strongly for cyclin D1 (CCND1), SMA, and vimentin − all of which are negative in the adjacent normal brain tissue(Tsikitis, Zhang, Edelman, Zagzag, & Kalpana, 2005). A small percentage of mice also develop tumors in the hind limb. The tumors have mutations in the second Smarcb1 allele thus fitting the classic Knudson two-hit model. Conditional interferon based Mx-Cre* Smarcb1 homozygous deletion, which circumvents the problem of embryonic lethality, leads to rapid formation of CD8+ lymphomas with rare instances of rhabdoid tumor formation. 1.4.2 SMARCB1 inactivated tumors SMARCB1 loss was first established in malignant rhabdoid tumors of the kidney and the atypical teratoid/rhabdoid tumors of the brain. As briefly mentioned earlier, rhabdoid is a descriptive term referring to their histology. It is a derivative of the term rhabdomyosarcomatoid which in turn was first used by Beckwith† and Palmer(Beckwith & Palmer, 1978) to describe a very unique variant of Wilm’s tumor in the National Wilm’s Tumor Study. They used the term to describe the resemblance of the tumor cells * According to authors, this system results in Cre-based excision in almost all tissues except the brain † An interesting side note, the Beckwith-Wiedemann Syndrome is named after the same author 20 to rhabdomyoblasts‡: abundant eosinophilic cytoplasm with single sometimes-eccentric vesicular and large nuclei that had a single prominent nucleolus. Rhabdomyoblasts not surprisingly are a particular feature of rhabdomyosarcoma. However, Beckwith and Palmer et al., realized that the aforementioned variant did not show evidence of primitive sarcomere formation when looked at with electron microscopy and thus was ultra-structurally distinct from rhabdomyosarcoma. They established a new pathology, clearly distinct from Wilm’s tumor or rhabdomyosarcoma, and named it malignant rhabdoid tumor of the kidney in reference to its light microscopic features, and aggressive clinical behavior(Haas et al., 1981). Shortly thereafter, they also described co-occurrence of renal malignant rhabdoid tumors with central nervous system(CNS) tumors(Bonnin, Rubinstein, Palmer, & Beckwith, 1984) in six children under two years of age. This phenomenon would later be described as Rhabdoid Tumor Predisposition Syndrome and linked to germ-line mutations in SMARCB1 where a second-hit led to tumor formation(Sevenet et al., 1999). In fact, 13.5% of all patients with malignant rhabdoid tumor of kidney have a CNS tumor without a clear temporal relation to the kidney tumor(Weeks, Beckwith, Mierau, & Luckey, 1989). The CNS tumors would be established as a new type of neoplasm that often occurred in the cerebellum. Because in addition to rhabdoid cells, the CNS tumors showed multiple differentiation patterns (neuroepithelial, epithelial, mesenchymal, etc.) within a single case yet lacked germ cell markers such as alpha fetoprotein, they were named atypical teratoid/rhabdoid tumors(Rorke, Packer, & Biegel, 1995). Interestingly both atypical teratoid/rhabdoid tumor and kidney malignant rhabdoid tumor were also reported to have a predilection for ‡ The common root of these terms, rhabdos, is Greek for “rod” in reference to skeletal muscle, which in histology has abundant cytoplasm and eccentric nuclei. 21 the male gender (Rorke et al., 1995; Weeks et al., 1989) who tended to do worse but as we saw with epithelioid sarcoma, rigorous statistical analyses may not necessarily support this notion. Uniquely, malignant rhabdoid tumor of the kidney shows a very diverse histological presentation with eight major groups (spindled, epithelioid, sclerosing, lymphomatoid, etc.) and sometimes there is a mixed pattern presentation(Weeks et al., 1989). Atypical teratoid/rhabdoid tumor on the hand has four prominent histologic subtypes (rhabdoid, primitive neuroectodermal-tumor like, epithelilal with glandular and squamous formations, and mesenchymal) which usually presents in a mixed fashion(Rorke et al., 1995). Metastasis of malignant rhabdoid tumor is mainly to the lungs while atypical teratoid/rhabdoid tumor commonly progresses to widespread CNS metastasis(Rorke et al., 1995; Weeks et al., 1989). Malignant rhabdoid tumor and atypical teratoid/rhabdoid tumor are highly aggressive with 80% lethality within a year of diagnosis(Hollmann & Hornick, 2011). A curious finding in a few cases of malignant rhabdoid tumor has been excessive parathyroid hormone production (and resulting hypercalcemia) that led Beckwith to suspect a neuroendocrine cell of origin which was not supported by staining for neural and neuroendocrine markers(Weeks et al., 1989). This is interesting as a group of ovarian tumors known as small cell carcinoma of the ovary are known to present with hypercalcemia in 50% of the cases and these have inactivation of SMARCA4, another member of the SWI/SNF chromatin remodeling complex(Jelinic et al., 2014; Kupryjanczyk et al., 2013; Ramos et al., 2014; Witkowski et al., 2014). Rare rhabdoid tumor predisposition syndrome(Schneppenheim et al., 2010), atypical teratoid/rhabdoid tumor(Hasselblatt et al., 2011) and familial immature ovarian 22 teratomas(Witkowski et al., 2013) with SMARCA4 loss but maintained SMARCB1 have also been described. In addition to malignant rhabdoid tumor of the kidney, extra renal sites such as the liver, the lip, the retroperitoneal space, the pericardium, the orbit and many other sites have been noted to develop tumors with features of rhaboid tumors including co-expression of EMA and vimentin and prominence of rhabdoid cells. These tumors became known as extra renal malignant rhabdoid tumor; however, Parham et al.. believed these to be a collection of multiple entities with diverse clinical features and different immunohistochemistry patterns(Parham, Weeks, & Beckwith, 1994). Some of these would later be recognized as their own entity with established loss of SMARCB1. Epithelioid sarcoma however has features that set it apart from these related entities. These are the unique pseudogranulomatous pattern, tendency to occur in older individuals, lymphogenous metastasis, slower growth, positivity for some vascular markers such as CD34 and ERG in half of cases, and negativity for SALL4 which is a germ cell marker positive in most malignant rhabdoid tumors and atypical teratoid/rhabdoid tumors. SMA expression is also more prominent in malignant rhabdoid tumors and atypical teratoid/rhabdoid tumors. Other tumors known to have lost SMARCB1 in most cases include cribriform neuroepithelial tumors, renal medullary carcinoma, adult sellar SMARCB1-deficient tumors (previously thought of as adult atypical teratoid/rhabdoid tumor variant), small cell undifferentiated variant of hepatoblastoma and gastrointestinal carcinoma(Agaimy, 2014; Hollmann & Hornick, 2011). Additionally 50% of epithelioid malignant peripheral nerve sheath tumors (MPNST), 10-40% of myoepithelial carcinoma, 28% of 23 undifferentiated pancreatic rhabdoid carcinoma, 16% of extraskeletal myxoid chondrosarcoma, 15% of collecting duct carcinoma, 3% of sinonasal carcinoma (which have been defined as a new entity named sinonasal basaloid carcinoma(Agaimy et al., 2014)), 0.7% of primary osteosarcomas, 0.6% of all primary bone tumors, and a variable portion of poorly differentiated chordomas also lose expression of SMARCB1 completely (Agaimy, 2014). Mosaic patterns of SMARCB1 loss have also been described in 70% of gastrointestinal stromal tumors (GIST), 74% of ossifying fibromyxoid tumors, and up to 50% of sporadic and familial schwannomatosis. Synovial sarcomas also have been noted to have a consistent pattern of reduced but not lost SMARCB1 expression(Kohashi et al., 2010). The study by Kohashi et al. examined 95 synovial sarcomas (with 85% of 47 tested cases showing a SS18-SSX fusion) and found reduced SMARCB1 levels in 70% of cases. They also noted higher prevalence but not exclusive loss of SMARCB1 with the biphasic form of synovial sarcoma. There was no prognostic indication with reduced levels of SMARCB1(Kohashi et al., 2010). Future studies would mechanistically link loss of SMARCB1 in synovial sarcoma with the fusion oncoprotein SS18-SSX(Kadoch & Crabtree, 2013). A significant portion of the aforementioned tumors show rhabdoid features and might have historically been grouped into the extra-renal malignant rhabdoid tumor category; however, the rhabdoid cell morphology is not a universal consequence of SMARCB1 loss. Examples include cribriform neuroepithelial tumor, collecting duct carcinoma, renal medullary carcinoma, epithelioid MPNST, myoepithelial carcinoma, synovial sarcoma, and ossifying fibromyxoid tumors that show little to no evidence of rhabdoid cells yet do have reduced SMARCB1 levels. 24 1.4.3 SMARCB1: a core SWI/SNF member SMARCB1 is one of the core members of the SWI/SNF chromatin remodeling complex. The members of SWI/SNF were first identified in genetic screenings of the yeast Saccharomyces cerevisiae. A group of genes were found to be needed for the expression of a central sucrose hydrolyzing enzyme and thus the mutants were referred to as sucrose non-fermentor (SNF)(Neigeborn & Carlson, 1984). Around the same time, another group of genes, with some overlaps with the SNF genes, were identified to be necessary for expression of the homothallic switching endonuclease and therefore mating-type switching (SWI)(Stern, Jensen, & Herskowitz, 1984). Mutations in individual SWI/SNF members led to a similar colony growth-deficiency phenotype as did mutations in other/all members (Peterson & Herskowitz, 1992). Furthermore, the stability of Swi3 (homolog of human SMARCC1 and SMARCC2) was reduced upon mutations in Swi1 (homolog of human ARID1A) or Swi2/Snf2 (homolog of human SMARCA4/SMARCA2) on Western blots(Peterson & Herskowitz, 1992). All members also independently localized to the nucleus. Because of these four lines of evidence, i.e. similar knock down phenotypes for all or single members, co-localization, co-regulation of target genes, as well as instability upon mutations of other members, it was suspected that the SWI/SNF proteins were part of a complex. Interestingly, knockdown of histones led to a reversal of the SWI/SNF mutant phenotypes of reduced growth and expression of genes known to be regulated by SWI/SNF (Peterson & Herskowitz, 1992). Thus from early on a link between SWI/SNF and histone regulation and transcription was established. In the early 1990s, Roger Kornberg§ and his group provided solid evidence § R. Kornberg also described the Mediator complex necessary for transcription of all class II genes for which he received the 2006 Noble Prize in Chemistry 25 through immunoprecipitation (IP) studies that indeed the SWI and SNF proteins (12 proteins) physically interacted with one another in yeast and that DNA was needed for the ATPase activity of the complex(Cairns, Kim, Sayre, Laurent, & Kornberg, 1994). This DNA-dependent activity was likely from Swi2/Snf2 as it has homologous sequences to other DNA-dependent ATPases. Interestingly, the sequence of the several DNA samples used did not seem to affect the activity significantly unless it was a monotonous repeat of deoxyinosine and deoxycytosine(Cairns et al., 1994). However, double stranded DNA (in supercoiled or open form) led to increased ATPase activity of the complex compared to single stranded DNA(Cairns et al., 1994). Yet none of the SWI/SNF members have DNA binding ability on their own. However there is evidence for sequence specific activity of the complex in vitro. Transfection of the SWI/SNF ATPase in a null background enhances transcription of Glucocorticoid Receptor (GR) promoters but does not do so for AP-1 (targets of FOS ATF and JUN), NFAT (immune response), HNF-1a (hepatocyte-specific genes), CMV, or EF1a promoters(W. Wang et al., 1996). Mutation screenings in Drosophila melanogaster looking for reversal of the mutant Pc (polycomb, homologous to human CBX2) homeotic phenotypes ** , identified brahma(brm) as one of 16 genes with such capability. The rest included mainly trithorax members and interactors of the mediator complex or RNA polymerase II(Kennison & Tamkun, 1988). Brahma was recognized as a homolog of Swi2/Snf2 and was also shown to interact with trx (trithorax, homologous to human KMT2A). Both trx and Swi2/Snf2 were necessary for activation of genes otherwise dominantly suppressed by Pc (Tamkun et al., 1992). Within a year of the drosophila studies, the human homologs of brm, ** Homeotic phenotype are misplacement of anatomic compartments 26 namely Brahma Related Gene 1 (BRG1 or SMARCA4) as well as the human BRM (also known as SMARCA2) were cloned by the Crabtree and Yaniv groups (Khavari, Peterson, Tamkun, Mendel, & Crabtree, 1993; Muchardt & Yaniv, 1993). Human homolog of SNF5 was also cloned and shown to bind to and be necessary for the integration of the Human Immunodeficiency Virus inegrase enzyme. Thus it was called Integrase Interactor 1 or INI1 (Kalpana, Marmon, Wang, Crabtree, & Goff, 1994) and would later be found to be the same as brahma-associated factor 47 KDa in weight (BAF47) or SMARCB1(W. Wang et al., 1996). Similar to the findings in yeast and drosophila, the human homologs of SWI/SNF could bind to each other and enhance the activity of the core enzymatic members SMARCA2 or SMARCA4. Biochemical studies showed a role of SWI/SNF in assembly and mobilization of nucleosome in the presence of ATP and thus SWI/SNF was established as a chromatin-remodeling complex (Yaniv, 2014). In addition to SWI/SNF, there are about 20 subfamilies of chromatin remodeling complexes in humans defined based on their core ATPases. Some of the better studied complexes other than SWI/SNF include INO80, CHD, and ISWI††(Narlikar, Sundaramoorthy, & Owen-Hughes, 2013). The SWI/SNF complex is defined by its core ATPase which is Swi2/Sn2 in yeast, brm in drosophila, and SMARCA4 or SMARCA2 in mammals. SMARCA4 and SMARCA2 are highly homologous to one another (>70%) and also to SWI/SNF ATPases of other species. Additional evidence for the high conservation of the complex across species is the capacity of human SMARCA4 to partially rescue the Swi2/Snf2- phenotype in yeast(Khavari et al., 1993). Moreover, the ATPase domains of yeast Swi2/Snf2 and drosophila brm are functionally interchangeable (Elfring, Deuring, McCallum, Peterson, †† ISWI (Imitation SWI), encoded by SMARCA1 or SMARCA5 in mammals, is the central ATPase of the ISWI family of complexes which includes the NURF(SMARCA1), CHRAC(SMARCA5), and ACF complexes. 27 & Tamkun, 1994). The assembly of the complex is still a mystery but is known to be ATP-independent (Peterson, Dingwall, & Scott, 1994). The members can still associate with each other in the absence of core ATPases SMARCA4 or SMARCA2(W. Wang et al., 1996). The molecular weight of the SWI/SNF complex is thought to be 2MDa. Liquid chromatography from yeast protein extracts has shown that the peak Western blot signal for the members occurs at approximately 2 MDa, however this is not the only size where prominent signal is seen. In fact smaller sizes as low as 600kDa show presence of multiple members (Peterson et al., 1994). Knockdown of individual members leads to loss of the peak signal at the 2MDa fraction, yet sub-2 MDa complexes still form (Peterson et al., 1994). Immortalized human T lymphocytes (Jurkat) show SMARCA4 signal mainly at the 2 MDa fraction (Khavari et al., 1993). The SMARCA4 bands in Jurkat cells (Khavari et al., 1993) seem to be much more limited to the 2MDa range compared with Swi2/Snf2 in yeast(Peterson et al., 1994). This could be because of the difference in species, but more likely it is a result of different chromatography conditions used. Regardless, the most prominent and stable form of the SWI/SNF complex across multiple species, given wild-type constituent members, is 2 MDa in size. Interestingly, the sum of the molecular weights of individual members of the largest described complex to date(Kadoch et al., 2013), which would have 15 members, is 1.4 MDa. Thus there might still be some critical members or combinations of existing members that have not been identified which would account for the 30% missing weight. In vitro studies show that in the presence of ATP, SWI/SNF mobilizes histones relative to DNA sequences providing access for restriction endonucleases (Francis, Saurin, Shao, & Kingston, 2001). This opening up of chromatin, which can be 28 dominantly overcome by the polycomb complex, is thought to enable transcription. There are about 2000 SWI/SNF complexes in a mammalian cell, and given that a much larger number of genes are normally expressed in differentiated cells, Wang et al. suggest that logically SWI/SNF would not be a mandatory component of transcription activation and maintenance machinery(W. Wang et al., 1996). Supportive of this hypothesis are cell lines such as the pheochromocytoma line SW13, which lack SWI/SNF yet are capable of transcription and cell division. Furthermore, only a third of the yeast genome is regulated by SWI/SNF(Sudarsanam, Iyer, Brown, & Winston, 2000). The exact mechanism of chromatin remodeling is a field of its own with a focus on ISWI and RSC complexes. Details would be beyond the scope of this thesis; however, the simplified version of one theory is an initial pulling of the DNA from one end, creation of an unstable intermediate, followed by addition of DNA from the other end (Narlikar et al., 2013). Another important feature is that both SMARCA2 and SMARCA4 possess bromodomains and hence recognize acetylated histone tail lysines, a mark associated with active chromatin. Although the early studies in drosophila and yeast mainly focused on the role of SWI/SNF in transcriptional activation, increasing evidence points towards the involvement of the complex in gene silencing as well. For example, both SMARCA4 and SMARCB1 are needed for the silencing of CD4 as well as activation of CD8 in cytotoxic T cell differentiation(Chi et al., 2002). Deletion of SMARCB1 in fibroblasts also seems to lead to more genes being activated than silenced(Isakoff et al., 2005). This is also seen in yeast where the inactivating function of SWI/SNF is independent of the SWI/SNF activated genes(Sudarsanam et al., 2000). One suggested mechanism is the recruitment of HDACs by SWI/SNF members to specific genomic sites which leads to removal of acetyl 29 groups from histone tail lysines leading to positively charged tails, and tighter histone DNA interaction. This in turn limits the DNA accessibility for transcriptional complexes. Such a mechanism is supported by the HDAC1-dependent silencing of CCND1 by SMARCB1 in the extra renal malignant rhabdoid tumor cell line MON(Zhang et al., 2002). Another conceptual hypothesis is that with the sliding open of histones at one site, there will be an inherent tightening and closing of chromain at other sites similar to movable beads on an abacus. Diversity of the SWI/SNF complex is well recognized. Of the 26 genes encoding composing members, different combination are found within a single cell type and across cell types. Figure 1-2 outlines some of the members and their association with specific cancer types. One key feature is that the complexes either possess SMARCA4 (BRG1) or SMARCA2 (BRM) as their core ATPases in a mutually exclusive form (W. Wang et al., 1996). Fruthermore, there are three core members that are normally always isolated with the core ATPase: these are SMARCB1 (or BAF47), SMARCC1 (or BAF155), and SMARCC2 (or BAF170). Two main types of the complex have been described: the BAF (BRG1 associated factors) and the PBAF (Polybromo BRG1 associated factors) (Figure 1-2). The BAF complex can possess either SMARCA4 or SMARCA2 and has the unique subunits of BRD9, ARID1A or ARID1B, BCL7 (A or B or C), and BCL11 (A or B). The PBAF complex on the other hand only possesses SMARCA4 as the ATPase, with BRD7, PBRM1 (polybromo 1) and ARID2 as the unique subunits(Wilson & Roberts, 2011). The BAF complex is likely 5 to 10 times more abundant than the PBAF(W. Wang et al., 1996) and can be present in the same cell line. PBAF also possesses a greater number of charged amino acids as it is separated from BAF by increasing the counterion 30 concentration in chromatography(W. Wang et al., 1996). Interesting switches between complex forms have been linked to cell states as well. During neuronal differentiation there is a microRNA mediated switch of ACTL6A with ACTl6B, PHF10 with DPF1, and SS18 with SS18L1 when cells transition from neural progenitors to mature neurons(Yoo, Staahl, Chen, & Crabtree, 2009). The complex found in embryonic stem cells also seems to lack BAF170 but possess two BAF155 units per complex(Ho et al., 2009). Abnormalities in SWI/SNF have fundamental developmental implications at the organismal level as well. Heterozygous mutations in the helicase domain of SMARCA2 have been linked to the Nicolaides-Baraitser syndrome which is characterized by sparse hair, distinctive facial morphology, distal limb anomalies and intellectual disability(Van Houdt et al., 2012). Furthermore, Coffin-Siris syndrome, defined by growth deficiency, intellectual disability, microcephaly, coarse facial features and hypoplastic nail of the fifth finger/toe is also associated with germline mutations in various SWI/SNF members(Tsurusaki et al., 2012). Moreover, other than SMARCB1 and SMARCA4, biallelic mutations in several other SWI/SNF members have been linked to cancer (Figure 1-2). In fact, recent sequencing data have also suggested that 20% of all malignancies have mutations in the SWI/SNF complex(Kadoch et al., 2013); however, whether all these mutations are indeed impactful is yet to be established. 1.4.4 SWI/SNF and cancer Interestingly, among all the chromatin remodeling complexes, only SWI/SNF seems to have a propensity for mutations linked to cancer(Kadoch et al., 2013). In all reported cases, this has been in a loss of function context. The mutations of individual members tend to be disease specific as well. Figure 1-2 illustrates a summary of 31 SWI/SNF members and cancers associated with their mutations. The lack of a link between other chromatin remodeling complexes, such as INO80 and ISWI, and cancer is curious. A lot has yet to be established on the specialty function of chromatin remodeling complexes but one possibility could be the lower abundance of SWI/SNF relative to some of the other chromatin remodelers such as the ISWI family(W. Wang et al., 1996). This could make SWI/SNF aberrations perhaps smaller in magnitude and more tolerable. However, homozygous knockout mouse models show that many but not all SWI/SNF members, including SMARCA4, SMARCB1, SMARCC1, SMARCC2, PBRM1, ACTL6A, ACTL6B, SMARCE1, SS18, BCL11A, and BCL11B are necessary for organismal development. Their homozygous deletions lead to death from early embryonic stages to a few days postpartum with various organ malformations. It is important to distinguish organismal survival from cell survival however, as obviously many cancer cells can not only tolerate loss of SWI/SNF members, but proliferate more with such loss. SWI/SNF seems to affect many pathways that could explain the oncogenicity of its aberrations. To simplify, these pathways can be grouped into two broad categories: stemness and specific oncogenic programs. Every cell differentiation state seems to have a unique SWI/SNF associated with it. SWI/SNF with two SMARCC1s and no SMARCC2s is found in embryonic stem cells(Ho et al., 2009). SWI/SNF with ACTL6B (BAF53b) leads to neuronal differentiation(Yoo et al., 2009) and the complex with DPF3(BAF45c) is unique to muscles(Lange et al., 2008). The complex plays a critical role in MyoD-mediated muscle differentiation (de la Serna, Carlson, & Imbalzano, 2001). SMARCA2, SMARCA4, ARID1A, and ARID1B have all been also implicated in osteoblast differentiation(Flowers, Nagl, Beck, & Moran, 2009; Nagl et al., 2005). Thus 32 it would not be surprising that a cell would lose its differentiation directionality with abnormalities in SWI/SNF. SMARCB1 overexpression on the other hand antagonizes OCT4 at its targets on a genome wide basis in the germ-cell tumor cell line NCCIT(You et al., 2013). Thus loss of SMARCB1 could mean an enhancement of the OCT4 pluripotency programs. SWI/SNF can also antagonize the polycomb complexes. A direct antagonism of the PRC1 and PRC2 vs. SMARCB1 was established in malignant rhabdoid tumor cell line MON at the promoter of the tumor suppressor CDKN2A(Kia, Gorski, Giannakopoulos, & Verrijzer, 2008). This opposition was later suggested to be a global event in mouse embryonic fibroblasts where loss of Smarcb1 led to a direct upregulation of Ezh2 as well(Wilson et al., 2010). Polycomb proteins, such as EZH2, contribute to the silencing of lineage specific genes and their dominance over SWI/SNF can lead to a stem-like state. Abnormalities in differentiation and stem cell-like features of SWI/SNF mutated tumors partly explain their malignancy. Specific tumorigenic pathways have also been associated with SWI/SNF. SMARCB1 is one of the strongest protein interactors of GLI1 and through direct binding, inhibits GLI1 hence downregulating its target genes. As a consequence the sonic hedgehog pathway is upregulated in atypical teratoid/rhabdoid tumors(Jagani et al., 2010). SMARCB1 also has control over cell cycle by regulating the expressions of CDKN2A and CCND1. Its loss leads to downregulation of CDKN2A(Kia et al., 2008) and upregulation of CCND1(Tsikitis et al., 2005) leading to active cell cycling and replication. In fact SWI/SNF can bind to RB itself and facilitate downregulation of RB targets such as E2Fs and CCND1(Trouche, Le Chalony, Muchardt, Yaniv, & Kouzarides, 1997). SMARCB1 loss also seems to lead to aberrant activation of Wnt signaling(Mora- 33 Blanco et al., 2014) and SMARCA4 can modulate transcription of Wnt target genes in endothelial cells(Griffin, Curtis, Davis, Muthukumar, & Magnuson, 2011). SMARCB1 enhances activation of MYC targets as well(Cheng et al., 1999). Additional tumorigenic pathways linked to SWI/SNF include Polo-like Kinase, Aurora A kinase, as well as Rho-A/ROCK1 control of cell migration(Kim & Roberts, 2014; Wilson & Roberts, 2011). Deep sequencing of rhabdoid tumors has revealed that, in many cases, loss of SMARCB1 without additional mutations can lead to tumorigenesis(Lee et al., 2012). However, mutations in SWI/SNF component are not always singular events. For instance, SMARCA4 mutations co-exist with TP53, CDKN2A, KRAS, NRAS, and LKB1 mutations in non-small cell lung cancer (Wilson & Roberts, 2011). SWI/SNF abnormalities do not have to be a primary driving event either. For instance, secondary loss of SMARCB1 in gliomas has been reported (Jeong, Suh, & Hong, 2014; Kleinschmidt-DeMasters, Birks, Aisner, Hankinson, & Rosenblum, 2011). TP53 mutations also accelerate tumor formation when SMARCB1 is lost(Isakoff et al., 2005). Recently, a dependence of SMARCB1 deficiency on Hippo signaling and upregulation of YAP1 in atypical teratoid/rhabdoid tumors have also been established(Jeibmann et al., 2014). 1.5 Other deregulated pathways in epithelioid sarcoma Thus far we have had an in depth discussion of epithelioid sarcoma and its link to SWI/SNF as well as the capacity of SWI/SNF mutations as oncogenic drivers. However, there are additional cancer pathways reported to be abnormal in epithelioid sarcoma. In a tissue microarray of 27 cases, 77% showed expression of epidermal growth factor receptor (EGFR), 90% of which expressed the active phosphorylated form of EGFR(Xie et al., 2011). Two epithelioid sarcoma cell lines, VAESBJ and Epi544, were shown to 34 have active EGFR under normal serum conditions. This activity was significantly increased with addition of exogenous EGF and led to increase in downstream AKT and ERK phosphorylation(Xie et al., 2011). Treatment of these lines with the EGFR inhibitor erlotinib reduced proliferation, and invasiveness. On the other hand, VAESBJ also revealed significant base line activity in MTOR and its downstream serine/threonine kinase RPS6KB1. This could at least be partly explained by the concurrent loss of PTEN expression in this line. The cell lines were both also sensitive to the MTOR inhibitor rapamycin. Epi544 was more sensitive to erolotinib, whereas VAESBJ was more sensitive to rapamycin. VAESBJ’s relative resistance to erlotinib can be explained by lack of PTEN and increased activity in some of the downstream targets of EGFR. Therefore, inhibition of both the EGFR and MTOR pathways was thought to potentially have some syngergism. The synergistic effect of rapamycin and erlotinib were established in both Epi544 and VAESBJ cell lines in colony formation and apoptotic assays. In vivo tumor growth inhibition and synergism were shown for VAESBJ as well(Xie et al., 2011). Another study confirmed the constitutive MTOR activity and RPS6KB1 phosphorylation in VAESBJ as well as in a newly established angiomatoid epithelioid sarcoma cell line Asra-EPS(Imura et al., 2014). Interestingly, similar to Epi544, Asra-EPS did not have significant PTEN downregulation yet it had a constitutively phosphorylated AKT, even without serum, similar to VAESBJ. This is unlike the Epi544 cell line where AKT phosphorylation would occur only after EGF addition(Xie et al., 2011). Xenograft models confirmed the sensitivity of Asra-EPS to MTOR inhibitors. However, as in VAESBJ and the study by Xie et al., MTOR inhibition would lead to AKT activation in Asra-EPS through a release of negative feedback. Such 35 mechanisms of MTOR inhibitor resistance were suggested to be driven through upstream receptor tyrosine kinases. Therefore, Imura et al. came to the same conclusion as Xie et al. in that in epithelioid sarcoma, MTOR inhibition should be combined with tyrosine kinase inhibitors to overcome resistance mechanisms. Using a proteome profiler array kit (R&D), from a selection of 49 phopho-tyrosine receptor kinases (22 families), Imura et al. found MET to be the only highly active receptor tyrosine kinase in both Asra-EPS and VASESBJ. EGFR activity seemed more limited to VAESBJ while the Wnt signaling co-receptor RYK was highly phosphorylated in Asra-EPS and much less so in VAESBJ. MET overexpression and phosphorylation found in both epithelioid sarcoma lines was absent in synovial sarcoma and human dermal fibroblast cell lines. The individual and synergistic effects of MTOR and MET inhibition were established in vitro and in vivo(Imura et al., 2014). MET is a receptor for the hepatocyte growth factor which is also usually expressed in epithelioid sarcoma tumors (Kuhnen, Tolnay, Steinau, Voss, & Muller, 1998). In adults, MET expression is limited to progenitor cells and is needed for healing and regenerating tissues such as that of the liver. Proliferative and potentially oncogenic pathways such as MTOR and STAT3 can also be activated by MET. CD109 is a negative regulator of TGF-β signaling that has also been found to be upregulated in epithelioid sarcoma cell lines. In an ALDHhigh stem cell-like subpopulation of the proximal epithelioid sarcoma cell line ESX, the CD109 upregulation was discovered through cDNA miacroarray analysis(Emori et al., 2013). It was later shown to be upregulated in two other epithelioid sarcoma cell lines, VAESBJ and FU-EPS-1‡‡ as well as other sarcomas but not in an ALDHhigh specific manner. CD109 ‡‡ VAESBJ is derived from spinal metastasis and FU-EPS-1 from a lymph node metastasis of EpS 36 correlated with poor outcome in soft tissue sarcomas in a multivariate analysis. However, the CD109 “stem population” in ESX seemed to have lower expression of SOX2, OCT4, and NANOG despite higher proliferative capacity(Emori et al., 2013). Interestingly, the original tumor from which ESX was derived had a heterozygous deletion of SMARCB1 and apparent evidence of some SMARCB1 protein expression. Given the reported aberrations in multiple oncogenic pathways in epithelioid sarcoma and the co-occurrence of SWI/SNF member inactivation with other mutations in gliomas and non-small cell lung carcinomas, we decided to explore the genomic landscape of epithelioid sarcoma looking for additional mutations that could be targetable. Major recent developments in next generation sequencing technologies also made such an approach a possibility. 1.6 Next generation sequencing and study of rare tumors Ever since the discovery of DNA and its association with human cancer, scientists and clinicians have dreamt of the possibility to scrutinize it base by base. The ability to sequence DNA, which quickly developed into a robust method by Sanger sequencing(Sanger & Coulson, 1975), was a solid step towards this goal. With the Human Genome Project establishing a map of the human genetic code and rapid advances in computer technology, everything seemed to be in place other than cost and efficiency. Billions of dollars and years of multi-institutional efforts would not make genome-wide nucleotide sequencing an accessible tool for scientists to ask questions on a regular basis and the limited resources were not earmarked for the study of rare specimens. The limitations of Sanger-sequencing were in the termination of polymerase reactions as well as in the need to separate the products of these reactions by gel or other 37 electrophoretic systems(Sanger, Nicklen, & Coulson, 1977). Additionally, preparation of sequencing libraries was necessary via transformation in E. coli or by an incredibly large number of separate PCR reactions. However, with massively parallel sequencing platforms, the first shortcoming was overcome by reversible fluorescent nucleotide addition and imaging (used in Illumina platforms) or through monitoring nucleotide addition via ion detection (used in Ion Torrent platforms (Life Technologies)) both achieved by cyclic manipulation of polymerase or ligase enzymes(Shendure & Ji, 2008). And the second shortcoming was resolved by in vitro library preparation via techniques such as emulsion PCR(Dressman, Yan, Traverso, Kinzler, & Vogelstein, 2003) (Ion Torrent) or bridge PCR on solid surfaces(Adessi et al., 2000; Fedurco, Romieu, Williams, Lawrence, & Turcatti, 2006) (Illumina). With these improvements, the sequencing cost and time requirements have been vastly reduced. Advancements in bioinformatics and the ability to more readily distinguish signal from noise have also increased the feasibility of large and small scale genomic studies. Thus today sequencing genomes and transcriptomes is more accessible and has become a reality for individual laboratories. We are in an incredibly exciting era of molecular medicine where a new “molecular microscope” in the form of massively parallel sequencing, also commonly referred to as next-generation sequencing or second generation sequencing, is giving rise to a whole new paradigm for the understanding of human diseases. 38 1.7 Thesis objective and chapter overview As the main goal of this thesis, we use next generation sequencing to examine epithelioid sarcoma samples looking for novel mutations. Our hypothesis is that additional targetable mutations that play a driving role in epithelioid sarcoma tumorigenesis can be identified. If SMARCB1 loss is the only driver, then focus of targeted therapeutics should be on this protein, and its impact on chromatin remodeling. We use genomic DNA from tumor and matched normal tissues but also tumor RNA. The RNA-sequencing enabled us to evaluate the gene expression patterns of epithelioid sarcoma. The outline of a typical analysis can be seen in Figure 1-3. We use our analyses to focus on targetable pathways and look into epigenetic modifying drugs as potential future therapies. Chapter 2 focuses on findings from the next generation sequencing study, including the mutational load of epitheloid sarcoma as well as several aberrations found in individual cases. We also use SNP6.0 arrays and look into copy number changes. Furthermore, we compare the expression of proximal vs. classic cases, showing distinct clustering and identifying the top differentially expressed genes. Lastly, we compared the expression of three epithelioid sarcoma lines with previously published RNA-sequencing data from 675 cancer cell lines. Chapter 3 follows the observation from SNP6.0 arrays and next generation sequencing that some epithelioid sarcoma cases do not show biallelic inactivation of SMARCB1 even though they still do not express the protein. This phenomenon has been described in the literature previously and several microRNA species have been suggested as a mechanism of inactivation. We take an in-depth look into three epithelioid sarcoma 39 cell lines and identify a cell line that maintains an intact allele of SMARCB1 without promoter methylation. We also examine and negate the suggested microRNA mechanism of inactivation in this in vitro model. Chapter 4 is an in vitro continuation of the findings in Chapter 1. As SMARCB1 mutation is the main driving event in epithelioid sarcoma, a Tet-on inducible SMARCB1 system is established in an epithelioid sarcoma line as well as in a malignant rhabdoid tumor cell line for comparison. EZH2 inhibitors are identified as drugs with similar effects as SMARCB1 induction at multiple loci and are shown to be effective in vitro. Lastly, we establish the maintenance of the SWI/SNF complex despite the loss of SMARCB1 and look into the potential of targeting this complex for therapy. Finally chapter 5 summarizes the findings of this work and attempts to put them in the context of understanding epithelioid sarcoma. It also elaborates on the possible future directions of this work with the most impactful potential. 40 Figure 1-1: Histology of epithelioid sarcoma (A) The image on the left depicts a pseudogranuloma of classic epithelioid sarcoma (EpS) where the epithelial tumor cells (T) are surrounded by reactive lymphocytes(R) with a central necrotic area (N). The image on the right is a phase contrast image of the proximal epithelioid sarcoma cell lines HSES which is spindled under normal culture conditions. However, we noted that when seeded at low density, over extended period of time, colonies with central epithelioid cells (seen to the right) appear. The phenomenon was density independent. It highlights the fluidity of this epithelioid sarcoma line to transition between an epithelioid and spindled morphology. (B) Histology and immunohistochemistry of epithelioid sarcoma. Tumor cells are epithelioid with prominent cytoplasm and large vesicular nuclei with prominent nucleoli. SMARCB1 stains negatively in tumor cells but positively in infiltrating lymphocytes. Tumor cells are also positive for both epithelial markers such as pan-cytokeratin and mesenchymal markers such as vimentin. Vimentin panCK SMARCB1 H&E A B N T R HSES 41 Figure 1-2: SWI/SNF mutations in cancer The SWI/SNF complex uses ATP to open up chromatin and lead to an active transcription site. However it is also capable of silencing genes. There are up to 15 members currently identified to be part of the complex. The two prominent distinct forms of the complex are known as BAF and PBAF (polybromro1 BAF). Here complex members with mutations identified in human cancers have been highlighted. It should be noted that not all these mutations have been confirmed to have a driver role in these tumors. However, all the tumors listed for SMARCB1 have confirmed complete or partial loss of the protein. Aka: also known as, CCC: clear cell carcinoma of ovary, EC: endometrioid carcinoma, CIN: cervical intraepithelial neoplasia, RCC: renal cell carcinoma, SCCOHT: small cell carcinoma of ovary hypercalcemic type, EpS: epithelioid sarcoma, MRT: malignant rhabdoid tumor, AT/RT: atypical teratoid/rhabdoid tumor, CRINET: cribriform neuroepithelial tumor, SBC: sinonasal basaloid carcinoma, RMC: renal medullary carcinoma, EMPNST: epithelioid malignant peripheral nerve sheath tumor, GIC: gastrointestinal carcinoma 42 Figure 1-3: Outline of study design A flow chart depicting the general design of a next generation sequencing study which we applied to epithelioid sarcoma (Chapter 2). Representative bioinformatics programs are in parenthesis and in bold. Further details can be found in the references (Amarasinghe et al., 2014; Ding et al., 2012; Ha et al., 2014; Ha et al., 2012; H. Li & Durbin, 2010; McPherson et al., 2011; Trapnell et al., 2012). For somatic mutations, tumor (T) and matched normal (N) samples, obtained from blood or adjacent normal skeletal muscle tissue, are used in whole genome (WGSS) or exome sequencing to look for somatic mutations and copy number changes (CN). Whole Transcriptome Shotgun Sequencing analysis (WTSS) of tumor samples will enable assessment of expressed mutations and fusions as well as expression patterns. Confirmation of the next-generation sequencing findings using a different platform such as Sanger sequencing to eliminate false positives would be the next step. Finally, if recurrence is observed in the discovery cohort, to understand the frequency and significance, analysis on a larger validation cohort of tumor samples should be completed. In verification and validation, the method of choice for hotspot mutations would be sequencing, for inactivating mutations sequencing or immunohistochemistry (IHC), and for fusions fluorescent in situ hybridization (FISH). 43 Table 1-1: Immunohistochemistry (IHC) findings in epithelioid sarcoma Further description of IHC findings. MRT: malignant rhabdoid tumor, EMT: epithelial mesenchymal transition. Asterisks indicate markers that distinguish epithelioid sarcoma and malignant rhabdoid tumor. Protein Staining Comment Reference Keratin + Epithelial marker (Miettinen, Fanburg-Smith, Virolainen, Shmookler, & Fetsch, 1999) EMA + Epithelial marker (Miettinen et al., 1999) Vimentin + Mesenchymal marker (Miettinen et al., 1999) CD34* + in ~50% Endothelial marker (Miettinen et al., 1999) ERG* + in ~50% Endothelial marker (Kohashi et al., 2015; Stockman et al., 2014) FLI1 + in ~90% Endothelial marker (Stockman et al., 2014) D2-40 + in 70% Lymphatic endothelial marker (Stockman et al., 2014) FVIIIRAg - Endothelial marker (Miettinen et al., 1999) CD31 - Endothelial marker (Stockman et al., 2014) CDH1 - EMT marker (Izumi et al., 2006) Dysadherin* + in ~50% Metastasis/EMT marker (Izumi et al., 2006) SMARCB1 - SWI/SNF protein (Hornick et al., 2009) PBRM1 - SWI/SNF protein (Klijn et al., 2014) Other core SWI/SNFs* + Epithelioid sarcomaSMARCA2 expression is different from MRT (Klijn et al., 2014); our results SALL4* - in classic, + in some proximal Germ cell tumors (Kohashi et al., 2015; Yoshida et al., 2015) CA125 + in ~ 80% Carcinoma marker (Kato et al., 2004; Sakharpe et al., 2011) S100 - Neuroectodermal marker (Chase & Enzinger, 1985) High Ki67 (>30%) + in ~55% Proliferation marker (Izumi et al., 2006) SMA / Desmin + in 15% of proximal Muscle markers (Hasegawa et al., 2001) CD56 + in 60% of proximal Neural, muscle, NK cell marker (Hasegawa et al., 2001) CD99 + in 25% of proximal Ewing’s sarcoma, leukocyte marker (Hasegawa et al., 2001) TP53 + in 80% Tumor suppressor (Sakharpe et al., 2011) 44 Table 1-2: Summary of karyotypic studies of epithelioid sarcoma. The bottom section of the table lists the aberrations reported in different studies and hence ones with stronger evidence. Karyotype Number of cases Reference(PMID) 37῀43,XY 1 15334553 43,XY 3 7685133 / 19866439 / 15334553 44,XY 1 19866439 45,XX 3 2724011 / 19866439 / 8143275 46,X,-X 1 15334553 46,XX 3 2724011 / 15334553 / 15541084 46,XY 6 15334553 / 9216728 / 19866439 / 15803217 / 15334553 47,XY 1 19866439 54῀73<3n>,XXY 1 15334553 61῀64,<3n>,XXYY 1 10526542 61῀67,XXY 1 8908166 62῀63<3n>,XXY 1 15334553 65 1 ATCC 66,XXYY 1 2432306 70-98 <4n>,XX 1 11150604 72 1 15803217 72῀79<4n>,XXYY 1 15803217 73῀78,<4n>XXYY 1 15803217 76῀77<3n>,XXX 1 15334553 78<4n>,XXYY 1 15803217 84,idem x2 1 15334553 90, XXYY 1 23915498 Abberation # of Recurrences Change 22q11 5 structural/deletion 7q31 4 structural/deletion 8q10 4 structural/gain 12q12 3 structural/deletion 12p13 2 structural/deletion 18p11 2 structural/deletion 21p11 2 structural / gain 22q12 2 structural/deletion 7q 2 deletion/gain 45 Chapter 2: Genomic Landscape of Epithelioid Sarcoma 2.1 Introduction Among the tumors with SMARCB1 loss, in depth genomic analysis with next-generation sequencing has been done on renal/extra renal malignant rhabdoid tumors, and atypical teratoid/rhabdoid tumors (Lee et al., 2012) as well as a sphenoid SMARCB1-negative tumor(Jamshidi et al., 2014) which likely represented a sinonasal basaloid carcinoma, one of the most recent additions to the SMARCB1-lost tumors (Agaimy et al., 2014). The results of these studies reveal that in both atypical teratoid/rhabdoid tumor and malignant rhabdoid tumors cases, which are pediatric tumors, as well as in sinonasal basaloid carcinoma, an adult tumor, genomic mutations other than at the SMARCB1 locus are rare. However, co-occurrence of SMARCB1 loss with other oncogenic mutations have been described (Jeong et al., 2014; Kleinschmidt-DeMasters et al., 2011). Additionally in a familial case of CRINETs, which are SMARCB1 negative pediatric brain tumors with much better prognoses than atypical teratoid/rhabdoid tumor, we discovered an NRAS Q61R mutation in a more aggressive chest sarcomatous tumor. This mutation was absent in the less aggressive brain tumor of the same patient (Fig 2-1D). Hence there is evidence that additional mutations compounding SMARCB1 alterations can affect the behavior of some tumors. Given these observations in such related tumors, and also because epithelioid sarcoma had not been explored at a genomic level previously, we set out to characterize the mutational landscape of this tumor. 46 2.2 Results 2.2.1 Description of the discovery cohort A summary of the samples collected and used in this study can be found in Table 2-1. The discovery cohort included four index tumor and normal pairs, which underwent whole genome sequencing. The tumors (but not paired normal tissues) also underwent whole transcriptome sequencing. Six additional samples (three of which were cell lines) underwent whole transcriptome sequencing. All samples were evaluated for copy number changes via SNP6.0 arrays. All tumor samples were confirmed to be SMARCB1 negative. Five additional tumor-normal pairs, initially recruited for the discovery cohort on the basis of an original diagnosis of possible or probable epithelioid sarcoma were recognized to in fact represent other entities, and were eliminated from the study. Most ended being cases of epithelioid hemangioendothelioma. All epithelioid sarcomas in the final data set had their diagnosis confirmed by two pathologists, including a musculoskeletal subspecialty pathologist, and all were confirmed to be SMARCB1 negative by immunohistochemistry and have a tumor cell content of at least 40%. 2.2.2 Copy number changes on tumor samples Eleven epithelioid sarcoma cases underwent SNP6.0 array copy number analysis. The samples were analyzed relative to a normal ploidy control and thus matched normal DNA was not needed for hybridization. The examination of DNA extracted from normal blood as a negative control revealed no copy number abnormalities and was confirmatory. A general look at the copy number profiles reveals an aneuploid background with several gains and losses in most cases (Fig 2-2). In fact, in addition to epithelioid sarcoma, the same methodology was used to analyze ovarian tumors and 47 undifferentiated sarcomas. Several cases of the juvenile granulosa cell tumors as well as undifferentiated sarcomas show a quiescent virtual karyotype unlike epithelioid sarcoma (Fig 2-2). This significant copy number instability immediately distinguishes epithelioid sarcoma from rhabdoid tumors (malignant rhabdoid tumors and atypical teratoid/rhabdoid tumors) and sinonasal basaloid carcinoma which possess quiet copy number profiles (Hasselblatt et al., 2013; Jamshidi et al., 2014; Lee et al., 2012). Our finding is supported by prior karyotypic descriptions of epithelioid sarcoma (Table 1-2). In examining Table 1-2, it is important to note that even though some of the studies have reported normal karyotypes (i.e. 46XY), many of the same papers have also observed aneuploidy (e.g. 63 XXY). Therefore, it is a strong possibility that some of the cells reported to represent epithelioid sarcoma cases with apparently normal karyotypes could actually be representing infiltrating normal cells rather than the malignant cell population, a recognized problem with classical karyotype analyses. For the four cases which had matched normal tissue (T3, T4, T5, T8), i.e. the discovery index cases, whole genome sequencing on tumor/normal pairs was used to evaluate copy number via TITAN. The findings were generally complementary to the SNP6.0 array results for the same samples; however, because of higher resolution of whole genome sequencing, we focused on the whole genome findings for these samples. Evidence for the limitations of SNP6.0 can be found in the case of the cell line VAESBJ which from SNP6.0 had neutral copy number for the locus of SMARCB1. A closer look at this cell line (to be described in detail in Chapter 3) reveals that there is a small deletion in exon 1 followed by a loss of heterozygosity. The estimated resolution of SNP6.0 arrays 48 is at 700bp on average (Bernardini et al., 2010; McCarroll et al., 2008) whereas the region encoding the entire 5’ untranslated region and exon 1 of SMARCB1 sum up to less than 400bp. A list of genes with prominent homozygous/heterozygous deletions or gains of copy number from the WGSS analysis of the discovery cohort is provided in Table 2-2. There were no focal amplifications. Several genes had slight increases in copy number the most significant being PDE4D, an enzyme involved in degradation of cyclic adenosine monophosphate (cAMP). This gain of copy number did not correspond with increased expression in WTSS analysis and hence was not pursued. Areas with deletions were more common than areas with gains of copy number. Among the genes with copy number loss, the most significant in terms of the degree of loss and recurrence were ERC1, BCL2L14, SMARCB1, WNK3 and KDM6A. ERC1 and BCL2L14 are located on 12p13 and SMARCB1 is on 22q11. Both of these regions have been implicated in prior karyotypic analyses (Table 1-2). In order to see if any of these regions had a driving role in epithelioid sarcoma we returned to the expression data from the cell lines. If indeed the homozygous deletions were recurrent and specific driving events, such as the case of SMARCB1, we expected to see a unique and significant loss of expression in the epithelioid sarcoma cell lines. Identifying cell lines with such abnormalities would be important for future experimentation as well. For this, we utilized the fragments per kilobase of exon per million fragments mapped (FPKM) values from the WTSS data and combined it with the publicly available data from the cell lines used in the ENCODE project. All were analyzed together using their raw data. Among the homozygously deleted candidates, SMARCB1 remained the main event of significance. WNK3 had low 49 expression in epithelioid sarcoma but not in a unique manner when compared to other lines (Table 2-4). Similarly KDM6A and ERC1 did not seem to have supportive data in the WTSS expression analysis. On the other hand, BCL2L14 had almost entirely lost expression in all three epithelioid sarcoma lines and this was confirmed by quantitative PCR (qPCR). However, the expression of this gene was comparably low across all lines examined including several other sarcomas, carcinoma and HEK293t cells (Ct values in high 30s). Even though deletion in the region of 22q11 encompassing SMARCB1 (Figure 2-2) was significant, the mode of deletion was not uniform across all cases. Three samples (T5, T6 and Epi544) showed homozygous deletion by SNP6.0 and/or WGSS, one sample had a prominent copy number loss on SNP6.0 but was categorized as heterozygously deleted by APOLLOH in WGSS analysis (T8), four samples had heterozygous deletions by SNP6.0 and/or WGSS (T3, T4, T7 and HSES) and three had neutral copy number by SNP6.0 (T1, T2, and VAESBJ). In summary, out of 11 samples evaluated by SNP6.0, 27% had homozygous deletion at SMARCB1, 46% had some degree of deletion and 27% had maintained normal copy number. Out of the four samples evaluated at the higher resolution of WGSS, three had some degree of loss and one had complete deletion of SMARCB1. This interesting lack of a homozygous loss of SMARCB1 at a copy number level in several cases of epithelioid sarcoma initiated the investigations of chapter 3. Among well-known cancer related genes, the tumor suppressor gene CDKN2A/B (encoding p16 and ARF) was homozygously deleted in one of the tumor specimens, T5, which interestingly also had a homozygous deletion of SMARCB1. Cell lines HSES and VAESBJ also show homozygous deletion of CDKN2A/B. Epi544, which has a neutral 50 copy number for CDKN2A, does not express the encoded protein (Figure 4-1), and thus CDKN2A in Epi544 has possibly undergone epigenetic silencing. This co-deletion of SMARCB1 and CDKN2A/B is mechanistically interesting. SMARCB1 is known to act through CDKN2A: when restored, SMARCB1 causes upregulation of CDKN2A which leads to cell cycle arrest. CDKN2A is necessary and sufficient for the effects of SMARCB1 restoration in the malignant rhabdoid tumor line G401(Oruetxebarria et al., 2004). Hence homozygous loss of SMARCB1, which leads to reduced CDKN2A expression, would be expected to be mutually exclusive from homozygous loss of CDKN2A as these would be two mutations redundantly hitting the same pathway. Additionally, based on Oruetxebarria et al., lack of CDKN2A would negate the anti-proliferative effects of SMARCB1 restoration. Our data, however, suggest that loss of CDKN2A and SMARCB1 in epithelioid sarcoma might confer independent proliferative advantages for the tumor cells and so can be selected to co-occur. Supportive of this is the inhibition of cellular proliferation in the CDKN2A-/- epithelioid sarcoma cell line VAESBJ upon SMARCB1 re-expression (Brenca et al., 2013). Other copy number changes of smaller magnitude and recurrence that involved previously reported oncogenic genes included slight gains in ALK (T3 and T8), EGFR (T7, T8, T4), JUN (T8), and MYC (T4). Evaluation of cell line expression data (Table 2-4) shows that VAESBJ is the only one of three epithelioid sarcoma lines with significant ALK up-regulation, yet significantly higher than any of the ENCODE comparison lines. EGFR is indeed highly expressed in all three epithelioid sarcoma lines, particularly in Epi544, but as it has not undergone a focal amplification, this event is likely not driven at 51 the DNA level. JUN and MYC do not display a significant aberration of expression in epithelioid sarcoma lines. 2.2.3 CDKN2A immunohistochemistry in epithelioid sarcoma In order to examine the frequency of concurrent loss of the CDKN2A and SMARCB1 protein products in epithelioid sarcoma, we examined a TMA of epithelioid sarcoma which has been previously published (Sakharpe et al., 2011). Of the 27 IHC evaluable cases of epithelioid sarcoma, 40% showed loss of CDKN2A (defined less than 10% of cells staining), 50% had some expression and 10% had high expression. The 60% of cases with at least some expression of CDKN2A are the group of interest as SMARCB1 is thought to directly regulate CDKN2A transcription(Kia et al., 2008). Therefore, its loss should reduce CDKN2A protein expression yet in epithelioid sarcoma there is an obvious discordance here. Perhaps in epithelioid sarcoma, the loss of SMARCB1 does not lead to reduced expression of CDKN2A, and mutations in CDKN2A can confer a separate proliferative advantage for the tumor cells (likely the case in T5 as well as the cell lines VAESBJ and HSES). It should be noted that although the dependence of CDKN2A expression on SMARCB1 has been well established in the MON and G401 malignant rhabdoid tuomr cell lines(Kia et al., 2008; Oruetxebarria et al., 2004), IHC examination shows 64% of malignant rhabdoid tumor and 32% of atypical teratoid/rhabdoid tumor positively stain for CDKN2A(Venneti et al., 2011). Thus the dependence of CDKN2A expression on SMARCB1 might be case specific. 52 However, co-deletion of CDKN2A and SMARCB1 was not observed in atypical teratoid/rhabdoid tumor or malignant rhabdoid tumor(Lee et al., 2012). 2.2.4 Somatic point mutations The unusual low rate of mutations in malignant rhabdoid tumor and atypical teratoid/rhabdoid tumor(Lee et al., 2012) was not seen in epithelioid sarcoma. Overall the rate of mutations was comparable to other adult cancers (Figure 2-4). There were no recurrent hot spot mutations or inactivating point mutations. Several samples (T1, T3, T4, T5, T6, T8, and HSES) underwent sequencing at both the DNA and RNA levels. A list of single nucleotide variants (SNVs) and small insertions/deletions (indels) found in common at the DNA and RNA sequencing for the same sample is provided in Appendix B. We did not find evidence for recurrent point mutations. Other than the mutations that were found in common between different platforms, and hence were very likely true positives, we looked at WGSS, Exome, and WTSS single nucleotide variant (SNV) and small insertions/deletion (indel) data separately to include mutations that might have been missed in one of the platforms. For SNVs, we focused on WTSS as point mutations that affect protein function would only be of significance if expressed. For truncations and indels, we focused on exome/WGSS data since the resulting mutations could potentially lead to nonsense-mediated RNA decay and would be missed in WTSS analysis. For WTSS SNVs, the following criteria were applied: alternate count ≥ 4 reads, probability greater than 0.8, lack of SNP predicted at position ± 1 bp, and elimination of low or neutral impact mutation. Out of the remaining SNVs, the following were previously reported as cancer mutations: NF2 (truncated in VAESBJ), BRCA2 R2520Q 53 (seen in T7). Several other genes were previously indicated in cancer, albeit with different mutations than the ones found in our cohort. These include ELN (predicted loss of protein in T5), SUZ12 (P74S in T3), CNOT3 (Q583P), and MSH6 (K519Q). Mutations occurring in more than 1 sample included the following: AHNAK (HSES, T8), CTGLF5 (T7, T8), GOLGA8A (T3, T7), KIF9 (VAESBJ, Epi544), LGALS8 (T4, VAESBJ), NBPF20 (T3, T7, T8), NEK3 (T3, T7, T8), POM121 (T3, T7), PZP (T3, T7), SCAMP1 (Epi544, T4, T5) and SYCP2L (Epi544, VAESBJ, T4). For exome and WGSS indel/truncating mutations, the following criteria were applied: when probability was available (in the case of truncating mutations) a minimum probability of 0.8 was considered, and mutations with no SNPs reported at position ± 1 bp were considered. Among genes previously reported in cancer the following had indel/truncating mutations: ABL1 (T3, T5) and ARID1A (T3, T7). CAMTA1 and ETV1 are mutations previously reported in fusions which were found to have indels in T8 and T1 respectively. ABL1 is an oncogene thus the mutation observed was likely an unreported polymorphism or of small significance. We later explored ARID1A expression via IHC in the validation TMA as discussed in Chapter 4. There were no recurrent hotspot mutations. There were individual cases of an NF2 mutation and a BRCA2 mutation. Possible ARID1A inactivation was explored further but was not found to be a recurrent event (Chapter 4). The pattern of mutations in some of the genes that seemed to recur did not reveal a clear loss of function (insertion/deletion or nonsense) or gain of function (hotspot) characteristic. To ensure that depth of coverage was not an issue, we examined the cell lines VAESBJ and Epi544 with the Ion Torrent AmpliSeq Cancer Panel that led to increased depth of coverage (100X-1000X depending 54 on region; Appendix A). This panel can identify point mutations in 46 cancer genes covering 739 mutations. No point mutations were identified in Epi544 nor VAESBJ. 2.2.5 Fusions in epithelioid sarcoma Without applying filtration criteria, samples had between 21 to 124 non-read through fusions. Of the predicted fusions, there were between 1 to 6 open reading frame fusions per sample. Open reading frame fusions are predicted to be transcribed with the fusion partners maintaining their normal reading frame and thus would be ideal to look for gain of function translocations rather than inactivation of genes. There were no recurrent open reading frame fusions. Circos plots are shown in Figure 2-3B. In comparison to sinonasal basaloid carcinoma, another adult SMARCB1 negative tumor, the epithelioid sarcoma genome seems to have less translocation stability (Figure 2-3). It is interesting that in both sinonasal basaloid carcinoma and epithelioid sarcoma, the 22q11 locus is a focal point for translocations. The susceptibility of this region for translocations might explain consequent deletions of SMARCB1. PCR across several of the predicted fusion breakpoints yielded a product in most cases examined confirming validity of deFuse predictions. Noteworthy were fusions involving SMARCB1 in two epithelioid sarcoma samples: one was an in-frame fusion with the SNF5 homology domain replaced by WASF2 (Fig 2-4C; case T1) and the other a fusion affecting the promoter region of SMARCB1 with an intronic region of PACSIN2 being involved (case T7). Both these samples had neutral copy number for SMARCB1 (Figure 2-2). Interestingly, FISH on T1 using break-apart probes confirmed 3’ end loss of SMARCB1 in one allele, yet also revealed 2 intact alleles in the majority of tumor cells (Fig 2-4C). The inactivation mechanism of the remaining alleles would again motivate the pursuits in Chapter 3. 55 Because of their solitary nature, we think most fusions identified are most likely interruptive. A list of fusions called by deFuse in the transcriptome data that passed the following criteria is provided in Table 2-4 (open-reading frame) and Appendix C (non-open reading frame): altsplice=N, cDNA_breakseqs_percident<0.5, breakseqs estislands percident<0.5, genome breakseqs percident<0.5, break adj entropy min>0.5, probability>0.5, and at least one of the breakpoints was in a coding region and the gene fusion partners were not adjacent. Comparing with the COSMIC Census list of cancer associated genes (http://cancer.sanger.ac.uk/cancergenome/ projects/census/), SMARCB1, KIF5B, PDE4DIP and STAT3 were called in the fusion data. KIF5B and PDE4DIP have been reported to be involved in translocations in non-small cell lung carcinoma (fused to RET or ALK) and in myeloproliferative neoplasms (fused to PDGFRB) respectively. PDE4DIP had interruptive fusions in both T3 and T5 with the last exon fused to adjacent intronic regions. Although not previously reported as cancer related, SRGAP2, MAFIP, DRAM1, HYDIN, TMEM184B and ACTG1 were involved in fusions in more than one case. These are all predicted to be non-functional fusions because of lost reading frame. There were several fusions involving mitochondrial (MT) genes as well. Due to lack of recurrent open reading frame fusions, we do not believe that epithelioid sarcoma has a driving fusion oncoprotein such as seen in other sarcomas like synovial sarcoma or epithelioid hemangioendothelioma. 2.2.6 Expression analysis We used two approaches to look at the expression profiling data. First was an unsupervised cluster analysis within our epithelioid sarcoma samples. This allowed us to look at expression sub-types within epithelioid sarcoma and to see whether proximal vs 56 distal have distinguishable expression programs. Second we used the data from our three cell lines and combined it with that of previously published WTSS data on 675 cell lines(Klijn et al., 2014). The purpose of the second analysis was to see if we could shed light on the cell of origin of epithelioid sarcoma. The unsupervised differential expression analysis showed separate clustering of three of four distal cases from the three proximal cases (Figure 2-5A). This could have two possible implications. Either these two subtypes, much like breast cancer subtypes, have very distinct expression profiles. Alternatively, the normal tissue within the tumor specimen led to the distinct clustering. Given the usually different sites of origin of these subtypes, and the intimate association of epithelioid sarcoma tumor cells with surrounding normal tissue, this is a strong possibility. In fact as seen in Table 2-1, certain samples were predicted to contain about 50% normal tissue. The higher expression of muscle-specific genes such as TTN, NEB, and DES in two of the distal cases is strongly supportive of the influence of normal tissue in the specimen on the overall expression profile. Yet one of the distal samples that clustered separately from proximal cases did not show up-regulation of these same genes; thus, the muscular component within the distal samples was not the only reason for the distinct clustering. There must be additional changes, whether from non-muscular normal tissue or inherent expression differences, that distinguish distal epithelioid sarcoma from proximal. Some of the genes that were differentially expressed between the proximal and distal clusters, excluding muscular components, were LOX, TGFBI, H19, AKAP6, THBS4, HSPB6, KRT6A, and ACHE. 57 For cell line expression comparison, we were first able to regenerate the original clustering by Kiljin et al. and then added our three lines HSES, VAESBJ and Epi544 into the same analysis. The epithelioid sarcoma lines all fell into the mesenchymal rather than the epithelial signature as used by Kiljin et al. Overall, the epithelioid sarcoma cell lines clustered close to one another, particularly Epi544 and HSES, which were always next to each another. VAESBJ clustered more distantly (Figure 2-6B). This is not particularly surprising as VAESBJ was originally derived from the bone marrow metastasis of a skin epithelioid sarcoma of a 41 year old male(Helson et al., 1995) and therefore was not established from primary tumor cases as was the source for HSES(Sonobe, Ohtsuki, Sugimoto, & Shimizu, 1997) and Epi544(Sakharpe et al., 2011). VAESBJ also behaves quite aggressively in tissue culture with prominence of very large epithelial cells that divide more rapidly than Epi544 or HSES. Interestingly, the cell lines with which VAESBJ clustered most closely were predominantly kidney or ovarian tumors. Epi544 and HSES on the other hand were closest to cell lines from a variety of other tissue types including kidney, pancreas, ovary, colorectal, and lung. Four of the six cell lines with which Epi544 and HSES most closely clustered have reported abnormalities in the SWI/SNF complex, three of them having lost expression of key SWI/SNF subunits (G402, MIA Paca-2, TYK-nu). Reports of SWI/SNF abnormalities within the cell lines related to VAESBJ were not as prominent although one is known to have lost SMARCC1 (SK-OV-3). Given the clustering of Epi544 and HSES with a variety of lines with known SWI/SNF deletions, one might presume that loss of SWI/SNF function has a significant impact on global expression. Such a hypothesis might not be too far fetched given the direct association of SMARCB1 loss with vimentin and cytokertain expression (Tsikitis et 58 al., 2005), which define SMARCB1 negative tumors as intermediate between epithelial and mesenchymal lineages. However, there are several well-established SMARCB1 and SWI/SNF abnormal lines that cluster distantly from Epi544 and HSES. Most prominent is the lymphoid line EJM that has lost SMARCB1 (homozygous deletion; http://www.cbioportal.org) yet it still remains within the lymphoid tissue cluster with a very distinct expression profile than that of any of the mesenchymal lines. Hence, although abnormalities in the SWI/SNF complex have some impact on global expression, their effect is much smaller than lineage defining profiles. 2.3 Discussion A surprising finding was the genomic complexity of epithelioid sarcoma. This tumor shows somatic point mutation rates comparable to many carcinomas, with a considerably higher mutational load compared to the SMARCB1 negative rhabdoid tumors (Figure 2-4). Epithelioid sarcoma is also not stable structurally with up to 124 rearrangements found in the transcriptome. This again is different from SMARCB1 negative sinonasal basaloid carcinoma (Figure 2-3). The genomic instability could perhaps be explained by loss of function of SWI/SNF. SWI/SNF is known to be involved in double stranded DNA repair(Park et al., 2006) as well as proper function of topoisomerase 2-alpha (TOP2A) in sister chromatid separation during anaphase (Dykhuizen et al., 2013); therefore, its malfunction could result in genomic instability. Perhaps because of the rapidity and early onset of rhabdoid tumors, there is no time for the accumulation of additional mutations, whereas in epithelioid sarcoma, a longer period of development allows the unstable phenotype to materialize. Similar to epithelioid sarcoma and unlike rhabdoid tumors, sinonasal basaloid carcinoma is a tumor of adults 59 yet genomic analysis of a single case showed a quiet genome(Jamshidi et al., 2014). Sinonasal basaloid carcinoma has not been studied at a genomic level in depth and study of additional cases would be needed to allow informative comparisons. Despite the relatively large number of structural and point mutations, few mutations were recurrent in epithelioid sarcoma. In fact, SMARCB1 abnormalities have the highest rate of recurrence. No well-known cancer mutations, such as those involving TP53, KRAS, etc. were observed; we used the COSMIC Census database as a comprehensive list of such genes to look for cancer-associated mutations. CDKN2A homozygous deletion was observed in a single tumor cases and two cell lines. Because many cell lines tend to acquire this particular mutation as a result of the augmented and rather non-representative growth conditions of tissue culture, loss of CDKN2A in cell lines may not define a mutational feature of the tumor under study. Using immunohistochemistry we found that 60% of epithelioid sarcoma cases had maintained CDKN2A expression. Whether the remaining 40% had lost expression because of genomic mutations or transcriptional down-regulation is not known. However, SMARCB1 regulation of CDKN2A seems to be decoupled. In the case of T5 as well as VAESBJ and HSES, loss of CDKN2A through mutations presumably gives an additional proliferative advantage to the epithelioid sarcoma cells. We found homozygous deletions and interruptive fusions to be the mechanisms of inactivation of SMARCB1. We did not observe any point mutations and in fact Sanger sequencing in validation samples also did not lead to the discovery of any point mutation. Interestingly, however, we did not find evidence of inactivation of all SMARCB1 alleles 60 in several epithelioid sarcoma cases. Perhaps a second epigenetic hit could serve to inactivate the remaining alleles in these cases, a topic we pursue in Chapter 3. Through transcriptional sequencing, we were able to obtain one of the first global expression profiles of epithelioid sarcoma. Proximal vs. classic subtypes showed distinct clustering in an unsupervised analysis; however, we believe that part of this separation is influenced by the infiltrating normal tissues that may be distinct in the proximal samples vs classic cases. Most importantly, however, we were also able to compare expression profiles of the clonal tumor cell line populations with that of a large collection of publicly available cell line WTSS data. This revealed several interesting features. First, that the epithelioid sarcoma lines cluster with the mesenchymal cell lines. Second, that HSES (proximal) and Epi544 (classic) had the closest association and were always grouped with each other in unsupervised analyses. Thus epithelioid sarcoma has a signature pattern that distinguishes it from other tumors. VAESBJ in contrast clustered at some distance, although overall it was still close to HSES and Epi544. VAESBJ may not be a representative case of epithelioid sarcoma as it was derived from a spinal metastasis and not from a primary tumor specimen, and interestingly it has a strong association with kidney and ovarian tumor cell lines. Overall our analysis led to a wealth of novel information and a broad genomic observation of the epithelioid sarcoma landscape. We use some of these observations in the following chapters while we also hope that our data could be built upon in the future to further characterize epithelioid sarcoma. We plan to make this data publicly available through the database of Genotypes and Phenotypes (dbGaP) of the National Center for Biotechnology Information. 61 2.4 Methods 2.4.1 Sample acquisition and nucleic acid extractions Use of human samples was approved by the institutional ethics review board under Clinical Research Ethics Board #H08-01411. All inter-institutional material transfer was done with appropriate transfer agreements. The discovery cohort samples were collected from the Vancouver Sarcoma Bank, Vancouver, BC as well as the Mount Sinai Hospital, Toronto, ON. Dr. Brendan Dickson, Mt. Sinai Hospital Toronto, did the pathology reviews, including subtype classification for cases from the Mt. Sinai bank. Dr. Torsten Nielsen performed the equivalent review for the Vancouver cases. The tissue microarray used for immunohistochemistry studies was published on previously(Sakharpe et al., 2011) and obtained from the MD Anderson Cancer Centre with the kind support of Dr. Alexander Lazar. Six tissue samples were also obtained from MD Anderson and were used for the Sanger sequencing of the SMARCB1 locus. Cryosectioning for all the collected cases was carried out and two unstained slides were made from each case collected. One was used for H&E based tumor content estimation and one for SMARCB1 immunohistochemistry to confirm the diagnosis. To extract nucleic acids, frozen samples were crushed in liquid nitrogen using mortar and pestle. Thereafter the resulting powder was immediately processed for DNA or RNA extraction with the DNeasy Blood and Tissue kit and miRNeasy Mini kit (Qiagen, Hilden, Germany). Genomic DNA integrity was checked on a 1% agarose gel to ensure no degradation, and RNA quality was assessed by 2100 Bioanalyzer (Agilent, Santa Clara, CA) via evaluation of the RNA Integrity Number (RIN). RNA was quantified by the 2100 bioanalyzer. DNA was quantified using the Qubit Quantification device (Life Technologies). 62 2.4.2 Deep sequencing Samples were processed at Canada’s Michael Smith Genome Sciences Centre for deep sequencing. Funding was obtained from the Terry Fox Research Institute, Frontiers in Cancer research program. Paired-end reads from all libraries were generated on an Illumina HiSeq2000 sequencer. Focused deep amplicon sequencing was performed using the AmpiSeq Cancer panel (v2.0) on the Ion Torrent PGM platform(Bashashati et al., 2013). A combination of established analytical pipeline and custom bioinformatics tools was used for identifying somatic nucleotide mutations, translocations and copy number aberrations. 2.4.3 Bioinformatics analysis Reads were aligned to the human reference genome (GRCh37, available from (Chase & Enzinger, 1985)) using BWA (0.5.7) (H. Li & Durbin, 2010). Reads from multiple lanes were merged and duplicate marked using Picard (v1.38) (Hasegawa et al., 2001). Copy number analysis was done using a Hidden-Markov model based approach via HMMCopy and APOLLOH v0.1.1 (Ha et al., 2012). Loss of Heterozygosity (LOH) was evaluated by APOLLOH and TITAN(Ha et al., 2014). SNVs and indels were identified using a probabilistic joint calling approach via SAMtools v0.1.17 (H. Li et al., 2009), MutationSeq v1.0.2 (Ding et al., 2012), and Strelka v0.4.6.2 (Saunders et al., 2012). Variants were annotated to genes using the Ensembl database (v69) (Flicek et al., 2013). RNA-Seq reads were analysed with Jaguar (Y. Butterfield, unpublished) to include alignments to a database of exon junction sequences and subsequent repositioning onto the genomic reference. For expression, RNA-Seq data was processed by Cufflinks (Trapnell et al., 2010) and the level of expression for each gene 63 normalized as fragments per kilobase per million observations (FPKM). Expressed variants were called with SNVMix2 (v0.12.1-rc1) (Goya et al., 2010; Shah et al., 2009) and SAMtools (v0.1.13) (H. Li et al., 2009). Differential expression analysis was done using outlier statistics and fold change comparison within the epithelioid sarcoma cohort. Raw data for the ENCODE lines was obtained from (http://genome.ucsc.edu/cgi-bin/). Raw data and clustering analysis for the cell lines was carried out as described by Kljin et al.(Klijn et al., 2014). Mutational rate analysis and collection of comparison data in figure 2-4 were done using MuTect and as reported previously (Lawrence et al., 2013). The data from all tumors other than epithelioid sarcoma is from The Cancer Genome Atlas (TCGA) (Lawrence et al., 2013). For data with genome sequencing, mutations rate in the coding region was calculated to compare to samples with exome sequencing. 2.4.4 Fluorescent in situ hybridization (FISH) FISH was performed on 4-μm thick full sections from FFPE blocks. The slides were baked overnight at 70˚C, deparaffinized in xylene, dehydrated with ethanol washes, incubated for an hour in 10mM Citric Acid at 80˚C, and treated with pepsin for 20 minutes at 37˚C. After washing and dehydration, nick-translated fluorescent probes were applied, followed by hybridization at 37˚C for 16hrs. The probes used were fluorescence-labelled RP11-71G19 to detect SMARCB1 and RP11-262A13 to mark chromosome 22. For breakapart studies we used RP11-1112A23 and CTD-2376E20 on centromeric end and RP11-80O7 and RP11-76E8 on the telomeric end of SMARCB1. DAPI II was applied to mark the nuclei before viewing under fluorescent microscopy. 64 2.4.5 PCR and Sanger sequencing SMARCB1 primers were the same as described previously (Boyd et al., 2008). When needed, PCR primers were designed using Primer3 (http://bioinfo.ut.ee/primer3-0.4.0/). We used Platinum Taq DNA polymerase High Fidelity (Life Technologies). Reactions were denatured at 94°C for one minute, and went through 35 to 45 cycles of PCR (94°C 30 sec, 64°C 30 sec, 72°C 30 sec) and five-minutes extension at 72°C. Purification of PCR products was done with ExoSAP-IT (Affymetrix, Santa Clara, CA). Sequencing was done with the ABI BigDye terminator v3.1 kit (Applied Biosystems, Waltham, MA). Products were sequenced on an ABI Prism 3130xl Genetic Analyzer (Applied Biosystems). 2.4.6 Immunohistochemistry Immunohistochemical analyses on standard 4-µm paraffin block tissue sections from FFPE blocks were performed on a semi-automated Ventana Discovery XT (Ventana Medical Systems, Tucson, AZ, USA). Standard CC1 antigen retrieval with a 2 hour primary incubation was done. For SMARCB1, the primary mouse BAF47 antibody from Becton Dickinson and Company (BD) Biosciences (New Jersey, USA) was used in a 1:50 dilution followed by a 16 minute incubation with pre-diluted HRP-conjugated anti-mouse secondary antibody. Histological images were obtained using a ScanScopeXT digital scanning system (Aperio Technologies, Vista, CA, USA). 65 Figure 2-1: Co-occurrence of NRAS hotspot and SMARCB1 loss in CRINET (A) The family tree of patients with CRINET which is a SMARCB1 negative tumor. Unlike atypical teratoid/rhabdoid tumor, most patients survive this brain tumor(Hasselblatt et al., 2009). This family had a germ-line duplication of SMARCB1 exon 6 (Ibrahim et al., 2011) which led to multiple members developing CRINET (black fill) as well as schwannomas. The only deceased member (crossed line), also developed an aggressive chest mass which was thought to be a sarcoma and was lethal. This more aggressive tumor was also SMARCB1 negative (B). Targeted deep sequencing of candidate cancer associated genes in this chest mass, revealed a hot spot mutation of NRAS at codon 61 (C). This mutation was confirmed by Sanger sequencing and found only in the chest mass and not in the (archival) brain tumor of the same patient. G401 is a malignant rhabdoid tumor line that was used for comparison (D). This shows that SMARCB1 mutations, unlike what has been observed in rhabdoid tumors, can co-occur with other oncogenic mutations which can possibly alter the behavior of a tumor. We hence set out to examine any abnormal pathway in epithelioid sarcoma in addition to SMARCB1. 66 Figure 2-2: Virtual karyotype of epithelioid sarcoma Copy number data for the epithelioid sarcoma samples. Blue and red signify loss and gain of copy respectively with the darker shades indicating greater loss. The upper panel shows genome wide copy number results for the epithelioid sarcoma samples (EpS) as well as for a series of ovarian tumors (OT), matched normal blood, and undifferentiated sarcomas (US) separated for comparison. The bottom panel is a zoomed-in image of the area of chromosome 22 covering SMARCB1 showing a unique prevalence of deletions in epithelioid sarcomas. Overall however, epithelioid sarcoma shows aneuploidy with multiple areas of copy number loss and gain. This is not the case for the normal blood, samples, several of the juvenile granulosa cell tumors (JGCT), undifferentiated sarcomas, or rhabdoid tumors (Lee et al., 2012). Samples T3, T4, T5, and T8 have the TITAN analysis on display whereas SNP6.0 analyses are used for all the remaining samples. The replacement of WGSS data for the four aforementioned samples was to increase resolution when WGSS was available. d= classic/distal epithelioid sarcoma, p= proximal epithelioid sarcoma. Red arrows indicate cases where fusions involving SMARCB1 were identified via WTSS. 67 Figure 2-3: Structural landscape of epithelioid sarcoma Epithelioid sarcoma shows evidence of significant structural aberrations. (A) is a circos plot of a case of the SMARCB1 negative tumor sinonasal basaloid carcinoma with few translocations(Jamshidi et al., 2014). Genomic translocations are shown with purple lines and transcriptomic fusions with orange. Epithelioid sarcoma on the other hand has many structural changes. Five samples including a cell line are shown here where the transcriptomic fusions are indicated by green lines (B). Dark green lines show open reading frame and light green lines indicate non-open reading frame fusions respectively. Sinonasal basaloid carcinoma, T1 and T5 show fusions to chromosome 22 at the region of SMARCB1 which could indicate the susceptibility of this region for translocations and subsequent copy number loss. T8 in particular has an abundance of fusions. This was the only case in the discovery cohort that had undergone chemotherapy before collection of the specimen. Some translocations involved SMARCB1 itself. In T1, SMARCB1 was fused to WASF2 with the SNF5 homology domain interrupted (C). IHC was negative as the antibody would pick up the lost section of the fused protein if it were produced. Breakapart FISH (red probe on the centromeric end corresponding to N terminus) on this case shows tumor cells that have lost the 3’ end (C terminus) of SMARCB1 in two alleles yet have maintained two fusion intact alleles. The FISH image shows a tumor cell to the left with a large nucleus and a normal cell to the right for comparison. This was a representative image. 68 Figure 2-4: Comparison of point (SNV) mutation rates of epithelioid sarcoma vs. other tumors. Epithelioid sarcoma (EpS) has a higher mutation rate than rhabdoid tumors. Numbers of cases are in parentheses. The epithelioid sarcoma case with lowest mutation rate is T3 which had the lowest tumor content estimate among epithelioid sarcoma samples. 69 Figure 2-5: Summary of mutations Summary of mutations found in epithelioid sarcoma. The most consistent event with definite effect on the indicated gene is SMARCB1 with frequent homozygous deletions and translocations. . VAESBJ has a homozygous deletion of exon1 only (asterisk) whereas all other indicated SMARCB1 deletions involve the entire gene. Heterozygous loss in chromosome 12p is the other event of recurrence and two select genes in this region shown. CDKN2A is another gene with evidence of homozygous deletion 70 Figure 2-6: Epithelioid sarcoma expression profiles. (A) Unsupervised clustering of the expression data between the proximal and classic/distal subtypes. (B) Clustering of the epithelioid sarcoma cell line data with that of 675 publicly available cell line data. To the left is an image of the overall clustering of all lines. To the right, there is a zoomed-in image with a focus on the specific epithelioid sarcoma lines (ESPR). It shows greater similarity between HSES and Epi544. VAESBJ clusters most closely with a series of kidney and ovarian tumor lines, whereas Epi544 and HSES cluster with a greater diversity of lineages. Four of six lines to the right of Epi544 and HSES, which are the closest cluster to these epithelioid sarcoma lines, have aberrations in SWI/SNF members. However, cell lines with SWI/SNF mutations are distributed throughout different lineages and are not limited to this cluster. 71 Table 2-1: Sample list Frozen tumor samples are numbered T1 through T8; five had matched normal germline tissue available. T1 had a proximal location (pubis) but was of ‘distal’ subtype on histology. In terms of the cell lines used, Epi544 was derived from a tumor of the foot; original histology showed admixed spindled cells thus it is considered to be representative of “classic” subtype. HSES was derived from a perineal tumor with prominent rhabdoid cells and considered proximal in subtype. VAESBJ was derived from a marrow metastasis of a skin-based epithelioid sarcoma; the original histology was not available but the cell line is characterized by very large and rapidly dividing epithelial cells. WGSS: whole genome sequencing data, WTSS: whole transcriptome sequencing data. Het: heterozygous. N/A: not available Type ID Matched Subtype Data NGS Coverage Tumour Content SMARCB1 Protein Status SMARCB1 Gene Status Frozen T1 N Distal Exome, WTSS 40%(Histology) - Fusion, No loss (SNP6.0) Frozen T2 N Distal Exome N/A - No loss (SNP6.0) Frozen T3 Y, Blood Distal WGSS(-N), WTSS X60(T), X30(N) 40%(APOLLOH), 40%(Histology) - Heterozygous deletion (WGSS), No loss (SNP6.0) Frozen T4 Y, Blood Distal WGSS(-N), WTSS X30(T), X30(N) 42%(APOLLOH), 50%(Histology) - Heterozygous deletion (WGSS), Het loss of ex1-3 (SNP6.0) Frozen T5 Y, Adj. Tissue Distal WGSS(-N), WTSS X60(T), X30(N) 70%(APOLLOH), 80%(Histology) - Homozygous deletion (WGSS), Heterozygous (SNP6.0) Cell Line Epi544 N Distal WTSS 100% - Homozygous deletion (SNP6.0) Frozen T6 N Proximal Exome, WTSS N/A - Homozygous deletion (SNP6.0) Frozen T7 Y, Adj. Tissue Proximal WTSS 80%(Histology) - No loss (WGSS), No loss (SNP6.0) Frozen T8 Y, Adj. Tissue Proximal WGSS(-N), WTSS X30(T), X30(N) 40%(APOLLOH), 66%(Histology) - Heterozygous deletion (WGSS), Heterozygous (SNP6.0) Cell Line HSES N Proximal WTSS, Exome 100% - Hemizygous, intact copy Cell Line VAESBJ N Proximal WTSS 100% - LOH, exon 1 deletion 72 Table 2-2: Genes with evidence of deletion in discovery cases. Gene T3 T4 T5 T8 Band DCAF12 Het Loss Hom Loss Neutral Het Loss 9p13.3 ERC1 Het Loss Het Loss Hom Loss Het Loss 12p13.33 CLEC4C Het Loss Hom Loss Het Loss Het Loss 12p13.31 NANOG Het Loss Het Loss Het Loss Hom Loss 12p13.31 SLC2A14 Het Loss Hom Loss Het Loss Het Loss 12p13.31 BCL2L14 Het Loss Het Loss Het Loss Hom Loss 12p13.2 C12orf70 Het Loss Hom Loss Het Loss Het Loss 12p11.23 IGLL1 Het Loss Het Loss Hom Loss Het Loss 22q11.23 C22orf43 Het Loss Het Loss Hom Loss Het Loss 22q11.23 RGL4 Het Loss Het Loss Hom Loss Het Loss 22q11.23 ZNF70 Het Loss Het Loss Hom Loss Het Loss 22q11.23 VPREB3 Het Loss Het Loss Hom Loss Het Loss 22q11.23 C22orf15 Het Loss Het Loss Hom Loss Het Loss 22q11.23 CHCHD10 Het Loss Het Loss Hom Loss Het Loss 22q11.23 MMP11 Het Loss Het Loss Hom Loss Het Loss 22q11.23 SMARCB1 Het Loss Het Loss Hom Loss Het Loss 22q11.23 DERL3 Het Loss Het Loss Hom Loss Het Loss 22q11.23 SLC2A11 Het Loss Het Loss Hom Loss Het Loss 22q11.23 GSTT2B Het Loss Het Loss Hom Loss Het Loss 22q11.23 DDTL Het Loss Het Loss Hom Loss Het Loss 22q11.23 DDT Het Loss Het Loss Hom Loss Het Loss 22q11.23 GSTT2 Het Loss Het Loss Hom Loss Het Loss 22q11.23 GSTT1 Het Loss Het Loss Hom Loss Het Loss 22q11.23 CABIN1 Het Loss Het Loss Hom Loss Het Loss 22q11.23 SUSD2 Het Loss Het Loss Hom Loss Het Loss 22q11.23 GGT5 Het Loss Het Loss Hom Loss Het Loss 22q11.23 SPECC1L Neutral Het Loss Hom Loss Het Loss 22q11.23 MTMR3 +1 gain Het Loss Het Loss Hom Loss 22q12.2 TTLL1 Neutral Hom Loss Het Loss Het Loss 22q13.2 PARVB +1 gain Het Loss Het Loss Hom Loss 22q13.31 CRLF2 Het Loss Hom Loss Neutral Het Loss Xp22.33 CSF2RA Het Loss Het Loss Neutral Hom Loss Xp22.33 GPKOW Het Loss Het Loss Neutral Hom Loss Xp11.23 FAM120C Het Loss Het Loss Neutral Hom Loss Xp11.22 WNK3 Het Loss Het Loss Neutral Hom Loss Xp11.22 USP4 Het Loss Hom Loss Neutral Hom Loss 3p21.31 NECAP1 Het Loss Hom Loss Het Loss Hom Loss 12p13.31 PLA2G6 Neutral Hom Loss Het Loss Hom Loss 22q13.1 73 Table 2-3: Fusions predicted via deFuse. These fusions had an open reading frame and met the filtering criteria outlined in the text. Gene 1 region Gene 2 region Gene 1 Gene 2 Expression gene 1 Expression gene 2 Probability Spanning reads Splitting reads ORF Sample coding coding SMARCB1 WASF2 342 2939 0.92 12 4 Y T1 coding coding APLP2 IGSF9B 241359 465 0.71 16 7 Y HSES coding coding ZNF74 GGA1 504 1959 0.95 28 24 Y HSES coding coding MT-ATP6 MT-ND1 339932 490689 0.62 46 40 Y HSES coding coding RFC5 APPL2 169 464 0.92 6 3 Y T6 utr5p coding PTPRS KIF5B 11158 4161 0.98 42 36 Y T7 coding coding CAPN7 SLC24A4 1017 252 0.89 5 5 Y T8 coding coding MT-ATP6 MT-ND1 69078 70470 0.69 17 19 Y VAESBJ coding coding DCUN1D5 TRPC6 1119 1366 0.98 46 42 Y Epi544 coding coding MT-ATP6 MT-ND1 48001 52951 0.72 5 6 Y T4 74 Table 2-4: FPKM values of epithelioid sarcoma lines compared to ENCODE lines. Mesenchymal Epithelial Lymphoid Myoblastic "Normal" Cancer Embryonic stem cell Epithelioid sarcoma Cell line name Epi544 HSES VAES-BJ Huvec Nhlf Hsmm Lhcn-m2 Mcf7 Hela Hepg2 Hct11-6 H1hes-c Nhek Gm128-78 VIM (mesenchymal marker) 1351.3 941.2 786 1354.2 1465.4 848.5 1070.3 1.7 736.2 0.2 0.8 231.3 40.2 96 Pan-KRT Average (epithlial marker) 4.5 6.4 4.8 2.2 0.8 0.9 1 12.8 7.4 2.2 9.9 2.5 67.4 0.4 CDH1 (EMT marker) 0 0 0.1 0.1 0.1 0.3 0.3 153.3 0.2 30.4 88.9 104.5 176.2 4.5 MYOD1 (muscular marker) 0 0 0 0 0 87.2 76.1 0 0.1 0 0.1 0.1 0.1 0.1 CD34 (Vascular marker) 364 9.8 0.2 14.6 0.5 3.3 2.1 0 0.1 0 0 0.3 0 0 CD48 (B-lymphoblast marker) 0 0 0 0 0 0 0 0 0 0.1 0 0.1 0 460.7 SMARCB1 0 1.5 0.2 45.6 50.5 31.6 72.7 55.7 23.3 53 61.7 69.7 35.1 91.4 GLI1 (inhibited by SMARCB1) 7.7 22.7 15.2 0.1 2.4 0.4 3.7 0.1 0.7 0.1 1.9 14.2 0.1 0.2 ERC1 16.5 5.9 17.7 12.2 13.1 14 9.4 8.2 7.4 4.7 9.3 9.7 8.3 1.7 BCL2L14 0 0 0 0.1 0 0 0 0.2 0.1 0.2 0.2 0.7 0.1 1 WNK3 0 0 1.7 0.4 0 0.2 0 0 0.1 0.1 0 3.9 0 0.1 KDM6A 7.5 5.5 6.9 5.2 7 4.6 3.7 9.3 9.5 9.6 18.1 28.5 14.1 14.4 75 Chapter 3: SMARCB1 Inactivation in Epithelioid Sarcoma 3.1 Introduction Genomic analysis of epithelioid sarcoma revealed that not every case had inactivation of all SMARCB1 alleles, despite the loss of protein expression. We did not find any point mutations in SMARCB1; fusions and loss of copy number were the main events affecting the SMARCB1 gene. A review of the literature showed that others have made similar observations. Kohashi et al. identified inactivating mutations in only 4 of 39 SMARCB1 negative epithelioid sarcoma cases(Kohashi et al., 2009). Flucke and colleagues also only found a point mutation in one of thirteen cases of epithelioid sarcoma, although they did not evaluate the cases for copy number alterations(Flucke, Slootweg, Mentzel, Pauwels, & Hulsebos, 2009). These authors suggested this lack of biallelic SMARCB1 inactivation in epithelioid sarcoma distinguishes it from malignant rhabdoid tumors. Interestingly, individual cases of renal medullary carcinomas(Calderaro et al., 2012) and atypical teratoid/rhabdoid tumor(Tsai et al., 2012) without a DNA-level explanation for protein loss have also been reported. Given the possibility of epigenetic silencing in epithelioid sarcoma, several groups compared the microRNA profiles of epithelioid sarcoma vs. malignant rhabdoid tumors and suggested possible microRNAs that could silence SMARCB1 in epithelioid sarcoma (Kohashi, Yamamoto, et al., 2014; Papp, Krausz, Stricker, Szendroi, & Sapi, 2014). However, these studies were not supportive of each other, each predicting a different set of microRNAs to be the culprit. Papp et al., had a rather indirect approach in proving their model because of a lack of an epithelioid sarcoma in vitro model. They used fibrosarcoma, colon adenocarcinoma and dermal fibroblast lines where they showed that one of the their predicted microRNA 76 species, miR206, could reduce SMARCB1 expression when over-expressed (Papp et al., 2014). We therefore set out to validate the previously reported findings and also to look for a potential in vitro model that could be used for future studies. 3.2 Results 3.2.1 Sanger sequencing and FISH on tumors Using a TMA that had been previously published on (Sakharpe et al., 2011), we performed FISH with probes for SMARCB1 and for the telomere of chromosome 22 to evaluate the frequency of homozygous deletion of SMARCB1 in epithelioid sarcoma. Among thirty examined cases of epithelioid sarcoma, four showed strong evidence for the maintenance of intact alleles. A large portion of cases showed a mix of tumor cells with homozygous and hemizygous deletion of SMARCB1 revealing some degree of intra-tumoral heterogeneity. We used stringent criteria to consider cases that had intact alleles. Firstly, greater than 40% of individually counted tumor cells for each case had to have one or more SMARCB1 alleles present. Secondly, the same cases, had to have ratio of total SMARCB1 probes to total telomere 22 probes that exceeded 0.4. And also since each case had two cores, results from both cores had to agree (Table 3-1). In a separate experiment, we sequenced 5 FFPE and 5 frozen samples of epithelioid sarcoma and sequenced SMARCB1 along the entire length of the coding sequence similar to the approach of Flucke et al. We did not find evidence for point mutations. 3.2.2 Cell line model: intact SMARCB1 allele in HSES Given that a certain portion of epithelioid sarcoma cases, whether it is 13% as per our observation or 90% as observed by Kohashi et al., seemed to lack a genetic explanation for complete loss of protein, we used the three cell lines that we have to 77 evaluate this possibility in more detail. From our SNP6.0 results as discussed in Chapter 2, Figure 2-2, we knew that Epi544 clearly had a homozygous deletion of SMARCB1. From the same results however, HSES seemed to have a heterozygous deletion and VAESBJ a neutral copy number. Nevertheless, as discussed earlier, SNP6.0 has a limited resolution. Hence to evaluate the possible presence of smaller deletions we used multiplex ligation dependent probe amplification (MLPA) of SMARCB1 in the epithelioid sarcoma lines with a rhabdoid tumor line, G401 used as negative control, and HEK293t cells and Control DNA (Sigma Male DNA) as positive controls. The results revealed that VAESBJ, as also found by others(Brenca et al., 2013), has a small homozygous deletion of exon 1 and the promoter region; Epi544 has homozygous deletion as predicted by SNP6.0 results; and HSES has an intact allele in a heterozygously deleted form (Figure 3-1A). This was also confirmed via simple PCR across different exons of SMARCB1 (Figure 3-1C). The next obvious possibility would be point mutations that could lead to loss of protein. However, we examined all exons and the 5’ and 3’ untranslated regions of HSES via Sanger sequencing and did not find any evidence for mutations. Yet all these methods could have missed a balanced translocation which could still lead to probe and PCR amplification in HSES. Hence we examined the cell lines with SMARCB1 probes described in chapter 2 which confirmed our findings. The breakapart FISH showed that indeed HSES did not have a translocation at the SMARCB1 locus (Figure 3-1 B) which is also supported by the lack of predicted fusions at this locus from the WTSS data. Thus HSES seems to be an ideal cell line model to address the reason for an observed loss of protein that is occurring despite the presence of an intact SMARCB1 allele. 78 To look at whether the loss of protein happens at a pre- or post-translational stage we performed qPCR on HSES and other cell lines for comparison (Figure 3-2). Transcript of SMARCB1 was present in HSES but at extremely low levels (about 2% of HEK293t cells for relative comparison). Despite the evidence that there is some transcription in HSES, due to the very low levels of mRNA we expect that the silencing occurs at the transcriptional stage and/or via greatly decreased mRNA stability. We used the Taqman qPCR (Life Technologies) which with an internal gene specific probe, has high specificity. Interestingly, VAESBJ, despite the loss of the promoter and exon1 regions, still revealed some transcripts -- though at a very low level, about 4% of the amount seen in HSES (Figure 3-2C). On the other hand, G401 which has complete deletion of the SMARCB1 locus has no detectable transcripts in this assay. 3.2.3 Methylation status A well-recognized epigenetic mechanism for silencing a gene is promoter CpG island methylation. We designed and optimized methylation specific PCR primers for the promoter region of SMARCB1 using completely methylated Jurkat DNA (NE Biolabs, Ipswich, MA) and HEK 293t extracted DNA as positive and negative controls respectively. We identified 3 pairs of primers covering a 400bp area in the CpG island upstream and within the 5’ untranslated region of SMARCB1. After bisulfite reaction, examination of HSES as well as cases in our discovery cohort (T1 and T7) that had evidence of allele maintenance showed no methylation (Figure 3-3 A). Yet methylation at distant sites that had some effect on SMARCB1 transcription remained a possibility. We therefore treated the HSES line with 5-aza-2′deoxycytidine (5-Aza-DC) for extended periods of time. We ensured that our treatment was effective by observing a down- 79 regulation in DNA methyltransferase 1 (DNMT1) (Figure 3-3 B). This long-term inhibition of DNA methylation did not lead to re-expression of the SMARCB1 protein; therefore, we did not find evidence for methylation as a silencing mechanism. To rule out protein degradation as a possible mechanism, we treated HSES with the proteasome inhibitor MG132 which did not restore the protein. We also examined histone deacetylation as a possible transcriptional inhibition mechanism; however, treatment with the histone deacetylase inhibitor romidepsin did not lead to re-expression of SMARCB1. Lastly, we considered covalent histone modification by the polycomb complex, i.e. histone three lysine 27 trimethylation, as a possible epigenetic silencing mark. Treatment with the EZH2 inhibitor GSK126 over 6 days at concentrations up to 10μM did not lead to SMARCB1 detection by Western blots hence we did not find support for this possibility either. 3.2.4 MicroRNA silencing of SMARCB1 Another possibility, suggested by two recent papers(Kohashi, Yamamoto, et al., 2014; Papp et al., 2014), was microRNA silencing of SMARCB1. One of the strongest microRNA predictions as a SMARCB1 regulator, both based on the seed sequence and from in vitro experiments(Dey, Gagan, & Dutta, 2011), is miR-206. Using qPCR and in comparison to human embryonic kidney cells HEK293t, the malignant rhabdoid tumor line G401 and the synovial sarcoma line SYO1, we found evidence for downregulation of one out of seven transcript targets of mir206. This was MEOX2, a homeobox gene thought to be involved in myogenesis (www.genecards.org). The other genes examined included CLCN3, FZD7, GJA1, MAP4K3, NFAT5 and RARB. While RARB was also downregulated in the epithelioid sarcoma lines, it was also not expressed in G401. Since 80 all of these genes are direct targets of miR-206 at least during miR-206 dependent myoblast differentiation(Dey et al., 2011), one would expect a more generalized pattern of downregulation of miR-206 targets if this microRNA was the cause of loss of SMARCB1 expression. However, the context specific and complex interactions of microRNAs meant that the possibility of microRNA silencing could not be ruled out. We therefore set out to measure microRNA levels directly using real-time PCR with the hypothesis that HSES, being the only epithelioid sarcoma cell line without biallelic inactivation of SMARCB1, would have a unique upregulation of one of the candidate microRNAs. miR206 and miR193a5p were among the more prominently suggested microRNAs suspected and both are predicted to target SMARCB1. We found that miR206 had low expression in HSES; however, miR193a5p had prominent expression although not in a unique manner. VAESBJ, which has mutated alleles of SMARCB1, has considerably higher levels of miR193a5p compared to HSES. Similarly, synovial sarcoma line SYO1 as well as myxoid liposarcoma line 1765/92 and clear sarcoma line Su-CCS have prominent levels of miR193a5p comparable to HSES, yet they express high levels of SMARCB1 (Fig 3-3C). These findings argue against a role for these suggested microRNAs in SMARCB1 inactivation in the tested models. We further expressed sponge antimiRs against all 4 suggested miRNA species: miR206, miR193a-5p, miR381-3p, and miR671-5p. The antimiRs had an mCherry marker to enable confirmation of transduction efficiency (Fig 3-3D). None of the miR inhibitors led to increases in levels of SMARCB1 — if anything the mRNA levels seemed to be further reduced§§. §§ Given the very low levels of expression of SMARCB1 in HSES, drawing conclusions on reduction detected on qPCR should be done with care. Miniscule insignificant events could reveal such a result given comparison to already very low levels. 81 3.3 Discussion The inactivation of SMARCB1 without DNA mutations is a phenomenon observed in several tumors including epithelioid sarcoma, atypical teratoid/rhabdoid tumor(Tsai et al., 2012) and renal medullary carcinoma(Calderaro et al., 2012). However, unique to epithelioid sarcoma, has been the higher frequency of such observations(Kohashi et al., 2009). There are significant disagreements in the literature with reported inexplicable inactivation occurring as frequently as in about 90% of cases(Kohashi et al., 2009) to 17% of cases (Sullivan, Folpe, Pawel, Judkins, & Biegel, 2013) to 10% of cases(Le Loarer et al., 2014). Using stringent criteria, we observed 13% of our cases to have maintained intact alleles and therefore we agree with the lower estimates. Point mutations are rare in epithelioid sarcoma (Flucke et al., 2009) and in our next generation sequencing and Sanger sequencing results we did not observe any. The majority of deletions observed in epithelioid sarcoma involve the entire gene and small deletions are less frequently observed(Sullivan et al., 2013). Large deletions are the main mutation form of SMARCB1 in epithelioid sarcoma followed by fusions (Chapter 2). Yet it is without question that there are certain cases that have maintained intact alleles that appear to have been silenced epigenetically. We found supportive evidence for this by identifying a proximal variant epithelioid sarcoma cell line, HSES, that has maintained an intact SMARCB1 allele. Detailed examination including sequencing of the gene as well as breakapart FISH and MLPA ruled out the possibility of SMARCB1 point mutations, small deletions or translocations in this line. We also performed methylation-specific PCR and did not find evidence for promoter methylation. Work by Papp et al. on epithelioid sarcoma tissue samples was supportive of our observation of lack of promoter 82 methylation(Papp et al., 2013). However, because we had identified an in vitro model, we could extend our findings by long term DNA methyltransferase inhibition, which still did not lead to protein re-expression meaning that methylation at distant sites did not cause the silencing of SMARCB1. Similarly we could not restore SMARCB1 protein via proteasome inhibition, EZH2 inhibition or broad-spectrum HDAC inhibition. However, from qPCR we noticed that most of the silencing does happen at the transcriptional level. There have been several microRNA suggested to be overexpressed in epithelioid sarcoma and lead to silencing of SMARCB1, the most prominent of which are miR-206 and miR-193a5p. None of these microRNAs were uniquely high in HSES, nor did microRNA inhibition via sponge antimiR vectors lead to an increase in mRNA levels of SMARCB1. Therefore microRNA does not explain the SMARCB1 loss in this model. Understanding the mechanism of SMARCB1 silencing in cases that have an intact allele is of value. This is because as an epigenetic phenomenon, it could be potentially reversible, which would restore the expression of the key inhibitor of epithelioid sarcoma tumor formation. Yet given the low frequency of this event and the rarity of epithelioid sarcoma, stratification of therapy might not be a practical possibility at the moment. However, because none of the common epigenetic mechanisms seem to be at work, this phenomenon is of great academic interest. Perhaps a novel silencing mechanism is involved that could also be occurring in more common cancers. Although we were unable to identify the mechanism of SMARCB1 silencing using any of several tested contemporary technologies, we hope that HSES can serve as an in vitro model in future studies attempting to understand novel mechanisms of gene silencing. 83 3.4 Methods 3.4.1 Multiplex Ligation-dependent Probe Amplification (MLPA) Pre-designed MLPA probe sets for SMARCB1 (MRC-Holland, Amsterdam, Netherlands) were used and post ligation amplified targets were analyzed on the 3130xl Genetic Analyzer (Applied Biosystems). 3.4.2 Methylation specific PCR Bisulfite conversion was done using Qiagen EpiTect Bisulfite kits (Qiagen). Methylation specific PCR primers were designed using MSPPrimer (Brandes, Carraway, & Herman, 2007) (http://www.mspprimer.org/) and three sets were optimized to cover the CpG island of SMARCB1. CpG Methylated Jurkat Genomic DNA (New England biolabs) was used as positive control. PCR reactions as described in chapter 2 were completed and the products were ran on 2% agarose gels containing SYBR green. 3.4.3 Cell culture, drug treatment and Western blots Cell lines were obtained from the original laboratories (HSES and Epi544) or ATCC (VAESBJ, G401) and grown in DMEM supplemented with 10% FBS. 5-Aza-DC (Sigma-Aldrich, St. Louis, MO), GSK126 (Cayman Chemical, Ann Arbor, MI), MG132 (Sigma-aldrich), and romidepsin (Tocris, Bristol, UK) were dissolved in DMSO and diluted to a minimum of 1:1000 from the stock before applying to cells. Adherent cells were scraped and collected in cold PBS, centrifuged at 4˚C for 5 minutes at 5000g after which the PBS was removed. The pellets were suspended in ice cold RIPA buffer supplemented with proteasome inhibitors, sodium orthovanadate, and PMSF (Santa Cruz Biotech., Santa Cruz, CA). The extracts were centrifuged at 15000g for 5 minutes and supernatant collected. Protein amounts were quantified via BCA assay (Santa Cruz) and 84 SDS-PAGE was carried out after addition of 4X sample buffer (Biorad, Hercules, CA) in 8-12% polyacrylamide gels. Primary antibodies used were SMARCB1 (BD Biosciences, Cat. 612111. 1:500), DNMT1 (NEB, M0231, 1:1000) and VIM (Santa Cruz Biotech., Cat. 5573, 1:1000). 3.4.4 qPCR Taqman gene and microRNA specific expression primers and probes were obtained (Life Technologies). Total RNA was extracted using miRNeasy kits (Qiagen) and 1 μg as measured by Nanodrop (Thermoscientific, Waltham, MA) was converted to cDNA by HiCapacity cDNA kit (Life Technologies). An internal control of ACTB/TBP and U6 was used for gene or microRNA expression analysis respectively. Real time PCR was performed on an Applied Biosystem 7900HT. The ddCT method was used to normalize expression, and results expressed relative to one of the samples (defined to have relative expression of 1.0). The error bars are calculated as the range given by 2-ddCt±s with “s” being the standard deviation of ddCT values from the three technical replicates. 3.4.5 Viral production and infection with sponge miRNA inhibitors Human lentiviral miRNA inhibitor vectors (Genecopoeia, MD, USA) with mCherry tags on an AM03 backbone were co-transfected with lentiviral packaging vectors pCMV-VSVG and pCMV-dR8.91 in HEK293T cells. The generated viral particles were collected after 48 hours and applied to HSES after filtration via 0.45μm filters. After 24 hours the media was changed with fresh media and the transduced cells were selected with hygromycin at 500μg/ml for 1 week. Transduction efficiency was evaluated by Axio-observer Inverted microscopes (Zeiss, Jena, Germany). 85 Figure 3-1: Evaluation for SMARCB1 mutations in epithelioid sarcoma cell lines (A) MLPA assay with multiple probes across the genome as well as at SMARCB1. Epi544 and G401 show complete loss of SMARCB1 while still amplifying at other genomic loci. VAESBJ lacks 5’UTR and exon 1 of SMARCB1. Although heterozygous, HSES has maintained all of regions of SMARCB1. (B) FISH showing heterozygous deletion of SMARCB1 in HSES as well as lack of translocation with a different set of breakapart probes(white arrows). Epi544 has homozygous deletion and VAESBJ has maintained two alleles both of which have lost exon1 and 5’UTR (A). (C) PCR across different exons of SMARCB1 confirms the maintenance of all exons and UTR in HSES. The SNP6.0 results for the cell lines as well as probe locations for SMARCB1 detection are indicated as well. 86 Figure 3-2: Transcriptional levels of SMARCB1 in epithelioid sarcoma cell lines (A) Taqman real-time PCR amplification plots showing the presence of SMARCB1 transcripts in HSES. Despite the loss of 5’UTR and exon 1, there is still some transcript detectable in VAESBJ; however, this amount is minute compared to HSES (B). 176592 is a myxoid liposarcoma line showing a much smaller Ct and hence greater abundance of SMARCB1 transcripts (A, left). ACTB was used as internal control to ensure equal loading of cDNA in all cases (A, right). Comparison to a more diverse collection of cell lines shows that HSES has very low abundance of SMARCB1 transcript which amounts to about 2% of transcript levels in HEK293t cells (C). A673: Ewing’s sarcoma line, SuCCS: Clear cell sarcoma line, Syo1: synovial sarcoma line. 87 Figure 3-3: Epigenetic mechanisms of SMARCB1 silencing in HSES (A) Methylation Specific PCR across of the promoter CpG island of SMARCB1 with 3 different primer sets reveals lack of methylation in HSES epithelioid sarcoma cell line. (B) Western blots show that prolonged inhibition of DNA methyltransferases with 5-Aza-dC does not lead to re-expression of SMARCB1 in HSES. (C) MicroRNA levels as evaluated by Taqman qPCR and normalized to U6 levels. (D) Upper two panels show a phase-contrast and mCherry fluorescent images of HSES cells after transduction with anti-miR206 vectors. Several miR-206 targets such as CLCN3, GJA1 are upregulated whereas SMARCB1 is not. Given the low levels of SMARCB1 transcripts in HSES, the down-regulation seen is likely a result of miniscule changes in RNA abundance in the post-transduction preparations. CLCN3 and GJA1 are expressed with relatively high mRNA abundance in HSES (Ct values under 35). 88 Table 3-1: FISH results on epithelioid sarcoma TMA Summary of FISH result using probes to detect SMARCB1 (Chapter 2 methods) for cores with sufficient tumor cells. Tumor cells were counted using matching H&E areas. Yellow highlight indicates cases where more than 40% of individually counted cells were observed to have one or no alleles. Pink indicates a total SMARCB1 to Telomere 22 probe ratio between 0.4-0.8 and blue indicated a ratio less than 0.4. There were 2 cores per case. 89 Chapter 4: In Vitro Studies and Epigenetic Modifying Compounds 4.1 Introduction Even though the mechanism of SMARCB1 silencing remains unclear in a portion of epithelioid sarcoma cases, our deep sequencing studies in Chapter 2 showed that SMARCB1 deletions are the most recurrent event in epithelioid sarcoma and hence a critical driver of oncogenesis. The effect of SMARCB1 loss has been studied in greatest depth in malignant rhabdoid tumors and atypical teratoid/rhabdoid tumors as discussed in Chapter 1. SMARCB1 has been found to affect many critical pathways in malignant rhabdoid tumor and atypical teratoid/rhabdoid tumor including cell cycle regulation, polycomb regulated genes, Wnt signaling, and sonic hedgehog. However, only one study thus far has taken a look at epithelioid sarcoma where they restored SMARCB1 in VAESBJ (Brenca et al., 2013) and observed inhibition of cellular proliferation in vitro and in vivo, albeit without examination of pathways affected. Important to note is that VAESBJ has lost CDKN2A via homozygous deletion, therefore, regulation of this gene as way of exerting control over the cell cycle by SMARCB1 is not happening. VAESBJ, although a readily available cell line, is perhaps not the most disease-specific cell line model to study epithelioid sarcoma as it is a metastatic line and our findings in Chapter 2 showed it to have a somewhat different expression profile than the other epithelioid sarcoma lines (Figure 2-6). The mechanisms by which SMARCB1 loss leads to oncogenesis in epithelioid sarcoma are unknown, and it is also not known whether these are similar to the observations made in malignant rhabdoid tumor and atypical teratoid/rhabdoid tumor. Furthermore, the status of SWI/SNF complex after the loss of SMARCB1 in epithelioid 90 sarcoma has not been examined. Perhaps if the residual complex is still existent, it could be targeted with a synthetically lethal approach similar to those applied in ARID1A mutant lines via ARID1B inhibition(Helming, Wang, Wilson, et al., 2014). In this chapter we discuss our creation of a SMARCB1 inducible system, which we use to gain insight into gene regulation by SMARCB1 in epithelioid sarcoma and also to examine potential therapeutics. Additionally, we examine the status of the residual SWI/SNF complex in epithelioid sarcoma and explore the possibility of using a synthetic lethality approach as therapy. 4.2 Results 4.2.1 Inducible expression of SMARCB1 We generated inducible cell lines from the malignant rhabdoid tumor G401 and the classic epithelioid sarcoma Epi544 lines. These cell lines were created with the Tet-on vectors from Clontech (Kyoto, Japan). SMARCB1 was subcloned using the Gateway recombination system (Life technologies) into the Plvx vector under the Tetracycline Responsive Element promoter. A second vector containing mutant E. coli tetracycline repressor sequence was concurrently transduced with either the plvx SMARCB1 or control vectors into the target cells. The resulting 4 cell lines were named G401_plvx_SMARCB1, G401_plvx_Control, Epi544_plvx_SMARCB1 and Epi544_plvx_Control. G401_plvx_SMARCB1 was validated using Western Blotting, which showed dose dependent SMARCB1 expression after the addition of doxycycline (Figure 4-1A). Immunocytochemistry showed SMARCB1 protein production in the plvx_SMARCB1 cells but not the plvx_Control cells after the addition of 0.5μg/ml of doxycycline confirming the Western blot findings (Figure 4-1B). From 91 immunocytochemistry we noted that more than half but not all the cells of the plvx_SMARCB1 lines would be induced to produce SMARCB1 protein. The G401_plvx_SMARCB1 cell line also showed decreased proliferation as measured by MTS assay over three days (Figure 4-1C). The observed effect was smaller than previously reported in the literature, perhaps due to different methodology used but more likely because of the smaller magnitude of exogenous SMARCB1 production (which was ideal for our purposes). As with prior reports, we noted an increase in CDKN2A levels after induction of SMARCB1. This effect was dose dependent: as the amount of SMARCB1 protein increased, so did the amount of CDKN2A (Figure 4-1D). We examined several other genes known to be regulated by SMARCB1 in G401 via qPCR and found expected changes (Figure 4-1E). These included downregulation of E2F1, and upregulation of CDKN1A. The Epi544_plvx_SMARCB1 line could also be successfully induced and this caused down regulation of CCND1 (Figure 4-1F). However interestingly, CDKN2A was not restored in Epi544 with the re-expression of SMARCB1. Epi544 was the only epithelioid sarcoma cell line without homozygous deletion of CDKN2A based on SNP6.0 results and lack of induction of CDKN2A would mean a stronger silencing mechanism that is in effect in epithelioid sarcoma. Accordingly, we did not observed the same extent of growth arrest in Epi544_plvx_SMARCB1 as in G401_plvx_SMARCB1. 4.2.2 Induction of SMARCB1 has regulatory effects similar to EZH2 inhibition Using these SMARCB1 inducible models, we examined a more comprehensive list of genes reported to be regulated by SMARCB1 in malignant rhabdoid tumors and atypical teratoid/rhabdoid tumors (Figure 4-2). Many of these targets are also known to 92 be regulated by the polycomb complex(Knutson et al., 2013). We noted that the regulation was different between G401 and Epi544. For instance, the downregulation of GLI1 was more readily observed in Epi544 whereas upregulation of PTPRK was much more significant in the G401 line (Figure 4-2). This points out cell line specific results upon restoration of SMARCB1. Given that many of these genes are also known to be regulated by the PRC2 complex, in which EZH2 serves as the catalytically-active histone methyltransferase, we treated the lines with the EZH2 inhibitor GSK126 for 7 days. We observed that several of the genes whose expression had changed upon SMARCB1 restoration also exhibited similar changes upon pharmacologic inhibition of EZH2. Certain genes, for instance CDKN1A in Epi544 and PTPRK in G401 showed large changes in transcript levels with the addition of SMARCB1, inhibition of EZH2 and with the combination of these two treatments (Figure 4-2). Claims to statistical significance were not be made here because these experiments were not run in parallel and did not have identical biological replicates. However, different doses of EZH2 inhibition and SMARCB1 induction, examined in separate sets of experiments showed similar results. With this observation we thought that EZH2 inhibitor compounds deserved further detailed examination in epithelioid sarcoma, especially given the published observation that they are effective in the malignant rhabdoid tumor line G401(Knutson et al., 2013). 4.2.3 EZH2i on epithelioid sarcoma For these studies we first examined our lines using the previously utilized GSK126 (Epizyme, Camebridge, MA) compound via MTS assays. We also used HEK293t as a SMARCB1 wild-type negative control and G401 as a GSK126 sensitive positive control(Knutson et al., 2013). We noticed that this compound led to reduced proliferation 93 in the G401 as well as HSES and Epi544 lines, while it did not do so for HEK 293t cells. Given this initial positive finding, we extended our experiments to include two additional EZH2 inhibitor compounds: EPZ005687 (Epizyme, Camebridge, MA) and UNC1999 (Sigma-Aldrich). Based on prior publications and our initial experiments, we realized that an extended period of treatment would be required to observe the effects of the EZH2 inhibitors. We therefore pretreated the cell lines for five to six days and then replated them, resupplying the drug treatment, and measured proliferation over the period of one week with Incucyte. The SMARCB1 mutant G401 malignant rhabdoid tumor line and all three epithelioid sarcoma lines were more sensitive than HEK293t (SMARCB1 wild-type) to UNC1999 (Figure 4-3). GSK126 also had a greater anti-proliferative effect on G401, Epi544 and HSES when compared to 293t cells. However, the VAESBJ epithelioid sarcoma cell line was resistant to this compound, behaving more similarly to HEK293t cells. Given that VAESBJ is a metastatic line, its behavior might indicate acquisition of a resistant state to EZH2 inhibition 4.2.4 The residual SWI/SNF complex in epithelioid sarcoma An interesting observation from our expression analysis in Chapter 2 was that, other than SMARCB1 and the neuron-specific members such as ACTL6B, all the other members of the SWI/SNF chromatin-remodeling complex maintained their expression in the epithelioid sarcoma lines. To explore further, we used the epithelioid sarcoma tissue microarray and carried out immunohistochemistry for several key SWI/SNF component members including ARID1A, SMARCA4 and SMARCA2. Our results confirmed the deep sequencing expression analysis showing that all the 20 immunohistochemically evaluable cases express these proteins (Figure 4-4). A recently published study made 94 similar observations but also noted lack of PBRM1 expression in epithelioid sarcoma (L. Li et al., 2014). The same study also indicates some loss of SMARCA2 in epithelioid sarcoma; however, their evidence is mainly supportive of loss of PBRM1. We found SMARCA2 to be strongly positive. This is of particular interest as malignant rhabdoid tumors are known to epigenetically silence SMARCA2 in 70% of cases and SMARCB1 re-expression induces SMARCA2 (Kahali et al., 2014). Of the 20 cases examined we did not find loss of SMARCA2 in any; Li et al. only found loss of SMARCA2 expression in 1 of 23 cases they examined. Yet again, in SMARCA2, we find another gene whose expression is different in most cases of malignant rhabdoid tumor and epithelioid sarcoma. We next moved to our cell line models. First we evaluated levels of PBRM1 with the hypothesis that perhaps the PBAF complex is lost in epithelioid sarcoma while the BAF complex might be maintained as PBRM1 is a unique subunit of the PBAF complex (Figure 1-2). We used a different antibody to that of Li et al. but found definite presence of PBRM1 in the VAESBJ cell line at levels comparable to HEK293t cells. Therefore, we suspect that the phenomenon of PBRM1 loss is not universal in epithelioid sarcoma. Next we wanted to assess whether the SWI/SNF could still assemble despite the loss of SMARCB1. We hence performed co-immunoprecipitation (Co-IP) on nuclear extracts from our lines and blotted for various SWI/SNF members (Figure 4-5). Interestingly, all the remaining members of SWI/SNF maintained interactions, not only in the malignant rhabdoid tumor line G401 but also in VAESBJ. This means that the complex does not go through complete degradation despite the loss of the major component encoded by SMARCB1. 95 4.2.5 Synthetic lethality in the residual complex Next the effect of targeting the residual complex was examined. This has been shown in ARID1A mutant lines where the mutant cell lines were more sensitive to ARID1B knockdown(Helming, Wang, Wilson, et al., 2014). A similar phenomenon has also been described for a SMARCA4 mutant line with SMARCA2 synthetic lethality (Hoffman et al., 2014) as well as in G401 where knockdown of SMARCA4 led to increased cell death(X. Wang et al., 2009). We therefore decided to examine an epithelioid sarcoma system to see if we could reproduce the same results using G401 as positive control. We observed that indeed G401 and VAESBJ were both more sensitive to siRNA of SMARCA4 (aka BRG1). This was true with different siRNA combinations involving two different miRNA species, a means to further increase specificity for the intended target. An important observation is the increase in the slope of the growth curve for HEK293t cells; in contrast, this slope decreases for G401 as well as VAESBJ lines (Figure 4-4 C). This increase in growth is supportive of SMARCA4 functioning as a tumor suppressor in many tumors given a wild type SWI/SNF background. However, loss of SMARCA4 in SMARCB1 aberrant SWI/SNF complexes leads to reduced proliferation, supporting the concept that there is an increased dependence in SMARCB1 mutant tumors on the remaining SWI/SNF complex. 4.3 Discussion In this chapter we report two novel, previously unreported observations about epithelioid sarcoma. First, that epithelioid sarcoma is susceptible to EZH2 inhibitors in vitro – possibly due to dysregulation of genes that were otherwise coregulated by the 96 SWI/SNF and PRC2 complexes. This susceptibility is greater with UNC1999, which inhibits both EZH2 and EZH1, than GSK126 (which is an EZH2 specific inhibitor). It is important to recall that links to the polycomb complex were critical in the identification of the SWI/SNF complex. In fact, Brahma (Brm) in Drosophila was identified as one of a select number of genes whose knockdown reverses the polycomb-loss phenotype. Brm was found to be necessary for the transcription of ectopically expressed genes that would have otherwise been silenced by the polycomb complex (Tamkun et al., 1992). The second novel observation was that epithelioid sarcoma cells maintain a residual complex which assembles even in the absence of SMARCB1. More importantly, the knock down of the core ATPase of the remaining SWI/SNF leads to a context specific proliferation inhibition and synthetic lethality. Additionally, we found support for SMARCA2 being predominantly expressed in epithelioid sarcoma which makes this a feature distinguishing between epithelioid sarcoma and malignant rhabdoid tumor. Although not explored in detail in our study, we did not find support for the loss of PBRM1 in the tested epithlioid sarcoma cell lines. Drug based inhibition of SMARCA4 is gaining traction. The crystal structure of its bromodomain has been mapped(Filippakopoulos et al., 2012). Development of compounds with increased efficacy in inhibiting bromodomains has also gone through major progress (Filippakopoulos et al., 2010). In fact an inhibitor of the SMARCA4 bromodomain, PFI-3, has already been developed. Therefore, we suggest two possibilities as targeted therapeutic options for epithelioid sarcoma exploiting the very mutation that gives rise to this tumor. Loss of SMARCB1 makes epithelioid sarcoma susceptible to EZH2/EZH1 inhibition as well to SMARCA4 inhibition. Given that 97 VAESBJ shows some resistance to EZH2 inhibitors but is still susceptible to SMARCA4 knockdown, SMARCA4 inhibition might be a more comprehensive and direct approach. 4.4 Methods 4.4.1 Vector cloning Vectors used for the generation of the tetracycline inducible lines were kindly supplied by Dr. Samuel Aparicio’s lab (BC Cancer Agency, Vancouver). These were modified from original vectors obtain from Clontech (Kyoto, Japan). The modified plvx vector had a Gateway cloning site introduced right after the TRE promoter. This allowed introduction of SMARCB1 via LR clonase reaction. The SMARCB1 sequence was obtained in a donor vector format from Genecopoeia (Rockville, MD). SMARCB1 has two transcript variants: a longer and a shorter form. We chose the longer transcript, which encodes nine additional amino acids. 50μl of MAX efficiency E. coli (Invitrogen) was transformed with addition of 1μl of the clonase reaction mix and incubation for 30 minutes on ice, heat shocked for 45 seconds at 42˚C and pre-incubated in Luria-Bertani medium for 1 hour. The bacteria were then plated on LB agar with ampicillin selection. The next day the cultures were selected and grown in 50-100ml LB media with ampicillin selection for midi-prep(Qiagen) extraction the following day. The amplified and isolated vectors were digested by restriction enzyme nucleases and run out on 1% agarose gels to ensure the success of the protocol. The insert was also Sanger sequenced to ensure no mutations were introduced during the process. The viral production and packaging vectors were the same as described in chapter 3 and underwent similar midi-prep amplification. 98 4.4.2 Viral transduction of the inducible system Similar to the methods described in Chapter 3, HEK293t cells were used for virus production. In short, the mutant tetracycline repressor coding vector and the plvx TRE vector containing SMARCB1 or no insert were transfected with the packaging vector d8.91 and vsvg in a 10:5:1 ratio using lipofectamine 2000 (Life). The supernatants were collected after 48 hours, filtered through 0.45μm filters and were both added to the target cells Epi544 or G401 (which were seeded at 30% confluency in 6-well plates before the addition of virus). After 3 days, concurrent puromycin selection at 1ng/μl and G418 selection at 800ng/μl was applied with wells of untransduced cells used as control to observe selection efficacy. The resulting cells were used for experiments. 4.4.3 Immunocytochemistry Cells were grown in chamber slides (Sigma-Aldrich) with the addition of 0.5μg/ml doxycycline for two days. The slides were then fixed in pre-chilled 100% methanol for 5 minutes followed by 4% paraformaldehyde addition for 10 minutes at room temperature. The slides were heated in antigen retrieval buffer at 95˚C for 10 minutes, before being washed in PBS and permeabilized with Tirton X-100. The cells were then blocked with 1% BSA PBST solution for 30 minutes and incubated with SMARCB1 (BD Biosciences, Cat. 612111) antibody diluted in BSA PBST for 1 hour. After washing with PBS, cells were incubated with HRP-conjugated anti-mouse secondary antibodies for 1 hour. DAB HRP (Thermoscientific. Wlatham, MA) substrate was applied to the slides. After dehydration with ethanol, the slides were examined. 99 4.4.4 MTS assay and Incucyte Cells were seeded at a concentration of 500 to 2000 cells per well in 96-well plates. For measurement of proliferation through mitochondrial activity, 40 μl of MTS solution (Promega, Madison, WI) was added and incubated for 3 hours at 37°C and 5% CO2. The colorimetric changes were then measured at 490 nm using Infinite M200 PRO Tecan microplate reader (Maennedorf, Switzerland). Three wells were used for replicates for each condition. For Incucyte proliferation experiments, cells were seeded at 500 to 2000 cells per well in 96-well plates, and plates were incubated in the Incucyte ZOOM® device inside a 37°C and 5% CO2 incubator. Four images per well were collected and confluency determined using the ZOOM® software using phase-contrast only conditions. 4.4.5 siRNA knockdown and immunoprecipitation siRNA against SMARCA4 were obtained from the Life technologies SureSelect collection (siRNA#1:s13141, siRNA#2:s13139). Various amounts of siRNA (10-50nM) and 3-9μl RNAiMax lipofectamine (Life technologies) were incubated with the target cells in 6-well plates. Optimal knock down was assesed via Western blot analysis 72 hours after siRNA addition(Figure 4-4B). Greatest knockdown without major cell death was achieved with 50nM siRNA and 8μl of RNAiMax lipofectamine. Transfections were done in reverse with trypsinized and counted cells added to a mixture of siRNA and lipofectamine that had been incubated at room temperature for at least 15 minutes. Nuclear extracts for immunoprecipitation were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific: 78833). Nuclear extracts were diluted with RIPA buffer supplemented with protease inhibitors (Santa Cruz). For each sample 250ul at 2.0mg/ml was prepared and incubated overnight with 2μg of either 100 SMARCC1 (Santa Cruz: 9746) or goat IgG (Santa Cruz:2028) antibodies at 4°C. The following day, protein G Dynabeads (Life Technologies: 10004D) were added and incubated at 4°C for 3 h. Beads were then washed three times with RIPA buffer and resuspended in reducing SDS gel loading buffer. SDS-PAGE was carried out in 7%-12% polyacrylamide gels. After transfer, each nitrocellulose membrane was incubated with primary antibodies in 5% w/v skim milk for two hours, washed with PBST, followed by incubation with HRP or fluorescent tagged secondary antibody (1:10000) for 1 hour. The membranes were visualized under LI-COR Odyssey CLx (LICOR, NE) or with G:Box Chemi XRQ (Syngene, Cambridge, UK) after addition of Supersignal Western Femto substrate (Thermoscientific). The primary antibodies used were SMARCC1 (Santa Cruz: 9746, 1:1,000); ARID1A (Bethyl Laboratories: A301-041A, 1:1,000); PBRM1 (Santa Cruz: 390095, 1:500); SMARCA4 (Santa Cruz: 17796, 1:500); SMARCC2 (BL: A301-039A, 1:3,000); SMARCD1 (BL: A301-595A, 1:3,000); SMARCE1/BAF57 (BL: A300-810A, 1:3,000); ACTL6A/BAF53A (BL: A301-391A, 1:3,000); SMARCB1 (BD Biosciences 612111, 1:500) and VIN/Vinculin (Santa Cruz: 5573, 1:1000). 4.4.6 qPCR The same methodology as described in section 3.4.4, described on page 84, was used. The taqman specific probes were CDKN1A (Hs00355782_m1), CDKN2A (Hs00923894_m1), EZH2 (Hs00544833_m1), GLI1 (Hs01110766_m1), PTPRK (Hs00267788_m1), and SMARCB1 (Hs00268260_m1). All probes were designed to cross exon boundaries to avoid genomic DNA amplifications. There were three technical replicates. Fold changes were calculated relative to the untreated samples (0.1% DMSO only, no GSK126, no Doxycycline induction). 101 Figure 4-1: Validations of doxycycline inducible cell lines. (A) Western blot showing different levels of induction of SMARCB1 in the G401_plvx _SMARCB1 cell line with various concentrations of doxycycline: samples were collected 48hrs post treatment. (B) Immunocytochemistry after induction for 48 hours in G401_plvx_ SMARCB1 with 0.5μg/ml doxycycline. The inset is the same experiment done in G401_plvx_ Control. (C) Reduction in proliferation as assessed by MTS assay in G401_plvx_ SMARCB1 with various concentration of doxycycline shown in μg/ml. (D) Dose dependent increase in SMARCB1 as well as CDKN2A by western blot in G401_plvx_SMARCB1 with addition of doxycycline. (E) Western blots after 48hr induction with doxycycline (10μg/ml). CDKN2A is induced in G401_plvx_SMARCB1 whereas CCDN1 is downregulated in Epi544_plvx_SMARCB1. (F) qPCR on some of the downstream targets of SMARCB1 in G401_plvx_SMARCB1 confirming expected downstream effects. 102 Figure 4-2: EZH2 inhibition and SMARCB1 induction on targets of SMARCB1 (A) qPCR of known SMARCB1 targets with the induction of SMARCB1 in G401_plvx_SMARCB1 line for 7 days (10ug/ml doxycycline) with or without the addition of the EZH2 inhibitor GSK126 (5μM). All fold changes are relative to untreated samples. (B) Same experiment in Epi544_plvx_SMARCB1. The results are different depending on parent cell line, however, CDKN1A shows upregulation with either EZH2 inhibition or SMARCB1 induction. CDKN2A induction occurs with both SMARCB1 addition and EZH2 inhibition in G401 but not in Epi544. Taqman qPCR methodology was used for this experiment as described in section 4.4.6. The error bars are based on the standard deviation of ddCt calculated from three technical replicates. 103 Figure 4-3: EZH1/2 inhibition in epithelioid sarcoma Confluence of wells are shown on the ordinate axis (replicates of 3 wells, within a 96 wells plate format, measured by IncucyteZOOM®). The abscissa axis demonstrates the time (hours) since initial seeding in the 96-well plates. Cells were pre-treated with the same concentration of UNC1999 for 5 days before the start of measurements. HEK 293t cells (with an intact SMARCB1) served as negative control. G401 served as positive control with known efficacy of EZH2 inhibition previously published by others. VAESBJ, HSES and Epi544, all epithelioid sarcoma cell lines, show increased susceptibility to EZH1/2 inhibition by UNC1999, particularly at 5μM. 104 Figure 4-4: Immunohistochemical analysis of SWI/SNF components in epithelioid sarcoma Images from the immunohistochemistry results for representative cases from the epithelioid sarcoma tissue microarray. Despite the SMARCB1 loss, SMARCA2, SMARCA4, and ARID1A are prominently expressed. Transcripts for all these SWI/SNF complex members were also present at high levels in the FPKM data discussed in Chapter 2. Images were taken under 10X original objective magnification. 105 Figure 4-5: Residual SWI/SNF and potential for synthetic lethality. (A) Co-immunoprecipitation using SMARCC1 as bait. In malignant rhabdoid tumor and epithelioid sarcoma lines, the complex members associate with each other despite the loss of SMARCB1. (B) Optimizing transfection conditions for knockdown of SMARCA4 with increasing siRNA concentration (Concentration range: 10nM to 50nM). (C) Knockdown of SMARCA4 leads to increased rate of growth (slope of curves) in SMARCB1-wild type HEK293t, whereas it reduces the growth rate of malignant rhabdoid tumor G401 and epithelioid sarcoma VAESBJ cell lines. Confluence of wells are shown on the ordinate axis (replicates of 4 wells, in 96 well plate format, measured by IncucyteZOOM®). The abscissa axis demonstrates the time since initial seeding in 96-well plates in hours. Knockdown condition 6 was used for this experiment (8μl lipofectamine and 50nM siRNA in 6-well format). 106 Chapter 5: Conclusions and Future Directions 5.1 Summary of findings To summarize the novel findings of this work, three main concepts were developed. Firstly, we describe for the first time, the genomic landscape of epithelioid sarcoma, comparing its mutation rate with other tumors including other SMARCB1 inactivated cancers. We show that epithelioid sarcoma has a higher mutation burden compared to rhabdoid tumors and has a mutation rate more comparable to some typical adult tumors such as cervical and esophageal carcinomas. SMARCB1 remains the most significant mutation in this tumor. Second, we characterize existing epithelioid sarcoma models further and establish new in vitro models. This includes the identification of an intact SMARCB1 allele in the cell line HSES. This is proof of principle for the observation that indeed when examined in detail, some epithelioid sarcoma cases maintain wild type alleles. Additionally, we make an inducible SMARCB1 restoration cell line, Epi544_plvx_SMARCB1. Lastly, we suggest two possible novel therapeutic options for epithelioid sarcoma: EZH1/2 inhibitors as well as SMARCA4 inhibition whose mechanisms of action are tumor specific. 5.2 Insights from the genomic landscape of epithelioid sarcoma Given the very low mutation rate observed in rhabdoid tumors and some of the other examined SWI/SNF mutated cancers, a conclusion might be drawn is that these tumors are genetically stable but because of SWI/SNF abnormality develop epigenetic (as opposed to genetic) instability. It might therefore be tempting to draw the conclusion that all tumors with SWI/SNF driving mutations will have low mutation rates. However, we clearly see that is not the case in epithelioid sarcoma. Additionally, there is an abundance 107 of translocations and copy number changes in epithelioid sarcoma. Given that SWI/SNF has been implicated in DNA repair machinery and the mechanics of cell division (Dykhuizen et al., 2013), this lack of genomic stability in epithelioid sarcoma makes biological sense. However, the low mutation burden of rhabdoid tumors would be unexplainable. Perhaps the rapid onset and aggressiveness of rhabdoid tumors do not allow enough time for tumor cells to accumulate mutations whereas the relatively more indolent growth rate in epithelioid sarcoma does. Yet the age of onset of these two tumors, rhabdoid vs epithelioid sarcoma, are quite different with epithelioid sarcoma typically presenting two decades later in life than rhabdoid tumors. This means that cells would have more time to accumulate mutations and indeed, we do not know at what point SMARCB1 is lost in the development of epithelioid sarcoma. Perhaps initiating tumor populations exist for a long period of time before the detection of tumor. On the other hand, the assumption that SMARCB1 loss enhances mutation instability in epithelioid sarcoma is not conclusive. Perhaps, it is the increase in mutational instability that manifests in the loss of SMARCB1. To support this concept, we find that several epithelioid sarcoma samples, as well as previously reported transcriptome data of other SMARCB1 negative tumors (Jamshidi et al., 2014), show a cluster of fusion breakpoints around the SMARCB1 locus. This could indicate a higher frequency of breaks at these positions and therefore a higher likelihood of loss of SMARCB1. A possible way to examine these possibilities would be to use our inducible systems and induce expression of SMARCB1 for extended periods of time. The cells pre- and post induction could be compared with one another in terms of mutational load and genomic landscape. 108 5.3 Lineage of epithelioid sarcoma The clustering of cell lines (Figure 2-6B) shows that epithelioid sarcoma does not resemble a specific tissue lineage. The cell lines with most similar expression patterns to the epithelioid sarcoma lines were established from kidney, pancreas, ovary, colorectal, and lung cancers. Some of these might come across as epithelial cancers. However, all had mesenchymal signatures as discussed by Klijin et al. 2014. Some of these lines have likely gone through epithelial-to-mesenchymal transition. On the other, epithelioid sarcoma lines might have gone through a de-differentiation process via SMARCB1 loss, or it is possible that they arise from mesenchymal stem cells that never fully mature because of the aberrant SWI/SNF complex. To address this issue we have initiated expression profiling studies using our inducible lines to see how the re-expression of SMARCB1 in Epi544 would change the positioning of this cell line on the cell line expression map. 5.4 SMARCB1 loss in epithelioid sarcoma The presence of intact SMARCB1 alleles has been an issue of considerable debate in the recent literature on epithelioid sarcoma(Kohashi et al., 2009; Le Loarer et al., 2014; Papp et al., 2013; Sullivan et al., 2013). We found that this phenomenon is indeed a real occurrence and not something that might have been missed because of the limitations of the evaluation methods on formalin-fixed paraffin-embedded tumor material. We were able to do so by identifying a cell line with an intact SMARCB1 allele, that enabled an in-depth look into the SMARCB1 allele given the unlimited supply of cells. HSES was assessed for point mutations, translocations, homozygous deletions, and small copy number changes at the SMARCB1 locus. DNA methylation was ruled out as a mechanism 109 of inactivation and furthermore, none of the suggested microRNAs in the literature (miR-206, miR193a-5p, miR381-3p, and miR671-5p) were found to be responsible. The majority of the gene silencing, however, occurs at a transcriptional level. Our preliminary results using Actinomycin D suggest that mRNA stability is not significantly compromised. Despite our attempts, we were not able to identify the cause of the silencing of the remaining SMARCB1 allele in HSES. More comprehensive epigenetic screening assessments for a comprehensive panel of histone marks, as well as cloning of the promoter of SMARCB1 from HSES into luciferase reporter systems and assessing for transcriptional activity by overexpression of common and specific transcription factors could be the next steps. The latter approach would allow the examination of specific trans regulators that might be abnormal in HSES cells and possibly a certain portion of clinical cases of epithelioid sarcoma. 5.5 Epigenetic modifiers as rational therapies for epithelioid sarcoma By using the inducible SMARCB1 cell lines, and considering previous knowledge in the rhabdoid tumor field, we were able to identify several genes that were regulated by SMARCB1 in epithelioid sarcoma. Some of these genes were also regulated by the polycomb complex, and inhibition of EZH2 led to a similar effect on their transcription as that of SMARCB1 restoration. EZH1/2 inhibition also showed specific efficacy in epithelioid sarcoma and malignant rhabdoid tumor lines compared to SMARCB1-wild type HEK 293t cells. Therefore, EZH1/2 inhibitors are a potential therapeutic option and this is important to follow up with in vivo and in further functional studies. On the other hand, we also examined another epigenetic modifying compound, the HDAC inhibitor romidepsin, which showed efficacy in epithelioid sarcoma lines as well. Mechanistically, 110 we found EGR1 and subsequent PTEN upregulation in HSES, but the evidence in the other two epithelioid sarcoma cell lines was not as significant (Appendix F). Recent studies have also found efficacy of EZH2 inhibitors in ARID1A mutant tumors(Bitler et al., 2015). ARID1A is a non-core member of the BAF SWI/SNF complex (Figure 1-2). They identified PIK3IP1, a negative regulator of the PI3K-AKT signaling, as a common target of ARID1A and EZH2 whose expression would be increased with either ARID1A re-expression or EZH2 inhibition. This is similar to what we observed with PTPRK and CDKN1A expression upon EZH2 inhibition or SMARCB1 restoration in G401 and Epi544 respectively. These findings highlight the therapeutic value of EZH2 inhibitors in a diverse set of SWI/SNF aberrant cancers. 5.6 Residual SWI/SNF complex in epithelioid sarcoma Despite the loss of SMARCB1 in epithelioid sarcoma, the SWI/SNF complex still forms and most members associate with one another. This residual complex has also been observed in other SWI/SNF aberrant tumors and could be exploited for therapy (Helming, Wang, & Roberts, 2014). Epithelioid sarcoma line VAESBJ seems to have increased dependence on the residual complex as the knockdown of SMARCA4 leads reduced proliferation; whereas, the same event causes increased proliferation in HEK293t cells. We believe this is a very promising finding. Synthetic lethality is a key concept in targeted therapeutics and epithelioid sarcoma seems to have increased susceptibility to inhibition of SMARCA4 because of its complementary function to the very mutation that drives this tumor – SMARCB1 loss. SMARCA4 inhibitors are being developed and some inhibitors of its bromodomain already exist. Evaluation of these compounds in vitro and in vivo would be the next obvious steps. Undermining the remnant SWI/SNF complex in 111 the epithelioid sarcoma and rhabdoid tumors is in theory one of the most direct ways of specifically targeting the tumor cells in these cancers. 5.7 Future directions Noting the different SMARCB1 regulation of target genes in G401 vs. Epi544 (Figure 4-2), it would be interesting to compare the differences in chromatin architecture caused by SMARCB1 re-expression in these cell lines. We have established collaborations that have started to look at nucleosome positioning pre- and post SMARCB1 induction via assay for transposase-accessible chromatin using sequencing (ATAC-seq). We hope to combine these results with RNA-seq analyses pre- and post SMARCB1 induction to gain insight into context specific role of SMARCB1 in chromatin remodeling and its relation to gene expression. Another interesting observation was the relative resistance of VAESBJ to the EZH2 specific inhibitor GSK126. This cell line was at the same time susceptible to EZH1/2 inhibitor UNC1999. As EZH2 inhibitors are becoming more recognized and available as cancer therapy drugs, the problem of resistance can be anticipated. Understanding the resistance of VAESBJ to GSK126, despite the fact that the aforementioned compound makes biological sense in inhibiting the growth of SMARCB1 negative tumors, will be of great value. RNA-seq anaylses with/without EZH2 and EZH1/2 inhibition in epithelioid sarcoma and malignant rhabdoid tumor lines, including VAESBJ, combined with the data on SMARCB1 restoration could shed some light on the susceptibility and sensitivity of SMARCB1 negative tumors to EZH1/2 inhibition. Due to the very nature of epithelioid sarcoma with close intimacy of the tumor cells with infiltrative lymphocytes and/or surrounding structures such as the cells of the 112 dermis, tenocytes, or fibroblasts, it is very difficult to find high tumor content in banked samples. The small size of epithelioid sarcoma primary tumors and slow rate of growth are additional challenges that make the acquisition of large quantities of nucleic acids from tumor cell-rich tissues difficult. Furthermore, at the time of this study, high quality nucleic acids of the appropriate amount could only be extracted via crushing of relatively large epithelioid sarcoma samples in liquid nitrogen. However, recent developments in sequencing technology mean that poorer quality nucleic acids from formalin-fixed paraffin embedded material can now be used for deep sequencing (Behjati et al., 2013). Hence sequencing macro-dissected (and therefore more pure) epithelioid sarcoma samples, and accessibility to a larger number of cases would be among advantages that future sequencing studies of epithelioid sarcoma could have. Larger studies will be necessary to confirm the observations of our study, particularly in expression differences between proximal and distal epithelioid sarcoma which will require a larger number of WTSS results than used in this study before any solid conclusions could be drawn. Additionally, studies of intratumoral heterogeneity would be of great interest in epithelioid sarcoma given the diversity of cellular morphologies observed and the complexity of the genomic landscape. 113 References Adessi, C., Matton, G., Ayala, G., Turcatti, G., Mermod, J. J., Mayer, P., & Kawashima, E. (2000). Solid phase DNA amplification: characterisation of primer attachment and amplification mechanisms. Nucleic Acids Res, 28(20), E87. Agaimy, A. (2014). The expanding family of SMARCB1(INI1)-deficient neoplasia: implications of phenotypic, biological, and molecular heterogeneity. Adv Anat Pathol, 21(6), 394-410. doi: 10.1097/PAP.0000000000000038 Agaimy, A., Koch, M., Lell, M., Semrau, S., Dudek, W., Wachter, D. L., . . . Hartmann, A. (2014). SMARCB1(INI1)-deficient sinonasal basaloid carcinoma: a novel member of the expanding family of SMARCB1-deficient neoplasms. Am J Surg Pathol, 38(9), 1274-1281. doi: 10.1097/PAS.0000000000000236 Amarasinghe, K. C., Li, J., Hunter, S. M., Ryland, G. L., Cowin, P. A., Campbell, I. G., & Halgamuge, S. K. (2014). Inferring copy number and genotype in tumour exome data. BMC Genomics, 15, 732. doi: 10.1186/1471-2164-15-732 Bashashati, A., Ha, G., Tone, A., Ding, J., Prentice, L. M., Roth, A., . . . Shah, S. P. (2013). Distinct evolutionary trajectories of primary high-grade serous ovarian cancers revealed through spatial mutational profiling. J Pathol, 231(1), 21-34. doi: 10.1002/path.4230 Beckwith, J. B., & Palmer, N. F. (1978). Histopathology and prognosis of Wilms tumors: results from the First National Wilms' Tumor Study. Cancer, 41(5), 1937-1948. Behjati, S., Tarpey, P. S., Presneau, N., Scheipl, S., Pillay, N., Van Loo, P., . . . Flanagan, A. M. (2013). Distinct H3F3A and H3F3B driver mutations define chondroblastoma and giant cell tumor of bone. Nat Genet, 45(12), 1479-1482. doi: 10.1038/ng.2814 Bernardini, L., Alesi, V., Loddo, S., Novelli, A., Bottillo, I., Battaglia, A., . . . Dallapiccola, B. (2010). High-resolution SNP arrays in mental retardation diagnostics: how much do we gain? Eur J Hum Genet, 18(2), 178-185. doi: 10.1038/ejhg.2009.154 Biegel, J. A., Zhou, J. Y., Rorke, L. B., Stenstrom, C., Wainwright, L. M., & Fogelgren, B. (1999). Germ-line and acquired mutations of INI1 in atypical teratoid and rhabdoid tumors. Cancer Res, 59(1), 74-79. Bitler, B. G., Aird, K. M., Garipov, A., Li, H., Amatangelo, M., Kossenkov, A. V., . . . Zhang, R. (2015). Synthetic lethality by targeting EZH2 methyltransferase activity in ARID1A-mutated cancers. Nat Med, 21(3), 231-238. doi: 10.1038/nm.3799 Bloustein, P. A., Silverberg, S. G., & Waddell, W. R. (1976). Epithelioid sarcoma: case report with ultrastructural review, histogenetic discussion, and chemotherapeutic data. Cancer, 38(6), 2390-2400. 114 Bonnin, J. M., Rubinstein, L. J., Palmer, N. F., & Beckwith, J. B. (1984). The association of embryonal tumors originating in the kidney and in the brain. A report of seven cases. Cancer, 54(10), 2137-2146. Boyd, C., Smith, M. J., Kluwe, L., Balogh, A., Maccollin, M., & Plotkin, S. R. (2008). Alterations in the SMARCB1 (INI1) tumor suppressor gene in familial schwannomatosis. Clin Genet, 74(4), 358-366. doi: 10.1111/j.1399-0004.2008.01060.x Brandes, J. C., Carraway, H., & Herman, J. G. (2007). Optimal primer design using the novel primer design program: MSPprimer provides accurate methylation analysis of the ATM promoter. Oncogene, 26(42), 6229-6237. doi: 10.1038/sj.onc.1210433 Brenca, M., Rossi, S., Lorenzetto, E., Piccinin, E., Piccinin, S., Rossi, F. M., . . . Modena, P. (2013). SMARCB1/INI1 genetic inactivation is responsible for tumorigenic properties of epithelioid sarcoma cell line VAESBJ. Mol Cancer Ther, 12(6), 1060-1072. doi: 10.1158/1535-7163.MCT-13-0005 Brennan, M. F., Antonescu, C. R., Moraco, N., & Singer, S. (2014). Lessons learned from the study of 10,000 patients with soft tissue sarcoma. Ann Surg, 260(3), 416-421; discussion 421-412. doi: 10.1097/SLA.0000000000000869 Cairns, B. R., Kim, Y. J., Sayre, M. H., Laurent, B. C., & Kornberg, R. D. (1994). A multisubunit complex containing the SWI1/ADR6, SWI2/SNF2, SWI3, SNF5, and SNF6 gene products isolated from yeast. Proc Natl Acad Sci U S A, 91(5), 1950-1954. Calderaro, J., Moroch, J., Pierron, G., Pedeutour, F., Grison, C., Maille, P., . . . Allory, Y. (2012). SMARCB1/INI1 inactivation in renal medullary carcinoma. Histopathology, 61(3), 428-435. doi: 10.1111/j.1365-2559.2012.04228.x Chase, D. R. (1990). Rhabdoid versus epithelioid sarcoma. Am J Surg Pathol, 14(8), 792-794. Chase, D. R., & Enzinger, F. M. (1985). Epithelioid sarcoma. Diagnosis, prognostic indicators, and treatment. Am J Surg Pathol, 9(4), 241-263. Chbani, L., Guillou, L., Terrier, P., Decouvelaere, A. V., Gregoire, F., Terrier-Lacombe, M. J., . . . Coindre, J. M. (2009). Epithelioid sarcoma: a clinicopathologic and immunohistochemical analysis of 106 cases from the French sarcoma group. Am J Clin Pathol, 131(2), 222-227. doi: 10.1309/AJCPU98ABIPVJAIV Cheng, S. W., Davies, K. P., Yung, E., Beltran, R. J., Yu, J., & Kalpana, G. V. (1999). c-MYC interacts with INI1/hSNF5 and requires the SWI/SNF complex for transactivation function. Nat Genet, 22(1), 102-105. doi: 10.1038/8811 Chi, T. H., Wan, M., Zhao, K., Taniuchi, I., Chen, L., Littman, D. R., & Crabtree, G. R. (2002). Reciprocal regulation of CD4/CD8 expression by SWI/SNF-like BAF complexes. Nature, 418(6894), 195-199. doi: 10.1038/nature00876 de la Serna, I. L., Carlson, K. A., & Imbalzano, A. N. (2001). Mammalian SWI/SNF complexes promote MyoD-mediated muscle differentiation. Nat Genet, 27(2), 187-190. doi: 10.1038/84826 115 Dey, B. K., Gagan, J., & Dutta, A. (2011). miR-206 and -486 induce myoblast differentiation by downregulating Pax7. Mol Cell Biol, 31(1), 203-214. doi: 10.1128/MCB.01009-10 Ding, J., Bashashati, A., Roth, A., Oloumi, A., Tse, K., Zeng, T., . . . Shah, S. P. (2012). Feature-based classifiers for somatic mutation detection in tumour-normal paired sequencing data. Bioinformatics, 28(2), 167-175. doi: 10.1093/bioinformatics/btr629 Dressman, D., Yan, H., Traverso, G., Kinzler, K. W., & Vogelstein, B. (2003). Transforming single DNA molecules into fluorescent magnetic particles for detection and enumeration of genetic variations. Proc Natl Acad Sci U S A, 100(15), 8817-8822. doi: 10.1073/pnas.1133470100 Dykhuizen, E. C., Hargreaves, D. C., Miller, E. L., Cui, K., Korshunov, A., Kool, M., . . . Crabtree, G. R. (2013). BAF complexes facilitate decatenation of DNA by topoisomerase IIalpha. Nature, 497(7451), 624-627. doi: 10.1038/nature12146 Elfring, L. K., Deuring, R., McCallum, C. M., Peterson, C. L., & Tamkun, J. W. (1994). Identification and characterization of Drosophila relatives of the yeast transcriptional activator SNF2/SWI2. Mol Cell Biol, 14(4), 2225-2234. Emori, M., Tsukahara, T., Murase, M., Kano, M., Murata, K., Takahashi, A., . . . Sato, N. (2013). High expression of CD109 antigen regulates the phenotype of cancer stem-like cells/cancer-initiating cells in the novel epithelioid sarcoma cell line ESX and is related to poor prognosis of soft tissue sarcoma. PLoS One, 8(12), e84187. doi: 10.1371/journal.pone.0084187 Enzinger, F. M. (1970). Epitheloid sarcoma. A sarcoma simulating a granuloma or a carcinoma. Cancer, 26(5), 1029-1041. Fedurco, M., Romieu, A., Williams, S., Lawrence, I., & Turcatti, G. (2006). BTA, a novel reagent for DNA attachment on glass and efficient generation of solid-phase amplified DNA colonies. Nucleic Acids Res, 34(3), e22. doi: 10.1093/nar/gnj023 Filippakopoulos, P., Picaud, S., Mangos, M., Keates, T., Lambert, J. P., Barsyte-Lovejoy, D., . . . Knapp, S. (2012). Histone recognition and large-scale structural analysis of the human bromodomain family. Cell, 149(1), 214-231. doi: 10.1016/j.cell.2012.02.013 Filippakopoulos, P., Qi, J., Picaud, S., Shen, Y., Smith, W. B., Fedorov, O., . . . Bradner, J. E. (2010). Selective inhibition of BET bromodomains. Nature, 468(7327), 1067-1073. doi: 10.1038/nature09504 Fisher, C. (1988). Epithelioid sarcoma: the spectrum of ultrastructural differentiation in seven immunohistochemically defined cases. Hum Pathol, 19(3), 265-275. Fisher, E. R., & Horvat, B. (1972). The fibrocytic deprivation of the so-called epithelioid sarcoma. Cancer, 30(4), 1074-1081. Flicek, P., Ahmed, I., Amode, M. R., Barrell, D., Beal, K., Brent, S., . . . Searle, S. M. (2013). Ensembl 2013. Nucleic Acids Res, 41(Database issue), D48-55. doi: 10.1093/nar/gks1236 116 gks1236 [pii] Flowers, S., Nagl, N. G., Jr., Beck, G. R., Jr., & Moran, E. (2009). Antagonistic roles for BRM and BRG1 SWI/SNF complexes in differentiation. J Biol Chem, 284(15), 10067-10075. doi: 10.1074/jbc.M808782200 Flucke, U., Slootweg, P. J., Mentzel, T., Pauwels, P., & Hulsebos, T. J. (2009). Re: Infrequent SMARCB1/INI1 gene alteration in epithelioid sarcoma: a useful tool in distinguishing epithelioid sarcoma from malignant rhabdoid tumor: Direct evidence of mutational inactivation of SMARCB1/INI1 in epithelioid sarcoma. Hum Pathol, 40(9), 1361-1362; author reply 1362-1364. doi: 10.1016/j.humpath.2009.04.018 Francis, N. J., Saurin, A. J., Shao, Z., & Kingston, R. E. (2001). Reconstitution of a functional core polycomb repressive complex. Mol Cell, 8(3), 545-556. Gallup, D. G., Abell, M. R., & Morley, G. W. (1976). Epithelioid sarcoma of the vulva. Obstet Gynecol, 48(1 Suppl), 14S-17S. Goya, R., Sun, M. G., Morin, R. D., Leung, G., Ha, G., Wiegand, K. C., . . . Shah, S. P. (2010). SNVMix: predicting single nucleotide variants from next-generation sequencing of tumors. Bioinformatics, 26(6), 730-736. doi: 10.1093/bioinformatics/btq040 btq040 [pii] Greenlee, R. T., Hill-Harmon, M. B., Murray, T., & Thun, M. (2001). Cancer statistics, 2001. CA Cancer J Clin, 51(1), 15-36. Griffin, C. T., Curtis, C. D., Davis, R. B., Muthukumar, V., & Magnuson, T. (2011). The chromatin-remodeling enzyme BRG1 modulates vascular Wnt signaling at two levels. Proc Natl Acad Sci U S A, 108(6), 2282-2287. doi: 10.1073/pnas.1013751108 Guillou, Donald R. (1999). Tumors of Uncertain Histogenesis. Rev Esp Pathol, 273(32), 273-274. Guillou, L., Wadden, C., Coindre, J. M., Krausz, T., & Fletcher, C. D. (1997). "Proximal-type" epithelioid sarcoma, a distinctive aggressive neoplasm showing rhabdoid features. Clinicopathologic, immunohistochemical, and ultrastructural study of a series. Am J Surg Pathol, 21(2), 130-146. Gustafson, P., & Arner, M. (1999). Soft tissue sarcoma of the upper extremity: descriptive data and outcome in a population-based series of 108 adult patients. J Hand Surg Am, 24(4), 668-674. Ha, G., Roth, A., Khattra, J., Ho, J., Yap, D., Prentice, L. M., . . . Shah, S. P. (2014). TITAN: inference of copy number architectures in clonal cell populations from tumor whole-genome sequence data. Genome Res, 24(11), 1881-1893. doi: 10.1101/gr.180281.114 Ha, G., Roth, A., Lai, D., Bashashati, A., Ding, J., Goya, R., . . . Shah, S. P. (2012). Integrative analysis of genome-wide loss of heterozygosity and monoallelic expression at nucleotide resolution reveals disrupted pathways in triple-negative breast cancer. Genome Res, 22(10), 1995-2007. doi: 10.1101/gr.137570.112 Haas, J. E., Palmer, N. F., Weinberg, A. G., & Beckwith, J. B. (1981). Ultrastructure of malignant rhabdoid tumor of the kidney. A distinctive renal tumor of children. Hum Pathol, 12(7), 646-657. 117 Hall, D. J., Grimes, M. M., & Goplerud, D. R. (1980). Epithelioid sarcoma of the vulva. Gynecol Oncol, 9(2), 237-246. Halling, A. C., Wollan, P. C., Pritchard, D. J., Vlasak, R., & Nascimento, A. G. (1996). Epithelioid sarcoma: a clinicopathologic review of 55 cases. Mayo Clin Proc, 71(7), 636-642. doi: 10.1016/S0025-6196(11)63000-0 Hasegawa, T., Matsuno, Y., Shimoda, T., Umeda, T., Yokoyama, R., & Hirohashi, S. (2001). Proximal-type epithelioid sarcoma: a clinicopathologic study of 20 cases. Mod Pathol, 14(7), 655-663. doi: 10.1038/modpathol.3880368 Hasselblatt, M., Gesk, S., Oyen, F., Rossi, S., Viscardi, E., Giangaspero, F., . . . Paulus, W. (2011). Nonsense mutation and inactivation of SMARCA4 (BRG1) in an atypical teratoid/rhabdoid tumor showing retained SMARCB1 (INI1) expression. Am J Surg Pathol, 35(6), 933-935. doi: 10.1097/PAS.0b013e3182196a39 Hasselblatt, M., Isken, S., Linge, A., Eikmeier, K., Jeibmann, A., Oyen, F., . . . Paulus, W. (2013). High-resolution genomic analysis suggests the absence of recurrent genomic alterations other than SMARCB1 aberrations in atypical teratoid/rhabdoid tumors. Genes Chromosomes Cancer, 52(2), 185-190. doi: 10.1002/gcc.22018 Hasselblatt, M., Oyen, F., Gesk, S., Kordes, U., Wrede, B., Bergmann, M., . . . Paulus, W. (2009). Cribriform neuroepithelial tumor (CRINET): a nonrhabdoid ventricular tumor with INI1 loss and relatively favorable prognosis. J Neuropathol Exp Neurol, 68(12), 1249-1255. doi: 10.1097/NEN.0b013e3181c06a51 Helming, K. C., Wang, X., & Roberts, C. W. (2014). Vulnerabilities of mutant SWI/SNF complexes in cancer. Cancer Cell, 26(3), 309-317. doi: 10.1016/j.ccr.2014.07.018 Helming, K. C., Wang, X., Wilson, B. G., Vazquez, F., Haswell, J. R., Manchester, H. E., . . . Roberts, C. W. (2014). ARID1B is a specific vulnerability in ARID1A-mutant cancers. Nat Med, 20(3), 251-254. doi: 10.1038/nm.3480 Helson, C., Melamed, M., Braverman, S., Traganos, F., Preti, R., & Helson, L. (1995). Va-es-bj - an epithelioid sarcoma cell-line. Int J Oncol, 7(1), 51-56. Ho, L., Ronan, J. L., Wu, J., Staahl, B. T., Chen, L., Kuo, A., . . . Crabtree, G. R. (2009). An embryonic stem cell chromatin remodeling complex, esBAF, is essential for embryonic stem cell self-renewal and pluripotency. Proc Natl Acad Sci U S A, 106(13), 5181-5186. doi: 10.1073/pnas.0812889106 Hoffman, G. R., Rahal, R., Buxton, F., Xiang, K., McAllister, G., Frias, E., . . . Jagani, Z. (2014). Functional epigenetics approach identifies BRM/SMARCA2 as a critical synthetic lethal target in BRG1-deficient cancers. Proc Natl Acad Sci U S A, 111(8), 3128-3133. doi: 10.1073/pnas.1316793111 Hollmann, T. J., & Hornick, J. L. (2011). INI1-deficient tumors: diagnostic features and molecular genetics. Am J Surg Pathol, 35(10), e47-63. doi: 10.1097/PAS.0b013e31822b325b 118 Hornick, J. L., Dal Cin, P., & Fletcher, C. D. (2009). Loss of INI1 expression is characteristic of both conventional and proximal-type epithelioid sarcoma. Am J Surg Pathol, 33(4), 542-550. doi: 10.1097/PAS.0b013e3181882c54 Howlader, N., Noone, A. M., Yu, M., & Cronin, K. A. (2012). Use of imputed population-based cancer registry data as a method of accounting for missing information: application to estrogen receptor status for breast cancer. Am J Epidemiol, 176(4), 347-356. doi: 10.1093/aje/kwr512 Ibrahim, G. M., Huang, A., Halliday, W., Dirks, P. B., Malkin, D., Baskin, B., . . . Hawkins, C. (2011). Cribriform neuroepithelial tumour: novel clinicopathological, ultrastructural and cytogenetic findings. Acta Neuropathol, 122(4), 511-514. doi: 10.1007/s00401-011-0872-5 Imura, Y., Yasui, H., Outani, H., Wakamatsu, T., Hamada, K., Nakai, T., . . . Naka, N. (2014). Combined targeting of mTOR and c-MET signaling pathways for effective management of epithelioid sarcoma. Mol Cancer, 13, 185. doi: 10.1186/1476-4598-13-185 Iossifides, I., Ayala, A. G., & Johnson, D. E. (1979). Epithelioid sarcoma of penis. Urology, 14(2), 190-191. Isakoff, M. S., Sansam, C. G., Tamayo, P., Subramanian, A., Evans, J. A., Fillmore, C. M., . . . Roberts, C. W. (2005). Inactivation of the Snf5 tumor suppressor stimulates cell cycle progression and cooperates with p53 loss in oncogenic transformation. Proc Natl Acad Sci U S A, 102(49), 17745-17750. doi: 10.1073/pnas.0509014102 Ishida, T., Oka, T., Matsushita, H., & Machinami, R. (1992). Epithelioid sarcoma: an electron-microscopic, immunohistochemical and DNA flow cytometric analysis. Virchows Arch A Pathol Anat Histopathol, 421(5), 401-408. Izumi, T., Oda, Y., Hasegawa, T., Nakanishi, Y., Iwasaki, H., Sonobe, H., . . . Tsuneyoshi, M. (2006). Prognostic significance of dysadherin expression in epithelioid sarcoma and its diagnostic utility in distinguishing epithelioid sarcoma from malignant rhabdoid tumor. Mod Pathol, 19(6), 820-831. doi: 10.1038/modpathol.3800599 Jagani, Z., Mora-Blanco, E. L., Sansam, C. G., McKenna, E. S., Wilson, B., Chen, D., . . . Dorsch, M. (2010). Loss of the tumor suppressor Snf5 leads to aberrant activation of the Hedgehog-Gli pathway. Nat Med, 16(12), 1429-1433. doi: 10.1038/nm.2251 Jamshidi, F., Nielsen, T. O., & Huntsman, D. G. (2015). Cancer genomics: why rare is valuable. J Mol Med (Berl). doi: 10.1007/s00109-015-1260-8 Jamshidi, F., Pleasance, E., Li, Y., Shen, Y., Kasaian, K., Corbett, R., . . . Yip, S. (2014). Diagnostic value of next-generation sequencing in an unusual sphenoid tumor. Oncologist, 19(6), 623-630. doi: 10.1634/theoncologist.2013-0390 Jawad, M. U., Extein, J., Min, E. S., & Scully, S. P. (2009). Prognostic factors for survival in patients with epithelioid sarcoma: 441 cases from the SEER database. Clin Orthop Relat Res, 467(11), 2939-2948. doi: 10.1007/s11999-009-0749-2 119 Jeibmann, A., Eikmeier, K., Linge, A., Kool, M., Koos, B., Schulz, J., . . . Hasselblatt, M. (2014). Identification of genes involved in the biology of atypical teratoid/rhabdoid tumours using Drosophila melanogaster. Nat Commun, 5, 4005. doi: 10.1038/ncomms5005 Jelinic, P., Mueller, J. J., Olvera, N., Dao, F., Scott, S. N., Shah, R., . . . Levine, D. A. (2014). Recurrent SMARCA4 mutations in small cell carcinoma of the ovary. Nat Genet, 46(5), 424-426. doi: 10.1038/ng.2922 Jeong, J. Y., Suh, Y. L., & Hong, S. W. (2014). Atypical teratoid/rhabdoid tumor arising in pleomorphic xanthoastrocytoma: a case report. Neuropathology, 34(4), 398-405. Kadoch, C., & Crabtree, G. R. (2013). Reversible disruption of mSWI/SNF (BAF) complexes by the SS18-SSX oncogenic fusion in synovial sarcoma. Cell, 153(1), 71-85. doi: 10.1016/j.cell.2013.02.036 Kadoch, C., Hargreaves, D. C., Hodges, C., Elias, L., Ho, L., Ranish, J., & Crabtree, G. R. (2013). Proteomic and bioinformatic analysis of mammalian SWI/SNF complexes identifies extensive roles in human malignancy. Nat Genet, 45(6), 592-601. doi: 10.1038/ng.2628 Kahali, B., Yu, J., Marquez, S. B., Thompson, K. W., Liang, S. Y., Lu, L., & Reisman, D. (2014). The silencing of the SWI/SNF subunit and anticancer gene BRM in Rhabdoid tumors. Oncotarget, 5(10), 3316-3332. Kalpana, G. V., Marmon, S., Wang, W., Crabtree, G. R., & Goff, S. P. (1994). Binding and stimulation of HIV-1 integrase by a human homolog of yeast transcription factor SNF5. Science, 266(5193), 2002-2006. Kato, H., Hatori, M., Kokubun, S., Watanabe, M., Smith, R. A., Hotta, T., . . . Aiba, S. (2004). CA125 expression in epithelioid sarcoma. Jpn J Clin Oncol, 34(3), 149-154. Kato, H., Ohta, S., Koshida, S., Narita, T., Taga, T., Takeuchi, Y., & Sugita, K. (2003). Expression of pericyte, mesangium and muscle markers in malignant rhabdoid tumor cell lines: differentiation-induction using 5-azacytidine. Cancer Sci, 94(12), 1059-1065. Kennison, J. A., & Tamkun, J. W. (1988). Dosage-dependent modifiers of polycomb and antennapedia mutations in Drosophila. Proc Natl Acad Sci U S A, 85(21), 8136-8140. Khavari, P. A., Peterson, C. L., Tamkun, J. W., Mendel, D. B., & Crabtree, G. R. (1993). BRG1 contains a conserved domain of the SWI2/SNF2 family necessary for normal mitotic growth and transcription. Nature, 366(6451), 170-174. doi: 10.1038/366170a0 Kia, S. K., Gorski, M. M., Giannakopoulos, S., & Verrijzer, C. P. (2008). SWI/SNF mediates polycomb eviction and epigenetic reprogramming of the INK4b-ARF-INK4a locus. Mol Cell Biol, 28(10), 3457-3464. doi: 10.1128/MCB.02019-07 Kim, K. H., & Roberts, C. W. (2014). Mechanisms by which SMARCB1 loss drives rhabdoid tumor growth. Cancer Genet, 207(9), 365-372. doi: 10.1016/j.cancergen.2014.04.004 Kleinschmidt-DeMasters, B. K., Birks, D. K., Aisner, D. L., Hankinson, T. C., & Rosenblum, M. K. (2011). Atypical teratoid/rhabdoid tumor arising in a 120 ganglioglioma: genetic characterization. Am J Surg Pathol, 35(12), 1894-1901. doi: 10.1097/PAS.0b013e3182382a3f Klijn, C., Durinck, S., Stawiski, E. W., Haverty, P. M., Jiang, Z., Liu, H., . . . Zhang, Z. (2014). A comprehensive transcriptional portrait of human cancer cell lines. Nat Biotechnol. doi: 10.1038/nbt.3080 Knutson, S. K., Warholic, N. M., Wigle, T. J., Klaus, C. R., Allain, C. J., Raimondi, A., . . . Keilhack, H. (2013). Durable tumor regression in genetically altered malignant rhabdoid tumors by inhibition of methyltransferase EZH2. Proc Natl Acad Sci U S A, 110(19), 7922-7927. doi: 10.1073/pnas.1303800110 Kohashi, K., Izumi, T., Oda, Y., Yamamoto, H., Tamiya, S., Taguchi, T., . . . Tsuneyoshi, M. (2009). Infrequent SMARCB1/INI1 gene alteration in epithelioid sarcoma: a useful tool in distinguishing epithelioid sarcoma from malignant rhabdoid tumor. Hum Pathol, 40(3), 349-355. doi: 10.1016/j.humpath.2008.08.007 Kohashi, K., Oda, Y., Yamamoto, H., Tamiya, S., Matono, H., Iwamoto, Y., . . . Tsuneyoshi, M. (2010). Reduced expression of SMARCB1/INI1 protein in synovial sarcoma. Mod Pathol, 23(7), 981-990. doi: 10.1038/modpathol.2010.71 Kohashi, K., Yamada, Y., Hotokebuchi, Y., Yamamoto, H., Taguchi, T., Iwamoto, Y., & Oda, Y. (2014). ERG and SALL4 expressions in SMARCB1/INI1-deficient tumors: a useful tool for distinguishing epithelioid sarcoma from malignant rhabdoid tumor. Hum Pathol. doi: 10.1016/j.humpath.2014.10.010 Kohashi, K., Yamada, Y., Hotokebuchi, Y., Yamamoto, H., Taguchi, T., Iwamoto, Y., & Oda, Y. (2015). ERG and SALL4 expressions in SMARCB1/INI1-deficient tumors: a useful tool for distinguishing epithelioid sarcoma from malignant rhabdoid tumor. Hum Pathol, 46(2), 225-230. doi: 10.1016/j.humpath.2014.10.010 Kohashi, K., Yamamoto, H., Kumagai, R., Yamada, Y., Hotokebuchi, Y., Taguchi, T., . . . Oda, Y. (2014). Differential microRNA expression profiles between malignant rhabdoid tumor and epithelioid sarcoma: miR193a-5p is suggested to downregulate SMARCB1 mRNA expression. Mod Pathol, 27(6), 832-839. doi: 10.1038/modpathol.2013.213 Kuhnen, C., Tolnay, E., Steinau, H. U., Voss, B., & Muller, K. M. (1998). Expression of c-Met receptor and hepatocyte growth factor/scatter factor in synovial sarcoma and epithelioid sarcoma. Virchows Arch, 432(4), 337-342. Kupryjanczyk, J., Dansonka-Mieszkowska, A., Moes-Sosnowska, J., Plisiecka-Halasa, J., Szafron, L., Podgorska, A., . . . Spiewankiewicz, B. (2013). Ovarian small cell carcinoma of hypercalcemic type - evidence of germline origin and SMARCA4 gene inactivation. a pilot study. Pol J Pathol, 64(4), 238-246. Lange, M., Kaynak, B., Forster, U. B., Tonjes, M., Fischer, J. J., Grimm, C., . . . Sperling, S. (2008). Regulation of muscle development by DPF3, a novel histone acetylation and methylation reader of the BAF chromatin 121 remodeling complex. Genes Dev, 22(17), 2370-2384. doi: 10.1101/gad.471408 Lawrence, M. S., Stojanov, P., Polak, P., Kryukov, G. V., Cibulskis, K., Sivachenko, A., . . . Getz, G. (2013). Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature, 499(7457), 214-218. doi: 10.1038/nature12213 Le Loarer, F., Zhang, L., Fletcher, C. D., Ribeiro, A., Singer, S., Italiano, A., . . . Antonescu, C. R. (2014). Consistent SMARCB1 homozygous deletions in epithelioid sarcoma and in a subset of myoepithelial carcinomas can be reliably detected by FISH in archival material. Genes Chromosomes Cancer, 53(6), 475-486. doi: 10.1002/gcc.22159 Lee, R. S., Stewart, C., Carter, S. L., Ambrogio, L., Cibulskis, K., Sougnez, C., . . . Roberts, C. W. (2012). A remarkably simple genome underlies highly malignant pediatric rhabdoid cancers. J Clin Invest, 122(8), 2983-2988. doi: 10.1172/JCI64400 Li, H., & Durbin, R. (2010). Fast and accurate long-read alignment with Burrows-Wheeler transform. Bioinformatics, 26(5), 589-595. doi: 10.1093/bioinformatics/btp698 Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., . . . Durbin, R. (2009). The Sequence Alignment/Map format and SAMtools. Bioinformatics, 25(16), 2078-2079. doi: 10.1093/bioinformatics/btp352 Li, L., Fan, X. S., Xia, Q. Y., Rao, Q., Liu, B., Yu, B., . . . Zhou, X. J. (2014). Concurrent loss of INI1, PBRM1, and BRM expression in epithelioid sarcoma: implications for the cocontributions of multiple SWI/SNF complex members to pathogenesis. Hum Pathol, 45(11), 2247-2254. doi: 10.1016/j.humpath.2014.06.027 Machinami, R., Kikuchi, F., & Matsushita, H. (1982). Epithelioid sarcoma. Enzyme histochemical and ultrastructural study. Virchows Arch A Pathol Anat Histol, 397(1), 109-120. McCarroll, S. A., Kuruvilla, F. G., Korn, J. M., Cawley, S., Nemesh, J., Wysoker, A., . . . Altshuler, D. (2008). Integrated detection and population-genetic analysis of SNPs and copy number variation. Nat Genet, 40(10), 1166-1174. doi: 10.1038/ng.238 McPherson, A., Hormozdiari, F., Zayed, A., Giuliany, R., Ha, G., Sun, M. G., . . . Shah, S. P. (2011). deFuse: an algorithm for gene fusion discovery in tumor RNA-Seq data. PLoS Comput Biol, 7(5), e1001138. doi: 10.1371/journal.pcbi.1001138 Miettinen, M., Fanburg-Smith, J. C., Virolainen, M., Shmookler, B. M., & Fetsch, J. F. (1999). Epithelioid sarcoma: an immunohistochemical analysis of 112 classical and variant cases and a discussion of the differential diagnosis. Hum Pathol, 30(8), 934-942. Miettinen, M., Wang, Z., McCue, P. A., Sarlomo-Rikala, M., Rys, J., Biernat, W., . . . Lee, Y. S. (2014). SALL4 expression in germ cell and non-germ cell tumors: a systematic immunohistochemical study of 3215 cases. Am J Surg Pathol, 38(3), 410-420. doi: 10.1097/PAS.0000000000000116 122 Mills, S. E., Fechner, R. E., Bruns, D. E., Bruns, M. E., & O'Hara, M. F. (1981). Intermediate filaments in eosinophilic cells of epithelioid sarcoma: a light-microscopic, ultrastructural, and electrophoretic study. Am J Surg Pathol, 5(2), 195-202. Modena, P., Lualdi, E., Facchinetti, F., Galli, L., Teixeira, M. R., Pilotti, S., & Sozzi, G. (2005). SMARCB1/INI1 tumor suppressor gene is frequently inactivated in epithelioid sarcomas. Cancer Res, 65(10), 4012-4019. doi: 10.1158/0008-5472.CAN-04-3050 Mora-Blanco, E. L., Mishina, Y., Tillman, E. J., Cho, Y. J., Thom, C. S., Pomeroy, S. L., . . . Roberts, C. W. (2014). Activation of beta-catenin/TCF targets following loss of the tumor suppressor SNF5. Oncogene, 33(7), 933-938. doi: 10.1038/onc.2013.37 Muchardt, C., & Yaniv, M. (1993). A human homologue of Saccharomyces cerevisiae SNF2/SWI2 and Drosophila brm genes potentiates transcriptional activation by the glucocorticoid receptor. EMBO J, 12(11), 4279-4290. Murray, P. M. (2004). Soft tissue sarcoma of the upper extremity. Hand Clin, 20(3), 325-333, vii. doi: 10.1016/j.hcl.2004.03.007 Nagl, N. G., Jr., Patsialou, A., Haines, D. S., Dallas, P. B., Beck, G. R., Jr., & Moran, E. (2005). The p270 (ARID1A/SMARCF1) subunit of mammalian SWI/SNF-related complexes is essential for normal cell cycle arrest. Cancer Res, 65(20), 9236-9244. doi: 10.1158/0008-5472.CAN-05-1225 Narlikar, G. J., Sundaramoorthy, R., & Owen-Hughes, T. (2013). Mechanisms and functions of ATP-dependent chromatin-remodeling enzymes. Cell, 154(3), 490-503. doi: 10.1016/j.cell.2013.07.011 Neigeborn, L., & Carlson, M. (1984). Genes affecting the regulation of SUC2 gene expression by glucose repression in Saccharomyces cerevisiae. Genetics, 108(4), 845-858. Oruetxebarria, I., Venturini, F., Kekarainen, T., Houweling, A., Zuijderduijn, L. M., Mohd-Sarip, A., . . . Verrijzer, C. P. (2004). P16INK4a is required for hSNF5 chromatin remodeler-induced cellular senescence in malignant rhabdoid tumor cells. J Biol Chem, 279(5), 3807-3816. doi: 10.1074/jbc.M309333200 Papp, G., Changchien, Y. C., Peterfia, B., Pecsenka, L., Krausz, T., Stricker, T. P., . . . Sapi, Z. (2013). SMARCB1 protein and mRNA loss is not caused by promoter and histone hypermethylation in epithelioid sarcoma. Mod Pathol, 26(3), 393-403. doi: 10.1038/modpathol.2012.190 Papp, G., Krausz, T., Stricker, T. P., Szendroi, M., & Sapi, Z. (2014). SMARCB1 expression in epithelioid sarcoma is regulated by miR-206, miR-381, and miR-671-5p on Both mRNA and protein levels. Genes Chromosomes Cancer, 53(2), 168-176. doi: 10.1002/gcc.22128 Parham, D. M., Weeks, D. A., & Beckwith, J. B. (1994). The clinicopathologic spectrum of putative extrarenal rhabdoid tumors. An analysis of 42 cases studied with immunohistochemistry or electron microscopy. Am J Surg Pathol, 18(10), 1010-1029. 123 Park, J. H., Park, E. J., Lee, H. S., Kim, S. J., Hur, S. K., Imbalzano, A. N., & Kwon, J. (2006). Mammalian SWI/SNF complexes facilitate DNA double-strand break repair by promoting gamma-H2AX induction. EMBO J, 25(17), 3986-3997. doi: 10.1038/sj.emboj.7601291 Perrone, T., Swanson, P. E., Twiggs, L., Ulbright, T. M., & Dehner, L. P. (1989). Malignant rhabdoid tumor of the vulva: is distinction from epithelioid sarcoma possible? A pathologic and immunohistochemical study. Am J Surg Pathol, 13(10), 848-858. Peterson, C. L., Dingwall, A., & Scott, M. P. (1994). Five SWI/SNF gene products are components of a large multisubunit complex required for transcriptional enhancement. Proc Natl Acad Sci U S A, 91(8), 2905-2908. Peterson, C. L., & Herskowitz, I. (1992). Characterization of the yeast SWI1, SWI2, and SWI3 genes, which encode a global activator of transcription. Cell, 68(3), 573-583. Ramos, P., Karnezis, A. N., Craig, D. W., Sekulic, A., Russell, M. L., Hendricks, W. P., . . . Trent, J. M. (2014). Small cell carcinoma of the ovary, hypercalcemic type, displays frequent inactivating germline and somatic mutations in SMARCA4. Nat Genet, 46(5), 427-429. doi: 10.1038/ng.2928 Rorke, L. B., Packer, R., & Biegel, J. (1995). Central nervous system atypical teratoid/rhabdoid tumors of infancy and childhood. J Neurooncol, 24(1), 21-28. Rorke, L. B., Packer, R. J., & Biegel, J. A. (1996). Central nervous system atypical teratoid/rhabdoid tumors of infancy and childhood: definition of an entity. J Neurosurg, 85(1), 56-65. doi: 10.3171/jns.1996.85.1.0056 Ross, H. M., Lewis, J. J., Woodruff, J. M., & Brennan, M. F. (1997). Epithelioid sarcoma: clinical behavior and prognostic factors of survival. Ann Surg Oncol, 4(6), 491-495. Sakharpe, A., Lahat, G., Gulamhusein, T., Liu, P., Bolshakov, S., Nguyen, T., . . . Lev, D. (2011). Epithelioid sarcoma and unclassified sarcoma with epithelioid features: clinicopathological variables, molecular markers, and a new experimental model. Oncologist, 16(4), 512-522. doi: 10.1634/theoncologist.2010-0174 Sanger, F., & Coulson, A. R. (1975). A rapid method for determining sequences in DNA by primed synthesis with DNA polymerase. J Mol Biol, 94(3), 441-448. Sanger, F., Nicklen, S., & Coulson, A. R. (1977). DNA sequencing with chain-terminating inhibitors. Proc Natl Acad Sci U S A, 74(12), 5463-5467. Saunders, C. T., Wong, W. S., Swamy, S., Becq, J., Murray, L. J., & Cheetham, R. K. (2012). Strelka: accurate somatic small-variant calling from sequenced tumor-normal sample pairs. Bioinformatics, 28(14), 1811-1817. doi: 10.1093/bioinformatics/bts271 Schneppenheim, R., Fruhwald, M. C., Gesk, S., Hasselblatt, M., Jeibmann, A., Kordes, U., . . . Siebert, R. (2010). Germline nonsense mutation and somatic inactivation of SMARCA4/BRG1 in a family with rhabdoid 124 tumor predisposition syndrome. Am J Hum Genet, 86(2), 279-284. doi: 10.1016/j.ajhg.2010.01.013 Sevenet, N., Sheridan, E., Amram, D., Schneider, P., Handgretinger, R., & Delattre, O. (1999). Constitutional mutations of the hSNF5/INI1 gene predispose to a variety of cancers. Am J Hum Genet, 65(5), 1342-1348. doi: 10.1086/302639 Shah, S. P., Morin, R. D., Khattra, J., Prentice, L., Pugh, T., Burleigh, A., . . . Aparicio, S. (2009). Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature, 461(7265), 809-813. doi: 10.1038/nature08489 Shendure, J., & Ji, H. (2008). Next-generation DNA sequencing. Nat Biotechnol, 26(10), 1135-1145. doi: 10.1038/nbt1486 Sonobe, H., Ohtsuki, Y., Sugimoto, T., & Shimizu, K. (1997). Involvement of 8q, 22q, and monosomy 21 in an epithelioid sarcoma. Cancer Genet Cytogenet, 96(2), 178-180. Soule, E. H., & Enriquez, P. (1972). Atypical fibrous histiocytoma, malignant fibrous histiocytoma, malignant histiocytoma, and epithelioid sarcoma. A comparative study of 65 tumors. Cancer, 30(1), 128-143. Stern, M., Jensen, R., & Herskowitz, I. (1984). Five SWI genes are required for expression of the HO gene in yeast. J Mol Biol, 178(4), 853-868. Stockman, D. L., Hornick, J. L., Deavers, M. T., Lev, D. C., Lazar, A. J., & Wang, W. L. (2014). ERG and FLI1 protein expression in epithelioid sarcoma. Mod Pathol, 27(4), 496-501. doi: 10.1038/modpathol.2013.161 Sudarsanam, P., Iyer, V. R., Brown, P. O., & Winston, F. (2000). Whole-genome expression analysis of snf/swi mutants of Saccharomyces cerevisiae. Proc Natl Acad Sci U S A, 97(7), 3364-3369. doi: 10.1073/pnas.050407197 Sullivan, L. M., Folpe, A. L., Pawel, B. R., Judkins, A. R., & Biegel, J. A. (2013). Epithelioid sarcoma is associated with a high percentage of SMARCB1 deletions. Mod Pathol, 26(3), 385-392. doi: 10.1038/modpathol.2012.175 Tamkun, J. W., Deuring, R., Scott, M. P., Kissinger, M., Pattatucci, A. M., Kaufman, T. C., & Kennison, J. A. (1992). brahma: a regulator of Drosophila homeotic genes structurally related to the yeast transcriptional activator SNF2/SWI2. Cell, 68(3), 561-572. Trapnell, C., Roberts, A., Goff, L., Pertea, G., Kim, D., Kelley, D. R., . . . Pachter, L. (2012). Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. Nat Protoc, 7(3), 562-578. doi: 10.1038/nprot.2012.016 Trapnell, C., Williams, B. A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M. J., . . . Pachter, L. (2010). Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol, 28(5), 511-515. doi: 10.1038/nbt.1621 Trouche, D., Le Chalony, C., Muchardt, C., Yaniv, M., & Kouzarides, T. (1997). RB and hbrm cooperate to repress the activation functions of E2F1. Proc Natl Acad Sci U S A, 94(21), 11268-11273. 125 Tsai, C. Y., Wong, T. T., Lee, Y. H., Chao, M. E., Lin, S. C., Liu, D. J., . . . Ho, D. M. (2012). Intact INI1 gene region with paradoxical loss of protein expression in AT/RT: implications for a possible novel mechanism associated with absence of INI1 protein immunoreactivity. Am J Surg Pathol, 36(1), 128-133. doi: 10.1097/PAS.0b013e3182348cc4 Tsikitis, M., Zhang, Z., Edelman, W., Zagzag, D., & Kalpana, G. V. (2005). Genetic ablation of Cyclin D1 abrogates genesis of rhabdoid tumors resulting from Ini1 loss. Proc Natl Acad Sci U S A, 102(34), 12129-12134. doi: 10.1073/pnas.0505300102 Tsurusaki, Y., Okamoto, N., Ohashi, H., Kosho, T., Imai, Y., Hibi-Ko, Y., . . . Matsumoto, N. (2012). Mutations affecting components of the SWI/SNF complex cause Coffin-Siris syndrome. Nat Genet, 44(4), 376-378. doi: 10.1038/ng.2219 Van Houdt, J. K., Nowakowska, B. A., Sousa, S. B., van Schaik, B. D., Seuntjens, E., Avonce, N., . . . Vermeesch, J. R. (2012). Heterozygous missense mutations in SMARCA2 cause Nicolaides-Baraitser syndrome. Nat Genet, 44(4), 445-449, S441. doi: 10.1038/ng.1105 Venneti, S., Le, P., Martinez, D., Eaton, K. W., Shyam, N., Jordan-Sciutto, K. L., . . . Judkins, A. R. (2011). p16INK4A and p14ARF tumor suppressor pathways are deregulated in malignant rhabdoid tumors. J Neuropathol Exp Neurol, 70(7), 596-609. doi: 10.1097/NEN.0b013e31822146ca Versteege, I., Sevenet, N., Lange, J., Rousseau-Merck, M. F., Ambros, P., Handgretinger, R., . . . Delattre, O. (1998). Truncating mutations of hSNF5/INI1 in aggressive paediatric cancer. Nature, 394(6689), 203-206. doi: 10.1038/28212 Wang, W., Cote, J., Xue, Y., Zhou, S., Khavari, P. A., Biggar, S. R., . . . Crabtree, G. R. (1996). Purification and biochemical heterogeneity of the mammalian SWI-SNF complex. EMBO J, 15(19), 5370-5382. Wang, X., Sansam, C. G., Thom, C. S., Metzger, D., Evans, J. A., Nguyen, P. T., & Roberts, C. W. (2009). Oncogenesis caused by loss of the SNF5 tumor suppressor is dependent on activity of BRG1, the ATPase of the SWI/SNF chromatin remodeling complex. Cancer Res, 69(20), 8094-8101. doi: 10.1158/0008-5472.CAN-09-0733 Weeks, D. A., Beckwith, J. B., Mierau, G. W., & Luckey, D. W. (1989). Rhabdoid tumor of kidney. A report of 111 cases from the National Wilms' Tumor Study Pathology Center. Am J Surg Pathol, 13(6), 439-458. Weiss, Sharon W., Goldblum, John R., & Enzinger, Franz M. (2001). Enzinger and Weiss's soft tissue tumors (4th ed.). St. Louis: Mosby. Wilson, B. G., & Roberts, C. W. (2011). SWI/SNF nucleosome remodellers and cancer. Nat Rev Cancer, 11(7), 481-492. doi: 10.1038/nrc3068 Wilson, B. G., Wang, X., Shen, X., McKenna, E. S., Lemieux, M. E., Cho, Y. J., . . . Roberts, C. W. (2010). Epigenetic antagonism between polycomb and SWI/SNF complexes during oncogenic transformation. Cancer Cell, 18(4), 316-328. doi: 10.1016/j.ccr.2010.09.006 Witkowski, L., Carrot-Zhang, J., Albrecht, S., Fahiminiya, S., Hamel, N., Tomiak, E., . . . Foulkes, W. D. (2014). Germline and somatic SMARCA4 mutations 126 characterize small cell carcinoma of the ovary, hypercalcemic type. Nat Genet, 46(5), 438-443. doi: 10.1038/ng.2931 Witkowski, L., Lalonde, E., Zhang, J., Albrecht, S., Hamel, N., Cavallone, L., . . . Foulkes, W. D. (2013). Familial rhabdoid tumour 'avant la lettre'--from pathology review to exome sequencing and back again. J Pathol, 231(1), 35-43. doi: 10.1002/path.4225 Wolf, P. S., Flum, D. R., Tanas, M. R., Rubin, B. P., & Mann, G. N. (2008). Epithelioid sarcoma: the University of Washington experience. Am J Surg, 196(3), 407-412. doi: 10.1016/j.amjsurg.2007.07.029 Xie, X., Ghadimi, M. P., Young, E. D., Belousov, R., Zhu, Q. S., Liu, J., . . . Lev, D. (2011). Combining EGFR and mTOR blockade for the treatment of epithelioid sarcoma. Clin Cancer Res, 17(18), 5901-5912. doi: 10.1158/1078-0432.CCR-11-0660 Yaniv, M. (2014). Chromatin remodeling: from transcription to cancer. Cancer Genet, 207(9), 352-357. doi: 10.1016/j.cancergen.2014.03.006 Yoo, A. S., Staahl, B. T., Chen, L., & Crabtree, G. R. (2009). MicroRNA-mediated switching of chromatin-remodelling complexes in neural development. Nature, 460(7255), 642-646. doi: 10.1038/nature08139 Yoshida, A., Asano, N., Kawai, A., Kawamoto, H., Nakazawa, A., Kishimoto, H., & Kushima, R. (2014). Differential SALL4 immunoexpression in malignant rhabdoid tumours and epithelioid sarcomas. Histopathology. doi: 10.1111/his.12460 Yoshida, A., Asano, N., Kawai, A., Kawamoto, H., Nakazawa, A., Kishimoto, H., & Kushima, R. (2015). Differential SALL4 immunoexpression in malignant rhabdoid tumours and epithelioid sarcomas. Histopathology, 66(2), 252-261. doi: 10.1111/his.12460 You, J. S., De Carvalho, D. D., Dai, C., Liu, M., Pandiyan, K., Zhou, X. J., . . . Jones, P. A. (2013). SNF5 is an essential executor of epigenetic regulation during differentiation. PLoS Genet, 9(4), e1003459. doi: 10.1371/journal.pgen.1003459 Zagars, G. K., Ballo, M. T., Pisters, P. W., Pollock, R. E., Patel, S. R., Benjamin, R. S., & Evans, H. L. (2003). Prognostic factors for patients with localized soft-tissue sarcoma treated with conservation surgery and radiation therapy: an analysis of 1225 patients. Cancer, 97(10), 2530-2543. doi: 10.1002/cncr.11365 Zhang, Z. K., Davies, K. P., Allen, J., Zhu, L., Pestell, R. G., Zagzag, D., & Kalpana, G. V. (2002). Cell cycle arrest and repression of cyclin D1 transcription by INI1/hSNF5. Mol Cell Biol, 22(16), 5975-5988. 127 Appendix A Ion-torrent Ampliseq Cancer Panel sequencing parameters for VAESBJ. 128 Appendix B Mutations called independently in multiple platforms. T1 Gene Mutation Cosmic Comment Prob CHD1 P275H, P275S N proximity of chromo-domain 0.61,0.54 (WTSS) LENG9 +CCA Intronless T3 Gene Mutation Cosmic Comment Prob AP000525.1 -CCT No genes, close to centromere of chromosome 22 CYP51A1 R446C N 0.61(WTSS), 0.96(WGSS) DCP1B +TGC middle of exon 7 of 9, no domains after insertion FNDC1 -CCGCACGACCACCA middle of exon 14 of 23 FRMD4A -CGCCCCCCG middle of exon 22 of 25, no known domain affected KHK R39C R39H R39H is a SNP 0.72(WTSS), 0.94(WGSS) KLHL12 R392H N In the middle of 6 Kelch motifs 0.68(WTSS) ,0.99(WGSS) SMPD1 P68S N 0.71(WTSS), 0.97(WGSS) SPG7 L630M N Within the peptidase domain 0.65(WTSS), 0.97(WGSS) WBSCR28 V91G N V91L is a SNP 0.55(WTSS), 0.17(WGSS) T5 Gene Mutation Cosmic Comment Prob AGAP10 -G N in exon 10 of 10 CMAHP +T N in exon8 of 14 MYO1D E773G Y, E773K in lung 1 case between calmoduling bindon motif and myosin tail, not in a specific domain 0.11(WTSS), 0.99(WGSS) 129 T5 Gene Mutation Cosmic Comment Prob NUP85 S638Y Y, S638P Liver 1 case 0.11(WTSS), 0.99(WGSS) PAXBP1 +T N close to end of exon 17 of 18 T6 Gene Mutation Cosmic Comment Prob GCNT4 L238W N 0.87 (WTSS) HNRPD Y297C N Not in any domains 0.49(WTSS) HSPA4L C782Y N783Y: 1 case (GI), F781L: 2 (Endometrium) Past the HSP domain 0.77(WTSS) KIAA1586 I295V N, E293D: 1 (Endometrium) 0.89(WTSS) MRC2 R1230H Y, 1 case of prostate Within a lectin domain (there are 8 such domains and a fibronectin type II domain) 0.57(WTSS) P2RY11 -AGTG N In the 2nd of 2 exons SUN1 M668I N Not in any domains 0.89(WTSS) TOX4 V351M Reported as a SNP Past the HMG-box 0.90 (WTSS) ZNF451 K499T N between 2 zinc finger domains but none directly affected 0.97(WTSS) T8 Gene Mutation Cosmic Comment Prob C10orf76 V306I Y, 1 case of GI 0.70(WTSS), 0.99(WGSS) CHD1L -A N In exon 17 of 23, past and not affecting the SNF2 and helicase domains KIAA0391 R450S N 0.71(WTSS), 0.99(WGSS) LIMK2 +C N early in exon 16 of 16 130 T8 Gene Mutation Cosmic Comment Prob PPFIA1 T835M N,1 case of S838L in breast 0.61(WTSS), 0.99(WGSS) SHROOM4 +TCC N Exon 5 of 9 SLC16A4 +A N in exon 10 of 10 HSES Gene Mutation Cosmic Comment Prob ABCD2 N81K N N81H is a SNP 0.83(WTSS) ADCY9 R264Q N Not in any domains 0.86 (WTSS) AHNAK K4384R N The gene has 5891 AA 0.96 (WTSS) CLIP1 T642M N, A641S in 1 case of lung Not in any domains 0.73 (WTSS) CLIP2 A500V N Not in any domains 0.57 (WTSS) EAPP D135G N Close to E2F associated phosphoprotein domain 0.72(WTSS) EPSTI1 +TCAGG N In exon 10 of 10. Alternate transcript lacks this exon GSPT2 G266D N, K265E reported in 1 case of GI In the GTP binding domain 0.62(WTSS) HLA-B Y140D N All surrounding AA are known SNPs, in domains alpha 1 and 2 0.86(WTSS) MSH6 K519Q N Between MutS domains I and II 0.94 (WTSS) PIAS3 L577V N Not in any domains 0.89(WTSS) RACGAP1 A468S N In the RhoGAP domain 0.85(WTSS) RARRES3 S107G N Not in any domains 0.38(WTSS) RPN1 Y209H N IN the Ribophorin domain 0.65 (WTSS) RPS6KB2 M331V N Between 2 kinase domains 0.61 (WTSS) SLC29A3 A466G N In the Nucleoside transporter domain 0.90(WTSS) 131 HSES Gene Mutation Cosmic Comment Prob SRCAP A271P N, S273P in 1 endometrial Not in any domains 0.97(WTSS) TICAM1 F397L N Not in any domains 0.93(WTSS) TRMT5 T465M N No in any domains, hot spot is AA385 with 6 cases on COSMIC 0.93(WTSS) URGCP L636P N Not in any known domains 0.95 (WTSS) WDR90 R511Q N Not in any domains 0.56(WTSS) 132 Appendix C Non-open reading frame fusions predicted using selection criteria described in Chapter 2. Gene 1 region Gene 2 region Gene 1 Gene 2 Expression gene 1 Expression gene 2 Spanning reads Splitting reads ID coding downstream CFL1 CFLP4 51744 0 8 8 HSES coding downstream CRYAB MAFIP 30303 992 24 10 HSES utr5p coding CBX3 C15orf57 603 1779 12 1 HSES intron coding RP11-439A17.7 SRGAP2 146 2045 22 18 HSES coding intron HNF1A CMKLR1 673 236 56 10 T6 coding intron RERE HERC2P3 1554 0 6 3 T6 coding intron RBCK1 PPM1H 848 141 11 6 T6 coding intron HNF1A CMKLR1 673 236 6 7 T6 coding intron PACSIN2 MMP11 2560 1478 35 37 T7 coding upstream TMEM184B MYCBP2 1880 2900 16 14 T7 coding upstream UTP14A RP3-353H6.1 1649 0 5 4 T7 utr5p coding AP1B1 C13orf38 1964 17 9 3 T7 coding upstream UTP14A RP3-353H6.1 1649 0 5 3 T7 intron coding RP4-565E6.1 HYDIN 151 1197 10 17 T7 coding utr5p LATS2 TMEM184B 1229 1880 49 49 T7 coding intron BAIAP2L2 LRCH1 496 1283 6 7 T7 coding upstream TMEM184B MYCBP2 1880 2900 5 1 T7 coding intron SRGAP2 SRGAP2P2 2761 0 15 15 T7 coding upstream AC116353.1 RPS26P8 184 0 16 19 T3 intron coding RP11-312O7.2 SRGAP2 84 4108 10 12 T3 intron coding RP4-565E6.1 HYDIN 232 955 10 14 T3 coding intron C8A LRRC42 52 1406 10 11 T3 coding intron PDE4DIP BX248398.1 15107 0 7 8 T3 coding downstream DRAM1 PAH 772 0 8 11 T3 coding coding SRY XG 82 2201 6 11 T8 intron coding RP11-259O2.3 ITGBL1 406 7509 5 4 T8 downstream coding RP11-312O7.2 SRGAP2 223 5332 15 17 T8 downstream coding MAFIP SCAF1 584 4037 5 2 T8 coding downstream POU2F1 AC092265.1 1782 0 7 7 T8 coding intron MTOR UBR4 4241 12209 14 18 T8 coding intron POLR2J3 TYW1 3228 1061 6 7 T8 133 Gene 1 region Gene 2 region Gene 1 Gene 2 Expression gene 1 Expression gene 2 Spanning reads Splitting reads ID coding coding MT-ND5 MT-CO3 102543 75476 110 91 VAESBJ coding intron GNB2L1 RP5-857K21.4 37445 27 13 13 VAESBJ coding intron HDLBP RP5-857K21.4 78649 27 9 12 VAESBJ intron coding C1RL-AS1 SMURF2 433 22021 9 26 VAESBJ coding intron PSMD13 RP5-857K21.4 5276 27 5 9 VAESBJ utr5p coding SPP1 DCBLD2 106552 87576 11 10 VAESBJ coding intron P4HB RP5-857K21.8 78511 0 9 6 VAESBJ utr5p coding MT2A TMBIM6 11643 33561 8 1 VAESBJ coding upstream ACTG1 NCRNA00273 151627 3 8 13 VAESBJ coding intron ACTB RP5-857K21.4 166219 27 26 37 VAESBJ coding coding CSTB HNRNPM 24591 14199 5 2 VAESBJ utr3p coding ZNF264 DRAM1 3098 8144 12 11 VAESBJ intron coding RP11-439A17.7 SRGAP2 606 7055 23 40 VAESBJ utr5p coding ACTB GANAB 166219 25833 6 6 VAESBJ utr5p coding MT2A COPS6 11643 8122 6 1 VAESBJ coding upstream MAP3K4 MIR1202 812 0 20 18 Epi544 coding intron ADAM9 AC005301.5 42291 31928 24 662 Epi544 coding upstream C8orf59 5S_rRNA.374 566 0 39 2 Epi544 coding upstream C8orf59 5S_rRNA.374 566 0 8 1 Epi544 coding downstream CRYAB MAFIP 34672 428 11 5 Epi544 coding intron ADAM9 AC005301.5 42291 31928 759 764 Epi544 intron coding RP11-439A17.7 SRGAP2 569 5778 24 35 Epi544 utr5p coding SH3D19 LRBA 8003 11867 8 11 Epi544 coding intron ADAM9 AC005301.5 42291 31928 19 27 Epi544 utr5p coding MLKL AC010531.1 705 113 17 23 Epi544 coding upstream ACTG1 NCRNA00273 90978 0 13 4 Epi544 coding intron ATP6V0C AL163011.1 13176 0 5 4 Epi544 coding coding AFF1 PTPN13 3124 1217 14 21 T4 intron coding RP11-439A17.7 SRGAP2 322 3210 13 15 T4 coding upstream DOCK5 AC084262.1 3458 0 10 17 T4 Coding coding SRGAP1 PAWR 1269 393 9 7 T4 coding downstream DTD1 TGFBR1 1048 3005 5 7 T4 coding coding DLG1 COPG 6114 7232 13 14 T5 coding downstream COL1A2 MAFIP 183761 681 6 4 T5 coding intron PDE4DIP BX284650.1 27388 0 8 11 T5 intron coding RP11-439A17.7 SRGAP2 487 6126 26 41 T5 134 Gene 1 region Gene 2 region Gene 1 Gene 2 Expression gene 1 Expression gene 2 Spanning reads Splitting reads ID coding intron PDE4DIP RP11-277L2.2 27388 427 10 15 T5 coding utr3p RNF123 STAT3 3150 19515 5 6 T5 coding intron DLG1 COPG 6114 7232 5 6 T5 coding coding DLG1 COPG 6114 7232 5 4 T5 135 Appendix D Cluster outlines for the differential expression analysis of cell lines (678 lines in total). All the epithelioid sarcoma lines are in cluster 3.75 which is shown in Figure 2-6B. 136 Appendix E List of top 30 differentially expressed genes between clusters 3.5 and 3.75 from Appendix C. logFC logCPM PValue FDR S100A14 -4.15 5.25 1.96E-25 TSPAN1 -3.39 6.55 2.20E-24 LAD1 -4.09 5.99 2.96E-23 COL6A1 2.52 12.02 1.27E-21 BSPRY -4.1 3.49 1.24E-19 COL1A1 4.03 13.92 1.30E-19 CEACAM6 -6.95 8.12 2.37E-19 CBLC -3.66 3.05 1.34E-18 MPZL2 -4.57 6.57 3.34E-17 COL6A2 2.61 11.9 4.29E-17 ESRP1 -3.78 5.73 1.19E-16 MMP2 3.75 10.96 2.43E-16 MAL2 -3.66 8.42 2.43E-16 EPCAM -3.53 7.64 2.44E-16 HOOK1 -3.39 6.36 3.59E-16 CCDC80 2.28 10.81 4.99E-16 ST14 -2.68 6.55 7.72E-16 ARHGAP4 -2.97 7.33 8.28E-16 CREB3L1 2.62 9.03 2.11E-15 RAB25 -3.39 3.83 2.36E-15 LCN2 -3.9 6.38 2.92E-15 ELF3 -3.05 7.67 7.56E-15 INADL -1.58 7.83 1.15E-14 CLDN3 -2.93 4.74 1.48E-14 PCSK9 -4.15 7.33 5.86E-14 IRF6 -3.27 5.23 1.35E-13 SPARC 2.91 13.27 2.10E-13 CLDN1 -2.41 9.83 3.59E-13 LOX 2.38 10.65 7.95E-13 C1orf116 -2.73 5.35 8.32E-13 137 Appendix F (Broad-spectrum) HDAC inhibitor romidepsin treatment (A). 5nM treatment leads to significant up-regulation of EGR1 and PTEN increase in HSES. Syo1 cells were used as positive control and HEK293t cells as negative control for comparison.