Translational Research in Well-Differentiated and
Dedifferentiated Liposarcoma

By David Thomas, FRACP, PhD; Ola Myklebost, PhD and Leonardo A. Meza-Zepeda, PhD

Introduction

The original aims of this International Collaborative Grant (ICG) as presented in the companion Experimental Plan were to:

1. Utilize state-of-the-art molecular genomics to map epigenetic and genetic landscape of WDLPS
2. Model and analyze the evolution of drug resistance, and to understand the process of de-differentiation common to this cancer
3. Develop useful preclinical models of WDLPS

This ICG created an international team of sarcoma researchers that has been working together for two years (the WDLPS Team). The collaboration included a Norwegian group led by Drs. Myklebost and Meza-Zepeda and an Australian group led by Dr. Thomas, with the early involvement in the US of Dr. Maki at Memorial Sloan-Kettering Cancer Center and Dr. Barretina at the Broad Institute. Not only has the program of work originally envisioned been substantially achieved, but changes in focus as the research unfolded have led to additional important results and open questions.

Results

The following summarises the findings of the research group under these aims.

1. Application of molecular genomics to map the genetic landscape of WDLPS

Genetic changes at a chromosomal scale are common in cancer. These include genomic instability, chromothripsis, and rearrangements such as the translocations characteristic of certain leukemias and sarcomas. A distinct form of cancer-associated chromosomal variation is the formation of ring or giant marker chromosomes (neochromosomes: NCs). NCs are an almost invariant feature of some cancers, including well- and de-differentiated liposarcoma (WDLPS), parosteal osteosarcoma and dermatofibrosarcoma protuberans.¹ While the oncogenic properties of cancer-associated NCs are not fully understood, gene amplification or structural gene rearrangements appear to be important. In WD/DDLPS, NCs contain amplified copies of genes including MDM2, CDK4, SAS, HMGA2 and YEATS4, whose over-expression is thought to drive or contribute to carcinogenesis.^2-6

The architecture of cancer NCs has not been previously described at single nucleotide resolution. We have undertaken the first sequence-level structural map of an entire flow-isolated cancer-associated NC. These analyses employ standard cytogenetic and fluorescence in situ hybridisation (FISH), array-based profiling, and flow cytometric isolation and massively parallel sequencing of WDLPS NCs, complemented by transcriptome sequencing. Taken together, these studies reveal a hitherto unrecognised structural complexity and shed insights into the mechanism of formation and pathogenesis of cancer NCs.

We have confirmed that the ends of the NC (telomeres) are acquired late in the evolution of the NC probably once it achieves a linear form. The telomeres are acquired by capture from native chromosomes, illustrated here for the GOT3 cell line by regions from chromosomes X and 1 (Figure 1A). Linearization of the NC seems to be associated with cessation of amplification and stabilization of the structure. Centromeric structures on the NC appear to use native centromeres, and once these are corroded by the breakage-fusion-bridge process outlined below, substitution by neocentromeres which assemble on non-alphoid regions of Chr1 and Chr6 which have become incorporated into the NC.

The NC core (outside the telomeric regions) is composed of about 31Mb of highly rearranged and amplified fragments of multiple native chromosomes. The final size of the core region is about 640 Mb. Approximately 250 genes are located in the amplified core, of which 181 appear to be highly amplified. There are over 500 distinct fragments, termed continuous genomic regions, which comprise the core. Detailed analysis of the breakpoints and fusions generated in the course of assembly indicates that the NC initially forms from chromosome 12, but subsequently incorporates additional material predominantly from chromosomes 1, 15, 13, and 16. The process of amplification involves a ring structure, which fails to separate following replication, forming a double ring. During cell division the double ring separates by random breakage, asymmetrically distributing into the two daughter cells. We believe that the cell that contains the greater amount of oncogenic material experiences a selective advantage in growth. The process thought to be responsible for rearrangement of the NC content is known as breakage-fusion-bridge,⁷ has been modeled, and generates a structure with most of the observed features of the NCs we have studied. The model we have developed is summarized in Figure 1B.

We have identified loss-of-function mutations in a key gene involved in resolution of the double Holliday junctions which are generated during replication of ring structures. We are currently investigating whether the mutation is required for the generation or pathogenesis of the NC. We have undertaken a detailed census of amplified genes on the NC, and correlated these with those amplified in primary liposarcoma. Subsequently, we have knocked down 50 of these amplified oncogenes, and identified a number of previously unrecognised potential oncogenes in addition to those previously reported.

2. Models of drug resistance: MDM2 and CDK4 antagonists

These studies aim to define genes that modify the sensitivity to inhibitors of MDM2 and CDK4. The outcomes of these experiments are to understand the molecular pathways involved in CDK4 signaling in WD/DDLPS, and to determine likely modes of resistance to inhibitors of this pathway in the clinic. The latter outcome will be particularly important because initial clinical studies of both agents singly have failed to demonstrate impressive activity (unpublished reports).

2.1 Resistance to Mdm2 antagonists

We have generated spontaneously resistant WD/DDLPS cell lines by prolonged culture in the presence of Nutlin, an MDM2 antagonist. Our experimental settings provide a reproducible model system for the selection of drug resistant clones derived from the Nutlin-sensitive parental 778 tumor cell population. Resistant variants are generated by continuous exposure to 2.5 µM Nutlin, empirically the optimal drug concentration for generating selective pressure for most cells, at the same time enabling the emergence of resistant clones among the surviving cells. Typically, a starting culture of 3x105 cells will, after a reproducible 7 week period, produce 1 to 3 spontaneously resistant subclones, which are then expanded further and maintained in the presence of Nutlin. We have undertaken profiling of the resistant lines, looking for genetic and epigenetic changes that are linked to drug resistance. We have made several findings of interest.

We have observed that resistant cells tend to demonstrate increased genetic instability. We quantified chromosomal aberrations of 778 parental and Nutlin-resistant cells via spectral karyotyping analysis. Drug resistant populations demonstrated karyotypic instability, and a higher background number of double strand breaks. Using transcriptional profiling, the resistant line has down-regulated genes involved in DNA repair, chromosome segregation and spindle checkpoint activity.

It appears that resistant lines demonstrate greatly increased methylation across the genome. This is consistent with the operation of epigenetic silencing mechanisms, which may contribute to drug resistance by inactivating genes required for sensitivity to Nutlin. To understand these, we have analysed gene expression profiles of resistant cells, identifying several hundred genes that are differentially expressed. These are currently being confirmed using more sensitive and specific techniques. A proportion of these genes demonstrate statistically significant methylation changes, consistent with epigenetic mechanisms of gene silencing. Several genes of interest are currently being investigated for functional contribution to drug resistance.

We have also deep sequenced the parental and resistant cell lines, looking for single nucleotide and small scale genetic changes that may be linked to drug resistance. These studies aim to identify genetically distinct subpopulations whose frequency changes under selection. We began by testing bioinformatic strategies to identify low-frequency populations, using synthetic datasets generated by combining the raw data from 10 independently derived cancer cell populations to develop the analytic tools that reliably detect low frequency genotypes (<5% population frequency).

These techniques have been applied to 12 independently derived parental and resistant clones, which have been allowed to expand for 16-17 population doublings. The goal for these studies was to apply population genetic methods for mapping clonal evolution of cancer cells, before and after passage through a selective bottleneck—in this case, MDM2 inhibition. The results are described in Figure 2.  One highly evolved clade, with the highest Nutlin resistance, has developed a mutation in TP53 (C238F, a known loss-of-function mutation that is involved in cancer development). This clearly explains the resistance in this clade, but leaves open the question of the mechanisms of resistance in the remaining clones, which also continue to evolve.

2.2 Resistance to CDK4 antagonists

We have developed a functional genome-wide screen for genes modifying sensitivity to CDK4 antagonists. This screen involves establishing conditions for drug selection of WD/DDLPS cell lines (449B, 778, T1000) using the Pfizer lead CDK4 inhibitor (CDK4i). After testing several compounds, this drug was chosen because it has established target specificity (ie, minimal off-target effects). A positive control (shRB1) has been generated to provide an estimate of assay conditions required, including drug concentrations and read-out, to detect the effects of loss of sensitivity to the CDK4i.

First and second round screens have been performed using the 449B cell line in the presence of the CDK4i. The first round studies have identified a subset of 300 genes whose knockdown appears to consistently result in some degree of drug resistance, analogous to RB1. These genes were taken into a second-round screen in which the siRNA "smartpools" were deconvoluted, and individual siRNAs tested for the ability to generate resistance. From this second-round screen, 13 genes have been selected for third-round screens against other cell lines (778, T1000). The results of these screens will be known in the next 3 months, and will be reported.

3. To develop useful preclinical models of WDLPS

Cell lines represent important in vitro models for studying disease mechanisms at the cellular level, and have been widely used as preclinical models for evaluation of new therapeutic drugs. Therefore, characterization of preclinical cell line models represent an important step towards better understanding the biology of different cancers, as well as identifying novel therapeutic strategies. To date, only a few immortal liposarcoma cell lines have been characterized extensively. In order to better understand the biology of liposarcomas and provide model systems for sarcoma research, we have characterized a panel of nine immortalized liposarcoma cell lines.⁸

The panel of cell lines we evaluated for proliferation, tumor-forming ability, stem cell and differentiation potential, as well as invitation and migration. Four out of 9 cell lines analyzed showed high proliferative capacity, measured by live cell imaging, meanwhile the remaining cell lines displayed a lower growth rate capacity (Figure 3a). The capacity of each cell line to form tumors was tested in vivo by injecting 1x10⁶ cells subcutaneously into the flanks of immunecompromised mice (NOD SCID). Five of the nine cell lines were capable to form tumors within a 6 month period (Figure 3b), meanwhile the remaining lines showed no or almost no tumor growth.  In addition, all cell lines were tested for the capacity to differentiate towards the adipogenic lineage, as well as for the expression of early (CEBPB, CEBPA and PPARG) and late (FABP) adipogenic markers. Only the cell line T778 showed strong expression of adipogenic markers, with levels similar to adipogenic differentiated human messenchymal stem cells (hMSC). The remaining cell lines showed marker levels similar to undifferentiated hMSCs. Interestingly, all cell lines, except T778, underwent spontaneous adipocytic differentiation when cultured in normal media at high density.

Re-expression of embryonic and pluripotency factors like OCT4, SOX2 and NANOG haves been associated with poorly differentiated and more aggressive cancers.⁹ All three markers were highly expressed in LPS141 and LISA-2, while moderately expressed in T449. Surprisingly, there was no correlation between the expression of these stem cell markers and colony forming ability, thought to reflect stem cell-ness. Expression of CD133 and Aldehyde dehydrogenase (ALDH) activity has been previously used to enrich for Cancer Stem Cells (CSCs) in liposarcomas.⁸ These, and other reported CSC markers were evaluated within the panel. All the lines contained a subpopulation of cells expressing CD133, varying from 4% to less than 1% for most of the cell lines. Only LPS141 displayed a significant subpopulation of cells with ALDH activity.  To assess the metastatic potential of liposarcoma cell lines, cells were subjected to migration and invasion assays. LPS141, T778 and FU-DDLS displayed a high capacity to migrate using a Boyden Chamber, while only LSP141 and FU-DDLS could efficiently invade through a layer of Matrigel. All the functional and phenotypic data are summarized in Figure 4. 

At present, the initial panel of liposarcoma cell lines has been extended to 12, and gene expression and exon mutation data is being generated by high-throughput sequencing for this extended panel of cell lines.

3.1 Genomic characterization of liposarcoma metastasis, a case study.

Using high-throughput genomic sequencing technology we are currently characterizing the transcriptome and genome of multiple metastatic lesion from a patient with a dedifferentiated liposarcoma. We have sequenced the exome of three independent metastases and identified somatic mutations within each. Each tumor contains between 30-40 somatic acquired mutations, of which only one third is shared among the three samples. Using the genomic exome data, DNA copy number and loss of heterozygosity has been deduced (Figure 5). Analysis of each metastasis shows similarities as well as differences among the samples analyzed, this most likely reflecting structural variation of the NCs and changes in subpopulation composition with the different tumor metastases. We are currently studying ways to define a sequence of events for the metastatic lesions. From the exome data we were also capable of identifying a number of genomic rearrangements, most of these located in amplified regions. A global overview of the genomic finding for the three tumors is presented in a Circos plot in Figure 5.

Several tumorgraft lines were established in immunodeficient mice, which will be important preclinical models to extend our studies of resistance to Nutlin and CDK4i, since this patient did not respond to either treatment.

Summary and Discussion

The inaugural LSSI ICG has proven a fertile vehicle for fostering collaborations across national boundaries, which focus upon a specific disease: well-/de-differentiated liposarcoma. What have we learnt?

Since the inception of this ICG, drugs targeting MDM2 and CDK4 have entered and completed initial trials. Although these data remain to be reported, it appears that neither class of compounds is proving a panacea—at least with each agent used singly. In this context, the studies here advance our knowledge of disease in two ways. First, they add important genomic and technical information to assist researchers in understanding the complex biology of WDLPS. The studies describing the architecture of the neochromosome, and proposing a model for its formation, represent significant contributions to elucidating this specific form of mutation, which is shared by several cancer types. In addition, the studies describing patterns of mutations in primary human WDLPS contribute to the catalog of mutations observed in the clinic, and begin to get at how these patterns vary within each patient. This will be important to understanding how these cancers evolve (see below). The studies focused on characterizing several cell lines are important because they provide researchers with detailed information about key tools that will be used across multiple studies in the future. It is important to note that there is a commitment by the ICG partners to make these tools available to researchers in the future.

The second area of contribution goes to drug resistance—perhaps the single greatest challenge to the use of targeted therapies in cancer. As the MDM2 and CDK4 inhibitors enter the clinic, it will be critical to understand why we don't see dramatic responses. Only through this iterative process will we achieve the ultimate goal: methods of controlling, and eventually eradicating, this disease. The resistance studies have a general and a specific purpose: to understand the various ways in which WDLPS fails to respond to targeted therapies; and to understand generally how the WDLPS genome morphs, or evolves, under the presumably intense selection pressure generated by these agents. The former addresses specific questions—in this case about mechanisms of resistance to CDK4 inhibitors; the latter addresses fundamental aspects of the evolution of cancer cells using population genetics approaches.

Finally, the ICG has established an excellent basis for future research, involving the partners to this project, and others. The meetings and discussions have been very fruitful over the past two years, and have extended beyond the 'nuclear family' of the ICG, including other relevant research groups working on WDLPS, as well as the myxoid LPS consortium supported by the Liddy Shriver Sarcoma Initiative. A communal web portal liposarcomaresearch.org has been made to enhance communication within this field. We hope that these discussions will form the basis of the next research projects in WDLPS.

Unanswered questions

The WDLPS ICG has raised many new questions that should be addressed. We offer the following as examples:

1. If the formation of the neochromosome is a unique and specific event, seen in a subset of human cancers, what defects in the mechanisms that maintain genome integrity permit this to happen? Presumably such defects are specific for these cancers, including WDLPS. If so, like the BRCA1/2 defects in breast cancer, treatments that focus on synthetic lethality may offer benefits.
2. Can we catalog the various genetic and epigenetic mechanisms that contribute to resistance to MDM2 and CDK4 antagonists? This will not only shed light on how these key pathways function in WDLPS, but may also complement the development of these classes of agents.
3. How does de-differentiation occur? In part this may relate to the plasticity of the WDLPS genome, so understanding how it changes under selection and with time will be critical to knowing why this tumor evolves from an almost benign entity into a highly aggressive neoplasm.

V10N1 ESUN Copyright © 2013 Liddy Shriver Sarcoma Initiative.

By David Thomas, FRACP, PhD; Ola Myklebost, PhD;
Jordi Barretina, PhD and Robert Maki, MD, PhD

Abstract

Well-differentiated and dedifferentiated liposarcomas (WD/DDLPS) comprise the single most common form of soft-tissue sarcomas. The oncogenes MDM2 and CDK4 have recently been shown to be important in WD/DDLPS, and the development of drugs that target these genes has created an urgent need to better understand the biology and genetics of WD/DDLPS. In this program of work, an international team of basic and translational researchers come together to address key questions facing the clinical and research community, including a detailed mechanistic knowledge of how they form, the role of the cancer stem cell, how resistance to MDM2 and CDK4 blockade might arise, and the development of animal models of this disease. These studies will accelerate the translation of an understanding of the biology of WD/DDLPS into clinical outcomes.

I. Background

Public Health

There are approximately 130,000 new cases of sarcoma world-wide each year.² Sarcomas comprise 10-20% of cancer in the young and the overall mortality is approximately 50%. An average of 17 life years per patient is lost due to sarcomas (three times greater than bowel, lung or breast cancer). Treatment is complex, often prolonged and costly, and surviving patients experience substantial life-long disability. Collectively, these features lead to a community impact (measured by disability-adjusted life years) equivalent to cervical cancer, or head and neck cancer.

Pathology and Clinical Features

Liposarcoma, the second most common type of sarcoma, forms four semi-distinct subclasses: Well-differentiated liposarcoma (WDLPS) comprises 40% of liposarcoma; dedifferentiated liposarcoma (DDLPS) comprises approximately 20%; myxoid liposarcoma comprises 30%; and pleomorphic liposarcoma the remainder.³ The median age at presentation of patients with WDLPS is 45-50 years of age.

Pathologically, WDLPS is often difficult to distinguish from the much more common benign lipoma. WDLPS is often mistakenly regarded as indolent, but in sites such as the retroperitoneum recurrence is the rule and the eventual mortality approaches 100%. Transformation to DDLPS occurs in about 25% of WDLPS, with distant metastasis in 15-20% of patients.⁴ DDLPS may be even more common, as it is often classified as malignant fibrous histiocytoma (MFH) [now termed undifferentiated pleomorphic sarcoma, UPS], another common category of high-grade soft-tissue sarcoma.⁵ Treatment of WDLPS/DDLPS is primarily surgical with or without radiotherapy, and current systemic therapies for WDLPS are toxic and have response rates under 25%.

Molecular Pathogenesis

WDLPS is almost unique in being ‘driven’ by focal, high-level amplification events, typically located in accessory ‘ring’ or giant rod chromosomes, and found in otherwise diploid karyotypes.⁶ Low-resolution cytogenetic studies have shown that WDLPS and DDLPS are characterized by high-level amplification of 12q13-q22 containing both CDK4 and MDM2, as well as a number of other sites.⁷ These have recently been mapped at very high resolution.⁸

Good evidence implicates MDM2 and CDK4 in the pathogenesis of WDLPS. MDM2 is located at 12q15, and is amplified or overexpressed in over 30% of sarcomas, over 80% of WDLPS, and 60% of myxoid LPS.⁹ MDM2 is an E3 ubiquitin ligase that binds and ubiquitylates p53, which leads to proteasomal degradation.¹⁰ MDM2 may also sterically block activating N-terminal serine/threonine phosphorylation of p53 sites by CHK2 and other DNA-damage activated kinases. p14ARF, which is also mutated in cancer, induces p53 levels by disrupting MDM2-p53 interactions. p53 transcriptionally activates genes involved in cell growth arrest and apoptosis, including the cyclin-dependent kinase inhibitor p21CIP1, and others. MDM2 has been reported to interact with non-p53 pathways, including the retinoblastoma tumour suppressor (Rb).¹¹ CDK4, located nearby MDM2 at 12q13, is co-amplified in over 80% of WDLPS with MDM2 amplification.⁹ The D-type cyclin-CDK4/6 complex phosphorylates and inactivates Rb, leading to the de-regulation of the G1/S-phase checkpoint. Co-amplification of MDM2 and CDK4 may therefore inactivate both the p53 and Rb pathways.

Two other genes located at 12q which have attracted interest include HMGA2 (a high-mobility group protein) and SAS (sarcoma amplified sequence, a member of the tetraspanin family).⁸ HMGA2 is systematically co-amplified with HMD2 in liposarcomas,¹² while transgenic mice ectopically expressing HMGA2 develop lipomatosis.¹³ During this amplification HMGA2 is fused to distant sequences frequently from other chromosomes, and loses its 3’ untranslated end, containing 7 sites for the Let-7 microRNAs, as well as exons 4 and 5.⁷ This leads to the overexpression of a truncated protein, lacking the c-terminal part which, among other genes, activates the sarcoma protooncogene SSX1 in mesenchymal stem-like cells.¹⁴ Remarkably, the amplification of CDK4, MDM2 and HMGA2 is also observed in parosteal osteosarcomas, another disease associated with cancer-associated neochromosomes.¹⁵ Parts of chromosome 1 (1q21-q24) are often amplified both in WD/DDLPS and parosteal osteosarcomas.¹⁶

It is notable that adipocytes and osteocytes are derived from a common stem cell, and transdifferentiation of DDLPS to osteogenic tumors is observed, at least in one case associated with amplification of 1q.¹⁸ The process of dedifferentiation is not well understood. Amplification of MDM2 and mutations in TP53 are mutually exclusive in WDLPS, but dedifferentiation may be associated with acquisition of p53 mutations.^{19, 20} Abnormal clones may evolve with large numbers of supernumerary markers, or new amplification events on chromosome 1q22-24.¹⁸ The oncogenic roles of these proteins remain unknown. Recent data suggest that the protooncogene c-JUN, which is a target for HMGA2,²¹ may play a role in dedifferentiation, although this is a subject of active research.^{22, 23} The understanding of processes that determine the balance between differentiation and self-renewal is critical for the development of new therapies.

Recently there has been a focus on the involvement of stem-like cells and stem cell phenotypes in cancer development, to a large extent fuelled by the tremendous strides in knowledge on normal stem cell biology (Note 1). Two main aspects of this are particularly important. First, many properties of stem cells are particularly dangerous in cancer cells, such as their immortality, their ability to detoxify and export xenobiotics, their increased repair capacity for DNA damage, and their relative quiescence. All these properties make them resistant to two of the three main arms of current standard therapies, chemotherapy and radiation. Second, the concept is that some tumors resemble a tissue in the sense that there is a small population of tissue stem cells that can renew the whole tissue. Thus, a successful therapy has to hit these cells, or else the tumor may seem to disappear, only to reappear as these cells repopulate it. Because most characterization of tumor properties is done on the average cancer cell, such a small, perhaps 0.1% population could have completely different and unknown properties.

Clinical opportunities in WD/DDLPS

As p53 is such a critical molecule in terms of cancer cell survival, it is not surprising that MDM2 has also become a target of interest in cancer.^10,24 A series of preclinical studies indicates the activity of the MDM2 antagonist Nutlin 3a in liposarcomas and other tumors with MDM2 overexpression.^25-28 The data from a limited number of cell lines also indicate that nutlin can be active in combination with chemotherapy in tumors that bear TP53 mutations, acting in an E2F1 dependent fashio.²⁵ A range of CDK4 antagonists are also in clinical development, most of which are in early phase, and most of which are fairly non-selective, having effects on a number of CDKs.²⁹ The first generation of non-selective CDK inhibitors was associated with significant toxicities, which may be ameliorated by the development of more selective antagonists.^29,30 The most advanced in clinical trials is PD0332991, which has been recently shown to inhibit proliferation of dedifferentiated liposarcoma cell lines, at least in vitro.⁸

The frequent co-amplification of MDM2 and CDK4 in WDLPS/DDLPS is of particular interest because it presents an opportunity for combined use of targeted therapeutics.^1,24,31 As a result, CDK inhibitors and MDM2 inhibitors have completed phase I studies and are being expanded to examine patients with wild type p53. The clinical trials become the logical extension of the preclinical work of this program, and the use of materials from patients treated with CDK4 or MDM2 inhibitors will be used to confirm findings from cell line and xenograft studies.

II. Structure and Context of This Research

A key objective of this research is to assemble a high quality team of clinicians and scientists with strong and relevant individual track records for the program of work outlined below. Each of the investigators currently work on WD/DDLPS. The leveraging of existing resources maximizes return on the Liddy Shriver Sarcoma Initiative investment, enables a greater scope for the research, and ensures that the work builds on strong, peer-reviewed foundations.

Robert Maki (MD, PhD) is Section Chief for Adult Sarcoma Medical Oncology at Memorial Sloan-Kettering Cancer Center (MSKCC). He has a special interest in new therapies for the treatment of soft tissue and bone sarcomas, and is focused on new drugs to treat metastatic disease. In particular, he is investigating new drugs to attack novel molecular targets for this rare and varied group of well over 70 or more types of cancer that arises from connective tissues. He is also interested in novel clinical trial designs that make testing of new agents more efficient and more effective. He is also involved in translational research in angiosarcomas, GIST, and other sarcomas in collaboration with Cristina Antonescu and others at MSKCC and at other centers.

Ola Myklebost (PhD) is Group Leader at the Department of Tumor Biology, Institute for Cancer Research, The Norwegian Radium Hospital, and Professor at the Department for Molecular Biosciences at the University of Oslo. He and his group are using functional genomics to characterise mesenchymal oncogenesis, which is the process leading to formation of malignant connective tissue tumours, or sarcomas. The Norwegian Radium Hospital is a national centre of competence in sarcoma diagnosis and treatment, and the research of Ola Myklebost is done in close collaboration with clinicians, both locally and within the Scandinavian Sarcoma Group. Recently, his group became part of European Network of Excellence. Myklebost has published 96 papers in international peer-reviewed journals, including EMBO Journal, Nature Biotechnology and Proceedings of the National Academy of Science USA.

Jordi Barretina (PhD) is a research scientist working in the Broad Cancer Program, with Dr. Levi Garraway. Jordi leads the Cancer Cell Line Encyclopedia (CCLE) project, a collaboration between the Broad, the Novartis Institutes for Biomedical Research (NIBR) and the Genomics Institute of the Novartis Research Foundation (GNF). His scientific focus is on applying genomic and functional tools to the systematic analysis of the cancer genome, with a special emphasis on translational research. Jordi carried out the Sarcoma Genome Project, which discovered several subtype-specific genomic alterations that define new targets for soft tissue sarcoma therapy (Nature Genetics, 2010).

David Thomas (FRACP, PhD) is currently a Victorian Cancer Agency Clinician Scientist at the Peter MacCallum Cancer Centre. He is the founding Chair of the Australasian Sarcoma Study Group, a member of the science task force for the Livestrong Young Adult Alliance, and a founding member of the World Sarcoma Network (see below). Dr Thomas’ major research interests include the survival gap in AYA cancer, giant cell tumour of bone, the molecular biology of osteosarcoma and liposarcoma, and the germline and somatic genetics of adult-onset sarcomas. Dr Thomas has published over 60 papers, including lead author papers in Molecular Cell, Journal of Cell Biology, Journal of Clinical Investigation, Nature Reviews Cancer, and Lancet Oncology.

It is anticipated that the studies outlined below will form part of a larger program of work whose ultimate goal is participation in a prospective clinical study that utilizes agents that target the key molecules (MDM2 and CDK4) at the centre of this program. These agents are already in early clinical development. In addition, more novel treatment approaches involving Let-7 may form part of the future therapeutic strategy available, and will be considered as evidence and opportunity arises. The studies outlined here represent a foundation for the future development of clinical trials, which will necessarily involve collaboration between the pharmaceutical industry and the clinical community. We hope that the World Sarcoma Network (WSN), a clinical inter-group structure bringing together multiple national trials groups, will provide a forum for development of the study and will support its conduct globally. The WSN comprises members of the US sarcoma group SARC, the EORTC, the French Sarcoma Group, in addition to other key members of the global sarcoma research community.

III. Aims

The overarching aim of this research is the development of a coordinated program of translational research focused on WD/DDLPS (see Figure 3). The specific aims are to:

1. Utilize state-of-the-art molecular genomics to map epigenetic and genetic landscape of WD/DDLPS
2. Model and analyze the evolution of drug resistance, and to understand the process of dedifferentiation common to this cancer
3. Develop useful preclinical animal models of WD/DDLPS.

IV. Experimental Program

1. Genomic studies in WD/DDLPS

We have recently generated one of the most detailed maps of genomic alterations (at the level of DNA copy number, gene expression and mutations) across diverse soft tissue sarcoma subtypes, providing potential subtype-specific targets for therapy.⁸ In the particular case of dedifferentiated liposarcoma, the subtype with the most samples analyzed, we performed an integrated structural and functional genomics analysis and found a number of genes significantly amplified, beyond the canonical amplifications of MDM2, HMGA2, and CDK4, likely to be additional driver genes.

We want to apply high-throughput genomic tools to clinical samples, particularly in the context of targeting MDM2 and CDK4. Oral inhibitors of various CDKs, e.g. CDK4/6, and CDK2/4 are in clinical development. One such CDK4/6 inhibitor, PD0332991, has completed phase I testing, with at least modest activity observed in patients with DDLPS. No biospecimens were obtained, so the effects on liposarcoma patients’ tumor remain unknown. A phase II study of PD0332991 has been approved by Pfizer, and is being performed at Memorial Hospital. This study will examine patients specifically with WDLPS and DDLPS who have received 0-1 lines of prior therapy, and will require pre and post treatment biopsies for participation in the study. We will take extra tumor biopsies at the time of pre and post treatment biopsies mandated by the study and hold these samples for examination of some of the markers of interest found in the preclinical and translational work noted in the prior sections of this application. For example it will be possible to examine cells for signs of cell cycle arrest and apoptosis, which should correlate with some of the genes up and down regulated by CDK4 and MDM2 inhibitors.

MDM2 antagonists from Johnson & Johnson and Roche have also completed phase I testing. At Memorial Hospital we have participated with three other centers in a phase I clinical trial of the Roche MDM2 inhibitor R7112, similar in concept to the Nutlin 3a compound used in some of the preclinical models of this application. A portion of this study will require biopsies be performed before and after therapy with the R7112 inhibitor. We will again take advantage of the access to patient tissue by performing an extra pass of the biopsy needle at the time of each procedure to procure material we can use for our own analyses independent of the clinical trial.

Materials from patients who are treated on clinical trials will consist of pre-existing banked material (frozen, paraffin) and normal tissue controls, part of the MSKCC tissue bank, as well as materials derived from clinical trials. For the MDM2 study, patients will have tumor samples collected before treatment and after a week of therapy, to look for acute changes while on therapy. Six to 10 patients will be treated at MSKCC on this study, and each will have paired biopsies. With a relatively long half-life of the agent involved (R7112), therapeutic levels will have been achieved so that a fair comparison can be made between pre and post treatment tissue with respect to the target at hand. It will not be possible to use these samples to determine resistance patterns of tumors, as resistance presupposes response to treatment, which may or may not occur on study. Similarly, patients on the CDK4 inhibitor study (again, 5-10 in number) will have pre and post treatment samples. Biopsies will be taken prior to and during the second week of therapy for this oral agent, again with a favorably long half life with respect to concerns of timing of biopsies relative to treatment dose. Where possible on the CDK4 inhibitor study, we will allow patients with resectable disease to allow them on treatment up to the day of a procedure, in order to obtain larger amounts of useful tissue. Otherwise it is anticipated we will only a have a 2-3 mm core of tissue taken at each biopsy time point. We are generally able to obtain viable paired material in about 80% of patients on such clinical trials at our institution.

We have access to the bioinformatics resources necessary to analyze the data that will be generated. Two bioinformaticians are available to work on the project at the Broad Institute, and the Institute for Cancer Research at the University of Oslo, 2.5EFT of bioinformatician time is available to the project. The Broad has a long track record of bioinformatics/computational analysis in cancer, and we’ll have access to a group of scientists affiliated with the Computational Biology and Bioinformatics group, that serves as a central resource on computational issues for the Broad's scientific staff.

1.1 Cataloguing of epigenetic changes during the progression from WD to DDLPS

Principal investigators: Maki/Barretina/Myklebost
The field of epigenomics is focused on the study of chemical modifications to DNA and histone proteins that modulate chromatin structure and genome function, impacting the regulation of gene transcription. Changes in the patterns of histone post-translational modifications have been implicated in several pathologies, including cancer.

To identify transcriptional and epigenetic mechanisms that might drive these tumors, we will analyze specific histone modifications in the epigenetic landscape of liposarcomas and related cell types by Illumina Infinium arrays from the same individuals. Specifically, genome-wide chromatin profiles of histone H3 trimethylated at lysine 4 (H3K4me3), lysine 27 (H3K27me3) and lysine 36 (H3K36me3) will be studied to identify promoters, sites of polycomb repression and coding and non-coding transcripts, respectively.³² Chromatin state maps will be examined in embryonic stem cells (ESCs), bone marrow/adipose tissue derived mesenchymal stem cells (MSCs), adipoblast and adipose tissue, and compared to dedifferentiated and well differentiated liposarcomas. Chromatin maps will be integrated with gene expression data in order to identify networks of genes and processes that drive liposarcoma development, as well as to identify epigenetic patterns that may resemble normal cell types in order to identify the cell type of origin (epigenetic molecular staging). Existing data from ESCs³³ and MSCs (i.e. Håkelien, Meza-Zepeda et al, unpublished) will be integrated to data generated by the project.

1.2 Cataloguing of genetic changes following exposure to MDM2 and CDK4 antagonists in vivo

Principal investigators: Maki/Barretina/Myklebost
The second set of studies will utilize samples from patients exposed to MDM2 and CDK4 antagonists in the clinic. Clinically these agents appear to induce a cytostatic rather than apoptotic response (Maki, unpublished observations). We need to understand tumor cell responses to inhibition of these pathways in clinical subjects. The biopsies will only be performed in patients in whom the risk is deemed acceptable, and will therefore pose minimal additional risk, since 2-3 passes are typically performed per biopsy already. The material will be flash frozen in liquid nitrogen and stored at -80C or below to maintain integrity of labile molecules such as RNA. At Memorial Hospital we have had an ~80% success rate in obtaining pre and post treatment biopsies from patients in clinical trials in this setting, providing the material we would expect to close the loop from basic science to translational research to the clinic and back to the translational and basic science laboratories. These samples will be subject to gene expression and epigenomic profiling to identify genetic changes (at the level of mutation or gene expression) associated with response or resistance to MDM2 or CDK4 antagonists in vivo.

1.3 Single nucleotide resolution mapping of WD/DDLPS neochromosome (NC)

Principal investigators: Thomas/Barretina
We believe that the key oncogenic events in WDLPS reside on their neochromosomes (NC). By using chromosome isolation techniques and single-nucleotide mapping, we aim to understand how the NC is formed and stabilized. Further, by integrating our high-resolution map of NC structure with transcriptome sequencing we will identify novel gain- or loss-of-function events in the WDLPS encoded by the NC. The aims of this project are to:

• Precisely map the linear architecture of the NC, including the structural chromosomal elements and their re-arrangements, and use this information to deduce the likely mechanism of assembly of the NC through analysis of breakpoints within the DNA sequence
• Characterize the functional domains of the chromosome - neocentromere/recruited centromere, telomeres – by analysis of the chromatinised NC
• Characterize the transcriptome derived from the NC, particularly with respect to any novel fusion proteins, in both cell lines and primary tumors
• Functionally characterize any novel fusion transcripts in our cell lines/primary cells using either knockdown with a fusion transcript-specific siRNA or overexpression. The consortium so far has experience with the detection of fusion transcripts by sequencing, using RNA-seq of liposarcoma (Broad Institute, Peter Mac) and osteosarcoma (Oslo) and cell lines, and are currently performing the validation of these by PCR. Many "private" (I.e. present in only one sample, so far) fusions are detected, but the studies will concentrate on those validated in several samples, not detected in normal samples, and also present in a panel of clinical samples. Our initial sequencing has identified 8 fusions present in the genomic and transcriptome sequence data, only one of which is in the correct orientation (ie, head-to-tail). This fusion (t4;12; CPZ-CPM) is in the correct orientation, but the 3’ fusion partner is not in a coding sequence (3’ UTR). To date, we have not identified candidate fusion genes for validation in primary tumors or functional characterization.

However, depending on apparent biological significance, a moderate number of validated fusion transcripts will be pursued in functional studies. These will be of two main types, knock-down, using siRNA or shRNA, and knock-in, using expression constructs. Fusion transcripts have the promise of cancer-specific knock-down, as they are not present in normal cells, but designing breakpoint-spanning siRNAs may be problematic due to limitations in the sequence context. We expect this to be possible at least in vitro, where requirements for specificity relative to normal cells is limited. Furthermore, if one of the partner genes is not, or only weakly expressed, this part of the fusion could be used in this setting.

The studies will be used in cell lines expressing the fusions in question (current screens utilize T778 and T449b, but can be extended to other lines), as they would be expected to be dependent on the candidate oncogene, and our immortalised but non-tumorigenic hMSC line (Tenstad et al. Lab on a chip 2010) will be used as the most relevant normal system, e.g. for the knock-n experiments.

In the first aim of the project, we have used flow sorting to purify neochromosome isoforms from well characterized WD/DDLPS cell lines, and have used SNP array-based CGH and massively parallel sequencing (MPS) to begin to develop a nucleotide level map of the linear architecture of the NC, including the structural and functional chromosomal elements, and to use this information to deduce the likely mechanism of assembly of the NC. A key strength of the chromosome isolation techniques used in this project is that the chromosome is isolated in a chromatinised form. This enables us to utilize domain-specific protein binding to further isolate DNA sequences associated with the telomeric and centromeric regions on the NC. In our preliminary data we have made significant progress in analysis of one NC isoform, and have proposed hypotheses related to the NC assembly, telomere and centromere capture, and identification of fusion boundaries within the NC structure, some of which are within the coding or regulatory sequences of known genes.

The second aim of the project involves applying MPS to analysis of the transcriptome of the same cells, with the knowledge of the DNA sequence of the NC informing our identification of aberrant transcription as derived from the NC, or the normal genome. All findings in the cell line models will be validated in a large panel of already collected primary WDLPS tumor samples. Finally, we describe an aim of functionally characterizing any aberrant transcripts using either knockdown in a tumor cell with a fusion transcript-specific siRNA, or overexpression of a fusion transcript in a normal cell. 

2. Cell biology of WD/DDLPS

Principal investigator: Myklebost/Maki

2.1 Stem cells and WD/DDLPS

WD/DDLPS make up a continuum of malignancies, ranging from almost benign lesions closely resembling adipose tissue, to aggressive undifferentiated and highly metastatic tumors. It seems likely that the cancer stem cell model will enable an understanding of and ability to manipulate the dynamics of self renewal and differentiation in sarcomas. The immature mesenchymal phenotype closely resembles that of invasive metastatic cells, and also in epithelial cancers, is closely associated with aggressive and metastatic disease.

Within this part of the project we will investigate stem cell aspects of WD/DDLPS, both to gain deeper understanding as a base for completely novel therapeutics, and because some of these pathways are connected to those of MDM2, p53, and CDK4, and give opportunities for synergistic treatments. In particular, the systematic coamplification of HMGA2 and MDM2³⁴ indicates a mechanistic connection. This project will be exploratory during this first phase.

It now appears that HMGA2 and Let-7 are important in many types of cancer, and that their balance determines whether a cell will be stem like, invasive and self-renewing, or differentiates into mature and less malignant cells. We therefore will investigate how overexpression of the truncated HMGA2 protein affects the phenotype of stem-like mesenchymal cells, and also to what extent Let-7-based therapy can be used to target LPS, at least those that do not contain the insensitive, truncated HMGA2 gene. Preclinical testing of let-7-based therapy in DDLPS in vitro and in vivo models will be done in collaboration with the microRNA therapy company Mirna Therapeutics (Austin, TX) and may support the tasks defined here. Another objective is high content screening for small molecular drugs affecting cellular let-7-levels, but this will require an extended budget and time.

Being part of a Cancer Stem Cell Innovation Centre (CAST), we have initiated studies of stem cell properties in LPS as a means to isolate and characterize subpopulations of such dangerous cells. Stem-like cells have been isolated by their ability to export xenobiotics (Hoechst dye-exclusion), by metabolic properties (aldehyde dehydrogenase activity), by specific surface markers, and by slow proliferation (label-retention), and subpopulations assayed for colony-forming ability in vitro, the best surrogate for the demanding and time-consuming animal studies, and the tumor-initiating ability of positive populations verified in immunodeficient mice. Such stem-like cells from other sarcoma xenografts have increased colony-forming ability, express higher levels of HMGA2, lower levels of Let-7, and have increased tumorigenicity (Myklebost et al. unpublished). To date, FU-DDLS1, GOT3, and T778 contain tumor-initiating cells that may be studied.

• The collection of DDLPS cell lines will be assayed for tumor-forming ability in NOD/SCID mice.
• Available stem cell phenotypes will be used to identify cancer stem cells in tumor-forming lines. Stem-like fractions will be assayed for colony- and tumor-forming abilities.Suitable markers for the isolation of CSCs from sarcomas have not been solidly defined, but we have made significant progress in this regard. We have successfully used CD133 in combination with ALDH to enrich colony-forming cells from liposarcoma. Utilizing as few as 100 of these CD133+ve/ALDH+ve cells we can establish tumors in immunodeficient (NOD/SCID) mice. Since we are studying engraftment in nude mice, we can  separate human cells from mouse by species-specific antibodies. In addition, shortly we will have a colony of EGFP-NOD/SCID mice, so that host cells can be directly identified through fluorescence.
• Cellular sub-populations containing stem-like cancer cells will be characterized by transcriptome, epigenome and genome profiling in search for specific markers and possible therapeutic targets.

The let-7 microRNA represses HMGA2 as well as a large number of growth-promoting genes, including ras, as well as a number of cell cycle genes, and it appears that let-7, being low in embryonic cells, maintains the differentiated state and protects mature tissues against oncogenic activation.

• Reporter constructs that allow subpopulations of cells to be isolated based on Let-7 levels are available, and improved versions are being made, which will be used to isolate cells based on their let-7 levels. The subpopulations of sarcoma cells with low let-7 have been shown to be more stem-like, and more such samples will be investigated and characterized further.
• The effect of let-7 mimics on DDLPS cells will be investigated, with emphasis on malignant (stem-like) phenotypes and differentiation propensity. These studies have direct preclinical potential, and part of these studies will be done in collaboration with Mirna Therapeutics.
• Cells transduced with the let-7 reporter will also at a later stage be used for high-content drug screens to identify let-7 agonists.

A complementary strategy, is focused on the molecular & cellular consequences of knock-down of HMGA2 in those cells where it is amplified and rearranged to escape repression by let-7. As yet, it is unclear whether the significance of loss of the 3’ end of the gene primarily is this escape from repression, or if the properties of the truncated protein contributes to the oncogenic phenotype. Our results so far indicate that the truncated protein does have significantly altered properties, at least in one mesenchymal stem cell model,¹⁴ but the consequences on tumor properties is as yet unclear.

Based on the assumption of oncogene addiction, we hope that knock-down of HMGA2 may have therapeutic potential, perhaps by counteracting stem cell programming. However, in any case these studies will provide insight into the pathways affected by this aberration, which we suspect interacts with that of MDM2 and p53. These studies will be done on available cell lines with amplified MDM2 and HMGA2, as well as our new, improved immortalized adipogenic mesenchymal cell line,³⁵ and both expression and epigenome profiling and phenotype assays will be performed.

• Lentivirus constructs containing inducible transgenes for wild type and truncated HMGA2 will be made, and used to transduce immortalized mesenchymal "stem cells" and the effect on cellular programming investigated by transcriptomic and epigenomic profiling, as well as differentiation potential and selected phenotype assays.
• DDLPS cells will either be transduced with lentivirus containing shRNA constructs targeting both wild type and truncated HMGA2, or transiently transfected with the same shRNAs, to knock-down endogenous HMGA2 levels. The phenotypes will be characterized with a focus on stem cell programming, differentiation, and stem cell markers, but global profiling will also be done.

2.2 Functional screens for genetic modifiers of sensitivity to CDK4 and MDM2 inhibition in WDLPS

Principal investigators: Thomas/Maki/Myklebost
Clinical studies into the use of MDM2 and CDK4 antagonists in patients with WD/DDLPS are under way (see below).  Resistance to MDM2 and CDK4 antagonists is likely to be an important clinical issue, based on experience with previous targeted therapies. It is therefore critical to begin studies to understand the mechanism of action of these agents in WD/DDLPS, so that we can anticipate likely mechanisms of resistance and develop biomarkers to predict response.

High-throughput functional genomics strategies are available to systematically map genetic modifiers of sensitivity to MDM2 and CDK4 antagonists. We will utilize the following resources to screen for genetic modifiers:

• We have established a 70,000 clone shRNA library in the T778 liposarcoma line, and already conducted a screen for genes whose silencing is associated with resistance to Nutlin 3a.
• We have excellent and representative cell models as described above to conduct the screens. These lines (T778, T449b, T1000, GOT3) recapitulate the genomic events seen in primary WD/DDLPS (Thomas, unpublished).
• These results will be mapped against the genomic profiles established in primary tumor studies to determine the relevance of resistance pathways in vitro to drug effects in human subjects (see section 1 above).

Characterization of MDM2-dependent pathways in liposarcoma. The following studies will utilize Nutlin 3a and CDK4 antagonists. A genome-wide screen for modifiers of sensitivity to MDM2-antagonists utilizes the Open Biosystems lentiviral hairpin library and positive selection for resistance to MDM2 and CDK4 antagonists in vitro. This library has already been introduced into the T778 cell line. The T778 cell line containing the lentiviral shRNA library has been subject to selection at 5M Nutlin 3a after 14 days. This dose has been selected as the minimal effective dose at which no spontaneous resistance is observed, assayed up to 42 days following selection of the parent line. At this concentration, nutlin 3a induces senescence and to a lesser degree apoptosis. In our assay development screen of 1000 shRNA clones, T778 clones expressing an shRNA to p53 demonstrated marked resistance to the anti-tumor effects of nutlin 3a.

These conditions have been used for the 70,000 shRNA screen, undertaken in 7 independent pools of 10,000 clones each. Resistant clones have been identified by amplification and sequencing of common sequences in the vector flanking the hairpin sequence. Our preliminary analyses indicate that p53-shRNA was enriched 8,000-fold in the population following selection, and ranked the 8th most enriched hairpin. Ontologic analyses of the top 100 shRNAs over-represented following selection indicates that several additional genes lie within the p53 pathway, consistent with expected results, including some not previously identified as linked to MDM2-p53 interaction. We therefore believe that the screen has successfully identified multiple putative resistance candidates which may mediate resistance in the clinic to MDM2 antagonists. These candidates will be subject to a secondary screen to verify the identified candidates using the shRNA vectors predicted to cause resistance from the primary screen. The efficacy of these hairpins will also be verified in T449B, GOT3 and T1000 cells (MDM2 amplified), as well as SW872 cells as a control (MDM2 normal copy number, p53 mutant). A tertiary screen of between 15-30 candidates will be undertaken using siRNA from the Dharmacon library, which will confirm the specificity of the gene predicted to be targeted in the primary and secondary screens.

Candidates validated in all three screens will be functionally characterized for their effects on p53 levels and activity, according to existing data on their role in regulating p53-dependent tumor suppressor activities. In addition, the role of identified modifiers/interactors in the tumours suppressor pathways activated by MDM2-blockade will be further characterized to determine whether the identified genes from part a) are transcriptionally activated downstream of p53; and to characterize the senescence-like transcriptional profile induced by activation of p53 in this system, and distinguish this from transcriptional profiles activated that lead to apoptosis. These studies make use of gene expression profiling of T778 cells treated with nutlin 3a at varying concentrations and at varying timepoints (Thomas, unpublished).

3. Animal models of WD/DDLPS

3.1 Generation of genetically defined models for WD/DDLPS

Principal investigator: Thomas/Myklebost
We aim to develop an in vivo model for WDLPS, by using inducible transgenes for CDK4 and MDM2. We have created two transgenic mice on a C57/Bl6 background. The transgenes are described in Figure 5, cloned into the ROSA26 locus. The UbiC promoter is strong and ubiquitously expressed, and should result in high levels of expression of both transgenes. The GFP and mCherry reporters will enable ready detection of transgene expression in vivo and in vitro. We will obtain fat-specific FABP4-Cre mouse from Jackson laboratories.

Murine embryonic fibroblasts will be generated from singly and compound heterozygote mice. These cultures will be subject to ex vivo activation by infection with viral vectors expressing the Cre recombinase. The activation of the transgene will be monitored by flow cytometric measurement of GFP and mCherry, and verified by Western blotting. The functional consequences of transgene activation on immortalization, cell proliferation, apoptosis, and differentiation in vitro will be assayed.

Mice will be crossed as single and compound heterozygotes onto a background of the FABP4-Cre mouse. The intention is to activate transgene expression in adipose tissue in vivo, and characterize the spontaneous evolution of tumors. These tumors will be evaluated in vivo by small animal PET and CT scanning, prior to isolation. The tumours will be analyzed histologically for resemblance to human disease, and then analyzed on differing genetic backgrounds (CDK4, MDM2, CDK4/MDM2) for differential gene expression and genomic profiles. To recapitulate stochastic activation of the transgenes, we will inoculate singly or compound heterozygotes with viral vectors expressing the Cre recombinase. Injections will be placed into orthotopic sites (perirenal fat pad, quadriceps), and the evolution of tumors monitored and characterized.

3.2 Establishment and validation of xenograft models for WD/DDLPS

Principal investigator: Myklebost/Thomas
Due to the heterogeneity of human tumors, a number of preclinical models will be required to capture some of the variation observed among patients. Part of our effort will therefore be on collecting and establishing an extensive panel of models, both in vitro cell lines, and in vivo human models in the form of patient samples transplanted directly to immunodeficient mice. Cell lines usually keep most of the genetic aberrations from their tumor of origin, but usually acquire new mutations and irreversible programming that severely affect their response to many drugs.^36-38 Although cell models are cost-effective, convenient, and very useful for mechanistic studies, xenografts are better predictors of the response in patients, due to their 3D growth pattern, interaction with host stroma, usually slower growth rate, and higher degree of heterogeneity. This work may be initiated during the proposed grant period, but would need additional funding to be completed. A similar effort is now nearing completion as part of the EuroBoNeT consortium, in which such a panel has been established for osteosarcoma.^{39, 40}

Currently we have 8-10 LPS cell lines (T1000, GOT3, T778, T449b, FU-DDLS-1, LPS141, DDLS8817, LP6, etc) and 7 direct tumorgrafts. Further direct LPS tumorgrafts will be initiated as part of this project. A problem is that the WDLPS type does not appear to grow as tumors even in the most immunocompromised mice. However, DDLPS does grow in some cases, and are likely to represent a large part of the biology of WDLPS. Furthermore, the clinical challenge is when the WDLPS dedifferentiate, so such models may anyway reflect the most important properties.

V. Future Directions

The WSN will be regularly informed of the progress of these studies. The WSN meets approximately twice per year (at the Connective Tissue Oncology Society (CTOS) meeting in November, and on one other occasion—either the American Society of Clinical Oncology (ASCO) conference or at European Society for Medical Oncology (ESMO) conference). The WSN is committed to supporting global approaches to translational and clinical research into sarcomas. This study constitutes a prime example of the sort of research supported by the WSN. We anticipate that future clinical studies arising from this project may involve either the WSN itself, or members of the WSN. Finally, we recognize that this project represents a beginning, rather than a final contribution to the study of WD/DDLPS. Our strategy for sustainability will be to utilize the results of the current program of work to leverage competitive grant funding for the project team. Funding sources which may support future research will include (but are not limited to) the European Union, National Institutes of Health and the National Health and Medical Research Council in Australia. Such future research will include building on newly identified candidate genes of importance, utilizing the preclinical models developed here.