$100K Grant Funds International Research on Rare Sarcomas

The Team

David Thomas, FRACP, PhD
and Mandy Ballinger, PhD

September 2, 2014 - The Liddy Shriver Sarcoma Initiative is pleased to award a $100,000 grant to David Thomas and Mandy Ballinger, two researchers working on the International Sarcoma Kindred Study (ISKS). This grant will allow the investigators to study eight especially rare forms of sarcoma within the established framework of the ISKS. These sarcoma subtypes include: angiosarcoma, chordoma, clear cell sarcoma, desmoplastic small round cell tumor, endometrial stromal sarcoma, epithelioid sarcoma, mixed mullerian mesenchymal tumor and perivascular epithelioid cell tumor (PEComa).

While the International Sarcoma Kindred Study was initially designed to study all sarcomas, the investigators involved know that research on the rarest sarcomas is critically important. Obtaining enough data and tissue samples to study these diseases is a major obstacle, and international collaboration is needed to move forward. The investigators explain: "Only by utilising a global resource will it be possible to subject rare cancers to meaningful analysis." The researchers hope to identify genes that shed light on why the rarest sarcomas form and how they develop.

Research on Rare Cancers

By David Thomas, FRACP, PhD and Mandy Ballinger, PhD

Video: Challenges in Sarcoma Research

Outcomes for patients with sarcoma are less than ideal, despite great progress over the past 30 years. Too many people die from these diseases, and too many survivors live with lifelong consequences of their treatment. In the absence of research, progress stagnates. Only by systematically studying sarcomas do we have a reasoned prospect for improving outcomes for individuals and families affected by sarcoma. This is the essence of research. Because sarcomas are rare, it is critical for clinicians and researchers to take every opportunity to learn from each case.

The Funding

The Initiative gratefully acknowledges the following generous support: donations made to angiosarcoma research in memory of Harvey Levitt and Nellie Vecchio, donations to chordoma research in memory of Alison Norcross and from the Michael Torrey family, donations to clear cell sarcoma research in memory of Steve Byrne and Nate Schull, donations to DSRCT research in memory of Jonathan Swartz, donations to endometrial stromal sarcoma research in memory of Karen Bloom, donations to epithelioid sarcoma research in honor of Andrew Moore and from the James & Lisa Curry Family, donations to MMMT research in memory of Bonnie Botwright, Sharon LaVelle, Joan Morse and Lilia Sousa, and donations to PEComa research in memory of Bill Sigler.

The International Sarcoma Kindred Study: Rare Sarcoma Initiative


There are over 130,000 new cases of sarcoma world‐wide each year, accounting for approximately 1‐3% of all malignancies.

Sarcoma has a strong genetic component. Sarcomas disproportionately affect the young, representing 20% of cancers in children and 10% of cancers <30 years of age. Early age of onset increases the likelihood that genetic factors are present. Sarcomas arise on average 20 years earlier than epithelial cancers. Sarcomas are more common in persons with recognized hereditary cancer syndromes, including retinoblastoma, Li‐Fraumeni syndrome (LFS), Gardner's syndrome, Werner's syndrome, neurofibromatosis type 1, and some immunodeficiency syndromes.1 Up to 33% of pediatric sarcomas are estimated to be associated with a significant family history of cancers.2,3 The risk of sarcomas in relatives of all children with sarcoma is increased over 6‐fold compared to age‐matched controls and, where a causal gene mutation is found, this risk increases to over 250‐fold.4 Although little data on familial risk are available in adult‐onset sarcoma, a Scandinavian population‐based study of over 800 patients showed that 20% of cases developed a second cancer at a median of 10 years after treatment, and that the hazard ratio for a second sarcoma was 17.6‐fold increased over the general population.5 The incidence of second sarcomas was increased almost 30‐fold in patients with a first sarcoma, and there was a 5.3‐fold increased risk of sarcomas in parents of patients with leiomyosarcoma.6 Because these data are entirely based on retrospective registries, they almost certainly underestimate actual genetic risk.

Current knowledge regarding genetic risk in adult‐onset sarcoma (90% of all sarcomas) is limited by the major ascertainment biases inherent in family cancer registry data. Such registries are over‐represented in common cancers with well‐defined genetic risk and where clinical interventions can modify that risk, exemplified by breast and bowel cancers.7 Existing familial sarcoma studies tend to be heavily biased towards pediatric populations,2,4,8-10 with a median age in one of the larger studies of 5 years.2 Pediatric sarcomas are biologically and pathologically different than adult‐onset sarcomas, with different genetic causes, treatments and cure rates. The median age of sporadic sarcomas is ~50 years.11 Finally, family history is an imperfect guide to genetic risk. Twenty-Five to 50% of sarcoma probands with germline TP53 mutations lack a family history in prior generations,3,12 similar to early onset breast cancer.13 These may be due to 'de novo' mutations, gonadal mosaicism in parents (with implications for siblings), or to recessive alleles. More precise identification of those patients at genetic risk for sarcoma are required to better understand the etiology of this rare disease and most importantly, to design novel treatment and effective prevention strategies.

These factors led to the formation of the International Sarcoma Kindred Study (ISKS). The ISKS has been recruiting since 2009, and has progressively increased the number of sites of accrual (Table 1).

Table I: Current Recruitment Progress
(datalock July 2014)

Table I

Our initial genomic studies have included over 730 participants affected by sarcoma. These have focused on the common sarcoma subtypes, but have included a subset of rarer subtypes of sarcoma.

Table II: Pathologic Sarcoma Subtypes in ISKS
and the subset of ISKS currently sequenced

Table II

We have used a targeted exome panel as the basis for screening the cohort, using an Agilent custom Halo platform (Table 2). A total of 104 genes have been sequenced, including genes linked to hereditary breast, bowel, and ovarian cancer syndromes as well as a subset linked to specifically to subtypes of sarcoma. From our experience, a minimum of 30 cases of each subtype will be required to draw any conclusions between individual subtypes and putatively pathogenic variation. The exact number depends on the molecular heterogeneity within each sarcoma subtype, and the assumption that pathogenic variation in individual genes will be enriched markedly in those subtypes.

Specific Rare Sarcoma Subtypes of Interest

The tumors that are the focus of this grant are rare, and typically present major problems clinically. They are often chemo- and radio-resistant. Familial patterns are not widely recognized, although like most sarcomas they affect a younger population than usually affected by cancer in our community. Accurate statistics on their frequency are not easily obtained, and there is little known about familial clustering.

Angiosarcomas comprise a spectrum of tumors bearing markers of endothelial differentiation. They may vary from relatively indolent to highly aggressive forms, and may arise in the head and neck in the elderly, in the liver in association with environmental exposures, and may also arise in women with a prior history of breast cancer within radiation fields. Angiosarcomas are genetically complex. It is likely that angiosarcomas represent a heterogeneous group of tumors. They are typically resistant to most forms of chemotherapy, although taxanes have recently shown some promise.

Chordoma is a rare tumor (affecting fewer than 1/million) typically arising in young adults in the midline, and most commonly in the pelvis. While cytogenetically diploid as a rule, they frequently carry mutations in brachyury gene. They are chemo- and radio-resistant.

Clear Cell Sarcoma (CSC) is also known as melanoma of soft parts and it may arise at any location. It is associated with translocations affecting the EWSR1 and ATF1 genes. It is a highly aggressive tumor, which is typically chemo- and radio-resistant.

Desmoplastic small round cell tumor (DSRCT) is a rare variant of primitive neuroectodermal tumor that usually affects young men and occurs most frequently in the abdomen. It is associated with a fusion between EWS and WT1. It has an extremely high lethality and presents a major clinical problem.

Endometrial Stromal Sarcoma (ESS) is a typically low-grade endometrial sarcoma, occurring in about 1/million of the population. There are no specific genetic anomalies associated with ESS. It is strongly estrogen dependent, but loses hormone sensitivity with increasing grade.

Epithelioid sarcoma (ES) comprises less than 1% of all sarcomas and is one of a group that express epithelial markers. They are genetically complex, although somatic mutations in SMARCB1 has been reported. This tumor frequently arises in the lower limbs, and spreads regionally. It is typically chemo- and radio-resistant.

Mixed Mullerian Mesenchymal tumor (MMMT) is also known as carcinosarcoma, and is most commonly a uterine sarcoma with epithelial and stromal elements. Risk factors include hormonal therapies and obesity. No specific genetic events are associated with MMMT, but the PI3K and TP53 pathways have been implicated. MMMT are highly malignant with a 5 year survival of 20%.

PEComa (tumors with perivascular epithelioid differentiation) are associated with tuberous sclerosis, although most are sporadic. They are more commonly benign than malignant. About 200 cases have been reported, more commonly in females than males. They arise most commonly in the abdomen. Genetically complex, they frequently show mutations in TSC2, which may have therapeutic implications.


Our hypothesis is that these distinct, rare subtypes of sarcoma will correspond to distinct genetic risk alleles. Our object is to encourage enrolment of the following sarcoma subtypes to the ISKS, and to undertake genomic analyses of these cases to identify subtype-specific enrichment in deleterious germline variation. Our goal is to sequence a minimum of 30 cases of each.

Research Plan

Targeted Recruitment

We have established an international, population-based case-control kindred study, whose goal is to understand the hereditary genetic basis and clinical risks associated with adult-onset sarcoma. ISKS Australia began recruitment in July 2009 and since then study sites have become active in India, France, New Zealand, USA, Canada, and the UK. Working together, each local study center will contribute data and samples from ISKS participants to the "Global Study Centre" located in Australia. These samples, with corresponding data, will then be freely available for participating ISKS members in order to achieve the objectives of each specific aim listed below. Each aim covers a different aspect of sarcoma risk and clinical outcome, and will best be answered through the continued enrollment of patients through ISKS.

The ISKS will encourage recruitment of these rare sarcoma subtypes through the ISKS network. For some of these subtypes (eg PEComa), it may not be feasible to recruit 30 cases over the next 2 years, but the goal will be to create a repository that will ultimately achieve the numbers to enable genomic studies in these rare diseases.

Genomic Analyses

An Agilent Haloplex custom capture panel will be used, as previously described above. This panel has excellent and robust operating characteristics, and has been extensively tested already. Sequencing will be performed on an Illumina HiSeq 2500. The details of the sequencing costs are outlined in the budget section below.

Bioinformatic Analyses

The power of next-generation sequencing of cancer gene panels is the identification of rare genetic variants that may contribute to cancer risk. Taking each group separately, we have two sources of controls: other sarcoma cases, and unaffected populations. The goal is to compare the frequency of pathogenic variants in individuals affected by these rare sarcoma subtypes with these control populations.

There are two statistical challenges. The first is that the case populations are very small (30 cases). The power of comparisons is limited by numbers, unless the influence of specific genes or pathways is very strong within that sarcoma subtype. This assumption is critical to the success of the approach being taken here. The second challenge is that it is unlikely that the comparisons can be made at the level oif individual alleles, since these are rare. For example, it is unlikely that a specific base substitution in gene x will be identified enough times within the 30 cases, such that the representation of that allele will be statistically enriched by comparison with even large numbers of controls.

Our approach to this problem is two-fold. First, we will bin individual pathogenic variants by each gene, and also at the pathway level. The numerator for comparisons will not be the allele, but rather the sum of pathogenic alleles for each gene, and if suitable, for multiple genes within a well-defined pathway. An example of the latter would be the BRCA complex, which is known to comprise BRCA1, BRCA2, BARD1, BRIP1, FANCD2, FANCI and RAD51. In this case, we will bin the total number of pathogenic alleles affecting genes within these genes, and compare that number with a) individuals within the ISKS cohort who do not have that sarcoma subtype; b) individuals who do not have cancer at all.

There are some important technical issues in relation particularly to the comparisons with individuals who do not have cancer at all. This comparison is important, because it may be the case that some of the genetic risk associated with any individual sarcoma subtype may be shared with other sarcoma subtypes, and so not be apparent in comparisons with the remainder of the ISKS cohort. However, if this is the case, we should still detect enrichment in pathogenic variation by comparison with a control population.

The power of statistical comparisons is strengthened by having good numbers of controls. Our controls include:

  1. the remaining ISKS cohort (which will ultimately include over 1000 other cases performed on the same platform as for the subtype of interest);
  2. ISKS controls (we have about 230 individuals who do not have sarcoma, who constitute an internal control set genotyped on the same platform as all cases);
  3. external datasets, usually much larger in size, which have been genotyped on different platforms:
    1. the NHBLI exome variant server dataset (4300 European American cases, ascertained for non-cancer phenotypes, genotyped using whole exome sequencing approaches);
    2. A dataset of 970 Australian individuals ascertained on bone mass phenotypes, and genotyped using whole exome sequencing;
    3. A dataset of almost 5000 elderly well Australians who have never had any cancer, who will be genotyped using whole genome sequencing.

The non-ISKS controls have been genotyped on different platforms, which introduces important biases in ascertainment of genetic variation. For example, coverage of some genes differs using whole exome, whole genome or targeted exome approaches. These biases, uncorrected, could influence representation of pathogenic variation in the comparisons. There are two methods for controlling for this. The first is the use of a number of controls, so that enrichment of pathogenic variation is measured independently against a number of control populations. Enrichment against any one control population not found in other controls would be regarded as suspect. The second is to measure heterogeneity between control populations, and also within the ISKS cohort as a whole, to ensure that ethnic stratification does not confound comparisons. The third is to use a metric to correct for ascertainment of pathogenic variation, using synonymous variation. Since the ability to detect synonymous or pathogenic variation ought to be the same within each platform, even if coverage of any individual gene may vary between platforms, the ratio of deleterious:synonymous variation ought to be reasonably constant. This ratio (termed the del:syn ratio) is used to correct for platform specific gene biases. Finally, we will mask out regions of the genome which are not covered to a similar level within any given set of comparisons.

Definitions of 'Pathogenic' Variation

Defining pathogenic variation is complex and challenging. Our approach is as follows:

  1. Quality metrics. Using a combination of read depth (>30 reads), automated quality scores (>200), and position of variant metrics, we have established parameters that ensure that 99% of variants called by the sequencing panel validate by Sanger methods.
  2. Rarity. We exclude variants that have been identified in either the ISKS cases or controls at more than 0.5% frequency, or in 1000 Genomes or the Exome Variant Server dataset.
  3. Reference to locus-specific databases for variants of any kind previously associated with a >2-fold increased cancer risk (Class I variants).
  4. Predicted effect on canonical reference transcript. The following changes are defined as automatically pathogenic: premature stop, frameshift, initiator codon mutations, or changes affecting splice donor or acceptor sequences (Class II variants).
  5.  Missense variants (the bulk of changes of interest). These are subject to in silico analysis for predicted functional impact.
  6. Any variant that otherwise meets the criteria above, for which there is evidence for neutrality, is excluded.


For any gene[s] in which we find an excess of pathogenic variation, we will source independent cases to validate those findings. These validation cases will be sourced through the recruitment sites from retrospective series, including the use of archival blocks of tumor tissue.

Impact Statement

This study will add significantly to the knowledge generated by the ISKS, and provide an impetus for selective enrichment in rarer subtypes of sarcoma. We anticipate that these rarer subtypes will be characterized by specific genetic risk markers, that are not shared by other sarcoma subtypes. These findings will shed light on the biology of these understudied sarcoma subtypes, and perhaps identify opportunities for therapy. In other situations, genetic risk markers (such as mutations in PTCH1 and BRCA1/2) have been used as biomarkers predicting responses to targeted therapy. We have already commenced a screening study (SMOC), which can be offered Australian individuals and their families identified to carry 'actionable' markers of genetic risk. Although a global study of this kind is beyond the scope of the current proposal, the establishment of a structured, research grounded intervention program provides evidence of translational intent, and will provide a foundation for future clinical programs suitable for the sarcoma population.

By David Thomas, FRACP, PhD
The Kinghorn Cancer Centre and Garvan Institute of Medical Research in Sydney, Australia

and Mandy Ballinger, PhD
Peter MacCallum Cancer Centre in East Melbourne, Australia


1. Zahm SH, Fraumeni JF, Jr. The epidemiology of soft tissue sarcoma. Semin Oncol. 1997 Oct;24(5):504-14. PubMed PMID: 9344316. Epub 1997/10/31. eng.

2. Hartley AL, Birch JM, Blair V, Kelsey AM, Harris M, Jones PH. Patterns of cancer in the families of children with soft tissue sarcoma. Cancer. 1993 Aug 1;72(3):923-30. PubMed PMID: 8334646. Epub 1993/08/01. eng.

3. Toguchida J, Yamaguchi T, Dayton SH, Beauchamp RL, Herrera GE, Ishizaki K, et al. Prevalence and spectrum of germline mutations of the p53 gene among patients with sarcoma. N Engl J Med. 1992 May 14;326(20):1301-8. PubMed PMID: 1565143. Epub 1992/05/14. eng.

4. Hwang SJ, Lozano G, Amos CI, Strong LC. Germline p53 mutations in a cohort with childhood sarcoma: sex differences in cancer risk. Am J Hum Genet. 2003 Apr;72(4):975-83. PubMed PMID: 12610779. Pubmed Central PMCID: 1180359. Epub 2003/03/01. eng.

5. Fernebro J, Bladstrom A, Rydholm A, Gustafson P, Olsson H, Engellau J, et al. Increased risk of malignancies in a population-based study of 818 soft-tissue sarcoma patients. Br J Cancer. 2006 Oct 23;95(8):986-90. PubMed PMID: 17008869. Epub 2006/09/30. eng.

6. Hemminki K, Li X. A population-based study of familial soft tissue tumors. J Clin Epidemiol. 2001 Apr;54(4):411-6. PubMed PMID: 11297891. Epub 2001/04/12. eng.

7. Manoukian S, Peissel B, Pensotti V, Barile M, Cortesi L, Stacchiotti S, et al. Germline mutations of TP53 and BRCA2 genes in breast cancer/sarcoma families. Eur J Cancer. 2007 Feb;43(3):601-6. PubMed PMID: 17224268. Epub 2007/01/17. eng.

8. Birch JM, Hartley AL, Blair V, Kelsey AM, Harris M, Teare MD, et al. Cancer in the families of children with soft tissue sarcoma. Cancer. 1990 Nov 15;66(10):2239-48. PubMed PMID: 2224780. Epub 1990/11/15. eng.

9. Hartley AL, Blair V, Harris M, Birch JM, Banerjee SS, Freemont AJ, et al. Multiple primary tumours in a population-based series of patients with histopathologically peer-reviewed sarcomas. Br J Cancer. 1993 Dec;68(6):1243-6. PubMed PMID: 8260380. Epub 1993/12/01. eng.

10. McIntyre JF, Smith-Sorensen B, Friend SH, Kassell J, Borresen AL, Yan YX, et al. Germline mutations of the p53 tumor suppressor gene in children with osteosarcoma. J Clin Oncol. 1994 May;12(5):925-30. PubMed PMID: 8164043. Epub 1994/05/01. eng.

11. Olivier M, Goldgar DE, Sodha N, Ohgaki H, Kleihues P, Hainaut P, et al. Li-Fraumeni and related syndromes: correlation between tumor type, family structure, and TP53 genotype. Cancer Res. 2003 Oct 15;63(20):6643-50. PubMed PMID: 14583457. Epub 2003/10/30. eng.

12. Chompret A, Brugieres L, Ronsin M, Gardes M, Dessarps-Freichey F, Abel A, et al. P53 germline mutations in childhood cancers and cancer risk for carrier individuals. Br J Cancer. 2000 Jun;82(12):1932-7. PubMed PMID: 10864200. Pubmed Central PMCID: 2363254. Epub 2000/06/23. eng.

13. Hopper JL, Southey MC, Dite GS, Jolley DJ, Giles GG, McCredie MR, et al. Population-based estimate of the average age-specific cumulative risk of breast cancer for a defined set of protein-truncating mutations in BRCA1 and BRCA2. Australian Breast Cancer Family Study. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 1999 Sep;8(9):741-7. PubMed PMID: 10498392.

  • Figure 1: Clonal evolution of tumors in different 'ecosystems' and the influence of SNPs.
    Adapted from Greaves and Maley 2012.
  • Table 1: TP53 mutation stratification by clinical risk criteria.
  • Figure 2: Participating Study Centers with participating local investigators.
  • Figure 3: Participant flow.