New Grant Funds Research on the Development of Leiomyosarcomas

An ESUN Announcement

The Liddy Shriver Sarcoma Initiative has awarded a $50,000 grant to fund promising leiomyosarcoma research by investigators at Brigham and Women’s Hospital and Harvard Medical School.

Leiomyosarcomas are rare aggressive tumors that can be difficult to diagnose. They do not respond well to radiation therapy or chemotherapy.

In the study, Drs. Bradley Quade and Marisa Nucci will look for the early genetic changes that lead to the development of these cancers. The doctors hope that their research will lead to better diagnostic testing and targeted drugs to treat leiomyosarcoma.

According to Dr. Nucci, the study could have significant impact. She says, "One of the main goals of the study is to identify a diagnostic marker that will enhance our ability, as pathologists and clinicians, to recognize smooth muscle tumors with recurrent or metastatic potential. This will greatly improve diagnostic accuracy and thus positively impact patient care."

Studying an Early Stage of Leiomyosarcoma

Benign uterine smooth muscle tumors (known as fibroids or leiomyomas) are a common human tumor; nearly 80% of women have at least one, and most women have several fibroids. In contrast, malignant smooth muscle tumors of the uterus are much less common, and they are often a diagnostic challenge. The researchers believe that certain fibroids, called atypical leiomyomas, are pre-cancerous. Studying these pre-cancerous tumors may provide much-needed clues about the development of leiomyosarcomas.

Dr. Quade explains, "This concept of a pre-cancer, while common in carcinomas of the breast, cervix, endometrium, and colon, is novel in sarcomas. If the concept is correct in this case, we will be able to study a very early stage of leiomyosarcoma."

According to Dr. Quade, finding this early stage cancer may be important because leiomyosarcoma, unlike many other sarcomas, is genetically very complex. He says, "When one studies the chromosomes of leiomyosarcoma, where the DNA is packaged, one finds both numerical and structural abnormalities, many of which vary from cell to cell, suggesting a great deal of 'genomic instability.' Thus, finding a 'leiomyosarcoma gene' is like looking for a genetic needle in a chromosomal haystack. By studying earlier stages, we hope that the size of the haystack is much, much smaller!"

Looking for Deleted Genes

The researchers will study a collection of atypical leiomyomas and leiomyosarcomas to see how often certain genes are deleted in the tumors and how those deletions affect the tumors' proteins.

The doctors hope to find the earliest and possibly the most important changes that lead to malignancy in smooth muscle cells. If they can identify those changes, they hope to design better, less subjective diagnostic tests and perhaps even more targeted anti-leiomyosarcoma drugs.

Meet the Investigators

Bradley Quade, MD, PhD: Both Dr. Nucci and I have studied smooth muscle tumors for many years, often as a team. Part of my interest stems from the fact that these tumors can be very difficult to classify, and we often cannot predict how they will behave. We call such tumors "smooth muscle tumors of uncertain malignant potential." It is very frustrating not being able to make a more certain diagnosis, and I hope that our studies will lead to improved tools to help us with our microscopic diagnoses. I also find this family of tumors to be fascinating because sarcomas are not thought to have precancerous lesions. Yet we recognize microscopic and biological intermediates such as atypical leiomyoma and disseminated peritoneal leiomyomatosis. Finally, at the most basic level, I am curious and would to know the answers to these very simple questions: Why are fibroids so common and (fortunately) leiomyosarcoma so rare? How are they alike? How are they different?

Marisa Nucci, MD: As a surgical pathologist with formal subspecialty training in gynecologic pathology, uterine sarcoma has been the focus of my research and clinical expertise for over 15 years. I was first drawn to this field by virtue of the fact that so little was known about the underlying mechanisms of tumor development.

Funding for this Study

This grant was funded, in part, by a very generous donation from Laura Somerville.

V10N1 ESUN Copyright © 2013 Liddy Shriver Sarcoma Initiative.

Deletion Of Type 4 Collagens And Smooth Muscle Sarcomagenesis

By Bradley Quade, MD, PhD and Marisa R. Nucci, MD

Abstract

Malignant smooth muscle tumors are characterized by complex numerical and structural chromosomal abnormalities. These abnormalities preclude discovery of an etiological gene based on positional cues. Several lines of evidence suggest that atypical leiomyoma, a histological variant with pleomorphic nuclei, may also be a biological intermediate, a pre-cancer. Further, we hypothesize that deletion of specific type IV collagen genes on the X chromosome is an important step in the development of leiomyosarcoma from atypical leiomyomas. To test these hypotheses, we will use a sensitive molecular biological tool to look for deletion of these collagen genes.

Overview

Uterine smooth muscle tumors span a spectrum from benign "fibroids," present in over three quarters of women of reproductive age, to the most common and very lethal pelvic sarcoma. Like many high grade malignancies, profound genomic instability found in leiomyosarcoma sharply distinguishes it from the minimal or absent instability found in leiomyomas. Although the high level of genomic instability hints at several potential mechanisms for malignant smooth muscle tumorigenesis, it also has been a significant barrier to discovering genes that drive sarcomagenesis using genomic-based positional methods. In between the benign leiomyoma and malignant leiomyosarcoma, we recognize histological intermediates, in which the prediction of their biological potential is very difficult. Study of these intermediates, however, may hold important clues to the early events in malignant smooth muscle neoplasia.

Taking a new tack, this project addresses the hypothesis that COL4A5 and COL4A6, genes on the X chromosome at band q22.3 encoding two α chains of type IV collagen found in extracellular matrixes surrounding smooth muscle cells, play an important role in uterine smooth muscle neoplasia. Four lines of evidence support our hypothesis that these genes act as tumor suppressors (Figure 1). First, inherited deletions cause clonal smooth proliferations in non-uterine smooth muscle tissues expressing α5(IV) and α6(IV). Second, their carboxy-terminal non-collagenous domains play a role in angiogenesis and tumor growth. Third, α5(IV) and α6(IV) polypeptides are specifically expressed in normal uterine smooth muscle, and α6(IV) loss is consistently observed in uterine leiomyosarcoma. Fourth and finally, we have observed a small somatic deletion involving parts of α5(IV) and α6(IV) in an index case in which an atypical leiomyoma progressed to leiomyosarcoma. Validation of this hypothesis by the proposed molecular and immunohistochemical methods hopefully will open new approaches to diagnose and treat leiomyomsarcoma.

Hypothesis and Specific Aims

Simply stated, the α5 and α6 chains of type IV collagen, expressed from genes arranged head-to-head at Xq22.3, regulate uterine smooth muscle proliferation, and deletion of one or both genes from the active X chromosome is a step towards development of malignant uterine smooth neoplasia. The Specific Aims to test the foundations of this hypothesis are:

1. Determine the frequency and location of COL4A5 and COL4A6 deletion(s) in uterine leiomyosarcoma, and compare them to the frequencies and locations of deletion(s) occurring in uterine leiomyoma, atypical leiomyoma, and Smooth muscle Tumors of Uncertain Malignant Potential (STUMPs).
2. Determine the X inactivation status of undeleted COL4A5 and COL4A6 alleles, providing mechanistic insights.
3. Compare genomic copy number status with protein expression in smooth muscle tumors using a tissue microarray and immunofluorescent microscopy or immunohistochemistry.
4. Determine the status of COL4A5 and COL4A6 genomic copy number and protein expression in two leiomyosarcoma cell lines obtained from ATCC as a prerequisite to the development of a faithful in vitro model for smooth muscle neoplasia.

Rationale: Background, Impact, and Preliminary Data

Uterine leiomyoma, also known as fibroids, are benign smooth muscle tumors (Figure. 2). They are the most common human neoplasm; ~80% of hysterectomy specimens, regardless of the indication for surgery, have at least one and usually multiple leiomyomas. In contrast, uterine leiomyosarcomas are uncommon, but highly malignant smooth muscle tumors. Survival rates for uterine leiomyosarcoma depend mostly on stage; the 5-year survival rate for tumors contained in the uterus (stage I) is 60%, but falls to 28%-15% after regional or distant metastasis (stage III or IV). Surgery is the mainstay of therapy, as leiomyosarcomas are quite refractory to radio- or chemotherapy. Pathological diagnosis primarily is based on histological recognition of atypia, proliferation, and a particular pattern of necrosis (Figure 2).

Although histological variants of leiomyomas are generally regarded as benign, our laboratory has shown that “atypical leiomyoma” have expression profiles and chromosomal abnormalities much closer to leiomyosarcoma than to myometrium and typical leiomyoma, suggesting that they may be both morphological and genetic intermediates on the path to malignancy (Figure 3).^1,2 Furthermore, it is not always possible to predict the clinical behavior of tumors with histological features close to the criteria for malignancy, and these are termed Smooth muscle Tumors of Uncertain Malignant Potential (STUMPs). Generally, the risk of local recurrence or metastasis following the diagnosis of STUMP is about 7%, but our unpublished studies suggest that addition of non-canonical criteria such as identification of atypical mitotic figures (reflecting genomic instability) or infiltrative borders with the adjacent myometrium (reflecting an aggressive biological phenotype). These additional criteria can significantly boost the prognostic power of STUMP such that 30% of tumors so-classified will recur or metastasize, which is less than but much closer to figure observed for leiomyosarcoma.

RB1 and TP53 are implicated in leiomyosarcoma pathogenesis, but discovery of additional genes by positional techniques has been hampered by high levels of genomic instability. In essence, the problem is that it is difficult to find the “genetic needle” in a vast and abnormal “chromosomal haystack”.

In summary, there is a great need to improve diagnosis and treatment of uterine leiomyosarcoma, but advances will only come from a deeper and more sophisticated understanding of leiomyosarcoma tumor biology.

Type IV collagen is a major constituent of the extracellular matrix known as basement membrane. While usually thought of as the adhesive substrate for epithelium, it also plays key roles in other tissues such as the glomerular membrane. Inherited mutations in the α3(IV), α4(IV), or α5(IV) chains result in hereditary nephritis, also known as Alport syndrome. Although highly homologous, six genes located on three different chromosomes code for distinct α(IV) chains. Each pair is arranged in a head-to-head fashion, such that their transcription start sites are in close proximity and transcription uses opposing strands. Three α(IV) chains spontaneously trimerize into protomeric tropocollagen subunits (Figure 4). Type IV tropocollagen does not assemble into fibrous polymers like prototypical type I collagen; rather, it forms a complex 3-D meshwork. Only three combinations (α1α1α2, α3α4α5, and α5α5α6) of the 56 potential heterotrimers occur in vivo. This specificity is mediated by the globular carboxy-terminal non-collagenous (NC1) domain. Tropocollagen formation begins with NC1 trimerization, followed by supercoiling of the interrupted triple helical domains. Outside the cell, the 3-D lattice is assembled by dimerization of NC1 trimers between two tropocollagens and tetramerization of amino-terminal 7S domains between four tropocollagens.

The first line of evidence to indicate that COL4A5 and COL4A6 on Xq22.3 play a role in regulating smooth muscle proliferation comes from the study of X-linked Alport syndrome (ATS). Most cases are due to mutation or deletions of COL4A5, but a subset of individuals also suffers from vulvar, esophageal and, occasionally, tracheobronchial leiomyomatosis.^3,4 Diffuse leiomyomatosis with Alport syndrome (DL-ATS) is a contiguous gene deletion syndrome that occurs when deletions of COL4A5 extend into the 5’ of COL4A6.^5,6 Genomic mapping of DL-ATS deletions defines a minimal overlap of 4.2-kb including exons 2, 1’ and 1 of COL4A6 and exon 1 of COL4A5, as well as their promoters and any shared regulatory elements in their overlapping 5’ regions.^7-11 Of note, somatic deletions also can be found in non-syndromic esophageal leiomyoma.¹¹ α5(IV) and α6(IV) immunofluorescent staining is lost in ATS-associated esophageal leiomyomatosis; the esophageal predilection is explained by localization of α5(IV) and α6(IV) chains in esophageal, but not in gastric or intestinal muscle.¹² Syndromic esophageal smooth muscle tumors are clonal proliferations.¹² Although the mechanism remains to be elucidated, loss of the shared promoter region and 5’ exons clearly are linked to smooth muscle proliferation, and α5(IV)2α6(IV) protein deficiency is marker of DL-ATS.

The second line of evidence recognizes a plausible relationship between tumor growth and type IV collagen NC1 domains. These fragments have been given other names in recognition of their non-structural properties: arrestin, canstatin, tumstatin, and hexastatin for α1(IV), α2(IV), α3(IV), and α6(IV) respectively.^13,14 NC1 domains liberated by proteolysis inhibit endothelial cell protein synthesis, proliferation, migration, and tumor growth. Free NC1 domains interact with endothelial and tumor cells by binding integrin adhesion receptors and triggering apoptosis,¹⁵ but whether this is the mechanism by which the NC1 domain inhibits tumor grow remains to be determined.

The third line of evidence is that α5(IV) and α6(IV) are both highly expressed in uterine smooth muscle, but consistently lost in uterine leiomyosarcomas. α5(IV) and α6(IV) proteins are found in the smooth muscle of numerous organs, including the uterus. ^16-18  GeneAtlas U133A mRNA expression profiling (Figure 5) shows that both genes are expressed at very high levels (nearly 30-fold above the median of all tissues surveyed) in the uterus. Figure 6 is a preliminary study we performed using immunofluorescence microscopy to illustrate high α5(IV) staining intensity in myometrium.

In a study comparing gene expression profiles of uterine leiomyosarcoma and leiomyoma relative to myometrium, we found that COL4A6 was increased by a factor of 2.08 in uterine leiomyoma, but was sharply decreased by a factor of 14.3 in leiomyosarcoma (ref. 2). Of note, this particular comparison generated the highest ANOVA p value in the study (p = 4.98x10-4), which indicates that COL4A6 expression was consistently reduced. COL4A5 also was reduced (p = 4.37x10-2) reduced, but only by a factor of 2.71 in leiomyosarcoma. Potentially indicating specificity, COL4A1 and COL4A2 mRNA levels in leiomyosarcomas were only half that of myometrium (p>0.05).

The fourth and final line of evidence comes from an unpublished case we are following. In brief, this individual presented with a “fibroid” that was removed by myomectomy, classified histologically as an atypical leiomyoma, and managed by conservative observation. Four years later, the patient returned with leiomyosarcoma. Array comparative genomic hybridization (aCGH) showed a 204,205-bp deletion with the exactly the same start and end points on Xq22.3 present in both the initial atypical leiomyoma and subsequent leiomyosarcoma, consistent with lineage continuity and tumor progression. This deletion includes exons 1 and 2 of COL4A5 and exons 1, 1’ and 2 of COL4A6, which is remarkably similar to the overlap region associated with DL-ATS (Figure 7).

Experimental Plan (by Specific Aim)

1) Determine the frequency and location of COL4A5 and COL4A6 deletion(s) in uterine leiomyosarcoma, and compare them to the frequencies and locations of deletion(s) occurring in uterine leiomyoma, atypical leiomyoma, and STUMP.

This Specific Aim is the heart of our proposal.

Multiple Ligation-dependent Probe Amplification (MLPA), one of a number of technologies we considered for determination of genomic copy number variation, was selected as the primary analytical method based on the high resolution and throughput, availability of resources interrogating the region of interest, and adaptability for evaluating DNA methylation status (for Specific Aim 2).^19-21 This technique has been used in over 350 publications investigating various aspect of cancer biology. In brief, two oligonucleotide half-probes are hybridized to adjacent sequences on the same genomic DNA strand of the target (Figure 8). Next, the 5’ and 3’ half-probes are ligated to form a single longer oligonucleotide. Thus, the abundance of ligated probe is proportional to abundance of the target DNA in each sample. Sequences complementary to a universal PCR primer set are located at the end of each half-probe, permitting PCR amplification of ligated MLPA probe. Multiplex analysis of up to ~40 loci is achieved by including “stuffer” sequences of variable lengths between the target hybridization and PCR primer sequences in one or both half-probes. Consequently, each amplicon in the multiplex reaction can be distinguished by size using capillary electrophoresis under denaturing conditions.

Our study collection consists of FFPE tissue blocks archived by the BWH Pathology Dept., which allows confirmation of pathological diagnoses on exactly the tissues to be studied and retrospective sample collection over many years. In addition to recording our original pathological diagnoses, we customarily review the histology (individually and blindly) from every case to verify that each has been correctly classified prior to further analysis. With our deep expertise, phenotypic classification of the cases we study is as accurate and complete as pathologically feasible. We anticipate analyzing ~200 smooth muscle tumors and corresponding myometrial controls.

Despite fragmentation that may limit other genetic analyses, DNA isolated from FFPE is suitable for MLPA because each oligonucleotide probe pair hybridizes to very small (<100-nt) genomic segments.

Our previous transcriptional analysis of smooth muscle tumors using the ANOVA test has taught us a number of lessons about the bioinformatic and statistical analysis of large and complex datasets. We feel well prepared for the analysis of this dataset; however, if greater statistical expertise is needed, it is readily available to us through the Biostatistical Consulting service of the Harvard Catalyst program, formally known as the Harvard Clinical and Translational Science Center.

2) Determine the X inactivation status of undeleted COL4A5 and COL4A6 alleles, providing mechanistic insights.

Tumorigenesis following gene deletion typically evokes thoughts of a mechanism based on loss of a tumor suppressor gene in two steps: germline mutation and deletion. Though women have two X chromosomes, one homolog is inactivated in part if not largely by DNA methylation. Thus, the first hit in the COL4A5/COL4A6 locus need not be a germline mutation (Figure 1). The second hit would be deletions identified by Specific Aim 1.

The MLPA assay can be modified to test for DNA methylation, one of the DNA modification associated with X inactivation.^21,22 For this Aim, we plan to design new half-probe oligonucleotides within the deleted region(s). One special feature of the target sequence(s) that we will select is presence of a HhaI restriction site (5’-GCG↓C-3’). HhaI cleaves DNA frequently because its 4-bp restriction site should occur randomly once every 256-nt. HhaI, however, does not cut DNA if the first cytosine is methylated on either strand (i.e., 5’-G5methylCG↓C-3’). Consequently, the target site will be cleaved by HhaI on the active X chromosome (XA), but not on the inactive X chromosome (XI). For the modified assay, restriction enzyme is added with ligase during the MLPA assay. Only target DNA on XI will result in probe amplification. If the deleted locus is on XI, the remaining target DNA on XA will be digested, and a difference in peak size will be noted with and without HhaI digestion. On the other hand, if the deleted locus resides on XA, the remaining target DNA on XI will resist digestion, and no difference in peak size will be noted between aliquots with and without HhaI digestion. We predict the later outcome, confirming functional COL4A5/COL4A6 nullozygosity due to silencing by DNA methylation of the XI allele and somatic deletion of the XA allele.

3) Compare genomic copy number status with protein expression in smooth muscle tumors using a tissue microarray and immunofluorescent microscopy or immunohistochemistry.

Nullozygosity for either COL4A5 or COL4A6 would result in absence of α5(IV) or α6(IV) protein, both of which have been observed in esophageal leiomyomatosis in DL-ATS using immunohistochemistry.¹² Monoclonal antibodies specific for α5(IV) and α6(IV) chains are available, and some are used in a pathology test for ATS using immunofluorescent microscopy of skin biopsies.

For this Aim, we plan to have our Specialized Histology Core construct a tissue microarray while the FFPE tissue plugs for DNA extraction are acquired for Aims 1 and 2. In such arrays, 3 cores from each tumor are placed in the array; and if the tumor has more than one histological phenotype, we will sample multiple areas to evaluate tumor heterogeneity. This resource will allow us to efficiently score staining performed under uniform conditions for a large number of tumors in only a few slides.

We have considerable expertise in both performing and scoring immunohistochemistry results. Other factors we considered in selection of potential assay methods were feasibility of translation into diagnostic testing and the insolubility of extracellular matrix collagens; immunohistological detection (Fig. 6) is superior to other solution-phase biochemical methods in both regards.

We also will obtain antibodies for nearby genes on Xq22.3 and stain the microarray to exclude the possibility of a positional effect produced by deletion of a regulatory element in the COL4A5/COL4A6 locus controlling a neighboring gene (a proposed hypothetical mechanism.²³ Finally, this tissue microarray will be a valuable resource for other studies.

4) Determine the status of COL4A5 and COL4A6 genomic copy number and protein expression in two leiomyosarcoma cell lines obtained from ATCC as a prelude to using them as an in vitro model for smooth muscle neoplasia.

In order to delve into the mechanism(s) responsible for suppressing COL4A6 expression in leiomyosarcoma and test the effect of replacing α6(IV), an in vitro cell culture model would be very helpful. Towards that end, we have obtained two human leiomyosarcoma cell lines, Hs 5.T and SK-LMS-1, derived from female genital tract tumors propagated by the ATCC biorepository. The methods in Aims 1-3 will be applied to map deletions in the COL4A5/COL4A6 locus. In place of the tissue microarray, we will perform immunofluorescence on cells cultured in chambers attached to glass microscopes (Lab-Tek). Characterization of type IV collagen expression by these cell lines will provide a gateway to a number of future tumor cell biological studies.

Future Plans

Based on the background information and preliminary studies, we are confident that our hypothesis will be proven true and open up a number of opportunities to study the pathobiology of uterine smooth muscle tumors. The first opportunity that we will pursue is developing COL4A5/COL4A6 genomic or α5(IV)/α6(IV) protein loss as a diagnostic marker. This effort will require correlation with clinical outcome and validation using an independent set of cases that we would collect by enlisting collaborators from the ranks of investigative gynecological pathologists across the U.S. The next opportunity will be pursuit of studies directed at understanding the mechanism by which deletion of type IV a chain(s) results in smooth muscle neoplasia. We envision designing cell biological experiments testing the effect of replacing α5(IV)/α6(IV) in a deficient cell line and dissecting the downstream pathway. Such replacement studies hopefully will lead to therapeutic trials targeting functional type IV collagen deficiency. In addition, the results of this project would justify creating a leiomyosarcoma mouse model, a prerequisite for deeper, rational scientific study of this devastating tumor.

V10N1 ESUN Copyright © 2013 Liddy Shriver Sarcoma Initiative.