Molecular characterization of a rare dedifferentiated liposarcoma with rhabdomyosarcomatous differentiation in a 24 year-old

Abstract

Aims:

The aim of this study was to identify potential driver genetic alterations in a dedifferentiated liposarcoma with rhabdomyosarcomatous differentiation.

Methods and Results:

A 24-year-old female underwent resection of an abdominal mass which on a previous biopsy demonstrated rhabdomyosarcomatous differentiation concerning for embryonal rhabdomyosarcoma (ERMS). Histologically the resected tumor displayed a high-grade sarcoma with rhabdomyosarcomatous differentiation in the background of well-differentiated liposarcoma consistent with dedifferentiated liposarcoma (DDLPS). Fluorescence in situ hybridization confirmed MDM2 amplification, as did array-based copy number profiling. Whole-exome sequencing revealed a somatic FGFR1 hotspot mutation and RNA-sequencing an LMNB2-MAP2K6 fusion only within the dedifferentiated component.

Conclusions:

This study represents an in-depth examination of a rare dedifferentiated liposarcoma with rhabdomyosarcomatous differentiation in a young individual. Additionally, it is also instructive of a potential pitfall when assessing for MDM2 amplification in small biopsies. Despite exhaustive analysis, mutation and gene copy number analysis did not identify any molecular events that would underlie the rhabdomyoblastic differentiation. Our understanding of what causes some tumors to dedifferentiate as well as undergo divergent differentiation is limited, and larger studies are needed.

INTRODUCTION

Soft tissue sarcomas can be classified into two main categories. One group consists of sarcomas with complex karyotypes and aneuploidy, while the other consists of tumors with recurrent translocations and/or point mutations with a simple karyotype. Since the latter group harbors recurrent genetic alterations, these features can be used for diagnostic purposes. Dedifferentiated liposarcoma (DDLPS) accounts for 18% of liposarcomas and was first characterized by Evans¹ in 1979 as a well-differentiated liposarcoma (WDLPS) adjacent to a cellular nonlipogenic sarcoma.² DDLPS and WDLPS have characteristic molecular signatures, in that they both have distinctive ring chromosomes.³ These rings contain amplified sequences of chromosome 12q13–21 where the proto-oncogenes MDM2, CDK4, and HMGA2 are located. In addition, DDLPS can show amplification of other genes including JUN and ASK1 on chromosomes 1p32 and 6q23, respectively.⁴ Other genes found amplified in DDLPS include DDIT3, PTPRQ, YAP1 and CEBPA.⁵ To date, it is not entirely understood from a genetic standpoint, how and why some liposarcomas dedifferentiate, while others do not.

Histologically, the dedifferentiated component of DDLPS resembles a high-grade undifferentiated pleomorphic sarcoma in 90% of cases.⁶ In the remaining cases, the dedifferentiated component can display divergent differentiation and exhibit elements of rhabdomyosarcoma, osteosarcoma, leiomyosarcoma, and angiosarcoma^7,8. It is important to recognize divergent differentiation within DDLPS, as other high-grade sarcomas often have a worse prognosis than DDLPS, and the cases with divergent differentiation will behave similar to conventional DDLPS^8,9. However, there appears to be an exception to this rule in DDLPSs with rhabdomyoblastic differentiation as they may have a worse prognosis.¹⁰

Rhabdomyosarcomas are tumors with skeletal muscle differentiation. There are four categories in the WHO Classification of Tumours of Soft Tissue and Bone.¹¹ These include alveolar rhabdomyosarcoma, embryonal rhabdomyosarcoma, pleomorphic rhabdomyosarcoma, and spindle cell/sclerosing rhabdomyosarcoma.¹¹ Here we report a case of DDLPS in a 24 year-old female that on initial biopsy, was thought to be an embryonal rhabdomyosarcoma (ERMS). However, at resection the specimen was revealed to be a DDLPS as it had areas diagnostic of WDLPS and the rhabdomyosarcomatous component showed amplification of MDM2. Additionally, after comprehensive molecular characterization, the tumor did not have the characteristic molecular features of ERMS. This case highlights an unusually early onset of DDLPS, compares the genetic changes of the well-differentiated and dedifferentiated components of a rare subtype of DDLPS, and illustrates potential pitfalls when assessing for MDM2 amplification.

MATERIALS AND METHODS:

Case:

A 24-year-old female presented to her primary care physician with right lower quadrant pain of two months duration. The work up eventually identified a 15 cm heterogeneous mass in the right retroperitoneum. A biopsy of the mass revealed a high-grade sarcoma composed of small round blue cells (figure 1). Immunohistochemically, the cells were positive for desmin and myogenin, while they were negative for S100, SOX-10, SMA, and CKAE1/3. A diagnosis of a high-grade sarcoma with rhabdomyosarcomatous differentiation was rendered, and given the patient’s age and myogenin positivity, ERMS was favored. At resection, the tumor was 31 cm in greatest dimension, with large areas of necrosis. The tumor had cystic and solid elements and was yellow and red in color. Microscopically the tumor consisted of the high-grade sarcoma identified on the previous biopsy, in addition to a background of WDLPS (figure 2). The WDLPS component was positive for MDM2 and CDK4 by immunohistochemistry and MDM2 amplification was defined by fluorescence in situ hybridization (FISH). The high-grade component was again positive for desmin and myogenin, but also showed MDM2 amplification. The resection margins were positive for both the well-differentiated and dedifferentiated components, and two months later she returned with numerous metastases throughout her abdominal cavity. She was then started on chemotherapy and radiation therapy, but despite aggressive measures, the patient passed away 10 months following the initial diagnosis.

Figure 1: Histologic features of the initial liposarcoma biopsy.

A-B. The initial biopsy was composed of small round cells with scant to moderate cytoplasm (100x and 200x, respectively). C. The cells were strongly positive for desmin (200x) and were also focally positive for myogenin (D) (200x).

Figure 2: Histologic features of the liposarcoma resection specimen.

A. The resection showed areas resembling a well-differentiated liposarcoma with large, pleomorphic cells (100x). B-C. In other areas the well-differentiated component was in adjacent to small round cells resembling the initial core biopsy (100x and 200x respectively). D. The small round cells were again positive for desmin (200x).

Oncoscan Copy Number Profiling:

Two separate DNA samples corresponding to the well-differentiated and dedifferentiated regions were extracted from formalin-fixed and paraffin-embedded (FFPE) tissue scraped from unstained tissue sections. These two histological components were identified and circled by a pathologist on a hematoxylin and eosin (H&E) stained slide. The well-differentiated, and dedifferentiated components were estimated to contain 70%, and 90% neoplastic cells, respectively. Corresponding regions of deparaffinized, unstained tissue sections were macrodissected from slides and subjected to separate DNA extractions using the QIAamp FFPE DNA Tissue Kit (QIAGEN). The two DNA samples were subjected to whole-genome copy number microarray-based analysis using the OncoScan FFPE Assay Kit (Thermo Fisher Scientific). OncoScan files were analyzed and interpreted using OncoScan Console v1.3 and Nexus Express for Oncoscan v3 software programs (BioDiscovery) and Chromosome Analysis Suite v4.0 (Thermo Fisher Scientific).¹² Copy number estimates inferred by the software did not account for tumor cell content of the tissue sections used for DNA extraction.

Whole-Exome and RNA-Sequencing:

DNA samples from the microdissected well-differentiated component, dedifferentiated component and matched normal tissue were subjected to whole-exome sequencing (WES) at Memorial Sloan Kettering Cancer Center’s (MSKCC’s) Integrated Genomics Operation (IGO) as previously described,^13,14 at a coverage of 118x, 121x and 98x for the well-differentiated, dedifferentiated and normal samples, respectively. Sequencing data were mapped onto the reference human genome GRCh37 using the Burrows-Wheeler Aligner (BWA, v0.7.15),¹⁵ and sequencing data analysis performed as previously described.^13,14 Mutations affecting hotspot codons were annotated according to Chang et al.¹⁶ The cancer cell fractions (CCFs) of all mutations identified were computed using ABSOLUTE (v1.0.6),¹⁷ using copy number information defined using FACETS,¹⁸ as previously described.^13,14

Total RNA extracted from the two different components was subjected to paired-end RNA sequencing (2× 50bp cycles) at MSKCC’s IGO, as previously described.¹⁹ Sequence read pairs for each case were aligned to the reference genome GRCh37 using STAR,²⁰ Bowtie2²¹ and BWA,¹⁵ and aligned read pairs supporting fusion transcripts from each aligned file were identified using INTEGRATE,²² STAR-Fusion,²³ MAPSplice,²⁴ and FusionCatcher.²⁵ Fusion genes and read-throughs present in six FFPE normal tissue samples subjected to RNA-Sequencing, in 287 normal samples from The Cancer Genome Atlas²⁶ or in FusionFilter were excluded to account for alignment and FFPE and alignment artifacts, and normal transcript variants.¹⁹ After filtering, we considered fusion genes as candidates if they had two or more chimeric junction reads, were in-frame and had a high oncogenic potential (> 0.85) as predicted by OncoFuse.²⁷

TERT promoter Sanger sequencing:

Sanger sequencing for the −124C>T and −146C>T hotspot sites of the TERT promoter was performed as previously described.²⁸ Sequences of the forward and reverse strands were analyzed using MacVector software (MacVector, Inc).

RESULTS:

Molecular Characterization:

Microarray-based genome-wide copy number analysis of the well-differentiated component of the tumor revealed multiple copy number gains within the long arm of chromosome 12 (12q) and no copy number alterations of other chromosomes. Within these copy number gains on 12q, focal regions of amplification (6–7 copies) were observed including CDK4 (12q14.1), HMGA2 (12q14.3), and MDM2 (12q15) (Figure 3). The dedifferentiated component displayed higher levels of chromosomal instability/ copy number alterations. In addition, the dedifferentiated region exhibited similar 12q copy number gains but with smaller, more focal regions with higher levels of CDK4 (12 copies), HMGA2 (12 copies), and MDM2 (15 copies) amplifications, suggesting clonal evolution and selection for higher-level amplification of these genes. The dedifferentiated cells also showed regions of copy number gains involving 1q, 3p, 3q, 7p, 9p, and 17p and losses involving regions of 6p, 14q, 15q, most of 21 and Xq (Figure 3). Additionally, alternating regions of gains and losses throughout chromosomes 11, 19 and 20 suggest chromosome shattering or other chromothripsis-like events.

Figure 3: Oncoscan copy number profiling of the well-differentiated and de-differentiated components of the liposarcoma.

Whole genome view of the well-differentiated component (A) and dedifferentiated component (B) with the top panels showing weighted log2 ratio (copy number) plotted on the Y axis and the chromosomes 1–22 on the X-axis. Colors demark different chromosomes. The well-differentiated component exhibits low-level amplifications within the long arm of chromosome 12 (12q) but no copy alterations in other chromosomes. The dedifferentiated component exhibits a higher degree of chromosomal instability/copy number alterations not restricted to chromosome 12. (C) Calculated copy number in the well-differentiated (top-purple) and dedifferentiated (bottom pink) plots with peaks representing regions with highest levels of amplification, including CDK4, HMGA2 and MDM2.

Whole-exome sequencing analysis revealed a low mutational burden, with 11 and 17 non-synonymous somatic mutations in the well- and dedifferentiated components of the liposarcoma, respectively (23 and 24 total somatic mutations, respectively; Supplementary Table 1). Three silent and four non-synonymous somatic mutations were shared between the two components; the non-synonymous mutations were predicted to be passenger/ non-pathogenic mutations: a PMEL p.S142C mutation, two C19orf47 mutations (p.S398N and p.G261E) and a CEP290 p.E428K mutation (Figure 4A). The remaining mutations were restricted to each of the components. Only the dedifferentiated component harbored a mutation affecting a cancer gene, a pathogenic FGFR1 p.N577K hotspot mutation. Consistent with the increase in copy number alterations, we observed that the C19orf47 p.G261E and CEP290 p.E428K mutations were subclonal in the well-differentiated component and clonal in the dedifferentiated component, providing evidence to suggest that progression from the well- to dedifferentiated component took place in this liposarcoma. Recurrent hotspot mutations in TERT promoter, associated with increased telomerase expression and decreased cell death and genomic instability, have been reported in more than 40 cancer types.^29,30 As the TERT promoter is not covered by WES analysis, we performed Sanger sequencing of the TERT promoter in the well- and dedifferentiated liposarcoma components, however no mutations at the hotspot positions −124 and −146 bp relative to the TERT translation start site were found (Supplementary Figure 1).

Figure 4. Somatic mutations and fusion genes identified in the dedifferentiated liposarcoma with rhabdomyosarcomatous differentiation.

A, Somatic mutations (left) and the cancer cell fractions of the mutations (right) identified in the well-differentiated (WD) and dedifferentiated (DD) components of the liposarcoma, subjected separately to whole-exome sequencing. Cases are shown in columns and genes in rows. Mutation types and cancer cell fractions are color-coded according to the legend, loss of heterozygosity is represented by a diagonal bar and clonal mutations by a yellow box. B, Schematic representation of the LMNB2-MAP2K6 fusion transcript identified by RNA-sequencing and validated by RT-PCR in the dedifferentiated component of the liposarcoma, depicting the exons and domains involved. The breakpoints of the 5’ and 3’ partner gene are represented as black vertical lines. Representative Sanger sequencing electropherograms of cDNA depicting the genomic breakpoints of the chimeric transcript is shown (bottom).

RNA-Sequencing analysis predicted three likely pathogenic fusion transcripts (i.e. in-frame, driver probability >90%, absent in normal; Supplementary Table 2), one of which was validated by orthogonal methods (Supplementary Table 2). The LMNB2-MAP2K6 fusion transcript was validated to be present in the dedifferentiated liposarcoma component (Figure 3B), where exon 1 of lamin B2 (LMNB2) is fused to exon 6 of mitogen-activated protein kinase kinase 6 (MAP2K6).

DISCUSSION:

There have been numerous reports of DDLPS with rhabdomyosarcomatous differentiation.^7,8,31–35 The rhabdomyoblastic component can resemble any type of rhabdomyosarcoma; however, in one study a pleomorphic rhabdomyosarcoma pattern was observed in 50% of cases of DDLPS with rhabdomyosarcomatous differentiation.⁸ Also, only two of 20 cases of DDLPS had a round cell morphology similar to ERMS. While studies have identified characteristic genetic alterations in both WDLPS/DDLPS and ERMS, there is very little information on cases with divergent differentiation. The molecular underpinnings of dedifferentiation in liposarcomas is not fully understood but it does appear that somatic copy number alterations, rather than activating point mutations, is the underlying mechanism of this phenomenon.⁵ There has been considerable interest in copy number alterations of genes that inhibit adipocytic differentiation including JUN, DDIT3, PTPRQ, YAP11, and CEBPA.^36–39 Regardless, no reproducible genetic events have been identified at this time.⁴⁰

The pathogenesis and molecular underpinnings of ERMS also remains poorly understood. In many cases of ERMS there is loss of heterozygosity on many loci within chromosome 11 including 11p15.5 which contains the IGF2, H19, CDKN1C, and HOTS genes.⁴¹ While our tumor did have many gains and losses of chromosome 11 in the dedifferentiated cells, there were no gains or losses of the previously mentioned genes commonly found in ERMS. Secondly, it has been shown that a combination of inactivation of the CDKN2A/B tumor suppressor and the p53 and Rb pathways in addition to activation of FGFR4, RAS, and Hedgehog signaling play vital roles in ERMS pathogenesis.⁴² Intriguingly in a study by Hatley et al., activation of Sonic hedgehog signaling in mice through expression of a constitutively active smoothened, frizzled class receptor (Smo) allele in adipocytes, was shown to give rise to a tumor very similar to ERMS in humans.⁴³ However, the human Smo resides on chromosome 7q32.1, which showed no copy number alterations in either tumor component in our case. The dedifferentiated cells of our tumor did show copy number losses of WNT5A and WNT11 (11q13.5), ligands that have been reported to be down regulated in ERMS.⁴⁴ Additionally, it has been hypothesized that impaired Wnt signaling may contribute to the impaired myogenic differentiation seen in cases of ERMS.⁴⁴ Thus the copy number alterations identified in our tumor may be a contributing feature for the dedifferentiated portion of our tumor to have features resembling ERMS.

Both the well-differentiated and dedifferentiated components shared multiple mutations, including one in PMEL, which encodes a premelanosome protein that participates in melanin production.⁴⁵ Thus, it is likely the dedifferentiated and well-differentiated components had a common origin. Based on the increase in the levels of copy number alterations from the well to the dedifferentiated component, and the observation that subclonal mutations found in the well-differentiated component became clonal in the dedifferentiated component provides evidence to suggest that progression from the well- to dedifferentiated component took place in this case. Of note, the dedifferentiated region of the tumor was found to have a mutation within the FGFR1 gene, which encodes fibroblast growth factor 1 (FGFR1), a receptor tyrosine kinase. FGFR1 belongs to a family of receptor tyrosine kinases that also includes FGFR2/3/4 and these receptors mediate cell migration, survival, proliferation and differentiation.⁴⁶ Activating mutations of FGFR4 have been implicated as oncogenes in rhabdomyosarcoma, with higher levels of FGFR4 indicating advance stage of cancer and poor survival.⁴⁷ However, there have been no such findings linking FGFR1 mutations and ERMS. Instead, Li et al⁴⁸ found FGFR1/3 mutations in liposarcomas, including DDLPS, and when present indicated a poor prognosis. Thus, in our case we were unable to find any genetic features that could definitively explain myogenic differentiation.

Our analysis did detect a likely pathogenic LMNB2-MAP2K6 fusion, which has never been reported in sarcomas. MAP2K6 is a member of the mitogen-activated protein kinase (MAPK) family and encodes for mitogen activated protein kinase kinase 6 (MKK6). MKK6 has been shown to be upregulated in numerous cancers including esophageal, stomach and colon cancers, and evidence suggests it may promote tumor progression.⁴⁹ Additionally, structural rearrangements of MAPK genes were recently shown to be present in a subset of Spitzoid neoplasms.⁵⁰ However, rearrangements involving MAP2K6 were not reported, but the authors pointed out the importance of identifying these alterations due to the potential of MAPK inhibitor therapy.

Our case also highlights a potential pitfall in the diagnosis of DDLPS, as other tumors can show MDM2 amplification. Of particular interest, rhabdomyosarcomas can show MDM2 amplification.^51–53 In one study, 11% of rhabdomyosarcomas demonstrated MDM2 positivity on immunohistochemistry.⁵³ Thus it is imperative that extensive tumor sampling be performed to identify areas of WDLPS. Secondly, another unusual feature of our case is the young age of or our patient as well as the highly aggressive clinical course of the tumor. WDLPS and DDLPS primarily arise in middle-aged adults with only very rare examples of them arising in childhood.⁵⁴ On the other hand, ERMS is much more common in children, with the vast majority occurring before the age of 10 years old.⁵⁵ In addition, despite the high grade morphology that can be present in DDLPS, an aggressive clinical course is only seen in a minority of cases.⁵⁴ The prognosis for rhabdomyosarcoma is dependent on age and histologic type, with embryonal rhabdomyosarcoma being an intermediate prognostic factor with an overall five year survival of 73%.^41,55 The patient in our case was considerably younger than typically seen in cases of WDL/DDLPS and slightly older than typically seen in ERMS.

We are continually accumulating more information about dedifferentiation within sarcomas; however, we still do not fully comprehend the mechanism in which some tumors dedifferentiate and others do not. Moreover, we also do not understand why some tumors undergo divergent differentiation as in our case. The DDLPS we presented had an unusually early onset, and could easily have been misinterpreted as an ERMS given the patient’s age and tumor morphology on biopsy. However, with adequate tumor sampling, areas of WDLPS were identified leading to a diagnosis of DDLPS. Despite extensive molecular analysis of the tumor, no specific genetic alterations associated with features remarkably similar to ERMS were found. There were copy number losses of WNT5A and WNT11, which has been identified in ERMS, but other than these findings, the tumor did not genetically resemble an ERMS. Additionally, a hotspot mutation within the FGFR1 gene as well as increased levels of genomic instability were identified in the dedifferentiated component, which has been linked to DDLPS progression.⁵ As we discover more patterns of copy number alterations in these tumors we may better understand how some tumors dedifferentiate and others display divergent differentiation. Thus, with more information we will be better able to risk stratify and potentially offer novel therapeutic targets to these patients.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 1. Representative Sanger sequence electropherograms of the TERT promoter region encompassing the −124C>T and −146C>T hotspot sites in the well- and dedifferentiated components and matched normal sample.

Supplementary Table 1

Supplementary Table 1: Non-synonymous somatic mutations identified in the well- and dedifferentiated liposarcoma components using whole-exome sequencing.

Supplementary Table 2

Supplementary Table 2: Candidate fusion transcripts identified by RNA-sequencing in the well- and dedifferentiated liposarcoma components.