Genetic characterization of adult primary pleomorphic uterine rhabdomyosarcoma and comparison with uterine carcinosarcoma

Abstract

Aims:

To characterize the genetic alterations in adult primary uterine rhabdomyosarcomas (uRMSs) and to investigate whether these tumors are genetically distinct from uterine carcinosarcomas (UCSs).

Methods:

Three tumors originally diagnosed as primary adult pleomorphic uRMS were subjected to massively parallel sequencing targeting 468 cancer-related genes and RNA-sequencing. Mutational profiles were compared to those from UCSs (n=57) obtained from The Cancer Genome Atlas. Sequencing data analyses were performed using validated bioinformatic approaches.

Results:

Pathogenic TP53 mutations and high levels of genomic instability were detected in the three cases. uRMS1 harbored a likely pathogenic YTHDF2-FOXR1 fusion gene. uRMS2 displayed a PPP2R1A hotspot mutation and amplification of multiple genes, including WHSC1L1, FGFR1, MDM2 and CCNE1, whereas uRMS3 harbored an FBXW7 hotspot mutation and an ANKRD11 homozygous deletion. Hierarchical clustering of somatic mutations and copy number alterations revealed that these tumors initially diagnosed as pleomorphic uRMSs and UCSs were similar. Subsequent comprehensive pathologic re-review of the three uRMSs revealed previously un-identified minute pan-cytokeratin-positive atypical glands in one case (uRMS3), favoring its reclassification as UCS with extensive rhabdomyosarcomatous overgrowth.

Conclusions:

Adult pleomorphic uRMSs harbor TP53 mutations and high levels of copy number alterations. Our findings underscore the challenge in discriminating between uRMS and UCS with rhabdomyosarcomatous differentiation.

Graphical Abstract

INTRODUCTION

Primary uterine rhabdomyosarcoma (uRMS) is an exceptionally rare cancer, even in pediatric patients,¹ and is subclassified into embryonal, pleomorphic, alveolar and spindle cell subtypes according to the 2020 World Health Organization Classification.² RMS is the most common pediatric sarcoma, with the embryonal subtype in the genitourinary tract representing the vast majority of lesions found in children. Embryonal RMS is also common in the cervix and corpus of adolescents and adults, while pleomorphic RMS is typically found in the uterine corpus of adults.^{3, 4} RMS can occur as a pure heterologous uterine sarcoma with aggressive behavior and poor prognosis.^{3, 5} More frequently, RMS can present as part of a biphasic neoplasm such as uterine carcinosarcoma (UCS) or adenosarcoma with sarcomatous overgrowth. The distinction between pure uRMS and rhabdomyosarcomatous differentiation in a biphasic tumor can be challenging when epithelial components are not readily seen in the latter when there is extensive sarcomatous overgrowth. Desmin, myogenin and myo-D1 are expressed in uRMS and in rhabdomyosarcomatous areas of UCS and adenosarcoma,³ although desmin staining alone may be observed in tumors without rhabdomyoblastic differentiation. Hormone receptors are negative in uRMS; but are less frequently expressed in adenosarcoma with sarcomatous overgrowth^{6, 7} and UCS.^{8, 9}.

Due to their rarity, very little is known about the genetics of adult uRMSs. In the pediatric population, germline and somatic mutations in DICER1 have been described in gynecologic embryonal RMS,¹⁰ and, as expected, PAX3-FOXO1 or PAX7-FOXO1 fusion genes have been identified in gynecologic alveolar RMS, consistent with the molecular features of alveolar RMSs from other anatomical sites.^{11, 12} The genomic features of UCSs, including those displaying extensive rhabdomyosarcomatous differentiation, have now been characterized. These tumors typically harbor recurrent mutations in TP53, PTEN, PIK3CA and chromatin remodeling genes as well as high levels of genomic instability.^13–16 Somatic mutations in DICER1 can also rarely occur in the rhabdomyosarcomatous component of UCS.^{17, 18}

Here we sought to characterize the genomic features of three uterine tumors in adult patients initially diagnosed as primary pleomorphic uRMSs, and to compare their repertoire of genetic alterations with that of UCSs, the main differential diagnosis of uRMS in this age group.

MATERIALS AND METHODS

Subject and Samples

Following approval by the Institutional Review Board of Memorial Sloan Kettering Cancer Center (MSKCC), primary uRMSs with available formalin-fixed paraffin-embedded (FFPE) tissue blocks were identified from the pathology archives of MSKCC (n=3). Samples were anonymized prior to the analyses. Representative hematoxylin & eosin (H&E)-stained slides and immunohistochemistry results were reviewed by expert gynecologic pathologists to confirm the diagnoses as well as to document the morphology and immunohistochemical findings. Following standard practice, uRMSs were sampled 1–2 blocks per cm of tumor.¹⁹

Immunohistochemistry

Myogenin and desmin were performed as part of the diagnostic work up as previously described.²⁰ Nuclear myogenin and cytoplasmic desmin and h-caldesmon expression of any extent and intensity in tumor cells was interpreted as positive. Pan-cytokeratin AE1/AE3 and p53 immunohistochemical analyses were performed after re-review of one case (Case 3) to evaluate for the presence of epithelial differentiation as previously described.^{21, 22} Cytoplasmic AE1/AE3 of any intensity in neoplastic cells was considered positive, and p53 expression pattern was considered aberrant if ≥80% of tumor cells showed strong nuclear or cytoplasmic p53 expression (overexpressed).²²

DNA and RNA sequencing

DNA samples derived from microdissected tumor and matched normal tissue were subjected to massively parallel sequencing of 468 cancer-related genes using the MSK-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) assay at MSK’s Integrated Genomics Operation (IGO) as previously described.^{23, 24} The median depth of coverage was 529x (range, 487 – 571x) for tumor and 263x (range, 207 – 318x) for normal samples. Sequencing data were analyzed as previously described.²³ In brief, somatic single nucleotide variants (SNVs) were identified using MuTect²⁵ and small insertions and deletions (indels) with Strelka (v.2.0.15)²⁶, Varscan2 (v.2.3.7)²⁷, Lancet (v1.0.0)²⁸ and Scalpel (v0.5.53).²⁹ SNVs and indels with a >1% global minor allele frequency in ExAC³⁰ were excluded. Copy number alterations (CNAs) and loss of heterozygosity were defined using FACETS.³¹ ABSOLUTE (v1.0.6)³² was employed to determine the cancer cell fraction of each mutation, as previously described.³³ Mutational signatures were inferred for cases with more than five SNVs using Signature Multivariate Analysis (SigMA) at default parameters,³⁴ as previously described.³⁵

RNA extracted from all three tumor samples was subjected to RNA-sequencing at MSK’s IGO using validated protocols, as previously described.³⁶ In brief, fusion genes were identified using INTEGRATE³⁷, deFuse³⁸ and FusionCatcher.³⁹ Fusion genes and read-through candidates that were detected in a set of 297 normal samples from The Cancer Genome Atlas (TCGA) or in six FFPE normal tissues were excluded to account for alignment artifacts and normal transcriptional variants.⁴⁰ Remaining in-frame candidate fusions were annotated using Oncofuse⁴¹ to define their oncogenic potential, as previously described.³⁶

Reverse Transcription-PCR

To validate the putative fusion genes identified by RNA-sequencing, total RNA was reverse-transcribed using SuperScript Vilo Master Mix (Life Technologies; Thermo Fisher Scientific), according to the manufacturer’s instructions. PCR amplification of 10 ng of cDNA was performed using specific primer sets designed based on the predicted fusion genes and breakpoints (YTHDF2, FOXR1; F1: 5’-AAAATGGATCTGTACATCAAAAGGA, R1: 5’-TCCTTATCAGGGTTGGGTTTT; F2: 5’-CCAAAAGGTCAAGGAAACAAA, R2: 5’-GGTTGGGCTCATAATCTGGA), and PCR fragments were sequenced on an ABI 3730 capillary sequencer according to the manufacturer’s protocol, as previously described.³⁶ Sequences of the forward and reverse strands were analyzed using MacVector (MacVector, Inc., Cary, NC, USA). All analyses were performed in duplicate.

Clustering analysis

For comparison with UCS, the updated whole-exome sequencing-derived MC3 data⁴² of UCS (n=57) from TCGA¹⁶ were retrieved from the Genomic Data Commons (GDC).⁴³ For comparison with distinct subtypes of primary RMSs, we retrieved the whole-exome sequencing-based mutation data from Seki et al, including six alveolar RMSs, three pediatric embryonal RMSs, one of mixed type and one RMS not otherwise specified (NOS).⁴⁴ Of the six alveolar RMSs, four harbored FOXO1 gene fusions.⁴⁴ Hierarchical cluster analysis, using complete linkage and Euclidian distance, was performed i) using the uRMSs and all 57 UCSs with or without rhabdomyosarcomatous differentiation from TCGA dataset,¹⁶ and ii) using the uRMSs and the 11 primary RMSs from Seki et al,⁴⁴ filtered for the 468 genes targeted by our sequencing panel.

RESULTS

Three cases initially diagnosed as primary pleomorphic uRMS with sufficient representative tissue for pathologic review and genetic analysis were identified. Patients were 74, 80 and 90 years of age at the time of uRMS diagnosis (Table 1). uRMS2 was FIGO stage I at the time of surgery with only superficial myometrial invasion, while uRMS1 and uRMS3 were stage IV and occupied most if not all of the endometrial cavity; all were exophytic and/or polypoid masses (Table 1). Histologically, the three tumors were composed of sheets of large spindled-to-epithelioid cells exhibiting marked nuclear pleomorphism, coarse chromatin and variable amounts of eosinophilic cytoplasm (Figure 1). Rhabdoid cells with abundant eosinophilic cytoplasm and eccentrically located nuclei were readily seen in the three tumors and were prominent in uRMS1 and uRMS3 (Figure 1). Brisk mitotic activity and necrosis were found in the three cases. Neoplastic cells displayed rhabdomyoblastic differentiation in the three cases, confirmed by the expression of desmin and myogenin (Table 1, Figure 1).

Table 1:

Clinicopathologic characteristics of tumors originally diagnosed as primary pleomorphic adult uterine rhabdomyosarcoma.

Sample ID	Age at diagnosis (y)	FIGO stage	Clinical group	Tumor size (cm)	Histologic subtype	Myogenin	Desmin	Mitosis (#/10 HPF)	Necrosis	Presence epithelial differentiation	AE1/AE3	Clinical Course
uRMS1	90	IV	IV	7.5	pleomorphic	positive	positive	17	Yes	No	No	Surgery, DOD <6 months
uRMS2	74	I	Ia	5.5	pleomorphic	positive	positive	30	Yes	No	No	Surgery, IVRT, length of follow-up <6 months
uRMS3	80	IV	IV	6.2	pleomorphic#	positive	NP	21	Yes	Two minute foci# (0.4 mm each)	Yes#	Surgery, chemotherapy*, DOD <6 months

Staging and clinical group according to Intergroup Rhabdomyosarcoma Study Group (IRSG).^{58, 59} DOD, dead of disease; HPF, high power field; IVRT, intravaginal radiation therapy; NP, not performed; y, years.

^#Following completion of molecular genetic analysis and extensive re-review of the pathology material, this case was favored to be a uterine carcinosarcoma with extensive rhabdomyosarcomatous overgrowth.

^*4 cycles of doxorubicin/cyclophosphamide.

Figure 1: Histologic and immunophenotypic features of tumors originally diagnosed as adult pleomorphic rhabdomyosarcomas.

Micrographs of representative hematoxylin & eosin stained sections of uRMS1–3 (top) and corresponding myogenin stained sections (bottom) are depicted. All tumors demonstrated sheets of spindled to epithelioid cells with marked nuclear pleomorphism; rhabdoid cells with brightly eosinophilic cytoplasm are readily seen in uRMS1 and uRMS3 (top). Nuclear myogenin labelling is present (bottom), confirming rhabdomyoblastic differentiation.

Massively parallel sequencing analysis revealed the presence of clonal pathogenic somatic TP53 mutations and high levels of genomic instability in the three cases analyzed (Figure 2). No other recurrent somatic mutations were identified in the three uRMSs studied (Supplementary Table 1). uRMS2 harbored a PPP2R1A p.S256F hotspot mutation and uRMS3 harbored a FBXW7 p.G423V hotspot mutation, genes that have been previously reported to be recurrently altered in uterine serous carcinomas and UCSs.^{16, 45, 46} uRMS2 harbored a PAK1 truncating mutation, and uRMS3 a PIK3CA in-frame indel (Figure 2A). Furthermore, uRMSs 1 and 2 harbored dominant aging-related mutational signatures (Figure 2A). uRMS3 did not have a sufficient number of mutations to perform mutational signature analysis.

Figure 2: Somatic mutations, copy number alterations and fusion genes detected in uRMS1, uRMS2 and uRMS3.

A) Somatic mutations (left) and cancer cell fractions of mutations identified (right) in uRMS1–3, color coded according to the legend. B) Chromosome plots of uRMS1–3 with the Log₂-ratios plotted on the y-axis and the genomic positions on the x-axis. Arrows highlight genes with amplification or homozygous deletion. C) YTHDF2-FOXR1 fusion gene detected in uRMS1 with exons and domains involved. The breakpoint of the 5’ and 3’ partner genes is represented as a black vertical line. The representative Sanger sequencing electropherogram is shown below. Single nucleotide variant (SNV).

These three uRMSs displayed complex patterns of copy number alterations, with a multitude of copy number gains and losses, and no recurrent amplifications or deletions. The fractions of the genome altered (FGA) in uRMS1 (uRMS1=55%) and uRMS2 (uRMS2=35%) were high, whereas uRMS3 displayed an FGA of 20%. Specifically, uRMS1 harbored amplification of MAX and PRKAR1A and a high-level gain of MYOD1. uRMS2 exhibited amplification in numerous genes, including MYC, FGFR1, WHSC1L1, CCNE1, MET and MDM2, and uRMS3 harbored an ANKRD11 homozygous deletion (Figure 2B, Supplementary Table 1).

RNA-sequencing analysis revealed the presence of a likely oncogenic in-frame fusion in uRMS1 composed of exons 1–4 of YTHDF2 (chr1:29064850) and exons 2–6 of FOXR1 (chr11:118849492; Figure 2C). YTHDF2 is thought to function in RNA binding and signal transduction,⁴⁷ whereas FOXR1 plays a role in transcription.⁴⁸ The Oncofuse driver probability score was 0.66. Both uRMS2 and uRMS3 were tested for the presence of the same YTHDF2-FOXR1 fusion using Reverse Transcription-PCR, and no YTHDF2-FOXR1 fusion was identified in either case.

Given that UCSs are characterized by the presence of recurrent TP53 mutations, high levels of genomic instability and dominant aging-related mutational signatures,^{15, 16, 49} we performed a hypothesis-generating exploratory analysis to ascertain how the genomic features of these three cases initially diagnosed as adult uRMSs would compare to those of UCSs. To this end, we performed hierarchical cluster analysis using the somatic mutations and gene copy number alterations restricted to the 468 genes in our sequencing panel for the uRMSs from this study and the 57 UCSs from TCGA.¹⁶ The uRMSs intermingled with the UCSs based on the somatic mutations and copy number alterations (Figure 3). We further noted that the four UCSs with extensive rhabdomyosarcomatous differentiation from TCGA (i.e. 60–90% of the tumor being heterologous RMS; TCGA-N5-A4R8, TCGA-N7-A4Y5, TCGA-N9-A4Q4, TCGA-NA-A4QW)¹⁶ did not form a separate cluster from the other UCSs, and also did not form a separate cluster with the uRMSs (Figure 3).

Figure 3: Hierarchical cluster analysis of tumors originally diagnosed as uRMSs from this study and UCSs from TCGA.

Hierarchical cluster analysis of somatic non-synonymous mutations A) and copy number alterations B) identified in the 468 cancer genes included in our targeted massively parallel sequencing assay (MSK-IMPACT) using complete linkage and Euclidian distance metric, including all the three cancers originally diagnosed as uRMSs from this study and 57 UCSs from TCGA. UCSs with extensive rhabdomyosarcomatous differentiation (i.e. 60–90%) are highlighted.

Given the similarities between uRMSs and UCSs in terms of their somatic genetic alterations, we performed a second extensive pathologic review of the three cases diagnosed as primary uRMSs. Upon re-review, two previously unidentified minute foci (407 µm and 427 µm) of an atypical epithelial proliferation was detected in uRMS3. Immunohistochemical analysis revealed that these foci expressed pan-cytokeratin AE1/AE3 (Figure 4). While the p53 expression patterns were not conclusive (Supplementary Figure S1), the proliferative appearance with prominent nucleoli and moderate nuclear atypia favored a minute carcinomatous component (Figure 4); however, unusual epithelial metaplasia overlying pure rhabdomyosarcoma could not be excluded. One could posit that uRMS3, rather than an uRMS, may be an UCS with extensive rhabdomyosarcomatous differentiation. No likely carcinomatous components were seen in the two other cases upon extensive re-review.

Figure 4: Histologic re-review and immunohistochemical analysis of uRMS3.

A) The rhabdomyosarcomatous component demonstrates markedly atypical pleomorphic cells with eccentrically located nuclei, abundant eosinophilic cytoplasm and B) myogenin expression consistent with rhabdomyosarcomatous differentiation. C) A minute focus of an epithelial proliferation with nuclear atypia and D) pan-cytokeratin AE1/AE3 expression favors a carcinomatous component.

DISCUSSION

Here we present the molecular evaluation of three adult uterine cancers initially diagnosed as primary pleomorphic uRMS, an exceptionally rare tumor whose diagnosis remains challenging. In fact, one of the cases originally diagnosed as uRMS was favored to be an UCS with extensive rhabdomyosarcomatous overgrowth after extensive histologic review and immunohistochemical analysis. Our study, however, demonstrated that primary uRMS harbor TP53 mutations and high levels of genomic instability. We further demonstrate that, similar to both pediatric and adult RMS from other anatomic sites, fusion genes, mutations in cell cycle-related genes and chromosomal instability appear to play a role in the pathogenesis of primary adult uRMS.

In one of the pleomorphic uRMSs in this study (uRMS1), we detected and validated a novel fusion gene, YTHDF2-FOXR1. FOXR1 encodes a member of the forkhead box (FOX) family of transcription factors, and PAFAH1B2-FOXR1 and MLL-FOXR1 fusions leading to overexpression of FOXR1 and oncogenic activation have been described in neuroblastoma.⁴⁸ Further studies are warranted to assess the frequency of the YTHDF2-FOXR1 in uRMS and whether it leads to activation of FOXR1 in this tumor type. PAX3/7-FOXO1-positive alveolar RMSs and embryonal RMSs generally have low levels of genomic instability and few recurrent amplifications⁴⁴ as compared to uRMS1 from this study, which in addition to the YTHDF2-FOXR1 fusion gene harbored a TP53 mutation and high levels of copy number alterations. One could posit that the YTHDF2-FOXR1 fusion is not the sole driver of this uRMS or that additional driver events are required for this disease. As a hypothesis-generating exploratory analysis, we also performed hierarchical cluster analysis using the somatic mutations identified in alveolar and embryonal RMSs from Seki et al.⁴⁴ and the three uRMSs studied here, which revealed that uRMSs formed a separate cluster distinct from alveolar or embryonal RMSs (Figure 5).

Figure 5: Hierarchical cluster analysis of tumors originally diagnosed as uRMSs from this study and alveolar and pediatric embryonal RMSs.

Hierarchical cluster analysis of somatic non-synonymous mutations identified in the 468 cancer genes included in our targeted massively parallel sequencing assay (MSK-IMPACT) using complete linkage and Euclidian distance metric, including the three cancers originally diagnosed as uRMSs from this study and 11 primary alveolar (ARMS), pediatric embryonal (ERMS), mixed and not otherwise specified (NOS) RMSs from Seki et al.⁴⁴ Tumor types and PAX3/7-FOXO1 fusion status are color-coded according to the legend.

The genetic profiles of the tumors in our study are similar to a previously published case of pleomorphic uRMS that was found to harbor PIK3CA and TP53 pathogenic mutations.³ There is also some overlap with the limited molecular data that exists regarding pleomorphic RMS of the soft tissues. Soft tissue pleomorphic RMS are characterized by complex karyotypes with numerical and unbalanced structural alterations, but no recurrent structural aberrations.⁵⁰ They usually show losses (10q23, 15q21, 3p, 5q32, 13), gains (1p22, 7p, 18/18q, 20/20p), and amplifications (1p21, 1q21, 3p12, 3q26, 4q28, 8q21, 22q), distinct from alveolar and embryonal subtypes,⁵¹ and are not associated with MDM2 amplification.^{52, 53} Genetic profiles of pleomorphic RMS are similar to undifferentiated pleomorphic sarcoma.⁵³ TP53 is often inactivated in pleomorphic RMS.^{54, 55}

Consistent with the histologic similarities between uRMSs and UCSs with extensive rhabdomyosarcomatous differentiation, our study illustrated the genetic similarities between these lesions, based on hierarchical cluster analysis using somatic mutations and gene copy number alterations. Epithelial differentiation in UCS may be extremely focal, as demonstrated by uRMS3, which may be a UCS after rare minute foci of malignant epithelial differentiation were identified upon pathological re-review. p53 immunohistochemistry, which is commonly positive in UCSs is not likely to be helpful in their distinction from uRMS, as these tumors harbored mutations in TP53. Consistent with uRMS3 being a UCS, this tumor harbored mutations in PIK3CA and FBXW7, which have been reported to be present in 35% and 28% of UCSs, respectively.⁴⁶ In fact, the high degree of molecular genetic overlap between uRMS and UCS (even those without rhabdomyosarcomatous elements) demonstrated in this study raises the tantalizing hypothesis that a subset of pleomorphic uRMSs may represent UCSs with extensive sarcomatous growth, where the sarcomatous component obliterates completely or near completely the carcinomatous component. Our results also suggest that uRMSs should be sampled sufficiently for pathology review (at least 1 block per cm of tumor), to exclude epithelial elements that may otherwise suggest UCS or adenosarcoma with sarcomatous overgrowth.

Nevertheless, the distinction between primary uRMS and UCS is still important as it may affect both treatment and prognosis. Whilst both tumor types show similarly low rates of 5-year survival for high-stage disease, there are significant differences in 5-year survival for uterus-confined disease. Localized RMS has a 5-year survival of 47%, whereas UCS with disease confined to the uterus has a reported 74% 5-year survival.^{5, 56} In addition, according to NCCN guidelines, UCS is typically treated with a uterine cancer regimen, while uRMS is treated with a sarcoma regimen extrapolated from non-uterine sites. Experience with treatment of adults with RMS is limited, but similar therapeutic regimens have been recommended for both adults and children with RMS.⁵⁷

This study has several limitations. The sample size of the study is small, given the rarity of uRMSs, and larger validation studies are warranted. Given the limited amounts of DNA available from these lesions, we restricted our sequencing analysis to 468 cancer-related genes. We cannot rule out, however, that there are other genes which may play a role in uRMSs and the use of whole-exome or whole-genome sequencing may identify additional driver alterations in these rare tumors. The panel utilized, however, covers most of the cancer genes reported to be mutated in RMSs and UCSs.

Taken together, we have shown here that adult pleomorphic uRMS harbors TP53 mutations and high levels of copy number alterations. In addition, gene fusions appear to play a role in the oncogenesis of a subset of uRMSs. Furthermore, our findings emphasize the challenge in discriminating between uRMS and UCS, suggesting that a subset of uRMS, particularly the pleomorphic type, may constitute UCS with extensive rhabdomyosarcomatous overgrowth obliterating the epithelial component.