Uterine Tumor Resembling Ovarian Sex Cord Tumor: A Distinct Entity Characterized by Recurrent NCOA2/3 Gene Fusions.

In the course of routine diagnostic RNA-Seq testing, which was performed to exclude the possibility of endometrial stromal sarcoma with sex-cord-like differentiation, an ESR1-NCOA3 fusion gene was identified in a patient with UTROSCT. A subsequent retrospective archival review identified three additional patients with UTROSCT (N=4). The average patient age was 53 years (range, 38–68 years). All tumors were uterine in location. The average size was 2.4 cm (range, 0.7–3.3 cm). Two tumors were surrounded by myometrium; one was centred in the myometrium and focally abutted endometrium; and, one was a radiologically polypoid and diagnosed on curettage, thus myometrial involvement could not be assessed. One of the tumors was circumscribed; two showed myometrial infiltration, one with tongue-like protrusions reminiscent of low-grade endometrial stromal sarcoma on low magnification. There was considerable morphologic heterogeneity amongst, and within, the tumors. They contained retiform, trabecular/cord, tubular and/or nested patterns. The tumor from Patient 1 was predominantly composed of small round cells with a reticular pattern, along with a secondary population of larger polygonal cells ( Figure 1 ). The predominant pattern in Patient 2’s tumor included broad anastomosing trabeculae/cords comprised of polygonal cells, along with more retiform areas ( Figure 2 ). The tumor of Patient 3 was almost exclusively composed of sheets and nests of small round-ovoid cells, with focal tubular and cord-like patterns ( Figure 3 ). Patient 4 had a biphasic tumor with spindle cell fascicles, and a minor component of interspersed tubules ( Figure 4 ). Overall, the cells ranged from round-polygonal-spindle shaped. The cytoplasm ranged from scant to ample and brightly eosinophilic. The nuclei were round-ovoid; occasionally they contained prominent clefts and/or angulation, vesicular chromatin and prominent small nucleoli. Mitotic activity was not conspicuous (0–1 per 10 HPFs [FD=0.55 mm]). Small nests of cells with abundant foamy cytoplasm, a recognized secondary finding,( 6 , 22 ) were noted in two cases. A single case contained lymphovascular invasion. None of the tumors showed necrosis. Patient 1 was noted to have adenomyosis (not shown). Each of the tumors tested showed diffuse immunoreactivity for calretinin, WT-1 (nuclear), and estrogen receptor. There was frequent expression of androgen receptor, keratin, muscle markers and Mart-1, though these tended to less extensive ( Table 1 ). None of the tumors were found to express HMB45. Immunohistochemistry for S100, CD10 and cyclin D1 was performed on a single case and found to be negative (not shown). The tumors with NCOA3 rearrangement involved either exon 14 or 15 (NCBI Reference Sequence: NM_181659.2 ), which was fused to ESR1 exon 3 ( NM_000125.3 ). Two cases with NCOA2 rearrangement involved exon 14 ( NM_001321703.1 ), which was fused with either ESR1 exon 3 ( NM_000125.3 ), or GREB1 exon 2 ( NM_014668.3 ). Rearrangement of ESR1 , NCOA2 and NCOA3 was independently confirmed by fluorescence in situ hybridization. All 4 cases showed significant up-regulation of ESR1 mRNA levels compared to >100 other sarcomas available on the same RNA-Seq platform ( Supplementary Figure 1 , Supplemental Digital Content 3 , http://links.lww.com/PAS/A685 ). There was no significantly increased expression of either NCOA2 or NCOA3 RNA compared to other tumor types (not shown). GREB1 was not represented on the fusion panel, so mRNA expression could not be assessed. A gene signature was also obtained of these 4 cases compared to the other tumors on the array, showing significant up-regulation of WT1, AR, HOXA10, HOXA11 and PBX1 ( Supplementary Figure 1 , Supplemental Digital Content 3 , http://links.lww.com/PAS/A685 ). By unsupervised hierarchical clustering of the RNA-Seq data, the four study cases clustered together, separate from all other sarcoma types. In order to investigate the potential transcriptional signature overlap with endometrial stromal sarcomas, we then included an additional group of endometrial stromal sarcomas with various gene fusions tested on the same platform. The repeat unsupervised hierarchical clustering showed that the study cohort grouped closely to certain molecular subsets of endometrial stromal sarcoma; namely, low grade endometrial stromal sarcomas with JAZF1-SUZ12 and JAZF1-PHF1 fusions, but not the others ( Figure 5 ).

A primary uterine tumor with a morphology and immunophenotype compatible with UTROSCT was reviewed by one of the authors (BCD), and tested by diagnostic RNA-Seq to exclude the presence of a gene fusion associated with endometrial stromal sarcoma. The result revealed an ESR1-NCOA3 fusion gene candidate. As a result, a retrospective archival review was performed for additional cases diagnosed as UTROSCT (Mount Sinai Hospital, 2010–2018). Each case was examined by RNA-Seq and the findings independently confirmed by fluorescence in situ hybridization (FISH). This study was performed following institutional Research Ethics Board approval. Formalin-fixed paraffin-embedded tissue blocks were cut at 4 microns and stained for calretinin, inhibin, desmin, smooth muscle actin, H-caldesmon, S100, keratin (AE1/AE3), WT-1, HMB45, MART-1, estrogen receptor and androgen receptor, using standard techniques ( Supplementary Table 1 , Supplemental Digital Content 1 , http://links.lww.com/PAS/A683 ). Positive on-slide controls were applied throughout. RNA was extracted from formalin-fixed paraffin-embedded tissue scrolls (10 micron sections, 3–4 per case) or tissue cut onto positively charged glass slides (4 micron section, 7–10 per case) using the ExpressArt FFPE Clear RNA Ready kit (Amsbio, Cambridge, MA). Total RNA was quantified using the Qubit RNA HS Assay Kit (ThermoFisher Scientific, Mississauga, ON). RNA-seq libraries were prepared following the manufacturer’s instructions using an input of 20–100 ng RNA and the TruSight RNA Fusion Panel (Illumina, San Diego, CA). The results were analysed using both STAR and BOWTIE2 aligners, and Manta and JAFFA fusion callers, respectively.( 18 , 19 ) In order to evaluate for similarities in transcriptional signatures, the expression profile of the study group was compared to 13 cases of molecularly confirmed endometrial stromal sarcoma ( JAZF1-SUZ12 , N=7; ZC3H7B-BCOR , N=7; BRD8-PHF1 , N=2; JAZF1-PHF1 , N=2). In addition, the expression profile was compared to over 100 other sarcoma cases that have been previously tested on the same RNA-Seq platform; this did not include other gynecologic neoplasms (e.g., ovarian sex-cord stromal tumors, leiomyosarcoma, Müllerian adenosarcoma). Fluorescence in situ hybridization for ESR1 , NCOA2 and NCOA3 was performed as previously reported in detail.( 20 ) The UCSC genome browser was used to design custom bacterial artificial chromosome (BAC) clone probes flanking the target genes ( http://genome.ucsc.edu ). These were ordered from the BACPAC Resources Center at the Children’s Hospital (Oakland, CA; https://bacpacresources.org ) ( Supplementary Table 2 , Supplemental Digital Content 2 , http://links.lww.com/PAS/A684 ).( 21 ) Each DNA BAC probe was labelled with fluorochromes by nick translation. Staining was performed using standard techniques, briefly: 4 micron formalin-fixed paraffin-embedded tissue sections were deparaffinized, pretreated, and then hybridized with denatured probes. The slides were incubated overnight, then rinsed, stained with 4’,6-diamidino-2-phenylindole (DAPI), mounted, and examined under a Zeiss fluorescence microscope (Zeiss Axioplan, Oberkochen, Germany).

Uterine tumor resembling ovarian sex-cord tumor is a rare neoplasm of unclear histogenesis, which is classified under the rubric of ‘endometrial stromal and related tumours’ in the World Health Organization classification scheme.( 9 ) Prior attempts to molecularly characterize these neoplasms have failed to show a relationship to either endometrial stromal tumors,( 13 – 15 ) or ovarian sex cord tumors.( 16 , 17 ) Following identification of an ESR1-NCOA3 fusion gene in an index patient, we investigated additional cases of UTROSCT in attempt to better characterize the morphologic, molecular and ontological nature of these tumors. Clement and Scully are credited with the first detailed description of UTROSCT. ( 5 , 23 ) In their series uterine tumors with sex-cord differentiation were divided into two groups: (Group I) tumors which are identical to endometrial stromal tumors but with focal epithelial-like differentiation resembling ovarian sex-cord tumors, and (Group II) tumors with a predominant or exclusive pattern resembling an ovarian sex-cord tumor.( 5 ) The relationship of Group I tumors to endometrial stromal sarcoma has recently been confirmed molecularly, where six cases were found to harbor PHF1 rearrangement.( 24 ) However, the nature of Group II tumors (UTROSCT) remains controversial. They have been proposed to originate from epithelial, stromal, or myoid elements; represent a distinct entity more closely related to ovarian sex-cord tumors;( 6 , 7 , 25 , 26 ) or, perhaps originate from an as yet unknown uncommitted cell with the capacity for multidirectional differentiation.( 11 ) Uterine tumor resembling ovarian sex-cord tumor is comprised of epithelioid cells that may assume a variety of architectural patterns (e.g., cords, nests, trabeculae, tubules, and sheets, as well as glandular and retiform patterns).( 6 , 10 , 11 , 13 , 26 , 27 ) A similar array of morphologies was encountered in our series; interestingly, one case was enriched with spindle cells in a fascicular pattern and a second case predominantly contained sheets of small round cells. These tumors have a polyphenotypic immunoprofile.( 7 , 11 ) They are generally positive for keratins (AE1/AE3, Cam5.2), calretinin, vimentin, WT-1 (nuclear/cytoplasmic), and hormone receptors (androgen, estrogen, and progesterone receptors); with more variable immunoreactivity for epithelial membrane antigen, inhibin, FOXL2, steroidogenic factor-1, desmin, smooth muscle actin, calponin, H-caldesmon, CD10, CD56, and Melan-A.( 6 , 10 , 11 , 17 , 26 , 28 ) To date, the molecular pathogenesis of UTROSCT remains to be elucidated. Following the incidental discovery of an ESR1-NCOA3 fusion gene in a patient with UTROSCT in the course of routine diagnostic RNA-Seq testing, we examined three additional cases by RNA-Seq (N=4) to further assess their molecular pathogenesis. Two patients were found to have ESR1-NCOA3 fusions genes, one had an ESR1-NCOA2 fusion gene, and one had a GREB-NCOA2 fusion product. Notably, NCOA2/3 fusions have previously been identified in uterine neoplasms. A GREB1-NCOA2 fusion gene was recently reported in a uterine tumor designated ‘sarcoma, not otherwise classifiable.’( 29 ) This tumor was centred in the uterine corpus and composed of spindle-polygonal cells, with immunoreactivity for keratin and hormone receptors. Interestingly, this case appears to show morphologic and immunohistochemical overlap with Patient 4 in our cohort, though our case additionally contained a tubular component facilitating classification as UTROSCT. In addition, in a series of 20 uterine adenosarcomas Piscuoglio et al . identified two cases with ESR1 rearrangement—one partnered with NCOA2 and the other NCOA3 .( 30 ) Interestingly, one of these cases was reported to contain sex-cord-like elements.( 30 ) Additional studies, with larger cohorts, are necessary to investigate the relationship between UTROSCT and other uterine neoplasms with NCOA gene fusions. Such studies would also be anticipated to further illuminate the diversity of potential fusion genes possible amongst these neoplasms (i.e., range of NCOA genes, and potential partners). ESR1 mRNA up-regulation was observed in all 4 of the cases in this series, including the GREB1-NCOA2 positive tumor. ESR1 encodes estrogen receptor 1, a ligand-dependent transcription factor.( 31 ) The NCOA1–3 genes encode for nuclear receptor co-activators 1–3, and belong to the steroid receptor coactivator p160/SRC family (murine SRC1–3 ), exerting pleiotropic roles on a wide spectrum of physiologic systems, including a critical role for SRC2 in progesterone-dependent uterine function in the mouse.( 32 ) Fusion genes involving NCOA2 have been reported in several soft tissue tumors, including: mesenchymal chondrosarcoma ( HEY1-NCOA2 fusion),( 33 ) soft tissue angiofibroma ( NCOA2 fused to various partners e.g., AHRR , GTF2I ),( 34 , 35 ) congenital spindle cell rhabdomyosarcoma ( NCOA2 fused with SRF, TEAD1, VGLL2 ),( 21 , 36 ) and rare examples of alveolar rhabdomyosarcoma ( PAX3-NCOA2 )( 37 ). The three members of the NCOA protein family have 50–55% sequence homology,( 37 ) thus it is perhaps not surprising these genes may substitute for one another in various fusion gene pairs. There is a precedent for this in alveolar rhabdomyosarcoma, where both NCOA1 and NCOA2 have the potential to pair with PAX3 ;( 37 , 38 ) and, in our cohort ESR1 was found to partner with both NCOA3 and NCOA2 . GREB1, growth regulation by estrogen in breast cancer, is an ESR1-upregulated protein that mediates estrogen activity. In humans GREB1 is expressed in all ESR1-expressing tissues within the reproductive tract.( 39 ) Furthermore, GREB1 is overexpressed in a variety of epithelial cancers (e.g., breast, ovary, prostate),( 40 ) and in ovarian cancer cell lines is associated with increased proliferation and induction of a mesenchymal morphology.( 39 ) In addition to ESR1 , the gene signature of UTROSCT included upregulation of Müllerian-related genes, such as AR and WT1 , as well as overexpression of transcription factors HOXA10, HOXA11 and PBX1 ( Supplemental Figure 1 , Supplemental Digital Content 3 , http://links.lww.com/PAS/A685 ). HOXA10 and HOXA11 are expressed in the uterus of both mice and humans.( 41 ) Pbx is an essential Hox cofactor and is required for Hox activity, they cooperatively promote cell proliferation and have been implicated as proto-oncogenes in human leukemia,( 42 ) and associated with epigenetic regulation in Ewing sarcoma.( 43 ) Although the 4 cases in our cohort clustered separately from other sarcoma types, they appeared to group closely to a subset of low grade endometrial stromal sarcomas, specifically those with JAZF1-SUZ12 and JAZF1-PHF1 fusions. Our sample size is admittedly limited; therefore, while intriguing, there is currently insufficient data to extrapolate a direct relationship between UTROSCT and low-grade endometrial stromal sarcoma. The projected ESR1-NCOA2/3 fusion oncoproteins retain the estrogen receptor domain (ESD) and the Zinc finger domain (ZF) encoded from the first 5 exons of ESR1 ; and the nuclear receptor coactivator (NRC) from the last 10 exons of NCOA2/3 . Similar breakpoints were reported in 2 cases of adenosarcoma, with the fusions retaining the first 3 exons of ESR1 and last 10 exons of NCOA2/3 .( 29 ) Furthermore, similar NCOA2 break points were also detected in other mesenchymal neoplasms with NCOA2 gene rearrangements, including mesenchymal chondrosarcoma, angiofibroma, etc. As uterine tissues typically show high ESR1 and GREB1 expression, a likely mechanism for the current translocations is hijacking of the active promoter region to dysregulate expression of the retained nuclear receptor coactivator domain of NCOA2/3. In summary, we report the molecular characterization of four cases of uterine tumor resembling ovarian sex-cord tumor. The cases were found to contain fusion genes involving ESR1-NCOA3 (N=2), ESR1-NCOA2 (N=1) and GREB1-NCOA2 (N=1). Our findings support the hypothesis that UTROSCT represents a distinct entity, and these results can be directly applied as a means of diagnostic confirmation. Further studies are necessary to characterize the breadth of molecular events possible in UTROSCT, and establish the relationship, if any, to other uterine neoplasms with NCOA rearrangement.

Sex-cord-like differentiation has been reported to occur as a secondary phenomenon in uterine neoplasms such as endometrial stromal sarcoma,( 1 , 2 ) and Müllerian adenosarcoma.( 3 , 4 ) In contrast, sex-cord-like differentiation is considered an intrinsic attribute of so-called ‘uterine tumor resembling ovarian sex-cord tumor’ (UTROSCT),( 5 – 7 ) a rare mesenchymal neoplasm of unclear histogenesis.( 8 ) The World Health Organization currently classifies uterine tumor resembling ovarian sex-cord tumor under the rubric of endometrial stromal tumors.( 9 ) These are uncommon neoplasms unique to the uterus, and rarely cervix.( 10 ) Tumors predominate in middle-aged women.( 7 , 10 , 11 ) Most patients present with symptoms of abnormal bleeding, pain and/or an enlarged uterus.( 5 – 7 ) Hysterectomy is generally curative,( 7 , 12 ) and most cases were thought to follow a benign course.( 9 ) A recent series, however, reported that 23.5% of patients developed metastases and 8.8% of patients died of their tumor; as a result, it has been suggested that these tumors are more appropriately considered of uncertain malignant potential.( 8 ) The molecular pathogenesis of UTROSCT has yet to be elucidated. In contrast to endometrial stromal sarcoma with sex-cord-like differentiation, UTROSCT lacks evidence of either JAZF1-SUZ12 fusion genes,( 13 , 14 ) or PHF1 rearrangement.( 15 ) Furthermore, these tumors lack FOXL2 and DICER1 mutations typical of ovarian adult granulosa cell tumor and Sertoli-Leydig cell tumor, respectively.( 16 , 17 ) Following the incidental discovery of an ESR1-NCOA3 fusion gene by RNA-Sequencing (RNA-Seq) in a patient with UTROSCT, we investigated 3 additional cases to better understand the molecular landscape in UTROSCT.