Ewing Sarcoma breakpoint region 1 (EWSR1) encodes EWSR1, a multifunctional RNA-binding protein that regulates the expression of genes in diverse cellular pathways.1–3 FUS, EWSR1 and TAF15 collectively comprise the ‘FET’ gene family, and FUS and EWSR1 are interchangeable fusion partners in some translocation-driven neoplasms.4 5 EWSR1 fusion proteins usually juxtapose its N-terminal transcriptional activation domain (encoded in exons 1–7) to the DNA-binding domain of its fusion partner.6 7

Ewing sarcoma was first described in 1918 and 1921 by Arthur Purdy Stout and James Ewing, respectively;8 9 in the early 1980s, its recurrent chromosomal translocation (11;22) (q24;q12) was discovered,8–11 and 10 years later, the EWSR1 gene was identified as the 22q12 fusion partner.1 Thereafter, in part based on the recurrent involvement of the 22q12 locus in translocations,12–14 it was discovered that EWSR1 is rearranged in clear cell sarcoma,15 angiomatoid fibrous histiocytoma,16 extraskeletal myxoid chondrosarcoma17 and hyalinising clear cell carcinoma,18 among others.

The identification of specific EWSR1 fusion genes has enabled the refinement of our tumour classification system. For example, the heterogeneous former ‘Ewing-like sarcoma’ category has been mapped more specifically based on the discovery of recurrent EWSR1 fusion partners PATZ1 and NFATC2. Here, we discuss round cell neoplasms with EWSR1::PATZ1 and EWSR1::NFATC2 fusions to highlight advances made possible by the discovery of specific EWSR1 fusion genes.

The EWSR1::PATZ1 fusion was initially described in isolated case reports and a small series of aggressive round-cell sarcomas, which showed infiltrative growth, high cellularity and marked nuclear atypia (figure 1A).19 20 More recent studies have identified cases that show more indolent behaviour and good outcomes following resection; such examples are cytologically bland, with multilobulated growth, prominent stromal collagen and microcystic change (figure 1B).21 22 In general, round cell neoplasms with EWSR1::PATZ1 fusion have a predilection for truncal and abdominopelvic sites, with a wide age range and no sex predilection. They co-express skeletal muscle proteins (desmin, myogenin and myo-D1) and neuroectodermal proteins (S-100, SOX10, GFAP and synaptophysin) by immunohistochemistry, presenting pitfalls for misdiagnoses of rhabdomyosarcoma and myoepithelial carcinoma, respectively. The initial discovery of aggressive sarcomas with EWSR1::PATZ1 fusion followed by the subsequent description of less aggressive tumours harbouring the same fusion might reflect sequencing bias; initial descriptions of tumours discovered via sequencing tend to overrepresent malignant cases, and there are several examples of tumour types that were described initially to be highly aggressive and found subsequently to be more indolent than initially thought.23 More studies are needed to correlate histological features with outcomes in tumours with EWSR1::PATZ1 fusion and to determine whether this entire tumour class should be considered malignant.

(A and B) Round cell neoplasms with EWSR1::PATZ1 fusion. (A) This sarcoma with EWSR1::PATZ1 fusion is an overtly malignant neoplasm with primitive, atypical nuclei and conspicuous mitotic activity. (B) This neoplasm with EWSR1::PATZ1 fusion shows bland, uniform neoplastic cells with round-to-ovoid nuclei. There are prominent microcystic spaces, a useful diagnostic clue when present. This tumour grew slowly during a 1.5-year preoperative interval, and this patient has remained without evidence of disease with 2 years of follow-up post-resection. (C and D) Round cell neoplasms with EWSR1::NFATC2 fusion. (C) This tumour shows uniform, epithelioid-to-round cells growing in cords and small clusters in a densely sclerotic stroma. This appearance is common in round cell neoplasms with EWSR1::NFATC2 fusion, and it raises the differential diagnostic consideration of sclerosing epithelioid fibrosarcoma. (D) This neoplasm with EWSR1::NFATC2 fusion shows regions of fascicular spindle cell morphology which are non-distinctive. Elsewhere, it shows regions with sclerotic stroma similar to those in (C), which facilitated the diagnosis. Images have been altered to improve color balance and brightness and to remove artifacts.

The EWSR1::NFATC2 fusion was first described in 2009, and it is now recognised that this fusion defines a neoplasm that occurs predominantly in bone and across a broad age range.24 25 Most neoplasms with EWSR1::NFATC2 fusion are composed of uniform round-to-epithelioid cells growing in sheets, lobules and cords in a hyalinised stroma; such examples raise the differential diagnostic considerations of sclerosing epithelioid fibrosarcoma or a myoepithelial neoplasm (figure 1C). Other examples that show significant cytologic atypia or fascicular spindle cell morphology are harder to recognise prospectively (figure 1D). Immunohistochemistry demonstrates variable expression of CD99, NKX2.2, EMA and, sometimes, keratins. Fluorescence in situ hybridisation (FISH) shows concomitant EWSR1 rearrangement and amplification, a specific finding among tumours with EWSR1 rearrangement.25 Although there are reported examples of aggressive behaviour, round cell neoplasms with EWSR1::NFATC2 fusion overall show indolent behaviour and mostly good outcomes following resection. More studies are needed to understand the clinicopathological spectrum and predictors of aggressive behaviour of this tumour type.24

As for EWSR1::PATZ1, identification of the EWSR1::NFATC2 fusion has enabled more precise classification and improved prognostication relative to the former heterogeneous ‘Ewing-like sarcoma’ category. These discoveries represent a microcosm of EWSR1-related discoveries: the discovery of EWSR1::SMAD3 led to a more specific definition of a group of fibroblastic tumours with significant local recurrence risk;26 EWSR1::POU2AF3(COLCA2) appears to define a polyphenotypic round cell sarcoma that might otherwise be confused with ganglioneuroblastoma or malignant peripheral nerve sheath tumour;27–29 and EWSR1::GFI1B is present in a subset of angiofibroma of soft tissue lacking AHRR::NCOA2.30 These success stories of the modern molecular era demonstrate how the identification of specific EWSR1 chimeric fusions has improved precision in tumour classification.

The preponderance of EWSR1 fusions in human neoplasia (table 1) presents pitfalls in diagnostic practice and has presented classification system conundrums as well. EWSR1 fusions are present in tumour types that show morphological overlap with one another, such as round cell sarcomas with EWSR1::PATZ1 fusion and Ewing sarcoma. The differential diagnosis of myoepithelioma and extraskeletal myxoid chondrosarcoma, which show significant morphological overlap, cannot be resolved with EWSR1 FISH because both tumour types commonly harbour EWSR1 fusions.17 31 In part for this reason and in part because NR4A3 can form oncogenes with other fusion partners,32 33 NR4A3 FISH (and not EWSR1 FISH) should be used to confirm the diagnosis of extraskeletal myxoid chondrosarcoma.34 In general, EWSR1 FISH should only be used as a confirmatory test after the differential diagnosis is appropriately constrained.

Selected common EWSR1 fusion partners and associated tumour types

The promiscuity of not just EWSR1 but also specific EWSR1 fusion genes has led to some confusion and controversy in the tumour classification system. One notable example is the controversy over whether so-called ‘adamantinoma-like Ewing sarcoma’, which harbours just the same EWSR1::FLI1 fusions as Ewing sarcoma,35 is a Ewing sarcoma variant or else represents a distinctive fusion-driven carcinoma with squamous differentiation.36 In contrast to Ewing sarcoma, adamantinoma-like Ewing shows nests of tumour cells with peripheral palisading, surrounding desmoplastic stroma and sometimes frank keratinisation,37 38 and it shows diffuse expression of keratins and p63/p40, as well as a more limited expression of CK5,5 an unusual immunophenotype for conventional Ewing sarcoma (figure 2). It also shows a different body site distribution than Ewing sarcoma, with a predilection for the head and neck, and limited data suggest that it is more indolent than Ewing sarcoma.36 39 Despite their differences in clinical behaviour, morphology and immunophenotype, some have argued that adamantinoma-like Ewing sarcoma is a Ewing sarcoma variant, largely on the basis of the EWSR1::FLI1 fusion that is otherwise specific for Ewing sarcoma. A recent study by Fritchie et al provided the strongest genetic evidence to date that these tumour types are distinct despite harbouring the same underlying gene fusion;40 methylation profiling of a series of conventional and adamantinoma-like Ewing sarcomas demonstrated two separate clusters on principle component analysis of methylation patterns, consistent with the presence of reproducible epigenetic differences between these tumour types. This result is perhaps not surprising, given the differential gene expression profiles reflected by immunohistochemistry for squamous epithelial markers. Ultimately, this study lends strong support to the notion that even the relatively specific EWSR1::FLI1 fusion of Ewing sarcoma occurs in other tumour types and therefore is not strictly disease-defining.

Adamantinoma-like Ewing sarcoma. (A) This adamantinoma-like Ewing sarcoma shows characteristic nested and trabecular growth, with intervening desmoplastic stroma. Scattered foci of keratinization (arrow) are sometimes identified in adamantinoma-like Ewing sarcoma. This tumour was found to harbour EWSR1 rearrangement by fluorescence in situ hybridisation. (B) Immunohistochemistry demonstrates the expression of keratins (shown) and p40, the latter of which would be unusual for conventional Ewing sarcoma. (C) Immunohistochemistry demonstrates diffuse NKX2.2 expression, a characteristic finding. Images have been altered to improve color balance and brightness and to remove artifacts.

There are other examples of non-specificity of particular EWSR1 fusion genes, most notably with CREB family members, CREB1 and ATF1, which are present in variable proportions of clear cell sarcoma,15 41 42 angiomatoid fibrous histiocytoma43 and malignant gastrointestinal neuroectodermal tumours.44 45 These tumour types show disparate clinical presentations and outcomes,46–48 as well as disparate morphology and immunohistochemical profiles, such that it is incontrovertible that they represent different tumour types. A recent detailed molecular analysis showed differential methylation profiles in these tumour types, as well as differential gene expression profiles by RNA sequencing.49 EWSR1::CREB1 and EWSR1::ATF1 fusions are being found in increasingly many tumour types, including hyalinising clear cell carcinoma,18 malignant mesothelioma,50 malignant epithelioid neoplasm of peritoneal surfaces with EWSR1/FUS-CREB family fusions51 and inflammatory and nested testicular sex cord tumours.52 53 Interestingly, some of these tumour types share architectural features, such as multilobular growth and intervening septa with dense peripheral lymphoplasmacytic infiltrates (figure 3). The presence of the same fusions in these tumour types suggests a shared biological mechanism for this common architecture, although we currently lack such detailed mechanistic understanding. Given their preponderance, it is not surprising that EWSR1/FUS-CREB family fusions have led to confusion in the tumour classification system in some instances; for example, it remains debatable whether primary pulmonary myxoid sarcoma and intracranial myxoid mesenchymal neoplasm represent site-specific variants of angiomatoid fibrous histiocytoma.54–56

Similar architecture in some tumour types harbouring EWSR1/FUS-CREB family fusions. (A) Angiomatoid fibrous histiocytoma, (B) Malignant epithelioid neoplasm of mesothelium-lined surface with EWSR1/FUS-CREB family fusions, and (C) Inflammatory and nested testicular sex cord tumours all show multilobular growth with intervening fibrous septa and peripheral lymphoplasmacytic inflammation (photograph in panel C) provided courtesy of Dr Andres Acosta, Indianapolis, Indiana, USA). Images have been altered to improve color balance and brightness and to remove artifacts.

More recently, the EWSR1::WT1 fusion was shown to lack specificity in mesenchymal neoplasia. For decades, it was considered pathognomonic for desmoplastic small round cell tumour (DSRCT).57 58 However, recent case reports and small series have called this notion into question.59–62 These studies have described unifocal tumours that lacked the nested architecture and desmoplastic stroma of DSRCT and that also exhibited strikingly different clinical behaviour—while DSRCT is rapidly progressive and almost uniformly fatal, these tumours behaved in a benign fashion. While more work is needed to define these other mesenchymal tumour types harbouring EWSR1::WT1, it is becoming clear that this fusion gene is also more promiscuous than previously thought.