Introduction

Since the late 1990s, molecular genetic techniques have accelerated developments in the field of soft tissue pathology, from the discovery that inflammatory myofibroblastic tumour is a neoplasm,1 to the delineation of gastrointestinal stromal tumour as a distinct diagnostic entity.2–4 Initially, these techniques provided objective reinforcement for the established, morphology-driven classification system. For example, the discovery of WWTR1::CAMTA1 fusions in epithelioid hemangioendothelioma (EHE), defined based on morphological features 30 years previously,5 confirmed their distinction from other vascular tumour types;6 7 the resulting immunohistochemical surrogate CAMTA1 enabled easy confirmation of the diagnosis in routine clinical practice.8 9 The discovery of NAB2::STAT6 fusions in solitary fibrous tumour confirmed its distinct molecular pathogenesis and enabled reconciliation with meningeal hemangiopericytoma in the WHO tumour classification system.10–12 Similarly, the discovery of FUS/EWSR1::CREB3L1/2 fusions in low-grade fibromyxoid sarcoma (LGFMS) and sclerosing epithelioid fibrosarcoma (SEF) definitively established the clinicopathological relationship between these tumour types,13 14 corroborating previously published, morphology-based reports of examples of hybrid LGFMS/SEF.15 16 These discoveries and others like them led to improvements in the classification system and diagnostic practice, and, in the case of EHE, they have provided a mechanistic basis to support clinical trials of novel therapeutic agents.17

The importance of these molecular genetic discoveries should not be understated. However, it is also important to recognise that the above-mentioned tumour types were identified originally by their distinctive morphological features, and their classification as distinctive tumour types in the WHO system was justified by their reproducible clinical behaviour: EHE is a sarcoma type that follows a distinctive clinical course, often with initially indolent behaviour but eventual progression,18 19 and LGFMS gives rise to metastases, often delayed by several years, in 50% of cases despite its bland appearance.20 The clinicopathological definition of these tumour types is consistent with the central thesis of the histopathology of tumours: tumour morphology is a surrogate of tumour biology and clinical behaviour, and therefore a morphology-driven classification system separates tumours into clinically meaningful categories. This concept applies to all recognised tumour types, including deceptively bland ones such as LGFMS as well as deceptively pleomorphic ones such as superficial CD34-positive fibroblastic tumour (SCD34FT). The use of molecular genetic testing and immunohistochemistry reinforces this morphology-driven system, providing tools to distinguish morphologically overlapping tumour types and providing objective evidence to support diagnoses.

In the last 10 years, next-generation sequencing technology has become more accessible, and in many institutions its application has become routine. This widespread adoption of sequencing has been an overwhelmingly positive development, improving diagnostic accuracy and increasing access to precision oncology.21 22 In academic centres, tumours not readily classified on morphological grounds are increasingly sequenced for diagnostic purposes. The sequencing of unclassified tumours has led to the identification of new tumour types on the basis of unique, previously undescribed molecular genetic features, marking a shift in the nature of tumour discovery: in the past, tumour types were discovered based on morphology, while now tumour types are being discovered primarily via molecular testing, with secondary histological evaluation to identify reproducible morphological features that might facilitate diagnosis. This shift has furthered our understanding, but it has introduced biases that have the potential to lead to overtreatment.

The influence of referral bias in the description of rare tumour types

Historically, rare tumour types have been described by expert subspecialist pathologists, who are in a unique position to identify tumours because of an affiliation with tertiary referral centres or because of high-volume personal diagnostic consultation practices. Because patients with aggressive disease are more likely to present at tertiary centres, and because aggressive tumours are more likely to be sent in diagnostic consultation, there is an inherent ‘referral bias’ in descriptions of new tumours types, with over-representation of aggressive behaviour. Due to this referral bias, descriptions of novel or rare tumour types tend to overstate the risk of local recurrence and metastasis. Similarly, tumour types identified by sequencing are generally discovered by pathologists working at tertiary referral centres, at which patients present due to recurrent and/or metastatic disease; however, sequencing at these centres is often driven by oncologists who are seeking therapeutic targets in locally advanced or metastatic tumours. Therefore, identification of tumour types via sequencing data has the potential to introduce an even more extreme bias towards aggressive behaviour.

Here, we discuss some recently defined tumour types, with an emphasis in some cases on the influence of the method used for their identification. We discuss tumour types that were recognised based on morphological features (SCD34FT, cutaneous clear cell melanocytic tumour with ACTIN::MITF fusion and pseudoendocrine sarcoma), as well as tumour types that were described initially based on their molecular findings (epithelioid and spindle cell rhabdomyosarcoma (ESRMS), round cell neoplasms with EWSR1::PATZ1 fusion, melanocytic tumour with CRTC1::TRIM11 fusion, clear cell tumour with melanocytic differentiation and MITF::CREM fusion and GLI1-altered mesenchymal neoplasms, including nested glomoid neoplasm). In the latter group, we discuss how subsequent morphological observations and clinical correlation have altered our understanding relative to initial descriptions, which in some cases over-represented malignant behaviour.

Tumour types discovered based on morphologySuperficial CD34-positive fibroblastic tumour

SCD34FT was initially recognised based on its unusual combination of superficial location, marked nuclear pleomorphism, paucity of mitotic activity and diffuse CD34 expression by immunohistochemistry.23 In the initial series of 18 cases, tumours were found to occur in the dermis and subcutis of adults (age range 20–76 years; median 38 years), with a predilection for the lower extremity (67%). Patients generally presented with longstanding, painless masses, consistent with an indolent, slow-growing tumour type. Despite its pleomorphism, which led to the misclassification as ‘low-grade undifferentiated pleomorphic sarcoma’ in some cases, SCD34FT behaved in a generally indolent fashion, with only one patient experiencing regional lymph node metastasis 7 years after resection.

Histologically, SCD34FT shows large neoplastic cells with glassy cytoplasm and bizarre nuclear pleomorphism (figure 1A,B). There are frequently admixed foamy histiocytes (figure 1C), and some examples show neoplastic cells with lipidized cytoplasm. Granular cell change is present in about 20% of cases, and it can be diffuse. Occasional examples are dominated by spindle cell morphology, rendering them more difficult to recognise (figure 1D). As the tumour name would suggest, CD34 is diffusely positive in nearly all cases, making it a useful diagnostic marker (figure 1E).

Figure 1

Superficial CD34-positive fibroblastic tumour (SCD34FT). (A) SCD34FT most commonly involves the dermis and superficial subcutis, where it presents as a slow-growing, painless mass. (B) Neoplastic cells show glassy cytoplasm and bizarre nuclear pleomorphism, and there is a paucity of mitotic activity. The combination of these features is the most helpful diagnostic clue. (C) Admixed foamy histiocytes are a common feature. (D) Some examples are dominated by spindle cell morphology, which makes them more difficult to recognise. (E) Immunohistochemistry demonstrates diffuse CD34 positivity, a feature that is present in essentially all cases. (F) Moderate to strong expression of CADM3 is 95% sensitive and 90%–95% specific for the diagnosis of SCD34FT among morphological mimics.

Around the same time as the description of SCD34FT, a molecular genetic study demonstrated the presence of PRDM10 gene fusions in a subset of tumours diagnosed as low-grade undifferentiated pleomorphic sarcoma.24 The 5′ fusion partners were MED12 and CITED2, the latter of which has been implicated in VGLL2::CITED2 fusions in congenital/infantile rhabdomyosarcoma.25 Other less common PRDM10 fusion partners have been reported, including ARHGAP32 and RAB30.26 A follow-up gene expression profiling study demonstrated that tumours with PRDM10 gene rearrangements overexpress CADM3, a cell adhesion molecule, and in vitro functional assays demonstrated that introduction of the CITED2::PRDM10 fusion gene into fibroblasts induced overexpression of CADM3.27

Subsequent studies established clinicopathological overlap between SCD34FT and neoplasms with PRDM10 gene fusion, suggesting that they are on a biological spectrum and that the latter might not be sarcomas.26 28–30 In two of these series, CADM3 expression was tested in tumours with and without PRDM10 fusions and was found to be positive in 95% of both groups, although the mechanism of CADM3 expression in tumours lacking PRDM10 rearrangement remains uncertain.29 30 Moderate to strong CADM3 expression was 90%–95% specific for SCD34FT among morphological mimics, making it a useful diagnostic marker (figure 1F). Anderson et al also demonstrated Wilms tumor protein (WT-1) expression to be 75% sensitive for SCD34FT, including all cases with proven PRDM10 rearrangement.30 Other, less specific markers that are commonly positive include AE1/AE3 (~70%) and desmin (~40%); these stains have the potential to present a pitfall for misdiagnoses of sarcomatoid carcinoma or pleomorphic leiomyosarcoma. The paucity of mitotic activity should mitigate against these pitfalls.

Overall, SCD34FT represents a tumour type that was identified based on distinctive histological features and was found to have a consistent clinical presentation and course. Subsequent studies corroborated the clinical findings in the seminal series, confirming that SCD34FT is indolent, with rare metastases to regional lymph nodes and no documented distant metastases. These studies also determined the genetics in a subset of cases and identified useful diagnostic markers. Given their overlapping clinical and morphological findings, it is now generally accepted that tumours with PRDM10 rearrangement represent a subset of SCD34FT, and the SCD34FT nomenclature is preferred to describe the tumour type.

Pseudoendocrine sarcoma

In contrast to SCD34FT, which is indolent but shows deceptively worrisome morphological features, pseudoendocrine sarcoma is a recently described malignancy with metastatic potential and deceptively bland morphology.31–33 Pseudoendocrine sarcoma occurs in older adults and shows a striking predilection for paravertebral soft tissue, where it exhibits locally destructive growth that can efface vertebral bones. Pseudoendocrine sarcoma has a roughly 40% local recurrence rate, possibly reflecting the difficulty in obtaining wide margins in its delicate anatomic site, and roughly 20% of tumours give rise to distant metastasis. To date, all metastases have been to the lungs, and no patients are known to have died of disease as yet.

Histologically, pseudoendocrine sarcoma is characterised by large lobules of monomorphic neoplastic cells with trabecular, nested or sheet-like growth (figure 2A,B). Neoplastic cells show epithelioid morphology, with moderate amounts of pale eosinophilic cytoplasm, round nuclei and speckled chromatin (figure 2C). The trabecular growth, monomorphic morphology and speckled, ‘salt-and-pepper’ nuclei are all reminiscent of well-differentiated neuroendocrine tumours. Other recurrent morphological features include extracellular hyaline globules (about half of cases) and psammomatous calcifications (about a third of cases) (figure 2D,E).

Figure 2

Pseudoendocrine sarcoma. (A) Pseudoendocrine sarcoma shows large lobules in which tumour cells grow in sheets, nests and trabeculae. In some cases, tumour lobules grow into and efface adjacent vertebral bones. (B) This example shows trabecular growth of monomorphic neoplastic cells. (C) Neoplastic cells exhibit moderate amounts of pale eosinophilic cytoplasm, uniform round nuclei and speckled chromatin. These cytomorphological features and nested architecture are reminiscent of well-differentiated neuroendocrine tumour. (D) Extracellular hyaline globules are present in about half of cases and are a useful diagnostic clue. (E) Psammomatous calcifications are present in about a third of cases. (F) Immunohistochemistry for β-catenin demonstrates diffuse nuclear positivity in >95% of tumours.

Despite its resemblance to well-differentiated neuroendocrine tumour, pseudoendocrine sarcoma lacks expression of keratins and neuroendocrine markers by immunohistochemistry. Instead, it expresses S100 and/or desmin (about 40% each); given this immunophenotype and its monomorphic appearance, pseudoendocrine sarcoma also presents a pitfall for the misdiagnosis of ossifying fibromyxoid tumour. Fortunately, these tumour types can be distinguished by immunohistochemistry for β-catenin, which is positive in >95% of pseudoendocrine sarcomas (figure 2F).31 This Wnt signalling activation is due to underlying CTNNB1 mutations, which to date have all been reported to occur in the DSGPhiXS ubiquitination motif that spans residues D32–S37.31–34 Mutations to this motif also predominate in solid pseudopapillary tumour of the pancreas and microcystic stromal tumour of the ovary, both of which show some morphological overlap with pseudoendocrine sarcoma.35–39

Pseudoendocrine sarcoma constitutes a tumour type that was identified prospectively based on morphology over a timeframe of 20 years, and analysis of this cohort demonstrated distinctive clinical and genetic features. Descriptions such as this one were the norm in soft tissue pathology prior to the advent of the sequencing era.

Cutaneous clear cell tumour with melanocytic differentiation and ACTIN::MITF fusion

Cutaneous clear cell tumour with melanocytic differentiation and ACTIN::MITF translocation (CCTMAM) was described in a series of seven cases by de la Fouchardière et al.40 Tumours in this series occurred in women across a wide age range (15–71 years), with a predilection for the extremities. Clinically, CCTMAM presented as dermal nodules, with infiltrative borders and subcutaneous extension in six of seven tumours. Histologically, CCTMAM shows large, clear neoplastic cells that exhibit sheet-like growth through the dermis (figure 3A). CCTMAM shows a characteristic pattern of collagen entrapment that is a useful diagnostic clue (figure 3B). Neoplastic cells show moderate nuclear atypia, with large nuclei and prominent, melanocyte-like nucleoli (figure 3C). Tumours variably express S-100 protein (~70%), as well as HMB-45 and melan-A in about half of cases each. MITF is diffusely positive (figure 3D), and, as the name implies, CCTMAM harbours translocations involving ACTB or ACTG1 (collectively, ‘ACTIN’) and MITF. These translocations are thought to drive MITF expression through a promoter swapping mechanism, like the ACTB::FOSB fusions in pseudomyogenic hemangioendothelioma or the ACTB::GLI1 fusions in pericytoma with t(7;12).41 42 Given the morphological and immunophenotypic findings in this tumour type, it is possible that CCTMAM represents a type of cutaneous PEComa; however, the expression of S-100 protein in most cases would be unusual for PEComa.

Figure 3

Cutaneous clear cell tumour with melanocytic differentiation and ACTIN::MITF fusion (CCTMAM). (A) CCTMAM shows a sheet-like dermal proliferation of large neoplastic cells with clear cytoplasm. (B) Entrapment of native dermal collagen bundles is a characteristic feature that represents a useful diagnostic clue. (C) Neoplastic cells show atypical, melanocyte-like nuclei, with prominent nucleoli. Note the infiltration through dermal collagen and around blood vessels. (D) Tumour cells show strong and diffuse MITF expression, consistent with constitutive activation driven by ACTIN::MITF fusion.

CCTMAM was identified in retrospective, morphology-based case searches, and its recurrent gene fusion confirmed that it represents a distinctive tumour type. Clinical follow-up was limited to two patients who did not experience recurrence in 7 or 17 years; nevertheless, since in general it is easier to obtain clinical follow-up for aggressive tumour types, it seems likely that CCTMAM is indolent.

Tumour types discovered based on sequencingEpithelioid and spindle cell rhabdomyosarcoma

ESRMS was initially identified through a retrospective gene expression profiling study of unclassified sarcomas, which found three rhabdomyosarcomas that harboured EWSR1::TFCP2 or FUS::TFCP2 fusions and that behaved in a highly aggressive fashion; all three patients, aged between 16 and 38 years, died of metastatic disease within 5 months.43

Although there is a tendency for retrospective, sequencing-based studies to over-represent malignant behaviour, in this tumour type the highly aggressive behaviour was borne out in follow-up, larger-scale studies. Since its initial description, >30 cases have been reported in the literature in a mixture of case series and case reports.44–54 ESRMS occurs across a wide age range (11–86 years, median 27 years) with a slight female predominance, and it has a predilection for craniofacial bones, with a minority of cases presenting primarily in soft tissue. ESRMS is clinically distinct from pleomorphic rhabdomyosarcoma and epithelioid rhabdomyosarcoma, both of which predominantly occur in the elderly,55 and it is distinct from other described spindle cell rhabdomyosarcoma subtypes, which encompass a heterogeneous group with disparate clinical behaviour depending on the genetic alterations.25 52 56–58 The prognosis of ESRMS is dismal, with most patients dying of disease within a year of presentation.

Histologically, ESRMS shows epithelioid and spindle cells with brightly eosinophilic cytoplasm and, usually, marked nuclear pleomorphism (figure 4A). ESRMS lacks morphologically evident rhabdomyoblasts, and it often shows more limited expression of desmin and skeletal muscle transcription factors than other rhabdomyosarcoma subtypes (figure 4B). ESRMS expresses AE1/AE3 in nearly all cases, presenting a pitfall for misdiagnosis of sarcomatoid carcinoma (figure 4C), and it expresses ALK in about 70%–80% of cases despite the lack of underlying ALK fusions (figure 4D).

Figure 4

Epithelioid and spindle cell rhabdomyosarcoma (ESRMS). (A) ESRMS is composed of pleomorphic epithelioid and spindle cells with brightly eosinophilic cytoplasm. In general, the differential diagnosis would include pleomorphic leiomyosarcoma and sarcomatoid carcinoma, and the immunohistochemical profile presents a pitfall for these misdiagnoses. (B) ESRMS can show more limited expression of desmin (pictured) and skeletal muscle transcription factors, in contrast to other rhabdomyosarcoma subtypes that tend to show strong and diffuse expression. (C) ESRMS expresses AE1/AE3 in nearly all cases, presenting a pitfall for the misdiagnosis of sarcomatoid carcinoma. (D) ESRMS expresses ALK in about 70%–80% of cases, a useful diagnostic marker when positive.

ESRMS is characterised by FUS::TFCP2 or EWSR1::TFCP2 fusions, with the former being slightly more common. The mechanism of ALK overexpression remains somewhat uncertain. Gene expression profiling has demonstrated consistent mRNA overexpression, at levels comparable to those seen in inflammatory myofibroblastic tumour.49 Most of the tumours that overexpressed ALK were found to harbour internal ALK gene deletions, the oncogenic mechanism of which remains unclear. Consistent ALK overexpression suggests there might be a role for targeted therapy in ESRMS, and there is a report of a patient whose unresectable disease was stable for 19 months on ALK inhibitor therapy.59

ESRMS is a tumour type identified based on molecular genetic studies, the aggressive behaviour of which has been corroborated on subsequent, large-scale studies. It represents a success of the sequencing era: it is a tumour type that would have been hard to identify based purely on morphology (although, in light of published studies, it is now prospectively recognisable), and gene expression profiling uncovered ALK inhibitor therapy as a potential treatment option in this highly aggressive sarcoma type.

Round cell sarcoma with EWSR1::PATZ1 fusion

The EWSR1::PATZ1 fusion was originally described in a round cell sarcoma in 2000.60 The next significant advance came in 2018,43 with a report of five cases that were tightly clustered by gene expression profiling but that showed significant variability in clinical presentation and tumour morphology. Subsequently, there were isolated case reports and small cases series of round cell sarcomas that harboured EWSR1::PATZ1 fusion and showed aggressive behaviour.61–64

Two recent, larger case series with clinicopathological correlation casted doubt on the notion that these tumours are uniformly sarcomas.65 66 Overall, the 28 described round cell neoplasms with EWSR1::PATZ1 fusion described in the literature occur in soft tissue sites across a wide age range (1–81 years) with no apparent sex predilection. Histologically, tumours can be separated into two groups—bland, monotonous neoplasms with low-to-moderate cellularity and a lack of nuclear atypia, and highly cellular neoplasms, some of which show marked nuclear atypia and high mitotic indices. Both of the recent larger series described bland tumours composed of lobules of neoplastic cells with scant cytoplasm, round-to-ovoid nuclei and vesicular chromatin (figure 5A,B).65 66 Stromal collagen and microcystic architecture were common features. In contrast, aggressive sarcomas with EWSR1::PATZ1 fusion showed more infiltrative growth, increased cellularity and higher-grade nuclear features (figure 5C).

Figure 5

Round cell neoplasms with EWSR1::PATZ1 fusion. (A–B) Morphologically low-grade neoplasm with EWSR1::PATZ1 fusion. (A) This indolent tumour, which grew from 5.0 cm to 5.5 cm in 1.5 years, shows lobular architecture with prominent microcystic spaces. (B) Neoplastic cells show scant, pale eosinophilic cytoplasm, with round-to-ovoid nuclei and vesicular chromatin. While this tumour is cellular, it lacks nuclear atypia or significant mitotic activity. (C) Morphologically high-grade neoplasm with EWSR1::PATZ1 fusion. This tumour shows high-grade nuclear atypia, including nuclear size variation, irregularly shaped nuclei and nuclear hyperchromasia. It is highly mitotically active. (D) Immunohistochemistry demonstrates positivity for rhabdomyoblastic markers desmin, myo-D1 (shown) and myogenin, as well as neuroectodermal markers SOX10, S-100 protein and GFAP.

By immunohistochemistry, neoplasms with EWSR1::PATZ1 fusion co-express neuroectodermal markers, such as SOX10, S-100 protein, GFAP and synaptophysin, and skeletal muscle markers, including desmin, myogenin and myo-D1 (figure 5D). Keratin positivity has also been described, and overall the immunophenotype and presence of EWSR1 rearrangement present the pitfall for misdiagnosis of myoepithelial carcinoma. The EWSR1::PATZ1 fusion occurs via intrachromosomal rearrangement on chromosome 22, as EWSR1 and PATZ1 are approximately 2 Mb apart; given the proximity of these genes, EWSR1 fluorescence in situ hybridisation is likely to be imperfectly sensitive.65 Secondary inactivation of CDKN2A/CDKN2B seems to be associated with aggressive behaviour.

Ultimately, round cell neoplasms with EWSR1::PATZ1 fusion show a biological spectrum that was not apparent in initial, small series and case reports, which over-represented aggressive behaviour. Given the presence of EWSR1::PATZ1 fusions in other tumour types, including renal cell carcinoma and central nervous system tumours,67–70 it is perhaps not surprising that this fusion can occur in tumours with disparate clinical courses. More work is needed to determine histological parameters that predict clinical behaviour in round cell neoplasms with EWSR1::PATZ1 fusion.

GLI1-altered mesenchymal neoplasms

GLI1 alterations were originally identified in mesenchymal neoplasms 19 years ago, in a tumour type termed ‘pericytoma with t(7;12)’ that was identified based on its consistent morphological features.42 71 Pericytoma with t(7;12) is characterised by ACTB::GLI1 fusion, and it was hypothesised that the promoter of the ubiquitously expressed ACTB gene drives expression of GLI1 and, thereby, tumourigenesis. Histologically, pericytoma with t(7;12) is characterised by lobules of spindle cells with bland, ovoid nuclei that show perivascular growth around prominent stromal blood vessels (figure 6A,B). There is often perivascular proliferation of tumour cells at the periphery of the tumour, similar to that seen in tumours in the myopericytoma family. Pericytoma with t(7;12) showed uniformly benign behaviour in the initial description of five cases.

Figure 6

GLI1-altered mesenchymal neoplasms. (A–B) Pericytoma with t(7;12). (A) Pericytoma with t(7;12) is characterised by lobules of neoplastic cells, as seen in this primary gastric example. (B) Within tumour lobules, neoplastic cells show diffuse growth, with a rich capillary network and perivascular accentuation, as seen in myopericytoma. Neoplastic cells show predominantly spindle cell morphology. (C–E) Nested glomoid neoplasm. (C) Nested glomoid neoplasm shows lobular architecture. (D) Within tumour lobules, there are nests of small, epithelioid neoplastic cells with bland, vesicular nuclei. (E) Like pericytoma with t(7;12), nested glomoid neoplasm shows perivascular proliferation of tumour cells, including bulging of tumour lobules into vascular lumina. (F) Immunohistochemistry for GLI1 is sensitive and specific for both pericytoma with t(7;12) and nested glomoid neoplasm. This pericytoma with t(7;12) shows nuclear and cytoplasmic positivity, the typical staining pattern in these tumour types.

Subsequently, MALAT1::GLI1 fusions and GLI1 amplification were described in a subset of gastric plexiform fibromyxomas.72 Around the same time, it was discovered that MALAT1::GLI1 fusions were present in gastroblastoma, a rare, biphasic, gastric tumour type that shows distinct epithelial and mesenchymal components.73 74 More recently, GLI1 fusions and amplification were described in two series of soft tissue neoplasms, some of which gave rise to distant metastases.75 76 Both series contained tumours with nested architecture and prominent capillary networks, while some of the GLI1-amplified tumours had other morphological appearances. With follow-up, three GLI1-rearranged tumours gave rise to lymph node metastases, and one gave rise to lung metastases.75 In the initial description of GLI1-amplified tumours, one patient with a high-grade spindle cell sarcoma developed lung metastases, while another with an overtly malignant-appearing epithelioid neoplasm developed local recurrence.76 Several follow-up series described GLI1-rearranged and GLI1-amplified tumours that occurred in a variety of body sites and organ systems,77–86 and one series raised the question of whether these tumour types, including malignant ones, all represented the same tumour type as pericytoma with t(7;12).77 Examining the data in these series, it is clear that nested morphology was common among tumours with GLI1 alterations, and that at least some tumours with nested morphology gave rise to metastases.75 77 86 However, the set of tumours that metastasised were enriched for other morphological patterns, including fascicular spindle cell sarcomas and/or overtly malignant-appearing epithelioid neoplasms.76 77 83 85 Lastly, there were morphologically benign, GLI1-altered neoplasms that were not nested and did not resemble pericytoma with t(7;12).83

The disparate range of morphological features among described GLI1-altered tumours raises the possibility that they represent different tumour types with overlapping genetic features. To address this issue, a recent series of GLI1-altered neoplasms approached them from a tumour morphology standpoint, describing a set of 20 morphologically consistent, nested tumours.87 These tumours were termed ‘nested glomoid neoplasms’ because histologically they showed some features reminiscent of glomus tumour, including lobular architecture, uniform round-to-ovoid nuclei and perivascular proliferation of tumour cells (figure 6C–E). However, in contrast to glomus tumour, nested glomoid neoplasms showed nested growth and lacked distinct cell borders. This set of morphologically uniform neoplasms behaved in an indolent fashion; 4 of 10 tumours with follow-up recurred (including previously unpublished follow-up, obtained after the series was published), and none metastasised. Most recurrences were delayed by several years. The one exception, a tumour that recurred at an interval of 3 months, showed a roughly 10-fold higher mitotic rate than the other tumours with available follow-up. Taken together with features in other reported cases in the literature, this finding suggests that conventional histological predictors of aggressive behaviour, such as high mitotic rate, nuclear atypia and necrosis, might have prognostic value in nested glomoid neoplasms and related GLI1-altered tumour types.

Ultimately, more work is needed to elucidate prognostic factors in GLI1-altered tumour types, and prospective identification of these tumours is necessary to facilitate their study. To this end, GLI1 immunohistochemistry was recently shown to be a useful diagnostic marker, with about 90% sensitivity and 98% specificity for GLI1-rearranged and amplified tumour types (figure 6F);88 this marker has the potential to identify bland, indolent tumours that might not be sequenced otherwise due to lack of clinical necessity. Given the indolent behaviour in published series of morphologically uniform pericytomas with t(7;12) and nested glomoid neoplasms,42 87 it seems possible that the series of GLI1-altered tumours identified via sequencing data might have over-represented malignant behaviour.

Melanocytic tumours with CRTC1::TRIM11 and MITF::CREM fusions

Cutaneous melanocytic tumour with CRTC1::TRIM11 translocation (CMTCT) was initially described by Cellier et al, based on discovery of identical CRTC1::TRIM11 fusions in two cutaneous tumours.89 Since its initial publication, there have been several additional studies and case reports of CMTCT,90–94 including a large series of 41 tumours,94 demonstrating that it shows distinctive morphological and clinical features. CMTCT most commonly occurs in the extremities across a wide age range. Histologically, it usually shows a well-circumscribed nodule in the dermis (figure 7A), a characteristic growth pattern that raises the differential diagnostic consideration of metastatic melanoma. Within the nodule, neoplastic cells grow in small nests and bundles and exhibit somewhat spindled cytomorphology, resembling cellular blue nevus or clear cell sarcoma (figure 7B,C). There is no significant nuclear atypia and no significant mitotic activity. Immunohistochemistry demonstrates expression of melanocytic markers, including SOX10, S-100 protein, MITF, HMB-45 and melan-A. While most examples have been indolent thus far, there are two reported patients who developed metastases, both to regional lymph nodes and one to the lung as well;91 94 therefore, CMTCT does appear to have malignant potential, although no patients have been reported to die of disease as yet.

Figure 7

Emerging cutaneous melanocytic neoplasms identified via sequencing data. (A–C) Cutaneous melanocytic tumour with CRTC1::TRIM11 fusion (CMTCT). (A) CMTCT presents as a circumscribed dermal nodule, a growth pattern that mimics metastatic melanoma. (B) Neoplastic cells show nested and bundled architecture with the nodule. (C) The cytomorphology resembles that of cellular blue nevus, with somewhat spindled neoplastic cells that exhibit melanocyte-like nuclei with prominent nucleoli. (D–F) Cutaneous clear cell tumour with melanocytic differentiation and MITF::CREM translocation (CCTMMC). (D) CCTMMC shows a nested, infiltrative proliferation of neoplastic cells, resembling clear cell sarcoma. (E) CCTMMC shows more cytoplasmic clearing than is typically seen in clear cell sarcoma. (F) Immunohistochemistry demonstrates diffuse nuclear expression of MITF (shown), as well as SOX10, S-100 protein, HMB-45 and melan-A.

The mechanism by which the chimeric CRTC1::TRIM11 protein drives melanocytic differentiation remains somewhat uncertain. CRTC1 is known to bind CREB1, a transcription factor that promotes expression of MITF, a master regulator of melanogenesis.95 96 It is known that the CRTC1::MAML2 fusion protein of mucoepidermoid carcinoma, which harbours similar breakpoints in CRTC1, can bind CREB1 in vivo.97 98 Therefore, it is hypothesised that the CRTC1::TRIM11 fusion protein could possibly drive the melanocytic phenotype through action on CREB1.94 The role of TRIM11, an E3 ubiquitin ligase, is uncertain. More work is needed to determine the precise mechanism of tumourigenesis. Ultimately, CMTCT represents a tumour type discovered via sequencing, for which the initial series accurately described the tumour behaviour as borne out in subsequent studies.

There is yet another emerging tumour type that shows melanocytic differentiation and harbours a unique gene fusion: clear cell tumour with melanocytic differentiation and MITF::CREM translocation (CCTMMC). To date, there are two reported cases of this tumour type, one on the dorsal hand of a man aged 37 years and the other on the occipital scalp in an infant.99 100 The former patient did not experience recurrence in 3 years of follow-up, while the infant, whose tumour was excised with microscopically positive margins, experienced recurrence at 9 months and no subsequent recurrence in over 2 years of additional follow-up. Histologically, CCTMMC shows a nested to diffuse proliferation of neoplastic cells with variably clear cytoplasm (figure 7D,E). The neoplastic cells show melanocyte-like nuclei, and both described examples showed regions with significant cytological atypia. By immunohistochemistry, CCTMMC expresses MITF (figure 7F), S-100 protein, SOX10, Mart-1 and HMB-45. Although neither reported case of CCTMMC gave rise to metastases, it is nonetheless possible that it might be a malignant tumour type. More reported examples with clinical follow-up are needed to make this determination.

MITF is a transcription factor in the same family as TFE3 and TFEB, and all of these proteins activate transcription on dimerisation.101 MITF, TFE3 and TFEB have been implicated as the 3′ fusion partners in various tumour types including EHE with YAP1::TFE3 fusion,102 renal cell carcinoma with TFEB fusion103 and clear cell tumour with ACTIN::MITF fusion (described above);40 in all of these tumour types, the chimeric fusion proteins retain the dimerisation domain. In contrast, the MITF::CREM protein does not retain this domain. Instead, the fusion protein of CCTMMC is hypothesised to rely on the 3′ CREM fusion partner, which is in the same ‘cAMP response element binding protein’ (CREB) family of transcription factors as ATF1 and CREB1.104 These transcription factors have many downstream target genes including MITF. It has been shown that MITF expression is driven by the EWSR1::ATF1 and EWSR1::CREB1 fusions of clear cell sarcoma, in which MITF is likely central to both tumourigenesis and the acquisition of a melanocytic phenotype.105 In the MITF::CREM fusion protein, CREM is hypothesised to function in a similar way, driving expression of MITF and the MITF::CREM fusion protein in a positive feedback loop.99