A Matched Molecular and Clinical Analysis of the Epithelioid Haemangioendothelioma Cohort in the Stafford Fox Rare Cancer Program and Contextual Literature Review

Simple Summary Epithelioid haemangioendothelioma (EHE) is an ultra-rare malignant vascular tumour with a prevalence of 1 per 1,000,000. It develops from endothelial cells, which are the cells that line all blood vessels in the body. Therefore, it typically expresses endothelial cell markers. It can also be identified through analysis of the genes. Two genes, WWTR1 and CAMTA1, are broken and fused together in 90% of cases. Alternatively, in approximately 10% of cases, the genes that are broken and fused together are YAP1 and TFE3. We analysed an Australian cohort of EHE patients to look for associations between genetic changes and clinical characteristics. EHE cases are typically refractory to therapies, and no anticancer agents are reimbursed for EHE in Australia; therefore, this detailed study of EHE adds important information to advance our understanding and help aid the design of treatment regimens in the future. Abstract Background: Epithelioid haemangioendothelioma (EHE) is an ultra-rare malignant vascular tumour with a prevalence of 1 per 1,000,000. It is typically molecularly characterised by a WWTR1::CAMTA1 gene fusion in approximately 90% of cases, or a YAP1::TFE3 gene fusion in approximately 10% of cases. EHE cases are typically refractory to therapies, and no anticancer agents are reimbursed for EHE in Australia. Methods: We report a cohort of nine EHE cases with comprehensive histologic and molecular profiling from the Walter and Eliza Hall Institute of Medical Research Stafford Fox Rare Cancer Program (WEHI-SFRCP) collated via nation-wide referral to the Australian Rare Cancer (ARC) Portal. The diagnoses of EHE were confirmed by histopathological and immunohistochemical (IHC) examination. Molecular profiling was performed using the TruSight Oncology 500 assay, the TruSight RNA fusion panel, whole genome sequencing (WGS), or whole exome sequencing (WES). Results: Molecular analysis of RNA, DNA or both was possible in seven of nine cases. The WWTR1::CAMTA1 fusion was identified in five cases. The YAP1::TFE3 fusion was identified in one case, demonstrating unique morphology compared to cases with the more common WWTR1::CAMTA1 fusion. All tumours expressed typical endothelial markers CD31, ERG, and CD34 and were negative for pan-cytokeratin. Cases with a WWTR1::CAMTA1 fusion displayed high expression of CAMTA1 and the single case with a YAP1::TFE3 fusion displayed high expression of TFE3. Survival was highly variable and unrelated to molecular profile. Conclusions: This cohort of EHE cases provides molecular and histopathological characterisation and matching clinical information that emphasises the molecular patterns and variable clinical outcomes and adds to our knowledge of this ultra-rare cancer. Such information from multiple studies will advance our understanding, potentially improving treatment options.


Introduction
Epithelioid haemangioendothelioma (EHE) is a rare, low-to-intermediate grade malignant tumour of vascular origin that may develop in the extremities, soft tissue, lung, bone, and liver. The name was first coined in 1982 by Weiss and Enzinger due to its overlapping features between a haemangioma and angiosarcoma. It is a very rare tumour with a prevalence of 1 per 1,000,000 in the general population [1].
The World Health Organisation (WHO) has recommended that EHE be grouped with malignant tumours [2]. It has an unpredictable clinical behaviour ranging from indolent to aggressively malignant, with a mean survival of 4.6 years, ranging from 6 months to 24 years [2]. Risk stratification models have been proposed to identify lesions at high risk for tumour progression, with tumour cellularity, tumour size, high mitotic figures, and a Ki67 score greater than 10% being the most significant indicators for poor prognosis [3,4].
EHE has no standard treatment regimen, and very few therapeutic options are available. Historically, liver resection and transplantation has been the only curative option for patients with hepatic EHE because of tumour multifocality, but is associated with variable outcomes [5]. However, an expanding oncology therapeutic landscape and advances in genomic tumour analysis may help direct the choice of potentially active treatments, such as those targeting vascular endothelial growth factor (VEGF) or immunotherapy. Herein, we review the current literature for EHE and report on a cohort of hepatic and extrahepatic EHE diagnoses based on histopathology and molecular studies. We aim to add to the current literature on this ultra-rare cancer so that we can better understand the behaviour and natural history of EHE in Australian patients, hopefully helping to better inform treatment-making decisions in the future.

Radiological Characterisation
Radiologic findings are often nonspecific and vary according to the site. Two characteristic computed tomography (CT) and magnetic resonance imaging (MRI) findings in hepatic EHE include the "lollipop sign", in larger lesions (>5 cm) due to bridging vein thrombosis with a rounded "head" and a tapering "tail" [6], and the "target sign", which is a lesion with a low intensity central area surrounded by a hyperintense rim, and is more likely to be found in smaller lesions (2-5 cm) [7]. Of note, benign-looking pulmonary calcification and hepatic capsular retractions are common findings in positron-emission topography (PET)/CT scans of pleural and hepatic EHE, respectively [8].

Histopathological Features
Morphologically, EHE is most often composed of epithelioid cells, organised in nests or cords, with eosinophilic to vacuolated cytoplasm, set in a distinctive myxohyaline stroma, and occasionally associated with haemorrhagic foci [9]. Blister cells with intraluminal erythrocytes, mild-to-moderate atypia, and a low rate of mitosis may also be present [10]. Immunohistochemical (IHC) stains show EHE consistently expresses endothelial cell markers (CD31, CD34 and ERG), and up to 40% of cases may be positive for cytokeratins [11]. EHE with a YAP1::TFE3 fusion has several distinct morphologic features, including voluminous cytoplasm, well-formed vascular lumens, solid growth, and minimal or no stroma [10]. A minority of cases show atypical or malignant features, such as increased cytological atypia and increased mitotic activity with or without necrosis, and may have morphologic overlap with epithelioid angiosarcoma [12].
Transcriptional co-activator with PDZ-binding motif (TAZ), encoded by the WWTR1 gene, regulates the activity of various transcription factors, including the Transcriptional enhancer associated domain (TEAD) transcription factors, which play roles in cell proliferation and apoptosis [18]. Nuclear translocation of TAZ is inhibited by upstream proteins in the Hippo pathway [18]. The fusion between TAZ and CAMTA1 results in constitutive activation of TAZ, hyperactivation of TEAD-based transcriptional programs, and upregulated cellular proliferation [19,20] which may be inhibited with drugs that modulate TEAD activity [21].
YAP1 is a transcriptional coactivator, also controlled by the Hippo signalling pathway, which also interacts with transcription factors, such as TEADs, to promote growth and inhibit apoptosis [18]. TFE3 belongs to the MiTF/TFE family of basic helix-loop-helix (bHLH) transcription factors and regulates lysosomal biogenesis and energy homeostasis [22]. The YAP1::TFE3 fusion protein also drives hyperactivation of TEAD-based transcription programs [19].
These different fusion genes give rise to diverse biologic behaviour. Some propose that YAP1-TFE3 fusion tumours should be classified as a distinct entity given their unique clinical and histopathologic characteristics in comparison to conventional EHE [9,23]. In addition, YAP1::TFE3 fused EHE tends to arise in younger patients and has a more favourable outcome than EHE characterised by TAZ::CAMTA1 fusions, with the 5-year overall survival being 86% versus 59%, respectively [23,24].
Next-generation DNA sequencing (IMPACT) and targeted RNA sequencing (Archer FusionPlex Custom Solid Panel) on cohorts of 18 and 49 patients have uncovered additional drivers; at least 22% of cases had alterations in cell cycle and epigenetic pathways, and/or loss-of-function alterations in the DNA damage response pathways. These included pathogenic variants in XRCC1/2, ERCC1, RB1, APC, FANCA, CDKN2A and CKDN2B and ATRX loss [25]. More than 50% of EHE tumours have secondary genomic variants, the presence of which may indicate more aggressive disease [24].

Clinical Behaviour
Notably, EHE has a variable clinical behaviour: patients with either solitary soft tissue or multifocal lung/liver disease may follow a relatively indolent course, whereas those with either pleural or lymph node involvement, regardless of their primary site or pathologic grade, follow a highly aggressive clinical course similar to a high-grade sarcoma [24]. In general, Mehrabi et al. found that the 1-, 3-, and 5-year survival rates for hepatic EHE, regardless of treatment, were 83.4%, 55.7% and 41.1%, respectively [5].
Several sets of predictors of clinical outcomes have been proposed. Clinical factors associated with shorter survival include: clinical baseline tumour-related pain (TRP), development of TRP during follow-up, baseline temperature, and development of fatigue during follow-up [26]; multifocality, nodal involvement, lung primary, and distant metastasis [24,27]. A proposed three-tiered risk assessment system using tumour size and histologic atypia (defined as high mitotic rate, tumour grade and coagulative necrosis) to stratify patients into low-risk, intermediate-risk, and high-risk groups has shown 5-year overall survival rates of 100%, 81.8%, and 16.9%, respectively [28]. Most recently, Li et al. have established and internally validated the first EHE nomogram prognostic model [29]. Based on the Surveillance, Epidemiology, and End Results (SEER) database, they recruited 512 EHE patients and calculated overall survival (OS) at 1, 5, and 10 years for all patients as 76.5%, 57.4%, and 48.2%, respectively. The age, tumour stage, degree of tissue differentiation, surgical treatment, chemotherapy, and radiotherapy were independent factors predicting prognosis [29].

Treatment and Management Principles
There is no universally agreed treatment strategy for EHE due to its rarity and lack of comprehensive molecular characterisation. However, a recent consensus strategy by experts and consumers has been published.
Based on analysis of the World Sarcoma Network database, all systemic treatments for sarcoma have limited activity in EHE. Seventy-three patients (33 treated with anthracyclinebased regimens, 11 with weekly paclitaxel, 12 with pazopanib, 15 with IFN-α 2b and 27 with other agents) were included, and none showed meaningful activity [34]. Systemic therapy should be reserved for patients with unresectable disease which is symptomatic and progressive. By contrast the experience with targeted therapies (Table 1), especially with VEGF(R) inhibitors and mTOR inhibitors, is slightly more encouraging, with the potential advantage of a more favourable toxicity profile. In particular, VEGF inhibition with bevacizumab is a promising therapy that exploits the vascular nature of EHE and is well tolerated [35,36].
There are currently no reported cases of delivering genomically guided therapies against any of the identified molecular targets. However, novel small molecules targeting the Hippo pathway are in early-phase trials (NCT04665206) [37]. The fusions identified in EHE are also thought to lead to activation of the MEK signalling pathway. Based on this, trametinib is under investigation in an ongoing phase II clinical trial (NCT03148275) [38]. Lenalidomide [43] A case report of one EHE patient • SD 39 months On treatment discontinuation, slight progression seen, responded to rechallenge.
Anlotinib [44] A case report of one EHE patient SD for more than 2 years Lenvatinib [45] A case report of one EHE patient PR for 6 months bridging liver transplant

DNA and RNA Sequencing and Analysis
DNA extracted from the FFPE tumour samples of two patients (#162 and #521) was sequenced using the TruSight Oncology 500 Panel. RNA extracted from FFPE tumour samples of five patients (#154, #368, #455, #503 and #521) was sequenced using the TruSight RNA Fusion Panel. Whole exome sequencing (WES) was carried out on DNA extracted from the FFPE tumour sample of one patient (#368), and whole genome sequencing (WGS) was carried out on DNA extracted from one snap-frozen tumour sample (patient #130). These tumour samples were matched with patient whole blood.
TruSight RNA Fusion Panel data were analysed by aligning the FASTQ files to the human reference assembly (GRCh37) using STAR aligner13 version 2.7.2b. Arriba [47]. Sample-specific read and coverage metrics were assessed using MultiQC [48].
See Supplementary Data for further details of sequencing and analysis.

EHE Cohort
Through the WEHI-SFRCP, we report a cohort of nine EHE cases with comprehensive histologic and molecular profiling. The diagnoses of EHE were made based on the histopathological examination and confirmed by IHC and, whenever possible, molecular profiling. See Tables 2 and 3 for a summary of the cohort.

Case #368
In October 2003, case #368, a man in his forties, presented with abdominal pain and weight loss (BMI 18) and was found to have deranged liver function tests and multi-focal, hypodense, ill-defined nodules in the right lobe of the liver as well as small bilateral pulmonary metastases on CTCAP. Liver biopsy demonstrated an infiltrating tumour, composed of strands of polygonal epithelioid and spindle cells, set within a fibrous stroma. The tumour cells showed moderate nuclear pleomorphism, hyperchromasia and scanty cytoplasm. Occasional tumour cells showed intra-cytoplasmic lumina that contained red  Table 3). A diagnosis of EHE was adopted based on its histologic appearance and IHC analysis.
hypodense, ill-defined nodules in the right lobe of the liver as well as small bilateral pulmonary metastases on CTCAP. Liver biopsy demonstrated an infiltrating tumour, composed of strands of polygonal epithelioid and spindle cells, set within a fibrous stroma. The tumour cells showed moderate nuclear pleomorphism, hyperchromasia and scanty cytoplasm. Occasional tumour cells showed intra-cytoplasmic lumina that contained red blood cells. IHC revealed strong and diffuse positivity for the endothelial cell markers CD34, CD31 and ERG, while it was negative for the epithelial marker pan-cytokeratin ( Figure 1a, Table 3). A diagnosis of EHE was adopted based on its histologic appearance and IHC analysis.   Table 3 and Supplementary  Table S1). Whole exome sequencing (WES) was carried out on DNA extracted from the same sample. Ploidy was found to be 3.7, indicating genome doubling had occurred, and 15% of the genome also displayed loss of heterozygosity (LOH). The tumour had a low tumour mutational burden (TMB). Somatic variants of uncertain significance (VUS) were identified in ADRBK1, SERPINB7, ABCA1, ABCC4 and ANK1 (predicted to be damaging by in silico predictors but have not been validated) ( Table 3).
Due to extensive hepatic disease and extrahepatic involvement, neither radical resection nor liver transplantation was feasible. Review of the literature suggested his disease may follow an indolent course, thus he underwent close clinical surveillance without intervention.
Two years later, in 2005, he developed progressive right hypochondrial pain with a confirmed radiologic progression. Chemotherapy (carboplatin and etoposide for six cycles) was started. Although the radiologic response was not impressive, he achieved durable pain relief. Subsequently, he developed secondary portal hypertension complicated by oesophageal varices which were effectively controlled by endoscopic banding. Radiological surveillance demonstrated ongoing stable disease (Figure 1c), and he developed atrophy of the right lobe of the liver, mainly segment V and VI, together with hypertrophy of the left lobe. He remains under active surveillance with 6-month clinical review and CT assessment without the need for systemic therapy.

Case #130
Case #130 was diagnosed in 2009 with metastatic EHE. The right medial calf was identified as the primary tumour, with metastatic deposits in the spine, lungs, liver and right humerus. Pathological review indicated the tumour comprised large epithelioid eosinophilic cells, some of which had intracytoplasmic vacuoles. In some areas, the cells were clustered around blood vessels, and some intracytoplasmic red blood cells were noted (Figure 2a). Accompanying fat was infiltrated by tumour and showed mild patchy chronic inflammatory infiltrate. IHC analysis indicated the tumour cells were positive for CD31, CD34 and ERG and negative for pan-cytokeratin ( Figure 2a, Table 3). Surveillance positron emission tomography (PET) scans continue to show overall stability of the metastatic disease ( Figure 2b).
Whole genome sequencing (WGS) was carried out on DNA extracted from the primary tumour, indicating a stable genome with very low TMB (0.12 mutations/Mb) with very few chromosomal rearrangements (Figure 2c). No pathogenic variants were identified; however, the less common YAP1::TFE3 fusion was identified (Figure 2d, Table 3). The breakpoints occurred in intron 1 of YAP1 (location 101,983,686 on chromosome 11) and exon 3 of TFE3 (location 48,896,661 on the X chromosome) so that a fusion protein containing half of the TEAD-interaction domain of YAP1 and the activation domain (AD), basic helix-loop-helix (bHLH) and leucine zipper (LZ) DNA-binding domains of TFE3 was generated (Figure 2d). Interestingly, a fusion not previously reported in EHE was also identified in case #130, with high confidence and more supporting reads than for the YAP1::TFE3 fusion. This fusion gene was likely to be a passenger. This novel gene fusion (CBX3::HECW1) would give rise to a fusion protein containing the Chromatin Organisation Modifier (Chromo) domain of CBX3 and the ubiquitin transferase (HECT) domain of HECW1 (Figure 2d, Table 3). YAP1::TFE3 fusion. This fusion gene was likely to be a passenger. This novel gene fusion (CBX3::HECW1) would give rise to a fusion protein containing the Chromatin Organisation Modifier (Chromo) domain of CBX3 and the ubiquitin transferase (HECT) domain of HECW1 (Figure 2d, Table 3).
These two cases exemplify the potential for EHE to follow an indolent course, with both having been followed closely for over 10 years without the need for subsequent intervention.  These two cases exemplify the potential for EHE to follow an indolent course, with both having been followed closely for over 10 years without the need for subsequent intervention.

Additional EHE Cases in the SFRCP
Six subsequent cases of EHE, #104, #154, #162, #455, #503 and #521, were pathologically assessed. A range of key pathologic features were observed, including plump epithelioid cells and/or spindle cells with enlarged irregular hyperchromatic nuclei, a dense fibrous stroma, and in some cases, apparent intravascular growth and associated mixed inflammatory infiltrate. The cellularity was higher at the periphery of the lesion, with a relatively pauci-cellular and hyalinised central zone. Scattered stromal calcifications were present (Figure 3a). The immunophenotype in all cases was positive for the typical endothelial cell markers CD31, ERG, and CD34 and negative for pan-cytokeratin (Figure 3b, Table 3).
CBX3::HECW1 fusion are shown to the left, with the resulting fusion protein shown to the right. IHC, immunohistochemistry; H&E, haematoxylin and eosin; pan-CK, pan-cytokeratin; PET, positron emission tomography.

Additional EHE Cases in the SFRCP
Six subsequent cases of EHE, #104, #154, #162, #455, #503 and #521, were pathologically assessed. A range of key pathologic features were observed, including plump epithelioid cells and/or spindle cells with enlarged irregular hyperchromatic nuclei, a dense fibrous stroma, and in some cases, apparent intravascular growth and associated mixed inflammatory infiltrate. The cellularity was higher at the periphery of the lesion, with a relatively pauci-cellular and hyalinised central zone. Scattered stromal calcifications were present (Figure 3a). The immunophenotype in all cases was positive for the typical endothelial cell markers CD31, ERG, and CD34 and negative for pan-cytokeratin (Figure 3b, Table 3). The WWTR1::CAMTA1 fusion was identified using RNA extracted from a pleural biopsy for #154, liver biopsy for #455 and liver lesions for #503 and #521 (Figure 4a,b, Table  3). These gene fusions would give rise to fusion proteins containing the WW domain of The WWTR1::CAMTA1 fusion was identified using RNA extracted from a pleural biopsy for #154, liver biopsy for #455 and liver lesions for #503 and #521 (Figure 4a,b, Table 3). These gene fusions would give rise to fusion proteins containing the WW domain of TAZ and the IPT/TIG domain and IQ calmodulin-binding motif of CAMTA1 (Figure 4a, TAZ and the IPT/TIG domain and IQ calmodulin-binding motif of CAMTA1 (Figure 4a,b, Supplementary Table S1). Another fusion not previously reported in EHE was also identified in case #503, with high confidence and more supporting reads than for the WWTR1::CAMTA1 fusion (Table  3). This novel gene fusion (FBN1::WWTR1) would give rise to a fusion protein containing the WW domain of TAZ only, and its function is unknown (Figure 4c, Supplementary  Table S1). This case shared pathological features of cases #368, #154, #455, #503 and #521, which also carried the WWTR1::CAMTA1 fusion, but uniquely demonstrated extension through sinusoidal spaces with atrophy of the intervening liver trabeculae.
DNA extracted from liver tumour samples for two of these six cases, #162 and #521, was analysed using the TruSight Oncology 500 (TSO500) Assay. Both cases had low TMBs, and no pathogenic variants were detected. A VUS in FBXW7 was identified at an allele Another fusion not previously reported in EHE was also identified in case #503, with high confidence and more supporting reads than for the WWTR1::CAMTA1 fusion (Table 3). This novel gene fusion (FBN1::WWTR1) would give rise to a fusion protein containing the WW domain of TAZ only, and its function is unknown (Figure 4c, Supplementary Table S1). This case shared pathological features of cases #368, #154, #455, #503 and #521, which also carried the WWTR1::CAMTA1 fusion, but uniquely demonstrated extension through sinusoidal spaces with atrophy of the intervening liver trabeculae.
DNA extracted from liver tumour samples for two of these six cases, #162 and #521, was analysed using the TruSight Oncology 500 (TSO500) Assay. Both cases had low TMBs, and no pathogenic variants were detected. A VUS in FBXW7 was identified at an allele frequency of 45.4% in #521 ( Table 3). Analysis of RNA extracted from #162 was unsuccessful, and there was insufficient tumour material for #104 to enable molecular analysis to be carried out.

Discussion
A review of the literature for EHE between 1984 and 2022 in PubMed, using EHE as a search criterion, revealed approximately 1800 reviews, case reports and retrospective studies. The results that discussed EHE mimics were considered irrelevant and were excluded from this paper [4,5,31,32,[61][62][63][64][65][66][67][68][69][70]. We have reported on the first Australian cohort of nine patients with EHE and delineate histologic and molecular features, clinical details and long-term follow up.
Molecular analysis in EHE is quite limited in numbers [9,24], and our matched clinical and molecular data provide an important addition to this knowledge base. The WWTR1::CAMTA1 fusion was identified in five of the six cases where RNA fusion panel sequencing or WGS was possible. This fusion frequency is consistent with others [71,72]. Errani et al. identified that this recurrent translocation was a consistent abnormality in all 17 EHE cases but not detected in any of the morphologic mimics of EHE, such as epithelioid haemangioma or epithelioid angiosarcoma [71]. Furthermore, in another study, WWTR1::CAMTA1 fusions were detected in 4/7 low-grade and 23/23 intermediate-grade EHE and again were absent in other EHE mimics [72]. Interestingly, Tanas et al., through deep transcriptome sequencing, reported that 87% to 89% of EHE tumours harboured the t(1;3)(q36;q25) translocation, while none of the other 118 EHE mimics showed that rearrangement [73].
The YAP1::TFE3 fusion was identified in one case (#130; Figure 2d), and this tumour demonstrated unique morphology compared to tumours harbouring the more common WWTR1::CAMTA1 fusion. This was in concordance with retrospective data from the USA, Italy, and World Sarcoma Network [34,74]. FBN1::WWTR1 and CBX3::HECW1, two novel fusions not previously reported in EHE, were identified in one case each and cooccurred with the WWTR1::CAMTA1 and YAP1::TFE3 fusions, respectively. While both novel fusions had more supporting reads than the pathognomonic fusions, they are likely to be passengers, and the function of the resulting fusion proteins is unknown. No known pathogenic gene variants or copy number changes were identified in the four out of nine cases analysed by WGS, WES or TSO500 panel screening. Case #521 had a VUS identified with an allele frequency of 45% in the gene encoding the F-box and WD repeat domain containing 7 (FBXW7) components of the SCF E3 ubiquitin ligase. FBXW7 is a critical tumour suppressor that is frequently mutated in human cancers [75]. Importantly, low expression of FBXW7 is correlated with invasion, metastasis and poor prognosis in a number of cancers [76]. If this variant is validated as pathogenic, this could explain the more aggressive nature of the tumour observed in this case (#521). Case #368 had a variant identified with a frequency of almost 100% in the ADRBK1 gene, which encodes the Gprotein-couple receptor kinase 2 (GRK2). Altered GRK2 expression has been observed in many human cancers and may play a role in angiogenesis [77]. Of particular interest, down-regulation of GRK2 levels in endothelial cells has been found to promote tumour growth [78]. This variant is not classified as pathogenic but is predicted to be damaging, using in silico predictors. While both cases #521 and #368 also had the common EHE fusion gene WWTR1::CAMTA1, these additional molecular features may indicate a more aggressive tumour phenotype. Indeed, the presence of secondary genomic variants has previously been proposed to indicate more aggressive EHE [25].
The immunophenotype in all cases was consistent with the known character of this vascular tumour, displaying positivity for the typical endothelial cell markers CD31, CD34 and ERG. In addition, tumours displayed CAMTA1 positivity if they harboured a CAMTA1 fusion and TFE3 positivity in the one tumour harbouring a TFE3 fusion. The tumour for case #154 also displayed positive TFE3 staining despite lacking a TFE3 rearrangement. This has been observed previously, indicating that TFE3 staining may occur irrespective of TFE3 rearrangement [11,79]. In our cohort, we were not able to associate variation in natural history of this disease with the underlying molecular fusions, although the single case with a YAP1::TFE3 fusion did appear to have relatively indolent disease compared to most of the cases with WWTR1::CAMTA1 fusions.
In our cohort, case #368 reported clinically significant symptomatic pain relief after combination carboplatin and etoposide chemotherapy. Pinet et al. reported a complete response to the same regimen in a patient with metastatic pleural EHE disease [80]. By contrast, liposomal doxorubicin [81,82], interferon α [83,84] and thalidomide [85] have shown limited clinical and radiologic responses. Notably, #368 survived 17 years with minimal symptoms, which could have been predicted by some recent nomograms. This patient's disease displayed few of the many known poor prognostic features (TRP and weight loss). Applying the prognostic nomogram-in retrospect-predicted a 30% chance of survival at 200 months [86]. This reinforces that proposed predictive models of an indolent course should be validated by international aggregation of cases to better predict patient outcome and avoid the toxicity of systemic therapy, which often fails to achieve disease response.

Conclusions
EHE is an ultra-rare, vascular tumour that often presents incidentally and has an unpredictable clinical course. Histologically, it can be mistaken for other benign and malignant vascular-epithelial tumours. However, it can be differentiated by IHC staining for CD31, CD34 and ERG, as well as by molecular testing. Our analysis of nine EHE cases harbouring the two gene-defining molecular abnormalities, WWTR1::CAMTA1 and YAP1::TFE3, supports previous reports that these abnormalities define different biologic entities. This group exemplifies the need to characterise the natural history of rare cancers, as well as identifying underlying genomic drivers. Advancing therapeutic progress for rare tumours requires a precision medicine approach using genomic sequencing and basket trial designs. In the case of EHE, a watch-and-wait approach should be considered, to spare patients with indolent disease from potentially unnecessary treatments. Further work is warranted to determine if the YAP1::TFE3 fusion predicts a less aggressive course than the WWTR1::CAMTA1 fusion.
Analysis of this cohort of EHE was achieved through the WEHI-SFRCP, facilitated through enrolment in the ARC Portal, providing matched clinical, molecular, and histopathological characterisation of rare cancers in the hope of improving treatment options. Cohort analyses such as this, achieved through national collaboration, are critical to advancing our knowledge of ultra-rare cancers.

Limitations of the Study
An obvious limitation of any study involving a rare cancer will always be cohort size. Considering the incidence of EHE is 1 per 1,000,000 population, our cohort of nine patients is quite impressive. While a small cohort size makes it difficult to draw significant associations, it is extremely important to make the results widely available so they can be added to the already presented data on this rare cancer. By combining the results of many small studies, we are able to advance our knowledge of this ultra-rare cancer.