Extraskeletal Ewing Sarcoma of the Extremities and Trunk: A Retrospective Analysis of a Mono-Institutional Series

Ewing sarcoma (EwS) is an undifferentiated, small round cell, malignant sarcoma that usually arises in diaphyseal and metaphyseal portions of long bones in adolescence [1], and about 20–25% of EwSs arise in soft tissue (extraskeletal Ewing sarcoma [EEwS]) [2]. The incidence of EEwS is about 0.4 per million, about ten times lower than its bony counterpart (bone Ewing sarcoma [BEwS]) [2], and represents less than 5% of all soft tissue sarcomas [3]. EEwS has identical morphological and genetic features to BEwS. Both are characterized by a proliferation of uniform, poorly differentiated, small round cells. Immunohistochemically, EEwS cells share with BEwS the diffuse and peculiarly strong positivity for CD99, a cell membrane molecule that regulates crucial biological processes, including cell migration and metastasis [4]. Nuclear positivity to NKX2-2 antibody is also characteristic [5]. From a genetic point of view, EEwS is characterized by the presence of an FET::ETS fusion, deriving from structural rearrangements between FET genes (EWSR1 or, rarely, FUS) with genes belonging to the ETS family of transcription factors (FLI1, ERG, ETV1/4, FEV1, E1AF) [1‒4, 6]. As for BEwS, the chimera is represented by EWSR1::FLI1 in most cases. However, although EEwS and BEwS appear identical from a morphological and genetic point of view, different clinical manifestations of bone and EEwS exist. Applebaum et al. [7] showed that EEwS patients tend to be older, female, non-white, and their tumors occur more often as axial primary sites. EEwS has a wider anatomical distribution [8], although the most common onset is observed in the extremities, trunk, and retroperitoneum [9]. It usually presents with swelling, palpable mass, local pain, and stiffness that may precede the diagnosis by several weeks or months. No associations between EEwS and environmental risk factors, drug exposure, or family history of cancer have been established so far [3]. EEwS is treated with regimens designed for BEwS, and the most recent recommendations [10] confirm a combined approach with systemic therapy (chemotherapy [ChT]) and local treatment (wide/R0 surgery and/or pre- or postoperative radiation therapy), according to tumor characteristics and clinical presentation [11, 12]. In the era of personalized medicine, the debate about whether this subgroup of tumors should be considered as a separate entity from BEwS or rather included in the group with fusion-driven soft tissue sarcoma is ongoing. Considering that EEwS is extremely rare, there is little published about it [13]. Available data are mainly from single-institution retrospective series or have been extrapolated from studies designed for BEwS, with a very limited number of adult patients. In addition, with the recent advent of new entities inside the undifferentiated small round cell tumor groups (round cell sarcoma with EWSR1-non-ETS fusions, CIC-rearranged sarcoma, sarcoma with BCOR genetic alterations), the diagnostic accuracy of these retrospective series may be called into question. The aim of the present study was to retrospectively analyze the clinical features, treatment outcomes, and transcriptional profiles of a homogeneous cohort of patients with genetically determined EEwS treated at a single institution to verify whether it is possible to ameliorate the diagnosis or treatment of patients.

This is a retrospective cohort study including patients with EEwS. After approval by the Local Ethics Committee, medical records of a series of patients affected by EEwS treated at our center from 1988 to 2022 were reviewed. We described clinicopathological features and therapeutic treatment in terms of local treatment (surgery +/− radiotherapy [RT]) and ChT. We acquired the information on clinical charts and pathology and imaging reports. Moreover, for a sub-cohort of patients included in the prognostic portion of the study, we analyzed the transcriptional profile along with BEwS samples taken from patients.

Patients with a proven diagnosis of EEwS were included in the study. Diagnosis was reviewed according to the 2020 WHO classification, and all cases were evaluated by an expert pathologist. Along with morphology and immunohistochemistry typical features, specific chromosomal translocations (EWS::FLI1 or EWR1 gene translocation) were researched in each case. Patients with no proven translocation (negative or not evaluable molecular confirmation) or no adequate tumor tissue available for molecular tests were excluded. The extraskeletal origin was evaluated on available radiological imagining. Only patients without any signs of bone involvement were included in the analysis. Patients with visceral or head-neck primary location were excluded.

The following clinicopathological variables were collected: patient gender and age, tumor site, depth and size, metastasis at diagnosis, baseline serum lactate dehydrogenase (LDH), type of initial presentation (primary, unplanned excision, or recurrent tumor), type of local treatment (surgery, RT only, or surgery plus RT), surgical margins status, systemic treatment, and histopathological response after induction ChT. Tumor depth was defined as superficial or deep based on the superficial fascia. Surgical margins were defined according to the residual tumor classification [14]. The evaluation of the histological response to ChT was evaluated according to the Bologna System. Patients with macroscopic foci of viable tumor (grade 1) after induction ChT were defined as “poor responders”, and patients with microscopic evidence of viable tumor (grade 2) or with no evidence of viable tumor (grade 3) were defined as “good responders” (GRs) [13, 15]. Patients were followed up with physical checkups and radiological imaging [16]. Clinical examination was scheduled every 3 months for the first 2 years, every 4 months during the third year, every 6 months for the fourth to the fifth year, and annually from the sixth to the tenth year. An MRI with contrast enhancement of the primary tumor site and a chest CT scan were performed at every follow-up. An abdominal CT scan with contrast enhancement was performed every 6 months for the first 2 years, every 8 months during the third year, and annually during the rest of the follow-up until the tenth year. After relapse, patients started a new follow-up cycle from the beginning. The median follow-up was 107 months (range 3–367). Eight patients out 35 had a follow-up shorter than 2 years, and two of them were lost at follow-up at 3 and 8 months, respectively. All cases were included in the survival analysis to avoid selection bias. All patients were treated according to protocols designed for bone EwS, all treatments were determined by a multidisciplinary team (orthopedic surgeon, radiotherapist, and oncologist). For patients not included in prospective studies, the treatment followed the same principles, that of Cht surgery and/or RT delivered with a multimodal approach, including preoperative induction Cht, local treatment, and postoperative Cht. Patients with metastases at diagnosis received local treatment on the site of metastases, when feasible, with surgery and/or RT. Modifications based on clinical presentation, patient age, and tumor location and size were proposed when necessary. During the study period, ChT protocols in use at the Rizzoli Institute included 5 or 6 drugs (doxorubicin, ifosfamide, vincristine, etoposide, cyclophosphamide, and actinomycin) [17‒19] Starting from late 90s, in high-risk patients (patients with metastases at diagnosis or patients with localized disease and poor response after induction ChT) aged up to 40 years old, high-dose ChT with autologous stem cells rescue was proposed as salvage therapy. ChT treatment did not differ among patients up to 40 years old. In older patients, treatment was decided on a case-by-case basis. Usually, all the drugs used for the younger patients were included, with a lower per cycle and cumulative dose. Surgery with wide margins was the elected local treatment in combination with preoperative or postoperative RT according to size and site of presentation, with a total dose of 50 Gy in 1.8–2 Gy fractions in the preoperative setting and up to 66 Gy in the postoperative setting.

We performed RNA sequencing in 18 Ewing cases, including 12 BEwSs and 6 EEwSs. The 6 EEwSs cases were included also in the prognostic portion of the study. Samples were selected out of the 35 cases using a non-probabilistic sampling method to obtain a representative sample sharing the same demographic characteristic of the larger population (age and gender distribution, localized tumor at onset, administration of same local treatment) and considering both specimen availability and high quality of the sample. Paired-end libraries were synthetized using the TruSeq RNA Library Prep Kit for Enrichment and loaded on NovaSeq 6000 platform (Illumina) and were then aligned to the GRCh38 human genome with the STAR algorithm [20]. The feature counts algorithm was used to assign aligned reads to genes [21]. Raw count was filtered to keep transcripts that had at least 50 reads for all samples, obtaining 18,436 genes. Then, we performed the normalization step and transformed the normalized count to variance-stabilized expression level (vst) using DESeq2 [22]. Unsupervised hierarchical clustering (HC) was generated using the Pearson correlation as distance measure from vst-transformed expression matrix and ward.D2 as clustering method [23]. Two-dimensional principal component analysis (PCA) was conducted using PCA tools via singular value decomposition [24, 25]. All analyses were performed using R environment version 4.2.2.

Patients’ characteristics were summarized with medians (along with interquartile range) and percentages for continuous and categorical variables, respectively. Overall survival (OS) was calculated from diagnosis to the date of death for any cause or last day of follow-up, set for April 2023. Local relapse-free survival (LRFS) and metastasis-free survival (MFS) were estimated as the temporal measures from diagnosis to the appearance of local relapses or distant metastasis, respectively, or death for cancer. OS, LRFS, and MFS distributions were estimated according to the Kaplan-Meier method. Differences in event risk between groups in the univariate setting were evaluated with the log-rank test. The hazard ratios and corresponding 95% confidence intervals (CIs) were calculated using the Cox proportional hazard regression model. A stepwise forward Cox regression procedure was used to fit regression models, including variable with a p value minor of 0.25. The statistical analysis was performed with Stata software version 18.0.

We retrieved 57 consecutive patients affected by EEwS treated at our clinic from 1977 to 2022. Twenty-two were excluded due to no proven translocation characteristic of EwS (negative or unavailable results or no tumor tissue available for molecular tests). We included 35 patients with molecularly confirmed diagnosis of EEwS performed between 1988–2022: 21 cases (60%) showed t (11;22) EWSR1::FLI1 translocation detected by RT-PCR, and 1 case (3%) showed t (11;22) EWSR1::FLI1 detected by massive parallel sequencing with Archer FusionPlex sarcoma, while in 13 cases (37%), only EWR1 gene translocation was detected by FISH analysis. In these last cases, the diagnosis of other soft tissue tumors characterized by EWRS1 translocations (i.e., myoepithelioma, myxoid liposarcoma, desmoplastic small round cell tumor) were excluded on the base of the morphological, immunohistochemical, and further molecular features of the neoplasms.

The main characteristics of the 35 included patients are shown in Table 1. The median age at diagnosis was 25 years (range 12–56), with most patients being over 18 years old (29/35). Of these patients, 43% were older than 30 and the 26% were older than 40. The male-to-female ratio was 2:1. Twenty-five patients were admitted due to primary tumor, 7 with recurrence tumor, and 3 after an unplanned excision. Primary tumor size was available for 27 of 35 cases. The median size was 8.75 cm (range from 2.5 to 24 cm). Most of the cases (80%) were deep-seated tumors. Tumor onset was mainly in the extremities (91%), followed by the trunk (11%). Altered LDH values at diagnosis were recorded in 37% (13) of patients. The elected surgery treatment was excision (72% of cases). Two patients (6%) were treated with amputation. Surgical margins were defined as R0 in 20 cases (57%) and R1+R2 in 6 cases (17%). Ten patients (29%) had metastases at the time of diagnosis. The sites involved were the lung (5/10), lung and lymph nodes (2/10), bone (1/10), parathyroid (1/10), and multiple locations (1/10). Tumor size was available for 8 out 10 metastatic patients. Six had tumors bigger than 8 cm (60%), and the other two tumors (20%) were 8 cm or less. In 80% (8/10) of the cases, metastatic tumors were deep.

Local treatment (including all therapies administrated locally prior or during surgery) consisted of surgery in 20 patients (57%), RT and surgery in 11 patients (31%), and RT only in 2 patients (6%). The last 2 patients (6%) were not locally treated. Thirty-three (94%) patients received ChT. The 2 patients not treated with ChT were recurrence excised and treated locally with RT. The histological response to ChT was evaluable in 26 patients, with 21 (80%) classified as poor responders (grade 1 BS) and 5 (20%) as GRs (grade 2/3 BS). Overall, 21 patients (60%) were treated with RT (either pre- or post-surgery). Ten patients (29%) had metastases at the time of diagnosis. The sites involved were the lung (5/10), lung and lymph nodes (2/10), bone (1/10), parathyroid (1/10), and multiple locations (1/10). Tumor size was available for 8 out of 10 metastatic patients. Six had tumors bigger than 8 cm (60%), and the other two tumors (20%) were 8 cm or less. In 80% (8/10) of the cases, metastatic tumors were deep.

To investigate the transcriptional profile of EEwS, we performed RNA sequencing on 6 out of the 35 EEwS patients and 12 BEwS cases. We used unsupervised HC analysis to cluster EwS cases. Based on the overall transcript expression (18,436 genes), we did not observe any differences in the transcriptional profile between EwSs arising in bone compared to soft tissue EwSs HC (shown in Fig. 1). The distinctive transcriptional profile of EEwS compared with BEwS was further explored using 2D PCA. The analysis confirmed that EEwS did not form a distinct cloud from the BEwS samples (shown in Fig. 2), further supporting the idea that the genetic landscape of EwS is independent from the tumor site of origin. Furthermore, we found the same result when using the 8,000 genes with the highest variance as input for unsupervised HC and PCA (shown in online suppl. Fig. 1a, b; for all online suppl. material, see https://doi.org/10.1159/000540613).

At the time of the last assessment, 14 (40%) patients had died of disease, 3 (9%) were alive with disease, and 18 (51%) were alive without evidence of disease. After a median follow-up of 107 months (range 3–367), the cumulative 5-year and 10-year OS rates were 67% (95% CI: 47–80) and 63% (95% CI: 44–77) (shown in Fig. 3a). In the group of patients with metastases at diagnosis (10/35 patients), 5-year OS was 56% (95% CI: 20–80). No significant difference was found between the metastatic and non-metastatic group of patients in terms of survival. Tumor recurrence (HR: 16, p = 0.004, 95% CI: 2.44–109), high LDH serum level (HR: 11.28, p = 0.006, 95% CI: 2–63), and size (HR: 7.54, p = 0.007, 95% CI: 1.7–32.6), were determined to be independent risk factors for OS (shown in Fig. 3b–d; Table 2). The 5-year and 10-year LRFS rates were 61% (95% CI: 43–75) and 55% (95% CI: 36–79), respectively (shown in Fig. 4a). Local relapse occurred in 6 patients after a mean follow-up of 17 months (range 4–65). Five of them underwent RT (1 neoadjuvant, 2 adjuvants, and 2 neo-/adjuvant). The 5-year and 10-year MFS rates were 53% (95% CI: 35–68) (shown in Fig. 5a). Distant metastasis occurred in 13 patients after a mean follow-up of 16 months (range 5–198). The independent risk factors associated to OS were confirmed to be associated also with LRFS and MFS (shown in Fig. 4 b–d, 5b–d; Table 2). Twenty-five patients out of 35 (69%) had localized disease. Median survival time corresponded to 115 months (range 6–367). The 5- and 10-year OS in this group were 70% (95% CI: 47–84) and 66% (95% CI: 42–81), respectively. No statistically significant risk factor was determined to be associated with OS (Table 3).

Patients were treated using a multimodality approach, including surgery, RT, and ChT. Considering the entire cohort, 33 (94%) patients received ChT: 21 patients (60%) were treated with ChT as first treatment, 3 patients (24%) received ChT after local treatment, and 9 patients (25%) received ChT pre- and post-local treatment. Four out of 33 patients were treated with high-dose ChT. Of the patients older than 40 (9/35), eight were treated with ChT, and 75% of them died of disease (6/8). The patients younger than 40 (26/35), 25/35 were treated with ChT, and 28% of them died of disease (7/25). The use of different protocols did not allow us to have enough patients to test for each kind of treatment.

EEwS is very rare, and there are only few reports describing its main features. The aim of this retrospective study was to further characterize EEwS, to verify if its transcriptional profile could help conform the therapeutic regimens between EEwS and BEwS, and to identify which risk factors could influence the prognosis and treatments of this disease. The novelty of this study is prevalently the transcriptomic analysis performed. Although the oncogenic and transcriptional effects of the EWSR1::FLI1 fusion have been extensively studied, EwS is characterized by paucity of genetic abnormalities [26]. Based on recent findings [27], at the single cell transcriptome level, the variation of the expression of EWSR1::FLI1 is a major source of heterogeneity in EwS revealing that cells with high expression of EWS::FLI1 display proliferation and strong oxidative phosphorylation, whereas EWS::FLI1 low cell subpopulations show a signature associated with hypoxia [28]. Thus, the variation in the expression levels of the genetic driver may contribute to the phenotypic heterogeneity of EwS cells, potentially posing an additional challenge for clinical management.

The retrospective design of the study, the small number of cases considered, and the long period of time covered are the main weaknesses of this study. However, the strength of this case series is the selected location (extremities and trunk) and the genetic diagnosis confirmation obtained in 100% of cases. Additionally, most published series have reported on EEwS in children and adolescents, while our series is unique for having the largest number of patients in the adult age group (>16). Despite these limitations, we support the use of a similar therapeutic regimen between EEwS and BEwS due to the similar transcriptional profile. However, further studies, using cells from primary tumors, will be needed to assess whether the composition of tumors in these different compartments influences the response to treatment and the prognosis of tumors. Moreover, we presented a robust outcomes analysis related to patients with EEwS, adding more information to the limited data available on this rare disease.

As a heterogeneous group of musculoskeletal malignancies, EES should be studied at the transcriptome level. Despite the efforts to define EwS either in cell lines or inpatient-derived specimens [28‒30], the characterization of its transcription profile has remained mostly uninvestigated. To overcome this limited knowledge, we included in this study the transcriptomic profiling of 18 Ewing primary tissue samples, including 12 BEwSs and 6 EEwSs.

Based on the transcriptome analysis, we have highlighted a similar transcriptional profile in EEwS and BEwS, independently from the tissue of origin. This finding supports the idea that EwS is a single entity driven by the oncogenetic chimera which guides the clinical heterogeneity of EwS patients.

This result also supports the inclusion of EEwS in the therapeutic regimens of BEwS. These findings add novelty of this research topic. From previous studies, EEws has always been described as an important subtype of EwS that may require different treatment strategies [7, 31, 32].

Moreover, including both subtypes in a unique therapy group would be helpful to implement the availability of data regarding these pathologies, and molecular signatures could be used to improve patient prognosis. However, considering the extreme rarity of EEwS, any further knowledge requires international studies with the collaboration of experienced referral centers.

Among the 35 patients analyzed, 5-year OS was 67% compared to 70% reported in published studies [33, 34]. This result is fairly low compared to most reports in the extremity [23, 35] and it is probably related to the low use of local neoadjuvant RT (11/35).

Consistent with the available literature, the male preponderance of EEwS was also observed in our cohort [2, 13]. However, this correlation was not statistically significant. Ahmad et al. [2] reported that EEwS is more frequent in adults. Our results were in line with this prevalence. The median age of incidence was 25 (range 12–56), and only 17% of our patients were younger than 18. The effect of age on the prognosis of EEwS remains controversial. Some studies have demonstrated that being older is an independent risk factor for worse prognosis [2, 13], while others reported that this correlation is not significant [13, 34]. In this case series, age was not found to be a predictive factor for survival. Index surgery turned out to be of paramount importance with recurrent tumors adversely predicting OS, LRFS, MFS. Tumor location is a debated prognostic factor. Tural et al. [33] described tumor location as a significant risk factor affecting OS. Other authors report better prognosis for extremity lesions [2, 13, 36]. We were not able to audit the impact of tumor location on prognosis because our series mostly comprised tumors of the extremities. While this restriction selected a more homogeneous group for analysis compared to previous studies [22, 37, 38] and added novelty to data already published, this might have reduced the relevance of the variable for prognosis.

Tumors were found to be mostly deep-seated (80%), and the size was mainly under 8 cm (48%). There were no significant differences in OS between subgroups of different tumor depth. Furthermore, in line with previous studies [34, 37, 38], we found that a tumor size >8 cm was an independent risk factor for local control and OS. Results were not confirmed in patients with localized disease, probably because of the smaller cohort of patients.

Several studies suggest a relationship between increased LDH serum level and neoplastic maintenance, progression, and invasion in different subtypes of cancer [38]. This relationship has also been investigated for EEwS, and, although the results are debated, it seems that serum LDH values can be proposed as a predictive prognostic factor [34, 39‒41]. In our cohort, LDH was assayed in 31 patients, and an increased serum level was significantly correlated with poor OS, LRFS, and DMFS, presumably mirroring initial high tumor burden.

The presence of metastasis at diagnosis was a negative prognostic factor, with a 55% 5-year OS compared to 70% in non-metastatic patients. Contrary to expectation and in contrast with previous data [35, 42‒45], this difference was not statistically significant. OS, LRFS, or MFS estimates were not correlated with the type of local treatment approach strategy (surgery ± RT) and surgical margins. The existing evidence is based on a retrospective study probably affected by selection bias, where most patients received surgery and where RT was only indicated for patients with less favorable prognosis.

Most of the patients (94%) were treated with ChT, but the use of different protocols did not allow us to have enough patients to test for each kind of treatment. Further investigation including more patients per ChT group would be needed to determine which therapeutic protocol would be optimal. Even if ChT-induced necrosis cannot be easily assessed in soft tissue sarcomas, histological response to neoadjuvant ChT was evaluated with the Bologna System score reported by Picci et al. [13]. The prognostic significance of this variable has been proven mainly in BEwS [13, 45]. In our study, ood ZZs had a better prognosis, but these results were not confirmed in multivariate analysis.

The similar transcription profile between EEwS and BEwS supports the inclusion of the two subgroups in the same therapeutic regimens. However, recurrent tumors, big size, and altered LDH values should be considered to sort patients with EEwS and feasibly distinguish prognostic subgroups. These findings would make a compelling reason to include additional centers to increase the power of this portion of the study and to validate a leading treatment strategy.

We are grateful to BIOTUM – member of the CRB-IOR – which provided biological samples.

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of the Rizzoli Orthopedic Institute of Bologna (protocol code: 0012075, CE AVEC:505/2023/Oss/IOR and date of approval: August 2, 2023). Consent was not requested as per terms of the law (Article 110(b) of the Italian privacy law (Italian Legislative Decree no. 101(1) of August 10, 2018), which indicates that informed consent for retrospective studies is not mandatory for a Scientific Institute for Research, Hospitalization, and Healthcare (IRCCS).

The authors have no conflicts of interest to declare.

This study was not supported by any sponsor or funder.

G.B. and R.L.: conceptualization; R.L. and G.B.: methodology, investigation, and project administration; R.L. and M.A.L.: software and data curation; R.L., G.B., K.S., M.G., S.C., D.M.D., T.I., and A.P.: validation; M.A.L. and E.S.: RNA sequencing library preparation and bioinformatic analysis; G.B., R.L., M.G., K.S., and F.M.: resources; R.L., G.B., M.A.L., and F.M.: writing – original draft preparation; R.L., G.B., A.P., M.G., D.M.D., K.S., T.I., G.T., and F.O.: writing – review and editing; G.B.: visualization; and G.B., R.L., and D.M.D.: supervision. All the authors have read and agreed to the published version of the manuscript.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author (R.L.) upon reasonable request.