Recently, Charville et al. [1] reported that EWSR1-rearranged fusion proteins mediate the expression of the paired-box transcription factor PAX7 in Ewing sarcoma. Based on an analysis of a published gene expression microarray dataset (accession code GSE60740), they state having identified PAX7 to be significantly overexpressed in Ewing sarcoma in comparison to CIC-DUX4-positive round cell sarcomas [1]. In that microarray analysis they compared CIC-DUX4-positive sarcomas with EWSR1-NFATc2-positive sarcomas, assuming that EWSR1-NFATc2-positive sarcomas belong to the family of Ewing sarcomas [1], which are typically characterized by EWSR1-ETS fusion oncogenes [2]. Accordingly, Charville et al. [1] summarized in a schematic EWSR1-FLI1-, EWSR1-ERG-, and EWSR1-NFATc2-positive sarcomas as “Ewing sarcoma,” referring to Szuhai et al. (2009) [3], and did not take into account more recent reports in the literature that EWSR1-NFATc2-positive sarcomas may constitute an own entity [2]. Comparison of the dataset (GSE60740) used by Charville et al. to a published transcriptome reference dataset of genetically defined EWSR1-ETS-positive Ewing sarcomas (GSE34620) [4] and 13 other malignancies that may constitute morphological mimics [5] shows that EWSR1-NFATc2-positive sarcomas do not cluster with any other analyzed tumor entity including EWSR1-ETS-positive Ewing sarcoma (Fig. 1a).

Fig. 1
figure 1

EWSR1-NFATc2-positive sarcomas exhibit highly divergent transcriptomes from EWSR1-ETS-positive Ewing sarcomas. a t-SNE cluster analysis of Ewing sarcoma (EWSR1-ETS-positive) and other morphological mimics. All gene expression data were generated on Affymetrix HG-U133-Plus2.0 microarrays and simultaneously normalized using RMA and brainarray CDF [5], yielding one optimized probe-set per gene [13]. Gene expression data were filtered for a minimum tumor purity of 40% as determined by the ESTIMATE algorithm [14]. Principal component analysis (PCA) using all genes represented on this microarray platform was done and visualized with the Rtsne Bioconductor package in R15. DLBCL diffuse large B-cell lymphoma, ARMS alveolar rhabdomyosarcoma. b Comparison of gene expression levels of FOXG1, NR5A2, SOX5, VDR, and TFAP2 in EWSR1-ETS-positive Ewing sarcomas (EWS), EWSR1-NFATc2-positive sarcomas, and CIC-DUX4-positive sarcomas (GSE34620 and GSE60740). Data are presented as dot plots. Horizontal bars indicate median expression levels. Whiskers indicate the interquartile range

Full size image

Furthermore, Charville et al. noted that (in addition to PAX7) the genes FOXG1, NR5A2, SOX5, VDR, and TFAP2 were highly overexpressed in EWSR1-NFATc2-positive sarcomas as compared to CIC-DUX4-positive sarcomas [1]. Interestingly, all these genes are similarly lowly expressed in EWSR1-ETS-positive Ewing sarcomas and CIC-DUX4-positive sarcomas (Fig. 1b). Collectively, these analyses strongly suggest that EWSR1-NFATc2-positive sarcomas constitute a tumor entity distinct from Ewing sarcoma [5]. Hence, the correctness of extrapolation from EWSR1-NFATc2-positive sarcomas to Ewing sarcoma remains to be substantiated.

Based on their disputable extrapolation from EWSR1-NFATc2-positive sarcomas to Ewing sarcomas, Charville et al. [1] performed immunohistochemical analyses of tissue samples, which were “diagnosed as Ewing sarcoma using a combination of morphologic, immunohistochemical, and molecular features.” For the majority of cases (102/103), which were all “PAX7-positive” by immunohistochemistry, no details or references were provided on how the diagnosis of Ewing sarcoma was molecularly confirmed. Only for one sample, which was classified as “PAX7-negative,” details were provided, but the diagnosis of Ewing sarcoma could not be molecularly confirmed [1]. The authors conclude from this sample set that “immunohistochemical detection of anti-PAX7 immunohistochemistry is a sensitive marker for Ewing sarcoma” [1]. Yet, it remains unclear whether the sensitivity of PAX7 is superior to that of the established Ewing sarcoma marker CD99 [6]. In their transcriptome analysis of EWSR1-NFATc2-positive sarcomas, Charville et al. noted that CD99 was not significantly overexpressed compared to CIC-DUX4-positive sarcomas [1], supporting the concept that EWSR1-NFATc2-positive sarcomas are distinct from Ewing sarcoma, which typically shows high and uniform CD99 expression [6]. The proposed utility of PAX7 as a marker for Ewing sarcoma might be diminished given its limited specificity: In a previous report Charville et al. showed very high expression of PAX7 in morphological mimics such as embryonal, alveolar, and pleomorphic rhabdomyosarcomas, as well as synovial sarcomas [7], and in the current report they provide first evidence for its high expression in EWSR1-NFATc2-positive sarcomas [1].

Though PAX7 has been shown to be highly expressed in morphological mimics not positive for EWSR1-regarrangements such as rhabdomyosarcoma and synovial sarcoma [7], Charville et al. explain the high expression of PAX7 in “Ewing sarcoma” by EWSR1-FLI1 binding to a GGAA-microsatellite sequence located 20 kb telomeric to the PAX7 gene [1], which shows epigenetic characteristics of an active enhancer element in published ChIP-Seq data [8].

However, binding of EWSR1-FLI1 to this DNA sequence, EWSR1-FLI1-dependent enhancer activity, and a true interaction of this potential enhancer with the PAX7 promoter were not confirmed experimentally. Yet, Charville et al. deduce from their approach that “these analyses provide evidence that a cis regulatory mechanism of EWSR1-FLI1 binding and enhancer activation leads to expression of PAX7 in Ewing sarcoma” [1]. In the corresponding figure, two additional and more proximal EWSR1-FLI1-bound DNA elements are shown (the first mapping to another GGAA microsatellite, the second to a canonical ETS-like binding motif) with EWSR1-FLI1-dependent epigenetic characteristics of active enhancers, which were not further investigated or discussed by the authors [1]. Given that there are thousands of such EWSR1-FLI1-bound GGAA microsatellites in the vicinity of genes in Ewing sarcoma tumors [8, 9], it appears questionable to conclude that any EWSR1-FLI1-bound enhancer actually represents the cis regulatory element controlling the expression of a certain nearby gene. Previous reports showed that such EWSR1-FLI1-bound GGAA microsatellites can loop to and accordingly regulate genes located in great distances (more than 300 kb) [8]. Thus, the potential enhancer reported by Charville et al. could control the expression of any gene in cis or trans. Taking into account that GGAA microsatellites underlie substantial germline variability and that the enhancer activity of EWSR1-ETS-bound GGAA microsatellites strongly depends on the number of consecutive GGAA-repeats [10,11,12], it seems improbable that a highly polymorphic enhancer-like DNA element should lead to a uniformly high expression of PAX7 in Ewing sarcoma as reported by Charville et al. [1]. Nevertheless, the authors conclude that EWSR1 fusion proteins are required for PAX7 expression in Ewing sarcoma by this mechanism. Given that ChIP-Seq analyses of Charville et al. are restricted to EWSR1-FLI1 [1], that there is as yet no convincing evidence that EWSR1-NFATc2 can bind to GGAA microsatellites as well, and that immunohistochemical analysis of PAX7 on EWSR1-NFATc2-positive sarcomas has not been performed, this conclusion is critical.

We deduce from the report of Charville et al. that particular care must be taken when analyzing published “omics”-data before conclusions may be drawn.