Activation of Wnt/β-Catenin in Ewing Sarcoma Cells Antagonizes EWS/ETS Function and Promotes Phenotypic Transition to More Metastatic Cell States

Ewing sarcomas are aggressive bone and soft tissue tumors with a high propensity for metastasis (1). Tumors are characterized by chromosomal translocations that generate tumor-initiating EWS/ETS fusion oncoproteins. The two most common translocations are t(11;22)q(24;12), which encode the EWS/FLI1 fusion protein in about 85% of cases, and t(21;22)(q22;q12), which encode EWS/ERG in 5%–10% of cases (1). Ewing sarcomas otherwise display few recurrent genetic mutations (2–4) and little is yet known about the molecular and cellular mechanisms that contribute to metastatic progression (1). Studies of Ewing sarcoma cells and tumors have, to date, identified several potential mediators of tumor relapse and metastatic progression, including activation of chemokine receptors (5, 6), tyrosine kinases (7), and WNT signaling (8, 9), as well as suppression of NOTCH signaling (10). Significantly, a recently described model of EWS/FLI1–induced transformation in primary murine cells also highlighted the Wnt/β-catenin axis as a key regulator of tumor initiation and progression (11). However, despite these findings, the contribution of Wnt/β-catenin signaling to Ewing sarcoma pathogenesis remains unclear.

Wnt/β-catenin signaling is required for both embryonic development and adult tissue homeostasis, and the pathway is often deregulated in epithelial cancers (12). Canonical Wnt ligands associate with frizzled protein and LRP5/6 protein coreceptor complexes at the cell membrane to activate Wnt signaling, resulting in inhibition of β-catenin degradation by the destruction complex, which includes the AXIN, APC, and GSK3β proteins. Stabilized β-catenin translocates to the nucleus where it can activate transcription by TCF/LEF transcription factors. The transcriptional targets of Wnt/β-catenin/TCF signaling are context- and cell type–dependent and their modulation can alter numerous cellular functions including proliferation, patterning, migration, and self-renewal (12, 13). In epithelial tumors, particularly colorectal cancer (14), inactivating mutations in genes encoding destruction complex proteins or activating mutations in β-catenin itself commonly deregulate the Wnt/β-catenin signaling pathway (14). Recent studies have also highlighted the possible significance of defects in the Wnt/β-catenin pathway in some sarcomas (15, 16); however, recurrent mutations have not been described in Ewing sarcoma (2–4). In addition to activating genetic mutations, Wnt/β-catenin signaling can also be aberrantly activated by deregulation of ligand–receptor interactions at the cell surface. In particular, high-level expression of R-spondin (RSPO) ligands and/or their receptors LGR4 and LGR5, have recently been shown to potentiate canonical Wnt signaling and promote tumorigenesis by inhibiting RNF43/ZNRF3–mediated turnover of the frizzled/LRP receptor complex (17–19). We recently showed that LGR5 is highly expressed by some Ewing sarcoma tumors and that LGR5⁺ cells potentiate Wnt/β-catenin signaling upon RSPO exposure (20). Moreover, our studies demonstrated increased expression of LGR5 in a group of rapidly progressive tumors, lending further support to the hypothesis that the Wnt/β-catenin pathway might contribute to an aggressive cellular phenotype (20).

In the current work, we evaluated activation of Wnt/β-catenin in Ewing sarcoma and defined the transcriptional and functional consequences of pathway activation in these tumors. Paradoxically, our findings reveal that activation of Wnt/β-catenin partially antagonizes EWS/ETS–dependent transcription and that this antagonism promotes phenotypic transition of tumor cells to cell states that promote migration and metastasis.

Ewing sarcoma cell lines were maintained in RPMI1640 media (Gibco) supplemented with 10% FBS (Atlas Biologicals) and 2 mmol/L l-glutamine (Life Technologies). Ewing sarcoma cell lines were kindly provided by Dr. Timothy Triche [Children's Hospital Los Angeles (CHLA), Los Angeles, CA; 2004] Dr. Heinrich Kovar (CCRI, St. Anna Kinderkrebsforschung, Vienna, Austria; 2010), and the Children's Oncology Group (COG) cell bank (cogcell.org; 2012). CHLA25, CHLA32, and STA-ET-8.2 were grown on plates coated with 0.2% gelatin. shA673-1C cells were kindly provided by Dr. Olivier Delattre (Institut Curie, Paris, France; 2014) and maintained as described previously (21). L-cells (ATCC CRL-2648) and Wnt3a L-cells (ATCC CRL-2647) were obtained from ATCC in 2011 (atcc.org) and were cultured in DMEM (Gibco) with 10% FBS. Cells were verified to be mycoplasma negative and identities confirmed by STR profiling every 6 months. Lentiviral production and transduction was performed as described previously (20), and the following plasmids were used: Addgene #24305 (7TGP), Addgene #24313 (EβP), Sigma TRCN0000230788 (shTNC-3), Sigma TRCN000015400 (shTNC-5), and UM-vector core pLentilox-luciferase/GFP (luc-tagged). Transduced cells were selected in puromycin (2 μg/mL).

Stably transduced 7TGP cells were stimulated with 1:1 RPMI:L-cell/Wnt3a–conditioned media (CM) ± 20 ng/mL recombinant RSPO2 (R&D Systems). Fluorescence was measured and quantified on an Accuri C6 cytometer (BD Biosciences) and cells sorted on the basis of GFP expression using the MoFlo Astrios instrument (University of Michigan Flow Cytometry Core). Peak GFP was detected at 48 hours and cells were sorted for gene expression studies at this time.

CHLA25-7TGP cells were stimulated with CM and FACS-sorted on the basis of GFP. Three biological replicates per sample were collected and three technical replicates for each of the samples were sequenced on the Illumina HiSeq 2000 (University of Michigan Sequencing Core). Fastq generation was performed using Illumina's CASAVA-1.8.2 software, analyzed for quality control using FASTQC, and aligned using Sailfish (22). For differential transcript expression analysis, RPKM was calculated and analyzed using Sailfish and the statistical R package, edgeR (23). RNAseq data have been deposited to GEO (GSE75859) and significant genes are listed in Supplementary Table S1. Gene ontology was determined using DAVID Bioinformatics Resources 6.7 (24). Overlapping datasets were identified using the Molecular Signatures Database v4.05 (MolSigDB) and genesets with false discovery rates (FDR) <0.05 (-log FDR >1.3) were considered significant (25). Gene-set enrichment analysis (GSEA) was performed using the GseaPreranked function of GSEA v2.1.0 software. To determine whether EWS/ETS targets were specifically enriched among Wnt-modulated genes, we performed χ² analysis as described previously (26). Wnt targets were validated by quantitative RT-PCR (qRT-PCR) using standard methods (20). Primer sequences are provided as Supplementary Table S2.

Sequential tumor tissue microarray (TMA) sections were probed with CD99 (BioCare Medical), β-catenin (BD Biosciences), and DAPI, and automated quantitative analysis (AQUA) was performed as described previously(27). RNA in situ hybridization (ISH) was performed using a commercially available kit and probes according to the manufacturer's instructions (Advanced Cell Diagnostics Inc.). Cells were grown on gelatin-coated coverslips and protein expression evaluated by standard techniques and immunofluorescence microscopy as described previously (20). Detailed protocols and antibody information are included in the Supplementary Methods.

Cells were lysed in 4% LDS (for FLI1 detection) or RIPA buffer (for ERG detection), and electrophoresis and transfer was performed using the Bio-Rad Mini-Protean Tetra System according to manufacturer's instructions. Primary antibodies FLI1 (Abcam EPR4646, 1:1,000) or ERG (Abcam #92513, 1:1,000) or GAPDH (Cell Signaling Technology, 1:5,000) were incubated in 5% milk overnight at 4°C, followed by incubation with secondary antibodies IRDye 800 and 680 (LI-COR, 1:5,000). Western blots were imaged and densitometry performed using the LI-COR imaging system and software.

A total of 1 × 10⁵ cells were added in serum-free RPMI to transwells containing 0.8-μm pores with serum-containing media in the bottom chamber. A total of 500 ng/mL of recombinant Wnt3a (R&D Systems) ± RSPO2 was added to the bottom chamber. After 24 hours, cells were fixed using a solution of 25% crystal violet and 25% methanol and membranes were imaged and then cells eluted using crystal violet in 10% acetic acid. Colorimetric absorbance (540 nm) was quantified on a BioTek plate reader.

Colony formation was assessed by plating a single-cell suspension of 1 × 10⁴ cells per well of 6-well plates in 0.35% noble agar on a layer of 0.5% noble agar. Colonies were stained with a solution of 0.005% crystal violet and counted three weeks later.

For subcutaneous tumor assays, 2.5 × 10⁵ stably transduced control (shNS) or tenascin C (TNC) knockdown (shTNC) cells were resuspended in PBS, diluted 1:1 in Matrigel, and injected subcutaneously into NOD/SCID mice (strain 394, Charles River Laboratories). Each mouse received an injection of shNS cells on the left flank and shTNC cells on the right flank, and tumor formation and growth was measured by calipers every other day. To assess lung engraftment, 1 × 10⁶ luciferase-labeled cells were injected via tail vein. Where indicated, cells were stimulated in vitro with either L-cell CM or Wnt3a CM+RSPO2 prior to injection. Satisfactory injection of viable cells was confirmed by detection of light emission in the lungs of recipient mice 30 minutes after cell injection (Xenogen IVIS bioluminescence system, Perkin Elmer). Imaging was repeated weekly and tumor burden quantified using Living Image software (Perkin Elmer). Mice were monitored for up to 7 weeks, at which point mice were euthanized. Tumor formation and location was confirmed by dissection with the help of Pathology Cores for Animal Research within the Unit for Laboratory Animal Medicine at the University of Michigan.

Outcomes were interrogated in a recently published, clinically annotated dataset from the COG (GSE 63157; ref. 6). Cox regression and Kaplan–Meier analyses were conducted for determination of overall survival and event-free survival relative to median and tertile of LEF1 expression. Log-rank tests were used to compare survival distributions.

Unless otherwise indicated, data are expressed as mean ± SEM from a minimum of three independent experiments. The data were analyzed using GraphPad Prism software by Student t test, χ², and Fisher exact tests, and P < 0.05 considered significant.

Ewing sarcoma cell lines activate Wnt/β-catenin signaling in response to exogenous Wnt3a and RSPO2 treatment and our prior studies revealed heterogeneity of response among cell lines (20). To better characterize the basis of the heterogeneous Wnt response, we exposed Ewing sarcoma cells to control L-cell–conditioned media (CM) or Wnt3a CM and assessed β-catenin localization in individual cells. While β-catenin was largely lacking under control conditions (Fig. 1A), stimulation with Wnt3a CM led to an increase in nuclear expression (Fig. 1B). Notably, robust nuclear β-catenin staining was evident in some cells but little to no β-catenin was observed in neighboring cells (Fig. 1B). To determine whether this heterogeneity resulted in heterogeneous activation of β-catenin–dependent transcription, we evaluated TCF transcriptional activity in GFP-reporter cells (7TGP). Although all cell lines showed an increase in the proportion of GFP-positive cells following Wnt3a stimulation, the responses were highly variable (Fig. 1C). Addition of RSPO2 potentiated the TCF-dependent transcriptional response but, again, heterogeneity persisted within and between cell lines (Fig. 1C). Thus, individual Ewing sarcoma cells display marked heterogeneity in their responsiveness to exogenous Wnt/β-catenin–activating ligands in vitro.

To determine whether Wnt/β-catenin signaling is similarly heterogeneously activated in vivo, we performed β-catenin AQUA immunofluorescence staining of a Ewing sarcoma TMA. The majority of over 50 evaluable biopsies showed no evidence of β-catenin expression; however, elevated cytoplasmic and nuclear β-catenin expression was detected in eight tumor samples (Fig. 1D). In these tumors, expression was limited to discrete subpopulations of tumor cells (Fig. 1E) confirming heterogeneity both within and between individual tumors.

Given that recurrent mutations in Wnt/β-catenin pathway genes have not been identified in Ewing sarcoma (2–4), and that RSPO2 signaling can augment the response of cells to Wnt3a stimulation (20), we hypothesized that heterogeneity in LGR5 expression may contribute to heterogeneity of β-catenin activation. To address this, we used ISH to measure LGR5 mRNA in tumor biopsies. Nonquantitative ISH-identified LGR5⁺ tumor cells in 8 of 26 evaluable tumors and cell-to-cell heterogeneity was again apparent (Fig. 1F). To determine whether LGR5⁺ tumor cells were more likely to activate β-catenin, we performed fluorescence-based, semiquantitative ISH on a TMA section adjacent to the section that was used for AQUA studies. This analysis affirmed that only a small proportion of tumors expressed high levels of LGR5 (Fig. 1G) and showed that the tumors with evidence of robust LGR5 expression were the same as those that displayed nuclear β-catenin staining (Fig. 1H). These data corroborate our in vitro studies (20) and lend support for the hypothesis that Ewing sarcoma cells that express high levels of LGR5 are more responsive to ligand-dependent activation of Wnt/β-catenin signaling.

We previously reported that LGR5 mRNA expression was increased in a small cohort of patients with unusually aggressive disease (20). Review of the clinical data for the current TMA samples showed that all tumors with evidence of β-catenin activation were obtained from patients who later relapsed. To more definitively address whether activation of Wnt/β-catenin signaling is associated with outcome, we turned to a recently published cohort of patients who were diagnosed with localized Ewing sarcoma and treated on COG therapeutic studies (6). To infer Wnt/β-catenin activation from the available gene expression data, we first needed to identify a biomarker of active β-catenin signaling. We transduced Ewing sarcoma cells with a constitutively active β-catenin construct (28) and measured induction of established Wnt/β-catenin transcriptional targets. Basal expression of LEF1 was low in control cells but more than 100-fold induced by activated β-catenin (Fig. 2A). In contrast, AXIN2 was more modestly induced and LGR5 was unchanged (Fig. 2A). Therefore, we selected LEF1 as our biomarker and discovered that, consistent with heterogeneity in β-catenin activation, levels of LEF1 were highly variable in the patient tumors (Fig. 2B). In this clinically homogeneous patient population, high expression of LEF1 was significantly associated with worse event-free (Fig. 2C) and overall survival (Fig. 2D).

Having established that Wnt/β-catenin activation is associated with worse clinical outcomes, we next sought to define why this might be. To achieve this we used 7TGP-CHLA25 reporter cells to compare the transcriptional profiles of Wnt/β-catenin activated cells with nonactivated controls (Fig. 3A). RNAseq analysis identified differential expression of over 1,000 transcripts in response to Wnt3a alone or in combination with RSPO2 (Supplementary Fig. S1A; Supplementary Table S1). AXIN2, LEF1, and NKD1 were among the previously established Wnt/β-catenin target genes that were also induced in Ewing sarcoma (Fig. 3B). Conversely, other established target genes were unaffected, demonstrating the cell-type specificity of this signaling axis (Fig. 3B). This differential Wnt response was validated by independent qRT-PCR analyses of multiple Ewing sarcoma cell lines (Supplementary Fig. S1B and S1C).

To gain functional insights into the Wnt/β-catenin transcriptional response, we compared our list of significant genes to published genesets (25). Strikingly, genes that were repressed 2-fold in Wnt/β-catenin activated cells overlapped most significantly with genes that are induced by EWS/FLI1 (Fig. 3C, top; ref. 29). Conversely, genes that were induced by Wnt/β-catenin activation overlapped most significantly with genes that were repressed by EWS/FLI1 (Fig. 3C, bottom). This remarkable overlap between EWS/ETS and Wnt/β-catenin targets in CHLA25 cells was highly significant and not merely a reflection of generalized transcriptional activation as demonstrated by specific enrichment of EWS/ERG target genes in the Wnt/RSPO–responsive geneset (Fig. 3D). To determine whether the overlapping genes were indicative of a more generalized inverse relationship between Wnt/β-catenin and EWS/ERG transcriptional activity, we expanded our analysis to include all genes that were significantly modulated by Wnt/β-catenin irrespective of fold change. GSEA of rank-ordered genes confirmed an inverse correlation between Wnt/β-catenin and EWS/ETS transcriptional targets (Fig. 3E and Supplementary Fig. S2A). Thus, transcriptional antagonism exists between EWS/ERG and Wnt/β-catenin in CHLA25 cells, corroborating a prior study of A673 cells, which reported an inverse relationship between EWS/FLI1 and TCF/LEF target genes (30).

To evaluate whether transcriptional antagonism between EWS/ETS and Wnt/β-catenin signaling exists in primary tumors, we performed GSEA on gene expression data from the aforementioned COG study. Genes that were positively correlated with LEF1 in these tumors were highly enriched for genes that are normally repressed by EWS/ETS fusions (Supplementary Fig. S2B). These data corroborate a prior study of gene expression in a different tumor dataset wherein a correlation was also discovered between levels of LEF1 and EWS/ETS–repressed targets (30). Together, these data provide compelling evidence that activation of Wnt/β-catenin antagonizes the transcriptional function of EWS/ETS fusions in Ewing sarcoma.

To gain insight into the potential mechanism of transcriptional antagonism, we assessed whether activation of Wnt/β-catenin signaling leads to changes in the levels of EWS/ETS fusions. To ensure that we compared expression of the fusions in truly Wnt/β-catenin–active and -inactive cells, we measured EWS/ETS levels after FACS sorting of 7GP cells. Significantly, reduced expression of EWS/FLI1 transcript was reproducibly seen in Wnt/β-catenin activated cells (Fig. 4A) and a trend to reduced protein expression was also observed (Fig. 4B). In contrast, no significant change in EWS/ERG levels was observed in Wnt-activated CHLA25 cells (Fig. 4C and D). These findings suggest that either Wnt/β-catenin can inhibit EWS/FLI1 transcription or that heterogeneity in expression of EWS/FLI1 prior to ligand exposure determines Wnt responsiveness. In support of this latter possibility, Navarro and colleagues (30) previously showed that cells with reduced expression of EWS/FLI1 were more responsive to Wnt3a, an observation that we have also now independently validated (Supplementary Fig. S3A–S3C).

We next sought to understand the functional consequences of Wnt/β-catenin–dependent antagonism of EWS/ETS activity. To achieve this, we focused on the 236 genes that were modulated by Wnt/β-catenin in CHLA25 cells and EWS/FLI1 in A673 cells (Fig. 4E; ref. 29). Hierarchical clustering of these genes confirmed both the opposing pattern of regulation and the potentiating effect of RSPO2 (Fig. 4E). Gene ontology analysis of the overlapping genes revealed that microtubule genes involved in cell-cycle regulation were most prominent among EWS/FLI1–induced/Wnt-repressed genes (Fig. 4E and Supplementary Fig. S3D). However, activation of Wnt/β-catenin does not lead to significant inhibition of Ewing sarcoma cell proliferation, at least in vitro (Supplementary Fig. S3E; ref. 20). Among EWS/FLI1–repressed/Wnt-induced genes, actin cytoskeleton genes were enriched (Fig. 4E). Thus, activation of Wnt/β-catenin signaling in Ewing sarcoma results in altered expression of EWS/ETS target genes that are involved in regulation of the microtubule and actin cytoskeletons.

Knockdown of EWS/ETS in Ewing sarcoma cells leads to induction of actin stress fibers, acquisition of a migratory cell phenotype, and enhanced tumor cell adhesion in the lung, phenotypes that are, in part, mediated by derepression of zyxin (ZYX; refs. 31, 32). Our observation that activation of Wnt/β-catenin leads to derepression of numerous cytoskeleton genes (Fig. 5A), including ZYX, led us to test whether activation of Wnt/β-catenin phenocopied EWS/ETS knockdown. Consistent with EWS/ETS loss-of-function, exposure of tumor cells to activating Wnt ligands led to the formation of actin stress fibers (Fig. 5B and Supplementary Fig. S4A), an increase in cell spreading (Supplementary Fig. S4B), and enhanced migration (Fig. 5C). Furthermore, podosomes, actin-rich cytoskeletal structures that are essential for cell migration (33, 34), were increased in Wnt/β-catenin–activated cells (Supplementary Fig. S4C). Similar results were obtained in cells that expressed constitutively active β-catenin, confirming that the Wnt-stimulated phenotype is, at least in part, mediated by β-catenin (Supplementary Fig. S4D and S4E). To test whether activated Wnt/β-catenin signaling promotes lung engraftment of tumors, Wnt-stimulated and nonstimulated cells were injected via tail vein and tumor engraftment monitored by bioluminescence imaging. Stimulation of A673 cells with Wnt3a +RSPO2 prior to injection resulted in a trend toward earlier onset of tumors (Fig. 5D), a higher rate of tumor formation overall (9/10 vs. 5/10 mice; Fig. 5E), and a significant increase in the formation of lung tumors (Fig. 5F). However, no statistically significant difference was seen in TC32 cells and all mice that developed tumors showed evidence of lung engraftment, irrespective of ex vivo Wnt stimulation (Supplementary Fig. S4F–S4H). Thus, β-catenin activation reproducibly activates cytoskeletal changes in Ewing sarcoma cells that are associated with enhanced cell migration in vitro and, in some cases, enhanced lung engraftment.

We next interrogated the list of Wnt/β-catenin–induced/EWS/ETS–repressed genes to identify genes that might contribute to migration and lung engraftment. In addition to ZYX, this analysis identified TNC (see Fig. 5A). TNC encodes tenascin C, a secreted matricellular protein that is essential for successful lung engraftment of breast cancer cells (35). In addition, TNC has also been implicated as a mediator of tumor metastasis in other tumor types and at multiple levels of the metastatic cascade (36). qRT-PCR and immunocytochemistry studies confirmed that TNC is induced by Wnt/β-catenin in Ewing sarcoma (Fig. 6A–C) and is also subject to EWS/FLI1–dependent repression (Supplementary Fig. S5A).

We next assessed the impact of TNC loss-of-function (Fig. 6D and E). Knockdown of TNC had no effect on cell proliferation in vitro but strongly inhibited anchorage-independent colony formation (Supplementary Fig. S5B and S5C). TNC knockdown cells showed diminished migration, both in standard culture conditions (Fig. 6F) and in response to Wnt/RSPO activation (Fig. 6G), demonstrating that Wnt/β-catenin–induced migration is, in part, mediated by TNC. With respect to in vivo tumorigenicity, TNC knockdown cells reproducibly showed a reduced capacity to form lung tumors following tail vein injection (Fig. 7A and B). Subcutaneous tumor formation was also abrogated in A673 cells following TNC knockdown (Fig. 7C) but was only mildly impacted in TC32 cells (Fig. 7D). We speculate that the effect of knockdown on subcutaneous engraftment was more pronounced in A673 than TC32 cells due to fact that the absolute levels of TNC expression in A673 cells were lower than in TC32 cells, both at baseline and following knockdown (Fig. 7D). Likewise, we hypothesize that high basal levels of TNC in TC32 cells might make them more amenable to lung engraftment than A673 cells, thus obviating a requirement for additional Wnt/β-catenin–dependent activation.

Together, these data demonstrate that TNC plays a key role in promoting Ewing sarcoma cell migration and lung engraftment. Activation of Wnt/β-catenin signaling in Ewing sarcoma cells leads to upregulated expression of TNC and this contributes to a more tumorigenic cell state. Overall these data support a model in which activation of Wnt/β-catenin leads to upregulation of metastasis-promoting genes that are normally repressed by EWS/ETS fusions and this, in turn, leads to an increased risk of metastasis and worse clinical outcomes (Fig. 7E).

In the current study, we investigated the impact of activated Wnt/β-catenin signaling on Ewing sarcoma. The combined results of these studies demonstrate that activation of canonical Wnt/β-catenin signaling, by ligands in the tumor microenvironment, results in more clinically aggressive disease. In keeping with the microenvironment being the source of Wnt/β-catenin activation in these tumors, recurrent somatic mutations in the pathway are rare (2–4, 15) and Wnt and RSPO ligands are highly expressed in developing bone, the predominant site of Ewing tumors in pediatric and young adult patients (37, 38). Mechanistically, our data show that the aggressive clinical phenotype of Wnt/β-catenin–activated tumors is mediated, at least in part, by a paradoxical antagonism of EWS/ETS–dependent transcription, which results in activation of metastasis-associated genes that are normally subject to EWS/ETS–dependent repression. In particular, we have shown that the extracellular matrix encoding gene, TNC, respectively, is induced by Wnt/β-catenin activation in Ewing sarcoma and that high-level TNC promotes cell migration, tumorigenicity, and lung engraftment.

Tumor cell heterogeneity can arise from genetic as well as from nonmutational mechanisms and nongenetic heterogeneity is emerging as a major contributor to tumor progression and relapse (39–41). As an example, reversible plasticity of tumor cells between proliferative and migratory states, such as occurs in epithelial-to-mesenchymal transitions, has been linked to metastasis (42, 43) and epigenetic plasticity of chromatin states promotes emergence of drug resistance (44). In concordance with the findings of Endo and colleagues (45), who described neurite outgrowth in response to Wnt/β-catenin activation, we also found that Ewing sarcoma cells display cytoskeletal plasticity in response to Wnt stimulation. However, our studies revealed that not all Ewing sarcoma cells respond equally to Wnt stimulation, and that responsiveness can vary widely even between neighboring tumor cells in vitro and in vivo. Numerous factors likely contribute to this heterogeneity of response. Expression of the Wnt-potentiating ligand, LGR5, was found to be heterogeneous on a cell-to-cell basis and we have previously shown that the presence of LGR5 directly impacts on the Wnt response when RSPO ligands are present (20). Variability in the availability of endogenous or exogenous ligands, as well as heterogeneous expression of Wnt-modulating receptors, could all impact on an individual cell's ability to activate β-catenin. In addition, other cell-intrinsic factors may be important determinants of the Wnt response. For example, G₂–M-phase is, at least in some contexts, most permissive for activation of Wnt signaling (46). Moreover, in colorectal cancer, nuclear β-catenin is heterogeneous among tumor cells, thus revealing that other factors contribute to pathway activation even when activating mutations are universally present (47). In lung adenocarcinoma, cells isolated from metastatic lesions were more Wnt responsive than their corresponding primary tumors, and aggressive phenotypes were enhanced by exposure to conditioned media from metastatic sites (48). Thus, variability in Wnt responsiveness among tumor cells can be dictated by the differences in the signaling microenvironment, as well as by differences in cell-intrinsic factors.

The results of our analyses of gene expression data from both Wnt3a/RSPO2-stimulated cells as well as primary tumor data revealed a striking transcriptional antagonism between Wnt/β-catenin/LEF1 and EWS/ETS. Our finding that Wnt stimulation inhibits EWS/ETS is concordant with results from an earlier study in which EWS/FLI1 was shown to inhibit β-catenin/TCF activity. Thus, the data from Navarro and colleagues and from our current work show that antagonism between these transcriptional axes is bidirectional. The mechanistic basis of this antagonism remains to be fully elucidated but the studies of Navarro and colleagues suggest that physical interactions between EWS/FLI1 and LEF1 inhibit transcription of TCF/LEF target genes in response to β-catenin activation (30). As LEF1 is strongly induced by Wnt/β-catenin in Ewing sarcoma cells, we speculate that this might lead to sequestration of EWS/ETS proteins by LEF1 and inhibition of EWS/ETS–dependent transcription. However, it should be noted that expression of many established EWS/ETS target genes was unaffected by Wnt/β-catenin activation, demonstrating that the inverse relationship between Wnt/β-catenin and EWS/ETS does not extend to all targets of the fusion. Further investigations are required to elucidate the molecular mechanisms that account for differential responses of different EWS/ETS target genes to Wnt/β-catenin activation. Given that EWS/ETS fusions effect changes to transcription through a variety of epigenetic mechanisms (49–51), it is likely that cooperation between EWS/ETS fusions, TCF/LEF transcription factors, and epigenetic regulators at target gene promoters and enhancers collectively determine the transcriptional output of Wnt/β-catenin signaling in Ewing sarcoma cells.

In summary, these studies provide compelling biological and clinical evidence that activation of Wnt/β-catenin signaling contributes to an aggressive cellular phenotype in Ewing sarcoma tumors. Moreover, they reveal a previously unappreciated degree of inter- and intratumoral heterogeneity in Wnt/β-catenin activation that is mediated by both cell autonomous (e.g., expression of LGR5) and non-cell–autonomous (e.g., availability of Wnt and RSPO ligands in the tumor microenvironment) factors. These findings support further investigation of Wnt/β-catenin pathway–targeted agents as potential adjuvant therapies for metastasis prevention in Ewing sarcoma patients.

No potential conflicts of interest were disclosed.

Conception and design: E.A. Pedersen, D.G. Thomas, E.R. Lawlor

Development of methodology: E.A. Pedersen, K.M. Bailey, D.G. Thomas, E.R. Fearon, E.R. Lawlor

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Pedersen, D.G. Thomas, R.A. Van Noord, J. Tran, R. Chugh, E.R. Lawlor

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Pedersen, R. Menon, K.M. Bailey, D.G. Thomas, H. Wang, P.P. Qu, A. Hoering, E.R. Lawlor

Writing, review, and/or revision of the manuscript: E.A. Pedersen, R. Menon, D.G. Thomas, H. Wang, E.R. Fearon, R. Chugh, E.R. Lawlor

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.A. Van Noord, E.R. Lawlor

Study supervision: E.R. Lawlor

Other (RNA and technical work): J. Tran

We would like to thank the Lawlor lab members for their helpful discussions and Dr. Patrick Grohar for FLI1 Western blot protocol.

This work was supported by the SARC Sarcoma SPORE NIH Grant U54 CA168512, an Alex's Lemonade Stand Foundation Innovation Award and Ruth L. Kirschstein NRSA F30CA183276 (E.A. Pederson). Additional support was provided by the University of Michigan's Cancer Center Support Grant (P30 CA046592) by the use of the following Cancer Center Core(s): flow cytometry, cellular imaging, sequencing, and vector.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.