EWS and RE1-Silencing Transcription Factor Inhibit Neuronal Phenotype Development and Oncogenic Transformation in Ewing Sarcoma

Identification of EWS-regulated genes and cellular processes in Ewing sarcoma

To characterize the function of EWS in Ewing sarcoma, we silenced EWS in A673 Ewing sarcoma cells (Fig. 1A) and performed high-throughput sequencing of RNA (RNA-seq) from control (luciferase) or EWS knockdown cells to identify EWS-regulated genes. Sequencing reads were mapped to Ensembl annotations (www.ensembl.org), and expression levels of genes based on the Ensembl annotation are shown in Supplementary File S1. Genes were ranked by the mean ± standard deviation of log-transformed FPKM (fragments per kilobase per million mapped reads) and shown as a heat map in Figure 1B. To gain insight into the functional significance of the differentially expressed genes, we performed DAVID functional annotation analysis (david.abcc.ncifcrf.gov) of 99 genes that pass the filter of a 5% false discovery rate (FDR) and log₂ ratio >1 or <–1. We found that these genes are associated with diverse functions, including those that have previously been indicated for EWS in non–Ewing sarcoma cells, such as a response to various cellular stresses, as well as previously unidentified functions including cell signaling, secretion, blood vessel development, and neuronal-related processes (Fig. 1C and Suppl. File S2). A subset of EWS–up-regulated and –down-regulated genes was randomly selected and validated by qRT-PCR) (Fig. 1D and 1E).

OPEN IN VIEWER

Figure 1.

Identification of EWS-regulated genes and functions in Ewing sarcoma. (A) EWS knockdown by shRNA. qRT-PCR analysis shows that the EWS transcript level decreased about 80% by EWS knockdown. Normalized fold change was calculated by determining the fold change of the EWS-RNAi condition relative to the control Luc-RNAi condition, with the data in each condition normalized to an internal housekeeping control gene GAPDH. Columns indicate the mean of 3 independent replicate experiments, error bars indicate the standard deviation, and asterisks indicate P < 0.05. EWS and tubulin (loading control) protein levels after control or EWS shRNA treatment are shown in the bottom panel. (B) Expression profiles for all detected and rank-ordered Ensembl genes are represented as a heat map. The FPKM values were mean-centered and normalized, with each row representing a different gene. The top 20 genes that either increase (left) or decrease (right) with increased EWS are shown. (C) Top 10 categories identified by DAVID functional annotation analysis of EWS-regulated genes. (D, E) RT-PCR validation of randomly selected EWS–up-regulated (D) or –down-regulated (E) genes. Normalized fold change was calculated by determining the fold change of the EWS-RNAi condition relative to the control Luc-RNAi condition, with the data in each condition normalized to an internal housekeeping control gene GAPDH. Columns indicate the mean of 3 independent replicate experiments, error bars indicate the standard deviation, and asterisks indicate P < 0.05.

EWS-regulated genes are differentially regulated by EWS/FLI

Because EWS/FLI has previously been shown to interfere with EWS functions,^12,13 we next sought to determine the relationship between EWS and EWS/FLI in regulating cellular processes in Ewing sarcoma. We performed RNA-seq following EWS/FLI silencing and compared the EWS/FLI-regulated transcriptional profile (Suppl. File S3) with that of EWS. We used VennMaster analysis (informatik.uni-ulm.de/ni/staff/HKestler/vennm) to identify genes commonly regulated by EWS and EWS/FLI. Fifty-three of the 99 EWS-regulated genes were found to also be regulated by EWS/FLI (P = 3.86345E-30) (Fig. 2A). Of these 53 EWS and EWS/FLI commonly regulated genes, 14 genes were up-regulated by EWS and down-regulated by EWS/FLI (P = 4.30758E-9), and 13 genes were down-regulated by EWS and up-regulated by EWS/FLI (P = 4.62419E-8), suggesting that the 2 proteins inversely regulate a significant subset of target genes (Fig. 2B). Interestingly, we also identified 7 genes that were up-regulated (P = 0.00897) and 19 genes that were down-regulated (P = 4.19203E-14) by both EWS and EWS/FLI (Fig. 2B), indicating that the 2 proteins also regulate a significant subset of genes in the same manner. We next performed Gene Set Enrichment Analysis (GSEA) using the EWS/FLI-regulated genes as the ranked list and the EWS–up-regulated or –down-regulated targets as the gene sets and applied a χ² test ²⁹ to establish enrichment of the EWS-regulated genes for both the EWS/FLI–up-regulated and –down-regulated genes. We found that the EWS–up-regulated genes cluster significantly (P ≤ 0.001) with both the EWS/FLI–up-regulated and –down-regulated genes and vice versa (Fig. 2C), suggesting a correlated regulation of genes by EWS and EWS/FLI in either the same or opposite directions. Notably, the correlation seemed to be stronger for the EWS–down-regulated genes as compared to the EWS–up-regulated genes. These results suggest that EWS and EWS/FLI differentially regulate genes and cellular processes in Ewing sarcoma.

OPEN IN VIEWER

Figure 2.

EWS/FLI differentially regulates EWS-regulated genes. (A) Venn diagram shows a significant overlap of the EWS- and EWS/FLI-regulated gene sets. (B) Venn diagrams showing significant overlapping gene sets between the EWS–up-regulated or –down-regulated and the EWS/FLI–up-regulated or –down-regulated gene sets. The P value was calculated by χ² analysis. (C) GSEA using EWS/FLI-regulated genes in A673 Ewing sarcoma cells as the rank-ordered data set and the EWS–up-regulated or –down-regulated targets as the gene set. The χ² test–derived enrichment scores and P values are shown for each end of the GSEA curve.

EWS inhibits a subset of REST target neuronal genes in Ewing sarcoma cells

Because Ewing sarcoma was found to display some neuronal features, we focused our study on the EWS-regulated genes in the category of neuronal-related processes. We noted a set of neuronal genes including chromogranin A (CHGA), cholinergic receptor nicotinic beta 2 (CHRNB2), pleckstrin and Sec7 domain containing (PSD), secretogranin III (SCG3), synaptotagmin IV (SYT4), and VGF nerve growth factor inducible (VGF). These 6 genes are involved in different aspects of neuronal functions^{30 -35} and are all down-regulated by EWS. Interestingly, elevated CHGA is an indicator for pancreas and prostate cancers. ³⁶ Detection of SCG3 and VGF transcripts is a prognostic biomarker for small cell lung cancer and large cell neuroendocrine carcinoma,^37,38 respectively. More importantly, a putative REST response element (neuron-restrictive silencer element [NRSE]) is found in the regulatory regions such as the promoter or 5′ untranslated region (5′-UTR) of each of these genes (Suppl. File S4). Most of these genes are known to be repressed by REST as well.^{39 -41}

To test if REST is involved in the repression of these neuronal genes in Ewing sarcoma, and to validate the RNA-seq data that EWS down-regulates the same subset of neuronal genes, we performed shRNA- and siRNA-mediated silencing of EWS or REST and analyzed the expression of these genes by qRT-PCR (Fig. 3A and Suppl. Fig. S1). Stable knockdown of EWS or REST by shRNA significantly increased the expression of each of these neuronal genes in both A673 and TC71 cells (Fig. 3A and Suppl. Fig. S1A). Transient knockdown of EWS or REST using siRNA (Suppl. Fig. S1B) generated the same pattern of increased expression for each of these neuronal genes (Suppl. Fig. S1C). These findings demonstrate the repression of a subset of neuronal genes by REST in Ewing sarcoma, confirm the EWS RNA-seq results, and suggest a direct connection between EWS and REST in the regulation of neuronal gene expression.

OPEN IN VIEWER

Figure 3.

EWS and REST repress neural gene expression in Ewing sarcoma cells. (A) qRT-PCR analysis showing that EWS and REST down-regulate the transcription of all 6 neural genes. Normalized fold enrichment was calculated by determining the fold change of each condition relative to the control Luc-RNAi condition, with the data in each condition normalized to GAPDH. Columns indicate the mean of the 3 independent replicate experiments, error bars indicate the standard deviation, and asterisks indicate P < 0.05. (B) Overexpression of EWS represses the neuronal genes of interest. Schematic representation of 3 EWS splice variants found in Ewing sarcoma (upper panel). Expression of indicated neuronal genes with the overexpression of each EWS variant was investigated by RT-PCR. Normalized fold enrichment was calculated by determining the fold change of each condition relative to the empty vector condition, with the data in each condition normalized to GAPDH. Columns indicate the mean of 3 independent replicate experiments, error bars indicate the standard deviation, and asterisks indicate P < .05. EWS protein expression was examined by Western blotting, shown in the lower panel.

Since knockdown of EWS increased the expression of the neuronal genes of interest, we next asked whether the overexpression of EWS would decrease the expression of these genes. There are 5 known alternative splice variants for EWS. We identified at least 3 variants during the process of cloning the EWS cDNA from A673 cells. Two variants, V2 and V3 (NM_005243.3 and NM_001163285.1), have previously been reported, and they differ only by 3 nucleotides in exon 11. The third variant, V6 (BankIt1574452 EWS JX977847), has not been identified previously. Variant V6 contains an alternative exon 9 (9b), which is 105 bases shorter than exon 9 (9a) that is found in V2 and V3 (Fig. 3B, upper panel). Following the enforced expression of each EWS isoform in A673 cells, we observed an overall repression of the neuronal genes tested to various extents (Fig. 3B, lower panel). It should be noted that the shRNA used to knock down EWS targets all 3 EWS variants and that EWS knockdown consistently increased the expression of the neuronal genes of interest (Fig. 3A). Taken together, these findings demonstrate that EWS indeed represses the set of neuronal genes tested in Ewing sarcoma cells.

EWS physically interacts with REST and binds to NRSE sites in the genome

Next, we examined the mechanism by which EWS and REST co-regulate these neuronal genes. One possibility is that EWS and REST regulate one another. However, knockdown of REST failed to change EWS levels and vice versa (Fig. 3A). EWS has previouisly been shown to regulate transcription by binding to proteins in the transcriptional machinery or to other transcription factors.^42,43 REST has also been shown to require co-factor interactions to mediate the repression of its target genes.^8,44 Therefore, we hypothesized that EWS and REST interact with each other to regulate the subset of neuronal genes. To test this hypothesis, we first asked whether EWS and REST physically interact in Ewing sarcoma cells. We immunoprecipitated endogenous EWS proteins from A673 cell lysates and tested for the presence of REST in the immunoprecipitates and vice versa. We found that EWS and REST co-immunoprecipitate in the reciprocal pull-downs (Fig. 4A). Domain mapping was then carried out by overexpressing flag-tagged full-length EWS, full-length EWS/FLI, which contains only the N-terminal portion of EWS, or Δ22 (a deletion mutant of EWS/FLI in which almost the entire EWS portion is deleted) (Fig. 4B) in HEK293 cells. After performing co-immunoprecipitation (co-IP), we found that full-length EWS and EWS/FLI, but not Δ22, were able to interact with REST (Fig. 4B). This indicates that the N-terminal domain of EWS is required for REST binding.

OPEN IN VIEWER

Figure 4.

EWS physically interacts with REST and binds to chromatin regions at or near NRSE sites. (A) Reciprocal co-IP of EWS and REST. (B) The N-terminus of EWS or EWS/FLI is responsible for binding to REST. Schematic representation of constructs used in the analysis is shown in the top panel. The number equals the numerical value of amino acid residues of the EWS protein. HEK293 cells were transfected with flag-tagged full-length EWS, full-length EWS/FLI, or EWS/FLI Δ22. Cell lysates were subject to immunoprecipitation with anti-REST antibody for endogenous REST. Normal mouse IgG was used as a negative control. Co-immunoprecipitates of flag-tagged proteins were detected by the anti-flag antibody. All co-IP experiments were performed 3 times. (C) EWS binds to the REST-repressed neural genes at or near their NRSE sites. ChIP-seq analyses were carried out as described previously. ⁴⁵ UCSC Genome Browser screenshots of the EWS ChIP-seq signal at indicated genes are shown. NRSE sites are indicated by arrows. (D) Mean EWS ChIP signal across all NRSE elements ±1 kb (left panel) and EWS-bound NRSE elements ±1 kb (right panel). Tags have been normalized for sequencing depth. (E) Overlap of EWS-binding sites across NRSE, SOX2, PAX6, and CTCF sites (dark gray bars). For comparison, overlap was also performed using randomly permuted EWS peaks (light gray bars). Asterisks indicate P < 0.001.

We next reasoned that if EWS and REST regulate these neuronal genes by mutual interaction, they would closely bind to chromatin at the set of neuronal genes. To examine whether EWS contacts chromatin at the NRSE sites present in the set of neuronal genes or at NRSE sites in the genome in general, we expressed epitope-tagged EWS in Ewing sarcoma EWS502 cells in which endogenous EWS/FLI had been silenced. We performed ChIP-seq analysis ⁴⁵ and observed EWS signals at regions 100 to 200 bp from the predicted NRSE sites at the neuronal genes of interest (Fig. 4C). We also observed EWS signals at other regions of some genes, which is consistent with previous findings that EWS regulates transcription and RNA splicing. We then examined EWS signals at computationally predicted NRSE sites (TTCAGCACCA/T/GC/ANGGACAGC /AG/AC/GC, N = 3-9). ⁴⁰ We found that the EWS signal was enriched and centered around NRSE sites (Fig. 4D, left panel) and that the signal intensity was even greater (Fig. 4D, right panel) when the analysis was limited to algorithmically selected EWS peaks (Zero-Inflated Negative Binomial Algorithm [ZINBA] ⁴⁶ ). As a validation of the signal-based enrichment of EWS at NRSE, we performed a permutation analysis of EWS binding at NRSE sites (Fig. 4E) and observed a significant association between EWS and NRSE. We also observed this relationship with the highly conserved insulator element CTCF but not the binding sites of 2 neuronal transcription factors SOX2 or PAX6 (Fig. 4E). These results suggest that EWS preferentially binds chromatin at or near NRSE sites in Ewing sarcoma cells. Taken together, our data indicate that EWS and REST interact with each other and bind chromatin at or near NRSE sites to repress the expression of the target neuronal genes.

EWS and REST inhibit the neuronal phenotype and tumorigenesis in Ewing sarcoma cells

Next, we sought to determine the biological function of the co-regulation of neuronal genes by EWS and REST. Because the depletion of EWS increases expression of the neuronal genes, we reasoned that EWS knockdown may induce a neuronal phenotype in Ewing sarcoma cells. We therefore examined the protein levels of 3 neuronal markers, β-III tubulin (TUBB3), neurofilament heavy polypeptide (NEFH), and microtubule- associated protein 2 (MAP2), following the knockdown of EWS. These 3 proteins have been previously used to assess neural differentiation in Ewing sarcoma cells.^17,47 We found that silencing of EWS in both A673 and TC71 cells resulted in an increased expression level of all 3 neuronal markers (Fig. 5A). REST knockdown also caused an increase in expression of the 3 neuronal markers, which is consistent with its role in inhibiting neural differentiation. Furthermore, we observed a significant morphological change of EWS or REST knockdown cells, as demonstrated by the increase in neurite outgrowth when the cells were exposed to low-serum medium and/or nerve growth factor (20 ng/mL) (Fig. 5B). In contrast, the control knockdown cells displayed only minor morphological changes when maintained under the same conditions of neuronal differentiation. These results suggest that both EWS and REST mediate repression of the neuronal phenotype in Ewing sarcoma.

OPEN IN VIEWER

Figure 5.

EWS and REST repress the neuronal phenotype and oncogenic transformation in Ewing sarcoma cells. (A) EWS or REST knockdown increases the expression of neuronal markers TUBB3, NEFH, and MAP2 in A673 and TC71 cells. (B) EWS or REST but not control knockdown cells display a neuronal phenotype when grown in media with 1% FBS and/or 20 ng/mL of nerve growth factor. (C) Growth curves of A673 cells with the overexpression of EWS V2, V3, or V6 or REST. (D) Soft agar assay with A673 cells overexpressing each of the 3 EWS splice isoforms or REST. Asterisks indicate P < 0.05.

Given that EWS is translocated in multiple cancers, and that REST has been implicated as a tumor suppressor in lung, breast, and colon cancers, we next tested whether EWS and REST inhibit tumorigenesis in Ewing sarcoma. We silenced EWS or REST by shRNA in A673 and TC71 cells and found that the cells grew similar in tissue culture and formed a similar number of colonies in soft agar as the control knockdown cells (Suppl. Fig. S2A-D). Because Ewing sarcoma cells are highly transformed at baseline, we reasoned that a decrease of the EWS or REST expression level may not result in a significant increase in colony formation. Therefore, we next enforced the expression of each of the 3 EWS isoforms or REST in A673 and TC71 cells. We also observed a similar growth rate of these cells in tissue culture compared to control cells (Fig. 5C and Suppl. Fig. S2E). However, expression of EWS V2 or V3 isoforms or REST resulted in a significant reduction in oncogenic transformation, as shown by diminished anchorage-independent growth in soft agar (Fig. 5D and Suppl. Fig. S2F). Interestingly, EWS V6 overexpression failed to inhibit colony formation in soft agar, suggesting that exon 9 in the V2 and V3 isoforms is necessary for EWS-mediated inhibition of anchorage-independent growth. These results indicate that EWS and REST inhibit the maintanence of oncogenic transformation in Ewing sarcoma.

Cell lines

Ewing sarcoma cell lines A673, TC71, and EWS502 were purchased from American Type Culture Collection (ATCC, Manassas, VA) and grown as previously described. ⁵⁶ Growth curve and soft agar colony formation assays were performed as previously described. ²

Constructs

To clone the 3xFlag EWS construct, full-length EWS was amplified by PCR using the cDNA library from A673 cells. EWS was then fused in-frame with the 3xFlag tag and cloned into the pQCXIN vector. Constructs for the 3 EWS isoforms were generated by the amplification of individual EWS cDNA for each isoform, followed by ligation into the pMSCV-neo vector. The pMSCV-hygro 3xFlag EWS/FLI and mutant Δ22 have been previously described. ⁵⁷

RNAi

A human EWS-specific 19-mer oligonucleotide, 5′-gactctgacaacagtgcaa-3′, that maps to nucleotides 1083 to 1102 of the EWS gene was inserted into the pMKO.1-puro vector for stable knockdown of EWS. The nucleotide for REST knockdown is 5′-gaaactttgaacaaggttt-3′. Vectors for control (Luc-RNAi) knockdown and production of retrovirus-expressing shRNA have previously been described. ¹ siRNA for transient knockdown of EWS or REST was purchased from Dharmacon (Thermo Scientific, Rockford, IL).

RT-PCR

RT-PCR was performed using the iScript SYBR green RT-PCR kit (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. Primer sequences are available in Supplementary File S5.

RNA-seq and data analysis

Total RNA was extracted from Ewing sarcoma cells and treated with DNase using the Qiagen RNeasy kit (Germantown, MD). mRNA was enriched by oligo-dT magnetic beads and was used to construct Illumina (San Diego, CA) sequencing libraries. The libraries were single-end sequenced on Illumina Genome Analyzer IIx for 36 cycles (EWS) or HiSeq 2000 for 50 cycles (EWS/FLI). Reads were mapped to the hg19 genome build with Casava (Illumina) for EWS or Novoalign (Novocraft, Petaling Jaya, Malaysia) for EWS/FLI. The RNA-seq analysis was carried out using USeq (useq.sourceforge.net) versions 8.1.5 for EWS and 8.3.9 for EWS/FLI. Sorted, mapped files were converted to PointData representation with the USeq Eland Parser application for EWS and SAM Parser for EWS/FLI (useq.sourceforge.net). The knockdowns were compared with control using either Defined Region Scan Seqs (EWS) or Overdispersed Region Scan Seqs (EWS/FLI) applications with default parameters and a gene reference file derived from a merger of Ensembl and RefSeq (using a USeq Merge UCSC Gene table). Differential expression was defined as EWS: FDR ≤ 0.05 and |log₂ fold change| ≥ 1.3 and EWS/FLI: FDR ≤ 1e-20 and |log₂ fold change| ≥ 2.

GSEA and χ² testing

EWS gene sets, both the up- regulated and down-regulated, were divided according to their corresponding log fold change in the EWS/FLI RNA-seq experiment. Gene set enrichment was determined using GSEA. ⁵⁸ To quantitatively establish enrichment of the EWS-regulated genes in the EWS/FLI–up-regulated and –down-regulated genes, we applied a χ² test as described previously ²⁹ and calculated the enrichment score separately for EWS/FLI–up-regulated and –down-regulated genes.

ChIP-seq and data analysis

ChIP-seq analysis was performed as described previously. ⁴⁵ NRSE, SOX2, PAX6, and CTCF motif locations were derived from MotifMap (motifmap.ics.uci.edu). ⁵⁹ EWS peaks were permuted 1,000 times over the mappable genome (hg19) and assayed for overlap with the NRSE, SOX2, PAX6, and CTCF sites.

Immunoprecipitation and immunoblotting

Immunoprecipitation and immunoblotting were carried out as described previously. ² Anti-TUBB3 (MAB1195) and anti-NEFH (AF3108) were from R&D Systems (Minneapolis, MN), anti-MAP2 (Ab32454) was from Abcam (Cambridge, MA), and anti-REST (sc-374611), anti-EWS (sc-48404), and normal mouse IgG (sc-2025) were from Santa Cruz Biotechnology (Santa Cruz, CA).