Hypoxia modulates EWS-FLI1 transcriptional signature and enhances malignant properties of Ewing's sarcoma cells in-vitro

Abstract

Hypoxia is an important condition in the tumor cell microenvironment and approximately 1-1.5% of the genome is transcriptionally responsive to hypoxia with hypoxia-inducible factor-1 (HIF-1) as a major mediator of transcriptional activation. Tumor hypoxia is associated with a more aggressive phenotype of many cancers in adults, but data on pediatric tumors are scarce. Since, by immunohistochemistry, HIF-1α expression was readily detectable in 18/28 primary Ewing's sarcoma family tumors (ESFT), a group of highly malignant bone-associated tumors in children and young adults, we studied the effect of hypoxia on ESFT cell lines in-vitro. Intriguingly, we found that EWS-FLI1 protein expression, which characterizes ESFT, is up-regulated by hypoxia in a HIF-1α-dependent manner. Hypoxia modulated the EWS-FLI1 transcriptional signature relative to normoxic conditions. Both synergistic as well as antagonistic transcriptional effects of EWS-FLI1 and of hypoxia were observed. Consistent with alterations in the expression of metastasis related genes, hypoxia stimulated the invasiveness and soft-agar colony formation of ESFT cells in-vitro. Our data represents the first transcriptome analysis of hypoxic ESFT cells and identifies hypoxia as an important microenvironmental factor modulating EWS-FLI1 expression and target gene activity with far-reaching consequences for the malignant properties of ESFT.

Introduction

Many tumors are profoundly hypoxic and frequently have a poorer prognosis than non-hypoxic tumors (1). The cellular adaptations of cancer cells to hypoxia affect transcriptional regulation (2), increase glucose uptake capacity and glycolysis (3), stimulate cell motility and invasion (4) and regulate apoptosis (5). Many of the phenotypes associated with hypoxic cancer cells are the consequence of altered gene expression (1). Genomic studies revealed a core set of consistently hypoxia-induced genes despite pronounced heterogeneity depending on the relative level and duration of hypoxia and the cell type used (6). This is likely due to the transcriptional activity of hypoxia-inducible factors (HIFs) that are stabilized and dramatically increased on the protein level as a central adaptive response to tumor hypoxia (7, 8). While HIF-1α is the best characterized inducer of gene transcription in hypoxic cells, additional signaling pathways can modulate gene expression in response to low oxygen (9, 10). HIF-1α is overexpressed in various types of cancer (11) and compelling evidence supports its role in tumorigenesis (12). Overexpression of HIF-1α-dependent genes has been associated with aggressive behaviour in human cancers in-vitro (13, 14) and in clinical specimens (15-17). Therefore, targeting HIF-1α is an exciting prospect for cancer therapy. Although the impact of hypoxia has been extensively studied in many adult malignancies, its role in pediatric tumors has remained largely unaddressed.

ESFT are characterized by the presence of EWS-ETS gene rearrangements. In 85% of cases, the gene fusion is between EWS and FLI. The EWS-ETS fusion proteins are necessary for the induction and maintenance of the ESFT malignant phenotype (for a review, see (18)). The ability of EWS-FLI1 to regulate gene transcription is central to its function in tumor formation and progression (19-21). Several studies have shown that the target gene repertoire of EWS-FLI1 varies according to the host cell type (22, 23). Genomic profiling experiments in ESFT cell lines upon modulation of EWS-FLI1 expression and comparisons to primary tumors have identified specific EWS-FLI1-regulated genes including, NKX2-2, NROB1, TRPM4, PPP1R1A, KMO, STYXL1 (EWS-FLI1 upregulated) and IGFBP3, LOX, GADD45B, DAB2 (EWS-FLI1 repressed) that discriminate ESFT from other sarcomas (24-29). These studies were done under normoxic in-vitro adherent cell-culture conditions that did not consider the influence of multicellular three-dimensional structure (30) and low oxygen that characterize in-vivo growth conditions of most solid tumors. Thus, the dramatic cellular changes associated with an hypoxic microenvironment have so far been largely ignored (31). We therefore aimed at unravelling the impact of low oxygen on ESFT biological properties under anchorage-independent and adherent growth conditions. Since EWS-FLI1 is the major driver of ESFT specific gene expression, we analysed the consequences of hypoxia for EWS-FLI1 and EWS-FLI1-regulated gene expression in ESFT cell lines grown as spheroids and as adherent monolayers. Our results identify hypoxia as a microenvironmental factor modulating EWS-FLI1 protein abundancy and transcriptional signature promoting anchorage-independent clonogenicity and invasiveness of ESFT cells.

Materials & Methods

Cell culture

ESFT cell lines SK-N-MC, TC71, WE68 and TC252 used in this study have been previously described (32). They were maintained in RPMI 1640 (Invitrogen, Germany) supplemented with 10% fetal calf-serum (PAA, Pasching, Austria) as monolayer-cells. Growing multicellular spheroids was done essentially as described elsewhere (30). Hypoxic and normoxic treatments of 10⁵ cells/ml were done for 16h (unless otherwise stated). For normoxic (21% O₂) growth, cells were placed in a Thermo-CO₂ incubator with a humidified-atmosphere containing 5% CO₂ (Thermo Electron Corp., Asheville, NC). Hypoxic treatment was achieved in a ProOx model 110 chamber (Biospherix, Redfield, NY) at 37°C flushed with a gas mixture of 5% CO₂/95% N₂. Oxygen concentration within the chamber was maintained at 1% using a ProOx 110 oxygen regulator (Biospherix, USA). The hypoxia-mimetic agents cobalt-chloride (CoCl₂) and Desferrioxamine (DFX) (Sigma, St. Louis, MO) were added to ESFT cells under normoxia at concentrations and for time periods indicated in the figures.

Immunohistochemistry

Immunohistochemical staining was done on 3 μm paraffin sections of molecularly confirmed ESFT on a tissue-microarray. Pretreatment for antigen retrieval was performed by pressure cooker. Antibody for HIF-1α was diluted 1:250. After blockage of biotin (by avidin-biotin) and peroxidase (by H₂O₂), staining was done on an automated immunostainer (Biogenex i6000, San Ramon, CA) using a standard labeled streptavidin-biotin method (UltraTek Reagent Dtection Kit, ScyTek, Logan, UT) followed by 3,3’-diaminobenzidine enzymatic development. Sections were counterstained with hematoxylin. Omission of the primary antibody as well as anti-isotype antibodies served as negative controls.

Plasmids and transfections

Plasmids encoding HIF-1α (pEF-Bos–HIF-1α) and mutant HIF-1α that is resistant to proteosomal degradation (pEF-Bos-ΔHIF-1α) were gifts from Dr. Murray Whitelaw (The University of Adelaide, Adelaide). Plasmid-based short-hairpin RNA to HIF-1α (pshHIF-1α) and a control plasmid encoding for a non-targeting shRNA was obtained from Dr. M. Vooijs (University Medical Centre, Utretch) (33). For plasmid transfections, SK-N-MC and TC252 ESFT cells were plated in 75-mm² culture flasks and grown to 60-70% confluence. Cells were transfected with 2 μg plasmid DNA using the LipofectAMINE Plus reagent (Invitrogen, Groningen, The Netherlands) according to the manufacturer's recommendations.

RNA preparation and Reverse Transcription-PCR (RT-PCR)

Total RNA was prepared with a Qiagen RNAeasy kit (QIAGEN) following the manufacturer's instructions. Reverse transcription, PCR amplification (using primers spanning the EWS-FLI1 fusion region), and 1% agarose gel electrophoresis were done according to standard protocols. Specific primer pairs used for amplification are listed in Table S9.

Immunoblot Analysis

Protein extracts were prepared using a modified RIPA lysis-buffer as described (34). 30-50 μg of protein samples were fractionated by 8% SDS-PAGE and immunoblotted. The FLI1 monoclonal-antibody 7.3 (recognizing EWS-FLI1 in ESFT) was kindly provided by O. Delattre (Institute Curie, Paris). Antibodies to the EWS N-terminus, HIF-1α and ß-actin were from Santa Cruz Biotechnology (Santa Cruz, CA), BD Transduction Laboratories (Lexinton, KY) and Abcam (Cambridge, UK), respectively.

Microarray analysis

Changes in gene expression profiles upon exposure of ESFT cells to hypoxia versus normoxia were followed on Human Genome U133-A2 and Human Genome U133 Plus 2.0 Arrays (Affymetrix, Inc., Santa Clara, CA). cRNA synthesis and GeneChip processing were performed according to standard protocols (Affymetrix, Inc., Santa Clara, CA). Bioinformatic analyses based on Affymetrix CEL files were performed in R statistical environment using Bioconductor packages (35). Affymetrix CEL files were normalized using the gcrma logarithm (36). For further analysis, probesets (ps) with very low expression across all arrays were excluded (Bioconductor package: panp P-value > 0.05) and, for each gene, one ps was selected by the criterion of maximizing the expression variation across samples.

To define functional categories of deregulated genes under hypoxia, we performed gene set enrichment analysis (GSEA) using the “pGSEA” package in the Bioconductor/R environment (37-39). Gene-wise log2 expression ratios (logFC) of hypoxia versus normoxia -and for the mean of their logFCs, were used as input for pGSEA. We tested three different gene sets in the MSigDB (http://www.broad.mit.edu/gsea/msigdb/ Cambridge, USA): curated gene sets from canonical pathways and experimental data (C2), DNA motif gene sets (C3), GO terms (C5). We added two more gene sets to the C2 group to study the effect of hypoxia on invasion-related genes and on EWS-FLI1 modulated genes (27). The invasion gene set was manually curated from the literature (40) . The output from GSEA and significance level of all tested groups are in supplementary tables (adherent cells: Table S2-4, cells grown as spheroids: Table S5-S7).

Matrigel Invasion Assays

Cell invasion was examined using Transwell chambers (Corning Incorporated, Life Sciences, NY, USA) with matrigel-coated 8μm pore membranes. TC252 and SK-N-MC cells (5×10⁵) were harvested, resuspended in serum-free media containing 0.1% BSA, added in triplicates to Transwell chambers, and allowed to invade through Matrigel toward complete media for 48h. Cells that invaded to the lower surface of the membrane were fixed using 4% paraformaldehyde and stained with 0,2% crystal-violet. Cells were photographed with a Zeiss Axiovert 40C microscope (Goettingen, Germany) and counted using the Image J software (NIH, USA), and the mean number of invasive cells was determined. The data were means (± SEM) from three replicate experiments each performed in triplicates.

Soft-agar assay

Cells were seeded in triplicates at 2×10⁴ cells/35mm-dish. After resuspension in 0.3% agar in 10% FCS and RPMI, cells were plated in 0.6% agar-coated dishes. A top layer containing 0.6% agar was then added. Plates were then incubated for 18h under normoxic conditions before hypoxia treatment was started. Cells were fed every 5 days by placing 3 drops of medium on the top layer. Colonies were microscopically counted after 14 days. Colony formation was examined at seven sites per well for a total of 21 fields in a minimum of six separate soft-agar experiments.

Stastistical analysis

When applicable, the data were analyzed by the unpaired t test with Welch's correction using the Prism 5 for Windows (Version 5.02) statistical software (GraphPad Software, Inc., La Jolla, CA ). Data shown in graphical format represents the means (± SEM) and a P value of <0.05 is considered statistically significant.

Results

EWS-FLI1 and HIF-1α protein expression is up-regulated under hypoxia in ESFT cells

To test for the relevance of hypoxia to ESFT, we screened for HIF-1α protein expression in paraffin-embedded primary tumor sections by immunohistochemistry on 28 EWS-FLI1 fusion-positive ESFT tumors. Strongly positive HIF-1α expression was observed in 39% (11/28) of the samples, positive expression was observed in 25% (7/28), while 26% (8/28) showed no detectable levels of HIF-1α protein expression (Figure 1A, upper panel). These results strongly indicate that HIF-1α protein is frequently stabilized in primary ESFT presumably as a consequence of tumor hypoxia.

Figure 1.

Hypoxia-inducible factor-1α (HIF-1α) is induced and EWS-FLI1 expression is augmented under hypoxia. (A, upper panel) HIF-1α is variably expressed in primary ESFT tissues. HIF-1α expression was analyzed by immunohistochemistry of sections from primary ESFT. Sections were initially subjected to standard H&E staining. The staining for HIF-1α was determined semi-quantitatively, comprising intensity of staining (negative (0), moderate (1), strong (2)) and the percentage of positive cells (0-10% (1); 10-50% (2); >50% (3)). Scores from intensity of staining and positive cells were added together receiving a final score of 0-5. A score of 0-1 was designated to be negative (-), a score of 2-3 positive (+), and a score of 4-5 strongly positive (++). Magnification, x400. Representative materials from three patients. (A, lower-panel) Both HIF-1α and EWS-FLI1 protein levels are enhanced by hypoxia in ESFT spheroids. (B) EWS protein level remained unchanged under hypoxia (Left-panel). The EWS antibody recognizes both EWS and EWS-FLI1 in ESFT. (Middle-panel) RT-PCR analysis of known hypoxia-inducible genes and EWS-FLI1 in TC252 cells. TC252 cells were cultured under hypoxia at the indicated time points and total RNA was reversed transcribed and tested for mRNA expression by semi-quantitative RT-PCR as described in Materials and Methods. ß-actin mRNA was assayed in parallel as an internal control for input RNA. A representative experiment of three performed is shown. (Right-panel) Graph showing Image-J quantification of RT-PCR bands. (C) Time-dependent HIF-1α and EWS-FLI1 protein expression changes under hypoxia. Both TC252 and SK-N-MC cells were cultured to 80% confluence and then subjected to hypoxia (1% O₂) for the indicated time-periods. (D) HIF-1α expression is induced by exposure to the hypoxia mimetic agents CoCl₂ and DFX in a dose-(upper-panel) and time-dependent (lower-panel) manner in TC252 cells. EWS-FLI1 expression is likewise enhanced in a time-dependent fashion by CoCl₂ and DFX. Cell lysates were subjected to immunoblotting analysis using anti-HIF-1α antibody and anti-FLI 7.3 antibody (which recognizes the EWS-FLI1 fusion protein in ESFT). The blot was reprobed with β-Actin antibody to control for protein loading and transfer efficiency.

To establish a model of hypoxia in ESFT cells, SK-N-MC, TC71, WE68 and TC252 cell lines were subjected to growth as multicellular spheroids under 1% oxygen and, for comparison, under normoxic conditions, and HIF-1α protein levels were analysed by immunoblot to monitor the induction of a hypoxic program. The ability of ESFT cells to form tight multicellular spheroids in liquid suspension culture was strongly increased at 20 hours of incubation under hypoxic versus normoxic conditions (data not shown). As demonstrated in Figure 1A (lower panel), HIF-1α expression was readily detectable after 16 hours of hypoxia treatment in all 4 cell lines tested. Surprisingly, EWS-FLI1 protein levels were also found consistently upregulated. Expression of the EWS-FLI1 fusion is controlled by the EWS upstream regulatory region (for a review, see (41)). However, hypoxia-induced EWS-FLI1 accumulation was restricted to the fusion protein and did not affect expression of the non-rearranged EWS allele as demonstrated representatively for the cell lines TC252 and SK-N-MC in Figure 1B (left panel). Focussing on these two cell lines we analysed the kinetics of HIF-1α and EWS-FLI1 expression (Figure 1C). HIF-1α was first detectable at 4h of hypoxia treatment, peaked between 8h and 16h and remained fairly constant thereafter. In contrast, EWS-FLI1 protein accumulation was transient, starting at 4h, peaking at 8h, and then rapidly dropping-off again. Similar results were obtained under adherent growth conditions using the hypoxia-mimetic agents CoCl₂ and DFX in TC252 cells. As shown in Figure 1D (upper panel), HIF-1α induction was dose-dependent. Using 200μM CoCl₂ or 100μM DFX, HIF-1α expression was already detectable at 2h of treatment preceding EWS-FLI1 accumulation which became visible at 4h (CoCl₂) and 6-8h (DFX), respectively (Figure 1D, lower panel). This result may suggest that the observed EWS-FLI1 accumulation is a consequence of HIF-1α induction. Semi-quantitative RT-PCR analysis of RNA from TC252 cells incubated at 1% oxygen for increasing time intervals revealed however that, like HIF-1α, EWS-FLI1 RNA expression remained unchanged, while transcript levels of known HIF-1α activated genes VEGF, Aldolase-C, Glut 1, CA9, and IGFBP3 increased under hypoxia (Figure 1B, right panel). This shows that hypoxia regulation of EWS-FLI1 occurs at a post-transcriptional level.

Involvement of HIF-1α in hypoxia-enhanced EWS-FLI1 protein expression

To assess whether the transient EWS-FLI1 protein accumulation under hypoxia is caused by HIF-1α, TC252 and SK-N-MC cell lines were left untreated or transfected with either no plasmid, or shRNA to HIF-1α, or a non-targeting shRNA construct, and were incubated at either 1% oxygen for 8h (Figure 2A) or in the presence of 200μM CoCl₂ for 8h (Figure 2B). As compared to control transfections, modulation of HIF-1α protein expression in shHIF-1α treated cells was consistently accompanied by loss of EWS-FLI1 accumulation. Vice versa, ectopic expression of HIF-1α or a degradation-resistant HIF-1α mutant enhanced EWS-FLI1 expression under normoxia, further arguing for the involvement of HIF-1α in hypoxia-augmented EWS-FLI1 expression (Figure 2C).

Figure 2.

Specific downregulation of HIF-1α by RNA-interference dramatically suppressed EWS-FLI1 protein accumulation under hypoxia. Knockdown of HIF-1α by the specific shRNA resulted in suppression of EWS-FLI1 expression in both TC252 and SK-N-MC cell lines both at 1% oxygen (A) and by CoCl₂ (B). (C) Expression of wildtype HIF-1α and a degradation-resistant HIF-1α cDNA in TC252 cells under normoxia resulted in a dose-dependent enhancement of EWS-FLI1 protein expression. Results are from a representative experiment of 3 performed.

Transcriptional response of ESFT cell lines to hypoxia

Since incubation in low-oxygen induced HIF-1α and, at least transiently, enhanced EWS-FLI1 expression, we analysed the consequences of hypoxia treatment on the overall ESFT transcriptome. TC252 and SK-N-MC cell lines propagated as adherent monolayers and anchorage-independent spheroids were subjected to different oxygen levels (1% versus 21% oxygen) for 16h and transcriptional profiles were investigated via microarray analysis. Supplementary Table S1 shows the top hypoxia-modulated genes (logFC > 1 or < -1 in both cell lines) in adherent (351 genes, Table S1a) and spheroid (280 genes, Table S1b) culture conditions. Pairwise comparison of the four lists of logFC values for all genes (5148 shared genes across experiments) showed moderate correlations (Pearson correlation coefficients: 0.31-0.48). However, when genes were filtered out that showed no strong regulation in any of the four experiments (-0.5 ≤ logFC ≤ 0.5) resulting in a list of 362 genes, the correlations rose to 0.74-0.86. This result indicates that genes which show regulation above noise in all experiments are overall consistently regulated independent of culture conditions and cell line. Functional analysis of deregulated genes in the four experiments was performed by GSEA whereby functionally defined gene sets were identified as significantly up- or down-regulated in the data. The output from GSEA and the significance level of all tested gene sets can be inspected in supplementary tables (adherent: Tables S2-4, spheroids: Tables S5-S7). Figure 3 (A, B) shows the top 20 deregulated gene sets for both culture conditions. The effect of hypoxia could be clearly identified, as functions/datasets related to hypoxia were among the most significant categories in all experiments. These included experimental datasets from the MSigDB (C2) where the effect of hypoxia was studied, but also hypoxia-related functions required by tumor cells to survive under hypoxic conditions including angiogenesis, vasculature development and glucose metabolism. Furthermore, the GSEA analysis strongly suggested a direct effect of HIF-1α as the gene set “HIF targets” (MSigDB-C2) is highly significant in all experiments and also gene sets with HIF-1 motifs in their promoter region (MSigDB-C3) are among the most significant in all experiments except for TC252 in adherent culture conditions (Tables S4, S7). Interestingly, the two gene sets describing an EWS-FLI1 transcriptional signature in ESFT (27) were among the most significantly modulated gene sets. Importantly however, the effect of hypoxia on these genes was overall counteracting EWS-FLI1, i.e. genes that are up-regulated by EWS-FLI1 show a strong tendency to be down-modulated by hypoxia and vice versa. This result is further confirmed by the GSEA analysis of the motif gene set (MSigDB–C3): genes with a GGAA/T core ETS binding-motif such as for ELK1, ERG, GABP, CETSP154, NRF2, are generally expressed at lower levels under hypoxia than under normoxia. As EWS-FLI1 binds to this motif as well, this result suggests that potential direct EWS-FLI1 targets might be down-modulated under hypoxia. To study the functional interaction between EWS-FLI1 and hypoxia in more detail, we intersected the data from the present study with data from gene expression targets of EWS-FLI1 in ESFT (27). In this data set, 344 and 237 genes were found to be consistently up- and down-regulated, respectively, in five ESFT cell lines and 59 primary tumors. This gene list was further filtered for genes showing a more than 2-fold gene expression change under hypoxia compared to normoxia. For the resulting data set, gene expression ratios for shRNA-mediated EWS-FLI1 knockdown (27) were plotted against gene expression ratios for hypoxic versus normoxic conditions. Figure 4 shows that for both cell lines and culture conditions there were more genes found in the upper left and lower right corners in which EWS-FLI1 expression and hypoxia have opposite effects. However, the level of antidromic change induced by hypoxia was lower than by EWS-FLI1 expression in most instances. Thus, the overall qualitative EWS-FLI1 signature was retained albeit quantitatively modulated by hypoxia. For a small number of genes, however, the two conditions have a synergistic effect (lower left and upper right corner). To gain more insight into antagonistic and synergistic functional interactions between EWS-FLI1 and hypoxia, EWS-FLI1 silencing experiments were performed under normoxic and hypoxic conditions. Focussing on genes whose expression changed more than two-fold upon EWS-FLI1 knockdown under normoxic conditions, Supplementary Figure S1 illustrates the relative effect of EWS-FLI1 knockdown under hypoxia. A moderate increase in differential gene expression for most EWS-FLI1 activated (orange)/repressed (blue) genes demonstrates that the synergistic effects of hypoxia are completely lost consistent with the rate limiting role of the fusion oncogene in the regulation of these genes. In contrast, the loss of differential gene expression with EWS-FLI1 knockdown under hypoxia of genes for which hypoxia had demonstrated an antidromic effect with EWS-FLI1 under normoxia (green and red), indicates that the hypoxia effect dominates over the EWS-FLI1 effect.

Figure 3.

Result from the GSEA analysis with the MSigDB gene sets C2 (left-panels) and C3 (right-panels). The rows in each heatmap show the 20 most significantly hypoxia-induced (red) and repressed (green) gene sets as determined by the mean of the z-scores for both tested cell lines (sorted by column “Both” in tables S3,4,6,7). The columns “SK” and “TC” indicate the individual z-scores in the SK-N-MC and TC252 cell lines when analyzed separately. The red-green color-coded z-score indicates the significance level. The most significantly repressed gene sets are found at the top, induced gene sets at the bottom of each heatmap. A: adherent culture condition, B: spheroid culture condition.

Figure 4.

Scatter-plots of the gene expression effect of hypoxia versus the effect of EWS-FLI1. Upper-panel: adherent, lower-panel: spheroid culture. Dots represent the log2 fold-change (logFC) values of genes for hypoxia versus normoxia (x-axis) and EWS-FLI1 control vs. knockdown (y-axis). Positive values on both axes indicate induction by hypoxia or EWS-FLI1 respectively, therefore genes with synergistic hypoxia/EWS-FLI1 effects are found in the lower left and upper right quadrants of the plots. For this plot, all genes in GS1 in Table S1 from Kauer et al (2009) were intersected with genes showing a logFC (hypoxia versus normoxia) of larger/smaller than +1/-1 in this study.

Serum-induced invasion of ESFT cells is enhanced by hypoxia

Further down on the list, but still highly significantly (P = 0.017) enriched in both cell lines (Supplementary Tables S3 and S6), GSEA identified “invasion”. We therefore asked, if hypoxia can enhance the invasive capacity of ESFT cells in-vitro. SK-N-MC and TC252 were cultured under normoxic and hypoxic conditions on matrigel-coated Transwell inserts, and cells that have invaded into the matrigel and migrated to the other side of the membrane were monitored after 48 hours. As shown in Figure 5A, both SK-N-MC and TC252 cultured under hypoxia exhibited significantly invasive capability to cross the matrigel barrier compared to the normoxia control enhanced (P = 0.0001 and 0.0063 for SK-N-MC and TC252, respectively).

Figure 5.

Hypoxia stimulates ESFT cell invasiveness and soft-agar colony formation. (A) ESFT cells cultured under hypoxia are highly invasive compared to their normoxia counterparts. A reconstituted basement membrane (Matrigel) assay was used to determine the invasion potential of two ESFT cell lines cultured under hypoxia versus normoxia. To estimate invasion, we counted invading cells (at magnification x50) after staining and making photographs. Representative images of Matrigel-invaded cells are shown on the right side for each condition. The box-plots report the means (± SEM) of the number of migrated cells in 5 selected fields. Experiments were performed in triplicates and repeated thrice with similar results. Statistical significance was determined using the unpaired t test. (B) Anchorage-independent growth was studied by soft-agar colony-formation assays. TC252 and SK-N-MC cells were incubated under hypoxic/normoxic conditions and vital colonies were photographed (scale bar: 50μm). Two representative fields at x50 magnification for each cell line were captured for each condition 14 days after seeding (inserts depict colony sizes at x100 magnification). The experiment was repeated thrice and statistical analysis was done using the unpaired t test.

Hypoxia promotes anchorage-independent colony formation in soft agar

The ability of cells to grow under anchorage-independent conditions is one of the hallmark characteristics of oncogenic transformation. To assess the influence of hypoxia on the colony-forming capacity of ESFT cells, SK-N-MC and TC252 cells were seeded into soft agar and grown for 2 weeks under hypoxic versus normoxic conditions (Figure 5B). While the number of colonies formed by both TC252 and SK-N-MC cells under hypoxia was only slightly and non-significantly increased (P = 0.8864 and P = 0.2913, respectively) colony sizes were markedly bigger than under normoxia. The median diameter of colonies obtained after 2 weeks of incubation for TC252 and SK-N-MC under normoxia was <5 and ≤20μm, respectively, but ≤30μm and >50μm, respectively, under hypoxia. These results suggest that while hypoxia only marginally increased clonogenicity of ESFT cells it strongly promoted growth under anchorage-independent growth conditions.

Discussion

Adapting to the hypoxic microenvironment is crucial for tumor cell growth and survival and is achieved largely by transcriptional activation of genes that facilitate short- and long-term adaptive responses (42, 43). The HIF-1α pathway is invariably activated under hypoxic conditions and orchestrates a complex transcriptional program geared towards enhancing cell survival in a transiently unfavourable microenvironment (44). We now report for the first time that a large fraction of primary ESFT accumulates HIF-1α protein presumably reflecting hypoxia induction in those tissues. Our in-vitro data demonstrate that ESFT cells adapt to hypoxia by redefining their transcriptome and acquiring a distinct hypoxic phenotype characterized by increased invasiveness and anchorage-independent growth.

There is ample experimental evidence that EWS-FLI1 is the major driver of altered gene expression in the pathogenesis of ESFT. In the light of hypoxia increasing malignant properties of solid tumors, our finding of a transient HIF-1α-dependent increase in expression levels of this central ESFT oncogene upon hypoxia treatment is intriguing. Although the exact mechanism of EWS-FLI1 protein accumulation in the absence of increased EWS-FLI1 transcription is not known, our results identify functional cross-talk between HIF-1α and the fusion oncogene. Protein accumulation in response to low-oxygen was restricted to EWS-FLI1 and neither observed for full-length EWS nor for an alternative EWS fusion to ERG (not shown). These results suggest that HIF-1α mediated EWS-FLI1 accumulation involves protein regulation at the C-terminal FLI1 moiety. However, if regulation of EWS-FLI1 stability or translation is mediated via post-translational modification of the protein itself or via regulation of a micro-RNA targeting the FLI1 3′UTR remains to be established. By switching EWS-FLI1 expression on and off in cell lines grown under normoxic conditions, previous studies identified a specific gene expression signature of the oncogene (25-27, 45). The data presented in this study indicate that this EWS-FLI1 signature is modulated under hypoxic conditions. Our results suggest that EWS-FLI1 and HIF-1α collaborate in both synergistic and antagonistic ways to regulate cellular metabolism under hypoxia in ESFT cells. Interestingly, the majority of EWS-FLI1 regulated genes showed opposite transcriptional effects in the two cell line models tested. Among them, a cell survival factor, IGFBP-3, which has previously been shown to be suppressed by EWS-FLI1 (46), was highly upregulated in ESFT cells under hypoxia. Concomitantly, we observed a downregulation of the pro-apoptotic gene BAX (data not shown). Together, these changes may contribute to the recently reported HIF-1α dependent protection of ESFT cells from anti-cancer drug- or tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-induced apoptosis (47). For a small proportion of EWS-FLI1 regulated genes (e.g. CENPF, IL1RAP, and c7orf44), hypoxia augmented the EWS-FLI1 effect. This synergistic effect may either be the consequence of the observed increase in EWS-FLI1 levels or of transcription factor cooperation between EWS-FLI1 and HIF-1α or any other hypoxia regulated transcription factor on the promoters of these genes. Cross-talk between HIF and transcription factors has already been demonstrated for p53, NF-ĸB and c-Myc, among others (48). Our data add EWS-FLI1 to the growing list of HIF-1α associates.

Functional annotation of the transcriptomic changes induced by hypoxia in ESFT cells by GSEA revealed a clear HIF-1α signature. In addition to classical HIF-1α-regulated energy-producing metabolic pathways, invasion-associated genes were among those upregulated under hypoxia. Consistent with this finding, in-vitro assays revealed an enhanced invasive capability of ESFT cells. Moreover, hypoxic conditions increased the clonogenic proliferation ability of ESFT cells in soft-agar. These findings suggest that hypoxia may be a major factor in the tumor microenvironment of ESFT contributing to the aggressive metastatic behaviour of the disease. While ESFT cell proliferation was markedly increased under anchorage-independent growth conditions, no influence on tumor cell proliferation was observed in adherent cultures (not shown). Consistent with an influence of in-vitro growth conditions on the tumor cell response to hypoxia, differences were observed in the transcriptomic response to hypoxia between ESFT cell lines kept as adherent and anchorage-independent cultures which are also reflected in variations of the interplay with EWS-FLI1. This is very intriguing considering the fact that anchorage-independent hypoxic conditions more closely mirror the in-vivo situation than that of standard adherent normoxic cultures. It should be noted that, although the overall ESFT specific EWS-FLI1 signature of modulated genes was maintained under hypoxic non-adherent conditions, a large number of EWS-FLI1 responsive genes was quantitatively changed in expression. These results, which provide novel insights into the composition of the hypoxia-regulated transcriptome in ESFT, should be considered when selecting EWS-FLI1 regulated genes for future therapeutic targeting strategies.