Integrative gene network and functional analyses identify a prognostically relevant key regulator of metastasis in Ewing sarcoma

Florencia Cidre‐Aranaz, Jing Li, Tilman L. B. Hölting, Martin F. Orth, Roland Imle, Stefanie Kutschmann, Giulia Ammirati, Katharina Ceranski, Martha Julia Carreño‐Gonzalez, Merve Kasan, Aruna Marchetto, Cornelius M. Funk, Felix Bestvater, Simone Bersini, Chiara Arrigoni, Matteo Moretti, Uwe Thiel, Daniel Baumhoer, Felix Sahm, Stefan M. Pfister, Wolfgang Hartmann, Uta Dirksen, Laura Romero‐Pérez, Ana Banito, Shunya Ohmura, Julian Musa, Thomas Kirchner, Maximilian M. L. Knott and Thomas G. P. Grünewald

The identification of cancer stemness genes is crucial to understand the underlying biology of therapy resistance, relapse, and metastasis. In pediatric tumors -despite being largely composed of undifferentiated stem celllike cells -mutations generally do not involve canonical stemness/metastasis-associated genes [1]. Ewing sarcoma (EwS), the second most common bone cancer in children and adolescents [2], is a highly aggressive malignancy associated with dismal outcome in metastatic disease [3], wherefore deciphering mechanisms of metastasis is imperative. EwS is characterized by a remarkably 'silent' genome with a single driver mutation generating an oncogenic fusion transcription factor (EWSR1-ETS) that promotes stem cell features [4,5]. Thus, EwS constitutes an ideal model to study how perturbation of a transcriptional network by a dominant oncogene can mediate metastasis.
To identify highly relevant EWSR1-ETS target genes involved in the stemness/metastasis axis we implemented a systems biology approach to analyze published transcriptome profiles of 5 EwS cell lines with/without shRNA-mediated knockdown of EWSR1-FLI1 or -ERG (< 20%) for 96 h [6] as described in Fig. 1a. This analysis yielded 348 differentially expressed genes (DEGs) being up-or downregulated (|log2 FC| ≥ 1.5) after knockdown of EWSR1-FLI1 or -ERG across all cell lines (Supplementary Table 1). We next filtered these DEGs for genes annotated with the significantly enriched gene ontology term 'Regulation of Cell Differentiation' using PantherDB (P = 4.93 × 10 − 12 , falsediscovery rate = 1.97 × 10 − 9 ). The resulting 76 DEGs (Supplementary Table 2), were subjected to network analysis for pathway, physical, and genetic interactions using Cytoscape and GeneMania (Fig. 1b, c). To identify clinically relevant key hubs within this network we focused on TFs (n = 11) that were highly interconnected (> 10 interactions) with a significant (P < 0.05, Mantel-Haenszel test) association with overall survival in a cohort of 166 EwS patients with matched gene expression and clinical data [7]. After adjustment for multiple testing (Supplementary Table 3), transcription factor 7 like 1 (TCF7L1, alias TCF3) emerged as the most promising candidate for functional follow-up, whose low expression was associated with poor patients' overall survival (nominal P = 0.0057; P = 0.0228 Bonferroni-adjusted)  (Fig. 1d). Next, we explored TCF7L1 protein levels in association with clinical outcome in tissue micro-arrays (TMAs) comprising 114 EwS tumors and found that low TCF7L1 protein expression was significantly (P = 0.035) associated with worse overall survival (Fig. 1e), supporting that low TCF7L1 upregulation of TCF7L1 (Supplementary Table 4), which was confirmed in 5 EwS cell lines by qRT-PCR (average ~ 6-fold upregulation) ( Supplementary Fig. 1b) and validated in A673 cells xenografts with/without conditional knockdown of EWSR1-FLI1 (A673/TR/shEF1 cells) at mRNA and protein levels ( Supplementary Fig. 1c).

Open Access
Prior reports suggested that TCF7L1 may have either oncogenic or tumor suppressor properties depending on the cellular context. Indeed, TCF7L1 expression has been linked to promotion of cell proliferation, tumor growth and sphere formation in breast cancer [8], colorectal cancer [9,10], acute lymphoblastic leukemia [11] and skin squamous cell carcinoma [12], while ectopic expression of TCF7L1 inhibits self-renewal of liver cancer stem cells [13]. To explore its role in EwS, we first analyzed the TCF7L1 expression pattern across 18 cancer entities using microarray data from the Cancer Cell Line Encyclopedia (CCLE) and an own study that compiled well-curated microarray data from the same cancer entities [14]. Surprisingly, TCF7L1 was very highly, but variably, expressed in EwS cell lines and primary tumors (Fig. 1f).
Weighted Gene Correlation Network Analysis (WGCNA) based on enriched gene-sets in TCF7L1-correlated genes in 166 EwS tumors showed that EwS tumors with low TCF7L1 expression were enriched in embryonic pathways (Fig. 1g). Although TCF7L1 is generally highly expressed in EwS, these data suggested that suppression of its transcription by EWSR1-FLI1 is associated with embryonic processes that may be linked with poor patient outcome. Thus, we generated two EwS cell lines (SK-N-MC and TC-71, with lowest TCF7L1 expression across the 5 EwS cell lines tested ( Supplementary Fig. 1a) with a doxycycline-(DOX)-inducible re-expression of TCF7L1 (Fig. 1h). Consistent with the hypothesis that TCF7L1 may be a EWSR1-ETS-repressed downstream transcription factor (with a potentially indirect regulation, possibly partially mediated by TCF7L1 promoter hypermethylation as depicted in Supplementary Fig. 1d-g), transcriptome profiling of two EwS cell lines after either knockdown of EWSR1-ETS or re-expression of TCF7L1 showed a highly significant (P = 2.57 × 10 − 165 or P = 5.34 × 10 − 272 ) overlap of concordant DEGs ( Supplementary Fig. 1e). Congruently, conditional re-expression of TCF7L1 significantly reduced cell proliferation (Fig. 1i), clonogenic (See figure on next page.) Fig. 1 Systems biology identifies TCF7L1 as a prognostically relevant EWSR1-ETS-regulated network hub whose re-expression inhibits tumorigenesis in vitro and in vivo. a) Workflow depicting a systems biology approach to identify DEGs regulated by EWSR1-ETS, involved in regulation of cell differentiation, functioning as highly interconnected TFs, and associated with overall survival in a cohort of 166 EwS patients. Number of genes represent remaining candidates after each filtering step. b) Network of EWSR1-ETS-regulated genes involved in regulation of cell differentiation. Genes are depicted as nodes (circles). Blue nodes represent TFs, node outline color show up-(green) or down-(red) regulation by EWSR1-ETS. Node border width represent strength of regulation by EWSR1-ETS (the thicker the border the higher the fold-change). Black dots surrounding the nodes represent interconnections with other nodes within the network. Connecting lines show three types of interconnection: physical (red), pathway (blue), genetic (brown). Line width represents strength of interconnection. c) Close-up image of highly interconnected TFs located at the center of the network (blue nodes). growth in two-dimensional (2D) cultures ( Supplementary  Fig. 2b), anchorage-independent and three-dimensional (3D) growth (Fig. 1j), and local tumor growth in a preclinical subcutaneous xenotransplantation mouse model (Fig. 1k). The xenografts exhibiting specific re-expression of TCF7L1 (DOX(+) group) ( Supplementary Fig. 2e,f) showed a reduced mitotic index (Fig. 1l), and a significant decrease of the proliferation marker Ki67 (Fig. 1m). Similar in vitro and in vivo experiments performed with both cell lines 'empty' controls exhibited no significant differences ( Supplementary Fig. 2a, c-d, g-i).
Since stemness features, such as elevated clonogenic capacity and anchorage-independent growth, are essential for circulating tumor cells to colonize and develop into clinically apparent metastases in distant organs, we reasoned that TCF7L1 may be linked to the metastatic process in EwS. In support of this hypothesis, transcriptome profiling, subsequent gene-set enrichment and WGCNA of TCF7L1-reexpressing SK-N-MC and TC-71 EwS cells uncovered that low TCF7L1 levels led to overrepresentation of gene-sets involved in cellular migration (Fig. 2a). Accordingly, re-expression of TCF7L1 resulted in significantly reduced transwell migration (Fig. 2b), invasion, and single-cell 3D migration in fibrin gel using advanced microfluidic chambers (Fig. 2c). Next, we employed an orthotopic spontaneous in vivo metastasis model by injecting TCF7L1 re-expressing cells into the proximal tibiae of immunocompromised mice. Remarkably, while there was no significant difference in local tumor growth in the limited space of the tibial plateau ( Supplementary   Fig. 3a,b), we observed a strong inhibition of macrometastatic spread to liver, lungs, and kidneys upon re-expression of TCF7L1, and significantly reduced micrometastatic burden in the same organs (Fig. 2d,e, Supplementary  Fig. 3c-e). In agreement with these in vivo findings, we observed that TCF7L1 was significantly (P = 0.016) lower expressed in EwS metastases (n = 7) compared to primary tumors (n = 118) in situ as defined by RNA-sequencing of tumor tissue (Fig. 2f). Strikingly, this result became even more significant when focusing on gene expression data of 4 matched pairs of metastases and primaries (P = 0.0078) (Fig. 2g).
To further investigate the mechanism of action of TCF7L1, we generated two EwS cell lines with conditional re-expression of different TCF7L1 deletion mutants for its major domains (ΔHMG: DNA binding domain; ΔCTNNB: β-catenin binding domain) and observed that only the ΔHMG mutants exhibited a normal clonogenic growth and migration capacity in vitro (Fig. 2h,i), and tumor growth in vivo (Fig. 2j). Moreover, further analysis of our WCGNA on TCF7L1-regulated signatures displayed in Fig. 2a highlighted several downregulated 'driver target genes' (aka leading-edge genes), including ANXA1, LMO7, SLC9A9, and TMEM71 as potential key mediators of TCF7L1 inhibition of the EwS migratory phenotype. These genes were selected for validation as they ranged among the top 10 downregulated genes by TCF7L1 and as their upregulation has been previously implicated in cancer progression [15][16][17][18]. As shown in Supplementary Fig. 3f, these genes are strongly downregulated upon re-expression of wildtype TCF7L1 in both EwS cell lines tested, but remain unaltered in the DNA-binding-domain deletion mutant (ΔHMG), further suggesting that DNAbinding of TCF7L1 is required to suppress the tumorigenic and migratory phenotype of EwS cells.

Conclusions
Previous studies proposed a binary model of EWSR1-FLI1 -high promoting proliferation and EWSR1-FLI1 -low promoting metastasis [19] following a traditional epithelial-mesenchymal-transition concept. However, our findings support an integrative concept of the 'migratory stem cell' [20]: EwS cells may reside in a 'metastable' state where all EwS cells of a tumor may be equipped with both proliferative and migratory capacities [21]. Upon different intrinsic/extrinsic cues (of which EWSR1-FLI1 may not be the only important component) these cells could underlie a certain accentuation.
In this context, we demonstrated that TCF7L1 is a critical mediator of metastasis in EwS, which may be utilized as a prognostic biomarker. Since we could detect a strong inverse correlation of TCF7L1 levels with patients' overall survival in both the mRNA-as well as in the TMA-cohort, we believe that the most straightforward possibility to translate TCF7L1 into a routine clinical setting would be the IHC detection and semiquantitative evaluation of TCF7L1 in primary biopsies. This technique is readily available, inexpensive, and the conditions of IHC staining of TCF7L1 in EwS tumors have been established in the current study, which would offer the possibility of further evaluating the prognostic value of TCF7L1 in ongoing and prospective clinical trials. In addition, it would be highly interesting to explore whether TCF7L1 may serve as a prognostic biomarker in other (EWSR1-rearranged) mesenchymal neoplasms, which is subject to future studies.
In summary, our study exemplifies the power of systems biology to decipher gene regulatory networks and to identify key players in the metastatic process, which may be highly relevant for, and translatable to, other oligomutated (pediatric) cancers.