Translational evidence for RRM2 as a prognostic biomarker and therapeutic target in Ewing sarcoma

Purpose Ewing sarcoma (EwS) is a highly aggressive bone- or soft tissue-associated malignancy mostly affecting children, adolescents, and young adults. Although multimodal therapies have strongly improved patients’ overall survival over the past decades, the development of prognostic biomarkers for risk-based patient stratification and more effective therapies with less adverse effects is stagnating. Thus, new personalized medicine approaches are urgently required. Experimental design Gene expression data of EwS and normal tissues were crossed with survival data to identify highly overexpressed, prognostically relevant, and actionable potential targets. RNA-interference and dose-response assays as well as tissue-microarray analyses were carried out to explore the functional role and druggability of a prominent candidate gene in vitro and in vivo, and to validate its suitability as a prognostic biomarker. Results Employing a multilayered screening approach, we discover ribonucleotide reductase regulatory subunit M2 (RRM2) as a promising therapeutic target and prognostic biomarker in EwS. Through analysis of two independent EwS patient cohorts, we show that RRM2 mRNA and protein overexpression is associated with an aggressive clinical phenotype and poor patients’ overall survival. In agreement, RRM2 silencing as well as pharmacological inhibition by the specific inhibitor triapine (3-AP) significantly reduces EwS growth in vitro and in vivo. Furthermore, we present evidence that pharmacological RRM2 inhibition by triapine can overcome chemoresistance against doxorubicin or gemcitabine, and synergize with cell cycle checkpoint inhibitors (CHEK1 or WEE1). Conclusions Based on the aggressive phenotype mediated by and the druggability of RRM2 our results provide a translational rationale for exploiting RRM2 as a novel therapeutic target in EwS and prompt further clinical investigations.


Main text
Ewing sarcoma (EwS) is an aggressive bone-or soft tissueassociated malignancy, characterised by the fusion oncoprotein EWSR1-FLI1 [1]. Over the past decades further therapeutic development for this devastating childhood tumour has remained relatively stagnant [2], especially for patients with metastatic or recurrent disease [3,4]. To develop more effective and specific treatment options we investigated potential therapeutic targets by exploring putative downstream genes of EWSR1-FLI1.
We took advantage of publicly available 'omics' data and filtered them in a multi-step approach (Fig. 1a): First, we interrogated a gene expression dataset comprising 50 primary EwS and 929 samples from 71 normal tissue types to identify overexpressed genes (min. log2 fold increase = 2) in EwS, which yielded 292 candidates (Fig. 1b, Supplementary Table 1). Second, we filtered for those genes whose overexpression was significantly negatively correlated with patients' overall survival in a dataset of matched gene expression and survival data of 166 EwS patients [5] that covered 280 of the 292 overexpressed genes (96%) (Fig. 1c), identifying 22 candidates (Supplementary Table 1). Third, we focused on druggable targets possessing kinase or other enzymatic functions for which specific inhibitors and their pharmacokinetic data were available, but were still not (pre) clinically tested in EwS. This survey identified ribonucleotide reductase regulatory subunit M2 (RRM2) as the single putative target with a prominently negative association with patients' overall survival (Fig. 1d). The ribonucleotide reductase (RNR) catalyses the conversion of ribonucleoside diphosphates to deoxyribonucleoside diphosphates, the rate-limiting process for de novo deoxyribonucleoside triphosphates synthesis. RNR is composed of two subunits, ribonucleotide reductase catalytic subunit M1 (RRM1) and either RRM2 or ribonucleotide reductase regulatory TP53 inducible subunit M2 (RRM2B) [6]. Notably, RRM2B is neither overexpressed in EwS nor negatively correlates with patients' outcome ( Supplementary Figs. 1a, cell line models, RRM2 was on average ~ ninefold higher expressed than RRM2B (P < 0.0001). These observations, together with the absence of a negative survival association of RRM2B in EwS ( Supplementary Fig. 1b), suggested that RRM2B, although being structurally similar to RRM2 [6], may play a subordinate role in EwS.
Prior reports suggested that RRM2 may contribute to the proliferative phenotype of EwS [7,8]. However, its role in primary EwS tumours remains unclear. To gain first insights into the biological function of RRM2 in EwS, we carried out gene ontology (GO) enrichment analysis of RRM2 co-expressed genes in 166 EwS tumours, which revealed that high RRM2 expression is closely correlated with cell proliferation-associated gene signatures (Fig. 1e), suggesting that high RRM2 expression may contribute to an aggressive clinical course by promoting tumour growth. Next, we analysed the potential association between RRM2 protein levels, clinicopathological prognostic factors, and clinical outcomes in tissue microarrays (TMA) from EwS tumours of 122 patients (Supplementary Table 2, Supplementary Fig. 2a). In agreement with the findings at the mRNA level (Fig. 1d), high RRM2 protein expression was significantly (P = 0.0095) associated with poor overall survival (Fig. 1f ). Correspondence analyses of individual cohorts and the joint-cohort (after exclusion of 6 samples (3.6%) from the mRNA-cohort that were in overlap with the TMA cohort) revealed that high RRM2 expression was significantly associated with metastatic disease at diagnosis (P = 0.0004) and occurrence of metastatic and/or local relapse (P = 0.0095; only available for the TMA cohort) (Supplementary Table 3), supporting that high RRM2 expression promotes an aggressive phenotype. Conversely, RRM2 inhibition by doxycycline (Dox)-inducible shRNA-mediated gene silencing inhibited proliferation and clonogenic growth of three EwS cell lines, and induced cell death in vitro (Supplementary Figs. 2b,c). Consistent with these functional experiments, transcriptome profiling upon RRM2 silencing in two EwS cells demonstrated downregulation of cell cycle and proliferation-associated gene signatures (Supplementary Fig. 2d). Similarly, RRM2 knockdown significantly reduced tumour growth of two xenografted EwS cells (Figs. 1g,h). This antineoplastic effect was accompanied by increased apoptosis and DNA damage, as assessed by immunohistochemistry for cleaved caspase 3 (CC3) and γH2A.X, respectively (Fig. 1i, Supplementary  Fig. 2e).
Generally, the activity of RNR can be blocked by irreversible RRM1 inhibition using gemcitabine, or by RRM2-specific inhibitors such as hydroxyurea or the more potent compound triapine (alias 3-AP) [6,9]. Although gemcitabine is used for palliative treatment of EwS patients, EwS tumours rapidly develop a relative resistance [10]. Consistently, we found that long-term treatment of EwS cells with ascending doses of either doxorubicin (A-673, ES7, EW-7, TC-71), gemcitabine (A-673, ES7, TC-71) or triapine (A-673) led to acquisition of relative resistance phenotypes in vitro (Supplementary Fig. 3a), where we noted a relatively fast and strong increase of the relative resistance towards gemcitabine (> 2,000-fold increase in IC 50 within ~ 6 weeks), compared doxorubicin (~ fourfold increase in ~ 28 weeks) and triapine (~ sevenfold increase in ~ 20 weeks) (Supplementary Fig. 3b), further suggesting that gemcitabine has limited potential for clinical treatment with curative intent. Thus, we focused on triapine for further functional analyses. First, to assess functional dependency of triapine on RRM2 expression, we performed drugresponse assays using triapine in EwS cells with/without RRM2 silencing, which demonstrated that knockdown of RRM2 led to a ~ twofold decrease of the IC50 for triapine in A-673 EwS cells, indicating that higher doses of the drug are required to fully block RRM2 activity in case of high RRM2 expression (Supplementary. Figure 3c). Such differential effect on sensitivity towards triapine was not observed in A-637 cells expressing a non-targeting control shRNA. Moreover, we observed an ~ twofold increase of RRM2 expression in triapine-resistant A-673 (A-673/ TR) compared to parental A-673 EwS cells, suggesting that RRM2 upregulation can be a potential mechanism for acquiring triapine-resistance in A-673 EwS cells (Supplementary Fig. 3d). Dose-response assays revealed that EwS cells were very sensitive towards triapine compared to osteosarcoma cells and non-transformed EwS patientderived mesenchymal stem cells (mean IC 50 values 0.35, 1.63, 101.63 µM, respectively) (Fig. 2a). Likewise, triapine treatment significantly reduced clonogenic growth of EwS cells at clinically relevant doses [11,12] (Fig. 2b). Interestingly, doxorubicin or gemcitabine resistant EwS cells (designated EwS/DR or EwS/GR, respectively) still retained triapine sensitivity (Fig. 2c), suggesting therapeutic potential of triapine for EwS refractory towards conventional chemotherapy. Strikingly, we could confirm a significant reduction of tumour growth by triapine treatment compared to controls in a subcutaneous xenograft model (Fig. 2d). Although triapine treatment was accompanied by weight loss (on average ~ 5% at the experimental endpoint) ( Supplementary Fig. 3e), no morphological changes of inner organs were observed including the gastrointestinal tract as assessed by histological analysis (Supplementary Fig. 3f ). Although it is interesting to note that EwS cells resistant to gemcitabine, which covalently binds and thus inactivates RRM1 [13], still retain sensitivity towards triapine (Fig. 2c), it should be noted that triapine may not be entirely specific for RRM2. Triapine presumably disrupts a tyrosyl free radical by labilising di-iron molecules on the small subunits of ribonucleotide reductase [6,14]. This proposed mechanism of action for triapine can clinically manifest as a reversible adverse effect such as methemoglobinemia, which is probably caused by the iron chelating effect of triapine, interrupting recovery cycles from methaemogloblin to haemoglobin [15]. To mitigate this toxicity, the small molecule COH29 has been developed that, upon binding to RRM2 subunits, interferes with the molecular interface of RRM1 and RRM2 subunits and thus inhibits its reductase function [14]. Yet, its clinical efficacy and safety remain to be investigated. Another approach for more specific RRM2 inhibition has been undertaken with antisense oligonucleotide-based techniques, exemplified by therapeutic silencing of RRM2 by GTI-2040, which, however, showed little clinical benefit in several clinical trials [16][17][18]. Hence, despite our data strongly support RRM2 as an actionable and valuable drug target in EwS, and triapine as a potential lead candidate drug for preferential RRM2 inhibition, the development of even more specific RRM2 inhibitors is desirable. We next explored effective drug combinations with triapine. Based on known functions of RRM2 in DNA synthesis and DNA repair [6] we examined combinatory applications of triapine with standard chemotherapeutics, doxorubicin, etoposide or vincristine, or poly ADP-ribose polymerase (PARP) inhibitors. Unexpectedly, we observed rather antagonistic effects (Supplementary Figure 4a). To identify rational combinations, we analysed integrated transcriptome profiles of two EwS cells upon RRM2 silencing and triapine treatment, revealing 263 commonly up-and downregulated genes (Supplementary Table 4). GO enrichment analysis demonstrated significant enrichment for cell cycleassociated processes, especially regulation of mitotic cell cycle-associated genes (Fig. 2e), which is consistent with the observation that RRM2 inhibition caused G1/S-phase cell cycle arrest [19,20]. Thus, we reasoned that RRM2 may synergise with checkpoint inhibitors targeting CHEK1 (checkpoint kinase 1) or WEE1 (WEE1 G2 checkpoint kinase), which were highly significantly (P < 0.0001) co-expressed with RRM2 in 166 EwS tumours (Fig. 2f ). In drug combination assays we observed a strong synergism between triapine and a CHEK1 inhibitor (CCT245737) or a WEE1 inhibitor (MK-1775) across four EwS cells (Fig. 2g,h, Supplementary Fig. 4b). Overall, these results provide a rationale for therapeutic combination of triapine with cell cycle checkpoint inhibitors. A recent study pointed out that the drug combination of hydroxyurea and a CHEK1 inhibitor (GDC-0575) can circumvent toxicities caused by the combination of gemcitabine and GDC-0575, which may imply a more manageable combinatory application through RRM2 inhibition and CHEK1 inhibitors [21].

Conclusions
Collectively, our results establish RRM2 as a promising actionable therapeutic target for EwS, even in chemotherapy-refractory cases, and suggest that the combination of triapine with cell cycle checkpoint inhibitors may be highly effective. Moreover, our integrative study of two independent cohorts provides evidence for RRM2 as novel and robust prognostic biomarker that can be readily assessed by immunohistochemistry in routine diagnostics. Thus, our findings may have immediate translational relevance for patients affected by this devastating disease.