FET fusion oncoproteins disrupt physiologic DNA repair networks and induce ATR synthetic lethality in cancer

The genetic principle of synthetic lethality is clinically validated in cancers with loss of specific DNA damage response (DDR) pathway genes (i.e. BRCA1/2 tumor suppressor mutations). The broader question of whether and how oncogenes create tumor-specific vulnerabilities within DDR networks remains unanswered. Native FET protein family members are among the earliest proteins recruited to DNA double-strand breaks (DSBs) during the DDR, though the function of both native FET proteins and FET fusion oncoproteins in DSB repair remains poorly defined. Here we focus on Ewing sarcoma (ES), an EWS-FLI1 fusion oncoprotein-driven pediatric bone tumor, as a model for FET rearranged cancers. We discover that the EWS-FLI1 fusion oncoprotein is recruited to DNA DSBs and interferes with native EWS function in activating the DNA damage sensor ATM. Using preclinical mechanistic approaches and clinical datasets, we establish functional ATM deficiency as a principal DNA repair defect in ES and the compensatory ATR signaling axis as a collateral dependency and therapeutic target in FET rearranged cancers. Thus, aberrant recruitment of a fusion oncoprotein to sites of DNA damage can disrupt normal DSB repair, revealing a mechanism for how oncogenes can create cancer-specific synthetic lethality within DDR networks.


Main
The DNA damage response (DDR) is a tightly regulated and redundant network that tailors specific repair complexes to address a diverse set of genotoxic insults. In cancer, genetic loss of certain DDR components creates tumor-specific vulnerabilities that can be therapeutically exploited 1 . The canonical example of DDR synthetic lethality is the use of poly-ADP ribose polymerase (PARP) inhibitors to target defective homologous recombination (HR) repair in BRCA1/2 mutant breast and ovarian cancers 2,3 . The broader applicability of DDR-directed therapies in cancers without genetic DDR alterations and whether oncogenes create vulnerabilities within DDR networks remains to be defined.
The FET family of intrinsically disordered proteins (FUS, EWS, TAF15) are frequent 5' oncogenic transcription factor (TF) fusion partners in a diversity of sarcomas and leukemias 4 . The most studied cancer in this class is Ewing sarcoma (ES), a pediatric bone tumor driven by the EWS-FLI1 TF fusion oncoprotein. Patients with relapsed or metastatic ES continue to have dismal outcomes despite maximally intense combination chemotherapy regimens 5 . Paradoxically, decades of clinical experience and laboratory testing of ES cancer cell lines have shown ES to be among the most chemo-and radiosensitive cancers [6][7][8][9] , at least initially. It has long been hypothesized that ES tumors harbor a DNA repair defect that explains their underlying sensitivity to DNA damaging therapies. The prevailing model is that ES belongs to a family of "BRCA-like" tumors that are functionally deficient in BRCA1 (due to sequestration of BRCA1 protein by RNA-DNA hybrid structures known as R-loops) and therefore defective in HR double-strand break (DSB) repair 10 . However, both xenograft studies 8 and clinical trials in ES patients 11 failed to demonstrate any benefit for PARP inhibitor monotherapy, in stark contrast to the impressive clinical responses seen across HR deficient BRCA mutant and BRCA-like cancers 12,13 . Thus, the precise nature of the DNA repair defect in ES remains uncertain. Native FET family proteins contain an N-terminal intrinsically disordered region (IDR) required for interactions amongst FET family members and a C-terminal domain with positively charged RGG (arginine-glycine-glycine) repeats, which mediate recruitment to DSBs via high affinity interactions with negatively charged poly-ADP ribose (PAR) molecules 14,15 . All 3 FET members are rapidly recruited to DSBs in a PARP-dependent manner 15,16 , where they undergo liquid-liquid phase separation that is thought to enable compartmentalization of DSB repair proteins 17,18 , though the specific role of FET proteins in DSB repair is not well defined. Interestingly, all oncogenic FET fusion proteins including EWS-FLI1 share a similar structure: the N-terminal IDR of the FET protein fused to the DNA binding domain of a transcription factor (e.g. FLI1), with loss of the Cterminal RGG repeats 19 .
In this study, we address the role of the oncogenic fusion protein EWS-FLI1 in regulating the DNA damage response in ES. Contrary to the current classification of ES as a BRCA-like tumor with defective HR 10 , we identify functional ATM deficiency as the principal DDR lesion in ES cells and define the mechanistic basis for EWS-FLI1 mediated DNA repair defects in ES. More broadly, our findings demonstrate how an oncoprotein can create tumor-specific vulnerabilities within the DDR network.

ES cells are dependent on HR factors for survival.
To address the uncertainty surrounding putative DNA repair defect(s) in ES, we set out to identify genes that modulate ES cell survival in response to doxorubicin, a DNA DSB-inducing agent and major component of current chemotherapy regimens 20 (Fig. 1A). We selected a CRISPR interference (CRISPRi) based screening approach given the concern of studying DSB repair phenotypes using an active Cas9 that generates DSBs 21 . Surprisingly, we found that key HR factors BRCA1 and PALB2 were essential for the growth of ES cells even in the absence of doxorubicin (Fig. 1B). Given the prevailing model that ES tumors are functionally HR-deficient, BRCA-like tumors 10 , the screen results were unexpected. We validated the finding of HR factor dependency in ES using two independent guide RNAs (gRNAs) against BRCA1 and PALB2 in the screening cell line (A673) and two additional ES cell lines TC71 and ES8 (Figs. 1C-E and S1A, B). In contrast, the same BRCA1 and PALB2 gRNAs had limited effects on cell growth in two non-ES cancer cell lines, which was consistent with the set of published CRISPRi screens [21][22][23] and suggested the observed dependency on HR factors may be specific to ES cells (Figs. S1C, D). To test the role of EWS-FLI1 in inducing this dependency, we utilized an ES cell line with a doxycycline-inducible shRNA against EWS-FLI1 24 (Fig. S1E). We observed that EWS-FLI1 knockdown rescued the growth defects caused by BRCA1 or PALB2 loss, confirming that the oncogenic fusion protein is necessary for the observed dependency on HR factors in ES cells ( Fig. 1F).
CRISPRi screening also identified genes whose loss sensitized ES cells to doxorubicin including LIG4, NHEJ1 (XLF), and 53BP1 (Fig. S1F). The presence of key canonical non-homologous end joining (c-NHEJ) genes as top chemo-sensitizer hits validated our experimental approach as this pathway is critical for repairing both drug and ionizing radiation (IR) induced DSBs and provides evidence for a functional c-NHEJ pathway in ES cells. The list of top chemosensitizer genes included Aurora Kinase A (AURKA), for which inhibitors are under clinical development 25 , and the E3 Ubiquitin Ligase RNF8, both of which could be targeted in combinatorial therapeutic approaches with doxorubicin-based treatment regimens. The final category of screen hits were genes whose loss promoted survival under high doses of doxorubicin (LD97), mirroring the residual disease state in ES patients (Fig. S1G). SLFN11 is a notable hit as loss of SLFN11 has been shown to promote chemotherapy resistance in multiple cancer subtypes including ES 26

EWS-FLI1 impairs resection-dependent DSB repair.
The absence of HR deficient genomic scars in ES tumors and paradoxical requirement of HR factors for ES cell survival prompted us to systematically re-examine how EWS-FLI1 impacts DSB repair pathway utilization (Fig. 3A). We posited that previous reports of defective HR in ES might alternatively be explained by a more general upstream defect in DSB repair. We utilized a set of well-established DSB repair reporter cell lines wherein expression of the I-SceI endonuclease induces a DSB within an interrupted GFP reporter cassette, such that utilization of a particular DSB repair pathway restores a GFP coding sequence enabling a quantitative readout of individual repair pathway efficiency 32 . Nucleofection of EWS-FLI1 into each reporter cell line was performed using a BFP-expressing, dual promoter plasmid, followed by expression of mCherry-I-SceI after 24 hours to induce a single DSB (Figs. S3A, B). The use of fluorescently tagged plasmids enabled detection of DSB repair specifically in cells that expressed both EWS-FLI1 (or empty vector) and I-SceI (Fig. S3A).
To examine HR repair, we utilized the DR-GFP reporter and observed a reduction in HR upon EWS-FLI1 expression, consistent with previous reports 10 (Fig. 3B). However, EWS-FLI1 expression also reduced the efficiency of MH-mediated alt-EJ repair (EJ2) and long-stretch MH We also evaluated c-NHEJ, a fast-acting DSB repair pathway which does not require endresection, using the EJ5 reporter system. We observed an increase in the usage of c-NHEJ upon EWS-FLI1 expression (Fig. 3E) consistent with our CRISPRi screen findings of an intact c-NHEJ pathway (Fig. S1F) and the increased frequency of small deletions observed in ES patient samples (Fig. 2C). We verified that these repair phenotypes were not the result of EWS-FLI altering cell cycle profiles or causing a cell cycle arrest (Figs. S3D, E). To assess how EWS-FLI1 regulates the transcription of DDR genes that control resection-dependent DSB repair, we analyzed published RNA-sequencing data in ES cells before and after EWS-FLI1 knockdown 34 ( Fig. S3F). Interestingly, we found that EWS-FLI1 either increased or had minimal effect on the expression of key genes involved in resection-dependent DSB repair (MRE11, CtIP, BRCA1/2, PALB2, RAD52, POLQ), and had no effect on expression of many c-NHEJ genes (LIG4, NHEJ1).
These data suggest that EWS-FLI's effect on DSB repair pathway utilization is not the result of transcriptional downregulation of key end-resection and resection-dependent DSB repair genes (e.g. BRCA1) by the fusion oncoprotein. In summary, we demonstrate that EWS-FLI1 does not induce isolated HR deficiency but instead compromises all three resection-dependent DSB repair pathways.

ATM activation and signaling is defective in Ewing sarcoma.
To explain our finding that EWS-FLI1 impairs multiple branches of resection-dependent DSB repair, we focused on the upstream regulation of the DDR and end-resection in ES cells (Fig.   3A). The DDR is a partially redundant signaling network regulated by three kinases, DNA-PK, ATM, and ATR, each of which control distinct aspects of DDR signal amplification and DSB pathway choice. ATM was a logical candidate since it promotes DNA end-resection and HR 35 , and ATM loss creates a synthetic lethal dependence on HR proteins 36,37 . We therefore tested whether EWS-FLI1 affects the activation and function of these three apical DDR kinases. Both ATM and ATR coordinate aspects of resection-dependent DSB repair (Fig. 3A) and ATM mutant tumors display synthetic lethality with ATR inhibitors, reflecting the vital compensatory role of the ATR signaling axis in the absence of ATM 38, 39 . We therefore tested if functional ATM deficiency caused by EWS-FLI1 induced a collateral dependence on the ATR signaling axis in ES given this emerging synthetic lethal relationship between the two DDR kinases. Indeed, ES cells displayed increased sensitivity to inhibitors of both ATR and its key downstream target CHK1, and inhibitor sensitivity was reversed by EWS-FLI1 knockdown in both cases (Figs. 4F and S4I). Overexpression of RNAseH1 which degrades R-loops, the major source of oncogeneinduced replication stress in ES 10 , did not alter the ATR inhibitor sensitivity of ES cells (Figs. S4J, K) suggesting that the molecular basis of ATR inhibitor response in ES cells may be a consequence of functional ATM deficiency. The collective findings establish an EWS-FLI1dependent specific impairment of ATM function in ES and resultant synthetic lethality with the compensatory ATR signaling axis.

Loss of native EWS phenocopies the DNA repair defects caused by EWS-FLI1.
How does EWS-FLI1 regulate DNA repair to create ATM defects? We hypothesized that The initial recruitment of EWS to laser micro-irradiated DSBs was unaffected by EWS-FLI1.
Instead, EWS-FLI1 expression resulted in the premature and rapid clearance of EWS from DSB stripes (Figs. 5A, B). To determine whether loss of native EWS function at DNA DSBs may contribute to EWS-FLI1-dependent DNA repair defects, we first characterized the role of native EWS in DSB repair using the pathway-specific DSB repair reporters. Analogous to EWS-FLI1 expression, native EWS knockdown decreased all three resection-dependent DSB repair pathways, HR, alt-EJ and SSA (Figs. 5C-E and S5B). Interestingly, c-NHEJ efficiency was also reduced upon EWS knockdown (Fig. 5F), highlighting differences between native EWS loss and EWS-FLI1 overexpression that may relate to the rapid kinetics of c-NHEJ repair and preserved initial recruitment of native EWS despite EWS-FLI1 expression.
We further assessed the role of native EWS in activation of the upstream DDR kinases ATM, DNA-PK, and ATR. Analogous to EWS-FLI1 expression, native EWS knockdown reduced activation of ATM itself (autophosphorylation) and to a greater extent, functional ATM signaling (phosphorylation of CHK2 and KAP1) upon IR, again with either no effect (DNA-PK) or increased (ATR) activation of the other two DDR kinases (Figs. 5G and S5C-G). We utilized a proximity ligation assay (PLA) to establish an IR-dependent interaction between pH2AX and EWS (Figs. 5H and S5H), further supporting a role for native EWS in coordinating early DSB repair responses.
Finally, knockdown of native EWS also decreased IR-induced pH2AX signal intensity by flow cytometry, but not pH2AX foci number, analogous to EWS-FLI1 overexpression (Fig. 5I). These data define the function of native FET family member EWS as a specific regulator of ATM activation and signaling. In total, we demonstrate that EWS-FLI1 causes premature loss of native EWS from DSBs and functional ATM defects that are phenocopied by native EWS knockdown.

EWS-FLI1 and other FET fusion oncoproteins are recruited to DNA double-strand breaks.
To determine a molecular basis for the ATM defects seen in ES, we asked whether the interaction between native EWS and EWS-FLI1 ( Despite these differences, the consequence of functional ATM deficiency in CCS was similarly increased reliance on compensatory ATR signaling, as CCS cells displayed EWS-ATF1dependent synthetic lethality with ATR inhibition (Fig. 6J). Taken together, our results support a model in which EWS fusion oncoproteins are aberrantly recruited to DSB repair sites and induce specific defects in ATM function.

Discussion
Here we detail a new mechanism for how oncogenes can create tumor-specific vulnerabilities in DNA damage repair networks: aberrant recruitment of an oncogenic fusion protein to sites of DNA damage causing disruption of normal DSB repair biology. We identify functional ATM deficiency as the principal DDR defect in ES and demonstrate oncogene-dependent synthetic lethality with ATR inhibition as a collateral dependency. These data provide an initial example of how DDRdirected therapies could be utilized in the broader set of cancers without genetic DDR alterations through improved understanding of how oncogenes interact with DNA damage networks.
The nature of the DNA repair defect in ES has been the subject of much debate based on the strong clinical and laboratory-based data demonstrating chemo-and radio-sensitivity 6-10 . Our discovery that ES tumors are not "BRCA-like", but instead functionally ATM deficient, may help explain the lack of clinical responses to PARP inhibitor monotherapy in ES patients 11,43 (unlike HR deficient BRCA mutant and BRCA-like cancers 12,13 ). ES cells are dependent on key HR genes for survival and direct analysis of ES patient tumors revealed none of the genomic hallmarks of HR loss. Instead of isolated HR deficiency, we show that EWS-FLI1 creates a broader defect in resection-dependent DSB repair and a specific impairment of upstream ATM activation and signaling. ES cells display increased reliance on a compensatory DDR kinase, ATR, and EWS-FLI1-dependent synthetic lethality with ATR and CHK1 inhibitors. These data nominate functional ATM deficiency as the principal DNA repair defect in ES.
We propose a model wherein the rapid recruitment of native FET proteins (e.g. EWS) are important for c-NHEJ, consistent with our findings upon native EWS knockdown and published reports on FUS's role in canonical repair 44 (Fig. 7). The FET proteins then coordinate local compartmentalization of DSB repair factors leading to ATM chromatin recruitment and activation, amplification of the ATM signaling cascade and phosphorylation of key substrates (e.g. CHK2), and slower resection-dependent repair. Our work raises important new questions as to how native FET proteins interact with DDR scaffolding proteins and how IDR-mediated phase separation might enable DSB compartmentalization and subsequent repair. In ES, we show that the EWS- Finally, what might be the selective advantage for TF fusion oncoproteins to also induce a DNA repair deficiency? We propose that by partially disabling physiologic DDR signaling, cells with FET rearrangements can tolerate high levels of DDR activation caused by transcription and replication stress induced by the oncoprotein itself and thus overcome an important barrier to cellular transformation. Whether other TF fusion oncoproteins (or oncogenes more generally) interact with and disrupt DDR signaling networks as part of tumorigenesis will be an intriguing topic for future work. In summary, our study provides a new mechanism for how oncogenes    Supplementary Table 1. For all panels, * denotes p< 0.05, ** denotes p < 0.01, ns denote not significant by one-way ANOVA with post hoc Tukey's HSD test.          After 72 hours, cells were trypsinized, counted and pooled every 3 days, and then replated in media without doxorubicin at the same density to maintain minimum 500× coverage of each sgRNA construct. Cells were viably frozen after 10 days and cell counts were used to determine actual lethal dose (LD) values. 2 biologic replicates were performed, and data combined.
Deep sequencing and data analysis were performed as described 21