Disruption of Microhomology-mediated End-joining in Ewing Sarcoma

SUMMARY

Ewing sarcoma (EwS) is a group of bone and soft tissue cancers in children and young adults. Since EwS cells have pronounced sensitivity to radiation and chemotherapy-induced DNA damage, the role of the oncoprotein, EWS-FLI1, in DNA repair is likely. Here, we demonstrate that EWS-FLI1 causes a defect in microhomology-mediated end-joining (MMEJ) repair. EWSR1 is a splicing factor that promotes the faithful splicing of the POLQ pre-mRNA, required for the expression of Polθ, a critical protein in the MMEJ pathway. Expression of EWS-FLI1, or depletion of EWSR1, causes increased POLQ exon 25 skipping, decreased Polθ expression, impaired MMEJ, and enhanced cellular sensitivity to inhibitors of the Fanconi Anemia (FA), HR, or NHEJ pathways, through the mechanism of synthetic lethality. Correction of POLQ exon 25 skipping restored Polθ expression and MMEJ activity in EwS. Inhibitors of the FA, HR, or NHEJ may therefore provide a targeted therapy for EwS patients.

Graphical Abstract

eTOC blurb:

Asada et al. show that EWS-FLI1 disrupts EWSR1-dependent POLQ pre-mRNA splicing, causing Polθ loss and MMEJ deficiency in Ewing sarcoma. This repair defect creates synthetic lethal vulnerabilities to FA, HR, and NHEJ pathway inhibition, revealing actionable therapeutic strategies.

INTRODUCTION

Ewing Sarcoma (EwS) is a group of aggressive bone and soft tissue cancers in children and young adults^1,2. EwS tumor cells have a chromosome 11;22 translocation, resulting in expression of the EWS-FLI1 fusion oncoprotein. Although EwS patients initially respond well to ionizing radiation and DNA-damaging chemotherapy, those patients with relapsed or refractory disease have poor prognosis, with a 5-year overall survival rate of approximately 10%³, highlighting an urgent need for novel therapeutic approaches.

Their pronounced sensitivity to ionizing radiation and DNA-damaging chemotherapy, coupled with a relatively low level of somatic mutational rate, suggests that EwS cells have an underlying defect in an error-prone DNA damage repair (DDR) pathway. Recent studies have begun to characterize abnormalities in DDR in EwS. Expression of the EWS-FLI1 fusion protein has been shown to promote R-loop formation and impaired BRCA1 recruitment to chromatin, suggesting a defect in the homologous-recombination repair pathway (HRD)⁴. Despite this preclinical observation, PARP inhibition, an established treatment for HR-deficient tumors, has proven to be ineffective in the treatment of EwS patients⁵, suggesting other more profound DDR defects, beyond the HRD phenotype, may underlie the cellular phenotype of EwS. Indeed, EWS-FLI1 expression influences multiple cellular processes that could affect DDR, including BAF complex-mediated transcription^6,7, condensate generation^8–10, and mRNA splicing¹¹, yet the precise mechanism by which EWS-FLI1 impairs specific DDR pathways remains unclear.

DNA double-strand breaks (DSBs) are the most critical types of DNA damage in cells. There are three main pathways for repairing DSBs - non-homologous end-joining (NHEJ), homologous recombination (HR), and microhomology-mediated end-joining (MMEJ). For a cancer cell with a defect in one DDR pathway, to survive and proliferate, DSBs must be repaired, and the cell must rely more on another pathway. The MMEJ pathway is mediated by DNA polymerase theta (Polθ), an error-prone polymerase encoded by the POLQ gene. Overexpression of POLQ confers chemoresistance, whereas depletion of Polθ results in hypersensitivity to irradiation and chemotherapy, which is a similar phenotype to EwS^12,13. To date, there are no reported human cancers with defects in the MMEJ pathway.

Here, we identify EwS as an MMEJ-deficient cancer driven by EWS-FLI1-mediated disruption of EWSR1-dependent POLQ splicing. This defect reduces Polθ expression and creates therapeutically actionable dependencies on FA, NHEJ, and HR repair pathways.

RESULTS

EWSR1 knock-out cells phenocopy MMEJ-deficient cells.

Previous CRISPR screens and DepMap analyses have demonstrated that loss of MMEJ genes is synthetic lethal in cells with an underlying defect in an HR gene^14,15. Interestingly, in these screens, knockout of the EWSR1 gene was highly similar to knockout of known MMEJ genes, including POLQ¹⁶, XRCC1¹⁷, APEX2¹⁸, and RHNO1¹⁹, in the killing of HR-deficient tumor cells (Fig. S1A). These findings are further supported by Repair-Seq analysis, in which EWSR1 and POLQ knockouts cluster together, suggesting a shared function in DNA damage repair²⁰. To investigate the function of EWSR1 in DNA repair, we generated EWSR1-knockout RPE TP53−/− cells (Fig. 1A). EWSR1 knockout resulted in hypersensitivity to ionizing irradiation, topoisomerase inhibitors, and ATR inhibitors, suggesting a defect in DNA damage repair (Fig. S1B–E). The drug sensitivity profile of EWSR1 knockout cells was similar to that of POLQ knockout cells, suggesting a loss of MMEJ¹².

Figure 1: EWSR1-knockout cells display MMEJ-deficient phenotypes through Polθ protein reduction.

(A) Immunoblotting of EWSR1 and GAPDH in CRISPR-edited RPE1 TP53−/− clones.

(B, C) MMEJ reporter analysis in U2OS I-SceI-based (B, n=9) and Cas9-based (C, n=4) MMEJ reporter cells transduced with sgRNA targeting control (NT), EWSR1, or POLQ. Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(D) Colony formation assay of wild-type and BRCA2-knockout DLD1 cells transduced with sgRNA targeting control (NT), EWSR1, or POLQ (n=3, left). Shown are representative images of the colonies (right). Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(E) Immunoblotting of Polθ, FLAG, EWSR1 and GAPDH in RPE1 TP53−/− 3xFLAG-POLQ knock-in cells transduced with Cas9 and sgRNA targeting control (NT), EWSR1, or POLQ.

(F) Immunoblotting of MMEJ proteins, EWSR1, and GAPDH in HEK293T 3xFLAG-POLQ single knock-in cells and HEK293T 3xFLAG-POLQ/2xHA-FKBP12^V36F EWSR1 double knock-in cells treated with DMSO or dTAG^V1 for indicated time.

(G) Immunofluorescence of D DNA damage-induced Polθ foci in V5-tagged endogenous Polθ-expressing cells with or without EWSR1 depletion (scale bars = 2μm). Statistical analysis was performed using Mann-Whitney test.

(H) MMEJ reporter (I-SceI-based) analysis in U2OS cells transduced with a control vector or POLQ cDNA, followed by transduction with Cas9 and sgRNA targeting control or EWSR1. Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(I) Colony formation assays of RPE1 TP53−/− wild-type and EWSR1-knockout cells transduced with control, EWSR1, or POLQ cDNA in the presence of topotecan (n=3; mean ± SEM).

The impact of EWSR1 loss on the activity of DNA double-strand break repair pathways in U2OS cells was evaluated, using established I-Sce-I-based reporter systems^21,22. EWSR1 knockout resulted in only a minor effect on HR activity (Fig. S1F). Strikingly, EWSR1 knockout significantly decreased MMEJ activity (Fig. 1B). EWSR1 loss reduced MMEJ activity in both I-SceI- and Cas9-based reporter assays and conferred resistance to the POLQ inhibitor, Novobiocin²³(Fig. 1C, S1G, H).

Recent studies suggest that EWSR1 knockdown is synthetic lethal with BRCA2 loss^14,15. To validate these results, we depleted EWSR1 in BRCA-proficient and BRCA-deficient cells. Like POLQ loss, EWSR1 loss decreased colony-formation in BRCA2-deficient DLD1 cells (Fig. 1D), as well as in the BRCA1-mutated breast cancer cell lines, MDA-MB-468 and HCC1937 cells (Fig. S1I, J). Dual-sgRNA competition assays in RPE1 TP53−/− Cas9 cells further confirmed that, like POLQ loss, EWSR1 loss is synthetic lethal with BRCA1/2 loss (Fig. S1K, L). Taken together, EWSR1 functions as an MMEJ gene.

EWSR1 promotes MMEJ by maintaining Polθ expression.

Having established that EWSR1 loss impairs MMEJ, we next examined the impact of EWSR1 loss on the expression of known MMEJ-related proteins. To monitor endogenous Polθ protein expression, we engineered RPE1 TP53−/− cells by knocking a 3xFLAG tag into the endogenous POLQ locus. Interestingly, EWSR1 knockout resulted in decreased Polθ expression (Fig. 1E, Fig. S2A) and increased radiosensitivity (Fig. S2B). To further evaluate the acute effect of EWSR1 loss on Polθ expression, we used dTAG-mediated degradation of endogenously tagged EWSR1 in endogenously 3xFLAG-POLQ-expressing HEK293T cells^24,25 (Fig. S2C–F). Acute EWSR1 depletion rapidly reduced Polθ expression but did not affect other MMEJ-related proteins, including APEX2, HMCES, FEN1, LIG3, PARP1, and XRCC1, nor HR-related proteins such as BRCA1 and BRCA2 (Fig. 1F, Fig. S2G). Consistently, EWSR1 knockout reduced the Polθ foci assembly upon irradiation-induced DNA damage (Fig. 1G). Since EWSR1, as well as FUS and TAF15, belongs to the FET family of proteins²⁶, we speculated that Polθ loss and MMEJ deficiency might be a shared consequence of the loss of FET family genes. However, neither FUS nor TAF15 loss affected the Polθ protein level or MMEJ activity (Fig. S2H, I).

The functional significance of the loss of Polθ on MMEJ activity in EWSR1-depleted cells was next examined. Transduction of the wild-type POLQ cDNA rescued MMEJ activity in EWSR1-knockout cells (Fig 1H). Moreover, POLQ cDNA transduction corrected the hypersensitivity to topoisomerase inhibitors, indicating that loss of Polθ protein underlies this phenotype in EWSR1-knockout cells (Fig 1I, Fig S2J). Like Polθ loss, EWSR1 loss increased unrepaired DNA damage in BRCA-deficient cells (Fig S2K). Collectively, EWSR1 promotes MMEJ via the maintenance of Polθ protein expression.

EWSR1 prevents POLQ exon 25 skipping.

To investigate the cause of reduced Polθ expression following EWSR1 loss, we first evaluated Polθ protein stability. Polθ stability was unaltered by acute EWSR1 depletion, and treatment with the proteasome inhibitor MG132 or the autophagy inhibitor chloroquine did not restore Polθ levels (Fig. S3A, B), suggesting a pre-translational mechanism.

We next performed RNA-seq analysis of wild-type and EWSR1-knockout RPE1 TP53−/− cells. Alternative splicing analysis revealed multiple exon-skipping events in EWSR1-knockout cells (Fig. S3C–G, Supplemental Table 1). Importantly, several POLQ exon-skipping events, including exon 7, exon 21, and exon 25, ranked among the top hits (Fig. 2A). Among these events, exon 25 skipping was the most prominent (Fig. 2B). This finding was further supported by a published EWSR1-knockdown RNA-seq dataset and validated by RT-PCR in multiple EWSR1-knockout clones (Fig. S3H–J). EWSR1 depletion also affected other exon-skipping events, including HYCCI exon 12 and ABI1 exon 10 (Fig. S3K).

Figure 2: Disruption of MMEJ activity in Ewing sarcoma is driven by POLQ exon skipping.

(A) Volcano plots of skipped exon (SE) events in EWSR1-knockout versus wildtype RPE1 TP53−/− cells defined by RNA-seq.

(B) Sashimi plots of the alternative splicing events within POLQ exon24–26 in wild-type and EWSR1-knockout RPE1 TP53−/− cells.

(C) Immunoblotting of V5 and GAPDH in U2OS endogenous V5-tagged Polθ expressing cells transduced with Cas9 and sgRNA targeting indicated POLQ exons.

(D) MMEJ reporter (Cas9-based) analysis in U2OS cells transduced with sgRNA targeting control (NT) or different POLQ exons (n=4). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(E) Immunoblotting of FLAG and GAPDH in HEK293T cells transfected with full-length or exon25-lacking 3xFLAG-POLQ cDNA, treated with cycloheximide (50 μg/mL) for 3–9 hours.

(F) MMEJ reporter analysis in HEK293T POLQ-knockout cells transfected with MMEJ reporter (Cas9-based) and sgRNA targeting the reporter together with control vector or indicated POLQ cDNA (n=3). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(G) Schematic of minigene splicing reporter construct (left). mNeonGreen-positive fraction in K562 reporter cells transduced with minigene splicing reporter, Cas9, and indicated sgRNAs (right, n=3). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(H, I) RT-PCR (H) and minigene reporter assay (I, n=3, mean ± SEM) showing POLQ exon 25 skipping levels across Ewing sarcoma (EwS), Osteosarcoma (OS), and Neuroblastoma (NB) cell lines.

(J, K) Immunoblotting (J) and Cas9-based MMEJ reporter (K, n=2; mean ± SEM) showing lower POLΘ levels and MMEJ activity in EwS cell lines compared to OS and NB cell lines.

(L) Schematic of the strategy for forced inclusion of POLQ exon 25 in A673 cells by CRISPR-Cas9-mediated knock-in.

(M, N) RT-PCR (M) and immunoblotting (N) comparing the POLQ exon 25 skipping and Polθ protein levels between the parental A673 cells and POLQ exon 25-skipping-corrected A673 clones.

(O) MMEJ reporter analysis in parental and POLQ exon 25-skipping-corrected A673 cells (n=4). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

To evaluate the impact of individual POLQ exon-skipping events on Polθ protein expression, we targeted exon 7, exon 21, or exon 25 by CRISPR-Cas9. Targeting each exon induced profound Polθ protein loss and reduced MMEJ activity (Fig. 2C, D). A cycloheximide chase assay further showed that a POLQ cDNA lacking exon 25 produced an unstable shorter Polθ isoform compared with full-length Polθ (Fig. 2E). Skipping of exon 7, exon 21, or exon 25 is predicted to cause premature termination and disrupt the helicase or polymerase domains (Fig. S3L), and POLQ cDNAs lacking exon 25 or exon 7 failed to rescue MMEJ deficiency in POLQ-knockout cells (Fig. 2F). Consistently, cycloheximide treatment increased the proportion of POLQ transcripts with exon skipping, indicating that these aberrantly spliced transcripts are subject to nonsense-mediated decay (Fig. S3M).

To further evaluate the effect of POLQ exon 25 skipping on MMEJ activity, we generated a mNeongreen-based minigene splicing reporter²⁷. In this reporter, mNeongreen coding sequence is interrupted by a cassette consisting of POLQ exon 25 flanked by partial sequences of introns 24 and 25 (Fig 2G). Interestingly, this minigene reporter recapitulated the POLQ exon 25 skipping driven by depletion of EWSR1, suggesting that EWSR1 prevents POLQ exon 25 skipping (Fig. S4E), possibly through direct binding to the POLQ pre-mRNA. Notably, higher levels of POLQ exon 25 skipping were observed in EwS cell lines, compared to the osteosarcoma and neuroblastoma cell lines (Fig. 2H, I). Consistently, Polθ protein and MMEJ activity in EwS cell lines were markedly lower than in osteosarcoma and neuroblastoma cell lines (Fig. 2J, K).

POLQ exon 25 skipping accounts for the MMEJ defect in Ewing sarcoma cells.

Since POLQ exon 25 skipping is a prominent feature of EWSR1-depleted cells and Ewing sarcoma cells, we next evaluated its impact on Polθ protein expression and MMEJ activity in the EwS cell line A673 and the EWSR1-depleted RPE1 TP53−/− cells. The POLQ mRNA has 30 exons, and exons 26–29 are not skipped in EWSR1-depleted cells and Ewing sarcoma cells (Fig. S8K).

To determine whether POLQ exon 25 skipping accounts for MMEJ deficiency, we corrected endogenous POLQ exon 25 inclusion in A673 and EWSR1-depleted RPE1 TP53−/− cells using CRISPR/Cas9-mediated insertion of a POLQ exon 24–30 cDNA cassette (Fig. 2L, S4A–D). This restored Polθ expression in A673 and EWSR1-depleted cells without affecting other exon-skipping events, including exon 7 skipping (Fig. 2M, N, S4E, F). Consistently, correction of POLQ exon 25 skipping restored MMEJ activity in A673 cells (Fig. 2O, S4G), confirming POLQ exon 25 skipping as the primary mechanism underlying reduced Polθ expression and impaired MMEJ.

The Prion-like domain of EWSR1 is required for the faithful splicing of the POLQ transcript.

To identify the functional domain of EWSR1 required for faithful POLQ splicing, we generated EWSR1 deletion mutants (Fig. 3A). Deletion of the prion-like domain (M1) or C-terminal RRG2 domain (M6) failed to rescue POLQ exon 25 skipping or restore Polθ expression (Fig. 3B, C, Fig. S5A). These mutants also failed to restore MMEJ activity, rescue hypersensitivity to topoisomerase or ATR inhibitors, or suppress synthetic lethality with BRCA2 loss in EWSR1-deficient cells (Fig. S5B–G). Because the RRG2 domain contains a nuclear localization signal, its deletion likely destabilizes EWSR1.

Figure 3: Prion-like domain–mediated assembly of the EWSR1 complex with FUBP1 and KHSRP/FUBP2 ensures faithful splicing of the POLQ transcript.

(A) Schematic of EWSR1 domains and deletion mutants.

(B, D) Immunoblotting analysis of FLAG, EWSR1, HA, V5, and GAPDH in RPE1 TP53−/− 3xFLAG-POLQ/2xHA-FKBP12^V36F-EWSR1 double knock-in cells transduced with empty vector, wild-type EWSR1, or the indicated EWSR1 mutants, treated with DMSO or 1μM dTAG^v1 for 24 hours.

(C, E) RT-PCR showing POLQ exon 25 skipping in indicated cells.

(F) MMEJ reporter (Cas9-based) analysis in U2OS cells transduced with sgRNA targeting control (NT) or EWSR1 together with empty vector, wild-type EWSR1, or the indicated EWSR1 mutants (n=4). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(G) Volcano plots comparing enriched proteins by IP-MS between wild-type EWSR1 and the EWSR1 37YS mutant.

(H) A Venn diagram showing the frequency of all the genes and positive hits with a prion-like domain identified by PLAAC.

(I) Immunoprecipitation followed by immunoblotting analysis showing wild-type EWSR1 preferentially bound to FUBP1, KHSRP, SMARCA4 and SMARCC1 compared to the EWSR1 mutants.

(J) FUBP1/KHSRP double-knockout cells exhibited reduced Polθ protein (upper) and increased POLQ exon 25 skipping (lower).

(K) MMEJ activity was profoundly decreased by FUBP1/KHSRP double-knockout in U2OS MMEJ reporter (Cas9-based) cells (n=4). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(L) Schematic of wild-type and YS-mutated EWSR1 or FUBP1.

(M) Immunoblotting showing wild-type FUBP1 preferentially bound to EWSR1 compared to the FUBP1 16YS mutant.

(N) MMEJ reporter (Cas9-based) analysis in U2OS cells transduced with indicated sgRNA(s) together with control or FUBP1 cDNA (n=4). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(O, P) Immunoblotting (O) and RT-PCR (P) in RPE1 TP53−/− 3xFLAG-POLQ/2xHA-FKBP12^V36F-EWSR1 double knock-in cells transduced with Cas9 and sgRNA targeting control (NT) or FUBP1 and KHSRP together with control vector, wild-type FUBP1, or the FUBP1 16YS mutant.

Subdivision of the prion-like domain showed functional redundancy among its regions (Fig. S5H, I). Because this domain promotes EWSR1 condensate formation, we next tested tyrosine-to-serine mutants²⁸. The EWSR1–7YS mutant partially failed to rescue POLQ exon 25 skipping and Polθ expression (Fig. 3D, E, Fig. S5J, K), whereas the condensate-defective EWSR1–37YS mutant abolished the ability of EWSR1 to prevent POLQ exon skipping. Consistently, both mutants failed to rescue MMEJ activity in EWSR1-depleted cells (Fig. 3F). Taken together, the prion-like domain of EWSR1 is required for faithful POLQ splicing and Polθ expression, suggesting that EWSR1 condensate formation is essential for these activities^6,28.

The prion-like domain of EWSR1 is critical for interaction with spliceosomes including FUBP1 and KHSRP.

To identify potential interacting proteins of EWSR1 via its prion-like domain, IP/mass spectrometry was next performed in HEK293T cells containing a biallelic FKBP12-V36F knock-in at the N-terminus of EWSR1 (Fig. 3G). The cDNAs encoding the control V5-tagged wild-type EWSR1-WT, or the EWSR1 37YS mutant, were transduced into the cells and, following dTAG^v1 treatment, cellular proteins were immunoprecipitated with the anti-V5 antibody. A previous study demonstrated that the prion-like domain of EWS-FLI1 interacts with BAF complex subunit components in condensates⁶. Consistent with this, wild-type EWSR1, but not the EWSR1 37YS mutant, preferentially bound to BAF subunits, including ARID1A, SMARCA2, SMARCA4, and SMARCC1 (Fig. 3G, I, Fig. S6A, Supplemental Table 2).

In addition, several RNA-binding proteins (RBPs) were identified as wild-type EWSR1-specific interactors. Intriguingly, some of these interactors also contain prion-like domains, further suggesting that EWSR1 assembles with functionally related RBPs through prion-like domain interactions to form a splicing complex (Fig. 3H). Among these interactors, FUBP1 and its paralog KHSRP/FUBP2 were of particular interest. Both are established splicing regulators with prion-like domains, previously shown to cooperate in controlling SMN2 exon inclusion^29,30, providing a strong rationale for their functional involvement in POLQ splicing. To test this, knocked out FUBP1 and KHSRP individually and in combination (Fig. 3J). Depletion of FUBP1 or KHSRP alone caused a minor reduction in Polθ protein expression or MMEJ activity. However, dual depletion of FUBP1 and KHSRP increased POLQ exon 25 skipping, reduced the Polθ protein level, and impaired MMEJ activity, indicating a functional redundancy between FUBP1 and KHSRP (Fig. 3 J, K, Fig. S6B). A FUBP1–16YS mutant, with sixteen tyrosine residues of its C-terminus region substituted with serine residues, failed to bind to wild-type EWSR1 and rescue the phenotype of FUBP1/KHSRP dual-knockout cells, suggesting that tyrosine residues of EWSR1 and FUBP1 are critical for their protein-protein interaction and for POLQ exon 25 inclusion (Fig. 3 L–P, Fig. S6C). Similarly, a KHSRP-17YS mutant failed to rescue the phenotype of FUBP1/KHSRP dual-knockout cells (Fig. S6D–G). Collectively, the formation of the EWSR1 complex, mediated by their prion-like domains, is critical for the faithful splicing of the POLQ pre-mRNA.

The EWSR1 complex binds to a cis-acting Exonic Splicing Enhancer (ESE) within exon 25 of the POLQ pre-mRNA.

To determine whether POLQ exon 25 of the POLθ pre-mRNA contains a cis-acting regulatory element capable of controlling POLQ exon 25 splicing, we utilized the dCasRx system^31,32 and the POLQ exon 25 minigene splicing reporter. We performed a screen (Fig. 4A) with the catalytically inactive dCasRx and with 96 sgRNAs, thereby tiling the POLQ exon 25 sequence and the surrounding upstream/downstream 1000-bp intron sequences. As we expected, sgRNAs targeting 3’splice site of exon 25 caused an increase in exon 25 skipping, as judged by an increase in mNeonGreen signal (Fig. 4B). In addition to this, a few sgRNAs directed against exon 25 sequences caused exon skipping in the minigene splicing reporter, indicating the presence of an exonic splicing enhancer (ESE) within the POLQ exon 25 pre-mRNA (Fig. 4B, C). These sgRNAs caused exon 25 skipping of the endogenous POLQ pre-mRNA, resulting in Polθ protein loss and MMEJ defect (Fig. 4, D–F).

Figure 4: Identification of an exonic splicing enhancer (ESE) within POLQ exon 25.

(A) Schematic of a tiling screen, using the dCasRx system together with control (NT) sgRNA or each of the 96 sgRNAs targeting POLQ exon 25 and flanking ±1000 bp intron in the cells carrying POLQ exon 25 minigene splicing reporter.

(B, C) Screen result (B, n=3, mean ± SEM) for the screen described in (A), and validation by flow cytometry analysis (n=6, top) and schematic of the sgRNAs targeting specific region POLQ exon 25 used in (C)-(I) (bottom) (Created in BioRender. Asada, S. (2026) https://BioRender.com/t52ig0q). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(D-F) Guiding of dCasRx to cis-regulatory elements increased POLQ exon 25 skipping in dCasRx expressing RPE1 TP53−/− minigene reporter cells, as measured by and RT-PCR (D), decreased Polθ protein levels (E), and decreased MMEJ activity (F). Statistical analysis for (F) was performed using one-way ANOVA (mean ± SEM).

(G) Schematic of the CARPID experiments for (H) and (I).

(H) Streptavidin IP followed by immunoblotting in (H) HEK293T cells transfected with BASU-dCasRx together with sgRNA targeting control (NT) or specific POLQ exon 25 regions, and (I) HEK293T 2xHA-FKBP12^V36F-EWSR1 knock-in cells transfected with control vector, wild-type EWSR1, or the EWSR1 37YS mutant together with BASU-dCasRx and sgRNA #7, treated with 1 μM dTAG^v1 for 24 hours before harvesting. Cells were treated with 200 μM biotin for 15 minutes before harvesting.

(J) A schematic showing that the EWSR1-FUBP1-KHSRP complex promotes POLQ exon 25 inclusion by binding to the ESE, which is disrupted by dCasRx (Created in BioRender. Asada, S. (2026) https://BioRender.com/2pqm35u).

(K) Schematic of the deletion mutants of POLQ exon 25 minigene splicing reporter used in (L) and (M).

(L-N) Minigene splicing reporter analysis in (L) RPE1 TP53−/− cells transduced with wild-type or deletion mutants of POLQ exon 25 minigene splicing reporters (n=5), (M) RPE1 TP53−/− cells expressing wild-type or indicated mutant forms of POLQ exon 25 minigene splicing reporter together with Cas9 and sgRNA targeting control (NT) or EWSR1 (n=4), and (N) RPE1 TP53−/− cells transduced with base-substituted POLQ exon 25 minigene splicing reporter mutants (n=4; mean ± SEM). Statistical analysis was performed using one-way ANOVA (mean ± SEM).

(O) Motif analysis by MEME showing that one of the top enriched EWSR1 peaks motif from two eCLIP-seq studies (GSE98836 and the ENCODE study) is identical or similar to the ESE of the POLQ exon 25.

(P) Schematic of the sgRNA and donor DNA sequence for CRISPR-mediated knock-in to disrupt the ESE of POLQ exon 25.

(Q, R) Increased POLQ exon 25 skipping as measured by RT-PCR (Q) and decreased Polθ protein level (R) in the ESE-disrupted RPE1 TP53−/− 3xFLAG-POLQ cells.

None of the mutagenesis used in this figure leads to a frameshift or generation of a stop codon.

Furthermore, utilizing the CARPID system³³, we confirmed that the EWSR1 complex (EWSR1, FUBP1, and KHSRP) is localized to this ESE of POLQ exon 25 (Fig. 4G, H). Acute depletion of EWSR1 led to delocalization of FUBP1 and KHSRP from the ESE (Fig. S6H). Importantly, the EWSR1 37YS mutant also failed to localize to this ESE (Fig. 4I). Either loss of the EWSR1 protein complex or disruption of this ESE with a sgRNA-directed dCasRx, caused exon 25 skipping (Fig. 4J). To further define the ESE, we generated several six-base-pair deletions within the exon 25 sequence of the POLQ exon 25 minigene splicing reporter (Fig. 4K). Deletions that disrupted the ESE sequence caused an increase in exon 25 skipping (Fig. 4L). EWSR1 depletion did not cause a further increase in exon 25 skipping in the deletion-mutant lacking the ESE, indicating that the EWSR1 complex promotes exon 25 skipping by binding to the ESE (Fig. 4M). Point mutagenesis of this regulatory region identified a six-base-pair ESE (GAGCTG) required for the prevention of POLQ exon 25 skipping (Fig. 4N). Consistently, analysis of the eCLIP datasets (GSE98836 and the ENCODE project) demonstrated that the ESE sequence matches one of the top EWSR1 RNA-binding motifs (Fig. 4O). Finally, disruption of the ESE sequence in the endogenous POLQ exon 25 resulted in increased POLQ exon 25 skipping and Polθ protein loss (Fig. 4P–R). Taken together, exon 25 of the POLQ mRNA contains a cis-acting ESE. The EWSR1 complex (EWSR1-FUBP1-KHSRP complex) is localized to this site where it stabilizes and promotes the inclusion of POLQ exon 25 in the final mature mRNA. Loss of EWSR1 binding to the ESE, or deletion of the ESE, promotes POLQ exon 25 skipping.

The EWS-FLI1 oncoprotein inhibits the formation of Polθ mitotic foci at DSB loci and disrupts MMEJ.

The t(11;22) chromosomal translocation not only eliminates one wild-type EWSR1 allele but also generates the oncogenic fusion protein EWS-FLI1 in EwS cells (Fig. S7A). Moreover, EWS-FLI1 is known to heterodimerize with EWSR1 and to disrupt its activity^6,34,35. We therefore hypothesized that this oncoprotein may directly or indirectly impair MMEJ activity. To test this hypothesis, the EWS-FLI1 oncoprotein was lentivirally expressed in U2OS-based MMEJ reporter cells and evaluated for MMEJ activity. EWS-FLI1 expression diminished the MMEJ activity in U2OS MMEJ reporter cells (Fig. 5A, Fig. S7B). Consistent with these results, EwS cell lines exhibited resistance to Polθ inhibitors, Novobiocin and ART558³⁶, compared with osteosarcoma cells U2OS (Fig. S7C, D). The influence of the EWS-FLI1 prion-like domain mutants, including EWS-FLI1 37YS, SYGQ1-FLI1, SYGQ2-FLI1, ΔSYGQ1, and ΔSYGQ2, on MMEJ activity was further tested (Fig. 5B). Expression of all these mutants, except the EWS-FLI1 37YS mutant decreased the MMEJ activity in the U2OS reporter cells, indicating functional redundancy in the prion-like domain of EWS-FLI1. Interestingly, the EWS-FLI1 37YS mutant, which failed to disrupt MMEJ activity (Fig. 5B, Fig. S7B), has been reported to be defective in condensate formation and in binding to wild-type EWSR1 protein^6,35. Moreover, expression of EWS-FLI1, but not the EWS-FLI1 37YS mutant, conferred hypersensitivity to topoisomerase inhibitors (Fig. S7E, F).

Figure 5: Expression of EWS-FLI1 perturbs POLQ faithful splicing and Polθ foci assembly, thus impairing MMEJ.

(A) MMEJ reporter analysis (I-SceI-based) in U2OS cells transduced with empty vector, wild-type EWS-FLI1, or the FKBP12^V36F-tagged EWS-FLI1, treated with 1 μM dTAGV1 (n=6) (n=6). Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(B) Schematic of the wild-type and mutant EWS-FLI (left). MMEJ reporter (I-SceI-based) analysis in U2OS cells transduced with empty vector, wild-type EWS-FLI1, or indicated mutant forms of EWS-FLI1 (right, n=6). Statistical analysis was performed using one-way ANOVA (mean + SEM).

(C) MMEJ reporter (I-SceI-based) analysis in U2OS, A673, RH1, and RD-ES cell lines transduced with Cas9 together with sgRNA targeting control (NT) or FLI1 (n=3). Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(D, E) RT-PCR (D) and immunoblotting (E) showing decreased POLQ exon 7/25 skipping and restored Polθ protein levels in EWS-FLI1-depleted A673 clones.

(F) Volcano plots showing skipped exon (SE) events in parental vs. EWS-FLI1-depleted A673 cells defined by deep RNA-seq.

(G) Immunoprecipitation followed by immunoblotting analysis showing wild-type EWS-FLI1 preferentially bound to EWSR1 compared to the EWS-FLI1 37YS mutant.

(H) Streptavidin IP followed by immunoblotting in HEK293T cells transfected with control vector or EWS-FLI1 together with BASU-dCasRx and control sgRNA or sgRNA #7. Cells were treated with 200 μM biotin for 15 minutes before harvesting.

(I) ChIP-qPCR assay showing EWS-FLI1 expression perturbed Polθ assembly at the DSB loci induced by I-SceI (n=2). Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(J, K) Immunofluorescence of endogenous V5-Polθ foci 4 hours post 2Gy irradiation in asynchronous and mitotically arrested U2OS cells expressing empty vector or EWS-FLI1 (J), and A673 cells transfected with sgRNA targeting control (NT) or FLI1 (K) (scale bars = 2 μm). Statistical analysis was performed using Mann-Whitney test.

To determine the effect of the EWS-FLI1 oncoprotein on MMEJ activity in EwS cells, we transduced an I-SceI-based MMEJ reporter into the EwS lines, including A673, RH1, and RD-ES. As expected, depletion of EWS-FLI1 from these cells by sgFLI1, as confirmed by western blot, restored MMEJ activity (Fig. 5C, Fig. S7G). These EwS cells, following EWS-FLI1 depletion, exhibited sensitivity to the Polθ inhibitor, ART558 (Fig. S7H), consistent with their restoration of MMEJ following EWS-FLI1 depletion. To further confirm the importance of the EWS-FLI1 oncoprotein on the MMEJ impairment, we generated stable EWS-FLI1-depleted and EWS-FLI1-wild-type A673 clones by CRISPR-mediated editing and subsequent single-cell isolation. EWS-FLI1-depleted A673 clones exhibited lower POLQ exon 7/25 skipping and higher POLQ protein expression compared with EWS-FLI1-wild-type clones (Fig. 5D, E). RNA-seq confirmed that POLQ exon 7/25 skipping is one of the most significant alternative splicing events driven by EWS-FLI1 expression (Fig. 5F, Supplemental Table 3). In addition to POLQ exon 7/25 skipping, additional exon skipping events, including HYCC1 exon 12 and ABI1 exon 10, are commonly induced by EWSR1-depletion and expression of EWS-FLI1 (Fig. S3K, S7I, Supplemental Table 4).

The minigene reporter assay showed that expression of EWS-FLI1 but not the EWS-FLI1 37YS mutant increased the POLQ exon 25 skipping, whereas depletion of EWS-FLI1 in EwS cells reduced the POLQ exon 25 skipping (Fig. S7J, K). Consistent with a previous study⁶, EWS-FLI1 preferentially binds to EWSR1, compared with the EWS-FLI1 37YS mutant (Fig. 5G). The CARPID system revealed that the EWS-FLI1 oncoprotein disrupts the localization of the EWSR1 complex to ESE without occupying the ESE (Fig. 5H), indicating that EWS-FLI1 inhibits EWSR1-mediated POLQ splicing through direct protein–protein interaction with EWSR1, rather than by competing for RNA binding. RNaseH1 expression did not affect EWS-FLI1-mediated MMEJ reduction, suggesting that this effect is independent of R-loop accumulation (Fig. S7L, M).

Recent studies indicate that DSB repair by MMEJ occurs during mitosis³⁷. Polθ is phosphorylated by PLK1 and assembles with TOPBP1, along with RHNO1 and the 9-1-1 complex, in Polθ foci^19,38. To assess the influence of EWS-FLI1 expression on the assembly of Polθ foci at DSB loci, we engineered U2OS MMEJ reporter cells with endogenously V5-tagged-POLQ by CRISPR/Cas9-mediated knock-in. After generation of adenoviral I-SceI-induced DSBs, we performed chromatin-immunoprecipitation (ChIP) with an anti-V5 antibody, followed by qPCR analysis. Overexpression of EWS-FLI1 in these cells disrupted the assembly of Polθ at DSB sites, as evaluated by ChIP (Fig. 5I) and further validated by immunofluorescence (Fig. 5J). Moreover, EWS-FLI1 overexpression increased γH2AX foci, consistent with the loss of MMEJ activity and the resulting accumulation of DSBs in mitosis. EWS-FLI1 overexpression can therefore reduce Polθ-mediated DSB repair in mitosis, at a time when Polθ-mediated MMEJ is highly active^19,38. EWS-FLI1 depletion in endogenously V5-tagged-POLQ expressing A673 cells restored Polθ foci assembly (Fig. 5K). Collectively, the EWS-FLI1 oncoprotein can reduce Polθ mitotic foci assembly, leading to a functional impairment of MMEJ in EwS cells.

Ewing Sarcoma tumor samples exhibit increased POLQ exon 25 skipping and MMEJ deficiency.

To validate potential MMEJ deficiency driven by POLQ exon 25 in tumor samples from EwS patient, we next analyzed whole genome sequencing and RNA-seq data of primary tumors from EwS and non-EwS patients. Since previous studies have revealed that deletions are classified as scars of MMEJ-mediated repairs if they are ≥3 bp in length with ≥2 bp of flanking microhomology^39–42, we first evaluated this established signature in patient-derived EwS tumors and other pediatric tumors including Osteosarcoma, Neuroblastoma, and Rhabdomyosarcoma. Strikingly, the frequency of deletions with microhomology was markedly lower in the EwS samples compared with non-EwS samples (Fig. 6A, S8A). Moreover, the frequency of templated insertion, which is another established evidence of MMEJ-mediated repair^43–46, was significantly lower in the EwS samples compared with non-EwS samples (Fig. 6B, S8B). These data indicate that EwS is an MMEJ-defective tumor. We also evaluated the mutational signature and small indels in patient-derived EwS tumors and other pediatric tumors. Although the level of HRD hallmarks, ID-6 and SBS signature 3, were overall low, which is consistent with a previous study⁴⁷, EwS tumors exhibited no ID-6 and a lower level of SBS signature 3, compared to other pediatric tumors, confirming that EwS is an HR-proficient tumor (Fig. 6C, S8C–J).

Figure 6. Low MMEJ activity and increased POLQ Exon 25 skipping in Ewing Sarcoma patient-derived tumor samples.

(A) Lower frequency of deletions with microhomology in Ewing sarcoma (EwS; n = 30) as compared to non-EwS tumors (n = 131).

(B) Lower frequency of templated insertions in Ewing sarcoma (EwS; n = 30) as compared to non-EwS tumors (non-EwS; n = 132).

(C) Lower proportion of SBS-signature 3 in Ewing sarcoma (EwS; n = 30) as compared to non-EwS tumors (non-EwS; n = 142).

(D) rMATS analysis showing higher rates of POLQ exon 25 skipping events in primary samples from Ewing sarcoma (EwS; n = 44) compared to non-EwS tumors (non-EwS; n = 454).

(E) RNA-seq data from the patient samples showing relatively lower POLQ mRNA expression in Ewing sarcoma (EwS; n = 83) than in non-EwS tumors (non-EwS; n = 749).

All statistical analysis was performed using Mann-Whitney test.

We next analyzed POLQ exon 25 skipping in tumor samples from patients with EwS and non-EwS including Neuroblastoma, Osteosarcoma, Wilms tumors and Rhabdomyosarcoma. EwS patient-derived tumor samples exhibited a higher frequency of POLQ exon 25 skipping than primary tumors from other pediatric solid tumors (Fig. 6D, S8K, L). Importantly, the same tendency was observed for POLQ exon 7 skipping (Fig. S8M), consistent with the RNA-seq results from the EWS-FLI1-depleted A673 cell line with (Fig. 5F). POLQ mRNA levels in EwS tumor samples were comparable to or lower than those in other pediatric solid tumors (Fig. 6E, S8N). Collectively, EwS tumor cells exhibit increased POLQ exon 25 skipping and a resulting defect in MMEJ.

Inhibition of the FA, NHEJ, or HR pathways efficiently kills the MMEJ-deficient Ewing sarcoma cells via synthetic lethality.

Recent studies have shown that targeting the Fanconi Anemia (FA) pathway is synthetic lethal with MMEJ deficiency^48–51. We therefore hypothesized that MMEJ-defective EwS cells would exhibit high dependence on the FA/BRCA signaling pathway. Indeed, depletion of FANCA, FANCI, or FANCL efficiently reduced colony-forming activity in the control A673 cells, whereas these deletions had only a minor impact on the EWS-FLI1-depleted A673 cells (Fig. S9A). Consistently, correction of POLQ exon 25 skipping in the A673 EwS cell line reduced the dependency on the FA/BRCA pathway (Fig. 7A). Moreover, correction of POLQ exon 25 skipping or depletion of the EWS-FLI1 oncoprotein reduced FANCD2 foci assembly upon Mitomycin C (MMC) treatment in EwS cells, whereas FLI1 depletion did not affect the activation of the pathway in U2OS cells (Fig. 7B, S9B–D). Since RBM39 degraders are potent FA pathway inhibitors, by inducing intron retention of the FANCD2 pre-mRNA⁵², we tested the effect of RBM39 degraders, E7820 and Indisulam, on the FA/BRCA pathway in the A673 cells. Consistently, RBM39 degraders decreased FANCA and FANCD2 protein expression in A673 cells (Fig. 7C). Intriguingly, RBM39 degraders killed EwS cells efficiently (Fig. 7D, E), and cell survival was rescued by correction of POLQ exon 25 skipping or depletion of the EWS-FLI1 oncoprotein (Fig. 7F, G, S9E–H). Moreover, the RBM39 degrader, Indisulam efficiently eliminated EwS cells in vivo (Fig. 7H–J). These data indicate that targeting the FA/BRCA pathway with RBM39 degraders is a potential therapeutic strategy for EwS.

Figure 7. POLQ exon 25 skipping confers sensitivity to inhibition of FA or HR in Ewing sarcoma cells.

(A) Colony-forming assay showing that the parental A673 cells exhibited higher dependence on FANCA, FANCI, FANCL, and BRCA2 genes compared with POLQ exon 25-skipping-corrected A673 clones and RPE1 TP53 wild-type cells (n=3). Statistical analysis was performed using two-way ANOVA (mean ± SEM).

(B) Immunofluorescence analysis showing FANCD2 foci formation was decreased by correction of POLQ exon 25 skipping in A673 cells. Cells were treated with 10 ng/ml mitomycin C (MMC) for 24 hours. Shown are the representative images (left, scale bars = 2um) and the number of FANCD2 foci (right). Statistical analysis was performed using one-way ANOVA.

(C) Immunoblotting analysis of the cell lysates from A673 cells treated with DMSO, 250 nM E7820 or 250 nM Indisulam for 72 hours. Treatment with E7820 and Indisulam decreased FANCA and FANCD2 protein expression.

(D, E) CellTiter-Glo assay showing that compared with U2OS, Ewing sarcoma cells exhibited higher sensitivity to RBM39 degraders, E7820 (D) and Indisulam (E).

(F, G) CellTiter-Glo assay showing that correction of POLQ exon 25 skipping in A673 cells renders resistance to RBM39 degraders, E7820 (F) and Indisulam (G).

(H-J) Efficacy of Indisulam in A673 cells-derived xenograft model. NSG mice (n = 10 mice/group) bearing luciferized A673 cells were treated with 25 mg/kg Indisulam intraperitoneally for 2 weeks. Tumor sizes were measured every 2–3 days (H), and mice survival was monitored (I). Tumors were also monitored by bioluminescence imaging (J). Statistical analysis was performed using Student’s t-test (on day19 and day23) for (H) and log-rank test for (I), respectively.

(K) Colony formation assay of RPE1 TP53−/− cells transduced with control vector, wild-type EWS-FLI1 or EWS-FLI1 37YS mutant with or without BRCA2 depletion (n=3, top). Shown are representative images of the colonies (bottom).

(L, M) CellTiter-Glo assay showing that compared with U2OS, Ewing sarcoma cells showed higher sensitivity to CDK12 inhibitors, THZ531 (L) and MFH290 (M).

(N, O) CellTiter-Glo assay showing that correction of POLQ exon 25 skipping in A673 cells renders resistance to CDK12 inhibitors, THZ531 (N) and MFH290 (O).

All graphs for the CellTiter-Glo assays show a representative result from at least three biologically independent experiments with similar results.

Since inhibiting two of three DSB repair pathways is synthetically lethal in cells^53,54, we asked whether inhibition of NHEJ or HR could also provide a therapeutic modality in MMEJ-deficient EwS cells. Previous studies have shown that NHEJ inhibitors are synthetic lethal with MMEJ loss^53,54. EwS cells were shown to be sensitive to DNA-PK inhibitors that target the NHEJ pathway⁵⁵. We confirmed that EwS cells are sensitive to Peposertib, a DNA-PKci inhibitor (Fig. S10A) and have an elevated baseline level of activated pDNA-PKc (Fig. S10B). Consistently, EwS patients-derived primary tumor samples have an elevated baseline level of pDNA-PKc compared with Osteosarcoma patients-derived tumor samples (Fig. S10C). EwS cells also exhibited high phospho-RPA, a replication stress biomarker, and were hypersensitive to the topoisomerase inhibitors, topotecan, and etoposide (Fig. S7A, S10D, E). Since etoposide is used in dose-intensive chemotherapy of EwS, inhibition of NHEJ could increase the therapeutic efficacy of etoposide in EwS cells. Indeed, the DNA-PK inhibitor, Peposertib, enhanced the activity of etoposide in two EwS cell lines (Fig. S10F, G). The EwS cell line A673, grown as a xenograft, was sensitive to Peposertib, in vivo, and the drug enhanced the activity of etoposide (Fig. S10 H–K).

EWS-FLI1- expressing, but not EWS-FLI1-YS mutant-expressing, cells are more dependent on BRCA2, confirming that the HR pathway is also a rational therapeutic target in MMEJ-deficient EwS cells (Fig. 7A, K, S9A). Previous studies have suggested that EwS cells are sensitive to HR inhibitors, perhaps through the mechanism of synthetic lethality. CDK12 is a known activator of HR repair, based on its ability to promote the transcription of long HR genes⁵⁶. Indeed, CDK12 inhibition is known to enhance the killing of EwS cells⁵⁷. Consistent with this report, EwS cell lines were sensitive to the CDK12 inhibitors THZ531 and MFH290 (Fig 7 L, M). Correction of POLQ exon 25 skipping or depletion of the EWS-FLI1 oncoprotein in the EwS cells restores MMEJ activity, resulting in acquisition of CDK12i resistance (Fig 7N, O, Fig. S9I–L).

Correction of POLQ exon 25 skipping in EwS cells rescued the hypersensitivity to RBM39 degraders and CDK12 inhibitors. These observations suggest that POLQ exon 25 skipping is a predictive biomarker for the sensitivity to these drugs, resulting from a cellular defect in MMEJ. Taken together, the sensitivity of the EwS cells to HR inhibitors results mechanistically from synthetic lethality between the HR and MMEJ pathways.

DISCUSSION

Despite their high sensitivity to DNA-damaging agents, EwS tumors have low somatic mutation burden and low ID6 mutational signatures^34,58–63, suggesting impairment of an error-prone repair pathway. Here, we identify EwS as an MMEJ-deficient cancer in which EWS-FLI1 disrupts EWSR1-dependent POLQ splicing, leading to POLQ exon skipping, reduced Polθ expression, and impaired MMEJ.

Splicing of specific mRNAs is regulated by RNA-binding proteins that can assemble in membrane-less condensates, including nuclear speckles⁶⁴. Many RBPs, including EWSR1, contain prion-like domains that promote such assemblies^6,65. Our findings suggest that EWSR1 functions as a selective splicing hub that cooperates with FUBP1 and KHSRP, two related splicing regulators, through prion-like domain–dependent interactions to promote POLQ exon 25 inclusion. The identification of a six-base-pair ESE within POLQ exon 25 further links this RBP complex to a defined cis-regulatory element required for Polθ expression and MMEJ activity.

Given that EWS-FLI1 oncoprotein overexpression induces extensive DNA damage⁶⁶, we reasoned that the EWS-FLI1 fusion oncoprotein could impair MMEJ activity in EwS cells. Beyond its established transcriptional role, EWS-FLI1 therefore acts as a dominant suppressor of EWSR1 splicing function and further impairs Polθ recruitment to mitotic DSBs. These dual effects provide a mechanistic explanation for MMEJ deficiency in EwS.

Collectively, our findings identify EwS as a cancer with a mechanistically defined Polθ-mediated MMEJ defect. This vulnerability creates dependencies on FA, NHEJ, and HR repair pathways, explaining the sensitivity of EwS cells to RBM39 degraders, DNA-PK inhibition, and CDK12 inhibition. POLQ exon 25 skipping may therefore serve as a biomarker for MMEJ deficiency and for therapeutic response to these repair-targeted strategies.

Limitations of the study

This study has several limitations. First, patient-derived tumor analyses are based on modest cohort sizes, limiting statistical power and generalizability. Therefore, larger prospective cohorts will be required to validate the clinical utility of POLQ exon 25 skipping as a bona fide biomarker for an MMEJ defect and for predicting sensitivity to FA, HR, and NHEJ inhibitors. Second, while the EWSR1–FUBP1–KHSRP complex is shown to localize to the POLQ exon 25 ESE to promote its faithful splicing, the structural basis of this RNA-protein interaction and the precise mechanism by which EWS-FLI1 disrupts the localization of the EWSR1 complex to the ESE remain to be determined. Whether condensate formation by the EWSR1 complex at the ESE is required for faithful POLQ splicing also remains a working hypothesis awaiting direct experimental validation.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Alan D. D’Andrea (alan_dandrea@dfci.harvard.edu).

Materials availability

Plasmids, cell lines, and other unique reagents generated in this study are available from the lead contact upon reasonable request, subject to institutional policies and completion of a materials transfer agreement where applicable.

Data and code availability

RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus under accession number GSE270149 and GSE318194. Publicly available datasets used in this study are listed in the key resources table.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
DNA Polymerase theta (POLQ) (F8D8T)	Cell Signaling Technology	Cat# 64708; RRID: N/A
γH2AX / phospho-H2A.X Ser139	Millipore / EMD Millipore	Cat# 05-636; RRID: AB_309864
LIG3 / DNA Ligase III	BD Biosciences	Cat# 611876; RRID: AB_399356
HMCES	Santa Cruz Biotechnology	Cat# sc-514238; RRID: AB_2813859
HMCES	Atlas Antibodies	Cat# HPA044968; RRID: AB_2679160
APEX2	Cell Signaling Technology	Cat# 74728; RRID: N/A
PARP1	Cell Signaling Technology	Cat# 9542; RRID: AB_2160739
CIP2A	Abcam	Cat# ab128179; RRID: N/A
DNA-PKcs	Cell Signaling Technology	Cat# 12311; RRID: AB_2797881
Phospho-DNA-PKcs Ser2056	Cell Signaling Technology	Cat# 68716; RRID: AB_2939025
TOPBP1	Abcam	Cat# ab2402; RRID: AB_303044
XRCC1	Cell Signaling Technology	Cat# 76998; RRID: AB_2936252
FEN1	Cell Signaling Technology	Cat# 82354; RRID: N/A
EWSR1	Cell Signaling Technology	Cat# 11910; RRID: AB_2616007
EWSR1	Santa Cruz Biotechnology	Cat# sc-28327; RRID: AB_675526
EWSR1	Life Technologies	Cat# A300-417A; RRID: AB_420957
BRCA1	Merck Millipore / Millipore	Cat# OP92; RRID: AB_10682944
BRCA2	Cell Signaling Technology	Cat# 10741; RRID: AB_2797730
FLI1	Abcam	Cat# ab133485; RRID: AB_2722650
FUS	Cell Signaling Technology	Cat# 67840; RRID: AB_2941069
TAF15	Cell Signaling Technology	Cat# 28409; RRID: AB_2798957
Vinculin	Cell Signaling Technology	Cat# 4650; RRID: AB_10559207
FLAG / M2	Sigma-Aldrich	F1804 RRID: AB_262044
HA-tag	Cell Signaling Technology	Cat# 3724 RRID: AB_1549585
HA-tag, clone 3F10	Roche	Cat# 11867431001 RRID: AB_390919
V5-tag	Cell Signaling Technology	Cat# 13202 RRID: AB_2687461
V5-tag	Abcam	Cat# ab15828 RRID: AB_443253
GAPDH	Cell Signaling Technology	Cat# 5174 RRID: AB_10622025
Histone H3	Cell Signaling Technology	Cat# 9715 RRID: AB_331563
FUBP1	Abcam	Cat# ab181111; RRID: N/A
KHSRP	Cell Signaling Technology	Cat# 13398 RRID: AB_2798208
SMARCA4 / BRG1	Cell Signaling Technology	Cat# 3508 RRID: AB_2193944
SMARCC1 / BAF155	Cell Signaling Technology	Cat# 11956 RRID: AB_2797776
MYC / c-Myc	Santa Cruz Biotechnology	Cat# sc-40 RRID: AB_627268
FANCA	Cell Signaling Technology	Cat# 14657 RRID: AB_2798558
FANCD2	Novus Biologicals	Cat# NB100-182 RRID: AB_10002867
α-Tubulin (11H10)	Cell Signaling Technology	Cat# 2125; RRID: AB_261964
Mach 3 rabbit HRP	Buocare Medical	Cat# RP531L
Biological samples
Patient primary tumor tissue sections (IHC analysis)	Mayo Clinic	N/A
Chemicals, peptides, and recombinant proteins
Olaparib	Selleck Chemicals	Cat# S1060
Novobiocin Sodium	Selleck Chemicals	Cat# S2492
Etoposide	Selleck Chemicals	Cat# S1225
VE822	Selleck Chemicals	Cat# S7102
Ro-3306	Selleck Chemicals	Cat# S7747
E7820	Selleck Chemicals	Cat# S6628
Indisulam	Selleck Chemicals	Cat# S9742
M3814	Selleck Chemicals	Cat# S8586
THZ0531	Selleck Chemicals	Cat# S6595
ART558	MedChemExpress	Cat# HY-141520
MFH-290	MedChemExpress	Cat# HY-153244
Nocodazole	Sigma-Aldrich	Cat# 1404
dTAGv1	Tocris	Cat# 6914
Biotin	Sigma-Aldrich	Cat# B4501
MG-132	Selleck Chemicals	Cat# S2619
Puromycin dihydrochloride	Sigma-Aldrich	Cat# P8833
Blasticidin S HCl	Thermo Fisher Scientific	Cat# A1113903
G418 Sulfate (Geneticin)	Thermo Fisher Scientific	Cat# 10131035
Crystal violet	Sigma-Aldrich	Cat# C0775
Alt-R S.p. HiFi Cas9 Nuclease V3	Integrated DNA Technologies	Cat# 1081060
Alt-R Cas9 Electroporation Enhancer	Integrated DNA Technologies	Cat# 1075915
Micrococcal Nuclease	Thermo Fisher Scientific	Cat# 88216
Penicillin-Streptomycin (P/S)	Gibco / Thermo Fisher Scientific	Cat# 15140122
Polybrene	Sigma-Aldrich	Cat# TR-1003
RIPA Buffer	Cell Signaling Technology	Cat# 9806
Protease/Phosphatase Inhibitor Cocktail (100X)	Cell Signaling Technology	Cat# 5872
Proteinase K	Thermo Fisher Scientific	Cat# 25530049
SYBR Safe DNA Gel Stain	Thermo Fisher Scientific	Cat#S33102
Mitomycin C	Sigma-Aldrich	Cat# M0503
Bluing Buffer	Dako / Agilent	Cat# CS702
Hematoxylin Solution, Mayer’s	Sigma-Aldrich	Cat# MHS16
Citrate Buffer	Dako	Cat# S1699
Hydrogen Peroxide	Sigma-Aldrich	Cat# H1009
EGTA	Boston BioProducts	Cat# BM-151
DMEM/F12	Gibco / Thermo Fisher	Cat# 10565042
DMEM	Gibco / Thermo Fisher	Cat# 11965118
RPMI 1640 Medium	Gibco / Thermo Fisher	Cat# 11875119
Opti-MEM I Reduced Serum Medium	Gibco / Thermo Fisher	Cat# 31985070
Fetal Bovine Serum (FBS)	Sigma-Aldrich	Cat# F2442
Critical commercial assays
CalPhos Mammalian Transfection Kit	Takara Bio	Cat# 631312
Lipofectamine LTX Reagent with PLUS Reagent	Thermo Fisher Scientific	Cat# 15338100
CellTiter-Glo Luminescent Cell Viability Assay	Promega	Cat# G7570
BCA Protein Assay Kit	Thermo Fisher Scientific	Cat# 23225
SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads)	Cell Signaling Technology	Cat# 9003
RNeasy Mini Kit	QIAGEN	Cat# 74104
RNase-Free DNase Set	QIAGEN	Cat# 79254
SuperScript IV First-Strand cDNA Synthesis System	Thermo Fisher Scientific	Cat# 18091050
CometAssay Kit (or Cometslides)	R&D Systems (Trevigen)	Cat# 4250-050-K
NEBNext Ultra II RNA Library Prep Kit for Illumina	New England BioLabs	Cat# E7770S
Liquid DAB+ Substrate Chromogen System	Dako	Cat# K3468
P3 Primary Cell 4D-Nucleofector® X Kit S	LONZA	Cat# V4XP-3032
SF Cell Line 4D-Nucleofector® X Kit S	LONZA	Cat# V4XC-2032
SE Cell Line 4D-Nucleofector® X Kit S	LONZA	Cat# V4XC-1032
Deposited data
Deep RNA-seq data (wild-type and EWSR1-knockout RPE1 cells)	This paper	GEO: GSE270149
Deep RNA-seq data (wild-type and POLQ exon 25-skipping-corrected A673 cells)	This paper	GEO: GSE318194
EWSR1 eCLIP-seq data	ENCODE / GEO	GEO: GSE98836
Patient tumor transcriptomic and WGS data	St. Jude Cloud	https://www.stjude.cloud/
Uncropped immunoblot, DNA electrophoresis and immunofluorescence data	This paper; Mendeley data	Mendeley Data, doi: 10.17632/85p36x9vnv.1
Experimental models: Cell lines
RPE1 TP53−/−	Lim et al.71	N/A
RPE1 TP53 wildtype	Lim et al.71	N/A
HEK293T	ATCC	Cat# CRL-3216
TC32	Children’s Oncology Group (COG) Cell Line Repository	N/A
TC71	Children’s Oncology Group (COG) Cell Line Repository	N/A
RH-1	DSMZ	Cat# ACC-493
RD-ES	ATCC	Cat# HTB-166
SK-ES1	ATCC	Cat# HTB-86
SK-PN-DW	ATCC	Cat# CRL-2139
U2OS	ATCC	Cat# HTB-96
143B	ATCC	Cat# CRL-8303
MNNG	ATCC	Cat# CRL-1547
KELLY	DSMZ	Cat# ACC-355
A673	ATCC	Cat# CRL-1598
CHLA-20	Children’s Oncology Group (COG) Cell Line Repository	N/A
SK-N-AS	ATCC	Cat# CRL-2137
SK-N-FI	ATCC	Cat# CRL-2142
HCC1937	ATCC	Cat# CRL-2336
HCC1937+BRCA1	Gift from Dr. Ralph Scully	N/A
K562	ATCC	Cat# CCL-243
MDA-MB-436	Gift from Dr. Geoffrey Shapiro	N/A
MDA-MB-436 + BRCA1	Gift from Dr. Geoffrey Shapiro	N/A
DLD1 Parental	Horizon Discovery	N/A
DLD1 BRCA2−/−	Horizon Discovery	Cat# HD 105-007
U2OS DR-GFP	Gift from Dr. Jeremy Stark	N/A
U2OS-MMEJ reporter (ver.1 / ver.2)	This paper	N/A
RPE1 TP53−/− EWSR1-knockout clones	This paper; generated from RPE1 TP53−/− cells	N/A
HEK293T POLQ-knockout cells	This paper; generated from HEK293T cells	N/A
A673 EWS-FLI1-depleted clones	This paper; generated from A673 cells	N/A
A673 EWS-FLI1-wild-type clones	This paper; isolated during genotyping of A673 EWS-FLI1-edited clones	N/A
HEK293T 3xFLAG-POLQ knock-in cells	This paper; generated from HEK293T cells	N/A
HEK293T 3xFLAG-POLQ/EWSR1-FKBP12V36F double knock-in cells	This paper; generated from HEK293T 3xFLAG-POLQ knock-in cells	N/A
RPE1 TP53−/− 3xFLAG-POLQ knock-in cells	This paper; generated from RPE1 TP53−/− 3xFLAG-POLQ knock-in cells	N/A
RPE1 TP53−/− 3xFLAG-POLQ/EWSR1-FKBP12V36F double knock-in cells	This paper; generated from RPE1 TP53−/− cells	N/A
U2OS V5-POLQ knock-in cells	This paper; generated from U2OS cells	N/A
A673 V5-POLQ knock-in cells	This paper; generated from A673 cells	N/A
A673 POLQ exon 25-skipping-corrected cells	This paper; generated from A673 cells	N/A
RPE1 TP53−/− POLQ exon 25-skipping-corrected cells	This paper; generated from RPE1 TP53−/− cells	N/A
RPE1 TP53−/− 3xFLAG-POLQ POLQ exon 25-ESEdisrupted cells	This paper; generated from RPE1 TP53−/− 3xFLAG-POLQ cells	N/A
Experimental models: Organisms/strains
NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ mice	The Jackson Laboratory	Cat# 005557
Oligonucleotides
sgRNAs used for gene knock out and knock-in, see Table S5	Integrated DNA Technologies	N/A
sgRNAs cloned into lentiviral vectors, see Table S5	This paper	N/A
sgRNAs used for dCasRx-based POLQ exon 25 tiling screen, see Table S5	This paper	N/A
Primers used for RT-PCR analysis of alternative splicing, see Table S5	This paper	N/A
Primers used for ChIP-qPCR, see Table S5	This paper	N/A
Recombinant DNA
psPAX2	Gift from Didier Trono; Addgene	Addgene plasmid #12260
pMD2.G	Gift from Didier Trono; Addgene	Addgene plasmid #12259
pLV-EF1a-IRES-Blast	Gift from Tobias Meyer; Addgene	Addgene plasmid #85133
pCDH-puro-EWS-FLI1	Gift from Jialiang Wang; Addgene	Addgene plasmid #102813
pLenti CMV Puro LUC (w168-1)	Gift from Eric Campeau and Paul Kaufman; Addgene	Addgene plasmid #17477
pCVL.SFFV.d14GFP.EF1a.HA.NLS.Sce(opt).T2A.TagBFP	Gift from Andrew Scharenberg; Addgene	Addgene plasmid #32627
lentiCRISPR v2	Gift from Feng Zhang; Addgene	Addgene plasmid #52961
lentiCRISPR v2-Blast	Gift from Mohan Babu; Addgene	Addgene plasmid #83480
pLKO5.sgRNA.EFS.GFP	Gift from Benjamin Ebert; Addgene	Addgene plasmid #57822
pLKO5.sgRNA.EFS.tRFP657	Gift from Benjamin Ebert; Addgene	Addgene plasmid #57824
pXR002: EF1a-dCasRx-2A-EGFP	Gift from Patrick Hsu; Addgene	Addgene plasmid #109050
pLentiRNAGuide_001 - hU6-RfxCas13d-DR1-BsmBI-EFS-Puro-WPRE	Gift from Neville Sanjana; Addgene	Addgene plasmid #138150
CARPID BASU-dCasRx	Gift from Jian Yan and Liang Zhang; Addgene	Addgene plasmid #153209
pCDH-puro-EWS-FLI1 37YS	This paper	N/A
pMYs-IRES-Blasticidin-EWSR1 WT	This paper	N/A
pMYs-IRES-Blasticidin-EWSR1 mutants	This paper	N/A
pMYs-IRES-Blasticidin-FUBP1 WT	This paper	N/A
pMYs-IRES-Blasticidin-FUBP1 mutants	This paper	N/A
pMYs-IRES-Blasticidin-KHSRP WT	This paper	N/A
pMYs-IRES-Blasticidin-KHSRP mutants	This paper	N/A
pMYs-IRES-Blasticidin-RNASEH1 WT	This paper	N/A
pMYs-IRES-Blasticidin-RNASEH1 mutants	This paper	N/A
lentiCRISPR v2-sgRNA constructs	This paper	sgRNA sequences are listed in Table S5
lentiCRISPR v2-Blast-sgRNA constructs	This paper	sgRNA sequences are listed in Table S5
pLKO5.sgRNA.EFS.GFP-sgRNA constructs	This paper	sgRNA sequences are listed in Table S5
pLKO5.sgRNA.EFS.tRFP657-sgRNA constructs	This paper	sgRNA sequences are listed in Table S5
dCasRx-BSD	This paper; derived from pXR002: EF1a-dCasRx-2A-EGFP	N/A
pLentiRNAGuide_001-sgRNA constructs for dCasRx/BASU-dCasRx	This paper; derived from Addgene plasmid #138150	sgRNA sequences are listed in Table S5
lentiCRISPR v2-BFP	This paper; derived from lentiCRISPR v2	N/A
POLQ exon 25 minigene splicing reporter	This paper; generated in pLV-EF1a-IRES-Neo backbone	N/A
Software and algorithms
ImageJ	NIH	https://imagej.net/
CellProfiler (v4.2.6)	CellProfiler project	https://cellprofiler.org/
CometScore 2.0	Rex Hoover	http://rexhoover.com/index.php?id=cometscore
FlowJo (v10.9.0)	Becton, Dickinson and Company	https://www.flowjo.com/
Sequest	Thermo Fisher Scientific	https://proteomicsresource.washington.edu/protocols06/sequest.php
Perseus (v2.0.11.0)	Max Planck Institute	https://maxquant.net/perseus/
limma (v3.54.2)	Bioconductor	Ritchie et al.75
ggplot2 / ggrepel R packages	CRAN	https://cran.r-project.org/
STAR aligner (v2.7.10b)	GitHub	https://github.com/alexdobin/STARs
rMATS (v4.2.0) / rMATS-turbo	GitHub	https://rnaseq-mats.sourceforge.io/
rmats2sashimiplot	GitHub	https://github.com/Xinglab/rmats2sashimiplot
Cutadapt (v2.5)	Martin	https://cutadapt.readthedocs.io/
Samtools (v1.9)	GitHub / SourceForge	https://samtools.github.io/
Pureclip (v1.3.1)	GitHub	https://github.com/skrakau/PureCLIP
Bedtools (v2.31.1 / v2.27.1)	GitHub	https://bedtools.readthedocs.io/
MEME suite (v5.5.5)	MEME Suite	https://meme-suite.org/meme/
AlphaFold3 / AlphaFold Server	Google DeepMind	https://alphafoldserver.com/
PyMOL (v3.0.4)	Schrödinger, LLC	https://pymol.org/
QuPath (v0.5.1)	University of Edinburgh	https://qupath.github.io/
GraphPad Prism (v10.2.2)	GraphPad Software	https://www.graphpad.com/
GATK (v4.2.4) (Joint Genotyper / Mutect2)	Broad Institute	https://gatk.broadinstitute.org/
bcftools (v1.14)	GitHub	http://samtools.github.io/bcftools/
SigProfilerMatrixGenerator (v1.2.13)	GitHub	https://github.com/AlexandrovLab/SigProfilerMatrixGenerator
SigProfilerAssignment (v0.0.21)	GitHub	https://github.com/AlexandrovLab/SigProfilerAssignment
Other
NuPAGE 4–12% Bis-Tris Gels	Thermo Fisher Scientific	Cat# NP0336BOX
NuPAGE 3–8% Tris-Acetate Gels	Thermo Fisher Scientific	Cat# EA0378BOX
NuPAGE MES SDS Running Buffer (20X)	Thermo Fisher Scientific	Cat# NP0002
NuPAGE Tris-Acetate SDS Running Buffer (20X)	Thermo Fisher Scientific	Cat# LA0041
NuPAGE Transfer Buffer (20X)	Thermo Fisher Scientific	Cat# NP00061
Fish Serum Blocking Buffer	Thermo Fisher Scientific	Cat# 37527
Dynabeads MyOne Streptavidin T1	Thermo Fisher Scientific	Cat# 65601
Dynabeads Protein G for Immunoprecipitation	Thermo Fisher Scientific	Cat# 10004D
Permount Mounting Medium	Fisher Scientific	Cat# SP15-100
Slide Brite Xylene Substitute	Parhelia / PolySciences	Cat# 23401-01
4D-Nucleofector System	Lonza	N/A
RS-2000 X-ray Irradiator	Rad Source Technologies	N/A
BD FACSAria III Cell Sorter	BD Biosciences	N/A
CytoFLEX Flow Cytometer	Beckman Coulter	N/A
Amersham Imager 600	GE Healthcare / Cytiva	N/A
ChemiDoc MP Imaging System	Bio-Rad	N/A
Motic EasyScan Pro 6	Motic	N/A
Parhelia Autostainer	Parhelia Biosciences	N/A
ZEISS Axio Observer	ZEISS	N/A

This paper does not report original code. All software and packages used for computational analyses, including sequencing alignment, differential splicing analysis, and motif analysis, are publicly available and are listed in the key resources table and STAR Methods. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines and culture conditions

RPE1 TP53−/− cells generated previously⁶⁷ and RPE1 TP53 wildtype cells were grown in DMEM/F12 containing GlutaMAX (Gibco) and supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich). HEK293T (ATCC, CRL-3216) cells were grown in DMEM (Gibco) supplemented with 10% FBS. A673 (ATCC, CRL-1598) were grown in DMEM (Gibco) supplemented with 10% FBS. TC32 and TC71 cell lines were obtained from children cancer repository and grown in RPMI (Gibco) supplemented with 10%FBS. RH-1 (DSMZ, ACC493) cells were grown in RPMI 1640 supplemented with 10%FBS, 1% P/S. RD-ES (ATCC, HTB-166) cells were grown in RPMI 1640 supplemented with 20%FBS, 1% P/S. SK-ES1 (ATCC, HTB-86) and SK-PN-DW (ATCC, CRL-2139) cells were cultured in DMEM (Gibco) supplemented with 10%FBS. Osteosarcoma cell lines, U2OS parental (ATCC, HTB-96) and carrying a DNA repair template, 143B (ATCC, CRL-8303), and MNNG (ATCC, CRL-1547) cells were grown in DMEM (Gibco) supplemented with 10%FBS. Neuroblastoma cell lines, KELLY (DSMZ, ACC-355), CHLA-20 (children cancer repository), SK-N-FI (ATCC, CRL-2142), SK-N-AS (ATCC, CRL-2137) were grown in RPMI (Gibco) supplemented with 10%FBS. The BRCA1-deficient cell lines HCC1937 (ATCC, CRL-2336) and BRCA1-complemented cells were grown in RPMI (Gibco) supplemented with 10%FBS. RH-1 (DSMZ, ACC493). K562 (ATCC, CCL-243) cells were grown in RPMI 1640 supplemented with 20%FBS, 1% P/S. MDA-MB-436 and MDA-MB-436+BRCA1 cells were a kind gift from Dr. Geoffrey Shapiro (Dana-Farber Cancer Institute) and grown in RPMI (GIBCO) supplemented with 10% fetal bovine serum. DLD1 BRCA2 −/− and DLD1 Parental were obtained from Horizon Discovery, Waterbeach, UK. U2OS cells with HR reporter (DR-GFP) were a kind gift from Dr. Jeremy Stark (City of Hope).

To establish the U2OS-MMEJ reporter cells, U2OS cells were lentivirally transduced with the I-SceI-based (ver.1) or Cas9-based (ver.2) MMEJ reporter and then incubated with puromycin. After selection with puromycin, a single clone was isolated and subsequently the clones were validated. All cells were cultured in an incubator maintained at 37°C, 5% CO₂ and a relative humidity of 95%.

Mouse xenograft studies

All animal experiments were conducted in accordance with Institutional Animal Care and Use Committee-approved protocol (#08–036) at Dana-Farber Cancer Institute. A673 and TC32 cells were lentivirally transduced with luciferase by using a pLenti CMV Puro LUC (w168–1) vector. One million A673 or two million TC32 cells expressing luciferase were subcutaneously injected into 7-week-old female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice. After confirming tumor engraftment, mice injected with A673 cells were treated with indisulam (25 mg/kg) or vehicle control once a day control via intraperitoneal injection 5 days a week for 2 weeks. Mice injected with TC32 cells were treated with peposertib (100 mg/kg) or vehicle control once a day via oral administration together with etoposide (8mg/kg) or vehicle control via intraperitoneal injection twice a week for 2 weeks. Tumors were measured every 2 to 3 days using an electronic caliper, and tumor volumes were calculated by using the formula L×W×W/2. Mice with a tumor of more than 20 mm in length or width were euthanized.

Human tumor datasets and samples

Patient tumor sample bulk transcriptomic sequencing data were obtained from St. Jude’s Cloud (https://www.stjude.cloud/)⁶⁸. For expression analysis, we used the mapped expression counts data; for the exon splicing data, we used the BAM files provided. For the expression counts data, all of the samples were uniformly processed and mapped to the same reference genome as part of the St. Jude’s Cloud resource, and as such, no correction between samples or cancer types was performed. To evaluate the POLQ expression, we first computed the TPM (transcript-per-million) counts for all samples. In this computation, we omitted genes annotated as pseudogenes in GENCODE v.31 and used the median gene length as the gene length for each gene symbol.

Mutational analysis was performed on the whole genome sequencing (WGS) data (HG38 gVCF and BAM file) from matched tumor-normal pairs of Ewing sarcoma (ES), Neuroblastoma (NB), Osteosarcoma (OS), and Rhabdomyosarcoma (RMS), obtained from St. Jude Cloud (https://www.stjude.cloud/). Sample identification and cross-contamination detection were performed as follows Short variants were called from the gVCFs data using GATK Joint Genotyper (version 4.2.4; https://www.biorxiv.org/content/10.1101/201178v3) and filtered for high-quality variants using bcftools (version 1.14) with these hard filters: “QD >= 7.5 && MQ >= 59.9 && MQ <= 60”.

Primary patient samples for IHC analysis were obtained from the Mayo Clinic (IRB approval #24–007702).

METHOD DETAILS

Compounds and antibodies

Chemical compounds, including Olaparib (#S1060), Novobiocin (#S2492), Etoposide (#S1225), VE822 (#S7102), Ro-3306 (#S7747), E7820 (#S6628), Indisulam (#S9742), M3814 (#S8586) and THZ0531 (#S6595) were purchased from Selleckchem. ART558 (#HY-141520) and MFH-290 (#HY-153244) were purchased from MedChemExpress. Nocodazole (#1404) was purchased from Sigma Aldrich. Chemicals were dissolved in DMSO and kept in small aliquots at −20°C or −80°C. The key sources of antibodies are: Polθ (Cell Signaling Technology, #64708), γH2AX (Millipore #05–636), LIG3 (BD Biosciences, #611876), HMCES (Santa Cruz Biotechnology, SC-514238), HMCES (Atlas Antibodies #HPA044968), APEX2 (Cell Signaling Technology, #74728), PARP1 (Cell Signaling Technology, #9542), CIP2A (Abcam, ab128179), DNA-PK (Cell Signaling Technology, #12311), Phospho-DNA-PKcs (Ser2056) (Cell Signaling Technology, #68716), TOPBP1 (Abcam, ab2402), XRCC1 (Cell Signaling Technology, #76998), FEN1 (Cell Signaling Technology, #82354), EWSR1 (Cell Signaling Technology, #11910 for WB), EWSR1 (Santa Cruz Biotechnology, sc-28327 for WB), EWSR1 (Life Technologies, A300417A for IP and RIP), BRCA1 (Merck Millipore, OP92), BRCA2 (Cell Signaling Technology, #10741), FLI1 (Abcam, ab133485), FUS (Cell Signaling Technology, #67840), TAF15 (Cell Signaling Technology, #28409), Vinculin (Cell Signaling Technology, #4650), FLAG (Sigma, F1804), HA (Cell Signaling Technology, #3724 for WB), HA (Roche, #11867431001 for IP and RIP), V5 (Cell Signaling Technology, #13202 for WB and IF), V5 (Abcam, ab15828 for ChIP), V5 (Abcam, ab27671 for IP and WB), GAPDH (Cell Signaling Technology, #5174), Histone H3 (Cell Signaling Technology, #9715), FUBP1 (Abcam ab181111), KHSRP (Cell Signaling Technology, #13398), SMARCA4/BRG1 (Cell Signaling Technology, #3508), SMARCC1 (Cell Signaling Technology, #11956), MYC (Santa Cruz Biotechnology, SC-40), FANCA (Cell Signaling Technology, #14657), FANCD2 (Novus Biologicals, NB100–182).

Generation of CRISPR-engineered cell lines

For generating EWSR1-knockout clones RPE1 TP53−/− cells, POLQ-knock out clones HEK293T cells and EWS-FLI1-depleted clones in A673 cells, a ribonucleoprotein (RNP) complex was formed by Alt-R^™ S.p. HiFi Cas9 and CRISPR-Cas9 sgRNA targeting EWSR1, POLQ, or FLI1 (Integrated DNA Technologies), respectively. For generating endogenously 3xFLAG-tagged POLQ knock-in HEK293T and RPE1 TP53−/− 3xFLAG-tagged POLQ knock-in (RPE1 TP53−/−^3xFLAG-POLQ) cells, EWSR1-FKBP12^V36F knock-in RPE1 TP53−/− and HEK293T cells, V5-tagged POLQ knock-in U2OS and A673 cells, or POLQ exon 25-skipping-corrected A673 and RPE1 TP53−/− cells, or POLQ exon 25-ESE-disrupted RPE1 TP53−/−^3xFLAG-POLQ cells, the Cas9/sgRNA RNP complex together with a template was delivered. Cells were resuspended in SE (U2OS cells), SF (HEK293T and A673 cells) or P3 (RPE1 TP53−/− cells) Nucleofector Solution with supplement (Lonza) and then mixed with the Cas9/sgRNA RNP complex and Alt-R^™ Cas9 Electroporation Enhancer (Integrated DNA Technologies), respectively. The Cas9/sgRNA RNP complex was delivered to the cells by 4D-Nucleofector^™ (Lonza). The media was changed the following day, and then 2 days after cells were seeded in a 96-well plate to isolate single clones. For generating POLQ exon 25-skipping-corrected cells, mScarlet positive cells were sorted by BD Aria III Cell Sorter (BD biosciences) before single cell isolation. Individual clones were tested for genomic editing analyses using immunoblotting and genomic PCR with subsequent Sanger sequencing. A673 EWS-FLI1-wildtype clones were also obtained during the genotyping together with A673 EWS-FLI1-depleted clones. All the sgRNA sequences used for CRISPR-mediated editing were provided in Supplemental Table 5.

Plasmids

psPAX2 and pMD2.G (Addgene #12260 and #12259) were gifts from Dr. Didier Trono. pLV-EF1a-IRES-Blast (pLV-Blast) (Addgene #85133) was a gift from Dr. Tobias Meyer.⁶⁹ pCDH-puro-EWS-FLI1 was a gift from Jialiang Wang (Addgene plasmid # 102813). pLenti CMV Puro LUC (w168–1) was a gift from Eric Campeau & Paul Kaufman (Addgene plasmid # 17477). pCVL.SFFV.d14GFP.EF1a.HA.NLS.Sce(opt).T2A.TagBFP was a gift from Andrew Scharenberg (Addgene plasmid # 32627). lentiCRISPR v2 was a gift from Feng Zhang (Addgene plasmid # 52961). lentiCRISPR v2-Blast was a gift from Mohan Babu (Addgene plasmid # 83480). pLKO5.sgRNA.EFS.GFP and pLKO5.sgRNA.EFS.tRFP657 were gifts from Benjamin Ebert (Addgene plasmid # 57822 and # 57824, respectively). pXR002: EF1a-dCasRx-2A-EGFP was a gift from Patrick Hsu (Addgene plasmid # 109050). pLentiRNAGuide_001 - hU6-RfxCas13d-DR1-BsmBI-EFS-Puro-WPRE was a gift from Neville Sanjana (Addgene plasmid # 138150). CARPID BASU-dCasRx was a gift from Jian Yan & Liang Zhang (Addgene plasmid # 153209).

EWS-FLI1 37YS mutant was cloned into pCDH-puro vector. EWSR1, FUBP1, KHSRP, RNASEH1, and their mutant forms were cloned into the pMYs-IRES-Blastcidin vector, respectively. sgRNAs for gene depletion were cloned into a lentiCRISPR-v2, lentiCRISPR v2-Blast, pLKO5.sgRNA.EFS.GFP, or pLKO5.sgRNA.EFS.tRFP657 construct as previously described⁷⁰. Of note, sgEWSR1 #2, sgFUBP1 #2 or sgKHSRP #2 target genomic DNA but not cDNA, so the simultaneous endogenous knockout and ectopic cDNA expression are feasible. dCasRx-BSD plasmid was generated by replacing EGFP with a Blasticidin resistance gene of pXR002: EF1a-dCasRx-2A-EGFP. sgRNAs for guiding dCasRx and BASU-dCasRx were cloned into a pLentiRNAGuide_001 - hU6-RfxCas13d-DR1-BsmBI-EFS-Puro-WPRE construct. lentiCRISPR v2-BFP vector was generated by replacing a puromycin resistance gene with a BFP of lentiCRISPR v2. All the sgRNA sequences used in this study were provided in Supplemental Table 5.

Virus generation

Lentivirus was generated by co-transfecting lentiviral constructs with packaging vectors, psPAX2 and pMD2.G, into HEK293T cells using CalPhos Mammalian Transfection Kit (Takara Bio) or using Lipofectamine LTX with Plus reagent (Thermo Fisher Scientific). Two days after transfection, supernatant containing virus was harvested and filtered through 0.45 μm filter. Viral transductions were performed in the presence of 8 μg/ml (for RPE1 TP53−/−, MDA-MB 436, DLD1, HCC1937, and HCC1954 cells) or 4 μg/ml (for other cell lines) polybrene (Sigma-Aldrich). For all virus generations and transductions media supplemented with heat-inactivated FBS were used. Retrovirus was generated by co-transfecting retroviral constructs with packaging vectors into HEK293T cells using CalPhos Mammalian Transfection Kit (Takara Bio). Two days after transfection, supernatant containing virus was harvested and filtered through 0.45 μm filter. Viral transductions were performed in the presence of 8 μg/mL polybrene (Sigma-Aldrich). Puromycin dihydrochloride (Sigma-Aldrich) or Blasticidin S HCl (Thermo Fisher Scientific) was used after infection wtih retrovirus or lentivirus for selection.

Colony formation assay

Cells were seeded in 6-well plates at 500–10,000 cells/well. The day after seeding, cells were exposed to each drug or x-ray. X-ray irradiation was performed using the RS-2000 Irradiator (Rad Source Technologies). After incubation, cells were washed with PBS and then fixed in a mixture of 50% methanol, 10% acetic acid, and 40% water for 20 minutes. The fixed cells were stained with 0.5% crystal violet (Sigma-Aldrich) in 20% methanol for 2–6 hours and then washed with water twice. The plates were imaged on GE Amersham Imager 600 using colorimetric transillumination setting. The area fraction of the colonies in each well was estimated using ImageJ. All graphs for the colony forming assays shown are mean ± SEM.

CellTiter-Glo assay

To test drug sensitivity, U2OS or Ewing sarcoma cells were seeded in 96-well plates at 1000–3000 cells/well. The day after seeding, the drug was added to the wells, and then the plates were incubated for 3–5 days. After incubation, ATP in each well was quantified using CellTiter-Glo Luminescent Cell Viability Assay (Promega). CellTiter-Glo assays were performed in technical triplicate if not otherwise specified in a figure legend. All graphs for the CTG assays shown are mean ± SEM.

Immunofluorescence

Mitotic synchronization was achieved using Ro-3306 (8 μM, 16 hours) followed by 30 minutes release in fresh medium. Then cells were incubated with 100 ng/ml nocodazole and subsequently irradiated. 4 hours after irradiation or in the absence of treatment, cells were fixed with 4% paraformaldehyde for 15 minutes at room temperature. For FANCD2 foci evaluation, cells were treated with 10 ng/ml MMC for 24 hours. Cells were then permeabilized with 0.3% Triton X-100 for 10 minutes on ice, followed by blocking with 3% non-fat milk for 1 hour at room temperature. The slides were stained with primary antibodies at 4 °C overnight. Afterward, they were stained with secondary antibodies for 1 hour at room temperature. The slides were scanned using fluorescence microscope. At least 100 cells were counted for each sample. Foci quantification was performed using software CellProfiler version 4.2.6.

Immunoblotting

Whole cell lysate was prepared by lysing cells in ice-cold RIPA buffer (Cell Signaling Technology) supplemented with Protease/Phosphatase inhibitor Cocktail (Cell Signaling Technology). Cell lysate was cleared by centrifuging the samples at 12000 rpm, 4 °C for 20 minutes. Supernatant was recovered and protein concentration was estimated using the BCA assay kit (Thermo Fisher Scientific). Protein samples were boiled in 2X Laemmli sample buffer (Bio-Rad) supplemented with 5% 2-mercaptoethanol or NuPAGE LDS sample buffer (Thermo Fisher Scientific) supplemented with NuPAGE sample reducing agent (Thermo Fisher Scientific). Samples were run in NuPAGE 4–12% Bis-Tris gels (Thermo Fisher Scientific) using NuPAGE MES SDS Running Buffer (Thermo Fisher Scientific) or NuPAGE 3–8% Tris-Acetate Gel (Thermo Fisher Scientific) using NuPAGE Tris-Acetate SDS Running Buffer (Thermo Fisher Scientific). After electrophoresis, samples were transferred onto a nitrocellulose membrane (Amersham). Membranes were blocked for 1 hour with Blocking Buffer, which was prepared by diluting Fish Serum Blocking Buffer (Thermo Fisher Scientific) with Tris-buffered saline containing 0.1% Tween20 (TBST). Membranes were incubated with primary antibodies diluted in Blocking Buffer overnight at 4 °C. Membranes were washed with TBST three times, 5 minutes each, incubated with secondary antibodies diluted in blocking buffer for 1hour, which were either horseradish peroxidase (HRP)-conjugated for chemiluminescent detection or fluorophore-conjugated for visualization using the LI-COR imaging system. After secondary antibody incubation, membranes were washed with TBST and imaged using chemiluminescence settings on GE Amersham Imager 600 or the LI-COR imaging system.

DNA repair template assay

DR-GFP cells were kind gifts from Dr. Jeremy Stark. MMEJ-GFP cells (ver.1 and ver.2) were generated by lentivirally transducing a cassette into Ewing sarcoma, Osteosarcoma, or Neuroblastoma cells using a pLV-EF1a-IRES-puro vector (addgene #85132) as previously described²¹. U2OS cells carrying a DNA repair template reporter (DR-GFP and MMEJ-GFP) or Ewing sarcoma cells carrying an MMEJ reporter were lentivirally transduced with Cas9 and sgRNA (co-expressing puromycin resistance gene). After puromycin or blasticidin selection, 40,000 DNA repair template reporter cells were seeded in 12-well plates and adenovirally or lentivirally (co-expressing BFP) transduced with I-SceI the following day. 48h after I-SceI transduction, GFP signals were analyzed by CytoFLEX (Beckman). The signals were normalized to control cells. For the Cas9-based MMEJ reporter assay, the cells were subsequently transduced with Cas9 and sgRNA targeting the reporter (co-expressing BFP). 72 hours after Cas9/sgRNA transduction, the proportion of GFP-positive fraction within BFP-positive fraction were analyzed by CytoFLEX (Beckman). For the Cas9-based MMEJ reporter assay using wild-type or POLQ-knockout HEK293T cells, the MMEJ reporter, Cas9, and sgRNA targeting the reporter (co-expressing BFP) together with control vector, POLQ WT cDNA, or POLQ mutant cDNA were transduced using Lipofectamine LTX with Plus reagent (Thermo Fisher Scientific). After 72-hour incubation, the proportion of GFP-positive fraction within BFP-positive fraction were analyzed by CytoFLEX (Beckman).

Comet assay

The alkaline comet assays were performed to detect both single and double-stranded DNA breaks. 1000 cells, suspended in PBS, were mixed with 50 μl melted low-melting agarose and pipetted onto the Cometslides (R&D Systems). Once the agarose solidified, the slides were immersed in a lysis solution for 18 hours to facilitate cell lysis. Following lysis, the slides were incubated in an alkaline unwinding solution for 1 hour to denature the DNA. Electrophoresis was performed at 21 V for 45 minutes in an alkaline electrophoresis solution. The slides were stained with SYBR green solution, scanned by fluorescence microscope, and analyzed using CometScore 2.0 software.

Chromatin immunoprecipitation (ChIP) assay

U2OS biallelic endogenous V5-POLQ expressing cells carrying an I-SceI based MMEJ reporter were adenovirally transduced with I-SceI in the presence or absence of 100 ng/ml nocodazole. 12 hours after I-SceI transduction, chromatin immunoprecipitation was performed using SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (Cell Signaling Technology). Protein-DNA cross-linking was performed with 1% formaldehyde and glycine solution. Chromatin was fragmented with Micrococcal Nuclease. The digested chromatin was incubated at 4 °C with each antibody for 4 hours. After the antibody reaction, ChIP-Grade Protein G Magnetic Beads were added to the samples, and the samples were incubated at 4 °C for 1–2 hours. For crosslinking reversal, each sample was incubated at 65 °C with Proteinase K overnight. After DNA purification, immunoprecipitated DNA was analyzed by qPCR using a primer set targeting the integrated MMEJ reporter locus. ChIP-qPCR was performed in technical triplicate. All the primer sequences for ChIP-qPCR used in this study were provided in Supplemental Table 5.

Reverse transcription PCR (RT-PCR)

Total RNA was extracted using the RNeasy Mini Kit (QIAGEN). Complementary DNA synthesis was performed using SuperScript IV First-Strand cDNA Synthesis Reaction (Thermo Fisher Scientific). The PCR products were separated in 2% agarose gel stained with SYBR Safe (Thermo Fisher Scientific) and imaged using ChemiDoc MP Imaging System (Bio-Rad). All the primer sequences for RT-PCR used in this study were provided in Supplemental Table 5.

Splicing minigene reporter assay

A minigene reporter carrying mNeongreen split by synthetic intronic sequence derived from POLQ intron 24/25 and exon 25 was generated as previously described²⁷. The synthetic intron was made up of 250 bp of the start of the POLQ intron 24, 1000bp of the end of the POLQ intron 24, the POLQ exon 25, 1000 bp of the start of the POLQ intron 25 and 250 bp of the end of the POLQ intron 25. The synthetic introns and an exon so generated were placed between the sequences encoding the N-terminus and C-terminus of split mNeongreen. The mScarlet-P2A-split mNeongreen (N-terminus)-synthetic introns/exon-split mNeongreen (C-terminus) construct was integrated into a pLV-EF1a-IRES-Neo vector (Addgene #85139) to generate the minigene reporter. Cells were lentivirally transduced with the minigene reporter. 48 hours after lentiviral transduction, the cells were enriched by using G418 (0.8mg/ul) for seven days. After antibiotic selection, mScarlet and mNeongreen signals in the reporter cells were analyzed by CytoFLEX (Beckman). For the flow cytometry analyses, mScarlet-positive fractions within living cells were gated. Thereafter, the proportion of the mNeongreen-positive fraction within mScarlet-positive fraction was analyzed by using Flowjo ver.10.9.0.

A screen for identification of cis-regulatory splicing elements

RPE p53−/− cells expressing the minigene splicing reporter were lentivirally transduced with dCasRx (co-expressing blasticidin resistance gene). Cells were treated with blasticidin for 96 hours. A control (NT) and 96 sgRNAs targeting POLQ exon 25, its upstream 1000nt (part of intron 24), or its downstream 1000nt (part of intron 25) were individually integrated into a pLentiRNAGuide_001 - hU6-RfxCas13d-DR1-BsmBI-EFS-Puro-WPRE vector (co-expressing puromycin resistance gene). The 96 sgRNAs were designed to target this region thoroughly. The dCasRx expressing cells were lentivirally transduced with each sgRNA and then treated with puromycin for 72 hours. After antibiotic selection, mScarlet and mNeongreen signals in the reporter cells were analyzed by CytoFLEX (Beckman) and then normalized to the control (NT). The screen result and sgRNA sequences in this study were provided in Supplemental Table 5.

CRISPR-assisted RNA–protein interaction detection method (CARPID)

The CARPID-based proximity labeling to identify binding proteins of POLQ pre-mRNA were performed as previously described³³. In short, HEK293T cells or HEK293T EWSR1-FKBP12^V36F knock-in cells were transfected with CARPID BASU-dCasRx (addgene #153209) together with an sgRNA using Lipofectamine LTX with Plus reagent (Thermo Fisher Scientific). 16h after transfection, media was changed with or without 1uM dTAG^v1. 24 hours after changing media, cells were treated with 200 μM biotin (Sigma) for 15min. Cells were lysed with Cell Lysis Buffer (Cell Signaling Technology, #9803) containing Protease/Phosphatase inhibitor Cocktail (Cell Signaling Technology). Cell lysates were incubated with Dynabeads^™ MyOne^™ Streptavidin T1 for 4 hours (Thermo Fisher Scientific), and then captured biotinylated proteins were subjected to immunoblotting analysis.

Mass spectrometry analysis

EWSR1-FKBP12^V36F knock-in HEK293T cells were seeded into 15 cm dishes and cultured overnight. The cells were transfected with pMYS-IB empty vector (EV) or pMYs-IB-EWSR1 wild-type or pMYs-IB-EWSR1 37YS plasmid. Cells were transfected with 30 μg of vector /dish using Lipofectamine LTX with Plus reagent. 16 hours after transfection, the media was changed with 1 μM dTAG^v1. The cells were harvested 40 hours after transfection and then were lysed with 40 mM Tris-HCl pH 8 buffer supplemented with 0.5% NP-40, 200 mM NaCl, 2 mM EDTA, Protease/Phosphatase inhibitor Cocktail (Cell Signaling Technology), and MG-132. The lysate was centrifuged, and the pellet was resuspended with 20 mM Tris-HCl pH 7.5 supplemented with 100 mM KCl, 2 mM MgCl2, 1 mM CaCl2, and Protease/Phosphatase inhibitor Cocktail and treated with Micrococcal Nuclease (Thermo Fisher Scientific) for approximately 30 minutes, followed by addition of 5 mM EGTA (final concentration) to stop digestion. After three pulses of sonication (30 s each), samples were centrifuged to collect the supernatants as chromatin fractions. Immunoprecipitation was performed using Anti-FLAG M2 antibody (Sigma-Aldrich) conjugated with Dynabeads Protein G for Immunoprecipitation (Thermo Fisher Scientific). Beads were washed with TGN-150 wash buffer containing 20 mM Tris-HCl pH 7.6, 150 mM NaCl, 3 mM MgCl2, 10% glycerol, and 0.01% NP-40. Beads were then boiled at 70 °C for 20 minutes in 1:1 diluted NuPAGE LDS Sample buffer (4X) (Thermo Fisher Scientific) to elute the proteins. The eluted proteins were applied to SDS-PAGE, and Coomassie-stained gel band samples were used for the analyses. Two independent biological replicate samples were analyzed by mass spectrometry.

The gel bands were cut into approximately 1 mm³ pieces. Gel pieces were then subjected to a modified in-gel trypsin digestion procedure. Gel pieces were washed and dehydrated with acetonitrile for 10 minutes and completely dried in a SpeedVac (Thermo Fisher Scientific). The gel pieces were rehydrated with 50 mM ammonium bicarbonate containing 12.5 μg/mL modified sequencing-grade trypsin (Promega) at 4 °C. After rehydration, the excess trypsin solution was removed and replaced with 50 mM ammonium bicarbonate solution. Samples were then incubated at 37 °C overnight. Peptides were later extracted with a solution containing 50% acetonitrile and 1% formic acid and dried in a SpeedVac. The samples were reconstituted in 5–10 μL of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid). A nano-scale reverse-phase HPLC capillary column was created by packing 2.6 μm C18 spherical silica beads into a fused silica capillary (100 μm inner diameter x ~30 cm length) with a flame-drawn tip. After equilibrating the column each sample was loaded via an EASY-nLC (Thermo Fisher Scientific). A gradient was formed, and peptides were eluted with increasing concentrations of solvent B (90% acetonitrile, 0.1% formic acid). As peptides eluted, they were subjected to electrospray ionization and then entered into an Orbitrap Exploris480 mass spectrometer (Thermo Fisher Scientific). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences were determined by matching protein databases with the acquired fragmentation pattern using the software program, Sequest (Thermo Fisher Scientific). All databases include a reversed version of all the sequences, and the data was filtered to between a one and two-percent peptide false discovery rate. Perseus (version 2.0.11.0)⁷¹ was used to preprocess the protein data. The protein intensities were normalized by the median and log2-transformed. Proteins with only one intensity value or with a sum of all total peptide counts of less than 5 were filtered out. For each sample, missing log2-intensity values were imputed using the normal distribution with width of 0.3 and down shift of 1.8. Histograms of log2-intensity values were inspected to confirm that the imputed values were at the lower end. Finally, limma analysis (version 3.54.2)⁷² was performed on the log2-intensity values to determine enriched and depleted proteins. Volcano plots were generated using ggplot2 and ggrepel R packages.

Next-generation sequencing

Library preparation and sequencing reactions were conducted at GENEWIZ, Lin/Azenta US, Inc. Deep RNA-seq (300 million reads/sample, N = 3 for each cell line) was performed for RPE1 TP53−/− (parental control), RPE1 TP53−/− EWSR1-knockout cells, A673 EWS-FLI1-wild-type cells, and A673 EWS-FLI1-depleted cells. RNA samples were prepared using RNeasy Mini Kit and RNase-Free DNase Set (QIAGEN). RNA samples were quantified using Qubit 2.0 Fluorometer, and the RNA integrity was checked with 4200 TapeStation. NEBNext Ultra II RNA Library Prep Kit for Illumina was used following the manufacturer’s recommendations (New England BioLabs). Briefly, mRNAs were initially enriched with Oligo d(T) beads. Enriched mRNAs were fragmented for 15 minutes at 94°C. First strand and second strand cDNA were subsequently synthesized. cDNA fragments were end repaired and adenylated at 3′ ends, and universal adapters were ligated to cDNA fragments, followed by index addition and library enrichment by PCR with limited cycles. The sequencing libraries were validated using Agilent TapeStation and quantified using Qubit 2.0 Fluorometer as well as by quantitative PCR (KAPA Biosystems). The sequencing libraries were clustered on flowcell lanes. After clustering, the flowcell was loaded on the Illumina NovaSeq instrument according to manufacturer’s instructions. The samples were sequenced using a 2×150 Paired End configuration. Image analysis and base calling were conducted by the Control Software (NCS). Raw sequence data (.bcl files) generated from the Illumina instrument was converted into fastq files and de-multiplexed using Illumina’s bcl2fastq (version 2.17) software. One mismatch was allowed for index sequence identification.

Several bioinformatics analyses were performed on the RNA-seq data to identify differentially expressed genes, differentially used exons, and differentially spliced events. For differential splicing analyses, STAR (version 2.7.10b)⁷³ was used to map RNA-seq reads and generate BAM files. Specifically, the 2-pass STAR was used to map the raw RNA-seq reads against the GRCh38/hg38 human reference genome, with the GENCODE human reference gene annotation (Release 44)⁷⁴. The first round of the 2-pass STAR was run to identify splice junctions, and more confidence junctions then were filtered by requiring them to have at least 5 total mapped reads. The second round of the 2-pass STAR was run using the filtered junctions to generate BAM files. Differential splicing analysis was performed using rMATS (version v4.2.0)⁷⁵. Lower quality events were filtered out by requiring each sample to have at least 35 total isoform-identifying reads. Significant differential alternative splicing events were called if they had FDR smaller than 0.05 and absolute delta percent-spliced-in (PSI) value of at least 0.05. Volcano plots for the different splicing events were generated using ggplot2 and ggrepel R packages. Sashimi plots were generated using the rmats2sashimiplot program (https://github.com/Xinglab/rmats2sashimiplot/).

Motif analysis of eCLIP-seq

The raw sequencing data of EWSR1 eCLIP-seq were obtained from ENCODE, and GEO (GSE98836). Cutadapt (version 2.5) was used to trim sequences of adapter sequences. The sequences were mapped using STAR (version 2.7.10b). Samtools (version 1.9) was used to filter the mapping results. Peak calling was performed using pureclip (version: 1.3.1). The top 3000 peaks from the binding region results were picked based on pureclip score. Bedtools (version 2.31.1) was used to pad the top peaks with 20bp. Motif analysis was performed using meme (version 5.5.5). The motifs from ENCODE data were ranked by E-values, and the motifs from GEO data were ranked by the number of sites.

AlphaFold analysis

AlphaFold3 and the AlphaFold Server⁷⁶ were used to model full-length (Polθ aa 1824–2579) and truncated (Polθ aa 1824–2390) Polθ polymerases domains in complex with two complementary 9 bp DNA molecules (5’-CCAATGACA-3’ and 5’-TGTCATTGG-3’). Models were visualized in Pymol V3.0.4 (Schrödinger) using the top-ranking predictions per seed. The full-length model was compared with a published crystal structure of the Polθ polymerase domain (PDB 6XBU)⁷⁷ to ensure its accuracy.

Computation of exon skipping rates in primary patient samples

To compute the proportions of reads with exons skipped in the sequencing reads data obtained from St. Jude’s Cloud⁷⁸, rMATS-turbo, a computational tool to quantify alternative splicing patterns from RNA sequencing data, was applied. To compute the proportion of reads that each exon was skipped in, the rMATS-turbo results inclusive of supporting reads counted from both junction and exon reads were used. Of note, for each exon, any sample that had no supporting reads for both the inclusive and skipped cases was omitted. All comparisons between distributions of skipped read proportions were performed using Mann-Whitney U tests.

Mutational Analysis of Tumors

Somatic variants were called from the BAMs of each matched tumor-normal pairs using GATK Mutect2 (version 4.2.4; https://www.biorxiv.org/content/10.1101/861054v1). Several samples were identified as potentially contaminated, lower quality samples and excluded. The passing SNPs and INDELs variants were further filtered with these more stringent hard filters: 1) total depth of both tumor and matched normal >= 20 reads; 2) tumor variant allele depth >= 5 (7 for INDELS); 3) tumor variant allele fraction >=0.075 (0.10 for INDELs); 4) normal variant allele depth <= 1; 5) median base quality by allele (MBQ) >= 27; 6) median mapping quality by allele (MMQ) >=55; 7) median distance from end of read (MPOS) >= 10 (>= 15 for INDELs; and 8) log10 likelihood ratio score of variant existing versus not existing (TLOD) >= 10.

Profiles of SNV mutations and INDELs were generated using SigProfilerMatrixGenerator (version 1.2.13; https://github.com/AlexandrovLab/SigProfilerMatrixGenerator). Mutational signature analysis was performed using SigProfilerAssignment (version 0.0.21; https://github.com/AlexandrovLab/SigProfilerAssignment). This involved fitting the mutational profiles to the COSMIC (https://cancer.sanger.ac.uk/signatures/) SBS signatures (SBS version 2) and INDEL signatures (ID version 3.3).

MMEJ-associated mutational events were defined as follows. Deletions were classified as MMEJ events if they were ≥3 bp in length with ≥2 bp of flanking microhomology, identified using the method of Taheri-Ghahfarokhi et al^41,42. Briefly, for each deletion, 50-bp flanking sequences were retrieved from the human reference genome (HG38) using bedtools (version 2.27.1; https://bedtools.readthedocs.io/). The deletion was then left-shifted iteratively until no further shifting was possible, and the microhomology length was defined as the longest identical subsequence between the deleted sequence and the immediately flanking right sequence. Templated insertions were classified as MMEJ events if the inserted sequence contained a template of ≥7 bp derived from the local 50-bp flanking sequence^16,45,46, detected by the Smith–Waterman local alignment algorithm⁷⁹. The insertions with 5+-bp templates covering at least 75% of the inserted sequence were designated as potential MMEJ events. The potential MMEJ insertions that had longest and closest templates were either 7+-bp or covering 85% of the inserted sequence, and adjacent by at most 2bp to the insertion position were excluded. For comparison between ES and other tumors, tumors with at least 30 INDELs, and the values of log₁₀(MMEJ insertions + 1) / log₁₀(insertions) were used.

Immunohistochemistry (IHC)

IHC was performed using a Parhelia autostainer. Slides were deparaffinized using Slide Brite xylene substitute (23401–01) followed by rehydration in ethyl alcohol. Endogenous blocking was performed using methanol hydrogen peroxide solution. Slides were heated for 30 minutes in citrate solution for epitope retrieval. Slides were incubated for 30 minutes in anti-DNAPKcs phosphor S2056 antibody at 1:150 followed by washing in wash buffer (Dako S3006) and incubation for 30minutes in Mach 3 rabbit HRP (Buocare Medical RP531L). Primary staining was performed by incubation in DAB (Dako K3468) for 10minutes. Counterstaining was performed using hematoxylin (Sigma-Aldrich MHS16) and bluing buffer (Dako C5702). Slides were then dehydrated in ethyl alcohol before coverslipping using Permount mounting medium (Fisher Scientific SP15–100). Slides were digitally imaged using a Motic Easyscan Pro 6.

Histological image analysis was performed using QuPath (version 0.5.1). Whole-slide images (WSIs) were imported into QuPath, and automated color deconvolution was applied to separate the hematoxylin and DAB (3,3′-diaminobenzidine) staining channels. The mean optical density (OD) of the DAB-positive areas was measured to assess staining intensity. For patients with multiple tissue sections, the mean OD across all sections for the same patient was used to represent the patient’s staining intensity.

Quantification and statistical analysis

Data were analyzed and visualized using GraphPad Prism (Version 10.2.2, GraphPad Software, LLC). For comparisons between two groups, either a two-tailed Student’s t-test or a Mann-Whitney U test was used. For comparisons involving more than two groups, one-way or two-way analysis of variance (ANOVA) was performed. For the in vivo survival analysis study, a log-rank test was performed. A p-value of less than 0.05 (p < 0.05) was considered statistically significant.

Supplementary Material

2

Table S1: Differential splicing changes in EWSR1 wild-type and EWSR1-knockout RPE1 TP53−/− cells, related to Figure 2.

3

Table S2: A list of proteins identified from EWSR1-IP mass spectrometry, related to Figure 3.

4

Table S3: Differential splicing changes in EWS-FLI1-wild-type and EWS-FLI1-depleted A673 cells, related to Figure 5.

5

Table S4: Differential splicing changes commonly seen in EWSR1 depletion and EWS-FLI1 expression, related to Figure 5.

6

Table S5: List of oligonucleotides used in this study and the result of the dCasRx-based screen, related to Methods and Figure 4.

Highlights:

• Ewing sarcoma is characterized by reduced Polθ expression and impaired MMEJ activity.

• The EWSR1/FUBP1/KHSRP complex prevents POLQ exon 25 skipping.

• The EWS-FLI1 oncoprotein disrupts EWSR1-dependent POLQ splicing.

• MMEJ deficiency makes Ewing sarcoma highly dependent on other DNA repair pathways.