CTC1-STN1-TEN1 controls DNA break repair pathway choice via DNA end resection blockade

Abstract

Antagonistic activities of the 53BP1 axis and the tumor suppressor BRCA1-BARD1 determine whether DNA double-strand breaks (DSBs) are repaired by end joining or homologous recombination. We show that the CTC1-STN1-TEN1 (CST) complex, a central 53BP1 axis component, suppresses DNA end resection by EXO1 and the BLM-DNA2 helicase-nuclease complex but acts by distinct mechanisms in restricting these entities. Whereas BRCA1-BARD1 alleviates the CST-imposed EXO1 blockade, it has little effect on BLM-DNA2 restriction. CST mutants impaired for DNA binding or BLM–EXO1 interaction exhibit a hyper-resection phenotype and render BRCA1-deficient cells resistant to poly(ADP–ribose) polymerase (PARP) inhibitors. Our findings mechanistically define the crucial role of CST in DNA DSB repair pathway choice and have implications for understanding cancer therapy resistance stemming from dysfunction of the 53BP1 axis.

Several DNA double-strand break (DSB) repair mechanisms exist, of which nonhomologous DNA end joining (NHEJ) and homologous recombination (HR) represent the major pathways (1). Which repair pathway cells use depends on antagonistic activities of the 53BP1 axis and the tumor suppressor BRCA1-BARD1 (2–5). Repair pathway choice occurs at the DNA end resection step of HR, with 53BP1 and associated factors dampening resection efficiency. DNA end resection is mediated by the 5′–3′ exonuclease EXO1 and DNA2 in association with one of the RECQ helicases, BLM or WRN (6–9). Upon adenosine triphosphate hydrolysis, DNA2 translocates on single-stranded DNA (ssDNA) until it reaches a ssDNA–double-stranded DNA (dsDNA) junction, where it incises the 5′-terminated ssDNA strand (10, 11). DNA end resection is up-regulated by BRCA1-BARD1 (8, 9). By contrast, the association of 53BP1 with DSBs leads to the recruitment of effectors including the heterotrimeric CTC1-STN1-TEN1 (CST) complex to favor the engagement of NHEJ as the DSB repair tool (12, 13).

Impairment of the 53BP1 axis renders BRCA1-deficient tumor cells resistant to cancer therapeutics, such as inhibitors of poly(ADP-ribose ) polymerase (PARPi), through restoration of DNA end resection and/or RAD51 loading that allows HR (12, 14, 15). Thus, understanding the mechanisms of action of 53BP1 and its associated factors on DNA end resection has important clinical implications. Here, we show direct action of CST on EXO1 and BLM-DNA2 and provide evidence that implicates BRCA1-BARD1 in alleviating the CST blockade on EXO1.

End resection proteins and CST

We purified EXO1, BLM, DNA2, BRCA1-BARD1, and RPA and reconstituted the end resection of a randomly labeled 2-kb dsDNA substrate as described (8, 16). CTC1, STN1, and TEN1 were coexpressed in insect cells, and the assembled CST complex was purified with a procedure that we have developed. The CTC1-STN1 (CS) and STN1-TEN1 (ST) subcomplexes were also expressed and purified (Fig. 1, A to C) with the same protocol, which is also suitable for obtaining CST mutants as we describe below.

Fig. 1. CST inhibits EXO1-mediated DNA resection.

(A) Schematic of the CST complex. OB, oligonucleotide binding fold; HTH, winged helix-turn-helix. (B) SDS–polyacrylamide gel electrophoresis (PAGE) analysis of purified CST, CS, and ST complexes. (C) CST, CS, and ST were analyzed by mass photometry. (D) Schematic of the EXO1 resection assay. (E) EXO1 resection of a 2-kb dsDNA substrate was performed for 10 min at 37°C in the presence or absence of the indicated amounts of CST, CS, or ST. (F) Quantification of (E); mean ± SD; one-way analysis of variance (ANOVA). (G) Whole-lane profiles of the indicated resection assays from (E). The single and double asterisks (* and **, respectively) indicate the relative position of resection products. Intensity was normalized by subtracting the lowest-intensity value from each lane. (H) Microscale thermophoresis (MST) was performed with CST, CST^m5, and CST^m10 by using 5′ Cy5-labeled 80-nt ssDNA. K_d, dissociation constant; WT, wild type. (I) EXO1 resection assay was performed with CST and the indicated DNA binding mutants. CST-alone lanes (no EXO1) used 100 nM protein. HD, heat denatured. (J) Quantification of (I); mean ± SD; one-way ANOVA.

EXO1 restriction by CST

We first determined the effect of CST on EXO1 (Fig. 1D) (16). CST strongly inhibited EXO1-mediated DNA end resection (Fig. 1, E to G, and fig. S1, A and B). Whereas the CS subcomplex was proficient in EXO1 restriction, the ST subcomplex was not, which thus revealed TEN1 as dispensable for CST action in vitro (Fig. 1, E to G). CST also inhibited yeast Exo1 and human EXO1^1–346, a truncated form that harbors the catalytic domain (fig. S1, C and D). CST inhibition of EXO1 was not influenced by RPA addition (fig. S1, E and F). We verified that CST has high affinity for telomeric ssDNA (3xTEL), which is consistent with previous reports that CST preferentially binds G-rich DNA (17), and exhibits nanomolar affinity for an 80-mer ssDNA of a randomized sequence (fig. S2, A to D). CST shows stronger binding to ssDNA or dsDNA with a 15-nucleotide (nt) 3′ overhang compared with dsDNA (fig. S2, E and F).

To assess the relevance of CST DNA binding, we constructed point mutants by referencing the available cryo–electron microscopy structure of CST bound to 3xTEL ssDNA (Protein Data Bank identifier: 6W6W) (18). We introduced mutations targeting crucial residues within CTC1 oligonucleotide binding (OB) folds F and G known to engage 3xTEL DNA (fig. S2, A and B). Although the CTC1-K1164E/K1167E and CTC1-R978E/N981D/Y983A mutants, designated as CST^m2 and CST^m3, respectively, were defective in binding 3xTEL, they exhibited a similar affinity for the 80-mer ssDNA substrate as their wild-type counterpart (fig. S2, C and D). We found that although CST^m2 and CST^m3 have the same affinity for DNA, only the m3 mutant is slightly impaired in EXO1 restriction (fig. S2, G and H). We next constructed two new CTC1 mutants by combining the m2 and m3 mutations (CST^m5) or with five additional point mutations in OB-F and OB-G (CST^m10) (fig. S2, A, I, and J) and showed that they are more impaired for DNA binding than the m2 and m3 variants, with m10 exhibiting a greater deficit (Fig. 1H and fig. S2D). Whereas CST^m5 retained residual capability in EXO1 restriction, CST^m10 was devoid of such activity (Fig. 1, I and J). The CST mutants were as proficient as the wild-type counterpart in EXO1 interaction (fig. S2K). Preincubation of CST with an unlabeled ssDNA greatly attenuated EXO1 restriction, which suggests that CST engages ssDNA in a cis fashion to down-regulate EXO1 activity (fig. S1, G and H). Altogether, the results above reveal that CST (i) physically interacts with EXO1, (ii) strongly restricts EXO1 activity in cis, and (iii) requires its DNA binding activity in EXO1 restriction.

Relief of EXO1 blockade by BRCA1-BARD1

The tumor suppressor complex BRCA1-BARD1 is an antagonist of the 53BP1 axis (4, 12), and recent studies have shown that it physically interacts with EXO1 and enhances its end resection activity (fig. S3, A and B) (8). By affinity pull-down, we found that CST physically interacts with BRCA1-BARD1 (Fig. 2A). These attributes of BRCA1-BARD1 prompted us to assess whether it would affect the CST-imposed blockade of EXO1. Under CST-imposed restrictive conditions, addition of BRCA1-BARD1 in an amount stoichiometric to EXO1 led to marked restoration of DNA resection (Fig. 2, B and C, and fig. S3, C and D). We next investigated whether BRCA1-BARD1 could restore EXO1 activity over blockade imposed by the catalytically dead EcoRI^E111Q (19). With a 2-kb substrate that harbored strategically placed Eco RI recognition sites at the DNA ends (fig. S3, E and F), we found strong inhibition of EXO1 activity by EcoRI^E111Q, but the addition of BRCA1-BARD1 could not alleviate this blockade (fig. S3, G and H).

Fig. 2. BRCA1-BARD1 alleviates the CST-imposed block of EXO1.

(A) CST interaction with BRCA1-BARD1 was examined by affinity pull-down. BRCA1^WT-BARD1^WT or BRCA1^ΔDBD-BARD1^ΔDBD were incubated with or without CST (Strep-tag on TEN1) before pull-down with Strep-Tactin Sepharose. Samples were analyzed by SDS-PAGE and Coomassie blue staining. S, supernatant; E, SDS eluate. (B) EXO1-mediated resection was measured in the presence or absence of BRCA1-BARD1 and/or CST. (C) Quantification of (B); mean ± SD; one-way ANOVA. (D) Resection assays were performed as in (B) by using BRCA1-BARD1 DBD deletions. BRCA1-BARD1 alone lanes used 10 nM protein. (E) Quantification of (D); mean ± SD; one-way ANOVA.

BRCA1 and BARD1 both bind DNA (20), with the BARD1 DNA binding activity being more important for EXO1 enhancement (fig. S3, A and B) (8). We tested BRCA1-BARD1 mutants deleted for either one or both of the BRCA1 and BARD1 DNA binding domains (ΔDBD mutations) for alleviation of the CST blockade. First, we verified by affinity pull-down that the BRCA1^ΔDBD-BARD1^ΔDBD double mutant is proficient in CST interaction (Fig. 2A). Whereas BRCA1^ΔDBD-BARD1 could alleviate the CST-imposed EXO1 blockade, BRCA1-BARD1^ΔDBD was partially defective, with the BRCA1^ΔDBD-BARD1^ΔDBD double mutant being strongly impaired for resection restoration (Fig. 2, D and E).

Restriction of BLM-DNA2 by CST

We investigated whether CST also affects the activity of BLM and DNA2. Whereas CST had little or no impact on BLM-mediated DNA unwinding (Fig. 3, A to C, and fig. S4A), it inhibited DNA cleavage by DNA2 (Fig. 3, D to F, and fig. S4, B and C). The CS subcomplex, but not the ST subcomplex, could restrict DNA cleavage by DNA2 (Fig. 3, E and F). The CST^m5 and CST^m10 mutants were both impaired for the ability to inhibit DNA2, with CST^m10 exhibiting a more pronounced functional deficiency (fig. S4, D and E).

Fig. 3. CST inhibits resection by BLM-DNA2.

(A) Schematic of the BLM helicase assay. (B) Testing of BLM helicase activity. Reactions with a 2-kb substrate were assembled by mixing 0.25 nM DNA, ±30 or ±60 nM CST, 200 nM RPA, and 15 nM BLM, followed by incubation at 37°C for the indicated times. (C) Quantification of (B); mean ± SD. (D) Schematic of the DNA2 flap endonuclease assay. (E) DNA2 activity was measured on 2.5 nM Y-shaped DNA in the presence or absence of 10 or 20 nM CST, CS, or ST for 10 min at 37°C. The red asterisk denotes the 5′ Cy5 label. Subcomplex reactions lacking DNA2 used 20 nM protein. (F) Quantification of (E); mean ± SD; one-way ANOVA. (G) Schematic of the BLM-DNA2 resection assay. (H) Resection of 2-kb dsDNA by BLM-DNA2 was measured with and without CST, CS, or ST. Time course reactions containing 0.25 nM DNA, ±60 nM CST or subcomplex, 200 nM RPA, 15 nM DNA2, and 15 nM BLM were performed at 37°C for the indicated times. (I) Quantification of (H); mean ± SD. (J) CST DNA binding mutants were measured for the ability to block BLM-DNA2 resection of 2-kb dsDNA. Reactions were performed as in (H). (K) Quantification of (J); mean ± SD.

Next, we determined the impact of CST on BLM-DNA2–mediated resection of the 2-kb substrate (Fig. 3G). Whereas CST exerted little negative impact on BLM-mediated DNA unwinding as indicated by the accumulation of near-full-length ssDNA, it strongly inhibited resection by BLM-DNA2 (Fig. 3, H and I). As revealed by testing the CS and ST subcomplexes, TEN1, but not CTC1, appears to be dispensable for BLM-DNA2 inhibition (Fig. 3, H and I). However, although CST^m5 and CST^m10 are impaired for the ability to attenuate EXO1 activity (Fig. 1, I and J) and the DNA flap endonuclease attribute of DNA2 (fig. S4, D and E), these CST mutants were nearly as proficient as wild-type CST in restricting BLM-DNA2 (Fig. 3, J and K). We also observed that both CST^m5 and CST^m10 interact with BLM as avidly as their wild-type counterpart (fig. S4F), and the functional consequence of this interaction will be analyzed below. Using the orthologous resection machinery from yeast, Sgs1-yDna2, we observed that CST does not affect DNA unwinding by Sgs1 and only slightly inhibits resection (fig. S4, G and H) (7, 21). Contrary to our findings with BLM-DNA2, CST^m10 could not restrict Sgs1-yDna2. Thus, CST has high specificity for BLM-DNA2, and restriction entails an attribute of CST distinct from its DNA binding function. Moreover, although BRCA1-BARD1 could efficiently alleviate the CST-imposed EXO1 blockade, it failed to overcome the inhibitory effect of CST on the nuclease activity of DNA2 or on DNA end resection by BLM-DNA2 (fig. S5, A to D).

Mechanism of BLM-DNA2 restriction

DNA curtain analysis was conducted to study the behavior of the CST-BLM-DNA2 ensemble (fig. S6A) (22). As we reported before (8), mCherry-BLM and green fluorescent protein (GFP)–DNA2 cotranslocate on ssDNA (fig. S6B). The addition of CST did not affect BLM-DNA2 cotranslocation (fig. S6B). We also observed co-migration of GFP-DNA2 and mCherry-CST in the presence of unlabeled BLM (fig. S6C). In congruence with the single-molecule analysis, affinity pull-down showed that CST associates with DNA2 only in the presence of BLM (fig. S6D). Taken together, our results suggest that CST does not prevent BLM-DNA2 complex formation or cotranslocation of BLM-DNA2 but instead acts by functionally altering BLM and DNA2 within the context of a higher-order complex that unwinds DNA normally but with a much-reduced potential for DNA strand cleavage.

BLM and EXO1 interaction domain in CTC1

Because DNA binding by CST is largely dispensable for BLM-DNA2 restriction (Fig. 3, J and K), we hypothesized that physical interaction between CST and BLM constitutes a key element of CST-imposed blockade (Fig. 4A). To test this premise, we used biochemical domain mapping and structural prediction tools to identify the CST-BLM interaction interface. By affinity pull-down, we found that CST interacts with the N-terminal region (residues 2 to 635) of BLM upstream of the helicase core (fig. S7A). We surmised that CTC1 harbors the BLM interface because we saw no interaction of the ST subcomplex with BLM (Fig. 4A). Deletion of OB folds ABC or ABCD of CTC1 yielded soluble CST^ΔABC and CST^ΔABCD mutants in insect cells amenable to purification (fig. S7, B to D). A published study reports that removal of the first 700 residues of CTC1 (similar to our CST^ΔABCD mutant) renders CST defective in binding a duplex DNA substrate with a 35-nt 5′ ssDNA overhang (23). In congruence with this finding, CST^ΔABC and CST^ΔABCD displayed a partial or strong deficiency in ssDNA binding, respectively (fig. S7E). Taken together, it is clear that aside from OB-F and OB-G (18), additional contacts with DNA are mediated through one or more of OB folds A, B, C, and D. Affinity pull-down of CST^ΔABC and CST^ΔABCD revealed that neither is capable of BLM interaction (fig. S7F), providing evidence that the BLM-interacting domain resides within OB fold A, B, or C. We also observed that CST^ΔABC and CST^ΔABCD are unable to restrict BLM-DNA2–mediated resection (fig. S8, A and B). Consistent with results shown earlier (fig. S4, D and E), neither CST^ΔABC nor CST^ΔABCD could inhibit DNA2 nuclease function (fig. S8, C and D). Both CST^ΔABC and CST^ΔABCD were impaired for EXO1 interaction, indicating that the OB-ABC region of CTC1 harbors an EXO1 interaction domain as well (fig. S8E). Further testing of CST^ΔABC and CST^ΔABCD showed that they are incapable of EXO1 restriction (fig. S8, F to I).

Fig. 4. Contribution of the CST-BLM interaction in the restriction of resection by BLM-DNA2.

(A) Affinity pull-down of BLM or DNA2 with CST or ST through the Strep-tag on TEN1. The indicated proteins were incubated together for 30 min at 4°C before the addition of Strep-Tactin Sepharose for an additional 30 min at 4°C. Proteins were eluted with SDS and analyzed by SDS-PAGE and Coomassie blue staining. S, supernatant; E, SDS eluate. (B) AlphaFold2-multimer prediction of CTC1 OB-ABC with the BLM N-terminal fragment. The CTC1 fragment is presented with electrostatic density (negative charges are highlighted in red and positive charges in blue). BLM residues (in cyan) that are predicted to interact with CTC1 are displayed. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; K, Lys; N, Asn; Q, Gln; R, Arg; and Y, Tyr. (C) CTC1 OB-B KRR and KQK mutations ablate BLM interaction. Affinity pull-down reactions were performed with the indicated proteins as in (A). (D) Functional deficit of CST mutants in BLM-DNA2 restriction. Resection assays were performed as above. (E) Quantification of (D); mean ± SD. (F) CST mutations have no impact on DNA2 nuclease inhibition. Lanes 9 and 10 used 20 nM of either OB-B mutant alone. (G) Quantification of (F); mean ± SD; one-way ANOVA. (H) Interaction of EXO1 with CST OB-B point mutants was determined by affinity pull-down. The indicated proteins were incubated at 4°C for 30 min before being mixed with GFP-Trap Agarose for protein capture. Proteins were eluted with SDS and analyzed by SDS-PAGE with Coomassie blue staining. (I) Testing CST mutants on EXO1 restriction. CST or the indicated mutant (50 or 100 nM) was incubated with 10 nM EXO1 and 0.25 nM 2-kb dsDNA for 10 min at 37°C. CST lanes lacking EXO1 used 100 nM complex. (J and K) Quantification of (I); total resection products (J) or intermediate products (K) were plotted to highlight a partial defect in EXO1 restriction; mean ± SD.

Importance of CTC1-BLM–EXO1 interaction

Results presented above show that CST^ΔABC and CST^ΔABCD are defective in DNA binding (fig. S7E) and BLM interaction (fig. S7F). We strived to isolate CTC1 mutants specifically impaired for BLM interaction to assess the role of CTC1-BLM interaction in resection control. For this, we performed a multimeric structure prediction of the salient CTC1 region (residues 1 to 546 encompassing OB folds A, B, and C) and BLM region (residues 177 to 385) using the ColabFold open-source tool to predict an interaction interface between CTC1 and BLM (24). The analysis identified a small acidic patch within BLM residues 290 to 300 that fits within a basic cleft on the surface of OB-B in CTC1 (Fig. 4B). Accordingly, we generated two compound point mutations—namely, CTC1^KRR (K290E, R292E, and R295E) and CTC1^KQK (K240E, Q241E, and K242E)—in two loops that make up the OB-B basic cleft (fig. S7G). We purified the two mutant CST complexes (fig. S7, H and I) and verified that both bind ssDNA as proficiently as the wild-type counterpart (fig. S7J). Salient analyses provided direct evidence that CST^KRR and CST^KQK are impaired for BLM interaction (Fig. 4C) and significantly attenuated for the ability to restrict end resection by BLM-DNA2 (Fig. 4, D and E) despite being able to efficiently inhibit DNA2-mediated DNA cleavage (Fig. 4, F and G). Thus, CTC1-BLM interaction underpins the efficacy of CST as a negative regulator of BLM-DNA2.

We found that CST^KRR and CST^KQK are partially impaired for EXO1 interaction (Fig. 4H) and exhibit a partial defect in EXO1 restriction as well (Fig. 4, I to K, and fig. S8I). Thus, as in the case with BLM-DNA2, interaction of CST with EXO1 is indispensable for the full inhibitory activity of CST on EXO1.

Hyper-resection phenotype of CST mutants

To determine the biological impact of our mechanistic findings, we first verified that loss of CTC1 would lead to a hyper-resection phenotype. Consistent with previous reports (12, 25, 26), RPE1 p53^−/− CTC1^−/− cells (fig. S9, A and B) displayed increased RPA foci in response to DNA damage induced by ionizing radiation, a reliable cellular readout for DNA end resection (fig. S10, A and B). Expression of GFP-CTC1 (fig. S9C) restored normal levels of DNA end resection (fig. S10, A and B). Because the CTC1^m5 and CTC1^m10 mutations strongly affect EXO1 restriction, whereas the CTC1^KRR and CTC1^KQK mutations impair resection by EXO1 and BLM, we determined whether these mutants would phenocopy CTC1 depletion in cells. Indeed, cells expressing any of the aforementioned CTC1 mutants (fig. S9D) failed to attenuate resection, as revealed by increased RPA foci in response to DNA damage, with the degree of hyper-resection phenotype in the order of KQK ≥ KRR > m10 > m5 (Fig. 5, A and B, and fig. S10C). We observed normal recruitment of all four CTC1 mutants to DSBs (Fig. 5A and fig. S10, D to F). By coimmunoprecipitation and proximity ligation assay, we verified that CTC1^KRR and CTC1^KQK are impaired for BLM and EXO1 interaction in cells, whereas the CTC1^m5 and CTC1^m10 are proficient in this regard (Fig. 5C and figs. S9E and S11, A to J). All four mutants retained the ability to associate with 53BP1 and two key components of the Shieldin complex, SHLD1 and SHLD2 (Fig. 5C and figs. S9E and S11, A to J).

Fig. 5. CST mutations impart a hyper-resection phenotype in cells.

(A) Representative images of RPA foci in S phase (cyclin A+) RPE1 p53^−/− CTC1^−/− cells expressing GFP-tagged CTC1 wild-type and mutants 4 hours after exposure to 5-Gy infrared (IR) light. Scale bar, 20 μm. DAPI, 4′,6-diamidino-2-phenylindole. (B) Quantification of (A) (N = 3; >100 cells per experiment; two-sided unpaired t test; black line represents median). (C) Interaction between CTC1 mutants and 53BP1, BLM, EXO1, SHLD1, and SHLD2. Lysates were collected 4 hours after 5-Gy IR exposure followed by coimmunoprecipitation and Western blotting. EV, empty vector; IP, immunoprecipitation. (D) Percentage of ssDNA formation at indicated distances from DSB1 4 hours after 4-OHT treatment in U2OS-AsiSI cells depleted of CTC1 and transfected with CTC1 mutants (data are mean ± SD.; two-sided unpaired t test). (E) Analysis of NHEJ and HR upon expression of CTC1 mutants with GFP-EJ7 and DR-GFP assays (data are mean ± SD of N = 3; two-sided unpaired t test). (F) Survival of RPE1 p53^−/− and RPE1 p53^−/− CTC1^−/− cells cotransfected with small interfering RNA (siRNA) against BRCA1 and GFP-tagged CTC1 or CTC1 mutants upon indicated concentrations of olaparib (data are mean ± SD of N = 3; non–regression curve analysis).

Next, we used the quantitative ER-AsiSI system to further define the impact of the CTC1 mutations on DNA end resection. In cells that express the AsiSI endonuclease fused to the estrogen receptor (ER-AsiSI), AsiSI is induced to enter the nucleus to create DNA breaks upon treatment with 4-hydroxytamoxifen (4-OHT) (27, 28). As expected, cells depleted of CTC1 (fig. S9, A and B) displayed a hyper-resection phenotype, which was corrected upon exogenous expression of wild-type CTC1 (figs. S9F and S12A). By contrast, cells expressing the CTC1^m5, CTC1^m10, CTC1^KRR, or CTC1^KQK mutant (fig. S9G) exhibited a hyper-resection phenotype at both break sites analyzed (Fig. 5D and fig. S12B). Consistent with results from RPA focus analysis, the CTC1^KRR and CTC1^KQK mutations induced a significantly stronger phenotype (Fig. 5D and fig. S12B). Loss of EXO1 partially suppressed the hyper-resection phenotype of CTC1 mutant cells, whereas depletion of BLM had a lesser effect (fig. S12, C to K), which could be due to the WRN helicase providing a parallel resection function (29).

Although TEN1 is dispensable for CST action on EXO1 (Fig. 1, E to G) and BLM-DNA2 (Fig. 3, H and I) in vitro, results from CRISPR/SpCas9 screens showing that TEN1 is needed for CST function in cells (30). We found that TEN1 depletion leads to enhanced DNA end resection in cells (fig. S13, A to E) because it abrogates CTC1 recruitment to DNA break sites (fig. S13A). Thus, TEN1 is necessary for the DSB recruitment of CST, although it is dispensable for the restriction of EXO1 and BLM-DNA2 activity in vitro.

Reacquisition of DNA end resection restores HR capacity in BRCA1-deficient cells (31–34). Similar to the loss of CTC1 (12), cells expressing any of the CTC1^m5, CTC^m10, CTC1^KRR, and CTC1^KQK mutants exhibited increased HR capacity and resistance to olaparib, a PARPi, upon loss of BRCA1 (Fig. 5, E and F, and fig. S14, A to D). Consistent with the severity of the hyper-resection phenotype observed, CTC1^KRR and CTC1^KQK mutant cells were more resistant to olaparib than cells expressing either of the DNA binding mutants.

Discussion

We have addressed how CST, a key effector within the 53BP1 axis, helps drive DSB repair pathway choice via direct restrictive action on EXO1 and BLM-DNA2 (fig. S15). CST acts via distinct mechanisms to attenuate DNA end resection by EXO1 and BLM-DNA2. Specifically, DNA binding by CST is required for EXO1 inhibition but largely dispensable for BLM-DNA2 restriction. By contrast, testing of the CTC1^KRR and CTC1^KQK mutants provides evidence that interaction of CST with BLM and EXO1 is needed for resection blockade, although the relative contributions of CST interactions with DNA and EXO1 in the restriction of EXO1 requires further investigation. Similarly, it will be of interest to delineate the biological importance of DNA binding by OB folds other than OB-F and OB-G (18). In congruence with these mechanistic findings, cell-based analyses have revealed a hyper-resection phenotype in cells expressing mutants defective in CTC1 DNA binding or BLM/EXO1 interaction. There is growing evidence that epigenetic silencing or mutational inactivation of the 53BP1 axis leads to partial restoration of DNA end resection and HR, resulting in PARPi resistance in BRCA-deficient tumors (35–39). Accordingly, BRCA1-deficient cells that express any of our CST mutations are resistant to PARPi. Although TEN1 does not affect EXO1 or BLM/DNA2 activity in vitro, it is essential for CST recruitment to DSBs in cells. CST recruits DNA polymerase α-primase to synthesize DNA and thus reduce the single-stranded nature of partially resected DNA ends to promote NHEJ (12, 40). However, biological evidence suggests that CST can also attenuate HR independently of DNA polymerase α-primase (41), highlighting the complexity of its role in DNA repair. Our findings not only clarify the mechanistic role of CST in DSB repair pathway choice but also are important for understanding cancer drug resistance of HR-deficient tumors.

Supplementary Material

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adt3034

Materials and Methods; Figs. S1 to S15; Table S1; References (43–49); MDAR Reproducibility Checklist

Data and materials availability:

The datasets for the findings in this manuscript are available here or as part of the supplementary materials. Source datasets are deposited in the Texas Data Repository (https://doi.org/10.18738/T8/J0AILA). Requests for reagents may be directed to the corresponding authors.