DNA-PK and the TRF2 iDDR inhibit MRN-initiated resection at leading-end telomeres

Telomeres replicated by leading-strand synthesis lack the 3′ overhang required for telomere protection. Surprisingly, resection of these blunt telomeres is initiated by the telomere-specific 5′ exonuclease Apollo rather than the Mre11–Rad50–Nbs1 (MRN) complex, the nuclease that acts at DNA breaks. Without Apollo, leading-end telomeres undergo fusion, which, as demonstrated here, is mediated by alternative end joining. Here, we show that DNA-PK and TRF2 coordinate the repression of MRN at blunt mouse telomeres. DNA-PK represses an MRN-dependent long-range resection, while the endonuclease activity of MRN–CtIP, which could cleave DNA-PK off of blunt telomere ends, is inhibited in vitro and in vivo by the iDDR of TRF2. AlphaFold-Multimer predicts a conserved association of the iDDR with Rad50, potentially interfering with CtIP binding and MRN endonuclease activation. We propose that repression of MRN-mediated resection is a conserved aspect of telomere maintenance and represents an ancient feature of DNA-PK and the iDDR.

Article https://doi.org/10.1038/s41594-023-01072-x absence of DNA-PK. As telomere fusions in Apollo-deficient cells are thought to be a consequence of a resection defect, we explored the possibility that loss of DNA-PK unleashes compensatory resection that restores the 3′ overhang. A role for DNA-PK as a repressor of resection would be in line with Ku70/80 blocking resection in yeast [26][27][28][29] . Indeed, after Apollo deletion, cells lacking Ku70 or DNA-PKcs did not show the decrease in the overhang signal observed in DNA-PK-proficient cells (Fig. 2a,b). By contrast, depletion of Lig3 or PolQ did not have this effect, and cells lacking alt-EJ showed the expected reduction in the 3′ overhang signal upon Apollo deletion (Fig. 2c,d), although leading-end telomere fusions were reduced. Similarly, and consistent with the telomere fusion phenotype, the lack of Lig4 did not affect the reduction in the overhang due to Apollo deletion (Extended Data Fig. 2a,b). Conditional deletion of Ku70 from otherwise wild-type mouse embryonic fibroblasts (MEFs) did not alter the 3′ overhang signal (Fig. 2e,f and Extended Data Fig. 2c,d), indicating that the effect of Ku70 on telomere resection is apparent only when Apollo is absent. Thus, the presence of DNA-PK appears to inhibit Apollo-independent resection at newly replicated leading-end telomeres. Without this inhibition by DNA-PK, 3′ overhang formation and telomere protection are generated independently of Apollo.

Nbs1 represses telomere fusions in the absence of DNA-PK and Apollo
To test how DNA-PK represses resection at telomeres, we targeted Nbs1 with CRISPR-Cas9 (Fig. 3a). Bulk targeting of Nbs1 did not induce telomere fusions in DNA-PK-null MEFs expressing Apollo (Fig. 3b,c). However, when DNA-PKcs and Apollo were both absent, Nbs1 targeting increased the frequency of leading-end telomere fusions (Fig. 3b,c), suggesting that DNA-PK represses resection by MRN. The frequency of telomere fusions induced by CRISPR-Cas9 targeting of Nbs1 in DNA-PKcs-null cells was lower than that observed in DNA-PK-proficient Apollo-knockout cells (Fig. 3b,c). It is possible that the CRISPR-Cas9 targeting was not sufficient to abort MRN-dependent resection in all cells; however, it is also possible that other nucleases promote resection of the blunt telomeres in the absence of DNA-PK, together or independently of MRN. Indeed, depletion of Exo1, which resects telomeres after replication in normal conditions 12 , did not induce telomere fusions (Extended Data Fig. 2e), but caused a significant reduction of the overhang signal when combined with Apollo deletion in DNA-PKcs-deficient cells (Extended Data Fig. 2f,g), suggesting that Exo1 contributes to the resection after it is initiated by MRN.

The iDDR of TRF2 inhibits MRN-CtIP at leading-end telomeres
The observation of DNA-PK preventing resection at blunt telomeres is surprising, given that DNA-PK promotes MRN-CtIP endonuclease activity at DSBs 30 , and it suggests that MRN-CtIP has different access to telomeres than to DSBs. As the iDDR of TRF2 interacts with MRN 6 , we asked whether the iDDR affects MRN-initiated resection at telomeres in cells lacking Apollo. We generated conditional Trf2 F/F cells expressing wild-type Trf2 (WT), the Trf2 ΔiDDR allele (ΔiDDR), the Trf2 F120A allele (F120A), whose product does not bind Apollo 7 , or a version of Trf2 containing both mutations (F120A ΔiDDR) ( Fig. 4a and Extended Data Fig. 3a,b). Importantly, removal of the iDDR completely mitigated the effect of Apollo loss, restoring the 3′ overhang signal and abolishing leading-end telomere fusions (Fig. 4b-e), although Chk2 phosphorylation increased ( Fig. 4a and Extended Data Fig. 3c). In addition, the deletion of the iDDR by itself caused a small reduction in proliferation and an increase of γ-H2AX foci at telomeres, as well as a trend toward greater telomeric overhang signals for reasons that remain to be determined (Fig. 4b,c and Extended Data Fig. 3d-f).
To establish whether the iDDR acts by controlling MRN, we expressed the same Trf2 alleles in cells lacking Nbs1 and examined the telomeric overhang signals and telomere fusions after TRF2 deletion. In accordance with the results presented above, in the Nbs1-proficient an unrelated Rad50-binding module, the inhibitor of the DNA-damage response (iDDR) motif 6 . At dysfunctional telomeres, the iDDR minimizes the accumulation of the DNA-damage factor 53BP1 (ref. 6), but its role at functional telomeres has not been established.
In amniotes, the 5′ exonuclease Apollo (also known as Dclre1b or SNM1B) has evolved a YxLxP motif in its carboxy terminus that allows it to bind to the TRFH domain of TRF2 through the interaction with a region surrounding F120 (refs. 7-11). TRF2-bound Apollo is thought to initiate 5′-end resection at leading-end telomeres to allow subsequent long-range resection by Exo1 (ref. 12). When Apollo is deleted or prevented from binding to TRF2 (for example, when TRF2-F120A is used to complement deletion of TRF2), leading-end telomeres do not regain their normal 3′ overhangs and are vulnerable to end joining [13][14][15][16][17] . Leading-end telomere fusions are abolished by deletion of Ku70, suggesting that they are mediated by c-NHEJ 13,15 . In addition, Apollo deficiency leads to activation of ATM signaling at a subset of telomeres, presumably the leading-end telomeres 16,17 . Because MRN is a requirement for ATM signaling [18][19][20][21] , MRN must be associated with the unprocessed leading-end telomeres.
Therefore, leading-end telomeres deprived of Apollo are a powerful tool to investigate how MRN can activate ATM signaling without initiating resection at blunt telomere ends. Here we show that the answer lies in the iDDR domain of TRF2, which prevents MRN-CtIP-dependent resection in vivo and inhibits its endonuclease activity in vitro while having no effect on MRN exonuclease activity. AlphaFold-Multimer modeling suggests that the inhibition is due to interaction of iDDR with the ATPase domain of Rad50, a mechanism analogous to the inhibition of MRX by Rif2. AlphaFold modeling also suggests that the iDDR-Rad50 interface overlaps with the binding site of CtIP, possibly explaining the inhibition of the CtIP-dependent endonuclease activity of MRN but not its exonuclease activity. We also show that resection at leading-end telomeres lacking Apollo is inhibited by DNA-PK, such that in the absence of DNA-PK, compensatory resection results in protected telomeres that do not undergo fusion. This result indicates that the previously noted dependence of telomere fusions on Ku70 is not related to the role of DNA-PK in c-NHEJ but is due to its ability to prevent resection when Apollo is absent. In agreement, we find that the telomere fusions in cells lacking Apollo are mediated by alt-EJ and are independent of c-NHEJ factor ligase IV (Lig4).

Alt-EJ mediates fusion of telomeres lacking Apollo
In agreement with previous reports that the leading-end telomere fusions do not occur at blunt newly replicated leading-end telomeres in cells lacking Ku70 (refs. 13,15), no telomere fusions were induced upon Cre-mediated deletion of Apollo from cells lacking Ku70, DNA-PKcs or both (Fig. 1a,b and Extended Data Fig. 1a-c). Although these results could be interpreted to mean that the leading-end telomeres are joined by c-NHEJ, this is not the case because the fusions were not dependent on Lig4 (Fig. 1c,d).
We therefore tested the role of alt-EJ in the telomere fusion events. Olaparib inhibition of poly(ADP-ribose) polymerase 1 and 2 (PARP1/2), which mediate the early steps of alt-EJ 22 , resulted in a significant reduction in the telomere fusions induced by Apollo deletion (Fig. 1e,f). Similarly, short hairpin RNAs (shRNAs) targeting Ligase III (Lig3) 23 or DNA polymerase theta (PolQ) caused a significant decrease in telomere fusions in cells lacking Apollo (Fig. 1g,h and Extended Data Fig. 1d,e). These data demonstrate that alt-EJ, rather than c-NHEJ, is a major mode of production of blunt telomere fusions and support previous reports indicating that, unlike c-NHEJ, alt-EJ can engage telomeres despite the presence of TRF2 (refs. 24,25).

DNA-PK loss restores telomere overhang in absence of Apollo
The finding that alt-EJ is responsible for the fusion of blunt telomere ends raised the question of why these fusions are not observed in the  Article https://doi.org/10.1038/s41594-023-01072-x cells, absence of the iDDR mitigated the 3′ overhang defect observed in absence of Apollo. By contrast, the deletion of the iDDR from TR2-F120A cells did not improve the processing of leading-end telomeres when Nbs1 was absent (Fig. 5a,b). Similarly, the leading-end fusions at telomeres lacking Apollo were diminished when the iDDR was removed from TRF2 in Nbs1-proficient cells. By contrast, in Nbs1-deficient cells, removal of the iDDR from TRF2 did not affect the leading-end telomere fusions owing to lack of Apollo recruitment (Fig. 5c,d). These results indicate that the iDDR of TRF2 acts through MRN.

No role for 53BP1
Because 53BP1 blocks the formation of excessively long 3′ overhangs at dysfunctional telomeres 23,31-33 , we tested whether 53BP1 also affected the formation of 3′ overhangs at telomeres lacking Apollo. However, in 53BP1-deficient cells with TRF2 deletion that expressed the F120A allele, there was a reduction of the 3′ overhang signal; for 53BP1-proficient cells, no reduction was observed in cells with the F120A ΔiDDR allele (Extended Data Fig. 4a). Furthermore, 53BP1 status had no effect on the reduction in the 3′ overhang after Apollo deletion (Extended Data Fig. 4b,c), indicating that the iDDR inhibits MRN-CtIP independently of 53BP1.

In vitro inhibition of MRN-CtIP endonuclease by the iDDR
Given that the iDDR inhibits MRN in a 53BP1-independent manner, we asked whether it directly affects MRN activity. MRN is an endonuclease that is activated by phosphorylated CtIP to nick the 5′ strand at protein-blocked DNA ends 34 3 ). Bottom, total telomeric signal (Denaturedtotal(TTAGGG) 3 ). The ssTTAGGG signal was normalized to the total telomeric DNA in the same lane. The normalized −Cre value for each cell line is set to 1, and the +Cre value is given relative to 1. b, Quantification of the relative overhang signal as detected in a for n = 4 independent experiments (indicated by different shades), with mean ± s.d. indicated. c,d, Telomeric overhang assay (c) and quantification (d) on SV40LT-immortalized Apollo F/F MEFs transduced with empty vector or shRNAs targeting Lig3 or PolQ and 108 h after Cre-mediated deletion of endogenous Apollo. n = 4 independent experiments. Data are presented as mean ± s.d. e,f, Telomeric overhang assay (e) and quantification (f) of Ku70 F/+ (n = 3) and two independent Ku70 F/F (n = 5) MEFs 96 h after Hit & Run Cre-mediated deletion of endogenous Ku70. n = 3 independent experiments. Data are presented as mean ± s.d. Statistical analysis by two-way ANOVA (b,f) and two-tailed unpaired t-test (d). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. See also Extended Data Figs. 1 and 2.
Article https://doi.org/10.1038/s41594-023-01072-x that is bound to DNA-PK. In this setting, MRN generates an endonucleolytic product of approximately 45 nt 30 . Indeed, purified TRF2 inhibited the formation of this product in a dose-dependent manner ( Fig. 6a-c). By contrast, the ΔiDDR motif TRF2 mutant did not have this effect, indicating that the iDDR is required for the inhibition of MRN endonuclease activity. MRN also exhibits a 3′-5′ exonuclease activity that is independent of CtIP 37,38 . In contrast with its role in inhibiting the endonuclease activity of MRN, the addition of WT or ΔiDDR TRF2 did not inhibit the exonuclease activity of MRN ( Fig. 6d-f). These results show that the iDDR has a specific effect on the endonuclease activity of MRN-CtIP but does not interfere with the DNA-end-binding or exonuclease activity of MRN. The effect of the iDDR is similar to that of Rif2 in budding yeast, which inhibits MRX endo-but not exonuclease activity 3,4 .

AlphaFold predicts the interaction between iDDR and RAD50
It has previously been shown that the iDDR of TRF2 pulls down Rad50 in co-immunoprecipitation experiments 6 . However, these experiments did not determine which subunit of MRN interacts with TRF2. We investigated the potential interactions of the human iDDR with MRN subunits using AlphaFold-Multimer 39 . AlphaFold-Multimer predicted an interaction between the iDDR and the globular ATPase domain of RAD50 ( Fig. 7a-c). The iDDR was predicted with high confidence (pLDDT value) in the top five ranked models (Fig. 7b). No interactions were predicted with MRE11, NBS1 or CtIP, and the iDDR was predicted with low confidence (predicted local distance difference test (pLDDT) value) in these models, suggesting that RAD50 binding orders the domain (Extended Data Fig. 5a). The predicted interaction of the iDDR with RAD50 is highly conserved in metazoans, including both vertebrate TRF2 and invertebrate TRF proteins (Extended Data Fig. 5b,c).
The telomere-binding proteins Rif2 and Taz1 of budding and fission yeast, respectively (Extended Data Fig. 6a), have recently been proposed to bind to Rad50 via their MRN(X)-inhibitory (MIN) domains (also called the BAT domain in Rif2) 3,5 . AlphaFold-Multimer predicted that Rif2 and Taz1 bind their cognate Rad50 proteins using the same interface in the ATPase domain used by the iDDR (Extended Data Fig. 6b,c). Despite this analogous binding site, the MIN domains of Rif2 and Taz1 do not show sequence homology and are predicted to be structurally different from the iDDR domain. In fact, the iDDR consists of a cluster of basic residues followed by acidic residues, which create a characteristic drop in local isoelectric point (pI) (Extended Data Fig. 6d,e; ref. 10). On the contrary, the MIN domains of Rif2 and Taz1 have a switched order of basic and acidic residues. This inversion is reflected in their predicted orientation on the surface of Rad50, with the N-terminal residues of the MIN domain pointing toward the coiled-coils, whereas in the iDDR, the C terminus is pointing towards the coiled-coils (Extended Data Fig. 6d). This inverted orientation suggests that these motifs may have evolved independently and that their similar function is due to convergent evolution (Extended Data Fig. 6f), and therefore that an MRN-inhibitory module is a highly selected and ancient feature of telomeres.

The iDDR is predicted to compete with CtIP for RAD50 binding
Data from budding yeast suggest that Rif2 inhibits the MRX endonuclease activity by outcompeting Sae2 at its Rad50-binding site 4,34 . Therefore, we used AlphaFold-Multimer to query whether the iDDR might similarly affect the interactions between CtIP and the MRN complex. The 'Sae2-like' C terminus of CtIP was predicted with high confidence to interact with the same surface of RAD50 that the iDDR of TRF2 also interacts with (Fig. 7d-f). Interestingly, this is the region where the RAD50S separation-of-function alterations cluster, which are deficient in CtIP-dependent endonuclease, but not in CtIP-independent exonuclease, activity [40][41][42] . This region of CtIP is highly conserved across metazoans, consistent with its potential role in regulating the MRN complex (Extended Data Fig. 7a), which is also highly conserved. Additionally, Sae2 and Ctp1 (the fission yeast homolog of CtIP) were predicted to bind to their respective Rad50 proteins in a similar manner (Extended Data Fig. 7b-d). Collectively, the data suggest that the iDDR of TRF2 can prevent CtIP from binding to RAD50, blocking the activation of MRN endonuclease activity (Fig. 7g).

Discussion
These results reveal the ability of the TRF2 iDDR and DNA-PK to independently inhibit all MRN-mediated resection at blunt-ended telomeres, although they resemble DSBs (Fig. 7h). When telomeres are replicated in the absence of Apollo, the leading-strand DNA-synthesis products lack the protective 3′ overhang and become vulnerable to alt-EJ. These fusion events depend on the ability of DNA-PK to prevent long-range resection. However, it has been puzzling why MRN-CtIP does not promote DNA end processing by removing DNA-PK from these telomeres, as it does at DSBs. Here we show that the iDDR of TRF2 blocks MRN-CtIP endonuclease activity in vitro and prevents MRN from acting at blunt leading-end telomeres in vivo. Our structure-prediction data suggest that the iDDR acts by preventing the formation of the MRN-CtIP complex. Because MRN-CtIP is not active at telomeres bearing the iDDR, DNA-PK can persist at these ends, leading to complete absence of resection. When the iDDR is removed, MRN-CtIP is active and can cleave DNA-PK off the telomere ends. Interestingly, when DNA-PK is absent, MRN can still engage the ends, leading to long-range resection, although the endonucleolytic activity of MRN-CtIP should be held in check by the iDDR. In both scenarios, resection is activated. W T Paradoxically, this leads to the formation of a protective 3′ overhang independently of Apollo. Importantly, we show that the inhibition of MRN endonuclease activity at eukaryotic telomeres is conserved, owing in part to convergent evolution. The iDDR is an ancient feature of the TRF subunit of shelterin. It was already present when metazoans emerged, prior to the gene duplication that created TRF1 and TRF2 and long before Apollo-TRF2 binding evolved 10 . By contrast, the genes involved in the ability of mammalian iDDR to minimize the accumulation of 53BP1 at dysfunctional telomeres, including BRCC3, RNF8, RNF168 and 53BP1 itself, are not conserved in all metazoans 43 . This argues that the iDDR evolved the ability to affect 53BP1 later as a secondary feature. The argument that inhibiting MRN is the original function of the iDDR is strengthened by the finding that the MRN/X inhibitory modules (called MIN or BAT) are found at telomeres in fungi. Remarkably, these fungal MIN modules interact with the same part of Rad50 that the iDDR interacts with. Because the MIN/BAT have no sequence similarity to the iDDR, are found in proteins that are not orthologous to TRF2 (for example, Rif2) and bind in an inverse orientation compared with the iDDR, we infer that the inhibition of MRN at telomeres represents an example of convergent evolution. Such convergent evolution speaks to the strong selective pressure to preserve the inhibition of MRN at telomeres, despite major changes in the  the presence of a 5′ 32 P-labeled DNA substrate. The gel is a representative example of two independent replicates. The blue arrow indicates the primary endonucleolytic cleavage product (~45 nt away from the end). d, Schematic of the MRN exonuclease assay. e,f, MRN endonuclease assay gel (e) and quantification (f) of n = 4 independent replicates in the presence of WT or ΔiDDR TRF2. Data are presented as mean ± s.d. Statistics by two-tailed unpaired t-test assuming a Gaussian distribution. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. Article https://doi.org/10.1038/s41594-023-01072-x telomere-associated proteins. Why the endonuclease activity of MRN needs to be inhibited at telomeres and/or whether iDDR contributes to direct inhibition of ATM activation at telomeres, as shown in yeast for the MIN/BAT 5 and suggested by high Chk2 phosphorylation in the presence of Trf2 F120A ΔiDDR , remains to be determined.
DNA-PK blocks the access of the long-range nucleases Exo1 and DNA2/BLM to DNA ends in vitro and suppress DNA end resection in vivo 30,44,45 . However, counterintuitively, the presence of DNA-PK also stimulates MRN-CtIP endonuclease activity in vitro 30 . These results are consistent with the essential role of MRN-CtIP in initiating resection    Article https://doi.org/10.1038/s41594-023-01072-x at ends blocked by DNA-PK, such as single-ended DSBs resulting from replication fork collapse 46 . Our data demonstrate that the iDDR inhibits MRN-CtIP from acting as an endonuclease, effectively blocking all MRN-CtIP-mediated resection at the DNA-PK-bound telomeres. Importantly, DNA-PK, which localizes at telomeres in normal conditions 47,48 , allows Apollo processing. However, in the absence of DNA-PK, telomeres can be processed independently of Apollo. This resection seems to at least partially depend on Nbs1. Therefore, we propose that MRN can promote the resection of blunt-ended telomeres when DNA-PK is absent, possibly owing to its ability to load other resection factors at the ends, as MRN itself has no 5′ exonuclease activity. We further propose that this loading activity of MRN does not require CtIP, explaining why it is not repressed by the iDDR. In agreement, MRN promotes loading of Exo1 and other long-range resection factors like BLM/DNA2 at DNA ends in vitro and enhances the processivity of Exo1 resection in the presence of RPA [49][50][51] . Consistent with this, Exo1 is involved in overhang generation in the absence of DNA-PKcs and Apollo, although probably only after the first step of resection is initiated. Whether other resection regulators in addition to MRN and Exo1 gain access to the telomere ends that lack DNA-PK remains to be determined. The data show that the leading-end telomeres in cells lacking Apollo become joined by PolQ-dependent alt-EJ (also called theta-mediated end joining, TMEJ) 52,53 . This result was unexpected because leading-end telomeres lacking Apollo are presumed to be blunt, whereas alt-EJ joins DSBs with 3′ overhangs. In alt-EJ, PolQ uses microhomologies to anneal 3′ overhangs and then executes templated DNA synthesis to create the substrate for Lig3-mediated ligation. Because PolQ is auto-inhibited at DNA ends with short or no 3′ overhangs 54 , it remains to be determined how it acts on the blunt leading-end telomeres. Conversely, it is prudent to ask why alt-EJ does not act on telomeres that have a 3′ overhang that can anneal to form 2 base pairs every 6 nt. Most likely, this is owing to the presence of the POT1 (Pot1a and Pot1b in mice) on the ssDNA, because alt-EJ has been shown to act at telomeres that lack POT1 (refs. 25,55). Finally, it is curious that the blunt leading-end telomeres formed in the absence of Apollo are not processed by c-NHEJ, despite their interaction with DNA-PK. Perhaps TRF2 inhibits c-NHEJ in S and G2 through Rap1 (ref. 56) or by binding to Ku70/80 (ref. 57). Such S-or G2-specific mechanisms of c-NHEJ repression may also account for the lack of c-NHEJ repair of telomere-internal DSBs created in S or G2 (ref. 24).

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-023-01072-x. Fig. 7 | The iDDR of TRF2 is predicted to compete with CtIP for binding to RAD50. a, Predicted aligned error (PAE) plot from AlphaFold-Multimer modeling of human RAD50-TRF2. Representative of five ranked models generated with default parameters. b, pLDDT plot showing per-residue confidence score across human TRF2 from the ranked RAD50-TRF2 models. c, Predicted structure of a human RAD50 dimer with a dimer of human TRF2. The RAD50 coiled-coils were truncated for the dimer model. Only the iDDR of TRF2 is shown. d, PAE plot from the AlphaFold-Multimer modeling of human RAD50-CtIP. Representative of five ranked models generated with default parameters. e, pLDDT plot showing per-residue confidence score across human CtIP from the ranked RAD50-CtIP models. f, Predicted structure of a human RAD50 dimer with the Sae2-like domain of human CtIP. The RAD50-CtIP monomer model was superimposed on the predicted structure of an RAD50 dimer from c. g, Model for inhibition of MRN endonuclease activity by the TRF2 iDDR. TRF2 competes with CtIP for binding to RAD50, which stimulates the endonuclease-active state of MRN. h, Model for leading-end telomere processing and protection mediated by TRF2 and DNA-PK. TRF2 promotes the 5′-resection of the leading-end telomere by recruiting Apollo. In the absence of Apollo, either owing to Apollo deletion or the lack of recruitment due to TRF2-F120A, the newly replicated leading-end telomere ends cannot be resected and undergo fusion mediated by alt-NHEJ. In the absence of Apollo and the iDDR domain of TRF2, MRN-CtIP initiates resection at DNA-PKbound leading-end telomeres, leading to telomere protection and the absence of fusions. When both DNA-PK and Apollo are absent, MRN can promote the resection of the free DNA ends, even in the presence of the TRF2 iDDR domain. See also Extended Data Figs. 5-7.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.

Immunoblotting
Cells were lysed in 2× Laemmli buffer at 5 × 10 3 cells μl −1 , and the lysate was denatured for 10 min at 95 °C before shearing with an insulin needle. Lysate equivalent to 1 × 10 5 cells was resolved using SDS-PAGE and transferred to a nitrocellulose membrane. Western blot was performed Article https://doi.org/10.1038/s41594-023-01072-x performed at 1 × 1 binning and multiple 0.2-μm Z-stacks using SoftWoRx software; images were deconvolved, and 2D-maximum intensity projection images were obtained using SoftWoRx software. Chromatid and chromosome-type fusions were analyzed using Fiji (1.0) software 63 after arbitrary assignment of red for both (TTAGGG) 3 -probes and green for both (CCCTAA) 3 -probes.
Semi-automated analysis and quantification of colocalization was performed using CellProfiler 64 with the following pipeline: image cropping to remove edge artifacts due to deconvolution; channel intensity rescaling to cover the full histogram range value; 'speckle features enhancement' to increase detection sensitivity and remove background or artifact aggregates; 'channel-wise primary objects identification' to detect individual nuclei and individual foci; and correlation of the foci coordinates in the different channels and with the respective nuclei to define colocalization events. Nuclei with fewer than 10 detected PNA foci were discarded.

RNA extraction and qRT-PCR
RNA was extracted from 1 × 10 6 cells using the RNeasy Mini Kit (Qiagen). Then, 500 ng of RNA was reverse transcribed using the First Strand cDNA Synthesis Kit (Thermo Scientific). SYBR Green PCR Master Mix (Applied Biosystems) was used for quantitative PCR. In-gel analysis of single-stranded telomeric DNA Mouse telomeric DNA was analyzed on Clamped homogenous electric field (CHEF) gels, as described previously 62 4 probe in Church mix (0.5 M sodium phosphate buffer pH 7.2, 1 mM EDTA, 7% SDS, 1% BSA). After three washes in 4× SSC and one in 4× SSC/0.1% SDS at 55 °C, the gel was exposed for 1 or 2 d, and the single-stranded telomere signal was captured by Typhoon Phospho-Imager. For the acquisition of the total telomere signal, the gel was denatured with 1.5 M NaCl/0.5 M NaOH for 1 h, neutralized with two washes of 1 h each in 0.5 M Tris-HCl pH 7.0/3 M NaCl, pre-hybridized for 30 min at 55 °C in Church mix and hybridized overnight at 55 °C with the same probe. The denatured gel was washed and exposed as described before. Quantification of the signals in each lane was done using ImageQuant software. After subtraction of the background, the single-stranded signal was normalized to the total telomeric DNA signal in the same lane. The indicated control value was set to 1, and all the other values were given as a percentage of it.
Protein purification and nuclease assay TRF2. Human TRF2 and TRF2-ΔiDDR were cloned into a modified pFastbac vector with a His 6 -MBP tag and a 3C protease cleavage site. Proteins were expressed in insect cells grown for 72 h after infection with baculovirus, collected, frozen in liquid nitrogen and stored at −80 °C. Cell pellets were thawed and homogenized in lysis buffer (40 mM Tris pH 8, 500 mM NaCl, 0.5 mM TCEP, 10% glycerol, 0.1% Tween-20, 1 mM PMSF, protease inhibitor cocktail (Roche)). Cells were pelleted at 18,000g and incubated with 1 ml Ni-NTA resin for 1 h. The resin was washed with buffer A (40mM Tris pH 8, 100 mM NaCl, 0.5mM TCEP, 10% glycerol) and proteins were eluted in buffer A and 200 mM imidazole. Proteins were incubated with 3C protease overnight at 8 °C and injected on a Hi-Trap Heparin column (Cytiva). After being washed extensively with buffer A, proteins were eluted in buffer B (40 mM Tris pH 8, 1 M NaCl, 0.5 mM TCEP, 10% glycerol). The most concentrated fractions were injected into a Superose 6 column (Cytiva) equilibrated in buffer A. Fractions were analyzed on SDS-PAGE for purity, pooled, concentrated to 1 mg ml −1 , aliquoted and flash-frozen in liquid nitrogen.
Endonuclease assay. The MRN-CtIP endonuclease assay was performed as described 30 , with the following modifications: MRN was preincubated with the noted concentrations of TRF2 or TRF2-ΔiDDR for 10 min on ice before addition of a separately prepared mixture containing DNA-PK, 5′ 32 P-labeled substrate DNA and NU7441. Endonuclease activity was initiated by addition of the remaining components, and CtIP and reaction products were assessed after 1 h at 37 °C on polyacrylamide gels containing 12% polyacrylamide, 20% formamide and 6M urea, and were imaged by phosphorimager.

AlphaFold-Multimer and evolutionary and structural analysis
AlphaFold-Multimer (v2.1.0) 39 was run locally on a GPU workstation using default parameters. A custom script (modified from AlphaFold Colab) was used to extract PAE and pLDDT information from the resulting pickle files, and structures were analyzed in PyMol (Schrodinger) and ChimeraX (UCSF). TRF2, TRF, CtIP and Rad50 protein sequences were obtained from PSI-BLAST searches of the non-redundant protein sequences (nr) database against human sequences (https://blast. ncbi.nlm.nih.gov/). Alignment of the sequences was performed using MUSCLE in SnapGene and formatted in Jalview. The CtIP alignment was visualized using the NCBI MSA Viewer (v1.22.2) and colored according to 'conservation.' Article https://doi.org/10.1038/s41594-023-01072-x

Quantification and statistical analysis
Quantification and statistical analysis were performed using Microsoft Excel and GraphPad, respectively, over three or more independent experiments, as indicated. For each CO-FISH analysis, at least ten metaphases per condition per experiment were scored. For in-gel analysis of single-stranded telomeric DNA, the normalized overhang signals were expressed for each cell line independently. Significance was assessed by calculating the P value using Kruskal-Wallis one-way ANOVA without Gaussian distribution assumption (to compare telomere fusions), unpaired t-test without Gaussian distribution assumption (to compare mRNA expression and telomere overhang signal in a single cell line) and two-way ANOVA without Gaussian distribution assumption (to compare telomere overhang signal from more than one cell line). P values ≤ 0.05 were considered statistically significant.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
All data generated or analyzed during this study are included in this article. Source data are provided with this paper. All raw images are available on FigShare (https://doi.org/10.6084/m9.figshare.23650509) or from the corresponding authors upon reasonable request. Source data are provided with this paper.