IntroductionTelomeres maintain genome stability by protecting native chromosome ends from the DNA damage checkpoint and repair pathways that act on double-strand breaks (DSBs)1,2. In most eukaryotes, telomeric DNA consists of tandemly repeated sequences bound by high-affinity telomere proteins. This repetitive protein occupancy is a distinctive feature of telomeres and is critical for their functions. Short telomeres, with less DNA-bound telomere proteins, are more susceptible to fusions, checkpoint activation, and telomere loss3,4,5. They are also more likely to be elongated by telomerase, which restores full protection6,7. The strong length-dependence suggests that, within telomeres, end-distal and end-proximal telomere proteins jointly contribute to telomere functions. However, it is unclear whether their contributions are equivalent and/or if end-proximal telomere proteins play specific roles due to their position at the forefront to encounter the DNA damage checkpoint and repair proteins that bind to DNA ends8,9.The NHEJ Ku70-Ku80 heterodimer (Yku70-Yku80 in S. cerevisiae, and referred to hereafter as Ku) is the first factor to bind DSBs ends in all eukaryotes10. Ku has a ring-shaped structure that encircles DNA double-stranded ends with high-affinity, but without sequence specificity10,11,12,13,14,15. Ku protects DSB ends from resection initiation and recruits the NHEJ-specific ligase 4 along with its associated protein partners. Together, they mediate end synapsis followed by end ligation, a process that requires Ku to translocate inward by approximately 25 base pairs on the DNA, thereby making the ends accessible to the ligase10,11,12,13,14,15,16,17,18. Paradoxically, Ku constitutively binds telomere ends and protects them from uncontrolled resection without engaging NHEJ, a feature conserved across yeast, mammals and plants19,20,21,22,23,24,25,26,27,28,29,30. How Ku’s roles in promoting NHEJ are so effectively suppressed at telomeres remains unknown. It is also unclear whether Ku and end-proximal telomere proteins compete or cooperate for DNA binding at telomere ends.In S. cerevisiae, the protein Rap1 directly binds telomere repeats with high affinity and protects telomeres against NHEJ-dependent fusions and uncontrolled end resection through multiple mechanisms4,24,31,32,33,34,35,36,37,38. The reliance on multiple pathways acting in synergy is a conserved feature of telomere end protection23,30,39,40,41,42,43. Each can provide significant, non-redundant and potentially similar levels of protection4,37,38 (Supplementary Fig. 1). Understanding the molecular bases of telomere stability requires identifying and deciphering each of these pathways. Here, we uncover a mechanism of telomere end protection that relies specifically on the end-proximal, telomere-bound Rap1 protein. This mechanism operates independently of telomere length and acts by restricting Ku’s inward translocation, thereby preventing promiscuous NHEJ repair at telomeres.ResultsNHEJ inhibition at a DSB by a single Rap1 moleculeLimiting Rap1 binding at telomere ends leads to frequent telomere fusions35,44,45 (Supplementary Fig. 2). To assess the potential contribution of the end-proximal telomere-bound Rap1 protein to telomere protection, we investigated whether a single Rap1 molecule, when bound close to a DSB, could significantly antagonise NHEJ repair. Within S. cerevisiae telomere sequences, the motif5’GGTGTGTGGGTGTG3’ matches the Rap1 consensus binding site32,46 (Fig. 1a and Supplementary Fig. 1A). Rap1 binds this sequence with subnanomolar affinity in vitro34,47 (Supplementary Fig. 3). We therefore tested whether a single copy of this Rap1 site could inhibit NHEJ in an assay where survival following continuous expression of a site-specific endonuclease serves as a proxy for NHEJ repair efficiency33,37 (Fig. 1b).Fig. 1: Inhibition of NHEJ by a single Rap1 site at a DSB.a Rap1 binding site at telomeres: Example of an S. cerevisiae telomere sequence (Rap1 sites in teal). b I-SceI NHEJ assay. Two inverted I-SceI sites were inserted at the URA3 locus. Survivors to continuous I-SceI expression eliminate fuse distal ends33,37. A Rap1 site was inserted near one I-SceI site. c NHEJ efficiency in wild-type (WT) and rif2∆ sir4∆ mutant cells with or without a Rap1 site at the break. Means from replicates, details and statistical analysis in Supplementary Table 1. d Distance between Rap1 and the break, (-) no insert, ( + ) 1 Rap1 site. d Sequences of the tested sites. Consensus from ref. 46 (R: G/A, K: G/T, Y: T/C, N: any base). Bases identical to the telomeric Rap1 site (teal); differing bases (black). Alternative sites from native telomere sequences (Supplementary Fig. 4). Other sites from HMR-E silencer and TEF2, RPS17, RPS11 and RPL14 promoters46. Negative controls: a mutated non-binding site47 and a 14-bp random sequence. e Divergent Rap1 sites from telomeres are less effective at blocking NHEJ. Sites 11 bp from the break. Cells lacking Rif2 and Sir4. Means from replicates, details and statistical analysis in Supplementary Table 1. f ChIP analysis of Yku80 and Rap1 binding at a DSB following 45 and 105 min of I-SceI induction. Data compare a Rap1 site (5’GGTGTGTGGGTGTG3’, teal) to a mutated site (5’GGAGTGTGGGAGTG3’, light grey). Cells were arrested in G1 to limit end resection. Quantification represents immunoprecipitated DNA (IP) relative to the input DNA (IN). Means from 2 independent cell cultures or more. OGG1 was used as a control locus. Source data are provided as a Source Data file.Full size imageThe insertion of a Rap1 site positioned 11 bp from the break led to significant inhibition of NHEJ repair (d = 11 bp, Fig. 1C). The strength of this inhibition was comparable in the presence or absence of Rif2 and Sir4, two Rap1 co-factors previously shown to inhibit NHEJ at telomeres4,33,36,37,38. This rules out any contribution of Rif2 and Sir4 in the NHEJ inhibition mediated by an isolated Rap1 molecule and supports the hypothesis that a previously unidentified mechanism must be at play. A Rap1 site located more than 30 bp from the end had no impact on NHEJ repair, indicating that the action of Rap1 on NHEJ is short-range (Fig. 1c, statistical analysis in Supplementary Table 1). Rap1 sites that diverge from the consensus Rap1 telomere site were less effective at blocking NHEJ when inserted at a break site (Fig. 1d, e and Supplementary Fig. 4; statistical analysis in Supplementary Table 1). Together, these results show that a single Rap1 molecule, bound to DNA ends, can significantly protect against fusion by NHEJ.Rap1 and Ku co-binding at DNA endsNext, we tested if a single Rap1 block NHEJ at a DSB by preventing Ku binding. Using a ChIP approach, we found that Rap1 binding 11 bp from the break did not inhibit Ku’s recruitment (Fig. 1f), which is consistent with the co-presence of Rap1 and Ku at telomeres in vivo. We next explored the conditions under which Rap1 and Ku could simultaneously bind the same DNA end in vitro. Binding experiments using purified proteins and Electrophoretic Mobility Shift Assays (EMSA) show that, individually, Rap1 and Ku bound specifically to a 17-bp DNA duplex containing a single Rap1 site and only a 1 bp extension downstream of the site (Fig. 2a and Supplementary Fig. 5A, B). However, when incubated together with the same DNA duplex, their binding appeared mutually exclusive (Fig. 2a and Supplementary Fig. 5C), suggesting that Rap1 binding on this DNA sterically hinders Ku’s ability to bind to the DNA end.Fig. 2: Ku binding on Rap1-DNA complexes.a Binding of Ku and Rap1 to a short DNA duplex containing a Rap1 site and a 1 bp downstream extension (1 nM). Unlabelled competitor DNA: (i) short linear duplex with a Rap1 site and a 1 bp downstream extension (titrates Rap1 and Ku, 200 nM), (ii) short linear duplex with a mutated site (mutated in Fig. 1) and a 1 bp downstream extension (titrates Ku only, 200 nM), circular plasmid with 16 Rap1 sites in tandem47 (titrates Rap1 only, 20 nM). (R) represents Rap1, (K) represents Ku. The experiment repeated three times. b Binding of Ku and Rap1 to DNA duplexes with downstream extensions ranging from 1 to 7 bp (1 nM). d refers number of base pairs separating the edge of the Rap1 site from the duplex end. The experiment repeated more than three times. c Representative titration of Ku binding to DNA duplexes with downstream extensions of 1 to 6 bp (1 nM) in the presence of excess Rap1 at a fixed concentration of 80 nM. d Representative titration of Ku binding to DNA in the absence of Rap1. The experiment repeated twice or more. e Quantified EMSA data showing Ku association on Rap1-bound DNA with a downstream extension of 5 bp (1 nM), as in (C). Means from 3 independent experiments. f Interpretative schematic illustrating the mutual exclusion or co-binding of Rap1 and Ku on short duplex DNA with a 1 or 5 bp extension downstream of the Rap1 site. Source data are provided as a Source Data file.Full size imageWe then investigated the length of DNA downstream of the Rap1 binding site that would be sufficient to support Ku binding to Rap1-bound DNA duplexes. DNA extensions of 4 bp or more shifted the equilibrium towards the formation of larger complexes, accommodating both Rap1 and Ku on the same DNA molecule (d ≥ 4 bp, Fig. 2b and Supplementary Fig. 5D). This suggests that these few base pairs downstream of Rap1 are sufficient to enable Ku binding to Rap1-DNA complexes, forming ternary complexes. When we progressively increased the concentration of both Rap1 and Ku (up to 80 nM each), these ternary complexes became predominant (Supplementary Fig. 6a). To better estimate the binding affinity of Ku on the Rap1-bound DNA ends, we fixed Rap1 concentration at 80 nM and increased progressively Ku concentration (Fig. 2c, to compare to Ku binding on free DNA shown in Fig. 2d). We observed that the association between Ku and Rap1-DNA complexes with a 5 bp extension predominates at Ku concentrations above 2 nM (Fig. 2c–e), suggesting an affinity in the low nanomolar range or less, close to the affinity of Ku for free DNA ends (Fig. 2d and Supplementary Fig. 3). On Rap1-bound DNA with shorter extensions of 3 bp or less, we note that Ku may still transiently bind at higher protein concentrations but dissociates during migration in the gel (Fig. 2b, c). Thus, the DNA length downstream of the Rap1 binding site is a key factor in determining Ku association on Rap1-DNA complexes. To corroborate these findings, we employed Mass Photometry. This method confirmed that Rap1 antagonises Ku binding on DNA duplexes with 2 bp downstream of the Rap1 site, but not on those with 6 bp, which is sufficient to allow Ku binding on Rap1-DNA complexes (Supplementary Fig. 6B).The DNA binding domain of Rap1 (Rap1DBD) comprises two Myb-like Helix-Turn-Helix subdomains (Myb1 and Myb2), each engaging with a 5-6 bp hemi-site32. Mutations in either hemi-site impaired both Rap1’s interaction with the DNA duplex and the formation of Rap1-DNA-Ku complexes (Supplementary Fig. 7A), indicating that Rap1 engages with both hemi-sites in these complexes and ruling out the possibility of a partial Rap1 interaction that might leave additional free DNA for Ku binding48. 3’ single-stranded overhangs of up to 12 nucleotides did not prevent Ku from binding to Rap1-DNA complexes (Supplementary Fig. 7B), suggesting that Ku is not obstructed by the short 3’ overhangs found at native yeast telomeres31. Collectively, these findings show that Ku associates with Rap1-bound DNA ends when there is a double-stranded DNA extension of 4 bp or more downstream of Rap1. With shorter DNA extensions, Rap1 and Ku binding tend to become mutually exclusive (Fig. 2f).Structural analysis of Rap1 ability to accommodate Ku at DNA endsThe apparent threshold length for Ku binding to Rap1-bound DNA ends (4 bp, Fig. 2) is unexpectedly short compared to the minimal number of base pairs typically bound by Ku, as seen in resolved structures (> 10 bp)10,11,13,14. To address this apparent paradox, we utilised molecular modelling and cryo-electron microscopy. Initially, we built a model of the ternary Rap1-DNA-Ku complex using the crystal structures of S. cerevisiae Ku13 and of S. cerevisiae Rap1DBD bound to a DNA duplex49. In this model, the 21 bp DNA duplex includes a 5 bp extension downstream of the 14 bp Rap1 site (see Methods). To perform cryo-EM, we incubated the same 21 bp DNA duplex with Rap1, followed by Ku and then purified and vitrified the complexes. The collected data allowed us to obtain a map of the ternary Rap1-DNA-Ku complex at an overall resolution of 3.1 Å. The initial model was further refined through molecular dynamics flexible fitting constrained by the cryo-EM map (Fig. 3a, b, Supplementary Figs. 8, 9, 10A, 11, 12A, Supplementary Table 2 and Supplementary Movie 1).Fig. 3: Structural analysis of Rap1’s ability to restrict Ku’s inward translocation on DNA.Cryo-EM maps and corresponding models in cartoon representation of the Rap1DBD-DNA-Ku ternary complex (a, b) and the DNA-Ku binary complex (c, d). Yku70 is shown in gold, Yku80 in cornflower blue, Rap1DBD in green, the G-rich DNA strand in light grey and the C-rich DNA strand in grey. Nucleotides within 4.0 Å of Ku are highlighted in red on both strands. a, c The cryo-EM maps are shown in mesh and coloured according to atom proximity in the refined structural model. At the thresholds used, electron density was missing for residues 194–211 of Yku70, 165–171, 287–301, 462–466 of Yku80 and residues 429–433, 480–506 and 598–601 of Rap1DBD in the Rap1DBD-DNA-Ku complex (a) and for residues 200–211 of Yku70, 95–104, 165–171, 263–270, 286–300 and 579–587 of Yku80 in the DNA-Ku complex (c).Full size imageThe resolved portion of Rap1 is restricted to its DNA binding domain, which aligns closely with the Rap1DBD-DNA crystal structure49 (Supplementary Table 2). Rap1 engages the DNA major groove via its two Myb-like subdomains, along with a loop at the C-terminal region of Myb2 (residues 565–601), which wraps around the DNA duplex and contacts the Myb1 subdomain, effectively clamping the molecule32,49 (Supplementary Fig. 11 and Supplementary Movie 1). The Ku structure in the ternary complex also closely resembles the S. cerevisiae Ku crystal structure13 (Supplementary Table 2 and Supplementary Fig. 13A). Ku’s interaction with DNA involves five base pairs at the DNA extremity (the extension downstream of the Rap1 site), which are engaged into the Ku beta-bridge (handle), in addition to seven more base pairs that also interact with Rap1 Myb1 subdomain (Fig. 3a, b, Ku-interacting nucleotides in red). This is permitted because the two proteins bind different sides of the same DNA double helix (Supplementary Fig. 14) and display a remarkable shape compatibility (Fig. 3a, b and Supplementary Fig. 11). Beyond the beta-bridge, Ku does not contact the DNA (Fig. 3a, b and Supplementary Fig. 11). In total, Ku engagement with 13 bp of DNA is consistent with previous estimates for Ku binding to duplex DNA10,11,13,14.Although Rap1 and Ku bind very closely to each other, contacts between Rap1 and Ku (distance