Main

DNA double-strand breaks (DSBs) pose a substantial threat to genome stability, as they disrupt chromosome integrity. Repair mechanisms, such as non-homologous end joining (NHEJ) and homologous recombination, are essential for restoring chromosome continuity by directly rejoining DSB-ends or using a donor homologous template1. However, before repair processes can occur, it is imperative to bring DSB-ends together, a task unlikely to be achieved through passive diffusion2. Instead, active DSB end-tethering mechanisms have been identified, and represent a critical step in preventing joining or recombination events between unrelated chromosome loci.

The mechanisms that facilitate DSB end-tethering were initially characterized in the budding yeast Saccharomyces cerevisiae3,4,5. The MRXMRN complex (Mre11–Rad50–Xrs2NBS1) is rapidly recruited to DSB-ends and plays an early role in end-tethering3,5,6. MRN has been proposed to serve a similar function in humans, preventing translocations7,8,9. In yeast, MRX nuclease activity is dispensable for DSB end-tethering. On the other hand, the Zn-hook ___domain and ATPase activity of Rad50 are essential, suggesting a physical bridging mechanism by MRX dimers3. In contrast, during later stages of repair, DSB end-tethering requires Exo1 exonuclease activity to reveal single-stranded DNA (ssDNA)4. However, the proteins responsible for physical bridging of DSB-ends during these late stages of repair remain unidentified.

Recent theoretical research has proposed a role for DNA loop extrusion in the tethering of DSB-ends2. Loop extrusion, a property associated with structural maintenance of chromosomes (SMC) family complexes, has emerged as a conserved mechanism for folding the genome10. Among the various SMC complexes, cohesin (comprising Smc1, Smc3, Mcd1Scc1 and Scc3STAG1/2) and the Smc5/6 complex are recruited to DNA-damage sites in both yeast and mammals11,12,13,14. In yeast, cohesin loading to DSBs involves a number of factors, including the cohesin loader Scc2/Scc4, Smc5/6 and DNA damage factors such as MRX, γH2A, Tel1ATM and Mec1ATR (refs. 13,14,15,16,17). DSBs also enrich cohesin throughout the genome18,19,20, contributing to tightening of sister chromatid cohesion (SCC)18,19,21,22,23, locally restricting homology search24 and aiding in DNA-damage checkpoint establishment20.

Given the involvement of cohesin and SMC5/6 in DSB response25,26, cohesin’s ability to bridge DNA molecules in vitro27, and gross chromosomal rearrangements in cohesin and SMC5/6 mutants14,28, we hypothesized that cohesin and/or Smc5/6 play a critical role in DSB repair by maintaining proximity between DSB-ends.

In this Article we combine genetic and live-microscopy-based approaches to demonstrate a cohesin-dependent DSB end-tethering mechanism involving Exo1 and Smc5/6. Furthermore, we show that cohesin compacts DSB-adjacent chromatin, beyond the compaction observed in G2/M cells. We expose oligomerization as a key mechanism for both MRX- and cohesin-dependent tethering through both disruption of protein–protein interactions in response to hexanediol treatment, and genetic loss of function mutants. Specifically, disruption of cohesin oligomerization through mutation in the Mcd1SCC1 subunit maintains compaction in the vicinity of DSB, but prevents DSB end-tethering. Finally, our real-time microfluidic assay demonstrates that cohesin is essential for efficient repair of DSBs, through its end-tethering capacity.

Results

Cohesin tethers DSB-ends

To assess the requirement of cohesin in tethering DSB-ends, we developed a microscopy-based assay in which LacO and TetO repeats were positioned either side of the endogenous Ho endonuclease cleavage site of Saccharomyces cerevisiae (Fig. 1a; ref. 29). Targeted by LacI–mCherry and TetR–GFP fusion proteins, these arrays allow for visualization of the regions flanking the DSB site. DSBs were induced by galactose treatment, which triggers GAL promoter-controlled expression of the Ho endonuclease (Extended Data Fig. 1a). In individual cells, we distinguished tethering or separation of DSB-ends based on the distance between the spot centres being less than or greater than 400 nm (Fig. 1b). This threshold was established by quantifying spot separation in the absence of DNA DSBs, where less than 5% of wild-type (WT) cells exhibited spots exceeding 400-nm separation (Extended Data Fig. 1b). We confirmed the assay’s sensitivity to detect previously described, early MRX- and late Exo1-dependent end-tethering pathways by imaging at 2 h and 4 h post-DSB induction. At 2 h post-DSB, WT and exo1Δ cells showed less than 10% separation, whereas cells lacking Mre11 displayed 31% separation (Fig. 1c and Extended Data Fig. 1c). At 4 h post-DSB, separation remained unchanged in WT cells but increased to 23% in exo1Δ cells. Importantly, double deletion of EXO1 and MRE11 led to a significant increase in end-separation compared to either single mutation, highlighting the presence of two DSB end-tethering pathways (Fig. 1d and Extended Data Fig. 1d).

Fig. 1: Cohesin tethers DSB-ends in the Exo1 pathway.
figure 1

a, A LacO/LacI–mCherry tag and a TetO/TetR–GFP tag were inserted at 5 and 7 kb from the Ho DSB site at the mating-type (MAT) locus, respectively. b, Example of cells with tethered or separated ends. The images are maximum intensity projections and representative of the independent experiments quantified in c and d. Signals are considered as separated when the distance between centres is more than 400 nm. Scale bars, 1 µm. c,d, Percentage of cells with separated ends in the indicated strains after 2 h (c) or 4 h (d) DSB induction. e, Examples of cells showing sister chromatid separation and DSB end-separation upon Smc1-AID auxin-mediated degradation in the absence or presence of DSB induction. The images are maximum intensity projections and are representative of the independent experiments quantified in f. Scale bars, 1 µm. f, Percentage of cells with separated ends in WT and SMC1-AID strains in the absence (−) or presence (+) of auxin after 2 h, 4 h or no DSB induction as indicated. g,i, Percentage of cells with separated ends in the indicated strains after 2 h (g) or 4 h (i) DSB induction. Data are presented as the mean of more than n = 3 independent experiments with N ≥ 150 for each strain and condition (c,d,fi). See source numerical data for detailed numbers. Overlaid black dots show the value of independent biological replicates. P values were calculated using a two-tailed unpaired t-test (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, not significant (NS) P > 0.05). Numerical P values are provided in Supplementary Table 4. In c, WT versus mre11Δ P = 0.0003, WT versus mre11Δ exo1Δ P = 0.0009. In d, WT versus exo1Δ P = 0.0009, WT versus mre11Δ P = 0.0003, exo1Δ versus mre11Δ P = 0.0053, mre11Δ versus mre11Δ exo1Δ P = 0.0126. In f, 2 h SMC1-AID − auxin versus SMC1-AID + auxin P = 0.0292, 4 h SMC1-AID − auxin versus SMC1-AID + auxin P = 0.0033, 4 h SMC1-AID + auxin versus SMC1-AID − auxin no DSB P = 0.0005, 4 h SMC1-AID + auxin versus SMC1-AID + auxin no DSB P = 0.0004. In g, mre11Δ SMC1-AID − auxin versus mre11Δ SMC1-AID + auxin P = 0.0128. In h, mre11Δ versus mre11Δ SMC1-AID + auxin P = 0.0068, mre11Δ SMC1-AID − auxin versus mre11Δ SMC1-AID + auxin P = 0.0095. In i, exo1Δ SMC1-AID – auxin versus exo1Δ SMC1-AID + auxin P = 0.1253.

Source data

To investigate cohesin in DSB end-tethering, we employed the auxin-induced degron (AID) system to deplete the Smc1 subunit30. Following a 1-h auxin incubation, Smc1 protein levels were substantially reduced and maintained at near undetectable levels for 4 h (Extended Data Fig. 2a). Depletion of Smc1 resulted in cells with separated sister chromatids (Fig. 1e) and impaired cell growth (Extended Data Fig. 2e), consistent with the essential role of cohesin in SCC. At 2 h post-DSB, a slight but significant increase in end-separation was observed (Fig. 1f). Strikingly, at 4 h post-DSB, ~30% of DSB-ends were untethered (Fig. 1f). To ensure that the increase in spot separation above 400 nm was due to the lack of DSB end-tethering and not due to the loss of cohesin-mediated chromatin folding, we quantified the percentage of cells with spots exceeding 400 nm upon Smc1 depletion in the absence of DSB. No significant increase in spot separation was observed when Smc1 was depleted in the absence of DSBs (Fig. 1f and Extended Data Fig. 1d), excluding an involvement of cohesin-mediated chromatin folding. Overall, these results reveal a cohesin requirement for DSB end-tethering.

Early cohesin tethering is independent of MRX and Exo1

Because both MRX and cohesin were required for early DSB tethering, we asked whether they act in the same pathway. Loss of Smc1 in mre11Δ cells significantly increased end-separation at 2 h post-DSB (Fig. 1g), suggesting that MRX and cohesin act in different pathways. Considering that Exo1 is not involved in early DSB tethering (Fig. 1c), our results suggest that, at 2 h post-DSB, cohesin acts independently of MRX and Exo1 to ensure DSB end-tethering.

Late cohesin tethering relies only on Exo1

At 4 h post-DSB, depleting cohesin in mre11Δ cells significantly increased end-separation (Fig. 1h), recapitulating the separation observed in mre11Δ exo1Δ cells (Fig. 1d). In contrast, depletion of Smc1 in exo1Δ cells did not further increase end-separation compared to exo1Δ cells (Fig. 1i). These findings suggest that cohesin functions with Exo1 to tether DSB-ends at 4 h post-DSB.

Altogether, our data indicate that cohesin supports DSB end-tethering at early and late time points through different pathways. Initially, cohesin tethering acts independently of the known MRX and Exo1 pathways, whereas later cohesin acts in the Exo1-dependent pathway.

Late DSB end-tethering requires de novo cohesin loading

Next, we interrogated whether cohesin present on chromosomes before DSB induction is sufficient for DSB end-tethering, or whether DSB-induced de novo cohesin loading is required. Accordingly, we arrested cells in G2/M phase using nocodazole, depleted Scc2 for 1 h to prevent de novo cohesin loading while maintaining pre-existing loops31, and induced DSB (Fig. 2a and Extended Data Fig. 2b). Smc1 depletion in nocodazole blocked cells lacking Mre11 resulted in increased separation at 2 h post DSB (Fig. 2b and Extended Data Fig. 3a), as was the case in the absence of nocodazole (Fig. 1g). In contrast, we observed no increase in end-separation upon depletion of Scc2 in nocodazole-arrested mre11Δ cells after 2 h DSB, showing that cohesin de novo loading is not required to tether DSB-ends at early time points (Fig. 2b and Extended Data Fig. 3a). After 4 h DSB induction, DSB end-separation significantly increased upon Scc2 depletion, reaching a similar level to that observed in Smc1-depleted and exo1Δ cells under the same experimental settings (Fig. 2c and Extended Data Fig. 3b). These results indicate that preformed cohesin loops participate in DSB end-tethering early post-DSB induction, but are not sufficient later, when de novo loading of cohesin is necessary for DSB end-tethering.

Fig. 2: Cohesin DSB end-tethering requires de novo cohesin loading but not sister chromatid cohesion.
figure 2

a, Schematic representation of an assay to determine DSB end-tethering in the absence of de novo cohesin loading. DSBs were induced after cells were blocked in G2/M with nocodazole for 3 h, and incubated with auxin or ethanol for a further 1 h. b,c, Percentage of G2/M blocked cells with separated ends in the indicated strains after 2 (b) or 4 h (c) DSB induction. d, Schematic representation of the assay used in e to determine DSB end-tethering in the absence of replication. G1 arrested cultures were incubated with auxin or ethanol for 1 h before release in galactose and auxin containing medium. In the absence of Cdc45, cells advance through the cell cycle upon DSB induction, and load cohesin onto chromosomes without undergoing replication. e,f, Percentage of cells with separated ends in the indicated strains after 4 h DSB induction. Data are presented as the mean of more than three independent experiments with N = 150 for each strain and condition (b,c,e,f). See source numerical data for detailed numbers. Overlaid black dots show the values of independent biological replicates. P values were calculated using a two-tailed unpaired t-test (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are provided in Supplementary Table 4. In b, mre11Δ SMC1-AID − auxin Noc versus mre11Δ SMC1-AID + auxin Noc P = 0.0054, mre11Δ SCC2-AID − auxin versus mre11Δ SCC2-AID + auxin P = 0.0886. In c, SMC1-AID + auxin Noc versus SMC1-AID − auxin Noc P = 0.0068. In e, SMC1-AID + auxin versus CDC45-AID + auxin P = 0.0025, SMC1-AID CDC45-AID − auxin versus SMC1-AID CDC45-AID + auxin P = 0.0092, SCC2-AID + auxin Noc versus SCC2-AID − auxin Noc P = 0.0032. In f, SMC1-AID SMC5-AID + auxin versus SMC1-AID SMC5-AID + auxin P < 0.0001, SMC5-AID + auxin versus SMC5-AID + auxin P = 0.0004.

Source data

DSB end-tethering does not require sister chromatid cohesion

Despite efficient cleavage of both sister chromatids in our assay (Extended Data Fig. 1a), which makes tethering of a cleaved chromatid by its sister unlikely, we aimed to confirm that DSB end-tethering was independent of SCC. In the absence of Cdc45, G1 cells progress to G2/M phase and load cohesin on chromosomes without firing replication origins and synthesizing sister chromatids32 (Fig. 2d and Extended Data Figs. 2d,e and 4). Depletion of Cdc45 did not disrupt DSB end-tethering at 4 h post-DSB induction (Fig. 2e and Extended Data Fig. 3c), indicating that the presence of a sister chromatid is not essential for DSB end-tethering. Additional depletion of Smc1 resulted in increased DSB end-separation, reaching levels similar to those observed in cells depleted of Smc1 alone. This indicates that cohesin can tether DSB-ends even in the absence of DNA replication and a sister chromatid.

Late cohesin DSB end-tethering requires Smc5/6

Previous studies have shown the importance of Smc5/6 in enriching cohesin at DSBs17. Accordingly, we depleted Smc5 in our DSB end-tethering assay (Extended Data Fig. 2c,e). At 4 h post-DSB induction, Smc5 depletion resulted in a significant increase in DSB end-separation (Fig. 2f and Extended Data Fig. 3d). Simultaneous depletion of Smc5 and Smc1 did not increase end-separation beyond that observed upon Smc1 depletion alone, indicating that cohesin and Smc5/6 function in the same pathway to tether DSB-ends.

Together, these findings unveil a series of events that ultimately result in cohesin-dependent DSB end-tethering. Initially, preformed cohesin loops and binding of the MRX complex tethers DSB-ends. Later, an Exo1-dependent pathway comes into play with the recruitment and de novo loading of cohesin, facilitated by Scc2/4 and Smc5/6, actively participating in the tethering of DSB-ends within individual chromatids.

Cohesin orchestrates compaction of DSBs flanking chromatin

Cohesin has been shown to form DNA loops and we hypothesized that this activity could contribute to cohesin-dependent DSB end-tethering. To gain insights into the behaviour of cohesin in the chromatin surrounding DSBs, we modified our assay to investigate chromatin compaction in a 48-kb region flanked by LacO–LacI–mCherry and TetO–TetR–GFP arrays, located 7 kb upstream of the DSB site (Fig. 3a). We measured the distance between these two signals in the presence or absence of DSB, to evaluate DSB-induced chromatin compaction. As the occurrence of a DSB triggers the DNA-damage checkpoint and a G2/M cell-cycle arrest, we treated all cells with nocodazole to ensure a fair comparison between DSB and no-DSB conditions (Fig. 3b).

Fig. 3: Cohesin compacts DSBs flanking chromatin.
figure 3

a, LacO/LacI–mCherry tags and a TetO/TetR–GFP tag inserted at 7 and 55 kb from the Ho DSB site at the MAT locus, respectively. b, Representative images illustrating spot separation in the presence and absence of DSBs for the experiments quantified in ce. Scale bars, 1 µm. c, Relative frequency of distances between the two tags in SMC1-AID-tagged budded cells treated with ethanol (+Smc1) or auxin (−Smc1). d, Relative frequency of distances between the two tags in nocodazole-arrested SMC1-AID-tagged cells treated with ethanol (+Smc1) or auxin (−Smc1) after 4 h and no DSB. e, Relative frequency of distances measured between the two tags in nocodazole-arrested PDS5-AID-tagged cells treated with ethanol (+Pds5) or auxin (−Pds5) after 4 h or no DSB induction. Shaded areas in ce represent Gaussian fitting (solid line) of n = 3 biological replicates with N ≥ 300 for each strain and condition, and a confidence interval of 95%. P values were calculated on the distance distribution using a two-tailed unpaired Mann–Whitney test. f, Percentage of cells with separated ends in the indicated strains after 4 h DSB induction. Overlaid black dots show the values of independent biological replicates. P values were calculated using a two-tailed unpaired t-test. Data are presented as the mean of more than three independent experiments with N ≥ 150 for each strain and condition (cf). See source numerical data for detailed numbers (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are presented in Supplementary Table 4. In c, −SMC1 versus +SMC1 P < 0.0001. In d, no DSB + Smc1 versus DSB + Smc1 P < 0.0001, DSB − Smc1 versus DSB + Smc1 P < 0.0001. In e, no DSB + Pds5 versus DSB + Pds5 P < 0.0001, DSB − Pds5 versus DSB + Pds5 P < 0.0001. In f, 4 h SMC1-AID − auxin versus SMC1-AID + auxin P = 0.0033, 4 h PDS5-AID − auxin versus PDS5-AID + auxin P = 0.0022.

Source data

We first examined the impact of cohesin chromatin folding in G2/M-arrested cells with no DSB. Cohesin depletion significantly increased the distribution of distances between spots (Fig. 3c and Extended Data Fig. 5a), showing that our assay detects the previously reported cohesin-dependent compaction of chromatin33,34. Following 4 h of DSB induction, we detected a significant reduction in spot distances compared to the no-DSB condition, indicative of a compaction of the DSB-flanking chromatin (Fig. 3d; black versus red). DSB-induced compaction was abolished upon depletion of Smc1, demonstrating that cohesin compacts DSB flanking sequences (Fig. 3d, orange and Extended Data Fig. 5b).

Pds5 facilitates DSB end-tethering but not genome compaction

If loop formation were at the basis of DSB end-tethering, the latter should be challenged by modulating loop expansion and turnover. To explore this, we tested the role of Pds5, a key factor in cohesin loop regulation. Pds5 depletion weakens loop boundaries, reduces defined chromosome contacts/loops, and generates much longer loops in regions such as those near centromeres35,36. DSB-induced chromatin compaction, although slightly affected, still occurred in the absence of Pds5 (Fig. 3e and Extended Data Fig. 5c–e). In contrast, Pds5 depletion increased end-separation at 4 h post-DSB, mimicking the effects of cohesin depletion (Fig. 3e and Extended Data Fig. 5f). These results imply that either the loops formed in the absence of Pds5 were not sufficient to support the function of cohesin in DSB end-tethering, or that cohesin tethers DSB-ends independently of loop formation, through another mechanism requiring Pds5. A recent study revealed an essential role of Pds5 in the oligomerization of multiple cohesin complexes37, opening the door for a role of Pds5-dependent cohesin oligomerization in DSB end-tethering.

Cohesin and MRX tethering use weak hydrophobic interactions

To investigate whether protein–protein interactions and cohesin oligomerization participate in DSB end-tethering, we used the aliphatic alcohol 1,6-hexanediol. Hexanediol has been instrumental in studying the liquid-phase separation and oligomerization properties of various proteins, including cohesin and proteins involved in the DNA-damage response38,39. Notably, it disrupts MRX and Rad52 foci (Extended Data Fig. 6a–d).

Treatment of cells with hexanediol 10 min before imaging at 2 h post-DSB, when tethering mostly relies on MRX, increased end-separation independently of Smc1 presence (Fig. 4a–c). Moreover, end-separation was not increased by hexanediol in the absence of Mre11 alone. These results suggest a role for weak hydrophobic interactions in MRX-dependent tethering. Strikingly, hexanediol-treated mre11Δ cells do not exhibit the separation levels observed in Smc1-depleted mre11Δ cells, with or without hexanediol treatment (Figs. 4c, 1g and Extended Data Fig. 6e). This finding aligns with a recent in vivo study in S. cerevisiae that demonstrated the resistance of a subset of topologically important cohesins to hexanediol treatment38. As hexanediol is known to disrupt protein–protein interactions, this further supports our finding that an Exo1-independent population of cohesin can tether DSB-ends that are formed within a cohesin loop (Fig. 1g).

Fig. 4: DSB-ends tethering by MRX or cohesin is sensitive to hexanediol treatment.
figure 4

a, Schematic representation of the assay to test DSB end-tethering sensitivity to hexanediol (Hex). b, Drop assay of strains plated on YPD after no treatment (NT), 10 min digitonin (D) or 10 min digitonin + hexanediol (D + H) treatment, incubated for 24 and 48 h at 30 °C. c,d, Percentage of cells with separated ends in the indicated strains treated with auxin or ethanol, and for 10 min with digitonin (−) or digitonin and 1,6-hexanediol (+), after 2 h (c) or 4 h (d) DSB induction. Data are presented as the mean of more than three independent experiments with N ≥ 150 for each strain and condition (c,d). See source numerical data for detailed numbers. Overlaid black dots show the values of independent biological replicates. P values were calculated using a two-tailed unpaired t-test (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are presented in Supplementary Table 4. In c, −Hex + Smc1 versus −Hex + Smc1 P = 0.0027, −Hex − Smc1 versus +Hex − Smc1 P = 0.0170, +Hex mre11Δ + Smc1 versus −Hex mre11Δ − Smc1 P = 0.0004. In d, −Hex + Smc1 versus −Hex + Smc1 P = 0.0027, −Hex − Smc1 versus +Hex − Smc1 P = 0.0170, +Hex mre11Δ + Smc1 versus −Hex mre11Δ − Smc1 P = 0.0004.

Source data

At 4 h post-DSB, hexanediol-treated control cells also exhibited untethering (Fig. 4d and Extended Data Fig. 6f). In line with hexanediol disruption of MRX-dependent tethering at 2 h, hexanediol and Smc1 depletion have additive effects at 4 h. In contrast to the 2-h time point, hexanediol increased end-separation in mre11Δ cells to levels comparable to cells depleted for both Smc1 and Mre11, suggesting disruption of cohesin-dependent DSB end-tethering (Fig. 4d). These results indicate that protein–protein interactions play a key role in DSB end-tethering by both MRX and cohesin pathways. Because hexanediol might affect proteins acting in these pathways other than MRX and cohesin, we aimed to directly test the relevance of oligomerization using specific mutants.

MRX has been shown to form oligomers in vitro, and disruption of these oligomers by a mutation of the hydrophobic interaction patch within the Rad50 head ___domain (rad50lo mutant40) led to the disappearance of DSB-dependent Mre11 foci in vivo. Because hexanediol also disrupts Mre11-GFP foci formation in our strain background (Extended Data Fig. 6c,d), we introduced this mutation in our tethering system. The rad50lo mutant protein is expressed at nearly WT levels in our strain background and remains proficient for NHEJ (Extended Data Fig. 7a,b). Notably, NHEJ is significantly increased in the rad50lo mutant compared to WT cells, consistent with the decreased resection previously described in this mutant40. Complementation of rad50Δ cells with rad50lo did not affect end-separation levels compared to WT at 2 h post-DSB (Fig. 5a and Extended Data Fig. 7c). Therefore, disrupting Rad50 head oligomerization is not sufficient to impair DSB end-tethering. However, the disruption of the MRX-dependent pathway of DSB end-tethering by hexanediol suggests that MRX may achieve this through oligomerization using other interfaces such as the Zn-hook3, or that oligomerization of other proteins acting with MRX is required (Fig. 5b).

Fig. 5: DSB-ends are tethered by MRX or cohesin oligomerization.
figure 5

a, Percentage of cells with separated ends in WT, rad50Δ and rad50Δ cells complemented with RAD50 or rad50-lo, after 2 h of DSB induction. b, Model for 2-h cohesin-mediated end-tethering. DSB-ends are primarily tethered by MRX assisted by cohesin-mediated loops. Cohesin tethering is independent of oligomerization and de novo loading. c, Percentage of cells with separated ends in MCD1-AID, and MCD1-AID strains complemented with MCD1 or mcd1Q266, in the absence (−) or presence (+) of auxin, after 2 h of DSB induction. Data are presented as the mean of more than three independent experiments with N ≥ 50 for each strain and condition. See source numerical data for detailed numbers. Overlaid black dots show the value of independent biological replicates. P values were calculated using a two-tailed unpaired t-test. d,e, Relative frequency of distances measured between two tags separated by 45 kb in MCD11-AID-tagged strains complemented with MCD1 (d) or mcd1Q266 (e), treated with ethanol or auxin and nocodazole after 4 h DSB induction. Shaded areas in d and e indicate the 95% confidence intervals of the Gaussian fitting (solid line) of three independent experiments, with N ≥ 300 for each strain and condition. P values were calculated on the distance distribution using a two-tailed unpaired Mann–Whitney test (d,e). f, Percentage of cells with separated ends in MCD1-AID, and MCD1-AID strains complemented with MCD1 or mcd1-Q266, in the absence (−) or presence (+) of auxin, after 4 h of DSB induction. Data are presented as the mean of more than three independent experiments with N ≥ 50 for each strain and condition. See source numerical data for detailed numbers. Overlaid black dots show the value of independent biological replicates. P values were calculated using a two-tailed unpaired t-test. g, Model for 4-h cohesin-mediated end-tethering. Cohesin mediates DSB end-tethering by oligomerization in a MRX-independent but Exo1-, Scc2-, Smc5/6- and Pds5-dependent manner (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are provided in Supplementary Table 4. In a, WT versus rad50Δ P < 0.0001, rad50Δ versus RAD50 P < 0.0001, rad50Δ versus rad50lo P = 0.0001, RAD50 versus rad50lo P = 0.4411. In c, MCD1-AID − auxin versus MCD1-AID + auxin P = 0.0100. In d, no DSB MCD1 + Mcd1 versus DSB MCD1 + Mcd1 P < 0.0001, no DSB MCD1 + Mcd1 versus DSB MCD1 − Mcd1 P < 0.0001. In e, no DSB mcd1Q + Mcd1 versus DSB mcd1Q + Mcd1 P < 0.0001, no DSB mcd1Q + Mcd1 versus DSB mcd1Q − Mcd1 P < 0.0001. In f, MCD1-AID − auxin versus MCD1-AID + auxin P = 0.0050, MCD1-AID MCD1 + auxin versus MCD1-AID mcd1Q266 + auxin P = 0.0127.

Source data

The cohesin subunit Mcd1 has been identified as a mediator of cohesin oligomerization, and a five-amino-acid insertion at Q266 in its regulation of cohesion and condensation (ROCC) ___domain has been shown to abolish cohesin oligomerization in vivo37,41. To test the role of cohesin oligomerization in DSB end-tethering, we complemented MCD1-AID cells with the mcd1-Q266 mutant in both our compaction and end-tethering strains (Extended Data Fig. 8a,b). In contrast to cells depleted for Mcd1, mcd1Q266 mutant cells did not exhibit separated DSB-ends at 2 h, indicating that early DSB end-tethering does not rely on cohesin oligomerization (Fig. 5c and Extended Data Fig. 8c), and probably relies on preformed loops (Fig. 5b). Critically, mcd1Q266 mutants exhibited strong DSB-dependent genome compaction at 4 h (Fig. 5d,e and Extended data Fig. 8d–f), indicating that cohesin is recruited to DSB sites and able to form chromatin loops. However, unlike complementation with MCD1, mcd1Q266 failed to restore DSB end-tethering to WT-like levels (Fig. 5f and Extended Data Fig. 8g), confirming the importance of cohesin oligomerization in DSB end-tethering (Fig. 5g).

Cohesin assists DSB repair by homology-directed mechanisms

Having identified cohesin’s role in DSB end-tethering, we questioned its importance for repair. We took advantage of our tethering system, which contains direct homologous repeats flanking the inserted LacO and TetO arrays (Fig. 6a). Following DSB induction and resection initiation, progressive formation of ssDNA away from the DSB causes loss of the dsDNA substrate that is necessary for the binding of the LacI–mCherry and TetR–GFP fusion proteins, leading to the gradual disappearance of fluorescent signals. Resection also unmasks the direct homologous repeats, which can anneal and be used to resynthesize the broken DNA strand. This restores chromosome continuity but results in loss of the genetic material that previously separated the homologous repeats used for repair. Following resynthesis, either the red or green signal reappears, depending on the repeats used for repair (Fig. 6a,b). Following repair, cells are released from the DNA-damage checkpoint and proceed through cell division (Fig. 6b and Supplementary Videos 14).

Fig. 6: Cohesin is required for efficient DNA DSB repair.
figure 6

a, Schematic representation of repair events after resection and disappearance of the spots followed by resynthesis of one spot. Black and grey arrows show direct repeats used for homologous recombination. b, Sequence of images showing the disappearance of both spots upon resection, and the reappearance of a green spot that is propagated to daughter cells at each division. The time post DSB induction is indicated on each frame. This sequence is representative of the repair events quantified in ch. Scale bar, 1 µm. c, Relative frequency of repair events corresponding to the resynthesis of a spot in rad52Δ, rad51Δ and SMC1-AID strains treated with ethanol (+Smc1) or auxin (−Smc1). Data are presented as means ± s.d. P values were calculated using a two-tailed unpaired t-test. d, Time taken for a spot to reappear, in rad52Δ, rad51Δ and SMC1-AID strains treated with ethanol (+Smc1) or auxin (−Smc1). The red line represents the median, and quartiles are represented by dashed lines. P values were calculated using a two-tailed unpaired Mann–Whitney test. Data are representative of N = 61 (rad52Δ), N = 54 (rad51), N = 77 (+Smc1) and N = 79 (−Smc1) cells from n = 2 independent experiments (c,d). e,f, Spot characteristics of +Smc1 (e) and −Smc1 (f), individual cells imaged every 10 min over 12 h after DNA DSB induction. Lines represent individual cell lineages, and each segment a time point. Colours indicate the presence of both spots (yellow), a red spot only (red), a green spot only (green) or no spots (grey). Data are representative of N = 77 (+Smc1) and N = 79 (−Smc1) cells from n = 2 independent experiments (e,f). g, Relative frequency of repair events corresponding to the resynthesis of a spot in the indicated strains treated with auxin. Cells in G1 phase upon induction were imaged. Data are presented as means ± s.d. P values were calculated using a two-tailed unpaired t-test. h, Time for a spot to reappear in the indicated strains treated with auxin. Cells in G1 phase upon induction were imaged. The red line represents the median, and quartiles are represented by dashed lines. P values were calculated using a two-tailed unpaired Mann–Whitney test. Data are representative of N = 45 (−Cdc45) and N = 47 (−Cdc45 − Smc1) cells from n = 3 (−Cdc45) and n = 2 (−Cdc45 − Smc1) independent experiments (g,h) (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are provided in Supplementary Table 4. In c, +Smc1 versus −Smc1 P = 0.0224. In d, +Smc1 versus −Smc1 P = 0.0307. In g, −Cdc45 versus −Cdc45 − Smc1 P = 0.0116. In h, −Cdc45 versus −Cdc45 − Smc1 P = 0.0004.

Source data

To assess repair events, we employed a microfluidics system to follow individual cells and image each fluorescent signal over a 12-h period after DSB induction. To validate our assay, we imaged cells lacking RAD52, which is essential for all homology-directed repair (HDR) events. In the absence of Rad52, no instances of spot reappearance were observed (Fig. 6c and Extended Data Fig. 9a). Conversely, the loss of Rad51, which impedes gene conversion and promotes single-strand annealing (SSA), led to an increase in repair events compared to the WT-like condition (SMC1-AID without auxin; Fig. 6c,e and Extended Data Fig. 9b), as previously reported42,43. This result suggests that inhibiting gene conversion, and favouring repair by SSA, leads to more detectable repair events in this assay, with unaltered repair kinetics compared to the WT-like condition (Fig. 6d). In contrast, Smc1 depletion resulted in a significant reduction in repair frequency, with delayed repair kinetics (Fig. 6c–f). This decrease in repair frequency was not caused by a resection defect (Extended Data Fig. 9c). To separate the dependence of repair events on SCC and DSB end-tethering, we employed Cdc45 depletion. Strikingly, repair still occurred upon Cdc45 depletion, whereas simultaneous depletion with Smc1 resulted in a severe decrease in repair frequency and kinetics compared to cells depleted of Cdc45 alone (Fig. 6g,h and Extended Data Fig. 9d–f). This result indicates that cohesin is necessary for DSB repair beyond its role in SCC. To specifically test the contribution of cohesin-dependent DSB end-tethering, we measured repair in the mcd1Q266 mutant. Unlike complementation with MCD1, which exhibited WT repair levels, mcd1Q266 cells exhibited decreased repair comparable to that observed upon Mcd1 depletion (Fig. 7a–f). Altogether, these results demonstrate that the specific function of cohesin in DSB end-tethering is essential for efficient repair between DSB-ends.

Fig. 7: Cohesin oligomerization is required for efficient DNA DSB repair.
figure 7

ae, Spot characteristics in MCD1-AID, and MCD1-AID strains complemented with nothing (c), MCD1 (a,b) or mcd1-Q266 (d,e), in the absence (+Mcd1) or presence (−Mcd1) of auxin for individual cells imaged every 10 min during a period of 12 h after DNA DSB induction. f, Relative frequency of repair events corresponding to the resynthesis of a spot in the indicated strains treated with auxin (+) or ethanol (−). Error bars represent s.d. P values were calculated using a two-tailed unpaired t-test. Data are representative of N = 101 (−Mcd1 + Mcd1WT), N = 76 (+Mcd1 + Mcd1WT), N = 77 (−Mcd1), N = 86 (+Mcd1 + Mcd1Q266) and N = 71 (−Mcd1 + Mcd1Q266) cells from n = 2 independent experiments (af) (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, NS P > 0.05). Numerical P values are provided in Supplementary Table 4. In f, MCD1-AID − auxin versus MCD1-AID + auxin P = 0.0149, MCD1 MCD1-AID − auxin versus MCD1 MCD1-AID + auxin P = 0.6439, mcd1Q266 MCD1-AID − auxin versus mcd1Q266 MCD1-AID + auxin P = 0.0406.

Source data

Discussion

Cohesin enrichment at DSBs has long been known11,12,13, with early studies highlighting the importance of cohesin for survival after DNA damage-inducing radiation11,13,44. Recent observations suggest that loop extrusion at DNA DSBs helps establish DNA-damage response-related chromatin modifications20, which ultimately organize DSBs into microdomains45. Moreover, SCC, which is increased in response to DSB18,19,21,22,23, prevents promiscuous repair events with distant loci24,28.

In addition to these functions, we reveal a DSB end-tethering role for cohesin. Cohesin’s first contribution, early after DSB formation, is independent of MRX and Exo1 and probably relies on cohesin-dependent genome looping (Fig. 5c), as predicted by recent theoretical work2. In most cells, tethering is ensured by the MRX complex, probably independently of cohesin, as suggested by the additive end-tethering defect observed in the absence of both Mre11 and cohesin. However, we cannot exclude that cohesin loops or chromatin folding contribute to keeping ends close before MRX binding, improving the efficiency of MRX tethering without being absolutely required.

Later, cohesin DSB end-tethering requires de novo cohesin loading, acts in cooperation with Exo1 and Smc5/6, is independent of SCC and loop formation, and relies on cohesin oligomerization (Fig. 5g). Importantly, our data provide a biological function to the recently described cohesin oligomerization mechanism37,38 that is independent of cohesin’s canonical roles in SCC and loop extrusion.

Our results support the existence of two populations of DSB-bound cohesin with separable functions (chromatin compaction and DSB end-tethering), through different modes of action (loop formation and oligomerization). One population of cohesin acts in a Pds5- and oligomerization-independent manner and compacts DSB-adjacent chromatin. This compaction may participate in DSB signalling though a loop extrusion-mediated spreading of histone H2AX phosphorylation20. A second population requires Pds5 and cohesin oligomerization, and tethers DSB-ends. What distinguishes loop-forming cohesin from DSB end-tethering cohesin, beyond the capacity to form oligomers, is unknown. However, that DSB end-tethering cohesin acts independently of MRX, which has been implicated in cohesin enrichment at DSBs12,15, suggests a different mode of recruitment to DSB-ends.

Our data support a role for Scc2-, Smc5/6- and Exo1-mediated ssDNA formation in recruiting or stabilizing DSB end-tethering cohesin. Scc2 and Smc5/6 were previously implicated in cohesin recruitment to DSB, but the formation of ssDNA by Exo1 is specifically required for cohesin-dependent DSB end-tethering. Given that dsDNA-bound cohesin can capture ssDNA46, the formation of ssDNA may directly assist cohesin recruitment. Bridging dsDNA with ssDNA could also be sufficient for DSB end-tethering. Alternatively, cohesin recruitment may occur independently of ssDNA, which could instead mediate the recruitment of Smc5/6. Indeed, Smc5/6 interacts with ssDNA through its hinge ___domain47,48, and stably associates with ss–dsDNA junctions48,49. Beyond recruiting tethering cohesin, Smc5/6 could also participate in DSB end-tethering through post-translational modification of proteins in the cohesin pathway. Indeed, Smc5/6 contains small ubiquitin-like modifier (SUMO) and ubiquitin ligase activities mediated by the Nse2/Mms21 and Nse1 subunits, respectively. The SUMO ligase activity of Smc5/6 targets numerous proteins implicated in genome organization, DNA replication and DNA repair50. Notably, substrates of Nse2/Mms21-mediated SUMOylation include subunits of the cohesin complex. Among these, the SUMOylation of Mcd1 is triggered by DSB induction and is required for DNA-damage-induced cohesion17. Given that interaction between SUMOylated proteins can lead to the formation of large protein aggregates51, SUMOylation of Mcd1 by Smc5/6 could participate in cohesin oligomerization and DSB tethering.

Our results, revealing cohesin’s role in DSB end-tethering, contrast with a previous report of a Hi-C approach suggesting that cohesin is dispensable for contacts between both sides of a DSB24. One plausible explanation for this discrepancy is rooted in the technologies used. Single-cell live-microscopy allows for detection of DSB-induced compaction beyond G2/M chromosome folding, and cohesin-dependent loss of end-tethering, both appearing below the detection threshold of the population-wide Hi-C approach24.

We also demonstrate that loss of Rad50 head oligomerization observed in vitro is not sufficient to disrupt MRX end-tethering in vivo. MRX oligomerization via both the Rad50 heads and coiled coils has been described in both yeast and humans40,52. Disruption of DSB end-tethering of the MRX pathway by hexanediol treatment suggests a crucial contribution of oligomerization in this process. Hexanediol could disrupt oligomerization of the MRX complex via interfaces other than the Rad50 heads, such as the Zn-hook3, or affect oligomerization of other complexes contributing to early DSB end-tethering, such as Nej1/Lif129.

Our results suggest that oligomerization of SMC complexes is a conserved and functionally relevant mechanism for maintaining genome integrity in response to DNA damage. Interestingly, hexanediol treatment disrupted MRX foci in response to DSBs, suggesting MRX at DSBs may form condensates. Although cohesin does not form detectable foci in response to DSBs in yeast, it has been shown to form phase separation condensates in vitro38. Thus, the relevance of phase separation in DSB end-tethering should be investigated using single-molecule microscopy.

Given the prevalence of chromosome translocations in cancer, and the role of DSB induction in cohesin-sensitive developmental processes such as V(D)J recombination53, our study gives further insights into how SMC complex dysregulation may lead to disease in the human population.

Methods

Strains and plasmids

The yeast strains used in this study are derivative of JKM179, JKM13954 or yKD80929, and were generated by polymerase chain reaction (PCR) gene targeting, plasmid transformation or cross (Supplementary Tables 1 and 2).

Media and growth conditions

Yeast strains were grown at 30 °C in glucose rich yeast extract–peptone–dextrose (YPD) medium, with appropriate antibiotic, or in synthetic medium (SD) lacking the appropriate amino acid. YPLGg medium containing 2% lactate, 3% glycerol and 0.05% glucose was used for DNA DSB induction, by addition of galactose (final concentration of 2%), to overnight (ON) cultures with an optical density at 600 nm (OD600) of 0.4–0.8, as in ref. 55. Conditional protein knockdown was achieved in AID-tagged strains by the addition of 3-indoleacetic acid (IAA) in EtOH to a final concentration of 2 mM (ref. 30) for 1 h before DSB induction by galactose addition (final concentration of 2%) for 2 or 4 h. 1,6-hexanediol treatment (final concentration of 10%) with 10 µg ml−1 digitonin was performed for 10 min before imaging.

Microscopy

Live-cell images were acquired using a wide-field inverted microscope (Leica DMI-6000B) equipped with adaptive focus control to eliminate Z drift, a ×100/1.4-NA immersion objective with a PriorNanoScanZ Nanopositioning Piezo Z Stage System, a complementary metal–oxide–semiconductor (CMOS) camera (ORCA-Flash4.0; Hamamatsu) and a solid-state light source (Spec-traX, Lumencore). The system was piloted by MetaMorph software (Molecular Device, v7.10.5). Images were acquired at indicated time points after DSB induction. Nineteen focal steps of 0.20 µm were acquired sequentially for green fluorescent protein (GFP) and mCherry with an exposure time of 50 ms using solid-state 475- and 575-nm diodes and appropriate filters (GFP-mCherry filter; excitation: double BP, 450–490/550–590 nm and dichroic double BP 500–550/600–665 nm; Chroma Technology Corp.). Images were processed using Fiji ImageJ software (National Institutes of Health, ImageJ2 v2.14.0/1.54f). Three-dimensional (3D) images were converted to 2D projections, from which xy coordinates of the most intense pixels were extracted. Distance analysis between the closest fluorescent signals in the mCherry and GFP channels was performed using an Rstudio script. When sister chromatids were separated, taking into account only the smallest distance underestimated separation. All images shown are maximal z projections of z-stack images.

Microscopy in microfluidic plates

CellASIC ONIX microfluidic plates (Y04C-02; MilliporeSigma) were used for long-duration movies. Ho was induced in YPLGg cultures with an OD600 of 0.5 by addition of galactose to a final concentration of 2%, and incubation at 30 °C for 30 min. After break induction, cultures were loaded into the microfluidic plate. The remaining culture was centrifuged at 900g for 3 min, and the conditioned medium was loaded into the microfluidic plate for flow over the cells for the duration of the experiment. After loading the plate, cell positions were defined, and images were acquired every 10 min for up to 24 h. Nineteen focal steps of 0.20 µm were acquired sequentially for GFP and mCherry, with an exposure time of 30 ms, using solid-state 475- and 575-nm diodes and appropriate filters (GFP-mCherry filter; excitation: double BP, 450–490/550–590 nm and dichroic double BP 500–550/600–665 nm; Chroma Technology Corp.). A single bright-field image on one focal plane was acquired at each time point with an exposure of 10 ms. For Cdc45 depleted strains, cells were loaded into the microfluidic plate immediately following galactose addition, and cells that were in G1 before DSB induction were imaged.

Monitoring DSB efficiency

Cells were grown in 2 ml of YPD ON. Cultures were then diluted in YPLGg, and grown to an OD600 of 0.5–0.8, then incubated with 2 mM IAA or EtOH for 1 h. Ho expression was then induced by addition of galactose to a final concentration of 2%. At 0, 1, 2, 4 and 6 h post DSB induction, ~4 × 107 cells were collected by 900g centrifugation for 5 min. DNA was extracted from cell pellets by Winston preparation. Samples were analysed by quantitative PCR with primers 1 kb upstream of the Ho site to analyse resection (200 nM), flanking the Ho site to determine DSB efficiency (450 nM) or targeting the OGG1 reference gene (200 nM). Supplementary Table 3 provides the primer sequences. Reactions were performed as in ref. 56. Each sample and no-template controls were run in triplicate, and reaction specificity was determined by melt curve analysis. Relative quantitation of resection and DSB efficiency reactions was achieved using the comparative Ct method57.

Western blot

Auxin-induced protein degradation of AID-containing strains was confirmed by western blot analysis30. Cells were grown in 2 ml of YPD ON. Cultures were then diluted in YPLGg, and grown to an OD600 of 0.5–0.8, and incubated with 2 mM IAA or EtOH for 1, 2 and 4 h. Approximately 4 OD600 of cells were collected by centrifugation at 900g for 5 min. Cells were washed in dH2O, and collected by centrifugation at 900g for 5 min. The supernatant was removed, and the cell pellets were frozen at −80 °C. Whole-cell extracts were prepared from the cell pellets using a standard trichloroacetic acid extraction protocol, then suspended in Laemmli buffer. Protein concentrations were determined by Bradford assay, and samples prepared for sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) by 5-min incubation at 90 °C. A 20-µg sample was migrated at 100 V for 1 h on 10% SDS polyacrylamide gels in standard running buffer. Nitrocellulose membrane transfer was performed using the iBlot transfer apparatus according to the manufacturer’s guidelines (Thermo Fisher). Membranes were washed with TBS-T, revealed by Ponceau staining, and blocked with 5% milk TBS-T for 1 h. The membranes were then incubated at room temperature with mouse primary anti-myc (1:1,000; anti-myc tag antibody (9E10) Abcam ab32), anti-Flag (1:4,000; ANTI-FLAG M2 antibody Sigma-Aldrich F3165), anti-Rad50 (1:1,000; PA5-32176 Invitrogen) and fluorescent secondary antibodies (1:10,000; goat anti-rabbit immunoglobulin G (IgG) (H + L) highly cross-adsorbed secondary antibody, Alexa Fluor Plus 800, Invitrogen A32735; goat anti-mouse IgG (H + L) highly cross-adsorbed secondary antibody, Alexa Fluor Plus 680, Invitrogen A32729; and goat anti-mouse IgG (H + L) highly cross-adsorbed secondary antibody, Alexa Fluor Plus 800 Invitrogen A32730) in 5% milk TBS-T for 1 h each. The membranes were developed by fluorescence using an Odyssey Clx system (LI-COR).

Flow cytometry

First, 0.5 OD600 of cells were fixed in ethanol (70%) and stored at −20 °C. Cells were pelleted, washed, and then incubated in sodium citrate pH 7.4, 50 mM with 0.25 mg ml−1 RNAseA for 1 h at 50 °C. Proteinase K was added to a final concentration of 2 mg ml−1, followed by incubation for a further 1 h at 50 °C. Cells were pelleted, then stained in a pH 7.4 50 mM sodium citrate solution containing 1 µM SYTOX Green nucleic acid stain (Invitrogen, S7020). Cells were sonicated, and flow cytometry was performed on a Novocyte cytometer (ACEA Bioscience). The data were analysed using FlowJo software (BD Biosciences, v10.10).

Statistics and reproducibility

For all experiments, no statistical method was used to predetermine sample size. Sample sizes are provided in the numerical source data or in the figure legends. The imaging data were excluded from analysis when a poor signal did not allow analysis. The experiments were not randomized. The investigators were not blinded to allocation during the experiments and outcome assessment. Quantifications and statistical analyses were done using GraphPad Prism v10.3.1. For the end-tethering assay, at least three independent experiments, each analysing more than 50 cells, were performed for each genotype, and statistical significance was determined by a two-tailed unpaired t-test. For the compaction measurements, distance data for at least 100 cells were sorted into 200-nm bins, and the bins of three independent experiments were fitted with a Gaussian curve using Prism software, with shaded areas representing a confidence interval of 95%. Statistical significance was determined on the distance distribution using a two-tailed unpaired Mann–Whitney test. Exact cell and biological replicates numbers are indicated in the figure legends or numerical source data. The statistical tests used are indicated in the figure legends and in Supplementary Table 4. P values that were <0.05 were considered statistically significant. Statistical differences are indicated by an asterisk (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, and non-significance by NS, P > 0.05). Numerical P values are indicated in the respective figure legends and listed in Supplementary Table 4.

Reporting summary

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