DNA Double-Strand Breaks Trigger Genome-Wide Sister-Chromatid Cohesion Through Eco1 (Ctf7)

See allHide authors and affiliations

Science  13 Jul 2007:
Vol. 317, Issue 5835, pp. 245-248
DOI: 10.1126/science.1140637

This article has a correction. Please see:


Faithful chromosome segregation and repair of DNA double-strand breaks (DSBs) require cohesin, the protein complex that mediates sister-chromatid cohesion. Cohesion between sister chromatids is thought to be generated only during ongoing DNA replication by an obligate coupling between cohesion establishment factors such as Eco1 (Ctf7) and the replisome. Using budding yeast, we challenge this model by showing that cohesion is generated by an Eco1-dependent but replication-independent mechanism in response to DSBs in G2/M. Furthermore, our studies reveal that Eco1 has two functions: a cohesive activity and a conserved acetyltransferase activity, which triggers the generation of cohesion in response to the DSB and the DNA damage checkpoint. Finally, the DSB-induced cohesion is not limited to broken chromosomes but occurs also on unbroken chromosomes, suggesting that the DNA damage checkpoint through Eco1 provides genome-wide protection of chromosome integrity.

A fundamental property of the eukaryotic chromosomes is sister-chromatid cohesion. Cohesion plays a crucial role in chromosome segregation (1) as well as post-replicative repair of double-strand breaks (DSBs) (2) and is mediated by a large ring-shaped complex, cohesin, and its associated protein, Pds5 (3, 4). In late G1 of budding yeast, cohesin is loaded onto chromosomes by the Scc2/Scc4 complex (5). This loading occurs around centromeres and at cohesin-associated regions (CARs) along chromosome arms (4). During S phase, Eco1 (also known as Ctf7) acts on the chromatin-bound cohesin complex to generate cohesion by an unknown mechanism (6, 7).

The generation of cohesion is limited to S phase in undamaged cells (810). This limitation cannot be explained by regulating the association of cohesins with chromosomes because they continue to load onto chromosomes at CARs in G2/M by an Scc2/Scc4-dependent mechanism (8). To explain this limitation, one model posits that cohesin can generate cohesion only by a replication-driven mechanism, facilitated by Eco1 (8) (herein called replication fork–driven cohesion model). The absence of a replication fork in G2/M prevents cohesion generation. However, cohesion is generated in G2/M upon irradiation (9) and was inferred to be mediated by cohesin loaded de novo around DSBs by a mechanism dependent on the DNA damage checkpoint (9, 11, 12). In the replication fork–driven cohesion model, a fork should be necessary for DSB-induced cohesion as well. Indeed, replication forks do occur at DSBs as part of the recombination-repair pathway.

To determine whether DNA replication is required for DSB-induced cohesion, we developed an assay to detect cohesion of specific regions on chromosomes in response to defined DSBs (Fig. 1A). This assay has three properties: First, formation of DSBs on the chromosomes is temporally and spatially controlled by placing the site-specific HO endonuclease under the control of an inducible promoter and by introducing two HO cleavage sites (HO-cs) on chromosome III (chr. III). One HO-cs is upstream of SRD1 and the second site is 60 kb away at MAT locus. After 1 hour of HO induction, chr. III (>90%) is broken into three pieces: the 60-kb SRD1-MAT fragment and two larger fragments (Fig. 1B). Second, cohesion of specific sites in the genome is detected by a cohesion reporter consisting of a tandem array of Lac operators (LacO) that can be visualized by LacI–green fluorescent protein (LacI-GFP). A single GFP spot in the cell indicates cohesion between the broken chromatid pairs, whereas two GFP spots indicate cohesion loss (Fig. 1A, right panel). Third, de novo DSB-induced cohesion is distinguished from the cohesion established during S phase. S-phase cohesion is established by the use of a thermo-sensitive cohesin subunit, mcd1-1 (referred to as S-cohesin). In G2/M, wild-type Mcd1 (also known as Scc1) is expressed concomitant with induction of DSBs, and subsequently S-cohesin is inactivated (Fig. 1, A and C).

Fig. 1.

DSB proximal cohesin generates sister-chromatid cohesion. (A) DSB-induced cohesion assay. See text and supporting online material (SOM) for details. (B) Chr. III cutting in EU3275 assayed by pulsed-field gel electrophoresis and Southern blotting. (C) Mcd16HA binding on SRD1-MAT fragment assayed by chromatin immunoprecipitation/real-time polymerase chain reaction. Gray squares: no DSB, black squares: DSB. Error bars indicate SD (n = 4). DSB-induced cohesion assay in (D) EU3275 and (E) EU3274 and EU3278. In (D) and (E) and subsequent figures, the genotype and location of the cohesion reporter are shown above each plot. Detailed information on strain genotypes are in table S1. a, active; i, inactive; ts, temperature-sensitive; HA, hemagglutinin. Error bars indicate SD (n = 3).

Using this assay, we asked whether cohesion can be generated in G2/M in proximity of the DSB by following the cohesion of the SRD1-MAT region. Without DSBs, when S-cohesin is active (30°C, permissive temperature), ∼20% of the chr. III sister chromatids are separated. As expected, upon inactivation of S-cohesin (37.5°C, nonpermissive temperature), >60% of chr. III sister chromatids separate (Fig. 1D). With DSBs, wild-type cohesin is loaded to the SRD1-MAT fragment (Fig. 1C), and this fragment retains cohesion upon the inactivation of S-cohesin (Fig. 1D). Both DSBs and wild-type cohesin are necessary, but neither is sufficient, to generate cohesion on the SRD1-MAT region in G2/M (Fig. 1E). These results show that DSBs can induce cohesindependent cohesion in G2/M, confirming that cohesion can be generated outside of S phase (9). In addition, these results show that as few as two DSBs are sufficient to induce cohesindependent cohesion in G2/M.

The first prediction from the replication fork–driven cohesion model is that DSB-induced cohesion should require recombination-dependent replication. To test this prediction, we deleted RAD52. Rad52 is a prerequisite for recombination-dependent DNA replication (13). In rad52Δ cells, the extent of DSB-induced cohesion on SRD1-MAT fragment is indistinguishable from that of RAD52 (Figs. 1D and 2A), indicating that recombination-dependent replication and/or DNA structures are dispensable for DSB-induced cohesion.

Fig. 2.

DSB-induced cohesion is independent of DNA replication but is dependent on the DNA damage response pathway. DSB-induced cohesion assay in (A) EU3286, (C) EU3291, and (E) EU3321. Error bars indicate SD (n = 3). Chromosome binding of G2/M-loaded cohesin in (B) EU3297, (D) EU3291, and EU3321. (F) DSB-induced cohesion assayed by fluorescence in situ hybridization in JH4112, JH4116, JH4115, and JH4117. See SOM for details.

The second prediction from the replication fork–driven cohesion model is that DSB-induced cohesion should occur only around DSBs and not on unbroken chromosomes. To test cohesion on unbroken chromosomes, we moved the cohesion reporter to chr. IV or chr. I while keeping the HO-cs on chr. III. We found that the loss of cohesion on chr. IV and chr. I upon inactivation of S-phase cohesin is prevented by the induction of DSBs on chr. III (Fig. 2C and figs. S1 and S2, A and B). This generation of genome-wide cohesion cannot occur by a replication-dependent mechanism because there is no ongoing DNA replication on unbroken chromosomes in G2/M.

How does a DSB in G2/M induce sisterchromatid cohesion both proximal to the lesion and genome-wide? In G2/M, in the absence of DSBs, cohesins load at CARs but cannot mediate cohesion (810). Upon DSBs, the DNA damage response pathway induces DSB-proximal cohesin loading (11). We posit that subsequently, cohesins loaded at CARs and DSBs becomes cohesive through the action of a critical factor that responds to cell cycle and DNA damage cues. In undamaged cells, this factor is active only during S phase. Hence, cohesins loaded at CARs in G2/M cannot generate cohesion. However, in response to DSBs, this factor is reactivated by upstream components of the DNA damage response pathway to generate cohesion both proximal to the DSB and at CARs genome-wide.

To identify the upstream components, we examined damage-induced cohesion in cells mutated for MRE11, MEC1/ATR (ataxia telangiectasia and Rad 3 related), TEL1/ATM (ataxia telangiectasia mutated), or H2AX. We were limited to analysis of genome-wide cohesion because only cohesin loading at CARs occurs independently of these factors (11) (Fig. 2D). In mre11Δ or mec1Δ cells, cohesion fails to form on chr. IVor on chr. XVI in response to DSBs on chr. III (Fig. 2, E and F). In contrast, neither Tel1 nor γ-H2AX is required for damage-induced cohesion (Fig. 2F and fig. S2C). Whereas all these factors are required for DSB-proximal cohesin loading, only Mre11 and Mec1 are also necessary for DSB-induced cohesion. Thus, the chromatin-bound cohesin complex is converted to a cohesive state by the DNA damage response pathway [also reported in (14)], presumably through a trans-acting factor.

A candidate for this trans-acting factor is Eco1 (Ctf7) because it is essential for cohesion establishment in S phase (6, 7). To determine whether Eco1 is also necessary for DSB-induced cohesion, we first subjected G2/M cells carrying both eco1-203 and mcd1-1 conditional alleles to a temperature shift (34°C), which inactivates only Eco1. After DSB induction, cells were subjected to a second, higher-temperature shift (37.5°C) to inactivate S-cohesin. We found that inactivation of Eco1 in G2/M results in a failure to generate DSB-induced cohesion (Fig. 3A and fig. S3, A and B). In addition, impairment of Eco1 in G2/M also compromises postreplicative repair, a process dependent on DSB-loaded cohesin (fig. S3C). Thus, Eco1 is necessary to generate sister-chromatid cohesion in G2/M in response to DSBs as well as in S phase, suggesting that DSB-induced and S-phase cohesion occurs by a similar mechanism. Furthermore, these results suggest that Eco1 is the cohesion factor that is reactivated in response to DSBs.

Fig. 3.

Eco1 is necessary to generate cohesion in G2/M and its acetyltransferase activity is required in response to DSBs. Cohesion assay in (A) EU3326, (B) EU3336, and (C) EU3307, EU3325, and EU3328. Arrow in (C) indicates over-production. Error bars indicate SD (n = 3). (D) Immunoblot showing Mcd16HA and Eco13HA with or without galactose (GAL) induction in EU3307 and EU3328.

One way of regulating Eco1 upon DSBs is through its C-terminal acetyltransferase (Ack) domain (15). Until now, the in vivo relevance of the Eco1 acetyltransferase domain has been elusive. It is dispensable for generating cohesion during S phase (16) (fig. S4A), suggesting that a distinct part of Eco1 is required for cohesion establishment. We asked whether the acetyltransferase activity of Eco1 is necessary for DSB-induced cohesion. We generated strains in which the sole copy of ECO1 is replaced by eco1R222G, K223G (eco1ack-) and monitored the cohesion of chr. III or chr. IV in response to DSBs. Similar to eco1-203, eco1 ack- cells are compromised for DSB-induced cohesion (Fig. 3B and fig. S4, B to D) and for postreplicative repair in G2/M (fig. S4E). These results show that Eco1 acetyltransferase activity is specifically required in G2/M to generate DSB-induced cohesion. Furthermore, they suggest that Eco1 has at least two distinct biological functions: One function converts the chromatin-bound cohesin complex to a cohesive state, and the acetyltransferase function activates directly or indirectly its cohesive function during G2/M.

Having established Eco1 as a critical factor for the generation of DSB-induced cohesion, we asked whether the failure to generate cohesion in undamaged G2/M cells results from limiting Eco1 activity. To test this hypothesis, we over-produced Eco1 during G2/M in the absence of HO-induced DSBs. Indeed, overproduction of Eco1 but not eco1ack- bypasses the requirement for DSBs to generate cohesion in G2/M (Fig. 3C). Both Eco1 and eco1ack- are present at similar levels (Fig. 3D), and eco1ack- encodes a functional protein because it complements the eco1-203 at nonpermissive temperature (fig. S4F). These results suggest that in G2/M, Eco1 acetyltransferase activity is limiting in undamaged cells and, in response to DSBs, this activity is elevated through the DNA damage checkpoint.

Here we show that the generation of cohesion in G2/M is Eco1 dependent but replication independent [also reported in (14)]. This contradicts the current model, which posits that cohesion generation can only occur in the context of DNA replication, and Eco1 (Ctf7) merely allows the replisome to slide through the cohesin ring in S phase (8). Rather, we suggest that Eco1 directly converts the chromatin-bound cohesin complex to its cohesive state. During S phase, Eco1 associates with replisome components (7, 17), and this allows Eco1 to establish cohesion before the sister chromatids separate (7, 18). We also show that the cohesive function of Eco1 requires its acetyltransferase domain in G2/M but not in S phase. We suggest that the cohesive function of Eco1 is inactivated after S phase either by inhibiting Eco1 directly or its accessibility to cohesin (Fig. 4A). Upon DSBs, the DNA damage checkpoint initiates a signal that induces cohesin loading around the DSB and augments the acetyltransferase activity of Eco1 (Fig. 4B). Eco1 in turn acetylates itself, cohesin subunits, or Pds5 (15), and thus overcomes the G2/M inhibition. Thus, like cohesion dissolution, cohesion establishment exhibits complex spatial and temporal regulation throughout the cell cycle, and Eco1 is the hub for this regulation.

Fig. 4.

(A and B) A model for cell-cycle and damage-induced regulation of sister-chromatid cohesion. Cohesin is represented in its noncohesive (circle) or cohesive (oval) state. (i) represents the inhibitor of cohesion generation in G2/M. See text for details.

The conservation of the Eco1 acetyltransferase domain suggests that it is critical for DSB-induced cohesion in all species. Defects in the DSB-induced cohesion pathway lead to genomic instability in humans (19). These defects have been attributed to improper repair of the broken chromosomes as a result of a failure to generate cohesion at the break site. The discovery that DNA damage response pathway activates cohesion on unbroken as well as broken chromosomes suggests that the role of this pathway extends beyond the repair of the break. Indeed, in G2/M, absence of DSB-induced genome-wide cohesion increases loss of unbroken chromosomes about threefold (fig. S5), suggesting that enhanced cohesion during checkpoint delay prevents precocious sister-chromatid separation. Further studies of genome-wide cohesion may reveal additional functions such as the prevention of rearrangements through ectopic recombination.

Supporting Online Material


Figs. S1 to S4

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article