|Home | About | Journals | Submit | Contact Us | Français|
DNA damage that blocks replication is bypassed in order to complete chromosome duplication and preserve cell viability and genome stability. Rad5, a PCNA polyubiquitin ligase and DNA-dependent ATPase in yeast, is orthologous to putative tumor suppressors and controls error-free damage bypass by an unknown mechanism. To identify the mechanism in vivo, we investigated the roles of Rad5 and analyzed the DNA structures that form during damage bypass at site-specific stalled forks present at replication origins. Rad5 mediated the formation of recombination-dependent, X-shaped DNA structures containing Holliday junctions between sister chromatids. Mutants lacking these damage-induced chromatid junctions were defective in resolving stalled forks, restarting replication and completing chromosome duplication. Rad5 polyubiquitin ligase and ATPase domains both contributed to replication fork recombination. Our results indicate that multiple activities of Rad5 function coordinately with homologous recombination factors to enable replication template switch events that join sister chromatids at stalled forks and bypass DNA damage.
The capability to replicate damaged DNA templates is crucial for living cells, as prolonged stalling or collapse of replication forks at DNA lesions can lead to cell death or genomic instability and cancer (Kolodner et al., 2002). In Saccharomyces cerevisiae, replication of damaged DNA requires the RAD6 epistasis group of genes (Lawrence, 1994). Genetic analysis identified two distinct DNA damage bypass pathways: translesion synthesis, error prone for most types of damage, but operating with high fidelity for UV lesions (Prakash et al., 2005), and the alternative error-free bypass, hypothesized to involve a template switch in which the blocked nascent strand uses the undamaged sister chromatid as a temporary replication template (Xiao et al., 2000). DNA damage bypass is associated with post-translational modifications of proliferating cell nuclear antigen (PCNA) (Ulrich, 2009). In the error-prone pathway, PCNA monoubiquitination on lysine-164 by the Rad6/Rad18 E2/E3 ubiquitin-conjugating complex activates translesion synthesis by damage-tolerant DNA polymerases (Waters et al., 2009). The alternative, error-free, template-switch bypass is controlled by Rad5, Mms2 and Ubc13 that form a second ubiquitin-conjugating complex (Chang and Cimprich, 2009). Rad5, through its RING-domain E3 activity, stimulates the Mms2/Ubc13-dependent synthesis of lysine-63-linked polyubiquitin chains onto monoubiquitinated PCNA (Hoege et al., 2002; Ulrich, 2009). Rad5 exhibits also a DNA-dependent ATPase activity through its helicase domain, which has homology to those of chromatin remodeling factors in the Snf2 family (Johnson et al., 1992). With limited physical data, it is not understood how the template switch bypass is induced and executed, and what roles the Rad5 activities play at stalled replication forks in vivo.
One model of template switch involves the regression of a blocked replication fork into a four-way “chicken foot” structure, possibly without any requirement for recombination proteins. Regressed forks have been visualized in vivo in mammalian cells in certain conditions (Higgins et al., 1976) and synthetic forked DNA structures can be reversed in vitro by Rad5 exclusively through its ATPase activity (Blastyak et al., 2007). However, there is no direct support for this model in budding yeast cells, in which regressed replication forks have been observed only in pathologic conditions (Sogo et al., 2002).
An alternative template-switch model entails a recombination-like invasion of the opposite undamaged sister chromatid by the blocked nascent strand either at a replication fork upon leading-strand stalling or behind the fork in the case of postreplicative filling of gaps formed opposite DNA lesions. Recombination would generate sister chromatid exchanges (SCE) and branched DNA structures with Holliday junctions between sister chromatids (Krogh and Symington, 2004; San Filippo et al., 2008). Recombination factors are required for gap-filling repair and for the formation of SCE and X-shaped DNA structures (X-DNA) in UV-treated nucleotide excision repair (NER) mutant cells, but their role is independent of Rad5 damage-bypass, and they are thought to contribute to lagging-strand lesion bypass (Kadyk and Hartwell, 1993; Neecke et al., 1999; Zhang and Lawrence, 2005; Gangavarapu et al., 2007). The X-DNA induced by methyl-methanesulfonate (MMS) exposure in sgs1 or top3 mutants has hemicatenane properties with no canonical Holliday junctions but rather ssDNA-containing sister chromatid junctions (Liberi et al., 2005). Hemicatenane formation is Rad18-dependent and thought to promote gap filling in the lagging strand behind the replication fork (Branzei et al., 2008). A template switch mechanism at stalled replication forks in the leading strand has not been identified, and possible roles for Rad5 and recombination factors in restarting replication at stalled forks are not known.
DNA damage bypass has been studied using mainly UV irradiation or MMS as DNA damaging agents. Although the corresponding lesions appear to slow the rate of replication-fork progression (Tercero and Diffley, 2001; Lopes et al., 2006), fork stalling has not been directly observed and linked to recombination-mediated restart of replication. Also, the natural occurrence of UV lesions has resulted in evolutionary selection for efficient NER and error-free translesion synthesis pathways (Prakash et al., 2005) with Rad5-dependent damage bypass playing only a lesser role in UV-damage survival. However, the intricate relationship between these repair and damage bypass pathways, and their coordination with DNA replication and recombination may be different for other types of DNA lesions.
Earlier we found that DNA alkylation by adozelesin induces site-specific fork stalling at a replication origin analyzed by two-dimensional (2D) gel electrophoresis (Wang et al., 2001). Adozelesin, a DNA alkylating drug, forms bulky adducts, which, unlike UV lesions, target the DNA minor groove (Swenson et al., 1982). This, together with its specificity for AT-rich DNA sequences (Weiland and Dooley, 1991), contributes to the site-specific fork stalling. The ability to directly detect stalled forks and other damage-induced DNA structures in replication intermediates through 2D-gel analysis can provide molecular evidence for DNA damage bypass mechanisms in vivo and help identify the roles of specific proteins. In the present work, we took advantage of this approach and used adozelesin, as well as MMS, to examine the role of Rad5-dependent damage bypass in DNA replication and the interplay with homologous recombination.
We show that Rad5 and recombination factors are required for complete replication of damaged chromosomal DNA and for the formation of X-DNA structures with Holliday junctions at forks stalled at replication origins. In the absence of Rad5 or recombination factors, stalled forks were defective in restarting replication and formed abnormal structures. The ATPase and ubiquitin ligase activities of Rad5 were both required for full damage bypass function. Our findings indicate that multiple activities of Rad5 coordinate replication template switch events at stalled forks through homologous recombination between sister chromatids.
An investigation of cell survival in a panel of yeast deletion strains defective in various DNA repair pathways, revealed that inactivation of Rad5-dependent DNA damage bypass or homologous recombination, conferred extremely high sensitivity to adozelesin when compared to inactivation of NER or translesion synthesis (data not shown). To determine whether the Rad5-dependent tolerance is related to replication over adozelesin-induced DNA lesions, we restricted DNA damage to S phase in cells progressing synchronously from a G1-block. Following a 1-hr drug exposure in S phase, cells were allowed to recover in drug-free medium for up to 18 hr (Figure 1A). We analyzed the dynamics of damaged chromosomal DNA replication in WT and rad5Δ cells by pulsed-field gel electrophoresis. The assay resolves linear chromosomal DNA from agarose-embedded cells but not DNA containing replication bubbles that is trapped inside the agarose plugs in the loading wells (Mesner et al., 2006). Accordingly, intact chromosomal DNA from G1-blocked cells was separated as individual bands, whereas, after 60 min of drug exposure in S phase, most of the chromosomal DNA from WT and rad5Δ cells was retained in the loading wells (Figure 1B), indicating ongoing DNA replication. This is consistent with the DNA content detected by flow cytometry, intermediate between 1C and 2C (Figure 1C). Chromosomal DNA from WT cells reentered the gel at 3 hr of recovery in drug-free medium, with progressively increasing amounts at later times (6 and 18 hr) due to subsequent duplications. This correlates with the flow-cytometry profiles as well as the microscopy and cell counting (Figures 1C, 1D and 1E) showing nuclear division and replicated 2C DNA at 3 hr, actively dividing cells at 6 hr and stationary cells with 1C DNA content at 18 hr.
In marked contrast to WT cells, most of the chromosomal DNA from rad5Δ cells remained in the loading well throughout the recovery period (Figure 1B), indicating that, in the absence of Rad5, damaged DNA replication was not completed. This in turn blocked nuclear division and cells failed to undergo mitosis, becoming progressively larger with time (Figures 1D and 1E). The apparently higher than 2C DNA content (Fig. 1C) at 6 and 18 hr is due partly to the dramatic increase in cell volume, which alters the optical transmission of the fluorescent signal in flow cytometry (Haase and Reed, 2002), and to the increasing mitochondrial DNA content as we did not observe higher-than-2C signals in similar experiments performed with rho0 rad5Δ mutants devoid of mitochondrial DNA (data not shown).
Similar to adozelesin, MMS also led to incomplete chromosomal DNA replication in rad5Δ cells after S-phase exposure (Figure S1). These results demonstrate that Rad5 is required for the completion of chromosome replication in the presence of alkylation DNA damage.
We previously showed that, in asynchronous WT cells exposed to adozelesin, replication-fork progression stalls at a specific site within a replication origin, ORI305 (Wang et al., 2001). The inability of rad5Δ cells to complete chromosomal DNA replication in the presence of adozelesin could be caused by prolonged fork stalling or fork collapse at such a site. Alternatively, Rad5 inactivation in the presence of DNA damage could inhibit replication origin activity.
To determine how adozelesin affects DNA replication at a molecular level in WT and rad5Δ cells progressing synchronously through S phase, we employed 2D-gel analysis of replication intermediates in a fragment containing early-firing origin ORI305 (Huang and Kowalski, 1993) (Figure 2A). Schematics of the signals resolved by this method in WT cells are illustrated in Figures 2B and 2C (top right panels).
Both WT and rad5Δ cells showed normal replication intermediates during the release from the G1 block in medium without drug (Figure 2B). ORI305 had a maximum of activity at 30 min post-release as indicated by strong bubble and late-Y signals, which decreased progressively at 45 and 60 min. A faint X signal was visible at 30 min in both WT and rad5Δ cells, representing hemicatenanes that form during normal replication (Lopes et al., 2003).
Upon release into S phase in the presence of adozelesin, the ORI305 firing was the same in WT and rad5Δ cells and similar to untreated controls in this time frame, indicating that the absence of Rad5 does not influence the origin activity. However, the presence of the drug induced a strong fork-stalling signal at ORI305 at the peak of the Y arc and an intensified the late-Y arc in both WT and rad5Δ cells (Figure 2C). The fork stalling, first seen at 30 min after the G1-block release, persisted at 45 and 60 min. The lack of an early-Y arc at 30 min indicates that the forks involved in stalling were generated locally and not at a distal origin. In WT cells, the drug also induced a strong vertical spike signal containing X-DNA. The X signal migrated similarly to the 2N linear DNA spot in the 1st dimension of electrophoresis, indicating fully replicated DNA structures containing joined sister chromatids. The X spike was most intense at its upper margin, which contains branched DNA structures with centered junctions, corresponding to the position of the fork-stalling signal in the middle of the fragment and proximal to the replication origin. Less intense signals on the X spike were associated with minor fork stalling signals on the late Y arc. The X-spike intensity increased from 30 to 45 min, along with a decrease in the bubble-arc signal, and remained strong also at 60 min. A Y to X arc signal, connecting the fork stalling signal to the upper margin of the X spike, was observed most strongly at 60 min. The convex shape of the high-rising Y to X arc resembles that of a bubble arc, as opposed to the linear-shaped signal generated by an incoming fork approaching a centrally-located stalled fork (Martin-Parras et al., 1991). In rad5Δ cells, the X-DNA was greatly reduced when compared to WT cells (Figure 2C), despite the presence of a similar fork-stalling signal. Also, the Y to X arc was eliminated, but not the faint early Y-arc due to passive replication of ORI305 by minimal incoming forks (60 min), implying that the Y to X arc contains Rad5-dependent replication intermediates elongating away from the origin following damage bypass and stalled fork restart to form fully duplicated X-DNA. Finally, the late-Y arc in rad5Δ cells showed an atypical concave shape (compared to WT), indicating the formation of abnormal replication fork structures.
We observed adozelesin-induced fork stalling also at another replication origin, ORI508 (Figure S2A). In rad5Δ cells, the fork stalling was again accompanied by deficient X-DNA formation, loss of the Y to X arc and an abnormal late-Y arc (Figure S2A). These results imply that, during adozelesin exposure, Rad5 promotes damage bypass through the formation of junctions between sister chromatids at stalled forks, giving rise to replication intermediates that form X-DNA molecules when fully duplicated. Damage bypass failure in rad5Δ cells leads to abnormal replication fork structures, possibly due to prolonged fork stalling.
The inability to complete the replication of damaged DNA and the reduction in X-DNA at fork stalling sites in rad5Δ cells, suggest that a Rad5-mediated template switch that joins sister chromatids is required to restart replication at stalled forks. To determine the fate of stalled replication forks following adozelesin exposure in S phase in WT and rad5Δ cells, we analyzed the replication intermediates during a subsequent recovery period in the absence of drug. The experimental outline was similar to that described in Figure 1A.
In WT cells, the replication intermediates at ORI305, including the fork-stalling signal and the X-DNA, still visible at 30 min of recovery, were reduced to a minimum level at 1 and 2 hr, indicating the completion of DNA replication in this region (Figure 3A), which preceded the completion of chromosomal DNA replication observed by PFGE at 3 hr (Figure 1B). As most WT cells entered a new S phase between 3 and 6 hr of recovery (Figures 1B, 1C and 1D), strong bubble and late-Y arcs reappeared at ORI305 at 4 and 6 hr with no fork stalling and a faint X signal (Figure 3A), a profile similar to that observed during an unperturbed S phase (Figure 2B). This suggests that adozelesin adducts were removed from DNA, and thus caused no more fork stalling. At 18 hr, the replication intermediates were again at minimum levels (Figure 3A), as cells accumulated in stationary phase with 1C DNA content (Figure 1C). 2D-gel analysis at early firing origin ORI508 revealed similar structures and kinetics (Figure S2A).
In contrast to WT cells, in rad5Δ cells, the fork-stalling signal at ORI305 was still present at 1 hr of recovery, along with the abnormal late-Y arc. Both the fork-stalling and the late-Y signals became progressively wider and less defined at 30 min to 1 hr of recovery (Figure 3A, brackets), indicating further alterations in the DNA structure at stalled forks in the absence of proper replication restart. At 2 hr of recovery, the replication intermediates were reduced to a low background level, which, in contrast with WT cells, was detected for up to 18 hr with no subsequent origin firing (Figure 3A). Again, a similar phenotype was observed at ORI508 (Figure S2A). These results indicate that replication forks stalled by DNA damage at early-firing origins cannot properly restart replication in the absence of Rad5 and the associated X-DNA, and progressively change into abnormal DNA structures.
Late-firing origins are repressed by the intra-S-phase checkpoint in the presence of DNA damage (Shirahige et al., 1998; Santocanale and Diffley, 1998). Accordingly, we detected no bubble arcs at late-firing origins ORI1412 and ORI501 in WT or rad5Δ cells after 1 hr of adozelesin exposure in S phase (Figures 3B and S2B). After the drug removal, in WT cells, the late origins showed activity at 30 min, 1 and 2 hr of recovery as cells completed the first S phase, as well as at 4 and 6 hr, during the subsequent S phase (Figures 3B and S2B). Interestingly, in rad5Δ cells, late origins also showed activity at 2, 4 and 6 hr of recovery (Figure 3B and S2B). Unlike early origin firing, the late origin activity was not accompanied by fork stalling in WT or rad5Δ cells, and it did not result in abundant X-DNA formation in WT cells.
Taken together, these results show that during recovery from DNA damage, Rad5 is required to form X-DNA structures and restart replication forks at early-firing origins but not at late origins, whose activity was repressed during the prior exposure to DNA damage, possibly allowing lesion repair.
The requirement of Rad5 for X-DNA formation during replication under genotoxic stress is not limited to adozelesin-induced damage. We observed a similar requirement during MMS exposure in S phase (Figure 4A). However, in contrast to adozelesin, MMS did not induce visible fork stalling at discrete sites within the ORI305-containing DNA fragment. This could be due to wide variations in size and DNA sequence specificity between MMS lesions and bulky adozelesin adducts leading to a different degree or distribution of DNA polymerase stalling and/or to different repair pathways.
To gain insight into the structure of the Rad5-mediated X-DNA molecules, we assayed whether they can be resolved under conditions that promote spontaneous branch migration of synthetic Holliday junctions in vitro (Panyutin and Hsieh P, 1994). After incubation of DNA preparations between the 1st and 2nd dimension of electrophoresis in branch-migrating conditions, the X-DNA structures induced by MMS and adozelesin in WT cells at 45 min after the G1-block release resolved as linear molecules (Figures 4A and 4B, arrows), indicating that they branch migrate. To further test the Holliday junction nature of these structures and to distinguish them from other types of branched DNA molecules described at replication origins like hemicatenanes (Lopes et al., 2003), we subjected the DNA intermediates to an in-gel digestion with RuvC Holliday junction resolvase (Dunderdale et al., 1994), prior to the 2nd dimension of electrophoresis (Zou and Rothstein, 1997). The enzymatic treatment resulted in the formation of linear molecules indicating that the damage-induced X-DNA contains Holliday junctions (Figure 4B). Inefficient RuvC activity during in-gel conditions and on DNA with bulky adozelesin lesions that also appear to hinder branch migration may account for partial linearization of X-DNA as could presence of RuvC-insensitive hemicatenanes, in addition to Holliday junctions. Finally, we also tested whether the X-DNA contains ssDNA, by incubation prior to the 1st dimension electrophoresis with mung bean ssDNA nuclease (Kowalski et al., 1976). The bubble and Y structures, including the stalled forks, were greatly reduced by this treatment. In contrast, the X-DNA was resistant (Figure 4B), indicating a relative absence of ssDNA regions as in Holliday junctions. In samples from cells not exposed to DNA damage, the faint X signal was sensitive to ssDNA nuclease treatment, confirming the presence of hemicatenanes (Figure S3). The insensitivity to ssDNA digestion distinguishes the Rad5-dependent X-DNA from hemicatenanes or other X-shaped molecules described to form in certain mutants after DNA damage and interpreted as hemicatenane-derived structures (Liberi et al., 2005). Taken together, these results indicate that X-DNA intermediates formed at sites of replication fork stalling contain genuine Holliday junctions.
The Holliday-junction structure of the Rad5-mediated damage-induced X-DNA molecules suggested that they represent recombination intermediates. To test this, we analyzed their presence in mutants that lack the main recombination factors Rad51 and Rad52. The X signal observed at ORI305 after adozelesin exposure in S phase was dramatically decreased in rad51Δ and rad52Δ mutants when compared to WT cells (Figure 5A), and was even lower than that found in rad5Δ cells (Figure 2C). The X signal decreased from 30 to 45 min and was similar to that observed in the absence of the drug (Figure S4A), suggesting that it contains hemicatenanes that form also during normal replication without any involvement of recombination factors (Lopes et al., 2003). Moreover, the Y to X arc was not present and the late-Y arc presented the same abnormal shape as in the rad5Δ mutant. We also tested cells lacking two other recombination factors, Rad54 and Rad55, and found that they were also lacked both X-DNA and a Y to X arc and presented a similar late-Y arc defect (Figure S4B). These data indicate that in WT cells, the X-DNA induced by adozelesin at stalled forks is formed through a recombination-dependent mechanism that is mediated by Rad5. To further test this association, we analyzed the replication of chromosomal DNA in rad51Δ and rad52Δ cells, by PFGE and flow cytometry as described in Figure 1. Like rad5Δ cells, the rad51Δ and rad52Δ mutants failed to complete chromosomal DNA replication after 18 hr of recovery following drug exposure in S phase, and presented a similar flow-cytometry profile (Figure 5B). In addition, like Rad5, Rad51 was also required for the restart of stalled forks, as indicated by the delayed clearance and the abnormal alteration of replication fork intermediates at ORI305 during drug-free recovery (Figure S4C). Moreover, the sensitivity to adozelesin damage in S phase of a rad5Δ rad51Δ double mutant was similar to that of rad5Δ single mutant, whereas the sensitivity of a rad14Δ rad5Δ double mutant was synergistic compared to either single mutant, indicating that the recombination factor Rad51, but not a NER factor, functions in the same pathway with Rad5 (Figure S4D). These results further indicate that forks stalled by DNA damage are restarted through a coordinated Rad5- and recombination-dependent mechanism.
Studies of Rad5 sequence and structure have identified two functional domains: a RING domain with E3 ubiquitin-ligase activity required for PCNA polyubiquitination, and a helicase domain with ATPase activity involved in DNA double-strand-break repair in vivo and regression of synthetic forked DNA structures in vitro (Chen et al., 2005; Blastyak et al., 2007).
To determine which of the two activities is required for the replication of damaged DNA, we constructed yeast strains with point mutations in the RAD5 gene that inactivate the E3 ubiquitin ligase activity (rad5-I916A) (Ulrich, 2003) or the ATPase activity (rad5-GAA) (Chen et al., 2005) or both activities in a double mutant (rad5-DM) (Figure 6A). We exposed these mutants to adozelesin in S phase and tested their survival, X-DNA recombination and capacity to complete chromosomal DNA replication. The rad5-GAA and rad5-I916A single mutants showed a similar degree of sensitivity to DNA damage in S phase, intermediate between that of WT and rad5Δ cells (Figure 6B). In rad5-DM, the concomitant inactivation of both Rad5 functions increased the sensitivity as compared to either single mutant, indicating that the two activities play, at least in part, independent roles for cell survival of DNA damage. However, the double mutant was less sensitive than the rad5Δ deletion mutant (Figure 6B), suggesting an additional Rad5 activity related to DNA damage tolerance in S phase. To rule out a possible non-specific loss of function in the rad5-DM mutant, we also tested a strain containing the rad5-GAA mutation combined with the deletion of MMS2, a gene essential for the Rad5 E3 polyubiquitination activity (Ulrich, 2009). The sensitivity of the rad5-GAA mms2Δ mutant was increased compared to that of either single mutants (Figure S5A), thus confirming the separate roles of Rad5-ATPase and polyubiquitination activities.
Furthermore, both rad5-GAA and rad5-I916A mutants showed defects in completing chromosomal replication and cell division after adozelesin damage compared to WT cells (Figures 1B and 1C), as assessed by PFGE and flow-cytometry analysis (Figure S5B). These defects were further enhanced in the rad5-DM double mutant. Importantly, all three rad5 point mutants presented reduced X-DNA formation and structural abnormalities on the late-Y arc at stalled replication forks during drug exposure in S phase (Figure 6C), similar to those observed in rad5Δ cells. These results indicate that, while the two Rad5 activities may play independent roles in DNA damage tolerance and replication completion, they function coordinately in resolving stalled replication forks and promoting efficient formation of X-DNA derived from sister chromatid recombination at stalled forks.
Despite important advances in understanding the DNA damage repair mechanisms essential for cell survival and maintenance of genetic stability, the mode of action of the Rad5-controlled error-free DNA damage bypass at stalled replication forks remained elusive and the interplay with recombination was unclear. Here we investigated the in vivo mechanism and found that Rad5 mediates the formation of recombination-dependent, X-DNA intermediates with Holliday junctions between sister chromatids at forks stalled near replication origins. Rad5 and recombination factors were required to resolve stalled forks, restart replication and complete chromosome duplication. Both the ATPase and the ubiquitin ligase activities of Rad5 were required for efficient DNA damage bypass at stalled forks, and also functioned independently but inefficiently. Our results indicate that multiple activities of Rad5 function coordinately with homologous recombination factors to join sister chromatids at stalled forks, enabling replication template switch events that bypass DNA damage.
DNA damage bypass mechanisms are required for the complete genome duplication in the presence of replication blocking lesions. One open question is whether the bypass of DNA damage is coupled with ongoing replication or whether it occurs behind replication forks. Previous studies have shown that single-strand gaps are formed on the nascent strands opposite template lesions during the replication of UV-damaged DNA (Prakash, 1981; Lopes et al., 2006). This would allow replication forks to carry on the bulk of DNA duplication with the gaps left behind being filled during late S or early G2 phase through postreplication repair mechanisms dependent on genes from the RAD6 epistasis group (including RAD5) and the RAD52 group. Whether gap formation and repair occurs for other types of DNA lesions is not known. Here we found that the inactivation of Rad5 or recombination factors results in extended persistence of replication bubbles in chromosomes following exposure to adozelesin or MMS in S phase (Figures 1B and S1). While not excluding the possibility of gap formation and repair, our data provide evidence that Rad5 and recombination factors act coordinately at stalled replication forks. First, in the absence of Rad5 or recombination factors, the replication fork stalling detected by 2D-gel analysis in early S phase during adozelesin exposure was associated with abnormal fork intermediates in the late-Y-arc region (Figures 2C, ,5A5A and S4B). Second, during recovery from DNA damage, rad5Δ and rad51Δ mutants showed a delayed clearance of stalled forks as well as increased abnormalities in replication fork intermediates (Figures 3A, S2A and S4C). In addition, simultaneous inactivation of recombination and Rad5-bypass results in a similar degree of cellular sensitivity to alkylating DNA damage in S phase as that of a rad5Δ single mutant (Figure S4D). These results indicate that Rad5 functions coordinately with recombination factors during DNA damage by restarting stalled replication forks.
We found that, like adozelesin, MMS also induces Rad5-dependent, branch-migrating, X-DNA structures in WT cells (Figure 4A). MMS is known to slow fork progression, likely reflecting fork stalling at many non-specific sites (Paulovich and Hartwell, 1995; Tercero and Diffley, 2001). Interestingly, unlike UV, MMS does not seem to induce single-strand gap formation during replication (Lopes et al., 2006). Our results with MMS are consistent with those obtained using adozelesin, revealing a requirement for Rad5-dependent bypass during the replication of damaged DNA.
Although Rad5-controlled DNA damage bypass has been previously characterized through genetic analysis (Xiao et al., 2000), its molecular aspects have remained unclear. We found that, at stalled replication forks, Rad5 promotes the formation of sister chromatid recombination products, detected by 2D-gel analysis as an X-DNA spike consisting of fully duplicated DNA molecules, through the action of Rad52-group factors. Rad52 mediates the association of Rad51, a key member of the group, with RPA-bound ssDNA to form ssDNA-Rad51 nucleofilaments, which invade homologous dsDNA and engage in strand exchange to form Holliday junctions through multiple mediators, including Rad54 and Rad55 (San Filippo et al., 2008). Earlier work on the postreplication repair of UV-damaged DNA suggested the involvement of the Rad52-group factors in lagging strand gap-filling, separate from the Rad5 pathway operating in a non-recombinational manner (Gangavarapu et al., 2007). Other studies showed that Rad18- and recombination-dependent hemicatenanes accumulate in high amounts only in certain genetic backgrounds (sgs1, top3, mms21) during prolonged MMS exposure (Branzei et al., 2008). Hemicatenanes contain ssDNA and no canonical Holliday junctions (Liberi et al., 2005) and have been interpreted as sister chromatid junctions assisting gap-filling repair (Branzei et al., 2008). Here we show that at forks visibly stalled at replication origins following DNA damage, Rad5- and Rad52-group mediated formation of an additional type of X-DNA structure in early S phase during replication initiation and in high amounts in WT cells. This X-DNA structure contains a Holliday junction as indicated by the branch migration potential, resolution by RuvC endonuclease, and lack of ssDNA. The presence of Y to X replication intermediates transitioning between stalled fork structures and fully replicated X-DNA implies that X-DNA junctions form during DNA damage bypass. Importantly, the formation of X-DNA is demonstrated to be accompanied by full chromosomal replication during recovery from DNA damage. Both rad5Δ and recombination mutants undergo post-treatment terminal G2-arrest due to incomplete DNA replication, and both present similar defects at stalled replication forks resulting in reduced X-DNA formation and deficient replication restart. Taken together, our results indicate that Rad5-dependent template switch events at stalled replication forks generate X-DNA containing products of sister chromatid recombination. Thus, while the Rad52-group factors appear to function in a postreplication repair pathway that is separate from Rad5 and serve to bypass lagging strand UV-lesions (Gangavarapu et al., 2007), our work provides evidence for the coordinate action of Rad5 and the Rad52-group in the recombinational bypass of leading-strand DNA adducts and the restart of forks stalled at replication origins.
We further expanded our investigation on the role of Rad5 in DNA damage bypass by dissecting the contributions of its specific domains. Previous work indicated a requirement for both the E3 ubiquitin-ligase and the ATPase activities of Rad5 for the postreplication repair of UV-damaged DNA (Gangavarapu et al., 2006), but their roles in DNA damage bypass at stalled replication forks are not known. We found that both Rad5 activities are required for resolving stalled forks and for full X-DNA formation during replication initiation, as well as for completion of chromosome replication and cellular tolerance of DNA damage. We envision that the Rad5 E3-ligase-dependent polyubiquitination of PCNA promotes damage bypass through a recombinational template-switch (Figure 7). This would involve limited unwinding of the blocked nascent strand followed by its recombination-factor-guided invasion into the undamaged sister chromatid and DNA synthesis past the damage site. The phenotypic similarities of the rad5 single-domain mutants and their inefficient function suggest that the Rad5 DNA-dependent ATPase activity residing in a helicase-like domain contributes to the efficiency of the E3-controlled template switch, in addition to its known independent role (see below). Unlike the ATPase function in vivo, the E3-ligase polyubiquitination of PCNA requires prior Rad18 monoubiquitination. Rad18 interacts with RPA at ssDNA regions (Davies et al., 2008), and Rad5 can bind to Rad18 and ssDNA (Ulrich, 2009). Taken together, these observations suggest the recruitment of Rad5 to ssDNA regions formed at stalled forks to promote PCNA polyubiquitination and activate the recombinational strand-invasion template switch.
The Rad5 ATPase activity is involved in the regression of synthetic fork structures in vitro (Blastyak et al., 2007). Based on this and on analogies with better studied prokaryotic systems it has been theorized that Rad5 promotes the regression of stalled replication forks with the formation of “chicken foot” structures. Regression of replication forks normally paused at a mating-type locus in fission yeast is visualized by 2D-gel analysis as a vertical spike rising up from the fork-pausing signal (Vengrova and Dalgaard, 2004). Although we did not observe such a signal at forks stalled by DNA damage, and “chicken foot” structures have only been observed under pathologic conditions in budding yeast (Sogo et al., 2002), we cannot exclude a Rad5-ATPase-dependent stalled-fork regression, which could be small in size or have a rapid turnover, thus limiting detection by 2D-gel analysis. We found that the Rad5 ATPase can function independently of the E3 activity, albeit at reduced efficiency, to form X-DNA and complete chromosome replication. ATPase-dependent fork regression may allow DNA lesion repair or DNA synthesis on the undamaged sister chromatid template, and the dsDNA end formed by nascent strand pairing may undergo a recombination-dependent recapture with Holliday junction formation and reconstitution of a functional replication fork. A free dsDNA end may also result from fork breakage by replication over lesion-induced single-strand nicks or by active fork cleavage. The Rad5-ATPase may be required for the dsDNA-end processing towards fork reassembly in a similar way it is involved in the repair of dsDNA breaks, independent of Rad18 or the E3-ligase activity (Chen et al., 2005). The ability of ATPase and E3 activities to contribute separately to DNA damage tolerance and replication completion in rad5-domain mutants indicates that Rad5 can mediate multiple mechanisms to bypass DNA damage. The cellular decision to use the ATPase alone or in combination with the E3 ligase may depend on the type and/or level of DNA damage. Overall, our results indicate that both the E3 ubiquitin-ligase and the ATPase activities of Rad5 function coordinately with homologous recombination factors to join the sister chromatids at stalled replication forks.
Formation of ssDNA at stalled replication forks due to functional DNA helicase-polymerase uncoupling, independently activates DNA damage bypass mechanisms and the intra-S-phase checkpoint (Byun et al., 2005; Davies et al., 2008). In WT cells, the bypass of DNA lesions and replication restart through Rad5 and sister chromatid recombination resolves stalled forks, likely limiting the amount of ssDNA formed during S phase. However, prolonged stalling of leading strand replication in the absence of the template-switch bypass mechanisms controlled by Rad5 and recombination factors would result in the persistence or possible extension of ssDNA at replication forks, which may account for the abnormal fork structures present at sites of stalled replication in the corresponding mutants. The persistence and aggravation of these abnormal structures during the recovery from DNA lesions suggests irreversible fork damage, a likely cause of the observed failure to complete chromosomal replication. The Rad5-dependent recombinational bypass of DNA damage is specific for early-firing origins with stalled replication forks. Unlike early origins, late origins were repressed from firing during exposure to DNA damage (Figures 3B and S2B), a known function of the intra-S-phase checkpoint (Shirahige et al., 1998; Santocanale and Diffley, 1998). Following removal of the damaging agent, late origins fired normally with no fork stalling, likely because the DNA damage was repaired during the recovery period prior to origin firing. The absence of fork stalling at late origins bypassed a need for Rad5-mediated sister chromatid recombination and fork restart.
Taken together, our results indicate that Rad5 mediates sister chromatid recombination at replication forks stalled by DNA damage, enabling the restart and completion of chromosome replication through the action of both the ATPase and the ubiquitin ligase functions. With orthologs of Rad5 identified as putative tumor suppressor genes (Unk et al., 2006; Motegi et al., 2008), better understanding of replicative DNA damage bypass mechanisms in model organisms like budding yeast may offer important clues about similar processes in mammalian cells and about their roles in maintaining genome integrity and preventing cancer.
Yeast strains used in this study (Table S1) were isogenic derivatives. Plasmids used for the construction of rad5 point-mutant strains were a gift from H. Ulrich. Standard protocols were employed for yeast genetic manipulations and culture. G1-synchronization of log-phase cultures was performed with α-factor, 150 nM (bar1 strains) or 6 μM (BAR1 strains) for 2.5 hr at 25°C. Treatments with adozelesin (U-73,975; gift of T. Beerman) and MMS and recovery from DNA damage were also performed at 25°C.
Flow-cytometry analysis of DNA content was performed as described (Paulovich and Hartwell, 1995) with the exception that cells were stained with 1 μM Sytox Green (Molecular Probes). For microscopy, cells fixed in 70% ethanol were washed and stained in phosphate-buffered saline with 0.2 μg/ml DAPI at 2×108 cells/ml. Overlay images of DAPI fluorescence and bright field were acquired on an Axioskop microscope (Carl Zeiss, Inc).
For PFGE, cells were embedded in 1% agarose and analyzed with a CHEF-DR II system (Bio-Rad) following the manufacturer’s instructions.
For 2D-gel electrophoresis, genomic DNA from 109 cells was isolated by CsCl gradient centrifugation, digested with EcoRI and FspI restriction endonuclease and analyzed as described (Huang and Kowalski, 1993). The sequences of the 32P-labeled DNA probes are available upon request. The radioactive signals were detected, analyzed and quantified with a STORM PhosphorImager and ImageQuant software (Molecular Dynamics). The quantified values presented are representative for the signals obtained in three experimental repeats.
Agarose slices with replication intermediates were incubated between the 1st and 2nd dimensions of electrophoresis in RuvC cleavage buffer (Zou and Rothstein, 1997) with or without 10 μg RuvC (MBL Intl Corp) for 5 hr at 37°C, or in branch-migrating buffer for 4 hr at 65°C as described (Panyutin and Hsieh, 1994). For ssDNA endonuclease treatment, replication intermediates were incubated prior to the 1st dimension electrophoresis with 1 U mung bean nuclease for 1 hr at 37°C.
This research was supported by National Institutes of Health grant GM30614 to D.K..
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.