Deacetylation of histone H3 K56, regulated by the sirtuins Hst3p and Hst4p, is critical for maintenance of genomic stability. However, the physiological consequences of a lack of H3 K56 deacetylation are poorly understood. Here we show that cells lacking Hst3p and Hst4p, in which H3 K56 is constitutively hyperacetylated, exhibit hallmarks of spontaneous DNA damage, such as activation of the checkpoint kinase Rad53p and upregulation of DNA-damage inducible genes. Consistently, hst3 hst4 cells display synthetic lethality interactions with mutations that cripple genes involved in DNA replication and DNA double-strand break (DSB) repair. In most cases, synthetic lethality depends upon hyperacetylation of H3 K56 because it can be suppressed by mutation of K56 to arginine, which mimics the nonacetylated state. We also show that hst3 hst4 phenotypes can be suppressed by overexpression of the PCNA clamp loader large subunit, Rfc1p, and by inactivation of the alternative clamp loaders CTF18, RAD24, and ELG1. Loss of CTF4, encoding a replisome component involved in sister chromatid cohesion, also suppresses hst3 hst4 phenotypes. Genetic analysis suggests that CTF4 is a part of the K56 acetylation pathway that converges on and modulates replisome function. This pathway represents an important mechanism for maintenance of genomic stability and depends upon proper regulation of H3 K56 acetylation by Hst3p and Hst4p. Our data also suggest the existence of a precarious balance between Rfc1p and the other RFC complexes and that the nonreplicative forms of RFC are strongly deleterious to cells that have genomewide and constitutive H3 K56 hyperacetylation.
GENOMIC stability is maintained by a complex interplay of DNA replication, repair, and checkpoint signaling. These processes play central roles in the maintenance of genomic stability but their mode of action in the context of chromatin is poorly understood. Newly synthesized histones deposited during DNA replication are acetylated at their N termini and in the core region (Jackson et al. 1976; Sobel et al. 1995; Hyland et al. 2005; Masumoto et al. 2005; Ozdemir et al. 2005; Xu et al. 2005; Ye et al. 2005). Lysine K56 acetylation is present on newly synthesized histone H3 and implicated in the DNA damage response (Hyland et al. 2005; Masumoto et al. 2005). This modification accumulates during S phase and is then removed either prior to or during mitosis (Masumoto et al. 2005) in a process regulated by the redundant yeast sirtuins, Hst3p and Hst4p (Celic et al. 2006; Maas et al. 2006). In hst3 hst4 mutants, K56 acetylation is observed in virtually 100% of histones throughout the cell cycle (Celic et al. 2006), although Sir2p does also contribute to deacetylation of H3 K56 in telomeric regions (Xu et al. 2007a).
Hst3p and Hst4p belong to a highly conserved family of NAD+-dependent protein deacetylases, known as the Sir2 protein family or sirtuins (Brachmann et al. 1995; Imai et al. 2000; Landry et al. 2000; Smith et al. 2000). The importance of K56 deacetylation is evident from the high level of genomic instability observed in hst3 hst4 cells. Cells lacking HST3 and HST4 show a plethora of chromatin-associated phenotypes (Brachmann et al. 1995) resulting from hyperacetylation of K56 in H3; mutation of K56 to arginine (K56R) suppresses nearly all these hst3 hst4 phenotypes (Celic et al. 2006; Maas et al. 2006).
hst3 hst4 cells also accumulate spontaneous suppressors at a high rate (Brachmann et al. 1995) and the majority of these suppressors appear to adapt to the high level of K56 acetylation rather than preventing acetylation (Miller et al. 2006). We show here that hst3 hst4 phenotypes are alleviated by overexpression of RFC1, encoding the large subunit of the clamp loader (Howell et al. 1994), supporting the notion that the inability to deacetylate K56 interferes with normal DNA replication. These phenotypes are also suppressed by inactivation of alternative clamp-loading complexes and by deletion of CTF4. We propose that CTF4, together with ASF1 (Celic et al. 2006; Recht et al. 2006), RTT109 (Schneider et al. 2006; Driscoll et al. 2007; Han et al. 2007), and HST3/HST4 define a K56 acetylation/deacetylation pathway important for the survival of replication-linked lesions induced by genotoxic agents or by collision of the replication fork with DNA-protein barriers that impinges upon clamp-loading complexes. We also show that cells lacking Hst3p and Hst4p activate a DNA damage checkpoint response due to the presence of chronic DNA damage. This is a direct consequence of K56 hyperacetylation. We show here that cells lacking Hst3p and Hst4p depend on a functional DNA damage checkpoint and a subset of repair factors for viability.
MATERIALS AND METHODS
Strains, growth conditions, and plasmids:
All strains used in this work are described in Table 1. They were generated by standard methods and grown under standard conditions unless otherwise noted. The pCEN-URA3-HST3 plasmid was previously described (Celic et al. 2006). YEP351/RFC1 is YEP351 (Hill et al. 1986) carrying an insert corresponding to chromosome XV coordinates 748644–752328 (high-copy library isolate carrying the RFC1 gene). Plasmid pJP16 is pCEN-LEU2-HHT2/HHF2. Derivative pCEN-LEU2-H3K56R was created by subcloning a SacI/XhoI insert from pDM18K56R (Park et al. 2002) into pRS415 (Sikorski and Hieter 1989).
High-copy suppressor screen:
The high-copy suppressor screen was done by transforming strain YCB828 (relevant genotype hst1 hst2 hst3 hst4 sir2 leu2 ura3) with a Saccharomyes cerevisiae genomic 2μ LEU2 library and selecting transformants on synthetic complete (SC) −Ura −Leu. Transformants were subsequently replica-plated on SC −Leu plates to segregate the resident CEN-URA3-HST3 vector and then replica-plated onto SC +5-FOA plates to select for the colonies that had lost the CEN-URA3-HST3 vector, but were able to support growth due to the presence of the library vector. Leu+, Ura− colonies were additionally tested through a plasmid segregation test, colony PCR to eliminate high-copy plasmids containing SIR2, HST3, and HST4, and finally through a retransformation assay. With this procedure we screened ∼20,000 Leu+ Ura+ colonies and isolated 82 5-FOA-resistant colonies. Twenty-five of these showed 5-FOA resistance dependent on the library plasmid. We obtained HST3 seven times, SIR2 six times, RFC1 three times, FKH1 two times, and UBP10 two times as high-copy suppressors.
Cell synchronization and FACS analysis:
Cells grown at 25° in YPD medium were arrested in G1 using 0.3 μm α-factor for 3 hr. Cells were released into the cell cycle by washing with 3–4 culture volumes of YPD and resuspending in fresh YPD medium with 0.1 mg/ml pronase (Sigma). Aliquots were collected at the indicated time intervals. DNA content was determined by flow cytometry with propidium iodide (Haase and Lew 1997).
Cells were processed as previously described (Pringle et al. 1991). Mouse anti-tubulin antibody (Sigma) was used at a 1:1000 dilution. Sheep anti-mouse secondary antibody (Amersham) was used at a 1:5000 dilution.
RNA isolation, Northern blot analysis, and microarray hybrdization:
Total RNA was isolated using the hot-acid phenol method. Probes for Northern blot were prepared by a random priming method using the Prime-It II kit (Stratagene). Total RNA was separated on 1% agarose-formaldehyde gel and hybridized to the probe using Ultrahybe hybridization solution (Ambion) according to the manufacturer's instructions. Microarray hybridization and data analysis were performed at the Johns Hopkins Microarray Core Facility (http://www.microarray.jhmi.edu). The raw data are deposited at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo); the accession number is pending. Whole-cell lysates for immunoblotting were prepared for SDS–PAGE using an alkaline method (Kushnirov 2000). For immunoblotting of Rad53p, lysates were prepared as described (Gardner et al. 2005). Depending on the experiment, lysates from 2.5 × 106 to 1 × 107 cells were resolved in SDS 4–20% (histones) or 8% (Rad53p) polyacrylamide gels and transferred to a PVDF membrane (Amersham). The blots were probed with rabbit polyclonal antibodies against the C terminus of histone H3 (Gunjan and Verreault 2003), K56-acetylated H3 (Masumoto et al. 2005), or Rad53p (Santa Cruz), followed by horseradish peroxidase (HRP)-conjugated antibody against rabbit IgGs (Amersham) and chemiluminescence detection (Amersham). As previously described (Masumoto et al. 2005), immunoblots to detect histone H3 K56 acetylation were performed with a 1000-fold molar excess of unacetylated peptide over the affinity-purified antibody to ensure that observed signals were specific for K56-acetylated histone H3.
Overexpression of Rfc1p, the large subunit of the DNA clamp loader, suppresses phenotypes of hst3 hst4 cells:
Considering that phenotypes of hst3 hst4 cells arise from K56 hyperacetylation and that the majority of spontaneous suppressors appear to adapt to the high levels of K56 acetylation (K56Ac), we have performed a genetic screen to isolate hst3 hst4 suppressors to uncover pathways that function aberrantly in the presence of K56 hyperacetylation, We conducted a high-copy screen for suppressors of the synthetic lethality of a hst1 hst2 hst3 hst4 sir2 strain (Brachmann 1996) and isolated a plasmid that contained a full-length RFC1 gene. Rfc1p is the large subunit of the “clamp loader,” which loads the PCNA clamp onto DNA during replication (Howell et al. 1994). The only other full-length ORF within this clone was the dubious ORF YOR218C, overlapping the 3′ end of RFC1. We introduced a deletion in the RFC1 ORF (from +230 to +1311) and showed that this clone lost the ability to suppress hst1 hst2 hst3 hst4 sir2 synthetic lethality (data not shown), confirming that RFC1 is indeed a high-copy suppressor of hst1 hst2 hst3 hst4 sir2 synthetic lethality. We subsequently tested whether overexpression of RFC1 suppresses the Ts phenotype of hst3 hst4 mutant cells. hst3 hst4 cells carrying the HST3 gene on a URA3-marked plasmid do not grow at 37° on 5-FOA medium. This lack of growth was suppressed by a high-copy plasmid carrying RFC1 (Figure 1A). In addition to the suppression of the Ts phenotype, overexpression of RFC1 partially suppresses the sensitivity of hst3 hst4 cells to several genotoxic agents. Overexpression of RFC1 suppressed hydroxyurea (HU), methyl methane sulphonate (MMS), camptothecin (CPT), and ultraviolet radiation (UV) sensitivity in a dose-dependent manner in hst3 hst4 cells (Figure 1B). We have previously shown that the Ts phenotype and sensitivity to genotoxic agents of hst3 hst4 cells are suppressed by the K56R mutation. This suggests that overexpression of RFC1 directly or indirectly counteracts K56 hyperacetylation in hst3 hst4 cells.
Suppression of the hst3 hst4 Ts phenotype by inactivation of alternative clamp loading complexes:
Rfc1p, together with Rfc2-5p, forms a heteropentameric complex called RFC that loads the homotrimeric PCNA ring onto DNA during replication (Tsurimoto and Stillman 1990; Fien and Stillman 1992; Cullmann et al. 1995). In addition to RFC, there are three “alternative” RFC-like complexes that share the Rfc2-5p subunits, but differ in the nature of their large subunit: Rad24p–RFC, Elg1p–RFC and Ctf18p–RFC. The function of these complexes has been linked to the DNA damage response and for Ctf18p, to sister chromatid cohesion (Shimomura et al. 1998; Green et al. 2000; Mayer et al. 2001; Naiki et al. 2001; Ben-Aroya et al. 2003). The Rad24p–RFC complex loads a specialized heterotrimeric (non-PCNA) clamp encoded by MEC3, RAD17, and DDC1, referred to as the 9-1-1 complex (Kondo et al. 1999).
We examined the effect of RAD24, ELG1, and CTF18 deletion on hst3 hst4 cells and found that deletions of the genes encoding large subunits of alternative clamp loaders efficiently suppress the Ts phenotype of hst3 hst4 cells (Figure 2A). In addition, deletion of MEC3, RAD17, and DDC1, encoding the subunits of the Rad24p-specific clamp and CTF8 and DCC1, which encode additional subunits of the Ctf18p–RFC complex (Bermudez et al. 2003) also suppress the Ts phenotype of hst3 hst4 cells (Figure 2A). We have observed suppression of the hst3 hst4 Ts phenotype by rad24Δ even in a rad9Δ background (Figure 2B). Rad9p is an important mediator of DNA damage checkpoints (Aboussekhra et al. 1996; De La Torre-Ruiz et al. 1998) that functions in parallel with Rad24p. We have previously shown that deletion of RAD9 increases UV sensitivity of hst3 hst4 cells (Brachmann et al. 1995) and reduces the DNA damage checkpoint response in hst3 hst4 cells (see Figure 6, A and B) indicating that spontaneous DNA damage in hst3 hst4 cells (see below) is partially recognized by Rad9p-mediated DNA damage checkpoint. The suppression of the hst3 hst4 Ts phenotype by rad24Δ even in a rad9 background suggests that inactivation of the Rad24p-clamp loader eliminates a requirement for RAD9-mediated checkpoint function by reducing K56 hyperacetylation-induced spontaneous DNA damage that is recognized by the Rad9p-mediated DNA damage checkpoint.
Deletion of CTF4 suppresses hst3 hst4 phenotypes:
The strongest suppression that we observed resulted from deletion of components of the Ctf18p–RFC clamp loader (Ctf18p, Ctf8p, and Dcc1p). The function of Ctf18p is related to that of Ctf4p, as both were genetically defined as chromosome transmission fidelity mutants (Spencer et al. 1990). Ctf4p is a replication fork-associated β-propeller protein (Jawad and Paoli 2002; Gambus et al. 2006; Lengronne et al. 2006) required for maintenance of genomic stability and sister chromatid cohesion (Kouprina et al. 1992; Miles and Formosa 1992; Hanna et al. 2001). This prompted us to test what effect ctf4Δ has in a hst3 hst4 background. Indeed, ctf4Δ strongly suppressed the Ts and partially suppressed the HU sensitivity phenotype of hst3 hst4 cells (Figure 3A). The strong suppression of hst3 hst4 phenotypes by ctf4Δ suggests that CTF4 may function directly in the K56Ac pathway. If this hypothesis is correct, deletion of CTF4 in K56R cells should not lead to any additional increase in sensitivity to DNA damaging agents (above the sensitivity observed with either single mutant) and we indeed observed very little increase in sensitivity in this double mutant (Figure 3B). The increase in sensitivity is at most fivefold, as determined by assaying phenotypes over a range of HU and MMS concentrations (data not shown). This result suggests that CTF4 and K56 acetylation may have some interdependent function in the maintenance of genomic integrity.
In striking contrast to the ctf4 deletion, a strong synergistic interaction was observed between the ctf18 and H3 K56R mutations (Figure 3B). These results suggest that CTF4 and K56 acetylation have a common function in the response to genotoxic agents, whereas CTF18 clearly acts via a separate pathway.
Analysis of K56 acetylation in hst3 hst4 suppressors:
Our results demonstrate strong genetic links between DNA replication clamp loaders, and the cohesion protein CTF4 on the one hand and the K56 acetylation/deacetylation pathway on the other. Overexpression of RFC1 could suppress hst3 hst4 phenotypes either by reducing K56 acetylation or by allowing hst3 hst4 cells to adapt to the high level of K56 acetylation. To determine which mechanism is in play, we analyzed K56 acetylation levels in hst3 hst4 cells carrying a high-copy RFC1 plasmid by immunoblotting with a K56Ac-specific antibody (Masumoto et al. 2005). Histone H3 K56Ac levels were equally high in hst3 hst4 cells carrying either a high-copy RFC1 plasmid or an empty vector (Figure 4A). Therefore, Rfc1p overexpression suppresses the phenotypes of hst3 hst4 mutant cells by allowing them to survive despite the persistence of K56 hyperacetylation. Next we examined the K56 hyperacetylation in ctf4 and other suppressors (ctf18, rad24, and elg1) of hst3 hst4 mutants and found that in all of these, K56 acetylation remained as high as in hst3 hst4 cells (Figure 4B). This suggests that CTF4 functions downstream of K56 acetylation and that mutations in CTF4 and RFC1 paralogs suppress hst3 hst4 phenotypes by allowing hst3 hst4 cells to adapt to constitutive K56 hyperacetylation.
Spontaneous DNA damage checkpoint activation in hst3 hst4 cells:
Although hst3 hst4 cells are sensitive to a wide spectrum of genotoxic agents that damage DNA during replication, this cannot be explained by a defect in the S-phase DNA damage checkpoint (Figure 5, A and B). Normal cells respond to DNA damage during S phase by slowing down DNA synthesis and spindle elongation, while cells defective in checkpoint functions progress through the cell cycle in the presence of damage with ultimately catastrophic consequences (Allen et al. 1994; Weinert et al. 1994; Navas et al. 1995; Paulovich and Hartwell 1995). We have analyzed DNA content in MMS-treated wild-type (WT) and hst3 hst4 cells. Wild-type and mutant cells were synchronized in G1 with α-factor, released into medium with or without 0.03% MMS, and DNA content was analyzed by fluorescence-activated cell sorting (FACS). hst3 hst4 cells slow down DNA replication in response to MMS treatment at a rate comparable to the wild type (Figure 5A), suggesting that the MMS-induced checkpoint is functional in hst3 hst4 cells. We next tested the effect of the replication inhibitor HU on cell-cycle progression and spindle elongation in hst3 hst4 and wild-type cells (Figure 5, B and C). Both wild-type and mutant cells did not elongate their spindle when treated with HU (Figure 5C). To gain further insights into the consequences of K56 hyperacetylation, we compared the genomewide transcriptional profiles in wild-type and hst3 hst4 cells. HUG1 and RNR3, two genes that are highly induced by DNA damage (Elledge and Davis 1990; Basrai et al. 1999), were the most highly upregulated genes in hst3 hst4 cells. Induction of HUG1 and RNR3 in hst3 hst4 cells suggests the activation of a chronic DNA damage response in the absence of exogenous damage. We have confirmed the microarray results by RNA blot analysis. In addition to hst3 hst4 cells showing strong upregulation of RNR3 and HUG1 (Figure 6A), we observed weaker induction of these genes in the hst3 single mutant (but no signal in the hst4 single mutant). This suggests that hst3 cells experience a low level of spontaneous DNA damage and that the double mutant is more severely affected. These results help explain synthetic fitness interactions observed between hst3 (but not hst4) mutants and several mutants affecting DNA metabolism (Tong et al. 2001; Suter et al. 2004; Pan et al. 2006) and suggest that HST3 has the more dominant role in regulation of genomic stability. This is also consistent with our observation that K56 acetylation is elevated in hst3 but not hst4 single mutants and is maximally elevated, to ∼100%, in hst3 hst4 double mutants (Celic et al. 2006). The checkpoint response to DNA damage or inhibition of DNA replication leads to Rad53p phosphorylation (Sanchez et al. 1996). In addition to upregulation of HUG1 and RNR3, Rad53p is hyperphosphorylated in normally growing hst3 hst4 cells (Figure 6B), further demonstrating activation of the checkpoint response in hst3 hst4 cells, presumably due to a form of spontaneous DNA damage.
RAD9 and RAD24 control two separate pathways required for induction of RNR3 and HUG1 and for Rad53p phosphorylation in response to DNA damage (Aboussekhra et al. 1996; De La Torre-Ruiz et al. 1998). We deleted RAD9 and RAD24 in hst3 hst4 cells and examined the levels of RNR3 and HUG1 mRNA and of Rad53p phosphorylation. Deletion of either RAD9 or RAD24 in hst3 hst4 cells resulted in reduction of RNR3 and HUG1 mRNA (Figure 6A) and Rad53p phosphorylation (Figure 6B) with RAD9 having the greater effect on HUG1 mRNA level and Rad53p phosphorylation than RAD24. Deletion of both genes showed a further reduction of, but did not completely abolish RNR3 expression, suggesting that there may be an additional pathway(s) required for residual upregulation of RNR3. Importantly, the lack of RNR3/HUG1 RNAs in hst3 hst4 rad9 rad24 mutant cells does not imply that deletion of RAD9 and RAD24 necessarily suppressed spontaneous DNA damage in hst3 hst4 mutant cells. These effects are likely due to the overlapping roles of Rad9p and Rad24p in activating Rad53p, which, in turn, is necessary for DNA damage-induced expression of RNR3 and HUG1. Induction of the latter genes is not observed in a histone H3 K56R mutant according to an expression study done on this mutant (Xu et al. 2005), suggesting the damage response of nonacetylatable chromatin is distinct from that of hyperacetylated chromatin.
K56 hyperacetylation is responsible for most if not all phenotypes observed in hst3 hst4 mutants (Celic et al. 2006). We next examined whether the presence of a single copy of a histone H3 gene with a K56R mutation would reduce Rad53p phosphorylation in hst3 hst4 mutants. Indeed, hst3 hst4 hht1K56R cells show reduced Rad53p phosphorylation, comparable to that of wild-type cells (Figure 6C). Even in hst3 hst4 hht2-hhf2 hht1K56R cells, in which the only source of histone H3 is H3K56R, the mutation reduced Rad53p phosphorylation, although not to the level of wild type, but only to the same level observed in a hht2-hhf2 hht1K56R strain, which itself shows a mildly elevated level of Rad53p phosphorylation; it is much less dramatic than in hst3 hst4 cells.
The hst3 hst4 mutant depends on the Mec1p-mediated checkpoint for viability:
In addition to recognition of DNA damage by the Rad9p- or Rad24p-dependent checkpoint, there is an additional checkpoint response that recognizes stalled replication forks (Navas et al. 1995). We investigated whether this checkpoint is activated in hst3 hst4 cells. To eliminate this DNA replication fork integrity checkpoint, we introduced the pol2-11 allele in hst3 hst4 mutant cells. This allele generates a truncated version of the DNA polymerase ε-catalytic subunit, which participates in leading strand replication (Pursell et al. 2007). The Pol2-11 protein is functional with respect to replication function at the permissive temperature, but allegedly loses a replication checkpoint function (Navas et al. 1995). We generated this mutation in hst3 hst4 mutant cells “covered” by a URA3-marked plasmid containing a wild-type copy of HST3. hst3 hst4 cells can readily lose the HST3 plasmid at the permissive temperature, because HST3 and HST4 are nonessential. If there is a synthetic lethality interaction between hst3 hst4 and the third gene, triple-mutant cells cannot grow on 5-FOA medium because they cannot lose the HST3 plasmid. After introducing the pol2-11 allele, the hst3 hst4 mutant cells became unable to segregate the HST3 plasmid (Figure 7A) at the permissive temperature, indicating that the triple-mutant combination is lethal and that the hst3 hst4 mutant potentially depends on a functional replication fork integrity checkpoint for viability. Although able to replicate at the permissive temperature, pol2-11 cells are likely to be somewhat deficient in DNA replication, on the basis of their FACS profile (Navas et al. 1995). Thus, the lethality of hst3 hst4 pol2-11 cells could result from sensitivity of hst3 hst4 cells to subtle perturbations in leading strand synthesis. Indeed, we have also observed lethality between hst3 hst4 and epitope-tagged alleles of otherwise wild-type replication proteins (Table 2). This indicates that hst3 hst4 cells are extremely sensitive to subtle perturbations in DNA replication that are well tolerated by wild-type cells or even the hst3 or hst4 single mutants. Lethality of hst3 hst4 pol2-11 cells was confirmed by generating a triply heterozygote diploid strain and performing tetrad analysis. In addition to the lethality of the triple mutant, we observe a synthetic growth defect between hst3 and pol2-11, but not between hst4 and pol2-11 (Figure 7A). This is further evidence that Hst3p plays the more prominent role in deacetylation of H3 K56. Mec1p is a central transducer of DNA damage signals, whether originating from breaks in DNA or stalled replication forks (Weinert et al. 1994). Mec1p activates Rad53p in response to DNA damage or replication blocks (Sanchez et al. 1996). This leads to the activation of the protein kinase Dun1p and transcriptional induction of numerous DNA repair genes (Zhou and Elledge 1993; Allen et al. 1994; Gasch et al. 2001). In a parallel pathway, Mec1p activates Chk1p, which leads to stabilization of the anaphase inhibitor Pds1p and arrest of the cell cycle at the metaphase–anaphase transition (Cohen-Fix and Koshland 1997; Gardner et al. 1999; Sanchez et al. 1999). Deletion of MEC1 in hst3 hst4 sml1 cells resulted in synthetic lethality (Figure 7B); this genetic interaction was confirmed by tetrad analyses. Surprisingly, deletion of RAD53 in hst3 hst4 cells did not result in lethality (Figure 7B), even though Rad53p is the direct target of Mec1p in both the DNA damage and DNA replication checkpoints (Sanchez et al. 1996; Sun et al. 1996). We also tested the effect of dun1Δ in hst3 hst4 cells and here results were mixed. We observed synthetic lethality in one strain background (the “FY” strains directly derived from S288C), but not in a related strain background (the “YPH” background derived from S288C by backcrossing into a different strain background (Kumar et al. 2003). The basis for these differences is unknown. In contrast to pol2-11, we did not observe synthetic fitness defects between hst3 and mec1 sml1 or hst3 and dun1 (in the FY background). Deletion of CHK1, which mediates the DNA damage response in parallel to RAD53, had no detectable effect on fitness of hst3 hst4 cells (Table 2). Similarly, elimination of the spindle or mitotic exit checkpoints by deletion of MAD2 and BUB2 (Gardner and Burke 2000) had no significant impact on hst3 hst4 cells (Table 2). The data presented suggest that although multiple pathways sensing DNA damage are activated in the absence of HST3 and HST4, the most important pathway required for survival of hst3 hst4 mutant cells is the DNA damage checkpoint mediated through MEC1.
hst3 hst4 cells require a subset of DNA repair proteins for viability:
The presence of spontaneous DNA damage in hst3 hst4 cells prompted us to examine genetic interaction between hst3 hst4 and various repair proteins. If hst3 hst4 cells require particular DNA repair pathways, one would expect to see genetic fitness or lethality interactions between hst3 hst4 mutations and those in the relevant DNA repair pathway. We have deleted several DNA repair proteins in hst3 hst4 strains. As described above, triple-mutant cells were grown on 5-FOA medium, allowing HST3 plasmid-free cells to grow. We observed synthetic lethality interactions between hst3 hst4 and rad52. Interestingly, hst3 hst4 cells do not require several other genes in the RAD52 epistasis group for viability, including RAD51, RAD54, RAD55, and RAD57 (Figure 8; Table 2).
hst3 hst4 cells require the MRX complex for viability (Figure 8; Table 2) and show synthetic lethality with mutations affecting all three members of this complex (xrs2, rad50, and mre11). Additionally, we observed synthetic lethality interactions with slx4 and srs2. These results demonstrate that hst3 hst4 require functional DNA repair for viability, consistent with the histone H2A S128 hyperphosphorylation observed in these cells (Celic et al. 2006), which suggests the presence of elevated levels of DNA double-strand breaks (DSBs). The lethality observed with a specific subset of repair genes suggests that hst3 hst4 cells are particularly susceptible to the absence of a specific repair pathway and hints at the existence of specific type(s) of DNA lesions caused by K56 hyperacetylation. Except for hst3 hst4 srs2, the triple-mutant lethalities that we observed can all be partially suppressed by a K56R mutation (Figure 8; Table 2).
The yeast sirtuins Hst3p and Hst4p are important for maintaining genomic stability and recent findings demonstrate that their role in regulating genomic stability is directly linked to regulation of histone H3 K56 deacetylation (Brachmann et al. 1995; Celic et al. 2006; Maas et al. 2006). Newly synthesized histone H3 molecules are acetylated at K56 and incorporated into DNA during S phase (Masumoto et al. 2005). K56Ac histone H3 incorporated into chromatin is then deacetylated in an Hst3p/Hst4p-dependent manner. Failure to deacetylate K56 has detrimental consequences for yeast cells and the resulting K56 hyperacetylation leads to accumulation of spontaneous damage and genomic instability. To gain insight into the consequences of K56 hyperacetylation, we performed a high-copy suppressor screen and isolated RFC1, the large subunit of the clamp loader that loads PCNA onto DNA during replication. Analysis of the K56 acetylation level indicated that, rather than causing a decrease in K56Ac, overexpression of RFC1 allowed cells to adapt to elevated levels of K56 acetylation. Similar results were observed upon deletion of CTF18, ELG1, and RAD24, which encode large subunits of alternative clamp loaders. Since yeast clamp loaders share four small subunits, Rfc2–5p, our suppression data suggest that persistent K56 acetylation negatively affects Rfc1p–RFC function. Suppression observed by deletion of alternative clamp loader large subunits would increase a pool of available small subunits and tip the equilibrium between different clamp loaders toward the formation of Rfc1p–RFC. Although deletion of CTF18, ELG1, and RAD24 suppressed the Ts phenotype of hst3 hst4 cells, we did not observe suppression of sensitivity to genotoxic agents in these deletion mutants, rather increased sensitivity of hst3 hst4 cells to genotoxic agents was observed (data not shown). We imagine that the increased availability of the small RFC subunits upon deletion of CTF18, ELG1, and RAD24 is sufficient to suppress the growth defect generated by K56 hyperacetylation.
The growth defect and the Ts phenotype of hst3 hst4 cells are caused, at least in part, by spontaneous DNA damage. On the other hand, treatment with genotoxic agents may create distinct DNA lesions that qualitatively differ from the consequences of K56 hyperacetylation. Under those conditions, the contribution of alternative clamp loaders to DNA repair and checkpoint signaling may be more important for the survival of hst3 hst4 cells than their ability to antagonize the Rfc1p–RFC. We have also found that deletion of CTF4 strongly suppresses the Ts phenotype of hst3 hst4 cells. In contrast to all of the other “knockout mutation” suppressors (ctf18, elg1, and rad24), only ctf4 suppressed sensitivity to HU, indicating a closer link between the response to K56 hyperacetylation and Ctf4p. For normal cellular growth, CTF4 was genetically defined as part of the K56 acetylation pathway (Collins et al. 2007), together with RTT109, ASF1, RTT101, MMS1, and MMS22. Our genetic analysis reinforces this notion and suggests that, for cellular resistance to genotoxic agents, K56 acetylation and CTF4 function together in a pathway that is parallel to and distinct from the CTF18 pathway. These pathways converge on the replication fork and promote molecular events that are necessary to rescue replication forks damaged by genotoxic agents. Ctf18p may not function strictly in parallel to Ctf4p, but may be partially controlled by Ctf4p as recruitment of Ctf18p to replication forks partially depends on Ctf4p (Lengronne et al. 2006). Ctf4p, a large β-propeller protein with many potential binding sites, could accommodate multiple functions. Considering that K56 acetylation levels are unchanged in hst3 hst4 ctf4 cells relative to hst3 hst4 cells, we believe CTF4 actually functions downstream of K56 acetylation, similarly to RTT101, MMS1, and MMS22, deletion of which does not affect K56Ac levels in hst3 hst4 cells (Collins et al. 2007). Ctf4p is a part of a large replisome progression complex (RPC) (Gambus et al. 2006) that includes the GINS complex (Kanemaki et al. 2003; Kubota et al. 2003; Takayama et al. 2003), Mcm2-7p helicase, Cdc45p, Tof1p–Csm3p complex, the histone chaperone FACT, Mcm10p, and Top1p. As part of the RPC that moves with replication forks, Ctf4p is ideally positioned to modulate replication fork integrity with the help of K56 acetylation. For instance, the loss of histone–DNA interactions mediated by H3 K56 acetylation (Masumoto et al. 2005; Driscoll et al. 2007) may facilitate the action of Ctf4p at damaged replication forks. Alternatively, Ctf4p itself or an associated protein may contain a “reading head” that directly binds to K56-acetylated nucleosomes at damaged replication forks. Rtt101p is a yeast cullin implicated in promoting replication through MMS-alkylated DNA and natural pause sites. It has been proposed (Collins et al. 2007) that Rtt101p functions in the same pathway as K56 acetylation by targeting a protein whose degradation is important to allow replisome progression through genomic regions that are inherently difficult to replicate. Our suppression analysis suggests that Rfc1p function is limiting in hst3 hst4 cells that have constitutive K56 acetylation throughout the genome. However, Rfc1p levels were not significantly affected in hst3 hst4 cells (data not shown). The K56 acetylation pathway may regulate, either directly or indirectly, Rfc1p complex formation rather than the actual protein level or the activity of the Rfc1p–RFC complex. An obvious consequence of negative regulation of Rfc1p–RFC by K56 hyperacetylation in hst3 hst4 mutants would be reduced loading efficiency of PCNA at replication forks and this may lead to defects in DNA replication and spontaneous DNA damage. However, we did not find evidence (by ChIP using anti-PCNA) that loading of bulk PCNA onto DNA was affected in hst3 hst4 cells. This may reflect the fact that Ctf18p–RFC also uses PCNA as a clamp. The nature of the clamp is unknown for Elg1p–RFC, but could be PCNA. Although overall PCNA loading appears unaltered in hst3 hst4 cells, either PCNA-associated proteins or posttranslational modifications of PCNA (Naryzhny and Lee 2004) may actually differ in the mutant cells. We hypothesize that the K56 acetylation pathway, together with Ctf4p assists Rfc1p–RFC in rescuing stalled or collapsed DNA replication forks resulting from lesions or protein barriers tightly bound to DNA. In wild-type cells, it is not known yet whether deacetylation of K56 happens immediately after fork passage or genomewide in G2. Since hst3 hst4 cells are viable, but extremely sick, the negative effect of K56 hyperacetylation on Rfc1p–RFC cannot be absolute; Rfc1p–RFC function is either modestly reduced overall or significantly reduced but only in specific genomic regions.
Another interpretation of these interactions is that even in wild-type cells, there is ongoing competition and a precarious balance between Rfc1p and the other RFC complexes. This functional antagonism between the different RFC complexes may interfere with smooth progression of replication forks. This is not a major problem for wild-type cells, but because hst3 hst4 mutant cells are acutely sensitive to subtle perturbations in DNA replication, the competition between the different RFC complexes is a serious threat to hst3 hst4 cells.
In any case, our data argue that hst3 hst4 cells replicate the genome under suboptimal conditions. This is consistent with the presence of spontaneous DNA damage in hst3 hst4 cells and their synthetic lethality observed specifically with mutations in genes implicated in DNA replication and repair. The lethality of hst3 hst4 cells occurs even with very subtle perturbations of DNA replication. For instance, a tagged but otherwise wild-type CDC45 allele that has no detectable phenotype in a wild-type cell is lethal in combination with hst3 hst4. In addition to synthetic lethality observed with DNA replication and repair genes, we have observed synthetic lethality between hst3 hst4 and some components of the DNA replication checkpoint. Interestingly, deletion of MEC1 results in synthetic lethality with hst3 hst4. Ironically, Hst3p is subjected to Mec1p-dependent degradation when cells are exposed to DNA damage (Thaminy et al. 2007). Thus it appears that Mec1p has multiple roles in the K56 acetylation/deacetylation cycle.
Interestingly, a recent report suggested that hst3 hst4 cells have a defect in sister chromatid cohesion (Thaminy et al. 2007). Conceivably, this defect could explain both the Ts phenotype and the genotoxic agent sensitivity of hst3 hst4 cells, since cohesion facilitates DNA double-strand break repair (Strom et al. 2004). However, the hst3 hst4 genetic interactions reported here are not fully consistent with this model. Ctf4p has been clearly implicated in sister chromatid cohesion (Hanna et al. 2001). Thus, the loss of Ctf4p would be expected to exacerbate the cohesion defect of hst3 hst4 cells but, contrary to this expectation, we find that CTF4 deletion rescues their Ts phenotype. Moreover, CTF4, CSM3, and TOF1 belong to the same epistasis group for sister chromatid cohesion (Xu et al. 2007b). However, while deletion of CTF4 suppresses hst3 hst4 phenotypes, deletion of other RPC subunits, like TOF1 and CSM3, actually results in synthetic lethality with hst3 hst4 (data not shown; Thaminy et al. 2007). Hopefully, a detailed molecular analysis of replisome architecture in cells lacking K56Ac or in hst3 hst4 cells that have constitutive K56 acetylation will reveal the detailed mechanism by which the cycle of K56 acetylation/deacetyation regulates genomic stability and whether or not this cycle is important uniformly throughout the genome.
The authors thank Steve Elledge for the pol2-11 construct, Boeke lab members for discussions, and Emerita Caputo for technical support. Research in A.V.'s laboratory is funded by Fonds de al Recherche en Santé Québec and the Canadian Institutes of Health Research (R0014340). This work was supported in part by a Roadmap grant from the National Institutes of Health (RR020839) to J.B.
Communicating editor: M. Hampsey
- Received March 7, 2008.
- Accepted May 7, 2008.
- Copyright © 2008 by the Genetics Society of America