Hst3 Is Regulated by Mec1-dependent Proteolysis and Controls the S Phase Checkpoint and Sister Chromatid Cohesion by Deacetylating Histone H3 at Lysine 56*

The SIR2 homologues HST3 and HST4 have been implicated in maintenance of genome integrity in the yeast Saccharomyces cerevisiae. We find that Hst3 has NAD-dependent histone deacetylase activity in vitro and that it functions during S phase to deacetylate the core domain of histone H3 at lysine 56 (H3K56). In response to genotoxic stress, Hst3 undergoes rapid Mec1-dependent phosphorylation and is targeted for ubiquitin-mediated proteolysis, thus providing a mechanism for the previously observed checkpoint-dependent accumulation of Ac-H3K56 at sites of DNA damage. Loss of Hst3-mediated regulation of H3K56 acetylation results in a defect in the S phase DNA damage checkpoint. The pathway that regulates H3K56 acetylation acts in parallel with the Rad9 pathway to transmit a DNA damage signal from Mec1 to Rad53. We also observe that loss of Hst3 function impairs sister chromatid cohesion (SCC). Both S phase checkpoint and SCC defects are phenocopied by H3K56 point mutants. Our findings demonstrate that Hst3-regulated H3K56 acetylation safeguards genome stability by controlling the S phase DNA damage response and promoting SCC.

Post-translational modification of histone proteins has been implicated in controlling many aspects of chromosome biology. Among the recently described histone modifications is acetylation of histone H3 on lysine 56 (Ac-H3K56) located at the lateral surface of the nucleosome core (1)(2)(3)(4). Ac-H3K56 appears in S phase on newly synthesized histone H3 prior to its deposition onto chromatin, and Lys 56 is deacetylated later in S phase following its deposition (5). Acetylation of H3K56 requires a histone chaperone, Asf1 (6,7), and is carried out by a recently characterized acetyltransferase Rtt109 (8 -11). Regulation of acetylation is critical, since cells with a mutant form of histone H3 that cannot be acetylated are DNA damage-and hydroxyu-rea (HU) 3 -sensitive. In the presence of DNA damage, persistent checkpoint-dependent H3K56 acetylation suggested an interplay between classical DNA damage checkpoint proteins (e.g. Mec1) and regulators of H3K56 acetylation.
In response to genotoxic stress, Mec1, acting as a DNA damage sensor, activates the DNA damage response network by phosphorylating Rad53. Rad9 serves as an adaptor (12) that facilitates Mec1-mediated phosphorylation of Rad53 throughout the cell cycle. In addition, and specifically during S phase, Mec1 is capable of efficiently phosphorylating Rad53 in a RAD9-independent fashion (13,14). The ability of Mec1 to transmit a DNA damage signal during S phase in the absence of RAD9 depends on several genes (e.g. TOF1, CSM3, and MRC1) that constitute an S phase-specific branch of the DNA damage response network (15). In addition to DNA damage signal transduction from Mec1 to Rad53, the members of the S phasespecific branch of the DNA damage response network play a separate role in ensuring the stability of the stalled replication forks (16,17) and enabling the establishment of faithful sister chromatid cohesion SCC (18,19). Their role in SCC appears to be distinct from DNA damage signaling, because mec1 mutants, although completely deficient in DNA damage-induced signal transduction, have intact SCC (19). S phase checkpoint proteins have been found to travel with the replication forks; in their absence, replication forks are prone to collapse and breakage (15), thus creating endogenous DNA damage during S phase and genome instability (20).
Two reports (21,22) implicated the functionally redundant NAD-dependent deacetylases Hst3 and Hst4 and their DNA damage and Mec1-mediated transcriptional control (22) as regulators of H3K56 acetylation. However, the published studies have not determined how H3K56 regulation contributes to genome stability. In this paper, we report that Hst3 has NADdependent deacetylase activity in vitro, and we confirm Hst3 functions in vivo during S phase as an Ac-H3K56 deacetylase. Furthermore, we show that the Hst3 protein level is regulated by Mec1-dependent proteolysis during genotoxic stress. Lack of Hst3-mediated regulation of H3K56 generates a defect in the DNA damage signal transduction from Mec1 to Rad53, increases formation of foci containing endogenous DNA damage during S phase, and impairs SCC. These observations suggest Hst3 and H3K56 acetylation safeguard genome stability by participating in the S phase-specific branch of the DNA damage response pathway.

EXPERIMENTAL PROCEDURES
Yeast Media and Strains-Yeast strains were grown in synthetic complete medium (SC) or selective synthetic drop-out medium containing 2% glucose or 2% galactose when mentioned. The yeast strains used in this study are listed in the supplemental materials (supplemental Table 1).
Cell Cycle Synchronization-For G 1 arrest experiments, cultures were grown at 30°C to logarithmic phase and arrested with ␣ factor (Zymo Research) at a final concentration of 5 M. Cultures were incubated with ␣ factor until at least 95% of the cells were G 1 -arrested. For G 2 /M arrest experiments, cells were arrested with 15 g/ml nocodazole. Cells were released from ␣ factor or nocodazole arrest into rich medium, and samples were taken at the indicated time points.
DNA Damage Sensitivity Assay-5-Fold serial dilutions of yeast cells were plated on rich media containing the indicated concentration of either HU or methyl methanesulfonate (MMS). Plates were photographed after 3-4 days of growth at 30°C. For MMS treatment in liquid culture, cells were grown to logarithmic phase, harvested, and resuspended in rich media containing 0.032% MMS.
Flow Cytometry-Cells were washed with distilled H 2 O, fixed in 80% ethanol at 4°C, and prepared using Sytox Green for DNA staining. Position in the cell cycle was assessed by flow cytometry analysis of DNA content. The DNA fluorescence was measured by using a FACScanR (BD Biosciences), and data analysis was done using FlowJo software.
Sister Chromatid Cohesion Assay-Strains were grown overnight at 30°C in YEPD containing 0.01% adenine. Cells were diluted to 2.5 ϫ 10 6 cells/ml in the same medium supplemented with 25 M CuSO 4 to induce GFP expression. The cultures were grown to logarithmic phase at 30°C and G 1 -arrested with 2.5 M ␣ factor, or G 2 /M-arrested with 15 g/ml nocodazole. After G 1 (␣ factor) or G 2 /M arrest (nocodazole), 1 ml of culture was harvested and fixed in 100 l of paraformaldehyde at room temperature for 15 min. Cells were washed in 1 M KHPO 4 , 50 mM sorbitol, sonicated, and analyzed by fluorescence microscopy.
Purification of Recombinant Protein and Deacetylation Assay-A histone deacetylase assay was performed using bacterially expressed and purified GST-Hst3 (aa 46 -377) and GST-Sir2 proteins as previously described (24). The GST-Sir2 expression plasmid was a gift from J. Boeke (25). The truncated Hst3 (aa 46 -377) and Sir2 proteins were expressed as GST fusion proteins from Amersham Biosciences pGEX plasmids in the Escherichia coli DH5-␣ strain. The cells were grown in LB with 100 mg/liter ampicillin at 37°C. When the A 600 reached 0.8 -1.0, 1 mM isopropyl 1-thio-␤-D-galactopyranoside was added. The cells were then induced at room temperature and harvested after 5 h of induction. The purification was performed according to the methods recommended by the manufacturer.
A histone deacetylase assay was performed using bacterially expressed and purified proteins as previously described (24). Histone H4 N-terminal peptide (SGRGKGGKGLGKGGAK-RHR; aa 2-20) and histone H3 peptide (GTVALREIRRFQKS-TELLIR; aa 44 -63) with the N-terminal amino group acetylated during peptide synthesis were acetylated chemically using the HDAC assay kit (Upstate Biotechnology). One microgram of purified truncated Hst3 or Sir2 proteins was incubated with [ 3 H]acetylated histone H3 or H4 peptide (20,000 cpm) with or without 1 mM NAD and with or without 5 mM nicotinamide in a 20-l reaction containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 1 mM dithiothreitol overnight at 30°C. The reaction was stopped by the addition of 5 l of 1 N HCl and 0.15 N acetic acid. Released [ 3 H]acetate was extracted with 500 l of ethyl acetate, and 250 l of the extract was subjected to scintillation counting.

RESULTS
Hst3 Is an NAD-dependent Deacetylase That Controls H3K56 Acetylation in Vivo-In order to determine whether Hst3 has NAD-dependent deacetylase activity, we expressed and purified a truncated form of Hst3 (Fig. 1A, left) that retained the conserved NAD-dependent deacetylase domain but lacked 45 N-and 70 C-terminal amino acids (retaining aa 46 -377) and used a chemically acetylated 20-aa peptide derived from the histone H3 globular domain or a tetraacetylated histone H4 N-terminal peptide as deacetylation substrates. We found that Hst3 deacetylates acetyl-K56 in vitro in an NAD-dependent fashion (Fig. 1A, top right). Hst3 also deacetylated a tetraacetylated histone H4 N-terminal peptide, and this NAD-dependent deacetylase activity was inhibited by nicotinamide (Fig. 1A, bottom right). These results demonstrate that Hst3 is an NAD-dependent histone deacetylase.
To determine whether HST3 deacetylates Ac-H3K56 in vivo, we used an antibody that recognizes Ac-H3K56 to monitor this protein modification in wild type and hst3 hst4 mutants in logarithmically growing cells, in ␣ factor-induced G 1 -arrested cells, and in synchronized cells following release from ␣ factor. We found hyperacetylation of H3K56 in cells lacking HST3 and HST4 in both G 1 -arrested and logarithmically growing cells (Fig. 1B). After release from ␣ factor, in wild type cells, H3K56 acetylation peaked at 50 min and then decreased, whereas the Ac-H3K56 level in strains lacking HST3 and HST4 was markedly elevated throughout the cell cycle, and its fluctuation was largely abolished (supplemental Figs. 7A and 8A). These results suggest that Hst3 deacetylates H3K56 in vivo. Furthermore, consistent with a requirement for the enzymatic activity of Hst3 in genotoxic stress resistance and Ac-H3K56 deacetylation, we found that point mutants in Hst3 known to abolish enzy-matic activity of NAD-dependent deacetylases (27) or treatment with nicotinamide phenocopied the DNA damage sensitivity and H3K56 hyperacetylation of HST3,4 deletion mutants (supplemental Fig. 7, B and C). These findings are consistent with those previously published by other groups (21,22).
Hst3 level, similarly to Ac-H3K56, fluctuates during the cell cycle, peaking at the end of S phase 60 min after release from ␣ factor, declining in mitosis, and is low in the subsequent G 1 phase (supplemental Fig. 7A). In order to determine whether the observed decline of Hst3 following S phase is mediated through proteasome degradation, cells were released into S phase from ␣ factor-induced G 1 arrest into medium containing the proteasome inhibitor MG132, and Hst3 level was monitored by Western blot. Proteasome inhibition led to an increase in Hst3 level and the appearance of multiple Hst3 forms with very low mobility, suggestive of polyubiquitylation (Fig. 1C, top). Probing of an Hst3 immunoprecipitate from a log phase culture in the presence of MG132 with a specific anti-ubiquitin antibody confirmed that low mobility forms of Hst3 correspond to polyubiquitylated forms of the protein (Fig. 1C, bottom). These results suggest that ubiquitin-mediated proteolysis contributes to cell cycle fluctuations in Hst3 levels.
Hst3 is down-regulated during DNA damage response in an MEC1-dependent fashion. The previously published observation that Ac-K56H3 persists in cells exposed to DNA-damaging agents in a DNA damage checkpoint fashion (2) led us to test the hypothesis that Hst3 level is regulated by genotoxic stress. Wild type and checkpoint-deficient mec1 cells were released from ␣ factor-induced G 1 arrest in media with or without 0.032% MMS. Hst3 levels were monitored using Western blots (Fig. 1D) and cell cycle progression by flow cytometry (supplemental Fig. 8A). We observed a marked difference in Hst3 kinetics between cells released in the presence or absence of MMS. Unlike cells released in medium lacking MMS, which showed an increase in Hst3 level that peaked at 60 min, cells released in MMS did not A (left), purification of truncated GST-Hst3. Because bacterial expression of full-length Hst3 was complicated by poor solubility, the fragment of Hst3 containing amino acids 46 -377 fused to GST was expressed in bacteria using a pGEX-T1-HST3 (aa 46 -377) bacterial expression plasmid, purified on glutathione beads, and eluted with 10 mM reduced glutathione. The samples were run on a 7.5% SDS-polyacrylamide gel and stained with Coomassie Brilliant Blue. Lane 1, protein molecular mass standard marker. Lane 2, GST construct. Lanes 3-5, elutions of GST-Hst3 (aa 46 -377). The arrow indicates the purified truncated Hst3. Right, chemically [ 3 H]acetylated histone H3 (top) and H4 (bottom) peptides were incubated with 1 g of purified recombinant protein GST, Sir2, or Hst3 (aa 46 -377) with and without 1 mM NAD at 30°C overnight without (gray bars) or with (black bars) 5 mM nicotinamide. NAD-dependent histone deacetylation was measured by counting the amount of tritiated acetate released from the peptide. B, hyperacetylation of H3K56 in hst3 hst4 deletion mutants. Strains expressing Hst3-TAP (14500), labeled WT, and strains lacking HST3 and HST4 (14299), labeled ⌬hst3,4, were grown to midlog phase and arrested in G 1 with ␣ factor. Samples were taken at the indicated cell cycle phase (log or G 1 ) and analyzed by Western blot using anti-Ac-H3K56 and anti-actin antibodies. C, ubiquitylation and proteasome-mediated degradation control Hst3 levels during the cell cycle. Upper panel, the HST3-TAP pdr5 strain (14611) was synchronized in G 1 with ␣ factor and released from the arrest in medium containing the proteasome inhibitor MG132. The pdr5 mutation prevents MG132 efflux. Samples were taken at the indicated time points and processed for Western blot analysis. An anti-TAP antibody was used to evaluate level and mobility of Hst3. Bottom, strain carrying GAL-HST3-TAP pdr5 (14608) or control HST3 pdr5 strain were grown in galactose to log phase and treated with MG132 for 3 h. Whole cell lysates were prepared using denaturing conditions and subjected to immunoprecipitation using an anti-TAP antibody. The immunoprecipitates from the HST3-TAPexpressing strain (IP) and from the control strain (IPc) were analyzed by Western blot (WB) using an anti-TAP antibody to visualize Hst3 and an anti-ubiquitin antibody to visualize polyubiquitylated proteins. D, Hst3 is regulated in a MEC1-dependent fashion during DNA damage response. WT (HST3-TAP sml1, strain 14611) and mec1 (HST3-TAP sml1 mec1, strain 14652) strains were grown to midlog phase, arrested in G 1 with ␣ factor, and released in the absence or presence of 0.032% MMS. Samples were taken at the indicated times and processed for Western blot analysis using anti-TAP antibody. Cell cycle progression was monitored by flow cytometry after staining DNA with Sytox (supplemental Fig. 8A). DECEMBER 28, 2007 • VOLUME 282 • NUMBER 52

Mec1-dependent Proteolysis of Hst3
show a significant increase in Hst3 level. Consistent with previous observations (2), MMS treatment was associated with increased levels of Ac-H3K56 (supplemental Fig. 8A). In contrast to wild type cells, MMS treatment did not alter Hst3 kinetics in checkpoint-deficient mec1 cells (Fig. 1D). This result links MEC1 with down-regulation of Hst3 protein in response to DNA damage. Importantly, down-regulation of Hst3 during genotoxic stress is critical for the ability of cells to survive genotoxic stress, since we found that overexpression of Hst3 resulted in MMS sensitivity comparable with that of cells lacking HST3 and HST4 (supplemental Fig. 8, B and C). This result is consistent with earlier reports by Maas et al. (22) and underscores the importance of tight control of Hst3 protein level during genotoxic stress.

Hst3 Is Subjected to Mec1-dependent Phosphorylation and Ubiquitylation and Proteasome-mediated Degradation during
Genotoxic Stress-Our observation that Hst3 is subjected to ubiquitylation and proteasome-mediated degradation led us to investigate the effects of proteasome inhibition on Hst3 level during genotoxic stress. Wild type and checkpoint-deficient mec1 cells were released into S phase from ␣ factor-induced G 1 arrest in medium containing MMS and MG132. Hst3 protein level was monitored by Western blot and S phase progression by flow cytometry. We found that proteasome inhibition led to increased Hst3 level during genotoxic stress; furthermore, most of the proteasome inhibitor-stabilized Hst3 had lower mobility relative to Hst3 in logarithmically growing cells ( Fig. 2A). Treatment with phosphatase increased the mobility of Hst3 obtained FIGURE 2. Hst3 is subjected to Mec1-dependent phosphorylation and ubiquitylation and proteasome-mediated degradation during genotoxic stress. A, the strains HST3-TAP sml1 (14611), labeled WT, HST3-TAP mec1 sml1 (14652), labeled ⌬mec1, and HST3-TAP mec1 tel1 sml1 (14804), labeled ⌬mec1 ⌬tel1, were grown to midlog phase, arrested in G 1 with ␣ factor, and released in medium containing MG132 with or without 0.032% MMS. Samples were taken at the indicated times and processed for Western blot analysis (left) and flow cytometry (right). Anti-TAP antibodies were used to measure levels of Hst3. Proteasome inhibition prevents genotoxic stress-induced decline in Hst3 level. The arrow points to a lower mobility form of Hst3 that is shown in Fig. 2B to correspond to phosphorylated Hst3. Rapid phosphorylation of Hst3 is dependent on Mec1, whereas the residual phosphorylation is Mec1-independent and reflects cell cycle progression rather than DNA damage response. B, Hst3 is phosphorylated during genotoxic stress. A whole cell lysate from cells treated with MMS and MG132 (80 min time point from Fig. 2A) was prepared under denaturing conditions and was treated with -phosphatase with and without phosphatase inhibitor mixture for 30 min. Hst3 mobility was evaluated by Western blotting using anti-TAP antibodies. Phosphatase treatment increases Hst3 mobility. C, Mec1 promotes ubiquitylation of Hst3 during genotoxic stress. The HST3-TAP sml1 strain (14611), labeled WT, and HST3-TAP mec1 sml1 strain (14652), labeled ⌬mec1, were arrested in G 1 with ␣ factor and released into medium containing 0.032% MMS and MG132. Samples were taken at the indicated times and processed for Western blot analysis. The anti-TAP antibody was used to analyze Hst3 level and mobility, and actin was probed as a loading control. The arrow in the upper panel indicates the phosphorylated form of Hst3; the polyubiquitylated forms of Hst3 (polyub Hst3) are indicated in the middle panel. D, genotoxic stress promotes MEC1-dependent proteolysis of Hst3. Logarithmically growing wild type (HST3-TAP sml1, strain 14611) and mec1 cells (HST3-TAP mec1 sml1, strain 4652), were resuspended in medium with or without 0.032% MMS. The samples were collected immediately before (time 0) or at the indicated time points following the addition of cycloheximide (35 g/ml) and analyzed for Hst3 and actin level. The intensity of the bands corresponding to Hst3 was quantified by densitometry and normalized to actin level, and the time Hst3/action curve was fitted using Prism software to calculate Hst3 half-life with and without MMS (wild type, 8.5 and 3.5 min; mec1, 9 and 8 min with and without MMS, respectively).
from MMS-and MG132-treated cells (Fig. 2B), demonstrating that the lower mobility forms of Hst3 under genotoxic stress correspond to phosphorylated protein. Thus, proteasome inhibition specifically stabilizes phosphorylated forms of Hst3. The phosphorylated form of Hst3 during genotoxic stress could not be detected without proteasome inhibition, presumably due to its short half-life. Phosphorylation of Hst3 appeared to be reduced in mec1 cells relative to wild type cells (Fig. 2A). Because mec1 cells exposed to MMS still exhibited residual phosphorylation of Hst3, we wanted to determine whether this residual phosphorylation is eliminated in cells lacking both MEC1 and TEL1 (Fig. 2A). Tel1 acts in parallel with Mec1 as a DNA damage sensor, and the two kinases often phosphorylate common substrates (15). Cells lacking both MEC1 and TEL1 exhibit the same degree of deficiency of Hst3 phosphorylation as mec1 cells ( Fig. 2A), suggesting that the observed residual phosphorylation is not Mec1/Tel1-dependent. In order to evaluate the possibility that MEC1-independent phosphorylation of Hst3 does not occur in response to DNA damage but instead reflects cell cycle progression over the course of the experiment, we released mec1 cells from ␣ factor arrest into medium containing MG312 only. In this condition, both wild type and mec1 cells exhibited Hst3 phosphorylation comparable with the residual phosphorylation observed in mec1 cells during genotoxic stress ( Fig. 2A); thus, MEC1-independent phosphorylation is not a response to genotoxic stress. We conclude that the rapid and almost complete phosphorylation of Hst3 that occurs during genotoxic stress is dependent on MEC1, whereas the residual phosphorylation of Hst3 observed in the absence of MEC1 is independent of genotoxic stress and depends instead on cell cycle progression.
Rapid Mec1-dependent phosphorylation of Hst3 during genotoxic stress, coupled with the requirement for Mec1 protein kinase for down-regulation of Hst3 protein level, led us to evaluate the possibility that Hst3 protein is targeted for proteasome-mediated proteolysis during genotoxic stress through Mec1-dependent phosphorylation. We found that phosphorylation of Hst3 in wild type cells is accompanied by the appearance of ubiquitylated protein forms that are barely detectable in mec1 cells, consistent with phosphorylationdependent ubiquitylation of Hst3 (Fig. 2C). To determine whether increased Mec1-mediated ubiquitylation of Hst3 during genotoxic stress promotes its instability, we next evaluated the effects of MMS on Hst3 protein stability in wild type and mec1 cells. We found that Hst3 protein has a half-life of 8.5 and 9 min in wild type and mec1 cells, respectively (Fig. 2D). Genotoxic stress decreases the half-life of Hst3 to 3.5 min in wild type cells, and this decrease appears to be dependent on Mec1, since we observe that Hst3 halflife does not significantly decrease in a mec1 mutant (9 min versus 8 min without and with MMS, respectively). These findings show that Hst3 level during genotoxic stress is regulated through Mec1-dependent proteolysis. Together, our results support the hypothesis that genotoxic stress leads to rapid Mec1-dependent phosphorylation of Hst3, its ubiquitylation, and subsequent proteasome-mediated degradation.
HST3 and HST4 Participate in the S Phase-specific Branch of the DNA Damage Response Network through Control of H3K56 Acetylation-The finding that Hst3 is expressed during S phase combined with the observation that hst3 hst4 mutants are sensitive to DNA-damaging agents led us to evaluate these mutants for sensitivity to HU. Because MMS sensitivity, created by the loss of HST3 and HST4, has been previously shown to be synthetic with the lack of RAD9 (28), we also tested hst3 hst4 rad9 triple mutants for sensitivity to HU. We found that these triple mutants have a growth defect and are sensitive to concentrations of HU that do not have an appreciable effect on growth of either hst3 hst4 double mutants or rad9 single mutants (Fig.  3A). This result demonstrates that loss of HST3 and HST4 makes cells vulnerable to replication stress in the absence of RAD9.
The synthetic phenotype of the lack of both HST3 and HST4 and of RAD9 under conditions of genotoxic stress is consistent with the properties of genes that control the S phase-specific branch of the DNA damage response pathway, such as MRC1, CSM3, or TOF1 (14,29). In order to determine whether lack of HST3 and HST4, similar to the lack of other S phase-specific checkpoint genes, creates a defect in Rad53 phosphorylation, we first evaluated hst3 hst4 double mutants for DNA damageinduced up-regulation and phosphorylation of Rad53. Rad53 phosphorylation, following MMS-inflicted DNA damage, was monitored by Western blot through the change in the electrophoretic mobility of HA-tagged Rad53. We found that cells lacking HST3 and HST4 have a modest defect in both Rad53 induction and MMS-induced Rad53 phosphorylation (Fig. 3B,  top and bottom). Importantly, this defect was further exacerbated in the triple hst3 hst4 rad9 mutants (Fig. 3B, bottom); the deficiency in DNA damage-induced phosphorylation of Rad53 in this triple mutant was comparable with that of mutants lacking MEC1. This result demonstrates that the combination of deletion of hst3, hst4, and rad9 causes a deficiency in the DNA damage response and supports the idea that Hst3 and Hst4 play an important role in the S phase-specific branch of the DNA damage response pathway. We also observed that the combined lack of HST3, HST4, and RAD9 creates a partial defect in delaying S phase progression in response to genotoxic stress (supplemental Fig. 9A). Deletion of the RTT109 gene, which encodes for an H3K56 acetyltransferase, also created synthetic sensitivity to HU (Fig. 3C) and synthetic Rad53 phosphorylation defect (Fig. 3D) with the loss of RAD9. Importantly, HST3 and HST4 loss of function phenotypes, including synthetic HU sensitivity with the loss of RAD9 (Fig. 3E), Rad53 phosphorylation defect (Fig. 3F), and S phase delay defects (supplemental Fig. 9B), were phenocopied by H3K56 point mutants, demonstrating that loss of H3K56 regulation is the critical event that leads to the observed checkpoint phenotypes in cells lacking HST3 and HST4. The observation that the nonacetylatable allele of histone H3 (K56R), the H3K56 mutant mimicking constitutive acetylation (K56Q), the deletion of HST3 and HST4, or the deletion of RTT109 all generated similar phenotypes suggests that reversible acetylation of H3K56 participates in the intra-S phase checkpoint. DECEMBER 28, 2007 • VOLUME 282 • NUMBER 52

HST3 and HST4 Loss of Function Mutants Accumulate
Endogenous DNA Damage during S Phase-During the analysis of the role of HST3 and HST4 in Rad53 phosphorylation, we observed an Rad53 phosphoshift in hst3 hst4 mutants even in the absence of treatment with MMS (Fig. 3B, top). Furthermore, the Rad53 phosphoshift generated by the loss of HST3 and HST4 was abolished by deletion of RAD9, consistent with the activation of the DNA damage response by endogenous DNA damage. In order to determine whether hst3 hst4 mutants, like mutants in other intra-S phase DNA damage response genes (e.g. TOF1 and MRC1), exhibit endogenous DNA damage during DNA replication (12, 22), we evaluated the spontaneous formation of DNA damage-associated checkpoint protein foci in hst3 hst4 cells during S phase (as determined by budding). Ddc2, a checkpoint protein that physically associates with Mec1, is recruited to sites of DNA damage, and Ddc2-GFP foci are observed at sites of DNA damage during checkpoint activation (20). We compared the frequency of cells with Ddc2-GFP foci in logarithmically growing wild type cells and hst3 hst4 double mutants. As previously reported in wild type (20), only a small fraction of cells, 5% of total cells or 9% of budded cells, which are enriched for cells in S phase, exhibited spontaneous Ddc2 foci (Fig. 4, A and B)  , and rad9 rtt109 RAD53-3HA (15029) strains were grown to midlog phase, harvested, and diluted in rich media or rich media containing 0.032% MMS. The Rad53-3HA level was detected by Western blotting with an anti-HA antibody. Actin was visualized, with an anti-actin antibody, as a loading control. E, H3K56 mutants and loss of RAD9 exhibit synthetic HU sensitivity. The strains used in these experiments are hht2-hhf2 hht1-hhf1 mutants. A wild type copy of HHT1-HHF1 (labeled WT) or an hht1 HHF1-K56Q (labeled K56Q) or hht1 HHF1-K56R (labeled K56R) allele was integrated at the trp1 locus. Yeast cells carrying a wild type copy of histone H3 (14598), an H3K56Q mutation (14596), an H3K56R mutation (14597), a wild type copy of histone H3 in a rad9 strain (14621), an H3K56Q mutation in a rad9 strain (14619), a histone H3K56R mutation in a rad9 strain (14620), and a wild type copy of histone H3 in a mec1 strain (14622) were diluted and spotted on YEPD medium in the absence or presence of 40 mM HU. The plates were incubated for 3 days at 30°C. All strains are deleted for sml1. F, H3K56 mutants and loss of RAD9 create a synthetic Rad53 phosphorylation defect. Strains with 3HA-tagged Rad53 and carrying a wild type copy of histone H3 (14647), an H3K56R mutation (14645), an H3K56Q mutation (14646), a wild type copy of histone H3 in a rad9 strain (14647), an H3K56R mutation in a rad9 strain (14648), an H3K56Q mutation in a rad9 strain (14649), and a wild type copy of histone H3 in a mec1 strain (14624) were grown to midlog phase, harvested, and resuspended in rich media in the absence or presence of 0.032% MMS for 45 min (top) or at the indicated time points (bottom). Rad53-3HA was visualized by Western blotting using an anti-HA antibody (top). An anti-actin antibody was used as a loading control (bottom). All strains are deleted for sml1. H3K56 mutations and deletion of RAD9 have a synthetic defect in MMS-induced Rad53 phosphorylation.
and HST4 was dramatically increased to 43 and 60%, respectively (Fig. 4A). Furthermore, many hst3 hst4 cells contained more than one Ddc2-GFP focus (Fig. 4B). This result demonstrates that loss of HST3 and HST4 creates spontaneous DNA damage during S phase, which is another feature shared with other S phase-specific DNA damage response proteins. Since our data strongly suggested a role of HST3 HST4 in the intra-S phase checkpoint, we decided to determine the relationship between HST3 HST4 and the S phase-specific DNA damage response genes TOF1, MRC1, and CSM3 through epistasis analysis. We found that loss of function of HST3 and HST4 creates synthetic HU sensitivity with the loss of function of CSM3 and synthetic lethality with the loss of TOF1 (Fig. 4C). The triple hst3 hst4 mrc1 mutant also exhibited slightly increased HU sensitivity relative to hst3 hst4 double mutant. Therefore, it appears that HST3 and HST4 belong to a separate epistasis group within the intra-S phase checkpoint pathway. This finding, together with the observation of synthetic phenotypes of MRC1 and TOF1 or CSM3 (30), indicates that intra-S phase checkpoint genes do not form a linear pathway.
HST3-and HST4-mediated Regulation of H3K56 Is Required for Efficient SCC-Because several of the genes that control the S phase-specific branch of the DNA damage network also exhibit defects in SCC (15), we sought to determine whether HST3 and HST4 were also required for SCC. To monitor SCC, we used strains expressing bacterial DNA-binding proteins (Lac repressor and TET repressor) fused to green fluorescent protein (GFP) and containing their respective DNA binding repeats (lactose and tetracycline operators) integrated at several different chromosomal locations. We first introduced HST3 and HST4 deletions, both individually and in combination, in a strain that has Lac operator repeats integrated at a site near the centromere of chromosome III (SBY 885). Cells were grown to midlog phase and then arrested in G 1 using ␣ factor or G 2 /M using nocodazole. As expected, two separate GFP signals were observed in only 5% of wild type cells arrested in G 2 /M, indicating that 95% of the sisters remain paired prior to anaphase (Fig. 5A).
As in wild type cells, in mutants lacking either HST3 or HST4, two GFP signals were observed in less than 6% of cells. In contrast, strains lacking both HST3 and HST4, arrested in G 2 /M, had 18% of cells with two GFP signals. The two GFP signals in hst3 hst4 double mutants were observed only in G 2 /M arrest and not in G 1 -arrested cells, indicating a defect in SCC and not a chromosome duplication. In order to determine whether deacetylase activity of Hst3 and Hst4 is required for faithful SCC, we evaluated SCC in wild type cells treated with 50 mM nicotinamide. We have shown earlier that nicotinamide inhibits NAD-dependent deacetylase activity of Hst3 in vitro (Fig.  1A) and leads to hyperacetylation of H3K56 in vivo (supplemental Fig. 7C). Nicotinamide treatment of wild type cells phenocopied the loss of SCC observed in hst3 hst4 double mutants (Fig. 5A), indicating that deacetylase activity of Hst3 and Hst4 is required for their function in SCC. Loss of SCC was also  (14141), hst3 hst4 (14299), csm3 (14460), hst3 hst4 csm3 (14814), mrc1 (14818), and hst3 hst4 mrc1 (14819) cells were assayed for HU sensitivity. 5-Fold serial dilutions were plated on YEPD with and without 40 mM HU. Plates were incubated for 3 days at 30°C. Right, tetrads were derived from a TOF1/tof1 HST3/hst3 hst4/hst4 diploid strain. The tetrads were dissected (spores from the same tetrad are arranged horizontally) and allowed to grow for 3 days at 23°C. No viable tof1 hst3 hst4 colonies were observed. DECEMBER 28, 2007 • VOLUME 282 • NUMBER 52 observed at loci on the arm of chromosome V and near a telomere on chromosome IV (Fig. 5B). The SCC defect of HST3 HST4 deletion mutants was phenocopied by histone K56Q and K56R mutants (Fig. 5C), thus demonstrating that Hst3 affects SCC through control of H3K56 acetylation.

Mec1-dependent Proteolysis of Hst3
Using conditional inactivation of Hst3 function with 50 mM nicotinamide in synchronous cells following the release from ␣ factor arrest, we found that Hst3 function in SCC is no longer required after S phase has been completed and SCC had been established (cohesion maintenance) but is required during S phase (Fig. 5D), consistent with a role of HST3 during SCC establishment.
In order to further characterize the role of HST3 and HST4 in SCC, we evaluated whether loss of HST3 and HST4 causes a synthetic interaction with mutations in MCD1, which encodes a component of the cohesin complex. We used a temperature- Logarithmically growing cells were arrested in either G 1 with ␣ factor or in G 2 /M with nocodozole, and lack of SCC was scored as the presence of pairs of fluorescent green spots. For the nicotinamide study, wild type cells (SBY 885) were grown in 50 mM nicotinamide to log phase before arresting in G 2 /M with nocodazole. B, sister chromatid cohesion was monitored along the arm of chromosome V using a GFP-tagged TET repressor that binds TET operator sequences integrated 35 kb from the centromere of chromosome V (ARM) in a wild type (YPH1477) and an hst3 hst4 double mutant strain (14587). These strains were derived from YPH1477. Sister chromatid cohesion was monitored near the telomere of chromosome IV (telomere) using a GFP-tagged LAC operator that binds LAC repressor sequences integrated near the telomere of chromosome IV in a wild type (SBY180) and an hst3 hst4 strain (14436). These strains were derived from SBY180. C, SCC was analyzed as in Fig. 5A using strains containing the centromeric LAC operator repeats and GFP-tagged LAC repressor and carrying a wild type histone copy (14625) or H3K56Q point mutation (14629) or H3K56R point mutation (14628). Strains were analyzed in G 1 (␣ factor) and G 2 /M (nocodozole). D, conditional inactivation of Hst3 and Hst4 activity with 50 mM nicotinamide (NA) demonstrates that Hst3 and Hst4 are required for establishment of SCC during S phase but are dispensable for SCC maintenance. The SBY 885 strain, containing LAC operator sequences next to the centromere of chromosome III, was arrested in G 1 with ␣ factor and released into medium containing nocodazole without or with nicotinamide. Nicotinamide (50 mM) was added at two different time points: immediately following release into S phase (t ϭ 0 min) or after S phase has been completed (t ϭ 80 min). Samples were collected every 20 min from time 0 to 160 min following release from ␣ factor and analyzed for DNA content and for SCC. When nicotinamide is added at the 80 min time point to cells that have already passed through S phase, untreated and nicotinamide-treated cells had similar frequency of double GFP signal for the duration of the experiment, which shows that Hst3,4 function is not required for maintaining sister chromatid cohesion. However, when cells were released from ␣ factor arrest into medium with nicotinamide at the beginning of S phase (t ϭ 0 min time point), the frequency of cells with double GFP signal in nocodazole arrest was higher than in control cells. This result suggests that inhibition of Hst3,4 function during S phase leads to failure to establish robust sister chromatid cohesion. E, synthetic interaction of a mutation in MCD1 and deletion of HST3 and HST4. Deletion of HST3 and HST4 lowers the maximal permissive temperature of temperature-sensitive mcd1-1 mutant. WT, wild type. sensitive allele of MCD1, mcd1-1, that fails to grow at 33°C due to loss of SCC. We found that hst3 hst4 mcd1-1 triple mutants fail to grow at 30°C, whereas hst3 hst4 double mutants and mcd1-1 single mutants are capable of growth at this temperature (Fig. 5E). This synthetic interaction further confirms the role of HST3 and HST4 in SCC.

DISCUSSION
During S phase, newly synthesized histone H3 is acetylated at Lys 56 and deposited into nucleosomes during DNA replication and becomes largely deacetylated by the G 2 phase of the cell cycle (2). We show that Hst3 deacetylates a Lys 56 -containing histone H3 peptide in an NAD-dependent fashion in vitro and provide strong support for the hypothesis that Hst3 and Hst4 are critical enzymes for deacetylating H3K56 in vivo. Consistent with previous reports (21,22), we show, first, that deletion of HST3 and HST4 leads to an increase in global H3K56 acetylation. Second, cell cycle-dependent oscillations in Ac-K56H3 are abolished in cells lacking HST3 and HST4. Third, point mutations that disrupt NAD-dependent deacetylase activities also lead to increased H3K56 acetylation and recreate the phenotypes of hst3 deletion mutants.
When cells are exposed to DNA damage, the Ac-H3K56 modification persists and has been shown to be retained specifically at sites of DNA damage in a checkpoint-dependent fashion (2). The finding that Hst3 is down-regulated when cells are exposed to DNA-damaging agents and that this down-regulation requires the DNA damage sensor Mec1 provides a mechanism for the persistence of Ac-H3K56 during genotoxic stress at the sites of DNA damage. Previous reports show that Hst3 protein levels appear to be down-regulated during genotoxic stress through a transcriptional mechanism (22). Our results demonstrate that during genotoxic stress Hst3 levels are also controlled through Mec1-dependent phosphorylation and sub-sequent ubiquitylation and degradation by the proteasome. Further studies are required to determine whether phosphorylation of Hst3 during genotoxic stress is carried out by Mec1 directly, by its downstream kinases, or both. Control of Hst3 protein stability is a particularly appealing mechanism, because, unlike transcriptional regulation, which can control only the global Hst3 level, checkpoint-regulated Hst3 protein stability may direct Hst3 protein degradation specifically to sites of DNA damage. This provides an explanation for the observed accumulation of Ac-H3K56 specifically at sites of DNA damage.
Perturbation of the H3K56 acetylation/deacetylation cycle phenocopies the loss of function mutations in S phase-specific DNA damage response genes (e.g. TOF1, CSM3, or MRC1) (14,19,29), including 1) a DNA damage signal transduction defect that acts in parallel with Rad9 to transmit the signal from Mec1 to Rad53 and a consequent defect in slowing down S phase when encountering DNA damage; 2) accumulation of endogenous DNA damage during S phase; and 3) defects in sister chromatid cohesion. Based on these observations and synthetic interactions with TOF1, MRC1, and CSM3, we propose that HST3,4-mediated control of H3K56 acetylation shares the functions of intra-S phase checkpoint genes in DNA damage signal transduction (14,26), stabilization of stalled replication forks (17), and establishment of sister chromatid cohesion (19). How does increased H3K56 acetylation at the sites of DNA damage contribute to DNA damage resistance? Our finding that Hst3-regulated acetylation of H3K56 plays a role in enabling Mec1 to phosphorylate Rad53 in the absence of RAD9 supports the model that reversible H3K56 acetylation and RAD9 act in parallel pathways for transmitting a DNA damage signal from Mec1 to Rad53 (Fig. 6, left). Considering that Rad9 serves as an adaptor protein that facilitates the recruitment of Rad53 to the sites of DNA damage and promotes Rad53 phosphorylation by Mec1 (12), we hypothesize that the analogous RAD9-independent Rad53 recruitment function may be mediated directly or indirectly by H3K56 acetylation. Ac-H3K56 retained at chromatin containing damaged DNA (Fig. 6, right) could facilitate the recruitment of DNA damage-signaling proteins, such as Rad53, or promote recruitment of other proteins required for stabilization of stalled replication forks. Local Ac-H3K56 deposited onto newly synthesized DNA following fork passage may play a similar role in the recruitment of factors required for the establishment of SCC. In the absence of regulated acetylation, such as in cells lacking HST3 and HST4 or in cells with H3K56Q, a mutation that mimics constitutive acetylation, Ac-H3K56 is distributed throughout chromatin, and cells therefore lose the ability to use this modification to direct FIGURE 6. Left, Hst3 and H3K56 act in parallel with Rad9 during DNA replication checkpoint. Right, model for targeted proteolysis of Hst3 and local Ac-H3K56 control in the S phase-specific DNA damage response and SCC. Ac-H3K56, targeted to the sites of DNA damage through regulated Hst3 proteolysis, serves to recruit DNA repair and cohesion factors. Global loss of Hst3 activity (e.g. HST3 deletion) leads to ubiquitous H3K56 acetylation, whereas excessive Hst3 activity (e.g. Hst3 overexpression) leads to global loss of Ac-H3K56. Both of these conditions disrupt specific targeting of Ac-H3K56 to the sites of DNA damage, and consequently, the efficient recruitment of DNA repair factors is abrogated. DECEMBER 28, 2007 • VOLUME 282 • NUMBER 52 DNA damage response proteins to sites of DNA damage. This model is consistent with the observation that unacetylatable histone H3 (H3K56R), complete loss of H3K56 acetylation (RTT109 deletion), and the histone H3 mutant that mimics constitutive acetylation (K56Q) all elicit phenotypes similar to that of loss of HST3 and HST4 and also with the observation that overexpression of HST3 results in a degree of DNA damage sensitivity similar to that of hst3 hst4 double mutants.

Mec1-dependent Proteolysis of Hst3
Since multiple S phase DNA damage checkpoint proteins have been shown to move along with the replication fork (16,29), it is possible that Hst3 also moves with the replication complex, where it controls the acetylation state of H3K56 and serves a role in S phase checkpoint signaling, stabilization of stalled replication forks, and establishment of SCC.