Bypassing the Requirement for an Essential MYST Acetyltransferase

Abstract

Histone acetylation is a key regulatory feature for chromatin that is established by opposing enzymatic activities of lysine acetyltransferases (KATs/HATs) and deacetylases (KDACs/HDACs). Esa1, like its human homolog Tip60, is an essential MYST family enzyme that acetylates histones H4 and H2A and other nonhistone substrates. Here we report that the essential requirement for ESA1 in Saccharomyces cerevisiae can be bypassed upon loss of Sds3, a noncatalytic subunit of the Rpd3L deacetylase complex. By studying the esa1∆ sds3∆ strain, we conclude that the essential function of Esa1 is in promoting the cellular balance of acetylation. We demonstrate this by fine-tuning acetylation through modulation of HDACs and the histone tails themselves. Functional interactions between Esa1 and HDACs of class I, class II, and the Sirtuin family define specific roles of these opposing activities in cellular viability, fitness, and response to stress. The fact that both increased and decreased expression of the ESA1 homolog TIP60 has cancer associations in humans underscores just how important the balance of its activity is likely to be for human well-being.

THE genetic information of DNA is packed into chromatin, which is predominantly composed of the H3, H4, H2A, and H2B core histone proteins that together with DNA form nucleosomes, the basic subunits of the genome (Kornberg and Lorch 1999). Chromatin structure regulates many cellular processes including gene expression, DNA replication, DNA damage repair, and recombination (Felsenfeld and Groudine 2003). Nucleosomes themselves are tightly regulated by mobilization and positioning that are mediated by ATP-dependent chromatin-remodeling machines (Rando and Winston 2012) and by multiple types of histone post-translational modifications that include acetylation and many other marks (Campos and Reinberg 2009).

Nucleosome acetylation is a highly dynamic modification that is promoted by HATs and removed by HDACs. HATs are classified by sequence into different families (Allis et al. 2007). ESA1/KAT5 belongs to the widely conserved MYST HAT family, named for its founding members (MOZ-YBF2/SAS3-SAS2-TIP60) (Lafon et al. 2007). Esa1 is an essential HAT in yeast (Smith et al. 1998; Clarke et al. 1999) and the catalytic subunit of two distinct multi-protein complexes: NuA4 and Piccolo (Boudreault et al. 2003). Notably, the human homolog of Esa1 is Tip60, which is also essential in vertebrates and has been linked to multiple human diseases (Squatrito et al. 2006; Avvakumov and Côté 2007; Lafon et al. 2007), thus increasing the relevance of gaining a deeper understanding of essential HAT functions.

Esa1 primarily acetylates H4 and H2A in vivo (Clarke et al. 1999; Lin et al. 2008) and regulates the expression of active protein-encoding genes (Reid et al. 2000; Lin et al. 2008). It plays a crucial role in cell cycle progression and ribosomal DNA (rDNA) silencing (Clarke et al. 1999, 2006) and is recruited to DNA double-strand breaks (DSBs) to promote damage repair by acetylating H4, an important step in the repair pathway (Bird et al. 2002; Tamburini and Tyler 2005). Esa1 also regulates replicative life span and autophagy by acetylating the nonhistone targets Sip2 (regulatory subunit of the Snf1 complex, the yeast AMP-activated protein kinase, or AMPK) (Lu et al. 2011) and Atg3 (autophagy signaling component) (Yi et al. 2012). Due to Esa1’s essential nature, to date, hypomorphic and conditional ESA1 alleles have been critical tools in defining its functions (Clarke et al. 1999; Decker et al. 2008; Lin et al. 2008; Mitchell et al. 2008, 2011). The esa1-414 allele is a frameshift mutation in codon 414 that leads to a C-terminal sequence change in 10 amino acids and truncation of the last 22 amino acids. Although esa1-414 cells have a growth rate similar to wild type at 30°, they are sensitive to DNA damage (Chang and Pillus 2009). The phenotypes of esa1-414 cells can be exacerbated by growth at elevated temperatures, conditions under which histone H4 acetylation is decreased and cells slow and/or stop growth by blocking cell cycle progression in G2/M (Clarke et al. 1999; Chang and Pillus 2009).

To oppose acetylation, Saccharomyces cerevisiae expresses three class I HDACs (Rpd3, Hos2, and Hos1), two class II HDACs (Hda1 and Hos3) (Yang and Seto 2008), and five class III HDACs or Sirtuins, including Sir2, Hst1, Hst2, Hst3, and Hst4 (Brachmann et al. 1995; Frye 2000). Rpd3 is active in two different complexes, Rpd3S and Rpd3L (Figure 1A), both of which deacetylate histones H3, H4, H2A, and H2B (Shahbazian and Grunstein 2007; Rando and Winston 2012), as well as many nonhistone proteins (Carrozza et al. 2005a,b; Keogh et al. 2005). Both complexes share a core defined by the Rpd3, Sin3, and Ume1 subunits (Carrozza et al. 2005b; Chen et al. 2012). Rpd3L is required for stress response and transcriptional silencing (Zhou et al. 2009; Ruiz-Roig et al. 2010). The Sds3, Dep1, Cti6, Sap30, Rxt2, Rxt3, and Pho23 subunits, in combination with the Rpd3 core, form the Rpd3L complex that is recruited to gene promoters by Ume6 and Ash1 subunits (Carrozza et al. 2005a; Zhou et al. 2009; Ruiz-Roig et al. 2010). The Rpd3S complex is formed by the Rpd3 core and the Eaf3 and Rco1 subunits (Carrozza et al. 2005b). Rpd3S is directed by Eaf3 to methylated histone H3K36, predominantly found at the 3′ end of coding regions, to repress cryptic transcription (Carrozza et al. 2005b). Rpd3 was recently reported in a third complex, Rpd3μ, along with the Snt2 and Ecm5 subunits (Figure 1A) (Shevchenko et al. 2008; McDaniel and Strahl 2013). This complex is enriched at promoter regions and has a role in the response to oxidative stress (Baker et al. 2013), but little else is yet known about its function.

Figure 1.

Disruption of Rpd3L bypassed the essential requirement for ESA1. (A) Functional organization of Rpd3 complexes with subunits shown to relative scale (adapted from Sardiu et al. 2009; McDaniel and Strahl 2013). Loss of RCO1 disrupts Rpd3S, whereas SDS3 or DEP1 deletions disrupt Rpd3L. Deletion of SNT2 disrupted Rpd3μ. (B) The esa1∆ sds3∆ and esa1∆ dep1∆ strains could lose the covering ESA1 plasmid at room temperature (RT), whereas only the SDS3 deletion bypassed esa1Δ at 30°. Serial dilutions of ESA1 (URA3)-covered strains: esa1∆ (LPY12205), esa1∆ rpd3∆ (LPY12207), esa1∆ sin3∆ (18032), esa1∆ sds3∆ (LPY16480), esa1∆ dep1∆ (LPY20385), esa1∆ sap30∆ (LPY20465), esa1∆ pho23∆ (LPY17027), and esa1∆ rco1∆ (LPY17029). Strains were grown on medium without uracil or with 5-FOA (2 mg/ml). An assay including a wild-type control is shown in Figure S1A. (C) Deletion of SNT2 was not a bypass suppressor of esa1∆. Assay as in Figure 1B, including ESA1 covered esa1∆ snt2∆ (LPY20664). (D) The esa1∆ sds3∆ strain was temperature and DNA damage sensitive. Four strains were assayed: wild type (LPY79), esa1∆ + pESA1 (12205), sds3∆ (LPY12959), and esa1∆ sds3∆ (LPY16595). Serial dilutions compared growth at 30°, 37°, and in response to UV light (60 J/m²) and the drugs HU (0.05 M) and CPT (7 μg/ml dissolved in DMSO). (E) Cell cycle profiles showed that the DNA content of the majority of the esa1∆ sds3∆ cells was in the 2C peak, indicative of a severe delay in progression through the G2/M phase of the cell cycle. (F) All tested acetylable lysine residues of the histone H4 N-terminal tail had low global levels of acetylation in the esa1∆ sds3∆ background. Protein extracts were immunoblotted against the isoform or protein-specified. (G) Deletion of SDS3 improved histone H4K8 and K12 acetylation when esa1-414 was inactivated by growth at 37°. See Figure S1, B and C, for quantification of immunoblots.

Genetic, biochemical, and genome-wide studies suggest that Rpd3 and Hda1 are the most relevant opposing activities to Esa1 (Vogelauer et al. 2000; Lin et al. 2008; Chang and Pillus 2009); yet important questions remain. Rpd3 has been directly identified as the deacetylase for some histone and nonhistone targets of Esa1 (Lu et al. 2011; Rando and Winston 2012; Yi et al. 2012), whereas genetic interaction networks point to deletion of HDA1 as the most prominent alleviating hub for NuA4 (Lin et al. 2008). Hda1 also has some overlapping functions with Rpd3 (Bernstein et al. 2000; Robyr et al. 2002). Finally, even though strong hypomorphic alleles of ESA1 have been studied (Clarke et al. 1999; Decker et al. 2008; Lin et al. 2008), residual Esa1 activity could mask the discovery of esa1∆ phenotypes. In this work we define the Rpd3L complex as the relevant opposing activity to Esa1’s essentiality through the identification of a gene deletion that allows ESA1 null cells to live. This bypass suppression validates and extends previous suggestions about the need for a balanced acetylation status for viability (Zhang et al. 1998; Lin et al. 2008). Importantly, however, we also demonstrate that modulation of acetylation states through specific genetic interventions can restore essential functions required for growth, cell cycle progression, and response to DNA damage and stress.

Materials and Methods

Yeast strains and plasmids

Yeast strains, plasmids, and oligonucleotides are listed in Supporting Information, Table S1, Table S2, and Table S3. The esa1-414 allele (Clarke et al. 1999) and the hst1∆2::LEU2 gene disruption (Smith et al. 2000)—for simplicity referred to herein as hst1∆—were previously described. All other mutants are null alleles constructed with standard methods. hst1∆ (Smith et al. 2000) was introduced into the W303 background by amplification of the LEU2 cassette from strain YCB232. The rpd3∆::LEU2 (DY1539), hda1∆::TRP1 (DY4891), and hos2∆::TRP1 (DY4549) parental strains were all generous gifts from D. Stillman. The parental sap30∆::kanmx strain (MAY8) was a generous gift from J. Heitman, and dep1∆::HIS3 (YJW678) was kindly provided by J. Workman. Silencing reporter and histone mutant strains were constructed as described (Chang and Pillus 2009), starting with strains developed in Roy and Runge (2000). All double and triple esa1∆ mutant strains were constructed with a covering plasmid (pLP796). All esa1∆ strains were further backcrossed to wild type to assess the tetrad segregation pattern, ruling out possibilities of aneuploidy or additional unlinked suppressors.

Growth assays

Dilution assays were performed as described (Chang and Pillus 2009) with fivefold serial dilutions from A₆₀₀ units of 5 ODs for 5-FOA counterselection assays and of 0.5 ODs for all other media. Camptothecin (CPT) sensitivity was assayed using 7 μg/ml in DMSO, added to YPAD (YPD plus adenine) plates buffered with 100 mM potassium phosphate to pH 7.5 (Nitiss and Wang 1988). The esa1∆ sds3∆ strains were extremely sensitive to standard concentrations of hydroxyurea (HU) (0.1–0.2 M) and CPT (20–30 μg/ml). They were also sensitive to the CPT solvent, DMSO (0.4%). For this reason, their growth was tested in less concentrated drug and solvent control plates (0.14%). The nature of the DMSO sensitivity has not been established (Zhang et al. 2013). UV damage was evaluated as described (Hampsey 1997). Images were captured after 2–5 days of growth.

Flow cytometry

Flow cytometry was done as described in Chang et al. (2012). Cells were fixed after growth at 30° on YPAD. A total of 30,000 propidium iodide-stained cells of each strain were analyzed with a FACSCalibur (BD).

Immunoblots

Whole-cell extracts were prepared by bead beating (Clarke et al. 1999). Nuclear fractions were prepared as in Kizer et al. (2006). Proteins were transferred to PVDF membranes (Hybond) and probed with the following: anti-H4K5Ac (1:5000 dilution, Serotec), anti-H4K8Ac (1:2000 dilution, Serotec), anti-H4K12Ac (1:2000 dilution, Active Motif), anti-H4K16Ac (1:2000 dilution, Upstate), anti-H3CT (1:10,000 dilution, Millipore), anti-H3K9,K14Ac (1:10,000 dilution, Upstate), anti-H4 (1:2000 dilution, Active Motif), and anti-β-tubulin (1:20,000) (Bond et al. 1986). All antisera except H4 were diluted in 2.5% milk in TBS–Tween. Anti-H4 was diluted in 5% BSA in TBS–Tween. Experiments were performed independently from two to five times.

Immunoblot quantification

Two to five independent experiments were quantified using the ImageQuant 5.2 program (Molecular Dynamics). The histogram peak function was applied to correct for background.

Plasmid end-joining assay

Assays with pRS316 were performed with established methods (Åström et al. 1999; Lee et al. 1999). Four independent experiments were performed, and P-values were calculated using the Student’s t-test.

Recombination assays

Performed as described (Clarke et al. 2006). Approximately 5000 colonies per isolate were counted in four independent experiments. When strains lost the rDNA::ADE2CAN1 marker inserted in the repetitive rDNA array, they became adenine auxotrophs and developed red pigment sectors. The number of half-red sectored colonies divided by the total number of colonies plated reported the frequency of cells in a population that had undergone a recombination event during the first cell division. The Student’s t-test was applied to assess significance.

DNA silencing assay

Expression of the CAN1 gene integrated in the rDNA cluster was detected as canavanine sensitivity when silencing was compromised (16 mg/ml) (Chang and Pillus 2009).

Results and Discussion

The discovery that ESA1 is an essential gene was significant because prior studies of single mutants of other HATs and HDACs had revealed important roles in gene expression, but not a role in viability. The powerful tool of genetic suppression has facilitated the characterization of conditional, non-null alleles of ESA1 (Biswas et al. 2008; Lin et al. 2008; Chang and Pillus 2009; Scott and Pillus 2010; Chang et al. 2012); however, much remains to be learned about its cellular roles.

Bypassing the requirement for an essential acetyltransferase

Although a condition bypassing the essential function of ESA1 has been alluded to in the literature (Natsume-Kitatani et al. 2011), its specific molecular identity has not been reported. Deletion of RPD3 is to date the best suppressor reported for esa1 alleles (Chang and Pillus 2009), yet rpd3∆ cannot bypass ESA1 since the esa1∆ rpd3∆ mutant is inviable (Chang and Pillus 2009). As Rpd3 is the catalytic subunit of both Rpd3L and S complexes, we asked if loss of components of either complex would eliminate the need for ESA1 while maintaining RPD3 and the activity that it encodes.

Double mutants were constructed, combining esa1∆ with deletions of specific subunits of the Rpd3L and Rpd3S complexes (Figure 1A), with the aid of an ESA1 wild-type gene borne on a URA3 plasmid. The gene deletions tested included SDS3, DEP1, SAP30, and PHO23, all complex-specific subunits of Rpd3L (Lechner et al. 2000); RPD3, the catalytic subunit of both complexes; SIN3, an Rpd3 core subunit; and RCO1, a specific subunit of Rpd3S (Carrozza et al. 2005a,b; Keogh et al. 2005). SDS3 and DEP1 deletions disassemble Rpd3L, whereas Rpd3S integrity is lost when RCO1 is deleted. Neither Pho23 nor Sap30 is essential for the stability of Rpd3L. Strains were tested for growth on 5-FOA, a compound toxic to cells expressing the URA3 gene marking the ESA1 plasmid. Growth of a double mutant indicated that the strain could live without the ESA1 plasmid and thus that the second mutation bypassed esa1∆’s lethality. Of the mutants tested, both SDS3 and DEP1 deletions bypassed the need for the covering ESA1 gene (Figure 1B). Thus, the Rpd3L complex opposes the essential function of ESA1, resulting in viable cells when both ESA1 and SDS3 (or DEP1) are deleted. Notably, bypass suppression was not seen upon deletion of PHO23 or SAP30, which encode more peripheral subunits of Rpd3L. Using the same approach, we also found that loss of the Rpd3μ complex (Figure 1A), through deletion of SNT2 (Baker et al. 2013), did not bypass esa1∆ (Figure 1C). In further characterization of bypass conditions, we focused on the SDS3 deletion as it proved to be a stronger suppressor of esa1∆ than dep1∆ (Figure 1B).

To determine if the SDS3 deletion suppressed phenotypes that had been defined for ESA1 alleles, the esa1∆ sds3∆ strain was evaluated for temperature sensitivity, DNA damage sensitivity, and cell cycle control. In these assays, the esa1∆ sds3∆ cells grew weakly at 30° and were severely temperature sensitive at 37° (Figure 1D). They were also sensitive to DNA damage induced by UV irradiation and by HU and CPT, compounds that induce DNA DSBs through distinct mechanisms (Figure 1D). The esa1∆ sds3∆ cells also proved to be sensitive to the DMSO solvent for CPT, mirroring sensitivity reported for other chromatin regulators (Zhang et al. 2013).

Flow cytometry revealed that whereas the sds3∆ and the esa1∆+pESA1 strains had a cell cycle profile similar to wild type, the esa1∆ sds3∆ strain grown at 30° showed a severe delay in progression through G2/M. The severity of this delay reflects the complete loss of ESA1, combined with the lack of the Rpd3L complex, which is also involved in cell cycle progression (Bernstein et al. 2000; Kaluarachchi et al. 2012) (Figure 1E).

To test the hypothesis that sds3∆ suppressed esa1∆ by restoring isoform-specific acetylation of Esa1 histone targets, whole-cell extracts were analyzed by immunoblotting (Figure 1F). When compared to wild-type strains, low acetylation levels were found for histone H4-modified lysines (Figure 1F). In contrast, acetylation levels of H3, primarily acetylated by other HATs such as Gcn5 and Sas3, remained relatively unaffected (Figure 1F). Notably, several independent esa1∆ sds3∆ strains constructed in the BY and W303 backgrounds had similar phenotypes. Thus, although sds3∆ could bypass the requirement for ESA1 for viability, the double mutant nonetheless had exacerbated phenotypes previously established for alleles of ESA1 in cell cycle control and H4 acetylation.

Based on the findings in Figure 1, we hypothesized that deletion of ESA1 caused a lethal global acetylation imbalance, since histone H3 acetylation levels remained high compared to those of H4 (Figure 1F). When SDS3 was deleted in combination with ESA1, this imbalance was partially alleviated because the Rpd3L complex did not further deacetylate the sparsely acetylated H4. This condition generated slightly higher H4 acetylation levels compared to those predicted for esa1∆ that allowed cells to live. As it was unlikely that the global histone acetylation balance in esa1∆ sds3∆ cells was fully restored, cell fitness remained seriously compromised.

Because the SDS3 deletion alone did not increase global H4 acetylation (Figure 1F), we took advantage of a strain expressing the esa1-414 conditional allele to test if loss of Rpd3L affected H4 acetylation in the absence of ESA1. The esa1-414 and esa1-414 sds3Δ strains were grown at 37° to further diminish esa1-414’s catalytic activity. Figure 1G shows that, when the SDS3 gene was deleted, the low levels of histone H4 acetylation were somewhat increased relative to esa1-414. This confirmed that deletion of SDS3 could promote higher acetylation of histone H4 in the absence of ESA1.

Critical histone residues that enhance esa1∆ sds3∆ fitness

To test the hypothesis that imbalanced acetylation interfered with the vitality of the esa1∆ sds3∆ strain, we sought to identify conditions that enhanced fitness. Initially, esa1∆ sds3∆ strains in combination with histone H4 and H3 lysine (K) to arginine (R) or alanine (A) mutations were constructed and analyzed (Figure 2A and Figure S2A). The arginine mutants were used to retain the positive charge of deacetylated lysines, whereas the alanine mutants were used as non-acetylable residues. Among the mutants, the H3K9R was the best suppressor of esa1∆ sds3∆ temperature and damage sensitivities (Figure 2A). We had earlier observed that the H3 acetylation levels remained high in esa1∆ sds3∆ cells (Figure 1F). Thus, it appeared that, if H3K9 cannot be acetylated due to the K-to-R mutation, the global acetylation imbalance can be eased, leading to improved growth under stress conditions. To pursue this idea, an esa1∆ sds3∆ H3K9Q mutant strain was tested as a proxy for cells with highly acetylated histone H3 in a low-H4-acetylation context. Indeed, this strain proved to be even sicker than cells with wild-type H3 and H4 (Figure 2B).

Figure 2.

Mutation of histone residues enhanced esa1∆ sds3∆ fitness. (A) The H3K9R mutant suppressed esa1∆ sds3∆ temperature and DNA damage sensitivity. An esa1∆ sds3∆ strain was constructed in which all copies of genes encoding H3 and H4 were deleted (hht-hhf∆) and covered with pJH33 (Ahn et al. 2005). Strains shown carried a plasmid either with wild-type H3 and H4 or with one mutated lysine in H4 or H3 as indicated. Lysines were mutated to arginine (R) as an unacetylated lysine mimic, to glutamine (Q) to mimic acetylation, or to alanine (A) as a non-acetylable residue. Plasmid retention was required for survival. Wild-type histones H3-H4 (LPY17368), H4K5R (LPY20430), H4K8R (LPY20431), H4K12R (LPY20432), H3K9R (LPY20388), H3K14R (LPY20389), H3K9A (LPY20387), and H3K14A (LPY20437) in the esa1∆ sds3∆ hht1-hhf1∆ hht2-hhf2∆ background were tested in serial dilutions. Hydroxyurea plates were grown at 30°. Strains were very sensitive to DMSO, the solvent for CPT, and thus this drug challenge is not shown (see text). (B) The H3K9Q mutant that mimics highly acetylated H3 in the esa1∆ sds3∆ background was sick. Serial dilution assays as above compared with the H3K9Q (LPY20661) mutant. (C) SET1 was required for survival of esa1Δ and esa1Δ sds3Δ in the absence of pESA1. Assay was as in Figure 1B testing esa1Δ sds3Δ set1Δ (LPY20775) and esa1Δ set1Δ (LPY20776) strains. (D) Lysine 4 of H3 was critical in the esa1Δ sds3Δ. The esa1Δ sds3Δ H3K4A mutant (LPY20903) was tested. (E) H4 acetylation mimics in esa1∆ sds3∆ strains were healthier than those expressing wild-type histones. Tested strains were H4K5Q, K8Q, K12Q (LPY20436) and H4K8Q, K12Q (LPY20435) in the esa1∆ sds3∆ hht1-hhf1∆ hht2-hhf2∆ background. Control assays demonstrating no effects on wild-type strains with the histone mutants are shown in Figure S2B.

In contrast to the H3K9 mutants, the H3K14R and A mutants remained sick (Figure 2A), a result that initially appeared at odds with the central importance of balanced histone acetylation. However, H3K14, but not K9, has a more widespread role in transcriptional regulation by interacting with the Jhd2 histone demethylase (Maltby et al. 2012). The H3K14A/R mutants result in loss of histone H3K4 trimethylation (Nakanishi et al. 2008), a modification that has a direct impact on chromatin structure through the recruitment of modifiers containing methyl-binding domains. Confirming the critical role of H3K4me in the esa1Δ sds3Δ strain, SET1, the gene encoding the H3K4 methyltransferase (Krogan et al. 2002), could not be deleted (Figure 2C) and the H3K4A mutant grew more slowly than the H3K14A mutant in esa1Δ sds3Δ cells (Figure 2D).

The H4K5R mutant partially suppressed the temperature and DNA damage sensitivity of the esa1∆ sds3∆ strain. This result may be considered relative to distinct roles of the Rpd3L and S complexes. Rpd3L is targeted to specific promoter regions (Carrozza et al. 2005a), whereas Rpd3S activity is more localized to coding regions by the H3K36-methyl mark from Set2 (Carrozza et al. 2005b; Drouin et al. 2010; Govind et al. 2010). The esa1∆ sds3∆ strains are expected to have lower acetylation at gene-coding regions, compared to promoter regions, since Rpd3S remains active in these cells. The H4K5R mutant may thus create a homogeneous non-acetylated H4K5 state that is less deleterious. Acetylation of H4K5, -8, and -12 is considered to be somewhat redundant (Dion et al. 2005); nonetheless, the major target of Esa1 acetylation is K5 of H4 (Smith et al. 1998; Clarke et al. 1999) (Figure 1F). Indeed, the H4K5R mutant was a good suppressor of esa1∆ sds3∆ whereas the H4K8R and K12R mutants remained sick. Finally, when H4K8 and K12 were simultaneously mutated to glutamine to mimic higher H4 acetylation levels, the esa1∆ sds3∆ temperature and damage phenotypes were rescued (Figure 2E).

HDAC deletions can improve esa1∆ sds3∆ fitness and acetylation

As a counterpoint to the site mutant studies, we asked if individual HDAC deletions could improve fitness of the esa1∆ sds3∆ strain. In this case, triple mutants were constructed with select members of each of the three HDAC classes. As before, the mutants were constructed using the covering ESA1 plasmid, followed by the 5-FOA challenge for plasmid loss (Figure 3A). Furthermore, to test if the Rpd3S complex could be lost after Rpd3L was eliminated, the integral Rpd3S-specific subunit RCO1 (Carrozza et al. 2005b) was also deleted (Figure 3A). Additional deletion of RPD3 or RCO1 led to inviability, consistent with the observation that the Rpd3S complex activity is necessary to bypass esa1∆. Deletion of HOS2 also interfered with bypass suppression, suggesting a significant role for HOS2 or the Set3 complex (Pijnappel et al. 2001) in esa1∆ sds3∆. In contrast, three of the HDAC triple mutants grew without the ESA1 gene. These were esa1∆ sds3∆ strains with deletions of genes encoding the type II HDAC Hda1 and the Sirtuins Sir2 and Hst1.

Figure 3.

Distinct functional interactions of esa1Δ sds3Δ with HDAC deletions. (A) Mutant strains expressing wild-type ESA1 from a URA3 vector were challenged on 5-FOA plates using serial dilutions to test for loss of the URA3 plasmid at RT and 30°. The esa1∆ sds3∆ strains mutated for HDA1, SIR2, or HST1 survived. In contrast, the bypass suppressor strain lost viability upon deletion of RPD3, RCO1, or HOS2, thereby demonstrating specificity for balancing of HDAC activities. Note that the added suppression of esa1∆ sds3∆ by additional HDAC deletions was strictly dependent on loss of SDS3 as the corresponding esa1∆ HDAC∆ mutants were inviable (Figure S3A). The surviving triple mutants could also cure additional esa1∆ sds3∆ phenotypes (see Figure 4 and Figure 5). Strains tested were esa1∆ (LPY12205), esa1∆ sds3∆ (LPY16480), esa1∆ sds3∆ hda1∆ (LPY17759), esa1∆ sds3∆ rpd3∆ (LPY 17714), esa1∆ sds3∆ rco1∆ (LPY17805), esa1∆ sds3∆ hos2∆ (LPY17848), esa1∆ sds3∆ sir2∆ (LPY17966), and esa1∆ sds3∆ hst1∆ (LPY17801). (B) Different HDACs improved acetylation at specific histone H4 residues. Extracts from strains in A without the ESA1 plasmid were probed for lysine-specific modifications. Additional controls included wild-type and esa1-414 strains. The esa1-414 strain was also grown at 34° to partially inactivate esa1-414. Strains tested were wild type (LPY79), esa1-414 (LPY4776), esa1∆ sds3∆ (LPY16595), esa1∆ sds3∆ hda1∆ (LPY17748), esa1∆ sds3∆ sir2∆ (LPY17997), and esa1∆ sds3∆ hst1∆ (LPY17900). Quantification of blots is shown in Figure S3, B and C. Histone H3 levels were modestly decreased in all bypass strains, whereas H4 levels remained relatively unaffected. The nature of the change of histone levels is not established; however, precedents for specific changes in histone levels have been reported under a variety of physiological or signaling conditions (Dang et al. 2009; Feser et al. 2010; Turner et al. 2010). In particular, loss of the Asf1 chaperone and other conditions leading to a G2/M block have been associated with decreased histone H3 levels (Feser et al. 2010).

When comparing global histone acetylation of esa1∆ sds3∆ and esa1-414 cells grown at 34° to partially inactivate Esa1, decreased H4 acetylation was observed in esa1∆ sds3∆ (Figure 3B). Acetylation increased at all H4 residues tested in the esa1∆ sds3∆ hst1∆ strain (Figure 3B). Compared to esa1∆ sds3∆, the esa1∆ sds3∆ hda1∆ strain also showed higher acetylation of H4 except at H4K16. The esa1∆ sds3∆ sir2∆ strain had higher levels of H4K16 acetylation relative to esa1∆ sds3∆; however, acetylation of H4K8 and K12 remained low (Figure 3B) with similar results obtained for both whole-cell lysates and isolated nuclei (not shown).

Specific contributions of individual HDACs in diverse cellular functions

Because the patterns of acetylation and overall growth differed among the triple-mutant combinations, we tested the idea that they might have distinct responses to specific cellular challenges. Notably, the HDA1 deletion improved high temperature growth at 34° and 37° (Figure 4A) and suppressed the cell cycle G2/M block of esa1∆ sds3∆ (Figure 4B). Loss of HDA1 also suppressed defective rDNA silencing of esa1∆ sds3∆ (Figure 4C).

Figure 4.

Deletion of HDA1 specifically suppressed the temperature sensitivity and the cell cycle progression and acetylation defects of esa1∆ sds3∆. (A) HDA1 and HST1 deletions suppressed the growth defect of esa1∆ sds3∆ at 30° and 34°. HDA1 and, to a lesser extent, SIR2, suppressed the temperature sensitivity of esa1Δ sds3Δ at 37°. Serial dilutions of strains studied in Figure 3B were plated on YPAD and grown at 30°, 34°, and 37°. Growth of single-mutant controls is shown in Figure S3D. (B) HDA1 deletion partially suppressed the G2/M block of esa1∆ sds3∆. Cell cycle profiles were analyzed by flow cytometry. (C) The HDA1 deletion suppressed the rDNA silencing defect of esa1∆ sds3∆. The location of the reporter gene within the rDNA repeat units is shown. The wild-type (LPY4767), esa1-414 (LPY4911), esa1∆ sds3∆ (LPY17959), esa1∆ sds3∆ hda1∆ (LPY18222), esa1∆ sds3∆ sir2∆ (LPY18573), and esa1∆ sds3∆ hst1∆ (LPY18210) strains contained the rDNA::ADE2 CAN1 reporter. Defective rDNA silencing led to expression of the CAN1 gene and sensitivity to canavanine. (D) HDA1 and HST1 deletions also suppressed H4 acetylation defects of esa1Δ sds3Δ when the strains were grown at 35°. Strains tested are listed in Figure 3B. (E) Specific double HDAC deletion strains were sensitive to high temperature and DNA damage. Growth conditions included 30°, 37°, UV 60 J/m², HU 0.1M, CPT 20μg/ml and DMSO 0.4%. Strains tested were wild type (LPY79), sds3∆ hos2∆ (LPY17724), sds3∆ rco1∆ (18226), sds3∆ rpd3∆ (LPY18595), sds3∆ hda1∆ (LPY17753), sds3∆ sir2∆ (LPY17969), and sds3∆ hst1∆ (LPY18112). The sds3∆ hda1∆ and sds3Δ hst1Δ strains were sensitive to temperature and DNA damage. Single-mutant controls are shown in Figure S3D. Deletion of ESA1 suppressed sds3∆ hda1∆ strain’s temperature sensitivity (A).

The immunoblot results in Figure 3B show that Hda1 opposed Esa1 function by deacetylating H4, whereas H3 acetylation remained high, as Hda1 is largely an H3 deacetylase. The same results were found at elevated temperatures (Figure 4D). Hda1 might oppose ESA1 function at overlapping Rpd3L targets that include cell cycle genes by deacetylating histones H3 and H4 (Bernstein et al. 2000). It is also possible that Hda1 deacetylates Esa1’s nonhistone targets, thus improving overall acetylation levels of these substrates. The chromatin landscape of esa1∆ sds3∆ hda1∆ cells includes higher acetylation of histone H4K5, K8, and K12 compared to esa1∆ sds3∆, likely resulting in a more balanced H3/H4 acetylation ratio at target genes important for progression through the cell cycle and silencing. This idea could explain Hda1’s prominent role in NuA4 synthetic interaction networks (Lin et al. 2008). Lysines 5, 8, and 12 of histone H4 are targets of the same histone-modifying complexes. Therefore, greater acetylation at these residues in the esa1∆ sds3∆ hda1∆ strain are consistent with their proposed overlapping roles in chromatin (Dion et al. 2005). H4K16 acetylation had a different and independent role (below).

Deletion of HST1 promoted restoration of H4 acetylation in esa1∆ sds3∆ (Figure 3B and Figure 4D). As HST1 is responsible for sensing and regulating cellular NAD+ levels (Bedalov et al. 2003), the esa1∆ sds3∆ hst1∆ strain may have deregulated NAD+ that negatively affects other Sirtuin HDAC functions. Suppression of the acetylation defect correlated with improved growth (Figure 4A), but not with suppression of cell cycle defects (Figure 4B), sensitivity to DNA damage, or 37° incubation. This suggests that increased H4 acetylation alone is not adequate to promote robust cellular function; perhaps H4 acetylation must instead be localized to specific chromatin regions, or increased acetylation of nonhistone substrates is required to suppress specific esa1∆ sds3∆ phenotypes.

The complex biology of balancing acetylation is seen further in Figure 4E. In the sds3∆ hda1∆ double mutant, where Rpd3L and Hda1 HDAC complexes were further disrupted, there was sensitivity to high temperature (Figure 4E). Indeed, the growth of esa1∆ was improved in triple mutants with the sds3∆ hda1∆ and sds3Δ hst1Δ deletions (Figure 4A), the double mutants of which were by themselves the sickest tested (Figure 4E), underscoring the complex nature of functional interactions between these acetylase and deacetylase activities.

Alleviating esa1∆ sds3∆ DNA damage sensitivity by deleting SIR2

When the triple-mutant strains were tested for DNA damage sensitivity, all HDAC deletions improved growth after UV irradiation (Figure 5A). Similarly, the triple mutants suppressed the extreme DMSO sensitivity of the esa1∆ sds3∆ strain (Figure 5A). This suppression suggested that the reduced growth in DMSO correlated with a severely affected chromatin state, as deletion of single chromatin regulators has been previously associated with DMSO sensitivity (Zhang et al. 2013).

Figure 5.

HDAC-specific effects on DNA damage sensitivity and repair. (A) SIR2 deletion broadly suppressed the DNA damage sensitivity of esa1∆ sds3∆. Loss of HDA1 and HST1 also suppressed UV and DMSO sensitivity in esa1∆ sds3∆. Serial dilutions of strains in Figure 4A were plated on YPAD and DNA damage plates. (B) The esa1∆ sds3∆ sir2∆ mutant down-regulated the NHEJ repair pathway. A plasmid end-joining assay was performed with strains analyzed in A and with sir2∆ (LPY18223) as an additional control. Data are expressed as NHEJ efficiency relative to wild type, after normalization to the corresponding transformation efficiency of the uncut plasmid. The repair efficiencies of esa1-414 and esa1∆ sds3∆ were identical, but they differed significantly from wild type and esa1∆ sds3∆ hda1∆ (p = 0.02), demonstrating that deletion of HDA1 enhanced NHEJ levels of esa1∆ sds3∆. The re-ligation efficiency of esa1∆ sds3∆ sir2∆ was statistically distinct from esa1∆ sds3∆ (p < 0.005), whereas the NHEJ efficiency of esa1∆ sds3∆ hst1∆ was higher than in esa1∆ sds3∆ (p = 0.01), demonstrating that HST1 deletion led to a further up-regulation of the NHEJ pathway. A previous report found that two different esa1 hypomorphic alleles had defective NHEJ (Bird et al. 2002). Results demonstrating enhanced NHEJ here could be due to allele-specific differences as seen in earlier studies comparing the severity of phenotypic defects and strength of suppression between different esa1 alleles (Decker et al. 2008; Chang et al. 2012). (C) The esa1∆ sds3∆ sir2∆ strain had high levels of mitotic recombination in the rDNA array. Recombination was evaluated using a marker-loss assay. The esa1-414 and esa1∆ sds3∆ strains had higher recombination rates than wild type (p < 0.05); however, esa1-414 was higher than esa1∆ sds3∆ (p = 0.024). Previous esa1-414 results were reproduced (Clarke et al. 2006). The esa1∆ sds3∆ hda1∆ strain was not distinct from wild type, demonstrating that HDA1 deletion suppressed the high recombination rates of esa1∆ sds3∆ (p = 0.03). The esa1∆ sds3∆ sir2∆ strain was statistically different from esa1∆ sds3∆ (p = 0.0009), whereas esa1∆ sds3∆ hst1∆ was not. In this case, loss of Rpd3L could not oppose the induced higher rDNA recombination rates in sir2∆ deletion, perhaps related to the concomitant ESA1 loss (Zhou et al. 2009) and/or to an Rpd3S-specific function at the rDNA. (D) A pseudodiploid state was not sufficient for DNA damage resistance in esa1∆ sds3∆. Wild-type (LPY79) and esa1∆ sds3∆ (LPY16595) MATα strains were transformed with TRP1 plasmids encoding the MATa or MATα locus and the vector alone as a control. The strains were compared to esa1∆ sds3∆ sir2∆::TRP1 (LPY17997) under DNA-damage-inducing conditions on trp− plates. Wild-type strains were unaffected by simultaneous expression of both MAT genes. The pseudodiploid esa1∆ sds3∆ strain remained sensitive to DNA damage when compared with the esa1∆ sds3∆ sir2∆ strain, where a pseudodiploid state was combined with global high levels of histone H4K16 acetylation. The slightly exacerbated HU sensitivity of esa1∆ sds3∆ upon MATα overexpression may relate to the previously reported synthetic lethality of rpd3∆ strains when MATα2 was overexpressed (Kaluarachchi et al. 2012).

In contrast to uniform improvements in the triple mutants after UV, only the SIR2 deletion suppressed sensitivity to DSBs induced by CPT and HU (Figure 5A). SIR2 had been previously linked to the DNA damage response (Tamburini and Tyler 2005; McCord et al. 2009), but we further considered that this suppression might also be coupled to the pseudodiploid state in SIR2 null cells caused by de-repression of the cryptic mating-type loci (Haber 2012).

DSBs can be repaired by either of two pathways: nonhomologous end joining (NHEJ) or homologous recombination (HR) (Chapman et al. 2012). NHEJ is error-prone as it predominantly ligates broken DNA ends (Betermier et al. 2014), whereas the HR pathway is more accurate as it uses an intact copy of the affected region as a template for repair. The a1-α2 repressor expressed in diploid and SIR2 null strains represses specific genes involved in the NHEJ pathway (Åström et al. 1999; Valencia et al. 2001). This is believed to increase HR efficiency, as diploid cells contain a second copy of each chromosome that can be used as a template for repair (Haber 2012). As with SIR2 null cells, the esa1∆ sds3∆ sir2∆ strain was non-mating, confirming that the cryptic mating loci were not silenced in haploid cells.

To investigate NHEJ repair efficiency, a plasmid re-ligation assay was used. The assay consists of transforming a linearized plasmid and counting the transformants that repaired the linear plasmid to convert it into a circular and thereby stably maintained plasmid (Åström et al. 1999; Lee et al. 1999). We found that esa1-414 and esa1∆ sds3∆ strains had increased NHEJ repair efficiency compared to wild type, one possible reason for their DNA damage sensitivity phenotypes (Figure 5B). Because in endpoint drug challenge assays the esa1Δ sds3Δ strain was orders of magnitude more sensitive to DNA damage than the esa1-414 strain (Figure 5A), we suspected that a severely altered chromatin structure in the bypass strain could influence other aspects of the repair process, such as the transcriptional response or the direct repair of the breaks. In contrast, the esa1∆ sds3∆ sir2∆ strain was defective in NHEJ repair to the same extent as sir2∆ (Figure 5B). This suggested that the esa1∆ sds3∆ sir2∆ strain had improved growth on CPT and HU because it was essentially repairing by HR, thereby avoiding mutations that could be introduced during the end-joining repair pathway.

SIN3 and RPD3 null strains are sensitive to DNA damage induced by phleomycin although they are not sensitive to HU, which, like CPT, is an S-phase-specific DSB inducer (Jazayeri et al. 2004). Such differential sensitivity is consistent with a role for Rpd3-Sin3 in NHEJ repair. By contrast, sds3∆ strains are sensitive to S-phase DSBs (Chang and Pillus 2009), suggesting that Rpd3L has a more prominent role in HR compared to Rpd3 as part of the core complex. This is consistent with our observations. The esa1∆ sds3∆ strain was very sensitive to CPT and had a higher rate of NHEJ compared to wild type. When the SIR2 gene was deleted in the esa1∆ sds3∆ strain, the NHEJ bias was shifted toward HR repair, making these cells resistant to S-phase-induced damage.

In an independent approach we analyzed recombination using reporter strains in which the ADE2 gene was inserted in the rDNA locus. Expression of ADE2 produced white colonies; however, chromosomal loss of ADE2 due to recombination between the rDNA repeats generated sectored red colonies. In quantifying the relative number of half-red-sectored colonies, we found that the esa1∆ sds3∆ sir2∆ strain showed high levels of rDNA recombination (Figure 5C). It is important to note that sir2∆ strains have similar high rDNA recombination levels (Gottlieb and Esposito 1989; Clarke et al. 2006); however, the high recombination rate is restricted to specific genomic areas (Su et al. 2000; Mieczkowski et al. 2007). In the esa1∆ sds3∆ sir2∆ strain, the genomic areas susceptible to recombination may be more dispersed than in sir2∆ cells as Rpd3 has a role in repressing mitotic recombination (Dora et al. 1999; Merker et al. 2008).

Finally, to test if the DNA damage resistance of the esa1∆ sds3∆ sir2∆ strain was due to a pseudodiploid state, we transformed a MATα esa1∆ sds3∆ strain with a plasmid-borne MATa locus. Enhanced suppression of the DNA-damage-sensitive phenotype was not observed (Figure 5D). This suggested that the DNA-damage-resistant phenotype of esa1∆ sds3∆ sir2∆ cells was not due simply to a pseudodiploid state but was more likely to result from enhanced levels of acetylation of Sir2 substrates, including H4K16 (Figure 3B), which is a mark of open chromatin that is involved in repair (Bird et al. 2002; Tamburini and Tyler 2005).

Conclusions

The human MYST gene TIP60, like its yeast homolog ESA1, is essential for viability (Gorrini et al. 2007) and contributes to the DNA damage response through mechanisms that are under active investigation (reviewed in Squatrito et al. 2006; Xu and Price 2011). Because of the deeply conserved characteristics of these genes and their other diverse functions, it is important to understand the nature of their essential functions. In this study, we took advantage of the powerful tool of genetic suppression (reviewed in Prelich 1999) to define an imbalance in dynamic acetylation and deacetylation as the factor underlying lethality in esa1∆ mutants. By restoring acetylation through fine-tuning the spectrum of active enzymes in the cell, it was possible to restore viability and successful responses to environmental stressors. It is also likely that keeping acetylation balanced through tight regulation of Tip60 activity is important in human cells, as altered Tip60 levels have been linked to both suppression and promotion of multiple types of cancer (Squatrito et al. 2006; Avvakumov and Côté 2007).

Special relationships between Esa1 and HDACs have been pointed to in earlier studies (Biswas et al. 2008; Lin et al. 2008; Chang and Pillus 2009) along with the discovery of other suppressors (Chang et al. 2012). However, none of these suppressors could bypass the need for ESA1. There has been some discussion about whether Esa1’s catalytic activity is its essential function (Smith et al. 1998; Decker et al. 2008), and there is structural and proteomic evidence that auto-acetylation and acetylation of nonhistone substrates are major contributors in the requirement for Esa1 (Yan et al. 2002; Lin et al. 2009; Lu et al. 2011; Yi et al. 2012). Our observation that the lethality of esa1Δ can be bypassed by the loss of a single HDAC complex supports the idea that ESA1’s essential function is its catalytic activity. Importantly however, it is that activity in the context of balancing the overall acetylation state of the cell that is critical.

Viability of the esa1∆ sds3∆ strain added emphasis to previous data implicating a significant functional relationship between Esa1 and Rpd3 (Biswas et al. 2008; Chang and Pillus 2009; Lu et al. 2011; Yi et al. 2012). The best-studied nonhistone targets of Esa1 are also deacetylated by Rpd3 (Lin et al. 2008; Lu et al. 2011; Yi et al. 2012), suggesting that these substrates may be part of an extreme acetylation imbalance that leads to death in esa1∆ cells. If nonhistone targets of Esa1 have a role in viability, at least some are also likely to be deacetylated by Rpd3L and consequently remain in a relatively balanced acetylation state in esa1∆ sds3∆ strains.

Our data support the idea that the inviability in esa1∆ cells may be caused in part by disproportionately high H3 acetylation compared to H4 acetylation, which is extremely low due to loss of Esa1, the primary nuclear H4 acetyltransferase (Figure 6, A and B). At the molecular level imbalanced acetylation is likely to influence multiple aspects of chromatin, including nucleosomal integrity, chromosomal stability, or enzymatic complex affinity for genomic targets. However, enhanced histone acetylation could not rescue all esa1Δ sds3Δ phenotypes, suggesting that balanced acetylation of both histone and nonhistone substrates of Esa1 is critical for robust growth in the absence of this essential enzyme. Supporting this idea, in proteomic studies >200 nonhistone substrates have been reported for NuA4 (Lin et al. 2009; Mitchell et al. 2013), and of these, at least 77 are encoded by essential genes. Substrates include components of the chromatin-remodeling complex RSC (e.g., STH1 and RSC8), proteins involved in ribosome biogenesis (e.g., NOP4 and DBP6), and cytoskeleton and spindle components (e.g., TUB2 and SPC42), whose acetylation state and role in viability in esa1Δ and esa1Δ sds3Δ remain to be tested.

Figure 6.

Bypassing an essential acetyltransferase by balancing dynamic acetylation. (A and B) ESA1 deletion resulted in a deleterious decrease in acetylation including a high H3:low H4 acetylation ratio. Imbalanced acetylation of nonhistone substrates of Esa1 is also likely to have a role in lethality of esa1Δ strains. (C) When the acetylation ratio was balanced by deletion of SDS3, esa1Δ cells were viable, although still compromised for growth. (D–G) Balancing acetylation of histone and nonhistone substrates to different extents further improved cell fitness and response to specific challenges. A promoter region and a coding region (blue) separated by the transcriptional start point (arrow) are represented for each genetic condition. Histones H2A and H2B are not shown for simplicity; likewise, only one of the tails of H3 and H4 are depicted. Nonhistone substrates are represented as a circle, a triangle, and a pentagon, and their acetylation is shown as red dots.

In esa1∆ sds3∆ cells, because the Rpd3L complex is not present to deacetylate targets shared with Esa1 and other HATs, the need for ESA1 can be bypassed (Figure 6C). However, the esa1∆ sds3∆ cells are very sick, a state that can be further improved by manipulating other chromatin modifiers and specific lysines in histones H3 and H4.

The fitness of the esa1∆ sds3∆ strain improves to differing extents depending on which additional chromatin modifier is affected, and the balancing act can be more or less robust depending on the target genes, chromatin regions regulated by the modifier, and/or histone and nonhistone substrates affected. One example includes a non-acetylable lysine 9 in histone H3 that significantly improved esa1∆ sds3∆ fitness. In these cells, H3 acetylation levels are lower and more balanced relative to low H4 acetylation (Figure 6D). Another example involves deletion of HDA1 that encodes a second HDAC, which increases acetylation of H4 and possibly of shared nonhistones substrates with Esa1 (Figure 6E). The esa1Δ sds3Δ sir2Δ strain is an example where fitness may additionally be influenced by cell-specific transcription. In this case, a pseudodiploid state is not sufficient, but may help promote error-proof repair of damaged DNA (Figure 6F). Finally, even though the esa1Δ sds3Δ hst1Δ strain has an improved H3/H4 acetylation ratio, it is not as fit as the esa1Δ sds3Δ hda1Δ strain, suggesting that in esa1Δ sds3Δ hst1Δ, acetylation of nonhistone substrates of Esa1 is not adequately balanced (Figure 6G). Together, the interplay of opposing acetylation activities can lead to a critical balance that promotes viability even under circumstances where death would normally result.