Expression profiling of S. pombe acetyltransferase mutants identifies redundant pathways of gene regulation

Background Histone acetyltransferase enzymes (HATs) are implicated in regulation of transcription. HATs from different families may overlap in target and substrate specificity. Results We isolated the elp3+ gene encoding the histone acetyltransferase subunit of the Elongator complex in fission yeast and characterized the phenotype of an Δelp3 mutant. We examined genetic interactions between Δelp3 and two other HAT mutants, Δmst2 and Δgcn5 and used whole genome microarray analysis to analyze their effects on gene expression. Conclusions Comparison of phenotypes and expression profiles in single, double and triple mutants indicate that these HAT enzymes have overlapping functions. Consistent with this, overlapping specificity in histone H3 acetylation is observed. However, there is no evidence for overlap with another HAT enzyme, encoded by the essential mst1+ gene.


Background
Chromatin is modulated in part by the pattern of different histone modifications, leading to the speculation that a "histone code" provides epigenetic information that facilitates chromosome activities [1,2]. Not surprisingly, the enzymes that modify chromatin have diverse roles in the cell that affect multiple DNA-dependent events. Acetylation was the first histone modification identified, and is associated with a variety of chromatin functions [2,3]. While some evidence suggests that acetylation changes association of DNA with the underlying nucleosomes [4], it also creates specific binding sites for proteins involved in a variety of DNA transactions [2,3]. Acetylation has long been linked to transcriptional activation, so that the histone acetyltransferase enzymes function generally as transcriptional activators [5].
Histone acetyltransferases are well conserved in eukaryotes and can be separated into multiple gene families based on primary sequence. The MYST family of histone acetyltransferases was named for founding members identified in yeast and humans, MOZ, YBF2/SAS3, SAS2 and TIP60 [6]. Most species contain multiple members of this family. The Kat5 group, the most conserved family members, includes SpMst1 (S. pombe), ScEsa1 (S. cerevisiae) and hTip60 (human), which acetylates residues in the tails of histones H2A and H4 [7]. These proteins are the catalytic HATs for the NuA4 complex in their respective organisms [8,9]. In the yeasts the Kat5 orthologues are the only MYST family HATs that are essential for viability [10,11].
In budding yeast there are two other MYST family members, ScSas2 and ScSas3, which were originally isolated with defects in silencing the mating locus [12,13]. ScSas2 is the catalytic HAT for the budding yeast SAS complex that acetylates H4K16 [14]. It antagonizes the histone deacetylase Sir2 [15]. ScSas3 is the catalytic HAT for the NuA3 complex in budding yeast [16] which specifically acetylates histone H3 [17]. In S. pombe by contrast there is only one additional MYST family member, SpMst2 [11]. Mst2 shows sequence similarity to both ScSas2 and ScSas3. However, its biological function, in which it antagonizes SpSir2 in telomere silencing, suggests that it is a functional homologue of ScSas2 [11].
Another conserved family of HATs is the Gcn5related N-acetyltransferase (GNAT) super-family consisting of Gcn5, Elp3, Hat1 and Hpa2, with Gcn5 (KAT2) and Elp3 (KAT9) as the most studied members [18]. The evolutionarily conserved Gcn5 is the catalytic subunit of several multi-subunit complexes including the SAGA co-activator complex [19]. Gcn5 acetylates multiple lysine residues of histones H3 and H2B and mediates both targeted and non-targeted (global) acetylation [20]. In S. cerevisiae, SAGA is recruited to promoter regions by transcription factors like ScGal4 and ScGcn4. Elongating RNA Pol II is associated with the Elongator complex that contains ScElp3 and has been shown to acetylate H3K14 and H4K8 in budding yeast [21]. Elp3 is highly conserved in structure and function between human and budding yeast [22].
While neither ScGCN5 nor ScELP3 is essential in budding yeast, a double mutation in Δgcn5 Δelp3 double mutant is significantly sicker than either single mutant [23], resulting in hypoacetylation of histone H3 in gene coding regions [24] and spreading of ScSir3 into sub-telomeric DNA [25]. Data from budding yeast suggest that ScGcn5 also overlaps with ScSas3, because a double deletion Δgcn5 Δsas3 is lethal [26]. This suggests that there are shared functions or targets of the NuA3, SAGA, and Elongator complexes in budding yeast, which implies that different HAT families may perform the same modifications. However, because the complement of HATs is somewhat different in S. pombe, it is not clear how general this overlap may be. SpGcn5 has been previously characterized for its effects on gene expression [27,28] and for chromatin binding genome-wide [29]. Surprisingly, relatively little change in gene expression profile is observed in gcn5 mutants despite a very high association with the chromatin. This suggests that SpGcn5 overlaps with other HATs.
Here, we examine evidence for redundancy between HAT enzymes in S. pombe. We report the initial characterization of fission yeast elp3 + and examine phenotypes associated with single, double, and triple mutations of Δmst2, Δgcn5 and Δelp3. We show that these enzymes affect expression of overlapping but nonidentical sets of genes, suggesting overlapping contributions to gene expression. Consistent with this, we observe biochemical evidence that these proteins share common histone substrates. We also observe that these phenotypes are distinct from the gene expression effects of a temperature sensitive allele of mst1, and suggest that multiple histone targets regulate gene activation in fission yeast.

Results
Characterization of Δelp3 and interaction with other HAT mutants.
In order to compare the roles of different histone acetyltransferases on transcription in fission yeast, we first characterized the disruption of the histone acetyltransferase gene elp3 + , the likely orthologue of the ScELP3 subunit of Elongator complex [21]. We found that S. pombe elp3 + is a non-essential gene, as is its budding yeast orthologue [30]. However, Δelp3 mutant cells have elongated morphology compared to wild type, indicative of cell cycle delay (Fig. 1A). The mean length of mononucleate cells is 6.7 μm in wild type, and 10.7 μm in Δelp3. For binucleate cells, wild type averages 9.8 μm compared to 15.9 μm in Δelp3. Thus Δelp3 is approximately 60% larger than wild type. Also as in budding yeast [30], fission yeast Δelp3 grows slowly due to a prolonged lag phase (Fig. 1B). Although the growth rate of Δelp3 cells once they have reached exponential phase is similar to wild type, they appear to enter stationary phase at a lower cell density. Given the presumed role of Elp3 as part of the Elongator complex, we presume that this reflects defects in one or more growth-associated transcription program.
We examined other phenotypes associated with Δelp3, and compared them to phenotypes associated with mutations in other non-essential histone acetyltransferases, Δmst2 and Δgcn5. We observed that Δelp3 has poor growth overall compared to either Δgcn5 or Δmst2 (Fig. 1C). Both Δgcn5 and Δelp3 showed slight temperature sensitivity. As reported previously [28], Δgcn5 is sensitive to salts (KCl or CaCl 2 ), but we observed no overlap in this phenotype with Δmst2 or Δelp3. We also assayed the sensitivities of these HAT mutants to different DNA damaging agents (Fig. 1D), since histone acetylation is necessary for proper repair from DNA damage [31]. The Δmst2 mutants are sensitive to hydroxyurea, which depletes nucleotides and results in replication fork stalling, and to high doses of the DNA alkylating agent MMS, which causes DNA damage [11]. We observed that Δgcn5 and Δmst2 show similar sensitivity to hydroxyurea. Consistent with our previous results, Δmst2 is modestly sensitive to low levels of MMS, but we observed no sensitivity in Δgcn5 or Δelp3. We also observed no sensitivity to the topoisomerase inhibitor camptothecin (which causes S phase specific DNA breaks) or UV irradiation. Δmst2 is also sensitive to the microtubule poison thiabendazole (TBZ), a phenotype typical of mutations that affect centromere function or chromosome segregation.
We next constructed double and triple mutants to see whether there was evidence for epistasis or synthetic growth defects under these conditions. All combinations were viable, although the triple mutant Δgcn5 Δelp3 Δmst2 was very slow-growing. The slow-growth phenotype of the triple mutant strain made it difficult to distinguish pathway specific effects from general sickness in this strain.
Previously, we assigned Mst2 as the likely ScSas2 orthologue, based on the similarity of their phenotypes, although it is also phylogenetically related to ScSas3. In budding yeast Δsas3 Δgcn5 is lethal [26], but in fission yeast Δmst2 Δgcn5 is viable. We assayed growth at 30°a nd 36°C for the single and double mutants. At 30°C, there was little effect observed. At 36°C, the Δgcn5 Δmst2 mutant displays a temperature sensitivity resembling Δgcn5, indicating that there is no overlap between Gcn5 and Mst2 with regards to heat stress. The Δgcn5 Δelp3 mutant, however, shows increased temperature sensitivity compared to the single mutant strains. The Δmst2 Δelp3 mutant displayed a phenotype intermediate between the parents, suggesting that the faster growth associated with Δmst2 in relation to wild type is also able to overcome the slow growth associated with Δelp3 (Fig. 1C).
We examined the salt sensitivity of the double mutants, because Δgcn5 has been associated with this phenotype [28]. The Δgcn5 Δmst2 double mutant is hypersensitive to salt compared to Δgcn5 alone (Fig.  1C), suggesting that for this function, these two HATs overlap. In contrast, the Δgcn5 Δelp3 double mutant has a phenotype similar to Δgcn5 alone suggesting that Elp3 does not contribute to proper salt stress response.
Since Δmst2 has been linked to damage sensitivity, we also examined the response to different DNA damaging agents (Fig. 1D). Δmst2 Δgcn5 cells also showed increased sensitivity to MMS and CPT treatment relative to the two single mutants, suggesting that these two HAT enzymes make redundant contributions to replication-specific DNA damage. Interestingly, Δelp3 modestly suppressed the sensitivity of Δmst2 to HU (Fig. 1D). All the double mutants showed some sensitivity to TBZ; however, Δgcn5 Δelp3 was more sensitive than the parent strains.

Genome wide expression studies
Previous studies showed while Gcn5 is bound to a high proportion of genes in S. pombe growing on rich media, it is requested for the expression of only a few of its binding targets [27,28]. This together with the synthetic and additive phenotypes observed here with the different mutant combinations suggests that HAT enzymes redundantly affect common pathways at the level of transcriptional regulation. However, it is also formally possible that they independently affect regulation of separate pathways that overlap downstream at the level of cellular function. To distinguish between these possibilities, we performed genome-wide analysis of expression using microarray technology to identify target genes that show HAT-specific changes in gene expression. We examined the effects on transcription upon deletion of the GNAT family HATs Gcn5 and Elp3 and the MYST HAT Mst2 both singly and in combination. We chose a very statistically rigorous cutoff of 3.25 log2 fold change for genes that we considered differentially expressed. This allowed us to be extremely confident that all results were biologically relevant and eliminated most false positives.
We found that there were very few genes whose expression was significantly changed by 3.25 log2 fold or more during asynchronous growth compared to wildtype cells in Δmst2, Δgcn5 or Δelp3 cells (Table 1). Of the small number of genes with increased expression, the RecQ-type DNA helicase SPBCPT2R1.08c, located in the sub-telomere domain (tlh2 + ; [32,33]) was the only significantly up-regulated gene in all three mutant strains. Consistent with previously published data we found that there was an enrichment of genes involved in mating and meiosis in Δgcn5 cells (Additional File 1). Interestingly, of the few genes with decreased expression in Δgcn5 and Δmst2, most are found within 150 kb of the ends of chromosome I or II, suggesting that cells require Gcn5 and Mst2 for expression of genes localized near the sub-telomeric region (Fig. 2D). The few genes that were differentially expressed in Δelp3 and Δgcn5 showed significant over-representation in several functional ontology groups. In agreement with prior reports [27] we observed that Δgcn5 mutants derepressed mating and meiosis genes. Specifically we found that gene ontology (GO) terms [34] related to mating and meiosis were found in the up-regulated genes, indicating that Gcn5 is required for repression of these genes. In Δelp3 mutants we found that genes significantly enriched in GO classes associated with transmembrane transport were down-regulated and those associated with ion transport were up-regulated (Fig 2, Additional File 1). Interestingly, mei2, which has been shown to be repressed by Gcn5 [27] was down-regulated in Δelp3 (Fig. 3). This may indicate that Gcn5 and Elp3 target the same pathway but with opposite effects. The few genes that were differentially expressed in Δmst2 mutants provided no over-represented GO classes.

Global gene expression changes in double mutants
We next asked whether the transcription profiles of the double mutants were simply additive to the single mutants, or whether they affected new genes not in common to the single mutant strains. We assayed changes in gene expression in the double mutants. The number of differentially expressed genes in the double and triple mutants was significantly increased compared to the respective single mutants for each strain tested (Fig. 2, Additional File 1, Additional File 2), suggesting overlapping regulation rather than an additive effect.
Only the helicase tlh2 + and a metabolic kinase, dak1 + were significantly up-regulated in all mutant combinations. The double mutants Δgcn5 Δelp3 and Δmst2 Δelp3 had the most overlapping up-regulated genes ( Fig.  2, Additional File 3). These genes included metal transporters and the sporulation specific gene spo6 + . We found that Δgcn5 Δmst2 cells had a greater number of up-regulated genes related to the mating pathway than Δgcn5 alone, suggesting that Gcn5 and Mst2 are functionally redundant in repressing sexual differentiation. This is also observed, although to a lesser extent, with the Δgcn5 Δelp3 mutant. However changes associated with Δmst2 Δelp3 were not significantly enriched in gene classes associated with mating or meiosis, except for spo6 + , which is essential for progression of meiosis II and sporulation. When we examined down-regulated genes in the double mutants, we again found the most overlap between Δgcn5 Δelp3 and Δmst2 Δelp3 cells, however there were more differentially expressed genes in Δgcn5 Δmst2. A large fraction of these genes were found to be sub-telomeric (Fig. 2D) which suggests that Mst2 and Gcn5 but not Elp3 function preferentially in the sub-telomeric regions. This is consistent with previous studies suggesting that Mst2 works at the telomere [11].

Functional redundancies between HATs
To determine whether the mutations affected distinct functional groups, we categorized the affected gene targets by gene ontology (GO) classification. As was apparent for the individual genes, the genes affected by the double mutants were more significantly enriched in a larger number of GO classes than for the single mutants. For GO classes enriched in down-regulated genes there were 23 terms for Δgcn5 Δelp3, 22 for Δgcn5 Δmst2, while Δmst2 Δelp3 had 10 enriched GO terms (Fig 2, Additional File 1). Curiously there is almost no overlap in enriched GO terms associated with genes down-regulated between double mutants, with the exception of cellular response to stress and response to stimulus shared between mst2 Δelp3 and Δgcn5 Δelp3. Therefore we conclude that Gcn5, Mst2 and Elp3 are functionally redundant in activating transcription. For GO categories enriched with up-regulated genes there were 7 categories for Δgcn5 Δelp3, 36 for Δgcn5 Δmst2, and Δmst2 Δelp3 had 3 enriched GO terms (Fig. 2). There are no overlapping enriched up-regulated GO categories between double mutants. When we examined the number of genes and the GO terms they represent in the double mutants compared to the single mutants, we conclude that Gcn5, Mst2 and Elp3 are functionally redundant in repressing transcription because the number of terms enriched in the double mutants cannot be explained by adding the single mutants' terms together. There were more significantly enriched genes in the double mutants than in the single mutants.

Validation of microarray results
We used quantitative PCR (qPCR) to confirm the microarray results. The ddCt method [35] was applied to measure relative fold differences between each strain and one wild type sample with act1 + used as the loading control. For each strain, we chose genes that were predicted by the microarray to have significant changes in transcription. The fold changes of the other two wild type samples relative to the first wild type sample was small (2-fold at the most) across all the loci. The fold changes of the other two control genes relative to act1 + were also small (2-fold at the most) across all the   strains. We found high degrees of congruence in gene expression as measured by RT-PCR and microarray analysis. (Fig 3).

Salt response in Δmst2 and Δgcn5 mutants
Since Δmst2 and Δgcn5 showed a synthetic phenotype in salt sensitivity, we performed expression profiling on Δmst2 and Δgcn5Δmst2 double mutant under KCl treatment and compared this to our previous results on Δgcn5 using the same platform [28]. The number of affected genes in Δmst2 increases during KCl induced stress but fewer genes are affected in Δmst2 compared to Δgcn5, consistent with the observation that Δmst2 alone has little effect on salt response. To find genes that require Gcn5 or Mst2 during KCl stress we compared the expression levels between Δgcn5 and Δgcn5 Δmst2.
We found 99 genes that were significantly down-regulated and 191 genes significantly up-regulated in the double mutant, compared to Δgcn5 single mutant (Additional File 4). About 80% of these genes were not significantly regulated in either of the single mutants (Fig 4). This suggests that the two HAT enzymes overlap in the regulation of salt-responsive pathways, although there was no significant overlap in particular GO pathways.

Histone acetylation levels in HAT mutants
Gcn5/SAGA is a known regulator of H3 acetylation, as is Elp3/Elongator [21,36]. The phenotypes associated with Δmst2 suggest it is the functional orthologue of ScSas2, but if this is the case, there is no obvious NuA3/ Sas3 homologue in S. pombe to overlap with Gcn5 activity, as Sas3 does in S. cerevisiae. Because more genes showed changes in expression in the double mutants than would be expected by a simple sum of those affected in the single mutants, we examined levels of histone H3 acetylation in the mutant strains to see whether there is evidence for overlapping specificity at the level of global histone acetylation.
As expected, we found H3 acetylation levels on H3K9, K14 and K18 were dramatically reduced in the Δgcn5 mutant (Fig 5). In Δelp3 and Δmst2 as well as in the Δmst2 Δelp3 double mutant there is only a modest decrease in H3K9 acetylation levels compared to wild type. In contrast, all strains lacking gcn5 show significant loss of H3K9ac (Fig. 5A), suggesting that Gcn5 is the major contributor to H3K9 acetylation. Similar results were observed for H3K18ac.
For H3K14 acetylation we found significant reductions of acetylation levels for each single mutant compared to wild-type cells. Strikingly, in the double mutant Δgcn5 Δmst2, the signal is much more strongly reduced (Fig.  5B). None of the deletion mutants showed any significant reduction in the overall level of histone H3 (Fig.  5D). Therefore we conclude that Gcn5 (GNAT family) and Mst2 (MYST family) each contribute to acetylate H3K14, which has been shown to be important for the response to salt stress [37]. Defective acetylation of H3K14 could explain the hypersensitivity to salt stress observed in the Δgcn5 Δmst2 mutant.

Genes differentially expressed by the essential HAT Mst1
Finally, we examined the essential MYST family HAT Mst1 using the temperature sensitive allele mst1 ts [38]. The mst1 mutant was grown at the semi-permissive temperature 30°C overnight and RNA was isolated for microarray analysis. There were many more differentially expressed genes in the mst1 mutant compared to any of the non-essential HATs studied here (Table 2). There were also many more up-regulated than downregulated genes in the mst1 mutant, suggesting that it functions in repression of gene expression. Consistent with this there were no GO categories that were significantly enriched in the down-regulated genes, however there were many that were enriched in the up-regulated genes (Additional File 1). These GO groups include metal and salt transporters, and other metabolic processes. When we compared these GO categories with those observed for the other mutants, we found that there was some overlap with the double mutant Δgcn5 Δmst2 and the triple mutant Δgcn5 Δelp3 Δmst2, suggesting general stress responses in these mutantsare deregulated (Fig 6).

Discussion
Fission yeast shares several conserved histone acetyltransferases with other eukaryotes. In the GNAT family, the Elp3 (KAT9) and Gcn5 (KAT2) enzymes are highly conserved [7]. In the MYST family, the Kat5 HAT Esa1/Tip60/Mst1 protein is also highly conserved. However, the relationships between the other MYST family members are less clear-cut. In budding yeast, two additional MYST proteins ScSas2 and ScSas3 have been characterized. ScSas2 antagonizes ScSir2 in telomere silencing as a component of the SAS complex. ScSas3, a component of NuA3 [16], overlaps with Gcn5 such that a double mutant Sc gcn5 sas3 is lethal [26]. By contrast, there is only one additional MYST protein in S. pombe, Mst2, which resembles both S. cerevisiae proteins in primary sequence. Like ScSas2, SpMst2 antagonizes SpSir2 in telomere silencing [11], while work from this study suggests that like ScSas3, SpMst2 acetylates histone H3K14. This works shows that SpMst2 functions similarly to both ScSas2 and ScSas3. To dissect out the relationship between these potentially overlapping HAT enzymes Mst2, Elp3 and Gcn5, we compared their phenotypic and transcriptional interactions in S. pombe.
First, we performed the initial characterization of the catalytic HAT, Elp3, which is the likely catalytic subunit of the Elongator complex in S. pombe. The strain elp3 is viable, although it suffers defects in overall cell growth. It shows a delay in entry into the cell cycle, and premature cell cycle exit (Fig. 1B). It should be noted that some of the phenotypes seen in Elp3 defective strains could result from defects in other functions that have been reported for the Elongator complex in addition to its role as a HAT [39,40]. We next investigated whether Δelp3 shows genetic interactions with mutations in other non-essential HAT genes encoded by gcn5 + and mst2 + . Double and triple mutants are all viable, although they show increasingly severe growth defects as more HAT genes are mutated. The viability of SpΔmst2 Δgcn5 contrasts with the lethality of ScΔsas3 Δgcn5 and suggests that Mst2 is not a simple equivalent of ScSas3. Alternatively, there may be another enzyme in fission yeast, which provides an additional degree of redundancy. We suggest that Mst2 fulfills many of the functions associated with both ScSas2 and ScSas3 (see below).  Phenotypic analysis suggests distinct effects of each HAT. For example, as shown previously Δgcn5 mutants are de-repressed for mating and meiosis functions, and are sensitive to certain salt-stress conditions [27,28]. Δmst2 mutants show modest sensitivity to DNA damaging agents such as HU and MMS. Δgcn5 Δmst2 double mutants show significantly increased salt sensitivity, and a modest increase in DNA damage sensitivity relative to single mutants. Thus, we conclude that in response to salt stress, Gcn5 and Mst2 have a substantial functional overlap. In contrast, Δgcn5 Δelp3 and Δmst2 Δelp3 showed less evidence for synthetic phenotypes, with only a modest increase in TBZ sensitivity in Δgcn5 Δelp3 relative to the parents (Fig. 1C &1D). In fact, Elp3 appears to be antagonistic to Mst2, because Δelp3 suppresses Δmst2 mutant sensitivity to high salt concentrations and HU. We speculate this could be due to slowing down the growth of Δmst2 cells, allowing time for the cells to repair any damage that may have occurred. Together, these data suggest that Gcn5 affects multiple pathways, some overlapping with Elp3, and others overlapping with Mst2. However, since the double mutants and even the triple mutants are viable, either additional HAT enzymes play a role, or the acetylation modifications controlled by these enzymes are not strictly essential for viability.
To determine whether the functional redundancy between HATs is at the level of gene expression, we performed transcriptional profiling using DNA microarrays of Δelp3, Δgcn5, and Δmst2. Similar to previous results with Δgcn5 [27,28] we found that few genes were affected by deletion of any single HAT gene under favorable growth conditions. Although HATs are presumed to be gene activators that facilitate transcriptional initiation or elongation [5], we found that gene expression was as likely to be increased as reduced in our mutant strains. Up-regulated genes could represent indirect targets, although many Gcn5-repressed genes are known to be bound by Gcn5 [37]. This suggests that Gcn5 may play a direct role in gene repression, which has been described previously [27]. It is thus possible that the other HATs studied here also play direct roles in transcriptional repression.
When we examined the double mutants, we found that their gene expression profiles were not simply the combination of the profiles from the two single mutants, but affected additional genes. This supports the hypothesis that the HATs play partly redundant roles in gene regulation, with multiple enzymes contributing to expression of common targets. Importantly, although they may affect common genes, the regulatory mechanisms may be distinct. In budding yeast, although ScGcn5 and ScElp3 affect transcription of the ScHsp70 genes ScSSA3 and ScSSA4, this occurs by different mechanisms because Gcn5 is required for transcription factor binding while Elp3 is required for proper Pol II elongation [41]. However, we have shown that in fission yeast SpGcn5 also has a role in transcriptional elongation [37] which could suggest a more direct mechanistic overlap for SpGcn5 and SpElp3 in this species.
Only one gene was significantly increased in all strains: the RecQ-type DNA helicase SPBCPT2R1.08c, located in the sub-telomere domain (Tlh2; [32,33]). We showed previously that Δmst2 increases silencing at the telomere and in telomere associated regions, consistent with a role for Mst2 in antagonizing Sir2. Loss of Mst2 results in a loss of overall H4 acetylation [11] and H3K9 acetylation in Δmst2 in a very limited region adjacent to the telomere 18 kb from the end. Interestingly, this region is telomere-distal to the tlh2 helicase gene located 13 kb from the end of the telomere chromosome II, which is highly up-regulated in the mutant. In contrast, genes at 27.5, 29 and 31 kb away from the telomere are down-regulated in Δmst2 (SPAC186.06, SPAC186.05 and SPBCPT2R1.02 respectively), while at 47 kb telomere distal (SPAC869.01), there is no change in expression levels. This suggests a gradient effect on transcription of genes in these regions.
The few genes that are down-regulated in Δgcn5 are localized near the ends of chromosome I and II which suggest that the repression might be due to spreading of heterochromatin by loss of H3K9ac. The combined effect of the double mutant may indicate a combination of effects associated with loss of both histone H3K9 and H3K14 acetylation. Clearly, there are specific regions at the telomeres that are regulated by specific HATs but more work needs to be done to fully understand the boundaries of the telomeres and the regulation between the highly heterochromatic region and the rest of the chromosome.
We used gene ontology (GO) classification to define the functional roles of the few genes that were differentially expressed in single mutants. We found that those in Δelp3 and Δgcn5, but not Δmst2 associated with a few distinct classes. Consistent with previous results [27], Δgcn5 mutants showed increased expression of specific sexual differentiation genes as well as enrichment for GO terms [34] related to mating and meiosis.
Gcn5 represses transcription of ste11 + [27], which is an important regulator of the mating pathway in response to nutrient limitation. Although we find very similar functional groups that are positively regulated in the absence of Δgcn5, we did not see evidence for induction of ste11 above our threshold level (Table 1).
Consistent with prior work, our study also found that mei2 + was up-regulated in Δgcn5 [27]; this was also seen for mst1 ts and to a lesser extent Δmst2 (Fig 3), indicating multiple inputs into this essential regulator of sexual development. The Δgcn5 Δmst2 mutant increased mei2 + and ste11+ mRNA levels above those observed in the single mutants, suggesting these two HATs function redundantly in repression. It is possible that they might also have non-histone targets in common since the repression of mei2 + expression by Gcn5 was suggested to be mediated by a histone independent mechanism [27].
In contrast, deletion of Δelp3 results in loss of mei2 + transcription and in the Δgcn5 Δelp3 and Δmst2 Δelp3 double mutants there is also a reduction of mei2 + transcription. Even the triple mutant shows lower levels of mei2 + than the Δgcn5 Δmst2 double mutant, suggesting that Δelp3 has an opposite effect on mei2 + transcription than Δgcn5. Interestingly transcription of SPAC1039.02, a putative phosphoesterase, shows a similar, but opposite, regulation pattern as it is down-regulated in Δgcn5, mst1 ts and Δmst2 while being up-regulated in Δelp3. Thus, for some targets, Gcn5 and Mst2 appear to overlap, while Elp3 appears to be antagonistic. This finding suggests that the common model linking HAT enzymes with gene activation is to simplistic. Dissecting the contribution of each HAT to gene expression or other effects will require future molecular studies mapping the histone acetylation and physical binding of HATs at different gene regions or their association with the transcription machinery.
To determine the molecular basis for the overlap of HAT targets, we examined histone H3 acetylation. Three acetylation sites on histone H3 are associated with gene expression: K9, K14, and K18 [42]. H3K9 is acetylated in euchromatin, but methylated in heterochromatin; consistent with other species, our data indicate that Gcn5 is the primary contributor to H3K9Ac because acetylation of this residue is strongly reduced in Δgcn5 but relatively unperturbed by the other HAT mutations.
H3K14 acetylation is also associated with gene expression [42]. Mutations that change the H3K14 residue display hypersensitivity to KCl and CaCl 2 induced stress in Δgcn5 cells [37]. Little or no H3K14 acetylation is observed in the Δgcn5 Δmst2 double mutant. Consistent with this, we observed that Δgcn5 Δmst2 is significantly more salt-sensitive than either single mutation. We conclude that these two HATs both contribute to H3K14 acetylation in response to stress. Importantly, this acetylation is not essential for viability in S. pombe. Data from budding yeast also indicates an overlap between ScSas3 and ScGcn5 in H3K14 acetylation [26]. Interestingly, the budding yeast data also suggest that Δsas3 is not lethal in combination with disruption of other SAGA subunits that are essential for Gcn5 activity against histones, and deletion of histone N-terminal H3 tails entirely is not lethal in budding yeast [26]. Thus, the essential redundant function of these enzymes may not be in histone modification. It is now well-established that HAT proteins acetylate substrates other than histones [43]. Perhaps it is one such substrate that relies on Gcn5 or Sas3 in budding yeast, but on a different enzyme in fission yeast. These results indicate that Mst2 has functions in common with both budding yeast enzymes.
Finally, we found that as expected there was little overlap between expression profiles in the non-essential HAT mutant strains, and cells with a temperature sensitive mutation of the essential HAT mst1 + . Mst1 is known to be required for DNA damage response and chromosome segregation [38]. However, we did not see a significant increase in GO terms related to these functions, which is consistent with data suggesting its effects are not mediated through the transcriptional program [44]. There was a significant overlap of regulated genes between Mst1 and the triple mutant Δgcn5 Δmst2 Δelp3, as half of the genes regulated in the triple mutant were also regulated in mst1 ts . When we compared the regulated GO terms from mst1 ts and Δgcn5 Δelp3 Δmst2 we found that these mutants shared terms related to metal and ion homeostasis, metal transports as well as regulation of meiosis. Since both strains show severe growth defects under the conditions employed, we suggest that these overlaps may represent the effects of general cellular stress rather than a particular transcriptional program.

Conclusions
We isolated and characterized the putative Elongator HAT Elp3, showing Δelp3 cells are delayed in the cell cycle relative to wild-type cells, and then compared it to other HAT mutants. Our study suggests, through analysis of single, double and triple mutants of MYST and GNAT family HATs, that histone acetyltransferases play partially redundant role in gene regulation, and that their role can be positive, or negative for the transcription of genes. We found evidence for significant redundancy between Mst2 and Gcn5 in regulation of stress response, and response to DNA damage. We found that different HAT families can share specificity for a common lysine residue in histone H3, thereby contributing to the modulation of gene expression. For example, we found that Gcn5 and Mst2 both acetylate histone H3 lysine 14. This work highlights the functional redundancies of different HAT families on transcription, salt response, and DNA damage repair. Future work will be required to investigate the direct contributions of these HATs to individual genes, as opposed to selected regulation of other transcriptional master regulators.

Strains, media, and manipulations
Strains used in this study are listed in Table 3. Strains were grown and maintained on yeast extract plus supplements (YES) or Edinburgh minimal media (EMM) with appropriate supplements, using standard techniques [45]. Matings were performed on synthetic sporulation agar (SPAS) plates for 2-3 days at 25°C. Double mutants were constructed by standard tetrad analysis or random spore analysis. G418 plates were YES supplemented with 100 μg/ml G418 (Sigma).

Growth Assays
Wild type and Δelp3 cells were grown to stationary phase >1 × 10 8 cells/ml and then diluted in YES media to four different concentration 1 × 10 4 , 1 × 10 5 , 4 × 10 5 and 1 × 10 6 cells/ml respectively and grown at 30°C. Cells were counted regularly for 48 hours generating growth curves to cover the beginning and end of the exponential growth.

Damage assays
Cells were grown to mid-log phase, serially diluted fivefold and spotted onto YES plates or YES plates were irradiated with the indicated doses of UV irradiation using a Stratalinker (Stratagene) or YES plates containing 10 or 12 μg/ml thiabendazole (TBZ), 5 mM hydroxyurea (HU), 7 μM camptothecin (CPT), 0.005% methyl methanesulfonate (MMS) and incubated for 2-3 days at 32°C. For the salt sensitivity assays cells were plated out in five-fold dilution series on rich media (YES) plates with or without KCl (1 M or 1.5 M) or CaCl2 (0.1 M or 0.25 M) at either 30°C or 36°C for 2 to 4 days.

RNA prep and Microarray protocols for Affymetrix and Eurogenetech
RNA was isolated from three independent isolates growing at 30-32 degrees Celsius using phenol/chloroform extraction [45], phase lock gels (Eppendorf) and resuspended into TE after ethanol precipitation. Biotinylated cRNA was prepared for Affymetrix array analysis from 5 ug Total RNA using the standard Affymetrix one-cycle target labeling protocol. Samples were assayed for gene expression using Affymetrix GeneChip Yeast Genome 2.0 arrays consisting of probe sets representing 5,031 genes in S. pombe based on Sanger (June 2004). Data was analyzed using the R statistical environment. Affymetrix CEL files were processed and normalized using RMA [46]. The linear modeling package Limma [47] was used to calculate p-values and derive gene expression coefficients and to identify differentially expressed genes. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [48] and are accessible through GEO Series accession numbers GSE17259 and GSE17262 http://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?token=dzmlrwoyekgmyp-k&acc=GSE17298. For microarrays using the Eurogenetech platform (salt sensitivity analysis), sample collection, processing and normalization was performed as in [28]. Genes were considered differentially expressed if the mutant/wt ratio was greater than 3.25 log 2 fold, with a p-value of 0.05 or less. Additional File 5 lists genes differentially expressed with a mutant/wt ratio of 1.7 log 2 fold, with a p-value of 0.05 or less.
qPCR Primers for the entire S. pombe genome were created by Chris Seidel using uPrimer. Total RNA in the amount of 2-4 ug was treated with DNase (Turbo DNase Treatment Ambion #AM1907) and 200-600 ng of the treated RNA was converted to cDNA by RT-PCR with the use of two primers. The concentration and sequence for the primers are 2 ng/uL final for the random primer pdN-15 and 20 ng/uL final for the anchored oligo-dT OdT-19N. A Corbett robot was used to load PCR reactions in 384-well plate format for analysis on an ABI-7900. Standard curves for the 9 primer pairs were made with 5 ten-fold dilutions of gDNA starting with 1000 ng/15 uL qPCR reaction and ending with 0.1 ng/15 uL qPCR reaction. The qPCR experiment consisted of measuring 9 loci in 24 samples three times each. 8 ng of the cDNA was used in each qPCR reaction without purification of the RT reaction. The final primer concentration was 500 nM each.

Venn diagram program, GO term analysis
Venn diagrams were made using VENNY (Oliveros, J.C. (2007) VENNY. An interactive tool for comparing lists with Venn Diagrams http://bioinfogp.cnb.csic.es/tools/ venny/index.html). Gene ontology classifications were classified at http://amigo.geneontology.org with the database filter set as GeneDB S. pombe with a threshold of maximum p value of 0.05 and minimum number of gene products of 2. We selected for publication only the biological process classifications.

Histone acetylation
Cell lysates was obtained as described by [49] with the following modifications; the volume of the cell cultures were increased 2.5 times to get enough material for 10 or more gels. Briefly cells were inoculated in YES and grown overnight (o/n) to 1 × 10 7 cells/ml in a total volume of 25 ml and harvested in room temperature (RT) by centrifugation, washed once in H 2 O then resuspended in 0.75 ml H 2 O and an equal volume of 0.6 M NaOH, incubated 10 minutes at RT. Cells were removed from the NaOH by centrifugation and dissolved in 175 μl of modified SDS-sample buffer (60 mMTris-HCl, pH 6.8, 4% β-mercaptethanol, 4% SDS, 0.01% BPB, and 5% glycerol). Approximately 15 μl were used to run on precast gradient gels (Bio-Rad, 161-1105). Transfer was performed using the iBlot™ Dry Blotting system (Invitrogen). After transfer the membranes were blocked in in PBS containing 3% milk for 30 min at RT followed by o/n incubation at 4°C with primary antibody (anti-histones); or blocked o/n at 4°C followed by incubation with anti-actin antibody, at a 1:12 000 dilution (ms ab8224, Abcam) 1 hour at RT. For detection of histones antibodies the following dilutions were used: H3