Multiple Mechanistically Distinct Functions of SAGA at the PHO5 Promoter

Abstract

Our previous studies have shown that the rate of chromatin remodeling and consequently the rate of PHO5 activation are strongly decreased in the absence of Gcn5 histone acetyltransferase activity. Using chromatin immunoprecipitation, we demonstrate that SAGA is physically recruited to the PHO5 promoter. Recruitment is dependent on the specific activator Pho4 and occurs only under inducing conditions. Spt3, another subunit of SAGA, also plays a role in PHO5 activation but has a function that is completely different from that of Gcn5. An SPT3 deletion severely compromises the PHO5 promoter and reduces the extent of transcriptional activation by diminishing the binding of the TATA binding protein to the promoter without, however, affecting the rate or the extent of chromatin remodeling. A gcn5 spt3 double mutant shows a synthetic phenotype almost as severe as that observed for an spt7 or spt20 mutant. The latter two mutations are known to prevent the assembly of the complex and consequently lead to the loss of all SAGA functions. The absence of the Ada2 subunit causes a strong delay in chromatin remodeling and promoter activation that closely resembles the delay observed in the absence of Gcn5. A deletion of only the Ada2 SANT domain has exactly the same effect, strongly suggesting that Ada2 controls Gcn5 activity by virtue of its SANT domain. Finally, the Gcn5 bromodomain also contributes to but is not essential for Gcn5 function at the PHO5 promoter. Taken together, the results provide a detailed and differentiated description of the role of SAGA as a coactivator at the PHO5 promoter.

Cells respond to environmental changes by reprogramming the expression of specific genes throughout the genome. The rate of transcription of a particular gene is determined by the interaction of diverse regulatory proteins with specific DNA sequences in its promoter. In addition to sequence-specific DNA binding proteins, a variety of large pleiotropic coactivator complexes also control gene expression. Among them are two classes possessing chromatin-modifying activities. One category, exemplified by the Swi/Snf chromatin-remodeling complex, brings about changes in nucleosomal structure and organization (for reviews, see references 9 and 44). A second class, best represented by the SAGA complex, targets histones and covalently modifies their N-terminal tails (for reviews, see references 42 and 48).

SAGA is a large multiprotein complex that contains Gcn5, a histone acetyltransferase (HAT) (21-23, 50) whose activity has been shown to be required for the transcription of a subset of genes in vivo (7, 24, 36, 46). However, the transcription of some genes, although Gcn5 independent, still requires SAGA; this finding suggests the existence of distinct SAGA activities not related to Gcn5 (38). Indeed, several lines of genetic and biochemical evidence have demonstrated that SAGA possesses multiple activities important for transcriptional regulation (21, 30, 47, 50). SAGA contains three Ada subunits (Ada1, Ada2, and Ada3), a subgroup of Spt proteins (Spt3, Spt7, Spt8, and Spt20), five TATA-binding protein-associated factors (Taf_II17, Taf_II25, Taf_II60, Taf_II68, and Taf_II90), and TraI (48). A functional analysis of SAGA subunits has revealed that Spt7, Spt20, and Ada1 are required for all SAGA functions (47), probably because of their critical importance for SAGA assembly (21, 50); however, more direct functions of these subunits in transcriptional regulation have not been reported. On the other hand, Spt3 and Spt8 have each been found to be required for a subset of SAGA functions (47, 50), and evidence has been provided that they are important in the context of TATA binding protein (TBP) function (10, 18). Recent work confirmed a critical role for Spt3 in recruiting TBP to the Gcn5-independent GAL promoters (11, 37).

Within SAGA, Gcn5 is contained in a trimeric subcomplex with Ada2 and Ada3; this comples also exists as a separate, so-called ADA complex in Saccharomyces cerevisiae (17, 21, 29, 41). It was shown that deletion of Ada2 results in the loss of Gcn5 from both Gcn5-containing complexes (21), suggesting a role for Ada2 in tethering Gcn5 to these complexes. In addition, there is evidence for a more direct role of Ada2 in regulating Gcn5 function in vivo (4, 14, 41, 45). It was previously reported that Ada2 contains a SANT domain, a motif found to be the hallmark of a number of proteins involved in transcriptional regulation, most of them being involved in chromatin-related activities (1). This finding also suggested that Ada2 may have more specific functions carried out by its SANT domain. Indeed, recent reports provided evidence for critical involvement of the Ada2 SANT domain in substrate recognition and the regulation of Gcn5 HAT activity (12, 51).

Another conserved protein domain that is also found in several transcription-related proteins that functionally interact with nucleosomes is the bromodomain (28). This domain was therefore proposed to provide structural interactions of proteins with nucleosomes. It has indeed been shown that bromodomains interact specifically with acetyllysine in the N-terminal tails of histones H3 and H4 (15, 32, 34, 43), leading to the suggestion that HATs, as well as other bromodomain-containing proteins, may be anchored to acetylated regions in chromatin through bromodomains (57). When the function of the Gcn5 bromodomain was examined in vitro, its removal reduced the HAT activity of SAGA on nucleosomal substrates but not on free histones (14).

In an effort to understand the various steps in transcriptional activation and the role of chromatin remodeling in this process, we have focused our attention on the yeast PHO5 promoter, which was shown to be a highly suitable model for that purpose (31, 53). PHO5 transcription is strongly regulated in response to the phosphate concentration in the medium. Under repressing conditions, i.e., in phosphate-containing media, Pho4, the dedicated transcriptional activator of the PHO genes, is negatively regulated by phosphorylation at three different levels. This covalent modification of Pho4 promotes its export from the nucleus, prevents its reimport into the nucleus, and abolishes interactions of Pho4 with the pleiotropic coactivator Pho2 (35), which assists Pho4 in binding to the PHO5 promoter (6). Under conditions of phosphate starvation, the phosphorylation of Pho4 is prevented, allowing strong induction of the PHO5 promoter. PHO5 activation is accompanied by a dramatic alteration of the chromatin structure at its promoter (53). Four positioned nucleosomes present in the repressed promoter undergo a remodeling event, resulting in a fully accessible promoter (3). Despite the fact that chromatin remodeling is a critical prerequisite for promoter activation (53), the full induction of PHO5 under conditions of phosphate starvation was found to be largely independent of Gcn5 HAT activity as well as of Swi/Snf chromatin-remodeling activity (24); this finding suggested that neither chromatin-modifying complex plays a significant role in the activation process. However, when the kinetics of PHO5 induction rather than the final level were examined, striking effects of Gcn5 HAT activity on the rate of chromatin remodeling at the promoter and the rate of PHO5 induction were found. Both processes are strongly delayed in the absence of Gcn5 activity. Evidence was provided to show that this delay is not a genome-wide phenomenon but instead is related to the chromatin structure of the PHO5 promoter (7).

To gain a better understanding of the roles of Gcn5 and of additional specific functions of SAGA in chromatin remodeling and transcriptional activation, we have investigated the functional importance of other SAGA subunits in PHO5 regulation. We demonstrate that the SAGA complex is specifically recruited to the PHO5 promoter under inducing conditions in a Pho4-dependent manner. We show that the PHO5 promoter system responds in a highly specific manner to the loss or impairment of different subunits and domains of SAGA and thereby has made possible the further delineation of their roles in transcriptional regulation.

MATERIALS AND METHODS

Yeast strains and media.

Yeast strains harboring null mutations in SPT3 (FY294), SPT7 (FY963), SPT20 (FY1096), GCN5 (FY1354), and GCN5 and SPT3 (FY1441) and their isogenic wild-type strain, FY86, were obtained from Fred Winston (Harvard Medical School, Boston, Mass.). The CY series of S. cerevisiae strains containing disruptions of GCN5 (CY563) and ADA2 (CY720) or a disruption of the Ada2 SANT domain, ADA2 with a deletion of positions 97 to 106 (ada2Δ SANT) (CY884), and their isogenic ADA2⁺ strain, CY733, were obtained from Craig Peterson (University of Massachusetts Medical School, Worcester). We constructed CY500 and CY501 from CY337, which was obtained from Craig Peterson. CY500 contains a deletion of the Gcn5 bromodomain (amino acids 377 to 440) (55), and in CY501 amino acids 371 and 372 of Gcn5 are mutated to alanine (55). Strain 8136 (Ada2-myc) was obtained from Kim Nasmyth.

Yeast strains were grown in yeast extract-peptone-dextrose-adenine (YPDA; high-phosphate conditions). For PHO5 induction, logarithmically growing cells were transferred to phosphate-free synthetic medium (54).

Functional assays and chromatin analysis.

The acid phosphatase assay was carried out as described before (26). All activity values shown are the averages of at least three independent measurements (standard deviations [SDs], 5 to 10%). The preparation of yeast nuclei (2) and restriction nuclease digestion of isolated nuclei (54) were carried out as described before. Restriction nuclease accessibility values were determined by phosphorimaging of the Southern blots and calculation of the ratio of the two fragments in each lane.

Chromatin immunoprecipitation analysis.

Fifty-milliliter yeast cultures with a density of 1 × 10⁷ to 2 × 10⁷ cells/ml were treated with 1% formaldehyde for 20 min at room temperature. Cross-linking was quenched by adding glycine to a final concentration of 125 mM. The cells were sedimented and washed two times with ice-cold 0.9% NaCl. They were resuspended in 400 μl of a buffer containing 50 mM HEPES, 10% glycerol, 1 mM EDTA, 150 mM NaCl, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride (pH 7.6) and treated with a French press (kindly provided by A. Böck) three times at a pressure of 1,100 lb/in². In this step, cells were broken and simultaneously the chromatin was sheared to an average size of 450 bp. Immunoprecipitation was performed as described before (52). Fifty microliters of antibody 9E11 (a gift from Kim Nasmyth) containing the cell culture supernatant was used for precipitating Ada2-myc. Immunoprecipitated DNA was analyzed by quantitative PCR with the following primer pairs: adjacent PHO5 region-A, 5′-GTGCAGTAGTAACTTATCAGTCCG-3′; adjacent PHO5 region-B, 5′-ATGTTCCTTGGTTATCCCATCGCC-3′; PHO5 promoter-A, 5′-ACGTGTGAGTGCCAAGGTTGTATC-3′; PHO5 promoter-B, 5′-TGGAAGTCATCTTATGTGCGCTGC-3′; PGK1-A, 5′-TCAAGTCCAAATCTTGGACAGAC-3′; and PGK1-B, 5′-CTTTTCTTCTAACCAAGGGGGTG-3′. PCR products were resolved on a 2.3% agarose gel, visualized with ethidium bromide, and quantified with an ImageMaster system (Amersham Pharmacia Biotech).

Immunoprecipitation of TBP was performed with 30 μl of an anti-TBP antibody (a gift from Kevin Struhl). Precipitated DNA was quantitatively measured in triplicate by using an ABI Prism 7000 sequence detection system with the following primers and TaqMan probes: PHO5-5′-adjacent region-A, 5′-CCTTTACCGTAATTTTCAATTGCTAA-3′; PHO5-5′-adjacent region-B, 5′-TCGCTTCTTCAACAGTGGTAAAAATA-3′; PHO5-5′-adjacent region-probe, 5′-CAATGTTCCTTGGTTATCCCATCGCCA-3′; PHO5 TATA-A, 5′-AAGCCATACTAACCTCGACTTAGCA-3′; PHO5 TATA-B, 5′-GGGTAAACATCTTTGAATTGTCGAA-3′; PHO5 TATA-probe, 5′-AACATCAGCGCTTATATAC-3′; PGK1-A, 5′-GAATGGCGGGAAAGGGTTTA-3′; PGK1-B, 5′-TGTTTGAAAGAGAGAGAGTAACAGTACGA-3′; and PGK1-probe, 5′-TACCACATGCTATGATGCCCACTGTGATC-3′.

RESULTS

The SAGA complex is recruited to the PHO5 promoter under activating conditions in a Pho4-dependent manner.

An earlier study from our laboratory showed that Gcn5 activity is required for a normal rate of chromatin remodeling upon induction of the PHO5 promoter, suggesting an association of Gcn5 with the promoter under these conditions (7). To confirm the physical recruitment of SAGA to the promoter, we examined the binding of SAGA subunit Ada2 by chromatin immunoprecipitation. The binding of Ada2 was measured under repressing and inducing conditions in wild-type and pho4 strains. To test how specific any SAGA interaction would be for the PHO5 promoter, we also assayed the binding of Ada2 to the adjacent upstream region and to PGK1. The results presented in Fig. 1 show that, under conditions of full induction, Ada2 was bound specifically to the PHO5 promoter, but only in the presence of Pho4. In the strain with a PHO4 deletion, no recruitment of Ada2 was observed. Therefore, SAGA recruitment correlates with promoter activation.

FIG. 1.

Pho4-mediated SAGA recruitment to the PHO5 promoter. Strain 8136, which contains 18-myc-tagged Ada2 and a wild-type PHO4 allele, and strain 8136Δpho4 were grown logarithmically in YPDA (+Pi) or phosphate starved for 4 h (−Pi). Acid phosphatase levels in strain 8136 increased from 22 U (+Pi) to 172 U (−Pi) under these conditions. For the chromatin immunoprecipitation analysis, chromatin was fixed by formaldehyde treatment and Ada2-myc was precipitated with anti-myc monoclonal antibody 9E11. The amounts of coimmunoprecipitated DNA determined by quantitative PCR were normalized to the respective input DNA and are shown in arbitrary units. The locations of the PCR fragments used for chromatin immunoprecipitation analysis of the PHO5 locus are shown above the nucleosomal organization of the repressed PHO5 promoter. Stable nucleosomes (solid circles), unstable nucleosomes (open circles) UASp1 (black box), UASp2 (grey box), and the TATA box (T) are indicated. As a further control, the PGK1 promoter was included in the analysis. Error bars show SDs.

SAGA provides more than Gcn5 HAT activity for the transcriptional activation of PHO5.

Three SAGA subunits (Spt7, Spt20, and Ada1) have been found to be required for all SAGA functions; this fact can be best explained by their critical importance for the assembly of the native SAGA complex (21, 47, 50). Due to the fact that the activation of the PHO5 promoter is not absolutely Gcn5 dependent, additional functions of SAGA may become apparent from an analysis of a mutant strain defective in SAGA assembly. We therefore examined the kinetics of PHO5 induction in spt7 and spt20 strains. As shown in Fig. 2, deletion of SPT7 or SPT20 had a much stronger effect than deletion of GCN5 and dramatically impaired the induction of the PHO5 promoter. After 8 h of induction, when a wild-type strain had almost reached maximal activity and a gcn5 strain had reached about 85% this value, the activities in spt7 and spt20 strains were in the range of only 15 to 20%. These results clearly demonstrate that other SAGA functions, in addition to Gcn5 HAT activity, are required for the efficient transcriptional activation of PHO5. Since Spt7 and Spt20 are essential for the integrity of the native SAGA complex, the observed additional function of SAGA cannot be directly ascribed to Spt7 or Spt20.

FIG. 2.

Disintegration of the SAGA complex by an SPT7 or SPT20 deletion abolishes PHO5 induction. The time course of PHO5 induction in wild-type (wt), gcn5, spt7, and spt20 strains was monitored by measuring acid phosphate activity at the indicated times after transfer of cells from phosphate-containing medium to phosphate-free medium.

Spt3 is required for a normal level of PHO5 promoter activation.

The examination of SPT7 and SPT20 deletion strains, which are defective in the assembly of the SAGA complex, showed that SAGA provides more than just Gcn5 activity to the PHO5 promoter (see above). The Spt3 and Spt8 subunits of SAGA were previously reported to play a role in TBP binding (10, 18), and Spt3 was found to be critically required for TBP recruitment and the consequent activation of the GAL gene promoters (11, 16, 37).Therefore, Spt3 was a possible candidate for an additional SAGA function during activation of the PHO5 promoter. We therefore examined the kinetics of acid phosphatase induction in an SPT3 deletion strain, a wild-type strain, and an spt7 strain. As shown in Fig. 3A, deletion of SPT3 strongly affected PHO5 induction. In the spt3 strain, the acid phosphatase levels at any time point, including overnight induction, were about 50 to 60% those in the wild-type strain, showing that the efficiency, rather than the kinetics, of transcriptional activation was affected. This result indicates that Spt3 fulfills a function in PHO5 activation distinct from that of Gcn5. Since Spt3 is not required for SAGA integrity (50), the observed effect can be ascribed specifically to Spt3 rather than being indirect due to SAGA disassembly.

FIG. 3.

Deletion of SAGA subunit Spt3 affects PHO5 activation in a manner distinct from that of Gcn5. The time course of PHO5 induction in wild-type (wt), spt3, and spt7 strains (A) or in gcn5, spt3, spt3 gcn5, and spt7 strains (B) was monitored by measuring acid phosphatase activity at the indicated times after transfer of cells from phosphate-containing medium to phosphate-free medium.

To determine whether Gcn5 would have a kinetic effect on PHO5 activation even in an spt3 strain, in which the PHO5 promoter is severely compromised, we went on to examine acid phosphatase induction in an spt3 gcn5 double mutant. It should be noted that this double deletion also does not bring about SAGA disassembly (50). The deletion of Gcn5 did indeed delay promoter activation even in an spt3 strain (Fig. 3B), confirming the idea that the two subunits have distinct and independent functions. It should also be noted that the phenotype of the spt3 gcn5 double mutant is almost as severe as that of an spt7 mutant (compare Fig. 3A and B), which lacks a SAGA complex altogether (21, 50), suggesting that Spt3 and Gcn5 together are responsible for most if not all of the SAGA functions at the PHO5 promoter. Taken together, the results presented in Fig. 3 show that the Spt3 component of SAGA has a significant and specific function in PHO5 promoter activation that is not directly related to that of Gcn5.

Spt3 is not required for chromatin remodeling but greatly improves TBP binding at the PHO5 promoter.

To better understand at what stage in the cascade of PHO5 activation Spt3 is involved, we examined the effect of an SPT3 deletion on chromatin opening. We used our standard assay to assess chromatin structure at the PHO5 promoter; this assay involves measuring the accessibility of a ClaI site located within nucleosome −2. The accessibility of this site changes dramatically upon promoter induction, from being almost fully protected under repressing conditions to nearly 100% accessible under inducing conditions (3). The ClaI site was analyzed in a wild-type strain, a gcn5 mutant, an spt3 mutant, and an spt3 gcn5 double mutant. After 4 h of induction in phosphate-free medium, the ClaI site was largely accessible (75%) in the wild-type strain (Fig. 4, lanes 1 and 2). Importantly, nearly the same level of accessibility was detected in the spt3 strain (Fig. 4, lanes 3 and 4), despite the fact that acid phosphatase activity in the spt3 strain was only about 50% that in the wild-type strain (Fig. 3A). In both the gcn5 strain (Fig. 4, lanes 5 and 6) and the spt3 gcn5 double mutant (lanes 7 and 8), the ClaI site was only about 15% accessible after 4 h of induction, consistent with the previous observation that the rate of chromatin opening is strongly delayed in a gcn5 strain (7). Taken together, these results demonstrate that the Spt3 component of the SAGA complex plays a stimulating role in the activation of the PHO5 promoter but, in contrast to Gcn5, does not contribute to chromatin opening, as judged by our accessibility assay.

FIG. 4.

The rate of chromatin remodeling at the PHO5 promoter is not affected by the absence of Spt3. Nuclei isolated from a wild-type (wt) strain, an spt3 strain, a gcn5 strain, and an spt3 gcn5 strain after 4 h of induction in phosphate-free medium were treated for 30 min with 50 U (lanes 1, 3, 5, and 7) or 200 U (lanes 2, 4, 6, and 8) of ClaI. In order to monitor the extent of cleavage, DNA was isolated, cleaved with HaeIII, analyzed on a 1.5% agarose gel, blotted, and hybridized with probe D (3). In all instances, the ratio of the lower band to the upper band reflects the relative accessibility of the ClaI site. The position of the ClaI site is shown above the nucleosomal organization of the repressed PHO5 promoter (see the legend to Fig. 1 for details).

Earlier genetic and biochemical studies suggested an interaction between Spt3 and TBP (16,18; reviewed in reference 39), and it was recently demonstrated that Spt3 was required to recruit TBP to the GAL1 promoter (11, 37). Therefore, we wondered whether Spt3 may have a similar effect on the PHO5 promoter. As shown in Fig. 5, TBP binding to the PHO5 promoter was strongly diminished in an spt3 strain, clearly indicating a role for Spt3 in the recruitment of TBP. The much lower occupancy of TBP at the PHO5 promoter in the absence of Spt3 can therefore account for the strong decrease in promoter strength seen in the spt3 strain.

FIG. 5.

Spt3 strongly increases TBP recruitment to the activated PHO5 promoter. Strains 8136 (wild type) and 8136Δspt3 were phosphate starved for up to 6 h. TBP occupancy was determined by chromatin immunoprecipitation analysis with an anti-TBP antibody. The amounts of coimmunoprecipitated DNA determined by quantitative PCR were normalized to the respective input DNA and are shown in arbitrary units. The locations of the PCR fragments used for chromatin immunoprecipitation analysis of the PHO5 locus are shown with respect to the nucleosomal structure of the repressed PHO5 promoter (3) (see the legend to Fig. 1 for details). As a control, the PGK1 promoter was included in the analysis. Error bars show SDs.

The Ada2 SANT domain affects the rate of chromatin remodeling at the PHO5 promoter.

In the multiprotein SAGA complex, Gcn5 forms a trimeric subcomplex with Ada2 and Ada3, raising the possibility that these two Ada subunits have some Gcn5-related function. It was previously reported that Ada2 was required for the assembly of Gcn5 into the SAGA complex (21), but more recent experiments showed that Ada2 plays a more active role in controlling the activity of Gcn5 and the ability of Gcn5 to recognize its nucleosomal substrate in vitro and in vivo (4, 14, 21, 41, 45) and that this function is linked to its SANT domain (12, 51).

To gain further insight into the functional importance of the Ada2 subunit and particularly of the Ada2 SANT domain in transcriptional regulation in vivo, we examined the role of Ada2 in the activation of the PHO5 promoter. To do so, we examined the kinetics of PHO5 induction by monitoring the rates of acid phosphatase production in an ada2 strain, an ada2ΔSANT strain, a wild-type strain, and a gcn5 strain. As shown in Fig. 6, the accumulation of acid phosphatase activity early in induction was strongly retarded in the ada2 strain compared to the wild-type strain, and the delay was almost as long as in the gcn5 strain. Importantly, there was no significant difference between the ada2 and ada2ΔSANT strains. Since it was shown that the Ada2 SANT deletion mutant used in this study (ada2Δ97- 106) permits the full assembly of a native SAGA complex including Gcn5 (12), the kinetic defect in transcriptional activation observed in this strain cannot be an indirect effect due to the loss of Gcn5 from the complex. Therefore, it can be concluded that the Ada2 subunit plays a direct and active role rather than being required only for Gcn5 assembly into the SAGA complex and that an intact SANT domain is required for this Ada2 function. Furthermore, the data presented in Fig. 6 show that the absence of this Ada2 function affects the kinetics of transcriptional activation to a degree similar to that seen in the absence of Gcn5 and that Ada2, unlike Spt3, must play a Gcn5-related role in the activation of the PHO5 promoter.

FIG. 6.

Deletion of the Ada2 subunit or its SANT domain causes a strong delay in PHO5 promoter activation. The kinetics of PHO5 induction in wild-type (wt), ada2, ada2Δ97-106 (ada2ΔSANT), and gcn5 strains were monitored by measuring acid phosphate activity at the indicated times after transfer of cells from phosphate-containing medium to phosphate-free medium.

Since we have demonstrated that the delay in PHO5 activation in a gcn5 strain occurs at the level of chromatin remodeling (7), the rate of chromatin opening at the PHO5 promoter, as measured by ClaI accessibility (see above), was also examined for an ada2ΔSANT strain, a wild-type strain, and an ada2 strain. As expected from the acid phosphatase activity measurements, ClaI accessibility in the ada2 and ada2ΔSANT strains was significantly lower than that in the wild-type t strain (Fig. 7). Analysis of the blots showed that accessibility in the wild-type strain was about 90%, much more than the 50 and 55% found for the ada2 and ada2ΔSANT strains, respectively (Fig. 7). Chromatin analysis therefore shows that Ada2 is involved in a chromatin-related SAGA activity at the PHO5 promoter, presumably by controlling Gcn5 activity. Moreover, the finding that the effects of deletion of ADA2 and deletion of the Ada2 SANT domain are almost the same shows the central importance of the SANT domain for Ada2 function.

FIG. 7.

The rate of chromatin remodeling at the PHO5 promoter is strongly decreased in ada2 and ada2Δ97-106 strains. Nuclei isolated from ADA2 (wild-type [wt]), ada2, and ada2Δ97-106 (ada2ΔSANT) strains after 4 h of induction in phosphate-free medium were analyzed for chromatin remodeling by the ClaI accessibility assay as described in the legend to Fig. 4. ClaI concentrations in lanes 1 to 6 correspond to those used in lanes 1 to 6 of Fig. 4.

The Gcn5 bromodomain contributes only modestly to SAGA function at the PHO5 promoter.

Another conserved motif found in most nuclear HATs, including Gcn5, as well as in other chromatin-modifying complexes is the bromodomain (28). However, deletion or mutation of the Gcn5 bromodomain had only a small effect on yeast SAGA function (14, 41, 50). We reasoned that the PHO5 promoter, which reacts in a highly specific way to the absence of Gcn5, may be an informative system for addressing the function of the Gcn5 bromodomain. We therefore examined the rate of PHO5 induction in strains containing Gcn5 with a deleted or mutated bromodomain. In both situations, a significant delay in the rate of promoter activation was observed, similar to but much less pronounced than the delay observed in the complete absence of Gcn5 (Fig. 8). Apparently, the Gcn5 bromodomain contributes to SAGA function but is not critical for Gcn5 HAT activity at the PHO5 promoter.

FIG. 8.

The Gcn5 bromodomain contributes to Gcn5 function at the PHO5 promoter. The kinetics of PHO5 induction in a wild-type (wt) strain, a gcn5 strain, and strains in which the Gcn5 bromodomain had been either deleted (gcn5Δbromo) or mutated (gcn5bromo-mut) (as described in Materials and Methods) were monitored by measuring acid phosphatase activity at the indicated times after transfer of cells from phosphate-containing medium to phosphate-free medium.

DISCUSSION

SAGA was originally identified and characterized as a protein complex providing Gcn5 HAT activity and, for some time, the entire complex was often simply referred to as a chromatin-modifying machine, although genetic and biochemical evidence had already suggested that SAGA provided more than just histone acetylation in the process of transcriptional activation. Indeed, recent studies with GAL gene promoters demonstrated very clearly that SAGA subunit Spt3 has a distinct regulatory function not related to Gcn5 activity (11, 37). This finding confirmed that SAGA possesses non-chromatin-related functions but also showed that the activation of some genes, although not significantly Gcn5 dependent, is strongly SAGA dependent, consistent with previous genome-wide expression analyses of the roles of different SAGA subunits (38). It was therefore important to investigate whether additional, distinct SAGA functions could also be demonstrated for other genes. The PHO5 promoter is a highly suitable system for such studies, since it is unusual in its response to the presence or the absence of the Gcn5 subunit of SAGA. Our main original finding that the absence of Gcn5 delayed chromatin remodeling during PHO5 activation but did not compromise its final level made it possible for us to search for additional functions of other SAGA subunits. Indeed, the results presented in this report clearly show that SAGA contributes to PHO5 activation through the cumulative actions of multiple subunits with mechanistically distinct functions.

Recruitment of SAGA to the PHO5 promoter.

Regulation of the PHO5 promoter in response to the phosphate concentration is achieved through a signal transduction pathway that impinges on the specific activator Pho4 (35). Under condition of phosphate starvation, Pho4 is able to bind the promoter and trigger the activation process. As a critical prerequisite for activation, the chromatin structure at the promoter undergoes a dramatic alteration (53). It was recently shown in our laboratory that efficient remodeling of the promoter chromatin structure requires Gcn5 HAT activity (7), indicating either global activity of SAGA or recruitment of the SAGA complex to the promoter. We show in this study that SAGA is indeed specifically recruited to the PHO5 promoter region but only under inducing conditions and in a Pho4-dependent manner (Fig. 1).

Similar to our results with Pho4, other acidic activators, such as Gal4, Gcn4, and VP16, have been shown to target SAGA to promoters (11, 33, 37, 56), suggesting that this ability is a shared property of acidic activators. Which SAGA subunit(s) is required for in vivo recruitment, on the other hand, is not clear. Earlier in vitro experiments had indicated interactions between the VP16 activation domain and SAGA subunits Ada2 (8, 49) and Spt20 (40). Recently, the Tral subunit was demonstrated to interact with several activators in vitro (13). The results presented in this report show that SAGA recruitment to the PHO5 promoter is not critically dependent on either Gcn5, Ada2, or Spt3, since the absence of any of these subunits results in only a moderate defect in PHO5 induction; in contrast, a much stronger phenotype is seen in spt7 and spt20 strains, in which the assembly of the entire SAGA complex is deficient (21, 50). The fact that SAGA is recruited to the PHO5 promoter through an interaction with Pho4 implies that the binding of Pho4 to the promoter precedes SAGA recruitment and consequent chromatin remodeling, which means that Pho4 must be able to bind to the nonremodeled promoter. For the PHO8 promoter, which is also activated by Pho4 in response to phosphate starvation, it has been directly shown that Pho4 binds to the promoter in the absence of either Gcn5 or Snf2, although both proteins are required for chromatin remodeling and activation of PHO8 (25). The critical Pho4 binding site at the PHO8 promoter is located in a hypersensitive, nucleosome-free region (5), thereby permitting the binding of Pho4 to the promoter prior to remodeling. In contrast to the situation for PHO8, chromatin remodeling and consequent activation of the PHO5 promoter require the cooperative binding of Pho4 to two sites, one being nonnucleosomal (UASp1) and the other being located within nucleosome −2 (UASp2). However, the binding of Pho4 to a nucleosomal site requires prior chromatin remodeling (53). This conundrum can be resolved by assuming that, at the PHO5 promoter, the binding of Pho4 to the nonnucleosomal site brings about SAGA recruitment, which then triggers the local remodeling of chromatin and allows the subsequent binding of Pho4 to the nucleosomal site, thereby promoting the remodeling process along the promoter.

The combined activities of Gcn5 and Spt3 are critically required for the transcriptional activation of PHO5.

SAGA contains three subunits, Spt7, Spt20, and Ada1, so-called core subunits, which are required for the integrity of the complex and therefore for all SAGA functions but which are not known to play any direct regulatory role (21, 47, 50). Deletion of any of the core subunits should therefore eliminate all SAGA functions and provide a worst-case scenario. By examining the kinetics of PHO5 activation in spt7 and spt20 strains in comparison to wild-type and gcn5 strains, we could indeed show that SAGA provides functions other than Gcn5 HAT activity for the transcriptional activation of PHO5 (Fig. 2). Obviously, because of the assembly defect, these additional SAGA functions cannot be directly ascribed to Spt7 and Spt20. An attempt to examine a possible direct function of Spt7 was made by mutating its bromodomain, but no significant phenotype was detected (20, 50). Quite recently, the results of an extensive analysis of Spt7 functions were published, and the role of Spt7 in SAGA assembly was further refined (58).

For two SAGA subunits, Spt3 and Spt8, a more specific role in transcriptional activation was previously demonstrated. They appear to control interactions between the TBP and the TATA box (10, 18, 39), and Spt3 was recently reported to be critically required for TBP recruitment and the subsequent activation of GAL gene promoters (11, 37). We have shown here that Spt3 is also involved in the transcriptional activation of PHO5 (Fig. 3A), although it is not critically required, as in the GAL gene promoters. The absence of Spt3 did not bring about a delay in PHO5 activation but reduced its final level, suggesting that the Spt3 function is quite distinct from that of Gcn5. Indeed, we have shown that Spt3 has no effect on chromatin remodeling at the PHO5 promoter (Fig. 4) but strongly affects the binding of TBP to the promoter (Fig. 5). These results are consistent with the previous finding that the absence of the TATA box has no effect on chromatin opening at the PHO5 promoter (19).

That the functions of Gcn5 and Spt3 are distinct and additive was further confirmed by examining a strain lacking both Gcn5 and Spt3. Such a double mutant showed a synthetic phenotype, and PHO5 activation was almost as strongly affected as in an spt7 or spt20 strain. These results, taken together, show that at the PHO5 promoter, the combined actions of two mechanistically distinct SAGA functions are critically required for efficient activation, and it is conceivable that no other major function of SAGA beyond Gcn5 and Spt3 is needed for PHO5.

The Gcn5 function in SAGA requires the Ada2 SANT domain and the Gcn5 bromodomain.

It was previously shown that the HAT activity of Gcn5 requires its association with Ada2 and Ada3 (4, 14, 21, 41, 45), but the mechanism by which these subunits contribute to the HAT activity of Gcn5 in vivo has remained unclear. Ada2 contains a SANT domain, a motif that is the hallmark of a number of proteins involved in transcriptional regulation (1), and recent reports have provided evidence for an involvement of the Ada2 SANT domain in substrate recognition and enzymatic catalysis by Gcn5 (12, 51). The functional importance of the Ada2 SANT domain in the transcriptional activation of a particular gene in vivo had been demonstrated up to now only for an HO-lacZ fusion gene, the expression of which was abolished by a SANT domain deletion (12). We show in this report that the activation of a native promoter under physiological conditions is strongly affected by a small deletion within the SANT domain of Ada2. The defect is very similar to the defect seen in a GCN5 deletion strain, although it is less pronounced. Moreover, we found that an intact SANT domain in Ada2 is required for chromatin remodeling at the PHO5 promoter, strongly suggesting that, unlike the function of Spt3, the function of Ada2 is related to Gcn5. The defect observed in an ADA2 deletion strain could be due, at least to some extent, to the loss of Gcn5 from the SAGA complex (21, 51). However, the SANT domain deletion used in this study has been shown not to alter the subunit composition of the SAGA complex (12). Our results are therefore perfectly consistent with the suggested role of the Ada2 SANT domain in controlling Gcn5 catalytic activity toward its nucleosomal substrate (12, 51).

Two subunits in SAGA, Spt7 and Gcn5, contain a bromodomain, another conserved motif found in most HAT activities and other activities dealing with chromatin (28). Bromodomains have been shown to serve as acetyllysine binding modules (15, 32, 34, 43). It was reported that the function of Spt7 did not require its bromodomain (20, 50) but that a deletion of the Gcn5 bromodomain had slight to moderate effects on transcription (14, 41, 50). Transcriptional activation by weak activators was significantly diminished in a Gcn5 bromodomain deletion variant, while the effect was smaller with strong activators (14). In addition, the HIS3 promoter was only weakly affected by a Gcn5 bromodomain deletion (50). In all of these situations, a deletion of the Gcn5 HAT domain had a more severe effect than a deletion of the bromodomain.

We show here that the rate of PHO5 induction is significantly delayed in strains with a deleted or mutated Gcn5 bromodomain. The effect is similar to what we have observed in the total absence of Gcn5, but it is much less pronounced (Fig. 8). The bromodomain therefore contributes to but is not essential for Gcn5 function at the PHO5 promoter. In light of the finding that bromodomains bind to acetyllysine residues, it was suggested that these interactions could enhance the binding of the HAT complex to acetylated chromatin, thereby facilitating, in turn, local acetylation by the HAT domain (57). A recent study by Hassan et al. (27) demonstrated a function of the Gcn5 bromodomain in anchoring SAGA on acetylated promoter nucleosomes. Our finding that the bromodomain significantly enhances Gcn5 function at the PHO5 promoter would fit this concept quite well. However, in our scenario the bromodomain appears only to fine-tune Gcn5 function and not to make an essential contribution to the recruitment of the entire SAGA complex, since its absence does not generate anything like an spt3 or spt7 phenotype.

Whole genomic mRNA analyses have shown that there are genes whose expression does not require Gcn5 function, yet they depend on other SAGA functions (38). As recently shown, an example of this group of genes is GAL1, which critically requires Spt3 yet is mostly Gcn5 independent (11, 37). PHO5, on the other hand, is the first gene for which both Gcn5 and Spt3 functions of SAGA have now been shown to be important for transcriptional regulation. The Gcn5 activity at this promoter is itself regulated in a complex way by other SAGA subunits and motifs. As a consequence, the idea that genes either depend or do not depend on SAGA has to give way to the concept that individual genes in the SAGA-requiring class depend on different SAGA functions in individual ways for their activation. The PHO5 promoter has proved to be unusually informative in dissecting these individual contributions of SAGA.