|Home | About | Journals | Submit | Contact Us | Français|
Transcriptional activators often require histone acetyltransferases (HATs) for full activity. The common explanation is that activators directly recruit HATs to gene promoters to locally hyperacetylate histones and thereby facilitate transcription complex formation. However, in addition to being targeted to specific loci, HATs such as Gcn5 also modify histones genome-wide. Here we provide evidence for a role of this global HAT activity in regulated transcription. We show that activation by direct recruitment of the transcriptional machinery neither recruits Gcn5 nor induces changes in histone acetylation yet can strongly depend on Gcn5 at promoters showing a high basal state of Gcn5-mediated histone acetylation. We also show that Gcn5 dependency varies among core promoters and is influenced by the strength of interaction used to recruit the machinery and by the affinity of the latter for the core promoter. These data support a role for global Gcn5 HAT activity in modulating transcription independently of its known coactivator function.
Activation of transcription in eukaryotes often requires modification of the chromatin template. Among the well-characterized chromatin-modifying enzymes, histone acetyltransferases (HATs) are known to regulate gene expression by adding acetyl groups to lysine residues within the N-terminal tails of histones (31). In the yeast Saccharomyces cerevisiae, one of the best known HATs is Gcn5, which is the catalytic subunit of the SAGA complex that acetylates primarily histones H3 and H2B (34, 39). Work in recent years has provided overwhelming evidence that this HAT-containing complex functions as a coactivator and directly participates in the transcription process by being specifically recruited by activators to the promoter of many genes (31). For instance, activators such as Gal4 and Gcn4 interact directly with the SAGA complex in vitro (4, 8, 16) and in vivo (2), and they recruit SAGA at target gene promoters in vivo (19, 22), an event that can occur independently of transcription (1, 19, 36) and that precedes and is required for the recruitment of the RNA polymerase II (Pol II) machinery at certain promoters (1, 5). Once recruited, the Gcn5 HAT subunit of SAGA locally acetylates histones, which is thought to facilitate transcription by loosening the chromatin structure or by generating specific binding sites for the recruitment of transcription factors (31). SAGA also serves an adaptor function to recruit the TATA-binding protein (TBP) through its Spt3 subunit (1, 22). Recent genome-wide location studies suggest that Gcn5-containing complexes may actually be recruited by regulatory proteins to the upstream activating sequence (UAS) of most active genes (29).
In addition to its targeted coactivator function, Gcn5 also acetylates histones genome-wide, a phenomenon affecting most nucleosomes in yeast (38) and referred to as global acetylation (reviewed in reference 21). This global activity results in a basal state of histone acetylation throughout the genome that varies among loci and over which targeted acetylation superimposes (15, 30, 37). While the importance of activator-targeted histone acetylation in transcriptional regulation has been extensively documented, the role of global acetylation remains largely elusive. A major complication in addressing this question stems from the ability of Gcn5 and other HATs to function in both a global and a targeted manner, thereby making it difficult to distinguish between the two activities. To circumvent this problem, we took advantage of the previous observation that artificial recruitment of the basal RNA Pol II machinery to a promoter suffices to stimulate transcription in the absence of any activator that could recruit chromatin-modifying complexes (for a review, see reference 27). We asked whether activation under these conditions from promoters showing a high basal state of Gcn5-mediated histone acetylation would depend on this specific HAT. Our results show that global acetylation by Gcn5 can indeed strongly affect activation and that it does so depending on the strength of interaction used to recruit the machinery and on the affinity of the latter for the core promoter. We discuss the general implications for current models concerning the role of histone acetylation in transcriptional regulation.
All proteins are encoded by single-copy plasmids marked with the URA3, TRP1, or LEU2 gene. The plasmids expressing wild-type TFIIB, TFIIB-Myc, and TFIIB-RFX from the native TFIIB promoter, and Max-RFX and VP16-RFX from the TBP promoter, have been described (13). As seen in Fig. Fig.4C,4C, TFIIB-Myc and Max-RFX are expressed at higher levels in the gcn5Δ strain from the DED1 promoter. The episomal RFX-dependent xPHO5 reporter constructs bearing a native or mutated (CGTA) TATA sequence are based on previously described plasmids (35). They contain a lacZ reporter gene under control of the PHO5 promoter region extending from nucleotides −392 to +3. A double-stranded synthetic oligonucleotide bearing a single RFX binding site was inserted into the BstEII site in the linker region between nucleosomes −1 and −2. The RFX-dependent xHIS3-lacZ allele has been described (13). xSNO3 and xCYC1 were generated from xHIS3-lacZ by replacing the HIS3 promoter region, including the upstream dA-dT element, with PCR fragments containing SNO3 (−212 to +17) and CYC1 (−244 to +5) promoter sequences downstream of a unique RFX binding site. Details of the plasmid constructions are available upon request.
The parental yeast strain is derived from KY320 (13) and contains a chromosomal TFIIB allele placed under control of the GAL1 promoter. As a consequence, cell growth on glucose requires expression of a plasmid-encoded TFIIB protein. Isogenic gcn5Δ and spt3Δ strains were obtained by homologous recombination using a kanamycin cassette as described previously (35). GCN5 was chromosomally tagged by adding a 3× hemagglutinin (HA) sequence to the C terminus of the coding region according to the method of Knop et al. (17). The xHIS3, xSNO3, and xCYC1 alleles were integrated at the HIS3 locus by standard procedures. Transformed yeast cells were selected on synthetic complete dropout (xPHO5) or Casamino Acids (xHIS3, xSNO3, and xCYC1) plates supplemented with galactose. Strains were then grown for 24 to 48 h in the same selective media but containing glucose to repress the chromosomal TFIIB gene. Because the xPHO5 construct retains the two upstream regulatory sequences present in the natural PHO5 promoter, all experiments with xPHO5 were performed by growing the cells in high-phosphate-containing medium to prevent nuclear translocation of Pho4, the major activator of the PHO5 gene, and its interaction with the pleiotropic transcription factor Pho2 (18). β-Galactosidase activity was assayed as previously described (13).
Western blot analysis using whole extracts from cells grown in glucose selective medium was performed as described previously (13). The membranes were probed with 1:5,000 anti-HA monoclonal antibody (clone 16B12; Covance) or 1:1,000 polyclonal anti-RFX1 antibodies (gift from Walter Reith, University of Geneva Medical School).
A detailed protocol for chromatin immunoprecipitation (ChIP) and quantitative PCR analysis can be found at http://www.mimo.unige.ch/STRUBIN_LABb.htm. Whole-cell extracts equivalent to about 2× 108 yeast cells were immunoprecipitated (IP) with antibodies to acetylated histone H3-K9 (7 μl) (Upstate Biotech), tetra-acetylated histone H4 (2.5 μl) (06-866; Upstate Biotech), core histone H3 (1 μl) (ab1791; Abcam), and anti-HA (2.5 μl). The recovered DNA and at least two standard dilutions of the input DNA were quantified in duplicate by real-time PCR using the SYBR green core kit (Eurogentec) and the ABI PRISM 7700 Sequence Detection System. Sequences of the oligonucleotide primers are available upon request. The relative IP value for a given locus is expressed in arbitrary units and was calculated as the ratio between the IP signal and the respective input DNA signal to correct for variation between different samples and primer pairs. Where indicated, the relative IP value of the tested locus was divided by that of the subtelomeric (TEL) region.
To assess the effect of global histone acetylation on transcriptional activation, we asked whether activation by recruitment of the basal machinery from a promoter showing a high basal state of histone acetylation would depend on a specific HAT. For this purpose, we selected the PHO5 promoter because of its reported high basal levels of histone acetylation (37). We introduced a single binding site for RFX, a transcriptionally inactive DNA-binding protein that we used in previous artificial recruitment experiments (13), in the linker region between nucleosomes −2 and −1 of the PHO5 promoter. The resulting xPHO5 synthetic promoter was used to direct the expression of an episomal lacZ reporter gene (Fig. (Fig.1A).1A). We then tested whether transcription from the xPHO5 promoter can be activated by recruitment of the general transcription factor TFIIB. Figure Figure11 shows that TFIIB tethered to the xPHO5 promoter by direct fusion to RFX efficiently stimulates transcription to levels comparable to those observed with an RFX variant bearing the potent VP16 activation domain (Fig. 1B and C). Activation also occurs, although at this promoter to a lower extent (see below), when TFIIB is recruited by RFX through a noncovalent and thus weaker interaction between the complementary dimerization domains of the c-Myc oncoprotein and its partner Max (Fig. 1B and C). Activation by TFIIB-RFX and TFIIB-Myc is significantly affected in cells expressing endogenous TFIIB (data not shown) and is reduced to nearly basal levels upon higher expression of wild-type TFIIB (Fig. (Fig.1C;1C; compare lane 5 to lane 6 and lane 7 to lane 8), whereas activation by VP16-RFX remains unchanged (Fig. (Fig.1C;1C; compare lane 3 to lane 4). This is consistent with the recruited TFIIB protein having to assemble with other basal factors and hence functioning as a component of the general transcription machinery, as reported previously (13). We conclude that recruitment of the basal transcription machinery leads to transcriptional activation from the xPHO5 promoter.
We next examined the basal acetylation state of histones H3 and H4 on the xPHO5 promoter by ChIP. Figure Figure2A2A reveals that both nucleosomes within the uninduced xPHO5 promoter show levels of histone H3 acetylation at lysine 9 (H3-K9) (Fig. (Fig.2A,2A, left panel) and histone H4 acetylation (right panel) that are similar to those observed at the transcriptionally active ACT1 promoter and much higher than the levels of acetylation detected within a repressed subtelomeric (TEL) region. Since histone H3-K9 is a primary target of Gcn5 HAT activity in vivo (34), we investigated whether basal H3 acetylation on xPHO5 is dependent on Gcn5. Figure Figure2A2A shows that this is indeed the case, as acetylation of histone H3-K9, but not of histone H4, is markedly reduced at the xPHO5 promoter in a gcn5Δ strain, whereas a roughly twofold decrease is observed at ACT1, consistent with previous studies (19). Taken together, these results indicate that under nonactivating conditions the xPHO5 promoter shows a high level of histone H3 acetylation that is largely dependent on Gcn5 and that recruitment of the basal transcription machinery to this promoter suffices to stimulate transcription, thus making the xPHO5 promoter construct suitable for further analysis.
Before assessing the role of global Gcn5 HAT activity in xPHO5 transcription, it was important to exclude the possibility that tethering TFIIB would result in the recruitment of HATs such as the Gcn5-containing SAGA complex to the promoter. Since the targeting of HATs to promoters can lead to an increase in histone acetylation, we first examined whether activation of xPHO5 by recruitment of TFIIB causes any changes in the acetylation state of histone H3-K9, which is acetylated by Gcn5 in vivo, and of histone H4, which is acetylated by a HAT other than Gcn5 (34) and therefore serves as a control. Since activation of the natural PHO5 promoter leads to a loss of promoter histones (3, 28), we also monitored the histone H3 content of the xPHO5 promoter using antibodies directed against core histone H3. As shown in Fig. Fig.2B,2B, there is little change in the levels of histone H3-K9 acetylation (Fig. (Fig.2B,2B, left panel) and histone H4 acetylation (Fig. (Fig.2B,2B, right panel) in the presence of Max-RFX alone or when transcription is stimulated by TFIIB-Myc and Max-RFX. By contrast, a modest but consistent decrease in histone H3-K9 and H4 acetylation is seen when xPHO5 transcription is activated by TFIIB-RFX (Fig. (Fig.2B).2B). This, however, is most likely due to a loss of histones as reported for the wild-type PHO5 promoter (3, 28), as a similar decrease is observed in the histone H3 content (Fig. (Fig.2C2C).
Recent studies demonstrate that the targeting of HAT-containing complexes at certain promoters results in increased histone acetylation only when transcription is compromised (35). We therefore repeated these experiments using a xPHO5 promoter construct bearing a deleterious mutation in the TATA box. As expected, transcription from the TATA mutant promoter was strongly attenuated; however, no increase in histone H3 and H4 acetylation was detected in the presence of TFIIB-Myc and Max-RFX or of TFIIB-RFX (data not shown). This contrasts with the enhanced histone H3 acetylation levels observed at a similar PHO5 TATA mutant promoter construct in response to the Gcn4 activator protein, which is known to recruit the Gcn5-containing SAGA complex (35). We thus conclude that preinitiation complex (PIC) formation leads to loss of histone-DNA contacts, most likely reflecting nucleosome depletion, but does not alter significantly the histone acetylation state at the xPHO5 promoter.
To directly test whether recruitment of TFIIB leads to recruitment of Gcn5 at the xPHO5 promoter, we performed chromatin immunoprecipitation of Gcn5. To that aim, the chromosomal GCN5 locus was replaced by an allele encoding a triple HA-tagged version of Gcn5, which fully substitutes for the wild-type protein (data not shown). Since the Gal4 activator is known to recruit SAGA to the galactose-inducible GAL genes (1, 5), we assessed Gcn5-HA occupancy at the GAL1 promoter as a positive control. Consistent with previous studies (1), background levels (compared to TEL) of Gcn5-HA are detected within the GAL1 promoter region when cells are grown in repressive medium containing glucose as a carbon source, and a strong increase in Gcn5-HA occupancy at the GAL1 upstream activating sequence, but not at the GAL1 core promoter, is observed upon galactose induction (Fig. (Fig.2D).2D). In contrast, no increase above background in Gcn5-HA occupancy is seen at the xPHO5 promoter upon activation by either TFIIB-RFX or the TFIIB-Myc and Max-RFX combination (Fig. (Fig.2D).2D). Taken together, these results indicate that PIC assembly at the xPHO5 promoter does not lead to changes in histone acetylation, and that Gcn5 (and presumably other HATs) does not reach the promoter in association with the RNA Pol II transcription machinery.
The above results suggest that any role of Gcn5 in transcription activation under these conditions will depend on its global, rather than targeted, HAT activity. We therefore compared the efficiencies of activation by recruitment of TFIIB in a wild-type and a gcn5Δ strain. Figure Figure3A3A shows that deletion of GCN5 has no effect on activation mediated by TFIIB-RFX (left panel). Strikingly, however, activation by TFIIB-Myc and Max-RFX is markedly reduced in the absence of Gcn5 (Fig. (Fig.3A,3A, left panel). By contrast, no decrease is observed in cells lacking Spt3, a SAGA subunit involved in stabilizing TBP binding to the TATA box (1, 22), consistent with the adaptor function of SAGA not being implicated (Fig. (Fig.3B).3B). The inability of TFIIB-Myc to respond normally to Max-RFX in cells lacking Gcn5 is not due to TFIIB-Myc being unstable or nonfunctional. Indeed, TFIIB-Myc is as efficient as wild-type TFIIB in responding to VP16-RFX in gcn5Δ cells (Fig. (Fig.3C),3C), and expression of Max-RFX and TFIIB-Myc at higher levels than those necessary in wild-type cells to achieve maximal activation does not bypass the need for Gcn5 (see below).
The difference in Gcn5 dependency between TFIIB-RFX and the TFIIB-Myc and Max-RFX combination suggests that the strength of interaction between TFIIB and RFX, and hence the affinity of TFIIB and the transcription machinery for the promoter, determines to what extent activation will depend on global Gcn5 HAT activity. According to this hypothesis, weakening the xPHO5 promoter should render activation by IIB-RFX dependent on Gcn5. The right panel in Fig. Fig.3A3A reveals that this is indeed the case: alteration of the TATA box in the xPHO5 promoter, although leading to a decrease in the absolute levels of transcription (data not shown), increases, if anything, the activation by TFIIB-RFX and yet renders activation strongly Gcn5 dependent. Thus, decreasing the affinity of the basal machinery for the promoter by either weakening the interaction between TFIIB and RFX or mutating the core promoter renders activation dependent on the global acetylation function of Gcn5.
To extend these findings beyond the episomal xPHO5 construct, and to investigate the role of the core promoter in Gcn5 dependency, we constructed additional RFX-dependent lacZ alleles that differ only in the identities of the core promoters. As core promoters, we chose HIS3, which contains an upstream dA-dT element and a single TATA box, SNO3, which is a TATA-less promoter, and CYC1, which contains two TATA boxes (see Fig. Fig.4B).4B). A single RFX binding site was introduced upstream of each of these core promoters, and the resulting constructs were integrated in the chromosomal HIS3 locus of isogenic wild-type and gcn5Δ strains. Examination of the histone H3 acetylation state at the promoter of one such allele, xCYC1, revealed a high level of acetylation under noninducing conditions that is largely Gcn5 dependent (Fig. (Fig.4A),4A), very similar to what was observed at the natural HIS3 locus under repressive conditions (19). We then tested for activation by recruitment of TFIIB at these promoters. Figure Figure4B4B shows that activation by TFIIB-RFX at the xHIS3 and xSNO3 promoters occurs independently of Gcn5, similar to what we found for the episomal xPHO5 promoter construct. At the xCYC1 promoter, however, TFIIB-RFX exhibits partial dependency on Gcn5. Activation by TFIIB-Myc and Max-RFX, despite reaching transcription levels similar to those observed with IIB-RFX, shows Gcn5 dependency at all three promoters and, strikingly, is strictly dependent on Gcn5 at the xCYC1 promoter. The absolute requirement for Gcn5 in TFIIB-Myc- and Max-RFX-mediated activation at the xCYC1 promoter prompted us to test whether increased expression of TFIIB-Myc and Max-RFX would facilitate PIC formation and thereby overcome the need for Gcn5. This is not the case, however, as activation remains inefficient in a gcn5Δ background even when TFIIB-Myc and Max-RFX are made in larger amounts than those needed in wild-type cells to reach maximal activation (Fig. (Fig.4C).4C). Altogether, these results indicate that activation by recruitment of the basal RNA Pol II machinery can strongly depend on the global acetylation function of Gcn5 and that it does so depending on the strength of interaction with the target in the transcription complex and on the nature of the core promoter sequences.
In the present study, we made use of an activator bypass approach to assess the role of Gcn5-mediated global histone acetylation in transcriptional regulation independently of its known coactivator function. We show that activation by artificial recruitment of the basal transcription machinery can strongly depend on Gcn5, which is consistent with previous studies (7, 33). Yet Gcn5 is not recruited to the promoter, and no changes in histone acetylation are detected upon activation. This indicates that Gcn5 (and presumably other HATs) does not associate with the basal machinery and thus points to an involvement of the global acetylation function of Gcn5 in the activation event. We also show that Gcn5 dependency varies among core promoters and is dictated, at least in part, by the affinity of the basal machinery for the core promoter. This suggests that global acetylation affects the activation process by facilitating PIC formation, either by increasing the affinity of one or more general transcription factors for the core promoter or by creating a more accessible chromatin environment. Surprisingly, the strongest requirement for the global HAT activity of Gcn5 was observed at the CYC1 promoter, which is occupied by TBP prior to activation (20, 25). This raises the possibility that histone acetylation may facilitate not only binding of TBP to the promoter, as previously demonstrated (7, 26, 32), but also association of the remainder of the basal machinery that reaches the promoter independently of TBP (13).
Transcriptional activation from the natural PHO5 promoter coincides with major changes in the chromatin structure, in the course of which nucleosomes are lost from the promoter (3, 28). Chromatin disassembly results from the binding of Pho4, the major activator of the PHO5 gene, as it occurs normally even at a TATA-deleted PHO5 promoter where the PIC is not assembled (9). Recent work reveals that Pho4-mediated nucleosome loss is preceded by a transient increase in histone H3 but not histone H4 acetylation (28), suggesting that targeted histone H3 acetylation by Pho4 is required for nucleosome eviction. Interestingly, we find that recruitment of the basal machinery by contact with TFIIB causes a similar loss of nucleosomes from a synthetic PHO5 promoter. This is consistent with the previous finding that targeting the RNA Pol II machinery through interaction with the Gal11 subunit of Mediator is sufficient for chromatin remodeling at the wild-type PHO5 promoter (11). Yet tethering TFIIB to the promoter neither recruits Gcn5 nor affects the acetylation state of histones H3 and H4, whereas in the Gal11 experiments chromatin remodeling was also observed in a strain lacking a functional Swi/Snf complex (11). This suggests that under certain conditions nucleosome depletion can also occur in the absence of targeted histone acetylation and chromatin remodeling activities and hence may be a direct consequence of transcription complex formation at the promoter. Perhaps the basal machinery competes with the histones in a dynamic fashion for access to the DNA. The recent finding that nucleosome loss occurs at many active promoters and is proportional to the transcriptional initiation rate is consistent with such a possibility (23).
There is ample evidence that in addition to recruiting chromatin-modifying activities, activators also function by directly facilitating recruitment of the transcriptional machinery to promoters (24). Our results therefore predict that global acetylation by Gcn5 (and perhaps Esa1; see below) must also affect gene expression under normal conditions. Consistent with this possibility, genome-wide ChIP studies revealed that many inducible genes are marked by histone acetylation even in their inactive state, presumably through the action of globally acting enzymes (30). That this basal state of acetylation can affect activator function is supported by the finding that targeted recruitment of the Rpd3 histone deacetylase complex leads to localized histone deacetylation and concomitant inhibition of a variety of activators, including those for which no evidence exists that they recruit HAT-containing complexes (7). This, together with the finding that at many promoters, including known HAT-dependent promoters, transcription activation is not associated with increased histone acetylation (6), is consistent with the notion that global acetylation may contribute importantly to regulated gene expression. An intriguing question, therefore, is which of the targeted or global HAT functions is most required at promoters showing a high basal state of histone acetylation and at which activator-mediated recruitment of HAT complexes is known to occur.
Although we did not specifically address this issue, it is very likely that global histone H4 acetylation mediated by Esa1 also plays a role in gene activation. Indeed, high basal levels of H4 acetylation are detected at many inducible promoters (30), and Esa1 is required for normal expression of genes showing constitutive association of Esa1 (12), or at which no changes in histone H4 acetylation are observed during induction (10, 28).
Interestingly, in the Rpd3 targeting experiments, weak activators tended to be more strongly affected by Rpd3-mediated histone deacetylation (7). Similarly, deleting one of the two Pho4 binding sites in the PHO5 promoter (14, 37), or weakening the strength of the Gal4 activation domain (33), renders activation by Pho4 and Gal4 strongly dependent on Gcn5. All these findings are strikingly similar to our observation that recruitment of TFIIB through a noncovalent interaction is more dependent on global Gcn5 HAT activity than when recruitment is achieved by direct fusion of TFIIB to a heterologous DNA-binding protein. We therefore propose that the strength of interaction between the activator and its target in the transcription complex dictates, at least in part, the extent to which activation will depend on the global acetylation function of Gcn5, and we suggest that the same is true for targeted acetylation. Since the basal acetylation state is likely to differ between loci, such a global role for Gcn5 HAT activity might indirectly participate in the specificity of activator function by differentially modulating activity of the same or different activators depending on the gene's chromosomal location and/or promoter architecture. Global acetylation may therefore add an additional level of complexity to the mechanisms that contribute to the enormous diversity of gene expression.
We are most grateful to Patrick M. Schaeffer for setting up the ChIP assay. We thank Tae-Young Roh and Keji Zhao for sharing unpublished results; Walter Reith for anti-RFX1 antibodies, helpful discussions and critical reading of the manuscript; and George Thireos for discussion and support.
This work was supported by grants from the Swiss National Science Foundation (3100A0-100785) and from the EU (HPRN-CT-2000-00087, supported by OFES no. 99.0754) to M.S.