GCN5 Dependence of Chromatin Remodeling and Transcriptional Activation by the GAL4 and VP16 Activation Domains in Budding Yeast

Abstract

Chromatin-modifying enzymes such as the histone acetyltransferase GCN5 can contribute to transcriptional activation at steps subsequent to the initial binding of transcriptional activators. However, few studies have directly examined dependence of chromatin remodeling in vivo on GCN5 or other acetyltransferases, and none have examined remodeling via nucleosomal activator binding sites. In this study, we have monitored chromatin perturbation via nucleosomal binding sites in the yeast episome TALS by GAL4 derivatives in GCN5⁺ and gcn5Δ yeast cells. The strong activator GAL4 shows no dependence on GCN5 for remodeling TALS chromatin, whereas GAL4-estrogen receptor-VP16 shows substantial, albeit not complete, GCN5 dependence. Mini-GAL4 derivatives having weakened interactions with TATA-binding protein and TFIIB exhibit a strong dependence on GCN5 for both transcriptional activation and TALS remodeling not seen for native GAL4. These results indicate that GCN5 can contribute to chromatin remodeling at activator binding sites and that dependence on coactivator function for a given activator can vary according to the type and strength of contacts that it makes with other factors. We also found a weaker dependence for chromatin remodeling on SPT7 than on GCN5, indicating that GCN5 can function via pathways independent of the SAGA complex. Finally, we examine dependence on GCN5 and SWI-SNF at two model promoters and find that although these two chromatin-remodeling and/or modification activities may sometimes work together, in other instances they act in complementary fashion.

Transcriptional activation in eukaryotes is generally initiated by the binding of an activator to the promoter. Activator binding sites may occur in nucleosome-free regions or may be incorporated into positioned nucleosomes (2, 14, 40, 73, 86). In the latter case, activator binding usually results in nucleosome perturbation (3, 58, 86). We have used yeast episomes having strongly positioned nucleosomes to show that transcriptional activators are capable of perturbing chromatin structure via nucleosomal sites in vivo even outside the context of a natural promoter (46, 66). This perturbation seems likely to result from stable activator binding, although this has not been demonstrated directly, and is strongly dependent on the activation domain (AD) (46, 66, 82). For example, GAL4 perturbs a positioned nucleosome containing its binding site in yeast to a much greater degree in cells grown in galactose medium, when it is in its activating configuration, than in glucose, when it is not, and the chimeric activator GAL4-estrogen receptor (ER)-VP16 similarly is much more effective at this perturbation in the presence of activating β-estradiol than in its absence (66). Consistent with these results, binding of transcription factors is increased by ADs in both yeast and mammalian cells to sites having uncharacterized chromatin structure (9, 70).

ADs could facilitate chromatin remodeling via nucleosomal binding sites by contacting components of the preinitiation complex (PIC), such as TFIIB or TATA-binding protein (TBP), and thereby stabilizing activator binding, or by recruiting complexes capable of remodeling or modifying chromatin. Considerable support for activators having both of these functions derives from in vitro and in vivo studies (53, 54, 56). Interactions with components of the PIC could enhance binding through cooperative protein-protein interactions, with additional free energy being provided by nonspecific interactions between PIC components and nearby DNA sequences (12, 72). We have recently shown that fusion of TBP to the DNA-binding domain (DBD) of GAL4 or LexA is not sufficient to remodel via a nucleosomal binding site or to remodel nucleosomes that are near accessible binding sites in vivo (63). However, whether TBP contributes to chromatin remodeling in the context of a true activation domain, or whether other PIC components could be involved in this function, remains to be tested.

ADs can recruit chromatin remodeling complexes, which use ATP to alter nucleosome structure in a way that increases accessibility of nucleosomal DNA (54). Recruitment of chromatin-remodeling complexes such as SWI-SNF and NURF can facilitate transcriptional activation of chromatin in vitro, probably by increasing accessibility to PIC components (1, 44, 54). Some genes that undergo chromatin remodeling upon activation in vivo depend on SWI-SNF but others do not, indicating functional redundancy among activities that contribute to chromatin remodeling (16, 22, 28, 30, 45). We have shown that activators can perturb chromatin via nucleosomal sites in swiΔ yeast cells, although some reduction in perturbation of nucleosome positioning was seen for GAL4-ER-VP16 in one case (62).

Another potential class of targets for ADs is chromatin-modifying complexes. Chromatin-modifying complexes characterized to date include enzymes, such as GCN5 in the SAGA complex and ESA1 in the NuA4 complex, that can acetylate the amino termini of histones (7). Other histone modifications that may contribute to transcriptional regulation also have been observed and presumably are due to additional complexes (11). GCN5, the prototypical histone acetyltransferase (HAT) (8), contributes to activation of a number of genes in yeast (23, 28, 31, 52, 55, 68). Recruitment of the SAGA complex greatly increases transcriptional activation in vitro (32, 67, 71), and a few examples in which chromatin remodeling depends on GCN5 in vivo have been reported (27, 28, 69). Whether GCN5 contributes to chromatin remodeling by activators via nucleosomal binding sites has not been reported and is the major focus of this paper.

Potential redundancy among chromatin-remodeling and -modifying complexes, already alluded to, has been investigated but is not well understood (6, 55, 57, 63, 68). Different activators may differentially recruit distinct complexes, and promoters may vary in their requirements for chromatin remodeling (1, 15, 32, 69). We have attempted to address this issue in this work by comparing chromatin remodeling dependence on GCN5 for ADs varying in type and in strength. We also compare requirements for several ADs at distinct promoters that require chromatin remodeling for their activation.

MATERIALS AND METHODS

Plasmids.

The chromatin reporter plasmid TALS was introduced into yeast as described previously (66). Plasmids expressing GAL4-ER-VP16, GAL4-GAL11, and GAL4-GCN4 have been described elsewhere (41, 62, 83). Plasmids RJR192 and RJR200 were a gift from M. Ptashne and are CEN plasmids that express truncated mini-GAL4s from the ACT1 promoter (81). The promoter and coding regions for these mini-GAL4s were subcloned into pRS416 and pRS414 after digestion with EcoRI and SaII. The β-galactosidase reporter gene, 17-CYC1-lacZ, contains a single GAL4 binding site upstream of the bacterial lacZ gene (62). The α-galactosidase expression reporter gene GAL10-MEL1 contains the four GAL4 binding sites and the TATA region from the yeast GAL10 promoter in front of the MEL1 coding sequence (62).

Yeast strains.

Strains used are listed in Table 1. GAL4 was disrupted by using a GAL4::HIS3 plasmid generously provided by Jon Swaffield and Stephen Johnston, and haploid a strains isogenic to GMy27 and PSY316 were constructed by two-step transplacement of Mata into the mating-type locus. Plasmids were transformed into yeast cells by a standard lithium acetate method (29). Cells were grown in dropout medium (Bio 101) with 2% glucose or 1.5% raffinose plus 1% galactose. To induce activation by GAL4-ER-VP16, β-estradiol (Sigma) from a 5 mM ethanol stock was added to a final concentration of 100 nM 3 to 4 h before cells were harvested for DNA purification or determination of enzyme activity. β-Galactosidase activity was assayed as described elsewhere (62) and reported in Miller units: 1,000× optical density at 420 nm [OD₄₂₀]/[OD₆₀₀ × time (min) × volume (ml)]. α-Galactosidase activity was reported similarly, as previously described (62).

TABLE 1.

S. cerevisiae strains used in this study

Strain	Genotype	Reference
PSY316	MATα ade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1	42
GMy27	MATα ade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1 gcn5Δ	42
PSY316a	MATaade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1	This study
GMy27a	MATaade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1 gcn5Δ	This study
GSY099	MATα ade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1 gal4Δ::HIS3	This study
GSY100	MATα ade2-101 ura3-52 leu2-3,2-112 his3-Δ200 lys2 trp1 gcn5Δ gal4Δ::HIS3	This study
FY24	MATα ura3-52 trp1 Δ63 leu2Δ1	78
FY1292	MATα his3Δ200 gcn5Δ::HIS3 leu2Δ1 ura3-52 arg4-12 trp1Δ63 lys2-173R2	59
FY1300	MATα his3Δ200 spt7 Δ402::LEU2 leu2Δ1 ura3-52 trp1Δ63 lys2-173R2	59
GSY1300	MATα his3Δ200 spt7 Δ402::LEU2 leu2Δ1 ura3-52 trp1Δ63 lys2-173R2 gal4Δ::HIS3	This study
CY296	MATagal4Δ::LEU2 lys2-801 ade2-101 leu2-Δ1 his3-Δ200 ura3-Δ99 trp1-Δ99	10
CY297b	MATα gal4Δ::LEU2 lys2-801 ade2-101 leu2-Δ1 his3-Δ200 ura3-Δ99 trp1 swi1Δ::LEU2	62

Analysis of plasmid chromatin.

Yeast cells (100 ml) were grown at 30°C to an OD₆₀₀ of between 0.6 and 1.3. Spheroplasts were prepared as previously described, and 300-μl samples were digested with 0 to 50 U of micrococcal nuclease (MNase) (Worthington) per ml for 5 min at 37°C (66). Cleavage patterns were consistent over these concentrations. Reactions were stopped with 55 μl of 5% sodium dodecyl sulfate (SDS)–5 mg of proteinase K per ml. Naked DNA controls were treated with SDS-proteinase K prior to digestion with MNase. Following cleanup with phenol and chloroform, aliquots were treated with RNase and digested with EcoRV. Cleavage patterns were visualized by indirect end labeling (50, 79). The samples were electrophoresed along with HaeIII-digested φX markers in a 1.2% agarose gel at 4 V/cm for 5 to 5.5 h. The DNA was transferred by capillary action to nylon membranes (Duralon UV: Stratagene), and Southern analysis was performed. Probes were EcoRV-to-XbaI fragments from TALS prepared by PCR. Densitometric scans were obtained using the public domain NIH Image program (developed at the National Institutes of Health and available at http://rsb.info.nih.gov/nih-image/).

Topoisomer analysis.

To prepare DNA, 100 μl of 5% SDS to 5 mg of proteinase K per ml was added to microcentrifuge tubes containing approximately 600 μl of glass beads. Then 10 ml of cell culture (A₆₀₀ = 0.6 to 1.2) was spun down, resuspended in 500 μl of 10 mM Tris–1 mM EDTA, transferred to the tubes with glass beads, and rapidly lysed by vortexing at top speed two times with a 1-min interval. Purified DNA was separated on 1.5% agarose gels with 40 μg of chloroquine diphosphate (Sigma) per ml in both gel and buffer at 2.5 V/cm for 18 to 20 h. The gel was blotted and probed as above. Quantitation of topoisomers was performed by PhosphorImager analysis (Molecular Dynamics), and the Gaussian centers of distribution were calculated (47).

Protein detection.

Western analysis was performed using whole-cell extracts. Two OD₆₀₀ units of cells (A₆₀₀ = 0.5 to 2) was added to 2 ml of 50 mM Tris (pH 7.5)–10 mM NaN₃ on ice, spun down, resuspended in 30 μl of ESB (2% SDS, 80 mM Tris [pH 6.8], 10% glycerol, 1.5% dithiothreitol, 0.1 mg of bromphenol blue per ml), and quickly transferred to a microcentrifuge tube for a 3-min incubation at 100°C. Samples were stored at −20°C. Prior to SDS-polyacrylamide gel electrophoresis, glass beads were added to reach the meniscus and the samples were vortexed at top speed for 2 min. An additional 70 μl of EBS was added, and the samples were heated to 100°C for 1 min. After standard SDS-polyacrylamide gel electrophoresis, the proteins were electroblotted to a Millipore polyrinylidene difluoride membrane. The blots were blocked with 5% blocking solution from Amersham's ECL kit or 5% powdered milk in TBS-T (20 mM Tris [pH 7.6], 137 mM NaCl, 0.1% Tween 20), incubated with primary antibody against the GAL4 DBD (Santa Cruz Biotechnology, Inc.) and secondary anti-rabbit immunoglobulin-G antibody, and developed as instructed by the manufacturer.

RESULTS

Dependence on GCN5 for chromatin remodeling by GAL4 and GAL4-ER-VP16 via nucleosomal binding sites.

In previous work, we have examined transcription factor binding to nucleosomal sites, using as a chromatin reporter the yeast episome TALS (4, 62, 66). This stably replicating episome harbors an α2-MCM1 operator, resulting in the plasmid being packaged into strongly positioned nucleosomes in yeast α cells (60). It also has a binding site for GAL4 located within nucleosome IV, in a region inaccessible to either Escherichia coli Dam methyltransferase or SssI methyltransferase expressed in yeast (34, 82). In spite of this inaccessibility, expression of either GAL4 or GAL4-ER-VP16 (an estrogen-dependent chimeric activator) under activating conditions (i.e., in galactose or in the presence of hormone) in yeast harboring TALS results in remodeling of the plasmid chromatin (66). This remodeling is detected as changes in the MNase cleavage pattern, in restriction endonuclease accessibility, and in plasmid topology. As the former two alterations are centered on the region of the nucleosome containing the GAL4 binding site, the simplest interpretation is that binding of these GAL4 derivatives perturbs the nucleosome containing the GAL4 binding site. When either GAL4 or GAL4-ER-VP16 is expressed under nonactivating conditions, or when other nonactivating derivatives of GAL4 are expressed, minimal perturbation of TALS chromatin is seen in yeast α cells (66).

One possible function for ADs in allowing binding of transcription factors to nucleosomal sites would be to recruit chromatin-modifying activities, such as GCN5. Such activities could, by modifying the histones, alter the equilibrium between histone binding and factor binding, therefore allowing stable occupancy by the factor. To test this possibility, we examined whether perturbation of TALS chromatin in yeast α cells by GAL4 and GAL4-ER-VP16 was dependent on GCN5. We first assessed perturbation of TALS chromatin by MNase digestion followed by indirect end labeling. Spheroplasts were prepared from cells harboring TALS and treated with various amounts of MNase, and MNase cleavages were determined relative to the unique EcoRV site (Fig. 1). In the absence of hormone, we observed MNase cleavage sites separated by protected regions about 150 bp in length, consistent with positioned nucleosomes on the TALS plasmid as observed many times previously (Fig. 1, lanes 1, 9, and 10 versus lanes 6 and 7) (34, 60, 66, 82). Addition of β-estradiol during growth of GCN5⁺ yeast induces a new cleavage site on the edge of nucleosome III and increased cutting at the edge of nucleosome IV (Fig. 1B, lanes 3 and 4; Fig. 1C). In contrast, when TALS and the GAL4-ER-VP16 expression vector were introduced into gcn5Δ cells, addition of hormone resulted in little or no change in the MNase cleavage pattern (Fig. 1B, lanes 9 to 12; Fig. 1C). Thus, changes in chromatin structure of TALS via a nucleosomal GAL4 binding site, and mediated by the VP16 AD, depend on GCN5.

FIG. 1.

GCN5-dependent nucleosome perturbation at a nucleosomal GAL4 binding site in the yeast episome TALS by ligand-activated GAL4-ER-VP16. (A) Schematic diagram of the TALS plasmid. Positioned nucleosomes present in yeast α cells are shown as ellipses, and the α2-MCM1 and GAL4 binding sites are shown as small rectangles. The TRP1 marker is also indicated. (B) Chromatin was prepared from yeast GCN5⁺ (PSY316) or gcn5Δ (GMy27) α cells harboring TALS and expressing GAL4-ER-VP16 in the presence or absence of β-estradiol (E2), as indicated, and digested with MNase at 0 (lanes 5, 8, and 13), 2 (lanes 1, 4, 9, and 12), 4 (lanes 6 and 7), or 5 (lanes 2, 3, 10 and 11) U/ml. MNase cleavage sites were mapped counterclockwise relative to the EcoRV site as indicated. The arrowheads indicate positions of cleavage sites that are enhanced in cells containing GAL4-ER-VP16 when hormone is present (lanes 3 and 4). Locations of nucleosomes II to V in unperturbed TALS and the α2-MCM1 operator (rectangle between nucleosomes IV and V) are indicated to the sides. The rectangle in nucleosome IV represents the GAL4 binding site. (C) Densitometric scans of lanes showing the perturbation elicited by GAL4-ER-VP16 in the presence of hormone in GCN5⁺ cells (lanes 1 and 4) but not in gcn5Δ cells (lanes 9 and 12).

This dependence is not seen for endogenous GAL4. The characteristic MNase cleavages induced in TALS by GAL4 in galactose are essentially identical to those induced by GAL4-ER-VP16 by addition of β-estradiol (66) and are also seen in gcn5Δ yeast cells (Fig. 2A and B). Since GAL4 is a stronger activator than GAL4-ER-VP16, we wondered whether the difference in their dependence on GCN5 might reflect their activator strength or was due to qualitative differences in their properties. To test this question, we constructed isogenic gal4Δ GCN5⁺ and gcn5Δ yeast strains and introduced expression vectors for mini-GAL4 proteins. These mini-GAL4s consist of the GAL4 DBD. (amino acid residues 1 to 100) fused to portions of the AD and have been shown to have weakened interactions with TFIIB and TBP that correlate very well with their strength as transcriptional activators (81). Figure 2C shows results of TALS remodeling by GAL4 (1–100+ 840–869) in GCN5⁺ and gcn5Δ yeast cells. Although remodeling of TALS chromatin is clearly seen in GCN5⁺ cells, it is not seen in gcn5Δ cells. Similar results were seen with the even more attenuated GAL4 (1–100+840–857), although remodeling was somewhat reduced even in GCN5⁺ cells for this derivative (data not shown and Table 2).

FIG. 2.

Weakening the GAL4 AD confers GCN5 dependence on its ability to remodel TALS chromatin. (A) Chromatin was prepared from yeast GCN5⁺ (PSY316) or gcn5Δ (GMy27) α cells harboring TALS and grown in medium containing glucose or galactose, as indicated, and digested with MNase at 0 (lane 2), 2 (lanes 3 and 9), 4 (lane 1), 5 (lanes 4 and 8), 20 (lanes 5 and 7), or 50 (lane 6) U/ml. MNase cleavage sites were mapped counterclockwise relative to the EcoRV site as in Fig. 1. The asterisks indicate positions of cleavage sites that are enhanced in cells grown in galactose (lanes 6 to 8). Locations of nucleosomes II to V in unperturbed TALS and the α2-MCM1 operator (rectangle between nucleosomes IV and V) are indicated to the right. The rectangle in nucleosome IV represents the GAL4 binding site. (B) Densitometer traces of lanes 5 and 6 from panel A. The arrows indicate the enhanced MNase cleavages that are indicated by the asterisks in panel A. (C) As in panel A, but chromatin was prepared from gal4Δ GCN5+ (GS099) or gal4Δ, gcn5Δ (GS100) yeast α cells harboring both TALS and the expression vector for GAL4(1–100+840–869). Samples were digested with 2 (lanes 1, 4, 5, and 8) or 5 (lanes 2, 3, and 5, 6) U of MNase per ml.

TABLE 2.

Change in topology of the TALS minichromosome induced by GAL4 derivatives in GCN5⁺ and gcn5Δ yeast haploid α cells

Activator	Linking no. change (avg ± SD)a
	GCN5+ (n)	gcn5Δ (n)
GAL4	0.299 ± 0.13 (18)	0.559 ± 0.31 (26)
GAL4-ER-VP16	0.643 ± 0.16 (11)	0.208 ± 0.14 (9)
GAL4(1–100+840–869)	0.296 ± 0.12 (7)	0.099 ± 0.10 (7)
GAL4(1–100+840–857)	0.184 ± 0.13 (7)	0.176 ± 0.17 (7)

^aAverage of n determinations, using at least three independent clones, between cells grown in glucose and galactose for all except for GAL4-ER-VP16, where the changes were between cells grown in the presence and absence of 100 nM β-estradiol. 

To examine further the dependence of TALS remodeling by GAL4 and derivatives on GCN5, we measured the topology of the TALS minichromosome in the presence of these activators under induced and uninduced conditions. In wild-type yeast, both GAL4 and GAL4-ER-VP16 alter TALS topology under inducing conditions, causing loss of an average of nearly one negative supercoil per plasmid molecule (4, 62, 66). Since each nucleosome in a closed circular plasmid induces one negative supercoil (24, 65), this is consistent with loss of about one nucleosome per plasmid nucleosome, in good agreement with the localized alteration in the MNase cleavage pattern. Furthermore, because the changes in topology can be measured with great accuracy, this provides a more quantitative assessment of remodeling of TALS chromatin.

Figure 3 shows the results of topological analysis of remodeling of the TALS minichromosome by GAL4, GAL4-ER-VP16, and GAL4(1–100+840–869) in GCN5⁺ and gcn5Δ yeast cells. Figure 3A shows an example of the raw data from which the quantitative measurements were derived. TALS DNA was rapidly purified from yeast cells grown in glucose or galactose, as indicated, and electrophoresed under conditions in which individual topoisomers are resolved. The uppermost band in each lane corresponds to nicked circular plasmid, and the distributions beneath are of individual topoisomers differing in linking number by unit increment, with more positively supercoiled topoisomers migrating faster. The downward shift in the topoisomer distributions from gcn5Δ cells grown in the presence of galactose relative to those grown in glucose is readily discerned. Quantitation of such data is summarized in Fig. 3B and Table 2. In agreement with the MNase cleavage data, remodeling by GAL4 is not decreased in gcn5Δ cells, whereas changes in topology induced by GAL4-ER-VP16 were substantially reduced and changes in topology induced by GAL4(1–100+840–869) were essentially eliminated. Western analysis demonstrated that the dependence of remodeling by GAL4-ER-VP16 on GCN5 was not due to reduced expression of this chimeric activator in gcn5Δ cells (Fig. 3C). Although we were unable to detect the mini-GAL4 proteins by Western blotting, these proteins were expressed under control of the ACT1 promoter, whose activity does not depend on GCN5 (31).

FIG. 3.

Topological analysis of TALS remodeling in GCN5+ and gcn5Δ, cells. (A) Topoisomer distributions of TALS in gcn5Δ (GMy27) cells grown in medium containing glucose or galactose, as indicated. The two pairs of lanes contain distinct TALS-containing clones. The uppermost band corresponds to nicked circular DNA, and the individual bands migrating below represent individual topoisomers, with faster-migrating bands being more positively supercoiled. (B) Quantitation of linking number changes conferred on TALS by GAL4, GAL4-ER-VP16, and GAL4(1–100+840–869) in GCN5⁺ (GMy27) and gcn5Δ (PSY316) haploid α and a cells, as indicated. Data are from at least three independent samples for each column. (C) Equal amounts of protein from GCN5⁺ (PSY316) and gcn5Δ (GMy27) cells expressing GAL4-ER-VP16 were subjected to Western blotting using antibody to the GAL4 DBD.

Nucleosome positioning on TALS in yeast haploid α cells is dictated by the α2-MCM1 complex acting in concert with the SSN6-TUP1 complex and the histone H4 amino terminus (13, 60, 61). As an association has recently been reported between the SSN6-TUP1 complex and histone deacetylase activities (77, 80), it seemed possible that the requirement for GCN5 in activator-dependent remodeling of TALS chromatin in α cells could reflect a requirement to overcome histone deacetylase activity. We therefore also examined activator-mediated changes in TALS topology in yeast haploid a cells. Although nucleosome positioning is weaker in TALS in a than in α cells, it is still nonrandom (34). Very little alteration in MNase cleavage pattern is seen in TALS in haploid a cells in the presence of GAL4 or GAL4-ER-VP16 under activating conditions (our unpublished results), but activator-dependent changes in topology are stronger than those seen in α cells (63). We found that as in haploid α cells, the change in TALS topology induced by GAL4 in haploid a cells was essentially independent of GCN5, whereas that induced by GAL4-ER-VP16 in the presence of hormone was reduced in gcn5 cells (Fig. 3B). Although this result does not refute the possibility that the dependence on GCN5 for TALS remodeling by GAL4-ER-VP16 and GAL4(1–100+840–869) in α cells is due in part to a requirement to overcome SSN6-TUP1-mediated repression, it indicates that a remodeling requirement apart from this is also likely. Taken together, these results indicate that chromatin remodeling via nucleosomal binding sites by GAL4-ER-VP16 and GAL4(1–100+840–869), but not by GAL4, depends strongly on GCN5.

Dependence of chromatin remodeling on the SAGA component SPT7 is less severe than on GCN5.

GCN5 is a component of both the SAGA and ADA complexes, each of which contains both shared and unique subunits (18, 26). Under conditions in which the SAGA complex was recruited by activators to enhance transcription of nucleosomal templates in vitro, the ADA complex was not, and the function of the ADA complex in yeast is not known (18, 71). To test further the dependence of chromatin remodeling via nucleosomal binding sites on GCN5, we therefore decided to examine TALS remodeling in spt7Δ yeast cells. SPT7 is a component of the SAGA complex but not of the ADA complex, and loss of SPT7 causes essentially complete disruption of the SAGA complex (26). Thus, if GCN5 functions entirely via the SAGA complex, spt7 phenotypes should be indistinguishable from, or possible more severe than, gcn5 phenotypes. However, if GCN5 also functions via the ADA complex, spt7 phenotypes may in some cases be less severe.

We first examined remodeling of TALS by GAL4 in spt7Δ cells by MNase digestion followed by indirect end-labeling analysis. Given that this remodeling did not require GCN5, it was not surprising to find that it appeared unaffected by loss of SPT7 (Fig. 4A, lanes 2 and 3). To examine the dependence of GAL4(1–100+840–869) on SPT7 for remodeling of TALS, we first constructed a gal4Δ spt7Δ strain and then introduced the expression vector for this mini-GAL4 into this strain along with TALS. MNase digestion showed no alteration in cleavage sites in TALS induced by GAL4(1–100+840–869) in galactose (Fig. 4A, lanes 7 to 10; cf. Fig. 2), indicating that remodeling of TALS by GAL4(1–100+840–869) requires the intact SAGA complex.

FIG. 4.

Dependence of TALS remodeling by GAL4 derivatives on SPT7. (A) Chromatin was prepared from yeast GAL4+ spt7Δ α (FY1300) or gal4Δ spt7Δ (GSY1300) cells expressing GAL4(1–100+840–869) and harboring TALS and grown in medium containing glucose or galactose, as indicated. Samples were digested with MNase at 0 (lanes 6 and 11), 2 (lanes 2, 3, 7, and 10), 5 (lanes 8 and 9), or 10 (lanes 1 and 5) U/ml. MNase cleavage sites were mapped counterclockwise relative to the EcoRV site as in Fig. 1. Lane 4 contains φX marker DNA digested with HaeIII. The arrowheads indicate positions of cleavage sites that are enhanced in cells grown in galactose (seen in lane 3 but not in lanes 9 and 10). Locations of nucleosomes II to V in unperturbed TALS and the α2-MCM1 operator (rectangle between nucleosomes IV and V) are indicated to the side. The rectangle in nucleosome IV represents the GAL4 binding site. (B) Densitometric traces of lanes 3, 2, and 10 (in descending order) from panel A. The arrows and vertical lines indicate enhanced MNase cleavage induced by GAL4, but not GAL4(1–100+840–869), in cells grown in galactose. (C) As in panel A, but cells were either SPT7 (FY24) or spt7Δ (FY1300) cells expressing GAL4-ER-VP16 and grown in the presence or absence of hormone, as indicated. Samples were digested with MNase at 0 (lanes 4 and 9), 2 (lanes 3, 5, and 8), 4 (lane 10), or 5 (lanes 1, 2, 6, and 7) U/ml. Enhanced cleavages induced in the presence of hormone are indicated by arrowheads in lanes 2, 3, 7, and 8. (D) Densitometric scans of lanes 3, 1, 7, and 6 (in descending order) from panel C. The arrows indicate enhanced MNase cleavages seen in the presence of hormone.

Surprisingly, when we examined remodeling of TALS by GAL4-ER-VP16, we found that in contrast to the dependence observed on GCN5, remodeling did not require SPT7 (Fig. 4C). To corroborate this result and to test the possibility that it could be due to a difference in strain background (as the spt7 deletion was in a different background than the gcn5 deletion used earlier), we examined TALS remodeling by topological analysis in congenic wild-type, spt7Δ, and gcn5Δ cells. The results (Fig. 5) confirm that remodeling of TALS by GAL4-ER-VP16 is essentially unaffected by loss of SPT7 but is substantially diminished by loss of GCN5. GAL4 was still able to alter TALS topology in gcn5Δ cells from the same strain background as the spt7Δ strain, as in the gcn5Δ strain used earlier (Fig. 5 and data not shown). TALS topology was unaffected by GAL4(1–100+840–869) under activating conditions in spt7Δ cells (data not shown).

FIG. 5.

Topological analysis of TALS remodeling in congenic spt7Δ, gcn5Δ, and wild-type (WT) cells. (A) Topoisomer distributions of TALS in gcn5Δ (FY1292), spt7Δ (FY1300), or wild-type (FY24) cells expressing GAL4-ER-VP16 grown in medium containing glucose in the presence or absence of β-estradiol or grown in medium containing galactose and harboring the GAL4 gene on a multicopy plasmid (lanes 5 and 6), as indicated. Lane pairs 1–2 and 3–4 contain distinct TALS-containing clones. The uppermost band corresponds to nicked circular DNA, and the individual bands migrating below represent individual topoisomers, with faster-migrating bands being more positively supercoiled. The centers of the topoisomer distributions are indicated by the solid circles (the centers are identical for lanes 1 and 2, for lanes 3 and 4, and for lanes 5 and 6.). (B) Quantitation of linking number changes conferred on TALS by GAL4 and GAL4-ER-VP16 in the same strains as in panel A. Data are from at least three independent samples for each column.

Transcriptional dependence on GCN5 depends on both promoter architecture and activator.

We have previously used two transcriptional reporter plasmids that differ in their promoter architecture (Fig. 6A) to show that activator dependence on the chromatin-remodeling complex SWI-SNF can vary with promoter structure (62). The reporter gene 17-CYC1-lacZ contains a single 17-bp binding site for GAL4 upstream of the CYC1 promoter sequence and the bacterial lacZ coding sequence. One of the two major TATA elements of the CYC1 promoter is in a region highly accessible to MNase (62). The other reporter consists of a fusion of the GAL10 promoter to the MEL1 coding sequence. This reporter gene retains the structure of the native GAL10 promoter, with four GAL4 binding sites in a nucleosome-free region and downstream sequences (including the TATA) packaged in a highly organized array of positioned nucleosomes (62).

FIG. 6.

Dependence on GCN5 and SWI-SNF for activation of 17-CYC1-lacZ and GAL10-MEL1 reporter genes by GAL4, GAL4-ER-VP16, GAL4-GCN4, GAL4-GAL11, and mini-GAL4s. (A) Schematic diagram of the two reporter genes. The fusions between the coding sequences (lacZ and MEL1) and the promoters are indicated by the jagged lines, the TATA elements are represented by the T's, and nucleosomes are indicated by ellipses. Fuzzy ellipses in the 17-CYC1-lacZ reporter gene indicate nucleosomes less well positioned than those present in the GAL10-MEL1 reporter gene (62). (B) Transcriptional dependence on GCN5 (strains PSY316 and GMy27). (C) Transcriptional dependence on SWI-SNF (strains CY296 and CY297b; data for GAL4, GAL4-ER-VP16, and GAL4-GAL11, indicated by asterisks, taken from reference 62). (D) Transcriptional dependence for activation by GAL4 and mini-GAL4s on GCN5 (strains GSY099 and GSY100). Each column represents data from independent determinations on at least three clones; absence of visible error bars indicates that the standard error was too small to be visible on the column.

To test whether dependence on GCN5 for activation depends on promoter structure, activator type, or both, we examined the ability of various ADs, as well as the holoenzyme component GAL11 (33), fused to the GAL4 DBD to activate transcription of these two reporters (carried on CEN plasmids) in GCN5⁺ and gcn5Δ yeast cells (Fig. 6B). We observed a general dependence on GCN5 for activation of 17-CYC1-lacZ for the three ADs tested, as well as for GAL4-GAL11. In contrast, activation of GAL10-MEL1 was nearly independent of GCN5 except for when GCN4 was used as the AD. Nearly the opposite results are seen for the SWI-SNF complex, with GAL10-MEL1 being strongly dependent, including when GAL4-GCN4 is the activator, whereas 17-CYC1-lacZ shows little dependence (Fig. 6C). These results show that promoter structure may dictate coactivator dependence in some cases, although some ADs may also show a stringent coactivator dependence, and they also confirm that GCN5 and the SWI-SNF complex sometimes operate in distinct pathways (6, 55, 57, 59, 68).

Since GAL4(1–100+840–869) depended on GCN5 for its ability to remodel TALS chromatin, whereas GAL4 expressed from the endogenous GAL4 gene did not, we were interested in determining whether GAL4(1–100+840–869) would also show a greater requirement for GCN5 than GAL4 for transcriptional activation. The 17-CYC1-lacZ reporter showed substantial dependence on GCN5 for both intact GAL4 and the two mini-GAL4s tested (Fig. 6D). For unknown reasons, we observed a somewhat stronger dependence on GCN5 for activation of 17-CYC1-lacZ mediated by GAL4 when the GAL4 gene was expressed from a plasmid than from the endogenous locus (compare Fig. 6D to Fig. 6A). More interestingly, the GAL10-MEL1 reporter gene showed a much stronger dependence on GCN5 for activation by the two mini-GAL4s than it did for GAL4 expressed from a plasmid. We conclude that although the strong GAL4 AD can probably recruit GCN5, most plausibly via the SAGA complex, it also likely recruits other activities that are redundant with GCN5 for chromatin remodeling and transcriptional activation.

DISCUSSION

Chromatin-modifying enzymes such as the HAT GCN5 have emerged as important participants in transcriptional regulation. Previous in vivo studies have shown that GCN5-mediated transcriptional activation requires GCN5 HAT activity and that acetylation of histone H3 is increased concomitant with GCN5-dependent activation by GCN4 in the vicinity of the GCN4 binding site (37, 38, 76), Similarly, activators can target the GCN5-containing SAGA complex to promoters in vitro, resulting in enhanced transcription and localized histone acetylation (67, 71, 75). Based on in vitro studies showing that histone acetylation can increase accessibility of nucleosomal binding sites (39, 74), it has been speculated that GCN5-mediated acetylation may facilitate binding of transcription factors by altering chromatin structure. However, this model is complicated by several observations: (i) nucleosome accessibility is more affected by histone H4 acetylation than histone H3 acetylation, although H3 is the principal histone target of GCN5 (74); (ii) the histone H3 amino terminus is not the only relevant in vivo target of GCN5 (85); and (iii) in some cases, access to nucleosomal binding sites is not increased by histone acetylation (1, 25). It therefore becomes important to examine whether factor accessibility is in fact increased in a GCN5-dependent manner in vivo.

In this work, we have used the yeast episome TALS as a chromatin reporter to test whether GCN5 contributes to chromatin remodeling via a nucleosomal GAL4 binding site by GAL4 and derivatives thereof. Remodeling of TALS by GAL4 derivatives requires a GAL4 binding site, and changes in chromatin structure measured by MNase cleavage are centered around the GAL4 binding site (66). Furthermore, in vivo footprinting by SssI methyltransferase shows that GAL4 is capable of binding to TALS chromatin (82). These data strongly suggest that chromatin remodeling of TALS reflects binding of GAL4 and derivatives to a nucleosomal site, with concomitant remodeling of chromatin. Since this remodeling depends on a functional AD, and GCN5-dependent acetylation and transcriptional activation by GCN4 require the GCN4 AD (37), it seemed reasonable that at least some activators might depend on GCN5 for remodeling of TALS chromatin and, by inference, for binding to nucleosomal sites.

We found that GAL4 was able to remodel TALS chromatin equally well in GCN5⁺ and gcn5Δ yeast cells, as assayed by MNase cleavage and changes in TALS topology. However, both GAL4-ER-VP16 and a truncated mini-GAL4 with a weakened activation domain (81) showed a strong dependence on GCN5 for remodeling of TALS. These results indicate that chromatin remodeling via nucleosomal binding sites can be enhanced by GCN5 in vivo and represent the first demonstration, to our knowledge, of dependence on GCN5 for chromatin remodeling by VP16 or by GAL4-derived ADs in vivo. Nucleosome positioning on TALS in yeast haploid α cells depends on recruitment of the SSN6-TUP1 repressor complex by the α2-MCM1 complex (13, 60). As the SSN6-TUP1 complex has recently been reported to recruit histone deacetylase activities (77, 80), it is possible that the dependence on GCN5 for remodeling TALS chromatin in yeast haploid α cells reflects in part a requirement to overcome histone deacetylase activity. However, dependence on GCN5 was also seen for changes in TALS topology induced by GAL4-ER-VP16 in the presence of hormone in haploid a cells (Fig. 3B), suggesting that an additional requirement for GCN5 in chromatin remodeling apart from this potential antirepressive effect is likely. Furthermore, the GAL4 binding site in TALS derives from the GAL3 promoter, but the downstream TATA element and transcription start site are absent. Thus, the AD-dependent chromatin remodeling of TALS occurs outside the context of a native promoter and is therefore separate from transcription per se. This is consistent with recent results that indicate that GCN5-dependent histone acetylation is separable from transcription (37) and further suggests that GCN5-dependent remodeling by transcriptional activators is also separable from transcription.

Unexpectedly, we found that remodeling of TALS by GAL4-ER-VP16 is affected less in spt7Δ than in gcn5Δ yeast cells. Previous work showed that activation and concomitant remodeling of chromatin structure at the PHO8 promoter depends equally on SPT7 and GCN5 (28). Since the SAGA complex is lost in spt7Δ cells (26), it was concluded that chromatin remodeling and transcriptional activation at the PHO8 promoter require the SAGA complex. In contrast, our results indicate that remodeling of TALS by GAL4-ER-VP16 may also occur through the GCN5-containing ADA complex (18). Interestingly, transcription of the TRP3 and HIS3 genes appears to depend more strongly on GCN5 than on SPT7 (21, 23). The extent to which the ADA complex functions in transcription and chromatin remodeling in wild-type cells, however, remains uncertain.

Additionally, we found that the role of GCN5 in transcription can vary between promoters, even when the same AD is used. Activation of the GAL10-MEL1 reporter gene by both GAL4 and GAL4-ER-VP16 showed little dependence on GCN5, whereas both of these activators showed substantial dependence on GCN5 at the 17-CYC1-lacZ promoter. In contrast to these activators, GAL4-GCN4 and the two mini-GAL4s examined showed strong GCN5 dependence at both test promoters. The dependence on GCN5 for the GCN4 AD is not surprising, as GCN4 has been shown to depend strongly on GCN5 for gene activation (5, 23, 48). The dependence on GCN5 for the mini-GAL4s, in contrast to wild-type GAL4, suggests that GAL4 is able to recruit other activities that are redundant with GCN5. By weakening the strength of the GAL4 AD, these other contacts are also likely weakened (81), unmasking the ability of this AD to be assisted by GCN5. Recruitment of GCN5 by GAL4 is also consistent with the increased acetylation of histone H3 observed upon activation of the GAL1 promoter (35). Similarly, previous work has demonstrated increased dependence on GCN5 for transcriptional activation by VP16 when this AD is truncated or mutated (64). Other activators, such as HAP4, might also be capable of recruiting GCN5 but not normally show dependence (5) because of recruitment of functionally redundant activities.

Dependence of activators on GCN5 could reflect direct recruitment, as the GAL4, VP16, and GCN4 ADs have been shown to interact with components of the SAGA complex (17, 43, 71), but it could also arise from untargeted histone acetylation by GCN5 (36–38, 85). This latter possibility is suggested by the dependence on GCN5 seen for transcription induced by GAL4-GAL11 (Fig. 6). This activation is believed to be due to direct recruitment of the holoenzyme (19), which has not been previously reported to result in recruitment of GCN5. Perhaps untargeted histone acetylation by GCN5 enhances the transcription observed in this activator bypass experiment by making the local chromatin more accessible to the general transcription machinery.

The dependence shown by GAL4 and GAL4-ER-VP16 on GCN5 at the two reporter genes is opposite that reported for SWI1 (62), where VP16 and GAL4 had little need for SWI1 to activate transcription from 17-CYC1-lacZ, but were largely inactive at GAL10-MEL1 in swi1Δ cells. The dependence on SWI-SNF for GAL10-MEL1 activation by GAL4 and GAL4-ER-VP16 that we reported earlier (62), and that shown here for GAL4-GCN4, is consistent with reports showing interactions between the SWI-SNF complex and ADs in vitro (49, 51, 84). The varying dependence on SWI-SNF and GCN5 seen for different activators at the two promoters is in general agreement with other recent reports showing that these two chromatin-modifying activities are used at overlapping but distinct sets of promoters (6, 55, 57, 59, 68). The parameters that govern which of these or other chromatin-modifying activities are needed at a particular promoter, and by a particular activator, remain to be worked out.

GCN5 has been shown to contribute to chromatin remodeling during activation of the HIS3 and PHO8 promoters (20, 28). Although activation and remodeling of the PHO5 promoter do not depend on GCN5 in wild-type yeast, the derepression and altered chromatin structure caused by loss of the PHO80/PHO85 repressor are strongly affected by loss of GCN5 (27). Interestingly, activation of modified PHO5 promoters having single binding sites for GAL4, PHO4, or GCN4 shows strong GCN5 dependence, indicating a promoter dependence similar to that seen for our 17-CYC1-lacZ reporter (27, 69). In contrast, little GCN5 dependence is seen when two GAL4 or two PHO4 binding sites are present (27). Perhaps the difference between activation via multiple and single activators at a promoter is analogous to the difference between the strong GAL4 activator and the weakened mini-GAL4 derivative in remodeling TALS chromatin.