In vivo growth defects of gcn5 mutants

The above HAT assays were done with bacterially expressed Gcn5p in the absence of other yeast proteins. Because yGcn5p exists in several large, multisubunit complexes (Georgakopoulos et al. 1995; Candau and Berger 1996; Grant et al. 1997; Pollard and Peterson 1997; Saleh et al. 1997), we set out to determine whether any of the above Gcn5p mutants were also defective in vivo. To this end, each mutant was subcloned, along with GCN5 5′ and 3′ transcriptional regulatory sequences, into a low copy ARS/CEN plasmid and transformed into a gcn5Δ strain. Gcn5p is involved in maximal expression of certain amino acid biosynthesis genes (Georgakapolous and Thireos 1992); thus, loss of GCN5 leads to poor growth under amino acid starvation. The ability of each allele to complement growth of the gcn5Δ yeast strain in minimal medium was then investigated.

Shown in Figure ​Figure2A2A is a comparison of growth of several representative yeast transformants on synthetic minimal medium. As predicted, wild-type GCN5 rescued growth under amino acid starvation, whereas the vector alone control exhibited poor growth under these conditions. With each of the above point mutants as the sole source of Gcn5p, several mutants caused clear gcn5⁻ phenotypes. Among these alleles, F221A caused the most profound defect followed by Y244A/E245A, D214A, G239A/Y244I, Y220A, I174A/I179A, and G187A/G189A. The last two double mutants caused subtle, but highly reproducible, growth phenotypes; such partial defects in the general control pathway can be augmented by applying amino acid analogs, such as ethionine or 3-aminotriazole to the medium (data not shown). The gcn⁻ phenotypes associated with the above mutants were unlikely to be attributable to insufficient expression or protein instability as Western analyses showed roughly equal amounts of Gcn5p (Fig. ​(Fig.3A;3A; data not shown). It is also unlikely that F221A or other in vivo defective mutants possess gross conformational abnormality as coimmunoprecipitation, using an anti-Ada2p antibody, demonstrates that several of our in vivo defective mutants, including F221A, are still able to interact with Ada2p (data not shown).

Open in a separate window

Open in a separate window

Open in a separate window

Figure 2

In vivo tests of functions of mutant Gcn5p. (A) Growth complementation test. Shown are representative clones plated to synthetic minimal medium and grown at 30°C for 3 days before pictures were taken. All mutants have been tested at least three times for reproducibility. Note that certain mutations, such as Y220A, I174A/I179A, and G187A/G189A, consistently showed retarded growth. However, with long-termed storage at 4°C after 30°C incubation was completed, colonies of these strains reached the size close to that of wild-type strains. (B) Growth curves of four representative clones grown in minimal medium. (C) Transcriptional activation potency of Gcn5p derivatives. gcn5 null cells were first double-transformed with the GAL4–VP16 and β-gal expression constructs (both are 2 μ plasmids) followed by different GCN5 alleles (CEN/ARS minichromosomes). The β-gal expression was measured from exponentially growing cells and presented here as the percentage of that activated by wild-type Gcn5p; each mutant has been assayed independently three to ten times. The relatively wide range of standard errors of certain samples is at least partially attributable to the copy number variation of the GCN5 plasmid (data not shown).

Open in a separate window

Figure 3

Overproduction of functional Gcn5p leads to hyperacetylation of histones in vivo. (A) Western analysis of overproduced Gcn5p. Yeast whole cell extracts (20 μg) prepared from strains overproducing various Gcn5p derivatives were resolved by SDS-PAGE (left) and blotted for Western detection with Gcn5p antibodies (right). Note that the endogenous Gcn5p was in a low abundance under our standard Western conditions. The overproduced GCN5 alleles are as follows: (lane 1) Wild type; (lane 2) vector; (lane 3) F221A; (lane 4) Y244A/E245A; (lane 5) G187A/G189A; (lane 6) I174A; (lane 7) L192A; (lane 8) G225A; (lane 9) T248A; (lane 10) wild type. Although there is a certain degree of expression variation of different Gcn5p derivatives, this variation is not likely sufficient to account for the dramatic difference in in vivo activity displayed by each allele. (B) Histones H3 and H4 are hyperacetylated in the presence of overexpressed functional Gcn5p derivatives. Purified core histones were resolved by 15% Triton–acetic acid–urea gel electrophoresis and visualized by silver staining. Acetylation ladders are marked on the side. Samples shown (left and right) were electrophoresed on separate gels. The same preparation of wild-type and vector samples were included on both gels for reference. Note that all three class 1 mutants showed a lower degree of H3 acetylation and subtle, but readily visible, effects on H4 acetylation.

To provide a more quantitative assessment of the growth defects of our gcn5 mutants, we compared the growth curves of four representative strains cultivated in the liquid minimal medium (see Fig. ​Fig.2B).2B). Consistent with the differential colony sizes observed on plates, the lack of functional GCN5 caused a significant lag of doubling time (cf. wild type, 132±2 min vs. vector control, 179±5 min and L192A, 126±5 min vs. F221A, 165±2 min). In addition, the gcn5Δ and F221A mutants appear to enter stationary phase at lower cell densities. Apparently, the combination of these two phenotypic defects associated with gcn5 loss-of-function mutations results in the significantly smaller colonies of these mutants shown in Figure ​Figure22A.

Growth defects of gcn5 mutants correlate with defects in transcriptional activation

A second functional test for Gcn5p is to measure the expression level of lacZ driven by the UASgal–CYC1 cis-acting element in vivo. In this construct, lacZ reporter gene expression is activated by Gal4–VP16 chimeric activator and requires functional Gcn5p and other Ada proteins (Candau et al. 1997; Wang et al. 1997). Gal4–VP16 activator and lacZ reporter constructs were cotransformed into a gcn5Δ strain followed by a second transformation with low-copy plasmids bearing specific GCN5 alleles. β-Galactosidase activity was then measured, and the results are summarized in Figure ​Figure22C.

It is clear that wild-type Gcn5p activated lacZ expression efficiently. As expected, F221A, the most affected mutant (see Fig. ​Fig.2A)2A) was unable to activate lacZ much above that by the vector alone control. Moreover, in complete agreement with each mutant’s ability to rescue growth in minimal medium (Fig. ​(Fig.2A),2A), all mutants that showed gcn5⁻ phenotypes activated lacZ less efficiently than those that supplied GCN⁺ function. Noteworthy is the G187A/G189A double mutant that exhibited the most subtle visible gcn5⁻ growth defect in the growth complementation assay. In the lacZ expression assay, this allele appears to represent a “threshold” activity. Mutants expressing lacZ at a higher level all function well in rescuing growth under amino acid starvation, whereas others that exhibit lower transcriptional activation cannot fully complement growth in minimal medium. GCN5 mutants that we have assayed for in vivo functional integrity are listed in Table ​Table1.1. Alleles in the first class show reproducible defects in vivo, whereas class 2 includes all mutants that functionally complement gcn5Δ phenotypes in vivo. Because these two in vivo functional assays (growth under amino acid starvation and UASgal–CYC1–lacZ expression) reflect transcription activated by at least two separate transcription activators (Gcn4p and Gal4–VP16, respectively), we conclude that Gcn5p-dependent activation is likely to use a common mechanism (i.e., histone acetylation at promoters; see below).

Overproducing functional Gcn5p leads to histone hyperacetylation

To provide in vivo evidence that GCN5 encodes a histone acetyltransferase, we examined GCN5 dosage-dependent changes in histone acetylation. Consistent with the finding that genes under the control of Gcn5p are likely to occupy a small portion of the yeast genome (Pollard and Peterson 1997), we find that deleting GCN5 does not reveal any significant changes in steady-state histone acetylation patterns (data not shown). To circumvent this problem, we overexpressed wild-type and mutant Gcn5p (Fig. ​(Fig.3A),3A), and analyzed the acetylation status of individual histones by Triton–acetic acid–urea gel electrophoresis, a gel system capable of separating histones based on the number of acetyl moieties that each histone isoform contains.

As shown in Figure ​Figure3B,3B, characteristic “ladders” indicative of increasing levels of acetylation are seen in both H3 and H4, and to a lesser extent, H2A and H2B (not shown) when wild-type Gcn5p is overproduced. In the vector control sample, only a very small percentage of H3 is acetylated and most of the acetylated H4 exists in a monoacetylated isoform. However, when histones, isolated from strains overproducing various Gcn5p mutants, were analyzed under the same condition, the ability of each allele to acetylate histones correlates precisely with the ability to supply normal GCN5 functions in vivo. For example, F221A, Y244A/E245A, and G187A/G189A (as well as Y220A and G239A/Y244I; data not shown), mutants that exhibit clear in vivo defects (see Fig. ​Fig.2),2), all fail to produce high levels of H3 and H4 acetylation (gel 1; Fig. ​Fig.3B).3B). Compared with control histones from wild-type Gcn5p overproduction strain, underacetylated histones H3 and H4, as indicated by the lower intensity of acetylated isoforms, are associated with these three gcn5 mutants.

Histones from several mutants that function nearly normally in our in vivo assays were also examined; these histones all show comparable levels of hyperacetylation to those generated by wild-type Gcn5p (gel 2; Fig. ​Fig.3B).3B). The three alleles shown in gel 2 (i.e., I174A, G225A, and T248A) were selected because they function normally in vivo (see Fig. ​Fig.2)2) despite an apparent reduction in in vitro HAT activity. Thus, in vivo overproduction of such mutants provides a good indicator as to whether the HAT activity can be restored in vivo, and if so, whether the GCN⁺ phenotype is correspondingly maintained. Together with transcriptional assay and growth complementation tests, these results suggest strongly that Gcn5p is indeed a HAT in vivo and that histones are physiological substrates of this enzyme. Most important, our results are consistent with Gcn5p HAT activity being required for transcriptional activation of target genes (see below).

HIS3 activation mediated by Gcn5p is linked to histone hyperacetylation

The above data suggest that maintaining the HAT activity in vivo is a prerequisite to transcriptional activation. To link directly the HAT activity of Gcn5p to the activation of specific target genes, we performed chromatin immunoprecipitation (ChIP) experiments (Braunstein et al. 1993, 1996) with low-copy GCN5 alleles, to evaluate the status of nucleosomal histone acetylation. We first analyzed nucleosomal acetylation in the promoter region of HIS3 gene whose expression requires a functional Gcn5p. Yeast gcn5Δ strains bearing various GCN5 alleles (CEN/ARS, GCN5 5′ and 3′ native sequences) were starved for amino acids, including histidine, to induce HIS3 expression. After formaldehyde cross-linking and chromatin solubilization, sonication-sheared oligonucleosomes were incubated with acetylation-specific H3 antibody for coimmunoprecipitation of DNAs that are associated with hyperacetylated H3 (and likely other core histones). After reversal of cross-links, the DNA was analyzed by slot-blot hybridization to determine whether sequences of interest were enriched in the hyperacetylated H3 fraction under a given condition.

Derepression of HIS3 and some other amino acid biosynthesis genes requires functional Gcn5p. As shown in Figure ​Figure4B,4B, Gcn5p is required for both constitutive and activated expression of HIS3 gene, a phenomenon reported earlier by others (Georgakopoulos and Thireos 1992). For constitutive expression, the absence of functional Gcn5p leads to a twofold decrease in transcription. Under conditions of amino acid starvation, where HIS3 expression is fully derepressed, the lack of functional Gcn5p (i.e., with the null or F221A allele) causes a threefold decrease of the transcriptional output of this gene. Importantly, in excellent agreement with the above Northern data, histone H3 is hyperacetylated when HIS3 is activated by functional Gcn5p (Fig. ​(Fig.4C).4C). Association of the HIS3 promoter with hyperacetylated nucleosomes increases threefold in wild-type and L192A alleles of GCN5⁺ strains under repressive conditions. Furthermore, with amino acid starvation, hyperacetylation of the HIS3 promoter increases sixfold in the presence of wild-type or L192A Gcn5p. In contrast, in either a repressive or derepressive situation, the F221A mutant behaves similarly to the vector control (i.e., this allele cannot provide normal Gcn5p functions in either transcriptional activation or histone acetylation).

Open in a separate window

Open in a separate window

Figure 4

Hyperacetylation of nucleosomal H3 is linked to the chromosomal copy of HIS3 activated by functional Gcn5p. (A) Source of the probes. The ORF probe was used only for Northern blot analyses. (B) Northern blot analysis of HIS3 transcription. The ratio of HIS3 expression, after normalization with ACT1 mRNA, was 1 : 1.8 : 2.5 : 0.9 : 6.0 : 19.4 : 19.3 : 8.2 (lanes 1–8). The generally lower expression of ACT1 under amino acid starvation is probably attributable to the complete lack of amino acids in the minimal medium during the 6-hr incubation before RNA extraction. (C) Chromatin IP results. Yeast cells were processed for immunoprecipitation of solubilized chromatin fragments using anti-H3.AcLys(9/14) antiserum. DNAs coprecipitated in these nucleosomal complexes were purified and analyzed by slot-blot hybridization. Relative IP efficiency (counts of the IP divided by those of the input) was obtained setting the value of the vector control slot as 1 in each hybridization result. The normalized ratio was derived by dividing the IP efficiency of each sample to that obtained when total genomic DNAs were used as a probe. (D) Chromatin IP using the antisera against unacetylated or hyperacetylated H3 shows comparable IP efficiency in different GCN5 backgrounds. Yeast extracts were subjected to ChIP using antisera against unacetylated (un.) or acetylated (ac.) H3 and the resultant slot-blots were first probed with HIS3 promoter sequence as above, followed by control probes, including ACT1 promoter and total genomic DNA, which showed essentially identical IP efficiency seen in Fig. ​Fig.4C4C (data not shown).

To rule out the possibility that the observed difference in IP efficiency by hyperacetylated histone H3 antibody among different GCN5 alleles is attributable to variation of formaldehyde cross-linking efficiency or to a strain-specific accessibility of the HIS3 locus to the antibody, we conducted the ChIP experiments using an antibody against unacetylated histone H3 (Braunstein et al. 1993). As shown in Figure ​Figure4D,4D, consistent with the notion that transcriptional repression is associated with histone underacetylation, the HIS3 promoter sequence is enriched in the unacetylated histone fraction when GCN5 is either deleted (vector) or replaced with the F221A allele (cf. un. and ac. columns in Fig. ​Fig.4D).4D). The GCN5 allele-specific, differential enrichment of HIS3 promoter by a specific antiserum strongly suggests that there is a dynamic change in nucleosomal acetylation within Gcn5p-mediated transcriptional activation.

Conserved amino acids are critical for Gcn5p functions

Among the mutations we have created and analyzed, F221A is the most severely damaging allele. It is inactive as a HAT in vivo and in vitro, and it fails to support normal Gcn5p-mediated transcriptional activation. Because this mutant is able to interact with Ada2p normally (not shown), as reflected by in vivo coimmunoprecipitation assays, it is most likely that the in vivo defects are attributable to its inability to catalyze histone acetylation. Consistent with this hypothesis, we observed a strong dominant negative phenotype of this allele with overproduction (data not shown), which suggests that this HAT catalytic mutant, with otherwise unaffected functions in vivo, can titrate away functional interacting proteins and hence brings about a gcn5⁻ phenotype. It is interesting that severe phenotypes are linked to this residue because phenylalanine is not likely to serve as an active site residue, although we point out that the catalytic mechanism used by Gcn5p is not known.

Some of our mutations in yGcn5p subdomains II and III, the regions likely to form an acetyl-CoA-binding motif, have been tested previously in human spermidine/spermine amino-acetyltransferase and yeast MAK3 (maintenance of killer) protein amino-terminal acetyltransferase (e.g., C177, I179, Q184, G187, Y188, G189, D214, F221, G225, and F226 of yGcn5p), and in general our in vitro HAT assay results are in excellent agreement with these studies (Lu et al. 1996). Domain IV, on the basis of our analyses, appears to be conserved only within the HAT family members. We speculate that this domain may be involved in histone substrate (amino-terminal tails) binding, although this hypothesis remains to be examined.

The in vitro HAT assays clearly show that Gcn5p, expressed and purified from bacteria, is sensitive to sequence alterations. There are two likely reasons for this observation. First, yeast Gcn5p are known to form complexes with several other subunits in vivo and in this form is better able to acetylate nucleosomal substrates (Grant et al. 1997; Pollard and Peterson 1997). None of these non-Gcn5p components were included in our in vitro HAT assays, and attempts to reconstitute nucleosomal HAT activity from individual components have so far been unsuccessful (Grant et al. 1997 and our unpublished data). It is likely that the association of accessory proteins with a Gcn5 mutant with, for example, an altered conformation could overcome or neutralize the mutation. Second, mutant Gcn5p may be less stable than the wild type, and therefore not able to fold during expression or purification. Others have observed mutant enzymes that maintain their activity in vivo, despite their altered stability in vitro, and in one case the mutations were made in the hydrophobic core of an enzyme (Serrano et al. 1992; Axe et al. 1996). One implication of our mutagenesis experiments is that many of the conserved residues in domains II and III form part of the conserved hydrophobic core of Gcn5p, and are not residues important specifically for catalytic function or interaction with Gcn5p-binding partners.

In a related study (see Wang et al. 1998), a number of alanine scan mutants in Gcn5p, many overlapping with our point mutants (including F221A), have been tested for both in vivo transcriptional activation and in vitro HAT activity. In general, our in vivo Gcn5p overproduction and histone hyperacetylation results agree well with their results assaying in vitro HAT activity using Gcn5p-containing complexes. These results are consistent with our notion that many mutants, when assayed in the recombinant and monomeric form, show in vitro enzymatic defects that can be “rescued” in vivo presumably by functioning in the context of large, multisubunit Gcn5/Ada or SAGA-type HAT complexes.

Gcn5p HAT activity is required for transcriptional activation

Our data suggest that recruitment of HATs to the promoter, possibly through interactions with activators, other coactivators, and components of the basal transcription machinery, leads to hyperacetylation of histones in the promoter region and transcriptional activation of target genes (Brownell et al. 1996; Wolffe and Pruss 1996; Wade and Wolffe 1997). In vivo phenotypes of Gcn5p mutants are tightly linked to the “potency” of each Gcn5p derivative to activate transcription of downstream genes and to acetylate histones. Importantly, the observed nucleosomal hyperacetylation at promoter region holds for two different transcription activators tested in this study. Collectively, these data provide compelling evidence that the HAT activity of Gcn5p is critically required for transcriptional activation of its target genes.

Our study also suggests that hyperacetylated histones may be confined to the promoter region of at least some genes in yeast (Fig. ​(Fig.5).5). This result differs from previous reports showing that the entire transcription units of some actively expressed loci are associated with hyperacetylated nucleosomes (Hebbes et al. 1994; O’Neill and Turner 1995). Such a discrepancy may be explained, at least in part, by the specific histone or site of acetylation that are under investigation. In our study, we used an antiserum that preferentially recognized a highly preferred site of Gcn5p-mediated acetylation in H3 (Lys-14; see Kuo et al. 1996). Depending on the substrate preference of the recruited HAT, histone hyperacetylation associated with transcriptional activation may escape detection from an antiserum with a different specificity.

Confinement of hyperacetylation to the promoter region is consistent with previous findings that tethering LexA–Gcn5p near a promoter can activate transcription (Georgakopoulos et al. 1995; Candau et al. 1997), suggesting that a “stationary” HAT is sufficient for facilitating transcription initiation by the basal transcription machinery on a nucleosomal template. It remains a formal possibility that another HAT, distinct from Gcn5p, conducts elongation-associated histone acetylation that spreads the hyperacetylation signal throughout the transcription unit. Recent reports of Gcn5p-independent HAT complexes in yeast (Grant et al. 1997) as well as the existence of an essential HAT in yeast (E.R. Smith, A. Eisen, J.C. Lucchesi, and C.D. Allis, in prep.) make this an intriguing possibility.

It is noteworthy that the three- to sixfold increase of association of HIS3 promoter with hyperacetylated histone H3 mediated by functional Gcn5p only reflects the minimal effects of Gcn5p in vivo. In the presence of functional Gcn5p [wild type, L192A, and G225A (not shown)], we observed consistently higher amounts of DNA associated with hyperacetylated histones even when ACT1 promoter (not shown) or bulk genomic DNA (Figs. ​(Figs.44 and ​and5)5) was used to probe the coprecipitated DNA sequences. It is tempting to speculate that Gcn5p is one of the major histone acetyltransferases by which histones are acetylated at specific lysine residues recognized by our antibody [anti-H3.AcLys(9/14)]. If this is the case, the relative IP efficiency we calculated (Figs. ​(Figs.44 and ​and5)5) would have been significantly underestimated.

Enzyme assays

Recombinant Gcn5p was expressed and purified from E. coli as described before (Kuo et al. 1996) with the following modifications: 20 mm imidazole was included in the Ni–NTA agarose binding reaction. Bound Gcn5p was washed three times for 10 min each [50 mm NaPi (pH 6.0), 300 mm NaCl, 10% glycerol] followed by a 5-min wash with the same buffer at pH 5.5. Gcn5p was then eluted with a 5-min incubation with pH 4.0 elution buffer; this elution step was repeated once before eluates were pooled. One-twentieth of the volume of 1 m Na₂HPO₄ was then added immediately to neutralize the pH. Gcn5p remains active in this buffer for at least several months when stored at −80°C.

HAT assays were conducted in 20-μl reactions containing 10-μg of free chicken histones, 0.1 μCi of ³H-acetyl CoA (4–6 Ci/mmole), 50 mm Tris HCl (pH 8.0), 10% glycerol, 1 mm DTT, and 10 mm n-butyrate. The reaction was incubated at 30°C for exactly 20 min. Approximately 15 ng of wild-type Gcn5p saturates the reaction; typically, 10 and 20 ng of each Gcn5p derivative was used to obtain the relative HAT activities. Each mutant has been purified and assayed independently for HAT activity at least three times.

Yeast histone preparation and analyses

Gcn5p overproduction was done by growing yeast strains bearing pMK144 series plasmids in synthetic complete −Ura, −Leu medium (to increase the copy number of pMK144) containing 200 μm of Cu₂SO₄ (to fully induce Gcn5p expression) until mid-log phase. Yeast histones were purified according to Edmondson et al. (1996).

The 15% Triton X-100–acetic acid–urea polyacrylamide gels (typically, 1.5 mm thick, 13×14 cm²) were used to separate yeast histones according to Mullen et al. (1989) except that gels were run at 250 V for 16 hr to obtain good resolution of isoforms of H2A, H2B, and H4. For better resolution of H3, the gel was run for 20 hr. Proteins were visualized by silver staining.

Transcriptional and growth complementation assays

Expression of lacZ and growth complementation assays were as described by Candau et al. (1997). gcn5Δ cells (yMK703) were transformed with each mutant derivative of Gcn5p on pRS414-GCN5. For colony size comparison, log-phase transformants in complete drop-out medium were plated to synthetic minimal medium and grown at 30°C for 3 days. In addition, late log-phase cells were inoculated into liquid minimal medium for growth curve analyses. For β-gal expression and measurement, yMK703 was transformed with GAL4–VP16_FA expression vector, UASgal–CYC1–lacZ plasmid, and different alleles of Gcn5p mutants. Log-phase transformants were harvested for β-gal measurement. Three to ten of each mutant transformants were analyzed.

We thank Shelley Berger for communicating data before publication and for providing us with strains, plasmids, and antisera. We acknowledge Ian Macreadie for sharing plasmids for overproducing Gcn5p. We are grateful to Martin Gorovsky, Beth Grayhack, and Sharon Roth for critical comments on this manuscript. M.H.K. thanks Beth Grayhack for constant encouragement and suggestions throughout this work and Martin Gorovsky for introducing him to the field of chromatin studies. This work is supported by the National Institutes of Health (NIH) (GM53512) to C.D.A. M.E.A.C. acknowledges support from The American Heart Association, NIH (Shannon Award GM52100), and a Cellular and Molecular Biology NIH Pre-doctoral Training Grant (P.J.).

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.