GST pull-down, Western blot, and Northern blot analyses.

Bacterial extracts containing GST proteins were prepared from transformants of Escherichia coli strain BL21 (30). Yeast whole-cell extracts (WCEs) were prepared (138) and GST pull-down assays were performed (30) as described previously. Samples were resolved on 4 to 12% bis-Tris NuPAGE gels (Invitrogen) with MOPS (morpholinepropanesulfonic acid) buffer according to the manufacturer's protocol. Proteins were transferred to nitrocellulose membranes and detected by immunoblotting. Rabbit polyclonal antisera have been described for Srb7p (47), Swp73p (15), Caf16p (75), and Taf40p (59). Purified rabbit polyclonal Sth1p antibodies were a generous gift from Brad Cairns. Myc-tagged proteins were detected with mouse monoclonal c-Myc (9E10) antibodies from Santa Cruz Biotechnologies (catalog no. sc-40) at a 1:500 dilution.

Yeast cultures for Western analysis of HA₃-Gcn4p were grown and induced as described above for the reporter gene assays. Extracts were prepared as described previously (29) except that additional protease inhibitors were added (Roche complete inhibitor used at a 1× concentration), and the cells were lysed in 14-ml tubes by using 10 cycles of 30 s of vortexing with 1.5 min of cooling on ice between cycles. Protein concentration was determined with a bicinchoninic acid protein assay kit (Pierce). Proteins were resolved on 10% bis-Tris NuPAGE gels (Invitrogen) in MOPS buffer at a constant voltage of 200 V for 1.5 h with the gel chamber submerged in ice. After transfer of the proteins to nitrocellulose, the membranes were cut and probed with rabbit polyclonal HA antibodies from Santa Cruz Biotechnologies (probe Y-11; catalog no. sc-805) at a dilution of 1:500 to detect HA₃-Gcn4p and with Gcd6p antibodies (14).

Cultures were grown and induced for Northern analysis as described above or in some cases by using the inducing conditions described below for ChIP assays. Total RNA was extracted and subjected to Northern analysis as described previously (92). Probes were prepared from PCR amplified yeast genomic DNA or as restriction fragments from plasmid clones by random prime labeling using the RediPrime II system (Invitrogen).

The array of coactivator subunits required for activation by Gcn4p varies at different promoters.

In an effort to confirm the conclusions reached above, we assayed a subset of the mutants for expression of a second Gcn4p-dependent reporter containing the 5′ noncoding region of HIS3 from position −450 to −3 (relative to the ATG start codon) fused to GUS coding sequences. As above, at least three independent transformants of each strain harboring the reporter plasmid were assayed for β-glucuronidase activity under inducing and noninducing conditions, and the mean values are listed in Table ​Table3.3. In all cases, the standard errors were less than 20% of the mean values. The HIS3-GUS reporter is induced three- to eightfold by SM treatment of WT cells (data not shown), and ~90% of its expression under these conditions is dependent on Gcn4p (Table ​(Table3).3). The residual HIS3-GUS expression in the gcn4Δ mutant can be attributed to the AT-rich element that confers Gcn4p-independent promoter activity (122). Because HIS3-GUS expression is ~5-fold more dependent on Gcn4p under inducing than noninducing conditions (Table ​(Table3),3), by comparing the relative impairment of reporter expression under these conditions, we could evaluate whether the coactivator mutants are defective for activation of HIS3-GUS by Gcn4p. The data are coded according to criteria similar to those employed above for the UAS_GCRE-CYC1-lacZ reporter (see the legend to Table ​Table33 for details). The results suggest that Hpr1p, the Cdc73p subunit of the Paf1 complex, RSC subunit Rsc1p, and multiple subunits of SAGA, SRB/MED, and CCR4/NOT are all required for full activation of the HIS3 promoter by Gcn4p (Table ​(Table33).

TABLE 3.

HIS3-GUS expression phenotypes of cofactor mutants

Relevant genotype	Cofactor complex	β-Glucuronidase activity (% of WT) fora:
		Uninduced HIS3-GUS	Induced HIS3-GUS
gcn4Δ	Activator	55	11
mbf1Δ	NAb	76	27
ada2Δ	ADA, SAGA	86	16
ada3Δ	ADA, SAGA	143	20
gcn5Δ	ADA, SAGA	29	6.4
spt8Δ	SAGA	289	84
spt3Δ	SAGA	141	29
ada1Δ	SAGA	149	20
ada5Δ	SAGA	235	32
spt7Δ	SAGA	46	9.1
tfg3Δ	Multiplec	113	58
rsc1Δ	RSC	15	4.2
rsc2Δ	RSC	23	49
swi2Δ	SWI/SNF	523	169
swi3Δ	SWI/SNF	43	25
snf5Δ	SWI/SNF	363	117
snf6Δ	SWI/SNF	147	82
snf11Δ	SWI/SNF	112	83
swp73Δ	SWI/SNF	437	129
cdc73Δ	Paf1 complex	117	42
hpr1Δ	Paf1 complex, THO/TREX	13	3.3
ccr4Δ	CCR4-NOT, Paf1 complex	25	10
not3Δ	CCR4-NOT	58	52
caf16Δ	CCR4-NOT	76	40
not4Δ	CCR4-NOT	67	51
caf40Δ	CCR4-NOT	58	33
caf130Δ	CCR4-NOT	56	46
caf4Δ	CCR4-NOT	86	54
caf17Δ	CCR4-NOT	97	64
dbf2Δ	CCR4-NOT	40	20
dhh1Δ	CCR4-NOT	16	12
not5Δ	CCR4-NOT	67	39
caf1Δ	CCR4-NOT	53	10
srb11Δ	SRB/MED, CCR4-NOT	140	77
srb9Δ	SRB/MED, CCR4-NOT	53	12
srb10Δ	SRB/MED, CCR4-NOT	227	73
pgd1Δ	SRB/MED	469	259
med2Δ	SRB/MED	97	43
rox3Δ	SRB/MED	92	21
sin4Δ	SRB/MED, Paf1 complex	253	119
ga111Δ	SRB/MED, Paf1 complex	145	55

Open in a separate window

^aAt least three independent transformants of each strain harboring the reporter plasmid pKN7 were grown under inducing and noninducing conditions (presence and absence of SM) as described in Materials and Methods, and β-glucuronidase activity was assayed in the cell extracts. Mean values were calculated from results that yielded standard deviations of 20% or less and expressed as percentages of the corresponding values measured in the WT parental strains. Results in italics are 200% of the WT value or greater. Underlined results are <68% of the WT value under inducing conditions and show reductions relative to the WT that are 2-fold (boldface with an asterisk), 1.5-fold (boldface only), or less than 1.5-fold greater under inducing than noninducing conditions.

^bNA, not applicable.

^cTfg3p is present in the SWI/SNF, TFIID, NuA3, and TFIIF complexes.

Interestingly, the spt3Δ mutant was impaired for HIS3-GUS induction (Table ​(Table3),3), even though it showed WT induction of UAS_GCRE-CYC1-lacZ (Table ​(Table1).1). Hence, the requirement for this SAGA subunit in activation by Gcn4p seems to vary between these two promoters. This last phenomenon also holds for the CCR4/NOT subunit Dbf2p, the SRB/MED subunit Srb9p, Rsc1p, and Mbf1p. The SRB/MED subunit Pgd1p showed the opposite behavior, as pgd1Δ cells were strongly impaired for UAS_GCRE-CYC1-lacZ induction (Table ​(Table1)1) but showed greater-than-WT induction of HIS3-GUS. Thus, Pgd1p may have a role in negative control that overrides its importance as a coactivator for Gcn4p at HIS3-GUS. This conclusion is consistent with previous results indicating dual positive and negative functions for Pgd1p at other promoters (101). Whereas the rsc1Δ mutation had a greater effect than rsc2Δ on HIS3-GUS expression (Table ​(Table3),3), the opposite was true for the UAS_GCRE-CYC1-lacZ reporter (Table ​(Table1).1). There are two forms of RSC that contain either Rsc1p or Rsc2p (16). Thus, it seems that Gcn4p utilizes both forms of RSC, but the relative importance of the Rsc1p- and Rsc2p-containing complexes depends on the promoter.

It was also important to analyze the effects of coactivator mutations on induction of authentic mRNAs by Gcn4p. Accordingly, we prepared total RNA from mutant and WT strains and conducted Northern analysis using probes specific for the Gcn4p target genes SNZ1, HIS4, ARG1, and ILV2 (94) and for ACT1, analyzed as an internal control. RNA was prepared from two independent cultures of each strain, and the Northern signals for the Gcn4p-regulated mRNAs were quantified and normalized to the ACT1 signals. Typical Northern data are presented in Fig. ​Fig.44 for the duplicate determinations of selected mutants to illustrate the reproducibility of these data. The mean normalized mRNA levels for the entire set of mutants are given in Table ​Table44.

Open in a separate window

FIG. 4.

Northern analysis of authentic Gcn4p target genes in a typical subset of deletion mutants. Total RNA was isolated for each strain under the inducing and noninducing conditions described in Table ​Table1,1, and equal amounts of RNA were subjected to Northern analysis, probing for ACT1, ARG1, HIS4, ILV2, and SNZ1 mRNAs. Adjacent lanes contain RNA samples isolated from two independent cultures for each strain. The hybridization signals were quantified with a PhosphorImager (Molecular Dynamics, with ImageQuant 5.2 software), and the values obtained for SNZ1 (A), HIS4 (B), and ARG1 (C) were normalized to the corresponding ACT1 signals. The resulting ratios calculated for the mutant strains were normalized to the ratio measured in the WT, and the normalized ratios are plotted in histograms beneath the corresponding lanes of each blot.

TABLE 4.

Northern blot analysis of cofactor mutants

Relevant genotype	Cofactor complex	Mean normalized signal ratio (% of WT)a
		SNZ1, induced	HIS4		ILV2		ARG1
			Uninduced	Induced	Uninduced	Induced	Uninduced	Induced
gcn4Δ	Activator	12b,c	67b,c	31b,c	78c	32c	51b,c	14b,c
mbf1Δ	NAd	63c		73c		68c		77c
ada2Δ	ADA, SAGA	70b	47b	43b			207b	96b
gcn5Δ	ADA, SAGA	124b,c	105b,c	98b,c	38c	71c	313b,c	143b,c
ada1Δ	SAGA	47c	90c	29c	69c	59c	915c	101c
ada5Δ	SAGA	45c		27c		45c		106c
spt3Δ	SAGA	68b	69b	52b			130b	86b
spt7Δ	SAGA	41b					458b	121b
swi2Δ	SWI/SNF	44b,c	139b	206b		109c		118b,c
swp73Δ	SWI/SNF	30b		32b
rsc1Δ	RSC	67c	118c	88c	88c	70b,c	198c	67c
rsc2Δ	RSC	144b,c	242c	63c	99c	124c	>600c	111c
not5Δ	CCR4-NOT	155c	118c	74c	85c	79c	91c	72c
caf1Δ	CCR4-NOT	76b,c	103c	77c	51c	62c	71c	48c
dhh1Δ	CCR4-NOT	65b,c	135b,c	102b,c	96c	56c	99c	41c
srb5Δ	SRB/MED	126b,c	307c	118c	152c	92c	>300b,c	148b,c
pgd1Δ	SRB/MED	62c	144b,c	67b,c	114c	72c	474c	57c
rox3Δ	SRB/MED	38b,c	129b,c	67b,c	83c	62c	105b,c	52b,c
sin4Δ	SRB/MED, Paf1	68c	125c	94c	91c	86c	379c	85c
gal11Δ	SRB/MED, Paf1	55c	78b,c	102b,c	118c	78c	213b,c	97b,c
med2Δ	SRB/MED	65c	76c	84c	168c	88c	171c	83c
cdc73Δ	Paf1 complex	98c		84c		77c		100c
ccr4Δ	CCR4-NOT, Paf1	57b,c	70b,c	49b,c	60c	51c	70c	44c
hpr1Δ	Paf1, THO/TREX	49b,c	59b,c	50b,c	54c	125c	65c	66c
gcn4Δ/gcn4Δ	Activator	16c	57c	23c	81c	26c	54c	7c
paf1Δ/paf1Δ	Paf1 complex	97c	327c	126c	120c	136c	2,770c	149c
snf5Δ/snf5Δ	SWI/SNF	75c	93c	85c	261c	162c	500c	104c

Open in a separate window

^aFor each value shown, at least two independent cultures of each strain were grown under inducing or noninducing conditions (presence or absence of SM) as described in Materials and Methods, and equal amounts of isolated total RNA were subjected to Northern analysis. The hybridization signals for SNZ1, HIS4, ILV2, and ARG1 were normalized to corresponding ACT1 signals. The resulting ratios calculated for the mutant strains were normalized to the ratio measured for the WT, and the mean normalized ratios are shown as a percentage of the WT value. The mean values were calculated from results that yielded standard errors of 30% or less. Results in italics are 200% of the WT value or greater. Underlined results are <68% of the WT value under inducing conditions and show reductions relative to WT that are ~2-fold (bold face type with an asterisk), 1.5-fold (bold face type), or less than 1.5-fold greater under inducing than noninducing conditions.

^bInduction as per the ChIP assay.

^cInduction as per the reporter assay.

^dNA, not applicable.

SNZ1 mRNA is an interesting case because it resembles the UAS_GCRE-CYC1-lacZ construct in showing very low levels of expression in noninducing conditions and a strong dependence on Gcn4p under inducing conditions (94). As shown in Fig. ​Fig.4A4A and Table ​Table4,4, mutations in multiple subunits of SAGA, SWI/SNF, and SRB/MED, Rsc1p (RSC), Dhh1p (CCR4/NOT), Ccr4p (CCR4/NOT and Paf1), Hpr1p (Paf1 and THO/TREX), and Mbf1p all lowered the induced level of SNZ1 mRNA to below 68% of WT levels. Surprisingly, deletion of GCN5 led to slightly greater than WT levels of SNZ1 mRNA. Although WT transcriptional induction of SNZ1 by Gcn4p requires the same coactivator complexes needed for high-level expression of the reporter constructs, the subunits most critically required differ somewhat among the SNZ1, HIS3-GUS, and UAS_GCRE-CYC1-lacZ promoters.

Efficient induction of HIS4 mRNA was dependent on multiple subunits of SAGA and SRB/MED and also required Rsc2p, Swp73p (SWI/SNF), Ccr4p, and Hpr1p (Paf1p and THO/TREX) (Fig. ​(Fig.4B4B and Table ​Table4).4). Surprisingly, deleting the ATPase subunit of SWI/SNF (Swi2p) led to greater-than-WT HIS4 expression, whereas inactivation of the Swp73p subunit significantly reduced HIS4 induction. It was shown recently that Swi2p can function as a repressor independently of other SWI/SNF subunits (82). Thus, perhaps Swi2p acts in this manner as a repressor at HIS4 and also in conjunction with other SWI/SNF subunits in activation and the former activity has the greater impact on HIS4 transcription. Again, the array of subunits of SAGA, SRB/MED, CCR4/NOT, SWI/SNF, and RSC critical for activation by Gcn4p differs somewhat between SNZ1 and HIS4: Swi2p, Rsc1p, Dhh1p, Med2p, Sin4p, and Gal11p are required at SNZ1 but not at HIS4, whereas Ada2p and Rsc2p are more important at HIS4 than at SNZ1.

ILV2 mRNA induction was substantially impaired by mutations in several subunits of SAGA and CCR4/NOT and in MBF1 and by deletion of ROX3 (SRB/MED) (Table ​(Table4).4). Thus, with the possible exception of SWI/SNF, high-level activation of ILV2 requires at least one subunit of the same coactivator complexes needed for activation of other Gcn4p-dependent promoters. We noted that a number of mutants with a SM^s phenotype displayed WT or greater levels of ILV2 mRNA (Table ​(Table4).4). Presumably, these mutants are defective for induction of one of the other four ILV genes rather than ILV2 (58).

The coactivator requirements for activation of ARG1 transcription are unique. As for certain other Gcn4p target genes, high-level induction was impaired by mutations in multiple subunits of CCR4/NOT and SRB/MED, Rsc1p, and Hpr1p (Table ​(Table44 and Fig. ​Fig.4C).4C). However, ARG1 induction showed no significant dependence on any SAGA subunit. Moreover, nearly all of the SAGA mutations led to higher ARG1 expression under noninducing conditions (Table ​(Table4).4). The latter data are consistent with the finding that SAGA is required for arginine-specific repression of ARG1 by the ArgR/Mcm1p repressor complex (107). The fact that SAGA mutations did not produce higher-than-WT ARG1 expression under inducing conditions (Table ​(Table4)4) may indicate that activation of this promoter by Gcn4p is SAGA dependent, as observed for all other Gcn4p target genes. In this view, the SAGA subunit deletions have offsetting positive and negative effects on transcription by simultaneously impairing repression by ArgR/Mcm1p and activation by Gcn4p. Considering the other coactivator mutations that derepressed ARG1 mRNA under noninducing conditions (Table ​(Table4),4), we suggest that RSC, SWI/SNF, SRB/MED, and the Paf1 complex may also be required for ARG1 repression. The derepressing effect of the paf1Δ mutation on ARG1 mRNA is particularly striking.

Coactivator mutants with SM^s phenotypes generally do not have reduced levels of Gcn4p.

A mutation may impair the activation of Gcn4p target genes by lowering the induced level of Gcn4p. To distinguish such mutants from those truly defective in activation, we measured the steady-state levels of an HA epitope-tagged version of Gcn4p by Western analysis in each SM^s mutant. The relevant strains were transformed with a single-copy plasmid expressing HA₃-Gcn4p from the native promoter (CEN/GCN4-HA₃), and WCEs were prepared following growth in the presence of SM to induce HA₃-Gcn4p expression. Western analysis using anti-HA antibodies showed that, as expected, HA₃-Gcn4p levels increased dramatically on treatment of WT transformants with SM (data not shown). In 20 of 27 mutants tested, the levels of HA₃-Gcn4p were greater than or equal to those in the WT strain. An example of the data obtained for one such mutant (ccr4Δ) is shown in Fig. ​Fig.5A5A (see Table ​Table55 for results on all mutants).

Open in a separate window

FIG. 5.

Western analysis of Gcn4p levels in Gcn⁻ mutants. Single-copy plasmid p2382 (CEN) or high-copy-number plasmid pHQ1239 (2μm) harboring the GCN4-HA₃ allele was introduced into the WT and Gcn⁻ deletion strains, and the WT strain was also transformed with empty CEN or 2μm vector. (A and B) Extracts were prepared from two transformants of each strain induced with SM. Two amounts (20 and 50 μg) of total protein, labeled as relative protein (Rel. Prot.) amounts of 1 and 2.5, respectively, were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and probed with Gcd6p antibodies and anti-HA antibodies. Lanes 1 and 2 contain samples from the WT strain bearing the empty vector. All Western blotting was carried out two or more times, and representative results are presented for the ccr4Δ (A) and not5Δ (B) mutants. All results are summarized in Table ​Table5.5. (C) Most mutants with lowered levels of HA₃-Gcn4p, such as the not5Δ strain, maintain their SM^s phenotypes after transformation with the high-copy-number GCN4-HA₃ plasmid (2μm). The not5Δ and gcn4Δ mutants and the WT strain containing p2382 (CEN), pHQ1239 (2μm), or empty vector were replica plated to SC−ILV medium containing 1 μg of SM per ml and incubated at 25°C for 3 days.

In the remaining seven mutants, we saw a significant reduction in the levels of HA₃-Gcn4p relative to the WT strain (Table ​(Table5),5), ranging from 20% (dhh1Δ) to 40% (ada1Δ, ada5Δ/spt20Δ, bdf1Δ, hpr1Δ, not5Δ, and swi3Δ) of the WT level. To determine whether the SM^s phenotypes of this class resulted from low levels of the activator, we increased expression of HA₃-Gcn4p by introducing GCN4-HA₃ on a high-copy-number plasmid (2μm/GCN4-HA₃). In five of these mutants (ada1Δ, ada5Δ/spt20Δ, dhh1Δ, hpr1Δ, and not5Δ), the levels of HA₃-Gcn4p equaled or exceeded that in the WT strain containing single-copy GCN4-HA₃, and the SM sensitivity remained unchanged. An example of this behavior is illustrated in Fig. 5B and C for the not5Δ mutant. We conclude that the SM^s phenotypes of these five strains cannot be accounted for by reduced levels of Gcn4p and that Gcn4p activation function is impaired. It is also noteworthy that hpr1Δ cells showed no activation defect at ILV2, that dhh1Δ did not impair HIS4 induction, and that not5Δ cells showed normal induction of SNZ1 mRNA (Table ​(Table4).4). Thus, the reductions in Gcn4p levels observed in these three mutants are not great enough to impair the activation of all Gcn4p-dependent promoters. Transformants of the bdf1Δ and swi3Δ mutants harboring the 2μm/GCN4-HA₃ plasmid had levels of HA₃-Gcn4p exceeding that of the WT strain bearing CEN/GCN4-HA₃; however, their SM^s phenotypes were partially complemented (Table ​(Table5).5). Thus, the activation defects in the bdf1Δ and swi3Δ strains may result partly from reduced Gcn4p expression.

Mutations in enzymes affecting chromatin structure generally do not disrupt activation by Gcn4p.

Since Gcn4p requires a number of complexes involved in chromatin modification, it was possible that any perturbation of chromatin structure might affect activation by Gcn4p. To address this possibility, we examined mutants with deletions of other nonessential HATs, histone deacetylases and ATP-dependent chromatin-remodeling enzymes for growth in the presence of SM. Among the known HAT mutants, only the gcn5Δ strain showed sensitivity to SM (Fig. ​(Fig.6A).6A). Interestingly, deletion of the HAT subunit of SRB/MED, Nut1p, or the transcription Elongator complex, Elp1p, had no effect on SM resistance, as did deletions of two other members of Elongator, Elp2p and Elp3p (Fig. ​(Fig.6A6A and data not shown). To address the possibility of redundant contributions of different HATs to activation by Gcn4p, we generated double deletions of various HATs with gcn5Δ or nut1Δ and assayed growth on SM. Double mutants containing a deletion of GCN5 displayed phenotypes nearly identical to that of the gcn5Δ single mutant (Fig. ​(Fig.6B6B and data not shown). The gcn5Δ sas3Δ strain was not viable, consistent with previous findings (53). Combining deletions of various HATs with the nut1Δ mutation also did not reveal any additive SM^s phenotypes (Fig. ​(Fig.6B6B and data not shown). Although histone deacetylases are normally associated with transcriptional repression, some deacetylase mutants show decreased transcription of specific genes (11); however, we detected no SM sensitivity for any deacetylase mutant (Table ​(Table11).

Open in a separate window

FIG. 6.

Deletions of chromatin-modifying enzymes do not generally produce Gcn⁻ phenotypes (A to C). The indicated mutant and WT strains were tested for sensitivity to SM as shown in Fig. ​Fig.1A1A.

As Gcn4p requires both SWI/SNF and the highly related RSC complex for WT activation (Fig. ​(Fig.2B),2B), we assayed deletions of the nonessential subunits of other chromatin-remodeling complexes. As shown in Fig. ​Fig.6C,6C, we observed no SM sensitivity in mutants lacking nonessential subunits of the ISWI complexes (isw1Δ, isw2Δ, and itc1Δ) and Chd1p, which functions as a homodimer and can affect transcription both positively and negatively (129). Thus, the SM^s phenotypes of the SWI/SNF and RSC mutants reflect a specific requirement for these complexes in activation by Gcn4p rather than a general perturbation of chromatin structure.

Gcn4p interacts with the RSC and CCR4-NOT complexes dependent on hydrophobic residues in the activation domain.

Previously, we showed that SAGA, SWI/SNF, and SRB/MED in WCEs bind specifically to a full-length GST-Gcn4p fusion, dependent on clusters of hydrophobic amino acids in the activation domain required for activation by Gcn4p in vivo (29, 30, 54, 93). Here, we performed similar experiments probing for subunits in the complexes newly determined to be required for Gcn4p activity: RSC, CCR4-NOT, and the Paf1 complex. Sth1p, the catalytic subunit of RSC, coprecipitated with WT GST-Gcn4p but not with the mutant protein containing 10 alanine substitutions in four hydrophobic clusters (10Ala) that destroy activation by Gcn4p in vivo (54) (Fig. ​(Fig.7A,7A, lanes 3 and 4). Similar results were obtained by probing for Swp73p of SWI/SNF and Srb7p of SRB/MED. As described above, Rsc1p and Rsc2p form mutually exclusive RSC complexes. To determine whether Gcn4p interacts with both forms of RSC, we used extracts from the rsc1Δ and rsc2Δ strains in pull-down assays. The Sth1p subunit in both extracts was precipitated by the WT but not by 10Ala GST-Gcn4p (Fig. ​(Fig.7A,7A, lanes 7 and 8 and lanes 11 and 12, respectively). Moreover, pull-down assays using extracts from strains in which either Rsc1p or Rsc2p is tagged with 13 copies of the Myc epitope show that WT GST-Gcn4p can precipitate both Rsc1p and Rsc2p (Fig. ​(Fig.7B,7B, lane 3). We conclude that both forms of the RSC complex can interact specifically with the Gcn4p activation domain in vitro, consistent with our data showing that both forms of RSC are required for WT activation of various Gcn4p-dependent promoters (Fig. ​(Fig.44).

Open in a separate window

FIG. 7.

GST-Gcn4p interacts specifically with the RSC and CCR4-NOT complexes but not with the Paf1 complex in cell extracts. Equal amounts of GST, GST-Gcn4p (GST-WT), and GST-Gcn4p containing 10 alanine substitutions in the activation domain (GST-10Ala) were incubated with WCEs from yeast strains grown in YPD medium. The GST proteins were precipitated with glutathione Sepharose, resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and detected by Western analysis. (A) Equal amounts of WCEs from WT (BY4741), rsc1Δ (strain 4686), and rsc2Δ (strain 5266) strains were incubated with the GST proteins. Blots were probed with antibodies against the proteins listed on the left of the upper three panels. The bottom two panels depict Ponceau S staining of total proteins. Input lanes contained 1% of the WCEs used in pull-down assays. (B) Same as panel A except that WCEs were from derivatives of WT strain BY4741 expressing Myc₁₃-Rsc1p (top three panels) or Myc₁₃-Rsc2p (lower three panels), detected with Myc antibodies. (C) GST proteins were incubated with yeast WCE from WT strain BY4741, and the pull-down assays were probed with antibodies against the proteins listed beside the three panels, subunits of CCR4-NOT (Caf16p), SRB/MED (Srb7), and TFIID (TAF11/TAF40). (D) The GST proteins were incubated with WCEs from strains derived from BY4741 expressing Myc₁₃-tagged forms of Dhh1p (DHH1-myc), Ccr4p (CCR4-myc), or Paf1p (PAF1-myc). The tagged proteins were detected with Myc antibody, and SRB/MED interaction was detected with Srb7p antibody. Input lanes contained 10% of the WCEs used in the pull-down assays.

The results of pull-down assays shown in Fig. 7C and D suggest that CCR4-NOT also interacts specifically with the Gcn4p activation domain but that the Paf1 complex does not. Probing the pull-down assays from a WT extract with antibodies against the Caf16p subunit of CCR4-NOT revealed a higher level of binding to the WT than to 10Ala GST-Gcn4p or GST alone, comparable to the results obtained for the SRB/MED subunit Srb7p (Fig. ​(Fig.7C,7C, lanes 2 to 4). As expected, the TFIID subunit Taf40p did not precipitate with GST-Gcn4p (30, 93). Similarly, using extracts from strains expressing Myc₁₃-tagged forms of Dhh1p or Ccr4p in pull-down assays, we observed specific binding by both of these CCR4-NOT subunits to WT GST-Gcn4p (Fig. ​(Fig.7D,7D, lanes 1 to 8). By contrast, Myc₁₃-tagged Paf1p did not interact with GST-Gcn4p dependent on the activation domain (Fig. ​(Fig.7D,7D, lanes 11 and 12). As Ccr4p is a component of CCR4-NOT and the Paf1 complex, it most likely precipitates with GST-Gcn4p as a component of the former complex.

Gcn4p requires subunits of the HAT complex SAGA for transcriptional activation in vivo.

Our finding that WT activation of the UAS_GCRE-CYC1-lacZ and HIS3-GUS reporters by Gcn4p requires the HAT Gcn5p agrees with previous results showing impaired induction of certain Gcn4p target genes (HIS3, ILV1, and TRP3) in a gcn5 mutant (40). By contrast, we and others found that HIS4, SNZ1, ARG1, and ILV2 transcription is Gcn5p independent (Fig. ​(Fig.9),9), and there is evidence that the Gcn5p requirement at HIS3 can be diminished by changes in the Gcn4p binding site (40). Since UAS_GCRE-CYC1-lacZ contains tandem copies of a Gcn4p binding site from HIS4 and is clearly Gcn5p dependent (Fig. ​(Fig.2),2), other elements in the HIS4 promoter may reduce its Gcn5p dependence compared to other Gcn4p target genes like HIS3.

We found that the nonessential subunits of SAGA, Ada2p, Ada3p, Ada1p, Ada5p/Spt20p, and Spt7p are required for WT activation of multiple promoters by Gcn4p in vivo (Fig. ​(Fig.9),9), consistent with previous reports on ada2 (10), ada3 (100), ada5/spt20 (80), and ada1 (52) mutants. Overall, the deletions of SPT7, ADA5/SPT20, and ADA1 had greater effects on Gcn4p activation of multiple promoters than did deletion of ADA2, ADA3, GCN5, or SPT3, whereas deleting SPT8 had little or no effect (Fig. ​(Fig.9).9). Similar results were reported previously concerning the relative effects of gcn5Δ, spt7Δ, ada5Δ/spt20Δ, spt3Δ, and spt8Δ mutations on HIS3 activation by Gcn4p (8). Moreover, others have shown that spt7Δ, ada5Δ/spt20Δ, and ada1Δ mutants display a broader range and severity of growth phenotypes compared to gcn5Δ, ada2Δ, ada3Δ, spt3Δ, and spt8Δ mutants (52, 80, 119), which may be attributable to a requirement for Spt7p, Ada5p/Spt20p (42), and Ada1p (119) for SAGA integrity. These and other genetic findings indicate that SAGA performs an important function beyond the HAT activity of Gcn5p, such as TBP recruitment (34, 67, 78, 108, 119).

We observed higher-than-WT expression of the UAS_GCRE-CYC1-lacZ and HIS3-GUS reporters in spt3Δ and spt8Δ mutants under noninducing conditions (Tables ​(Tables11 and ​and3),3), in accordance with previous observations that HIS3 and TRP1 mRNA levels are elevated in such mutants (8). The negative function of SAGA at the latter genes has been attributed to inhibition of TBP binding. At ARG1, Gcn5p-dependent histone H3 acetylation, most likely in the context of SAGA, is required for transcriptional repression by ArgR/Mcm1p and is correlated with reduced TBP binding to the promoter (107).

The smaller HAT complex known as ADA shares Ada2p, Ada3p, and Gcn5p with SAGA, but ADA uniquely contains Ahc1p. The deletion of AHC1 had no effect on activation of the ILV genes (judging from the WT SM resistance of the mutant or UAS_GCRE-CYC1-lacZ (Table ​(Table1),1), consistent with the finding that purified ADA does not interact with Gcn4p in vitro (132). Because the ADA complex is unstable in the absence of Ahc1p (32), it seems likely that Ada2p, Ada3p, and Gcn5p promote activation by Gcn4p in the context of SAGA and not ADA.

Mutants lacking any one of seven nonessential HATs besides Gcn5p had little or no defect in activation of ILV genes by Gcn4p, as judged by their WT resistance to SM, including the HATs found in the SRB/MED and NuA3 coactivators, Nut1p and Sas3p, respectively (Fig. ​(Fig.6A).6A). Even in strains lacking Gcn5p, we observed no increase in SM sensitivity upon deleting Nut1p (Fig. ​(Fig.6B).6B). As the gcn5Δ sas3Δ strain is inviable, these two HATs could make overlapping contributions to activation by Gcn4p. Indeed, the modest effects of the sas3Δ and yer049Δ mutations on UAS_GCRE-CYC1-lacZ induction (Table ​(Table1)1) are consistent with a minor role for NuA3 in activation of this reporter. However, neither mutant has an SM^s phenotype, and Gcn4p could not recruit NuA3 to chromatinized templates in vitro (132).

In vitro, Gcn4p can interact with NuA4, containing the essential HAT Esa1p (132). In addition, depletion of Esa1p in vivo reduced histone H4 acetylation at HIS3 and HIS4, along with other promoters analyzed in parallel. Except for ribosomal protein genes, however, Esa1p depletion was not associated with reduced transcription. Consistently, we observed no SM^s phenotype in temperature-sensitive esa1 mutants at semipermissive growth temperatures (data not shown). Therefore, Gcn5p is the only HAT with an established in vivo function in transcriptional activation by Gcn4p.

Gcn4p requires SWI/SNF and RSC but not the ISWI chromatin-remodeling factors for transcriptional activation at certain promoters in vivo.

We confirmed our previous finding that Gcn4p requires the ATP-dependent chromatin-remodeling complex SWI/SNF for transcriptional activation in vivo. Except for SNF11, deletions of all SWI/SNF subunits produced SM^s phenotypes and defects in UAS_GCRE-CYC1-lacZ induction comparable to those observed in the swi2Δ/snf2Δ mutant, which lacks the ATPase subunit. It was shown that deletions of SWI2/SNF2, SWI3, SNF5, or SNF6 affect the integrity of the SWI/SNF complex; however, Swi2p/Snf2p was still found in a high-molecular-weight complex of ca. 1 MDa that lacked Swi3p, Snf5p, and Snf6p in these mutants (99). Interestingly, the Swi2p/Snf2p-containing subcomplex present in a snf5Δ mutant was defective for binding to the SUC2 promoter in vivo, implicating Snf5p (or one of the other subunits lacking in this subcomplex) in recruitment of SWI/SNF by activators. Consistently, Gcn4p was cross-linked to Snf5p, Swi1p, and Swi2p/Snf2p in vitro (95), suggesting that interactions with multiple noncatalytic subunits may contribute to SWI/SNF recruitment by Gcn4p.

It was surprising that only Swi3p was required for activation of HIS3-GUS (Table ​(Table3),3), particularly since it was shown previously that transposon insertions in SWI2/SNF2, SWI1, and SWP73 all impaired induction of this reporter by Gcn4p in another strain background (93). Perhaps the deletion library background contains a genetic modifier of the activation defects conferred by certain swi/snf mutations. Mutations in Swi2p, Snf5p, and Swp73 did result in derepressed HIS3-GUS expression (Table ​(Table3),3), and the swi2Δ/snf2Δ strain had derepressed HIS4 mRNA levels (Table ​(Table4)4) under noninducing conditions. Thus, activation defects in these mutants may be obscured by offsetting defects in a repression mechanism.

Our results showed that the Rsc2p subunit of RSC is required for full activation by Gcn4p of UAS_GCRE-CYC1-lacZ, HIS3-GUS, and HIS4, whereas activation of HIS3-GUS, SNZ1, and ARG1 was reduced by deletion of RSC1 (Fig. ​(Fig.9).9). Rsc2p is more abundant than Rsc1p (16), and we found that a 2μm/RSC1 plasmid can suppress the SM^s phenotype of the rsc2Δ strain (data not shown). Thus, the SM^s phenotype of rsc2Δ cells may reflect the fact that the Rsc1p complex is not abundant enough to support full activation by Gcn4p at one of the ILV promoters. However, the strong defect in HIS3-GUS, SNZ1, and ARG1 activation in rsc1Δ cells suggests that the Rsc1p complex is required to alter the chromatin structure at these genes in a way that cannot be performed by the Rsc2p complex.

It may seem surprising that Gcn4p requires both SWI/SNF and RSC for full activation of certain promoters, including UAS_GCRE-CYC1-lacZ, SNZ1, and HIS4. Whereas both complexes are recruited by Hir1p and Hir2p to the HTA1/HTB1 histone genes, SWI/SNF functions as a coactivator while RSC recruitment was correlated with repression (26, 97). Not all ATP-dependent chromatin-remodeling complexes are required for activation by Gcn4p, however, as deletion of the ATPase subunits of the ISWI complexes and of Chd1p had no effect on activation of ILV genes by Gcn4p (Fig. ​(Fig.6C6C).

Gcn4p shows differential requirements for Gal11 module subunits in SRB/MED.

The Gal11 module of SRB/MED, containing Gal11p, Med2p, Pgd1p, and Sin4p, has been implicated as a target of activators through biochemical analysis of mutant mediator complexes. SRB/MED complexes purified from pgd1Δ, med2Δ, or gal11Δ strains are devoid of two or all three of these subunits, and they exhibit quantitative reductions in transcriptional activation by Gcn4p in vitro (71, 88, 98). Consistently, the purified pgd1Δ mediator complex, lacking Gal11p and Med2p, failed to bind recombinant Gcn4p, and both recombinant Gal11p and Pgd1p can interact with Gcn4p in vitro. However, apart from the med2 insertion described previously (93), deletions of GAL11, PGD1, and MED2 were reported to have little or no effect on activation by Gcn4p in vivo (88, 98). By contrast, in the strain background employed here, deletions of all three genes led to significant defects in activation by Gcn4p. Thus, our genetic data support the idea that the Gal11 module provides an important binding site for Gcn4p in SRB/MED.

The purified sin4Δ mediator is devoid of all four Gal11 module subunits (27) and was shown to be more defective than the SRB/MED complexes purified from pgd1Δ, med2Δ, or gal11Δ strains in promoting transcriptional stimulation by Gcn4p in vitro (88). These results suggested that Sin4p is critically required for activation by Gcn4p. Surprisingly, transcriptional activation of HIS3 (56), HIS3-GUS, UAS_GCRE-CYC1-lacZ, HIS4, ARG1, and ILV2 (Fig. ​(Fig.9)9) was not diminished in the sin4Δ mutant. To account for this discrepancy between in vitro and in vivo findings, it could be proposed that the absence of Gal11p, Pgd1p, and Med2p from the sin4Δ mediator is an artifact of purification that does not prevail in vivo and that the latter proteins can support strong activation by Gcn4p in the absence of Sin4p in SRB/MED (89). Another possibility is that Sin4p harbors both positive and negative regulatory functions, and elimination of its negative function can overcome the absence of the entire Gal11 module at certain promoters in the context of in vivo chromatin structures. Indeed, the sin4Δ mutation led to derepression of UAS_GCRE-CYC1-lacZ, HIS3-GUS, and ARG1 expression under noninducing conditions (Tables ​(Tables1,1, ​,3,3, and ​and4)4) and was shown to restore partial function to the CYC1 promoter lacking the native UAS (56).

The activation defects conferred by the rox3Δ mutation are comparable to those produced by deletion of GAL11, PGD1, or MED2 (Fig. ​(Fig.9);9); however, Rox3p does not appear to be a component of the Gal11 module (71, 76, 88). Moreover, ROX3 is an essential gene in other strain backgrounds (89). Perhaps Rox3p provides a second binding site for Gcn4p in SRB/MED. Srb5p and Srb2p, and presumably the subcomplex comprised of these proteins and containing Srb4p and Srb6p (60), also contribute to activation by Gcn4p in vivo (Table ​(Table1).1). There is evidence that Srb5p facilitates a step in gene activation, possibly CTD phosphorylation, subsequent to activator recruitment (71).

Srb10p and Srb11p constitute a kinase-cyclin pair that is thought to impede assembly of the RNA PolII holoenzyme prior to promoter binding through phosphorylation of the CTD (89). Recently, Srb10p was also implicated in rapid degradation of Gcn4p (22). However, Gcn4p target genes were not constitutively induced in an srb10Δ mutant (51). Moreover, we observed a Gcn⁻ phenotype for the srb10Δ mutant in our genetic background (Fig. ​(Fig.9).9). Presumably, the positive effects of Srb10p on Gcn4p function outweigh its role in Gcn4p degradation.

Evidence that the Paf1 complex functions in Gcn4p-mediated transcriptional activation.

We found that Paf1p is required for WT SM resistance and induction of the UAS_GCRE-CYC1-lacZ reporter, whereas the Cdc73p subunit of the Paf1 complex was required for normal induction of the HIS3-GUS reporter. However, we observed no defects in activation of authentic Gcn4p target genes in paf1Δ or cdc73Δ cells (Fig. ​(Fig.9).9). As the Paf1 complex has been implicated in transcriptional elongation (117), it might be required by Gcn4p only for transcriptional elongation through reporter constructs containing bacterial coding sequences. The same argument could apply to the hpr1Δ mutant, shown previously to be defective for transcriptional elongation through lacZ (18), as this strain shows very low reporter expression but only a modest defect in activation of authentic Gcn4p target genes (Fig. ​(Fig.9).9). However, both the cdc73Δ and hpr1Δ mutants impaired reporter gene expression to a greater extent under inducing than noninducing conditions, suggesting that they contribute to the mechanism of activation by Gcn4p. This observation, plus our finding that Paf1p was recruited by Gcn4p to ARG1 (Fig. ​(Fig.8),8), leads us to suggest that the Paf1 complex is a physiological coactivator required by Gcn4p for WT induction of a subset of target genes that harbor transcribed sequences that impede elongation.

Hpr1p belongs to the THO/TREX complex, also involved in elongation, mitotic recombination (18, 19), and nuclear export of mRNA (112, 121). We observed no SM sensitivity in mutants lacking other THO/TREX complex subunits, including Tex1p, Mft1p, and Thp2p (Table ​(Table1).1). However, it is possible that hpr1Δ impairs an overlapping function of the Paf1 and THO/TREX complexes in elongation that is critical for activation by Gcn4p. We did not detect any requirement for the Elongator complex or for elongation factors encoded by SPT4 (Spt complex) or DST1 (TFIIS) for activation of ILV genes by Gcn4p, judging from the WT SM resistance of the corresponding mutants (Table ​(Table1).1). Thus, Gcn4p may depend primarily on the functions of Hpr1p in the Paf1 or THO/TREX complexes for stimulating transcriptional elongation at its target genes.

Gcn4p requires the CCR4-NOT coactivator, which possesses cytoplasmic deadenylase activity.

There is a large body of evidence implicating the CCR4-NOT complex in transcriptional activation and repression, including the requirement for Ccr4p in transcriptional activation by Adr1p, its physical association with Srb9p, -10p, and -11p, and its genetic and physical interactions with TBP and associated factors (see the introduction). Our ChIP results provide the first evidence for activator-dependent recruitment of CCR4-NOT to an induced promoter. These results are significant because CCR4-NOT also functions in the cytoplasm as an mRNA deadenylase (130, 131), with Ccr4p itself serving as the catalytic subunit (20, 130). Deadenylation should diminish translation of an mRNA and also render it susceptible to decapping and 5′-3′ exonucleolytic degradation (7, 110). Thus, mutations that merely destroy the deadenylase activity should not lower transcript levels in the manner we observed for several Gcn4p target mRNAs in ccr4Δ cells (Fig. ​(Fig.9).9). Previously, others provided strong evidence that factors involved in mRNA processing (79) or export from the nucleus (72, 121) become associated with the nascent transcript. Perhaps recruitment of CCR4-NOT by Gcn4p allows subunits of this complex to associate with the nascent transcript and accompany it into the cytoplasm to control deadenylation, translation, and degradation of the mRNA.

MBF1 is required for activation of a subset of Gcn4p-dependent genes.

Previous results indicated that Mbf1p serves as an adaptor between the DNA binding domain of Gcn4p and TBP. Deletion of MBF1 eliminated HIS3 and HIS5 mRNAs in vivo, and point mutations in Gcn4p that reduced binding to Mbf1p and a point mutation in Mbf1p that reduced interaction with TBP both lowered HIS3 mRNA levels (127). The mbf1Δ mutation in our strain background did not produce SM sensitivity or impair UAS_GCRE-CYC1-lacZ expression. However, expression of the HIS3-GUS reporter was lower in the mbf1Δ strain, in agreement with previous findings, and induction of SNZ1 and ILV2 was also impaired. Together, the results indicate a promoter-specific role for Mbf1p in Gcn4p-activated transcription.

We thank Andres Aguilera, Brad Cairns, Marion Carlson, Jan Fassler, Bruce Futcher, Leonard Guarente, Young-Joon Kim, Hannah Klein, Lawrence Myers, Akira Sakai, Fred Winston, Rick Young, and Richard Zitomer for plasmids; Brad Cairns, Rick Young, Clyde Denis, and Tony Weil for generous gifts of antibodies; Maria Pia Cosma, Min-Hao Kuo, and Rohinton Kamakaka for advice on ChIP assays; Judith Jaehning for sharing unpublished data on the Paf1 complex; members of the Hinnebusch and Dever labs for helpful suggestions; and Bobbie Felix, Jane Lin, and Felecia Johnson for help in preparation of the manuscript.

M.J.S. and H.Q. contributed equally to this work.