Dynamic Chromatin Alterations Triggered by Natural and Synthetic Activation Domains*

The activation domains (ADs) of transcription activators recruit a multiplicity of enzymatic activities to gene promoters. The mechanisms by which such recruitment takes place are not well understood. Using chromatin immunoprecipitation, we demonstrate dynamic alterations in the abundance of histones H2A, H3, and H4 at promoters of genes regulated by the HSF and Gal4 activators of Saccharomyces cerevisiae . Transcriptional activation of these genes, particularly those regulated by HSF, is accompanied by a significant reduction in both acetylated and unacetylated histones at promoters and may involve the transient displacement of histone octamers. To gain insight into the function of ADs, we conducted a genetic screen to identify polypeptides that could substitute for the 340-residue C-terminal activator of HSF and rescue the temperature sensitivity caused by its deletion. We found that the ts (cid:1) phenotype of HSF(1– 493) could be complemented by peptides as short as 11 amino acids. Such peptides are enriched in acidic and hydrophobic residues, and exhibit both trans -activating and chromatin-modifying activities when fused to the Gal4 DNA-binding domain. We also demonstrate that a previously identified 14-amino acid histone H3-binding module of human CTF1/NF1, which is similar to synthetic ADs, can substitute for the HSF C-terminal activator in conferring temperature resistance and can me-diate the modification of promoter chromatin structure. Possible mechanisms of AD function, including one involving direct interactions with histones, are discussed. (cid:3) -Galactosidase Assay— The activity of (cid:3) -galactosidase in liquid cell cultures (grown in synthetic complete 2% glucose-containing medium lacking tryptophan) was measured as described elsewhere (34). Computer Analysis of DNA and Protein Sequences— DNA sequence analysis and calculation of the isoelectric point of synthetic peptides was conducted using Vector NTI software. The calculation of protein secondary structure was done using the nnPredict algorithm of Vector NTI.

Despite major advances in characterizing components of the eukaryotic transcription machinery, the mechanisms by which these components are recruited to gene promoters are poorly understood. Activation domains (ADs) 1 of gene-specific transcription factors are critical for these recruitment steps. ADs are believed to work by directly contacting coactivators, components of the mediator complex and basal transcriptional machinery, and polymerase II holoenzyme itself (reviewed in Refs. [1][2][3][4][5][6]). These interactions have been previously shown to occur in vitro or in whole cell lysates. In certain cases, the biochemical data have been buttressed by mutational analyses showing a correlation between the change in the affinity of these contacts and transcriptional activity in vivo (7)(8)(9)(10)(11)(12). Recent studies indicate that recruitment of coactivators is often sequential and gene-specific (13)(14)(15). Despite these advances, there is little clarity of how multiple factors are recruited to a given promoter or what determines their order of recruitment.
An important quality of ADs is that they can substitute for one another not only between transcription factors of a given species, but also between transcription factors from such diverse eukaryotes as human, fungi, and plants. This suggests some uniformity in the mechanism(s) by which ADs work and implies that ADs share sequence or structural homology. In fact, although most ADs fall into one of three categories, acidic, glutamine-rich, or proline-rich, there is no consensus sequence or clear homology, even within a particular class. Moreover, ADs often lack any discernable secondary structure (1, 16 -18). The conservation of mechanism in the absence of any sequence or structural homology is paradoxical and reflects a gap in our understanding of transcriptional regulatory mechanisms.
Recruitment of ATP-dependent nucleosome remodeling and histone modifying enzymes to gene promoters is especially important since chromatin remodeling is a fundamental prerequisite to transcriptional activation. It has been shown that yeast Swi/Snf and histone acetyltransferase-containing complexes interact in vitro with a variety of transcription activators, including Gal4, Gcn4, Swi5, and Gal4-VP16 (7)(8)(9)(10). Recruitment of the SAGA complex by ADs of Gal4 and Gal4-VP16 has also been demonstrated in vivo (19,20). Recent investigations employing photo-cross-linking label transfer methodology identified the Tra1 subunit of SAGA and three subunits of the Swi/Snf complex (Snf5, Swi1, and Swi2/Snf2) as primary targets of a number of ADs (11,12). In these studies a strong correlation was observed between the level of transcription and strength of AD-target interactions for mutated ADs. Yet the multiplicity of targets capable of interacting with the same AD suggests the existence of a hierarchy of recruitment steps in vivo, and how such a hierarchy might be established is unknown.
Heat shock factor (HSF) is a key transcriptional activator of stress-responsive genes in yeast. It is also responsible for establishing the constitutive DNase I hypersensitive chromatin structures within the promoter regions of at least two genes, HSP82 and HSC82 (21,22). As such, it is likely that HSF recruits nucleosome-remodeling and histone-modifying activities to target promoters using its ADs. Yeast HSF is comprised of a highly conserved core, consisting of DNA-binding and trimerization domains, and two activation domains, one located at the N terminus and the other at the C terminus (Fig.  1). The N-terminal activator (NTA) has been shown to be critical for the transient heat shock response of the cell, while the C-terminal activator (CTA) is important for sustained stimuli, and its deletion results in lethality at elevated temperatures (23,24). As HSF has been shown to bypass the requirement for a number of key PIC components, including TFIIA, TAF17, Srb4, Srb6, Kin28, and the C-terminal domain of the large subunit of polymerase II (25)(26)(27)(28)(29)(30), the targets of its ADs may extend beyond the typical repertoire of most yeast activators (27).
In this work, we use chromatin immunoprecipitation (ChIP) to provide evidence for dynamic alterations of chromatin structure at heat shock promoters upon induction. Similar changes in histone content are observed at two promoters regulated by the Gal4 activator, GAL1 and GAL7. For both sets of genes, chromatin remodeling depends on the presence of an AD in the corresponding gene-specific activator. Using a genetic screen, we found that the temperature sensitivity associated with deletion of the HSF CTA can be alleviated by substituting short synthetic peptides enriched in hydrophobic and acidic residues for the large, native C-terminal domain. These peptides also function as ADs when tethered to the heterologous Gal4 DNAbinding domain. At least some of them can trigger histone modifications at Gal4-regulated promoters. The nature of these peptides suggests that their interacting targets likely are hydrophobic and basic, pointing toward histones as possible targets. Consistent with this notion, we found that the ts Ϫ phenotype resulting from a CTA deletion can be rescued by fusing short histone-binding modules, derived from either human CTF1/NF1 or the mouse liver-specific activator HNF3, to HSF(1-493).
Cultivation Conditions-Saccharromyces cerevisiae strains were cultivated at 30°C to early log phase (1-2 ϫ 10 7 cells/ml) in rich YPD broth supplemented with 0.04 mg/ml adenine. Heat shock induction was achieved by transferring 110 ml of cultures to a vigorously shaking 39°C waterbath; once the temperature had risen to 39°C, incubation was allowed to continue for an additional 15 min or 4 h. For GAL gene induction, cells were cultivated in rich medium as above, except containing 2% galactose rather than 2% glucose.
ChIP-In vivo cross-linking/chromatin immunoprecipitation was conducted essentially as described (32), except that PCR amplification was for 23 cycles, not 25. Our procedure differs from other yeast histone ChIP protocols (14,38) in that the wash steps following incubation with antibodies included (i) 1% Triton rather than 0.1% SDS and (ii) either 500 mM NaCl or 250 mM LiCl instead of 150 mM NaCl. Linearity of the PCR was confirmed through use of 3-fold serial dilutions of template DNA. Polyclonal anti-histone H4 (tetra-acetylated isoform), anti-histone H3 (unacetylated isoform), anti-histone H3 (Lys-14 mono-acetylated isoform), and anti-histone H2A (residues 88 -97) antibodies were obtained from Upstate Biotechnology (Lake Placid, NY). Promoterspecific primers used in this study were as follows: HSP12, ATACGC- Band intensities were quantified on a Storm 860 PhosphorImager (Molecular Dynamics) utilizing ImageQuant 1.11 software. To calculate the relative abundance of a given promoter DNA present in an immunoprecipitate (IP), we used the following formula: Q promoter X ϭ IP promoter X /Input promoter X . Abundance of each test promoter is expressed relative to that of PHO5, which served as a normalization control (32).
Library Construction and Screening-The library cloning vector pAME3000 (Fig. 4A) was constructed by insertion of a truncated version of HSF1 (positions Ϫ1018 to ϩ1479, encompassing the HSF1 promoter and amino acids 1-493 of the coding region) into the polylinker of pRS314. To facilitate cloning, KpnI and ClaI linkers were appended during PCR amplification of HSF1. In addition, a unique BglII site was engineered just 5Ј of the truncation site followed by several stop codons in different reading frames. The plasmid construction was confirmed by restriction digestion analysis and by sequencing of junction points. The BglII site was used to insert fragments (mean size ϳ2 kb) generated by Sau3AI digestion of yeast genomic DNA. The library contained 7000 independent clones with at least 70% clones containing inserts. The plasmid library was isolated from bacteria and transformed into PS128, selecting for Trp ϩ prototrophs. The transformation plate, containing ϳ10,000 individual clones, was replica-plated onto two 5-fluoroorotic acid-containing plates, one of which was incubated at 30°C and the other at 37°C.
Plasmid Constructions-Gal4 fusions were constructed as follows. Fragments of yeast genomic DNA from pAME3000 derivatives were PCR-amplified using primers flanking the BglII cloning site and containing synthetic BamHI and SalI and, using this sites, were recloned into the pGBD C-1 vector (31) maintaining the reading frame and stop codons. Resulting constructs were sequenced to confirm the intact reading frame and stop codon position. The CTF1/NF1 histone-binding module fusion was constructed by insertion of a synthetic oligonucleotide (complemented by opposite strand oligonucleotide) bearing the CTF1/NF1 DNA sequence with the natural Sau3AI site at the 5Ј-end and a synthetic NotI site after the stop codon at the 3Ј-end into pAME3000 using the corresponding sites. The resulting construct was sequenced to confirm the intact reading frame and position of the stop codon. Fusion of the HNF3 histone-binding module to HSF(1-493) was constructed similarly by in-frame fusion of a PCR-amplified HNF3 DNA fragment encoding amino acids 421-468.
Northern Analysis-Northerns were performed as previously described (33) using the following gene-specific probes: HSP12, PCR product spanning positions ϩ3 to ϩ296; HSP26, PCR product spanning positions ϩ67 to ϩ251; HSP82, synthetic oligonucleotide spanning positions ϩ2071 to ϩ2270; HSP104, PCR product spanning positions ϩ2528 to ϩ2726; ACT1, PCR product spanning positions ϩ984 to ϩ1125. Coordinates listed are relative to the ATG start codon. All  ␤-Galactosidase Assay-The activity of ␤-galactosidase in liquid cell cultures (grown in synthetic complete 2% glucose-containing medium lacking tryptophan) was measured as described elsewhere (34).
Computer Analysis of DNA and Protein Sequences-DNA sequence analysis and calculation of the isoelectric point of synthetic peptides was conducted using Vector NTI software. The calculation of protein secondary structure was done using the nnPredict algorithm of Vector NTI.

Histone Displacement Takes Place at Promoters of Inducible
Heat Shock Genes-The expression of yeast heat shock genes is principally under the control of HSF. However, the extent to which their promoter chromatin structure is remodeled upon transcriptional induction is unclear. On the one hand, DNA footprinting has demonstrated the presence of constitutive, promoter-associated DNase I hypersensitivity at HSF target genes that is unaffected by heat shock (35)(36)(37). Yet on the other, chemical cross-linking experiments suggest that substantial histone displacement may occur upon heat shock at these same promoters (32,38). To help resolve this discrepancy, we performed ChIP experiments using antibodies to both acetylated and unacetylated histone isoforms. To ensure quantitative, reproducible results the PHO5 promoter was co-amplified in each PCR to serve as a loading and amplification control. The promoter chromatin structure of PHO5 is well characterized; under phosphate-rich conditions it consists of an array of four stable, sequence-positioned nucleosomes (39,40). As shown in Fig. 2, A-H, the abundance of tetra-acetylated H4 and both unacetylated and mono-acetylated H3 is drastically reduced at each heat shock promoter following a 15-min heat shock. These histone rearrangements are dynamic, since during a chronic heat shock (4 h), when transcriptional attenua-tion is seen (Fig. 2, I-L, striped bars), histones are once again detected (Fig. 2, A-H). Indeed, there is a striking inverse relationship between promoter histone content and transcriptional activity.
To extend the analysis to H2A, we performed ChIP with an antibody raised against its globular domain (residues 88 -97), and thus theoretically insensitive to either N-or C-terminal tail modification. As shown in Fig. 3, the 15-min heat impulse induced a significant reduction of H2A content within the promoters of HSP12, HSP26, and HSP82; following a 15-min return to non-stressful conditions (recovery), the pre-heatshock abundance of H2A was restored. Taken together with the H3 and H4 ChIP data, we conclude that dynamic alterations occur in HSP promoter chromatin as a function of transcription and that such alterations may involve the displacement and reinstatement of histone octamers in response to activating and attenuating conditions, respectively.
Short Synthetic Peptides Can Functionally Substitute for the Large CTA of HSF-The dramatic chromatin modifications seen at HSF target promoters are a likely consequence of the activity of the ADs of HSF. It was shown previously that deletion of the HSF CTA causes a ts Ϫ phenotype and diminishes expression of heat shock genes (23,41,42). The mechanism by which the CTA enhances transcription and participates in chromatin remodeling at target promoters is unknown. To gain insight into its function, we substituted the sequence encoding the 340-amino acid CTA with random yeast genomic DNA fragments. Our goal was to select those protein domains whose in-frame fusion with HSF(1-493) suppresses the ts Ϫ phenotype, thereby complementing the loss of the CTA. A library of plasmids bearing the 3Ј-truncated hsf1(1-493) gene fused to Sau3AI-generated genomic DNA fragments (mean size ϳ2 kb) was constructed (Fig. 4A). This was transformed into an hsf1⌬::LEU2 strain (PS128) whose viability is maintained by an episomal HSF1 ϩ gene harbored on a URA3-CEN vector. Using the technique of plasmid shuffling, the original HSF1-containing plasmid was substituted with those from the library and transformants were screened for growth at the elevated temperature.
It was anticipated that this approach would lead to the isolation of the natural targets of the CTA (coactivators and 10-fold serial dilutions were spotted onto rich YPD medium and grown at the indicated temperatures for 2 days. WT, PS128 derivative in which an HSF1 ϩ /TRP1/CEN expression vector was substituted for HSF1 ϩ /URA3/CEN by plasmid shuffling; 3000, isogenic to WT except hsf1(1-493) (bearing pAME3000); 3002-3096, isogenic to 3000 except bearing episomal hsf1(1-493) fusions (described in Table I).
components of the basic transcription machinery), thereby bypassing the requirement for the CTA (in analogy to the "artificial recruitment" model (1)), or to the isolation of ADs derived from other activators (e.g. see Ref. 43). Either way, we anticipated that productive fusions would be rare events. Instead, complementation of the ts Ϫ phenotype was readily obtained. One percent of library transformants survived at the elevated temperature (data not shown). In contrast, there were no survivors at 37°C when PS128 was transformed with the episomal hsf1(1-493) gene, ruling out homologous recombination between hsf1 alleles as contributing to the revertant phenotype. To confirm plasmid-borne complementation of the ts Ϫ phenotype, plasmids were isolated from randomly selected clones. Each one was retransformed into the original PS128 strain, and the HSF1-containing plasmid counterselected as above. As assessed by a spot dilution assay, all transformants were able to grow at elevated temperature, although at different rates (Fig. 4B).
Identification and Characterization of Fragments Complementing the ts Ϫ Phenotype of HSF(1-493)-The inserts of the complementing plasmids were sequenced and their sequences searched against the S. cerevisiae genome data base. Remarkably, the isolated fragments have little if any relationship to transcription. Moreover, some of the cloned fragments were fused out of frame (Table I, clones 3024 and 3021), i.e. encode fortuitous peptides. The amino acid sequences specified by these DNA fragments underlined yet another remarkable observation, namely the short length of the encoded peptides. Some as short as 11 or 12 residues (Table I,  While amino acid sequence analysis failed to identify any consensus, it did reveal a tendency for negatively charged residues alternating with hydrophobic residues. Although all cloned fragments encode peptides with a strong total negative charge, the complementation is weaker for those fragments containing fewer hydrophobic and more positively charged residues (clones 3085 and 3096). Another interesting feature of the isolated peptides is the absence of any predicted secondary structure (see "Experimental Procedures"). These findings are highly reminiscent of those of Ptashne and co-workers (44,45), who previously identified synthetic ADs encoded by E. coli genomic DNA fragments or synthetic oligonucleotides (see "Discussion").
Synthetic Activators Have trans-Activation Potential Irrespective of Promoter or Transcription Factor Context-Even though the polypeptides described in Table I complemented the loss of the CTA, it is possible that the presence of these short fragments unmasked a cryptic activation domain in HSF . To address whether these polypeptides possess intrinsic activation potential, DNA molecules encoding them were ligated in-frame to Gal4(1-147), and these new constructs were transformed into the strain PJ69 -4A. This strain originally was constructed for yeast two-hybrid screens (31) and contains three reporter genes, each under the control of a different Gal4-regulated promoter (Fig. 5A). The resultant PJ694A transformants were assayed for GAL reporter gene expression. As shown in Fig. 5, B and C, polypeptides that complemented the HSF(1-493) ts Ϫ phenotype also acted as activators of the GAL7-lacZ and GAL1-HIS3 reporters. Moreover, the activation potential of each fragment is similar in the context of either Gal4(1-147) or HSF(1-493), indicating that these polypeptides can act as activation domains independent of the DNA-binding domain to which they are tethered. For the GAL2-ADE2 reporter, which confers the most stringent test of transcription (31), only clones 3002 and 3024 supported viability (Fig. 5D), underscoring their strength as activators.
GAL Gene Promoters Undergo Chromatin Modification Dependent on Native or Synthetic Activation Domains-In the context of HSF, the function of the synthetic ADs is obscured by the presence of the NTA (Fig. 1), which is capable of transactivation, particularly in response to acute heat shock (23). Gal4(1-147), which has little or no intrinsic activation potential (44) (Fig. 5B, strain 3000), permits a more straightforward test of remodeling activity. To examine the activity of Gal4(1-147) fused to the potent 3002 synthetic AD, ChIP analysis was employed as above. At the GAL1 and GAL7 promoters in a Gal4 ϩ strain, there is a significant decrease in the relative abundance of both unacetylated and acetylated isoforms of H3 following galactose induction (Fig. 6, A and B). These results indicate that Gal4 catalyzes the alteration of promoter-associated nucleosomes, although to a lesser degree than HSF (Fig.  2). Strikingly, the synthetic AD 3002, derived from a gene whose product is not related to transcription regulation, shows a similar capability at both promoters (Fig. 6, C and D). Thus, depletion of both acetylated and unacetylated H3 isoforms accompanying transcriptional induction can be induced by simple polypeptide modules that are normally not part of either transcription factors or their coactivators.
Histones Are Potential Targets of Activation Domains-The above data are consistent with a model in which chromatin modifications are central to the function of HSF and that its CTA plays a role in such modifications. Furthermore, short acidic/hydrophobic polypeptides bypass the requirement for the CTA in conferring temperature resistance. The acidic/hydrophobic nature of these peptides implies that their interacting partners could be basic and hydrophobic, raising the possibility that histones at gene promoters are among the targets of ADs. Indeed, the human CTF1/NF1 transcription factor bears a histone H3-binding module within its transactivation region (46). Analogous histone-interacting activity has been suggested for the yeast activator Hap1 (47) and was recently demonstrated for the mouse liver-specific activator, HNF3 (48). The CTF1/ NF1 histone-binding module has the sequence DPAGIY-QAQSWYLG and bears a striking resemblance to the synthetic activators isolated in our genetic screen: (i) it is short (14 amino acid residues); (ii) it has an acidic pI (3.0); (iii) it has six hydrophobic residues; and (iv) it lacks any predicted secondary structure. To determine whether the CTF1/NF1 histone-binding module could complement the ts Ϫ phenotype of HSF(1-493), we fused it to the C-terminally truncated HSF in place of the natural CTA. The strain expressing this new fusion protein grows at 37°C (Fig. 7), demonstrating that a natural AD of HSF can be functionally substituted by a histone-binding module. Similar complementation of temperature sensitivity, although to a lesser degree, was detected with the HNF3 histone-binding module (Fig. 7). The growth rate of CTF1(486 -499) and HNF3(421-468) fusions at 37°C is less robust than that of WT and resembles the 3022 and 3085 fusions, respectively (Fig. 4B).
To test the ability of the CTF1/NF1 module to remodel chro-matin, it was fused to the Gal4 DNA binding domain, and the histone composition of GAL1 and GAL7 promoters in yeast strains bearing either this fusion construct or Gal4(1-147) alone was analyzed. Remarkably, the 14-residue CTF1/NF1 module has the ability to initiate chromatin changes as indicated by a reduction in both unacetylated and diacetylated isoforms of H3 at GAL7 and a reduced level of unacetylated H3 at GAL1 (Fig. 8, B and C). The extent of histone alteration is modest yet reproducible for two different promoters and is consistent with the low level of trans-activation of GAL7/lacZ ( Fig. 8A; compare with Gal4(1-147)-3002 in Fig. 5B). The changes in histone composition nonetheless parallel those mediated by Gal4(1-147)-3002 (Fig. 6, C and D). These results, coupled with the ability of HSF(1-493)-CTF1/NF1 to rescue the ts Ϫ phenotype of HSF(1-493), demonstrate that the histone-H3-binding module can participate in transcription induction and trigger chromatin changes at disparate yeast gene promoters. They suggest that histone binding activity can be an important characteristic of AD function in vivo and imply the existence of a transcription activation mechanism involving AD-histone interactions.

Promoter-specific Transcription Factors Catalyze Dramatic Chromatin Modifications of Target Gene Promoters-
The results of our study indicate that upon transcription induction, there is significant chromatin alteration in the promoters of heat shock-and galactose-responsive genes. For heat shock genes the effect is particularly striking as the abundance of both acetylated and unacetylated histone isoforms is dramatically reduced. Quite significantly, abundance of H2A, monitored using an antibody insensitive to either N-or C-tail modifications, drops markedly as well. These data are consistent with recent ChIP analyses demonstrating that HSF-targeted promoters are uniquely associated with gross chromatin alterations, possibly including octamer displacement upon heat shock induction (32,38). Another explanation for these findings is that upon induction, heat shock gene promoters are uniquely occluded by components of the transcription initiation complex, rendering them inaccessible to antibodies. This interpretation seems unlikely since it invokes a fundamental difference in PIC assembly and/or structure at activated heat shock versus nonheat shock gene promoters (38). Moreover, we have previously shown that the abundance of promoter-associated histones is essentially unaffected by heat shock at a heterochromatic (SIRsilenced) version of the HSP82 promoter where heat shockinduced transcription is still very robust (32).
Our data for the GAL1 and GAL7 promoters, indicating a reduction in both hypo-and hyperacetylated histones upon transcriptional activation by wild type Gal4, are at variance with previous studies that revealed an increase in the abundance of either acetylated H3 (14) or unacetylated H4 (38). It is possible that the discrepancy lies in methodology. Here we have used multiplex PCR with an internal loading control in each reaction, obviating possibility of sample to sample variation. Moreover, the normalization control (PHO5) is transcriptionally inert under all conditions employed and thus ensures that differences that we detect are real and not a consequence of the vagaries of sample recovery. Moreover, our wash steps following incubation with antibodies are more stringent than those done in either of the previous two studies (14,38) and thus are more likely to remove chromatin fragments that are less avidly bound to antibodies. Finally, there may be intrinsic differences in the extent of chromatin remodeling at GAL1 and GAL7 stemming from differences in strain background.
It seems reasonable to suggest that chromatin remodeling, at least at HSF-and Gal4-regulated promoters, may be the product of two consecutive steps: (i) transition of histones from a hypo-to a hyperacetylated state and (ii) actual octamer displacement. We suggest that less robust chromatin remodeling occurs at Gal4-regulated promoters, reflecting differences in the potency of Gal4 and HSF ADs (see the introduction). Synthetic ADs are probably deficient in one or both steps, leading to partial chromatin remodeling (Fig. 6, C and D; Fig. 8, B  and C).
The 340-amino Acid C-terminal Domain of HSF Can Be Functionally Substituted by Synthetic Polypeptides as Short as 11 Amino Acids-To address the basis for HSF-catalyzed chromatin modifications, a genetic screen was designed for polypeptides that could complement the loss of the HSF CTA. The results of this screen showed that such peptides are encoded at an extremely high frequency (1% of the genomic library). These peptides are characterized by their short size, by the fact that they are not derived from transcription-related factors, and by the fact that they do not have any clear consensus sequence. Yet they are invariably enriched in acidic and hydrophobic amino acid residues. These qualities are highly reminiscent of synthetic ADs isolated from an E. coli genomic library or synthetic oligonucleotides fused to the Gal4 DNA binding domain (44,45,49). Thus, in different transcription factor and promoter contexts, short unstructured polypeptides enriched in acidic and hydrophobic residues are able to activate gene expression. The excess of hydrophobic and acidic amino acid residues suggests that interacting targets of such ADs are hydrophobic and basic. The histones are the most abundant nuclear proteins with such properties. As they are invariably present on gene promoters, they represent potential targets of ADs.
Transcriptional Activation in Yeast Can Be Stimulated by the Histone-binding Module of Human CTF1/NF1-Using the yeast two-hybrid assay, histone H3 has been previously identified as a target of the human CTF1/NF1 transcriptional activator (46). Using a similar methodological approach, histone H3 has also been implicated as a potential target of yeast Hap1 (47). A recent study likewise identified a histone-binding domain of mouse HNF3 as important for its function in remodeling chromatin (48). For CTF1/NF1, interactions with histone H3 were characterized both in vivo and in vitro, and the regions of strongest contact were mapped to a 14-amino acid module at the C terminus of CTF1/NF1 and the central core of H3 (amino acids 30 -136) containing the histone-fold domain. Here we demonstrate that both the CTF1/NF1 and HNF3 histone-binding modules at least partially complement the HSF(1-493) ts Ϫ phenotype. Moreover, we show that the CTF1/NF1 module exhibits transcriptional and chromatin-modifying activities when fused to Gal4-(1-147). Consistent with histone binding being a critical activity of CTF1/NF1 is the observation that transactivation in yeast by naturally occurring splice variants of CTF1 occurs only with those variants retaining the H3binding module at the C terminus (50,51).
The Targets of Activation Domains in Context of Direct and Indirect Recruitment-Our data indicate that substantial chromatin modifications occur at a number of gene promoters and that such alterations are mediated by both natural and synthetic ADs. Our observations certainly can be explained in terms of direct recruitment of nucleosome-remodeling and histone-modifying activities by these ADs. Nevertheless, we would like to suggest an alternative mechanism, which may apply in certain activator and/or promoter contexts. The idea is that at least some ADs have low affinity for histones and that this affinity is sufficient to epigenetically "mark" promoter nucleosomes and prime them for subsequent remodeling by histone-modifying and nucleosome-remodeling complexes, which are thereby recruited indirectly (see Fig. 9). The low affinity of AD-target interactions follows from the results of our and previous genetic screens for synthetic ADs. Indeed, natural ADs can be readily substituted with short synthetic ADs (this study and Refs. 44,45). The notion that these synthetic domains are (i) represented in 1% of random DNA sequences (this study and Ref. 44) and (ii) lack common consensus sequence and predictable secondary structure tends to argue against the high specificity of their interactions with targets. In fact, even homopolymeric glutamine or proline stretches can function as ADs (52), and the strong VP16 AD retains its functionality when chemically synthesized from non-natural D amino acids (53).
Chromatin remodeling at gene promoters appears to be a necessary prerequisite for the initiation of eukaryotic gene transcription. Most suggested mechanisms of this chromatin remodeling imply direct recruitment of coactivators by ADs. Recent reports indicate that the primary target of certain acidic ADs is the SAGA complex (19,20) with its largest subunit Tra1 directly interacting with ADs (12). Yet another study suggests that the Snf5, Swi1, and Swi2/Snf2 subunits of the Swi/Snf complex are the primary targets of VP16 and other acidic activators (11). Previously the VP16 AD and other acidic ADs were shown to interact with multiple components of the transcription initiation complex. The multiplicity of suggested AD targets makes it difficult to imagine a discriminatory mechanism for AD-target interactions. Consistent with the idea of indirect recruitment of nucleosome-remodeling and histone-modifying activities is the demonstration of non-targeted, global chromatin remodeling dependent on Gcn5-, Esa1-or Rpd3-containing complexes (54 -56) and non-targeted catalytic nucleosome remodeling by SWI/SNF (57). The recent demonstration that binding of Swi/Snf complex is stabilized by hyperacetylated nucleosomes (58) suggests a mechanism by which an epigenetic mark per se may be sufficient to recruit chromatin-modifying activities. Sequence-specific factors could focus these global activities by causing local distortion of chromatin structure.
A low specificity of interaction between ADs and their targets fits well with our model because distortion of the nucleosome, unlike the creation of new specific bonds implicit in the direct recruitment model, is likely to be achieved by interference with different histone-DNA and histone-histone bonds forming the nucleosome (59). Thus, nucleosome distortion does not presume a specific AD sequence but rather a general motif (acidic and hydrophobic) based on the nature of histones as the nucleosome components. These low-specificity interactions could readily occur in an indirect recruitment model since AD and histones are brought into close proximity by the DNAbinding domain critical for activator function. In contrast, in the direct recruitment model, the AD targets, which are often scarcely represented in the cell, must be physically brought to the promoter by selective interactions with ADs. These targets compete not only with each other but also with nucleoplasmic components that often have no relation to transcription. An appealing aspect of our model is that it helps to resolve the paradox of interchangeability of ADs and preservation of AD function among different biological kingdoms. Indeed, histones are among the most highly conserved proteins in eukaryotes. Direct AD-histone interactions may underlie, in part, the conservation of AD function among eukaryotic phyla and the ease with which they can be exchanged.
The indirect and direct recruitment mechanisms are not necessarily mutually exclusive. Many DNA-binding proteins, including TBP and TFIIB, are basic and have hydrophobic regions. ADs have been shown to interact and stabilize these PIC components at gene promoters increasing the frequency of transcription initiation. Importantly, CTF1/NF1 interacts with histone H3 within its histone-fold domain (46). This motif is widely represented in transcription-related factors including the histone-like TAFs, subunits of the SAGA and P/CAF histone acetyltransferases, and components of the ISWI and CH-RAC ATP-dependent chromatin remodeling complexes (60 -63). Such broad representation of histone-fold motifs in PIC components and nucleosome-modifying complexes on the one hand and the implication that promoter-specific activators may interact with the histone fold motif of H3 on the other suggests the possibility that direct recruitment of these complexes by ADs and interactions with histones could have a similar basis. Indeed, the VP16 AD has been shown to directly interact with the histone-like TAFs (64). It seems reasonable to suggest that the indirect and direct recruitment mechanisms coexist and complement each other in vivo and that the degree to which one or the other plays a role in remodeling is specified by features of a given promoter. This and other aspects of the indirect recruitment model are conjectural and require experimental verification. Such experiments are currently in progress.