A Gal4-ς54 Hybrid Protein That Functions as a Potent Activator of RNA Polymerase II Transcription in Yeast*

The bacterial ς54 protein associates with core RNA polymerase to form a holoenzyme complex that renders cognate promoters enhancer-dependent. Although unusual in bacteria, enhancer-dependent transcription is the paradigm in eukaryotes. Here we report that a fragment ofEscherichia coli ς54 encompassing amino acid residues 29–177 functions as a potent transcriptional activator in yeast when fused to a Gal4 DNA binding domain. Activation by Gal4-ς54 is TATA-dependent and requires the SAGA coactivator complex, suggesting that Gal4-ς54functions by a normal mechanism of transcriptional activation. Surprisingly, deletion of the AHC1 gene, which encodes a polypeptide unique to the ADA coactivator complex, stimulates Gal4-ς54-mediated activation and enhances the toxicity of Gal4-ς54. Accordingly, the SAGA and ADA complexes, both of which include Gcn5 as their histone acetyltransferase subunit, exert opposite effects on transcriptional activation by Gal4-ς54. Gal4-ς54 activation and toxicity are also dependent upon specific ς54 residues that are required for activator-responsive promoter melting by ς54in bacteria, implying that activation is a consequence of ς54-specific features rather than a structurally fortuitous polypeptide fragment. As such, Gal4-ς54represents a novel tool with the potential to provide insight into the mechanism by which natural activators function in eukaryotic cells.

Promoter specificity of bacterial RNA polymerases (RNAPs) 1 is conferred by sigma factors that bind the core RNAP to generate the ␣ 2 ␤␤Ј holoenzyme. In addition to 70 , which is required for transcription of most bacterial genes, there are several alternative sigma factors that typically promote transcription of genes involved in physiologically related pathways. 54 was initially identified as an alternative sigma factor, although subsequent studies revealed that 54 , like 70 , is involved in transcription of physiologically diverse genes (1,2).
All forms of bacterial RNAP bind promoter DNA in at least two steps: formation of a closed promoter complex followed by isomerization to an open complex where the two strands of DNA encompassing the start site are "melted" (3). These two steps are generally coupled such that the holoenzyme becomes stably bound to promoter DNA only upon open complex formation (3,4). However, this process is fundamentally different at 54 -dependent promoters, where the 54 holoenzyme binds promoter DNA to form a fully stable closed complex (5). Isomerization from a closed to open complex requires ATP hydrolysis and occurs only upon contact of a remote activator protein with the 54 holoenzyme (6). The uncoupling of promoter recognition from promoter melting enables 54 to respond to activators bound at distal sites (7). 54 is structurally distinct from 70 and is the only sigma factor that is not a member of the 70 family (8). Unlike 70 , 54 is a modular protein comprising at least three functional domains: the N-terminal region regulates DNA melting and transcriptional activation (7, 9 -12), the central region binds core RNAP (13,14), and the C-terminal region binds promoter DNA (11,15). 54 includes several structural characteristics found in eukaryotic transcription factors, including a glutamine-rich region near the N terminus, an acidic region that overlaps a hydrophobic heptad repeat near the central region, and a helixturn-helix motif within the C-terminal promoter binding region. Despite overall structural disparity, 54 and the 70 family of proteins share a region of sequence similarity involved in RNAP binding, located at the end of a heptad repeat within the central region of 54 (13).
The mechanism of transcription initiation by the 54 holoenzyme is reminiscent of the mechanism of initiation by eukaryotic RNA polymerase II (RNAP II). Eukaryotic transcriptional activators bind cognate enhancer sequences and stimulate transcription by interacting, either directly or indirectly, with components of the general machinery. In addition to RNAP II, the general transcription factors include the TATA-binding protein (TBP), TFIIB, TFIIE, TFIIF, and TFIIH (16). These factors are either assembled de novo at the core promoter or are recruited in association with RNAP II as a complex, in either case forming a stable closed complex over the core promoter. Transcription initiation is dependent upon subsequent isomerization to an open promoter complex in an ATP-dependent manner (17).
Activation of RNAP II promoters involves at least two steps, including recruitment of TFIID (TBP plus TBP-associated factors or TAFs) and recruitment of the RNAP II holoenzyme (18). Strong evidence in support of "activation by recruitment" comes from tethering experiments where the DNA binding domain of a transcription factor is fused to a component of the core machinery. The resulting hybrid proteins stimulate transcription, independent of an activator (19 -23). The reciprocal experiment, where an activation domain is fused to a compo-nent of the core machinery, does not bypass the activator requirement (24).
The core transcription factors that bypass the activator requirement when fused to a DNA binding domain include TFIIB (25). In an effort to define the mechanism by which a Gal4-TFIIB hybrid protein activates transcription, we attempted to identify TFIIB amino acid replacements that would strengthen the Gal4-TFIIB activator. As part of this study, we made the fortuitous discovery that fusion of a portion of Escherichia coli 54 to the Gal4 DNA binding domain is a potent activator in yeast. Here we describe the structural and genetic requirements for Gal4-54 -mediated activation.

RESULTS
Identification of a Gal4-54 Fusion Protein as a trans-Activator in Yeast-Artificial recruitment of TFIIB to promoter DNA stimulates transcription (25). Activation is dependent upon TFIIB residues that interact with RNAP II/TFIIF, suggesting that when tethered to promoter DNA, TFIIB stimulates transcription by recruiting the RNAP II holoenzyme complex; alternatively, TFIIB is a component of the holoenzyme complex such that the holoenzyme is recruited to the promoter by the Gal4 DNA binding domain. We reasoned that contact points between TFIIB and other components of the preinitiation complex might be identified as amino acid replacements that affect the activation potential of TFIIB when fused to the Gal4 DNA binding domain. TFIIB derivatives were generated by polymerase chain reaction amplification of the SUA7 gene under error-prone conditions, followed by ligation of amplified DNA behind the Gal4 DNA binding domain. The resulting Gal4-IIB library was introduced into yeast strain CBY14a, which carries chromosomally integrated LYS2::UAS GAL1 -HIS3 and URA3::UAS GAL1 -lacZ reporters (26). A single transformant exhibiting 3-aminotriazole-resistance (3-AT r ), because of HIS3 overexpression, was identified. This strain also turned intensely blue on X-gal indicator medium, indicating elevated UAS GAL1 -lacZ expression.
Plasmid DNA was recovered from the 3-AT r strain and sequenced. Surprisingly, sequence analysis revealed that the DNA fragment fused to Gal4 did not correspond to SUA7. Instead, comparison of this sequence with the databases identified a fragment of the E. coli rpoN gene, which codes for 54 . The insert corresponds to the rpoN open reading frame, encoding amino acids 29 -194 of the 478 residue 54 protein ( Fig. 1). This rpoN DNA fragment is flanked by SalI and BglII restriction sites, which are the same sites included in the SUA7 amplification primers for cloning purposes. Therefore, rpoN DNA appears to have been cloned from contaminating E. coli chromosomal DNA in our SUA7 plasmid preparation.
Suppression of Gal4-54 - (29 -182) Toxicity by Mutations in Components of the SAGA Coactivator Complex-The Gal4-54 -(29 -182) construct is toxic in yeast, a result reminiscent of the toxicity of Gal4-VP16 in yeast (32). Gal4-VP16 toxicity is thought to be a consequence of activator-mediated titration of transcription factors, or "squelching" (32). This effect was exploited by Guarente and co-workers (32,33) to identify components of the SAGA coactivator complex, a multifunctional coactivator that includes the Gcn5 histone acetyltransferase, as well as ADA and SPT proteins. These components are organized into distinct subcomplexes that exert differential effects on transcription (34 -38).
To identify the genetic requirements for activation by Gal4-54 , we asked if Gal4-54 -(29 -182) toxicity is suppressed by ada2⌬, ada3⌬, or gcn5⌬ deletions. Although the original Gal4-54 - (29 -194) construct had a mildly adverse effect on cell growth, the Gal4-54 -(29 -182) construct was extremely toxic, yielding colonies only upon prolonged incubation (Fig. 4). However, the toxic phenotype was fully suppressed by ada2⌬, ada3⌬, or gcn5⌬ deletions. The slow-growth phenotype associated with Gal4-54 -(29 -182) was restored in the ada2⌬, ada3⌬, or gcn5⌬ mutants by plasmid-borne ADA2, ADA3, or GCN5, respectively, confirming that suppression is caused by the indicated gene disruptions (data not shown). These results estab- FIG. 1. Structure of E. coli 54 . A, full-length 54 includes at least three functional domains: region I is required for activator-dependent promoter isomerization; region II binds RNAP; and region III binds the upstream core promoter element centered at position Ϫ24 (see text). Gal4-54 - (29 -194) includes portions of regions I and II. B, amino acid sequence of 54 - (29 -194). Sequence alignments of 54 with the S. cerevisiae Rpb5 subunit that is common to RNAP I, II, and III and with E. coli 70 are shown. Amino acids conserved between 54 and Rpb5 or 54 and 70 are highlighted in bold lettering.
lish that Ada2, Ada3, and Gcn5 are required for the toxicity associated with Gal4-54 -(29 -182) and implicate the SAGA complex as a cofactor in toxicity.
For this set of experiments, strain YMH171 and the indicated isogenic derivatives harboring the lacZ reporter plasmid were assayed for ␤-galactosidase activities. Compared with 1700 units of activity in the normal strain, activity was reduced 32-fold, to 54 units, in the isogenic gcn5⌬ deletion strain (Fig.  5). Activity was diminished in the ada2⌬ and ada3⌬ strains to 130 and 140 units, respectively, corresponding to 12-to 13-fold effects on activation. The spt7⌬ deletion had the most severe effect, diminishing activity 400-fold to 4.4 units. These results demonstrate that transcriptional activation by Gal4-54 is differentially dependent upon ADA, SPT, and Gcn5 components of the SAGA complex. This effect is consistent with previous reports that Spt7 is required for all SAGA-mediated functions, whereas other subunits, including the Gcn5 histone acetyltransferase, are required for a subset of SAGA functions (34, 37-39).

Gal4-54 Function Is Dependent upon 54 -specific Features-
The requirement for physically disparate regions for maximal activation by Gal4-54 suggests that activation is because of 54 -specific features and not to a fortuitous consequence of a peptide fragment that is structurally appropriate for activation. To test this possibility more directly, we generated specific amino acid replacements in the N-terminal region of 54 shown previously to affect enhancer-dependent promoter melting in E. coli. Syed and Gralla (9) found that extensive, random substitutions within the N-terminal 40 residues of 54 were tolerated without loss of function. However, no replacements were uncovered at positions Leu-33, Glu-36, and Leu-37, suggesting that these positions are critical for transcription in vivo (9).
Subsequent site-directed mutagenesis demonstrated that L33R, E36K, and L37R replacements are nonfunctional in vivo: L33R and L37R are specifically defective for activator-dependent, open complex formation, whereas E36K appeared to affect reinitiation. Because these residues are included in the Gal4-54 -(29 -177) activator, we constructed L33R, E36K, and L37R replacements and scored their effects on Gal4-54 -(29 -177)-mediated toxicity. The L31R and E32K replacements, which had no effect on transcription in E. coli, were generated as controls. Remarkably, the L33R, E36K, and L37R derivatives relieved the toxicity conferred by Gal4-54 -(29 -177), yielding a population of Leu ϩ transformants, whereas the L31R and E32K derivatives displayed the same toxicity as Gal4-54 -(29 -177) (Fig. 8). Thus, the toxicity conferred by Gal4-54 -(29 -177) is dependent upon specific amino acid residues that are also critical for promoter melting by 54 in bacteria. These results suggest that activation by Gal4-54 in yeast reflects a normal function of 54 in bacteria.
There is a notable difference between relief of Gal4-54 toxicity conferred by the cis-acting amino acid replacements and the trans-acting SAGA mutations. Whereas ada2⌬, ada3⌬ and gcn5⌬ deletions yielded transformants of uniform colony size (Fig. 4), the L33R, E36K, and L37R replacements yielded heterogeneous transformants (Fig. 8). A similar effect on toxicity is apparent for activation-defective derivatives of Gal4-VP16 (33). This effect precluded accurate quantification of the effects of the L33R, E36K, and L37R replacements on activation by Gal4-54 -(29 -177). Although each of these replacements (but neither L31R nor E32K) diminished activation, the magnitude of the effect varied in proportion with relief of toxicity defined  (29 -194) or the Gal4 DNA binding domain alone (Gal4Ϫ), was introduced into strain W303-1A or its ahc1⌬ derivative. Transformants were plated on ϪTrp selection medium and incubated for 3 days at 30°C. In contrast to the effect of deleting SAGA components (Fig. 4), deletion of the ADAspecific AHC1 gene does not suppress Gal4-54 -(29 -182) toxicity and enhances Gal4-54 - (29 -194)  Transformants were plated on ϪLeu selection medium and incubated for 3 days at 30°C. Comparison of Gal4-54 -(29 -177) WT, L31R, and E32K with the Gal4 DNA binding domain alone demonstrates the toxicity associated with by Gal4-54 - (29 -177). By contrast, the 54 L33R, E36K, and L37R replacements, shown previously to render 54 defective in activator-responsive promoter melting in E. coli (9), relieved toxicity. by transformant colony size (data not shown). Nonetheless, the replacements that affect promoter isomerization by 54 in E. coli clearly correlate with effects on cell growth by Gal4-54 in yeast. DISCUSSION In this study we define a fragment of E. coli 54 that functions as a potent transcriptional activator when fused to a Gal4 DNA binding domain in yeast. The initial Gal4-54 - (29 -194) construct displays ϳ60% of the activity of the Gal4-VP16-(412-490) activator, and the activities of these two hybrid activators were nearly indistinguishable when the C-terminal 17 residues of Gal4-54 - (29 -194) were deleted. Gal4-54 -mediated activation is TATA-dependent and also requires the SAGA histone acetyltransferase coactivator complex, with the relative effects of ada2⌬, ada3⌬, gcn5⌬, and spt7⌬ deletions on Gal4-54 -mediated activation comparable with their effects on Gal4-VP16mediated activation. Thus, Gal4-54 -(29 -177) is a potent activator of RNAP II transcription that is dependent upon a common RNAP II core promoter element and a general coactivator that facilitates chromatin remodeling by catalyzing histone acetylation.
Yeast contains a second Gcn5 histone acetyltransferase complex, designated ADA, which is distinct from SAGA (29). Deletion of the AHC1 gene, which encodes a subunit unique to the ADA complex, affects Gal4-54 activity in a manner opposite to deletion of genes encoding SAGA components: ahc1⌬ enhanced activation by Gal4-54 and not only failed to suppress Gal4-54 -(29 -182) toxicity, but rendered the Gal4-54 -(29 -194) construct toxic. Although ahc1⌬ was reported previously to have no effect on Gal4-VP16-mediated activation (29), we found that ahc1⌬ stimulates activation by Gal4-VP16 and Gal4-54 , in both cases by about 2.2-fold. Although we do not understand the basis for this discrepancy, it is clear that the SAGA and ADA complexes can be functionally distinct with respect to their effects on transcriptional activation, despite having a common histone acetyltransferase subunit.
The structure of 54 can be divided into at least three functional domains (Fig. 1). One is the C-terminal DNA binding domain (region III), encompassing residues 329 -478, which enables 54 to bind the promoter element centered at position Ϫ24 (11,15). The second domain (region II) interacts with core RNAP and is located between residues 120 and 215 (13,14). The most functionally complex domain encompasses the Nterminal 50 amino acids (region I) and plays a critical role in regulating the ATP-dependent isomerization of the 54 -DNA promoter complex (7, 9 -12). Accordingly, the Gal4-54 -(29 -177) activator includes sequence elements that affect RNAP binding and promoter isomerization (Fig. 1), suggesting that either or both of these activities might be responsible for Gal4-54 -mediated activation.
With one exception, progressive deletions from either end of the 54 moiety of Gal4 BD -54 -(29 -194) diminish activation function (Fig. 3). The exception is a 1.7-fold increase in activity associated with deletion of the C-terminal 17 amino acids. Sequence comparisons with the databases revealed that this region includes 8 of 10 residue identity to a conserved sequence within yeast Rpb5 (Fig. 1), a subunit common to RNAP I, II, and III, and homologous to a subunit of archeal RNA polymerase (40,41). The only region of structural similarity between 54 and 70 is also located in this region and binds core RNAP (13). Limited sequence similarities between bacterial factors and the GTFs, including TBP, TFIIB, and both subunits of TFIIF, have been noted previously, consistent with the proposed distribution of factor activities among the GTFs in eukaryotic cells (42)(43)(44)(45). The sequence similarity between 54 and Rpb5 suggests that distribution of factor activities ex-tends to a subunit of eukaryotic RNAPs. Whether there is functional significance to this sequence similarity and how it might account for the deleterious effect of the C-terminal 17 residues of Gal4 BD -54 - (29 -194) remains to be investigated.
Does the ability of Gal4-54 to function as an activator in yeast reflect a normal activity of 54 or does Gal4-54 simply include a polypeptide fragment that is structurally appropriate for activation? This is a valid concern because a small percentage of random E. coli DNA sequences are known to encode transcriptional activators when fused to a DNA binding domain (46). Two observations suggest that Gal4-54 activation reflects a normal function of 54 . First, activation by Gal4-54 is dependent upon a subfragment of 54 that encompasses 150 amino acids: deletions from either the N-or C-terminal region of 54 within Gal4-54 -(29 -177), spanning parts of two functional domains of 54 (Fig. 1), dramatically reduced reporter gene activation (Fig. 3). This compares to random bacterial activator sequences that are generally short, in several cases less than 20 amino acids in length (46). Second, specific amino acid replacements that render 54 defective in activator-responsive promoter melting in E. coli suppress the toxicity of Gal4-54 -(29 -177) in yeast (Fig. 8). Conversely, adjacent amino acid replacements that are without effect on 54 have no effect on Gal4-54 - (29 -177). Thus, the activity of Gal4-54 in yeast appears to reflect, in part, a normal function of 54 in bacteria.
The Gal4-54 activator might stimulate transcription by directly recruiting RNAP II to the core promoter. However, we were unable to detect a physical interaction between Gal4-54 and purified RNAP II by a co-immunoprecipitation assay (data not shown), even though RNAP II interacts with TFIIB and the Ssu72 protein in the same assay (47). Earlier studies also failed to detect a direct interaction between 54 and bacterial RNAP in a co-immunoprecipitation assay that detected interaction of RNAP with 70 and 32 (48). Despite the presence of a domain in Gal4-54 -(29 -177) implicated in RNAP binding, we have no evidence that Gal4-54 stimulates transcription as a consequence of direct interaction between 54 and RNAP II.
It is striking that the same amino acid replacements in 54 that adversely affect promoter melting in E. coli relieve the toxicity of Gal4-54 in yeast, an effect that correlates with Gal4-54 activation potential. This result is consistent with the premise that the Gal4-54 activator is dependent upon a normal function of 54 and suggests that Gal4-54 stimulates RNAP II transcription by enhancing or stabilizing open complex formation. Thus, Gal4-54 might not activate transcription by recruiting RNAP II or other components of the initiation complex, but instead by facilitating promoter melting. Furthermore, the SAGA requirement for activation by Gal4-54 suggests that SAGA might also play a role in promoter isomerization. One possibility is that Gal4-54 binds and stabilizes the fork junction that forms at the interface of double and single stranded DNA, as proposed recently to account for activatorinduced promoter melting by 54 in bacteria (49,50). Further characterization of Gal4-54 has the potential to provide novel insights into the mechanism by which natural activators function in eukaryotic cells.