Alterations in the GAL4 DNA-binding Domain Can Affect Transcriptional Activation Independent of DNA Binding*

The GAL4 protein belongs to a large class of fungal transcriptional activator proteins encoding within their DNA-binding domains (DBD) six cysteines that coordinate two atoms of zinc (the Zn2Cys6 domain). In an effort to characterize the interactions between the Zn2Cys6 class transcriptional activator proteins and their DNA-binding sites, we have replaced in the full-length GAL4 protein small regions of the Zn2Cys6 domain with the analogous regions of another Zn2Cys6 protein called PPR1 an activator of pyrimidine biosynthetic genes. Alterations between the first and third cysteines abolished binding to GAL4 (upstream activation sequence of GAL (UASG)) or PPR1 (upstream acitvation sequence of UAS) DNA-binding sites and severely reduced transcriptional activation in yeast. In contrast, alterations between the third and fourth cysteines had only minor effects on binding to UASG but led to substantial decreases in activation in both yeast and a mammalian cell line. In the crystal structure of the GAL4 DBD-UASG complex (Marmorstein, R., Carey, M., Ptashne, M., and Harrison, S. C. (1992) Nature356, 408–414), this region is facing away from the DNA, making it likely that there exists within the GAL4 DBD an accessible domain important in activation.

The GAL4 protein is an 881-amino acid transcriptional activator of the GAL and MEL1 genes in Saccharomyces cerevisiae. The mechanism by which GAL4 activates transcription is conserved because GAL4 also activates transcription in plant, Drosophila, and mammalian cells if GAL4 DNA-binding sites (UAS G ) 1 are present in the promoter of appropriate reporter genes (reviewed in Ref. 1). GAL4 is a member of a large family of fungal transcriptional activators that contain within their DNA-binding domains a conserved cysteine-rich region encoding six cysteine residues that coordinate two atoms of zinc (the Zn 2 Cys 6 domain). The DNA recognition sequences to which these activators bind have been identified and all have in common a CGG or related triplet in each of the two symmetrically opposed half-sites (2)(3)(4). The sequences between the trip-lets as well as the spacing of the triplet sequence are highly variable, indicating that these variables may in large part determine specific recognition. Based on the extent of homology between the Zn 2 Cys 6 activators, it was hypothesized that the Cys-rich region performs a function common to all members of this class of proteins and the region immediately adjacent, which encodes divergent sequences, determines DNA binding specificity (5). We previously demonstrated that this model is essentially correct (6). All but one (Lys 23 ) of the 28 amino acids in the Zn 2 Cys 6 region of GAL4 can be replaced by the analogous sequences of PPR1, an activator of pyrimidine biosynthetic genes (7), without changing DNA binding specificity. In contrast, replacing the 14 amino acids immediately adjacent to the Zn 2 Cys 6 region resulted in switching the DNA binding specificity to that of PPR1. Similar results were obtained when hybrids of LAC9, PPR1, and PUT3 were used (8,9). Recent structural characterizations of GAL4 (10), PPR1 (11), and PUT3 (12) complexes with their respective UAS were consistent with the biochemical characterization and revealed that the Zn 2 Cys 6 regions act as CGG-binding modules whose spacing and thus specificity is determined by the linker regions immediately adjacent to the Zn 2 Cys 6 region.
In an effort to further characterize the interactions between Zn 2 Cys 6 class transcriptional activators and their DNA-binding sites, we have revealed a region within the conserved Zn 2 Cys 6 module of GAL4 that, when structurally altered, drastically affects transcriptional activation in yeast and a mammalian cell line but has only minor effects on DNA binding. Within the crystal structure of GAL4-UAS G , this region is accessible to the environment and thus may make a contact crucial for activation.

EXPERIMENTAL PROCEDURES
Construction of GAL4 DNA-binding Mutants-Production of singlestranded DNA, site-directed mutagenesis of full-length GAL4, cloning into the multicopy plasmid YEp351, and sequencing were performed as detailed earlier (6). GAP71 and GAP72 were constructed from GAP1 and GAP2 using oligonucleotide 4PPR1-11. GAP100 and GAP110 were made from GAL4 and GAL4 Gln 23 , respectively, using oligonucleotide 4PPR1-13. GAL4 Gln 23 was made using oligonucleotide 4PPR1-12. The GAL4, GAP100, GAP110, GAP71, and GAP72 amino acid sequences 1-10 were replaced with the analogous sequences from PPR1 by replacing the wild-type BamHI-SphI fragment with the BamHI-SphI fragment from GAP3N/pUC119 to give the corresponding N derivative. The GAP3N mutant was made from GAP3 (6) using the oligonucleotide 4PPR1-8. Plasmids used for transfection into the COS-1 cells were made by amplifying the DNA-binding domains with oligonucleotides PAL-30 and PAL-32, cutting with BamHI and XhoI, and cloning into the unique BglII and XhoI sites of the plasmid pSGVP (13). All constructions were confirmed by sequencing. The sequence of all oligonucleotides used in this study are available upon request.
Synthesis of Proteins-Fragments of DNA encoding amino acids 1-147 of the DNA-binding domains of GAL4 or GAPs were amplified by the polymerase chain reaction using Taq DNA polymerase and the GAL4 or GAP genes on the plasmid YEp351 as starting templates for * This work was supported in part by a National Institutes of Health grant (to S. A. J.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: UAS, upstream activation sequence; polymerase chain reaction. The DNAs were amplified in the presence of oligonucleotide PAL-4, which defines the 3Ј end, and oligonucleotides that incorporate the T3 RNA polymerase recognition site into the 5Ј end (PAL-1 for GAL4 N-terminal amino acids or PAL-2 for hybrids containing PPR1 amino acids at the N terminus). The PPR1 coding region amino acids 25-171 was amplified with PAL-2 and PAL-6. Transcription reactions were performed using T3 RNA polymerase, and translation reactions were performed in rabbit reticulocyte lysates according to the supplier's instructions (Promega). The same RNA was used to program synthesis in two parallel reactions, one containing [ 35 S]methionine (Amersham Pharmacia Biotech) and the other containing cold methionine. The 35 S-labeled proteins were analyzed on denaturing SDS 15% polyacrylamide gels (19:1, acrylamide:bis-acrylamide). Prestained low molecular weight protein standards were from Stratagene. Electrophoretic Mobility Shift Assays-Assays of protein-DNA interactions were performed as described earlier (6). All experiments were repeated at least twice with similar results.
Yeast Culturing and ␣-Galatosidase Assays-Yeast culture conditions and ␣-galactosidase assays were performed as described (14). Protein concentrations of the yeast extracts were determined using the BCA protein assay reagent (Pierce) and bovine serum albumin to generate the standard curve. All assays were performed with at least three independent transformants with standard deviations less than 20%. For illustrative purposes, the activities of the GAL4-PPR1 hybrids are presented as percentages of the activity of wild-type GAL4.
Determination of Protein Levels in Yeast-All plasmids were tested for level of production of GAP proteins in the B80 S strain (␣ GAL4 GAL80 S -100 ura3-52 leu2-3 112 MEL1) by assaying growth on galactose (6).
Culturing and Transfection of COS-1 Cells and CAT Assay-Cells were transfected using the calcium-phosphate precipitation method, extracts were prepared, and CAT activity was measured using standard methods (15).

RESULTS
To gain insight into the sequences important in determining DNA binding specificity as well as structure of the Zn 2 Cys 6 domain, mutants of full-length GAL4 were made by replacing different regions of the GAL4 Zn 2 Cys 6 domain and N-terminal flanking region with the analogous sequences from PPR1. The sequence changes in the Zn 2 Cys 6 region of GAL4 were grouped into two subregions ( Fig. 1). Subregion A is highly conserved in the Zn 2 Cys 6 class of activator proteins (Fig. 1A, bottom). There are three amino acid differences between GAL4 and PPR1 within this region. Lys 17 and Lys 18 in GAL4 (10) or Lys 40 and Lys 41 at the equivalent positions in PPR1 (11) make specific contacts with the CGG motif within the UAS G or UAS u , respectively. Subregion B is less well conserved and includes three amino acid differences clustered between the third and fourth cysteines.
Replacement of the two subregions had pronounced but contrasting effects on transcriptional activation in yeast and DNA binding in vitro. Replacing the three amino acids in subregion A of GAL4 with those of PPR1 produced the GAP100 protein, which had severely decreased ability to activate MEL1 (Fig.  1A). Using in vitro transcribed and translated proteins in electrophoretic mobility shift experiments, binding of GAP100 to UAS G was not detected (Fig. 1B). In our DNA binding experiments with GAL4 and mutants that exhibited UAS G binding, two DNA-protein complexes were observed: a major complex (bottom arrowhead) and a slower-migrating, more labile complex that was not consistently observed (top arrowhead). Both complexes contained GAL4 DBD, as shown by supershift experiments using anti-GAL4 DBD antibodies (data not shown). Compared with earlier studies of interactions between GAL4 and UAS G (16), the lower complex is most likely a dimer of the GAL4 DBD bound to 1 UAS G . The upper complex may then be a GAL4 DBD dimer complexed to an additional protein. In addition a minor faster migrating GAL4-UAS G complex may represent a degradation product of GAL4 DBD. Replacement of subregion A and an additional amino acid in subregion B (Gln23) to give GAP110 abolished the ability to activate MEL1.
Loss of UAS G binding was not replaced by UAS u binding because none of the hybrid proteins examined in this study were able to bind to UAS u (data not shown). These data indicate that FIG. 1. DNA binding and transcriptional activation properties of mutants of GAL4 encoding changes within the Zn 2 Cys 6 region. A, sequence and properties of GAL4-PPR1 hybrid proteins. The Zn 2 Cys 6 (Cys-rich) region has been divided into subregions A and B. Sequences that differ from GAL4 are shown in the GAP and PPR1 lines. Dots represent identical amino acids in GAL4, GAP, and PPR1 proteins. Asterisks represent the invariant Cys residues. The numbers represent the position in GAL4 or PPR1 of the first amino acid shown. Activation of the MEL1 gene is expressed as a percentage of GAL4 activation. Also shown is the summary of DNA binding behavior to UAS G and UAS u . The consensus sequence of the Cys-rich and flanking regions is based on amino acids that are conserved in Ͼ33% of sequences from 79 Zn 2 Cys 6 fungal proteins (2). B, binding to UAS G by GAL4-PPR1 hybrid proteins. Extracts (5 l) containing the designated proteins synthesized in vitro were incubated with 32 P-labeled UAS G , and the complexes were resolved on nondenaturing polyacrylamide gels. C, SDS-polyacrylamide gel analysis of the GAL4-PPR1 hybrid proteins labeled with 35 S-labeled methionine. The position of an 18-kDa protein standard relative to the 35 S-labeled proteins is shown. the structures of subregion A of GAL4 and PPR1 are not interchangeable.
In contrast to the subregion A changes, replacement of the three amino acids in subregion B (GAP71) or subregion B plus four adjacent C-terminal amino acids (GAP72) had less drastic or no effects on UAS G binding, respectively (Fig. 1B). Despite significant binding to UAS G , GAP71 and GAP72 were severely compromised in their ability to activate MEL1. The defects in activation by the GAL4-PPR1 hybrids were not unique to the MEL1 gene, which is under control of one UAS G , but were also observed when a reporter gene under control of two UAS G s from the GAL1-GAL10 promoter region was assayed (data not shown). Large differences in the ability of the different GAL4-PPR1 hybrids to bind UAS G could not be attributed to differences in protein expression because expression of all of the hybrids examined in this study varied by no more than ϳ4-fold (Fig. 1C). Differences in in vivo expression and stability of the hybrid proteins were tested in a GAL80 S mutant strain that cannot grow on galactose because of constitutive repressive effects of a mutant of GAL80. Titration of the GAL80 S protein by either wild-type GAL4 or any of the GAL4-PPR1 hybrid proteins allowed equal growth on galactose, indicating that expression of the proteins in vivo was indistinguishable. The empty vector or the vector expressing PPR1, which does not bind GAL80, did not allow growth (data not shown). Taken together, these results show that alterations within subregion A drastically affect DNA binding, whereas alterations within subregion B preferentially affect activation.
The region N-terminal to the Zn 2 Cys 6 region is highly variable in both length and sequence between the Zn 2 Cys 6 family members. Only the amino acids immediately adjacent to the Zn 2 Cys 6 region have been conserved to an appreciable extent (Fig. 1A, bottom). Very little is known about the role of this region in supporting the structure of the Zn 2 Cys 6 domain or in DNA binding. Replacing the GAL4 N-terminal sequences with those from PPR1 had different effects on DNA binding and transcriptional activation depending on which GAL4 subregions contained PPR1 sequences. Replacement of the GAL4 N-terminal region with the analogous sequences from PPR1 in wild-type GAL4 (GAL4N) had no detectable effect on DNA binding, as shown earlier (6), but did have a small negative effect on activation of MEL1 (Fig. 1A). Replacing the N-terminal sequences of GAP100 with that of PPR1 to give GAP100N did not appreciably change the ability of the protein to activate or bind DNA. In contrast, addition of the PPR1 N-terminal sequence enabled the hybrids containing one (GAP110N) or three (GAP71N and GAP72N) PPR1 amino acids in subregion B to activate MEL1 better (2.7-51-fold). The ability of the hybrids containing PPR1 N-terminal sequences to bind DNA was unaffected. Minor differences in binding could be explained by differences in the amount of expressed protein. These data indicate that PPR1 amino acids in the N-terminal region were able to enhance the activation activities of hybrids containing PPR1 amino acids in subregion B but not subregion A. Despite this enhancement, all of the hybrid proteins containing changes in subregion B were moderately to severely defective in transcription activation.
To determine whether mutations in subregion B also affect transcriptional activation in a mammalian context, the DBDs from the GAL4-PPR1 hybrids ( Fig. 2A) were fused to the VP16 activation domain (13). The expression plasmids were cotransfected into a green monkey kidney cell line, COS-1, with a plasmid encoding the reporter gene CAT under control of one UAS G (Fig. 2A). Compared with GAL4-VP16, the GAL4N-VP16 exhibited a minor decrease in ability to activate the reporter gene (Fig. 2B). In contrast, the GAP72-, GAP72N-, GAP71-, and GAP71N-VP16 proteins activated the CAT gene to only a fraction of that activated by GAL4-VP16. Notably, the pattern of activation in COS-1 cells paralleled the activation in yeast, i.e. GAP71 activated less than GAP72 and both mutants were partially corrected by the PPR1 N-terminal sequences. Thus, the structural defects in subregion B that affect activation in yeast also have parallel effects on activation in a mammalian cell line. DISCUSSION In an effort to define the amino acids responsible for conferring DNA binding specificity and protein stability on Zn 2 Cys 6 class proteins, we have identified a short sequence within the GAL4 DBD that, when altered, results in a preferential decrease of transcriptional activation in yeast and a mammalian cell line but only minor effects on DNA binding. All of the proteins that exhibited this phenotype possessed at least four PPR1 amino acids replacing the analogous GAL4 amino acids between the third and fourth cysteines. In addition, the Nterminal region containing PPR1 amino acids was shown to be important in restoring the ability of hybrids containing PPR1 amino acids in subregion B to activate transcription. This could be only partly attributed to increased UAS G binding. The close proximity of the N terminus and subregion B in the GAL4-UAS G co-crystal structure (10) supports the idea that the compensatory effects of the N-terminal sequences are through direct interaction with subregion B. Although the structure of the Zn 2 Cys 6 modules from GAL4, PPR1, and PUT3 are similar (10 -12) and can be exchanged between the proteins without altering DNA binding specificity (6,8,9), our work indicates that replacing amino acids within the Zn 2 Cys 6 module of GAL4 with analogous amino acids from PPR1 can lead to two types of defects. Alterations of amino acids in proximity (e.g. GAP100) to Lys 17 and Lys 18 that make direct contact with the CGG triplets within the UAS G can disrupt DNA binding, possibly by disrupting the intramolecular interactions within the module. Alterations within the amino acids between the third and fourth cysteines can preferentially disrupt transcriptional activation without adversely affecting DNA binding.
The subregion B mutants characterized in this study are similar to the positive control (pc) mutants first isolated in repressor (17), which bind DNA with high affinity but are defective in transcriptional activation. pc-type mutants have been identified in other transcriptional activators that coordinate zinc in their DBDs including the yeast ADR1 protein (18,19) and the human glucocorticoid receptor (20). The mutational changes in ADR1 and human glucocorticoid receptor map to the amino acids in the C-terminal half of the first or second zinc fingers, respectively, in regions analogous to subregion B in GAL4. Although these regions do not share any significant sequence or structural homologies with GAL4, a common pclike mutation occurs in which Lys 23 in GAL4, Arg 115 in ADR1, or Arg 488 in glucocorticoid receptor was changed to Gln. Examination of the environment of this region in crystal structures of Zif268 (21) structurally similar to ADR1, (22) and glucocorticoid receptor (23) reveals that at least some of the amino acids that were altered in the pc-like mutants are not making contact with the DNA and are accessible to the surrounding environment. In the GAL4-UAS G co-crystal structure (10) subregion B is also exposed to the environment (Fig. 3). Three of the four amino acids within this region (Ser 22 , Glu 24 , and Lys 25 ) are directed away from the DNA. Lys 23 forms a hydrogen bond with a phosphate in the UAS G (10). The amino acids between the third and fourth cysteines of PPR1 (11) and PUT3 (12) also do not make contact with the DNA and are directed out toward the surrounding environment based on the protein-UAS crystal structures. Taken together, these data indicate that a number of DNA-binding transcriptional activator proteins that coordinate zinc possess a region within the DNA-binding domains important in determining optimal gene activation.
How do mutations in GAL4 subregion B lead to defects in transcriptional activation? One possibility is that GAL4 pc mutants simply have an altered conformation that inhibits the activation domain. Examples of this mechanism comes from work on pc mutants of the glucocorticoid receptor (24) and another Zn 2 Cys 6 protein, HAP1 (25), which indicated that appropriate transcription factor-UAS interactions are necessary for optimal transcriptional activation.
Another possibility is that subregion B or a region affected by changes in subregion B defines an interactive site with another protein important in transcriptional specificity. Physical interactions have been characterized between coactivators EIA (26) or SWI3 (27) and the DBDs of transcriptional activators. A number of proteins exhibit genetic or biochemical properties that make them candidates for interacting with the GAL4 DBD. A mutant of one protein GAL11, called GAL11P, potentiates transcriptional activation by proteins containing the GAL4 DNA-binding domain and weak transcriptional activation domains from other transcription factors (28). The positive effect of GAL11P is reversed by a GAL4 mutation (Lys 20 to Glu) in close proximity to the subregion B amino acids. However, there is no evidence for a direct interaction between wild-type GAL11 and the GAL4 DBD. Secondly, subregion B may be interacting with components of the nucleosome. Mutations in histones H3 (29) and H4 (30) allow hyperactivation or reduced activation, respectively, of a number of GAL4-regulated genes. In this scenario, the GAL4 subregion B might operate like a histone plow, as proposed earlier (5). Lastly, other regions of GAL4 itself could be interacting with subregion B. The conserved region between the DBD and activation domains in Zn 2 Cys 6 class protein (or middle homology region) has been proposed to bridge interactions between the Zn 2 Cys 6 module and the activation domain and to enhance DNA binding specificity (2). In a recent study, however, DNA binding and transcriptional activities of GAL4 mutants that lack the middle homology region were not adversely affected (31).
Intensive investigation is presently focused on identifying the protein target(s) of the activation domains from different types of transcriptional activators. Our work demonstrates that the GAL4 DBD possesses a region that determines optimal transcriptional activation. Thus, to completely understand the mechanism of transcriptional activation by intact GAL4 in yeast or by GAL4 activation domain hybrids in mammalian cells, it will be necessary to determine the interactions made between the DBD and other proteins. Selection of suppressors of GAP71 or GAP71N activation defects in yeast should identify the protein(s) important in determining transcriptional specificity of GAL4 and allow for determination of conservation in higher eukaryotes.

FIG. 3. Subregion B of GAL4 is directed away from the UAS G .
The structure of the DNA-binding domain of GAL4 (residues 1-65) bound to DNA looking down the DNA helix axis. The DNA is shown in gray in stick representation, and the protein is shown as a ribbon diagram with one monomer depicted in blue and the other in purple.
The zinc atoms are shown as yellow spheres, and the side chains of residues 22-25 belonging to subregion B of each monomer are shown in ball and stick form. The figure was drawn with the program MOL-SCRIPT (32) and rendered with RASTER3D (33,34).