The Transcriptional Activators BAS1, BAS2, and ABF1 Bind Positive Regulatory Sites as the Critical Elements for Adenine Regulation of ADE5,7 *

Adenine repression of the purine nucleotide biosynthetic genes in Saccharomyces cerevisiae involves down-regulation of the activator protein BAS1 or BAS2 by an unknown mechanism. To determine the minimal cis-acting requirements for adenine regulation, hybrid promoter constructs were made betweenADE5,7 promoter fragments and a CYC1-lacZreporter. A 139-nucleotide fragment containing two BAS1 binding sites was sufficient to confer adenine regulation on theCYC1-lacZ reporter. Analysis of deletion and substitution mutations led to the conclusion that the proximal BAS1 binding site is both necessary and sufficient for regulation, whereas the distal site augments the function of the proximal site. By performing saturation mutagenesis, we found two essential regions that flank the proximal site. An ABF1 consensus sequence is within one of these regions, and mutations that impaired in vitro ABF1 binding impaired promoter activity in vivo. A second region is AT-rich and appears to bind BAS2. No substitution mutations led to high level constitutive promoter activity as would be expected from removal of an upstream repression sequence. Our results indicate that ABF1, BAS1, and BAS2 are required for ADE5,7 promoter function and that adenine repression most likely involves activator modification or a negative regulator that does not itself bind DNA.

Adenine repression of the purine nucleotide biosynthetic genes in Saccharomyces cerevisiae involves downregulation of the activator protein BAS1 or BAS2 by an unknown mechanism. To determine the minimal cis-acting requirements for adenine regulation, hybrid promoter constructs were made between ADE5,7 promoter fragments and a CYC1-lacZ reporter. A 139-nucleotide fragment containing two BAS1 binding sites was sufficient to confer adenine regulation on the CYC1-lacZ reporter. Analysis of deletion and substitution mutations led to the conclusion that the proximal BAS1 binding site is both necessary and sufficient for regulation, whereas the distal site augments the function of the proximal site. By performing saturation mutagenesis, we found two essential regions that flank the proximal site. An ABF1 consensus sequence is within one of these regions, and mutations that impaired in vitro ABF1 binding impaired promoter activity in vivo. A second region is AT-rich and appears to bind BAS2. No substitution mutations led to high level constitutive promoter activity as would be expected from removal of an upstream repression sequence. Our results indicate that ABF1, BAS1, and BAS2 are required for ADE5,7 promoter function and that adenine repression most likely involves activator modification or a negative regulator that does not itself bind DNA.
The de novo synthesis of purine nucleotides requires 10 enzymatic steps to form the first purine nucleotide, inosine monophosphate. IMP is converted to either AMP or GMP in two steps. The products of at least 13 genes are required for this synthesis, since mutations in any one of these lead to an adenine requirement (1). Most of these genes encode the biosynthetic enzymes that participate in specific steps of the pathway (ADE1, ADE2, ADE4, ADE5,7, ADE6, ADE8, ADE12, ADE13), others encode enzymes required to produce additional substrates necessary to complete the pathway (ADE3, ASP5), and the function of one gene product is unknown (ADE9). Strains with mutations in two additional genes, BAS1 and BAS2, have a partial adenine requirement (2). In these latter mutants, expression of the biosynthetic enzymes is low, indicating a positive regulatory role for these gene products (3,4).
Expression of the adenine biosynthetic genes is repressed when cells are grown in the presence of adenine (3)(4)(5)(6). This adenine-mediated repression occurs at the transcriptional level (7). ADE gene transcription is unregulated in bas1 or bas2 mutant strains (3). The expression of BAS1 and BAS2 is not regulated by adenine levels (8,9), however, indicating that adenine repression occurs by down-regulating the activator functions of the BAS1 or BAS2 proteins.
BAS1 and BAS2 were identified as transcriptional activator proteins required for the basal expression of the HIS4 gene (2). BAS1 binds to DNA using an amino-terminal myb motif (8). This tryptophan-rich motif is repeated three times in the BAS1 protein (8,10). BAS1 binds to two sites in the HIS4 promoter (8) and to two sites in each of the promoters for the ADE2 and ADE5,7 genes (3). The consensus sequence derived from these six BAS1 binding sites is TGACTC. Interestingly, this hexanucleotide sequence forms the core of the binding site for GCN4 protein as well, although flanking nucleotides differently affect the binding affinity of these two proteins (10,11). Whereas GCN4 has a preference for the sequence RRTGACT-CATTT (R represents A or G; Ref. 11), none of the known BAS1 sites or any predicted sites in other ADE gene promoters have either an A nucleotide at the 3Ј base following the conserved hexanucleotide core or any other sequence conservation.
BAS2 binds to DNA via an amino-terminal homeodomain that is closely related to the engrailed protein of Drosophila (8). The BAS2 binding site at HIS4 has been mapped to an A ϩ T-rich repeat, TTAA (8). The results of electrophoretic mobility shift assays suggested that BAS2 binds to the ADE2 and ADE5,7 promoters; however, no specific binding sites were identified (3). In addition to participating with BAS1 in transcriptional activation of HIS4 and the ADE genes, BAS2 (also known as PHO2 and GRF10) stimulates transcription of PHO5 (12) and HO (13). High level transcription of HO requires BAS2 in conjunction with SWI5 (13,14). Binding of BAS2 and SW15 to the HO promoter is cooperative in vitro (13,15). BAS2 also interacts with PHO4 to induce PHO5 transcription under phosphate starvation conditions (16). Binding of PHO4 is BAS2-dependent and restricted to derepressing conditions as shown by in vivo footprinting experiments (17). There is evidence that BAS2 and PHO4 physically interact at the promoter and that this interaction is regulated by phosphate (18).
Because expression of the genes that are activated by BAS1 and BAS2 is lower when cells are grown in adenine excess, adenine antagonizes the activation function of one or both of these proteins. Excess adenine could inhibit their ability to interact with one another, to bind to DNA, or to interact with components of the transcriptional machinery. Down-regulation of BAS protein function could also occur by covalent modification or through binding of a negative regulator, analogous to binding of the repressor GAL80 to the activator GAL4 (12). Alternatively, a DNA binding repressor might interact with negative control sites in the promoter, as described for the MIG1-TUP1-SSN6 complex that mediates the carbon catabo-* 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 U.S.C. Section 1734 solely to indicate this fact. lite repression of GAL genes (12).
To define more clearly the cis-and trans-acting regulatory elements required for adenine regulation, we performed an extensive analysis of the ADE5,7 promoter. First, we identified the minimal sequences from ADE5,7 sufficient to confer adenine-regulated activation (UAS ADE5, 7 1 function) to a heterologous CYC1-lacZ reporter lacking its native UAS CYC . We then subjected the minimal UAS ADE5,7 to extensive mutagenesis by making successive 3-nucleotide substitutions across this element to identify precisely the critical nucleotides contained therein. Our results indicate that of the two BAS1 binding sites at UAS ADE5,7 , the gene-proximal copy is the more critical, being both necessary and sufficient for adenine-regulated promoter activity. Two additional regions were essential for UAS ADE5,7 function. One region, located 5Ј to the critical BAS1 site, is a binding site for ABF1. The other, located 3Ј to the critical BAS1 site, is an extended A ϩ T-rich element that appears to be a Bas2 binding site. We found that none of the nucleotide substitutions in the UAS ADE5,7 led to constitutively derepressed expression, the phenotype expected from the loss of a repressor binding site. Thus, our results suggest that the repression of UAS ADE5,7 function in adenine-replete cells involves a modification of one of the three transcriptional activator proteins, BAS1, BAS2, or ABF1, rather than binding of a repressor.
E. coli cells were grown in LB medium supplemented with 100 g/ml ampicillin. S. cerevisiae cells were grown in synthetic dextrose (SD) medium (20) supplemented with 0.5 mM arginine and 0.3 mM histidine. Adenine was added to 0.15 mM in the cultures grown under repressing conditions.
Oligonucleotides and Plasmids- Table I lists the oligonucleotides that were used as polymerase chain reaction (PCR) and sequencing primers for cloning and as probes for DNA binding assays as described

5Ј-CGTCCTCGAGCGCATATTCACGGGTCCTGGGAGATCCAGGAC-3Ј
below. Deletion of an XhoI fragment that contained the UAS CYC1 from plasmid pLG699Z (21) generated pLG699Z⌬XhoI. Plasmid pCM81, a derivative of pLG699Z⌬XhoI, contains an oligonucleotide that replaces the unique XhoI site with adjacent BglII and XhoI sites (22). Fragments of the ADE5,7 promoter were generated by the PCR using plasmid pYEADE5,7(5.2R) (23) and the primers listed in Table I, as described in  the relevant figure and table legends. PCR reactions were performed using Taq polymerase (Perkin-Elmer) under the following conditions: 30 cycles of denaturing at 95°C for 2 min, annealing at 55°C for 1 min, and chain extension at 72°C for 30 s. PCR fragments were digested with either XhoI or XhoI and BglII, separated from primers by electrophoresis, purified, and inserted into pLG699Z⌬XhoI or pCM81. The ligated products were transformed into competent E. coli by electroporation (19). Bacterial transformants were screened for insertions by colony hybridization (24), by restriction analysis (24), or by whole cell PCR (25). The nucleotide sequence of the vector-insert junctions was verified in all constructs by sequence analysis using primer RO-26, which corresponds to positions Ϫ160 to Ϫ142 (relative to the ATG codon) of the CYC1 gene. Plasmids were transformed into yeast cells using lithium acetate (26) and plated on medium lacking uracil. Plasmid pR116 carries an ADE5-lacZ fusion and was constructed by first introducing a 3.0-kb BamHI fragment carrying the E. coli lacZ gene (27) into the BamHI site located in the ADE5,7 gene of plasmid pYEADE5,7(5.2R) (23) to produce plasmid pR111. The ϳ8-kb SalI to BspEI fragment containing the ADE5-lacZ fusion from pR111 was subcloned into the SalI and XmaI sites of a modified form of pRS316, lacking the ␣-fragment of lacZ, to produce pR116.
Plasmid pR173 contains a fusion of the bacterial glutathione Stransferase (GST) gene with the full-length coding sequence of BAS1 and was constructed in three steps. First, an oligonucleotide duplex formed between RO-97 and RO-98 was inserted at the BamHI site of pGEX-5X-3 (Pharmacia Biotech Inc.) to yield pR171. The 5Ј end of the gene was generated by PCR using oligonucleotides RO-107 to RO-108 and pCB286 (2). The PCR fragment was cleaved with NdeI and BglII and ligated into the NdeI and BamHI sites of pR171 to yield pR172. The remainder of the BAS1 gene was cloned as a 3.5-kb BamHI fragment from pCB286, ligated into the same site in pR172 to yield pR173.
Plasmid pR175, encoding the GST-BAS2 fusion, was constructed in two steps. First, an oligonucleotide duplex formed between RO-103 and RO-104 was inserted at the BamHI and EcoRI sites of pGEX-5X-3 to yield pR174. Plasmid pCB841 (2), which carries BAS2, was partially digested with EcoRI to generate a 2.5-kb fragment that was ligated into the EcoRI site of pR174 to yield pR175. Plasmid M2025 carrying the (His) 6 -BAS2 fusion was a gift of D. Stillman (15).
␤-Galactosidase Assays-Transformants to be assayed for ␤-galactosidase activities were inoculated in 5 ml of SD medium supplemented with 0.5 mM arginine, 0.3 mM histidine and 0.15 mM adenine and cultured for ϳ42 h. Each saturated culture was diluted 1:50 in 25 ml of fresh medium, with and without adenine supplementation, and grown for 5 h with shaking at 30°C. Cells were harvested by centrifugation and frozen at Ϫ20°C overnight. ␤-Galactosidase assays were performed using whole-cell extracts (22).
Oligonucleotide Probes for Gel Mobility Shift Assays of DNA-binding Proteins-To generate labeled ADE5,7 DNA fragments, oligonucleotides RO-41 and RO-136 were end-labeled with polynucleotide kinase and [␥-32 P]ATP and were separated from unincorporated label by passage over a G25 spin column. PCR reactions were performed as described above using the labeled oligonucleotides as primers and plasmids containing the appropriate ADE5,7 sequences as templates to amplify the region between Ϫ211 and Ϫ145 (relative to the ADE5,7 start codon). The templates were pR224, carrying the wild-type ADE5,7 fragment, or selected plasmids carrying mutated ADE5,7 promoter fragments, constructs numbered 5,7,10,11,13,14,19,20,27,29,34,35,39, and 43 as listed in Table IV. After PCR, a portion of each sample was separated by electrophoresis, and concentrations of the PCR products were estimated by ethidium bromide staining in comparison with known concentrations of duplex oligonucleotides of approximately the same length.
Oligonucleotides RO-118 and RO-119, encoding ARS1, and RO-120 and RO-121, encoding a mutant ARS1, were prepared by end labeling with polynucleotide kinase and [␥-32 P]ATP (24), removing the unincorporated ATP on G25 spin columns, and annealing equimolar amounts of each by heating to 65°C and cooling slowly to room temperature.
DNA fragments used as competitors were prepared as described above for the radiolabeled PCR probes, omitting the end labeling and purification steps. To generate the fragment carrying mutations in both regions III and IV, RO-136 and FZP20 were used as primers in a PCR reaction. The product of this PCR reaction was then used as template in another round of PCR using RO-136 and FZP21 as primers. The PCR fragment was inserted into pCM81 and subjected to sequence analysis to verify the correct sequence.
Electrophoretic Mobility Shift Assays of DNA Binding by ABF1-Yeast strain AY862 was grown to midlog phase and harvested by centrifugation. Cells were broken using glass beads in buffer A (25 mM Hepes, pH 7.7, 50 mM KCl, 10% glycerol, and 0.5 mM EDTA). Cell lysates were clarified by centrifugation at 20,000 ϫ g. Labeled DNA probes for electrophoretic mobility shift assays (EMSAs) were either a 32-base duplex oligonucleotide containing the ABF1 binding site from ARS1 or radiolabeled PCR fragments containing ADE5,7 sequences, prepared as described above. DNA binding assays were performed in 20 l with 10 g of protein in a whole cell lysate and 10 fmol of DNA in a buffer containing 25 mM Hepes, pH 7.7, 150 mM KCl, 5 mM MgCl 2 , 1 g of poly(dI⅐dC), and 1 g of sonicated calf thymus DNA. Binding reactions were incubated at room temperature for 30 min and separated by gel electrophoresis using 6% polyacylamide gels in 22.3 mM Tris borate, pH 8.3, 0.5 mM EDTA. The gel was pre-electrophoresed for 15 min prior to loading samples. Supershift experiments were performed by adding either preimmune serum or immune serum raised against partially purified ABF1 to the binding reactions. The antisera, kindly provided by Bruce Stillman (28), were diluted in buffer A as indicated in the legend to Fig. 3. Competitions with unlabeled DNA fragments were performed using wild-type or mutant ARS1-containing oligonucleotides, or with PCR-derived fragments of the ADE5,7 promoter prepared as described above.
Electrophoretic Mobility Shift Assays of DNA Binding by BAS1-Bacterial strains containing the GST-BAS1 fusion construct or the pGEX empty vector were inoculated at a 1:100 dilution from saturated cultures into 50 ml of LB containing ampicillin. Cultures were grown for 2.5 h, and IPTG was added to 0.5 mM for induction of the GST proteins. Cells were harvested by centrifugation after an additional 2.5 h of growth, frozen in a dry ice/ethanol bath, and stored at Ϫ80°C. Pellets were thawed on ice and resuspended in buffer A containing 10 mM 2-mercaptoethanol and a mixture of protease inhibitors: 1 g/ml aprotinin, 0.5 g/ml leupeptin, 1 g/ml antipain, 1 g/ml chymostatin, and 1 mM phenylmethylsulphonyl fluoride. Cell extracts were prepared by sonication with four 15-s pulses in an ice water bath and by three cycles of freezing in a dry ice/ethanol bath and thawing in water. Cellular debris was removed by centrifugation at 20,000 ϫ g for 15 min. Extracts were loaded onto a glutathione S-Sepharose column with a 2-ml bed volume equilibrated in buffer A. GST control and GST-BAS1 fusion proteins were eluted from the column with 10 mM reduced glutathione in buffer A. Electrophoretic mobility shift assays were performed in 10 l using 0.25 g of purified protein and 10 fmol of wild-type or mutant DNA fragments in a buffer containing 25 mM Hepes, pH 7.7, 50 mM KCl, 0.5 mM EDTA, 1 mM dithiothreitol, 0.5 g of poly(dI⅐dC), and 10% glycerol (8). Unlabeled DNA fragments were added as competitors in a 200-fold molar excess. Samples were fractionated by electrophoresis on 6% polyacrylamide gels, as described above for ABF1 binding.
Electrophoretic Mobility Shift Assays of DNA Binding by BAS2-GST-BAS2 was prepared and used in EMSAs as described above for GST-BAS1. The (His) 6 -BAS2 protein was prepared for use in EMSA experiments as described (13,15) with the following modification. Crude cell lysates (1.5 mg) from bacteria transformed with GST-BAS2 plasmid pM2025 (13) or the empty vector were prepared and added to 200 l of Ni 2ϩ charged His-Bind resin (Novagen). The fraction that eluted with 300 mM imidazole was dialyzed against buffer: 20 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 100 mM NaCl, 0.5 mM dithiothreitol, 10% glycerol, and the protease inhibitors described above. Bovine serum albumin as a carrier protein was added to a final concentration of 0.75 mg/ml before storage at Ϫ80°C. Electrophoretic mobility shift assays were performed using 1.6 l of these purified protein fractions.

Transcription of ADE5,7 and Its Repression by Adenine Both
Depend on BAS1 and BAS2-We began our study of adenine regulation by analyzing expression of an ADE5,7-lacZ fusion in various strains grown in the presence and absence of exogenous adenine. We assayed wild-type and mutant strains lacking BAS1, BAS2, or GCN4. We found that ADE5,7 promoter activity under derepressing conditions (without Ade medium) and repressing conditions (with Ade medium) was lost in strains containing the mutant alleles bas1-2 or bas2-2 but was not significantly affected in a strain containing the mutant gcn4 allele (Table II). These results indicated that BAS1 and BAS2 are both essential for transcriptional activation of ADE5,7 under derepressing conditions, and they are consistent with the idea that the adenine repression involves down-regulation of BAS1 or BAS2 (3). In addition, they show that GCN4 is dispensable for ADE5,7 promoter activity and its regulation by adenine under these growth conditions (7). Table II suggest that the ability of BAS1 and BAS2 to activate ADE5,7 transcription is completely inhibited in adenine-replete cells. It was not known, however, whether ADE5,7 promoter activity is dependent on any additional positive regulators that might be subject to adenine repression or whether repression involves a DNA binding repressor. To address these issues, we set out to identify the cis-acting sequences required for adenine-regulated promoter activity by determining the smallest DNA fragment from ADE5,7 sufficient to confer adenine-repressible UAS function on a CYC1-lacZ reporter (21). Various fragments of the ADE5,7 promoter were generated by PCR and inserted ϳ185 nucleotides upstream of the CYC1 transcription initiation sites where the UAS CYC normally resides (Fig. 1). These constructs were introduced into four different yeast strains carrying wild-type or mutated alleles of GCN4, BAS1, and BAS2, and ␤-galactosidase expression was assayed after growing transformants in minimal medium containing or lacking adenine. As expected, expression of ␤-galactosidase in transformants of each strain bearing the parental CYC1-lacZ construct lacking UAS CYC1 was very low (ϳ10 units) and essentially unaffected by adenine supplementation or mutations in any of the regulatory genes (Table III, line 19). The ADE5,7 promoter fragments exhibited a wide range of expression from the CYC1 promoter in the wild-type strain, and in many cases promoter activity was repressed by adenine in the medium. As seen with the authentic ADE5,7 promoter (Table II), expression depended upon the wild-type alleles of BAS1 and BAS2 but not upon GCN4 (Table  III). In addition to these adenine-regulated constructs, promoter fragments with 3Ј end points at positions Ϫ183 (Table  III, lines 3 and 7) or at Ϫ160 (Table III, lines 2 and 10) exhibited significant activity that was independent of GCN4, BAS1, and BAS2 and was largely unaffected by adenine in the medium. Fig. 1 presents the ␤-galactosidase activities of the constructs analyzed in Table III after subtracting the promoter activity that was independent of BAS1, BAS2, and GCN4 (see Fig. 1 legend). The constructs containing both the distal and proximal BAS1 binding sites (3) exhibited the highest expression under derepressing conditions (Fig. 1A, lines 1, 2, 4, 5, and 6). In addition, each of these constructs showed substantial repression by adenine, with repression ratios ranging from 8.4 to 44. One notable difference among this group of constructs was observed in the comparison between pR136 and pR148, where removal of the 12 nucleotides between Ϫ145 and Ϫ133 led to a roughly 5-fold increase in expression under both growth conditions (compare lines 4 and 5). We attribute this difference to the removal of a negative element that is either located between positions Ϫ133 and Ϫ145 or was formed by the novel junction between position Ϫ133 in the ADE5,7 fragment and the CYC1 promoter sequences. Because the magnitude of adenine repression was not diminished by removal of the Ϫ133 to Ϫ145 interval, no additional studies were carried out on these sequences. Truncating the ADE5,7 fragment from position Ϫ145 to Ϫ160 (compare lines 5 and 6) led to a decrease in UAS ADE5,7 function by roughly a factor of 2, suggesting that a positive element resides within this interval. As shown below, the Ϫ145 to Ϫ160 interval contains a binding site for BAS2.

Identification of a Minimal ADE5,7 Promoter Fragment with UAS Function-The results in
The two constructs lacking the gene-proximal BAS1 site but containing the distal site (Fig. 1A, lines 3 and 7) showed very low unregulated activity, suggesting that the proximal BAS1 site is critical for UAS ADE5,7 function. In accord with this idea, three constructs containing the proximal site but lacking the distal BAS1 site (lines 8 -10) showed substantial promoter activity that was adenine-repressible. As a group, these last three constructs showed weaker UAS ADE5,7 function than the constructs containing both BAS1 sites, suggesting that the presence of two BAS1 sites confers greater UAS ADE5,7 function than occurs with the proximal site alone. As above, removal of positions Ϫ160 to Ϫ145 from pR145 decreased UAS ADE5,7 function, consistent with the presence of a positive element in this region.
The results obtained with constructs pR151, pR150, and pR149 (lines [11][12][13] revealed that the region adjacent to the proximal BAS1 site is not sufficient for high level UAS ADE5,7 function. The very low expression shown by these constructs suggested that a critical positive element resides between positions Ϫ211 and Ϫ195. Although they exhibited very low promoter activity under derepressing conditions, it appears that all three constructs are adenine-repressible. To confirm the different contributions of the two BAS1 binding sites to UAS ADE5,7 function, we introduced point mutations into each BAS1 site in construct pR134, containing the ADE5,7 promoter fragment from position Ϫ233 to Ϫ160. A mutation that alters BAS1 binding site from TGACTC to TGAATTC was introduced into one or both of the BAS1 sites in pR134. An equivalent substitution was shown to impair in vitro binding of BAS1 to its sites in the ADE2 promoter (3). The mutation in the distal BAS1 binding site reduced expression by a factor of about 4 but did not completely eliminate adenine regulation (Fig. 1B,  pR137). In contrast, a mutation in the proximal site virtually eliminated both promoter activity and adenine regulation (pR138). The doubly mutated construct pR140 also showed very low expression, similar to pR138. Comparable results were obtained when these point mutations were introduced singly into pR136, which contains the ADE5,7 fragment extending from position Ϫ133 to Ϫ233 (Fig. 1B, lines 5-7). These findings agree with the results of the deletion analysis in Fig.  1A in showing that the two BAS1 sites are not equivalent; the proximal site is essential for promoter activity and adenine repression, whereas the distal site only augments UAS ADE5,7 function.
Identification of Multiple Positive Regulatory Elements in UAS ADE5,7 -Given the absence in the literature of defined regulation of an ADE5,7-lacZ fusion Yeast strains were transformed with plasmid pR116 containing an ADE5,7-lacZ fusion on a URA3 CEN plasmid and were grown in SD medium with and without adenine (see "Experimental Procedures"). The bas1-2 and bas2-2 deletion alleles have been described (2). Extracts prepared from each transformant were assayed for ␤-galactosidase activity. The specific activity of ␤-galactosidase is expressed as nmol binding sites for BAS2 at ADE5,7 and ADE2 and our identification of a positive element located between the two BAS1 binding sites, we decided to map all the DNA elements in the UAS ADE5,7 critical for promoter function and adenine regulation by saturation mutagenesis. To accomplish this goal, we introduced random substitutions into the 22 successive groups of three nucleotides that span the 67-nucleotide ADE5,7 fragment in pR145. This construct was chosen because it contains the smallest fragment that conferred high level promoter activity that was strongly repressed by adenine (Fig. 1A, line 9).
The presence of a single BAS1 binding site in this construct simplified our analysis of the minimal cis-acting requirements for UAS ADE5,7 function. In all, 59 clustered substitution mutations were assayed for their effects on ␤-galactosidase expression in BAS1 BAS2 GCN4 strain AY854, under both adeninerepressing conditions and derepressing conditions. The results shown in Table IV indicated that substitutions in four regions, defined by constructs 3-19, 26 -36, 39 -43, and 47-52, reduced expression from that given by the parental construct. In all cases, expression was decreased to a greater extent under derepressing versus repressing conditions, leading to a reduction in the repression ratio. Fig. 2 summarizes the results obtained under derepressing conditions for representative samples taken from Table IV, normalized to the level of expression seen for the parental wild-type construct. It can be seen that mutations in three of the four regions severely impaired expression under derepressing conditions (boxed sequences I-III), whereas mutations in the fourth region decreased expression by only about 40 -50%. Mutations outside of these four regions had little (Ͻ25%) or no effect on expression. Region I is defined by mutations in clusters numbered 2-6, spanning positions Ϫ208 to Ϫ194 (Fig. 2), that lowered expression under derepressing conditions from 120 units to 6 -16 units. These were the only clustered substitutions mutations that also substantially decreased expression under repressing conditions, lowering it from 10 units to 3-5 units (Table IV, lines 3-19). Thus, most of the mutations in region I reduced the magnitude of adenine repression but did not abolish it, decreasing the repression ratio from ϳ7 (measured for the parental construct) to values of 2-4-fold (Table IV). Region I is coincident with the interval between Ϫ211 and Ϫ195 whose deletion led to a dramatic reduction in UAS ADE5,7 function. Thus, the results of deletion and substitution mutations agree in suggesting that region I stimulates UAS ADE5,7 function by roughly an order of magnitude, without being absolutely required for regulation by adenine.
The sequence in region I conforms exactly to the consensus binding site for ABF1 (29) (5Ј-CGTNNNNRRYGAY-3Ј in which R represents any purine, Y represents any pyrimidine, and N represents any nucleotide (30)), and the data in Table IV are consistent with the idea that region I functions as an ABF1 binding site. Mutations in clusters 2-3 and 5-6 alter the conserved positions in this bipartite consensus sequence and have a large effect on expression, whereas mutations in cluster 4 alter the nonconserved positions in the consensus sequence (NNR) and have a relatively weak effect on promoter activity (Table IV).
Region II is defined by mutations in clusters 10 -13, spanning positions Ϫ184 to Ϫ173 (Fig. 2), of which the most severe decreased promoter function under derepressing conditions by FIG. 1. Deletion and substitution mutations defining the UAS ADE5,7 . The top section shows a schematic of the 139-nucleotide fragment from the ADE5,7 promoter with nucleotide positions numbered relative to the initiation codon. Hatched and solid boxes represent the distal and proximal BAS1 binding sites, respectively, that contain the core sequence TGACTC (3). The distal site is located between Ϫ217 and Ϫ212, and the proximal site is located between Ϫ184 and Ϫ179. A, constructs used in the deletion analysis. PCR fragments with 5Ј end points at positions Ϫ271, Ϫ233, Ϫ211, or Ϫ195 and 3Ј end points at Ϫ183, Ϫ160, Ϫ145, or Ϫ133 were generated using oligonucleotides listed in Table I and are indicated by the lines shown below the schematic. These fragments were inserted in place of UAS CYC1 in plasmid pLG669Z⌬XhoI, and the resulting constructs were introduced into yeast strain AY854 (wild type). ␤-Galactosidase activities in whole cell extracts were determined after growing the transformants under derepressing (ϪAde) or repressing conditions (ϩAde) and are listed in Table III. These results have been corrected for UAS activity that is not attributed to the three transcriptional activators GCN4, BAS1, or BAS2 by subtracting the activity found in strain AY862 (gcn4 bas1 bas2) and are shown to the right of each construct. In cases where the corrected value was a negative number, it is enclosed by parentheses. NA, not applicable. B, constructs used to analyze mutations in the BAS1 binding site. A point mutation from TGACTC to TGAATTC in the core sequence is represented by an X. about 10-fold. There was little change from the wild-type level under repressing conditions (Table IV, lines 26 -36). A complete loss of adenine regulation distinguishes the mutations in this region from mutations in the other three regions. The first two clusters in region II, TGA and CTC, alter the core sequence of the proximal BAS1 binding site described previously at ADE5,7 (3) and identified above as being critical for UAS ADE5,7 function. As expected, all six substitutions in these two clusters essentially abolished expression and adenine repression. Mutations affecting the next two clusters in region II were also very deleterious, including the substitution of TCA for GTG at cluster 12 and GGA for TCC at cluster 13 ( Fig. 2 and lines 34 and 35 in Table IV). We propose that all of the mutations in the 12-nucleotide sequence comprising region II impair UAS ADE5,7 function by reducing BAS1 binding to the ADE5,7 promoter.
Region III consists of clusters 15 and 16, spanning Ϫ169 to Ϫ164 (Fig. 2). The most severe mutations in this region decreased expression from 120 units to 13-40 units under derepressing conditions but had little or no effect under repressing conditions (Table IV, lines 39 -40, 42, and 43). The sequence and location of this region, TAATAA, resembles the binding site described for BAS2 at HIS4 (8). The sequence of region IV comprising clusters 18 -20, CAATAATGA, is related to that of region III except that two of the three TAA repeats are degenerate. Mutations in region IV have the weakest effects, decreasing expression under derepressing conditions by 50% or less (Table IV, lines 47-52). On the basis of their sequence and juxtaposition 3Ј to the BAS1 site, we conclude provisionally that regions III and IV function as strong and weak binding sites, respectively, for BAS2.
None of the clustered substitutions in the UAS ADE5,7 led to high, unregulated promoter activity. The greatest level of expression under adenine-repressing conditions was observed for mutations in clusters 19 and 21 (Table IV, lines 50 -51 and 53-56). The ϳ2-fold increase above the repressed values elicited by these mutations clearly does not account for the 6.7-fold repression ratio seen with the parental construct. This result is noteworthy because it suggests that adenine repression does not involve binding of a repressor protein to an upstream repression sequence. If such a repressor existed, its binding site would have to coincide with one of the four positive control sites we identified in UAS ADE5,7 where, as shown below, ABF1, BAS1, or BAS2 binds. Given that only mutations in the BAS1 site completely abolished adenine regulation, a hypothetical repressor would probably have to compete with BAS1 for binding to its site in region II of UAS ADE5, 7 .
Region I in UAS ADE5,7 Functions as a Binding Site for ABF1-Results from the mutational studies described above suggest that region I of UAS ADE5,7 functions as a binding site for ABF1. ABF1 is an abundant, essential yeast protein first shown to bind DNA at autonomous replicating sequences (ARSs) involved in replication (31). Later, ABF1 was shown to interact with telomeres, the silent mating cassettes, and the promoters of various metabolic genes (29,30,32). To demonstrate that region I is a binding site for ABFI, we performed EMSAs to detect DNA binding. Whole cell extracts containing ABF1 were prepared from a gcn4 bas1 bas2 yeast strain (AY862) and incubated with radioactively labeled oligonucleotide probes containing the ABF1 binding site from ARS1 or UAS ADE5,7 . As shown in Fig. 3A, the extract contains a binding activity for both the ARS1 (lane 2) and ADE5,7 (lane 6) oligonucleotides. The addition of a 100-fold molar excess of unlabeled ARS1 oligonucleotides eliminated binding to the labeled ARS1 probe (lane 3) and to a large extent in the case of the ADE5,7 probe (lane 7). A 200-fold molar excess of unlabeled ARS1 oligonucleotide abolished binding to the ADE5,7 probe (data not shown). The tighter binding to ADE5,7 is consistent with the observation that the ABF1 binding site in the ADE5,7 fragment has greater homology to the derived consensus sequence for ABF1 binding than does the ARS1 sequence. The experiment was repeated using oligonucleotides containing a double point mutation in the ARS1 sequence (33) as the competitor DNA. This mutant oligonucleotide failed to compete with either of the labeled probes for protein binding (Fig. 3,

TABLE III
Analysis of UAS ADE5,7 function conferred on a CYC1-lacZ fusion by different fragments from the ADE5,7 promoter Plasmids containing either no insert (line 19, vector pLG699Z⌬XhoI) or containing various fragments from the ADE5,7 promoter (lines 1-18, plasmids pR133 to pR152) were introduced into strains AY854 (GCN4 BAS1 BAS2), AY860 (GCN4 bas1-2 bas2-2), AY957 (gcn4 BAS1 BAS2), and AY862 (gcn4 bas1-2 bas2-2), and the transformants were assayed for ␤-galactosidase activities after growth under repressing (ϩAde) and derepressing (ϪAde) conditions, as described in Table II (GCN4, BAS1, and BAS2) was transformed with each of the constructs listed and assayed for ␤-galactosidase as described for Table  II. Expression from each construct was assayed in duplicate on at least two independent transformants, and the values shown have standard errors less than 30%. Plasmid pCM81 is the empty vector and pR224 is the parental wild-type construct containing ADE5,7 sequences Ϫ211 to Ϫ148 inserted between the BglII and XhoI sites in pCM81. The remainder of the constructs are identical to pR224 except for the mutations listed in column 4 ("Mutant sequence"). The wild-type sequence for each cluster is listed in column 3 in the row with the first mutation in that cluster. Column 5 lists the number of changes necessary to convert the wild-type sequence to the mutant one. The boldface type corresponds to the regions discussed in the text: region I, clusters 2-6; region II, clusters 10 -13; region III, clusters 15-16; region IV, clusters 18 -20. The 5Ј-most nucleotide in each cluster is as follows: cluster 1, Ϫ211; cluster 2, Ϫ208; cluster 3, Ϫ205; cluster 4, Ϫ202; cluster 5, Ϫ199; cluster 6, Ϫ196; cluster 7, Ϫ193; cluster 8, Ϫ190; cluster 9, Ϫ187; cluster 10, Ϫ184; cluster 11, Ϫ181; cluster 12, Ϫ178; cluster 13 produced a supershift of the DNA-protein complexes, whether the labeled probe contained ADE5,7 sequences (Fig. 3B, lanes [3][4][5] or ARS1 sequences (data not shown). These results prove that ABF1 binds to the ADE5,7 fragment under these in vitro conditions. To demonstrate that the mutations in the presumed ABF1 binding site in the UAS ADE5,7 that impair promoter activation in vivo also reduce ABF1 binding in vitro, we performed EMSA using as probes either the wild-type UAS ADE5,7 fragment or mutant fragments bearing selected clustered substitutions described in Table IV. Fig. 3C shows that ABF1 binds with nearly equal efficiency to the wild-type ADE5,7 fragment (lanes 1 and 9), to a fragment bearing mutations with no effect on in vivo UAS ADE5,7 function (lane 5), and to fragments containing mutations that impair UAS ADE5,7 but lie outside of the presumed ABF1 binding site (lanes 6 -8 and 14 -16). In contrast, ABF1 did not bind to fragments containing mutations in the conserved positions of the consensus ABF1 binding site in UAS ADE5,7 (lanes 2, 4, 10, 12, and 13). Moreover, mutations that alter the central nonconserved positions of the consensus binding site had little or no affect on ABF1 binding (lanes 3 and 11), consistent with the known sequence requirements for ABF1 binding. We performed complementary experiments in which mutant and wild-type unlabeled fragments were compared for their ability to compete with labeled wild type ADE5,7 probe for binding of ABF1 (data not shown). These results established a perfect agreement between the effects of mutations in the ABF1 consensus binding site on ABF1 binding in vitro and UAS ADE5,7 function in vivo. We conclude that ABF1 binds to region I in vivo and is responsible for the positive role of this sequence in the UAS ADE5, 7 .
Region II in UAS ADE5,7 Functions as a 12-Base Pair Binding Site for BAS1-We proposed that mutations in region II, including the six nucleotides located 3Ј to the TGACTC core, which are not conserved among different BAS1 binding sites, destroy UAS ADE5,7 function by interfering with binding of BAS1. To verify this, we investigated whether mutations in all four of the triplets composing region II interfere with BAS1 binding to the ADE5,7 fragment in vitro. We purified a GST-BAS1 fusion protein expressed in E. coli and used it in EMSA experiments with wild-type or mutant ADE5,7 probes. As shown in Fig. 4A, GST-BAS1 bound to the ADE5,7 fragment (lane 3), whereas GST alone did not (lane 2). Mutations in clusters 10 -13 reduced or eliminated the ability of GST-BAS1 to bind to the ADE5,7 DNA (Fig. 4B, lanes 5-8), whereas mutations outside of region II had no effect on binding ( lanes  2-4, 9, and 10). Mutations in the TGACTC core sequence (lanes 5 and 6) and in the adjacent hexanucleotide sequence GTGTCC (lanes 7 and 8) Table IV on the effects of mutating successive triplets in the 67-nucleotide sequence from position Ϫ211 to Ϫ148 of the ADE5,7 promoter. The wild-type sequences of the triplet clusters numbered as shown in Table IV are given along the bottom of the graph with the clusters assigned to regions I-IV enclosed in separate boxes (see text for details). The ␤-galactosidase expression measured under derepressing (without Ade) conditions for selected mutants (numbers 2, 3,9,11,13,18,20,21,23,26,30,34,35,37,39,43,46,49,50,52,54, and 57 as listed in Table IV) is expressed as a percentage of that given by the parental wild-type construct pR224. ADE5,7 . A, binding of ABF1 in whole cell extracts to radiolabeled oligonucleotide probes was analyzed by EMSA. Lanes 1-4 show the results obtained with a probe containing the ABF1 binding site in ARS1; for lanes 5-8, the probe contained nucleotide positions Ϫ211 to Ϫ145 from the ADE5,7 promoter. For lanes 1 and 5, no yeast extract was added; for all others, 10 g of extract from strain AY862 was added. For lanes 3 and 7, a 100-fold molar excess of the unlabeled ARS1 oligonucleotide was used as competitor; for lanes 4 and 8, the same amount of a mutant unlabeled ARS1 oligonucleotide was used as competitor. The mutation in the ARS1 fragment is a two-nucleotide substitution, replacing AGCATTTTTGACG with AGCATTTTTGGTG (underlines indicate the changes from the ABF1 consensus), shown previously to abolish ABF1 binding (33). B, formation of a supershifted complex with ABF1 antibodies. ABF1-DNA complexes were formed using the labeled ADE5,7 probe as described in A, and no serum was added to the reaction (lane 2), 1 l of a 1:10 dilution of preimmune serum was added (lane 3), or 1 l of a 1:100 dilution or of a 1:10 dilution of a polyclonal serum prepared against partially purified ABF1 (28) was added (lanes 4 and 5, respectively), prior to electrophoresis. C, binding of ABF1 to wild-type and mutant fragments from ADE5,7. Labeled oligonucleotides identical to the ADE5,7 probe describe above (lanes 1 and 9) or containing point mutations in the clusters indicated along the top were employed in EMSAs as described above for A. The mutations in clusters 2, 3, 5, and 6 alter conserved positions in the predicted ABF1 binding site and correspond to the mutations introduced in constructs shown as numbers 5, 7, 13, 14, and 19 in Table IV. The mutations in cluster 4 (corresponding to constructs 10 and 11 in Table IV) alter only nonconserved positions in the ABF1 site. Cluster 7 lies outside the ABF1 binding site, and the mutations analyzed at these positions had no effect on UAS ADE5,7 function (construct 20 in Table IV). The mutations analyzed in clusters 10 -13, 15, and 16 led to reduced UAS ADE5,7 function but altered clusters outside of the ABF1 binding site constructs 27,29,34,35,39, and 43 in Table IV). sequence (constructs 26, 30, and 32 in Table IV) showed strongly diminished or no in vitro binding of GST-BAS1 (data not shown).

FIG. 3. In vitro binding of ABF1 protein in yeast extracts to UAS
To confirm these results, we compared the ability of unlabeled mutant fragments present in 200-fold molar excess to compete with the labeled wild-type fragment for binding to GST-BAS1. These results, shown in Fig. 4C, completely mirrored those presented in Fig. 4B, in that only the mutations in clusters 10 -13 of region II diminished the ability to compete with the wild-type ADE5,7 fragment for binding to GST-BAS1. Moreover, the mutations in the TGACTC core sequence (lanes 10 and 11) had a greater effect than did those in the adjacent hexanucleotide sequence (lanes 12 and 13). We conclude that region II comprises a 12-bp binding site for BAS1 and that all of the mutations in this region reduce UAS ADE5,7 function by decreasing BAS1 binding to the promoter.
Evidence That Regions III and IV in UAS ADE5,7 Function as BAS2 Binding Sites-Based on their A ϩ T-rich sequence and location 3Ј to the BAS1 binding site, we suggested above that regions III and IV are binding sites for BAS2. To test this idea, we carried out EMSA experiments using a GST-BAS2 fusion protein expressed in E. coli to determine whether mutations in clusters 15 or 16 reduced binding of BAS2 to the ADE5,7 fragment. GST-BAS2 bound specifically to the wild-type ADE5,7 fragment (Fig. 4A, lane 4), whereas mutations in cluster 15 or 16 of region III significantly reduced binding of this interaction (Fig. 5A, lanes 1, 5, and 6). In an effort to verify these results, we carried out EMSA experiments comparing wild-type and mutant competitors using a purified polyhistidine-tagged form of BAS2 expressed in E. coli (13,15). We saw small reductions in the ability of fragments containing mutations in clusters 15 and 16 or clusters 18 and 19 relative to the wild-type ADE5,7 fragment to compete with the labeled wildtype probe for (His) 6 -BAS2 binding (data not shown). We then combined these mutations in the same fragment and compared it with the wild-type fragment for ability to compete. As shown in Fig. 5B, we found that significantly greater amounts of the fragment containing mutations in clusters 15-19 were required to achieve the same degree of competition for (His) 6 -BAS2 binding to ADE5,7 DNA. These results support the idea that sequences in regions III and IV function as binding sites for BAS2. The fact that mutations in clusters 15 and 16 have only modest effects on BAS2 binding in vitro but substantially re-duce UAS ADE5,7 function in vivo suggests that the in vitro binding reactions do not duplicate some important aspect of the in vivo interaction of BAS2 with the ADE5,7 promoter.

DISCUSSION
Evidence That Binding Sites for BAS1, BAS2, and ABF1 Are the Critical Elements of the UAS ADE5,7 -Our results define the minimum promoter elements required for adenine-regulated transcription of ADE5,7 and provide a detailed picture of the spatial relationships among the DNA-binding proteins that interact with these sequences. These results provide the first complete description of a UAS from an adenine biosynthetic gene. Wild-type promoter activity and adenine repression require (i) two binding sites for BAS1 that are separated in the promoter by about 20 nucleotides, (ii) a single binding site for ABF1 located immediately 3Ј to the distal BAS1 site, and (iii) two A ϩ T-rich elements located immediately 3Ј to the proximal BAS1 site to which monomers of BAS2 appear to bind, all within a region of only ϳ65 bp (Fig. 6). A 52-bp fragment containing a single binding site for each protein was sufficient to confer adenine-repressible activation, UAS ADE5,7 function, on the CYC1 basal promoter.
The conclusion that regions I and II function in UAS ADE5,7 as binding sites for ABF1 and BAS1, respectively, is based on the strong correlation between the effects of mutations in these sequences on protein binding in vitro and on UAS ADE5,7 function in vivo. The results of clustered substitution mutagenesis of region II provides in vivo evidence that BAS1 recognizes between 4 and 6 bp 3Ј to the highly conserved TGACTC core element present in all known BAS1 sites. Similarly, Høvring et al. (10) showed that BAS1 recognizes the G residues on both strands of the sequence CGG located immediately 3Ј to the TGACTC core at the BAS1 binding site in the HIS4 promoter. This particular sequence is not conserved in either of the BAS1 binding sites at ADE5,7 or ADE2. Based on our findings, we suggest that sequence differences in the 4 -6 bp on the 3Ј side of the TGACTC core element lead to differences in BAS1 binding affinity.
Our proposal that regions III and IV function as BAS2 binding sites is reasonable given that BAS2 is required for ADE5,7 promoter activation and that these are the only cis-acting elements in addition to the BAS1 and ABF1 sites that are critical for UAS ADE5,7 function in vivo. The BAS2 binding sites have FIG. 4. In vitro binding of a GST-BAS1 fusion protein to UAS ADE5,7 . A, binding of GST fusion proteins to a radiolabeled ADE5,7 probe (described in Fig. 3 Table IV) led to reduced UAS ADE5,7 function and lie outside of the BAS1 binding site but within the ABF1 binding site. The mutation in cluster 7 (number 20 in Table IV) had no effect on UAS ADE5,7 function and lies outside of the binding site for BAS1. The mutations in clusters 10 -13 alter the proximal TGACTC and the adjacent six nucleotides to the 3Ј side (numbers 27, 29, 34, and 35 in Table IV). The mutations in clusters 15 and 16 (numbers 39 and 43 in Table IV) decrease UAS ADE5,7 function and lie outside of the BAS1 binding site. C, competition by unlabeled ADE5,7 fragments for binding of GST-BAS1. EMSAs were performed as for A above using radiolabeled ADE5,7 DNA, GST-BAS1, and various unlabeled DNAs as competitors. For lane 1, no competitor was added; for lanes 2-10 competitors as indicated along the top of the figure were added a in 200-fold molar excess. The mutants were the same as described for B.
not been identified in any other ADE gene promoter; however, the high A ϩ T content of the region III and IV sequences and their location 3Ј of the proximal BAS1 binding site are both reminiscent of the juxtaposed BAS1 and BAS2 binding sites at HIS4 (8). The identification of the BAS2 binding sites at ADE5,7 is not definitive because mutations in regions III and IV that greatly impaired activation in vivo had relatively small effects on DNA binding in vitro, and it was necessary to combine multiple substitutions in these sequences before achieving a significant reduction in BAS2 binding. One possible explanation for these results is that purified BAS2 protein may not bind to ADE5,7 DNA with high sequence specificity. Perhaps specific binding of BAS2 to regions III and IV is stabilized in vivo by its interaction with BAS1 bound at the adjacent sequences at positions Ϫ184 to Ϫ173 (Fig. 6). As discussed below, we recently obtained strong evidence for complex formation by BAS1 and BAS2 (9). Although no cooperativity was detected between BAS1 and BAS2 in binding to the HIS4 promoter (8), the BAS2 site defined at HIS4 (repeats of TTAA; Ref. 8) differs from that proposed here for ADE5,7 (TAATAA and CAATAA). It is possible that BAS2 has a greater affinity for the HIS4 sequences and does not require cooperativity with BAS1 to bind there with high specificity.
We recently obtained evidence for interactions between BAS1 and BAS2 by showing that a LexA-BAS1 fusion protein can recruit BAS2 to a promoter containing lexA operator se- FIG. 5. In vitro binding of purified recombinant BAS2 to UAS ADE5,7 . A, binding of purified GST-BAS2 to radiolabeled oligonucleotide probes was analyzed by EMSA. Lane 1 shows the results obtained with a probe containing ADE5,7 sequences from Ϫ211 to Ϫ145 as described above in Fig. 4. For lanes 2-10, probes contain mutations in the clusters indicated at the top. The mutations in clusters 3, 5, and 10 -13 (numbers 7, 13, 27, 29, 34, and 35, respectively, in Table IV) reduce UAS ADE5,7 function but lie in regions for binding ABF1 and BAS1. The mutation in cluster 7 does not affect UAS ADE5,7 function (number 20 in Table IV). Mutations in clusters 15 and 16 (numbers 39 and 43 in Table IV) reduce UAS ADE5,7 function and lie in an A ϩ T-rich region located to the 3Ј side of the proximal BAS1 binding site. B, competition by wild-type or mutated ADE5,7 fragments for binding to (His) 6 -BAS2. Binding reactions were performed with purified (His) 6 -BAS2 and the radiolabeled ADE5,7 fragment described for A. The top part shows the results obtained when wild-type DNA, as described for A, was added as competitor, and the lower part shows the results obtained when a multiply mutated DNA was added as competitor. The multiply mutated ADE5,7 fragment carries changes in clusters 15 and 16 (TAATAA to ATCTCC) and in clusters 18 and 19 (CAATAA to GACCCG). For lanes 1 no (His) 6 -BAS2 was added, and for lanes 2 no competitor DNA was added to the reaction. Increasing concentrations of competitor DNA in 10-, 20-, 40-, 80-, 160-, and 320-fold molar excess was added to the samples in lanes 3-8, respectively. C, graphical representation of the data from panel B. Plotted is the percentage of bound probe versus an increasing concentration of competitor DNA. The closed circles represent the wild-type competitor, and the open circles represent the multiply mutant competitor. The amount of binding when no competitor was added is set at 100%. The amount of competitor DNA required to see a 50% loss in binding is at 27-fold molar excess of the wild-type DNA and at 160-fold molar excess of the multiply mutated DNA. quences (9). This interaction at the lexA operator occurred even when the DNA binding activity of BAS2 was destroyed by mutations in its homeodomain. These mutations impaired activation of ADE5,7 transcription, however, indicating that the DNA binding activity of BAS2 was required for its binding to the native ADE5,7 promoter. To explain these observations, we propose that BAS1 cannot bind to the ADE5,7 promoter as effectively as LexA-BAS1 binds to the lexA operator and that an interaction between BAS1 and BAS2 is required for these proteins to bind with high affinity to their adjacent sites at ADE5, 7. We propose that the BAS2 binding sites at ADE5,7 are relatively weak and that cooperativity with BAS1 is required for stable binding of BAS2 to regions III and IV. Similarly, stable binding of BAS1 at the ADE5,7 proximal site depends on cooperativity with BAS2. In contrast, both proteins can interact more effectively with their respective binding sites at HIS4, and cooperativity is not required for promoter occupancy.
Possible Functions for ABF1 and the Distal BAS1 Binding Site in the ADE5,7 Promoter-Our results showed that the two BAS1 binding sites in the ADE5,7 promoter are not equivalent. Mutations in the distal site led to an approximately 6-fold loss in expression but no obvious impairment of adenine regulation. In contrast, the proximal BAS1 binding site was absolutely required for promoter activity and regulation. Given that BAS1 and BAS2 must interact to form a potent activation complex (9), we propose that juxtaposition of BAS2 binding sites on the 3Ј side of the proximal BAS1 site allows formation of the functional BAS1-BAS2 complex. The fact that the distal BAS1 site is flanked by an ABF1 site rather than a BAS2 binding site may explain the failure of the latter to support UAS ADE5,7 function without the proximal site. In addition to ADE5,7, the genes ADE2, ADE4, ADE1, and ADE8 have two TGACTC core sequences that are potential BAS1 binding sites, and it was shown that BAS1 binds in vitro to both sites at ADE2. The proximal TGACTC sequence is more important for activation at ADE2 (3,4,6), whereas the distal site is more critical at ADE4 (34). It will be interesting to learn whether the more potent sites at these genes are uniquely flanked by BAS2 binding sites, in the manner we suggest for ADE5,7.
At ADE5,7, the two BAS1 binding sites are separated by 33 nucleotides, which is 3.1 helical turns. The two TGACTC core sequences in the other ADE genes are also separated by an integral number of helical turns (with the exception of the ADE1 gene), placing the BAS1 binding sites on the same face of the DNA helix. McBroom and Sadowski (35) have shown that ABF1 is able to bend DNA 120°around itself. If this occurred at the ADE5,7 promoter, then binding of ABF1 could bring the distal and proximal BAS1 sites into close proximity. Perhaps the distal BAS1 site has a higher affinity for BAS1, and the bend in the DNA introduced by ABF1 allows BAS1 bound at the distal site to stabilize binding of the BAS1-BAS2 complex to the proximal site. Alternatively, the bend in the DNA might allow BAS1 bound at the distal site to cooperate with the BAS1-BAS2 complex in promoting assembly of the transcription initiation complex, although it is incapable of carrying out this function alone.
Because the distal BAS1 site is less important than the ABF1 site for high level activation, ABF1 must perform a function distinct from promoting interactions between BAS1 molecules bound at the distal and proximal sites at ADE5,7. Perhaps ABF1 is required to mark the region as a promoter, and by bending the DNA keeps UAS ADE5,7 free of nucleosomes. It has been proposed that RAP1 binding in the vicinity of BAS1 and BAS2 binding sites performs precisely this function in the HIS4 promoter (36). The proximity of the distal BAS1 site and the ABF1 site at ADE5,7 (Fig. 6) raises the possibility that binding of BAS1 and ABF1 is mutually exclusive. If so, ABF1 may be displaced by binding of BAS1 to the distal site in the course of assembling the transcription initiation complex. Additional in vitro studies using purified BAS1, BAS2, and ABF1 will be required to test this idea and to address the possibility that BAS1-BAS1, BAS1-BAS2, or BAS2-BAS2 interactions stabilize the binding of these proteins to UAS ADE5, 7 .
The Mechanism of Adenine Repression of BAS1 and BAS2 Function-Our mutational analysis provided no evidence for a DNA binding repressor that would mediate adenine repression of ADE5,7 transcription. Mutations in the binding site for such a repressor would be expected to produce constitutively derepressed expression, whereas all of the mutations we generated led to reduced UAS ADE5,7 function. While it is possible that a repressor protein binds to UAS ADE5,7 with the same sequence requirements as ABF1, BAS1, or BAS2, this seems unlikely. Moreover, the fact that only mutations in the BAS1 site completely abolished adenine repression probably indicates that the hypothetical repressor would have to compete with BAS1 for binding to its proximal site. A simpler explanation for our findings is that adenine regulation occurs by covalent modification of BAS1 or BAS2, or the binding of a negative regulator to one of these two proteins, that inhibits their ability to interact with DNA or activate transcription. In addition, because BAS2 regulates expression of promoters unrelated to purine biosynthesis, it seems most likely that BAS1 is the target of this negative regulation. Our results using LexA-BAS1 fusions eliminated the possibility that the DNA binding domain of BAS1 is the target of adenine repression (9). Instead, we found that complex formation between BAS1 and BAS2 unmasks the latent activation function of BAS1 and is down-regulated by adenine. It is possible that the activation function of the BAS1-BAS2 complex is also negatively regulated by adenine. Several mechanisms can be proposed to explain how the ability of BAS1 to interact with BAS2 is negatively regulated, including phosphorylation, binding of a negative regulator, or prevention of its entry into the nucleus. Efforts are under way to distinguish between these possibilities and to identify other trans-acting factors that might be involved in this regulatory mechanism.
FIG. 6. Model for protein binding at ADE5,7. Sequence of the upper strand of ADE5,7 DNA from position Ϫ228 to Ϫ145 is shown in the rectangle. The shaded ovals represent two molecules of BAS1 bound at the distal (to the left) and proximal (region II) binding sites. The open oval represents ABF1 bound at region I. The two hatched ovals represent BAS2 bound at the strong (region III) and weak (region IV) sites. Binding sites for each protein are indicated by the same markings as the proteins. BAS1 and BAS2 are shown interacting with one another at the proximal BAS1 binding site (see text for details).