Evidence That Intramolecular Interactions Are Involved in Masking the Activation Domain of Transcriptional Activator Leu3p*

The Leu3 protein of Saccharomyces cerevisiae regulates the expression of genes involved in branched chain amino acid biosynthesis and in ammonia assimilation. It is modulated by α-isopropylmalate, an intermediate in leucine biosynthesis. In the presence of α-isopropylmalate, Leu3p is a transcriptional activator. In the absence of the signal molecule, the activation domain is masked, and Leu3p acts as a repressor. The recent discovery that Leu3p retains its regulatory properties when expressed in mammalian cells (Guo, H., and Kohlhaw, G. B. (1996) FEBS Lett. 390, 191–195) suggests that masking and unmasking of the activation domain occur without the participation of auxiliary proteins. Here we present experimental support for this notion and address the mechanism of masking. We show that modulation of Leu3p is exceedingly sensitive to mutations in the activation domain. An activation domain double mutant (D872N/D874N; designated Leu3-dd) was constructed that has the characteristics of a permanently masked activator. Using separately expressed segments containing either the DNA binding domain-middle region or the activation domain of wild type Leu3p (or Leu3-dd) in a modified yeast two-hybrid system, we provide direct evidence for α-isopropylmalate-dependent interaction between these segments. Finally, we use the phenotype of Leu3-dd-containing cells (slow growth in the absence of added leucine) to select for suppressor mutations that map to the middle region of Leu3-dd. The properties of nine such suppressors further support the idea that masking is an intramolecular process and suggest a means for mapping the surface involved in masking.

The Leu3 protein of yeast belongs to a class of transcriptional regulators whose members are characterized by a Zn(II) 2 -Cys 6 binuclear cluster in their DNA binding domains (1,2). Many of them also have the ability to be modulated, i.e. to respond to metabolic signals. Well studied examples of modulation include Gal4p, which is activated ("unmasked") by galactose in a process that involves changes in the interaction between Gal4p and an auxiliary protein known as Gal80p (3,4), and Hap1p, the activation of which by heme is thought to be brought about by dissociation of cellular factors that allow the DNA binding domain to become functional and its activation domain (AD) 1 to become exposed (5,6). Leu3p is modulated by ␣-IPM, 1 an intermediate in leucine biosynthesis (7). Leu3p binds to UAS L elements in the promoters of genes involved in the biosynthesis of branched chain amino acids (8,9) and, surprisingly, in ammonia assimilation in yeast (10). The finding that the GDH1 gene is regulated by Leu3p and ␣-IPM (10) led to the hypothesis that ␣-IPM serves as a signal molecule that communicates the status of amino acid biosynthetic activity (as represented by the leucine pathway) to more central points of nitrogen metabolism. It is of interest in this context that the LEU3 gene itself is controlled by Gcn4p (11).
Leu3p binds to target DNA irrespective of the presence or absence of ␣-IPM (9). In the absence of ␣-IPM, Leu3p represses gene expression below the basal level seen in leu3 null cells (9,12). In the presence of ␣-IPM, it causes strong activation. The zinc cluster part of Leu3p's DNA binding domain is located between amino acids 37 and 67 of the 886-residue protein (11,13,14), and a putative dimerization domain is found between amino acids 85 and 99 (14). A region responsible for transcriptional activation has been identified within the C-terminal 30 amino acids of Leu3p (14,15). Truncated Leu3p molecules lacking the AD still bind to DNA. They are totally devoid of activation potential and act as repressors (9,15,16). Deleting as many as 600 amino acids from the middle of Leu3p leaves DNA binding and transcriptional activation functions intact but eliminates modulation (16,17). Such molecules have a constitutively high activation potential that usually exceeds that of modulated forms of Leu3p. The implication of these results is that Leu3p's middle region contains information that is essential for modulation. In dealing with the question of modulation, it may be helpful to break the modulation process down into discrete steps, e.g. the binding of the modulator ␣-IPM, conformational changes caused by the binding of ␣-IPM, and the eventual exposure of the AD that allows it to interact with components of the transcription apparatus. Our current view is that the unmasking of the AD as well as its masking in the absence of ␣-IPM occur without the participation of auxiliary proteins. The strongest support for this notion comes from recent observations made when full-length yeast Leu3p was expressed in mammalian cells (18). It was found that the behavior of Leu3p in mouse cells was almost indistinguishable from that seen in its native environment; Leu3p was stable but inactive when ␣-IPM was absent and caused an apparent severalfold repression of reporter gene expression. When ␣-IPM was present in the cell culture medium, Leu3p was converted into a strong activator. Since mammalian cells do not synthesize branched chain amino acids, it appeared highly unlikely that they would elaborate Leu3p-specific factors for masking, unmasking, or ␣-IPM interaction.
In this study, we have used biochemical and molecular genetic approaches to advance our understanding of the modulation process. We show that modulation is exquisitely sensitive to mutational changes in the AD of Leu3p. Using wild type Leu3p and a mutant form with drastically intensified masking, we show that the AD and the remainder of Leu3p interact and that this interaction depends on ␣-IPM. Isolation of intragenic suppressors of the slow growth phenotype of the masking mutant shows the usefulness of this approach for identifying regions or individual residues of Leu3p that are involved in masking.

MATERIALS AND METHODS
Strains and Growth Conditions-The Saccharomyces cerevisiae strains used were DBY746 (MAT␣ leu2-3 leu2-112 trp1-289 ura3-52 his3-⌬1; YGSC), DK1 (MAT␣ leu3-⌬::LEU2 leu2-3 leu2-112 trp1-289 ura3-52 his3⌬1; see below), and XK157-3C (MAT␣ leu3-⌬2::HIS3 trp1-289 ura3-52 his3-⌬1; (16)). The latter two contain total leu3 deletions. DK1 was constructed as follows. Yeast shuttle vector pRS305 (19) was digested with ScaI and BsaI. A fragment containing the LEU2 gene was recovered and inserted into plasmid pGB12 (9) that had been cut with BamHI and rendered blunt-ended with T4 DNA polymerase. The resulting plasmid, pDW1, contained one copy of the LEU2 gene flanked by 5Ј-and 3Ј-noncoding sequences of LEU3. The orientation of LEU2 was the same as the original orientation of LEU3 at this locus. Plasmid pDW1 was digested to completion with SphI and SacI. Among the fragments generated was a 4.3-kbp fragment containing the LEU2 gene and LEU3-flanking sequences. The digestion mixture was used for integrative transformation of strain DBY746, selecting for Leu Ϯ . (Since the desired transformants carried an intact LEU2 gene, but no LEU3 gene, the phenotype was expected to change from Leu Ϫ (DBY746, no growth in the absence of added leucine) to Leu Ϯ (slow growth, about 40% that of wild type cells, Ref. 9).) Transformants that also had the His Ϫ , Ura Ϫ , and Trp Ϫ phenotype were collected and designated DK1. Their genotype was further confirmed by PCR using purified yeast chromosomal DNA and primers complementary to the LEU2 coding region and regions of the LEU3 locus flanking the LEU2 gene. Unless stated otherwise, yeast cells were grown on SD medium (20) supplemented with the required nutrients. A supplement of 2 mM leucine plus 1 mM each of valine and isoleucine was routinely used to generate low intracellular ␣-IPM levels. (The presence of valine was found not to be obligatory, and valine was therefore omitted in some experiments.) High ␣-IPM levels were generated by supplementing the medium with 0.2 mM leucine. Cells were grown at 30°C and harvested at an A 600 of about 1.
Mutagenesis of the AD of Leu3p-The majority of the mutants was generated by degenerate oligonucleotide incorporation. This approach required the construction of an expression plasmid containing a LEU3 gene with engineered restriction sites that would allow precise cassette exchange of the AD. To achieve this, new NgoMI and PmeI/SmaI sites were created by site-directed mutagenesis. E. coli CJ236 cells were transformed with plasmid pRS316-LEU3 which was constructed by cloning a LEU3-containing fragment into pRS316 (19). (The fragment extended from Ϫ561 to about ϩ4700 relative to the A at the beginning of the open reading frame of LEU3 (11).) Transformed cells were grown to mid-log phase and superinfected with helper phage M13K07 in the presence of uracil. Phage particles were collected and single-stranded DNA isolated according to a protocol from Bio-Rad. Equimolar amounts of two phosphorylated primers (5Ј CATCATGGCCGGCTGGGAT-AAC-3Ј for the 5Ј side of the AD and 5Ј-CCCAAGGTTTAAACCCGGGT-TCTTTTTTTGCG-3Ј for the 3Ј side of the AD (NgoI and PmeI/SmaI sites underlined)) were then used to mutagenize pRS316-LEU3. Note that changes made to generate the new restriction sites did not alter the amino acid sequence of Leu3p. Potential mutants were screened for the presence of an additional SmaI site, and positives were sequenced using a double-stranded DNA cycle sequencing kit from Life Technologies, Inc. A 1-kbp fragment of DNA containing new NgoMI and PmeI sites was amplified by PCR using appropriate primers. A BlnI-SmaI piece was excised from the fragment and cloned into BlnI/SmaI-digested pPC62H/86T-LEU3 from which two existing NgoM1 sites had been removed, resulting in plasmid pYHA. It contains unique NgoMI and PmeI sites that define the AD of the Leu-3 protein. Its LEU3 gene is flanked by ADC1 promoter and terminator sequences, respectively. Plasmid pPC62H/86T-LEU3 had been constructed from centromerecontaining plasmid pPC62H/86T (a gift from E. Taparowsky, Purdue University) and pT7-LEU3 (a gift from J.-Y. Sze of this laboratory). pT7-LEU3 was digested with PstI and SmaI, and the 3.1-kbp, LEU3containing fragment was inserted into pPC62H/86T that had also been cut with PstI and SmaI.
To mutagenize the AD of Leu3p, a 93-base pair oligonucleotide, 5Ј-ATGGCCGGCtgggataactgggaatctgatatggtttggagggatgttgatattttaat-gaatgaatttgcgttcaatcccaaGGTTTAAACC-3Ј (NgoMI and PmeI sites underlined) was synthesized (Integrated DNA Technologies, Coralville, IA) such that the misincorporation rate at the positions shown in lowercase letters was 4.5% (i.e. the proportion of each of the three non-native nucleotides was 1.5%) or, in a separate experiment, 2.7% (the proportion of each of the three non-native nucleotides was 0.9%). The lowercase letters correspond to amino acid positions 861-885 of Leu3p (11). The oligonucleotide mixture was incubated, heated to 85°C, then cooled to 4°C over a 1-h period. Nucleotides and Klenow enzyme were added to the annealed mixture which was then incubated on ice for 5 min, at 23°C for 5 min, and at 37°C for 20 -30 min. The extended and now double-stranded DNA was digested with NgoMI and PmeI. The final product was a mixture of monomeric DNA fragments. These were ligated into plasmid pYHA that had been digested with NgoMI and PmeI. The molar ratio of fragment to plasmid was approximately 3. Aliquots of 10 ng of ligation mixture DNA were used to transform E. coli DH5␣ cells. DNA was purified from 4 ml of overnight cultures using a QIAprep spin column. Sequencing was done by the double-stranded DNA cycle sequencing procedure (Life Technologies, Inc.) using a [ 33 P]ATP-end-labeled primer (5Ј-CCCGTTACAACTACAATC-3Ј). pYHA plasmids carrying mutations in the NgoMI-PmeI region of LEU3 were used to transform yeast strain XK157-3C/pYB1 (leu3 null; pYB1 contains a LEU2Ј-lacZ reporter gene (9)). Yeast cells were transformed either with the help of a transformation kit (Zymo Research, Orange, CA) or by the lithium acetate procedure (21). Transformants were plated on SD medium (20). Single colonies were suspended in 10 ml of SD medium supplemented with 2 mM leucine plus 1 mM isoleucine and grown at 30°C for 24 -30 h. Aliquots of the subcultures were then inoculated into 10 ml of SD medium supplemented with either 0.2 mM leucine (for high intracellular concentrations of ␣-IPM) or 2 mM leucine plus 1 mM isoleucine (for low ␣-IPM concentrations). To determine reporter gene activity, harvested cells were resuspended in 0.1 M sodium phosphate buffer, pH 7.0, containing 10 mM KCl, 1 mM MgSO 4 , and 50 mM ␤-mercaptoethanol. Aliquots of the suspension (made up to a total volume of 1 ml) were mixed with 20 l of a 0.1% solution of sodium dodecyl sulfate and 50 l of chloroform and vortexed vigorously for 15 s. The ␤-galactosidase activity was then measured following the procedure of Miller (22).
Several mutations in the AD of Leu3p were generated by site-directed mutagenesis, as described (15,23). The two sets of data, i.e. those obtained by the degenerate oligonucleotide method and those obtained by site-directed mutagenesis, were normalized with respect to the wild type controls.
Construction of the Leu3-dd (D872N/D874N) Mutant-Following the procedure of Kunkel et al. (24), uracil-containing single-stranded DNA from pRS316-LEU3 was mutagenized using an oligonucleotide (5Ј-GTT-TGGAGGAACGTTAATATTTTAATG-3Ј) that contained AAC and AAT triplets in place of the native GATs. Plasmid DNA isolated from several E. coli colonies was then used to transform a leu3 null strain (XK157-3C). Colonies growing slowly on leucine Ϫ plates were identified, and the original plasmid preparations were sequenced. There was excellent correlation between slow growth and the D872N/D874N double mutation. Plasmids of this type were designated pRS316-LEU3dd.To transfer the LEU3dd DNA to a plasmid with a different marker and to place the gene behind the ADC1 promoter, a cassette exchange was performed between pRS316-LEU3dd and pPC62H/86T-LEU3, as follows: pRS316-LEU3dd was used as template for a PCR reaction to synthesize a DNA fragment extending from the SalI site to the end of the LEU3 gene. The PCR primers were 5Ј-CCAACAGAAGACATACGGA-3Ј (for the 5Ј end of the fragment) and 5Ј-GTAGCACCGCGGTCATTACATA AC-3Ј (for the 3Ј end of the fragment; KspI restriction site underlined). The PCR product was digested with SalI and KspI and then inserted into pPC62H/86T-LEU3 cut with the same enzymes. The resulting plasmid was designated pPC62H/86T-LEU3dd. A derivative encoding a Leu3-dd protein from which residues 174 -773 were deleted was created by digesting pPC62H/86T-LEU3dd with SalI and AvrII, followed by Klenow enzyme treatment and re-ligation. It was designated pPC62H/86T-LEU3dd⌬12.
Modified Yeast Two-hybrid System-The DNA binding part of the two-hybrid system was designed to contain the extended DNA binding region of Leu3p (DB, residues 1-173) and the adjacent "middle region" (MR, residues 174 -773). It was expressed behind the ADC1 promoter. An appropriate centromere-containing plasmid was constructed by digesting pPC62H/86T-LEU3 (see above) with AvrII, filling in the overhangs with T4 DNA polymerase, and re-closing the plasmid with T4 DNA ligase. This created an in-frame stop codon at amino acid position 775 of Leu3p and replaced the arginine at position 774 with a serine. The resulting plasmid was designated pPC62H/86T-DB-MR.
The activation domain constructs for use in the two-hybrid system were also expressed behind the ADC1 promoter. Plasmid pNLVP16 (a gift from E. Taparowsky, Purdue University) served as starting material for pVP-LEU3-WT-AD. A pair of primers was used to clone, by PCR, a fragment encoding the AD of VP16. The primers 5Ј-CTGAGCTATTC-CTGCAGTAGTGAAGAG-3Ј (5Ј end primer) and 5ЈTCGACGGATC GACCTAGGACCCGGGGAA-3Ј (3Ј end primer) were designed to contain a PstI site and an AvrII site, respectively (underlined). The PCR product was digested with PstI and AvrII and then ligated into plasmid pPC62H/86T-LEU3 that had also been cut with PstI and AvrII, thus fusing the VP16 AD sequence to the N terminus of and in-frame with the extended Leu3p AD sequence (residues 774 -886). This yielded pPC62H/86T-VP-LEU3-WT-AD. To transfer the VP-LEU3-WT-AD sequence to plasmid pRS423 (multicopy, different marker; ref. 25), pPC62H/86T-VP-LEU3-WT-AD was digested with ApaI and PvuII. A 3-kbp ApaI-PvuII fragment containing the VP-LEU3-WT-AD sequence was ligated into pRS423 that had been cut with ApaI and SmaI. The resulting plasmid was designated pVP-LEU3-WT-AD. Plasmid pVP-LEU3-dd-AD was constructed in the same way except that pPC62H/ 86T-LEU3-dd (see above) was used instead of pPC62H/86T-LEU3. To construct pVP, a plasmid coding for the VP16 AD only, the above PCR product was first digested with AvrII, filled in with T4 DNA polymerase, then digested with PstI. The digested PCR product was inserted into plasmid pPC62H/86T-LEU3 that had been cut with SpeI, filled in, and then cut with PstI. The resulting plasmid, pPC62H/86T-VP, contained an in-frame stop codon behind the VP16 AD sequence. Next, the VP sequence was transferred to pRS423 in the way described above, yielding pVP. The DNA sequence of all junction regions and of the entire PCR-synthesized region of the VP16 AD was confirmed using the Sequenase version 2.0 sequencing kit from Amersham Corp.
Transformation of yeast cells was performed by the lithium acetate method (21). The recipient strain DK1 was transformed first with the reporter plasmid pYB1 (9) and then with pPC62H/86T-DB-MR. The resulting doubly-transformed strain was then further transformed with either pRS423 (control), or pVP, or pVP-LEU3-WT-AD, or pVP-LEU3dd-AD. The transformants were purified and single colonies from different isolates were used to inoculate 2 ml of SD medium supplemented with 1 mM each of leucine, valine, and isoleucine. After the pre-cultures had grown to saturation, cells that originated from the same colony were used to inoculate 10 ml of SD medium supplemented with either 0.2 mM leucine (if high intracellular concentrations of ␣-IPM were desired) or 4 mM leucine and 2 mM each of valine and isoleucine (if low ␣-IPM concentrations were desired). Preparation of cell-free extracts and determination of ␤-galactosidase activity were done as described above (see "Mutagenesis of the AD of Leu3p").
Random Mutagenesis of the Leu3p Middle Region (MR)-To facilitate the identification of mutants (suppressors of the Leu3-dd phenotype), the MR (encompassing residues 172-772) was divided into three subregions defined by naturally existing restriction sites. The first subregion (SubRI) extended from the SalI to the SpeI site (corresponding to residues 172-469); the second subregion (SubRII) from the SpeI to the NdeI site (residues 470 -607); the third subregion (SubRIII) from the NdeI to the AvrII site (residues 608 -772). The subregions were subjected to mutagenic PCR (26) separately. The pairs of primers used for SubRI, SubRII, and SubRIII, respectively, were 5Ј-CCAACAGAAGA-CATACGGA-3Ј plus 5Ј-TTTCCAGCACTTTGGGAGG-3Ј, 5Ј-AAGTCA-ATTGGAGATTAGTC-3Ј plus 5Ј-TACCTCCACCTTCCTTTTG-3Ј, and 5Ј-GACGTTTAATGCCTCAGTT-3Ј plus 5Ј-GTGTCCTTGATGTCTGT-AG-3Ј. The PCR products (pools of mutated DNA fragments) were digested with the appropriate restriction enzymes and inserted into pPC62H/86T-LEU3dd that had been cut with either SalI and SpeI (SubRI), or SpeI and NdeI (SubRII), or NdeI and AvrII (SubRIII). Thus, any one Leu3p molecule contained only one mutated subregion at most. XK157-3C/pYB1 cells (leu3 null with the LEU2Ј-ЈlacZ reporter) were transformed with ligation solutions containing either SubRI, SubRII, or SubRIII mutants. The transformed cells were plated on SD medium (20) and selected for significantly increased growth rates. Cell-free extracts were prepared and ␤-galactosidase activities were measured as described above ("Mutagenesis of the AD of Leu3p"). Mutants with elevated ␤-galactosidase activities were isolated, and the appropriate subregion was subjected to DNA sequence analysis.
To determine the phenotype of the MR mutants in the context of a Leu3p molecule with a wild type AD, wild type subregions were replaced with the corresponding mutated subregions by cassette exchange. For example, mutated SubRII's were isolated by cutting mutated pPC62H/86T-LEU3dd molecules with SpeI and NdeI and inserting the SpeI-NdeI fragments into pPC62H/86T-LEU3 that had been cut with the same enzymes. The resulting plasmids contained a mutated SubRII in an otherwise wild type LEU3 gene.
Electrophoretic Mobility Shift Assays and Western Blots-Whole-cell extracts used in electrophoretic mobility shift assays and Western blots were prepared using the glass bead method (27). Specifically, cells from 10 ml cultures were harvested at an A 600 of about 1, washed once, resuspended in 300 l of lysis buffer (0. For electrophoretic mobility shift assays, whole-cell extract was incubated for 15 min at 30°C with 25 mM HEPES-NaOH buffer, pH 7.9, containing 80 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 4 mM dithiothreitol, 5% (v/v) glycerol, 1 g of poly(dI-dC)⅐poly(dI-dC), 40 g of bovine serum albumin, 1.4 ng of 32 P-5Ј end-labeled UAS LEU -30-mer DNA (9), and 280 ng of non-labeled, non-binding UAS LEU -24-mer DNA (9) in a total volume of 40 l. The solution was then applied to a pre-electrophoresed 4% non-denaturing polyacrylamide gel. Electrophoresis was performed for 2.5 h at 30 mA in buffer consisting of 90 mM Tris base, 90 mM H 3 BO 3 , and 2 mM EDTA. Gels were dried and autoradiographed. Western blotting was performed on 15% polyacrylamide gels containing 0.1% sodium dodecyl sulfate. For immunoblotting, the enhanced chemiluminescence kit from Amersham was used, following the supplier's protocol. VP16 AD, VP16-Leu3-WT AD, and VP16-Leu3dd AD were detected using LA2-3 antibody (anti-Gal4-VP16 rabbit serum, gift from S. Triezenberg, Michigan State University).

RESULTS
The Modulation Function of Leu3p Is Very Sensitive to Mutational Alterations of the AD-Earlier experiments had defined the AD of Leu3p as being contained within the C-terminal 30 residues (14,15). This region is sufficient to cause transcriptional activation not only in yeast but also in mammalian cells. Leu3p molecules lacking the AD are inactive. Full-length Leu3p is subject to metabolic modulation, requiring the presence of ␣-IPM to be transcriptionally active (9,12). To gain a better understanding of the contribution of individual amino acids of the AD to activation and modulation, we extensively mutagenized the AD, both by random and site-directed mutagenesis methods. The results are shown in Table I. In this table, the permutated sequence of the AD is followed by two columns that show the activity of a LEU2-lacZ reporter gene at high and low intracellular ␣-IPM concentrations; the third column shows the ratio of these numbers (modulation ratio). Since yeast does not take up ␣-IPM from the medium, "low" and "high" levels of ␣-IPM were established by supplementing the growth medium either with an excess of branched chain amino acids (conditions that severely diminish the production of ␣-IPM (9)) or with a limiting amount of leucine. Under both conditions, the specific activity of ␤-galactosidase was close to 10 when Leu3p was absent. This value therefore represents a basal level of expression of the reporter gene. The long upper part of Table I shows the effect of single amino acid mutations on the activation and modulation functions of Leu3p. Twenty of the 26 C-terminal residues were found to have been mutated at least once. Remarkably, mutations at 16 of the 20 positions had a significant effect on modulation. Mutations causing lower modulation ratios were in the majority and mapped to both the N-terminal side (positions 861-864, 866, 867, 869) and the C-terminal side (positions 875, 879, 880, 882-884) of the AD. In all of these cases, the main reason for lower modulation ratios was a sharp increase in reporter gene activation at low ␣-IPM levels. Given that this effect was produced by mutations in so many different places, it is very unlikely that it was brought about by improved, i.e. tighter, binding of ␣-IPM to Leu3p; more likely, the effect was due to impaired masking. Most of these mutants also showed a substantial rise in activation potential at high ␣-IPM levels (with a disproportionate increase at low ␣-IPM levels). The strongest increase in activa-tion potential was seen with mutant proteins that also had the lowest modulation ratios (W864A, S866P, S866Y, V869F, V869A, F882Y, P884⌬, P884R, and P884A), suggesting that loss of modulation allowed these molecules to approach their maximal inherent activation capacity.
A significant increase in the modulation ratio was caused by mutations at positions 872, 874, 883, and 885. Of these, the D872X and D874X mutations (including a seemingly conservative D872E change) caused a particularly sharp, 5-11-fold decline of the activation potential at low ␣-IPM levels. The activation potential at high ␣-IPM levels also declined but to a lesser extent (3-5-fold). This decline was not due to instability of the mutant proteins, as judged by the results from electrophoretic mobility shift assays ( Fig. 1A; see Ref. 23). These results are consistent with the idea that the mutations at   Table II. The Leu3-dd protein had lost essentially all of its activation potential. Again, the loss of the activation potential was not caused by protein instability (Fig. 1B). To learn more about the intrinsic activation potential of the AD of Leu3-dd, we modified Leu3-dd by deleting the 600 amino acid residues between positions 173 and 774. The same deletion, when performed on wild type Leu3p (designated Leu3-⌬12), had resulted in a highly active, non-modulatable Leu-3 molecule (16). Table II shows that the Leu3-dd-⌬12 protein also regained substantial activation potential which in this case amounted to about one-third of that of wild type Leu3p. This value must be considered minimal since Leu3-dd-⌬12 appeared to be somewhat unstable. Like Leu3-⌬12, Leu3-dd-⌬12 had lost essentially all modulatability. These results strongly suggest that the inability of full-length Leu3-dd to activate was not caused by an elimination of the activation function; they are consistent with the idea that the D872N/D874N mutation caused a severe defect in modulation that keeps the Leu3-dd AD in a masked configuration (e.g. through stronger interactions with a complementary surface) and thus prevents it from interacting with the transcription machinery.
Our random mutagenesis of the AD of Leu3p also yielded a number of nonsense mutations, creating stop codons at various points along the AD. As shown in the lower part of Table I, deleting the eight C-terminal residues of Leu3p (mutant ⌬879 -886) totally eliminated the modulation response while the activation potential was fully retained. This result is consistent with the effect of point mutations in this region, exemplified by F882Y, P884D, P884R, and P884A. Shortening the AD further by deleting two or three additional residues actually stimulated the activation potential. A decrease of the activation potential to about 25% wild type was noted when the 17 C-terminal residues were deleted (Leu3p-⌬870 -886). Deleting 23 C-terminal residues (⌬864 -886) resulted in an apparently complete loss of the activation potential. It is likely, however, that Leu3p-⌬864 -886 retained just enough potential to overcome the repression of reporter gene expression that is typically seen with Leu3p molecules that are totally devoid of activation potential. Finally, deleting 26 residues from the C terminus (Leu3p-⌬861-886) created a protein that did cause repression of reporter gene expression, indicating that the activation function had been lost. Again, electrophoretic mobility shift assays showed that low activation potentials were not due to loss of mutant protein (Fig. 1C). For example, the level of Leu3-⌬864 -886 protein (not an activator) was at least as high as that of Leu3p-⌬876 -886, ⌬877-886, or ⌬879 -886 (all good activators).
Direct Evidence for Interaction between the AD and the Remainder of Leu3p-The above data show that the region of Leu3p identified as AD is also intimately involved in modulation. Earlier experiments had revealed that deletion of part or all of the middle region of Leu3p, defined as extending from residue 174 to residue 773, also led to a loss of modulation; by contrast, deletion of a region adjacent to the AD (residues 774 -854) had practically no effect on modulation (15). Since other experiments, e.g. the recent observation that masking of the AD is not affected when LEU3 is expressed in mammalian cells (18), had led to the conclusion that masking probably does  a Specific activities of ␤-galactosidase. The reporter gene was LEU2-lacZ under the control of its own promoter. See "Materials and Methods" for construction of mutants and growth conditions. The data are the average of three independent experiments. The experimental error was Ͻ25%.
c Ratio of activity at high over that at low ␣-IPM levels.
not require extraneous factors, we wondered whether there was direct interaction between the Leu3p AD and the remainder of the molecule. To answer this question, we turned to the yeast two-hybrid system in which the interaction of two proteins, separately fused to a DNA binding domain and a transcriptional activation domain, results in the formation of a functional transactivator and subsequent activation of a reporter gene (28). Since Leu3p was a DNA binding protein itself, we chose the following experimental setup. The two segments of Leu3p to be tested for interaction, expressed from separate plasmids, consisted of residues 1-773 and 775-886, respectively. The former contained the DNA binding domain (DB) plus the middle region (MR) of Leu3p and was used directly as the DNA binding part of the two-hybrid system; the latter contained the AD of Leu3p and an apparently functionless connecting peptide (15). Interaction between the two segments, expected to occur at low ␣-IPM levels, would by definition create a silent activator since the AD would be masked. To be able to recognize interaction, we therefore fused the AD of the herpes simplex virus protein VP16 (designated VP) to the N terminus of the 775-886 segment. We expected that interaction between the Leu3p AD and the remainder of Leu3p at low ␣-IPM levels would recruit VP to the UAS L -containing promoter and activate the reporter gene. At high ␣-IPM levels, the two segments would not interact, VP would not be recruited, and the reporter gene would not be activated. An important additional consideration was that interaction between the two separated segments of Leu3p would likely be considerably weaker than interaction between the same regions in the intact molecule since the contact between the regions would be diffusion controlled and no longer directed by the shape of the intact protein. For this reason, we also included the 775-886 segment of the Leu3-dd mutant protein (containing the Leu3-dd AD) in the analysis. If the interactions leading to AD masking were indeed stronger for Leu3-dd than for wild type Leu3, we would expect interaction of the severed Leu3-dd AD with the remainder of Leu3p also to be stronger. The experimental design is illustrated in Fig. 2. All of the above expectations were fulfilled by experimental results. Fig. 3 shows that a low basal level of reporter gene expression occurred when the DB-MR segment of Leu3p was expressed by itself; changing the intracellular ␣-IPM concentration had no effect (column pair 1). The same result was obtained when the DB-MR segment of Leu3p was co-expressed with the VP16 AD (column pair 2). When the DB-MR segment was co-expressed with the VP-Leu3-WT AD segment, a weak but statistically significant (2-fold) activation of the reporter gene was observed at low but not at high ␣-IPM concentrations (column pair 3). Finally, when the DB-MR segment was coexpressed with the VP-Leu3-dd AD segment, a strong, approximately 11-fold activation occurred at low (but not at high) ␣-IPM levels (column pair 4). The effects were specific for the Leu3 ADs since no activation was seen with VP alone. Specificity was also implied by the fact that reporter gene activation depended on the ␣-IPM concentration. That is, the DB-MR segment showed no sign of interacting with the Leu3 ADs when the ␣-IPM concentration was high; the very same DB-MR segment did interact with the Leu-3 ADs when the ␣-IPM concentration was low. The ␣-IPM dependence was as predicted, i.e. activation was seen when the ␣-IPM level was low, a condition which should promote AD-MR interaction. The results are consistent with the idea that the Leu-3 AD interacts with the remainder of the Leu3p molecule and that this interaction is intensified when aspartate residues 872 and 874 are mutated to asparagines (see also Fig. 7). Importantly, the results also demonstrate that Leu3-dd is capable of interacting with ␣-IPM.
Proof that the constructs used in this experiment were stably expressed in the cells was obtained from electrophoretic mobility shift assays and Western blots (Fig. 4, A-C).

Intragenic Suppression of the Leu3-dd Mutation as a Means to Identify Residues in the Middle Region (MR) of Leu3p That
Are Potentially Involved in AD Masking-Cells expressing the Leu3-dd protein grew very slowly on plates lacking leucine (Fig. 5), probably because of a repressive effect by Leu3-dd on the expression of LEU2 and possibly other LEU genes. Similar growth behavior and concurrent in vivo repression of LEU2 expression had earlier been observed with mutants of Leu3p that carried a stop codon at position 772 or 812 (9,29). The slow growth of Leu3-dd-containing cells provided an opportunity to select for suppressors. We argued that if the phenotype exhib- FIG. 2. Illustration of modified two-hybrid experiment. The "bait" molecule consists of the DB-MR portion of Leu3p (DNA binding and middle regions, residues 1-773). It is shown bound to the UAS L element of the promoter of the reporter gene (LEU2-lacZ). Three potential "prey" molecules are shown: VP, not expected to interact with the bait; a Leu3p WT AD-VP fusion, expected to interact weakly; and a Leu3-dd AD-VP fusion, expected to interact strongly with the bait. Interaction is expected to occur only at low intracellular ␣-IPM concentrations. See text for further details.

FIG. 3. Modified yeast two-hybrid experiment.
The yeast strain used in this experiment was a leu3 null strain. Cells were transformed sequentially with plasmids carrying the reporter gene (LEU2Ј-ЈlacZ) and the DB-MR segment of Leu3p, respectively. They were then further transformed with either a control plasmid (column pair 1), a plasmid carrying the VP16 AD only (pVP; column pair 2), a plasmid carrying a VP16 AD-Leu3p WT AD fusion (pVP-LEU3-WT-AD; column pair 3), or a plasmid carrying a VP16 AD-Leu3-dd AD fusion (pVP-LEU3-dd-AD; column pair 4). Transformed cells were grown under conditions that generated either high or low intracellular ␣-IPM levels, and the specific ␤-galactosidase activity was measured. Assays were performed in quadruplicate and averaged (error Ͻ10%). Three independent colonies were used for each pair of growth conditions. The error bars indicate the standard deviation of the mean for these three independent experiments. See "Materials and Methods" and text for further information.
ited by the Leu3-dd cells was indeed caused by unusually strong interactions between the AD and the MR of Leu3-dd, second-site mutations should be found that lessen these interactions, returning the Leu-3 molecule to a wild type-like or constitutively active mode and thereby leading to increased growth rates. To test this idea, we performed mutagenic (errorprone) PCR on three MR fragments of Leu3p that consisted of residues 172-469, 470 -607, and 608 -772, respectively. The mutated fragments were inserted into a cloned LEU3-dd gene by cassette exchange, and plasmids carrying a mutated LEU3-dd gene were then used to transform leu3 Ϫ yeast cells that contained a LEU2-lacZ reporter gene (9). Fast-growing transformants were isolated and analyzed for ␤-galactosidase activity. Mutants of interest, i.e. those that caused reporter gene activation (at high ␣-IPM levels) corresponding to at least 25% that caused by wild type Leu3p, were collected and subjected to DNA sequence analysis. A batch of nine such mutants is shown in Table III (mutants 1A-9A). Suppression of the phenotype of the Leu3-dd mutation was evident from the activation potential of the mutants, which ranged from 29 to 118% wild type value (Leu3-dd's activation potential is Ͻ2% wild type). The modulation ratios of the suppressors also covered a wide spectrum, ranging from relatively large values (mutants 1A-4A and 6A-8A) to a value that was closer to normal (mutant 5A). The behavior of mutants 1A-8A is readily explained by assuming that, to a varying degree, their Leu3 molecules regained the ability to expose their AD in response to ␣-IPM. The modulation ratio of one mutant (9A) was close to 1. The phenotype of this mutant (strong activation potential, essentially no response to ␣-IPM) was identical to that of cells containing constitutive Leu3p molecules with a permanently active (unmasked) conformation. The behavior of all nine mutant proteins is therefore consistent with a relaxation of tight AD masking interactions present in the Leu3-dd parent molecule. If the masking interactions in Leu3-dd proceeded through the same (or very similar) contact regions as in wild type Leu3p, one would expect wild type protein containing the above mutations also to show diminished AD-masking capabilities. This was indeed found to be the case. When the MR mutations present in mutants 1A-9A were introduced into wild type Leu3 (by the same cassette exchange that was used to introduce them into Leu3-dd), it was found that all nine mutant proteins (1B-9B, right half of Table III) had become essentially constitutive (modulation ratios Ͻ2). At the same time, their activation potential was either significantly above or close to that seen with wild type Leu3p. These results strongly suggest that the underlying mutations occurred in residue(s) that are important for the masking of the AD. For most of the MR mutants listed in Table III, identification of individual critical residues is not possible at this point since the mutants are the result of multiple (double, triple, or quadruple) residue changes (see FIG. 4. Demonstration that the fragments used in the modified two-hybrid system were stably expressed. A, electrophoretic mobility shift assays using whole-cell extracts as protein source and a UAS L 30-mer (9) as probe.  (20) for 3 days at 30°C. The leu3 null strain XK157-3C, transformed with pYB1 (9) was used throughout. It was transformed additionally either with pPC62H/86T-LEU3, a plasmid that generates full-length wild type Leu3p (WT LEU3), or with pPC62H/86T-LEU3dd, a plasmid that generates full-length Leu3-dd (LEU3dd).
legend of Table III). There is, however, a single mutation (K664E, mutant 1) that points to lysine in position 664 as a residue potentially involved in the masking process. The region of Leu3p to which all of the sequenced mutations map is shown in Fig. 6.

DISCUSSION
In this paper, we have taken a first step toward understanding the mechanism by which Leu3p responds to ␣-IPM. The functional change brought about by ␣-IPM is quite dramatic. At low concentrations of the metabolite or in its absence, Leu3p acts as a repressor, causing a 4-to 5-fold drop in reporter gene expression below the level observed in cells lacking Leu3p (12,14). When the intracellular ␣-IPM concentration rises above a threshold value, which is probably in the 10th-millimolar range (12,18), Leu3p becomes a strong activator of gene expression. This transition is thought to be accompanied by a conformational change that somehow allows the sole AD of Leu3p, located near the C terminus, to interact with elements of the transcription machinery. The form of Leu3p that represses gene expression (by an as yet unknown mechanism, see Ref. 30) is also the form that assumes a masked configuration. However, masking and repression are clearly different processes since repression is seen with relatively small segments of Leu3p (e.g. the Leu3p-(17-147) peptide, Ref. 14), whereas masking requires much larger and different segments of Leu3p. How is masking accomplished? In the case of Gal4p, the best-studied member of this class of proteins, masking is achieved by specific interaction between Gal4p and the negative regulator Gal80p. In the presence of galactose, the Gal4p-Gal80p interaction is altered, and Gal4p's AD becomes available for transcriptional activation (3). The masking of the Leu3p AD likely proceeds in a different manner. The main argument against the participation of a Gal80p-like protein in the masking of Leu3p's AD has come from the observation that tight masking of Leu3p takes place when the LEU3 gene is expressed in cultured mammalian cells (18). Masking is reversed and an active form of Leu3p is generated upon addition of ␣-IPM to the cell culture medium. In those experiments, expression of the LEU3 gene was directed by the human cytomegalovirus major intermediate early promoter. No other yeast-specific genes were present. Since the leucine biosynthetic pathway is absent from mammalian cells and such cells do not normally contain a Leu3p-type regulator, they would not be expected to contain a specific Leu3p-masking factor either. It is important to note in this context that expression of LAC9 of Kluyveromyces lactis (a Gal4p homolog) in mammalian cells produced a protein that was not masked unless GAL80 was co-expressed (31). The argument that Leu3p does not require a separate masking factor is further supported by the observation that, in an in vitro transcription system using yeast wholecell extract from Leu3p-deficient cells, purified Leu3p was unable to out-titrate a presumptive masking factor and stayed in a masked mode even at relatively high concentrations, as long as ␣-IPM was absent (12). Finally, the demonstration in this paper that the Leu3 AD can directly interact with the remainder of the protein provides a very strong argument for the self-contained nature of Leu3p with respect to its modulation.
If masking of the AD of Leu3p is achieved intramolecularly, the next question is which parts of the Leu3p molecule are involved in this process. Recent domain-swap experiments with the serine/threonine-responsive activator Cha4p of yeast have shown that the extended DNA binding domain of Leu3p (encompassing residues 1-173) is not required for modulation by ␣-IPM. 2 Also, deleting a region adjacent to the AD of Leu3p (residues 774 -854) had no effect on modulation (15). In the present work, we therefore focused on the AD and the MR of Leu3p as potentially holding the key to the modulation process.
Our attempt to understand modulation was aided significantly by the construction of the Leu3-dd mutant. This mutant not only proved useful in the two-hybrid experiment but also led to the isolation of modulation-related mutants that map to the MR of Leu3p. It is therefore important to inquire about the mechanism by which the D872N/D874N mutation keeps the regulator in a virtually inactive state. A priori, at least three possibilities to explain the behavior of Leu3-dd might be considered: (i) the D872N/D874N mutation drastically reduces the binding of ␣-IPM to Leu3p; (ii) the double mutation does not affect ␣-IPM binding but essentially quells a conformational change that, in wild type Leu3p, is a consequence of ␣-IPM binding and eventually leads to the exposure of the AD; or (iii) the double mutation allows the AD to interact more strongly with the remainder of the molecule. There are several observations that argue against the first possibility. First and foremost, Leu3-dd still responds to changes in the ␣-IPM concentration; a strong dependence on ␣-IPM was clearly evident in the two-hybrid experiment with cleaved Leu3-dd. Second, if the D872N/D874N mutation had caused severe impairment of ␣-IPM binding, it would be quite improbable that a relatively large number of second-site suppressor mutations would be found that would repair a damaged ␣-IPM binding pocket. Also, if the mutations leading to suppression of the Leu3-dd phenotype did so by repairing the ␣-IPM binding pocket of Leu3-dd, those same mutations would not be expected to create generally strong and nearly ␣-IPM-independent activators when introduced into wild type Leu3p. Yet that is what is observed (Table III). Turning to the second possibility, it is also unlikely that the D872N/D874N mutation interferes with a conformational change caused by ␣-IPM. If this were the case, one would expect the intramolecular masking interactions (which by definition occur before any ␣-IPM-induced conformational change could take place) to be the same with wild type Leu3p and Leu3-dd. Yet these interactions, as observed in the two-hybrid experiment, are much stronger with Leu3-dd. We therefore conclude, in accordance with the third possibility and based on the evidence from the two-hybrid experiment, that the behavior of Leu3-dd is a consequence of stronger interactions between the AD and the remainder (very likely the MR) of Leu3p. In Leu3-dd, these interactions are so strong that the normal exposure of the AD following ␣-IPM binding cannot occur. A different behavior results when the AD of Leu3-dd is severed from the rest of the molecule. The now diffusion-controlled interactions are weaker than those in the intact Leu3-dd molecule, yet are still strong enough to resemble those of wild type Leu3p. A schematic representation of interactions proposed to take place is shown in Fig. 7. The observation that the Leu3-dd suppressor mutations studied so far not only reverse the behavior of Leu3-dd but also lead to essentially permanent unmasking of the AD of wild type Leu3p suggests that residues identified in this way are also involved in the normal masking process. The question of whether a given residue participates directly in masking (e.g. by making contact with the AD) or has an indirect effect (e.g. by stabilizing a configuration favorable for masking) will be difficult to answer in the absence of structural information. However, we think that Lys-664 is a good candidate for direct participation because the K664E mutation (Table III) causes a change in side chain size and a drastic change in side chain chemistry, yet the mutated protein is very active, making it seem unlikely that the mutation caused a gross conformational change.
We now turn to the portion of Leu3p that is presumed to interact with the MR, i.e. the AD. This region is remarkably sensitive to mutation with respect to the modulation function, suggesting that important secondary or tertiary structural configurations are present. CD spectroscopy has indicated a propensity for ␣-helical structure in the 859 -886-residue region, 3 and it is possible that the intactness of helical structure is important for efficient masking. However, drastic effects on modulation are seen both when the effect of a mutation on ␣-helical stability is expected to be strong (e.g. F882Y) and when it is expected to be minor (e.g. S866Y). Strong effects on modulation are also seen when Pro-884 (which is very likely not part of an ␣-helix) is mutated. This indicates that a need to conserve ␣-helical structure per se is not sufficient to explain the sensitivity of the modulation function to mutation. Another interesting feature of the AD is the presence of two types of amino acid residues: those that loosen and those that tighten the masking interactions. Mutations that appear to loosen the interactions the most are found at positions Trp-864, Ser-866, Val-869, Phe-882, and Pro-884. It is reasonable to assume that those same residues facilitate masking in wild type Leu3p. Residues Asp-872 and Asp-874, which are located in the center of the AD and which, when mutated, increase masking efficiency, would not be expected to contribute much to masking in wild type Leu3p and might even antagonize masking. This pattern of residues that either favor or oppose (or are neutral toward) masking might be important for achieving a physiologically desirable balance between open and closed forms of Leu3p.
In striking contrast to their effect on the modulation function of Leu3p, mutations in the AD have a much smaller effect on FIG. 7. Schematic illustration of interactions between the activation domain and the middle region of Leu3p. A, interactions within wild type Leu3p. Three major regions of Leu3p are shown: the extended DNA binding region (DB, residues 1-173), the middle region (MR, residues 174 -773), and the activation domain (WT, residues 855-886). The "flexible arm" between the MR and the activation domain (residues 774 -854) is apparently irrelevant for function (15). In the absence of ␣-IPM, the activation domain is masked by interaction with the MR, and the reporter gene is not activated. When ␣-IPM is present at sufficiently high concentrations, the MR changes its conformation to MR*, unmasking the activation domain and allowing reporter gene activation. B, interactions between the MR and a severed wild type activation domain (WT) fused to the activation domain of VP16 (VP) via the flexible arm. In the absence of ␣-IPM, the activation domain still interacts with the MR, albeit much more weakly than in the intact molecule. VP is recruited to the promoter, and the reporter gene is (weakly) activated. In the presence of ␣-IPM, MR changes to MR*, the activation domain-VP fusion is released and diffuses away, and the reporter gene is not activated. C, interactions within intact Leu3-dd. In the absence of ␣-IPM, the dd activation domain is masked by interacting strongly with the MR; the reporter gene is not activated. In the presence of ␣-IPM, dd is still sufficiently masked to prevent activation of the reporter gene. D, interactions between the MR and a severed dd activation domain fused to VP via the flexible arm. In the absence of ␣-IPM, dd still interacts with the MR (more strongly than the WT activation domain in B). VP is recruited to the promoter, and the reporter gene is activated. In the presence of ␣-IPM, the conversion of MR to MR* weakens the interactions such that the dd-VP fusion is released; the reporter gene is not activated. Shading indicates relative strength of non-covalent interactions. the activation potential of the protein. This is most evident from the deletion analysis (Table I). A Leu3p molecule lacking 17 of the 26 residues of the AD still has about 25% the activation potential of full-length Leu3p. It should be noted that the remaining short region still contains three of the original six acidic residues, as well as four hydrophobic residues. In view of the well-known fact that even random (acidic) sequences can bestow transcriptional activation potential upon DNA binding regions to which they are fused (32), the apparent promiscuity of the Leu3p AD does not come as a surprise.
Although this work has provided important evidence supporting intramolecular interactions as a mechanism for masking the AD of Leu3p, we are not yet able to determine whether these interactions occur in cis or in trans. Since Leu3p very likely acts as a dimer (33,34), masking could in theory be achieved either by intra-monomer or by intra-dimer interactions. Experiments addressing this question are underway.