Functional characterization of an inositol-sensitive upstream activation sequence in yeast. A cis-regulatory element responsible for inositol-choline mediated regulation of phospholipid biosynthesis.

A repeated element, the inositol-sensitive upstream activation sequence (UASINO), having the consensus sequence, 5'-CATGTGAAAT-3', is present in the promoters of genes encoding enzymes of phospholipid biosynthesis that are regulated in response to the phospholipid precursors, inositol and choline. None of the naturally occurring variants of the UASINO element exactly recapitulates the consensus (for review, see Carman, G. M., and Henry, S. A. (1989) Annu. Rev. Biochem. 58, 635-669 and Paltauf, F., Kolwhein, S., and Henry, S. A. (1992) in Molecular Biology of the Yeast Saccharomyces cerevisiae (Broach, J., Jones, E., and Pringle, J., eds) Vol. 2, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). The first six bases of the UASINO element are homologous with canonical binding motif for proteins of the basic helix-loop-helix (bHLH) family. Two bHLH regulatory proteins, Ino2p and Ino4p from yeast, were previously shown to bind to promoter fragments containing this element. In the present study, an extensive analysis of UASINO function has been conducted. We report that any base substitution within the putative bHLH binding site resulted either in a dramatic reduction or in a complete obliteration of UASINO function as tested in an expression assay in vivo. Base substitutions in the 5' region that flanks the 10-base pair repeat, as well as sequences within the repeat itself at its 3' end outside the bHLH core, were also assessed. The two bases immediately flanking the 5' end of the element proved to be very important to its function as a UAS element as did the two bases immediately 3' of the bHLH core motif. Substitutions of the final two bases of the original ten base pair consensus (i.e. 5'-CATGTGAAAT-3') had less dramatic effects. We also tested a subset of the altered elements for their ability to serve as competitors in an assay of Ino2p x Ino4p binding. The strength of any given sequence as a UASINO element, as assayed in vivo, was strongly correlated with its strength as a competitor for Ino2p x Ino4p binding. We also tested a subset of the modified UASINO elements for their effects on expression in vivo in a strain carrying an opi1 mutation. The opi1 mutation renders the coregulated enzymes of phospholipid synthesis constitutive in the presence of phospholipid precursors. All elements that retained some residual UASINO activity when tested in the wild-type strain were constitutively expressed at a level comparable with the wild-type derepressed level when tested in the opi1 mutant.(ABSTRACT TRUNCATED AT 400 WORDS)

vivo in a strain carrying an opi1 mutation. The opi1 mutation renders the coregulated enzymes of phospholipid synthesis constitutive in the presence of phospholipid precursors. All elements that retained some residual UAS INO activity when tested in the wild-type strain were constitutively expressed at a level comparable with the wild-type derepressed level when tested in the opi1 mutant. Thus, UAS INO appears to be responsible for OPI1 mediated repression, as well as Ino2p⅐Ino4p binding. Furthermore, each of the identified functions of the UAS INO element appears to have the same sequence specificity, and all require the presence of the intact bHLH motif, suggesting that transcriptional activation, repression, and Ino2p⅐Ino4p binding are all components of a single regulatory mechanism.
In Saccharomyces cerevisiae, a large number of phospholipid biosynthetic enzyme activities show a common pattern of regulation (1). These enzymes are fully derepressed in the absence of the soluble phospholipid precursors, inositol and choline. They are partly repressed in the presence of inositol alone and fully repressed in the presence of inositol plus choline. Choline alone appears to have little effect on the expression of these coregulated enzymes, which include the cytoplasmic enzyme, inositol-1-phosphate synthase, product of the INO1 gene (3). The membrane-associated activities that catalyze the sequence of reactions leading to synthesis of phosphatidylcholine via methylation of phosphatidylethanolamine are also regulated in response to inositol and choline. The structural genes encoding a number of these enzymes have been cloned and coordinate regulation in response to inositol and choline has been shown to occur at the transcriptional level (for review, see Refs. 1 and 2).
The structural genes encoding the enzymes discussed above also respond to a common set of regulatory genes, including INO2, INO4, and OPI1 (1,2). Cells bearing an ino2 or ino4 mutation exhibit inositol auxotrophy due to inability to derepress the INO1 gene (4). The ino2 and ino4 mutants also exhibit reduced phosphatidylcholine synthesis due to the inability to derepress enzymes of phosphatidylcholine biosynthesis (5). In contrast, cells bearing an opi1 mutation constitutively overexpress the products of the coregulated genes (6,7). The INO2 and INO4 gene products, Ino2p and Ino4p, both possess the basic helix-loop-helix (bHLH) 1 motif similar to that seen in members of myc family of oncogene and protooncogene proteins (8 -10). Ino2p and Ino4p have been shown to form a heterodimer (10 -12) that binds to a repeated element with the consensus sequence, 5Ј-CATGTGAAAT-3Ј, which is found in the promoters of all of the coregulated genes. The OPI1 gene product contains a leucine zipper motif and polyglutamine stretches (13), motifs that are common to many DNA binding proteins. However, the mechanism of interaction of the OPI1 gene product with the promoters of the coregulated genes of phospholipid biosynthesis has not yet been elucidated.
The first six bases at the 5Ј end of the 10-base pair (bp) consensus element, 5Ј-CATGTGAAAT-3Ј, are homologus to the canonical bHLH binding site, 5Ј-CANNTG-3Ј (2). Naturallyoccurring variants of this 10-bp element have been shown to function as an inositol-sensitive upstream activation sequence (UAS INO ) (for review, see Refs. 2 and 6). However, none of the naturally occurring copies of the repeated element is identical to the 10-bp consensus sequence, 5Ј-CATGTGAAAT-3Ј, and the various genes containing this element exhibit widely disparate repression ratios in response to inositol and choline. In the present study, we have conducted a systematic analysis of the role of individual bases within the 10-bp consensus sequence and its 5Ј-flanking region. We have also assessed the relationship between UAS INO function in vivo and the binding of the Ino2p⅐Ino4p complex in cell extracts. We have also examined the sequence specificity for repression in response to inositol and choline and compared it with the response to the opi1 regulatory mutation. In this report, we demonstrate the functional importance of the bHLH element within the 10-bp UA-S INO . We have also assessed the function of the bases within the 10-bp consensus at its 3Ј end but outside the bHLH motif, as well as bases flanking the 10-bp repeat on its 5Ј side. We demonstrate that the sequence specificity for the OPI1 dependent response, UAS INO activity, and Ino2p⅐Ino4p binding, are virtually identical, suggesting that they are all components of a single regulatory mechanism.
X-gal plates were identical in composition to complete media except that vitamin-free yeast nitrogen base was replaced with 0.1 MKH 2 PO 4 , pH 7.0, 15 mM (NH 4 ) 2 SO 4 , 0.8 mM MgSO 4 , 2 mM FeSO 4 , 75 mM KOH, and 0.04 g/liter X-gal. These plates were used to score transformants containing the heterologous CYC1-lacZ reporter gene fusions.
Recombinant DNA Methods-Standard recombinant DNA methods were carried out as described by Maniatis (15). DNA sequencing was performed using the Sequenase kit (U. S. Biochemical Corp.).
Plasmids and Their Constructions-The parental plasmids used were as follows: pJH304 (16), which was constructed by removing a 430-bp XhoI restriction fragment from the CYC1-UAS of pLG669z, and pNG22 (17). Plasmid pNB101 was constructed by deleting the CYC1-UAS of pNG22 by digestion with XhoI and religating. The CYC1-lacZ expression vector pNB404 used in this study was derived from pJH304 and pNB101 as shown in Fig. 1. The BamHI/EcoNI fragment containing the CYC1-promoter region without the UAS elements of pJH304, which contains a single XhoI site, was exchanged for the CYC1 promoter region (⌬UAS) of plasmid pNB101 as shown, which contains a polylinker at XhoI, to yield plasmid pNB404.
Cloning the Oligonucleotides-Synthetic oligonucleotides used in this work are described in Table I and were obtained from Operon Technologies, Inc. To construct the plasmids pNB501, pNB502, pNB307 and pNB308, the oligonucleotides listed in Table I were cloned into the Klenow-filled XhoI and SalI sites of pNB404 (see Fig. 1). To produce the remaining plasmids listed in Table I, oligonucleotides whose sequences are given in the table were cloned in to the EagI and SalI sites of the pNB404 expression vector.
The complementary pairs of single-stranded synthetic oligonucleotides (1-2 g each) were separately dissolved in 10 mM Tris-HCl, pH 8.0, and phosphorylated by the polynucleotidekinase reaction at 37°C, followed by ethanol precipitation. DNA was dissolved in 10 mM Tris-HCl buffer, pH 8.0. Two opposite strands of each desired fragment were combined (1 g each), heated to 85°C for 10 min and slowly cooled to room temperature (25°C). Annealed oligonucleotides were then ligated into the expression vector pNB404. Verification of inserts and flanking sequences was carried out by DNA sequencing, as described above.
Yeast Transformation and ␤-Galactosidase Assay-Yeast transformations were carried out by the lithium acetate method as described previously (16). Several individual yeast transformants were picked randomly for each plasmid construct and were used for ␤-galactosidase assay. Plasmids containing the 2 origin of replication are known to be mitotically unstable. Therefore, the transformants were used immediately, as soon as colonies retrieved from the transformation experiments were large enough to inoculate into liquid culture. Yeast transformants were grown in 25 ml of complete synthetic medium in the absence (I Ϫ C Ϫ medium) or presence of inositol and choline (I ϩ C ϩ medium) and were harvested at exponential growth phase. Pelleted cells were washed, resuspended in 1 ml of ␤-galactosidase assay buffer composed of 0.1 M Tris-HCl buffer, pH 8.0, 20% glycerol, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride, and frozen at Ϫ70°C. Cells were broken by vortexing with glass beads (0.5 mm diameter). ␤-Galactosidase activities in yeast transformants were determined, as described previously (16), except that aliquots were removed at 5, 10, and 15 min of incubation. ␤-Galactosidase units are defined as (A 420 / min/mg of total protein) ϫ 1000. Two to three independent yeast transformants were assayed for each of the plasmid constructs given in Table  I, and all assays were performed in duplicate.
Electrophoretic Mobility Shift Assay-Whole cell extracts from wildtype strain W303-1A, grown in medium lacking inositol and choline (I Ϫ C Ϫ ), were prepared as described previously (14). Standard binding reactions were carried out in a 20-l volume containing 50 g of whole cell extract proteins, 4 mM Tris-HCl, pH 8.0, 100 mM KCl, 4 mM MgCl 2 , 1 mM dithiothreitol, 4% glycerol, 1 g of poly(dI-dC), 5000 cpm of radiolabeled DNA. The DNA used for the standard binding reaction, Template B, a DNA fragment from the INO1 promoter region that spans nucleotides Ϫ259 to Ϫ154 has been described previously (10,14). In DNA competition assays, 5 M of double-stranded oligonucleotide competitor was included in the binding reactions along with Template B DNA. A concentration of 5 M was selected empirically by testing a range of concentrations using oligonucleotides NB508 (Table I) and JML9/10, both of which contain the consensus element that gives maximum UAS INO activity (Table II). The level of competition observed with the NB508 oligonucleotide was indistinguishable from the level of competition observed for the oligonucleotide, JML9/10 (Table I). JML9/10, which contains the same 10-bp consensus sequence as pNB508 embedded in different flanking sequence, has previously been shown to be an effective competitor of Ino2p⅐Ino4p binding (10,11). The residual binding observed for the Ino2p⅐Ino4p complexes on radioactive Template B when competitors NB508 and JML9/10 were used at a concentration of 5 M, thus, defines the maximal activity for any potential competitor employed in this study. Also included as a control in some competition assays was oligonucleotide JML13/14 (Table I), which contains a G instead of a C in the first base of the 10-bp consensus sequence. This oligonucleotide was previously shown to be completely unable to complete Ino2p⅐Ino4p binding on Template B (10,11). This inactive oligonucleotide, and the equally inactive nonspecific poly(dI/dC) control, produced no detectable reduction in binding of Ino2p⅐Ino4p complexes as compared with the standard binding assay without the addition of a competitor (Fig. 2).
Protein-DNA complexes were resolved on 4% polyacrylamide gels and were visualized by autoradiography. Immediately 5Ј to the Putative HLH Core Binding Site in UAS INO -The 10-bp element having the consensus sequence, 5Ј-CATGTGAAAT-3Ј, that was originally identified as the UAS INO includes the sequence CAT-GTG at its extreme 5Ј end. This portion of UAS INO is homologus to the consensus binding site, CANNTG, for bHLH proteins (10,13).

Effect of Mutations in the Region
A systematic modification of the four bases immediately 5Ј to the 10-bp element (and, consequently, the putative HLH binding site, as well) was carried out, and the data are shown in Table II. The four bases immediately 5Ј to the putative bHLH binding site were initially set as 5Ј-CGTN-3Ј followed by the 10-bp consensus element: i.e. 5Ј-(CGTN)CATGTGAAAT-3Ј. All four possible base substitutions for N were constructed and tested (plasmids pNB501-pNB504; Table II) for UAS activity by inserting them into the expression vector pNB404 and transforming them into the wild-type yeast strain, as described in under "Materials and Methods." The empty vector, pNB404, gave very low background expression of ␤-galactosidase under the growth conditions employed in this study (Table II).
When the transformants containing modified UAS INO elements were grown under derepressing growth conditions (i.e. absence of the phospholipid precursors inositol and choline; I Ϫ C Ϫ condition on Table II), constructs pNB502 and pNB503, containing a G or a C in the 5Ј position, immediately flanking the consensus element, were 3-fold more active than constructs pNB501 and pNB504, containing an A or a T. Consequently, G was used in this position in all subsequent constructs. Modification of the next adjacent base, i.e. 5Ј-(CGNG)CATGTGAAAT-3Ј, also had a measurable effect. Constructs pNB502 and pNB505, containing a T or an A in this position, were some-  Table II) and derepressing conditions than constructs pNB506 and pNB507 containing a G or a C. Again, G was selected as the standard flanking sequence for subsequent constructs. Substitutions in the next two bases 5Ј to the element, i.e. 5Ј-(NNGG)CATGTGAAAT-3Ј had much smaller effects (compare constructs pNB508 and pNB509 with pNB507; Table II). However, construct pNB508 containing the sequence 5Ј-(CCGG)CATGTGAAAT-3Ј gave the highest overall level of expression under derepressing growth conditions. Expression driven by this element was subject to over 10-fold repression in response to inositol and choline. Therefore, in all of the subsequent experiments described below, in which sequences within the 10-base pair UAS INO element were systematically modified, we employed the 5Ј-flanking sequence, 5Ј-CCGG-3Ј.
Effect of Mutations in the Sequence within the Putative bHLH Binding Site-As mentioned above, the putative bHLH binding site is located at the 5Ј end of the 10-bp consensus element and has the sequence 5Ј-CATGTG-3Ј. For each of the bases in the putative bHLH binding domain, all four possible nucleotide substitutions were tested (Table II, part B). Con-struct pNB508, which retained the consensus sequence (i.e. the 10-bp element 5Ј CATGTGAAAT-3Ј), was by far the most active and, as noted above, expressed 536 units of ␤-galactosidase activity when the wild-type yeast transformant carrying it was grown in the absence of inositol and choline. Construct pNB603, containing an A as the first base in the 5Ј position of the 10-bp element (i.e. 5Ј-AATGTGAAAT-3Ј), expressed approximately 40 units of ␤-galactosidase activity under derepressing growth conditions, and this activity was repressed to about six units in the presence of inositol and choline. A construct (pNB604) that, upon sequencing, proved to contain a second substitution, namely a T in place of a G in the sixth position in the element, in addition to an A in the first position, was completely inactive. Construct pNB601, containing a G as the first base in the 5Ј end of the consensus element, and pNB602, which has a T, both exhibited very little ␤-galactosidase activity under repressing or derepressing growth conditions.
The second base from the 5Ј end of the 10-bp consensus element is an A; (i.e. 5Ј-CATGTGAAT-3Ј). Any substitution in this second position resulted in 10-fold or greater reduction in ␤-galactosidase expression under derepressing growth condi- To construct plasmids pNB501, pNB502, pNB307, and pNB308, the nucleotide sequence of the given oligonucleotide was cloned between the Klenow filled-in XhoI and SalI sites of pNB404 Fig. 1. For the remainder of the plasmids, the given double-stranded synthetic oligonucleotides were cloned between the EagI and SalI sites of the same plasmid. The 10-bp consensus element is given in capital letters. The bases sequentially modified in each construct, as described in the text, are given in boldface.
A. Substitutions in 5Ј flanking region Substitutions in putative bHLH core Oligonucleotides NB601-NB620 (plasmids pNB601-pNB620) contain the same 5Ј-and 3Ј-flanking sequence as NB508 (pNB508) but have substitutions in the putative bHLH core consensus sequence, CANNTG, as shown in Table II. C. Substitutions in 3Ј end of 10-base element Oligonucleotides NB301-NB306 (plasmids pNB301-pNB306) have the same 5Ј-and 3Ј-flanking sequence as NB508 but have substitutions in the 3Ј end of the 10-base pair element outside the bHLH core sequence, as shown in Table II. Oligonucleotides NB307 and NB308 have the same 5Јand 3Ј-flanking sequence as NB502.  Synthetic double-stranded oligonucleotides were constructed in which the presumptive 10-bp consensus and its 5Ј-flanking sequence were altered one base at a time. The presumptive 10-bp consensus sequence is written in capital letters, 5Ј-flanking sequence in small letters, and substitution mutations in boldface. For ␤-galactosidase determinations, wild-type and opi1 transformants carrying the individual plasmids were grown in synthetic medium and assayed as described under "Materials and Methods." Oligonucleotides used for construction of the plasmids were, in some cases, also tested for ability to complete Ino2p/Ino4p binding in gel shift assays (Fig. 2). A qualitative assessment of this data is indicated in the final column. The strongest binding competitors, equivalent to the consensus oligonucleotide (NB508) used to construct plasmid pNB508 are designated ϩϩ. Oligonucleotides having slightly lower ability than NB508 to serve as competitors, are listed ϩ. Oligonucleotides having much reduced, but still detectable activity to compete are designated Ϫ/ϩ. Those showing no ability above the control (dI/dC) to compete Ino2p/Ino4p binding are designated Ϫ. Examples of the actual binding assays are shown in Fig. 2. I Ϫ C Ϫ , synthetic complete medium lacking inositol and choline; I ϩ C ϩ synthetic complete medium supplemented with 75 M inositol and 1 mM choline. NT, not tested.

Plasmid
Sequence inserted a Data for construct pNB502 is repeated in section C for purposes of comparison to data for constructs pNB307 and pNB308 which have the same 5Ј-flanking sequences as pNB502.
b Data for construct pNB508 is repeated for purposes of comparison, at the start of sections B and C.
sponse to inositol and choline. Modification of the fourth base (i.e. CATGTGAAAT) to a C or a T (constructs pNB611 and pNB612) left little ␤-galactosidase expression above background (13 and 11 units, respectively; Table II). Placement of an A into the fourth position, on the other hand, produced construct pNB613, which expressed 67 units of ␤-galactosidase activity under derepressing conditions, and this residual expression was repressed nearly 10-fold upon the addition of inositol and choline to the growth medium. Another construct (pNB614) that was found, upon sequencing, to have substitutions in the third and fourth positions (a C and an A, respectively) gave much lower activity than either of the single substitutions (compare data in Table II for pNB614 to pNB613 and pNB610). All modifications of the fifth base (i.e. modifications of the consensus; CATGTGAAAT) also resulted in reduced expression. Construct pNB616 containing a G in this position had the lowest activity. All three such constructs (i.e. pNB615, pNB616, and pNB617) exhibited reduced expression compared with the consensus element, but the residual expression was repressed approximately 2-3-fold in response to inositol and choline. Likewise, all constructs (pNB618 -620) containing substitutions in the sixth base exhibited very reduced expression compared with pNB508. Constructs (pNB619 and pNB620) containing an A or a T as the sixth base appeared to be essentially inactive, while construct pNB618, containing a C, expressed 28 units of ␤-galactosidase activity under derepressing conditions. Construct pNB618 was also repressed a little over 2-fold in response to inositol and choline.
Thus, all six bases that comprise the putative core binding site for bHLH proteins appeared to be essential for optimal UAS INO function. Of the 20 constructs tested that contained base substitutions within this motif, none retained more than 67 units of ␤-galactosidase expression compared with 536 units for the consensus element in pNB508. However, all of the elements containing modification of the bHLH motif that retained 20 units or greater expression of ␤-galactosidase activity were capable of some repression in response to inositol and choline.
Effect of Mutations in the Region 3Ј to the Putative HLH Binding Site-Systematic modification of the bases immediately 3Ј to the putative core bHLH binding site was also carried out. Any change from the consensus sequence in the two bases immediately flanking the putative bHLH element (i.e. the seventh and eighth bases of the 10-bp element, 5Ј-CATGT-G(AAAT)-3Ј), resulted in a substantial reduction in expression (Table II,  The strongest level of competition, and the resulting reduction in Template B binding, is observed for oligonucleotide NB508 and for JML9/10 (as previously shown in Refs. 9 and 10). Oligonucleotide JML13/14, having a modified sequence that places a G into the first 5Ј position of the 10-bp consensus sequence, is unable to compete (as previously shown in Refs. 9 and 10). Addition of poly(dI/dC) also has no effect on Ino2p⅐Ino4p binding. For ease of visual comparison, the lanes in each panel having no competitor are labeled Ϫ; lanes having NB508 as a competitor are marked as ϩ; those having poly(dI/dC) competitor are marked *. only 15 and 17 units of ␤-galactosidase activity, respectively, under derepressing conditions, were both repressed to some degree in response to inositol and choline.
The final two bases at the 3Ј end of the 10-bp consensus element were tested less rigorously. The sequence that was employed for the two constructs tested (pNB307 and pNB308) had a different flanking sequence at the 5Ј end than other constructs discussed above. Consequently, their activity levels should not be compared with pNB508 but rather to pNB502 (5Ј-(CGTG)CATGTGAAAT-3Ј), which has the same 5Ј-flanking sequence as pNB307 and pNB308. Construct pNB502 expressed only 280 units of activity under derepressing growth conditions compared with 536 units for pNB508 (Table II, part  A).
In construct pNB307, a G has been substituted for the A in the second to last position at the 3Ј end of the 10-bp consensus element, 5Ј-(CGTG)CATGTGAAGT-3Ј. This construct expressed 150 units of ␤-galactosidase under derepressing growth conditions and was repressed to 10 units of activity in the presence of inositol and choline. Thus, construct pNB307 expressed a lower level of ␤-galactosidase than pNB502, which expresses 280 units. In construct pNB308, the final T in the consensus sequence has been replaced by a G, (i.e. 5Ј-(CGTG)CATGTGAAAG-3Ј). Construct pNB308 expressed 450 units of ␤-galactosidase activity under derepressing growth conditions and 53 units of activity under repressing conditions, a level of ␤-galactosidase activity under both repressing and derepressing conditions that is nearly 2-fold higher than the consensus construct pNB502, which has the same 5Ј-flanking sequence as pNB308.
Effect of the opi1 Regulatory Mutation on Expression of Constructs Containing Modified UAS INO Elements-The opi1 mutation renders the constitutive expression of INO1 and all of the other coregulated genes of phospholipid biosynthesis (1). However, the mechanism of interaction of the opi1 gene product with the INO1 promoter is unknown, and no previous studies have addressed the promoter sequence specificity for the genes that exhibit altered expression in response to the opi1 mutation. Therefore, a subset of the constructs described above was tested for expression in an opi1 regulatory mutant background. The expression data for the opi1 transformants are shown on Table II (parts A-C) alongside the comparable data for the wild-type transformants. In all instances in which expression is retained for a particular construct in the wildtype strain in the absence of inositol and choline, a comparable level of expression was retained in the opi1 mutant strain. Constructs which were expressed at minimal or background levels in the wild-type strain, under depressing conditions, were also expressed at comparable low levels in the opi1 strain under both derepressing and repressing growth conditions. This expression of each construct was essentially constitutive in the opi1 mutant at a level comparable with, or in some cases, slightly higher than the level observed under derepressing conditions in the wild-type strain.
One construct, pNB617, which contained a substitution of a C for the T as the fifth base of the 10-bp element, was 3 times more active in the opi1 strain than in the wild-type strain under derepressing conditions. Other than the case of this one construct, the opi1 mutation rendered constitutive any level of expression that was associated with a particular construct in the wild-type strain under derepressing condition. Thus, the sequence specificity for regulation in response to the OPI1 gene appeared to be essentially identical to the sequence specificity for UAS INO Table II was assessed using a competition assay as described previously (10,11,16). In this assay, a piece of the native INO1 promoter, described under "Materials and Methods," was used in electrophoretic band mobility shift assays (12), as shown in Fig. 2. The results were highly reproducible. Note, for example, the effect of competition by NB508 on each of the three panels on Fig. 2, representing three separate experiments. Oligonucleotide NB508 was capable of reducing the binding of the Ino2p⅐Ino4p complexes on the fragment of native INO1 promoter to an extent that is indistinguishable from that observed with the JML9/10 oligonucleotide ( Fig. 2A, compare lanes 4 and 10). Oligonucleotide JML9/10 was previously demonstrated to be an effective competitor (9,10). Oligonucleotides NB505 and NB506, containing the 10-bp consensus element but possessing different 5Ј-flanking sequences than oligonucleotide NB508, appeared to be as effective as NB508 in competing Ino2p⅐Ino4p binding ( Fig. 2A, lanes 8 (NB505) and 9 (NB506) compared with lane 10 (NB508); and Fig. 2C, lane 5 (NB505) compared with lane 6 (NB508)). Oligonucleotide NB503 (Fig. 2A, lane 6), which also retains the 10-bp consensus sequence was also active as a competitor, but it appeared to be a slightly less effective competitor than NB508. Oligonucleotide NB504, which had the lowest UAS INO activity of any of the elements with a 5Ј-flanking sequence modification (Table II), also had little if any ability to compete Ino2p⅐Ino4p binding ( Fig. 2A, lane 7). Thus, the relative ability of each of these oligonucleotides to compete Ino2p⅐Ino4p binding correlates generally with its strength as a UAS INO element (Table II, part A). Among the oligonucleotides that retained the 10-bp consensus and functioned as a strong UAS INO element, only NB503 appeared to have slightly reduced ability to serve as a competitor of Ino2p⅐Ino4p binding.
Finally, among the oligonucleotides having substitutions in the 3Ј region flanking the putative bHLH binding motif, NB301 ( Fig. 2A, lane 11), NB305 (Fig. 2C, lane 7), NB307 (Fig. 2C, lane  3), and NB308 (Fig. 2C, lane 4), were tested. Among this group of oligonucleotides, only NB308 was as effective in competing Ino2p⅐Ino4p binding as NB508. Oligonucleotide NB301 (Fig.  2A, lane 11) appeared to have little or no activity as a competitor, which is not surprising since pNB301 also has very low UAS activity (Table II, part C). NB307 and NB305 both appeared to have retained a little ability to serve as competitors to Ino2p⅐Ino4p binding, but they were clearly much less active than NB508 of Fig. 2 (compare lane 6 with lanes 3 and 7 in  panel C). It is not surprising that NB305 is a weak competitor since it also has relatively low UAS activity. However, pNB307 expresses an intermediate level of ␤-galactosidase activity (Table II) and is considerably more active as a UAS INO element than construct pNB305, yet NB307 appears no more active as a competitor than NB305.
Overall, with the exception noted above, the binding competition assay revealed a strong correlation between the ability of a particular oligonucleotide to serve as a competitor of Ino2p⅐Ino4p binding and its ability to serve as a UAS INO element. This correlation was particularly strong for substitutions within the bHLH core motif, where essentially all of the base changes eliminating the consensus produced a simultaneous reduction in UAS INO function and inability to compete Ino2p⅐Ino4p binding. In the 5Ј-and 3Ј-flanking regions, however, some base substitutions resulted in intermediate levels of UAS INO function and Ino2p⅐Ino4p binding. DISCUSSION The element, 5Ј-CATGTGAAAT-3Ј, was first identified as a repeated sequence within the promoters of the coregulated genes of phospholipid synthesis during computer-aided homology searches (1,2). The sequences within the INO1 promoter that contain this element were shown to be capable of forming specific complexes with the products of the INO2 and INO4 regulatory genes (10,11). The INO2 and INO4 gene products are members of the bHLH class of DNA binding proteins (9,18) and together form a heterodimer (11,12) that is required for derepression of INO1 and other coregulated genes encoding enzymes of phospholipid biosynthesis (2).
In the present analysis, we have shown that the putative core bHLH binding site (CATGTG) contained within the previously identified UAS INO , is absolutely required for its function as a UAS element. Likewise, the putative bHLH binding site is essential for Ino2p⅐Ino4p binding as measured by a competition assay. With respect to UAS INO function, as measured by the ability to support ␤-galactosidase expression in an expression vector, all substitutions within the bHLH core, without exception, were detrimental. None of the elements that had substitutions in the bHLH motif retained more than 67 units of ␤-galactosidase activity compared with 536 units for the consensus element. However, those elements that retained any residual expression of ␤-galactosidase above approximately 5% of the activity of NB508 (i.e. above about 20 -25 units of ␤-galactosidase) were still repressible. All elements having substitutions within the bHLH core, which were tested in the DNA binding competition assay, exhibited a substantial reduction in Ino2p⅐Ino4p binding when compared with the consensus oligonucleotide NB508. Thus, the putative bHLH binding site is crucial for both Ino2p⅐Ino4p binding and UAS INO function.
The data reported here also demonstrated that sequences flanking the canonical bHLH binding site in UAS INO are also critical to its function. In the region flanking the 5Ј end the 10-bp consensus element, 5Ј-CATGTGAAAT-3Ј, the presence of a C or a G in the first two positions (i.e. 5Ј-NNCATGTGAAAT-3Ј) enhanced the activity of the element. At the 3Ј end within the 10-bp consensus, an A was the optimal base in each of the first two positions immediately 3Ј to the bHLH core binding site (i.e. 5Ј-CATGTG(AA)AT-3Ј). Curiously, however, T, which is the consensus base for the 3Ј end of the originally identified 10-bp UAS INO element (i.e. 5Ј-CATGTGAAAT-3Ј), appears to be less active than a G in this position. It is clear, however, that sequences immediately 5Ј and 3Ј to the bHLH consensus element play a major role in the overall activity of UAS INO . Likewise, there was a general correlation between the level of UAS INO activity and the ability to serve as a competitor for Ino2p⅐Ino4p binding. The correlation between binding activity and UAS INO function, however, did not appear to be quite as strong for substitutions in the 5Ј-and 3Ј-flanking regions as it did within the bHLH core motif itself.
Flanking sequence-dependent alteration of binding affinity has been shown previously for regulatory proteins of the bHLH family (8, 19 -21). Using an in vitro binding site selection assay, Blackwell et al. (22) demonstrated that c-Myc-Max proteins, members of a mammalian bHLH protein family, bind not only to canonical motifs (CACGTG or CATGTG) flanked by variable sequences, but also to noncanonical sites. However, all of the noncanonical sites to which Myc-Max proteins bind are composed of fixed internal dinucleotide sequence (CG or TG) in the context of particular variations in the CANNTG consensus (22). UAS INO conforms to the canonical CATGTG bHLH motif. In S. cerevisiae, other proteins of bHLH family, including the PHO4 activator of yeast phosphatase genes and yeast centromere promoter-binding factor Cpf1/Cbf1 have been identified (23)(24)(25). Both of these factors bind the core sequence, CACGTG. It is interesting that the modified UAS INO sequence found in oligonucleotide NB610 (CACGTGAAAT), which resembles the sequence to which these two other yeast factors bind, also retains one of the higher levels of activity among the elements in this study with substitutions within the bHLH core. It is possible that the base substitution in pNB610 may have reduced the specificity of the element, thus allowing it to acquire some ability to respond to other bHLH regulatory proteins. This could explain the relatively high residual expression of this element under derepressing conditions, as well as its relatively low repression ratio in response to inositol and choline.
Finally, the data presented in this report demonstrated that the UAS INO element is involved in the regulatory response to the OPI1 gene product. Recently, it has been demonstrated that UAS INO is also involved in responding to the global regulatory element, SIN3 (26). In the case of the OPI1 gene, previous work had established that fragments of the INO1 promoter containing one or more native UAS INO elements were constitutively expressed in the opi1 genetic background (16). However, no DNA binding activity has been definitively associated with the OPI1 gene product 2 in contrast to the evidence establishing that the Ino2p⅐Ino4p complex binds directly to UAS INO elements in the INO1 promoter (10,11). In the present study, we demonstrated that all of the modified UAS INO elements that expressed 20 or more units of ␤-galactosidase activity under derepressing conditions in wild-type transformants were also expressed constitutively in the opi1 mutant. These results demonstrate that repression in response to the OPI1 gene product is mediated by UAS INO . Furthermore, the sequence specificity for the repression response that is mediated by the OPI1 gene product appears to be concordant with sequence specificity for UAS INO activity and Ino2p⅐Ino4p DNA binding. These results suggest that Ino2p⅐Ino4p binding and transcriptional activation, as well as repression in response to inositol and choline in the presence of an active OPI1 gene product, are all components of a single regulatory mechanism that functions through UAS INO .