Overlapping Positive and Negative GATA Factor Binding Sites Mediate Inducible DAL7 Gene Expression in Saccharomyces cerevisiae *

Allantoin pathway gene expression inSaccharomyces cerevisiae responds to two different environmental stimuli. The expression of these genes is induced in the presence of allantoin or its degradative metabolites and repressed when a good nitrogen source (e.g. asparagine or glutamine) is provided. Three types of cis-acting sites and trans-acting factors are required for allantoin pathway gene transcription as follows: (i)UAS NTR element associated with the transcriptional activators Gln3p and Gat1p, (ii)URS GATA element associated with the repressor Dal80p, and (iii) UIS ALL element associated with the Dal82 and Dal81 proteins required for inducer-dependent transcription. Most of the work leading to the above conclusions has employed inducer-independent allantoin pathway genes (e.g. DAL5 and DAL3). The purpose of this work is to extend our understanding of these elements and their roles to inducible allantoin pathway genes using theDAL7 (encoding malate synthase) as a model. We show that eight distinct cis-acting sites participate in the process as follows: a newly identified GC-rich element, two UAS NTR, two UIS ALL, and threeURS GATA elements. The two GATA-containingUAS NTR elements are coincident with two of the three GATA sequences that make up the URS GATAelements. The remaining URS GATA GATA sequence, however, is not a UAS NTR element but appears to function only in repression. The data provide insights into how these cis- and trans-acting factors function together to accomplish the regulated expression of the DAL7 gene that is observedin vivo.

Inducible nitrogen catabolic genes in Saccharomyces cerevisiae respond to at least two different types of signals. One is a dominant, general signal that monitors the overall nitrogen supply of the cell. The availability of readily transported and metabolized nitrogen sources result in high nitrogen catabolite repression (NCR) 1 (1,2). Alternatively, low concentrations of good nitrogen sources or the presence of only slowly transported and/or metabolized nitrogen sources in the environment of cells results in low NCR. Genes encoding proteins involved in the transport and metabolism of poor nitrogen sources are NCR-responsive, i.e. they are expressed at low levels under conditions of high NCR and at higher levels under conditions of low NCR; the latter is also referred to occasionally as derepression (1,2). The second signal is pathway-specific and derives from the presence of a particular nitrogen source (e.g. allantoin or one of its metabolites) (3). Responses to these two physiological signals are integrated at transcription resulting in fine regulation of catabolic gene expression that ranges from high to low as needed to exploit most effectively the prevailing environmental nitrogen supplies for the needs of the cell.
Allantoin catabolic pathway gene expression has been a useful model with which to study the nature of nitrogen regulatory signals and the cell's detection of and response to them (see Ref. 4 for a comprehensive review of the allantoin pathway literature; shorter literature reviews covering contributions from the range of investigators in the field of yeast GATA factors and nitrogen regulation per se appear in the introductions of Refs. 5 and 6). These studies identified three types of cis-acting elements and trans-acting factors associated with allantoin pathway genes (7). The cis-acting element mediating NCR-sensitive transcriptional activation is UAS NTR (UAS Nitrogen-Regulated) (9,10), consisting of two separated dodecanucleotides each with the sequence GATAA at its core (10). UAS NTR has been shown to be both necessary and sufficient for NCR-sensitive DAL gene transcription (11); Gln3p and Gat1p are required for this transcription (12). The Magasanik group (13,14) reported that antibody against a Gln3p peptide containing the GATA family zinc finger motif precipitated a synthetic DNA fragment containing seven repeats of a 32-bp GLN1 promoter fragment containing a GATA sequence from crude extract; this led them to suggest that Gln3p bound to GATA sequences. Their observation was consistent with the finding that the deduced Gln3p sequence contains a zinc finger motif homologous to the mammalian GATA-binding family of transcription factors (15). By using electrophoretic mobility shift assays (EMSAs), direct binding of Gln3p to UAS NTR sequences has been demonstrated (16,17). Multiple UAS NTR homologous sequences are situated upstream of all allantoin pathway genes, and these sequences have been shown to account for NCR-sensitive, Gln3pdependent expression of the inducer-independent DAL5 and DAL3 genes (8,10,11,18,38).
The cis-acting element mediating inducer responsiveness of the allantoin pathway genes is the dodecanucleotide element, UIS ALL (Upstream Induction Sequence) (7,19). The inducer to which proteins associated with the UIS ALL respond is the last intermediate of the allantoin pathway, allophanate, or its nonmetabolized analogue, oxalurate (20). The allophanate-inducible DAL genes contain one or two copies of UIS ALL (7,21). The DAL81/DURL/UGA35 and DAL82/DURM gene products are required for this inducer responsiveness (22)(23)(24)(25). Dal82p has been shown to be the UIS ALL DNA-binding protein whose binding to DNA is independent of inducer (21). Whereas Dal82p appears to be a pathway-specific regulatory element, Dal81p functions more broadly, being required for induced expression of the UGA (␥-aminobutyrate catabolic pathway) and AGP1 (tryptophan uptake) genes as well as those of the allantoin pathway (26,27).
A third type of cis-acting element and cognate transcription factor down-regulate DAL gene expression; they are URS GATA and Dal80p, respectively (28,29). The DAL80 locus was first identified genetically (28). Mutations in this gene increase allantoin pathway-inducible gene expression in the absence of inducer to the same level observed in a wild type strain grown with inducer. This observation led to the suggestion that Dal80p functions to reduce inducible gene expression when inducer is absent (28). Dal80p was subsequently found to perform a similar function for the inducible UGA genes (33,34). However, Dal80p also down-regulates inducer-independent DAL gene expression 2-20-fold (18), suggesting a more general physiological function than simply maintaining inducible gene expression at low levels. Gln3p and Dal80p were proposed to be opposing regulators of most NCR-sensitive nitrogen catabolic genes (30). This proposal has been subsequently supported by data from other laboratories (31). The deduced amino acid sequence of Dal80p contains a zinc finger motif homologous to the one in Gln3p (32,34). Prompted to test the implication of this homology, we demonstrated Dal80p to be a DNA-binding protein whose optimal binding site, URS GATA , consists of two GATAAG-containing sequences separated 15-35 bp, oriented tail-to-tail or head-to-tail (29). The requirement of two GATAcontaining sequences in a Dal80p-binding site correlates with the fact that Dal80p molecules have been recently shown to dimerize via a leucine zipper motif in their C-terminal domains using two-hybrid assays (35). The structural similarity of UAS NTR and URS GATA sites led to the suggestion that Dal80pbinding sites might be Gln3p-binding sites as well (18). This suggestion was supported by the demonstration that the GATA sequences of the UGA4 and DAL3 UAS NTR s that bind Gln3p also bind Dal80p (18). The fact that Gln3p can bind to a single GATA sequence while Dal80p cannot explains why some genes, such as DAL5, are Gln3p-dependent but largely immune from regulation by Dal80p (4).
The bulk of our understanding of allantoin pathway gene regulation and the role played by Dal80p down-regulating Gln3p transcriptional activation has been derived from studies with inducer-independent genes such as DAL5 and DAL3 (4). Our objective here is to extend these studies to the inducible allantoin pathway genes whose expression is not only NCRsensitive like that of DAL5 and DAL3, and Dal80p-regulated like that of DAL3, but also inducer-responsive. The inducible DAL7 gene is the one we investigated. The results obtained delineate which of the many DAL7 upstream sequences that are homologous to known transcription factor binding sites actually participate in DAL7 expression. They further identify a cis-acting site not previously known to mediate allantoin pathway gene transcription and suggest a crude picture of how Dal80p, Gln3p, and Dal82p might work together to regulate inducer-dependent DAL7 expression.

EXPERIMENTAL PROCEDURES
Strains and Media-The strains used in this work are listed in Table  I. When correlating data from this and other work, it is important to recognize that the fold induction (induced level divided by the basal level of activity) for and Dal80p regulation of DAL7 expression is highly strain-dependent. It ranges from as much as 20 -30-fold in strains such as RH218 (7) to as little as 2-4-fold in the case of strains derived from ⌺1278b. 2 The source(s) of these differences are not known. The media used in this work were all standard formulations.
Transformation and ␤-Galactosidase Assay-Transformation of S. cerevisiae and the procedures used to assay ␤-galactosidase have been described earlier (36). The cultures were grown (2% glucose, 0.17% YNB (Difco), and 0.1% of the indicated nitrogen source) for 16 -24 h to a cell density of A 600 nm ϭ 0.4 to 0.8 as measured on a Zeiss, Gilford, or Spectronic Genesys 5 spectrophotometer. After the desired cell density was reached, 10-ml samples were harvested and assayed (36).
Statistics of the ␤-galactosidase measurements and their associated error (normally 5-10%) have been analyzed in detail (36). In this regard, it is important to emphasize that data (absolute values) in Fig.  2 cannot be quantitatively compared with those in Figs. 3 or 4 even though some of the same plasmids were used. This limitation derives from the use of different spectrophotometers and different strains; the spectrophotometers mentioned above, as is true for all spectrophotometers, differ significantly in their light scattering characteristics. These differences can range up to 2-fold depending upon the density of the sample. Although it is inappropriate to compare the absolute values between separate figures, critical comparisons of values contained within a figure and the patterns of regulation observed from one experiment (figure) to another are justified (36).
Plasmid Constructions-The heterologous expression plasmids used in Figs. 2-4 were constructed using double-stranded, synthetic oligonucleotides (the coordinates are indicated) that were extended by the addition of five bases of the SalI and EagI restriction endonuclease sites to their 5Ј and 3Ј termini, respectively. These synthetic fragments were then cloned into plasmid pNG 15 (10) digested with the same enzymes. All constructs were sequenced before being used.
The native DAL7 promoter fragments used in Figs. 9 and 10 were created using PCR-based methods and cloned into the BamHI site of plasmid pVAN2 (Fig. 1) which generated plasmid pRR352, the wild type parent plasmid. Plasmid pVAN2 was derived from plasmid pHP41 (37) and does not contain any of the CYC1 promoter sequences. The method of creating mutations in the promoter region involved using mutagenic primers during PCR. The heat-stable DNA polymerase used in these reactions was PWO polymerase from Roche Molecular Biochemicals. Reaction conditions were standard and described by the manufacturer. All of the PCR-generated DNA fragments were cloned into the BamHI site of plasmid pUC18. This vector was chosen to facilitate cloning and DNA sequencing of the fragments. The sequences of all of the PCRgenerated DNA fragments were determined prior to being used. The mutations introduced into the promoter fragments are listed in Table  II. The mutant fragments were identical to the wild type parent except at the positions listed; those positions carried the indicated substitution mutations.
EMSAs-EMSAs were performed as described earlier (29) including preparation of the double-stranded oligonucleotides, Klenow-mediated fill in reactions to generate [ 32 P]dCTP-labeled probes, preparation of the Dal80p extracts, binding reactions, and electrophoresis. All oligonucleotides contained SalI and EagI overhangs at their 5Ј and 3Ј ends, respectively, to provide targets for the fill-in reactions and the ability to clone the fragments. In cases where a particular GATA sequence was mutated, the wild type sequence starting with the 5Ј-guanosine of the GATA was changed to a MunI restriction site. For example, if the wild type sequence was 5Ј-GATAAG-3Ј (or 5ЈGATAGT-3Љ), it was converted to the sequence 5Ј-CAATTG-3Ј. (See Table III for the mutated bases and their coordinates in the DNA fragments used in the EMSAs.) DNA fragments DAL3-5, containing the three clustered GATAs from the DAL3 gene, and DAL3-35, a synthetic fragment containing two GATA elements in tail-to-tail orientation, have been described in detail elsewhere (29).  tively acting DAL7 elements that participate in physiologically relevant heterologous reporter gene expression, UAS NTR and UIS ALL (7). However, these studies were unable to address adequately how these cis-acting elements and the trans-acting proteins associated with them cooperate to mediate inducer-dependent, Dalp80-regulated DAL7 expression. The inadequacy derived, in part, from a lack of knowledge about the biochemical function of Dal80p and clear identification of which of the cis-active element homologous sequences actually participated in DAL7 expression (7). When early 5Ј deletion data (derived from a DAL7-lacZ fusion plasmid (7)) are analyzed from the perspective of sequences homologous to those of known cisacting elements, there are instances in which single deletions potentially removed more than one cis-acting element. The region between positions Ϫ290 and Ϫ221, relative to the ATG, is not only an area where our information is incomplete but also one in which existing deletions have the greatest effect upon expression (7).

Upsteam Activation Sequences That
To characterize further the regions most responsible for DAL7 expression, we cloned a synthetic fragment, containing sequences from Ϫ320 to Ϫ199, into a heterologous expression vector (plasmid pJD98, see "Experimental Procedures"); sequential 5Ј deletions were then used to delineate the cis-acting elements (Fig. 2). The first two deletions, to Ϫ300 and Ϫ266, had little effect upon ␤-galactosidase production (Fig. 2, plasmids pJD98, pJD95, and pJD68). This argues that the UAS NTR homologous sequence between positions Ϫ281 and Ϫ286 (5Ј-ATTATC-3Ј) does not demonstrably function as a UAS element in this fragment. However, deletion of the next 12 bases (to position Ϫ254) decreased reporter gene expression 3-4-fold in the presence or absence of inducer (plasmid pJD72). This region contained a sequence homologous to UIS ALL and has been shown to bind Dal82p (21). These results argue that this UIS ALL sequence participates in reporter gene expression.
When the above data and those published earlier (7) are considered together, the only region of plasmid pJD68 not analyzed is between positions Ϫ244 and Ϫ230. To remedy this, we synthesized a mutant form of plasmid pJD68 in which a GC-rich sequence, 5Ј-CCGCGG-3Ј, at positions Ϫ240 to Ϫ235 was mutated. This particular sequence was chosen because it is a GC-rich inverted repeat, both characteristics of multiple transcription factor binding sites. This alteration affected reporter gene expression in two ways (Fig. 2, compare plasmids pJD68 and pJD92). First, basal and induced levels of ␤-galactosidase production decrease 110-and 12-fold, respectively. Second, inducer responsiveness increases 10-fold relative to the wild type; there is a 2-fold response with plasmid pJD68 and 20-fold with plasmid pJD92 (PROϩ level divided by the PRO level).
We next analyzed the potential cis-acting sites in plasmid pJD68 using point mutations that did not otherwise alter the number of bases or spacing of putative elements relative to the heterologous TATA elements and mRNA start sites. As shown in Fig. 3, these mutations decreased reporter gene expression in cells growing in glucose/proline medium 2-60-fold. A more modest result (13-fold maximum) was observed with glucose/  (10). The resulting plasmids were then transformed into wild type strain M1682-19b. Following growth to a cell density of A 600 ϭ 0.4 to 0.8 in yeast nitrogen base (1.7 g/liter; Difco) medium containing 0.1% proline and 2% glucose as sole nitrogen and carbon sources, respectively the cultures were harvested for enzyme assays. Inducer, oxalurate, was provided at a final concentration of 25 mM (PROϩ). Data from plasmid pJD68 were duplicated in the lower part of the figure to facilitate its comparison with that from plasmid pJD92. Ϫ252 to Ϫ263, GAAAGTTGCGGT GAcctTTGtgta pRR375 Ϫ245 to Ϫ249, GATAG tgacc pRR371 Ϫ235 to Ϫ240, CCGCGG gacgtc pRR376 Ϫ217 to Ϫ228, GAAAATTGCGTT GAcctTTtggaa pRR380 Ϫ204 to Ϫ208, TTATC ggtca pRR377 Ϫ171 to Ϫ175, TTATC ggtca pRR372 Ϫ163 to Ϫ167, GATAA tgacc a Underlined bases indicate those that were mutated.
Ϫ196 to Ϫ148 a The first base and the 4-base overhang of a SalI restriction site were added to the 5Ј end of each oligonucleotide. Similarly, 5 bases of an EagI site were added to the 3Ј end. asparagine medium. Mutation of the 5Ј UIS ALL element (plasmid pJD152) decreased reporter gene expression about 2.5fold, which is similar to the 4-fold decrease observed in Fig. 2. When the 5Ј-most UAS NTR homologous sequence was mutated, lacZ expression decreased about 2-fold (plasmid pJD156). Mutation of the GC-rich element (plasmid pJD92) has already been described, but it is worth noting that of all the mutations examined in Fig. 3, it was the one with the strongest effect, decreasing expression more than 50-fold. Mutation of the 3Ј UIS ALL decreased lacZ expression nearly 11-fold (plasmid pJD153). Similarly, mutation of the 3Ј UAS NTR decreased expression over 20-fold (plasmid pJD155). The combined roles of the UIS ALL elements were central to expression since mutating both of them reduced reporter gene expression 30 -35-fold (plasmid pJD154). It is important to note that these data demonstrate that UIS ALL plays a significant role in basal level as well as induced expression since no inducer was present in these experiments. Furthermore, mutating the 3Ј UIS ALL and UAS NTR elements resulted in 4 -10-fold less reporter activity than mutating the homologous 5Ј elements (compare plasmids pJD155, pJD153, pJD156, and pJD152).
Up to 5-fold less activity was observed with each of the mutant plasmids when asparagine was provided in place of proline as the nitrogen source (Fig. 3), indicating that reporter gene expression supported by plasmid pJD68 and its derivatives was NCR-sensitive. However, this NCR sensitivity was less than that observed when the intact gene (39) or a DAL7-lacZ fusion plasmid were analyzed (7); the earlier work was performed with strains that are different from the ones used here. Occurrences of the lowest NCR sensitivities (plasmids pJD92, pJD155, and pJD154) were coincident with those in which there was the least activity with proline as nitrogen source. In this regard, plasmids pJD92 and pJD154 generated a puzzling result. The UAS NTR element has been demonstrated by several laboratories to be the one upon which NCR sensitivity depends (7,18,19). Therefore, one would a priori have expected that expression supported by plasmids pJD92 and pJD154 to be as NCR-sensitive as wild type plasmid pJD68, or more so because (i) their UAS NTR elements are intact and (ii) elements supporting NCR-insensitive expression have been inactivated by mutation thereby decreasing their contribution to the overall results (Fig. 3). We cannot presently determine whether this lack of NCR sensitivity derived from a requirement of functional UIS ALL and GC-rich elements for the UAS NTR elements to carry out their normal function or, alternatively, that the mutant constructs in plasmids pJD92 and pJD154 supported insufficient expression to obtain a meaningful result. We tend to favor the latter explanation.
Next we compared reporter gene expression supported by the collection of plasmids used in Fig. 3 in wild type and gln3⌬ mutant strains (Fig. 4). ␥-Aminobutyric acid was used as the nitrogen source in this experiment because even though proline is the nitrogen source (among those we normally use) generating the least NCR, gln3 mutant strains grow very poorly when provided with proline (13). Therefore, note that all values derived from cells growing with ␥-aminobutyric acid as nitrogen source were 2-4-fold lower than with proline (compare Figs. 3  and 4). The Gln3p dependence of reporter gene expression was about 5-fold for most constructs (Fig. 4). Plasmids pJD156 and pJD155, both of which contain mutations in the UAS NTR elements, on the other hand, yielded opposite results. The Gln3p dependence of plasmid pJD156 increased approximately 2-fold relative to wild type, whereas that of plasmid pJD155 decreased by about the same amount. The source of this behavior has not been identified, although the effect may be due to the participation of Gat1p in transcription supported by one or both of the UAS NTR sites.
In sum, the data presented above suggest that at least five putative cis-acting elements participate in gene transcription supported by plasmid pJD68. Two observations, however, were unexpected. (i) These constructs supported lacZ expression that was less inducer-responsive than that supported by the entire DAL7 promoter (7); this derived from the elevated basal levels of activity in the constructs. (ii) The high basal level activity was largely dependent upon the GC-rich element. This is best seen when comparing the results with plasmids pJD68 and pJD92 (Fig. 3). Although mutating the GC-rich element dramatically decreased the overall levels of transcription, that which remained was far more inducer-responsive (Fig. 2).
Dal80p-binding Sites Upstream of DAL7-Dal80p has been shown to play a pivotal role in allophanate/oxalurate-induced allantoin pathway gene expression. This is evidenced by the observation that dal80 mutants exhibit a similar high level of DAL expression in the absence of inducer as when the wild type is assayed with inducer present (28). Searching for cis-acting sites required for Dal80p-regulated expression using the approach described in Figs. 2-4 is ill-advised because Dal80p binds to some of the same sequences as the transcriptional activator, Gln3p (18). Therefore, each deletion and substitution mutation would potentially generate two changes as follows: transcriptional activation and repression would both decrease. We avoided this complication by using an EMSA rather than comparing reporter gene expression in wild type and dal80 mutant strains. To identify Dal80p-binding site(s) upstream of DAL7, we synthesized several large DNA fragments containing promoter sequences previously shown to participate in control of DAL7 expression (Ref. 7 and Figs. 2-4 of present work) and used them as competitors for binding of a well characterized DNA probe (DAL3-5) containing a Dal80p-binding site. As shown in the upper panel of Fig. 5, fragment DAL7-3 (nucleotides Ϫ292 to Ϫ214) was ineffective as a competitor of DAL3-5 DNA for Dal80p binding, indicating the absence of a strong Dal80p-binding site (Fig. 5, upper panel, lanes G-M); DNA fragment DAL3-35 (29) was used as a positive control (upper panel, lanes A-G). Although the orientation and spacing of the GATA sequences on the DAL7-3 fragment appeared suitable for Dal80p binding (tail-to-tail, 31 bp apart), the 3Ј-GATA possessed the sequence GATAG. In contrast, DNA fragment DAL7-2 (nucleotides Ϫ239 to Ϫ156) was an effective competitor of DAL3-5 DNA for Dal80p binding (Fig. 5, lower panel, lanes   A-G). DNA fragment DAL7-1 (nucleotides Ϫ183 to Ϫ122) was also able to compete with DAL3-5 DNA for Dal80p binding but less well than the DAL7-2 fragment (lower panel, lanes G-M). DNA fragment DAL7-4 (nucleotides Ϫ265 to Ϫ196) was an ineffective competitor of DAL3-35 DNA for Dal80p binding (Fig. 6, lanes G-M); DNA fragment DAL3-3 (29) was used as the positive control (lanes A-G). The result with fragment DAL7-4 was not unexpected because only one of the sequences contained GATAA (the other was GATAG) and they are oriented head-to-head.
The DAL7-2 and DAL7-1 fragments each contained three GATA sequences, two contiguous GATAs (schematic at the top of Fig. 5, elements D and E) at one end of each fragment and a single GATA at the other (elements C or F). Based on earlier work defining the Dal80p-binding site (24), we predicted that the two GATA sequences at the 3Ј end of DNA fragment DAL7-2, or the 5Ј end of fragment DAL7-1 (Fig. 5, elements D and E) would perhaps be too close to one another to function as a strong Dal80p-binding site. If this were true, we hypothesized that either one or both of them, in combination with a single GATA sequence at the opposite end of the fragment, formed the Dal80p sites observed with the DAL7-2 and DAL7-1 fragments. To test this, we prepared variants of the DAL7-1 and DAL7-2 fragments in which each of the three GATA sequences was mutated; these fragments were then tested as competitors of the DAL3-5 DNA fragment for Dal80p binding (Figs. 7 and 8).
As shown in Fig. 7, fragment DAL7-8 (containing elements C and D) and DAL7-9 (containing elements D and E) were both competitors of the DAL3-5 fragment for Dal80p binding, with DAL7-9 being somewhat better than DAL7-8 (Fig. 7, upper  panel, compare lanes A-G with G-M). The effectiveness of fragments DAL7-9 and DAL7-8 as competitors correlates with data from the DAL3 gene demonstrating that two GATA sequences oriented tail-to-tail bind Dal80p better than a headto-tail orientation (29). We also expected that DNA fragment DAL7-10 would compete less well than fragment DAL7-9 because the GATA sequences it contained were only 8 bp apart (elements D and E, oriented tail-to-tail); this was confirmed experimentally (Fig. 7, compare data with DNA fragments, DAL7-8, 7-9, and 7-10). A smaller DNA fragment, containing only the closely spaced GATAs (elements D and E), behaved similarly to fragment DAL7-10 (DAL7-11, Fig. 7, lower panel,  lanes G-M).
We next assayed the GATA sequences contained on DNA fragment DAL7-1; recall that fragment DAL7-1 was a less effective competitor than fragment DAL7-2 (Fig. 5). In contrast to the data obtained with mutant alleles of fragment DAL7-2, the only DAL7-1 mutant allele that retained high effectiveness FIG. 5. Competition EMSAs between DNA fragments derived from the DAL7 promoter and a standard radioactive DAL3 fragment for Dal80p binding. The top portion of the figure is a schematic representation of the DAL7 promoter that includes the fragments (with end points indicated relative to the ATG) that were used as competitors. The radioactive probe was DNA fragment DAL3-5, containing a well characterized Dal80p-binding site (29). Lanes E-A and I-M contained increasing amounts of the unlabeled oligonucleotide as indicated above the autoradiogram. Lanes F and H have no added unlabeled competitor, and lane G has no protein extract added to the reaction mixture. Specific Dal80p-DNA complexes are indicated by arrows. as a DAL3-5 competitor was the one in which the isolated 3Ј-GATA (element F) was mutated (Fig. 8, lower panel, lanes  G-M, fragment DAL7-7). Mutation of either of the two closely spaced GATA sequences (elements D or E) significantly decreased the ability of the mutant fragments to act as effective competitors (Fig. 8, upper panel, lanes G-M, and lower panel,  lanes A-G). The DAL7-7 and DAL7-1 fragments were similar competitors of the DAL3-5 fragment for binding to Dal80p (compare Fig. 8, lower panel, lanes G-M with Fig. 5, lower  panel, lanes G-M). In other words, the closely spaced GATA sequences that were in common with fragment DAL7-2 (elements D and E) were more effective competitors than either of them paired with the isolated GATAG at the 3Ј end of fragment DAL7-1 (element F) and accounted for the overall ability of fragment DAL7-1 to serve as competitor. Whereas this observation might seem somewhat at odds with data derived from assaying fragment DAL7-2, the effectiveness of a Dal80p-binding site is determined not only by spacing and orientation of the GATAs but also by their specific sequences. Although the GATA sequences of DNA fragment DAL7-1 possess acceptable orientations and spacings, one of the two GATAs possesses the sequence GATAG. Together, these data identify the GATA elements C-D, C-E, and D-E (Fig. 8) as Dal80p-binding sites, URS GATA s, upstream of DAL7.
Effects of Promoter Mutations in the Context of a Complete DAL7 Promoter-The data described above in combination with earlier studies (7) yield a reasonably consistent picture of the cis-acting elements functioning in DAL7 expression. However, none of the mutations used in any of the DAL7 studies has ever been analyzed in the context of an intact DAL7 promoter. Although the contribution of cis-acting elements contained on small promoter fragments to the overall lacZ expression supported in a heterologous expression vector can be determined with certainty, their relation(s) to DAL7 expression in situ cannot be as confidently deduced. This problem was addressed by systematically mutating the putative cis-acting elements in an intact, wild type DAL7 promoter (Fig. 9). In each case, substitution mutations were used to ensure that the relative positions of the remaining promoter elements were not otherwise altered. Furthermore, the plasmids were CEN-based to eliminate conceivable "dose response" or transcription factor "titration" effects. Mutation of the 5Ј-most GATA sequence (plasmid pRR373) resulted in 25% less transcription in the presence of inducer compared with wild type (Fig. 9, plasmid pRR373 versus plasmid pRR352) but also doubled basal level expression. Overall, the effects of this mutation were relatively modest and consistent with observations made in Fig. 3. Mutation of the 5Ј-most UIS ALL element decreased induced expression 5-fold, uninduced expression 3-fold, and the inducer response from about 20-to 13-fold (plasmid pRR374 versus plasmid pRR352). These data correlate with those in Figs. 2 and 3, demonstrating that this element participates in DAL7 expression. Although inducer responsiveness is decreased in this mutant plasmid, it is still present, arguing that a single UIS element is sufficient to elicit this physiological response, FIG. 7. Competition EMSAs between DNA fragment DAL3-5 and mutant fragments derived from the DAL7-2 fragment for Dal80p binding. Reactions conditions and lane designations were as described in Fig. 6. The top portion of the figure diagrams the DAL7 promoter and the oligonucleotide fragments used in the experiment. ϫ indicates the mutated elements (see Table III).
FIG. 8. Competition EMSAs between DNA fragment DAL3-5 and mutant DNA fragments derived from the DAL7-1 fragment for Dal80p binding. All parameters were as described in Fig. 7. albeit at a more modest level than when two UIS ALL elements are present.
Mutation of the next GATA sequence (plasmid pRR375) yielded an unexpected result (Fig. 9). Based on data derived from the isolated promoter fragments (plasmids pJD68 and pJD156, Fig. 3), one would have expected mutation of this element to have a negative effect on reporter gene expression. However, in the context of the entire DAL7 promoter, it did not appear to participate in transcription (Fig. 9, plasmid pRR375)). A similar surprise was observed in the case of GCrich element. We had concluded from data derived with isolated fragments that this element played a central role in transcription, because it was the one that generated the most drastic decreases in transcription when mutated (plasmids pJD68 and pJD92, Figs. 2 and 3). Mutating the element in an otherwise intact promoter, however, decreased induced expression by only half (Fig. 9, plasmid pRR371); one can conclude that it is a participant in DAL7 transcription but probably not the central one. The elements downstream of the GC-rich element that are most important to DAL7 expression are the UIS ALL and UAS NTR elements at positions Ϫ228 to Ϫ217 and Ϫ208 to Ϫ204. Mutation of either element decreased basal and induced expression 4 -8-fold (Fig. 9, plasmids pRR376 and  pRR380). As in the case when the 5Ј UIS ALL element was mutated, the remaining lacZ expression in the 3Ј uis all mutant is inducible (Fig. 9, plasmid pRR376).
Previous experiments did not evaluate the function of either of the two contiguous GATA sequences oriented in opposite directions. Since early 5Ј deletions extending to position Ϫ205 no longer supported gene expression (7), the priority of studying these elements was low. However, since both GATAs participated in Dal80p binding, we tested each for their participation in transcriptional activation. Mutation of the 5Ј-most of these GATA sequences (corresponds to element D in Figs. 4 -8) decreased uninduced and induced expression levels 3-6-fold ( Fig. 9, plasmid pRR377). Mutation of the 3Ј-most of these GATA sequences, however, was largely without effect (plasmid pRR372). Therefore, the 3Ј-most of these GATA sequences (element E in Figs. 4 -8) participates predominantly in Dal80p binding rather than transcriptional activation.
Finally, we compared transcription supported by each of the above plasmids in wild type and dal80 mutant strains with proline as nitrogen source (Fig. 10). Reporter gene transcription supported by all but two of the plasmids increased 2-5-fold when DAL80 was disrupted. The two plasmids that lost their ability to respond were pRR380 and pRR371. In plasmid pRR380 the GATA sequence at position Ϫ208 toϪ204 was mutated; this is the UAS NTR that was most important to the expression of DAL7 (Fig. 9). Plasmid pRR371, on the other hand, carried a mutation of the GC-rich sequence. The loss of one of the two UAS NTR GATA sequences that participate in gene expression can be reasonably suggested to account for the data in plasmid pRR380. Similar reasoning cannot be used in the case of results derived with plasmid pRR371. However, it is important to note that reporter gene expression supported by plasmids pRR371 and pRR380 was below or barely above the background level observed in the parent vector, plasmid pVAN2.

DISCUSSION
The data presented above document the participation of eight cis-acting elements in the regulated expression of DAL7 (Fig. 11). Transcriptional activation depends upon a GC-rich element, two UIS ALL , and two GATA-containing UAS NTR elements. Transcriptional repression (i.e. Dal80p-binding), on the other hand, depends upon three URS GATA elements; two of the three GATA sequences associated with the URS GATA elements are also common to the DAL7 UAS NTR s.
Some of the proteins associated with the three types of cisacting elements have been identified as follows: UIS ALL elements bind Dal82p (21), which is required for induction of allantoin pathway gene expression (22,40). The GATA-containing UAS NTR elements bind Gln3p and are also thought to bind Gat1p as well (12,14,16,42). Although Gat1p-binding remains to be experimentally demonstrated, the protein contains a GATA-binding zinc finger motif similar to the one in Gln3p (12,36,43,44), and both proteins are known to be  Table II). Coordinates above the map are relative to the ATG of the coding sequence. transcriptional activators through which NCR operates (16). We speculate that the role of the UIS ALL -Dal82p complex may be to facilitate, or stabilize, Gln3p and/or Gat1p binding. Supporting this idea is the observation that a UIS ALL site placed close to a UAS NTR site, containing point mutations in it, will suppress those mutations (19); suppression requires Dal82p but not Dal81p or inducer. Whether this interaction is at the level of a Gln3p-Dal82p complex or further along the pathway of preinitiation complex formation is not known at present. It is interesting, however, that the distance between the functional 5Ј UIS ALL and UAS NTR elements is the same as between the 3Ј UIS ALL and UAS NTR elements (Fig. 11).
It should be noted that UIS ALL and Dal82p do not appear to be unique in their ability to function synergistically with Gln3p and/or Gat1p; the GC-rich element and the protein(s) presumed to be associated with it also appear to function in this capacity. Similar relationships have been observed between the UAS NTR and the Rap1p and Abf1p sites in the CAR1 promoter (36), the GATA element and the Put3p site of the PUT1 promoter (45), and the UAS NTR and Abf1p sites of CAR2 promoter. 3 Finally, we turn to the URS GATA elements that bind Dal80p, the repressor protein responsible for maintaining inducible DAL gene expression at low basal levels in the absence of inducer, and for down-regulating those genes (e.g. DAL3) whose expression is not inducer-dependent. Only three of six possible DAL7 GATA sequences appear to participate in Dal80p binding. Although the two 5Ј-most GATAs (Fig. 11, elements A and B) possess the appropriate orientation and spacing, they did not demonstrably bind Dal80p; the 3Ј of these two GATAs possess the sequence GATAG. Binding of Dal80p to the two GATAs at positions B and C did not occur because they are oriented head-to-head. DAL80p, on the other hand, does bind to the three GATA sequences at positions C, D, and E (Fig.  11). To the extent that the ability of a DNA fragment to serve as an effective competitor in an EMSA is a reflection of its binding affinity, Dal80p binding to these sites is, in order of decreasing affinity, C-E, C-D, and D-E (Fig. 11). We found earlier that a tail-to-tail orientation was the most favorable for Dal80p binding (29), a conclusion that correlates with the present data. It also makes sense that the D-E-binding site is the weakest site because the spacing is at the limit of the requirements determined in the DAL3 work (29). Finally, Dal80p binding to the D-F and E-F element pairs does not occur even though the orientation and spacing requirements are met, presumably because the F element does not possess the sequence GATAA but GATAG. Although participating as a Dal80p-binding site, the E element does not demonstrably function in transcriptional activation. To our knowledge, this is the first example of a GATA sequence that possesses this characteristic.
This work identifies a new cis-acting participant in DAL7 expression, the GC-rich sequence. Whereas this element plays an important role in transcription supported by the DAL7 promoter fragment contained in plasmid pJD68, its role appears somewhat less prominent or masked when assayed in the context of the intact DAL7 promoter. Earlier results may be consistent with the suggestion that the GC-rich sequence participates in gene expression in that the footprint covering the UIS ALL sequence extends to the GC-rich sequence as well ( Fig.  4 and see Ref. 19). We are not, however, suggesting that available data point to the existence of an interaction between proteins associated with the GC-rich and UIS ALL elements. Detailed analysis of the CAR1 gene (encoding arginase) also identified a positively acting GC-rich sequence similar to the one identified here (36). Whether these sequences represent the same cis-acting element or are just coincidentally similar sequences is not clear.
A model describing how the above cis-acting elements and trans-acting factors accomplish regulated expression of DAL and perhaps other inducible nitrogen catabolic genes was proposed some time ago (7,36,46). According to this model, NCRsensitive, inducer-independent transcriptional activation is mediated by UAS NTR , Gln3p, and the homologous Gat1p. Dal80p limits the interaction of GATA-containing UAS NTR elements with Gln3p by competing with it for binding to the GATA elements. We speculate that, when it is possible to perform the experiments, the same will be found to be true for Gat1p binding. In other words, the positively acting Gln3p/ GAT1p contribution to DAL gene transcription is balanced by the negative action of Dal80p; the metaphor of a see-saw was used in the original formulation of this model (7,36,46). The implication is that the effect of Dal80p impeding Gln3p and/or Gat1p binding is greater than its facilitation by Dal82p in the absence of inducer, and hence the DAL genes are expressed at only basal levels. Dal82p and Dal81p, associated with the cisacting element UIS ALL , shift the balance in the direction of elevated expression in an inducer-dependent manner (7). This hypothesis makes the prediction that gene expression should be influenced by the intracellular levels of both Dal80p and Gln3p/Gat1p. Recent evidence has shown this to occur. 4 This model of inducer-independent transcriptional repression acting in opposition to inducer-independent activation, with the balance being shifted by inducer-dependent transcriptional activation, has also been proposed to account for NCRsensitive, inducer-dependent regulation of the CAR genes (36,46). Although the cis-and trans-acting elements are different in the case of the CAR genes, their operation follows the same formal model as the allantoin pathway genes (36). Published studies of the NCR-sensitive, inducible UGA genes similarly fit this perspective (30).
Finally, GATA sequences situated contiguously head-to-tail or tail-to-tail are a rather common feature of NCR-sensitive gene promoters that respond to Dal80p regulation. In addition to DAL7, such contiguous GATAs are found in many promoters, e.g. DAL3, DUR1,2, DUR3, UGA4, CAN1, GAP1, and MEP2. They are rarely found in NCR-sensitive genes that are not Dal80p-regulated (e.g. work suggest that this array of contiguous GATA sequences is responsible for Dal80p regulation of Gln3p/Gat1p in the cases where it exists. In this regard, as more nitrogen catabolic promoters are dissected in detail, it will be interesting to assess how widely GATA sequences that participate in Dal80p binding but not transcriptional activation occur, as is the case for the E element in Fig. 11. In a technical context, this work has pointed out one of the limitations of heterologous expression vector systems in identifying promoter elements and analyzing their functions. We identified three instances in which the characteristics of a potential cis-acting site carried on a cloned promoter fragment are quite different than when the same site is contained within an intact promoter. For example, transcription supported by the promoter fragment carried in plasmid pJD68 appears less NCR-sensitive than when the entire promoter is assayed (10). This behavior may be explained by suggesting that (i) the GC-rich element is responsible for a greater proportion of the transcription supported by plasmid pJD68 than by an intact promoter, and (ii) the GC-rich mediated transcription is not NCR-sensitive. An analogous situation occurs when assessing the contribution of the GC-rich element to gene expression. When the GC-rich element is carried on plasmid pJD68, it is the major contributor to transcriptional activation. In contrast, this element seems to play a more limited role when present in the native promoter. Finally, the UAS NTR -homologous GATA sequence at positions Ϫ249 to Ϫ245 (element B) appears to contribute to reporter gene transcription supported by plasmid pJD156 in that ␤-galactosidase decreased 53% when it was mutated (Fig. 4, plasmid pJD156). However, when the sequence was mutated in an otherwise native DAL7 promoter, a much more modest effect was observed (Fig. 9, plasmid pRR375).