Mapping the DNA Binding Domain of the Zap1 Zinc-responsive Transcriptional Activator*

The Zap1 transcriptional activator ofSaccharomyces cerevisiae plays a major role in zinc homeostasis by inducing the expression of several genes under zinc-limited growth conditions. This activation of gene expression is mediated by binding of the protein to one or more zinc-responsive elements present in the promoters of its target genes. To better understand how Zap1 functions, we mapped its DNA binding domain using a combined in vivo and in vitro approach. Our results show that the Zap1 DNA binding domain maps to the carboxyl-terminal 194 amino acids of the protein; this region contains five of its seven potential zinc finger domains. Fusing this region to the Gal4 activation domain complemented a zap1Δ mutation for low zinc growth and also conferred high level expression on a zinc-responsive element-lacZ reporter. In vitro, the purified 194-residue fragment bound to DNA with a high affinity (dissociation constant in the low nanomolar range) similar to that of longer fragments of Zap1. Furthermore, by deletion and site-directed mutagenesis, we demonstrated that each of the five carboxyl-terminal zinc fingers are required for high affinity DNA binding.

The Zap1 transcriptional activator of Saccharomyces cerevisiae plays a major role in zinc homeostasis by inducing the expression of several genes under zinc-limited growth conditions. This activation of gene expression is mediated by binding of the protein to one or more zinc-responsive elements present in the promoters of its target genes. To better understand how Zap1 functions, we mapped its DNA binding domain using a combined in vivo and in vitro approach. Our results show that the Zap1 DNA binding domain maps to the carboxyl-terminal 194 amino acids of the protein; this region contains five of its seven potential zinc finger domains. Fusing this region to the Gal4 activation domain complemented a zap1⌬ mutation for low zinc growth and also conferred high level expression on a zinc-responsive element-lacZ reporter. In vitro, the purified 194-residue fragment bound to DNA with a high affinity (dissociation constant in the low nanomolar range) similar to that of longer fragments of Zap1. Furthermore, by deletion and sitedirected mutagenesis, we demonstrated that each of the five carboxyl-terminal zinc fingers are required for high affinity DNA binding.
Zinc is an essential trace element necessary for the growth of all organisms. Its nutritional importance can be illustrated by the large number of proteins that require zinc for their function. For example, there are over 300 known zinc metalloenzymes (1), and it has been estimated that as many as 1% of the genes in the human genome encode proteins with C 2 H 2 zinc finger motifs (2). In contrast, zinc excess can be toxic to cells so mechanisms of controlling intracellular zinc levels are critical to cell viability. In Saccharomyces cerevisiae, zinc homeostasis is largely regulated at the transcriptional level. The Zap1 transcriptional activator directly controls the expression of a number of genes in response to zinc (3). For example, in zinc-limited cells, Zap1 induces the expression of a high affinity zinc transporter encoded by the ZRT1 gene and a low affinity transporter encoded by ZRT2 (4,5). Zap1 also controls the transport of stored zinc out of the vacuole by regulating the expression of yet another transporter encoded by the ZRT3 gene. 1 In addi-tion to these important zinc transporters, DNA microarray analysis indicated that perhaps as many as 46 total genes in the yeast genome are directly regulated by this factor. 2 Clearly, Zap1 is a critical determinant of zinc homeostasis and the ability of cells to thrive under zinc-limiting conditions. Zap1 binds in a sequence-specific manner to an 11-base pair sequence called the zinc-responsive element (ZRE). 3 This element is found in one or more copies in the promoter regions of Zap1 target genes. Experimental evidence (8) and computeraided motif analysis of Zap1 target gene promoters 2 have indicated that the sequence 5Ј-ACCTTNAAGGT-3Ј is the preferred recognition site for this protein. Determining how Zap1 recognizes and binds to this sequence is a critical issue in understanding how Zap1 functions as a transcriptional activator. Zap1 contains seven potential C 2 H 2 zinc finger domains, and previous studies suggested that only a subset of these may be required for ZRE binding in vitro (8). In this study, we use a combined in vitro and in vivo analysis to map the DNA binding domain of Zap1 and determine which of its zinc fingers contribute to ZRE recognition and binding.
Plasmid Constructions-All plasmids generated for this study were confirmed by DNA sequencing. The ZAP1 open reading frame was PCR-amplified with primers containing added 5Ј and 3Ј SalI sites. This fragment was inserted into SalI-digested pGAD424 (CLONTECH) to create pGAD-Zap1 1-880 . Other GAD-Zap1 fusions were created similarly with 5Ј primers containing an added EcoRI site and 3Ј primers containing a BamHI site. The fragment containing GAD-Zap1 ⌬553-686 was generated by two-step overlapping PCR (10) and cloned into pGAD424 as described above. pGAD424 allows expression of cloned inserts from the ADH1 promoter. To allow expression from the GAL1 promoter, the fragments encoding the Gal4 activation domain-Zap1 fusions were subcloned into pRS316-GAL1 (11) by gap repair (12). Expression from the GAL1 promoter using the GEV ␤-estradiol-responsive system was performed as described by Gao and Pinkham. 4 Cells were co-transformed with plasmid pGEV-HIS3, which expresses a hy-* This work was supported by National Institutes of Health Grant GM58265. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  brid activator protein, GEV, that contains the Gal4 DNA binding domain, the human estrogen receptor hormone response domain, and the VP16 activation domain. The activity of the GEV protein was induced by adding 100 nM ␤-estradiol. The c-Myc epitope-tagged ZAP1 allele was expressed from plasmid pYef2-mycZap1. 5 The Gal4 DNA binding domain fusion plasmid pGBD-Zap1 552-880 was generated by PCR-amplifying the ZAP1 open reading frame encoding residues 552-880 using primers with added 5Ј EcoRI and 3Ј BamHI sites. The resulting fragment was inserted into BamHI-EcoRI-digested pMA424 (14). This allele and its mutant derivatives are expressed from the ADH1 promoter. ZAP1 truncate and the zinc finger mutant alleles were subcloned from pMA424 vectors into pGEX4T-1 (Amersham Pharmacia Biotech) using EcoRI and SalI sites to generate NH 2 -terminal glutathione S-transferase (GST) fusions.
Immunoblot and Immunofluorescence Analysis-Protein extracts were prepared as described previously (9), and protein concentrations were measured by the method of Bradford (15). Samples (5 g of protein/lane) were fractionated by SDS-PAGE and then transferred to nitrocellulose. Blots were incubated with monoclonal anti-Gal4 DNA binding domain antibodies (CLONTECH) or anti-Vph1 vacuolar ATPase 100-kDa subunit (Molecular Probes), washed, incubated with goat anti-mouse IgG antibody coupled to horseradish peroxidase (Pierce), and then detected by enhanced chemiluminescence (Amersham Pharmacia Biotech). Indirect immunofluorescence was performed as described previously (9) using anti-GBD primary antibody (CLON-TECH) and Alexa 488-conjugated goat anti-mouse IgG secondary antibody (Molecular Probes).
␤-Galactosidase Assays-Cells were grown for 15-20 h to mid-exponential phase (A 600 ϭ 0.3-0.7) in LZM supplemented with 5 M ZnCl 2 . ␤-Galactosidase activity was assayed as described by Guarente (16) and is expressed in units calculated as follows (⌬A 420 ϫ 1000)/(min ϫ ml of culture used ϫ absorbance of the culture at 600 nm). The ZRE-lacZ reporter used was pDG2 in which a ZRE fragment was inserted into a minimal CYC1 promoter lacking its normal upstream activation sequences (8).
Protein Expression and Purification-GST-Zap1 fusion proteins were expressed in BL21(DE3)pLysS cells pre-grown at 37°C to an A 600 of 0.4 in the presence of 0.5 mM ZnSO 4 . Following transfer to 30°C, expression was induced with isopropyl-1-thio-␤-D-galactopyranoside and the cells were incubated for an additional 4 h. Cells were harvested by centrifugation and washed in 0.25 M sucrose. The pellet was resuspended in buffer (20 mM NaH 2 PO 4 , 3.6 mM KH 2 PO 4 , pH 7.3, 280 mM NaCl, 5.4 mM KCl, 10% glycerol, 5 mM dithiothreitol) containing lysozyme for sonication. The lysate was clarified by ultracentrifugation (100,000 ϫ g, 30 min, 4°C) and the proteins were purified by affinity chromatography using glutathione (GSH)-Sepharose (CLONTECH). Eluted protein was concentrated in VIVA spin concentrators (6000 -8000-kDa cut-off), dialyzed against sonication buffer to remove GSH, and analyzed by 12% SDS-PAGE. The protein was estimated to be Ͼ85% pure, and 12 liters of cells typically yielded approximately 8 mg of protein.
Where specified, the GST-Zap1 538 -880 fusion was again bound to GSH-Sepharose and incubated with thrombin (0.1 unit g Ϫ1 protein in sonication buffer) for 6 h at room temperature. Following cleavage of the NH 2 -terminal GST fragment, the eluted Zap1 protein obtained was of the expected molecular mass and found to be Ͼ95% pure by gel electrophoresis. To obtain active protein, GSH affinity-purified GST-Zap1 538 -880 was further purified using a DNA affinity column. Biotinylated ZRE-1 and ZRE-2 oligonucleotides (see below) were annealed and bound to 5 ml of streptavidin resin (Pierce). Zap1 protein was bound to the resin and washed on ice in low salt buffer (3 mM NaH 2 PO 4 , 0.5 mM KH 2 PO 4 , pH 7.3, 42 mM NaCl, 0.8 mM KCl, 10% glycerol, 5 mM dithiothreitol). The protein was eluted with 0.75 M NaCl and dialyzed against sonication buffer.
Determination of Zn 2ϩ Stoichiometry-Amino acid analysis (Beckman 6300) was used to determine protein concentration following hydrolysis in 5.7 M HCl for 24 h in vacuo at 110°C. Simultaneous readings of iron, copper, and zinc were obtained by inductively coupled plasma emission spectroscopy using a Perkin-Elmer Optima 3100XL instrument. Zn 2ϩ levels were also measured by atomic absorption spectrometry (Perkin Elmer AAnalyst 100). Protein samples in sonication buffer were diluted in deionized water as needed prior to analysis.
The standard 15-l EMSA reaction contained hybridization mix (65 mM KCl, 0.2 mg ml Ϫ1 bovine serum albumin, 20 mM Tris-HCl, pH 7, with 20% glycerol and 0.04% IGEPAL CA-630; Sigma) with end-labeled oligonucleotides and protein in sonication buffer. Following incubation at room temperature for 15 min, the samples were applied to a 6% polyacrylamide nondenaturing gel and electrophoresed for 1.5 h at 30 mA. Gels and running buffer contained 1ϫ Tris borate buffer, pH 8, and they were pre-electrophoresed for 1 h at 30 mA. A reaction lacking protein was used as a "free probe" control. Dried gels were viewed by autoradiography. Results from electrophoretic mobility shifts assays were shown to be identical following incubation of binding reactions for 1 h, 30 min, and 5 min. Therefore, these reactions were judged to be at equilibrium.
Measurement of Apparent Dissociation Constants (K d ) -A series of 15-l reactions were prepared containing different concentrations of protein in hybridization mix and 32 P-radiolabeled oligonucleotides maintained at a constant DNA concentration. Reactions were allowed to reach equilibrium by incubation at room temperature for 15 min. The protein-DNA complex was then separated from free DNA by EMSA. Phosphorimages of dried gels were obtained and quantified using QUANTITY ONE software. The percentage of complex formation (as determined by the loss of free DNA) was plotted against concentration on a logarithmic scale to determine the apparent K d .

RESULTS
In Vivo Mapping of the Zap1 DNA Binding Domain-The functional and structural domains of Zap1 that were predicted from its primary amino acid sequence (3) are diagrammed in Fig. 1. Based on the distribution of acidic residues, this protein was predicted to have two potential activation domains designated AD1 and AD2. Zap1 also contains seven zinc finger motifs that fit the minimal consensus sequence C-X 2-4 -C-X 12 -H-X 3-5 -H. Our previous results suggested that the region containing the five COOH-terminal zinc fingers was sufficient for sequence-specific Zap1 DNA binding (8). A Zap1 fragment including only amino acids 687-880, i.e. zinc fingers ZnF3-ZnF7, produced in vitro by coupled transcription and translation reactions behaved in EMSA and DNase I footprinting assays in a similar fashion to the wild type protein. It was unclear from these studies, however, whether this domain retained the full DNA binding activity of the intact protein. As a first in vivo test of this hypothesis, the Zap1 687-880 region was fused to the Gal4 activation domain (GAD) and tested for its ability to complement a zap1⌬ mutation. The GAD fusion approach allows assessment of DNA binding ability without requiring Zap1 activation domain function. Furthermore, the GAD fragment used contains the VP16 nuclear localization sequence to facilitate trafficking of the fusions to the nucleus should the still un-mapped Zap1 localization signal be deleted.
Wild type and zap1⌬ mutant cells grow equally well on zinc-supplemented media (Fig. 2), but the zap1⌬ mutant is unable to grow under zinc-limiting conditions. When the fulllength Zap1 protein (in this experiment, a functional allele with six NH 2 -terminal c-Myc epitope tags) was expressed in the zap1⌬ mutant from the GAL1 promoter, wild type growth in low zinc was restored. Expression of the Gal4 activation domain Zap1 687-880 fusion (GAD-Zap1 687-880 ) also strongly complemented the mutant defect. These results supported the in vitro data suggesting that ZnF3-ZnF7 were sufficient for Zap1 DNA binding.
We noted that the zap1⌬ strain expressing GAD-Zap1 687-880 grew slightly slower than the same strain expressing fulllength Zap1 (data not shown). Therefore, as an independent test of these complementation results, we compared the ability of the GAD-Zap1 687-880 and other GAD-Zap1 fusions to activate expression of a ZRE-lacZ reporter gene in a zap1⌬ mutant (Fig. 3). These assays were conducted using cells grown in low zinc to avoid the repressive effect of zinc on Zap1 activity. In all cases, these fusions were found to complement the zinc-limited growth defect of the zap1⌬ mutant (data not shown). Although wild type cells showed high level lacZ reporter expression, the zap1⌬ mutant transformed with the vector alone showed only low levels of ␤-galactosidase activity. The zap1⌬ mutant expressing either the full-length Zap1 protein or amino acids 552-880 fused to the Gal4 activation domain showed similarly high levels of expression, i.e. ϳ80% of wild type. To test if ZnF1 and/or ZnF2 are important for DNA binding, the GAD-Zap1 642-880 fusion was constructed in which these two fingers are deleted. Although expression is somewhat diminished relative to the larger fusion proteins, this allele retained high levels of expression suggesting that these fingers are not necessary for DNA binding. The small reduction in expression is most likely caused by the removal of a Zap1 activation domain located within the 552-642 region. 5 Thus, deletions eliminating the NH 2 -terminal 640 amino acids of Zap1 can bind DNA in vivo. In contrast, the GAD-Zap1 687-880 fusion produced only about 20% of wild type expression levels. These results are consistent with at least two models. First, the ZRE binding activity of this Zap1 fragment may be partially impaired by deletion of the region from amino acids 642-686. Alternatively, fusion of the Gal4 activation domain close to the Zap1 DNA binding domain may inhibit the function of either domain perhaps through steric interference. An additional fusion, GAD-Zap1 ⌬553-686 in which the 642-686 region in question was deleted from the full-length protein, was constructed to address these models. This fusion showed wild type lacZ expression, indicating that the region of Zap1 from 642 to 686 is not required for DNA binding activity. Thus, the reduced ability of GAD-Zap1 687-880 to complement the growth defect of the zap1⌬ mutant is due to reasons other than deletion of important DNA

FIG. 3. Activity of Zap1-Gal4 activation domain fusions in vivo.
The indicated regions of Zap1 were fused to the Gal4 activation domain under the control of the GAL1 promoter and tested for their ability to activate expression of a ZRE-lacZ reporter in DEY1538 (zap1⌬). Expression was compared with ZRE-lacZ activity of chromosomal ZAP1 (YM4271, WT). All strains were co-transformed with pGEV-HIS3 allowing ␤-estradiol-regulated expression. Cells were grown to exponential phase in LZM supplemented with 5 M ZnCl 2 and 100 nM ␤-estradiol prior to ␤-galactosidase assay. A representative experiment is shown, and the error bars indicate 1 S.D.
FIG. 1. Potential domain structure of Zap1. The predicted activation domains (hatched boxes) and zinc finger domains (ZnF1-ZnF7, filled boxes) are shown. The sequences of the potential Zap1 zinc finger domains are shown below. The conserved Zn 2ϩ ligand residues are boxed, the normally hydrophobic "fingertip" amino acids are underlined, and the locations of the presumed ␤1, ␤2, and ␣-helix structures are shown. The amino acids relative to the start site of the ␣-helix are numbered Ϫ1 to 7, and the histidine to glutamine substitutions in the mZnF mutations are indicated. binding domain function.
Assessing the Role of Zap1 Zinc Fingers in DNA Binding by Site-directed Mutagenesis-To determine which of the Zap1 zinc fingers are required for specific ZRE binding, we used site-directed mutagenesis to convert the two conserved histidine Zn 2ϩ ligands in each finger to glutamines (Fig. 1). These mutations impair Zn 2ϩ binding in these sites and therefore interfere with proper formation of the zinc finger structures. The mutations were introduced into a fusion allele in which the amino acid 552-880 region of Zap1 was fused at its amino terminus to the Gal4 DNA binding domain (GBD). This domain was included to provide both an epitope tag and a nuclear localization signal for these proteins, and for future studies of Zap1 zinc responsiveness. When expressed in zinc-limited zap1⌬ mutant cells, GBD-Zap1 552-880 induced a high level of expression from the ZRE-lacZ reporter (Fig. 4A). Mutation of zinc fingers 1 (mZnF1) or 2 (mZnF2) individually or both ZnF1 and 2 together (mZnF1/2) resulted in no loss of expression, supporting the hypothesis that these domains are not required for DNA binding. In fact, lacZ expression in each of these mutants was increased by 50 -100% relative to the corresponding wild type protein fragment, suggesting that these mutations may increase activation domain function under these conditions. Similar mutations in ZnF3, ZnF4, ZnF5, ZnF6, or ZnF7 resulted in complete loss of ZRE-lacZ expression.
Although the simplest interpretation of these results is that each of these zinc fingers is required for DNA binding, a number of other explanations were also possible. For example, these mutant proteins may be unstable and fail to accumulate. Immunoblot analysis of total protein samples prepared from these strains demonstrated that this is not the case (Fig. 4B). All of the GBD-Zap1 proteins accumulated to similar levels. We did note that the nonfunctional GBD-Zap1 fusions also accumulated as lower mobility forms that may represent homo-or FIG. 4. In vivo analysis of the Zap1 zinc finger mutants. Strain DEY1538 (zap1⌬) expressing the vector (pMA424), GBD-Zap1 552-880 , or its mutant derivatives containing histidine to glutamine substitutions in the indicated zinc finger domains were assayed for their ability to activate expression of a ZRE-lacZ reporter (A), fusion protein accumulation by immunoblotting (B), and subcellular location of GBD-Zap1 proteins (C). In panels B and C, V ϭ vector-transformed cells, WT ϭ GBD-Zap1 552-880 , and mZnF1, mZnF2, etc., are shown. In panel B, the arrow indicates the expected GBD-Zap1 fusion and the asterisks mark the lower mobility forms. The positions of molecular mass markers and a Vph1 loading control blot are also shown. In panel C, cells expressing the indicated proteins were viewed by Nomarski optics or epifluorescence. DAPI was used to stain the nucleus and the GBD-Zap1 protein was detected by immunofluorescence. The blue fluorescence of DAPI staining was changed to red, and the DAPI and GBD-Zap1 images were overlaid using ADOBE PHOTOSHOP (Merge). Yellow color in the merged images indicates colocalization of the markers. Cells were grown to exponential phase in LZM supplemented with 100 nM ␤-estradiol and 5 M ZnCl 2 (panel A) or 1000 M ZnCl 2 (panels B and C) prior to analysis. The higher zinc concentration greatly facilitated the immunoblot and immunofluorescence experiments owing to the poor growth of some of the strains in low zinc. heteromultimeric complexes of the protein that are not disrupted by SDS-PAGE. Another explanation for the inactivity of these mutant proteins is that they fail to accumulate in the nucleus despite the presence of the Gal4 nuclear localization signal. To test this hypothesis, we examined their intracellular distribution by indirect immunofluorescence microscopy using an anti-GBD antibody. Little cell-associated fluorescence was observed in untransformed cells or vector-only controls, whereas the GBD-Zap1 fusions were easily detectable (Fig. 4C, data not shown). The wild type (Fig. 4C), mZnf1, mZnf2, and mZnF1/2 (data not shown) GBD fusions clearly localized in the nucleus. Mutants mZnF3 and mZnF7 were largely present in the nucleus, and fluorescence was also visible in the cytoplasm as punctate spots of apparently aggregated protein. In the mZnF4, mZnF5, and mZnF6 mutants, the proteins were mostly visible as cytoplasmic aggregates and little nuclear staining was observed. We can conclude from this analysis that ZnF3 and Znf7 are likely required for DNA binding; these proteins accumulate in the nucleus in substantial amounts but fail to activate transcription. However, it was not possible to assess if mZnF4, mZnF5, or mZnF6 could bind to a ZRE, i.e. their failure to activate in vivo could simply be due to the inability to accumulate in the nucleus. Finally, it is intriguing that all of the mutant forms of Zap1 that are seen as cytoplasmic aggregates also appear in lower mobility forms on SDS-PAGE, suggesting that their may be a link between these observations.
In Vitro Analysis of Zap1 DNA Binding-To complement this in vivo analysis, in vitro DNA binding studies using purified recombinant Zap1 protein were performed. Various Zap1 truncate and mutant proteins were expressed in E. coli as NH 2terminal GST fusions and purified using GSH-Sepharose affinity chromatography. Attempts to purify full-length Zap1 were unsuccessful because of the formation of insoluble inclusion bodies. However, GST-Zap1 538 -880 , GST-Zap1 611-880 , and GST-Zap1 687-880 were isolated, and each contains the functional DNA-binding domain (ZnF3-ZnF7) implicated by our in vivo studies. These GST-Zap1 fusion proteins were largely soluble in E. coli. The Zn 2ϩ content of GST-Zap1 538 -880 was found to be approximately 5 mol eq (Table I) Binding of these purified Zap1 proteins to a functional ZRE was analyzed by EMSA. Each protein bound specifically to a ZRE-containing oligonucleotide (Fig. 5). ZRE sequence specificity was demonstrated by the lack of protein-DNA complex formation using a mutant DNA oligonucleotide lacking the intact ZRE element (M2, Ref. 8) (data not shown). The apparent dissociation constant (K d ) for the GST-Zap1 538 -880 fusion protein-ZRE complex was determined to be approximately 2 nM (Fig. 5A, Table I). Protein-DNA complex formation was not lost with addition of Յ0.3 mg ml Ϫ1 poly(dI/dC) (data not shown). GST-Zap1 538 -880 protein forms two complexes with DNA in EMSA (Fig. 5B), perhaps indicative of limited proteolysis that does not affect DNA binding. Alternatively, we may be observing different oligomeric states of the Zap1 protein and this is currently under investigation. Removal of the GST fragment by thrombin cleavage of the GST-Zap1 fusion did not greatly alter the affinity of Zap1 for the ZRE duplex (Table I). Furthermore, the affinity was also not altered by further purification of the fusion on a ZRE-DNA affinity column. Thus, the GST-Zap1 fusions appear to be predominantly active for DNA binding.
The two other Zap1 truncates also showed high affinity ZRE binding interactions. Both GST-Zap1 687-880 (Fig. 5, C and D) and GST-Zap1 611-880 (Table I)  The high affinity ZRE binding of Zap1 687-880 containing only ZnF3-ZnF7 suggested that mutation of fingers 1 and 2 would not disrupt DNA binding. A mutant form of GST-Zap1 was purified in which zinc fingers 1 and 2 contained the same histidine to glutamine substitutions as were analyzed in vivo. The GST-Zap1 552-880 mZnF1/2 mutant protein bound DNA with a similar degree of ZRE specificity and affinity as did the wild type GST-Zap1 538 -880 protein (Fig. 5, E and F, Table I).
Mutation of any one of zinc fingers 3-7 resulted in the loss of this specific interaction with ZREs. These mutant proteins bound the ZRE duplex with a low affinity in the high nanomolar to low micromolar concentration range indicative of only low affinity interactions. Most of these zinc finger mutant proteins bound on average one less Zn 2ϩ atom per molecule than the wild type protein (Table I). The reason why mZnF6 and mZnF7 may retain a 5 mol eq Zn 2ϩ stoichiometry is not yet clear. Perhaps mutation of these fingers allow for Zn 2ϩ binding to alternate ligands in these domains.
Finally, two Zap1 truncates lacking portions of the protein's COOH terminus were purified and evaluated. These two truncates remove either 55 (GST-Zap1 552-825 ) or 83 (GST-Zap1 552-797) COOH-terminal residues. Both of these proteins exhibited low affinity for the ZRE duplex with apparent K d values in the micromolar concentration range. GST-Zap1 552-825 and GST-Zap1 552-797 bind ZREs only at protein concentrations Ͼ20 nM, and, in each case, complex formation was abolished with the addition of 0.04 mg ml Ϫ1 poly(dI/dC) (data not shown).

DISCUSSION
The fully functional DNA binding domain of Zap1 maps to the carboxyl-terminal 194 amino acids of Zap1, i.e. residues 687-880. Fusion of this region to the Gal4 activation domain complemented a zap1⌬ mutation for growth under zinc-limiting conditions and conferred expression from a ZRE-lacZ reporter gene in vivo. In vitro, a purified 687-880 fragment and a much longer fragment with amino acids 538 (or 552)-880 bound to a ZRE-containing duplex oligonucleotide with high affinity. The dissociation constant of this interaction was estimated to be in the 0.8 -4 nM range.
Region 687-880 contains five of the seven zinc finger domains found in Zap1 (Fig. 1). All seven of the Zap1 fingers fit the zinc finger consensus sequence and are predicted to form the ␤␤␣ structure found for other such domains. Zinc finger proteins bind to DNA through a generally conserved docking arrangement with each finger ␣-helix fitting into the major groove of the DNA double helix (17,18). Residues Ϫ1, 2, 3, and 6 (numbering with respect to the start of the ␣-helix) typically make key base contacts that are responsible for the sequence specificity of the interaction. Thus, each zinc finger commonly recognizes a 4-base subsite that can overlap the recognition site of an adjacent finger by 1 base. These considerations raise a question regarding which fingers of Zap1 contact DNA. Mutagenesis studies reported here demonstrate that all five zinc fingers are required for DNA binding, yet the 11-base pair ZRE sequence predicts that at most only three or four of these are involved in site-specific interactions. For example, three fingers in transcription factor IIIA are sufficient to recognize an 11-base pair binding site (19). Structural studies of the GLI zinc finger protein may explain this paradox (18). Like Zap1, GLI contains five tandem zinc fingers that are required for DNA binding. Although GLI zinc fingers 2 through 5 bind in the major groove, finger 1 does not contact DNA but rather makes extensive protein-protein interactions with finger 2. For Zap1, a similar structure may exist with ZnF4 -ZnF7 making major groove contacts and ZnF3 making protein-protein contacts analogous to GLI finger 1. Alternatively, Zap1 ZnF3 may make nonspecific DNA contacts with the phosphate backbone.
The hypothesis that Zap1 has a protein-DNA complex structure similar to that of GLI is supported by examining the Reactions containing increasing amounts of protein ranging from pM to M, with a constant radiolabeled ZRE concentration, were analyzed by EMSA to determine apparent dissociation constants (K d ) as measured at 50% protein-DNA complex. A free probe (FP) control lacking protein was also analyzed. The binding isotherms plots (upper panels) were generated by quantifying phosphorimages (lower panels) using QUANTITY ONE software. Proteins analyzed were GST-Zap1 538 -880 (panels A and B), GST-Zap1 687-880 (panels C and D), and GST-Zap1 552-880 mZnF1/2 (panels E and F). A representative experiment for each protein is shown. sequences and lengths of the spacer regions between the fingers. Adjacent zinc fingers known or predicted to make major groove contacts are typically separated by a linker with the consensus sequence (T/S)-G-E-(R/K)-P (6,17,18). Sequences that precisely match this consensus separate Zap1 ZnF5 and ZnF6, and ZnF6 and ZnF7, whereas a closely related sequence, SKYKP, separates ZnF4 and ZnF5 (Fig. 1). The presence of these conserved linker sequences suggests that ZnF4 -ZnF7 make major groove contacts. A very different linker in terms of both sequence and length, LTRGKSE, separates ZnF3 and ZnF4, suggesting that ZnF3 does not bind in the major groove. The preceding discussion assumes that Zap1 binds to DNA as a monomer, as is commonly found among zinc finger proteins. However, if Zap1 binds as a dimer, as is predicted by the palindromic nature of the ZRE (7,8), perhaps as few as two fingers of each monomer make major groove contacts. The other fingers may form a dimer interface (7) or make phosphate backbone DNA contacts.
Significant progress has been made recently in devising a protein-DNA recognition code for zinc finger proteins like Zif268 (13). Such a code, in which the identity of DNA contacting residues can be predicted from the nucleotide sequence of the recognition site, and vice versa, would be extremely useful for designing zinc finger proteins that bind specifically to target DNA sequences. Attempts to apply the current recognition codes to predict which fingers of Zap1 bind to which bases of the ZRE have yet to provide unambiguous results.
Surprisingly, our results demonstrate that Zap1 ZnF1 and ZnF2 are not required for high affinity ZRE binding. Either deletion of these two finger domains or site-directed mutagenesis such that they could no longer bind Zn 2ϩ failed to reduce the ability of Zap1 to function in vivo or bind to DNA in vitro. Furthermore, our results suggest that ZnF1 and ZnF2 bind Zn 2ϩ with lower affinity than do the other five Zap1 finger domains. GST-Zap1 538 -880 contains 5 mol eq of Zn 2ϩ when purified from E. coli and mutation of ZnF1 and ZnF2 had no effect on this stoichiometry. One explanation for their apparent lower affinity for Zn 2ϩ is suggested by the sequence of these two fingers. Most zinc fingers have a hydrophobic residue located in a turn between ␤2 and the ␣-helix that provides part of a stabilizing hydrophobic core at the "tip" of each finger. For Zap1 ZnF3-ZnF7, this residue is a phenylalanine (Fig. 1).
However, ZnF1 has a cysteine and ZnF2 has a glycine in the corresponding position, and neither of these residues would provide this important hydrophobic interaction.
What role then do these two fingers play in Zap1 function? One potential function is as Zn 2ϩ sensors; Zap1 binds to DNA and activates transcription under low zinc conditions, but this activity is repressed in high zinc. Although the precise mechanism of this regulation is unknown, we postulate that Zap1 is the primary zinc sensor and that binding of the metal to lower affinity sites in the protein repress its activation function. The apparently low affinity of ZnF1 and ZnF2 for Zn 2ϩ suggests that these two fingers may be regulatory rather than structural components. In this regard, it is intriguing that ZnF1 and ZnF2 are found within an activation domain of Zap1; perhaps, binding of Zn 2ϩ to these fingers may control the activity of this domain. The observation that ZnF1 and ZnF2 mutations increase ZRE-lacZ expression (Fig. 4A) are consistent with this model. Ongoing studies are addressing this hypothesis and further characterizing the roles of the zinc fingers in Zap1 function.