Autoregulation of the yeast copper metallothionein gene depends on metal binding.

The yeast CUP1 gene product, copper metallothionein, acts to repress the basal transcription of its own structural gene. By creating a series of truncation and amino acid substitutions in CUP1, we show that the ability of the protein to autoregulate is directly correlated to its ability to bind and detoxify copper. These results support a model in which metallothionein controls the level of free intracellular copper available to interact with positive transcription factors. In addition, mutations in chemically equivalent cysteine residues were functionally dissimilar, suggesting that partial sites in the molecule are critical for the formation of the sulfur-metal cluster.

The yeast CUPl gene product, copper metallothionein, acts to repress the basal transcription of its own structural gene. By creating a series of truncation and amino acid substitutions in CUPl, we show that the ability of the protein to autoregulate is directly correlated to its ability to bind and detoxify copper. These results support a model in which metallothionein controls the level of free intracellular copper available to interact with positive transcription factors. In addition, mutations in chemically equivalent cysteine residues were functionally dissimilar, suggesting that partial sites in the molecule are critical for the formation of the sulfur-metal cluster.
There are two natural variants of the yeast Saccharomyces cereuisiae with respect to copper resistance: those sensitive to 0.3 mM CuS04, designated cupl and those resistant to this level of copper, designated CUP1 R. This difference in copper sensitivity is mediated by the CUPl locus, which is present in a single copy in cupl strains but is tandemly reiterated in CUPIR strains (Fogel and Welch, 1982). Copper resistance is not only afforded by amplification of the CUPl locus, but also by induction of CUPl gene transcription by the addition of exogenous copper to the growth medium of yeast cells (Karin et al., 1984;Butt et al., 1984a). Deletion studies of the CUPl promoter have defined a duplicated positive regulatory sequence necessary for copper induction as well as a putative negative regulatory element .
Sequence analysis of the CUPl gene predicts its product to be a 61-amino acid, cysteine-rich protein which resembles mammalian metallothioneins (Karin et al., 1984;Butt et al., 1984a). Structural studies performed on mammalian metallothionein have revealed that the protein exists in different conformational states depending on which metal is bound to it. NMR and x-ray crystallography studies have shown the Cd,Zn form of the protein to be a two-domain structure, a carboxyl-terminal portion containing 11 Cys residues and binding to 4 metal ions and an amino-terminal domain containing 9 Cys residues and binding to 3 metal ions (Boulanger et al., 1983;Furey et al., 1986). Metal reconstitution studies have shown that the binding of these metals to each domain * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported by Public Health Service National Research Service Award 1 F32 GM 10958-01 from the National Institute of General Medical Sciences. PreseEt address: Laboratory of Viral Diseases, National Inst. of Allergy and Infectious Diseases, Bethesda, MD 20892.
is cooperative and that the four-metal cluster is filled first (Nielson and Winge, 1983). In contrast to the above results, additional studies have shown that mammalian metallothionein is capable of binding to 12 Cu or Ag ions with the amino-terminal domain being filled first (Nielson and Winge, 1984). Recent in vitro experiments have shown that yeast metallothionein, like its mammalian counterpart, is capable of binding to a variety of heavy metals with different stoichiometries (Winge et al., 1985). The yeast protein binds 8 Cu ions or 4 Cd ions, again suggesting that it exists in different configurations depending on the metal ligand. The structure of yeast metallothionein has not been studied in detail, and as yet no intermediates of metal binding have been isolated.
Yeast, unlike mammalian cells, provides an experimentally accessible system to study structure-function relationships for a metal binding protein. Not only does yeast genetics allow mutated forms of CUPl to be studied in a background lacking the endogenous gene, but yeast cells have also been shown to be a convenient source from which metallothionein can be isolated in large quantities (Winge et al., 1985). Two functions for the CUPl protein have been determined by constructing yeast strains in which the CUPl locus is deleted from the chromosome (cupl A) . The first is copper detoxification, as such yeast strains are extremely susceptible to copper poisoning. The second is autoregulation; deletion strains transformed with a plasmid containing the CUPl promoter fused to the structural gene for Escherichia coli galactokinase (galK) express this enzyme activity at a high constitutive level. However, if the CUPl gene is present in the cell, either chromosomally or episomally, the basal level of CUPl transcription is repressed, and the fusion product is seen to be copper-inducible. Two mechanisms have been proposed to explain these results: either the CUPl protein represses transcription of its own structural gene by binding to the promoter region, or it simply sequesters copper away from as yet undefined factors which require copper to activate CUPl transcription. If the first model were true, it might be possible to isolate CUPl mutants that bind copper but fail to autoregulate or vice versa. In contrast, the second model predicts a direct correlation between copper binding and autoregulation in all mutants.
To distinguish between these models, we have generated a series of mutants in the CUPl coding sequences. These include four carboxyl-terminal truncations of the CUPl protein, four pairs of cysteine-to-serine substitutions, and a double lysine-to-alanine substitution in the most conserved portion of the molecule. Characterization of these mutants has elucidated two key points: 1) autoregulation occurs by a copper chelation mechanism rather than a direct DNA binding mechanism, in that there is a direct correlation between copper detoxification and CUPl gene transcription; and 2) the cys-teine residues are not functionally equivalent.

MATERIALS AND METHODS
Plasmids and Strains Used-The plasmid YEp36 has been described previously (Butt et al., 198413) and consists of the yeast 2p and LEU2 sequences, pBR322 sequences, and a 1.1-kb' fragment of DNA containing the CUP1 coding and flanking sequences. The plasmid 336ASst is a derivative of YEp36 in which an approximately 400-base pair SstII fragment was deleted between two SstII sites which were introduced 5' and 3' to the CUPl coding sequence. The plasmid RC4 has also been described previously (designated pYSK in Butt et al., 1984a) and contains a CUPZ::galK fusion gene on a vector that contains the TRPl gene as a selectable marker and ARSl and CEN3 sequences to promote stable low copy number replication.
(It0 et al., 1983) into strain 55.6B (Mata, trpZ-Z, leu2-3, ku2-ZZ2, Preparation of the Lysine-to-Alanine Substitutions-The double lysine-to-alanine substitution was constructed by oligonucleotidedirected mutagenesis as follows: an approximately 1.1-kb BanHI fragment of DNA containing the CUPZ coding and flanking sequences was gel-purified and mixed in a ligation reaction with BamHI-cleaved, phosphatase-treated M13mp8 DNA. The ligation reactions were transfected into E. coli strain JMlOl and plated. Plaques were picked and single-stranded DNA prepared by standard procedures (Sanger et al., 1980) and subjected to sequence analysis to determine the orientation of the insert. Single-stranded DNA was prepared from a clone which contained the CUP1 noncoding strand and annealed to an oligonucleotide with the sequence 5'-TCTGAAGAAACCg-GgGTCATGCTGCTCT-3'. This oligonucleotide contains four mismatches with the wild-type CUP1 template (underlined above) creating a SstII site in the DNA and converting lysine residues 54 and 55 of CUPl to alanine residues. Annealing and extension reactions and purification of closed-circular duplex DNA was performed as described by Winter et al. (1982). The closed-circular duplex DNA was then transfected into E. coli strain JMlOl and DNA from the the desired base changes. After purification, single-stranded DNA resulting plaques screened by restriction analysis for those containing from a clone containing an SstII site was also subjected to sequence analysis to assure that no extra base changes were present in the CUPZ coding sequence. Duplex DNA from these phage was isolated, digested with BamHI, and mixed in a ligation reaction with BamHIcleaved, phosphatase-treated YEp36 DNA. The ligation mixture was transformed into E. coli strain HBlOl and the resulting clones screened by restriction analysis for the presence and orientation of the insert. The resulting plasmid is identical to YEp36 except for the base changes in the CUPZ coding sequence.
Preparation of Truncation and Cys-to-Ser Mutants-The preparation of M13mplO vectors containing premature translational termination codons and the cysteine-to-serine changes in the CUPl coding sequence have been described previously (Wright et al., 1986). These mutations were subcloned by cleaving the mpl0 DNA with SstII and mixing in a ligation reaction with 336ASst digested with SstII and phosphatase-treated. The orientation of the inserts was determined to be the same as the CUPZ gene in YEp36 by restriction enzyme analysis.
These mutations were further subcloned onto single copy vectors as follows: the YEp36 derivatives were digested with BamHI, and the resulting 1.1-kb fragment containing the CUPl coding and flanking sequences was gel-purified and mixed in a ligation reaction with RC4 digested with BglII. The ligation reactions were digested with BglII before transformation into E. coli strain MC1061. The resulting clones were screened by restriction analysis to determine the presence and orientation of the insert. These manipulations created a series of plasmids that contain the yeast TRPZ-ARSl sequence, pBR322 sequences, a tribrid fusion consisting of the CUPZ promoter driving the g a K structural gene and containing the CYCZ 3'-flanking sequences, and a 1.1-kb fragment containing the CUP1 structural gene and flanking sequences.
RNA Blot Anulysis-Yeast cultures were grown in synthetic dextrose medium, induced with copper, and RNA extracted as previously described  except that total cellular RNA was used. Blots were probed with one or more of the following 32P-labeled probes: a 700-base pair XbaI/KpnI fragment containing the CUPl coding sequences; a 1-kb EcoRI fragment containing the galK coding The abbreviation used is: kb, kilobase.
sequences; a 2.5-kb BglII fragment containing the yeast LEU2 coding sequences; or a plasmid containing the yeast PYKZ coding sequences.
Galuctokinuse Assays-Yeast cultures were grown in synthetic dextrose medium, induced with 0.05 mM copper, and galactokinase assays were performed as described previously .

RESULTS
We used oligonucleotide-directed mutagenesis in a CUP1-lac2 construct (Wright et al., 1986) to generate three classes of mutants: 1) carboxyl-terminal truncation mutations in which translational stop codons were introduced into the CUPl coding sequence at positions 17,32,44, and 57 (see Fig.  1); 2) amino acid substitutions in which pairs of cysteine residues at positions 17 and 19,32 and 34,44 and 46, and 57 and 58 were converted to serine residues; and 3) a double lysine-to-alanine substitution in the carboxyl-terminal hexapeptide sequence Lys-Lys-Ser-Cys-Cys-Ser which is highly conserved in all metallothioneins (Butt et al., 1984a).
All mutations were transformed into a yeast strain lacking a chromosomal copy of CUPl (cuplh), allowing us to study copper detoxification and regulation of the mutant proteins in a background lacking the wild-type protein.
Copper Detoxification by Mutants-The CUPl coding and flanking sequences from all of the mutations were transferred from the M13 vectors on which they were constructed to the high copy number yeast vector YEpl3, which contains pBR322 sequences, the yeast LEU2 gene, and sequences from the yeast 2p circle. The resulting plasmids were transformed into a cuplA yeast strain. To determine the ability of the mutant CUPl genes to support metallothionein synthesis, [35S]cysteine-labeled cell lysates from the yeast transformants grown in the presence of copper were run on 20% polyacrylamide gels. The CUPl gene product was identified by its mobility and by its intense labeling with this isotope . The results from this analysis (data not shown) showed that the stop 44 and stop 57 mutants and all of the substitution mutants were expressed at approximately the same level as the wild-type CUPl gene. The peptides produced from the shortest truncations, stop 17 and stop 32, were not visible on this gel; whether this resulted from instability of these proteins or whether these peptides were simply not retained on the gel was not determined. Next, the transformants were streaked onto a series of plates containing different concentrations of copper. As can be seen in Fig. 1, the degree of copper protection, which presumably reflects the ability of the mutant proteins to bind to copper, varied over a wide range. The truncations showed an essentially linear response with the longest, stop 57, conferring wild-type protection t o copper; the two shorter truncations, stop 17 and stop 32, conferring very little resistance; and stop 44 conferring an intermediate resistance. The amino acid substitutions also followed a pattern in their ability to impart copper resistance. The substitutions at the carboxyl terminus of the molecule did not greatly affect the ability of the protein to detoxify copper. In contrast, the substitutions at the amino terminus of the molecule, in particular the ser 32/34 double mutant, were noticeably more copper-sensitive than yeast transformed with the wild-type CUPl gene. These results suggest that all of the Cys residues do not participate equally in copper binding since these mutants, which all contain the same number of Cys residues, have different degrees of copper resistance. The double lysine-to-alanine substitution in the conserved hexapeptide was functionally equivalent to the wild-type with respect to copper detoxification. Therefore, this conserved hexapeptide sequence, at least in yeast, does not seem to play an important role in the ability of metallothionein to bind metal. Cells used to prepare the RNA were grown to log phase in synthetic dextrose medium and then either incubated in the presence (+) or absence (-) of 0.5 mM CuSO, for 1 h at 30 "C. The transformant containing the wild-type CUPl gene is designated YEp36, other strains contain mutant CUP2 genes, as designated in Fig. 1, on this same vector. After hybridization to the CUPl probe, the blot was washed and subsequently hybridized to a probe for the yeast LEU2 sequences (data not shown). Scanning laser densitometry of this blot revealed that approximately the same amount of RNA was present in each lane of the gel. Also, since the LEU2 gene is present on the multicopy vector, this hybridization suggests that there is no significant variation in copy number of this plasmid in the various transformants.
Regulation of Mutant Gene Expression-We next analyzed RNA from yeast cells transformed with the wild-type or mutant plasmids to investigate the regulation of CUPl mRNA. As can be seen in Fig. 2, cells containing the wildtype CUPl gene (YEp36) express CUP1 mRNA at a low basal level which is induced 10-20-fold by the addition of exogenous copper to the growth medium. In general, cells transformed Stable, low copy number replication of this vector is maintained by the ARSl and CEN3 sequences; 3"termination signals are provided by yeast CYCZ sequences. The wild-type or mutant CUP1 genes were inserted into RC4 to create the plasmids RCG24 and mutant derivatives.
with the mutant CUPl genes exhibited an inverse relationship between the basal level of CUPI mRNA and the ability to detoxify copper. For example, the stop 17 mutant protein, which confers little or no copper detoxification, also showed the highest level of basal CUPl mRNA expression. Conversely, the double lysine-to-alanine substitution (ala 54/55), which had wild-type ability to detoxify copper, also had a normal low basal level of CUP1 mRNA. The one exception to this pattern was the stop 32 protein which conferred no copper resistance yet had a low basal level of expression. However, it was difficult to interpret this result because the induced level of expression in this mutant was also low, presumably reflecting instability of the mRNA encoded by the mutant gene.
Use of a Fusion Gene to Quantitate Autoregulation-To more carefully quantitate the ability of the mutant proteins to regulate CUPI gene transcription, we tested their effects on the expression of a CUPI ::galK fusion gene (Butt et al., 1984a). For this purpose, the mutated genes were subcloned into a low copy number yeast vector containing CEN3 sequences and a fusion gene between the CUPl promoter and the E. coli galactokinase coding sequences (Fig. 3). The ad-vantage of this strategy is that the effect of each mutated protein is measured in trans on the same transcription unit rather than on different, mutated genes. Another, advantage was the ability to monitor the copper detoxification of the mutant proteins on a low copy number vector. The copper resistance results are shown in Fig. 1 and agree qualitatively with the results obtained with the high copy number vectors. As expected, however, these cells are all more copper-sensitive than cells transformed with the multicopy vectors.
The basal galactokinase activity in extracts of cells grown in the absence of added copper and transformed with the above plasmids are shown in Table I. In addition, galactokinase activity was measured in a strain transformed with a plasmid containing only the CUP1::galK fusion (RC4); no structural gene for CUPl is present in these cells. In agree-  ment with previous results , galactokinase activity is expressed a t a high level in these cells, even in the absence of added copper in the medium. Cells in which the wild-type CUP1 gene is also present on the fusion plasmid We also analyzed the steady-state levels of galactokinase mRNA in the various transformants. As can be seen in Fig.  4, the basal levels of galK message correlate well with the levels of galactokinase enzyme activity. In particular, the stop 32 mutant shows a high basal level of galK mRNA which confirms the result obtained by assaying galactokinase activity but contrasts with the extremely low basal level of CUPl mRNA expressed from the stop 32 mutant gene. A possible explanation for this is that the base mutations introduced into the CUPl coding sequence of this mutant causes the mRNA to be labile and therefore gives the appearance of a low basal message level.

DISCUSSION
Using oligonucleotide-directed mutagenesis we have introduced four truncation mutations, four pairs of cysteine-toserine substitutions, and a double lysine-to-alanine substitution into the CUPl gene product. Characterization of these mutants for their ability to detoxify copper and down-regulate CUPl basal transcription has elucidated two points: 1) autoregulation occurs via a copper chelation mechanism; and 2) the cysteine residues of yeast metallothionein do not participate equally in metal detoxification and, presumably, metal binding.
Previous experiments had shown that the CUPl gene product is necessary for transcription from the CUPI promoter to be reduced to a low basal level. However, these experiments did not address the mechanism of this regulation. Two models which were proposed were either that the CUPl protein is a transcriptional repressor (direct or indirect) or that the CUPl protein simply controls the levels of intracellular copper by chelation. The latter model requires that other transcriptional factors bind to the CUPl promoter to stimulate CUPl transcription in the presence of copper. We reasoned that if we impaired the ability of yeast metallothionein to bind copper by removing cysteine residues from the protein, we could discriminate between these two models. For example, if copper chelation is the autoregulatory mechanism, then a decreased ability to bind copper will necessarily lead to a decreased ability to regulate transcription. On the other hand, if the CUP1 protein modulates transcription by a direct DNA binding mechanism, we might expect to isolate mutants that separate the ability to bind copper and to autoregulate.
The results obtained with the various mutations described above support the copper chelation model of autoregulation. The protein truncations showed a linear response in ability to detoxify copper as well as ability to cause repression of CUPl basal transcription. The longest truncation peptide, containing 10 Cys residues, was the best able to confer copper resistance and restore autoregulation. The two shortest truncations, neither of which could confer an appreciable degree of copper resistance, showed an approximately equal inability to repress basal transcription. The chelation model of autoregulation is perhaps best supported by the Cys-to-Ser substitution mutations since these are specific for the residues known to participate in metal binding. In all cases the ability to detoxify copper correlated directly with the ability of the protein to cause repression of transcription from the CUP1 promoter.
Additional evidence for the copper chelation mechanism comes from the result with the double lysine to alanine substitution in the conserved Lys-Lys-Ser-Cys-Cys-Ser sequence. It had been previously demonstrated that monkey metallothionein can complement both copper detoxification and autoregulation in a cuplA yeast strain . Because monkey and yeast metallothionein share little sequence homology, and because the monkey and yeast metallothionein control sequences are unrelated, this result was interpreted to favor the copper chelation model. However, because the yeast and monkey genes share the conserved hexapeptide, an alternate explanation was that this sequence was responsible for DNA binding. Our demonstration that a mutant in this sequence exhibits normal regulation excludes this latter hypothesis.
A surprising result from the experiments presented here is that proteins which contain the same number of cysteine residues have different abilities to detoxify copper and to regulate transcription. It has been proposed that copper ions would bind to yeast metallothionein in a symmetrical fashion such that all cysteine residues would participate equally in a trigonal metal binding geometry. However, our results suggest that the cysteine residues may not be functionally equivalent as substitutions at the amino terminus of the molecule are less able to detoxify copper and repress basal transcription than substitutions at the carboxyl terminus of the molecule. One possibility is that the amino terminus serves as a nuclea-tion center to which copper ions bind and thereby initiate the folding of the rest of the molecule.