A Mutational Study in the Transmembrane Domain of Ccc2p, the Yeast Cu(I)-ATPase, Shows Different Roles for Each Cys-Pro-Cys Cysteine*

Ccc2p is homologous to the human Menkes and Wilson copper ATPases and is herein studied as a model for human copper transport. Most studies to date have sought to understand how mutations in the human Menkes or Wilson genes impair copper homeostasis and induce disease. Here we analyze whether eight conserved amino acids of the transmembrane domain are important for copper transport. Wild-type Ccc2p and variants were expressed in a ccc2-Δ yeast strain to check whether they were able to restore copper transport by complementation. Wild-type Ccc2p and variants were also expressed in Sf9 cells using baculovirus to study their enzymatic properties on membrane preparations. The latter system allowed us to measure a copper-activated ATPase activity of about 20 nmol/mg/min for the wild-type Ccc2p at 37 °C. None of the variants was as efficient as the wild type in restoring copper homeostasis. The mutation of each cysteine of the 583CPC585 motif into a serine resulted in nonfunctional proteins that could not restore copper homeostasis in yeast and had no ATPase activity. Phosphorylation by ATP was still possible with the C583S variant, although it was not possible with the C585S variant, suggesting that the cysteines of the CPC motif have a different role in copper transport. Cys583 would be necessary for copper dissociation and/or enzyme dephosphorylation and Cys585 would be necessary for ATP phosphorylation, suggesting a role in copper binding.

Copper is an essential element for living organisms because it is a cofactor for many enzymes. However, copper can also be toxic because of its redox properties and its ability to form free radicals. Therefore, all living cells from bacteria to plants and mammals have developed a fine control of their intracellular copper (1). In mammals, two P-type ATPases are responsible for delivering copper into the Golgi to newly synthesized proteins that need it as a cofactor, and also for expelling copper out of the cell, when necessary. The latter function is a response to an increase in intracellular copper and can be seen as a detoxification function when it occurs in the liver or an assimilation function when it occurs in the intestinal epithelium. In humans, the copper ATPases were named Menkes and Wilson ATPases after the related genetic diseases. The Menkes ATPase is expressed in almost all cells and mutations in its gene ATP7A result in a systemic lack of copper and severe neurological disorders; the Wilson ATPase is mainly expressed in hepatocytes and mutations in its gene ATP7B result in an overload of copper in the liver and cirrhosis (2). In yeast, there is one homologous P-type ATPase named Ccc2p, which delivers copper into the Golgi (3,4).
All these copper ATPases belong to the P-type ATPases family (5), and more precisely to the P1-type subfamily, which is specific for heavy metal transport (6). The P1-type ATPases have a membranous domain made of 8 putative transmembrane segments and in the 6th transmembrane segment (TM6), 1 a conserved CPC/H motif (see Scheme 1 for Ccc2p). The cytoplasmic domain is made of two loops, the largest one containing the catalytic site, i.e. ATP-binding and phosphorylation sites. The N terminus comprises 1 to 6 metal-binding domains, each of them containing the CXXC consensus sequence, specific for heavy metal binding. Functional aspects of heavy metal transport by P1-ATPases have been gathered by various studies on prokaryotic cadmium, zinc, or lead ATPases (Staphylococcus aureus CadA (7), Escherichia coli ZntA (8,9), Listeria monocytogenes CadA (10)), prokaryotic copper ATPases (Enterococcus hirae CopA and CopB (11,12), E. coli (13) and Archeoglobus fulgidus CopA (14)), and eukaryotic P1-type ATPases (for a review on mammalian copper ATPases, see Ref. 15), the latter all being copper ATPases except in plants (16) (see Axelsen's data base). 2 All these studies suggest that despite their N-terminal domain, the P1-type ATPases follow the same catalytic cycle as that of the P2-type ATPases (Scheme 2). For instance, the prokaryotic cadmium, zinc, or lead ATPases ZntA (18) and CadA (10) were found to be active after truncation of their N-terminal domain, suggesting that the N-terminal domain does not have an essential role in the ionic transport mechanism across the enzyme.
The copper ATPases translocate copper using the energy of ATP and are phosphorylated during their catalytic cycle (11, 13, 14, 19 -21). As depicted in Scheme 2, the ATPase is phos-phorylated by ATP (step 2) after copper has bound to its membrane transport site(s) (step 1), whereas in the absence of copper, it cannot be phosphorylated by ATP, but rather by P i (step 4) (22). The ability of the metal to induce phosphorylation by ATP is a specific property of P-type ATPases, which allows to check whether or not the transport sites are filled with the metal to be transported. This property has been used to identify amino acids belonging to the transport sites of the Ca 2ϩ -ATPase (23).
Thus far there have not been many studies of the copper ATPase transmembrane domain, except for the conserved CPC motif of the Wilson ATPase (24,25), the Menkes ATPase (26), the Caenorhabditis elegans copper ATPase (27), E. coli CopA (28), and E. hirae CopB (29) and a few other mutations found in patients (25,30,31) or in the toxic milk mouse (32)(33)(34)(35). Assuming that among the 8 predicted transmembrane segments, some amino acids are arranged so as to coordinate copper, as described for calcium in the Ca 2ϩ -ATPase three-dimensional structure (36), we have modified some of the highly conserved amino acids that could potentially form or participate at the copper transport site(s) and have evaluated the consequences of these modifications.
In this report, the functional consequences of changing 8 conserved amino acids in TM3, TM4, TM5, and TM6, among them each cysteine of the CPC motif, were studied both in vivo by complementation assays in a ccc2-⌬ yeast strain (i.e. a yeast strain in which the CCC2 gene was disrupted), and in vitro by measuring Ccc2p ATPase activity and phosphorylation by ATP on membrane preparations from Sf9 cells, after overexpression. We show here that (i) overexpressed Ccc2p in Sf9 cells has a measurable Cu(I)-dependent activity of 20 nmol/mg/min at 37°C; (ii) all our mutations affected copper homeostasis in yeast, although to different extents; (iii) the mutation of each 583 CPC 585 cysteine impaired yeast growth as efficiently as the mutation of the phosphorylation site; (iv) each CPC cysteine is necessary to ATPase activity and (v) only Cys 585 is necessary to ATP phosphorylation. Therefore, modification of the second cysteine in the CPC motif impairs ATPase activity because it impairs ATP phosphorylation, whereas modification of Cys 583 , the first cysteine, does not impair ATP phosphorylation. Thus, modification of Cys 583 impairs other steps in the cycle. In other words, these two cysteines do not play the same role in the catalytic cycle.

EXPERIMENTAL PROCEDURES
Wild-type and Mutated CCC2 Gene-The genomic DNA from S. cerevisiae strain 288c was used to amplify the full-length CCC2 gene. The PCR product (3,015 bp) was subcloned into the pSP72 plasmid (Promega) using the StuI/XbaI sites provided by the PCR primers. The resulting plasmid was designated pSP72CCC2 and the DNA sequence was confirmed by dideoxynucleotide sequencing (Genome Express). The CCC2 gene was then mutated to generate the desired codon changes corresponding to the following mutations: D337A, D374A, T375A, T541A, C583S, C585S, T591A, T593A, and the non-phosphorylating D627A mutant. The pSP72CCC2 was used as template to produce all the mutants by the QuikChange TM site-directed mutagenesis kit (Stratagene). The mutated fragments were sequenced and subcloned into the CCC2 gene in replacement of the wild-type fragment in the pSP72CCC2 vector. Each mutant was further used and subcloned in the 2 plasmid, using the EcoRI/AvrII sites. This plasmid, which contains the LEU2 selection factor and the PMA1 promoter, is denoted YEp181PALEH or YEp and used for expression in S. cerevisiae. The mutants were also subcloned in the pFastBac1 plasmid in the EcoRI/PstI sites for expression using the baculovirus/Sf9 system (Invitrogen).
CCC2 Gene Disruption-The chromosomal CCC2 gene was disrupted by modifying the pSP72CCC2 plasmid with an insertion of the URA3 gene into the CCC2 coding sequence. This was done by excising with BamHI the URA3 gene from the YDpU vector (37) and inserting the 1154-bp fragment into the unique BamHI site of CCC2. The 2549-bp SpeI/HindIII fragment containing the URA3 gene flanked at the 5Ј and 3Ј ends by 584 and 811 bp of CCC2 was used to transform the W303-1A S. cerevisiae strain. Yeast transformants were selected on Drop Out medium without uracil and their genomic DNA was prepared according to Ref. 38 for further checking by Southern blotting.
Expression in Yeast and Complementation Assay-The transformed yeast strains were selected on Drop Out medium without uracil or leucine. They were grown on a yeast nitrogen base (YNB, without Cu,Fe, Q-BIOgene) supplemented with amino acids, except for leucine, and 1 M CuSO 4 , 50 M NH 4 (FeSO 4 ) 2 ⅐6H 2 O, and 1 mM ferrozine for 5 days at 30°C (25). Protein expression in yeast was assessed by immunodetection using anti-Ccc2GB, a polyclonal rabbit antibody raised against the following peptide: 660 CIKATESISDHPVSKAIIRY. Anti-Ccc2GB was used at 1:5000, and the protein was visualized using a chemiluminescence kit (Roche Applied Science). The same growth medium as described above was used for complementation in the presence of high copper (500 M CuSO 4 ) or in the presence of high iron (350 M NH 4 (FeSO 4 ) 2 ⅐6H 2 O).
Expression in Sf9 Cells and Membrane Fraction Preparation-Cultures of 120 ml (2 ϫ 10 6 cells/ml) were infected with recombinant baculovirus containing the wild-type or mutant CCC2 genes. Infection with the wild-type baculovirus was taken as control. After 3 days, the membrane fraction was prepared as described before (39) except for eliminating CaCl 2 and adding 5 mM dithiothreitol. At the end, the membranes were suspended in 2 ml of storage buffer (100 mM MOPS-KOH, pH 7.0, 40% glycerol, 5 mM dithiothreitol), frozen in liquid nitrogen, and stored at Ϫ80°C for further use. Protein concentration (ϳ20 mg/ml) was determined using the DC protein assay (Bio-Rad). Protein expression was assessed by SDS-PAGE and quantified by densitometric analysis of the Coomassie Blue-stained gel (TotalLab, Amersham Biosciences). Ccc2p was also detected using anti-Ccc2GB.
Phosphorylation by [␥-32 P]ATP-50 g of membrane fraction were suspended in 200 l of 20 mM bis-Tris propane (pH 6.0), 200 mM KCl, 5 mM MgCl 2 , according to Ref. 21. Unless otherwise specified, no copper was added to this medium. Radioactive [␥-32 P]ATP was added to a final concentration of 5 M (ϳ1500 cpm/pmol), and the reaction mixture was incubated on ice for various periods of time as indicated in the figure legends. The reaction was stopped by addition of 50 l of ice-cold 1 mM NaH 2 PO 4 in 50% trichloroacetic acid. Samples were centrifuged for 20 min at 18,000 ϫ g and 4°C. The protein pellet was washed once with ice-cold water and resuspended in 40 l as described in Ref. 21 before loading onto an acidic polyacrylamide gel (40). The gel was exposed to SCHEME 1. Schematic representation of Ccc2p, the yeast Cu(I)-ATPase. On the cytoplasmic side, the N-terminal domain contains two consensus sequences CXXC, here CSAC and CGSC, which bind Cu(I). The large loop contains the P-type ATPase consensus sequence DKTGT, where Asp 627 is the phosphorylated amino acid. Other consensus sequences are also indicated. White letters indicate the amino acids that were modified in this work into alanine or serine. SCHEME 2. Ccc2p catalytic cycle based on that of the Ca 2؉ -ATPase. E denotes Ccc2p; and E-P, a phosphoenzyme; cyt, cytoplasmic; lum, Golgi lumen; n, unknown number of transport sites.
PhosphorScreen and analyzed on a PhosphoImager (Amersham Biosciences) to measure the intensity of the phosphorylated bands. Isotopic dilution and ADP sensitivity experiments were carried out by adding non-radioactive ATP or ADP to the sample at various times after the reaction with 5 M [␥-32 P]ATP had started. The subsequent steps were the same as those described above. The effect of hydroxylamine was investigated on samples precipitated in trichloroacetic acid after phosphorylation. After the first centrifugation and washing in ice-cold water, a second centrifugation was followed by incubation for 30 min in 1.1 M hydroxylamine (pH 5.5) (41). The samples were then prepared for the acidic gel as above.
Phosphoenzyme Quantification-Each acidic gel was stained with Coomassie Blue to evaluate the relative amounts of loaded proteins in each lane and normalize [␥-32 P] incorporation to the total loaded proteins. The protein band densitometry was analyzed by the Scion Image program. 3 This procedure allowed to quantify phosphoenzyme levels within the same gel, as shown for Ccc2p WT phosphorylation kinetics in Figs. 4 and 5. Each experiment was repeated 3-6 times with different membrane preparations and the maximal value on each gel, i.e. the Ccc2p WT phosphoenzyme level measured at 1 min was taken as 100%.
The next step in the quantification was to find some means to compare the phosphorylation kinetics of different variants of Ccc2p. This was achieved by further using the 100% reference defined above for Ccc2p WT. Therefore, for each variant studied, all samples were analyzed on the same gel together with a sample of Ccc2p WT phosphorylated for 1 min.
ATPase Activity-ATPase activity was measured according to the method described by Grubmeyer and Penefsky (42) with some modifications for adaptation to our system (41). Unless otherwise specified, the assay buffer was 20 mM MOPS (pH 6.0), 100 mM KCl, 5 mM MgCl 2 , 100 M ascorbate, 0.2% SDS-␤-D-maltoside, 4% glycerol, and 10 mM NaF, the latter being necessary to inhibit Sf9 cell phosphatase activity. Note that no copper was added to this buffer. Membrane fractions (0.25 mg/ml) were first incubated for 30 min on ice in 20 mM MOPS (pH 6.0), 100 mM KCl, 5 mM MgCl 2 , 100 M ascorbate, 10 mM NaF, and 20 M thapsigargin to inhibit Sf9 cell Ca 2ϩ -ATPases. The reaction was started by adding 50 l of the membrane suspension to 200 l of the assay buffer supplemented with 1 mM [␥-32 P]ATP and kept at 37°C; it was stopped after various periods of time by the addition of HCl-activated charcoal (0.1 M). The 32 P i released was measured in an aliquot of the supernatant obtained after centrifugation of the charcoal suspension for 25 min at 4°C and 18,000 ϫ g. Spontaneous hydrolysis of [␥-32 P]ATP was measured simultaneously by mixing the membranes and the acid before adding [␥-32 P]ATP. P i release increased linearly during the first 10 min; therefore, measurements were generally stopped after 5 min.
Inhibition of ATPase Activity by Bathocuproine Disulfonate (BCS) and/or Bicinchoninic Acid (BCA) and Reactivation by Copper-Membrane fractions (0.25 mg/ml) were incubated on ice with copper chelators in the following medium: 20 mM MOPS (pH 6.0), 100 mM KCl, 5 mM MgCl 2 , 10 mM NaF, 20 M thapsigargin, 250 M BCS, and 100 M BCA. After 30 min, the membrane suspension was spun down at 18,000 ϫ g for 15 min at 4°C to remove the chelators, and the pellet was suspended in the assay buffer 20 mM MOPS (pH 6.0), 100 mM KCl, 5 mM MgCl 2 , 0.2% SDS-␤-D-maltoside, 4% glycerol, and 10 mM NaF for ATPase activity measurement in the presence of 1 mM [␥-32 P]ATP at 37°C. ATPase activity was measured 5 min later as described above. ATPase reactivation was obtained by incubating the pellet after the removal of the chelators for 30 min on ice in 20 mM MOPS (pH 6.0), 100 mM KCl, 5 mM MgCl 2 , 10 mM NaF, 0.2% SDS-␤-D-maltoside, 4% glycerol, 100 M ascorbate, and 10 M CuSO 4 . It was then equilibrated at 37°C and 1 mM [␥-32 P]ATP was added to start the reaction assay, which was stopped as described above.

RESULTS
The Choice of the Amino Acids-To determine which amino acids may participate in the coordination of copper in the transmembrane domain, we first selected the acidic and polar residues of this domain, which are conserved in most of the Cu ϩ -ATPases. Among them, we chose those that are identical in at least 10 of the 12 eukaryotic Cu ϩ -ATPases known today. Table I shows the predicted transmembrane segments TM3 to TM6 of the eukaryotic Cu ϩ -ATPases. 2 The amino acids that were mutated in this work are in bold; they were replaced by alanines, except for the CPC motif in which the cysteines were changed to serines. The Cys to Ser substitution was chosen because it eliminates the thiol that is a preferential ligand for copper and also because a simple sulfur to oxygen change is hoped to have little effect on the structure of the protein.
Complementation of ⌬CCC2 (ccc2-⌬ Yeast Strain) by Different CCC2 Variants-The modified proteins were studied in vivo using a complementation assay that detects whether or not copper homeostasis is impaired. In yeast, lack of Ccc2p impairs copper incorporation into the multicopper oxidase Fet3p and, therefore, the iron high affinity import complex 3 Available at www.scioncorp.com.  (43). Such a phenotype was evidenced in a medium deprived from copper and iron that allows the parental wild-type strain W303-1A to grow but not ⌬CCC2 (Fig. 1A). Supplementing this medium with iron restored the growth of both strains that can take iron up through a lowaffinity carrier (Fig. 1B); such recovery was also observed by supplementing the medium with copper (Fig. 1C). Therefore, in these experiments, the ⌬CCC2 strain behaved as expected (43), namely it could only grow in iron-or copper-rich media. This ⌬CCC2 strain was further used for complementation studies based on the observation that the re-introduction of the CCC2 gene on a multicopy expression vector allowed recovery of the wild-type phenotype.
The results of complementation experiments are shown in Fig. 1D. Each strain is a ⌬CCC2 strain transformed with YEp, a multicopy expression vector containing the wild-type or a mutated CCC2 gene, as described under "Experimental Procedures." The cells transformed with the empty YEp had the ⌬CCC2 phenotype, whereas the cells had the wild-type phenotype when they expressed Ccc2p WT (compare Fig. 1, A and the first two lines in D). D627A is a ⌬CCC2 strain expressing a modified Ccc2p that is unable to be phosphorylated by ATP because Asp 627 , which is phosphorylated by ATP during enzymatic cycling has been changed to an Ala (see in Scheme 1, the DKTGT consensus sequence). D627A is therefore taken as a reference for non-functioning Ccc2p. All other strains express a modified Ccc2p, which has one point mutation in the transmembrane domain, as explained above.
As expected, expression of D627A does not allow the growth of ⌬CCC2 in copper-and iron-limited medium; neither does the expression of C583S or C585S, the two proteins in which one of the cysteines of the TM6 CPC motif has been modified to a serine (Fig. 1D). Expression of all the other proteins was able to restore yeast growth, but only partially as none of them reached the level of Ccc2p (Fig. 1D). Copper homeostasis was therefore modified by every single mutation tested here, but to various extents. For each mutation, such modification can be due to different reasons: low expression level, mis-localization of Ccc2p, or incorrect insertion into the Golgi membrane. On the other hand, if we assume that Ccc2p is correctly expressed, folded, and addressed to the Golgi membrane, then the mutation is probably impairing protein activity. The first assumption was checked by immunodetection using anti-Ccc2GB, which showed that all variant proteins were expressed although with some variations in the expression levels ( Fig. 2A). Nevertheless, these variations did not correlate the observed impairment in yeast growth. For instance, both C583S and C585S completely impaired yeast growth, although C585S was expressed in higher amounts and C583S in lower amounts than Ccc2p. Mutation-induced inactivation of Ccc2p was therefore checked by further studying Ccc2p enzymatic cycle.
Wild-type and Modified Ccc2p Expression in Sf9 Cells-Modified or wild-type Ccc2p expression levels in the transformed ⌬CCC2 yeast strains did not allow functional analysis by ATPphosphorylation assays, given the background due to phosphorylation of yeast native proteins. Therefore, modified and wildtype Ccc2p were expressed in Sf9 cells via baculovirus, and their expression was checked by SDS-PAGE of membrane extracts (Fig. 2B). Overexpression of wild-type Ccc2p is clearly visible when compared with the membrane extract from cells infected with the wild-type baculovirus (C 0 ). As expected, the additional band migrated at the same level as sarcoplasmic reticulum Ca 2ϩ -ATPase 1a (110 kDa, not shown). In all Sf9 membrane preparations, Ccc2p, wild-type as well as modified, accounted for about 7% of the total membrane proteins, except for the C583S protein, which accounted for less than 1% (Fig. 2,  B and C). Thus, the expression of the C583S protein was low in both yeasts and Sf9 cells. A single band was immunodetected by anti-Ccc2GB in membrane preparations from Sf9 cells expressing Ccc2p or the modified proteins (Fig. 2C).
The Wild-type Ccc2p Expressed in Sf9 Cells-The membrane fraction containing the wild-type Ccc2p was used to evaluate Ccc2p ATPase activity and phosphoenzyme formation from ATP. ATPase activity was measured as 32 P i release from [␥-32 P]ATP, as described under "Experimental Procedures." The total activity displayed by C 0 was 96 Ϯ 4 nmol/mg/min, compared with 116 Ϯ 7 nmol/mg/min when the cells had expressed Ccc2p. This specific copper ATPase activity became dominating in the presence of thapsigargin, a sarcoplasmic reticulum Ca 2ϩ -ATPase specific inhibitor (44) and sodium flu- In all four panels, the first column corresponds to 6 ϫ 10 4 cells and the next to 10and 100-fold dilutions, respectively. oride, a phosphatase inhibitor (45). In their presence the activities were 11 Ϯ 1 and 29 Ϯ 1 nmol/mg/min for C 0 and Ccc2p, respectively. Ccc2p ATPase activity was also found to be sensitive to the copper concentration as shown in Fig. 3A. Taking the value measured without added copper as reference, there was a 10% increase with 1 M and 25% inhibition with 10 M added copper, confirming previous observations on purified CopA and CopB (12,46). Furthermore, copper chelation should also inhibit Ccc2p ATPase activity. BCS and BCA, two Cu(I) chelators, were used as described under "Experimental Procedures." As shown in Fig. 3B, we found that the mixture of 250 M BCS and 100 M BCA was the most efficient in inhibiting ATPase activity, which was inhibited to about 35% of the control and recovered up to 80% of its control value once the chelators were removed (see Fig. 3A in gray). The recovery was higher in the presence of 10 M copper than without added copper, because after removal of the chelators their concentrations were not negligible as compared with contaminating copper. These experiments therefore confirmed the sensitivity of Ccc2p ATPase activity to Cu(I). Thus, Ccc2p seems to behave like a P-type ATPase (Scheme 2) because it is activated by low copper concentrations (step 1) and inhibited by high copper concentrations (step 3).
Therefore, another way to observe whether or not the transport sites have bound copper is to measure the phosphoenzyme formation from ATP (step 2 in Scheme 2). C 0 and membrane fractions containing D627A or Ccc2p were mixed with [␥-32 P]ATP as described under "Experimental Procedures," the reaction was stopped 15, 45, or 60 s later, and samples were loaded onto acidic gels. C 0 and D627A membranes displayed a stable and small phosphorylation signal at 110 kDa, whereas Ccc2p displayed a strong and transient signal (Fig. 4A). Ccc2p phosphoenzyme was sensitive to hydroxylamine, which cleaves acylphosphate bonds (Fig. 4B). The sensitivity of the phosphoenzyme to the copper chelators and its recovery in the presence of copper were also checked, as shown in Fig. 4B. The chelators inhibited 70% of the phosphorylation by ATP reaching about the same level as D627A, and the recovery was up to 80%, as found for the Wilson disease ATPase (21).
To characterize Ccc2p further, phosphoenzyme formation was followed for 10 min and Ccc2p was found to have a maximum labeling at 60 s and to dephosphorylate within the next 3 min because of ATP consumption (Fig. 5, graph and gel 1). Such a decay is not only due to Ccc2p but also to other proteins present in the Sf9 membrane preparations that also consume ATP. Another way to observe Ccc2p cycle is to add an excess of unlabeled ATP and follow the dephosphorylation rate of the labeled phosphoenzyme during Ccc2p cycling. This was done after 1 min phosphorylation with 5 M [␥-32 P]ATP by diluting the radioactive ATP into 50 M non-radioactive ATP (Fig. 5,  graph and gel 2). Ccc2p label disappeared within the first min, faster than without isotopic dilution, confirming the transient nature of the phosphorylated intermediate. Phosphoenzyme measurements on C 0 and D627A show the background level (Fig. 5, graph and gels).
The Ccc2p Variants-The Ccc2p variants were phosphorylated for various times to compare their kinetics to those of Ccc2p and D627A during the first minute, the time necessary to reach the maximum level of phosphoenzyme with Ccc2p (Fig.  6A). Longer time scales up to 10 min was explored to evaluate the contribution of the dephosphorylation in each Ccc2p variant (Fig. 6B).
For each Ccc2p variant, the results in Fig. 6A show the mean of three to six experiments performed on different membrane preparations. Phosphorylation kinetics of Ccc2p and D627A during the first minute are shown as bold lines in Fig. 6A. Apart from C585S, which was not phosphorylated at all, all modified proteins were phosphorylated, although to various extents between those of Ccc2p and D627A.
The phosphorylation studies on the long time scale show that all the variant phosphoenzymes had a slower decay than the wild-type phosphoenzyme. The various phosphoenzyme kinetics were simply analyzed as the sum of an exponential rise and an exponential decay, with no attempt to attribute any rate constant to the true phosphorylation or dephosphorylation steps. Given such a simple description, the variants can be separated into three groups. The first group includes D337A, D374A, T375A, T541A, and T593A, which all reached their higher phosphoenzyme level at 1 min, as did Ccc2p, the second group includes those which reached that level later, namely T591A and C583S, and the third group includes C585S, which could not be phosphorylated at all, as the non-phosphorylating D627A. For the sake of simplicity, only one variant from each group is shown in Fig. 6B, together with Ccc2p.
In the first group, the maximum phosphoenzyme levels varied from 40 to 90% as compared with the 100% maximum level of Ccc2p, which was chosen as the reference (see "Experimental Procedures"). As we found that the expression levels in Sf9 cells were about 7% of the total membrane proteins for all the first group variants (data not shown), differences in the expression level cannot account for these variations, unless the proportion of active protein among these 7% is different. The same remark holds for the complementation studies, which showed for instance that D374A was more efficient than T375A in restoring yeast growth (Fig. 1D), although there was more T375A expressed than D374A ( Fig. 2A). Therefore, differences in the protein functionality as those measured in Fig. 6, A and B, may be responsible for the different capacities to restore yeast growth.
In the second group, T591A and C583S were both phosphorylated to a higher extent than Ccc2p: 120% was reached within 3 min for T591A and C583S phosphorylation stabilized after 10 min (see below). This can be seen in Fig. 6C, which shows 32 P labeling of Ccc2p after 1 min and C583S after 1 and 10 min phosphorylation. C583S labeling after 10 min was clearly higher than that of Ccc2p after 1 min, although C583S was less expressed than Ccc2p. C583S labeling after 10 min was sensitive to hydroxylamine. Incidentally, the fact that C583S was still phosphorylated after 20 min indicates that there was still some ATP available for phosphorylation after all that time, despite the probable existence of other ATP-consuming proteins in Sf9 membranes. A simple explanation for the high phosphorylation level is that copper dissociation and/or enzyme dephosphorylation (steps 3 and 4 in Scheme 2) are slowed in C583S, the slow decay of the phosphoenzyme suggesting a slow turnover, and therefore impairment of copper transport and ATP consumption. In the third group, C585S was not at all phosphorylated although it was highly expressed in both systems (Fig. 2).  1, 4, and 7), 45 s (lanes 2, 5, and 8), and 60 s (lanes 3, 6, and 9). B, inhibition by hydroxylamine (hx) and copper chelators (BCS ϩ BCA), and reactivation from the chelatorinduced inhibition by addition of 10 M CuSO 4 . Phosphorylation was measured after 1 min and quantified as described under "Experimental Procedures." The CPC Variants-The failure of the two CPC-modified proteins to restore copper transport in yeast (Fig. 1D) was confirmed biochemically by measuring their ATPase activities that were about the same as the residual activity measured with the non-phosphorylating D627A protein or with C 0 (Fig.  7). As both CPC-modified proteins were inactive, although C583S was phosphorylated, another series of experiments were performed to analyze which step was modified by the mutation. First, the ability of C583S to cycle was evaluated by isotopic dilution after 1 min of phosphorylation by 5 M [␥-32 P]ATP, as done for Ccc2p in Fig. 5. Fig. 8A shows that after 1 min phosphorylation, the response of C583S to isotopic dilution was the same as that of D627A and can therefore be attributed to the background sensitivity to isotopic dilution and not to C583S cycling. The background response was slow when compared with that of Ccc2p. Then, the isotopic dilution experiment was repeated on C583S and D627A after 10 min phosphorylation by 5 M [␥-32 P]ATP to ensure full phosphorylation of the C583S variant and evaluate the background (Fig. 8B). Isotopic dilution induced a slow decrease in the labeling of D627A with a similar rate to the one measured after 1 min phosphorylation (Fig. 8A), indicating that although the background phosphorylation increased during 10 min from 42 to 65%, its response to isotopic dilution was not changed. For the C583S variant, the kinetics were biphasic. A fast phase was observed that had about the same rate constant and amplitude as that of the background. The rest of the labeling was stable and attributed to the C583S phosphoenzyme, which did not cycle.
Finally, ADP sensitivity of the phosphoenzymes was measured to determine whether there was any copper bound to the phosphoenzymes. According to the P-type enzymatic cycle, a copper-bound phosphoenzyme in the so-called energy-rich conformation (Cu n E ϳ P in Scheme 2) will react with ADP to synthesize ATP and consequently dephosphorylate very quickly, as opposed to the enzyme cycling rate. These experiments were performed as for the isotopic dilution except that 10 mM ADP was added instead of ATP and the reaction was stopped after shorter times. Dephosphorylation was completed within the first 10 s, as shown for Ccc2p and variants that all reached the same final level of 20% (Fig. 9). For D627A and C583S, when ADP was added after 1 min phosphorylation, the amount of phosphoenzyme that was sensitive to ADP was the same as the amount that dephosphorylated under ATP isotopic dilution (Fig. 8A), but the reaction was much faster. Therefore, the background phosphorylation was sensitive to ADP. The same remark holds for Ccc2p, suggesting that under our conditions, all the phosphoenzyme was in the Cu n E ϳ P form. Finally, when ADP was added to C583S after 10 min phosphorylation, the effect was not only much faster but also much higher than the effect of isotopic dilution, as even the stable phosphoenzyme is dephosphorylated upon ADP addition. Thus, the C583S phosphoenzyme was ADP-sensitive, as was Ccc2p.

DISCUSSION
The aim of this work was to study amino acids of the Ccc2p transmembrane domain, chosen because of their high conservation among the eukaryotic copper ATPases (Table I). We did both complementation studies to check the effect of Ccc2p modifications on yeast growth under copper-and iron-limiting conditions, and biochemical studies on membrane preparations after heterologous expression in Sf9 cells to verify whether or not Ccc2p functioning was affected by each of these modifications. Our results show that among the selected amino acids, the most important were the cysteines of the CPC motif that were found to play different roles in Ccc2p functioning.
Ccc2p ATPase Activity-For the purpose of the biochemical studies, we made an Sf9 cell membrane preparation containing active Ccc2p that allowed us to measure a Cu(I)-dependent ATPase activity of 20 nmol/mg/min at 37°C. It was inhibited by Cu(I) chelators and high Cu(I) concentrations (Fig. 3A), confirming what was shown for a purified prokaryotic copper ATPase (46). These two properties agree with the assumption that the Ccc2p catalytic cycle follows Scheme 2. According to the P-type ATPases catalytic cycle, E should have a high affinity for cytoplasmic copper, whereas E-P should have a low affinity for copper, allowing dissociation in the Golgi lumen.
The ATPase activity measured here is similar to that reported for a mouse copper ATPase in liver microsomes (1.5 mol/mg/h (47)), the only report to date of any eukaryotic copper ATPase activity. This value of 20 nmol/mg/min, together with the fact that Ccc2p accounts for 7% of total protein in the membrane preparation (Fig. 2B), and M r 110,000 gives a turnover rate of 0.5/s. This is a slow turnover when compared with that reached by P2-type ATPases, for instance, 20/s for native sarcoplasmic reticulum Ca 2ϩ -ATPase at 28°C (39) or 14/s for native gastric H ϩ ,K ϩ -ATPase at 25°C (48), and even 50 -100/s for the Na ϩ ,K ϩ -ATPase expressed in yeast at 37°C (49). We do not know now whether this slowness is an intrinsic property of Cu ϩ -ATPases, as suggested by the 3/s turnover estimated from purified A. fulgidus CopA (14) or due to the fact that we assumed all Cu ϩ -ATPases expressed in Sf9 cells to be active.
The D337A, D374A, T375A, T541A, T591A, and T593A Variants-Although each of the following amino acids, D337, D374, T375, T541, T591, and T593, are present in all or almost all eukaryotic Cu ϩ -ATPases, their replacement by an alanine had no major functional consequences. Still, none of them was as efficient as the wild-type Ccc2p in restoring copper homeostasis (Fig. 1D). Biochemical studies showed that they could all be phosphorylated by ATP and displayed a slower phosphoenzyme decay than that of Ccc2p (Fig. 6, A and B). Some mutations of the sarcoplasmic reticulum Ca 2ϩ -ATPase, which were first shown to have mild effects on Ca 2ϩ transport, as for instance, the N768A mutation (50), are now known to belong to the Ca 2ϩ transport site (36). In addition, small incidences induced by mutation are expected if the peptidic backbone and not the lateral chain of the concerned amino acid is involved in ion binding. Therefore, the variants studied here may be of interest and further studies are needed to analyze at which step of the Cu ϩ -ATPase cycle (Scheme 2) these amino acids are important. The D374A mutation, for instance, was expected to have an important effect because Asp 374 is in 11 eukaryotic Cu ϩ -AT-Pases (Table I) and is replaced by a Glu in the 32 prokaryotic Cu ϩ -ATPases having a CPC motif. 2 In addition, a point mutation in ATP7B changing Asp to an Asn was found to induce Wilson disease (51).
The CPC Motif Variants-The CPC motif is highly conserved among the P1-type ATPases, especially the Cu ϩ -ATPases, except for some histidine-rich prokaryotic ATPases such as E. hirae CopB, which have a CPH motif instead (5,52). 2 The proline of the CPC motif is also conserved in P2-type ATPases and has been assigned an important role in the Ca 2ϩ -ATPase, as it untwists the M4 helix in the region where one of the two Ca 2ϩ binds (36). In addition, point mutations changing the CPC motif to RPC in ATP7A and to CPY in ATP7B were found to induce Menkes (53) and Wilson diseases (54). All these findings suggest that the CPC motif is important for copper transport.
A few articles report on in vivo studies of the CPC motif variants of different copper ATPases. The CAC variants of the Wilson and Menkes ATPases, as well as the SPS variant of the Wilson ATPase were not able to restore the growth of a ⌬CCC2 yeast strain (24 -26). C. elegans Cua-1 failed in restoring yeast growth when its CPC motif was modified to CPA (27); E. hirae CopB failed in restoring the cell resistance to copper, when its CPH motif was modified to SPH and the variant protein had a low phosphoenzyme level even after 8 min and no ATPase activity (29); E. coli CopA also failed in restoring resistance to copper, when its CPC motif was changed to CPA, CPH, or APC, and the variant proteins were not phosphorylated (28). Our biochemical studies showed that none of the two CPC cysteine variants displayed any ATPase activity (Fig. 7), and that their phosphorylation by ATP was either not possible in the case of the CPS (C585S) protein, or very slow in the case of the SPC (C583S) protein (Fig. 6, A and B). These results explain why both variants were as inefficient as D627A in restoring yeast growth. In addition, they also suggest that Cys 583 and Cys 585 play different roles in Ccc2p enzymatic cycle (Scheme 2). Interestingly, some of these Cys to Ser modifications in the CPC motif can be found in native P1-type ATPases. According to the protein data banks, the SPC motif can be found in a few Cd 2ϩ /Zn 2ϩ /Pb 2ϩ /Co 2ϩ -ATPases but not in copper ATPases. The CPS motif is only found in one copper ATPase from Helicobacter pylori (17), however, in this ATPase, the CPS motif is immediately followed by a cysteine, as if the presence of a second cysteine were compulsory in native copper ATPases.
The replacement of the Cys by Ser was chosen because the substitution of an oxygen to the sulfur of the Cys should reduce the affinity for copper without inducing many changes in the structure. Thus, one possibility is that the C585S mutation has reduced the affinity for copper to such an extent that under our conditions, there was no copper bound to the CPS protein and consequently no phosphoenzyme formed from ATP. Conversely, the fact that the SPC protein was phosphorylated by ATP and sensitive to ADP (Figs. 6, 8, and 9) suggests that the affinity of the non-phosphorylated enzyme for copper was not changed by the C583S mutation (or that the change was not detectable in the presence of contaminant copper) or that Cys 583 is not essential for copper binding. From the results shown in Figs. 6, 8, and 9, we can assume that the second cysteine (Cys 585 ) is essential for copper binding at the transport site(s) (step 1 in Scheme 2) and that the first cysteine (Cys 583 ) is essential for copper dissociation from the phosphoenzyme and/or dephosphorylation (steps 3 and 4 in Scheme 2). Therefore, the two cysteines in the CPC motif are not equivalent.
The results shown here were mainly obtained in the presence of contaminant copper, therefore we can expect more information varying the copper concentration. One of the major difficulties with this comes from working with copper that seems to be occluded in Ccc2p. For instance, lengthy incubation with copper chelators is necessary to inhibit Ccc2p, a difficulty that was also encountered with the Wilson ATPase and discussed by Lutsenko and co-workers (21). Further work on the transmembrane domain of Ccc2p should enable us to determine which amino acids contribute to the copper transport site(s) and the number of transport sites, because these characteristics are still unknown for P1-type ATPases.