Specific copper transfer from the Cox17 metallochaperone to both Sco1 and Cox11 in the assembly of yeast cytochrome c oxidase

The assembly of the copper sites in cytochrome c oxidase involves a series of accessory proteins including Cox11, Cox17 and Sco1. The two mitochondrial inner membrane proteins Cox11 and Sco1 are thought to be copper donors to the Cu B and Cu A sites of cytochrome oxidase, respectively, while Cox17 is believed to be the copper donor to Sco1 within the IMS. In this report we show Cox17 is a specific copper donor to both Sco1 and Cox11. Using in vitro studies with purified proteins, we demonstrate direct copper transfer from CuCox17 to Sco1 or Cox11. The transfer is specific as no transfer occurs to heterologous proteins including bovine serum albumin and carbonic anhydrase. In addition, a C 57 Y mutant of Cox17 fails to transfer copper to Sco1, but is competent for copper transfer to Cox11. The in vitro transfer studies were corroborated by a yeast cytoplasm expression system. Soluble domains of Sco1 and Cox11, lacking the mitochondrial targeting sequence and transmembrane domains, were expressed in the yeast cytoplasm. Metallation of these domains was strictly dependent on the co-expression of Cox17. Thus, Cox17 represents a novel copper chaperone that delivers copper to two proteins.

these phenotypes are not reversed in yeast by the addition of exogenous copper. The role of Cox11 in Cu B site formation was suggested by the observation that CcO isolated from Rhodobacter sphaeroides cox11∆ cells lacked Cu B and Mg(II) but contained the other cofactors (10). Yeast lacking Cox11 are devoid of CcO activity (11). Cox11 is also an intrinsic IM protein tethered by a transmembrane helix. A fourth protein Cox19 is implicated in copper metallation of CcO by virtue of its limited sequence similarity with Cox17 (12). Yeast cells lacking Cox19 are respiratory deficient, but this phenotype is not reversed by the addition of exogenous copper salts.
The roles of Cox17, Sco1 and Cox11 in copper metallation of CcO were substantiated by the observation that each protein is a Cu(I) binding protein and that mutations that abrogate in vivo function attenuate Cu(I) binding (11,(13)(14)(15). Cox17 is a soluble proteins that localize within the IMS and also within the cytoplasm (8). The dual localization of Cox17 suggested that it served as a copper ion shuttle to the mitochondrion. However, we recently demonstrated that Cox17 is fully functional if tethered to the IM by a heterologous transmembrane domain (16).
The pathway of copper ion delivery to CcO has not been resolved. Sco1 was proposed to mediate the transfer of Cu(I) ions from Cox17 to the Cu A site in Cox2 (17). In support of this model is the observation that Sco1 and Cox2 interact (18). Cox17 has not been shown to interact with any of these proteins, so its role as a copper donor to Sco1 and/or Cox11 remains unsubstantiated. Cox19, like Cox17, is a small soluble protein within the IMS, so it may also have a copper transfer function. However, nothing is known of which step in the CcO assembly is affected by Cox19.

Materials and Methods
Yeast strains. A wild-type W303 strain (MAT a, ade2-1, his3 -1,15, leu2,3,112, trp1-1, ura3-1) and isogenic cox17∆ variant (6) were used for all experiments. Yeast strains were cultured on plates or in liquid medium in either complete medium (CM) lacking uracil for pYEF2 selection and lacking leucine for pRS315 selection. Cells were cultured with glucose, raffinose or galactose as carbon sources as specified in the text. DNA transformations were performed using a lithium acetate protocol.
Plasmids. SCO1 corresponding to codons 95-296 fused to a 5' sequence encoding a His-purification tag was amplified from pTNDW4 (15) by PCR with NotI and EcoRI restriction sites added to the 5' and 3' ends, respectively. Likewise, mutant SCO1 encoding the same segment with the mutation (C 148 A,C 152 A) was amplified from pTNDW4-148,152 with NotI and EcoRI terminal sites. The truncated SCO1 genes were subcloned into pYEF2 containing the GAL1 promoter and CYC1 terminator for transformation of W303 cells (25). The plasmid is designated YEp-GAL1-SCO1. A fusion of thioredoxin and COX11 lacking the first 300 base pairs of the ORF was PCR amplified from the pET32a vector pHCDW4 (11) and subcloned into pYEF2 creating plasmid YEp-GAL1-COX11. DNA encoding a Cox11 mutant with a triple alanine substitution of Cys111, Cys208 and Cys210 was amplified as a thioredoxin fusion from an existing mutant plasmid for subcloning into pYEF2. The construction of YCp-MET25-COX17 and YCp-MET25-SCO2/COX17 was described previously (16). Maximum expression with the MET25 promoter occurs in medium lacking methionine, whereas minimal expression occurs in medium containing 0.67 mM methionine (5x the normal Met level). The Sco2/Cox17 fusion contains the 5'104 residues of Sco2 containing the N-terminal mitochondrial import sequence and transmembrane domain of Sco2 fused to Cox17. YEp-MET25-COX17-1 was constructed by insertion of COX17-1 encoding the C 57 Y variant of Cox17 in pRS426 (26). The correct sequences of all plasmids generated in this study were confirmed by sequencing.
Protein purification. Recombinant Cox17 was purified as described previously (Heaton et al., 2001). E. coli BL21 (DE3) transformants with COX17 on a pAED3 vector were used for the Cox17 purification using conventional chromatography on an AKTA Fast Protein Liquid Chromatography unit (Amersham Pharmacia). The C 57 Y mutant variant of Cox17 was purified from cells harboring the COX17-1 mutant. Recombinant Sco1 and mutant Sco1 (C 148 A,C 152 A double substitution) were purified from BL21 (DE3) transformants harboring pTNDW2 (His-tagged SCO1 or mutant SCO1) as described previously (Nittis et al., 2001). The Sco1 molecules purified were soluble truncates lacking the N-terminal 95 residues that form the mitochondrial target sequence and transmembrane domain. Nickel-NTA Superflow (Qiagen) was used for the purification of the His-tagged Sco1 proteins. Apo-Sco1 was isolated from E. coli transformants induced in the absence of added copper to the growth medium. Under these conditions, Sco1 is recovered with less than 0.1 mol equiv. bound copper. E. coli strain BL21(DE3) harboring pHCDW4, a pET32a (Novagen) vector containing COX11 lacking the first 300 base pairs, was used for the purification of thioredoxin-Cox11 as a His-tagged fusion.
Ni-NTA (Qiagen) column chromatography was used as described previously (11). Apo-Cox11 was isolated from bacterial transformants induced in the absence of added copper to the growth medium. Yeast transformants with YEp-GAL1-SCO1 or YEp-GAL1-COX11 were cultured in raffinose medium to a OD 600nm of 0.6. Galactose was then added to induce expression of the His-tagged Sco1 or Cox11. Cells were harvested after 5 h for preparation of lysates by use of a French press and subsequent Ni-NTA purification from clarified samples. The overexpression of the soluble domain of Sco1 or Cox11 did not cause impaired mitochondrial respiration.
Copper transfer assay. For in vitro studies, CuCox17 and mutants were purified from E. coli, concentrated and stored anaerobically in 50 mM Tris pH 7.4, 100 mM NaCl, 1 mM dithiothreitol.
ApoSco1 or apoCox11 purified from E. coli were reduced in 50 mM dithiothreitol and subsequently desalted into transfer buffer (20 mM Tris-HCl pH 7.0, 100 mM NaCl). The transfer was initiated by dilution of the Cox17 into the transfer buffer followed by the addition of the apo-protein. Emissions were monitored (with a 350 nm band width filter) between 350 nm and 700 nm after excitation at 300 nm. Excitation and emission slits were set at 5 and 20 nm, respectively.
After analysis transfer mixtures were separated by either affinity purification by Ni-NTA chromatography or by size exclusion on a Superdex 75 equilibrated in 20 mM Tris-HCl pH 7.0, 100 mM NaCl. Ni-NTA chromatography was carried out by the addition of Ni-NTA resin to the protein mixture followed by a 10 min incubation at 4°C with gentle agitation. Resin was removed by centrifugation and the resulting supernatant was designated unbound. The resin was washed once in 1 volume of transfer buffer before elution in 1 volume of 20 mM Tris pH 7.0, 250 mM imidazole (designated as bound). Protein content was monitored by absorbance at 280 nm with an elution buffer blank. Copper content was measured by atomic absorption spectroscopy.
Assays. The copper concentration of the protein samples was measured using a Perkin Elmer (AAnalyst 100) atomic absorption spectrophotometer or a Perkin-Elmer Optima (3100XL) ICP spectrometer. Protein was quantified by amino acid analysis after hydrolysis in 5.7 N HCl at 110ºC in vacuo on a Beckman 6300 analyzer. Optical absorption spectroscopy was carried out on a Beckman DU640. Luminescence was measured on a Perkin Elmer LS55 fluorimeter with excitation set at 300 nm.

Results
The luminescence of CuCox17 was used initially to monitor in vitro copper transfer to acceptor proteins. The emission of the copper conformer of Cox17 near 580 nm arises from thiolate-coordinated Cu(I) ions shielded from solvent quenching (14). The target molecules tested were recombinantly expressed, soluble domains of Sco1 and Cox11 lacking their transmembrane segments. We had previously shown that the soluble domain of each protein was competent for specific Cu(I) binding, and that the Cu(I) conformer of each protein was non-luminescent (11,15).
The addition of either apo-Sco1 or apo-Cox11 to a solution of CuCox17 led to an attenuation in luminescence consistent with either quenching of the CuCox17 emission or copper ion transfer ( Fig. 1A). The attenuation was complete within a 5 min assay period. Two studies suggested that the rapid loss of emission did not arise from quenching. First, the addition of the non-luminescent CuSco1 to a solution of CuCox17 did not result in any time-dependent change in emission ( Fig.   1A). Second, the addition of either BSA or carbonic anhydrase did not attenuate luminescence (data not shown).
The attenuation of CuCox17 emission by the addition of apo-Sco1 or apo-Cox11 was suggestive of Cu(I) ion transfer to this protein. Copper transfer was confirmed, by the purification of Sco1 or Cox11 from the incubation mixture and quantitation of bound copper. The initial His-tagged variants of Sco1 and Cox11 contained only residual copper prior to incubation with CuCox17. After incubation with CuCox17 and re-purification by Ni-NTA chromatography, both Sco1 and Cox11 were recovered with bound copper ( Fig. 2A). Cox17 was not retained on the Ni-NTA column and did not stably associate with either Sco1 or Cox11. The eluates contained only Sco1 or Cox11 (Fig. 2C). Thus, the copper recovered in the eluate was associated with either Sco1 or Cox11. Residual copper in the unbound fraction was associated with Cox17.
The specificity of the transfer reactions was tested in two ways. First, carbonic anhydrase was used as a nonspecific apo-protein. Binding of Cu(I) by carbonic anhydrase results in a nonfunctional enzyme that cannot be activated by exogenous Zn(II) (27). Even after a prolonged incubation of CuCox17 with apo-carbonic anhydrase, the enzyme was fully activated by the addition of Zn(II) to the solution suggesting that Cu(I) ion transfer had not occurred (data not shown). Second, the C 57 Y mutant Cox17 encoded by the non-functional COX17-1 allele was used as a test of transfer specificity. We previously demonstrated that C 57 Y Cox17 binds Cu(I) normally, exists in homodimer/tetramer complexes and when targetted to the mitochondria was still non-functional (13,16). The addition of apo-Sco1 did not attenuate the luminescence of the C 57 Y Cox17 (Fig. 1B). In contrast, the emission of both WT and C 57 Y Cox17 molecules was attenuated by the addition of apo-Cox11 (Fig. 1B). Recovery of Sco1 after incubation with C 57 Y Cox17 did not reveal any appreciable transferred copper, whereas recovery of Cox11 after a similar incubation with C 57 Y Cox17 revealed copper ion transfer (Fig. 2B). These results suggests that the C 57 Y substitution abrogates a specific interface for Sco1.
The concentration dependency of the copper transfer reactions was determined. Titrations using a fixed CuCox17 concentration and increasing the level of the apo-proteins to excess conditions revealed that only a subset of the Cox17-bound Cu is transferred to Sco1 as monitored by the CuCox17 emission (Fig. 3A, 3B). Even after the addition of excess apo-Sco1, residual Cu(I) luminescence persisted in Cox17. Only approximately 40% of the Cox17-bound Cu(I) is transferred regardless of the excess of Sco1 used. The incomplete copper transfer from Cox17 arises from an inherent property within Cox17 and not Sco1. Doubling the Cox17 concentration doubles the amount of Sco1 that can be metallated. Concerning the metallation of Sco1, the nature of the titration curve suggests that copper transfer to Sco1 appears maximal near 1 mol eq. (Fig.   3A). This observation is corroborated by recovery of Sco1, using the affinity tag, after a transfer reaction with limiting Sco1 levels. The resulting Sco1-bound copper content was near 1 mol equiv.
In a second series of titrations using excess Sco1 and varying the CuCox17 concentration, we observed a concentration dependency for copper transfer from Cox17 at limiting Cox17 levels ( Fig.   3B). At concentrations below 10 µM Cox17, transfer to Sco1 appears impaired. One possibility is the known concentration-dependency in Cox17 oligomerization from dimers to tetramers (K D near 20 µM) (14). Dimeric Cox17 species may be less efficient than tetrameric species in the copper transfer reaction. Cox17 incubated in the presence and absence of apo-Sco1 was fractionated by gel filtration and the distribution of Cox17 oligomers was assessed by western analysis of the column fractions (Fig. 4A). In the absence of Sco1, Cox17 was distributed in species representing tetramers, dimers and monomers as was shown previously (14) (Fig. 4B). After incubation with apo-Sco1, less tetrameric Cox17 species were observed (Fig. 4B).
Similar titrations of CuCox17 to apo-Cox11 revealed that only a subset of the Cox17bound Cu was also transferred to Cox11 (Fig. 3A). Approximately 55% of the Cox17-bound Cu(I) was transferred to Cox11. Mixtures of Sco1 and Cox11 did not result in greater depletion of Cox17-bound Cu(I) (data not shown). The stoichiometry of Cu(I) bound to Cox11 appeared to be 0.5 Cu(I) per monomer. Titration of Cox11 with varying CuCox17 concentrations showed more efficient copper transfer relative to Sco1 at lower Cox17 levels (Fig. 3B). Cu binding to Cox11 appears to saturate in this assay at a copper stoichiometry near 0.5 mol equiv. This stoichiometry was confirmed by recovery of Cox11 using the affinity tag These in vitro studies suggested that Cox17 is a specific copper donor to both Sco1 and Co-expression of the mutant His-SCO1 and COX17 resulted in the recovery of Sco1 with less than 0.1 mol equiv copper bound. Cox17 did not co-purify with Sco1 in the single step affinity purification step, suggesting that the two proteins do not stably interact.
The Cox17-mediated metallation of Sco1 likely arises from direct copper transfer from Cox17. To confirm that cytoplasmic Cox17 contains bound copper, we expressed a His-tagged Cox17 gene under the GAL1 promoter on a YEp vector. Induction of COX17 expression by transferring the cultures to galactose medium for 5 h resulted in appreciable accumulation of Cox17 in the cytoplasm as shown by western analysis (Fig. 5C). Recovery of the protein by Ni-NTA chromatography revealed mean copper stoichiometries between 0.3 and 0.5 in multiple experiments.
Thus, a fraction of the Cox17 molecules are copper-loaded in cells cultured in standard growth medium.
As mentioned, in vitro copper transfer studies with apo-Sco1 revealed no apparent copper transfer from the C 57 Y mutant Cox17. To test whether this mutant was capable of copper transfer in the yeast cytoplasm, double transformants containing both YEp-GAL1-SCO1 and YEp-MET25-COX17-1 (encoding C 57 Y Cox17) vectors were cultured in raffinose followed by 5 h in galactose as mentioned above. Sco1 purified from lysates showed a copper content near 0.2 mol equiv., a value similar to cells not overexpressing Cox17 (Fig. 5A). The expression level of C 57 Y Cox17 was lower than the overexpressed Cox17 used in the above experiments (Fig. 5C). As mentioned above, reduction of the WT Cox17 levels by culturing cells in medium containing excess methionine did not appreciably lower copper metallation of Sco1, but the Cox17 protein levels were now comparable with the level of the C 57 Y mutant protein (data not shown). Thus, the C 57 Y Cox17 is an ineffective copper donor in both the in vitro and in vivo systems.
In the absence of Cox17 over-expression, the copper content of the recovered Sco1 is 0.2 as mentioned, potentially due to low levels of endogenous, cytoplasmic Cox17. We did not test the metallation of Sco1 in cox17∆ cells as they grew poorly in raffinose medium. To test whether the limited Sco1-bound copper arose from endogenous cytoplasmic Cox17, we used cox17∆ cells containing a SCO2/COX17 chimeric protein. This chimera is functional and resides solely within the mitochondrial IM (16). YEp-GAL1-SCO1 and YCp-ME25-SCO2/COX17 cotransformants were cultured in the usual conditions (raffinose followed by 5 h in galactose). Sco1 recovered from these cells had a copper content of <0.1 mol equiv. suggesting that copper ion metallation of Sco1 is dependent on the Cox17 protein under normal physiological growth conditions (Fig. 5A).
The cytosolic assay was repeated using YEp-GAL1-COX11 transformants in the presence and absence of YCp-MET25-COX17. In the absence of co-expression of Cox17, Cox11 was recovered from yeast cytosol using Ni-NTA chromatography. The purity of the recovered Cox11 was insufficient for determination of the binding stoichiometry, so the samples were further purified by gel filtration on a S200 matrix. Purified Cox11 was recovered with 0.1 mol equiv. bound copper from cells cultured in synthetic medium without any supplemental copper (Fig. 5B). Recovery of Cox11 from cells cultured in medium supplemented with 0.8 mM CuSO 4 resulted in 0.4 mol equiv. copper bound. The protein eluted from the S200 gel filtration column in a volume consistent with a dimeric species. We reported previously that Cox11 purified from E. coli forms a stable dimer (11).
Co-expression of GAL1-COX11 and MET25-COX17 chimeric genes in cells cultured without supplemental copper results in enhanced copper metallation of Cox11 (Fig. 5B). The copper content increased from 0.1 mol equiv. in the absence of Cox17 to nearly 0.5 in the presence of Cox17. This stoichiometry is 2-fold less than that observed when COX11 is expressed in E. coli cells cultured in copper-supplemented medium (11). The copper ions in the dimeric Cox11 purified from E. coli are closely spaced (11), so the repeated recovery of Cox11 from the yeast cytosol with only 0.5 mol equiv. bound copper may suggest a species exists with a single copper ion bound at the dimer interface. When a mutant Cox11 was induced in which Cys111, Cys208 and Cys210 were substituted by alanines, no appreciable copper binding was observed in the recovered protein (Fig. 5B). Thus, copper binding by Cox11 is dependent on this Cys motif. Consistent with this result, we showed previously that recombinant Cox11 purified from E. coli as a thioredoxin fusion the observed Cu(I) binding was associated with Cox11 and not the thioredoxin (11).
Whereas the C 57 Y Cox17 was an ineffective copper donor to Sco1 either in vitro or in vivo, C 57 Y Cox17 was nearly as efficient as WT Cox17 as a copper donor to Cox11 in the yeast cytosol assay system (Fig. 5B). Cox11 purified from YEp-GAL1-COX11 transformants containing YEp-MET25-COX17-1 (C 57 Y variant) contained 0.4 mol equiv bound copper confirming that C 57 Y Cox17 is an efficient copper donor to Cox11 but not Sco1.
The efficient copper transfer by the C 57 Y mutant to Cox11 but not Sco1 implies that the substitution perturbs transfer to only one target. The specificity of the mutant copper transfer is consistent with Cox17 docking with each target molecule in distinct interfaces. Cox17 did not form stable interactions with either Sco1 or Cox11 in the studies. To determine whether a weak interaction of Cox17 with either Sco1 or Cox11 existed, the purified Sco1 and Cox11 samples were tested by western analysis for co-purification of low levels of Cox17. Cox17 was not detected in either purified Sco1 or Cox11 samples. Cox17 may only transiently interact with these target molecules. In attempt to stabilize a putative interaction between Cox17 and Sco1, we co-expressed COX17 and a mutant SCO1 encoding a C 148 A, C 152 A double mutation of the two Cu(I) binding cysteine residues (15). No co-purification was observed of Cox17 and the Sco1 mutant lacking the ability to bind Cu(I) (data not shown).

Discussion
Cox17 was postulated to function as a copper donor to Sco1 as an intermediate step in the copper metallation of the Cu A in Cox2 during assembly of cytochrome c oxidase (7,17, 29). This postulate was based on the suppression of the respiratory defect of a cox17∆ strain by overexpression SCO1 (7) and the fact that Sco1 is a copper-binding protein (15,17). In the present studies we directly confirm this postulate and extend the functional significance of Cox17 to show for the first time that it is also a copper donor to Cox11. Like Sco1, Cox11, is a copper-binding protein suggesting that copper metallation of Sco1 and Cox11 is an intermediate step in the transfer of copper to the Cu A site in Cox2 and Cu B site in Cox1, respectively (11).
Two assays were used to confirm the specificity of the Cox17-mediated copper transfer to both Sco1 and Cox11. First, in vitro studies with purified proteins showed physical copper transfer from CuCox17 to either Sco1 or Cox11. Specificity of this copper transfer reaction was indicated by the absence of transfer to heterologous proteins including bovine serum albumin and carbonic anhydrase. In addition, the C 57 Y mutant of Cox17 failed to transfer copper to Sco1, but was competent for copper transfer to Cox11. Second, the in vitro transfer studies were corroborated by a yeast cytoplasm assay. Soluble domains of Sco1 and Cox11 lacking the mitochondrial targeting sequence and transmembrane domains were expressed in the yeast cytoplasm in the presence and absence of Cox17 co-expression. Recovery of either Sco1 or Cox11 revealed that copper metallation of either protein was dependent on either growth of cells in medium containing high exogenous copper ions or co-expression of Cox17. The Cox17-mediated metallation of Sco1 or Cox11 occurred in yeast cultured in non-copper supplemented medium. Residual copper binding was observed in Sco1 recovered from wild-type yeast containing chromosomally-encoded Cox17.
This residual copper metallation of Sco1 arises from the low, endogenous levels of Cox17 within the cytoplasm (8). Copper metallation of Sco1 was nearly abolished when purified from cells containing Cox17 tethered to the mitochondrial inner membrane. Specificity in the yeast cytoplasm assay was further demonstrated by a lack of copper transfer to Sco1 by the C 57 Y Cox17 mutant but copper transfer by the mutant to Cox11 in agreement with the in vitro assay.
Yeast harboring C 57 Y Cox17 are respiratory deficient and lack appreciable cytochrome c oxidase activity (6). The mutant protein fails to accumulate to a significant level within the mitochondrial IMS in the presence or absence of a heterologous IMS targeting sequence (13, 16).
The recombinant C 57 Y Cox17 protein binds Cu(I) normally and forms similar tetrameric conformers as the wild-type protein (14,16). Thus, the non-functionality of the protein does not arise from defective copper ion binding.
The present data shows that C 57 Y Cox17 is defective only in copper transfer to Sco1. The ability of the mutant protein to transfer copper to Cox11 suggests that Cox17 may have different interaction interfaces for Sco1 and Cox11. We reported recently that the C-terminal segment containing Cys57 may adopt an amphipathic helix (16). This segment of the protein may be an important interface for Sco1, but not Cox11. We were unable to detect any stable interactions between Cox17 and Sco1 or Cox11 with in vitro or yeast cytoplasm assays by evaluating co-purification, implying that its interactions with Sco1 and Cox11 may be transient. CuCox17 transfers Cu(I) ions to both Sco1 and Cox11 for subsequent donation to Cox2 and Cox1, respectively. The postulated Sco1-mediated metallation of Cox2 is likely to occur after extrusion of the Cox2 soluble domain into the IMS by Cox18 (31). This transfer may occur upon physical interaction between Sco1 and Cox2 as shown previously (18). It is unclear at this time whether both copper ions in the Cu A site are derived from CuCox17. Studies are underway to extend this yeast cytoplasm assay system as a nonnative, cellular environment to evaluate copper transfer from Sco1 to Cox2 and to test the function of Cox19.     YEp-GAL1-SCO1 or YEp-GAL1-mSCO1 encoding a Sco1 C 148 A,C 152 A mutant were cultured in raffinose to a OD 600nm of 0.6. Galactose was added to induce expression of His-tagged Sco1.

Figure Legends
Cells were harvested after 5 h and soluble lysates were fractionated by Ni-NTA chromatography.
The copper binding stoichiometry of the purified Sco1 is shown. The YEp-GAL1-SCO1 transformants were cultured in standard synthetic medium or in medium containing 0.8 mM CuSO 4 (designated +Cu). Cells co-transformed with YCp-MET25-COX17 or YEp-MET25-COX17-1 encoding C 57 Y Cox17 and cultured in synthetic medium (lacking additional copper) were