Cox11p Is Required for Stable Formation of the CuBand Magnesium Centers of Cytochrome c Oxidase*

Assembly of the core subunits of theaa 3-type cytochrome c oxidase in mitochondria and aerobic bacteria such as Rhodobacter sphaeroides requires the association of three subunits and the formation of five to seven metal centers. Several assembly proteins are required for the late stages of oxidase assembly in eukaryotes; some of these are also present in Rb. sphaeroides. To investigate the role of one of these proteins, Cox11p, the mitochondrial-like oxidase of Rb. sphaeroides was overexpressed and purified from cells that lacked cox11, the gene for Cox11p. The oxidase that assembled in the absence of Cox11p lacked CuBat the active site and contained greatly reduced amounts of metal at the magnesium/manganese-binding site between subunits I and II. This inactive oxidase, however, did contain hemes a anda 3, CuA, and all three subunits. These results indicate that Cox11p is required at a late, perhaps final, step in the assembly of cytochrome oxidase, most likely the insertion of CuB. Oxidase which assembled in a strain with a low copy number of cox11 appeared nearly wild type, suggesting that Cox11p is required in substoichiometric amounts for its role in oxidase assembly.

Assembly of the core subunits of the aa 3 -type cytochrome c oxidase in mitochondria and aerobic bacteria such as Rhodobacter sphaeroides requires the association of three subunits and the formation of five to seven metal centers. Several assembly proteins are required for the late stages of oxidase assembly in eukaryotes; some of these are also present in Rb. sphaeroides. To investigate the role of one of these proteins, Cox11p, the mitochondrial-like oxidase of Rb. sphaeroides was overexpressed and purified from cells that lacked cox11, the gene for Cox11p. The oxidase that assembled in the absence of Cox11p lacked Cu B at the active site and contained greatly reduced amounts of metal at the magnesium/manganese-binding site between subunits I and II. This inactive oxidase, however, did contain hemes a and a 3 , Cu A , and all three subunits. These results indicate that Cox11p is required at a late, perhaps final, step in the assembly of cytochrome oxidase, most likely the insertion of Cu B . Oxidase which assembled in a strain with a low copy number of cox11 appeared nearly wild type, suggesting that Cox11p is required in substoichiometric amounts for its role in oxidase assembly.
Cytochrome c oxidase, the terminal electron acceptor of the respiratory chain in mitochondria and many aerobic bacteria, is a membrane-bound protein complex. Two mitochondrial-like cytochrome c oxidases are found in the cytoplasmic membranes of Rhodobacter sphaeroides and Paracoccus denitrificans (1,2). The structure of the P. denitrificans oxidase is nearly identical to that of the three-subunit catalytic core of the mitochondrial oxidase (3)(4)(5)(6).
Each bacterial oxidase monomer contains two heme A molecules, three copper atoms, and one magnesium atom (7,8). Two copper atoms form the Cu A site in subunit II that initially accepts an electron from reduced cytochrome c (9,10). The electron is transferred sequentially to heme a and then to the heme a 3 -Cu B active site where molecular oxygen is reduced to water (11). Both heme centers are located in subunit I. Heme a has two histidine ligands, whereas heme a 3 is bound by a single histidine residue, leaving the sixth coordination position avail-able for binding O 2 . Cu B is ligated by three histidines and is located 4.5-5.2 Å from the iron of heme a 3 (3,4,6,8,12). A non-redox active magnesium center is present at the interface of subunits I and II (4,8,13). At least two residues from subunit I and two from subunit II are necessary for coordination of this metal (8,(12)(13)(14). Manganese can replace magnesium at this site without altering oxygen reduction activity (13).
The assembly of the multisubunit oxidase with its metal centers apparently occurs in an ordered sequence; most of the details, however, remain unresolved (15)(16)(17). In Saccharomyces cerevisiae, genetic analysis indicates that more than a dozen genes are involved in holoenzyme assembly (18,19). Orthologs of some of these genes are found in bacteria, suggesting that they are required for assembly of the core subunits and the metal centers (20,21). Two such genes, cox10 and cox11, are present in the Rb. sphaeroides operon that encodes subunits II and III of the oxidase (22). The importance of cox10 and cox11 for cytochrome oxidase assembly is underscored by the finding that orthologs of these genes are present in the human genome (23,24).
In S. cerevisiae, cox10 and cox11 mutants lack an optically detectable cytochrome c oxidase even though all of the structural subunits are synthesized (25,26). The product of cox10 (Cox10p) has farnesyltransferase activity and catalyzes the conversion of protoheme to heme O, a precursor of heme A (27,28). Because heme A is required for function and optical detection of the aa 3 -type oxidase, the phenotype of the yeast cox10 mutant is readily understood. Based on the similar phenotypes of the cox10 and cox11 mutants in yeast, it has been suggested that Cox11p also functions in heme A biosynthesis (25,26), specifically in the formylation of the heme (29).
Cox11p is not found in isolated cytochrome oxidase, and its sequence (22) does not reveal any homology to proteins of known function. Sequence analysis predicts that Cox11p of Rb. sphaeroides is a 20.7-kDa protein with a single transmembrane helix and a large globular domain (22). Immunological studies in yeast have shown that Cox11p is localized to the mitochondrial inner membrane (26). Our investigation into the function of Cox11p has revealed an essential role for this protein in the formation of the Cu B center. In addition, Cox11p was necessary for proper alignment of hemes a and a 3 and stability of the magnesium/manganese center. The presence of Cox11p was not required for heme A biosynthesis, heme A insertion, Cu A assembly, and association of the subunits of the oxidase. Experiments altering the copy number of cox11 suggested that Cox11p was required in only substoichiometric amounts relative to the structural subunits.
Bacterial Growth and Oxidase Purification-Rb. sphaeroides strains were grown in Sistrom's media (30) as 500-ml cultures in 2-liter baffled flasks with vigorous shaking at 32°C and supplemented with 1 g/ml tetracycline and 50 g/ml streptomycin and spectinomycin as necessary. To obtain cytochrome oxidase with manganese bound at the magnesium center, the media were supplemented with 0.7 mM MnSO 4 (13). Cells were harvested at late exponential phase (optical density at 660 nm of 1.0 -1.6) and stored at Ϫ80°C. Cytoplasmic membranes were prepared as described previously (1) and the oxidases were purified on Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) as in Zhen et al. (31) with the following exceptions. Dodecyl maltoside was added to a final concentration of 3% to solubilize the purified membranes, and the Ni 2ϩ -nitrilotriacetic acid-agarose slurry was added to the membrane solution at 1 ml per 1 mg oxidase. The purified proteins were frozen in liquid N 2 and stored at Ϫ80°C.
Activity Assays-Oxygen reduction assays were performed at 25°C in a mixture of 2 pmol of wild-type or Cox11 oxidase or 40 -60 pmol ⌬Cox11 oxidase, 50 mM potassium phosphate, pH 6.5, 0.5 mM EDTA, 0.1% dodecyl maltoside, 3 mM ascorbate, 0.6 mM tetramethyl-p-phenylenediamine, and 1 mg/ml phosphatidylcholine. The reactions were initiated by the addition of 0.5-100 M horse heart cytochrome c. Oxygen consumption rates were measured using a Hansatech DW1 oxygen electrode unit, a CB1-D3 control box, and Minirec software (Hansatech) on an IBM-compatible computer. The data were imported into Origin 5.0 (Microcal) for subtraction of the slow rate of non-enzymatic reduction of O 2 by ascorbate and calculation of initial slopes. Turnover numbers were calculated as described by . V max values were obtained by fitting plots of turnover numbers versus cytochrome c concentration to a hyperbolic function.
Optical Spectroscopy, Heme Analysis, and Metal Analysis-Optical spectra were recorded at 25°C using a Hitachi U-3000 UV-visible spectrophotometer. Oxidase samples were reduced using solid sodium dithionite in 50 mM KH 2 PO 4 , pH 7.2, 1 mM EDTA, 0.1% dodecyl maltoside, and the concentrations were determined using the extinction values of Vanneste (33). The heme content of the oxidases was determined optically as in Hosler et al. (1) except that oxidase concentrations were determined from the absorbance at 280 nm using a derived extinction coefficient (⑀ 280 -312 nm ϭ 358 mM Ϫ1 cm -1 ). Oxidase samples were prepared for metal content analysis by washing the proteins extensively in 10 mM Tris-HCl, pH 8.0, 40 mM KCl, 25 mM EDTA and then reducing the EDTA concentration to 1 mM in a 50-kDa Ultrafree-4 filtration device (Millipore). Samples were concentrated to 25 M, based on their ␣-band absorbance (peak around 605 nm), and shipped to the University of Georgia Chemical Analysis Laboratory for inductively coupled plasma atomic emission spectroscopy.
Plasmids-High level expression of wild-type mitochondrial-like cytochrome c oxidase of Rb. sphaeroides was driven by the expression vector pRKpAH1H32. This plasmid was a derivative of the broad host range vector pRK415-1 (34) and is similar to pRK-pYJ123H (31), except that the coxI and coxII/III operons are transcribed in opposite directions in pRKpAH1H32. The EcoRI-HindIII fragment of pRKpAH1H32 that contained both operons was inserted into pUC19 to create a smaller plasmid termed pAH1H32.
An in-frame deletion in cox11 removed the codons for Thr-38 through Arg-139. Two MluI sites were introduced into pAH1H32 by site-directed mutagenesis (QuickChange, Stratagene) to create pLAJ198. With numbering from Ade of the initiation codon ATG of Rb. sphaeroides cox11 (22) the five nucleotide mutations were G114A, G115C, T118G, G414A, and G417C. The mutations were confirmed by dideoxy DNA sequencing (Sequenase, Amersham Pharmacia Biotech). The 300-base pair MluI fragment was then deleted by standard techniques (35) to create pLAJ200. The EcoRI-HindIII fragment of pLAJ200 was inserted into pRK415-1 to create pRKpLAJ200, which was identical to pRKpAH1H32 except for the cox11 deletion.
A control plasmid was used to test for effects on expression due to mutagenesis. The EcoRI-HindIII fragment of pLAJ198 was cloned into pRK415-1 to create pRKpLAJ198. This plasmid retained cox11 with two amino acid alterations (G39R and F40V) generated by the creation of the upstream MluI site. After conjugation (36) into Rb. sphaeroides strain YZ200, which lacks the genomic copy of the operon containing cox11 (31), it was found that pRKpLAJ198 directed high level expression of normal cytochrome aa 3 , similar to the overexpression vector pRKpAH1H32. Thus, the mutagenesis procedures themselves did not appear to have affected normal synthesis of cytochrome oxidase.

RESULTS
Expression and Purification of Oxidase Forms-In order to examine the role of Cox11p in oxidase assembly, the mitochondrial-like oxidase of Rb. sphaeroides was overexpressed in the presence of different copy numbers of cox11. A derivative of pRKpAH1H32, a plasmid containing both cox operons with their natural promoters, was prepared that contained a 300base pair in-frame deletion in cox11. This ⌬cox11 plasmid, termed pRKpLAJ200, was conjugated into wild-type Rb. sphaeroides 2.4.1 (37) and YZ200, a strain in which the genomic copy of the coxII/III operon has been deleted (31), for overproduction and purification of cytochrome c oxidase. The enzyme isolated from 2.4.1 (wild type) cells harboring the ⌬cox11 plasmid was termed Cox11 (for "low cox11") because it was assembled in a strain with a single copy of cox11 (in the 2.4.1 genome) and about five plasmid-borne copies of coxI, coxII, cox10, and coxIII. Correspondingly, ⌬Cox11 oxidase was isolated from cells with no copies of cox11 by introducing the same ⌬cox11 plasmid into YZ200 (⌬coxII/III). The purified wild-type oxidase used in this study was prepared from strain YZ300, which is deletion strain YZ200 containing pRK-pYJ123H (31).
The level of expression of ⌬Cox11 and Cox11 was compared with that of the wild-type oxidase (Fig. 1). Because ⌬Cox11 and Cox11 result from expression of the same plasmid (pRK-pLAJ200) in two different host strains (YZ200 and 2.4.1), expression of the wild-type oxidase from the corresponding plasmid, pRKpAH1H32, was measured in both strains. Compared with the expression of the wild-type oxidase in the same host strain, Cox11 accumulated to 100% (Fig. 1, 3rd and 4th columns), whereas ⌬Cox11 accumulated to approximately 50% (Fig. 1, 1st and 2nd columns). When cox11 is deleted in yeast, no cytochrome oxidase can be detected by optical spectroscopy (26). Thus, a mutation that causes a catastrophic phenotype in yeast had a less severe defect in Rb. sphaeroides. The reduction in oxidase expression observed in the absence of cox11 was completely reversed by a single copy of cox11 (Fig. 1), suggesting that Cox11p was not required in stoichiometric amounts with the structural subunits for oxidase assembly.
Spectral and Heme Analysis-The absolute reduced spectrum of purified ⌬Cox11, but not of Cox11, revealed a 3-4 nm blue shift in both the ␣ (602 versus 605 nm for wild type) and Soret (441 versus 444 nm) peaks (Fig. 2). Because the ␣ band is primarily due to absorption by heme a, whereas the Soret band includes roughly equal contributions from both hemes a and a 3 (33,38), the observed shifts indicated altered environments around both heme centers in ⌬Cox11. Indeed, resonance Raman analysis (not shown) revealed a disturbed heme a 3 environment in ⌬Cox11, similar to mutant oxidases containing alterations around heme a 3 (39). In contrast, there was no evidence for disturbance of the heme centers in Cox11 by optical or resonance Raman spectroscopy.
The pyridine hemochrome spectrum of ⌬Cox11 was indistinguishable from that of heme extracted from the purified wildtype or Cox11 oxidases (Fig. 3), indicating that ⌬Cox11 contained bona fide heme A. This argued against a role for Cox11p in heme A synthesis. Furthermore, the heme A content of ⌬Cox11 was nearly the same as that of the wild-type oxidase (Table I), indicating that Cox11p was not required for insertion of heme A during oxidase assembly.
Subunit Association-SDS-polyacrylamide gel electrophoresis analysis of purified ⌬Cox11 and Cox11 revealed that both protein complexes contained all three of the structural subunits (Fig. 4). This showed that Cox11p was not essential for association of the oxidase subunits.
O 2 Reduction Activity-The O 2 reduction activity of cytochrome c oxidase purified from strains with varying amounts of cox11 is listed in Table I. The ⌬Cox11 oxidase was unable to reduce oxygen efficiently. Measurable activity (80 s Ϫ1 ) using nearly saturating amounts of cytochrome c (90 M) was observed in only one of three ⌬Cox11 preparations. The high activity of Cox11 (Table I) indicated that substoichiometric amounts of Cox11p were sufficient for assembly of an active oxidase. This was consistent with the high expression of Cox11 (Fig. 1).
EPR Spectroscopy and Metal Content-Because both of the heme centers were present in ⌬Cox11, the loss of activity suggested that one of the other metal centers was disrupted. Electron paramagnetic resonance spectroscopy (EPR) is a sensitive method for examining the environments of Cu A and heme a of oxidized cytochrome oxidase, and it is also used to detect the absence of Cu B (40). The heme a 3 and Cu B centers are normally EPR-silent due to spin coupling of the metals. However, in the absence of Cu B a strong signal for five-coordinate, high spin heme a 3 is seen at approximately g ϭ 6 (41). The ⌬Cox11 spectrum (Fig. 5) showed a signal for high spin heme a 3 at g ϭ 6 with an amplitude similar to oxidases known to lack Cu B (41). This strongly suggested that ⌬Cox11 lacked Cu B . In contrast, signals in the g ϭ 2 region of the EPR spectrum showed that ⌬Cox11 contained wild-type amounts of a normal Cu A center.
The extent to which Cu B was lost from ⌬Cox11 was determined by inductively coupled plasma atomic absorption spectroscopy. The wild-type and Cox11 oxidases gave Cu:Fe values of about 1.5 (Table I) as expected for an enzyme with three coppers (Cu B and the di-copper Cu A center) and two hemes (7). The Cu:Fe value of ⌬Cox11 was 0.99 (Table I), suggesting complete loss of Cu B . As predicted from the heme A content, iron levels were normal in all three of the oxidases (data not shown).
In addition to loss of Cu B , the EPR spectra showed significant disturbance of the heme a environment of ⌬Cox11 and, to a lesser extent, of Cox11. The wild-type heme a signal at g ϭ 2.83 (g z ) was replaced by a broad signal around g ϭ 3 in ⌬Cox11 (Fig. 5). Based on previous analysis of the heme a signals of the mitochondrial-like oxidase of Rb. sphaeroides, this change indicated disruption of the hydrogen bonds to the histidine ligands of heme a in ⌬Cox11 (1). The blue-shifted optical spec-trum of ⌬Cox11 (Fig. 2) was consistent with this interpretation. The loss of strong hydrogen bonds to the heme a ligands will cause the electron density on the heme iron to decrease, thus forcing the electronic transitions of heme a in ⌬Cox11 to occur at higher energy (see Ref. 1). The heme a signals of ⌬Cox11 (g z ϭ 2.98 and g y ϭ 2.30) were also diminished in intensity due to incomplete oxidation of heme a in resting ⌬Cox11 (data not shown).
In order to examine the magnesium/manganese center of cytochrome oxidase, ⌬Cox11 and Cox11 were prepared from bacteria grown in media supplemented with manganese salts. This allows substitution of manganese, which is EPR-visible, for magnesium at its binding site between subunits I and II (8,12,13,42). Comparison of the EPR spectra of ⌬Cox11 and Cox11 (Fig. 6) showed that much less manganese was inserted into the oxidase in the absence of Cox11p. EPR spectra of acid-extracted Mn 2ϩ (not shown) indicated that ⌬Cox11 bound approximately 15% of the manganese of Cox11. Similarly, metal analysis showed that reduced levels of magnesium (29 Ϯ 11% of the normal oxidase) were incorporated into ⌬Cox11 purified from cells grown in normal media. FIG. 1. Accumulation of ⌬Cox11, Cox11, and wild-type oxidases in the membrane of Rb. sphaeroides. The relative expression level of the three oxidase forms was determined from reduced (dithionite) minus oxidized (ferricyanide) spectra of dodecyl maltoside-solubilized membranes. The wild-type (WT) oxidase was expressed from pRKpAH1H32, whereas ⌬Cox11 and Cox11 were expressed from pRK-pLAJ200. The host strains were YZ200 (⌬coxII/III; 1st and 2nd columns) and 2.4.1 (wild type; 3rd and 4th columns). The aa 3 /b value is the ratio of the amplitude of the cytochrome aa 3 peak (605-650 nm) and the amplitude of the cytochrome b peak (560 -574 nm). Data from three (1st column) or four (2nd to 4th column) independent cell growths were averaged. Error bars show standard deviation.

DISCUSSION
Previous work in yeast identified Cox11p as one of several proteins required for the post-translational stages of the assembly of cytochrome c oxidase (26,29). Here, we have further defined the role of Cox11p by characterizing the oxidase that was synthesized in its absence. Due to the presence of an alternative cytochrome c oxidase in Rb. sphaeroides (43,44), it is possible to grow strains without the mitochondrial-like oxi-dase; this permits the study of inactive oxidases. Cells in which most of the cox11 gene was deleted produced a partially assembled oxidase, termed ⌬Cox11, that contained all three structural subunits, both hemes a and a 3 , and the Cu A center. Contrary to a previous proposal based on experiments in yeast (26,29), the presence of the heme A centers in ⌬Cox11 argued against a role for bacterial Cox11p in heme A synthesis. Roles for Cox11p in the association of the oxidase subunits, insertion of heme A, and the formation of the Cu A center were also eliminated. The principal defects identified in ⌬Cox11 were diminished accumulation in the membrane, the absence of Cu B , a 70 -85% decrease in magnesium/manganese content, and altered heme environments.
The complete absence of Cu B indicated that Cox11p normally functions in the insertion of this metal. In ⌬Cox11, the magnesium/manganese center was incompletely formed, and perturbation of the heme centers was observed. Therefore, the insertion of Cu B is likely to facilitate the stable formation of other structures in subunit I. This seems reasonable because the metal centers of subunit I are close together. For example, Cu B and the magnesium atom are only 13.5 Å apart, and imidazole   side chains of ligands from each of these centers come within 7 Å (6,8).
Cox11p may be a copper chaperone specific for the Cu B center of cytochrome oxidase. Specific copper chaperones are required for the assembly of metalloprotein copper centers. For example, CopY, a repressor in Enterococcus hirae, accepts copper from CopZ but not from other copper chaperones (45). Sequence analysis of Cox11p orthologues from several species reveals a conserved CFCF motif that resembles the metalbinding regions of other metal chaperones (45)(46)(47)(48). However, a role for Cox11p in formation of Cu B does not require it to bind copper. It is also possible that by interacting with subunit I, Cox11p permits a conformation receptive for copper delivery by a different protein. The final position of the Cu B center is not solvent-accessible and is within the transmembrane region of subunit I (3,4,6,8). Thus, the involvement of an integral membrane protein such as Cox11p in the assembly of this site was not unexpected.
The level to which ⌬Cox11 accumulated in the bacterial membrane was one-half that of the normal oxidase that assembled in the presence of Cox11p (Fig. 1). An oxidase lacking Cu B would be predicted to have reduced stability at the heme a 3 -Cu B center. This, in turn, could lead to more rapid degradation of the oxidase resulting in a lower steady-state concentration in the membrane. From a different perspective, it is significant that ⌬Cox11 did accumulate to relatively high levels. This raises the intriguing possibility that ⌬Cox11 may be a normal, relatively stable assembly intermediate. Thus, the insertion of Cu B may be a late step in oxidase assembly, following heme A insertion and the association of subunit I with subunits II and III.
In yeast, the deletion of cox11 eliminates the accumulation of optically detectable cytochrome oxidase in the mitochondrial membrane (26). The different results seen in yeast and Rb. sphaeroides are likely a consequence of the complexity of the mitochondrial oxidase. In the mitochondria, association of several nuclear-encoded accessory subunits follows assembly of the catalytic core (17). Despite the near structural identity of the simpler bacterial oxidases and the mitochondrially encoded core subunits of the eukaryotic oxidase, the latter enzyme is not structurally stable in the absence of the accessory subunits (49). In this respect, bacteria are likely to be of greater utility in elucidating the mechanisms of the assembly of the catalytic core of cytochrome oxidase.
In summary, the results presented here suggest that ⌬Cox11 has a primary defect in Cu B insertion, with secondary effects being the loss of metal from the magnesium/manganese center and misalignment of the hemes. We propose a role for Cox11p in the insertion of Cu B into the aa 3 -type cytochrome c oxidase at a late stage in assembly.