COX16 Encodes a Novel Protein Required for the Assembly of Cytochrome Oxidase in Saccharomyces cerevisiae *

We have characterized Cox16p, a new cytochrome oxidase (COX) assembly factor. This protein is encoded byCOX16, corresponding to the previously uncharacterized open reading frame YJL003w of the yeast genome.COX16 was identified in studies of COX-deficient mutants previously assigned to complementation group G22 of a collection of yeast pet mutants. To determine its location, Cox16p was tagged with a Myc epitope at the C terminus. The fusion protein, when expressed from a low-copy plasmid, complements the mutant and is detected solely in mitochondria. Cox16p-myc is an integral component of the mitochondrial inner membrane, with its C terminus exposed to the intermembrane space. Cox16 homologues are found in both the human and murine genomes, although human COX16 does not complement the yeast mutant. Cox16p does not appear to be involved in maturation of subunit 2, copper recruitment, or heme A biosynthesis. Cox16p is thus a new protein in the growing family of eukaryotic COX assembly factors for which there are as yet no specific functions known. Like other recently described nuclear gene products involved in expression of cytochrome oxidase, COX16 is a candidate for screening in inherited human COX deficiencies.

Cytochrome oxidase (COX), 1 the terminal enzyme of the mitochondrial respiratory chain, catalyzes the transfer of electrons derived from sugars, fats, and amino acids to molecular oxygen. The mammalian enzyme consists of 13 subunits and has a well characterized structure (1). The three largest subunits, which are encoded in mitochondrial DNA, form the catalytic core of the enzyme, with subunits 1 and 2 binding the prosthetic groups required for electron transfer. There are two heme A molecules, both of which are located in the hydrophobic interior of subunit 1 and, based on their spectral properties, are denoted hemes a and a 3 . In addition, there are two copper atoms that form the binuclear Cu A site in subunit 2 and a single copper atom (designated Cu B ) located adjacent to the heme a 3 site in subunit 1 (2). The nuclear genome contributes the remaining 10 subunits, which are thought to play a primarily structural role.
In contrast to the wealth of information about the structure of cytochrome oxidase, the mechanism by which the holoenzyme complex is assembled remains unclear. Studies in the yeast Saccharomyces cerevisiae have provided much of the information regarding the proteins involved in the COX assembly pathway. Some of these proteins are involved in recruiting copper to mitochondria and COX (3)(4)(5). Others have been implicated in heme A biosynthesis (6,7) and transport and maturation of the mitochondrially encoded subunits (8 -10). Despite these advances, the precise functions of a number of COX assembly factors remain unknown (11)(12)(13). It is also unclear whether all of the proteins required for COX assembly have been identified. This information is important for understanding the mechanism of COX assembly and for elucidating the genetic basis of human COX deficiencies (14,15).
Here we report the identification and characterization of Cox16p, a new COX assembly factor encoded by a previously uncharacterized yeast open reading frame (YJL003w; Gen-Bank TM accession number Z49278). Mutations in COX16 result in a failure to complete assembly of cytochrome oxidase, and cox16 mutants are respiration-deficient. Cox16p does not appear to function in mitochondrial copper homeostasis or in the synthesis of heme A. The phenotype of cox16 mutants also argues against a role of Cox16p in processing of Cox2p or membrane insertion of the mitochondrially encoded subunits of COX. Like many of the other COX assembly factors, Cox16p is an integral component of the mitochondrial inner membrane and appears to exist in a higher molecular weight complex. Recent additions to the data base reveal that COX16 has homologues in Schizosaccharomyces pombe, as well as in the murine and human genomes. Human COX16 is thus a candidate gene for human COX deficiencies in which mutations in the known COX assembly factors have not been identified.

MATERIALS AND METHODS
Strains and Media-The genotypes of the strains used in this study are indicated in Table I. The composition of media for growth and analysis of yeast strains has been described elsewhere (16).
Cloning and Disruption of COX16 -The yeast genomic library used to clone COX16 was constructed from partial Sau3A fragments of nuclear DNA of the respiratory-competent haploid strain S. cerevisiae D273-10B/A1. C25/U1, a mutant from complementation group G22 (8), was transformed with the yeast genomic library as described previously (11), and the COX16 gene identified by isolating subclones capable of conferring respiratory competence on C25/U1. The COX16 gene was disrupted by insertion of a 1.1-kb URA3 fragment at the internal HindIII site and transformation of the respiratory-competent haploid strain, W303-1A, with the linear fragment containing the disrupted gene and flanking sequences. The transformation yielded the cox16 mutant aW303⌬COX16.
Construction of pMGL5 and the COX16-myc Fusion-To overexpress a Cox16p-myc fusion protein, we first constructed pMGL5, a yeast/ Escherichia coli shuttle plasmid containing a Myc epitope tag in the backbone of the multicopy vector YEp351 (17). Briefly, a 3.3-kb Nar1-AatII fragment of YEp351 containing the LEU2 marker and the 2-m origin of replication, was used to replace a 3.8-kb Nar1-AatII fragment in YCpmyc111 (a gift from Dr. Troy Harkness, modified from Gietz and Sugino (18)). This converted the CEN plasmid into an episomal plasmid. To make the Myc fusion, COX16 was amplified by PCR, using primers that introduced a PstI site 216 nucleotides upstream of the start codon (5Ј-gttattagactgcagatacacttcc-3Ј) and a SmaI site two nucleotides upstream of the termination codon (5Ј-cgttttgaatgttcccgggcattc-3Ј). The 416-bp PCR product was ligated to pMGL5, resulting in an in-frame fusion of Cox16p to the Myc epitope at the C terminus. The same COX16 fragment was also fused to the Myc sequence in the CEN plasmid YCpmyc111. Both constructs were verified to be in-frame by automated sequencing (LiCor, Lincoln, NE).
Miscellaneous Methods-Transformation of yeast was carried out by the method of Schiestl and Gietz (19). Yeast strains were grown to stationary phase in YPGal, and mitochondria were isolated as described previously (20). For analysis of in vivo labeled mitochondrial translation products, strains were grown in YPGal and labeled with [ 35 S]methionine in the presence of cycloheximide (21). The labeling reaction was terminated after 15 min by addition of an excess of 80 mM cold methionine and 80 g/ml puromycin (0 time). Samples of the cultures were incubated at 30°C and collected at 15 and 30 min, pelleted, and resuspended in 75 l of a mixture containing 1.8 M NaOH, 1 M ␤-mercaptoethanol, and 0.01 M phenylmethylsulfonyl fluoride. Proteins were precipitated by addition of an equal volume of 50% trichloroacetic acid. The mixture was centrifuged, and the pellet washed once with 0.5 M Tris, twice with water, and resuspended in 50 l of sample buffer (22). Labeled mitochondrial proteins were separated on a 12.5% polyacrylamide gel containing 6 M urea and 6% glycerol, and the gel was dried prior to autoradiography.
Mitochondrial preparations used for the inner membrane localization were isolated as described by Glick (23). Density gradient centrifugation was carried out using mitochondria (3 mg) from aW303⌬COX16/myc1, which were extracted with 1% deoxycholate and 0.5 M NaCl and loaded onto 2.4 ml of 7-20% sucrose gradients (11). Lactate dehydrogenase (0.3 mg) was added to the extract as an internal standard, along with 2.5 mg hemoglobin. Gradients were chilled at 4°C for 1 h before addition of the extract and, following loading of the samples, centrifuged in a Beckman Optima TLX tabletop ultracentrifuge (Beckman Coulter, Fullerton, CA) for 12 h at 54,000 rpm. Protein concentrations were determined by the method of Lowry et al. (24). Cytochrome oxidase activity and mitochondrial cytochrome spectral analyses were carried out as described by Tzagoloff et al. (25). Mitochondrial proteins were separated by SDS-PAGE with 12% gels (22), 15% gels with glycerol (26), or 16.5% gels containing 6 M urea (27). Western analysis was carried out as described previously using antibodies against subunits of cytochrome oxidase or Sco1p (28) or commercial antibodies to the Myc epitope tag (Sigma). Immunoblots were visualized using enhanced chemiluminescence (Amersham Biosciences).

RESULTS
cox16 Mutants Are Defective in Cytochrome Oxidase Assembly-S. cerevisiae strains C25 and E699 of complementation group G22 of a pet mutant collection (8) are respiration-deficient due to a specific loss of cytochrome oxidase. Spectral analysis of mitochondria isolated from C25 and E699 show a partial or complete loss of the 605-nm cytochrome aa 3 peak, respectively (Fig. 1A). Since the respiratory defect is complemented by 0 mutants, the cytochrome oxidase lesion stems from recessive mutations in a nuclear gene, which has been designated as COX16.
As in other COX assembly-defective mutants (20), the steady state concentrations of subunits 1, 2, and 3 are very low in the cox16 mutant (Fig. 1B). With the exception of subunit 5, which is slightly reduced, the other nuclear-encoded subunits are present at normal concentrations. In vivo labeling of C25 in the presence of cycloheximide to block cytoplasmic protein synthesis, indicated that the three mitochondrially encoded subunits of cytochrome oxidase are correctly expressed (Fig. 1C), although labeling of Cox2p appeared to be slightly less than in wild type. To assess if Cox16p participates in export of the Cox2p precursor, the size and stability of this mitochondrial gene product was compared in different COX null mutants sharing the same nuclear background. The results of the in vivo pulse-chase labeling experiment confirmed that only processed Cox2p is detected in the cox16 null mutant aW303⌬COX16 after the 15-min pulse (Fig. 2). Since proteolytic cleavage of the presequence requires export of the N-terminal transmembrane domain to the intermembrane space (29), these results exclude Cox16p from having a function in membrane insertion of the N-terminal domain of the precursor. The pulse-chase results also make it unlikely that Cox16p is involved in membrane insertion of the C-terminal domain of Cox2p. By comparison, the turnover of Cox2p in a cox18 mutant was greatly increased compared with the wild type or other COX mutants (e.g. cox15). Cox18p is required for the export of the C-terminal domain of Cox2p (30), and the rapid degradation of the protein probably occurs as a consequence of exposure of the C terminus to proteolytic enzymes of the inner membrane and/or matrix. In contrast, Cox2p turnover was similar in the cox15 and cox16 mutants during the 30-min chase (Fig. 2). Since Cox15p functions in heme A synthesis (31), the turnover rate of Cox2p in cox15 mutants is unlikely to be due to a problem with membrane insertion, and the decreased amount of Cox2p in cox16 and cox15 mutants is likely related to the general block in COX assembly. The reduction of newly synthesized Cox1p observed in the three mutants is a hallmark of most COX assembly mutants examined to date. 2 Analysis of mitochondrial hemes in both the C25/U1 and cox16 null mutant revealed the presence of heme A and heme O (data not shown) in amounts seen with most other COX assembly mutants (7), suggesting that Cox16p is not involved in heme A biosynthesis. Finally, the respiratory deficiency of C25/U1 and the cox16 null mutant were not rescued by supplementation of the growth medium with copper, calcium, iron, magnesium, manganese, or zinc (data not shown). This result does not, however, exclude Cox16p from having a function in 2 A. Barrientos   transport or insertion of one of the heavy metals known to be associated with COX.
Cloning and Disruption of COX16 -To identify the gene responsible for the cytochrome oxidase deficiency of G22 mutants, C25/U1 was transformed with a yeast genomic library, and uracil prototrophic clones were checked for growth on non-fermentable carbon sources. A transformant, C25/U1/T1, rescued both the uracil auxotrophy and respiratory deficiency and was used to obtain the recombinant plasmid pG22/T1, which contained an insert of genomic DNA ϳ7 kb in length. This plasmid restored respiration when back transformed into the mutant. Subcloning generated the construct pG22/ST2, which conferred respiratory competence to C25/U1, and sequencing of this construct identified the gene responsible for the complementation. Comparison of the putative COX16 open reading frame to the current databases revealed that the COX16 gene is identical to open reading frame YJL003w on chromosome X (GenBank accession number Z49278) in the yeast genome.
COX16 was disrupted to ascertain that the respiratory deficiency of G22 mutants is caused by mutations in this gene. The mutant allele was obtained by inserting a 1-kb HindIII fragment containing the URA3 gene at an internal HindIII site of COX16 in pG22/ST2, yielding pG22/ST4. The respiratory-competent haploid strain W303-1A was transformed with a linear BglII-SphI fragment of pG22/ST4. An uracil-independent and respiration-deficient transformant (aW303⌬COX16) was determined by Southern blot analysis of genomic DNA to have the URA3 insertion in COX16 (not shown). The biochemical phenotype of aW303⌬COX16 was ascertained to be similar to that FIG. 3. The COX16 gene encodes a small, acidic protein. A, a hydropathy plot of Cox16p predicts a single membrane-spanning segment. B, the primary amino acid sequence of Cox16p is shown with the potential mitochondrial targeting sequence (as determined by the P-SORT program) denoted by the bar above the sequence. The predicted transmembrane domain is indicated by the bar underneath the sequence.

FIG. 1. Spectral analysis and steady state levels of COX subunits in cox16 mutants.
A, spectra of mitochondrial cytochromes. Mitochondria were prepared from the two wild type strains D273-10B/A1 (D273) and W303-1A (W303) and from the cox16 mutants C25, E699, and aW303⌬COX16 (⌬COX16) and extracted at a protein concentration of 5 mg/ml with potassium deoxycholate (11). Difference spectra of the oxidized (potassium ferricyanide) versus reduced (sodium dithionite) extracts were recorded at room temperature. The position of cytochrome aa 3 is marked. B, Western analysis of COX subunits. Mitochondria were isolated from the wild type W303-1A (W303), the mutant aW303⌬COX16 (⌬COX16), aW303⌬COX16/myc1 (myc1) and aW303⌬COX16/myc2 (myc2), which are the cox16 mutant transformed with the COX16-myc fusion in a high-copy and a CEN plasmid, respectively. Subunits 1, 2, and 3 of cytochrome oxidase (Cox1, Cox2, Cox3) were analyzed by separating 10 g of mitochondrial protein on 12% polyacrylamide gels. The nuclear-encoded subunits (Cox4; Cox5; Cox6; Cox7, 7a, 8) were detected by separating 20 g of mitochondrial protein on a 16.5% polyacrylamide/6 M urea gel. Following transfer to nitrocellulose, the blots were probed with subunit-specific antibodies and visualized using enhanced chemiluminescence. C, the wild type strains D273-10B/A1 and the mutant C25 were grown in YPGal and labeled in vivo with [ 35 S]methionine in the presence of cycloheximide (46). Mitochondria were isolated and separated by SDS-PAGE on a 7.5-15% linear polyacrylamide gel. The identities of the labeled proteins are marked in the margin: ribosomal protein (Var1); subunit 1 (Cox1), subunit 2 (Cox2), and subunit 3 (Cox3) of cytochrome oxidase; cytochrome b (Cytb); subunit 6 (Atp6), subunit 8 (Atp8) and subunit 9 (Atp9) of ATPase. of G22 mutants. It displays a selective absence of cytochromes aa 3 and a loss of cytochrome oxidase activity. The cytochrome oxidase deficiency, combined with the lack of complementation of aW303⌬COX16 by C25 and E699, constitutes strong evidence that the mutations in these strains are allelic with the cox16 disruption.
Cox16p consists of 118 amino acid residues with a predicted mass of 14.1 kDa. The sequence includes one potential transmembrane domain (Fig. 3). The presence of a mitochondrial targeting sequence at the N terminus (Fig. 3B) is predicted by the P-Sort program (//psort.nibb.ac.jp/helpwww2.html). The amino acid composition of Cox16p is biased toward acidic residues, and analysis of the putative mature protein reveals a pI of 5.0 (//ca.expasy.org/tools/pi_tool.html). Cox16p does not reveal any homology to proteins of known function in the most recent databases, nor does it appear to have any identifiable functional domains that might provide clues about its function.
Expression and Mitochondrial Location of a cox16-myc Fusion Protein-The location of Cox16p was studied with a Myctagged protein. Because of the presence of a potential cleavable N-terminal signal sequence, the Myc epitope was added at the C terminus. Growth of transformants, expressing the tagged protein from either a high copy episomal (aW303⌬COX16/ myc1) or low copy CEN plasmid (aW303⌬COX16/myc2) on nonfermentable carbon sources (ethanol/glycerol), was indistinguishable from that of transformants expressing an untagged version of Cox16p. Restoration of growth on these substrates correlated with the recovery of cytochrome oxidase. This was evident from the spectra (not shown) and Western analysis of COX subunits (Fig. 1B). The concentrations of the three mitochondrially encoded subunits and subunit 5 in the transformants are similar to the wild type levels, regardless of the vector used to express the Myc-tagged Cox16p (Fig. 1B).
A low molecular weight protein in mitochondria from aW303⌬COX16/myc1 and aW303⌬COX16/myc2 is detected with the Myc antibody (Fig. 4A). This band is absent in the post-mitochondrial supernatant fraction from these strains and is also undetectable in wild type and mutant mitochondria. Based on its migration, this protein has an apparent mass of 21 kDa, which is 6 kDa larger than the combined size of native Cox16p and the additional 1 kDa contributed by the Myc epitope. The reason for the discrepancy, which may be even larger if Cox16p has a cleavable signal, is not clear but is probably related to anomalous binding of SDS by this acidic protein.
Almost all of the COX assembly factors identified to date are constituents of the mitochondrial inner membrane, as would be expected from their involvement in assembly of a hydrophobic multimeric complex of this membrane. The prediction of a membrane-spanning domain in Cox16p suggested that it might be an integral membrane protein. This is confirmed by the solubility properties of the protein. Titration of aW303⌬COX16/myc1 mitochondria with deoxycholate in the presence of 0.5 M NaCl showed that extraction of Cox16p-myc requires a minimum of 0.25% detergent (Fig. 4B).
Western analysis of intact mitochondria and mitoplasts from aW303⌬COX16/myc1 revealed Cox16p to be a constituent of the mitochondrial inner membrane as it is only found in the pelleted, mitoplast fraction (Fig. 4C). Treatment with proteinase K caused Cox16p to be degraded in mitoplasts but not mitochondria. The decreased signal seen in the proteinase Ktreated mitochondria is probably due to a subpopulation of mitochondria with damaged outer membranes. Similar results were obtained when mitochondria and mitoplasts were probed with antibody against Sco1p, an inner membrane protein protruding into the intermembrane space (32). Under these conditions, subunit 5 is protected from digestion by proteinase K in mitoplasts (data not shown).
Some of the COX assembly factors characterized to date

FIG. 4. Cox16p is a mitochondrial inner membrane protein.
A, Western analysis of mitochondria (Mit) and postmitochondrial supernatant (PMS) fractions from the wild type strain W303-1A (W303), the cox16 mutant aW303⌬COX16 (⌬COX16), and the transformants aW303⌬COX16/myc1 (myc1) and aW303⌬COX16/myc2 (myc2), expressing a Myc-tagged Cox16p from a high copy plasmid and from a CEN plasmid, respectively. Mitochondrial (20 g) or postmitochondrial supernatant fractions (40 g) were separated in 12% polyacrylamide gels and transferred to nitrocellulose, and the blot probed with a monoclonal antibody specific to the Myc epitope tag. B, mitochondria from aW303⌬COX16/myc1 were extracted in the absence (Ϫ) or presence (ϩ) of 0.5 M NaCl and increasing concentrations of deoxycholate (DOC). Following centrifugation at 100,000 ϫ g, the pellet (containing mitochondrial membrane with non-extracted protein) and supernatant (containing proteins extracted from the mitochondrial membrane) fractions were normalized back to the starting volume of mitochondria and separated on a 15% gel and analyzed as described in A. C, mitochondria with intact outer membranes were isolated and converted to mitoplasts by the method of Glick and Pon (47). The equivalent of 1 mg of mitochondrial protein starting material was incubated in the presence or absence of 0.016 mg of proteinase K for 30 min, and each fraction analyzed for the presence of the Myc-tagged Cox16p or Sco1p (mitochondrial copper transport). MT ϭ mitochondria; MP ϭ mitoplasts; p ϭ pellet; S ϭ supernatant; PK ϭ proteinase K; ϩ or Ϫ ϭ presence or absence of proteinase K. appear to be part of higher molecular weight homo-or heteroligomeric complexes (11,21,33,34). The native molecular mass of Cox16p-myc was estimated, by sedimentation of a 1% deoxycholate extract of mitochondria in a 7 to 20% linear sucrose gradient, to be ϳ84 kDa for both aW303⌬COX16/myc1 and aW303⌬COX16/myc2 (data not shown).
Cox16p Has Mammalian Homologs-Searches of current protein and expressed sequence tag databases indicate that Cox16p appears to have a human (HSCP203; accession number NP_057552) and a murine (accession number NP_079737) homologue. The human cDNA was originally identified in a screen for novel proteins expressed in CD34ϩ hematopoietic stem/progenitor cells (35). Human COX16 is located on the long arm of chromosome 14, in the interval 14q22.1-14q24.3 (LOC51241). Fig. 5 presents an alignment of the Cox16 proteins from yeast, humans, and mice. This analysis reveals that the highest sequence conservation is in the region of the transmembrane domain and the C-terminal half. The four Cox16 proteins shown in Fig. 5 share 24% identity and 40% conserved residues.
To test if the human gene can complement the yeast cox16 mutant the cDNA for human COX16 was amplified by PCR from a HeLa cell cDNA library. The cDNA was cloned into a yeast expression vector containing the ADH1 promoter and terminator in a YEp351 backbone, 3 allowing constitutive expression of proteins from cDNAs with their own start codons. This construct (pCOX16H/ST1) did not complement the respiratory deficiency of the aW303⌬COX16 parent strain even after prolonged incubation on a non-fermentable carbon source (7 days), whereas the yeast COX16 gene restores growth on ethanol/glycerol after one night. These results indicate that the human gene is homologous, but not orthologous, to yeast COX16.

DISCUSSION
The yeast COX16 gene product reported in this study is a mitochondrial protein with an essential function in COX assembly. The absence of cytochrome oxidase in cox16 mutants is accompanied by a loss of the cytochrome aa 3 spectral signal and a marked reduction in the steady state concentrations of the mitochondrial-encoded but not nuclear-encoded subunits of COX. The exception is Cox5p, which is also reduced in the cox16 mutant but not in other previously studied COX mutants. It is unclear whether this observation is relevant to the function of Cox16p.
Like most COX assembly factors characterized to date, Cox16p is an integral component of the mitochondrial inner membrane. The protease protection experiments indicate that the C terminus of Cox16p-myc is exposed to the cytoplasmic side of the inner membrane. This orientation implies that the hydrophobic sequence, proximal to the N terminus, acts as a stop-transfer signal (36). The location of the N-terminal transmembrane domain combined with the orientation of Cox16p in the inner membrane implies that the active region is in the hydrophilic C-terminal half and is located in the intermembrane space. At present, all the COX assembly factors with this orientation have Cox2p as their target. Sco1p is inferred to function in maturation of the Cu A site (28), the Imp protease is responsible for removal of the Cox2p N-terminal presequence (37), and Cox20p is a chaperone of the Cox2p precursor that is required for the proteolytic processing step (10). The presence of stable and mature Cox2p in the mutant excludes the involvement of Cox16p in export or proteolytic processing of the Nterminal extension. Attempts to relate Cox16p to maturation of a metal center in the enzyme have also been negative. Copper and other heavy metals, known to be present in COX, do not rescue cox16 mutants when added to the growth medium. This result does not, however, rule out an indirect role for Cox16p in the acquisition of any one of these metals by the enzyme.
Studies of the yeast Saccharomyces cerevisiae have shed considerable light on the widely differing roles that the dozen or more currently known nuclear-encoded factors play in COX assembly (15). Homologous proteins for most of these assembly factors have been identified in many other genomes, including human (38 -41). The yeast model has also facilitated the attribution of some autosomal recessively inherited human COX deficiencies to mutations in genes coding for proteins involved in heme A biosynthesis (42), copper homeostasis (41,43), and still other poorly understood aspects of COX expression (44). The human COX16 homologue does not complement a yeast cox16 mutant, which may indicate a different function for human COX16 or may be due to the heterologous context. The inability of a mammalian gene to exert its function in a yeast background is not without precedent. Only about one-third of the human COX assembly factors identified to date complement corresponding yeast mutants (38,40). We tend to think, therefore, that human COX16 has the same function as yeast Cox16p. The discovery of a new COX assembly factor provides another candidate gene for sequencing in patients with autosomal recessively inherited COX deficiencies without identified mutations.