Heme A is not essential for assembly of the subunits of cytochrome c oxidase of Rhodobacter sphaeroides.

The aa(3)-type cytochrome c oxidase of Rhodobacter sphaeroides, a proteobacterium of the alpha subgroup, is structurally similar to the core subunits of the terminal oxidase in the mitochondrial electron transport chain. Subunit I, the product of the coxI gene, normally binds two heme A molecules. A deletion of cox10, the gene for the farnesyltransferase required for heme A synthesis, did not prevent high level accumulation of subunit I in the cytoplasmic membrane. Thus, subunit I can be expressed and stably inserted into the cytoplasmic membrane in the absence of heme A. Aposubunit I was purified via affinity chromatography to a polyhistidine tag. Copurification of subunits II and III with aposubunit I indicated that assembly of the core oxidase complex occurred without the binding of heme A. In addition to formation of the apooxidase containing all three large structural proteins, CoxI-II and CoxI-III heterodimers were isolated from cox10 deletion strains harboring expression plasmids with coxI and coxII or with coxI and coxIII, respectively. This demonstrated that subunit assembly of the apoenzyme was not an inherently ordered or sequential process. Thus, multiple paths must be considered for understanding the assembly of this integral membrane metalloprotein complex.

Cytochrome c oxidase (complex IV), the terminal member of the electron transport chain in mitochondria, is essential for respiration. A closely related aa 3 -type oxidase is found in members of the ␣ subgroup of the proteobacteria, including Rhodobacter sphaeroides. Defective assembly of the bacterial aa 3 -type oxidase does not inhibit growth because of the presence of alternative terminal oxidases in the cell (1)(2)(3). Thus, assembly of cytochrome c oxidase can be studied in vivo in R. sphaeroides. A variety of human diseases, including myopathies and neuropathies, are associated with oxidase deficiencies that result from a failure of the multisubunit oxidase to properly assemble (reviewed in Ref. 4).
The subunit composition of the aa 3 -type oxidase of R. sphaeroides is ideal for studying assembly of the catalytic core complex. The bacterial oxidase is composed of only four subunits (5), whereas the mitochondrial oxidase has up to 13 different polypeptides encoded by both nuclear and mitochondrial genes (6). The three largest subunits of the bacterial oxidase are integral membrane proteins homologous to the subunits synthesized in mitochondria (7)(8)(9). Both prokaryotic and eukaryotic oxidases contain tightly bound cofactors that are necessary for electron transport (10,11). Two molecules of heme A, sixcoordinate heme a and five-coordinate heme a 3 , and a copper atom (Cu B ) are in subunit I; two additional copper atoms form the Cu A site in subunit II (5,(12)(13)(14).
Heme A is derived from heme B (iron protoporphyrin IX) with heme O as a probable intermediate (15). Conversion of the vinyl group at position 2 of the heme B porphyrin ring to a hydroxyfarnesyl moiety by the farnesyltransferase, Cox10p, forms heme O (16,17). Hemes A and O both contain the hydrophobic farnesyl group but are readily distinguished by optical spectroscopy because heme A has a formyl rather than a methyl group at position 8 of the porphyrin ring.
Many soluble hemoproteins require heme to achieve a native fold but assume a partially folded state in the absence of heme (Ref. 18 and references therein). The role that heme A plays in the folding and assembly of the integral membrane aa 3 -type oxidase complex is not well defined. An early investigation using fetal rat liver concludes that apocytochrome c oxidase fails to accumulate in the absence of Ѩ-aminolevulinic acid, a precursor of all hemes (19). Similar observations have been made in Paracoccus denitrificans (20) and Saccharomyces cerevisiae (21) cox10 mutants. In the yeast mutants, subunit I is efficiently translated but is nearly absent in steady-state mitochondria (21). These studies suggest that heme A is necessary for insertion and/or stability of subunit I in mitochondrial membranes. In fact, significant amounts of free subunit I are observed in bacterial membranes when heme A is present (22,23). Subunit I is relatively abundant (72% of control) in mitochondria of patients with a cox10 mutation encoding a partially active protein (24). Taken together, these studies raise the question of what role heme A plays in the assembly of cytochrome c oxidase.
In this report, the ability of subunit I to accumulate in the R. sphaeroides membrane in the absence of heme A was studied. An affinity tag on subunit I was used to isolate partially assembled forms of cytochrome c oxidase from cells lacking the genes for the assembly factors Cox10p and Cox11p (8). Heme A was absent because of the cox10 deletion; Cox11p is necessary for the formation of Cu B (25). The ability of subunit I lacking heme A to associate with the other members of the protein complex was investigated by expressing subunit I with subunits II and/or III.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-Plasmids with various combinations of the genes for R. sphaeroides cytochrome c oxidase subunits I, II, and III (see Table I) were constructed using standard molecular techniques (26). Plasmid pYJ123H contains the coxI-His gene and the coxII/III operon (coxII, cox10, cox11, and coxIII) in pUC19 (27). Deletions of 4.3-or 3.3-kb 1 SmaI fragments from pYJ123H were made, respectively, to create plasmids containing only coxI-His (pJH310) or coxI-His and * This work was supported by National Institutes of Health Grant GM56824 (to J. P. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Plasmid pMB307 is related to pYJ123H but does not contain coxIII (23). Site-directed mutagenesis was used to change two nucleotides to introduce a SmaI site 20 nucleotides 3Ј of the cox10 initiation codon. The resulting plasmid was named pMB307a. Deletion of the 1.6-kb SmaI fragment from pMB307a left coxI-His and coxII on opposite strands, each under the control of its natural promoter. This plasmid was designated pWA300. The addition of a 956-base pair SmaI fragment from pMB301 (see below) to pWA300 created pAH123⌬10,11, which contains coxI-His on one DNA strand and coxII and coxIII behind the coxII promoter on the opposite strand. The plasmid pMB301 was created by ligating the 956-base pair SmaI fragment from pYJ100 (27) into the multiple cloning site of pBC SK ϩ (Stratagene). The expression vectors pRKpWA300 and pRKpAH123⌬10,11 were derived from pWA300 and pAH123⌬10,11, respectively, by placing 3.7-or 4.6-kb EcoRI-HindIII fragments into pRK415-1. To create the control plasmid containing only coxII and coxIII, a 1.15-kb SalI fragment internal to coxI-His was deleted from pAH123⌬10,11 removing greater than 60% of the coxI open reading frame and creating pLH339. The 3.5-kb EcoRI-HindIII fragment of pLH339 was then placed into pRK415-1 to form pLH347. Each of the pRK415-based plasmids was conjugated into R. sphaeroides strain YZ200 by the published method (29).
Bacterial Growth and Oxidase Purification-R. sphaeroides cells were grown in minimal medium to the late exponential phase (25). Protein was purified by affinity chromatography on Ni-NTA-agarose from cytoplasmic membranes solubilized in N-dodecyl-␤-D-maltoside as previously described (25) except that the concentration of KCl in the purification buffers was 150 mM rather than 40 mM KCl. This change in the salt concentration did not significantly affect the ability to purify the oxidase subcomplexes.
Electrophoresis and Immunoblots-Electrophoresis was conducted in 14% polyacrylamide gels containing sodium dodecyl sulfate and 6 M urea (30). The proteins were visualized by rapid staining with Coomassie Blue (31). For immunoblots, the proteins were transferred to nitrocellulose and probed essentially as described (32). For detection of subunit I, a 1:5000 dilution of a monoclonal antibody against polyhistidine (Sigma) was used with colorimetric detection. For detection of subunit II, a polyclonal primary antibody raised against subunit II of the R. sphaeroides aa 3 -type oxidase 2 was used at a 1:400 dilution with protein A-125 I as the secondary reagent.
Cell Labeling-R. sphaeroides cells were grown in minimal medium to the exponential phase (OD 660 ϭ ϳ0.3). The cells were then washed with minimal medium prepared with chloride rather than sulfate salts and incubated in this medium for 4 h to decrease the pool of free methionine and cysteine. Radiolabeled methionine and cysteine were then added to the culture (1 mCi/50 ml) in the form of EasyTag EX-PRESS (PerkinElmer Life Sciences). Growth was rapidly stopped after 10 min by the addition of 0.01% chloramphenicol, 2% NaN 3 , 15 mM EDTA and placing the cells on ice. The culture was then divided into six equal portions for processing. After washing the cells, the extracts were prepared by vortexing in the presence of glass beads and the protease inhibitor, phenylmethylsulfonyl fluoride. Membranes were isolated by centrifugation at 350,000 ϫ g for 20 min. Washed membranes were solubilized in N-dodecyl-␤-D-maltoside, and the polyhistidine-tagged subunit I was bound to and eluted from Ni-NTA-agarose in a batch process in a microcentrifuge tube using a procedure similar to that used for column purification. The entire eluted fraction was prepared for denaturing gel electrophoresis, and fixed gels were exposed to a phosphor screen and analyzed using a Storm 860 PhosphorImager and ImageQuant software (Molecular Dynamics).

Isolation of Subunit I Lacking Heme
A-The necessity of heme A for steady-state accumulation of subunit I in the cytoplasmic membrane of R. sphaeroides was investigated using a strain (YZ200) with a genomic deletion of the coxII/III operon containing the coxII, cox10, cox11, and coxIII genes (27). Because the product of the cox10 gene, Cox10p, is required for heme A synthesis (33), YZ200 membranes had only b-and c-type cytochromes (Fig. 1). A low copy, broad host range expression vector containing the gene for subunit I with a carboxyl-terminal His 6 tag was used to express subunit I in YZ200 in a form that could be readily purified using Ni-NTA affinity chromatography. As expected, the cytochrome a peak (ϳ606 nm) was absent from the reduced minus oxidized membrane spectrum (Fig. 1, CoxI) because of the absence of Cox10p. However, a polypeptide was isolated that migrated on a denaturing polyacrylamide gel like wild-type subunit I and was identified as the product of the coxI-His gene using an antibody against the His 6 tag (Fig. 2, lane 2).
The ability to purify the subunit I polypeptide from a strain with a cox10 deletion demonstrated that the expression and membrane insertion of this protein was independent of heme A production. The possibility that another type of heme was substituted for heme A was addressed by measuring the heme content of the isolated protein using the pyridine hemochrome method (34). Heme B was identified in minor molar amounts (about 5% that of total protein) in protein isolated from the strain expressing CoxI. However, a similar amount of heme B was also found in a control experiment using membranes from strain YZ200 harboring only the parent vector (pRK415-1). Thus, heme B was likely to have been present as a minor cytochrome b contaminant that bound nonspecifically to the metal chelating resin. The other prosthetic group of subunit I, Cu B , was likely to be absent because of the deletion of the cox11 gene (25). Thus, the isolated product was presumed to be aposubunit I, i.e. lacking both hemes and Cu B . Throughout this report, the term "aposubunit I" will refer to the product of the coxI gene expressed in cells lacking cox10 and cox11.
Association of Aposubunit I with Subunits II and III-The ability of R. sphaeroides aposubunit I to associate with subunits II and III was ascertained using a copurification assay. Membrane proteins from cells expressing plasmid-borne coxI-His, coxII, and coxIII genes (CoxI-II-III; Table I) were subjected to affinity chromatography on Ni-NTA-agarose. The eluted material contained all three subunits, which were identified by gel electrophoresis and immunoblotting (Fig. 2, lanes 5 and 8).
Because only aposubunit I contained the His 6 tag, the isolation of subunits II and III via histidine affinity chromatography indicated formation of a complex containing all three subunits. Subunits II and III were not retained on Ni-NTA-agarose when 2 J. P. Hosler and S. Ferguson-Miller, unpublished data.
FIG. 1. Absence of the aa 3 -type oxidase in cox10 deletion mutants. Dithionite reduced minus ferricyanide oxidized spectra of membranes of R. sphaeroides strain YZ200 harboring no plasmid (YZ200), pRK415-1 (Vector), pWA302 (CoxI), or pRKpAH1H32 (WT) were recorded at room temperature using a Hitachi U-3000 UV-visible spectrophotometer. The spectra were normalized to the magnitude of the cytochrome b peak. The peaks labeled a, b, and c are attributed to a-, b-, and c-type cytochromes, respectively. they were expressed from CoxII/III (Table I) in the absence of polyhistidine-tagged subunit I (data not shown). Thus, the copurification of subunits II and III with aposubunit I demonstrated that formation of the three-subunit oxidase complex was independent of heme A insertion into subunit I. This CoxI-II-III complex was termed "apooxidase" because, in addition to the lack of heme A, metal analysis showed substoichio-metric amounts of copper (data not shown). This suggested that there was little or no Cu A in subunit II. Cu B was presumed to be absent because of the cox11 deletion (25).

Expression of Aposubunit I and Extent of Complex Formation-
The isolation of the apooxidase was a novel result, but the significance of this observation to the normal assembly of cytochrome c oxidase remains to be determined. To begin to address this question, the extent of complex formation was compared in strains that differed only by the presence or absence of cox10 and cox11 genes on the expression plasmid. The cells were labeled in vivo with 35 S-containing amino acids. Complexes containing the product of the coxI-His gene were isolated from cell membranes using Ni-NTA-agarose. Expression of the subunit I polypeptide and the extent of complex formation were evaluated after 10 min of cell labeling (Fig. 3). Aposubunit I was expressed to a level at least equal to that of subunit I in the wild-type strain (Fig. 3, A and B). All three subunits were detected from both the control strain (WT) and the strain lacking heme A (Fig. 3A, CoxI-II-III). Because of the low subunit II signal, attributed to a relatively small number of methionines and cysteines, subunit III was chosen for the purpose of quantitation. Approximately 30% of the newly translated subunit I from cells containing heme A (WT) and 10% of the aposubunit I from cells lacking heme A (CoxI-II-III) had associated with labeled subunit III (Fig. 3C). Because of the possible existence of an unlabeled pool of subunit III available for assembly, these numbers were taken to be a conservative estimate of the extent of complex formation. One possible explanation for the observed difference between formation of the apooxidase and the holooxidase control is a decreased affinity of subunit III for subunit I in the absence of heme A.
The accumulation of the normal oxidase (WT) and apooxidase (CoxI-II-III) to comparable levels in the bacterial membrane was indicated by the ability to isolate similar amounts of these proteins from steady-state cell cultures (data not shown). Generally, however, the yield of apooxidase was about 20% of the yield of normal oxidase from the same number of cells. These low recoveries may have resulted from an instability of the apooxidase in vitro; aposubunit I showed a tendency to precipitate. The smaller scale and more rapid processing of the pulse-labeled cultures may have minimized the effects of any such instability. Because a complex of aposubunit I with subunits II and III formed to a level of at least 30% relative to formation of a similar complex containing heme A, the role of  Table I, the vector pRK415-1 (column 1), or pRKpAH1H32 (WT; 25). Four separate gels were used for electrophoresis with loading of samples normalized to the percentage of material eluted from the affinity column. Less of the normal oxidase (WT) was used to give approximately equal signal intensities. Following electrophoresis, detection of the oxidase subunits was by immunoblotting (CoxIp and CoxIIp) or by staining with Coomassie (CoxIIIp and  lanes 7 and 8). Complexes containing the polyhistidine-tagged subunit I (CoxIp) were isolated using a metal chelating resin. A, denaturing SDS-polyacrylamide gel electrophoresis was used to separate the isolated complexes into their component subunits (CoxIp, CoxIIp, and CoxIIIp). Subunit II (CoxIIp) was found in multiple bands because of posttranslational processing events (27). The image shown is from a PhosphorImager. B, the amount of subunit I from six replicate samples of each strain was determined (mean Ϯ S.D.). Band intensities were normalized to cell number based on optical density of the cell cultures. C, the relative amount of labeled subunit III isolated was calculated using the same samples as for panel B. Subunit III band intensities were multiplied by 2.46 to account for the number of methionines and cysteines: 32 in subunit I and 13 in subunit III. Using these normalized values, the amount of subunit III as a percentage of the amount of subunit I in the same lane was determined (mean Ϯ S.D., n ϭ 6). aposubunit I in the assembly of the normal oxidase deserves consideration.
Assembly Order-The assembly pathway for mitochondrial cytochrome c oxidase is believed to be sequential or ordered (35,36). To directly determine the minimal requirements for the association of subunit I with the other core subunits, copurification of subunit II or subunit III with aposubunit I was tested. R. sphaeroides cells expressing coxI-His and coxII or coxI-His and coxIII were studied. Both CoxI-II and CoxI-III heterodimers were isolated from cells lacking heme A (Fig. 2, lanes  3 and 4). This result demonstrated that there was no obligate order of assembly for the protein subunits because both subunits II and III were competent to associate with subunit I in the absence of heme A. DISCUSSION In this study, assembly of the three catalytic core proteins of cytochrome c oxidase was investigated using the aa 3 -type oxidase from R. sphaeroides. Both aposubunit I and a threesubunit apooxidase were isolated from strains lacking heme A. These proteins inserted into the cytoplasmic membrane and accumulated to significant levels (at least 30%) compared with their heme A-containing counterparts. Thus, the insertion of heme A was not a prerequisite for subunit assembly of the oxidase. Conversely, subunit association is not required for insertion of one molecule of heme A (23). This raises the possibility that heme A insertion may be able to occur at multiple steps in the assembly process.
A dependence on order for the association of aposubunit I with the other structural subunits was investigated using strains expressing two of the subunits in the absence of heme A. Two distinct subcomplexes containing aposubunit I (CoxI-II and CoxI-III) were isolated (Fig. 2, lanes 3 and 4). Because aposubunit I could associate with either subunit II or subunit III in genetically manipulated strains, it was concluded that there was no obligate order of subunit assembly. Thus, the possibility of multiple paths must be considered in determining the assembly process of cytochrome c oxidase in wild-type cells. Insertion of the prosthetic groups also appears to be unordered because heme A is incorporated into subunit I in the absence of other subunits (23) but is not required for subunit assembly of several apooxidase forms (i.e. CoxI-II, CoxI-III, and CoxI-II-III).
A "state diagram" is a way of illustrating a finite set of possible assembly paths where some states (subcomplexes) lie on multiple paths. There are 32 theoretical oxidase assembly intermediates that contain subunit I, considering only the three metal centers within subunit I and the three large structural subunits. These possible partially assembled oxidase forms are arranged in Fig. 4 based on the number of prosthetic groups in subunit I and the number of protein subunits such that assembly proceeds from left to right and components are added one at a time.
From this study and others (23,25), a few of these oxidase forms (shown as schematics) have been isolated from steadystate membranes, and the logical paths between them can be inferred. Assuming that none of these purified subcomplexes lie on dead-end paths, the conclusion must be made that assembly of the holoenzyme is not a strictly ordered process. For example, the addition of subunit III can be the initial step (to form CoxI-III) or the final step between I aa 3 Cu B -II and the normal oxidase, I aa 3 Cu B -II-III. Note that some subcomplexes (e.g. I-II-III) lie on multiple putative paths. The existence of the majority of the theoretical states is currently undetermined, and the extent to which any given subcomplex or putative path exists in wild-type bacteria remains to be investigated. Therefore, the partially assembled forms of the oxidase that have been isolated from genetically manipulated bacteria may or may not be bona fide assembly intermediates.
The deletion of cox10 and cox11 in R. sphaeroides did not prevent the accumulation of aposubunit I in bacterial cytoplasmic membranes or its association with subunits II or III (Fig.  2). Consistent with these results, subunit I with no heme A accumulates to a low level in mitochondria of a yeast strain with a disruption of the gene for Cox10p (21). This suggests that prokaryotes and eukaryotes have similar mechanisms for assembly of the cytochrome c oxidase core subunits. Furthermore, the gene for Cox11p was unaltered in the yeast strain (21), indicating that the genetic alteration critical for the ob- FIG. 4. Theoretical oxidase subcomplexes. Possible assembly intermediates or "states" of oxidase (see text) are shown. Subunits are designated I, II, and III with the metal-containing cofactors of subunit I shown as subscripts. Oxidase forms that have been isolated in this or previous studies (23,25) are shown as schematics with labels in bold type. Subunits I, II, and III are represented by rectangles with no shading, shading, and cross-hatching, respectively. The bars indicate the two hemes, and the circles represent Cu B . The Cu A site of subunit II has been omitted for simplicity. Arrows illustrate the most direct paths connecting characterized subcomplexes which differ by a single component.
served phenotypes in the present study was the cox10 deletion. This is consistent with our observation that an apooxidase (CoxI-II-III) was isolated from a R. sphaeroides strain containing the wild-type cox11 gene and a deletion of cox10 (data not shown). The posttranslational loss of significant amounts of subunit I in yeast (21) could be due to in vivo degradation or, as indicated in the present study, could be explained by an in vitro instability during sample preparation or by structural alterations that reduce the ability of the epitope to be recognized by the antiserum. These explanations are also relevant to the report of a P. denitrificans strain with a deletion of the genes corresponding to coxII, cox10, and cox11 that does not express immunologically detectable subunit I (20). It is likely that modest overexpression of the gene for subunit I and isolation of the protein, as opposed to analysis in the membrane, significantly enhanced our ability to detect aposubunit I in this study.
In conclusion, heme A is not essential for the expression, membrane insertion, or association of the core complex proteins of cytochrome c oxidase in R. sphaeroides. Furthermore, the assembly of this heterooligomeric integral membrane complex need not follow a linear or ordered pathway. The structural similarities between the bacterial and mitochondrial oxidases suggest that these findings may extend to assembly of the human oxidase. A "state model" has been proposed as a paradigm for describing the multiple possible assembly paths.