Purification of a Cytochrome bc1-aa3 Supercomplex with Quinol Oxidase Activity from Corynebacterium glutamicum IDENTIFICATION OF A FOURTH SUBUNIT OF CYTOCHROME aa3 OXIDASE AND MUTATIONAL ANALYSIS OF DIHEME CYTOCHROME c1*

The aerobic respiratory chain of the Gram-positive Corynebacterium glutamicum involves a bc1 complex with a diheme cytochrome c1 and a cytochrome aa3 oxidase but no additional c-type cytochromes. Here we show that the two enzymes form a supercomplex, because affinity chromatography of either strep-tagged cytochrome b (QcrB) or strep-tagged subunit I (CtaD) of cytochrome aa3 always resulted in the copurification of the subunits of the bc1 complex (QcrA, QcrB, QcrC) and the aa3 complex (CtaD, CtaC, CtaE). The isolated bc1-aa3 supercomplexes had quinol oxidase activity, indicating functional electron transfer between cytochrome c1 and the CuA center of cytochrome aa3. Besides the known bc1 and aa3 subunits, few additional proteins were copurified, one of which (CtaF) was identified as a fourth subunit of cytochrome aa3. If either of the two CXXCH motifs for covalent heme attachment in cytochrome c1 was changed to SXXSH, the resulting mutants showed severe growth defects, had no detectable c-type cytochrome, and their cytochrome b level was strongly reduced. This indicates that the attachment of both heme groups to apo-cytochrome c1 is not only required for the activity but also for the assembly and/or stability of the bc1 complex.

Corynebacterium glutamicum is a non-pathogenic aerobic soil bacterium that has gained considerable interest because of its use in large scale biotechnological production of L-glutamate and L-lysine (1) and because of its emerging role as a model organism for the Gram-positive bacteria with high GϩC content (2), which include a number of important pathogens, in particular Corynebacterium diphtheriae and Mycobacterium tuberculosis. In this context, the respiratory chain of C. glutamicum was also analyzed in recent years, both genetically and biochemically. It is composed of several dehydrogenases that transfer electrons to an intramembrane pool of menaquinone-9 (3) and at least two branches for reoxidation of menaquinol, one consisting of a cytochrome bd-type quinol oxidase (4) and the second one consisting of menaquinol:cytochrome c oxidoreductase (cytochrome bc 1 complex) and cytochrome aa 3 oxidase (5-7). The latter one is of primary importance for growth, because mutants lacking either the bc 1 complex or cytochrome aa 3 have severe growth defects (5) (see also Fig. 2).
The dehydrogenases include a non-proton-pumping NADH dehydrogenase encoded by the ndh gene (8,9), malate:quinone oxidoreductase encoded by the mqo gene (8,10), and succinate dehydrogenase encoded by the sdhCAB genes (Cgl0370, Cgl0371, Cgl0372). Succinate oxidase activity was shown to be inhibited by an uncoupler, indicating that electron transfer from succinate to menaquinone requires the electrochemical proton potential across the cytoplasmic membrane (11).
The cytochrome bd oxidase was purified and shown to consist of two subunits of 56 and 42 kDa encoded by cydA and cydB, respectively. It was proposed that this oxidase is predominant during the stationary phase of growth (4). The cytochrome bc 1 complex is encoded by the qcrCAB genes ( Fig. 1) for cytochrome c 1 , Rieske iron-sulfur protein, and cytochrome b, respectively (5,6). The protein sequences deduced from these genes revealed a number of differences to classical representatives of the bc 1 complex, such as an extension of about 120 amino acids at the C terminus of cytochrome b and the presence of three putative transmembrane helices in the N terminus of the Rieske iron-sulfur protein rather than only one. Most remarkably, cytochrome c 1 was found to have two CXXCH motifs for covalent heme attachment, suggesting that it is a diheme ctype cytochrome (5,6). Purification of cytochrome c 1 confirmed the presence of two heme groups in the protein (6). Upstream of qcrC, the genes encoding subunit II (ctaC) and III (ctaE) of cytochrome aa 3 oxidase were identified (5,6), as was as an additional open reading frame located in between these two genes, which was designated ctaF in the course of this work (Fig. 1). Compared with "classical" subunit II representatives, CtaC of C. glutamicum contained an insertion of about 30 amino acid residues in the substrate binding domain, which was proposed to play a role in the interaction with cytochrome c 1 (7). The gene encoding subunit I of cytochrome aa 3 was found to be located separately at a different genomic site (5,7). Cytochrome aa 3 oxidase was isolated by conventional chromatographic techniques as a complex consisting of CtaD, CtaC and CtaE (7). HPLC 1 and mass spectrometry of the isolated heme of subunit I indicated that it is presumably heme a S , in which the farnesyl group (C 15 H 25 Ϫ) of heme a is replaced by a geranylgeranyl side chain (C 20 H 33 Ϫ). Subunit II contains a lipoprotein signal sequence, and in fact Cys-29, whose thiol group might be diacylglycerated, was identified as the N-terminal residue of the mature protein (7).
Staining of proteins separated by SDS-PAGE for covalently bound heme indicated that there is only a single c-type cytochrome with an apparent mass of 31 kDa present in C. glu-tamicum wild type (5,6). This protein was missing in the mutant strain 13032⌬qcr, which lacks the qcrCAB genes, confirming that it represents cytochrome c 1 (5). The absence of additional c-type cytochromes in C. glutamicum indicated that the second heme group of cytochrome c 1 takes over the function of a separate cytochrome c in electron transfer to cytochrome aa 3 oxidase. Such a function would require an intimate contact between cytochrome c 1 and the Cu A electron entry site in subunit II of cytochrome aa 3 , and therefore we suggested that the bc 1 complex and cytochrome aa 3 might form a supercomplex (5). In this work, we were able to prove the existence of such a supercomplex by using a very gentle method for its purification, i.e. affinity chromatography with the StrepTag II/StrepTactin system (12). Moreover, a fourth subunit of cytochrome aa 3 oxidase was identified, and it was shown that incorporation of both heme groups into cytochrome c 1 is essential for the assembly and/or stability of the entire bc 1 complex.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Culture Conditions-C. glutamicum strain ATCC 13032 (13) and derivatives were cultivated aerobically in Erlenmeyer flasks at 120 rpm and 30°C in brain heart infusion medium (BHI; Difco) with 2% (w/v) glucose or in CGXII minimal medium with 4% (w/v) glucose as carbon and energy source (14). When appropriate, 25 g of kanamycin/ml was added. The C. glutamicum strains and the plasmids used in this study are listed in Table I. For all cloning purposes, Escherichia coli DH5␣ (Invitrogen) was used and routinely grown at 37°C in LB medium (15). When appropriate, 50 g of kanamycin/ml or 100 g of ampicillin/ml was added.
Recombinant DNA Work-All enzymes for recombinant DNA work were obtained either from Roche Diagnostics or New England Biolabs. The oligonucleotides used in this study were obtained from MWG Biotech (Ebersberg, Germany) and are listed in Table II. Standard methods were used for the cloning procedures (15).
For the purification of the cytochrome bc 1 complex, a QcrB derivative with a C-terminal StrepTag II (12) was constructed as follows. The entire ctaE-qcrCAB gene cluster including the putative promoter region was amplified by PCR (Expand high fidelity PCR system; Roche Diagnostics) using a reverse primer that included the codons for the StrepTag II (WSHPQFEK) preceded by two alanine codons. The resulting 5.0-kb fragment was cloned into the E. coli-C. glutamicum shuttle vector pJC1 using the XbaI and SalI restriction sites introduced by the primers. The resulting plasmid pJC1-qcrB St encoded a QcrB derivative with ten additional residues at the C terminus (calculated mass, 61.1 kDa). A CtaD derivative with a C-terminal StrepTag II for the purification of the cytochrome aa 3 complex was constructed similarly except that only the monocistronic ctaD gene with its native promoter was amplified by PCR. The resulting 2.0-kb fragment was cloned via XbaI restriction sites into pJC1 yielding pJC1-ctaD St . The modified CtaD protein contained ten additional residues at the C terminus (calculated mass, 66.3 kDa). Each of the two plasmids was transferred into the C. glutamicum strains 13032⌬ctaD and 13032⌬qcr by electroporation as described (16). For the synthesis of cytochrome c 1 derivatives defective in covalent binding of either the N-terminal or the C-terminal heme group, site-directed mutagenesis of qcrC was performed by a two-step PCR procedure according to Higuchi et al. (17). For that purpose, a 2.0-kb XbaI-ApaI fragment from pJC1-qcrB St was cloned into pUC-BM20. The resulting plasmid pBM20-QXA served as template for mutagenesis with the universal primers M13-forward/-reverse and the mutagenic primers C67S-forward/-reverse and C177S-forward/-reverse, respectively. The products obtained after crossover PCR were digested with XbaI and ApaI and cloned into pUC-BM20, yielding pBM20-QXA-C67S and pBM20-QXA-C177S. The presence of the desired mutations and the absence of additional mutations were confirmed by DNA sequence analysis (18). The mutated 2.0-kb XbaI-ApaI fragments were exchanged against the corresponding wild-type fragment of pJC1-qcrB St yielding pJC1-qcrB St -C67S and pJC1-qcrB St -C177S, respectively, which were transferred into C. glutamicum strain 13032⌬qcr. In the cytochrome c 1 variants encoded by these plasmids, residues Cys-67 and Cys-70 or Cys-177 and Cys-180 are replaced by serine residues. In-frame deletion mutants of C. glutamicum were constructed as described previously using crossover PCR and the suicide vector pK19mobsacB (5). The deletions were verified by PCR and by Southern blot analysis (data not shown).
Preparation of Cell Membranes-Cells (10 g wet weight) were suspended in 15 ml of 100 mM Tris-HCl buffer, pH 7.5, containing 5 mM MgSO 4 and 10 mg/ml lysozyme. After 45 min of incubation at 37°C, 1 mM of the protease inhibitor phenylmethanesulfonyl fluoride was added, and the cells were disrupted by five passages at 207 mega Pascals through a French pressure cell (SLM Aminco). Cell debris was removed by centrifugation at 27,000 ϫ g for 20 min, and the supernatant was ultracentrifuged at 150,000 ϫ g for 90 min. The pellet containing the cytoplasmic membrane fraction was washed in 100 mM Tris-HCl, pH 7.5, and centrifuged again at 150,000 ϫ g for 90 min. Then the membranes were resuspended in a small volume of the same buffer containing 10% (v/v) glycerol and stored at Ϫ20°C.
Purification of the Strep-tagged Protein Complexes-Washed membranes were adjusted to a protein concentration of 5 mg/ml in 100 mM Tris-HCl, pH 7.5, containing 50 g/ml egg white avidin (Sigma). The membrane proteins were solubilized by adding n-dodecyl-␤-D-maltoside (Biomol, Hamburg, Germany) from a 10% (w/v) aqueous solution to a final ratio of 2 g of dodecyl maltoside/g of protein. After 45 min of incubation on ice with slow stirring the sample was ultracentrifuged at 180,000 ϫ g for 20 min. The supernatant was applied to a StrepTactin-Sepharose column with a bed volume of 2 ml (IBA, Göttingen, Germany) equilibrated with 100 mM Tris-HCl buffer, pH 7.5, containing 0.025% (w/v) dodecyl maltoside. The column was washed with 9 ml of a buffer containing 100 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM MgSO 4 , and 0.025% (w/v) dodecyl maltoside. Specifically bound proteins were eluted with the same buffer supplemented with 2.5 mM D-desthiobiotin (Sigma) and 10% (v/v) glycerol.
Protein Identification by Peptide Mass Fingerprinting-Proteins separated by SDS-PAGE and stained with Coomassie Blue were identified as described previously (19) by peptide mass fingerprinting after in-gel digestion with trypsin (Promega) or with cyanogen bromide (20). If required for reliable identification, post-source decay analysis of selected peptides was carried out (19). Peptide mass lists were used to search a local digest data base of 3312 C. glutamicum proteins, provided by the Degussa AG (Frankfurt, Germany).
Enzyme Assays-N,N,NЈ,NЈ-Tetramethyl-p-phenylenediamine (TMPD) oxidase activity was measured spectrophotometrically at 562 nm in air-saturated 100 mM Tris-HCl buffer, pH 7.5, containing 200 M TMPD, at 25°C. For the calculation, an extinction coefficient of 10.5 mM Ϫ1 ⅐cm Ϫ1 was used (7). One unit of activity is defined as 1 mol of TMPD oxidized per min. Quinol oxidase activity was measured as oxygen consumption in a magnetically stirred 2-ml chamber with a Clark-type oxygen electrode (Rank Brothers, Cambridge, United Kingdom). The chamber was thermostatted at 25°C and filled with 1 ml of 50 mM air-saturated sodium phosphate buffer, pH 6.5, supplemented with 200 M dimethylnaphthoquinol (DMNH 2 ). After recording the rate of autoxidation, the measurement was started by adding the protein sample. One unit of activity refers to 1 mol of O 2 reduced per min. Dimethylnaphthoquinone (DMN) was obtained initially as a gift from A. Kröger (Frankfurt, Germany) and later synthesized by mild oxidation of 2,3-dimethylnaphthaline with chromium(VI) oxide as described by Kruber (21). DMNH 2 was formed by adding a few grains of sodium borohydride and sodium dithionite to a 5 mM solution of DMN in 50% ethanol. Cytochrome c oxidase activity was measured spectrophotometrically at 550 nm with bovine heart cytochrome c (Sigma) or yeast cytochrome c (Sigma) as described (7). One unit of activity refers to 1 mol of cytochrome c oxidized per min.
Miscellaneous-Protein concentrations were determined with the bicinchoninic acid protein assay (25) using bovine serum albumin as the standard. SDS-PAGE was carried out as described (26) except that the samples were incubated at 40°C for 30 min before loading. Staining of c-type cytochromes in polyacrylamide gels was performed with 3,3Ј,5,5Ј-tetramethylbenzidine (27). For Western blotting, proteins were separated by Tricine-SDS-PAGE (28) and electroblotted onto a polyvinylidene difluoride membrane (Immobilon P; Millipore) using the semidry method according to Schä gger and von Jagow (29). Strep-tagged QcrB was detected using streptavidin-alkaline-phospha- Primers for the construction of pJC1-ctaD St and pJC1-qcrB St In-frame deletion of ctaD gene encoding subunit I of cytochrome aa 3 oxidase 5 13032⌬qcr Deletion of the qcrCAB genes encoding the three subunits of the bc 1 (12). 3 Oxidase-To purify the cytochrome bc 1 complex and cytochrome aa 3 oxidase by affinity chromatography, plasmids pJC1-qcrB St and pJC1-ctaD St were constructed encoding QcrB and CtaD proteins elongated with a C-terminal StrepTag II, respectively. Plasmid pJC1-qcrB St contained the entire ctaE-qcrCAB gene cluster under control of its presumed native promoter and was able to complement the severe growth defect of C. glutamicum strain 13032⌬qcr, which contains a deletion of the chromosomal qcrCAB genes (Fig. 2). Plasmid pJC1-ctaD St contained the ctaD gene with its promoter region and complemented the growth defect of C. glutamicum strain 13032⌬ctaD (Fig. 2). Reduced minus oxidized difference spectra of the complemented strains ⌬Q-B St (13032⌬qcr with plasmid pJC1-qcrB St ) and ⌬C-D St (13032⌬ctaD with plasmid pJC1-ctaD St ) revealed a wild-type-like pattern with cytochromes of the a-, b-, and c-type (data not shown) whereas those of strains 13032⌬qcr and 13032⌬ctaD lacked cytochrome c and cytochrome a, respectively (5). Thus, pJC1-qcrB St and pJC1-ctaD St allowed the synthesis of a functional bc 1 complex and of a functional cytochrome aa 3 oxidase, respectively, and the presence of the StrepTag II did not interfere with the activity of the two respiratory complexes.

Construction and Functional Analysis of Strep-tagged Variants of Cytochrome b (QcrB) and of Subunit I (CtaD) of Cytochrome aa
Isolation of a Cytochrome bc 1 -aa 3 Supercomplex-For the purification of the bc 1 complex and cytochrome aa 3 oxidase, membranes of the complemented strains ⌬Q-B St and ⌬C-D St were isolated, and the proteins obtained after solubilization with dodecyl maltoside were subjected to affinity chromatography on StrepTactin-Sepharose. After washing, specifically bound proteins were eluted with desthiobiotin and analyzed by SDS-PAGE. Surprisingly, the protein pattern observed in the eluates from strains ⌬Q-B St (Fig. 3, lane 3) and ⌬C-D St (Fig. 3,  lane 2) were highly similar and contained eight protein bands of identical apparent mass. The protein of 24 kDa (P24) was not only copurified with QcrB St but also appeared in some prepa-rations obtained with CtaD St (data not shown). The identity of the proteins indicated in Fig. 3 except for the 17-kDa protein (P17) was determined by peptide mass fingerprinting using in-gel digestion with trypsin or cyanogen bromide and matrixassisted laser desorption ionization-time of flight mass spectrometry. Because the bands with an apparent mass of 52 and 29 kDa were found to consist of two different proteins at a time, 10 proteins were identified in total. Besides the known subunits of the bc 1 complex (QcrA, QcrB, QcrC) and of cytochrome aa 3 oxidase (CtaC, CtaD, CtaE), the four additional proteins with an apparent mass of 29 kDa (P29), 24 kDa (P24), 20 kDa (P20) and 19 kDa (P19) were assigned to the hitherto hypothetical proteins Cgl2664, Cgl2226, Cgl2017, and Cgl2194, respectively.
The successful protein identification clearly showed that the eluate both of strain ⌬Q-B St (Fig. 3, lane 3) and of strain ⌬C-D St (Fig. 3, lane 2) contained the three subunits of the bc 1 complex (QcrA, QcrB, and QcrC) and the three subunits of cytochrome aa 3 oxidase (CtaD, CtaC, and CtaE). The fact that these proteins were copurified irrespective of whether the purification was performed via QcrB St or CtaD St strongly indicated that the bc 1 complex and cytochrome aa 3 oxidase form a supercomplex in C. glutamicum.
Heme Contents and Enzymatic Activities of the bc 1 -aa 3 Supercomplex-Reduced minus oxidized difference spectra of the supercomplex purified either via QcrB St or via CtaD St showed that both preparations contained cytochromes of the a-, b-, and c-type but in different ratios (Fig. 4). The calculated contents of heme a, heme b, and heme c were 1.6, 6.1, and 2.8 mol/g of protein in the QcrB St complex and 4.2, 2.6 and 3.0 mol/g in the CtaD St complex. These values cannot be fit into a simple ratio of small integers, indicating that the preparations are stoichiometrically heterogeneous. In both cases the strep-tagged subunit was most abundant, i.e. cytochrome b in the QcrB St complex and cytochrome a in the CtaD St complex. Thus, the bc 1 -aa 3 supercomplex was partially dissociated despite the gentle method used for purification.
A functional association of bc 1 complex and cytochrome aa 3 oxidase should possess quinol oxidase activity. Using the menaquinol analogon DMNH 2 as substrate, such an activity could be measured polarographically not only with membrane fractions but also with the purified supercomplexes as summarized in Table III. The turnover number decreased during the purifica- Purification of the bc 1 Complex and of Cytochrome aa 3 Oxidase as Single Complexes-To purify the bc 1 complex as a single complex rather than as a supercomplex, plasmid pJC1-qcrB St was transferred to C. glutamicum 13032⌬ctaD. The resulting strain ⌬C-B St had the same growth defect as strain 13032⌬ctaD because of the absence of CtaD and formed both wild-type and strep-tagged QcrB. Purification of strep-tagged proteins from dodecyl maltoside-solubilized membranes by StrepTactin affinity chromatography resulted in two dominant proteins that were identified as cytochrome b (QcrB) and Rieske iron-sulfur protein (QcrA; Fig. 3, lane 4). In addition, minor amounts of P24 were enriched. Most remarkably, cytochrome c 1 (QcrC) was not present in this preparation, indicating that the interaction between QcrC and the two other subunits of the bc 1 complex is quite weak.
For the purification of cytochrome aa 3 as a single complex, a similar approach was applied as described above for the bc 1 complex, i.e. plasmid pJC1-ctaD St was transferred into strain 13032⌬qcr. The resulting strain ⌬Q-D St had the same phenotype as strain 13032⌬qcr and formed both wild-type and streptagged CtaD. The eluate obtained after StrepTactin affinity chromatography of dodecyl maltoside-solubilized membranes contained four proteins (Fig. 3, lane 1), which were identified as CtaD, CtaC, CtaE, and P19. The TMPD oxidase activity of the cytochrome aa 3 oxidase preparation was 0.34 units/mg, corresponding to a turnover number of 1.1 TMPD oxidized/aa 3 s Ϫ1 . The 10-fold decreased TMPD oxidase activity compared with the supercomplexes is because of the absence of cytochrome c 1 . The cytochrome c oxidase activity with bovine heart cytochrome c and yeast cytochrome c was 0.35 and 0.28 units/mg, respectively. This corresponds to turnover numbers of 1.2 and 0.9 cytochrome c oxidized/aa 3 s Ϫ1 .
Evidence for a Fourth Subunit of Cytochrome aa 3 by Phenotypic Analysis of Mutants Lacking P29, P24, P20, or P19-To determine the relevance of proteins P29, P24, P20, and P19 for respiration and formation of the bc 1 -aa 3 supercomplex, the corresponding genes were deleted in-frame from the chromosome of C. glutamicum, resulting in strains ⌬Cg2664, ⌬Cg2226, ⌬Cg2017, and ⌬Cg2194, respectively. The former three strains showed no obvious phenotype regarding growth in rich medium and the formation of a-, b-, and c-type cytochromes (data not shown). Apparently, proteins P29, P24, and P20 are not essential for the formation and activity of the bc 1 -aa 3 branch of the respiratory chain, and the functional significance of the interaction between these proteins and the bc 1 -aa 3 supercomplex remains to be elucidated.
In contrast, deletion of the gene Cgl2194 encoding P19 led to a similar phenotype observed previously for the 13032⌬ctaD strain. Growth on rich medium agar plates was strongly impaired (data not shown), cytochrome a was almost absent in the spectrum of dithionite-reduced cells, and the level of cytochrome c 1 was markedly lower than in the wild-type (Fig. 5). Consequently, the P19 protein is essential for the formation of an active cytochrome aa 3 oxidase. P19 was enriched with the supercomplex and the isolated cytochrome aa 3 oxidase, but not with the isolated bc 1 complex (Fig. 3), showing that copurification is because of an interaction with the cytochrome aa 3 subunits. Based on these data, P19 has to be regarded as a fourth subunit of the C. glutamicum cytochrome aa 3 oxidase.
Protein P19 is composed of 143 amino acids and has a predicted mass of 15.5 kDa. It contains three hydrophobic regions extending from residues 7-27, 40 -60, and 97-130, which presumably form three or four transmembrane helices. The first transmembrane helix may be part of a signal peptide. As shown in the alignment in Fig. 6, the primary sequence is well conserved in other species of the actinomycetales including C. diphtheriae (68% identity), mycobacteria (38 -39%), Streptomyces coelicolor (39%), and Thermobifida fusca (33%). In all these organisms the corresponding gene is located immediately downstream of ctaC or a ctaCD gene cluster in the case of S. coelicolor and T. fusca and presumably is cotranscribed with these genes. This further supports the previous suggestion that the P19 homologues represent a fourth subunit of cytochrome aa 3 oxidase in the actinomycetes. Therefore, the corresponding genes were named ctaF.
Necessity of Heme Incorporation into Cytochrome c 1 for Assembly and/or Stability of the bc 1 -aa 3 Supercomplex-According to our previous proposal that the second heme group of the C. glutamicum diheme cytochrome c 1 is involved in electron FIG. 4. Reduced minus oxidized difference spectra of the bc 1aa 3 supercomplex purified from the C. glutamicum strains ⌬C-D St and ⌬Q-B St . The supercomplexes purified by affinity chromatography on StrepTactin-Sepharose were reduced with dithionite or oxidized with ferricyanide before recording the spectra at room temperature (protein concentration 0.5 mg/ml).

TABLE III
Purification of a cytochrome bc 1 -aa 3 -supercomplex from membranes of the C. glutamicum strains ⌬Q-B St and ⌬C-D St The quinol oxidase activity was determined with DMNH 2 as substrate by measuring the oxygen consumption rate with a Clark-type oxygen electrode. One unit corresponds to 1 mol of O 2 consumed per min. The turnover numbers (TN) are calculated as electrons transferred per cytochrome aa 3 per second, except for the number in parentheses, which was calculated as electrons transferred per cytochrome b per second. ND, not determined. transfer from the first heme group of c 1 to the Cu A center of cytochrome aa 3 oxidase, both heme groups should be essential for the activity of the bc 1 -aa 3 branch of the respiratory chain.
To test this assumption, both cysteine residues in each of the two CXXCH heme binding motifs of QcrC were converted to serine residues by site-directed mutagenesis of plasmid pJC1-qcrB St . The effects of these mutations were analyzed after transformation of strain 13032⌬qcr with the resulting plasmids pJC1-qcrB St -C67S and pJC1-qcrB St -C177S, respectively. Both mutant strains showed strongly impaired growth similar to strain 13032⌬qcr (Fig. 7), indicating the absence of a functional bc 1 complex. The membranes of the two strains did not contain c-type cytochromes as judged by heme staining of SDS gels (Fig. 8A) and reduced minus oxidized difference spectra of membranes (data not shown). Obviously, both heme groups of cytochrome c 1 are essential for respiration via the bc 1 -aa 3 branch of the respiratory chain, and no stable monoheme intermediate can be formed during the maturation of QcrC if the incorporation of the other heme group is blocked. Western blot analysis with streptavidin-alkaline phosphatase conjugate was performed to check whether the disturbed cytochrome c 1 maturation in the mutant strains also influences the QcrB St content of the cytoplasmic membranes. As shown in Fig. 8B, both QcrC mutant strains had strongly decreased QcrB St levels of less than 10% compared with strain ⌬Q-B St as estimated from the signal intensities. This indicated that the presence of holo-cytochrome c 1 is highly important for the assembly and/or stability of the entire bc 1 complex. Besides QcrB St , which was unequivocally identified with a sample of the purified QcrB St complex, additional proteins were detected by the streptavidin-alkaline phosphatase conjugate (Fig. 8), which represented the biotinylated proteins pyruvate carboxylase (not shown) (30) and the ␤-subunit of acyl CoA carboxylase (AccBC; see Ref. 31). The cytochrome a level of the cytochrome c 1 mutants was unchanged compared with the control strain ⌬Q-B St (data not shown), indicating that synthesis of cytochrome aa 3 is independent of an intact bc 1 complex.

DISCUSSION
Identification of a Cytochrome bc 1 -aa 3 Supercomplex in C. glutamicum-In the present study we show that the bc 1 complex and cytochrome aa 3 oxidase of C. glutamicum are organized in a supercomplex with quinol oxidase activity. The approach used to isolate this supercomplex involved the modification of one subunit with a StrepTag II and subsequent affinity purification with StrepTactin-Sepharose. Similar approaches were used previously, e.g. to systematically define protein complexes in yeast (32,33). Although this procedure can lead to the accidental copurification of proteins, this was not the case here, because all subunits of the bc 1 complex and of cytochrome aa 3 were isolated both with strep-tagged QcrB and with strep-tagged CtaD. The lack of evidence for a supercomplex in the previous purification of either cytochrome c 1 (6) or cytochrome aa 3 oxidase (7) shows that the interactions are relatively weak and require a very gentle purification procedure for preservation. Although we could also isolate the bc 1aa 3 supercomplex using a hexahistidine-tagged QcrB and Ni 2ϩchelate affinity chromatography (data not shown), the StrepTag II/StrepTactin system proved to be superior in our hands.
The formation of a bc 1 -aa 3 supercomplex with quinol oxidase activity is not unique to C. glutamicum. In fact, such complexes were purified from several bacteria, i.e. Paracoccus denitrificans (34), the thermophilic Bacillus PS3 (35), or the thermoacidophilic archaeon Sulfolobus sp. strain 7 (36). In Bradyrhizobium japonicum, a bc 1 -c M -aa 3 complex was isolated from aerobically grown cells but not characterized for its quinol oxidase activity (37). From bacteroids of B. japonicum a complex of cytochrome bc 1 and a cb-type cytochrome oxidase, most probably cytochrome cbb 3 (38,39), was isolated (40). It displayed cytochrome c oxidase and TMPD oxidase activity but no quinol oxidase activity, presumably because of the lack of the Rieske iron-sulfur protein.
A supramolecular organization of complexes III and IV was also shown in yeast and bovine mitochondria (41, 42). Thus, quinol oxidase supercomplexes were detected in Gram-positive and Gram-negative eubacteria, in archaea, and in eukaryotes, indicating that this highly organized state is a general feature rather than a specific character of certain species.
Identification of Subunit IV of Cytochrome aa 3 Oxidase-In contrast to the previous purification of cytochrome aa 3 oxidase from C. glutamicum by conventional column chromatography, which resulted in the isolation of subunits I, II, and III (7), our preparation contained an additional protein (CtaF) encoded by the gene downstream of ctaC (Fig. 3). The identification of this protein as a fourth subunit rests on the observation that a C. glutamicum mutant lacking ctaF showed the same growth defect as a ctaD deletion mutant, and like in this strain cytochrome a was almost undetectable. Although CtaF is essential for the formation of a functional cytochrome aa 3 oxidase, it is presumably not required for catalytic activity, because the turnover numbers of the four-subunit complex were in the same range as those of the three-subunit complex (7). Therefore, CtaF is probably involved in the assembly and/or stabilization of cytochrome aa 3 oxidase.
The composition of four subunits is common within the heme-copper family of bacterial terminal oxidases. Subunit IV (CtaH) of cytochrome aa 3 oxidase from P. denitrificans consists of a single transmembrane helix residing in a cleft between subunits I and III (43). Deletion of the ctaH gene had no consequences for the integrity of the complex and its spectral and enzymatic properties (44). Subunit IV (CyoD) of the botype ubiquinol oxidase from E. coli consists of three transmembrane helices and is located between subunits I and III. The third helix is in contact with helix VII of subunit I in the vicinity of the Cu B -heme a 3 binuclear center (45). Deletion analyses indicated that subunit IV is essential for the synthesis of the functional bo 3 oxidase complex and for the Cu B binding to the binuclear center, although it can be removed in vitro without a loss of the enzymatic activity (46). Subunit IV (QoxD) of the cytochrome aa 3 menaquinol oxidase from B. subtilis, like CyoD of E. coli, consists of three transmembrane helices. A mutant lacking the qoxD gene was reported to have decreased respiratory activity and proton pumping activity (47).
CtaF of C. glutamicum shows no significant sequence similarity to CtaH of P. denitrificans, CyoD of E. coli, QoxD of B. subtilis, and CtaF of B. subtilis, and homologs of these proteins were absent in the C. glutamicum genome. In contrast, all actinomycetes with known genome sequence contain homologs of C. glutamicum CtaF (Fig. 6), and the corresponding genes are always clustered with ctaC. Thus, CtaF represents the first member of subunit IV of cytochrome aa 3 oxidase in this group of bacteria.
Electron Transfer between the bc 1 Complex and Cytochrome aa 3 -The identification of a bc 1 -aa 3 supercomplex with quinol oxidase activity supports the assumption that the second heme group of the diheme cytochrome c 1 transfers electrons from the first heme group to the Cu A center in subunit II of cytochrome aa 3 oxidase. The question whether further proteins are involved in this process remains open at present. The proteins P29, P24, and P20 can certainly be excluded, because C. glutamicum mutants lacking these proteins showed no obvious growth defects. In the case of subunit IV (CtaF) of cytochrome aa 3 oxidase, a role in electron transfer is also very unlikely (see above). However, the copurified protein P17 could not yet be identified, and therefore the effect of its absence on the formation and activity of the supercomplex could not be tested.
Besides two covalently bound heme groups, cytochrome c 1 of C. glutamicum has another unusual property, i.e. its weak interaction with the Rieske iron-sulfur protein and cytochrome b. Neither of these two proteins was copurified during the isolation of cytochrome c 1 by conventional chromatographic methods (6), and vice versa, the preparation isolated via streptagged cytochrome b lacked cytochrome c 1 (Fig. 3, lane 4). The presence of QcrC in the supercomplex, purified either via streptagged QcrB or via strep-tagged CtaD, therefore must be because of an interaction with cytochrome aa 3 oxidase or requires interaction with both the bc 1 complex and cytochrome aa 3 oxidase. Further studies are needed to discriminate between these possibilities. Formation of Holo-cytochrome c 1 Is Essential for Assembly and/or Stability of the bc 1 Complex-Mutation of either the N-terminal or the C-terminal CXXCH motif in cytochrome c 1 to SXXSH led to a severe growth defect in the corresponding mutants similar to strain 13032⌬qcr (Fig. 7), which was because of the absence of holo-cytochrome c 1 and drastically reduced levels of cytochrome b. This shows that if any monoheme cytochrome c 1 is formed, it must be rapidly and completely degraded. Similar results have been reported for the diheme cytochrome c subunit (FixP) of the cbb 3 oxidase in B. japonicum (48). The strong effect of the cytochrome c 1 mutations on the cytochrome b level showed that holo-cytochrome c 1 is not only required for electron transfer but also for the maturation and/or stabilization of the entire bc 1 complex. The purification of an apparently stoichiometrical complex of cytochrome b and Rieske iron-sulfur protein from strain ⌬C-B St argues against a role of cytochrome c 1 in stabilization but certainly does not exclude this possibility.
The data from C. glutamicum are in accordance with results from other bacteria, i.e. that deletion of the cytochrome c 1 gene in P. denitrificans (49), Rhodobacter capsulatus (50), and B. japonicum (51), as well as mutation of the heme binding site of B. japonicum cytochrome c 1 (37), caused degradation of cytochrome b and, if tested, also of the Rieske iron-sulfur protein.
According to these data the current model for bc 1 complex maturation predicts that formation of holo-cytochrome c 1 is an early and essential requirement for assembly of the whole complex (52).