Assembly and function of the cytochrome cbb3 oxidase subunits in Bradyrhizobium japonicum.

The Bradyrhizobium japonicum cbb-type cytochrome oxidase, which supports microaerobic respiration, is a multisubunit enzyme encoded by the genes of the fixNOQP operon. We investigated the contribution of the individual subunits to function and assembly of the membrane-bound complex. In-frame deletion mutants of fixN, fixO, and fixQ, and an insertion mutant of fixP were constructed. All mutants, except the fixQ mutant, showed clearly altered absorption difference spectra of their membranes and decreased oxidase activities, and they were unable to fix nitrogen symbiotically. The presence of the individual subunits was assayed by Western blot analysis, using subunit-specific antibodies, and by heme staining of the c-type cytochromes FixO and FixP. These analyses led to the following conclusions: (i) FixN and FixO are necessary for assembly of the multimeric oxidase, (ii) FixN and FixO assemble independently of FixP, and (iii) FixQ is not required for complex formation and, therefore, does not seem to be an essential subunit. The possible oxidase biogenesis pathway involves the formation of a primary core complex consisting of FixN and FixO, which allows the subsequent association with FixP to form the complete enzyme.

Terminal oxidases are the ultimate components of respiratory chains catalyzing the four-electron reduction of molecular oxygen to water. The bacterial respiratory system is usually branched and comprises several terminal oxidases, allowing bacteria to grow at different oxygen tensions. Most of the bacterial oxidases belong to the so-called superfamily of hemecopper oxidases which is divided into two groups, the quinol oxidases and the cytochrome c oxidases (for a review, see Ref. 1). Recently, a new type of heme-copper oxidase with regard to subunit composition and content of prosthetic groups has been discovered, the cbb 3 -type oxidase. Genes encoding this oxidase were isolated first from several rhizobial species (2)(3)(4)(5) and called fixNOQP as they supported symbiotic nitrogen fixation. Meanwhile, homologous genes have also been identified in other Gram-negative bacteria (6,7). The cbb 3 -type oxidase has been purified from several species and characterized biochemically (8 -12). It acts as a cytochrome oxidase, not as a quinol oxidase, and contains three major subunits: a membrane-integral b-type cytochrome (subunit I) with a low-spin heme and a high-spin heme-Cu B binuclear center, and two membrane-anchored c-type cytochromes. Interestingly, there is no Cu A -containing subunit, in contrast to conventional cytochrome c oxidases (8,9).
In Bradyrhizobium japonicum, the fixNOQP-encoded cbb 3type oxidase is expressed only under microaerobic and anaerobic conditions (2). The 61-kDa fixN gene product has up to 14 putative transmembrane helices and binds heme B (2,11). The fixO and fixP genes code for membrane-anchored c-type cytochromes of 28 and 32 kDa, respectively. FixO is a monoheme cytochrome c and FixP probably a diheme cytochrome c. FixQ is a 54-amino-acid polypeptide that is thought to be membranebound by its hydrophobic N-terminal half. Having a K m for O 2 as low as 7 nM, the B. japonicum cbb 3 -type oxidase supports respiration under oxygen-limiting conditions and is, therefore, required for nitrogen fixation in symbiosis with soybean (Glycine max L. Merr) (2,11).
Correct folding and assembly of the individual subunits of a multimeric membrane protein complex may involve a specific nucleation and a precise pathway. Little is known about the assembly of oxidase complexes in bacteria. Minagawa et al. (13) reported the absence of the Escherichia coli cyoABCD gene products when only one of the subunit genes was deleted. Studies on assembly of the Paracoccus denitrificans aa 3 -type oxidase revealed that subunit I was not present in a subunit II gene deletion mutant, whereas subunit III was not absolutely essential for the formation of a subunit I⅐II subcomplex (14,15). Using the B. japonicum cbb 3 -type oxidase system as an example, we attempted to elucidate the role of its individual subunits in assembly and biological function.
Recombinant DNA Work-Standard procedures were used for cloning, Southern blotting and hybridization (18). Chromosomal DNA of B. japonicum was isolated as described previously (19). DNA hybridization probes were radioactively labeled using the nick-translation technique (18). DNA sequence analyses were performed using the chain-termination method (20) and the equipment for automated DNA sequencing (Sequencer model 370A and fluorescent dye terminators from Applied Biosystems, Foster City, CA).
Marker Replacement Mutagenesis and Genetic Complementation-Relevant plasmids used or constructed throughout this work are listed in Table I. For complementation of the fixNOQP operon deletion mutant Bj4503 (11), we constructed plasmid pRJ4504 that contained the entire fixNOQP operon, orf141, and most of orf277 gene (see Fig. 1). The fixN in-frame deletion construct pRJ4526 lacks a 1.2-kilobase pair PvuII-NsiI fragment. To construct the fixO in-frame deletion, a 0.34kilobase pair Eco47III fragment was excised, resulting in pRJ4518. The fixQ in-frame deletion was obtained in plasmid pRJ4528 by eliminating a 0.12-kilobase pair BsaI fragment whose ends were made blunt, and adding a XhoI linker (octamer) to regenerate the ribosome binding site of the downstream gene fixP. The plasmids pRJ4504, pRJ4526, pRJ4518, and pRJ4528 are pSUP202pol6K derivatives.
All of the newly fused sites at which remote DNA was joined after in-frame deletion were confirmed by sequencing. The plasmids pRJ4526, pRJ4518, and pRJ4528 were cointegrated into the Bj4503 chromosome by homologous recombination. Tetracycline-resistant clones were selected (single crossing over) and their DNA examined by Southern blot hybridization for the correct genomic structure. In the case of Bj4526, plasmid cointegration occurred at the fixG locus, i.e. downstream of the operon deletion, whereas the plasmids pRJ4518 and pRJ4528 cointegrated within the orf277orf141 locus, i.e. upstream of the operon deletion.
Construction of Translational lacZ Fusions-The lacZ gene of pNM482Xb was fused to the 7th codon of fixO at the BbsI site (made blunt) that was preceded by either an intact fixN gene or a fixN deletion (described above), resulting in pRJ4569 and pRJ4545, respectively. The fixQЈ-ЈlacZ fusion (to the 15th codon of fixQ) was constructed by ligating at the BsaI site (made blunt) the lacZ gene of pNM481Xb (21). This fusion was constructed either downstream of intact fixNO genes, or downstream of the ⌬fixN or ⌬fixO deletions (described above), yielding pRJ4542, pRJ4543, and pRJ4550, respectively. A translational fusion of the lacZ gene of pNM481X (22) to the 7th codon of fixP was generated at the EcoRI site, yielding pRJ3603. To position the fixPЈ-ЈlacZ fusion into the mutated ⌬fixN, ⌬fixO and ⌬fixQ operons (described above), the lacZ gene of pNM481Xb (21) was cloned into the EcoRI site of fixP, resulting in pRJ4539, pRJ4538, and pRJ4540, respectively. Plasmids pRJ4569, pRJ4545, pRJ4542, pRJ4543, pRJ4550, pRJ4539, pRJ4538, and pRJ4540 are the respective pSUP202pol6K derivatives.
All fusion sites were confirmed by sequencing. While pRJ3603 was cointegrated into the chromosome of the wild type B. japonicum 110spc4 at the fixNOQ locus, the plasmids pRJ4538, pRJ4539, pRJ4540, pRJ4542, pRJ4543, pRJ4545, pRJ4550, and pRJ4569 were cointegrated into the operon deletion mutant Bj4503 by conjugation. Tetracycline-resistant clones were selected and confirmed by Southern blot hybridization. Thus, strains Bj4538, Bj4539, Bj4540, Bj4542, Bj4543, Bj4545, Bj4550, and Bj4569 were obtained that contain the corresponding plasmids integrated at the homologous position in the chromosome.
Antibodies-MalEЈ-ЈFixO and MalEЈ-ЈFixP hybrid proteins were used as suitable antigens. The malEЈ-ЈfixO and the malEЈ-ЈfixP gene fusions (to the 7th codon of fixO and the 6th codon of fixP) were constructed at the BbsI site (made blunt) in fixO or at the EcoRI site in fixP with the malE gene of pMal-cRI plasmid, resulting in pRJ4532 and pRJ3620, respectively. The correct fusion sites were confirmed by sequencing. The MalEЈ-ЈFixO and MalEЈ-ЈFixP proteins were expressed and purified according to the instructions of the manufacturer (New England Biolabs, Schwalbach/Taunus, Germany). The purified fractions were separated by SDS-polyacrylamide gel electrophoresis (PAGE), and the proteins were stained with 0.01% Coomassie Brilliant Blue R250. Gel pieces with the visualized hybrid proteins were excised and used to immunize a New Zealand White rabbit. The MalEЈ-ЈFixP antiserum was used without purification. The MalEЈ-ЈFixO antiserum was purified following the method of Smith and Fischer (23). Antiserum directed against three FixN-specific peptides (PEP1 from position 61 to 73: RYFERPAALPPAE; PEP2 from position 366 to 381: TLSGAWD-KLRTDPVLR; PEP3 from position 533 to 549: RVGEAEVQMPVALQ-PAE) was purchased from TANA Laboratories LC. (Houston, TX). The antiserum against cytochrome c 1 had been described previously (24). Enzymatic Assays-␤-Galactosidase activity was measured from 100 l samples of three independent cultures as described by Miller (25). N,N,NЈ,NЈ-Tetramethyl-p-phenylendiamine (TMPD) 1 oxidase activity was measured with whole cells as described previously (26). Cytochrome c oxidase activity with horse heart cytochrome c as an electron donor was measured as described by Gerhus et al. (27).
Cell Fractionation-B. japonicum cells were harvested in the late exponential growth phase. Membrane fractions were isolated as described elsewhere (26) with the following modification: the membranes were solubilized in 50 mM Tris-HCl, pH 7.5, 1 mM phenylmethylsulfonyl fluoride and 1% dodecyl maltoside (Sigma), and the ultracentrifugation step was repeated. The solubilized oxidase complex remained in the supernatant. Protein concentration, SDS-PAGE according to Laemmli (28) or Schä gger and von Jagow (29), and heme stains were performed as described previously (30,31). Spectroscopy of membrane fractions was performed as described elsewhere (32).
Western Blotting-Proteins were separated by SDS-PAGE and electroblotted onto Hybond-C nitrocellulose (Amersham Corp., Buckinghamshire, UK) or polyvinylidene difluoride (Millipore Corporation, Bedford, MA) filters. The samples were not boiled before loading. For the detection of FixN, proteins were blotted with a semidry blotting apparatus (Semi-Phor , Hoefer Scientific Instruments, San Francisco, CA) with the following buffer system: the cathode buffer contained 25 mM Tris base, 192 mM glycine and 0.06% SDS; the anode buffer 25 mM Tris base, 192 mM glycine and 20% methanol. The transfer of proteins separated in a 1 mm thick gel was carried out at 0.4 mA/cm 2 during 2 h. Cross-reacting proteins were detected by the biochemoluminescence method according to the instructions of the manufacturer (Boehringer Mannheim).
Plant Infection Test-Plant infection tests with soybean (G. max L. Merr) were performed as described previously (19). Nitrogen fixation activity was determined by the acetylene reduction assay (33).

Mutagenesis of the Individual Genes in the fixNOQP
Operon-To investigate the role of each of the four genes in function and biogenesis of the multisubunit enzyme complex, we constructed in-frame deletion mutations in fixN, fixO, and fixQ and used a fixP insertion that had been generated previously (2) (Fig. 1). The idea was to create nonpolar mutations in the first three genes that would not affect transcription and translation of the downstream genes in the operon. In the case of fixN, 394 codons (corresponding to amino acid positions 110 -504) were deleted. From fixO, 114 internal codons (corresponding to amino acid positions 85-199) were removed. In the fixQ deletion the last 39 codons of the gene (amino acid positions 16 -54) were removed. To restore the Shine-Dalgarno sequence of the downstream fixP gene, and to bring the fixQ stop codon in-frame, an XhoI-linker had to be inserted (see "Experimental Procedures"). All of the constructs contained the fixN promoter region (between orf141 and fixN) 2 that drives transcription of the mutated operon. The ⌬fixN, ⌬fixO and ⌬fixQ deletion plasmids were cointegrated into the chromosome of the mutant strain Bj4503 (11) in which the entire fixNOQP operon is deleted (Fig. 1). To make sure that a functional oxidase can be expressed from such cointegrates, a plasmid containing the entire wild type fixNOQP operon was also cointegrated into Bj4503. The resulting strain Bj4504 served as a positive control. Since fixP is the last gene of the operon, an aphII cassette was inserted into it, and the insertion was transferred into the wild type chromosome by marker replacement (strain Bj3618; Control for Nonpolarity of the In-frame Deletions-Expression of the genes downstream of the in-frame deletions was tested. The fixNOQ ϩ fixPЈ-ЈlacZ fusion plasmid (pRJ3603) was cointegrated at its homologous site into the chromosome of the wild type, whereas all other fusion plasmids (Table II) were cointegrated at their homologous sites into the chromosome of the operon deletion strain Bj4503. Thus, expression of the fixOЈ-ЈlacZ fusion was tested cotranscriptionally with either an intact fixN gene or the ⌬fixN deletion. Expression of the fixQЈ-ЈlacZ fusion was tested cotranscriptionally with fixNO ϩ , ⌬fixN-fixO ϩ , or fixN ϩ ⌬fixO. Expression of the fixPЈ-ЈlacZ fusion was tested cotranscriptionally with fixNOQ ϩ , ⌬fixNfixOQ ϩ , fixN ϩ ⌬fixOfixQ ϩ , or fixNO ϩ ⌬fixQ. ␤-Galactosidase activity of microaerobically grown cells was measured because the fix-NOQP operon was found to be maximally expressed under microaerobic growth conditions (data not shown). Table II shows that the fixOЈ-ЈlacZ, fixQЈ-ЈlacZ, and fixPЈ-ЈlacZ fusions were expressed to a similar extent. Thus, the in-frame gene deletion mutations ⌬fixN, ⌬fixO, and ⌬fixQ neither disturbed transcription nor translation of the downstream genes. Notably, the fixPЈ-ЈlacZ fusion in the ⌬fixQ mutant led to a similar ␤-galactosidase activity as that in the wild type, proving that the reconstructed fixP ribosome binding site was functional. The results from these experiments allowed us to conclude that the single-gene mutants were suitable tools to study the role of the individual gene products in respiration and complex assembly.
Presence of the Subunits and Cofactors in the Oxidase-We first looked for the presence or absence of cbb 3 -type oxidase subunits in the ⌬fixN, ⌬fixO, ⌬fixQ, and fixP::aphII mutants by Western blot analysis (Fig. 2), using antibodies specific for FixN, FixO, and FixP (see "Experimental Procedures"). B. japonicum cells were grown microaerobically, and membrane fractions were isolated. The cross-reaction of blotted membrane proteins with immunoglobulins specific for FixN or FixO gave an identical pattern (Fig. 2): the FixN and FixO proteins were present in the wild type (lane 2), in the complemented strain with the entire fixNOQP operon (lane 4), and in the ⌬fixQ and fixP::aphII mutants (lanes 7 and 8). By contrast, they were absent in the operon deletion mutant (lane 3), and in the ⌬fixN and ⌬fixO mutants (lanes 5 and 6). In no case was FixN protein present while FixO protein was absent, or vice versa. The FixP protein pattern behaved in the same way, except that FixP was also absent in the fixP::aphII mutant. Antibodies directed against the membrane-bound cytochrome c 1 were used as a positive control to check if similar amounts of protein were loaded from each strain. Cytochrome c 1 was indeed detected in about the same amounts in all strains except in the bc 1 Ϫ mutant (lane 1). Interestingly, FixN, FixO, and FixP were present at normal levels in this mutant, indicating that the cytochrome bc 1 complex is not required for the presence of the cbb 3 oxidase in membranes.
The membrane protein fractions of the same strains were also tested for heme C incorporation into the FixO or FixP proteins. Proteins were separated by SDS-PAGE and stained for covalently bound heme. Again, the complemented strain served as a positive control. Heme staining revealed four membrane-bound cytochromes c (Fig. 3, lanes 2, 4, and 7): a strongly staining 32-kDa protein that had been identified previously as FixP, the diheme cytochrome c (2) and which was also seen in the bc 1 Ϫ mutant (lane 1); the 20-kDa CycM protein (32,34) that was present in all of the strains (lanes 3, 5, 6, and 8 included), except in the bc 1 Ϫ mutant (lane 1); the 28-kDa FixO protein (lanes 1, 2, 4, 7, and 8; the lower of two bands) which almost comigrates with cytochrome c 1 (lanes 2-8; the upper of the two bands). A comparison between the Western blot (Fig. 2) and heme staining analyses (Fig. 3) clearly showed that the upper band of the 28-kDa double band must be attributed to cyto-chrome c 1 , whereas the lower one represents FixO. This finding is corroborated by the fact that the 28-kDa heme staining double band was completely missing in a bc 1 Ϫ /fixO Ϫ double mutant (data not shown). The presence or absence of FixO and FixP in heme stains completely concurred with their presence or absence in Western blots. Hence, heme C is covalently bound to FixO and FixP in the complemented strain, the bc 1 Ϫ mutant, and the ⌬fixQ mutant; in addition, FixO is also detectable in the fixP::aphII mutant.
The presence of heme B in the oxidase was investigated spectroscopically. Solubilized membrane fractions were used in these experiments. As shown in previous work (11), peaks characteristic for cytochrome b at 558 nm and for cytochrome c at 552 nm were detected in wild type membranes of microaerobically grown cells. The interpretation of dithionite-reduced minus air-oxidized difference spectra of membranes is somewhat complicated by the fact that a number of different b-and c-type cytochromes are present whose absorption maxima overlap in the range of 550 to 560 nm. The spectrum of the operon deletion mutant complemented with the entire fixNOQP operon looked very similar to a wild type spectrum ( Fig. 4;  Bj4504). The mutants showed either weak (Bj4528 ⌬fixQ), intermediate (Bj3618 fixP::aphII) or severe defects (Bj4503 ⌬fix-NOQP, Bj4526 ⌬fixN, Bj4518 ⌬fixO) in cytochrome b and c contents. The spectra of the ⌬fixN and ⌬fixO mutants were almost identical to those of the operon deletion mutant, with a strong decrease and a slight red shift of the cytochrome c peak and a partial decrease of b-type cytochromes. The cytochrome b-specific shoulder was resolved in these mutants into two components, one at 558 nm and the other at 563 nm. The shape of the ⌬fixQ mutant spectrum was similar to that of the wild type; however, the cytochrome c peak appeared somewhat decreased. The fixP::aphII mutant exhibited a spectrum with almost normal amounts of cytochrome b, but strongly decreased levels of cytochrome c. The results indicate that the B-type hemes of FixN were present in the ⌬fixQ and fixP::aphII mutants.
In summary, the results from Western blots, heme stains and visible light spectroscopy implied that (i) complementation of an operon deletion with the wild type fixNOQP operon was successful; (ii) FixN, FixO and FixP were absent in the ⌬fixN and ⌬fixO mutants; (iii) the ⌬fixQ mutation had no effect on the presence of the FixN, FixO and FixP subunits except perhaps a slight reduction in the amount of FixP (Fig. 2, lane 7, and the difference spectrum in Fig. 4); and (iv) FixN and FixO, but not FixP, were present in the fixP::aphII mutant.
Role of the fixNOQP-encoded Proteins in Microaerobic Respiration and Symbiotic Nitrogen Fixation-The functional consequences of the different mutations on respiration was tested    Bj4526 (⌬fixN, lane 5), Bj4518 (⌬fixO, lane 6), Bj4528 (⌬fixQ, lane 7), and Bj3618 (fixP::aphII, lane 8) . Lanes 1, 3, 5, 6, and 8 contained  approximately 200 g, the other lanes 100 g of protein. In the operon deletion and in the ⌬fixN, ⌬fixO, and fixP::aphII mutants, a new, weakly heme-staining band of unknown identity appeared at approximately 26 kDa. The apparent molecular masses of the proteins (kDa) are shown on the right. by using TMPD as an artificial electron donor to measure cytochrome c-dependent oxidase activity from microaerobically grown cells with an oxygen electrode (Table III). The ⌬fixQ strain had a similar TMPD oxidase activity as the wild type. The operon deletion mutant (Bj4503) showed a 25% residual oxidase activity, indicating that the cbb 3 oxidase is responsible for about 75% of the total oxidase activity present in cells grown under microaerobic conditions. The ⌬fixN, ⌬fixO and the fixP::aphII mutants also showed decreased oxidase activities (18, 24.5, and 38%, respectively, as compared with the wild type). The fixP::aphII insertion mutant consistently had a slightly higher oxidase activity than the ⌬fixN and ⌬fixO deletion mutants. It thus appeared as if the FixN and FixO subunits alone are responsible for a small amount of TMPD oxidase activity. Interestingly, the cytochrome bc 1 Ϫ mutant showed almost wild type activity, suggesting that in the absence of the bc 1 complex, the cbb 3 -type oxidase can still oxidize TMPD as a substrate. In addition to the tests shown in Table  III we measured cytochrome c oxidase activity with reduced horse heart cytochrome c as electron donor, and similar results as with TMPD were obtained (not shown).
As shown previously (2), the cbb 3 -type oxidase is essential for symbiotic nitrogen fixation because it supports energy conservation under the extremely microaerobic conditions that prevail in soybean root nodules. Thus, we analyzed the ability of the single-gene deletion mutants to fix nitrogen in symbiosis (Fix phenotype; Table III). Soybean seedlings were inoculated with the B. japonicum mutant strains, and nitrogenase activity was measured as acetylene reduction activity 23 days after infection. While the fixNOQP ϩ strain and the ⌬fixQ mutant were Fix ϩ , the ⌬fixN, ⌬fixO and fixP::aphII mutants showed between 0 and 5% residual Fix activity. Nodules infected with the latter mutant strains were greenish inside, indicating the absence of functional leghemoglobin. Our results suggest that the FixN, FixO and FixP proteins, but not the FixQ protein, are essential to energetically support nitrogen fixation in root nodules.

DISCUSSION
To study the pathway for assembly of multisubunit membrane complexes is inherently difficult. Lack of a single subunit often causes complete absence of the entire complex because membrane proteins may be rapidly degraded unless they find their correct partner subunit(s) to become stabilized by proteinprotein interaction. For examples, the assembly of the F 1 F 0 -ATPase of E. coli relies on the presence of both ␣ and ␤ subunits (35), and cytochrome c 1 of the Rhodobacter capsulatus and P. denitrificans ubiquinol-cytochrome c oxidoreductase is required to maintain the stability of the other two subunits (cytochrome b, Rieske iron sulfur protein) (27,36).
Here we report on studies to elucidate how assembly of the B. japonicum cbb 3 -type cytochrome oxidase takes place. The approach was to knock out separately each of the four genes in the fixNOQP operon, and then we assessed the presence or absence of the other gene products in the cytoplasmic membrane. It was mandatory to construct in-frame deletions and ensure that genes located downstream of the deleted gene were expressed at normal levels. In fact, none of the deletions showed any polarity effects on the downstream genes.
The mutations analyzed here affected cytochrome cbb 3 formation to different extents, allowing us to draw conclusions with respect to subunit composition, function and assembly. Surprisingly, the ⌬fixQ mutation did not cause any severe deficiencies even though (i) a fixQ-like gene is conserved in all organisms examined for the presence of fixNOQP-like operons (2-6) and (ii) fixQ was shown here to be transcribed and translated. Hence, it is clear that the 54-amino acid FixQ protein is not essential for the formation and function of the cbb 3 oxidase complex. However, this does not exclude the possibility that FixQ is bound to the oxidase, perhaps even in substoichiometric amounts. The question of whether or not FixQ is part of the complex will be answered only when substantial quantities of highly purified cytochrome cbb 3 have become available. The partially purified oxidase preparation reported recently (11) was not of sufficient quantity to identify a protein of the expected molecular mass of FixQ, i.e. approximately 6 kDa. In this context, we find it intriguing that the recent three-dimensional structure determination of P. denitrificans cytochrome aa 3 has uncovered the presence of an unprecedented 60-amino acid peptide whose functional role is also not known (37).
The FixN and FixO proteins, as opposed to FixQ, are absolutely essential for the function and assembly of cytochrome cbb 3 . Neither FixN and FixO nor the FixP protein were detectable in membranes isolated from ⌬fixN and ⌬fixO mutants. This supports previous biochemical data obtained with purified FIG. 4. Visible absorption difference spectra (dithionite-reduced minus air-oxidized). Membrane proteins (0.5 mg ml Ϫ1 ) solubilized with dodecyl maltoside from the following B. japonicum strains were tested: Bj4504 (fixNOQP ϩ by complementation), Bj4503 (⌬fix-NOQP), Bj4526 (⌬fixN), Bj4518 (⌬fixO), Bj4528 (⌬fixQ), and Bj3618 (fixP::aphII). cbb 3 oxidase from B. japonicum and two Rhodobacter species (8,9,11) which showed that it is a three-subunit enzyme complex consisting of FixN-, FixO-, and FixP-like proteins. Interestingly, in the B. japonicum fixP insertion mutant we detected almost wild type levels of the FixN and FixO proteins, whereas only FixP itself was absent. Moreover, the mutant had wild type levels of cytochrome b, but a decreased level of cytochrome c probably due to the absence of FixP. Notably, FixO was shown by heme-staining after SDS-PAGE to be present as mature holocytochrome c in the fixP mutant. The results are reminiscent of studies with the cytochrome bc 1 complex, in which cytochromes b and c 1 were still detectable in a mutant defective in the gene for the Rieske iron-sulfur protein (36). Taken together, our results led us to propose an ordered biogenesis pathway for the cbb 3 -type cytochrome oxidase. After translation of the individual subunit polypeptides from the fixNOQP mRNA, FixN and FixO are probably the first to be inserted into the membrane to form an apparently stable FixNO core complex. This primary nucleation step occurs also in the absence of FixP and, therefore, precedes assembly of the FixNO core with the FixP protein. The fact that not even traces of FixN plus FixP and FixO plus FixP are detectable in ⌬fixO and ⌬fixN mutants, respectively, suggests that the subunits are degraded extremely rapidly when assembly in the membrane is not possible for lack of cognate, stabilizing protein partners.
In the aerobic respiratory chain of B. japonicum the bc 1 complex appears to be a nucleation center for the formation and stable maintenance of the subsequent electron transport cytochromes CycM and aa 3 , which are thought to be organized in a bc 1 ⅐CycM⅐aa 3 supercomplex (24,30,32). By contrast, the Fix-NOP proteins do assemble also in the absence of, and therefore independently from, the bc 1 complex.
The phenotypical analyses of the mutants also help toward understanding the organization of the microaerobic electron transport chain in B. japonicum. It is now well established that under microaerobic conditions electrons are transferred from ubiquinol to the bc 1 complex that is also used in the aerobic respiratory chain (30,32). The Fix Ϫ phenotype of a fixNOQP mutant has been interpreted to mean that the oxidase encoded by these genes participates in the microaerobic branch of respiratory electron transport (2). This is strongly supported by the high affinity for O 2 (K m ϭ 7 nM) of this oxidase (11). Moreover, the structural predictions made for the FixN protein imply that it corresponds to a heme-copper oxidase subunit I, the component where oxygen is reduced. The two other components, FixO and FixP, are likely to mediate electron transfer between the bc 1 complex and the FixN oxidase subunit. Midpoint redox potentials have been determined for subunits of the homologous R. capsulatus cbb 3 -type oxidase. Values of 265 mV and 320 mV were determined for the 32-and 28-kDa c-type cytochromes, the ccoP and ccoO gene products (8,9). CcoP, and by analogy FixP, is therefore likely to mediate electron transfer between the bc 1 complex and CcoO (FixO). This would be a role comparable to that of the membrane-bound CycM protein in the aerobic respiratory branch of B. japonicum (32,34). Since FixO is closely associated with FixN, it is the best candidate to donate the electrons directly to FixN. In the case of the mi-croaerobic respiratory branch, the electrons may be transferred from the bc 1 complex to FixP, and from there to FixO and FixN, where they are finally used for the reduction of molecular oxygen to water.