Assembly of Respiratory Complexes I, III, and IV into NADH Oxidase Supercomplex Stabilizes Complex I in Paracoccus denitrificans*

From the ‡Zentrum der Biologischen Chemie, Universitätsklinikum Frankfurt, D-60590 Frankfurt, Germany, the §Institut für Biochemie, Universität Frankfurt, D-60439 Frankfurt, Germany, the ¶Department of Molecular and Experimental Medicine, The Scripps Research Institute, La Jolla, California, and the Viikki Drug Discovery Technology Center, Department of Pharmacy, University of Helsinki, FIN-00014 Helsinki, Finland

With mitochondria, eukaryotic cells possess specialized organelles dedicated mainly to oxidative phosphorylation. Inner mitochondrial membranes of higher eukaryotes like mammalia are highly enriched for the four respiratory chain complexes, NADH:ubiquinone oxidoreductase (complex I), succinate: ubiquinone oxidoreductase (complex II), ubiquinol:cytochrome c oxidoreductase (complex III), and cytochrome c oxidase (complex IV), which generate an electrochemical potential across this membrane. F 1 F O -ATP synthase (complex V) uses this electrochemical proton gradient to synthesize ATP (1).
Bacterial respiratory complexes are located in the cytoplasmic membrane. Paracoccus denitrificans, a Gram-negative soil bacterium, uses an electron chain for aerobic growth that comprises a full complement of mitochondrial respiratory complexes I-IV plus ATP synthase. In context of the endosymbiotic theory (2), proteobacteria like P. denitrificans are therefore regarded as likely ancestors of present day mitochondria (3,4). In addition to its "canonical" respiratory chain, respiration in P. denitrificans is characterized by many branching points. Alternative oxidases (ba 3 quinol oxidase and cbb 3 cytochrome oxidase) allow growth under strongly varying oxygen concentrations. Additionally, Paracoccus may use a variety of electron donors (e.g. hydrogen) or nitrate as terminal electron acceptor, which leads to high environmental flexibility (5).
Bacterial complexes usually comprise a substantially lower number of subunits, e.g. 14, 3, and 4 subunits for P. denitrificans complexes I, III, and IV (6 -9), respectively, compared with 46, 11, and 13 subunits for the corresponding bovine complexes (10 -13). Association of respiratory chain complexes to supercomplexes was observed in mitochondrial (14,15) and bacterial respiratory chains (16 -21). Although stable interactions of bacterial complexes III and IV to form quinol oxidases have been reported, participation of complex I in supercomplex formation escaped detection so far. This may largely be due to a pronounced detergent sensitivity of complex I from various bacteria. So far, complex I could only be isolated from four bacteria: Escherichia coli (22), Rhodothermus marinus (23), Klebsiella pneumoniae (24), and Aquifex aeolicus (25).
We found that digitonin can retain a supramolecular assembly of NADH oxidase from P. denitrificans and isolated for the first time chromatographically a complete "respirasome" comprising complexes I, III, and IV in a 1:4:4 stoichiometry that is suitable for detailed structural and functional analyses.
Enzymatic Analyses-All of the enzymatic assays were performed at 30°C.

FIG. 1. Separation of respiratory chain complexes and supercomplexes by BN-PAGE.
A, P. denitrificans membranes were solubilized using a digitonin/protein ratio of 2 (g/g), and the native membrane protein complexes were separated using a 4 -13% acrylamide gradient gel. Band a was identified as NADH oxidase supercomplex (I 1 III 4 IV 4 ) containing complexes I, III, and IV in a 1:4:4 stoichiometry (cf. "Results"). Bands b and c also contained a core structure of tetrameric complex III associated with four or two copies of complex IV but missed complex I. Bands d and e correspond to complete tetrameric complex III and a subcomplex missing the Rieske iron sulfur protein and cytochrome c 552 , respectively. V, complex V or ATP-synthase; IV, individual complex IV. Oxidative phosphorylation complexes I-V from bovine heart mitochondria were used for molecular mass calibration (not shown; Ref. 34). B, Western blot after two-dimensional resolution by Tricine-SDS-PAGE using a 10% acrylamide gel. A mixture of antibodies against subunits of complex III (cyt. c 1 , cytochrome c 1 ; cyt. b, cytochrome b), of complex IV (Cox I, subunit 1; Cox II, subunit 2), and against cytochrome c 552 (cyt. c 552 ) were used to localize individual respiratory chain complexes and supercomplexes. Black arrows mark subunits cytochromes b and c 1 of the minor amounts of dimeric complex III. Nonspecific interaction of the antibody mixture with the ␣ and ␤ subunits of ATP synthase is indicated by arrowheads on a white background. For immunodetection of complex I in band a see Fig. 5A.

RESULTS AND DISCUSSION
Identification and Purification of Respiratory Chain Supercomplexes-P. denitrificans membranes were solubilized by digitonin and separated by BN-PAGE. A major band representing monomeric complex V (M app ϭ 530 kDa) and several additional bands with apparent masses in the range 460 -1900 kDa were detected (Fig. 1A, bands a-e). Assignment of these bands was possible after SDS-PAGE in a second dimen-sion and immunological detection (Fig. 1B). Band d contained cytochromes c 1 and b and the Rieske iron sulfur protein (separate identification not shown), suggesting that complex III was fully assembled. Some cytochrome c 552 was bound in addition, indicating that cytochrome c 552 can bind not only to complex IV (described below) but also to complex III. Band e was identified as a subcomplex of complex III missing the Rieske iron sulfur protein and cytochrome c 552 . The apparent masses of bands d and e from BN-PAGE suggest that fully assembled complex III and also the subcomplex are tetrameric in P. denitrificans. Minor amounts of dimeric com-  plex III, which is the minimal structural unit of cytochrome bc 1 complexes (45)(46)(47)(48), were identified by immunological detection (Fig. 1B, arrows).
Assignment of bands a-c was facilitated by the analysis of chromatographically isolated NADH oxidase supercomplex that was found to correspond to band a. It contained monomeric complex I, four copies of complex III, four copies of complex IV, and approximately two molecules of cytochrome c 552 (described below). The smaller complexes b and c were devoid of complex I (see Fig. 5A). They contained tetrameric complex III and four and two copies of complex IV, respectively, as deduced from the apparent mass in BN-PAGE (Table I) and densitometric analyses of Coomassie-stained two-dimensional gels (not shown). Because other species containing one or three copies of complex IV were not detected, this suggested that dimeric complex IV is bound to the supercomplexes, although both the structural and functional units have been described as monomers (49).
Evidence for similar supercomplexes was obtained in comparable experiments using Triton X-100 and dodecylmaltoside. However, a stable NADH oxidase supercomplex could only be purified from digitonin-solubilized membranes.
Chromatographic purification, as described under "Experimental Procedures," involved three major steps: solubilization/ centrifugation, hydroxylapatite chromatography, and gel filtration. The elution profile from the hydroxylapatite column and corresponding BN-PAGE is shown in Fig. 2A. More than 70% of the applied protein including ATP synthase (complex V) passed the column unbound. Yet all of the respiratory chain complexes were almost quantitatively bound and eluted together. The enzymatic activity elution profile of a subsequent gel filtration and the corresponding BN-PAGE (Fig. 2B) indicated that the NADH oxidase associate of complexes I, III, and IV eluted first, followed by assemblies of complexes III and IV and individual complexes V, II, and IV. Table II summarizes the purification of this NADH oxidase supercomplex. Complex II clearly was not associated with NADH oxidase. Only fractions 4 -6 from gel filtration were used for further spectrometric and enzymatic analyses.
NADH oxidase supercomplex might comprise some protein constituents in addition to components of complexes I, III, and IV (molecular mass ϭ 1.55 MDa) that have not been identified so far. In this case this potential extra mass should total approximately 350 kDa, because the apparent mass of the holocomplex in BN-PAGE was approximately 1.9 MDa. The cytochrome b and cytochrome aa 3 contents of the novel NADH oxidase supercomplex preparation were 0.71 and 0.80 mol/g, respectively. This is far below the theoretical values of 2.6 or 2.1 mol/g based on a stoichiometry of 4 mol/mol supercomplex for both cytochromes and the calculated mass (1.55 MDa) or the apparent mass from BN-PAGE (1.9 MDa), respectively. This substantial deviation can only be partly explained by protein impurities, because the subunit composition of NADH oxidase isolated chromatographically or by BN-PAGE was very similar, except for some additional bands present in the chromatographic preparation (Fig. 3). However, it seems more likely that the protein elution profile of Fig. 2B reflected an unknown non-protein contaminant most prominent in fractions 5-9, which interfered with the Lowry protein determination method. Digitonin and lipids can be excluded as this unknown non-protein contaminant, because digitonin and extracted Paracoccus lipids were tested not to interfere with the Lowry method. However, less than 1% digitonin in the sample does considerably interfere with the Biuret and Bradford methods, and more than 1% digitonin also interferes with the BCA method.
Ratio of Complexes in NADH Oxidase Supercomplex-The ratios of complexes were determined by fluorometric quantification of flavin mononucleotide as a marker for complex I and from pyridine hemochrome spectra (Table III). Calculation of the ratio is based on ratios of one flavin mononucleotide/complex I, two hemes a/complex IV, two hemes b, and one c-type heme (cytochrome c 1 )/complex III (50). Cytochrome c 552 concentration was calculated from the total heme c content (cytochromes c 1 plus c 552 ) considering that cytochrome c 1 concentration should equal the cytochrome b concentration, which is half the heme b concentration. Complexes I, III, and IV were found to be present in a 1:4:4 ratio. Cytochrome c 552 was substoichiometric to complex IV (approximately 1.6 Ϯ 0.8/4). Assuming an initial 1:1 ratio, this would indicate that 40 -80% of total cytochrome c 552 was lost during isolation.
Enzymatic Activities of Isolated NADH Oxidase Supercomplex-Turnover number of cytochrome c oxidase was 160 Ϯ 15 s Ϫ1 (n ϭ 3) in isolated NADH oxidase, compared with 168 Ϯ 15 s Ϫ1 with digitonin-solubilized membranes assuming that all spectral absorption at 603 nm in the membranes used was due to cytochrome aa 3 . These matching data seemed to indicate that alternative enzymes like ba 3 quinol oxidase and cbb 3 oxidase were not present in considerable amounts in the aerobically grown cells used. Turnover number of DBH:cytochrome c oxidoreductase (complex III) was 95 Ϯ 14 s Ϫ1 (n ϭ 3) in isolated NADH oxidase complex. Spectral absorption of complex III at approximately 560 nm in membranes may be superimposed by several other b-type cytochromes, e.g. of succinate dehydrogenase (51). Therefore, complex III activity in membranes cannot be quantified reliably. NADH:DBQ oxidoreductase (complex I) turnover number in digitonin-solubilized membranes could not be calculated on the basis of the extractable flavin mononucleotide in membranes (0.096 mol flavin/g; n ϭ 6), because extractable flavin apparently originated also from proteins other than complex I, and calculated numbers were approximately 70 s Ϫ1 compared with 186 Ϯ 26 s Ϫ1 for the isolated NADH oxidase complex. However, by comparing deamino-NADH:DBQ oxidoreductase activity and deamino-NADH:hexamino-ruthenium oxidoreductase activity, a stable catalytic activity of the hydrophilic part of complex I that is not altered by a potential dissociation of the hydrophobic part, it was possible to test whether the complex I turnover number was altered or unchanged during isolation of NADH oxidase (52). Deamino-NADH was used instead of NADH to exclude potential alternative dehydrogenases, although alternative dehydrogenases do not seem to exist in P. denitrificans (53). The deamino-NADH:hexaminoruthenium/deamino-NADH:DBQ ratio was 3.9 Ϯ 0.2 in membranes and 4.6 Ϯ 0.3 in isolated NADH oxidase supercomplex, indicating that no considerable changes occurred during isolation.
NADH:ubiquinone oxidoreductase (complex I) activity of isolated NADH oxidase supercomplex was 2.24 Ϯ 0.08 mol min Ϫ1 mg Ϫ1 . This is the maximal activity theoretically obtainable for coupled activities of complexes I, III, and IV, because activities of complexes III and IV were significantly higher because of the 4-fold excess of the corresponding enzymes to complex I (see turnover numbers above and Table III). NADH: cytochrome c oxidoreductase (complexes I and III) activity of the same protein sample (1.51 Ϯ 0.01 mol min Ϫ1 mg Ϫ1 ) indicated almost optimum electron transfer from complex I to complex III, in agreement with high endogenous phospholipid and quinone contents required for electron transfer. Based on the Lowry protein determination, the phospholipid content decreased from 980 Ϯ 80 nmol/mg in membranes to 290 Ϯ 10 nmol/mg in isolated NADH oxidase, whereas the ubiquinone content increased from 2.7 Ϯ 0.3 nmol/mg in membranes to 6.7 Ϯ 0.8 nmol/mg in isolated NADH oxidase. The 8-fold increase of the quinone/phospholipid ratio seems to indicate specific enrichment of ubiquinone in the NADH oxidase supercomplex. A ubiquinone/supercomplex ratio of 10 mol/mol was calculated based on the Lowry protein determination. Considering a potential interference of a non-protein contaminant with the Lowry method as described above, the ratio can approach a value of 30 mol/mol. However, specific NADH oxidase activity (complexes I, III, and IV) was reduced by 74% to 0.53 Ϯ 0.07 mol min Ϫ1 mg Ϫ1 . Similarly, DBH oxidase (complexes III and IV) was reduced by 70% to 1.19 Ϯ 0.01 mol min Ϫ1 mg Ϫ1 compared with 4.07 Ϯ 0.12 mol min Ϫ1 mg Ϫ1 for DBH:cytochrome c oxidoreductase (complex III). This seems to indicate that the transfer of electrons between complexes III and IV was impeded, presumably by partial loss of cytochrome c 552 , as described above.
Assembly into NADH Oxidase Supercomplex Stabilizes Complex I-Digitonin is one of the mildest detergents known. However, it was not clear whether retention of the complex I integrity in isolated NADH oxidase was due to the particular detergent properties or to an additional stabilizing role of supercomplex formation. To address this question, we compared respiratory chain activities in digitonin-solubilized membranes and the stability of complex I from wild-type and mutant strains during chromatography and BN-PAGE.
Complex III activity was not adversely affected by the loss of cytochrome c 552 or complex IV in the corresponding mutant strains (Fig. 4). However, the loss of complex III in the complex III mutant strain led to considerable reduction of complex I and IV activities, indicating that complex III is required for assembly/stability of both complexes. A moderate reduction of complexes I and IV was also observed in the mutant strain lacking cytochrome c 552 , indicating that also cytochrome c 552 favors assembly/stability of the two complexes.
Complex I activity was substantially lower in all mutant strains compared with wild type, which might be taken as a hint for impeded assembly or stability of complex I in these membranes. Although all complex I activities were stable for 60 min after solubilization using low digitonin/protein ratios, reduced stability of complex I in the mutant strains lacking complexes III or IV became apparent from the subsequent analyses.
Attempts to isolate complex I from mutant strains lacking complexes III or IV led to an almost complete loss of NADH: DBQ oxidoreductase activity when the same isolation protocol as for parental strain was applied (results not shown). Even omission of the hydroxylapatite step that requires use of low Triton X-100 concentrations for protein elution and direct gel filtration using digitonin was not successful. These findings seem to indicate that complex I from these mutant strains is sufficiently stable at the low digitonin/lipid ratios used for solubilization, but stability is reduced under the conditions of chromatography. Because the presence of only one of the two complexes in the corresponding mutant strains could not protect complex I from inactivation during chromatography, we conclude that complex I requires assembly into a complete NADH oxidase complex for optimal stability.
Reduced stability of complex I in these mutant strains was also apparent from the complete dissociation of complex I under the conditions of BN-PAGE. Following two-dimensional SDS-PAGE and electroblotting, a mixture of specific antibodies was used to identify the location of assembled complexes and dissociated subunits. Using parental strain Pd1222 (Fig. 5A), the antibodies identified supercomplexes a, b, and c and individual complexes III and IV, but intact individual complex I was not present. Dissociated complex I subunits NQO 3 and 1 were found at the running front of BN-PAGE, i.e. at the righthand side of the two-dimensional gel. At present we cannot discriminate between two possibilities. The dissociated subunits might either originate from individual complex I if larger fractions of it were in equilibrium with supercomplexes b and c. On the other hand, almost all complex I might initially assem-ble into NADH oxidase supercomplex, which partially disintegrates during BN-PAGE.
Using the complex III deletion strain (Fig. 5C), individual complex IV was detected but no assembled complex I. This seems to indicate that all complex I that was functional after solubilization by digitonin dissociated under the conditions of BN-PAGE. Analysis of the complex IV deletion strain (Fig. 5D) led to a similar result, except that stable tetrameric complex III was identified.
The proteins involved in the interaction of complexes III and IV are not known. Tightly bound cytochrome c 552 that possesses a transmembrane anchor (31) initially was regarded as a candidate linker protein, because it binds preferentially to complex IV, and minor amounts were also found associated with complex III. Analyzing the strain carrying an inactivated cycM gene coding for cytochrome c 552 (Fig. 5B) indicated that this electron carrier was not essential for  formation of supercomplexes a, b, and c. However, the amounts of supercomplexes were reduced, and the amounts of dissociated subunits of complex I were increased, indicating that cytochrome c 552 favors supercomplex formation and indirectly stabilizes complex I.
Summarizing, we conclude that detergent-labile complex I from P. denitrificans is protected by supercomplex formation. Deletion of complexes III and IV causes decreased complex I contents in membranes, suggesting altered assembly/stability of complex I that is not assembled into a complete NADH oxidase supercomplex also in membranes. However, detergentstable NADH dehydrogenases, like E. coli complex I, do not require supercomplex formation. In fact, E. coli does not possess complex III. Similarly, complexes I from the yeast Yarrowia lipolytica, and the hyperthermophilic eubacterium A. aeolicus are rather stable (25,54), and no respiratory chain supercomplexes could be detected. Structural stabilization of a labile membrane protein complex seems to be a major function of supercomplex formation, in addition to substrate channeling, which is easily envisaged for the interaction of complexes III and IV in P. denitrificans via cytochrome c 552 but also seems to play a role for the interaction of complexes I and III via ubiquinone, as recently reported for the bovine respirasome (55).