The Elusive Third Subunit IIa of the Bacterial B-Type Oxidases: The Enzyme from the Hyperthermophile Aquifex aeolicus

The reduction of molecular oxygen to water is catalyzed by complicated membrane-bound metallo-enzymes containing variable numbers of subunits, called cytochrome c oxidases or quinol oxidases. We previously described the cytochrome c oxidase II from the hyperthermophilic bacterium Aquifex aeolicus as a ba 3-type two-subunit (subunits I and II) enzyme and showed that it is included in a supercomplex involved in the sulfide-oxygen respiration pathway. It belongs to the B-family of the heme-copper oxidases, enzymes that are far less studied than the ones from family A. Here, we describe the presence in this enzyme of an additional transmembrane helix “subunit IIa”, which is composed of 41 amino acid residues with a measured molecular mass of 5105 Da. Moreover, we show that subunit II, as expected, is in fact longer than the originally annotated protein (from the genome) and contains a transmembrane domain. Using Aquifex aeolicus genomic sequence analyses, N-terminal sequencing, peptide mass fingerprinting and mass spectrometry analysis on entire subunits, we conclude that the B-type enzyme from this bacterium is a three-subunit complex. It is composed of subunit I (encoded by coxA2) of 59000 Da, subunit II (encoded by coxB2) of 16700 Da and subunit IIa which contain 12, 1 and 1 transmembrane helices respectively. A structural model indicates that the structural organization of the complex strongly resembles that of the ba 3 cytochrome c oxidase from the bacterium Thermus thermophilus, the IIa helical subunit being structurally the lacking N-terminal transmembrane helix of subunit II present in the A-type oxidases. Analysis of the genomic context of genes encoding oxidases indicates that this third subunit is present in many of the bacterial oxidases from B-family, enzymes that have been described as two-subunit complexes.


Introduction
In aerobic respiration of prokaryotic and eukaryotic organisms, the reduction of molecular oxygen to water is catalyzed by terminal oxidases which are integral membrane multi-subunit enzymatic complexes belonging to the heme-copper oxidases superfamily [1]. These respiratory enzymes, which also pump protons across membrane, have been named cytochrome c oxidases or quinol oxidases, depending on the electron donor that they oxidize. A classification of these enzymes into three families was proposed by Pereira et al. [2]: i) type A (mitochondriallike oxidases), ii) type B (ba 3 -type oxidases) and iii) type C (cbb 3 -type oxidases). More recently, Hemp and Gennis identified five novel oxygen reductase families found exclusively in Archaea [3]. The subunit composition of types A and B enzymes differs from one oxidase to another but these heme-copper oxidases always contain the catalytic subunit (subunit I) and a smaller subunit named subunit II. Subunit I is an integral membrane protein composed of at least 12 transmembrane helices, containing the heme-copper binuclear centre Cu B and its immediate electron donor, a low-spin heme. Subunit II of oxidases A and B is made of a transmembrane domain and a peripheral one in which the copper Cu A centre is located. This subunit binds the electron donor (reduced cytochrome c, but also other metalloproteins) and transfers electrons to subunit I [2].
Aquifex aeolicus (A. aeolicus) belongs to the Aquificales order of Bacteria and is a hyperthermophilic chemolithoautotrophic microorganism using molecular hydrogen as electron donor and molecular oxygen as electron acceptor in the presence of a sulfur compound. The Aquifex genome sequence analysis revealed the presence of genes coding for enzymes potentially involved in oxygen reduction: two cytochrome c oxidases and a cytochrome bd quinol oxidase [4,5]. Based on sequence comparison, Pereira et al. proposed that cytochrome c oxidase I from A. aeolicus (putatively encoded by coxA 1 , coxB and coxC) belongs to type A and has all the specific residues of the D-and K-protons channels, whereas cytochrome c oxidase II (encoded by coxA 2 and coxB 2 ) is a B-type enzyme with specific residues forming a K-channel homologue for proton translocation [2]. Our previous work on A. aeolicus membranes showed the presence of the bd quinol oxidase and the cytochrome c oxidase II when the bacterium grows with molecular hydrogen, molecular oxygen and elemental sulfur [5]. We have recently described a new multienzyme supercomplex carrying a sulfide oxidase-oxygen reductase activity that contains the sulfide quinone reductase (Sqr), the dimeric bc 1 complex and the cytochrome c oxidase II [6]. We demonstrated that this latter enzyme is a ba 3 -type oxidase and that it is composed of at least the catalytic subunit I and the small subunit II [6], in agreement with its classification in the B family, as proposed by Pereira [2]. However, a novel small subunit belonging to the type B ba 3 cytochrome c oxidase from Thermus (T.) thermophilus, called subunit IIa, was discovered in 2000 with the resolution of the crystal structure of this complex, implying that this B-type enzyme is thus a three-subunit oxidase [7,8]. This novel subunit has 34 amino acid residues and forms a single helix across the membrane, corresponding in space to the first transmembrane helix of subunit II of the A-type cytochrome c oxidases [2,7,8].
Concerning the ba 3 cytochrome c oxidase from A. aeolicus, there were some points that were intriguing to us: on the one hand the very small theoretical molecular mass of subunit II (about 9000 Da), and on the other hand the lack of a transmembrane segment for this protein as found in the NCBI database (accession NP_214505.1). Moreover, the presence of a third subunit in the T. thermophilus enzyme prompted us to characterize in more detail the subunit composition of the ba 3 cytochrome c oxidase from A. aeolicus. We now demonstrate the existence of a third small helical subunit IIa in the oxidase complex belonging to the B family in A. aeolicus. This enzyme is thus only the second in which this subunit has been biochemically evidenced. Furthermore, we propose that the presence of this small protein is common in type B of oxidases, at least in the bacterial organisms.

A. aeolicus genomic sequence analysis
Subunit II: detection of an error in the genomic sequence. We previously showed that cytochrome c oxidase II, belonging to the family B of oxidases, is synthesized in the membranes of A. aeolicus in our growth conditions and that it is a ba 3 enzyme [5,6]. We decided to check the genomic deoxyribonucleic acid (DNA) sequence for a potential error of sequencing. As in the oxidase cluster grouping 7 genes, a relatively large non-coding nucleotide sequence stands between coxA 1 and coxB 2 genes ( Figure 1, nucleotides 1541640 to 1542360), this region has been carefully analyzed (with potential open reading frame (ORFs) detection and sequence comparison). A probable insertion of one nucleotide at the position 1542190 caused a frame-shift. We confirmed this error in the genomic sequence by sequencing the A. aeolicus DNA located between oligonucleotide pairs (centred on positions 1541997 and 1542342, Figure 1). This led to the identification of a possible new start for the coxB 2 gene, coding for a putative longer protein (extended at its N terminal part with 64 amino acid residues) with a theoretical molecular mass of 16700 Da which is compatible with the experimental one (see next section). A BLAST search using the corrected protein sequence indicates high similarity with proteins, annotated cytochrome c oxidase subunit II, from bacteria belonging to Aquificales order (Hydrogenivirga sp., Thermocrinis albus, and Hydrogenobacter thermophilus, 74%, 61% and 56% identity respectively) (Supplementary Figure S1). Some similarities were also found with other bacterial and archaeal proteins, including the characterized ba 3 cytochrome c oxidases from T. thermophilus (31% identity) [7], Rhodothermus (R.) marinus (36% identity) [9] and Geobacillus (G.) stearothermophilus (36% identity) [10,11]. Sequence comparison of subunit II shows that: (1) length of A. aeolicus subunit II is in the same range of oxidases from the B-family, and (2) amino acid residues involved in the binuclear copper Cu A centre coordination (His 96, Cys 131, Cys 135, His 139, Met 142) are conserved in the A. aeolicus sequence. Moreover, a prediction of secondary structures for the corrected subunit II sequence indicates the presence of only one transmembrane helix which constitutes the domain of insertion of the protein in the membrane as described for subunit II from B-type enzymes (instead of two helices found in the A-type oxidases) [2,7,8,10,12] ( Figure S1). This transmembrane domain was previously not described because of the error in the genomic sequence [2].
Subunit IIa: detection of an additional subunit. The gene coding for the small subunit IIa (cbaD) of T. thermophilus is located directly upstream from that for the subunit II in the operon coding for the ba 3 oxidase [8]. We thus decided to search for a potential third subunit gene that would be located upstream the one of subunit II, in the genome of Aquifex. Directly upstream of coxB 2 gene, we found a small ORF encoding a putative protein of 41 amino acid residues with a theoretical monoisotopic molecular mass of 5104.46 Da ( Figure 1, nucleotides 1541953 to 1542075). A search with the BLAST program using this sequence indicates similarities with very few proteins: two proteins from the Aquificales Hydrogenobacter thermophilus and Thermocrinis albus and one annotated cytochrome c oxidase subunit IIa from the bacterium Meiothermus silvanus (in cox operon in the three cases). This potential small subunit is thus probably the third subunit, subunit IIa of the ba 3 cytochrome oxidase. We compared, with a sequence alignment, the novel A. aeolicus sequence of subunit IIa with the one from subunit IIa from T. thermophilus (34 amino acid residues). This shows 60% of residues with similar chemical properties (including 30% identity) between the two peptides ( Figure 2). The prediction indicates that this small subunit is probably a transmembrane helix. Taken together, all these results support the existence of a small helical third subunit in the ba 3 cytochrome c oxidase from A. aeolicus.
Experimental evidence for the A. aeolicus subunits II and IIa and structural model of the complex As the ba 3 cytochrome c oxidase from A. aeolicus cannot be separated by purification from its tightly bound partners (bc 1 complex and sulfide quinone reductase), we worked with the entire supercomplex in search of evidence for the presence of the putative subunit IIa, and to confirm the new N-terminal sequence of the subunit II.
A separation of the supercomplex subunits on a 10-20% denaturing gel did not allow us to visualize a small protein around 5 kDa, as was also the case for the same subunit from T. thermophilus [8]. Subunit II migrated on this gel at a position corresponding to a protein of slightly less than 17 kDa (Figure 3 A), corresponding to the correct molecular mass. After transfer of the protein from this gel band, N-terminal sequencing of the first 12 amino acid residues allowed to confirm the N terminal sequence (MDRAEKTGLTLA-) of subunit II determined from the genomic sequence (shown in Figure 3 B). In addition, a second protein was clearly detected in the same sample with a distinct Nterminal sequence, except for the first Met, (MNEKHXXXEF-, X being an undetermined residue) matching the N-terminal part of the proposed subunit IIa ( Figure 2). This corroborates the existence of this additional protein in the A. aeolicus oxidase and shows the strong interaction between subunits II and IIa as they are not dissociated after migration on a sodium dodecyl sulfate (SDS) gel. The unexpected migration of the complex subunit IIsubunit IIa at this size might be explained by the hydrophobic nature of the proteins, as previously pointed out [13]. The analysis of the protein band at 17 kDa by peptide mass fingerprinting after trypsin digestion showed assignment of ionic peptides spanning more than 75% of the subunit II sequence, including some in the extended N-terminal part (upstream from the previously proposed N-terminal) (Figure 3 B), whereas ions from subunit IIa were not detected, probably because of the highly hydrophobic nature and the very small size of this protein.
The molecular masses of both subunit II and subunit IIa were determined by Matrix-Assisted Laser Desorption/Ionization-Time-of-flight (MALDI-TOF) mass spectrometry using the intact supercomplex solution. As expected, the mass spectrum revealed several species corresponding to the constituents of the supercomplex (data not shown). In reflectron and positive mode, an ion has been measured at m/z 5105.99 and could be assigned to the subunit IIa (theoretical m/z 5105.47, Dm = 103 ppm). None of the other supercomplex components has a molecular mass that can be assigned to this latter one. Two other masses are also observed: one at m/z 5088.87, a 17 Da lower mass which could correspond to this subunit without one -NH 3 group, the other observed at m/z 5147.99, a 42 Da heavier mass which could correspond to an addition of an acetyl group (Figure 4 A). The ion precursor at m/z 5105.99 was enough intense to be fragmented in a lift mode, and a continuous series of y ions could unambiguously elucidate the Figure 2. Multiple sequence alignment of putative bacterial and archaeal oxidase subunits IIa. These proteins correspond to putative or characterized ba 3 cytochrome c oxidases. Residues identical in all sequences are shown in red (except the Tyr from the Salinibacter sequence), strongly similar residues in green. The Proline residue present in all sequences in the motif PX 1-3 GT/A is shown in orange. Grey shaded residues are involved in a putative transmembrane helix (Transmembrane region prediction with TMHMM program, http://www.cbs.dtu.dk/services/TMHMM-2.0/). Grey shaded residues in the Thermus sequence form a transmembrane helix in the three-dimensional structure of the enzyme (PDB 1XME). The alignment has been made with ClustalW. Complete names of organisms and accession numbers of sequences used to create the alignment are given in Table 1 These experiments confirmed that subunit II is longer than previously annotated and that a third subunit is present in the ba 3 oxidase from A. aeolicus.
A model of the three-dimensional structure of A. aeolicus ba 3 cytochrome oxidase has been calculated with SWISS-MODEL protein modeling program using the structure of the ba 3 cytochrome oxidase from T. thermophilus as a template (PDB entry 1XME) [14]. An overall view of the structure is presented in Figure 5 A. The complex is composed of three protein subunits, I, II and IIa.
Whereas T. thermophilus ba 3 oxidase subunit I is composed of 13 transmembrane-spanning helices, Aquifex subunit I consists of only 12 transmembrane helices. Heme B and the binuclear centre heme [A 3 .Cu B ] reside in subunit I. Hemes and copper ions were included in the structural model from the crystal structure of the ba 3 oxidase from T. thermophilus in which the A-type heme corresponds to the heme A s . The heme A s , first isolated in the SoxB-type terminal oxidase from the thermoacidophilic archae Sulfolobus acidocaldarius, is a A-type heme in which the farnesyl chain is replaced by a geranylgeranyl group [15]. Unlike T. thermophilus, A. aeolicus seems to contain heme A in the membrane, as shown for Aquifex pyrophilus [15]. Therefore, the heme A s has been modified to a heme A in the structural model of the A. aeolicus ba 3 oxidase ( Subunit II has only one transmembrane helix which is in close contact with subunit I. The periplasmically oriented polar domain of subunit II contains the binuclear copper Cu A site. The Cu A site is symmetrical with respect to the two histidine (His 96 and His 139) and the two cysteine (Cys 131 and Cys 135) ligands ( Figure  S1).
The third subunit of the complex, subunit IIa, consisting of 41 amino acid residues, forms one transmembrane helix. This helix,  also found in the structure of the ba 3 oxidase from T. thermophilus, is superimposable with the first transmembrane helix of subunit II from the A-type aa 3 cytochrome oxidase from Paracoccus denitrificans, but with opposite polarity [8] (Figure 5 A and B).

Presence of the subunit IIa in B-type cytochrome c/quinol oxidases
To determine whether subunit IIa is present in the majority of enzymes in the B family, or on the contrary, whether it is rare, we searched for a gene potentially encoding this subunit IIa in the genomes of various micro-organisms. We have investigated the existence of the peptide IIa, at the genomic level, in (1) B-type putative prokaryotic oxidases [12,16,17,18], and in (2) biochemically characterized ba 3 oxidases (R. marinus, Geobacillus sp. and Natronomonas pharaonis) [9,10,11,19,20]. In each case, we retrieved the sequence of the gene located directly upstream of the subunit II gene of the oxidase. When present, it is annotated as hypothetical protein in most cases, apart from the one in Meiothermus, which is annotated cytochrome c oxidase subunit IIa, those in Methylococcus, Natronomonas and Haloferax, annotated cbaD subunit, and the one found in Natrialba, annotated halocyanin. When such a gene coding for a putative peptide IIa was not present, we directly analyzed the DNA upstream the subunit II gene to find a potential small ORF, possibly not detected at the time of automatic genome annotation. We analyzed 36 oxidase genes region, found 21 putative subunit IIa sequences and detected 6 previously unannotated ORFs (Table 1). Sequence alignment confirms that the closest sequences to the A. aeolicus sequence belong to organisms in the Aquificales order such as Hydrogenivirga, Thermocrinis and Hydrogenobacter (Figure 2). It seems that archaeal subunits (53 to 78 amino acid residues) are longer than the bacterial ones (34 to 71 residues). Although sequence similarities between archaeal and bacterial organisms are low, two residues are conserved in all the 27 sequences (Gly 14 and Trp 30, A. aeolicus numbering), except that Glycine is absent in the sequence of Magnetospirillum and Tryptophan is replaced by Tyrosine in that of Salinibacter. In addition, the motif PX 1-3 GT/A (Pro 12, A. aeolicus numbering) appears to be present in most cases ( Figure 2). It seems that this sequence motif is absent from subunit II from A and B-type oxidases possessing two transmembrane helices. The role of this recurrent sequence pattern is unknown.
Secondary structure prediction indicates that all these subunit IIa sequences contain one transmembrane helix, like that in T. thermophilus ( Figure 2 and Table 1). Our analysis suggests that the helical peptide is present in many (75%) of the oxidase complexes studied. All these enzymes possess a one-helix transmembrane domain in subunit II. However, this subunit IIa was not detected when subunit II is predicted to have two membrane-spanning helices (small number of analyzed oxidases [12], Table 1). This confirms the idea that this transmembrane helix IIa is structurally equivalent to the missing helix of subunit II [7,8]. Although we did not analyze all the B-type cytochrome oxidases available in the genomic data, it seems that this third subunit IIa is widespread in the bacterial ba 3 -type enzymes.
All operons analyzed in this study have the same genomic organization as found for Thermus [8] and Aquifex ba 3 oxidases, with at least three genes encoding the oxidase, for the large subunit I, for the small subunit II and for a third subunit IIa (Figure 1). An exception is the enzyme from Methylococcus capsulatus, where one gene (coding for a hypothetical small protein) is located between the genes encoding the putative subunit IIa and the subunit II. It should be noted that archaeal oxidases seem to contain at least one more subunit than the bacterial B-type enzymes, as illustrated by the purified oxidase from N. pharaonis [19,20].

Discussion
Cytochrome c oxidases are essential enzymes in the respiration of aerobic organisms. The A-family of heme-copper oxidases is the largest and the best studied, whereas few type B enzymes have been purified and fully characterized. We previously proposed that the cytochrome c oxidase II from A. aeolicus (B-family) is a ba 3 -type enzyme [6]. We show in this work that it is in fact composed of three subunits, subunit I, subunit II and subunit IIa. The name coxIIa can be attributed to the new gene coding for the subunit IIa. We moreover propose that the small hydrophobic subunit IIa is present in a large number of oxidases in the B family. Aquifex is the second bacterium, after T. thermophilus, for which the existence of a third subunit in this type of oxidase has been experimentally demonstrated. Indeed, a ba 3 oxidase was purified from R. marinus and from G. stearothermophilus and described in both cases as two-subunit enzymes, although we show here that coxIIa gene is present in the genome of Rhodothermus and Geobacillus sp. C56T3 (this strain is homologous to G. stearothermophilus, with more than 93% identity between subunits II of the B-type oxidases from both organisms) (Figure 2 and Table 1). In Archaea, although subunits I and II were detected in the ba 3 oxidase from Aeropyrum pernix [21], the enzyme from N. pharaonis contains in addition a smaller subunit that is lost during purification [19,20]. This underlines the difficulty of detecting subunit IIa in these complexes at the protein level, probably because of its high hydrophobic nature and its very small size. Indeed, it is not visible on a denaturing gel and is not amenable to identification by MALDI TOF after trypsin digestion. Moreover, the fact that the gene encoding subunit IIa is not always annotated in genomes makes it hard to detect at the genomic level.
Our analysis of the genetic environment of some oxidase subunit II genes suggests that subunit IIa is a general feature of the bacterial B-type oxidases that possess only one transmembrane spanning helix. These enzymes are thus three-protein complexes, subunits IIa and II being, as previously postulated, the structural equivalents of subunit II of type A enzymes [7,8,22]. A gene fission, located between the two transmembrane helices, of the gene coding for the two-helices subunit II or in contrary a fusion of the genes coding for the two subunits II and IIa (Figure 5 B) might have occurred during evolution of the oxidases. A systematic and comprehensive study proposed that there is a correlation of gene fission with thermophily and that split genes might reflect an adaptation to high temperature [23]. However, our present study points that subunit IIa is potentially present in organisms that are phylogenetically diverse and that this presence is not limited to thermo or hyperthermophilic species (Table 1). This study shows also that some bacterial oxidases from family B are inserted in the membrane by two transmembrane helices, like the type A enzymes (Table 1), as previously proposed in [12]. In these cases, no subunit IIa has been found. In Archaea, type B oxidases appear to be more diverse in terms of composition, size and nature of subunits than in Bacteria. A recent study, based on sequences analysis, proposes a reclassification of the archaeal B-type heme-copper oxidases in new families [3]. According to these authors, subunit IIa seems to be present in archaeal oxidases of only one family of this type of enzymes [3]. It is noteworthy that this small protein IIa occurs as a fusion protein with halocyanin in Haloarchaea, the blue copper halocyanin being described as the redox partner of the ba 3 cytochrome oxidase in N. pharaonis [3,20,24]. Indeed, in Halobacterium and Haloarcula, the equivalent of subunit IIa is found at the Cterminal extremity of a protein annotated halocyanin (which gene is directly upstream the one for subunit II), the motif ProXaaGlyThr (Xaa is an unspecified amino acid) and the Trp residue ( Figure 2) being conserved ( Figure S2). Thus in prokaryotes, the membranespanning helix IIa appears to be present in a large number of oxidases either as an independent subunit (most of B-type oxidases, Table 1), or as a protein fusion with subunit II of oxidases (A-type oxidases and some B-type oxidases as in Table 1) or as a protein fusion with halocyanin (B-type enzymes in some Haloarchaea). This is reminiscent of the case of the ''long'' versus ''split'' cytochrome b from bc complexes. The majority of cytochromes b in Bacteria consists of a common ''long'' protein of eight membrane-spanning helices, whereas cyanobacteria and Firmicutes contain two separate proteins, cytochrome b 6 and SUIV, corresponding to the N and C-terminal parts of the ''long'' cytochrome b respectively [25]. Moreover, in a few Firmicutes, SUIV is fused to a c-type cytochrome subunit [25]. Our preliminary analyses seem to indicate that there is no sequence similarity between subunit IIa and the transmembrane segment of the B and A-type oxidases (data not shown). A more in-depth investigation needs to be carried out to determine the evolutionary history of the subunit IIa of the type B oxidase complexes in Bacteria and Archaea.
Over-expression of only the two subunits I and II of the ba 3 -type oxidase from G. stearothermophilus led to an enzymatically active enzyme with proton pumping activity [11], which suggests that subunit IIa may not to be essential for the ba 3 cytochrome oxidase activity. Nevertheless, as it is a homologous expression and in view of the difficulty to detect subunit IIa in the complex, we cannot absolutely rule out the possibility that a fraction of the purified oxidase could contain subunit IIa. The fact that the presence of two helices (separate or fused to another protein) in the transmembrane region of subunit II of heme-copper oxidases, a well preserved feature through evolution, suggests that it contributes to the stability, positioning, or interaction of the various subunits within the complex or of the whole complex in the membrane. Small subunits, predicted to form a single transmembrane helix, have been described in eukaryotic membrane protein respiratory or photosynthetic complexes [26][27][28][29][30]. In most cases, they are proposed to have a function in assembly, stabilization or dimerization of the complexes. Such a subunit can be found also in the bacterial bc 1 complex of Rhodovulum sulfidophilum where it seems to play a stabilizing role [31]. Two additional small subunits containing a single transmembrane helix each were recently identified in the cyanobacterial NADH dehydrogenase complex [32]. Phylogenomic analysis highlighted the high level of redundancy of oxidase in prokaryotes [3,16,18]. As shown by biochemistry and bioinformatics, numerous prokaryotes possess both aa 3 -and ba 3type oxidases [12,17,18,21,33,34]. Based on genome analysis, A. aeolicus possesses two putative cytochrome c oxidases (Figure 1) [5]. The terminal cytochrome oxidase produced in the growth conditions used here is the ba 3 -type enzyme. If the second uncharacterized cytochrome c oxidase is expressed in our cultures, it must be at concentrations an order of magnitude lower than that for ba 3 oxidase. As we assumed that a very low oxygen concentration is available during the growth of A. aeolicus (probably a few mg/L, at least at the end of growth in our conditions), it seems logical to us that the ba 3 terminal oxidase is predominantly expressed. Indeed, under low aeration conditions, the preferred oxygen reduction pathway in T. thermophilus is the ba 3 branch and it has been proposed that the ba 3 enzyme, with a high affinity to O 2 , is evolutionary selected to work under low oxygen tension [35][36][37]. It is also the case in Archaea [17,21]. The aim is now to grow A. aeolicus with various controlled oxygen concentrations in the medium to investigate the presence of both cytochrome c oxidases in the membranes of the bacterium, as well as the one of the bdtype quinol oxidase.

N-terminal sequence determination
The N-terminal amino acid sequence was determined after SDS gel electrophoresis and transfer of proteins onto polyvinylidene difluoride (PVDF) membrane for 45 min at a current intensity of 5 mA.cm 22 using a semi-dry electrophoretic transfer unit (Biometra). Sequence degradation was performed on an Applied Biosystems Procise 494 microsequencer equipped by a high-pressure liquid chromatography for quantitative determination of phenylthiohydantoin derivates.

DNA sequencing
To identify a potential error of A. aeolicus genome sequencing, PCR was used to generate a 325 bp DNA fragment located between the coxA 1 and coxB 2 genes (positions 1541988 to 1542007 and 1542333 to 1542353) using the (59-GTCGGGTACAATAG-CATTTT-39 and 59-CGTGAATATGTCCACTACGGA-39) oligonucleotide pairs. The DNA fragment was sequenced using the Sanger method (Beckman Coulter Genomics). The nucleotide sequence and annotation have been deposited to the European Nucleotide Archive under accession number FR846386.

MALDI-TOF mass spectrometry
1 mL of a 22 mM solution of intact proteins was directly spotted onto a MALDI stainless steel target plate and an equal volume of a saturated solution of matrix sinapinic acid (40% CH 3 CN in water, 0.1% Trifluoroacetic acid (v/v)) was added (n = 3). Then, mixtures were allowed to dry at room temperature. Data acquisition was operated on MALDI-TOF mass spectrometers Ultraflex II and Microflex II from Bruker Daltonics using the Flex control software. The methods used was either the linear positive mode with a pulsed ion extraction for studying proteins in the range of 1-30 kDa and 10-80 kDa, either the reflectron positive mode for characterization of the subunit IIa in the 1-10 kDa range. External mass calibration was carried out on the Protein calibration standard I or Peptide calibration standard (Bruker Daltonics). In linear mode, mass spectra were treated by a Gaussian smoothing. MS/MS analysis on the ion precursor at m/ z 5105.99 was performed in a lift mode on the MALDI-TOF/ TOF Ultraflex II mass spectrometer.
For the analysis, oxidases described as belonging to the B family were used [12,16,17,18]. We used also sequences of oxidases found with a BLAST search (http://www.ncbi.nlm.nih.gov/ sutils/genom_table.cgi) using the subunit II of A. aeolicus. For these latter sequences, a multiple sequences alignment of all the subunits I from these enzymes indicated the presence of the four conserved residues proposed to be involved in the K channel of B-type oxidases (Tyr 226, Tyr 237, Ser 300, Thr 303, A. aeolicus numbering) [2]. When not annotated, genes encoding putative subunits IIa in micro-organisms were searched in the DNA genomic sequence, directly upstream subunit II gene, using Translate tool (ExPASy Proteomics Server, http://us.expasy.org/ tools/dna.html) to detect the potential ORFs and TMHMM Server v. 2.0 to predict the transmembrane helices in the putative proteins. Potential small proteins were considered as putative subunits IIa only if they contain a transmembrane domain (otherwise they were not considered).

Molecular modeling
Since there is no crystallographic structure determined for A. aeolicus ba 3 cytochrome oxidase, a homology model was built using SWISS-MODEL protein modeling program (version 8.05) which is available from the ExPASy website [40,41]. The modelization of the three subunits of the complex is based on the modified sequence of A. aeolicus genes cytochrome oxidase II (see section A. aeolicus genomic sequence analysis). The T. thermophilus ba 3 cytochrome oxidase X-ray crystal structure (protein data bank entry 1XME) was used as a template, the sequences of A. aeolicus ba 3 oxidase subunits being at least 60% similar to those of the enzyme from T. thermophilus. Hemes and copper ions from the crystal structure were included in the model. The procedure was carried out separately for subunit I, subunit II and subunit IIa.
T. thermophilus subunit I contains 13 transmembrane helices [7,8]. Search for membrane-spanning helices using the TMHMM server in the Aquifex subunit I revealed only 12 transmembrane helices. As a consequence, the sequence of Thermus subunit I used for modeling did not contain the 46 C-terminal amino acids.
The homology model was checked for correct stereochemistry by performing a Ramachandran analysis. The final model has more than 95% of the residues in the allowed regions. The resulting theoretical model was displayed and analyzed with WebLab ViewerLite 5.0 software (Accelrys Inc.). Figure S1 Multiple sequence alignment of some bacterial and archaeal oxidase subunits II. These proteins correspond to putative or characterized ba 3 cytochrome c oxidases. Residues identical in all sequences are shown in red, strongly similar residues in green and weakly similar residues in blue. Grey shaded residues in the Aquifex sequence are involved in a putative transmembrane helix (TMHMM Server v. 2.0, Prediction of transmembrane helices in proteins). Boxed residues in the Thermus sequence form a transmembrane helix in the three-dimensional structure of the enzyme (PDB 1XME). The Histidine and Cysteine residues ligands of the Cu A site are shown in bold (His 96, Cys 131, Cys 135 and His 139, A. aeolicus numbering). The alignment has been made with ClustalW. Complete name of organisms and accession numbers of sequences used to create the alignment are given in Table 1.