Electron Transfer and Binding of the c-Type Cytochrome TorC to the Trimethylamine N-Oxide Reductase in Escherichia coli *

Reduction of trimethylamine N-oxide (E′0(TMAO/TMA) = +130 mV) in Escherichia coli is carried out by the Tor system, an electron transfer chain encoded by the torCAD operon and made up of the periplasmic terminal reductase TorA and the membrane-anchored pentahemicc-type cytochrome TorC. Although the role of TorA in the reduction of trimethylamine N-oxide (TMAO) has been clearly established, no direct evidence for TorC involvement has been presented. TorC belongs to the NirT/NapC c-type cytochrome family based on homologies of its N-terminal tetrahemic domain (TorCN) to the cytochromes of this family, but TorC contains a C-terminal extension (TorCC) with an additional heme-binding site. In this study, we show that both domains are required for the anaerobic bacterial growth with TMAO. The intact TorC protein and its two domains, TorCN and TorCC, were produced independently and purified for a biochemical characterization. The reduced form of TorC exhibited visible absorption maxima at 552, 523, and 417 nm. Mediated redox potentiometry of the heme centers of the purified components identified two negative midpoint potentials (−177 and −98 mV) localized in the tetrahemic TorCN and one positive midpoint potential (+120 mV) in the monohemic TorCC. In agreement with these values, thein vitro reconstitution of electron transfer between TorC, TorCN, or TorCC and TorA showed that only TorC and TorCC were capable of electron transfer to TorA. Surprisingly, interaction studies revealed that only TorC and TorCN strongly bind TorA. Therefore, TorCCdirectly transfers electrons to TorA, whereas TorCN, which probably receives electrons from the menaquinone pool, is involved in both the electron transfer to TorCC and the binding to TorA.

expressed at a low level (6). Each of these systems comprises a periplasmic terminal reductase and a membrane-anchored ctype cytochrome. The membranous Me 2 SO reductase, which is constitutively expressed in anaerobiosis, is also capable of TMAO reduction (7).
The TorCAD system is encoded by the torCAD operon, and its expression is under the control of the TorS/TorR two-component regulatory system, which mediates the response to the presence of TMAO in the medium (8). The torA gene encodes the periplasmic molybdoreductase TorA (5). According to sequence homologies, TorA belongs to the Me 2 SO reductase family, a group of periplasmic enzymes that contain a molybdenum cofactor as a single prosthetic group and are capable of utilizing TMAO and/or Me 2 SO as electron acceptors (9). In the past few years, structures have been published for several members of this family (10 -13). They are organized in four domains surrounding a bismolybdopterin guanine dinucleotide cofactor that is consequently buried in a deep depression of the protein surface. The insertion of the molybdenum cofactor into the TorA apoprotein is a cytoplasmic event (14) that probably involves the TorD cytoplasmic chaperone (15).
It was previously shown that the first gene of the tor operon, torC, encodes protein TorC, a 46-kDa pentahemic c-type cytochrome (5,16). TorC is anchored to the inner membrane by a sequence of about 20 hydrophobic residues, while a globular domain containing five hemes faces the periplasm. Heme binding to the five consensus CXXCH motifs takes place in the periplasm by a mechanism involving the c-type cytochrome maturation machinery encoded by the ccm genes (17,18). From sequence comparisons, TorC has been included in the NirT/ NapC class of membrane-anchored multiheme c-type cytochromes. This family originally was composed of tetrahemic c-type cytochromes of about 20 kDa that were involved in periplasmic nitrite and nitrate reduction (19). The N-terminal region of TorC, which is homologous to NirT and NapC, possesses the four heme binding sites, whereas the C-terminal part, which is present specifically in the Me 2 SO/TMAO respiratory systems, carries the fifth heme motif (5). Recently, two members of the NirT/NapC family, NapC and the TorC homologue DorC, have been characterized, and the redox potential values determined for the four hemes of NapC and the five hemes of DorC were all negative (20,21).
In TMAO and Me 2 SO reduction systems, pentahemic cytochromes proved to be essential for the electron transfer and thus appeared to be the intermediate components between the membrane quinone pool and the terminal reductase (6,22). Since it has been shown in E. coli that TMAO reduction involves menaquinones (23), TorC should receive the electrons from the menaquinone pool (EЈ 0(menaquinol/menaquinone) ϭ Ϫ74 mV) and transfer them to the active site of TorA, where TMAO reduction (EЈ 0(TMAO/TMA) ϭ ϩ130 mV) occurs.
Interestingly, a recent genetic study showed that TorC in its unprocessed form is a negative regulator of the torCAD operon (18). Indeed, apo-TorC seems to inhibit the kinase activity of the TMAO sensor TorS by an unknown mechanism. This negative autoregulation probably means that the maturation of TorC is the limiting step for the Tor system biogenesis.
This paper provides in vivo and in vitro evidence that the mature form of TorC is directly involved in the electron transfer to the terminal reductase, TorA. We also show that TorC exhibits four negative redox potentials and a positive one corresponding to the fifth heme. The latter group is responsible for the direct electron transfer to the catalytic site of TorA. In addition, TorC binds TorA, and this interaction involves the N-terminal domain of TorC.

Bacterial Strains, Plasmids, Media, and Growth Conditions
The bacterial strains and plasmids used in this work are listed in Table I. To maintain selection for plasmids, antibiotics were added at the following concentrations: ampicillin, 100 g/ml; chloramphenicol, 10 g/ml. For the biochemical study, cells were grown anaerobically at 37°C on Luria broth medium (24). The concentration of arabinose or glucose, added to the growth medium, is detailed for each experiment under "Results." The growth kinetics studies were performed as described by Gon et al. (6).

Constructions Leading to TorC, TorC N , and TorC C Production
To synthesize the membrane His tag TorC, the complete torC coding sequence was amplified from MC4100 chromosomal DNA by polymerase chain reaction, using an oligonucleotide (C5Hc) that corresponds to an EcoRI site followed by the 5Ј coding sequence of torC and a primer (Cct) that corresponds to a SmaI site followed by a sequence encoding a C-terminal His 6 tag and the complementary sequence of the 3Ј-end of torC (Fig. 1A).
To synthesize the N-terminal domain of TorC (TorC N , positions 1-194 relative to TorC) that includes the first four heme binding sites of TorC (Fig. 1B), the corresponding coding sequence was amplified by polymerase chain reaction using oligonucleotide C5Hc and a primer that corresponds to a SmaI site followed by a His 6 tag coding sequence and the appropriate torC internal sequence.
To synthesize the periplasmic C-terminal domain of TorC (TorC C , from position 198 to 390) that contains the fifth heme of TorC (Fig. 1C), the corresponding torC coding sequence was amplified by polymerase chain reaction using an oligonucleotide that contains an EcoRI site followed by the TorT signal peptide (18 codons) coding sequence (28) and the appropriate torC sequence and primer Cct. Detailed information on the primer sequences is available on request.
The purified polymerase chain reaction products were digested by the appropriate restriction enzymes and ligated into the corresponding cloning sites of the expression vector pBAD24 (27) downstream from the arabinose-inducible P BAD promoter to give pBC, pBC N , and pBC C ( Table  I). All of the polymerase chain reaction products and the fusion sites were confirmed by sequencing. DNA preparations were carried out with high pure DNA isolation kits from Roche Diagnostics. Transformations were performed according to the method of Chung and Miller (29). Plasmids pBC, pBC N , and pBC C were digested with NheI and HindIII, and the resulting DNA fragments encoding the modified His tag TorC proteins were cloned just downstream of torD, into the compatible sites XbaI and HindIII of plasmid pTorAD (15), leading to plasmids pTorADC, pTorADC N , and pTorADC C , respectively.

Purification of TorA, TorC, TorC N , and TorC C
The purification of the TMAO reductase (TorA) was performed from the periplasmic fraction of cells DSS401/pEC86. TorA protein enrichment was achieved as follows. The periplasmic fraction of strain DSS401/pEC86, grown anaerobically in the presence of TMAO (30 mM), was prepared by the lysozyme-EDTA procedure (30) and precipitated with ammonium sulfate 60% (w/v). The pellet was resuspended into Tris-HCl 40 mM buffer (pH 7.6) and then loaded on a Mono Q HR 5/5. The TMAO reductase active fractions, obtained from a 50 -500 mM NaCl gradient were then dialyzed, pooled, concentrated by 60% (w/v) ammonium sulfate precipitation, and loaded on a Superdex G200 HR 10/30 column. At this stage, the active fractions contained 3 mg of protein, and the purity of the preparation was checked on Coomassie Bluestained gel.
Purification of TorC and TorC N was performed from the membrane fraction of cells MC4100, containing both plasmids pEC86 and either pBC or pBC N , grown anaerobically in the presence of 0.0005% arabinose. Membrane fraction, prepared as described by Silvestro et al. (4), was incubated for 1 h at 4°C in 20 mM phosphate buffer (pH 7.4) with 500 mM NaCl containing 2% Triton X-100. After centrifugation at 14,000 rpm for 30 min at 4°C, the fraction containing the solubilized cytochrome c, TorC or TorC N , was diluted two times with the same buffer containing 0.2% Triton X-100 and 2 mM imidazole and loaded onto a HiTrap TM chelating Ni 2ϩ column (Amersham Pharmacia Biotech). The proteins were eluted with a step gradient of imidazole from 20 to 500 mM. Both TorC and TorC N cytochromes were eluted with 150 mM imidazole.
TorC C , the soluble C-terminal domain of TorC, was purified from the periplasmic fraction of cells MC4100 containing both pEC86 and pBC C grown under anaerobic conditions in the presence of 0.1% arabinose. The periplasm was diluted two times into 20 mM phosphate buffer (pH 7.4) with 500 mM NaCl, 2 mM imidazole, and the purification onto a HiTrap TM chelating Ni 2ϩ column was performed as described for TorC and TorC N .

Analytical Procedure
Protein analysis was carried out using 12.5 or 10% SDS-PAGE. After electrophoresis, the presence of hemes within TorC and derivatives was revealed by staining the gel for peroxidase activity using 3,3Ј,5,5Јtetramethylbenzidine as described by Thomas et al. (31); then Coomassie Blue staining of total proteins was performed (32). Protein concentrations were estimated using the technique of Lowry et al. (33).

Kinetics of Electron Transfer
The kinetic assays were achieved using anaerobic cuvettes, typically filled with phosphate buffer (20 mM, pH 7.4), TMAO (50 mM), and either TorC C (10 M), TorC N (5 M), or TorC (5 M) previously reduced by dithionite. Before adding TMAO reductase (350 units⅐ml Ϫ1 ), the cuvette mixture was flushed with argon (10 min). The oxidation rate of the cytochrome was determined from the absorption band at 552 nm.

Interaction Experiments
Interaction Studies on PAGE under Native Conditions-TorA (5 M) and either TorC (5 M), TorC N (5 M), or TorC C (10 M) were incubated for 30 min at room temperature in 20 mM phosphate buffer (pH 7.4), Triton X-100 0.05%. The interactions were analyzed by 10 or 12.5% PAGE under native conditions. For identification, the proteins were either stained for heme with 3,3Ј,5,5Ј-tetramethylbenzidine or were transferred to a Hybond TM ECL TM nitrocellulose membrane. The protein complexes were immunodetected with anti-His antibodies (Invitrogen). The ECL-Western blotting system was used as recommended by the supplier (Amersham Pharmacia Biotech).
Interaction Studies by Biosensor Experiments-The BIAcore apparatus (Amersham Pharmacia Biotech) was used to analyze, in real time, the binding between TorA (M r 94,000) and TorC (M r 46,000), TorC C (M r 23,000), or TorC N (M r 24,500), by using surface plasmon resonance. All experiments were carried out at 25°C. TorA, in 10 mM sodium acetate buffer (pH 5), was immobilized on a CM5 sensor chip (BIAcore) through amine coupling (3000 resonance units). TorC, TorC N , and TorC C were diluted in 20 mM phosphate buffer (pH 7.4), 0.05% Triton X-100 and injected using a constant flow rate of 10 l/min. The resulting sensorgrams were evaluated using the biomolecular interaction analysis evaluation software (BIAcore) to calculate the kinetic constants of the complex formation.

RESULTS
The Intact TorC Protein Is Required for TMAO Respiration-To establish that the c-type cytochrome TorC is required for electron transfer to the terminal enzyme TorA, we cloned the torAD genes alone (15) or together with torC under the control of the P TAC promoter of the plasmid pJF119. The resulting pTorAD and pTorADC plasmids were introduced into strain LCB504, which is unable to reduce TMAO. Indeed, this strain carries an interposon in the beginning of the first gene of the torCAD operon and a deletion of the entire dmsABC operon (6). The strains carrying either pJF119 or pTorAD grew very slowly and at the same rate under anaerobic conditions in the presence of TMAO (Fig. 2). In contrast, when plasmid pTorADC was introduced into strain LCB504, the recombinant strain exhibited a significant growth rate in the presence of TMAO (Fig. 2) and produced the characteristic odor of volatile TMA, which results from reduction of TMAO. These results confirmed the hypothesis that TorC is essential for TMAO reduction.
The TorC protein belongs to the NapC/NirT family of c-type cytochrome because its tetrahemic N-terminal domain (hereafter called TorC N ; Fig. 1B) is homologous to the NapC protein and contains four heme-binding sites. In addition, TorC contains a C-terminal domain (hereafter called TorC C ; Fig. 1C) with a single heme-binding motif. The C-terminal extension is found exclusively in TorC/DorC systems (35). To determine whether one or both of the two domains are required for the TMAO reduction process, we cloned DNA fragments encoding either one of these domains behind torAD. The resulting plas-mids pTorADC N and pTorADC C were introduced into strain LCB504, and the growth rate was monitored. The recombinant strains containing pTorADC N did not grow significantly in the presence of TMAO (Fig. 2), and the cells containing pTorADC C grew very slowly only after 36 h of incubation. These results strongly suggest that both domains of the TorC protein are required for the TMAO respiratory process.
Overproduction and Purification of TorC and of Its N-and C-terminal Domains-To study the biochemical characteristics of the TorC protein, we constructed recombinant plasmids that encode either TorC; TorC N , which corresponds to the N-terminal domain of TorC; or TorC C , which corresponds to the second half of the TorC protein (Fig. 1). DNA fragments encoding TorC, TorC N , and TorC C were cloned into the pBAD24 vector, giving rise to plasmids pBC, pBC N , and pBC C with His tags added at the C-terminal extremity of each protein. Moreover, since the heme incorporation occurs in the periplasm, we introduced the TorT signal peptide in front of TorC C to assure its periplasmic localization.
Overproduction of mature c-type cytochrome, and especially multiheme c-type cytochromes, is difficult in E. coli, due to the limited capacity for synthesis of the c-type cytochrome of this organism. To overcome this problem, we used a strain containing the plasmid pEC86 that carries the c-type cytochrome maturation ccm genes and thus improves production of c-type cytochromes (17). Nevertheless, the amount of mature cytochrome produced is rather low even in the presence of this plasmid, and the expression of cytochrome-encoding genes has to be controlled to prevent saturation of the maturation machinery. Accordingly, the TorC genes were cloned into the pBAD24 vector under the control of the P BAD promoter for controlled expression of TorC proteins. Plasmids pBC, pBC N , and pBC C were then introduced into strain MC4100 carrying the compatible plasmid pEC86. TorC and TorC N expression was induced with 0.0005% arabinose because concentrations higher than 0.001% apparently overloaded the heme delivery machinery and led to degraded proteins. Higher levels of the mature TorC C protein were obtained by using 0.1% arabinose, probably because TorC C is a monoheme protein. Heme-staining proteins of 47 and 24 kDa were present in membrane extracts of the strains containing pBC and pBC N , respectively, indicating that TorC and TorC N were matured (data not shown). Since these proteins were exclusively located in the membrane fraction (data not shown), detergent Triton X-100 was used to solubilize them. A 23-kDa heme-staining protein was detected in the periplasmic extract of the strain containing pBC C (data not shown). The detergent-solubilized fractions containing TorC or TorC N or the periplasmic fraction containing TorC C were loaded onto Ni 2ϩ columns, and the His-tagged proteins were eluted from the column at about 150 mM imidazole. Analysis by SDS-PAGE revealed the presence of TorC, TorC N , and TorC C as a single major band on gels stained for heme-dependent peroxidase activity (Fig. 3A). No other c-type cytochrome copurified with the proteins and no apparent degradation occurred. The same gels stained with Coomassie Blue (Fig. 3B) revealed one major band; moreover, this band cross-reacted with anti-His antibodies in Western blots (data not shown), confirming that His-tagged derivatives of TorC had been purified.

Characterization of TorC by Optical Spectroscopy Shows That TorC Exhibited Positive and Negative Heme Redox Potentials-
The UV-visible spectrum of purified TorC shows a Soret peak at 411 nm in the oxidized state and peaks at 417 (Soret), 523, and 552 nm in the dithionite reduced form (Fig. 4A). These peaks are characteristic of c-type cytochromes and confirm that TorC is a c-type cytochrome as predicted by the presence of c-type heme binding sites (CXXCH) in the amino acid sequence. TorC was totally reduced by dithionite, but it was only partially reduced by ascorbate (Fig. 4A). These results indicate that at least one of the TorC hemes exhibits a positive redox potential, whereas the remaining hemes display more negative redox potentials. Similar UV-visible spectra were obtained with membrane extracts from strain DSS401 containing native TorC after reduction with ascorbate or dithionite (Fig. 4B), indicating that the purified His tag TorC cytochrome has spectral properties similar to those of the unmodified native TorC protein.
To determine which domain of TorC contains a positive redox potential center, we carried out UV-visible spectra with the purified TorC N and TorC C domains (Fig. 4, C and D). Fig. 4C shows that the hemes of TorC N are totally reduced by dithionite but not by ascorbate. In contrast, Fig. 4D shows that the unique heme of TorC C is totally reduced by both ascorbate and dithionite. These findings strongly suggest that the fifth heme of TorC has a positive potential, whereas the first four hemes exhibit negative potentials.
Determination of the Redox Potentials of the TorC Protein-To define more precisely the heme potentials, a mediated redox potentiometry was carried out on the purified TorC protein. This analysis revealed three distinct redox type centers, one at Ϫ177 mV (two hemes), one at Ϫ98 mV (two hemes), and one at ϩ114 mV (one heme) (Table II). These potentials are in agreement with the previous spectral evidence for one positive heme potential in the TorC C domain and four negative heme potentials in the TorC N domain. Mediated redox potentiometry of TorC C (Table II) yielded a value of ϩ120 mV, in complete agreement with the spectral evidence for a positive heme in the TorC C domain of TorC (Fig. 4D) and with the heme potential values obtained with the intact TorC protein.
Transfer of Electrons from TorC to the TMAO Reductase TorA-In the model of the Tor electron transfer pathway, TorC transfers electrons from the menaquinones, which are embedded in the cytoplasmic membrane, to the periplasmic TMAO reductase TorA. To demonstrate directly that TorC can act as an electron shuttle from the inner membrane to TorA, we studied electron transfer between purified TorC and TorA. TorC was previously reduced by dithionite under anaerobic conditions (Fig. 4A). Oxidation of TorC, followed at 552 nm ( Fig. 5) was specifically dependent upon TMAO reductase, since no oxidation occurred before the addition of TorA (data not shown).
Based on the redox potentials of TorC N (Ϫ177 and Ϫ98 mV) and TorC C (ϩ120 mV) and on EЈ 0(TMAO/TMA) (ϩ130 mV), we hypothesized that electrons are transferred from TorC N to TorC C and from TorC C to TorA. In this model, TorC C is the direct electron donor to TorA. Purified TorC N and TorC C were tested as electron donors for TMAO reductase. The monoheme TorC C domain was oxidized by the addition of TorA, whereas the tetrahemic TorC N domain was not (Fig. 5). Although oxidation of TorC C was slower than that of TorC, these results indicate that the electron donor for TorA is most probably located in the C-terminal part of TorC. When a mixture of purified TorC N and TorC C was tested, the electron transfer was slightly higher than that of TorC C alone but slower than that of TorC. These results indicate that TorC N also plays a role in the electron transfer pathway.
Binding of TorC to TorA Mainly Involves the TorC N-terminal Domain-The above results with purified components imply a direct interaction between these proteins. To characterize the binding between TorC and TorA, the formation of a TorC-TorA complex was studied under nondenaturing conditions by native PAGE and by BIAcore experiments. TorA and TorC were mixed together and loaded in the same conditions onto two polyacrylamide gels. After electrophoresis under native conditions, one gel was stained for heme (Fig. 6A, lanes 1 and  2), and a Western blot with anti-His antibodies was performed with the second gel (Fig. 6A, lanes 3 and 4). Significantly, the migration of TorC was retarded in the presence of TorA (Fig.  6A, compare lanes 1 and 3 with lanes 2 and 4, respectively). This result suggested that TorC can bind to TorA. To confirm this result, we studied the interaction between TorA and TorC by using the BIAcore technique (surface plasmon resonance). For this purpose, purified TorA protein was coupled to the dextran matrix of a sensor chip, and TorC was injected into the TorA-containing sensor chip. The sensorgram reflected the association and dissociation of two proteins, indicating that TorC directly interacts with TorA (Fig. 7A) in agreement with the above result.
The interaction of the individual domains with TorA was also studied by the same two approaches. Surprisingly, no binding between TorC C and TorA was detected on a native gel stained for heme or by Western blot with anti-His antibodies. Indeed, whatever the conditions used, TorC C migrated at exactly the same position in the presence and in the absence of TorA (Fig.  6B, compare lanes 1 and 3 with lanes 2 and 4, respectively). Similarly, no binding between TorC C and TorA was detected by the BIAcore procedure (Fig. 7B). As a control, we coupled TorC C to the dextran matrix and injected TorA, and again, no formation of a complex corresponding to the association of TorC C and TorA was detected (data not shown). When TorC N and TorA were mixed and loaded onto a polyacrylamide gel under nondenaturing conditions, TorC N migration was retarded by the presence of TorA (Fig. 6C), suggesting that the N-terminal domain of TorC is sufficient for TorA binding. The fact that TorC N can bind to TorA was confirmed by using the BIAcore technique. Indeed, an increase in the amount of recovered resonance units was observed when the TorC N protein was   injected (Fig. 7C). These results clearly show that the TorC N domain interacts with TorA.
The association rate constants (k on ) and the dissociation rate constant (k off ) of TorC and TorC N to TorA were determined at four different concentrations between 2.5 and 10 M (data not shown). TorC N exhibited a K d value of 4.5⅐10 Ϫ8 M (Table III). Significantly, the sensorgram of intact TorC binding to TorA exhibits at least two steps leading to the determination of two constants, K d1 ϭ 1.7⅐10 Ϫ8 M and K d2 ϭ 3.0⅐10 Ϫ6 M (Table III). These two steps could correspond to the fixation of TorC N and TorC C , but TorC N is mainly responsible for the binding of intact TorC to TorA because TorC C alone does not bind TorA. The low affinity between TorC C and TorA could also explain the poor catalytic efficiency in electron transfer to TorA exhibited by TorC C compared with that of the intact TorC protein (Fig. 5). DISCUSSION This study has revealed that TorC is made up of two domains of similar size, corresponding to its N-and C-terminal half (TorC N and TorC C , respectively) ( Figs. 1 and 3). As the other members of the NirT/NapC family with which it shares sequence homologies, TorC N contains a membrane anchor segment followed by four heme-binding sites (5,20). Production of the isolated TorC N domain resulted in a stable mature protein.
Unfortunately, and in contrast to what was described previously for NapC (20), the removal of the membrane anchor of TorC and TorC N led to an unstable protein (data not shown). The second domain, TorC C contains the fifth heme-binding motif of TorC (Fig. 1C) and presents sequence homologies with the corresponding region of pentahemic cytochromes specifically involved in TMAO/Me 2 SO respiration of various bacteria (6,35). Although the fifth heme binding site is located at the C-terminal extremity (positions 329 -333 relative to TorC), we were unable to produce a stable protein starting from position 285 of TorC, i.e. just upstream from the fifth heme binding site (data not shown). The region located between the fourth and the fifth heme binding sites probably plays an essential role in the stability and/or the folding of the TorC C domain.
The comparison between the midpoint redox potential values obtained for the soluble form of the Paracoccus denitrificans NapC (n ϭ 1, Ϫ235, Ϫ207, Ϫ181, and Ϫ56 mV) (20) and TorC N , the N-terminal domain of TorC (n ϭ 2, Ϫ177 and Ϫ98 mV) ( Table II), highlights that they are all negative and in the same range. This result confirms that NapC and TorC N are highly related. It is also striking that, for TorC N , each of the midpoint redox potentials corresponds to two heme centers. This feature could be representative of a symmetry in the N-terminal domain of TorC. This hypothesis agrees with the previous proposal of Roldan et al. (20), that the tetrahemic domain of NirT/NapC c-type cytochromes comes from a gene duplication leading to two related diheme subdomains. If true, then the redox potential of the first and third hemes of TorC or that of the second and the fourth ones should be very similar.
The midpoint redox potential described by Shaw et al. (21) for purified DorC, a TorC pentahemic homologue, is in the same range of values (Ϫ276, Ϫ185, Ϫ184, Ϫ128 mV, and Ϫ34 mV) as those detailed above for NapC. Surprisingly, all the potentials exhibited by DorC are negative, while a positive one has been detected for TorC. Indeed, we have clearly shown that the redox potential of the fifth heme is about ϩ120 mV, and the presence of a positive heme center has been observed in both the isolated TorC C domain and the intact TorC protein (Table  II). Moreover, this result has been confirmed by a study performed directly on native TorC using membrane extracts (Fig.  4B). These results are also in agreement with the fact that ascorbate reduced partially TorC and TorC C completely (Fig.  4). Positive midpoint redox potentials are often associated to His-Met-ligated hemes (36). The comparison of the sequences surrounding the fifth heme-binding site of the pentahemic cytochrome family reveals the presence of one highly conserved residue of methionine (position 353 in the TorC amino acid sequence) (35). This residue is a good candidate for the axial ligand, although a conserved histidine residue (position 340) might also play such a role. As shown by the in vitro electron transfer experiments, the FIG. 8. Model of binding and electron transfer between TorC and TorA. In our model, in a first step, TorC N , the N-terminal membrane-anchored domain of TorC, binds the periplasmic TorA protein (1). Then TorC C is correctly positioned to transfer electrons directly to the catalytic site of TorA, allowing the reduction of TMAO in TMA (2). The proposed electron transfer pathway is thus from menaquinones to TorC N , then from TorC N to TorC C , and finally to TorA.  TorC C domain donates the electrons directly to TorA (Fig. 5). This finding fits well with the positive midpoint redox potential exhibited by the TorC C heme (EЈ 0 ϭ ϩ120 mV) and that of the reduction reaction of TMAO (EЈ 0(TMAO/TMA) ϭ ϩ 130 mV). The rate of electron transfer is slightly but significantly enhanced by the presence of the tetrahemic domain TorC N in the reconstitution system. The enhanced rate in the presence of TorC N probably means that TorC N transfers the electrons to TorC C , since TorC N alone cannot directly feed TorA with electrons (Fig. 5). This result, together with the fact that the presence of both TorC N and TorC C domains are required for bacterial growth with TMAO as a sole exogenous electron acceptor (Fig.  2), support a model in which TorC N receives the electrons from the menaquinone pool and then transfers them to TorC C , which gives them to TorA. In the Nir and Nap respiratory systems, the TorC N homologues (NirT and NapC) and their associated diheme cytochromes (NirB and NapB, respectively) constitute distinct proteins (19,37). In the case of TorC, the TorC N tetrahemic domain is fused to the TorC C monohemic domain, and it is clear from our experiments that the intact TorC protein is more efficient than a mixture of the two isolated domains.
Although the simplest model of interaction between TorC and TorA would have been that TorC C binds TorA, since it transfers the electrons to TorA, we have shown by two different approaches that TorC C does not significantly bind TorA, whereas TorC N binds TorA efficiently (Figs. 6 and 7). The apparent equilibrium dissociation constant obtained by the analysis of the BIAcore data is similar for TorC N and TorC, although a second K d with a higher value can also be calculated for TorC (Table III). This latter might correspond to a weak interaction between TorC C and TorA that takes place after the TorC N -TorA binding, allowing the electron transfer. Based on these findings, we proposed a model in which the tetrahemic domain of TorC transfers the electrons to TorC C and binds TorA in such a way that TorC C is correctly positioned to transfer them to TorA (Fig. 8). This may also explain why TorC N and TorC C are fused. Finally, the binding of the TorC N domain to TorA raises the question of a possible binding of DorC and also of the tetrahemic homologues of TorC to their associated terminal reductase.