Trimethylamine N-oxide (TMAO) is reduced to the volatile compound trimethylamine (TMA) by at least two respiratory systems, the TorCAD and the DmsABC systems (5, 29). The torCAD operon, which encodes the periplasmic Tor system, is induced in the presence of TMAO by the TorS-TorR two-component regulatory system (25), whereas the membranous dimethyl sulfoxide (DMSO) reductase system, encoded by the dmsABC operon, is synthesized constitutively in anaerobiosis (5). The reason for the presence in the same host of several systems dedicated to a common substrate is still unclear, but one possibility is that they allow the cell to better adapt to changing environmental conditions during the different growth phases.

The terminal reductase of the inducible Tor pathway, TorA, encoded by the torA gene, is a periplasmic molybdoenzyme of 90 kDa (29) that is thought to receive electrons from the membrane pool of menaquinone through the TorC protein. TorC, which is encoded by the first gene of the torCAD operon, is a pentahemic c-type cytochrome anchored to the inner membrane by its N-terminal extremity (21, 29). The last gene of the tor operon, torD, encodes a cytoplasmic protein of 23 kDa, which is a private chaperone that is required for TorA assembly (35). TorD interacts with TorA at an early stage of TorA synthesis, probably before the insertion of the molybdenum cofactor. After folding and cofactor insertion, TorA is translocated across the inner membrane by the Tat (twin arginine translocation) system (40, 41, 52). Except for the c-type cytochromes (49), the metalloproteins located into the periplasm are transported by this pathway. The N-terminal signal sequences of the metalloenzymes translocated by the Tat mechanism are recognized by their length (>30 amino acids), the twin-arginine motif RRXFL, and their hydrophobicity (3, 11).

According to sequence homologies and to its biochemical properties, TorA belongs to the DMSO reductase family (20, 22). In this family of molybdoenzymes, three groups of enzymes can be defined, as follows: (i) the specific TMAO reductases encoded by torA of E. coli or Shewanella massilia, which do not have any S-oxide reductase activity and can only reduce TMAO as a natural substrate (13, 23); (ii) the TMAO- DMSO reductases like the DmsA subunit of the E. coli membranous DMSO reductase or the periplasmic DorA and DmsA enzymes of Rhodobacter capsulatus and Rhodobacter sphaeroides, respectively, which are able to reduce a wide range of N- and S-oxide compounds, including DMSO and TMAO (42, 45); and (iii) the biotin d-sulfoxide reductases like BisC from E. coli and the biotin sulfoxide (BSO) reductase from R. sphaeroides, which are cytoplasmic enzymes involved primarily in the recycling of biotin from BSO (33, 34). An in vitro study showed that the R. sphaeroides enzyme can also poorly reduce other N- and S-oxides like TMAO and DMSO (17, 34). Although these three groups of enzymes share sequence homologies, a major difference is that enzymes of the first two groups are involved in anaerobic respiratory processes while the cytoplasmic BSO reductase enzymes are not.

Recently, it has been proposed that a newly characterized gene, bisZ, encodes a second BSO reductase in E. coli (12). Genetic and biochemical evidence showed that the bisZ product was responsible for the 4% background BSO reductase activity observed in a bisC mutant. Moreover, the bisZ gene encodes a protein that exhibits 62% sequence identity with the bisC product. We demonstrate in this paper that, in contrast to BisC, the bisZ gene product is periplasmically located. Furthermore, this enzyme has an extended substrate specificity which includes, in addition to BSO, other N- and S-oxides, such as TMAO. Significantly, it exhibits greater catalytic activity with TMAO than with BSO. Accordingly, we propose to rename bisZ as torZ. We also show that torZ (bisZ) is the second gene of the torYZ (yecK bisZ) operon and that torY encodes a c-type cytochrome homologous to TorC. Finally, TorY and TorZ constitute a respiratory system.

Chemicals. The N- and S-oxide compounds were purchased from Sigma or Aldrich except for d-biotin d-sulfoxide, which was synthesized from biotin as described by Melville (28).

Bacterial strains, plasmids, and growth conditions. The bacterial strains and plasmids used in this work are listed in Table 1. To maintain selection for plasmids or to select for transductant strains, we used antibiotics as follows: ampicillin, 50 μg/ml; chloramphenicol, 10 μg/ml; kanamycin, 50 μg/ml; and spectinomycin, 25 μg/ml. For the biochemical study, cells were grown anaerobically at 37°C on Luria-Bertani (LB) media. The concentration of arabinose or of glucose, added in the growth medium, is detailed for each experiment in Results. Otherwise, growth of E. coli was performed under anaerobic conditions in 3-ml full-cap cuvettes at 37°C with a minimal salt medium (MSM) derived from that described by Bilous and Weiner (4). It contained K₂HPO₄ (3.5%); KH₂PO₄ (1%), (NH₄)₂SO₄ (0.5%), MgSO₄ (0.05%), CaCl₂ (0.015%), Na citrate (0.3%), casein acid hydrolysate (0.15%; Difco), and thiamine hydrochloride (0.002%, pH 7). The MSM was supplemented with 0.5% glycerol as a nonfermentable carbon source and with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) when necessary. Electron acceptors were used at a concentration of 50 mM. The MSM was inoculated at a dilution of 1% with cells grown overnight in LB broth (supplemented with ampicillin when necessary), centrifuged, and resuspended in the same volume of MSM. Growth was monitored in the same full-capped cuvettes at 600 nm. Values (about 0.15 after 24 h) obtained from a control curve with cells in MSM supplemented with glycerol but without electron acceptor were subtracted from the experimental values.

TABLE 1 Bacterial strains and plasmids used in this study

DNA manipulations. DNA was prepared with the High Pure DNA Isolation kit from Boehringer Mannheim. Plasmid preparation, restriction endonuclease digestions, and DNA purification and ligation were carried out as described by Sambrook et al. (39). Transformations were performed according to the method of Chung and Miller (9). PCR amplification was carried out using standard procedures according to the supplier's instructions. Reverse transcriptase-PCR (RT-PCR) was performed with the Promega Access system. The oligonucleotides used were as follows: a, 5′-ATTAGAGGAATCTATGCGAGGGAAAAAACG-3′; b, 5′-GTGGACAGTTCCCTGATATTCCTC- 3′; c, 5′-CGCTATGGCAATACGCTGAAAAACTTG-3′; d, 5′-TAACAATTGACCATGATCAGGGAGGAAGTTATGACATTAAC-3′; and e, 5′-CCGCCGCCGTAGACTGTAAGGAAT-3′. One microgram of total RNA prepared with the High Pure RNA Isolation kit (Boehringer Mannheim) was denatured at 94°C for 2 min in the presence of either primers a and b, primers c and e, or primers d and e. Immediately afterward, reverse transcription and 35 cycles of PCR amplification were carried out according to the supplier's protocol.

Construction of plasmids. To create plasmids ptorZ and ptorYZ, we used PCR to generate DNA fragments corresponding to the torZ and the torYZ coding sequences with, in both cases, an upstream MfeI site and a downstream SmaI site. After enzymatic hydrolysis, the PCR products were cloned into the compatible EcoRI and SmaI sites of pJF119EH (16), yielding plasmids ptorZ and ptorYZ. In these plasmids, torZ and torYZ are under the control of the P_tac promoter. To create the pBtorYZ, the same torYZ PCR product was cloned into the EcoRI and SmaI sites of pBAD24 (19). In this construct, the torYZ genes are under the control of the arabinose promoter. To create plasmid ptorY, the torZ gene was partly deleted from plasmid ptorYZ after an EcoRI digestion, followed by an intramolecular ligation. To create plasmid ptorCAD, we used PCR to generate a promoterless torCAD DNA fragment. This purified PCR product was then cloned into the pPCR-script vector (Stratagene), according to the supplier's protocol, to yield pPStorCAD. The PstI-SacI fragment from pPStorCAD was cloned into the same sites on pJF119EH, resulting in plasmid ptorCAD, in which the entire tor operon is under the control of the P_tac promoter. To create plasmid pPTorYZ, we amplified the torY promoter region by PCR (from position −283 to position +170 relative to the first nucleotide of the initiation codon of torY gene). The DNA fragment was blunted using the blunting kit from Takara and introduced into plasmid pGE593 (14), previously linearized by SmaI, thus placing the lacZ gene under the control of the putative torY promoter. All the PCR products and fusion sites were confirmed by sequencing, except for the torCAD coding region of ptorCAD which was subjected to a PCR amplification of only 13 cycles in order to minimize the number of possible mutations.

Construction of strain LCB504. The mutation of strain DSS401 (Δdms Km^r) was transferred to LCB502 by P1 transduction, resulting in a torC dms strain (LCB504). Integration of the Km^r gene at the correct position on the chromosome was verified by PCR.

β-galactosidase assays. Strain LCB504 carrying pPTorYZ was grown anaerobically at 37°C in LB medium alone or supplemented with 50 mM TMAO, DMSO, or BSO. β-galactosidase activities were measured according to the method of Miller (30) from culture samples that were taken during the exponential- and the stationary-growth phases.

Preparation of subcellular fractions. Crude extracts were prepared by disrupting the cells in a French press as described by Iobbi-Nivol et al (24). The periplasmic fractions were prepared according to the sucrose-lysosyme-EDTA procedure described by Osborn et al. (32). Membranous and cytoplasmic fractions were obtained from the spheroplasts after disruption in a French press and ultracentrifugation as detailed by Silvestro et al. (44).

Enzyme purification. TorZ (BisZ) was purified from the periplasm fraction obtained from 4 g of LCB620/pBtorYZ cells grown anaerobically in the presence of 0.1% arabinose and ampicillin. The periplasm was dialyzed to remove the sucrose and applied to an ion-exchange Q Sepharose, HiLoad 16/10 column (Pharmacia-Biotech). The fractions obtained from a 0 to 0.5 M NaCl linear gradient elution were assayed for TMAO reductase activity, and active fractions were analyzed by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) (26). Two fractions, possessing also the highest specific activity, presented only two Coomassie blue-stained bands each. The major one, corresponding to more than 80% of the stained proteins, had a calculated molecular mass of about 90 kDa, whereas the minor band was slightly smaller. These fractions were then tested for TMAO and DMSO reductase activities on SDS gels according to the procedure described by Pommier et al. (35). The single active band obtained for both the TMAO and DMSO activities corresponded to the major Coomassie blue band. The two active fractions were pooled and used for analytical experiments. A total quantity of 1.76 mg of protein was obtained with a specific activity of 709 μmol of TMAO reduced per min per mg of protein.

Analytical procedures. TMAO reductase activity was measured spectrophotometrically at 37°C by following the oxydation of reduced benzyl viologen at 600 nm coupled to the reduction of TMAO. Each of the compounds tested (see Table 4 for a list) was prepared at a final concentration of 1 or 2 M and used at least at eight different concentrations with three to four duplicates. Kinetic parameters were determined as previously described (23).

TABLE 4 Kinetic parameters with different electron acceptors catalyzed by the torZ (bisZ) product

The N-terminal sequence of TorZ (BisZ) was determined by Edman degradation (model 470a; Applied Biosystems) after electroblotting of TorZ onto polyvinylidene difluoride membrane (50).

The plasmid pEC86, carrying all the ccm genes, whose products are involved in the maturation mechanism of the c-type cytochromes, was introduced into the LCB620/pBtorYZ strain to increase the amount of mature TorY (2). The presence of covalently bound hemes in c-type cytochromes was revealed by staining for peroxidase activity using 3,3′,5,5′-tetramethylbenzidine (TMBZ) (48). Protein concentration was estimated by the procedure of Lowry et al. (27).

Gene organization and putative products of the bisZ locus. The torZ (bisZ) gene is located at kb 1955 on the chromosome of E. coli. The analysis of the DNA sequences surrounding this gene revealed the presence of a 1-kb open reading frame that we propose to call torY (formerly designated yecK), located immediately upstream from torZ (bisZ) and transcribed in the same orientation (Fig. 1). The distance between torY (yecK) and torZ (bisZ) is only 24 bp, while a noncoding region of 387 bp separates torY (yecK) from the upstream cutC gene which is transcribed in the same orientation and ends with a probable transcriptional terminator. The next gene downstream from torZ (bisZ) (yecP) is transcribed divergently. To determine whether the torY (yecK) and torZ (bisZ) genes are organized in a single transcription unit, RT-PCR was performed using RNA extracted from a strain grown anaerobically and appropriate oligonucleotide pairs that hybridize to regions in the beginning or within torY (yecK) and within torZ (bisZ), as shown in Fig. 1. The PCR synthesis of a DNA fragment that overlaps the region between the end of torY (yecK) and the beginning of torZ (bisZ) strongly suggests that these genes are organized in an operon. Amplification of DNA fragments that cover the beginning of torY (yecK) indicates that the transcriptional start of this operon is, as expected, upstream from the ATG start codon of torY (yecK). The intergenic region located between the cutC and the torY (yecK) genes was cloned upstream from the promoterless lacZ gene of the multicopy plasmid pGE593 to yield plasmid pPTorYZ. Unfortunately, the β-galactosidase activity measured from strain MC4100 carrying pPTorYZ was very low (10 to 20 Miller units) under all the growth conditions tested (see Materials and Methods). This result suggests either that the torYZ (yecK bisZ) operon is always expressed at a very low level or that the conditions for inducing this operon have not been discovered.

FIG. 1 (a) Physical map of the torYZ operon region of E. coli. The large arrows show the locations and the orientations of the reading frames. The primers (a, b, c, d, and e) used for the RT-PCR are indicated. (b) Analysis of tor gene transcription by RT-PCR (more ...)

Nucleotide analysis indicates that a strong ribosome binding site (GAGGA) is located 9 bp upstream from the start codon of both torY (yecK) and torZ (bisZ). The sequence of the torY (yecK) gene product (TorY) revealed five consensus heme binding sites (CXXCH); thus, TorY is likely to be a pentahemic c-type cytochrome. The presence of a hydrophobic segment of 20 residues located in the N-terminal extremity of TorY suggests that TorY is anchored to the inner membrane. TorY is homologous to TorC and DorC (36 and 37% identity, respectively), and, like TorC and DorC, it seems to contain two domains: an N-terminal tetrahemic domain presenting homologies with the members of the NirT family and a C-terminal monohemic domain which appears only in c-type cytochromes involved in TMAO and DMSO respiratory systems (13, 31, 37).

As previously described by del Campillo-Campbell and Campbell (12), the primary amino acid sequence of the torZ (bisZ) gene product shares homologies with the DMSO reductase family of the molybdoenzymes (20). The best score in the homology search was observed with the cytoplasmic BSO reductase of E. coli (62% identity) (33). Moreover, the torZ (bisZ) product exhibited high similarity scores with periplasmic molybdoenzymes like DorA from R. capsulatus, DmsA from R. sphaeroides, TorA from E. coli, and TorA from S. massilia (46, 46, 42, and 38% identity, respectively) (13, 29, 43, 53). Interestingly, a significant difference between the sequences of the torZ (BisZ) product and BisC, which was not previously pointed out, is that the N-terminal part of TorZ (BisZ) shows the characteristics of a signal peptide specifically found in molybdoenzymes such as TorA, DorA, and DmsA, (Fig. 2) (3). The presence of a motif RRXFI, close to the classical twin-arginine motif (RRXFL), followed by a long hydrophobic segment and by the consensus AXA cleavage site led us to propose that, in contrast to the cytoplasmic BisC protein, the product of torZ (bisZ) is exported to the periplasm.

FIG. 2 Amino acid sequence alignment of the N-terminal part of the DMSO-TMAO reductases of R. sphaeroides (RsDmsA) (53) and R. capsulatus (RcDorA) (43), the TMAO reductases of S. massilia (SmTorA) (13) and E. coli (EcTorA) (29), the product of the H. influenzae (more ...)

This analysis allowed us to conclude that the two genes, torY (yecK) and torZ (bisZ), which appear to be organized in a single transcription unit, should encode a membrane-bound pentahemic cytochrome and an unexpected periplasmic molybdoprotein, respectively. The homologies observed between the components of the electron transfer chain of both the Tor and Dor respiratory systems and the products of this new operon suggested to us that it may encode a novel respiratory system reducing N- and/or S-oxide compounds.

torYZ (yecK bisZ) operon encodes a respiratory system. To test our hypothesis that the TorY and TorZ proteins are involved in an electron transfer pathway similar to that of the TorCAD system, a strain that was unable to reduce TMAO was constructed. This strain, LCB504, carries an interposon in the beginning of the first gene of the torCAD operon and a deletion of the entire dmsABC operon (Table 1). As expected, strain LCB504, containing the pJF119 expression vector, grew at an extremely slow rate under anaerobic conditions with TMAO as the only exogenous electron acceptor (Fig. 3). When plasmid ptorCAD, carrying the torCAD operon under the control of the P_tac promoter, was introduced into strain LCB504, the recombinant strain exhibited a high growth rate in the presence of IPTG. Introduction of plasmid ptorYZ, carrying the torYZ (yecK bisZ) operon under the control of the P_tac promoter, into strain LCB504 also yielded a recombinant strain with similar IPTG-dependent growth rate (Fig. 3). In the absence of IPTG, no significant growth was observed with either recombinant strain (Fig. 3). It is noteworthy that production of the characteristic odor of volatile TMA occurred under condition of rapid growth, indicating that TMAO was reduced to TMA in both experiments. Therefore, when expressed to a certain level, the torYZ (yecK bisZ) operon, like torCAD, allows the strain to use TMAO as a substrate for anaerobic respiration.

FIG. 3 Anaerobic growth profiles of E. coli strain LCB504 carrying plasmid ptorYZ or plasmid ptorCAD. The LCB504 recombinant strains were grown in MSM in the presence of 1 mM IPTG (unless otherwise indicated) and of a 50 mM concentration of either TMAO (strains (more ...)

Because the torZ (bisZ) gene product was previously shown to encode a BSO reductase enzyme (12), we also tested the growth of strain LCB504 carrying ptorYZ when BSO was added to the medium. Again, growth was observed, but it did not reach either the rate or the yield obtained with TMAO, indicating that TMAO serves as a better electron acceptor than BSO for the terminal reductase under these conditions (Fig. 3). This result confirms that BSO can be reduced by the TorZ (BisZ) enzyme. In contrast, the expression of the plasmid born torCAD operon allowed no bacterial growth in the presence of BSO (Table 2), suggesting that the specificity of TorA and TorZ are quite different.

TABLE 2 Anaerobic growth of LCB504 carrying either ptorYZ or ptorCAD with N- and S-oxide compounds

As shown previously, unlike the membranous DMSO reductase and the Rhodobacter DorA enzyme, TorA is unable to reduce DMSO (23, 42, 43, 45). As expected, when the strain LCB504 harboring plasmid ptorCAD was grown in minimal media in the presence of DMSO, no growth was detected (Table 2). In contrast, when the strain carried the ptorYZ plasmid, the bacteria grew quite well (Fig. 3), confirming that the TorYZ system has a substrate specificity broader than that of TorCAD. To extend this analysis, we compared the effect of various N- and S-oxide substrates on the bacterial growth of strain LCB504 carrying either ptorYZ or ptorCAD (Table 2). Expression of torYZ allows E. coli to grow anaerobically on a wider range of substrates than does expression of torCAD.

To establish that the TorY cytochrome is required for electron transfer to the TorZ (BisZ) terminal enzyme, we cloned the torZ (bisZ) gene alone under the control of the P_tac promoter. The strain carrying the ptorZ plasmid did not grow in the presence of TMAO, DMSO, or BSO (Fig. 3 and data not shown). The torY (yecK) product is thus required in the respiration pathway. Similarly, the torZ (bisZ) gene of ptorYZ was inactivated, leading to plasmid ptorY, and, as expected, no significant growth was allowed (Fig. 3).

This set of in vivo experiments clearly shows that, in contrast to the BisC enzyme, the torZ (bisZ) product together with the torY (yecK) product is involved in an anaerobic respiratory system. Moreover, this system can use several exogenous electron acceptors, including TMAO, BSO, and DMSO. As TMAO seems to be the most efficient substrate for this respiratory system, we conclude its operon should be called the torYZ operon.

TorYZ respiratory system is made up of a membrane-anchored cytochrome and a periplasmic reductase. For the production of both the torY (yecK) and torZ (bisZ) products, the genes were cloned under the strict control of the P_bad promoter into the pBAD24 vector. The resulting plasmid, pBtorYZ, was then introduced into strain LCB620 and grown in the presence of either arabinose or glucose for the induction or repression, respectively, of torYZ expression. First, to confirm the predictions deduced from the sequence analysis about the cellular location of TorY as well as the presence of heme in it, various fractions of the strain, carrying both plasmids pBtorYZ and pEC86, grown in presence of 0.2% arabinose or 0.2% glucose were tested by SDS-PAGE followed by heme staining. In the membrane fraction of cells grown in the presence of arabinose, a cytochrome of about 40 kDa was observed (Fig. 4). It is distinguishable from TorC by its slightly smaller size, and it is absent when glucose is added to the growth media or in the soluble extracts of the bacteria (Fig. 4 and data not shown). Only five c-type cytochromes have been described previously for E. coli: two of them are involved in the periplasmic nitrate respiratory system (Nap), two others are involved in the nitrite reduction pathway (Nrf), and one, TorC, belongs to the TMAO respiration system (21). This study shows that the E. coli genome contains a gene, torY, able to encode a sixth c-type cytochrome. According to the results obtained from a search in the Colibri databank using the heme binding motif as a pattern, no other multihemic cytochrome seems to be encoded by the E. coli genome.

FIG. 4 Detection of c-type cytochromes by TMBZ staining. Membranous proteins (50 μg) from cells carrying pBtorYZ grown anaerobically at 37°C in LB medium supplemented with 0.2% glucose (lane 3) or supplemented with 0.2% arabinose (more ...)

Results obtained during the growth experiments led us to test the torZ (bisZ) gene product for in vitro TMAO reductase activity. As expected, a significant TMAO reductase activity was measured in the supernatant of cells carrying pBtorYZ grown in the presence of 0.1% arabinose (5.5 μmol/min/mg of protein), whereas when 0.1% glucose was added to the growth media or when the strain contained only the vector, no significant activities were observed (0.05 and 0.02 μmol/min/mg of protein, respectively). These results support the idea that TorZ (BisZ) is a soluble enzyme.

To distinguish the location of this protein in the cell, periplasmic and cytoplasmic fractions were prepared from cells grown in the presence of arabinose. A total of 80% of the TMAO reductase activity was recovered in the periplasmic fraction of the cell (Table 3). This result is in agreement with the presence of a putative signal sequence in the N-terminal part of TorZ (BisZ). To determine the position of the cleavage site, the N-terminal extremity of TorZ (BisZ) was sequenced after purification of the protein from the periplasmic fraction. The obtained sequence (EEKGGKIL) corresponds to positions 31 to 39 of the deduced amino acid sequence and follows the consensus AXA cleavage site, as indicated in Fig. 2. Usually, the periplasmic metalloenzymes possessing the twin-arginine motif are transported across the inner membrane by the Tat system (41). To show that translocation of TorZ (BisZ) involves the Tat pathway, a tatC strain (B1LK0) lacking the pore-forming protein, was transformed by plasmid pBtorYZ. We observed that, in this recombinant strain grown with arabinose, the TMAO reductase activity in the periplasm is about 16 times lower than that in the wild type (Table 3). Moreover, most of the activity is found in the cytoplasm in the tat strain. Therefore, the transport of this enzyme across the membrane is dependent on the Tat system.

TABLE 3 Repartition of TMAO reductase activity in strains carrying the pBtorYZ plasmid

Taking these results altogether, it appears that, in contrast to the BisC protein, the torZ (bisZ) product is located in the periplasm and involved with TorY in a respiratory process. All these findings strengthen the idea that this new respiratory system resembles the TorCAD system. However, the fact that the TorYZ system allows growth in the presence of S-oxides suggests that the catalytic properties of TorZ (BisZ) are quite different from those of TorA (23).

Catalytic properties of TorZ (BisZ). The kinetic study was performed on the purified product of torZ (bisZ) as described in the Materials and Methods. The compounds tested were structurally related to TMAO or DMSO and have been previously used to determine the specificity of TorA and the membranous DMSO reductase of E. coli (23, 45).

As shown in Table 4, the best catalytic efficiency (V_max/K_m), which takes into account both the affinity of the enzyme for the substrate and the substrate turnover, is obtained for TMAO, 4-methylmorpholine N-oxide, and BSO, in decreasing order. The details of the analysis indicate that although the K_m obtained with TMAO is higher than that obtained with BSO, the catalytic efficiency measured with TMAO is more than 2 times higher than that measured with BSO. This is due to the fact that the V_max measured with TMAO is about 10 times higher than that observed with BSO. We can then conclude that TMAO is a more efficient substrate than BSO. This conclusion accommodates the differences in growth rate observed in the presence of TMAO, BSO, and 4-methylmorpholine N-oxide (Fig. 3 and Table 2).

Except for TMAO and 4-methylmorpholine N-oxide, which both appear to be good substrates for TorZ (BisZ), the N-oxide compounds tested in this experiment are not very efficiently reduced by TorZ (BisZ). Nitrate and nitrite, which are widespread N-oxide compounds, have also been tested, but neither was reduced at a significant rate by TorZ (BisZ) (Table 4). These results are reminiscent of those obtained during the kinetic study of TorA (23). Nevertheless, in contrast to TorA, the torZ (bisZ) product is capable of sulfoxide reduction. Indeed, the kinetic parameters determined for BSO, tetramethylene sulfoxide (TMSO), and dl-methionine sulfoxide indicate that TorZ (BisZ) can reduce these compounds, and among them, BSO is the best substrate. However, the TMAO reduction catalytic efficiency is never reached with any S-oxide. DMSO was so weakly reduced that kinetic parameters for the enzyme could not be determined without ambiguity. This was surprising, since a significant growth rate was observed with this compound (Fig. 3 and Table 2). To investigate this apparent discrepancy, a competition experiment was carried out to determine whether DMSO can bind the active site of the enzyme. The thermodynamic parameters of the TMAO reductase activity of TorZ (BisZ) were modified when DMSO (11.4 mM) was added in the assay (K_mTMAO^DMSO = 41 mM, V_maxTMAO^DMSO = 120 M s⁻¹). Therefore, DMSO, which is a weak substrate for the enzyme in vitro, acted as a competitive inhibitor towards TorZ. These results fit a mixed-alternative-substrate model (data not shown) (15) and emphasized the difference in substrate specificity between this enzyme and TorA since DMSO does not compete with TMAO in the catalytic site of TorA (23).

During this study, we have demonstrated that the torYZ (yecK bisZ) operon of E. coli encodes a new respiratory system. This system is made up of a membranous pentahemic c-type cytochrome and a periplasmic molybdoreductase, and it is closely related to the Tor and Dor respiratory systems of E. coli and Rhodobacter species (29, 31, 51). The association of a pentahemic membrane-bound cytochrome with a periplasmic enzyme seems to be the hallmark of respiratory systems capable of reducing either specifically TMAO (the Tor system of E. coli and of S. massilia) or TMAO and related N- and S-oxide compounds (the Dor system of Rhodobacter species). Accordingly, the best substrate among those tested for the TorYZ is TMAO. However, our in vitro and in vivo studies revealed that the specificity of TorZ is different from that of either TorA or DorA. In contrast to TorA, TorZ can reduce other N- and S-oxide compounds such as BSO, and in contrast to DorA, the best substrate for TorZ is not a sulfoxide compound but TMAO.

To avoid any confusion, we proposed to rename BisZ as TorZ because this enzyme is a better TMAO reductase than a BSO reductase and because it is located in the periplasm, whereas BisC (the higher TorZ [BisZ] homologue) is a cytoplasmic protein which is not involved in a respiratory process. Our data are in agreement with those of del Campillo-Campbell and Campbell (12), showing that TorZ can reduce BSO quite efficiently, but, from our results, it is probable that reduction of BSO takes place mainly in the periplasm rather than in the cytoplasm. It would be interesting to test whether TorZ can also reduce BSO when located in the cytoplasm. If this happens, the electron donor could be different from TorY which, most probably, faces the periplasm.

In E. coli, two homologous nitrate respiratory systems, nitrate reductase G and nitrate reductase Z, have been described as having issued from a duplication of ancestral genes (6, 24). For a long time, the narZYWV operon encoding nitrate reductase Z was supposed to be expressed constitutively at a very low level. However, a recent study showed that expression of this operon is induced in the early stage of the stationary phase of cell growth in a RpoS-dependent manner (8), while the nitrate reductase G is synthesized in anaerobiosis when a high concentration of nitrate is available (10, 36, 46). The torYZ operon originates probably from a duplication of the torCAD genes. In the case of the TorYZ system, no obvious regulation has been highlighted so far. In particular, the expression of the torYZ operon was very low whatever the growth phase and did not increase in the presence of TMAO, DMSO, or BSO. One can imagine that either the torYZ operon is constitutively expressed at a very low level or the induction conditions of its expression are still unknown. This raises the question of the existence of an unknown inducer for the torYZ operon which might be the best natural substrate of this respiratory system. The fact that the specificity of the TorYZ system is not exactly that of the TorCAD or the DmsABC system of E. coli suggests that the TorYZ system has evolved from the TorCAD system to play a specific role in E. coli respiration. It would be interesting to see whether related bacteria, such as Salmonella species, contain a TorYZ homologue and, if present, how this system is expressed. In this line of thought, a gene homologous to torZ (bisZ) was found in the chromosome of Haemophilus influenzae (47). This putative gene has been called bisZ, but the presence of a potential signal sequence in the N-terminal extremity of its deduced amino acid sequence indicates that it encodes a periplasmic protein (Fig. 2). Moreover, the presence before it of a gene encoding a pentahemic cytochrome homologous to TorC and to TorY raises the question of a possible respiratory role of this bisZ product.

An obvious difference between the torYZ and the torCAD operons is that a torD gene homologue is not found in the former. TorD is a private chaperone for the terminal reductase TorA, and its absence in a torD strain led to a significant decrease in the quantity of the TorA protein compared to that observed with a wild-type strain, but even in such a mutant strain, 30% of the active TorA protein is present in the periplasm (35). Our result indicates that TorZ is synthesized and folded in a torCAD strain. Thus, TorD is not required for TorZ maturation. Therefore, TorZ folding involves either no chaperone or a TorD homologue whose gene is located elsewhere on the chromosome. Interestingly, we have found several genes which could encode TorD homologues on the E. coli chromosome (unpublished results). We are investigating whether one of these TorD homologues plays the role of a TorZ chaperone.

We gratefully acknowledge J. Demoss for critical review of this manuscript. We are indebted to M.-T. Giudici-Orticoni for her help during the kinetic study, P. Brun for advice in the preparation of BSO, T. Palmer for providing strain B1LK0, and J. Weiner for providing strain DSS401. We thank M. Lepelletier and D. Cazeilles for technical assistance.

This work was supported by grants from the Centre National de la Recherche Scientifique and the Université de la Méditerranée and by an MRT fellowship to S.G.