TorD, A Cytoplasmic Chaperone That Interacts with the Unfolded Trimethylamine N-Oxide Reductase Enzyme (TorA) in Escherichia coli *

Reduction of trimethylamine N-oxide (TMAO) in Escherichia coli involves the terminal molybdoreductase TorA, located in the periplasm, and the membrane anchored c type cytochrome TorC. In this study, the role of the TorD protein, encoded by the third gene of torCADoperon, is investigated. Construction of a mutant, in which thetorD gene is interrupted, showed that the absence of TorD protein leads to a two times decrease of the final amount of TorA enzyme. However, specific activity and biochemical properties of TorA enzyme were similar to those of the enzyme produced in the wild type. Excess of TorD protein restores the normal level of TorA enzyme, and also, leads to the appearance of a new cytoplasmic form of TorA on SDS-polyacrylamide gel electrophoresis using gentle conditions. This probably indicates a new folding state of the cytoplasmic TorA protein when TorD is overexpressed. BIAcore techniques demonstrated direct specific interaction between the TorA and TorD proteins. This interaction was enhanced when TorA was previously unfolded by heating. Finally, as TorA is a molybdoenzyme, we demonstrated that TorD can interact with TorA before the molybdenum cofactor has been inserted. As TorD homologue encoding genes are found in various TMAO reductase loci, we propose that TorD is a chaperone protein specific for the TorA enzyme. It belongs to a family of TorD-like chaperones present in several bacteria, and, probably, involved in TMAO reductase folding.

In Escherichia coli, the main anaerobic respiratory pathway responsible for reduction of trimethylamine N-oxyde (TMAO) 1 to TMA (trimethylamine) requires the products of the torCAD operon which, in anaerobiosis, is induced in the presence of TMAO or related compounds via the two-component regulatory system, TorS/TorR (1,2).
TMAO reductase (TorA), the terminal reductase of this system, is a 97-kDa molybdoprotein encoded by the torA gene (3,4). Based on sequence homologies, TorA belongs to the Me 2 SO/ TMAO reductase family which includes the M 2 SO/TMAO reductases from Rhodobacter capsulatus and Rhodobacter spha-eroides (5) and TMAO reductase of Shewanella species. 2 These enzymes share several properties: (i) they are all molybdoenzymes located in the periplasm of the bacterium and (ii) in each case, it has been proposed that a membrane anchored pentahemic c type cytochrome feeds electrons to the terminal enzyme (6,7). In E. coli, this cytochrome, TorC, is encoded by torC, the first gene of the torCAD operon (4,8). While the role of the TorC and TorA proteins is well documented, the role of the third protein, TorD, predicted in the torCAD operon is not (4). However, the presence of TorD homologue proteins, not only in R. capsulatus and R. sphaeroides Tor systems (6,7), but also in Shewanella species 2 is intriguing and suggests a similar role for this protein in these systems. As TorD contains two small hydrophobic segments, at its amino and carboxyl ends, respectively, it was proposed to be a membranous b type cytochrome involved in the electron transfer pathway for TMAO reduction (4,9).
In this study, we show that TorD is not a membrane-bound protein but rather a cytoplasmic protein able to interact specifically with the unfolded TorA protein. We propose that TorD belongs to a chaperone family specific for certain molybdoproteins.
Purification of Proteins-The purification of TMAO reductase was performed from the periplasmic fraction of MC4100 cells (initial activity: 11 mol of TMAO reduced/min/mg of protein). The TMAO reductase was purified to homogeneity by DE52 ion exchange chromatography, Mono Q HR 16/10, and preparative electrophoresis (5.5% polyacrylamide). 1 mg of pure TMAO reductase was obtained with a specific activity of 250 mol of TMAO reduced/min/mg of protein. The protein FA, product of the mobA gene, was purified as described by Palmer et al. (12). Enrichment in TorD protein was achieved as follows: the supernatant fraction (16 mg of proteins in 40 mM Tris-HCl, pH 7.6, 1 mM benzamidine-HCl) of strain LCB641/ptorD, grown anaerobically in the presence of IPTG, was loaded on a Sephadex G-75 column. The presence of TorD (Ͼ60% of total proteins) was tested by SDS-PAGE. The band of the expected size for TorD was submitted to amino-terminal sequencing after an electrotransfer on polyvinylidene difluoride membrane. The sequencing was performed using an Applied Biosystem apparatus.
Analytical Procedures-Benzyl viologen (BV)-TMAO reductase activity was measured at 37°C by a spectrophotometric technique (11) based on the oxidation of reduced benzyl viologen at 600 nm coupled to the reduction of TMAO. TMAO reductase activity in polyacrylamide gel was revealed as described above, except that methyl viologen was used, instead of benzyl viologen.
Quinone-TMAO reductase activity assay was as described for nitrate reductase activity assay (13), except that menadiol (menaquinone analog) was used instead of duroquinol. Oxidation of menadiol by extract containing TMAO reductase was coupled to oxidation of NADH by the diaphorase, which catalyzes NADH-menaquinone oxidoreduction.
The amount of TMAO reductase antigen, present in the extracts, was determined by rocket immunoelectrophoresis (14) using a polyclonal antiserum specific for TMAO reductase (100 l). SDS-PAGE was performed using 7.5 or 15% polyacrylamide gels (15). The electrophoretic transfer of proteins to nitrocellulose or membranes and immunodetection with anti-TMAO reductase were carried out as described by Tawbin et al. (16). ECL-Western blotting system was used as recommended by the supplier (Amersham). Protein concentrations were estimated by the technique of Lowry et al. (17).
Biosensor Analysis-The interaction between TorA and TorD proteins was investigated with a biomolecular interaction analysis biosensor based analytical system (BIAcore; Pharmacia). All experiments were performed at 25°C. The TorD containing fraction was dialyzed against 10 mM acetate buffer to reach pH 3. The protein was then immobilized on a sensor chip CM5 (Pharmacia biosensor) through amine coupling. The carboxylic acid groups of a dextran matrix were activated with 70 l (10 l/min) of a mixture of 0.2 M N-ethyl-NЈ(3dimethylaminopropyl)-carbodiimide hydrochloride and 0.05 M N-hydrosuccinimide (18). TorD protein was injected during 8 min (10 l/min), resulting in approximately 2000 resonance units of immobilized protein and the reaction was stopped by the injection of 70 l (10 l/min) of 1 M ethanolamine hydrochloride in order to transform the remaining active esters into amides. This procedure allows the TorD protein to be covalently coupled to the carboxyl dextran-modified gold surface via the exposed amino groups. TMAO reductase, bovine serum albumin, or ␤-galactosidase (0.4 mg/ml) were injected using a constant flow rate of 10 l/min. When required, purified proteins were heated for denaturation during 10 min at 75°C and then placed at 4°C until injection under the same conditions. The sensorgram study was performed using the biomolecular interaction analysis evaluation software (Pharmacia).
Activation of the Inactive TorA Protein in mob Crude Cell Extracts-The activation was performed with either a fixed amount of TMAO reductase apoenzyme of mob strain (40 mg/ml, 200 l) and increasing quantities of purified FA protein (1 g/l) or, alternatively, with increasing amounts of TMAO reductase apoenzyme of mob strain and a fixed amount of the supernatant fraction of strain LCB620 (TorA Ϫ ) grown anaerobically (31 mg/ml, 100 l). The assay mixtures were incubated at 37°C for 60 min. The reaction was stopped by cooling the samples on ice.
DNA Manipulations-Standard procedures for plasmid preparation, restriction endonuclease digestions, agarose gel electrophoresis, DNA purification, and ligation were as described by Sambrook et al. (19). Transformations were performed by the method of Chung and Miller (20). Chromosomal DNA was prepared as described previously (4). PCR (polymerase chain reaction) amplifications were performed as described in Jourlin et al. (2).
Construction of Plasmids pTorAD, pTorD-The torAD or torD coding sequence was amplified by PCR using oligonucleotides Am1 or Dr1 and Db1 and chromosomal DNA preparation from MC4100 cells. The sequence of Am1 (5Ј-ACGGCCAATTGAAGGAAGAAAAATAATGAACA-ATAACG) corresponds to a MunI site followed by the 5Ј sequence of torA. The sequence of Dr1 (5Ј-TCGAATTCAGGAGGTGAAATCATGAC-CACGCTG) corresponds to an EcoRI site, followed by an efficient ribosome-binding site and the 5Ј sequence of torD. The sequence of Db1 (5Ј-CTTGGATCCTTATCTGTTTTGGTGGTCGCAC) corresponds to a BamHI cleavage site, followed by sequence complementary to the very 3Ј-end of the torD coding sequence. The purified PCR products were digested by the appropriate restriction enzymes and introduced into the corresponding cloning sites of the pJF119EH vector (21) to yield plasmid pTorAD or pTorD.
Construction of a Chromosomal torD::⍀ Mutant-The insertion of the ⍀ interposon into torD gene was performed as described for the construction of the torS::⍀ mutant (2). The 2-kilobase SmaI fragment, containing the ⍀ interposon (Sp r ) from plasmid pHP45⍀ (22), was cloned into plasmid pTorAD, linearized at the NheI site located at the beginning of torD gene. The resulting plasmid pTorAD⍀ was digested by ClaI and PstI restriction enzymes and the 3.8-kilobase fragment, which contained the ⍀ interposon and flanking regions, was inserted into the temperature-sensitive Cm r replicon, pMAK705 (23), digested by the same restriction enzymes. The resulting plasmid (pMD⍀) was tested for the presence of the ⍀ fragment. It was introduced into strain MC4100 at 30°C. The cells were then incubated at 43°C to allow integration of the plasmid by homologous recombination into the chromosomal torD gene. Cells were grown at 30°C during 48 h without antibiotics then plated on spectinomycin containing media. 21 Sp r Cm s clones were selected. To ascertain the location of the ⍀ insertion in the chromosomal torD gene, we performed PCR amplifications with chromosomal DNA of 5 clones as template and with a primer located either at the beginning of the torA sequence (Am1) or downstream from the insertion site (Db1), together with a ⍀ primer (2) complementary to both extremities of the ⍀ interposon. In all cases, the ⍀ interposon was correctly located within the torD gene. The absence of the plasmid was also checked in cells grown at 30°C. One clone was retained and called LCB641.

RESULTS AND DISCUSSION
Effect of Absence of TorD on the Tor System-In vitro, TMAO reduction by TorA can be measured using a quinone homologue (menadione) or benzyl viologen (BV) as electron donor. Reduction of TMAO with quinone involves all the components of the electron transfer chain from the membranous quinone pool to the periplasmic terminal reductase. As a result, this activity represents an electron transfer closer to the in vivo mechanism than the BV-TMAO reductase activity involving only the terminal enzyme and the artificial electron donor.
Quinone-TMAO reductase activity was performed on crude extract of strain LCB641 (TorD Ϫ ) grown under inducing conditions. The results reported in Table I shows that a quinone-TMAO reductase electron transfer does occur in strain LCB641. This activity is half of that measured in the cognate wild type strain (Table I). When determined by immunoelectrophoresis rocket, the amount of TorA protein in the TorD Ϫ strain is also decreased two times, compared with the wild type strain, indicating that the TMAO reductase specific activity is similar in both TorD Ϫ and wild type strains. The existence of a relevant quinone-TMAO reductase activity in the TorD Ϫ mutant strain rules out the possibility for the protein TorD to be a constituent of the electron transfer chain. This is in agreement with the production of TMA in vivo by strain LCB641 during the growth in anaerobiosis in the presence of TMAO.
BV-TMAO reductase activity was measured on crude extract of strain LCB641 (TorD Ϫ ). Table I shows that the total TMAO reductase activity in this strain is at least two times lower than in the wild type MC4100 strain. Here again, specific activity is equivalent for both strains.
70% of the total BV-TMAO reductase activity of strain LCB641 (TorD Ϫ ) is recovered in the periplasmic fraction of the a Crude extracts were performed in exactly the same conditions from 1 g of cells of each strain. b TMAO reductase activity was measured as described under "Experimental Procedures" using BV or quinone analog (Quinone) as electron donor.
c The amount of TorA was estimated by immunoelectrophoresis rocket.
d The specific activity is calculated with the quinone analog as electron donor. cells as described previously for the wild type (3). As expected, the amount of TMAO reductase protein in the periplasm is two times lower than the wild type counterpart (Fig. 1). No differences are observed between the TorA protein, synthesized in the mutant, and the wild type strain with regard to the SDSelectrophoretic pattern (Fig. 1A) and molecular weight. K m for TMAO and 4-methylmorpholine N-oxide, as well as substrate specificity, using various N-and S-oxide compounds, are presented in Table II. The results clearly indicate that the high substrate specificity of the TMAO reductase is not affected, when torD gene is inactivated (24).
Upon induction of the torCAD operon in a wild type strain, the pentaheme c type cytochrome TorC is synthesized and anchored to the inner membrane of the cell (4,8). Using a specific coloration of heme proteins we checked the presence of the TorC protein in the membrane fraction of strains LCB641 (TorD Ϫ ), compared with the wild type MC4100. Although the precise quantification of TorC cannot be ascertained, no obvious differences appeared between the two strains (data not shown). Thus, TorD is not involved in the processing or the stability of the TorC cytochrome. Altogether these results show that the absence of TorD only affects the final TMAO reductase concentration in the cell.
Complementation of Strain LCB641(TorD Ϫ ) by pTorD-The torD coding sequence was amplified by PCR and cloned as described under "Experimental Procedures." The resulting plasmid pTorD was then introduced into strain LCB641 (TorD Ϫ ). Strain LCB641/pTorD was grown anaerobically in the presence of TMAO and IPTG. The concentration of TorA protein in this strain, estimated by immunoelectrophoresis rocket, was restored to the wild type level (Table I and Fig. 1B).
Quinone-or BV-TMAO reductase activity measured in crude extract of this strain represents at least 90% of the wild type activity (Table I). Moreover, as for the wild type strain, 70% of BV-TMAO reductase activity as well as 70% of TorA protein were found in the periplasmic fraction of LCB641/pTorD cells. Complementation of the mutated strain LCB641 (TorD Ϫ ) by plasmid pTorD restores both TorA protein concentration and TMAO reductase activity to the wild type level.
TorD, Which Is Located in the Cytoplasm, Is Neither a Hemebinding Protein nor a Transcriptional Regulator-As the TorD sequence exhibits at least two hydrophobic segments at the amino and carboxyl ends of the protein, a membranous location has been proposed. To determine unambiguously in which compartment of the cell the TorD protein is located, the membranous, periplasmic, and cytoplasmic fractions of LCB641/pTorD cells overproducing the TorD protein were analyzed by SDSpolyacrylamide gel electrophoresis. Surprisingly, a thick protein band with an apparent molecular mass close to that calculated from the amino acid sequence (22.5 kDa) was detected only in the cytoplasmic fraction ( Fig. 2A). This band was absent in the soluble fraction of a strain lacking this plasmid (Fig. 2B). The NH 2 -terminal amino acid sequence (MTTLTAQQIA) of the protein containing band was determined, confirming that this protein corresponds to the TorD protein. These results clearly indicate a cytoplasmic location for the TorD protein.
In an earlier study (4), TorD was tentatively proposed to be a b-type cytochrome since spectroscopical evidence for the existence of such a species has been presented previously in the TMAO reduction pathway (9). This hypothesis can now be discarded for three majors reasons: (i) the cytoplasmic location of the TorD protein (Fig. 2), (ii) no heme-binding protein was detected in the cytoplasmic fraction of strain LCB641/pTorD  when analyzed by both TMBZ staining of SDS-PAGE and low temperature spectroscopy (data not shown), and (iii) the physiological electron transfer from a menaquinone analogue to the terminal enzyme still occurs in the TorD-defective strain LCB641 (Table I).
As TorD is located in the cytoplasm, we have also investigated the possibility that TorD is a transcriptional regulator of the torCAD operon. ␤-Galactosidase activity was measured in strain LCB620 grown in anaerobiosis and in the presence of TMAO. In this strain harboring a chromosomal torA-lacZ fusion, the torD gene, which is the last gene of the torCAD operon, is probably not expressed. When plasmid pTorD was introduced in this strain and IPTG added to the growth medium, the ␤-galactosidase activity remained unchanged. In addition, when bandshift assays were performed using the torCAD promotor and the TorD protein partially purified, no retardation of the DNA fragment was observed (data not shown). This strongly suggests that TorD has no affinity for the promotor sequence of the torCAD operon.
Overproduction of TorD Protein Modifies Cytoplasmic TorA Electrophoretic Mobility-As described previously (3), 70% of TorA is located in the periplasm of the cell whereas only about 25% is found in the cytoplasmic fraction. Study of TorA electrophoretic pattern of a strain overproducing TorD was carried out loading non-heated samples on 7.5% SDS-polyacrylamide gel. These conditions allow the TMAO reductase activity to be revealed directly on the gel and makes the detection of different folding states of this protein possible. As shown on Fig. 3, MC4100 and LCB641/pTorD periplasmic fractions present a similar pattern. The BV-TMAO reductase active band is the band recognized by the anti-TorA-specific serum. A major difference is noted concerning the electrophoretic pattern of TorA present in the cytoplasmic fraction. Indeed, although one active band similar to that found in the periplasmic fraction is present in the wild type cytoplasm, this band is missing in the LCB641/ pTorD cytoplasm. In this last case, one active band is detected at a different position. Western immunoblot showed that the protein corresponding to this active band is recognized by the serum. Moreover, the response to the antibodies is stronger than that of the periplasmic protein (Fig. 3). Overproduction of TorD protein in the cell leads to a modification of the cytoplasmic TorA electrophoretic pattern suggesting that the protein can exist in different active conformational states. We then propose that TorD is a specific chaperone for TorA, and, that overproduction of TorD results in an unusual folding of the entire cytoplasmic TorA protein.
TorD Interacts Directly with TorA-If TorD is a TorA-specific chaperone then a binding between TorD and TorA is expected at least when TorA is unfolded. Interaction between these two proteins should be detected using an apparatus dedicated to study protein-protein contact, the BIAcore system (see, for example, Ref. 25). For this purpose, partially purified TorD protein was immobilized on the dextran matrix (see "Experimental Procedures"), while TorA was used in either a native or an unfolded conformation.
An increase in the amount of recovered resonance units was observed after the native TorA protein was injected (Fig. 4A, 2, time Ͼ 440 s). A stronger signal (2.5 times more efficient) was obtained when the TorA protein was previously heated (Fig.  4A, 1). When the pH was changed from 5 to 7, the response was similar but the signal amplitude was less important (data not shown). Identical experiments were performed using an equivalent fraction originating from cells lacking the TorD protein.
None of the tested TorA samples presented a specific interaction with the activated matrix (Fig. 4B, 3). The experiment was completed by checking the specificity of TorD toward denatured TorA protein using two purified proteins, bovine serum albumin and ␤-galactosidase. In both cases, no specific interactions were observed even when the proteins were previously heat denatured (Fig. 4B, 1 and 2). This result clearly shows that TorD interacts with unfolded TorA protein. Together with the in vivo effect of TorD, this strongly suggests that TorD is a TorA-specific chaperone.
Overproduction of TorD in a Molybdenum Cofactor Defective Strain, mobA, Enhances Apoenzyme TorA Stability-The insertion of the molybdenum cofactor into TorA apoenzyme is a cytoplasmic event taking place before the molybdoenzyme  1 and 2) and 50 g of unheated cytoplasm protein (lanes 3 and 4) samples from strain MC4100 (lanes 1 and 3) and LCB641/pTorD (lanes 2 and 4) were loaded on SDS-PAGE 7.5%. Panel A, methyl viologen-TMAO reductase activity. Panel B, Western blot using serum anti-TorA as in Fig. 1A. crosses the inner membrane by a Sec-independent mechanism (26). The question arises whether TorD is able to interact with a TorA protein devoid of molybdenum cofactor. The processing of synthesis, insertion, and maturation of the molybdenum cofactor is a complex pathway involving the products of five transcriptional units (the mo-genes). The last step of this process requires the presence of protein FA, the mobA gene product. It has been shown that the amount of TorA apoprotein present in mo-strains is 3-4 times lower compared with the quantity observed in a wild type strain (11). The plasmid pTorD was introduced in a mobA mutant in which all the molybdoenzymes are devoid of molybdenum cofactor. The amount of TorA apoprotein was estimated in a crude extract of the mob/pTorD strain by immunoelectrophoresis rocket. As shown on Fig. 5, the amount of TorA apoprotein is increased when the torD gene is overexpressed, and reaches the level of enzyme present in the wild type strain. We can therefore conclude that TorD acts on TorA before the insertion of the molybdenum cofactor into the apoprotein and, as a result, probably protects the TorA apoprotein against degradation.
In Vitro Reactivation of TorA When the torD Gene Is Overexpressed-Molybdoapoenzymes from a mobA strain can be activated in vitro in the presence of either the purified protein FA or a crude extract containing FA protein. Since TorD can interact with TorA independently of the presence of the molybdenum cofactor, we analyzed the effect of TorD overproduction in this reactivation system. When the purified FA or FA-containing supernatant fraction (FA ϩ TorA Ϫ strain) was used to reactivate TorA apoprotein, activity was restored to the maximum value of 0.2-0.3 mol of reduced TMAO/min/mg of protein (Fig. 6, A and B). The same experiment was performed using a mobA (FA Ϫ )/pTorD strain instead of the mobA strain. The activity, in this case, was higher since the maximum value obtained was about 0.8 -1 mol/min/mg of protein, explained by the higher quantity of TorA in this strain (Fig. 6, A and B). In the experiment described on Fig. 6C, the amount of TorD protein varies according to the amount of FA protein. Despite the increased quantity of TorD protein, the level of apoTorA activation is not enhanced. Accordingly, the effect of the overproduction of TorD is only observed when both TorD and apo-  /l). The systems were incubated 60 min at 37°C, before BV-TMAO reductase activity was measured. B, activation system containing 100 l of supernatant of LCB620 strain (31 mg/ml) and increasing quantity of the apoTorA containing strains mobA (E) and mobA/pTorD (q). C, TMAO reductase activity is measured in the activation systems which are performed with 200 l of mobA strain (30 mg/ml) and increasing volume of LCB620 (30 mg/ml) (⌬) or LCB620/pTorD (30 mg/ml) (OE) supernatants. TorA are produced in the same cell.
These results indicate that most of the TorA inactive apoprotein present in the mobA/pTorD strain can be activated. The presence of a high amount of TorD protein probably decreases the turnover of the inactive TorA enzyme. Therefore, we propose, that TorD interacts with TorA in an early stage of the enzyme synthesis.
TorD Is Part of a Chaperone Family-When the TorD protein was first mentioned (4), no homologous protein could be found in data banks. Since the release of the sequences of proposed operons encoding the Me 2 SO/TMAO reductases from R. sphaeroides and R. capsulatus (DmsA and DorA, respectively), a gene encoding TorD homologous protein (DmsB and DorD, 26 -27% of identity with TorD, Fig. 7) was found. Recently, the TMAO reductase system in a Shewanella species was also shown to contain a TorD-like protein sharing 34% identity with E. coli TorD 2 (Fig. 7). Starting from the complete sequence of the Hemophilus influenzae genome, we also found a TorD homologous protein. The structural gene (Hi1044) is close to the dms operon encoding a membranous Me 2 SO reductase enzyme. Interestingly, in E. coli, an open reading frame (EcYcac) located downstream from the dms operon and transcribed in the opposite orientation could also encode a TorD homologue. These proteins share an equivalent size (from 199 to 226 amino acids) and show at least two highly conserved regions (position 121-136 and 161-192 of E. coli TorD). We propose that all these proteins are part of a chaperone family specific for Me 2 SO/TMAO molybdoenzymes (27).
In the absence of TorD, the TorA protein presents wild type properties but the amount of the protein is decreased compared with the wild type strain. A tentative explanation is that TorA is capable to fold on its own but with a less efficient ratio. Alternatively, another chaperone, which could be either a general molecular chaperone or a specific chaperone (for example, EcYcac), takes over the folding of TorA. Surprisingly, Moncey et al. (28) have recently proposed that R. sphaeroides DmsB protein possesses an essential role in the TMAO respiratory pathway. Experiments are in progress to confirm that the TorD homologues play the same role toward their cognate enzyme as that of the E. coli TorD protein.
The presence of specific chaperone for molybdoenzymes is probably widespread in bacteria, since a nitrate reductase chaperone has been recently described for E. coli (29,30). Although the NarJ nitrate reductase chaperone shares similar size with TorD, no apparent homologies could be detected between the two primary structures.