Enzymatic and physiological properties of the tungsten-substituted molybdenum TMAO reductase from Escherichia coli


  • Jean Buc,

    1. Laboratoire de Chimie Bactérienne, UPR9043, IFRC Biologie Structurale et Microbiologie, CNRS, 31 chemin Joseph Aiguier, 13402 Marseille cedex 20, France.,
    Search for more papers by this author
    • †The first two authors contributed equally to this work

  • Claire-Lise Santini,

    1. Laboratoire de Chimie Bactérienne, UPR9043, IFRC Biologie Structurale et Microbiologie, CNRS, 31 chemin Joseph Aiguier, 13402 Marseille cedex 20, France.,
    Search for more papers by this author
    • †The first two authors contributed equally to this work

  • Roger Giordani,

    1. Laboratoire de Botanique Cryptogamie et Biologie Cellulaire, PAON, EA 864, Faculté de Pharmacie, Université de la Méditerranée, 27 Bd Jean Moulin, 13385 Marseille cedex 05, France.,
    Search for more papers by this author
  • Mirjam Czjzek,

    1. Laboratoire Architecture et Fonction des Macromolécules Biologiques, IFRC Biologie Structurale et Microbiologie, CNRS, 31 chemin Joseph Aiguier, 13402 Marseille cedex 20, France.
    Search for more papers by this author
  • Long-Fei Wu,

    1. Laboratoire de Chimie Bactérienne, UPR9043, IFRC Biologie Structurale et Microbiologie, CNRS, 31 chemin Joseph Aiguier, 13402 Marseille cedex 20, France.,
    Search for more papers by this author
  • Gérard Giordano

    1. Laboratoire de Chimie Bactérienne, UPR9043, IFRC Biologie Structurale et Microbiologie, CNRS, 31 chemin Joseph Aiguier, 13402 Marseille cedex 20, France.,
    Search for more papers by this author

Gérard Giordano. E-mail giordano@ibsm.cnrs-mrs.fr; Tel. (+33) 4 91 16 42 12; Fax (+33) 4 91 71 89 14.


The trimethylamine N-oxide (TMAO) reductase of Escherichia coli is a molybdoenzyme that catalyses the reduction of the TMAO to trimethylamine (TMA) with a redox potential of + 130 mV. We have successfully substituted the molybdenum with tungsten and obtained an active tungsto-TMAO reductase. Kinetic studies revealed that the catalytic efficiency of the tungsto-substituted TMAO reductase (W-TorA) was increased significantly (twofold), although a decrease of about 50% in its kcat was found compared with the molybdo-TMAO reductase (Mo-TorA). W-TorA is more sensitive to high pH, is less sensitive to high NaCl concentration and is more heat resistant than Mo-TorA. Most importantly, the W-TorA becomes capable of reducing sulphoxides and supports the anaerobic growth of a bacterial host on these substrates. The evolutionary implication and mechanistic significance of the tungsten substitution are discussed.


Tungsten (W) and the chemically analogous metal molybdenum (Mo) are minor, yet equally abundant, elements on this planet (Johnson et al., 1996; Kletzin and Adams, 1996). The essential role of molybdenum in biology has been known for decades, and molybdoenzymes are ubiquitous. It is only recently that a biological role for tungsten has been established in prokaryotes, although not as yet in eukaryotes. The best characterized organisms that appear to be obligately tungsten dependent are hyperthermophilic archaea. Attempts to grow various organisms in the presence of tungstate resulted in the production of either inactive Mo-enzymes lacking any metal or W-substituted molybdoenzymes with little or no catalytic activity (Kletzin and Adams, 1996). Thus, in spite of the high degree of chemical similarity between these two elements, organisms are highly selective.

Four distinct types of tungstoenzymes have been purified from various organisms. These are formate dehydrogenase (FDH), formylmethanofuran dehydrogenase (FMDH), acetylene hydratase (AH) and aldehyde-oxidizing enzymes such as carboxylic reductase (CAR) and aldehyde ferredoxin oxidoreductase (AOR) (Kletzin and Adams, 1996). With the exception of acetylene hydratase, all of the tungstoenzymes known so far catalyse reactions of very low potential (≤ −420 mV) and have all been purified from strictly anaerobic organisms. To catalyse a reaction of extremely low redox potential, tungsten should be ‘preferred’ over molybdenum. Nevertheless, for some of these tungstoenzymes, there is a molybdoenzyme counterpart that catalyses the same chemical reaction (Kletzin and Adams, 1996). On the contrary, molybdoenzymes are more suitable for catalysing reactions with relatively high redox potential, and these molybdoenzymes do not have tungsto-containing counterparts. W-substituted molybdoenzymes are generally inactive because of the lower reduction potential of the tungsten site with respect to the molybdenum site (Kletzin and Adams, 1996).

The trimethylamine N-oxide reductase (TMAOR; EC of Escherichia coli is a periplasmic molybdoenzyme composed of two identical subunits of 90 kDa, which is encoded by the torA gene of the torCAD operon (Silvestro et al., 1989; Méjean et al., 1994). It catalyses the reduction of trimethylamine N-oxide (TMAO) to trimethylamine (TMA) and functions as a component of an anaerobic respiratory chain, which provides energy for bacterial growth (Barrett and Kwan, 1985). This enzyme can use only TMAO as a substrate, although its biosynthesis is induced by both TMAO and dimethyl sulphoxide (DMSO) (Iobbi-Nivol et al., 1996). Under anaerobic growth conditions, E. coli synthesizes a membrane-bound DMSO reductase (DMSOR) composed of DmsA, DmsB and DmsC (Bilous et al., 1988). In contrast to the TMAO reductase, the DMSO reductase can use both DMSO and TMAO as substrates. The redox potential for the reactions (CH3)2SO + 2H+ + 2e → (CH3)2S + H2O and (CH3)3NO + 2H+ + 2e → (CH3)3N + H2O are + 160 mV and + 130 mV (at pH 7) respectively.

In this paper, we report a successful in vivo tungsten substitution of the molybdoenzyme TMAOR. Compared with the molybdoenzyme, the substituted tungstoenzyme is more heat resistant and is capable of catalysing the high potential conversion of not only TMAO/TMA but also DMSO/DMS. Moreover, the substituted enzyme confers on the host cells the capacity of using DMSO as substrate for their anaerobic growth, implying an important physiological advantage. We thus provide the first example showing that tungsten can replace molybdenum in catalysing reactions at high redox potential.


Identification of tungstate-dependent TMAO reductase

In E. coli, molybdate and tungstate are taken up mainly through a periplasmic binding protein-dependent ABC transport system encoded by the mod operon (Maupin-Furlow et al., 1995). Deficiency in the Mod transporter reduces the intracellular molybdenum concentration to less than 10% of the wild-type level (Scott and Amy, 1989). The poor intracellular molybdenum availability results in a pleiotropic effect on the activities of molybdoenzymes. This defect, however, can be phenotypically suppressed by the addition of molybdate in the growth medium (Campbell et al., 1985). In this case, molybdate is taken up through the sulphate transport system (Rosentel et al., 1995), and the intracellular molybdenum concentration in the mod mutants is restored to the level of a wild-type strain. Likewise, tungstate is taken up through the same ABC transporter as molybdate when it is present at lower concentration and via the sulphate transport system when present at higher concentration. We exploited the particular phenotype of the mod mutant and observed recently that tungstate is capable of conferring a TMAOR activity on the mod mutant (Santini et al., 1998). In addition, by following the β-galactosidase activity that was expressed from the torA::MudII(lacZ, KanR) fusion, we observed that neither molybdate nor tungstate affects the expression of torA.

This apparent tungstate-dependent restoration of TMAOR activity might result from contamination of the growth medium by molybdate. In fact, the sodium tungstate used may contain less than 0.002% molybdenum. We therefore compared the activation curves of the TMAOR of the mod mutant by molybdate or tungstate. With such an eventual contamination, 100-fold more tungstate would be required in order to attain the same level of activation as we observed (data not shown). We also measured the tungsten and molybdenum contents in the periplasms of the mod mutant grown with molybdate or tungstate. The mod–torA double mutant was used to measure the background level of molybdenum and tungsten, which was unrelated to the only TMAOR encoded by torA (also called TorA) of E. coli. The TorA protein content in these fractions was determined by immunoelectrophoresis. When the cells were grown in the presence of molybdate, about 601 μg l−1 molybdenum was found in the periplasm, which corresponds to 1.3 atoms of Mo mol−1 TorA. This value is in good agreement with the idea that each molecule of TorA binds one molybdopterin (Czjzek et al., 1998). No tungsten was detected in this fraction. When the cells were grown with tungstate, about 165 μg l−1 tungsten was found in the periplasm, giving a ratio of 1.2 atoms of tungsten mol−1 TorA. The content of tungsten is about 14% that of the molybdenum, which is highly consistent with our previous observation that the amount of TMAOR in the periplasm of tungstate-grown cells was about 10% that of the mod mutant grown with molybdate (Santini et al., 1998). Most importantly, no molybdenum was detected in the periplasm prepared from the mod mutant grown with tungstate. Therefore, the TMAOR activity seems to be contributed by a tungstoenzyme.

The tungstate-dependent TMAOR activity might be related to either the TMAOR encoded by torA or the presence of another unknown tungstoenzyme. We therefore assessed the correlation between the TMAOR activity and the enzyme TorA by rocket immunoelectrophoresis (Fig. 6B1). Using this technique, TorA is specifically immunoprecipitated by an antiserum raised against this protein, and TMAOR activity is revealed by activity staining. The enzymatically active arcs detected in the periplasmic fraction overlapped exclusively with the immunoprecipitates, whether the cells were grown with molybdate or tungstate. In addition, interruption of the torA gene resulted in a simultaneous disappearance of both the tungstate-dependent TMAOR activity and TorA-related immunoprecipitates. These results confirm the authenticity of the tungstate-dependent TMAOR as the product of the torA gene. The authenticity was proved further by purification of the TMAO reductase.

Figure 6.

. Visualization of TMAOR and DMSOR activities. Periplasmic (5 μg of protein each, except 20 μg for lane 4) fractions prepared from the wild-type strain (lane 1), the torA mutant (lane 2) and the mod mutant restored by molybdate (lane 3) or tungstate (lane 4) were analysed by a native PAGE (10%, A) or immunoelectro- phoresis with antiserum against TorA (B). TMAOR and DMSOR activities were visualized by activity staining using TMAO (A1 and B1) or DMSO (A2 and B2) as electron donors.

Purification of TMAO reductase from the mod mutant grown with tungstate

Using our standard TorA purification protocol, we purified 0.5 mg of TMAOR to near homogeneity from 38 g of cells of the mod mutant grown with tungstate (Fig. 1). The TMAOR was purified 65-fold with a specific activity of 41 μmol of TMAO reduced min−1 mg−1 purified enzyme in an overall yield of 15%. It had a molecular size of 90 kDa and was specifically recognized by the antiserum against TorA (Fig. 1). These results confirm that the purified TMAOR is the torA gene product. The metal content of the purified enzyme was then analysed by inductively coupled plasma emission spectroscopy (ICPES). The results revealed the presence of 0.89 atoms of tungsten mol−1 purified enzyme, but no detectable molybdenum was found in the purified enzyme. Therefore, we have successfully substituted the molybdenum with tungsten to obtain an enzymatically active TMAO reductase. We called this tungsten-substituted enzyme W-TMAOR or W-TorA, compared with Mo-TMAOR or Mo-TorA that are used to describe the molybdo-TMAO reductase.

Figure 1.

. Purification of W-TorA. The mod mutant was grown with tungstate under anaerobic conditions until stationary phase. A periplasmic fraction (P) was prepared, and W-TorA was purified by chromatography on DEAE-cellulose (DE52) and MonoQ (HR5/5) ion exchange columns. Fractions eluted before (lanes 1–3) or after (lanes 5–8) the peak fraction (lane 4) obtained from the MonoQ column were analysed by SDS–PAGE (10%, A) or by immunoelectrophoresis with antiserum against TorA (B). TMAOR activity was visualized by activity staining using TMAO (B1), and proteins were detected by Coomassie blue staining (A1 and B2).

Catalytic properties of the tungsten-substituted TMAO reductase

Before studying the catalytic properties, we compared the kinetic parameters of the TMAORs present in the periplasm with the corresponding purified enzymes that were prepared from the wild-type strain or from the mod mutant grown with either tungstate or molybdate. The TMAOR present in the periplasm was quantified by rocket immunoelectrophoresis using the purified enzyme as reference. In all cases, the TMAORs present in the periplasmic fractions showed identical kinetic characteristics to the purified enzymes (Fig. 2). Therefore, the following experiments were performed using the TMAORs present in the periplasmic fractions.

Figure 2.

. Kinetic comparison between purified TMAORs (open symbols) and those present in the periplasm (closed symbols). The square, circle and rhombus represent the enzymes from the wild-type strain, Mo-TorA and W-TorA respectively. The curves were fitted to the Michaelis–Menten equation. The concentration of reduced benzyl viologen was about 0.12 mM in all the experiments to avoid inhibition.

The initial rate of TMAO reduction was measured as a function of TMAO concentration at several fixed concentrations of benzyl viologen (BV). The result obtained with the enzyme prepared from the wild-type strain MC4100 is shown in Fig. 3A. The corresponding double reciprocal plots produced parallel lines (Fig. 3B), indicating that the apparent specificity constant of each substrate is independent of the concentration of the other. This behaviour implies a substituted enzyme type of mechanism, which was observed previously for another molybdo-enzyme, the nitrate reductase (Buc et al., 1995). This kind of catalytic behaviour leads to the following rate equation:

Figure 3.

. Dependence of the rate on the two substrate concentrations for the wild-type strain enzyme. A. Plot of rate against the concentration of TMAO at concentrations of reduced benzyl viologen at 0.122 (open square), 0.083 (closed rhombus) and 0.0156 mM (open circle), with curves calculated from eqn 1 and with the parameter values shown in Table 1. B. Double-reciprocal plots of data presented in (A).


where v is the initial rate at total enzyme concentration [E]total, kcat is the catalytic constant, Km,TMAO is the Michaelis constant with respect to TMAO, K0.5,Donor is a non-Michaelis constant at half-saturation concentration of electron donor at saturating TMAO, [Donor] and [TMAO] is the concentration of electron donor and TMAO respectively. Because the reaction does not follow Michaelis–Menten kinetics with respect to benzyl viologen, two molecules of the one electron donor benzyl viologen are required by the stoichiometry of the reaction.

The primary plots of the enzymes prepared from the mod mutant grown with tungstate or molybdate were the same as those of the wild-type strain (data not shown). Therefore, the mechanisms used by the three kinds of TMAORs are identical. The parameter values shown in Table 1 were obtained by fitting the data of the molybdo-TMAOR of the wild-type strain and those of the W-TorA and Mo-TorA from the mod mutant to eqn 1. The Km,TMAO and K0.5, Donor for the Mo-TorA of the wild-type strain were similar to those of the molybdate-restored Mo-TorA of the mod mutant, while the kcat of the former was about twice that of the latter. Most importantly, the Km, TMAO and K0.5, Donor of the W-TorA decreased by about fourfold and twofold with respect to those of the Mo-TorA (Table 1). These results indicate an increase (twofold) in the catalytic efficiency (kcat/Km) of the W-TorA, albeit a decrease of about 50% in its kcat compared with Mo-TorA. It should be noted that the optimal temperature for W-TorA is much higher than that for Mo-TorA (see below). Therefore, the lower level of kcat reported here for W-TorA assayed at the standard temperature (37°C) is partly a consequence of measuring the activity at a temperature too far from the optimal temperature of the enzyme.

Table 1. . Comparison of kinetics parameters of wild-type, molybdenum- (Mo-TorA) and tungsten- (W-TorA) restored TMAORs. a. Catalytic efficiency.Thumbnail image of

The pH dependence of the enzyme activity is shown in Fig. 4. Both W-TorA and Mo-TorA were active over a broad pH range. At pH values above 8.5, the activities were essentially zero. Maximal activity for Mo-TorA was observed at pH values of 5 and 5.5. Interestingly, the optimal pH value for W-TorA was 5, and a significant decrease in its activity was found at 5.5.

Figure 4.

. Dependence of the rate of W-TorA and Mo-TorA on pH. Enzyme rate was measured under the standard assay conditions with various pH ranges as described in Experimental procedures. Mo-TorA and W-TorA are represented by the open squares and closed circles respectively. The data presented are the average of those obtained from at least five independent experiments.

The temperature dependence of the activities of Mo-TorA and W-TorA followed the Arrhenius equation between 20°C and 70°C (Fig. 5B). Activation energies were 319 and 409 J mol−1 for Mo-TorA and W-TorA respectively. Maximal activity of Mo-TorA was reached at 60°C and remained the same at 80°C, which was the highest temperature limit in our enzyme assay system. Interestingly, the activity of W-TorA kept increasing in the temperature range, and the maximal level did not seem to be reached even at 80°C (Fig. 5A). The requirement of higher temperature for reaching the optimal activity of W-TorA is consistent with its higher activation energy compared with Mo-TorA. Interestingly, W-TorA was also more resistant to high temperature than Mo-TorA. The two enzymes were incubated at 80°C for various times, and their activities were measured. Whereas the activity of Mo-TorA decreased drastically (50%) after 4 min of incubation, and only about 3% of the activity was left after 90 min of incubation, the activity of W-TorA showed a slow progressive decrease with increase in the incubation time, and more than 50% activity remained after 90 min of incubation (Fig. 5C).

Figure 5.

. Effect of temperature on W-TorA (open circles) and Mo-TorA (closed squares) activities. A. Dependence of the rate of W-TorA and Mo-TorA on temperature. Enzyme essay was performed under standard conditions at various temperatures. B. Corresponding Arrhenius plots. C. Thermostability of W-TorA and Mo-TorA. The enzymes were incubated at 80°C for various times (min) as indicated before the enzyme assay.

An increase in the ionic strength of the reaction medium showed a similar effect to the temperature inactivation on the activities of Mo-TorA and W-TorA. The activity of Mo-TorA decreased drastically with a slight increase in the ionic strength, and about 50% of its activity was lost in the presence of 0.1 M NaCl (data not shown). Only about 4% of the initial activity was left in the presence of 2 M NaCl. However, the activity of W-TorA decreased more slowly, and it remained higher than the activity of Mo-TorA at a concentration of NaCl above 0.1 M. At 2 M NaCl, about 15% of the initial activity was found. Therefore, W-TorA is more resistant to an increase in ionic strength than Mo-TorA.

Substrate specificity of W- and Mo-TMAO reductase

The DMSORs of Rhodobacter capsulatus and Rhodobacter sphaeroides are molybdoenzymes and homologous to TorA of E. coli. In contrast to TorA, they are capable of catalysing the reduction of both DMSO and TMAO. We therefore investigated whether the tungsten-substituted TorA can also use DMSO and other sulphoxide substrates.

The periplasmic fractions of the wild-type strain, the mod mutant grown with tungstate or molybdate and the torA mutant were resolved on a native gel. Reduction of TMAO and DMSO was visualized by activity staining. As expected, an active band was found in the periplasms of the wild-type strains and of the mod mutant grown with either molybdate or tungstate when TMAO was provided as a substrate (Fig. 6A1). This band was recognized specifically by the antiserum against TorA (Fig. 6B1), and it was absent from the torA mutant. Therefore, the TMAO reduction must be catalysed by the TMAOR encoded by torA. Intriguingly, the tungsten-substituted TorA was also capable of catalysing the reduction of DMSO when the same native gel was analysed using DMSO as a substrate (Fig. 6A2). These results were confirmed by rocket immunoelectrophoresis. Only the immunoprecipitates of the periplasm of the mod mutant grown with tungstate showed the activity of DMSO reduction (Fig. 6B2). Therefore, substitution of molybdenum by tungsten confers on the TMAOR the capacity of catalysing DMSO reduction. It should be mentioned that W-TorA was the only enzyme that catalysed the reduction of DMSO when the periplasms (Fig. 6A2) or crude extracts (not shown) of the mod mutant grown with tungstate were analysed by activity staining and enzyme assays. Tungsten could not substitute for the molybdenum in the molybdocofactor of DMSOR encoded by the dms operon, or the substitution would lead to an inactive DMSOR. This is consistent with the previous observation that the addition of tungstate in the culture inhibits DMSOR activity (Rothery et al., 1995).

Kinetic parameters of W-TorA were determined and compared with those of Mo-TorA. When six N-oxide derivatives were used as substrates, Mo-TorA showed similar catalytic properties (Table 2) to the wild-type TMAOR (Iobbi-Nivol et al., 1996). However, the affinities of W-TorA for these substrates increased three- to 100-fold for five of the six substrates, while the kcat kept the same or slightly decreased compared with Mo-TorA. As a consequence, the catalytic efficiency of W-TorA was similar to that of Mo-TorA for TMAO and 4-methylmorpholine N-oxide, and increased by one or two orders of magnitude for picoline N-oxide and for pyridine N-oxide. Hydroxylamine N-oxide remained the worst substrate for both W-TorA and Mo-TorA.

Table 2. . Comparison of substrate specificity of wild-type, molybdenum-(Mo-TorA) and tungsten-(W-TorA) restored TMAORs. The kinetic parameter values are obtained by fitting the data to the Michaelis–Menten equation. Benzyl viologen reduced concentration is 0.12 mM for all experiments; Mo-TorA and W-TorA concentrations are 3.52 nM and 1.64 nM respectively.a. Data from Iobbi-Nivol et al. (1996).ND, not detectable.Thumbnail image of

As expected, Mo-TorA could not use the three sulphoxide derivatives as substrates. Interestingly, W-TorA was capable of reducing all three sulphoxide derivatives. The affinities of W-TorA for these substrates were about 10−5 M, which was about 10-fold lower or similar to those of W-TorA or Mo-TorA for TMAO respectively. Tetramethylene sulphoxide is the most closely related molecule to TMAO, and it was found to be the best sulphoxide substrate for W-TorA.

In E. coli, the membrane-bound constitutive Mo-DMSO reductase is genetically distinct from the inducible Mo-TMAO reductase. It is noticeable that the kcat with DMSO as substrate of this enzyme (Simala-Grant and Weiner, 1996) is twofold higher than that of the W-TorA.

Physiological significance of tungsten-substituted TMAO reductase

TMAOR and DMSOR enable bacteria to grow anaerobically in a minimal medium using TMAO and DMSO as electron acceptors. We therefore assessed the possibility of W-TorA supporting anaerobic growth of the mod mutant on DMSO. As expected, the mod mutant could grow with neither TMAO nor DMSO as electron acceptors in the absence of molybdate or tungstate. The addition of 1 mM molybdate to the growth medium restored both TMAOR and DMSOR activities and conferred on the mod mutant an anaerobic growth on TMAO or DMSO with generation times of 3 h 30 min and 4 h 30 min, respectively, and a similar maximal cell density of 1.7 × 109 cells ml−1 of culture. In the presence of 1 mM tungstate, the generation time of the mod mutant grown with TMAO or DMSO as electron acceptors was 8 h and 9 h 30 min, and maximal cell density was 0.6 × 109 and 0.4 × 109 cells ml−1 of culture respectively. In addition, the torA mutation abolished the slow but significant tungstate-dependent growth in the mod–torA double mutant on TMAO or DMSO. These results are consistent with our observation above that the substitution by tungstate could only lead to an active TMAOR, which can catalyse the reduction of both TMAO and DMSO, but it was unable to restore the activity of the membrane-bound DMSOR encoded by the dms operon. Taken together, these results strongly suggest that the tungsten-substituted TMAOR can interact successfully with other components of a respiratory chain and thus support anaerobic growth of the bacteria on TMAO or DMSO. As shown above, the amount of W-TorA in the periplasm is about 10% that of Mo-TorA, and the kcat of the former with TMAO as substrate is reduced twofold compared with the latter. These findings might explain the reduction in the growth rate and the maximal yield of the mod mutant grown on TMAO with tungstate compared with molybdate.


Recently, dramatic progress has been made in the physiological, biochemical, enzymatic and structural study of tungstoenzymes. The most intriguing questions, however, remain open: why some microorganisms have chosen to use tungsten rather than molybdenum at the active site of key enzymes; why some can use both elements; and why the majority of life forms on this planet use molybdenum exclusively. Studies with synthesized chemical analogues of the active sites of W- and Mo-containing enzymes have provided insight into the fundamental differences in the chemistry of equivalent Mo and W complexes. The W complexes are very oxygen sensitive, show strongly enhanced thermal stability and possess much lower redox potentials compared with Mo complexes (Johnson et al., 1996). From these differences, it has been predicted that tungsten could only be used by biological systems to catalyse low-potential reactions under anaerobic conditions, and significant catalytic rates would only be observed at high temperatures. Indeed, studies with hyperthermophilic archaea lend credence to the notion that tungsten is used exclusively at the active sites of enzymes that catalyse reactions with low redox potential and high temperatures (Johnson et al., 1996; Mukund and Adams, 1996; Vorholt et al., 1997). Our results presented here fit most of these predictions and are in contradiction with others. For example, W-substituted TorA is more resistant to high temperature and high salt concentration and more sensitive to high pH than Mo-TorA. Likewise, the enhanced thermal stability of the W-active site may account for the observation that W-TorA is kinetically slower than Mo-TorA in oxygen atom transfer reactions, as in the TMAO/TMA conversion. Moreover, W-TorA is more sensitive to exposure to oxygen than Mo-TorA (data not shown). Assuming that primitive forms of life emerged from a very hot, acidic environment free of oxygen, the W-containing isoenzymes thus represent less evolved forms or ancestors of the Mo-containing isoenzymes. The Mo-isoenzymes might be catalytically more efficient than the W-isoenzymes, and they took advantage and replaced the W-isoenzymes by natural selection during evolution.

Investigations with synthesized analogues and with tungstoenzymes, except acetylene hydratase in which the catalytic function of tungsten is unknown, confirm that they catalyse reactions of extremely low potential (≤ −420 mV). However, this notion is challenged now by our observation that W-TorA is capable of catalysing the conversion of TMAO to TMA or DMSO to DMS with redox potentials of + 130 mV and + 160 mV respectively.

Finally, we come to the question of how the substitution of molybdenum in the active site by tungsten can change the substrate specificity of the enzyme. The TMAOR of E. coli and the recently discovered homologous TMAOR of Shewanella massilia can only use TMAO and its N-oxide analogues as substrates (Iobbi-Nivol et al., 1996; Dos Santos et al., 1998). However, the two DMSORs of R. sphaeroides and R. capsulatus, which are homologous to the TMAORs and show similar crystal structures to the TMAOR of S. massilia, are capable of using not only DMSO but also TMAO as substrates. The comparison of the crystal structure of the TMAOR from S. massilia with those of the two DMSORs in oxidized or reduced form and in complex with DMSO has provided important insight with respect to substrate specificity (Schindelin et al., 1996; Schneider et al., 1996; McAlpine et al., 1997; Czjzek et al., 1998; McAlpine et al., 1998). Although the overall fold of the structures of the three enzymes is essentially the same, significant differences were observed between the structure of the DMSORs and that of the TMAOR. The crystal structure reveals that DMSO is bound in the short channel connecting the base of the active site cleft to the molybdenum ion and has replaced a water molecule that is present in the oxidized enzyme of R. capsulatus (McAlpine et al., 1998). The short channel is lined by the side-chains of aromatic residues. This short channel and the aromatic and hydrophobic residues lining it are perfectly superimposable on the structure of the oxidized TMAOR entrance (Czjzek et al., 1998). However, when moving out of the active site cleft towards the wide entrance, substantial differences in the number of charged residues and the nature of the charges have been encountered. In addition, Tyr-114 of the DMSORs, conserved in all of them, is absent from the active site of the TMAOR. In the crystal structures of the DMSORs, this Tyr-114 is hydrogen bonded via its OH group to one of the oxo-groups, which in turn is a ligand of the molybdenum ion. Schindelin et al. (1996) have therefore proposed that Tyr-114 could play an important role as proton donor in the ultimate transfer of an oxygen atom from the DMSO molecule into water.

As this tyrosine is not present and not replaced by any equivalent residue in the active site of the TMAOR of S. massilia (Czjzek et al., 1998), the oxygen transfer might be more difficult and render the utilization of DMSO as a substrate by this enzyme unfavourable. The absence of the partially conserved Tyr-114, together with the observations of other substantial differences, such as the change in the charges in the tunnel-like entrance of the TMAOR, have been pointed out as possible factors to account for the differences in substrate specificity, when comparing DMSORs and TMAORs (Czjzek et al., 1998).

As W-TorA and Mo-TorA show the same pI (not shown), electrophoretic mobility on a native gel and identical behaviour during the purification, the substitution of molybdenum by tungsten does not seem likely to lead to a significant structural change in TorA. The difference in substrate specificity provoked by the molybdenum–tungsten substitution may thus be based on the variation of two factors: small structural adjustments in the close environment of the active metal centre and, therefore, the substrate binding site and the nature of the metal centre itself. The co-ordination of molybdenum or tungsten by pterins to constitute the active centre must lead to chemically different compositions with a strongly decreased redox potential in the case of the tungstoenzyme (Johnson et al., 1996). This second factor could play an essential role in the catalytic mechanism.

The S2′ group of the Q pterin has been proposed as another potential candidate for participation, via Ser147, in proton transfer (Schindelin et al., 1996). Although the pterins may not participate directly in the actual electron transfer, their oxidation state-dependent interactions with the molybdenum or the tungsten could influence the electronic properties of the metal, thereby providing some of the necessary requirements for catalysis. The low redox potential of a tungsten site implies a possibly different co-ordination sphere of the metal centre in W-TorA compared with Mo-TorA. It is worth mentioning that, unlike molybdoenzymes, tungsten in aldehyde oxidoreductase is not co-ordinated by an amino acid of the protein (Chan et al., 1995). Our results showed that the restoration of TorA activity of the mod mutant by tungstate is not as efficient as that by molybdate (see above; Santini et al., 1998), suggesting a possible difficulty in the construction of the tungsten site by the molybdocofactor biosynthesis and incorporation machinery. In fact, TMAOR is the only molybdoenzyme in which molybdenum can be functionally substituted by tungsten in E. coli. In addition, we observed that DMSO did not have a competitive effect on the binding of TMAO to Mo-TorA and that the tungsten substitution resulted in a decrease in Km of W-TorA for TMAO compared with Mo-TorA. Therefore, the tungsten substitution does indeed lead to a change in the substrate binding site.

Taken together, the Tyr-114 may be a determining factor for the substrate specificity of TMAORs and DMSORs. The low potential and altered co-ordination situation in the case of W-TorA could circumvent the dependence on the tyrosine residue and allow W-TorA to use DMSO as a substrate. This speculation should be checked by structural analysis of the tungsten substitute TMAO reductase.

Experimental procedures

Strains, plasmids and growth conditions

The strains mod RK5209 (Stewart and MacGregor, 1982), LCB620 (torA::Mud II(lacZ, KanR)) (Méjean et al., 1994) and the double mod–torA mutant (Santini et al., 1998) are derivatives of MC4100 (F′lacΔU169 araD139 rpsL150 thi flbB5301 deoC7 ptsF25 relA1 ).

Bacteria were grown routinely in Luria–Bertani (LB) medium. Anaerobic growth was achieved in LB medium supplemented with glucose (0.2%) and TMAO (0.1%) in stoppered bottles or tubes filled to the top. As required, sodium molybdate (1 mM), or sodium tungstate (1 mM) was added. Cultures were incubated at 30°C. Precultures were inoculated from a single colony and used as a 100-fold dilution.

Preparation of subcellular fractions

Periplasm and spheroplasts were prepared by the lysozyme/EDTA/cold osmoshock method as described previously (Santini et al., 1998). Spheroplasts were washed once and disrupted by two passages through a French press. Cell debris was discarded by centrifugation at 18 000 × g for 15 min. The supernatant was further centrifuged twice at 120 000 × g for 90 min. The supernatant was saved as the cytoplasmic fraction. The pellet obtained from the first ultracentrifugation was washed in 40 mM Tris-HCl (pH 7.6) containing 1 mM benzamidine-HCl and 15% sucrose. The resulting pellet was washed once and saved as the membrane fraction.

Enzyme assays and kinetic studies

TMAOR and DMSOR activities were measured at 37°C by following spectrophotometrically the oxidation of reduced benzyl viologen at 600 nm coupled to the reduction of TMAO or DMSO as described previously (Iobbi-Nivol et al., 1996). One unit of specific activity was defined as 1 μmol TMAO or DMSO reduced mg−1 protein min−1. Alternatively, TMAOR and DMSOR activities were visualized on native polyacrylamide gels or rocket plates by a staining method that is based on a methyl viologen-linked TMAO or DMSO reduction (Silvestro et al., 1989). β-Galactosidase activity was assayed in cells treated with SDS and chloroform by following the hydrolysis of ONPG at 420 nm according to Miller (1972). Proteins were estimated according to the Lowry technique (Lowry et al., 1951). Steady-state kinetic studies were performed with a Hitachi U-2000 spectrophotometer connected to a PC-compatible computer as described previously (Buc et al., 1995), except that TMAO or DMSO was used as substrate. To study pH effect on TMAOR activities, three buffers (sodium succinate pH 4.0–6.0, sodium phosphate pH 6.0–8.0 and Tris-HCl pH 7.5–9.0) were used. Overlap of the pH ranges allowed elimination of the buffer effect.

Electrophoretic and immunological procedures

Proteins were separated by polyacrylamide gel electrophoresis on 10% acrylamide gels without (native PAGE) or with SDS (SDS–PAGE) (Laemmli, 1970). The proteins present on the native gel were immobilized onto a polyvinylidene difluoride (PVDF) membrane, and TMAOR was analysed by immunoblot, using the ECL method according to the manufacturer's instructions (Amersham). Rocket immunoelectrophoresis with antiserum against TorA was performed as described previously (Silvestro et al., 1989). Isoelectrofocusing electrophoresis was performed on a pH 3–9 IEF gel according to the manufacturer's instructions (Pharmacia).

Purification of TMAO-reductase

E. coli strain RK5209 (mod ) was grown anaerobically in LB medium supplemented with glucose (0.2%), TMAO (0.1%) and sodium molybdate (1 mM) or sodium tungstate (1 mM). The periplasmic fraction from 38 g (wet mass) of cells was applied on a DEAE-cellulose (DE52) ion exchange column (120 ml) that had been pre-equilibrated with 0.2 M Tris-HCl (pH 8) buffer. The column was washed with three column volumes of the same buffer. TMAOR was eluted after the passage of 0.1 M NaCl in the same buffer. The active fractions were pooled and diluted fivefold with water. The sample was applied to a DEAE-cellulose column (40 ml) equilibrated with 40 mM Tris-HCl (pH 7.6) plus 0.05 M NaCl. The column was then washed with three column volumes of the same buffer. Bound protein was eluted with a linear gradient of NaCl from 0.05 M to 0.45 M in the same buffer. The active fractions were combined and dialysed for 4 h against a large volume of 0.04 M Tris-HCl (pH 7.6) buffer and applied to a MonoQ column (HR5/5) equilibrated with the same buffer. The column was then washed with 10 column volumes of the same buffer, and elution was performed by a linear gradient of NaCl from 0 to 0.45 M in the same buffer. The TMAOR was collected as a single active peak at a NaCl concentration of 0.35 M.

Determination of tungsten and molybdenum

The purified TMAOR or periplasmic fractions were dialysed twice for 1 h against a large volume of water. Distilled nitric acid (12 N) was added to the preparation with the aim of releasing the elements for analysis after organic substance destruction. Tungsten measurements were performed by argon plasma emission spectroscopy using a Jobin-Yvon J38. Molybdenum measurements were effected by electronic atomic absorption spectroscopy with a Zeeman effect apparatus Hitachi Z8200. The apparatus was calibrated with tungsten and molybdenum standards prepared in the same way as the samples.

Analysis of data

The Newton–Gauss method was used to fit experimental data to the appropriate equation by non-linear regression according to Cleland (1967).


  1. †The first two authors contributed equally to this work


We are grateful to J. Pommier and M. Cigna for advice regarding TorA purification, to V. Méjean and C. Iobbi-Nivol for valuable discussions, to B. Guigliarelli for reading the manuscript, and to A. Chanal for help with the artwork. We thank J. P. Ambrosi for assistance with ICPES. This research was supported by a grant from the Centre National de la Recherche Scientifique to UPR9043.