S - and N-Oxide Reductases

VICTOR W. T. CHENG AND JOEL H. WEINER*

[SECTION EDITOR: VALLEY STEWART]

Posted August 31, 2007

Department of Biochemistry and Membrane Protein Research Group, University of Alberta, Edmonton, Alberta T6G 2H7, Canada

*Corresponding author. Mailing address: Department of Biochemistry and Membrane Protein Research Group, University of Alberta, Edmonton, Alberta T6G 2H7, Canada. Phone: (780) 492-2761, Fax: (780) 492-0886, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

Escherichia coliand Salmonella species are capable of respiratory growth under a diverse set of environmental circumstances. Inner membrane dehydrogenases transfer the reducing energy stored as chemical potential from low redox potential donors (carbon compounds such as succinate (+0.03 V), glycerol 3-phosphate (−0.19 V), lactate (−0.19 V), formate (−0.43 V), and hydrogen (−0.42 V)) to the membrane-soluble ubiquinone or menaquinone quinone pool (Q-pool) (145). Terminal reductases re-oxidize the quinone-pool coupling electron flow to high potential electron acceptors such as O₂ (+0.82 V), nitrate (+0.42 V), fumarate (+0.03 V), dimethyl sulfoxide (DMSO) (+0.16 V), and trimethylamine N-oxide (TMAO) (+0.13 V). This process results in the generation of a transmembrane proton electrochemical gradient or proton motive force (PMF) (90). The energy stored in this electrochemical gradient works directly to power important cellular processes such as motility and active membrane transport, or it is used to synthesize ATP (using F₀F₁ ATP synthase) which can be used to energize a wide variety of important cellular processes. The PMF can be generated either vectorially by direct pumping of protons across the E. coli cytoplasmic membrane or through a redox-loop mechanism.

ORGANIZATION OF THE DMSO AND TMAO RESPIRATORY CHAINS

The DMSO and TMAO respiratory chains consist of a primary dehydrogenase and a terminal reductase that are functionally linked through menaquinone-8 (MQ) and demethylmenaquinone-8 (DMQ) (each quinone has eight isoprenoid side chains) (144). At least 15 primary dehydrogenases have been identified in E. coli (145), and while most of these enzymes will support anaerobic growth with DMSO or TMAO as the terminal reductant, growth is normally observed with glycerol 3-phosphate or l-lactate as reductant. Formate and hydrogen can also be used if appropriate carbon sources (i.e., amino acids) are available (34). The respiratory chains terminating with either reductase, but not the terminal reductases themselves, generate the essential PMF across the cytoplasmic membrane (17). Terminal reductases that lack PMF capacity in general have their proton-consuming active site facing the periplasm; this is the case for TMAO reductase and possibly for DMSO reductase (Fig. 1). Respiratory chain enzymes that do not generate a PMF directly are "electron sinks" whose sole function is to reoxidize the Q-pool such that upstream dehydrogenases can continue to function and generate an electrochemical gradient (17).

Fig. 1A schematic representation of the DMSO and TMAO reductases. MQ and MQH₂ are the oxidized and reduced forms of menaquinol-8. FS1 to FS4 are the four [4Fe-4S] clusters in DmsB. The topology of DmsAB is shown directed to the periplasm, but see the text for further discussion.

While the dehydrogenase and menaquinone are common to both electron transfer chains, the architectural organization of the terminal reductases differs (Fig. 1). TMAO reductase is a soluble periplasmic molybdoprotein (TorA) that receives electrons via a transient interaction with a pentaheme c-type cytochrome (TorC) tethered to the membrane. TorC receives electrons directly from the MQ/DMQ pool (144). DMSO reductase is a trimeric, complex iron-sulfur molybdoenzyme in which a membrane anchor subunit (DmsC) communicates with the MQ/DMQ pool and funnels electrons through a polyferredoxin electron transfer subunit (DmsB) to the catalytic site in DmsA. Both TMAO reductase and DMSO reductase have a molybdo-bis(molybdopterin guanine dinucleotide) (Mo-bisMGD) molecule at their active centers.

DMSO REDUCTASE

DmsABC

The ability to reduce DMSO to dimethyl sulfide (DMS) (Fig. 2), a major intermediate in the global sulfur cycle, is widespread in both prokaryotes and eukaryotes (7, 157). It is believed that DMS generated by degradation of dimethylsulfoniopropionate (DMSP) or by reduction of DMSO accounts for 50% of the biologically synthesized reduced sulfur in the environment and plays an important role in global climate control (71, 79).

Fig. 2The reaction catalyzed by DMSO reductase.

In the presence of hydrogen, formate, or glycerol 3-phosphate as electron donor and DMSO as electron acceptor, and the absence of oxygen and nitrate, the enzyme DMSO reductase (DmsABC) is expressed to allow E. coli to respire anaerobically (Fig. 1). The E. coli DMSO reductase is a heterotrimeric protein comprising a molybdopterin-containing catalytic subunit (DmsA), a ferredoxin-like electron transfer subunit (DmsB), and a membrane anchor subunit (DmsC) that interacts with the MQH₂/MQ pool. Its enzymatic function involves the oxidation of MQH₂ to MQ, the transfer of two electrons from the membrane domain to the molybdenum active site, and the reduction of DMSO to DMS.

The DMSO Reductase Operon

Three open reading frames are in the DMSO reductase operon located at 940 kbp (20 minutes) on the E. coli chromosome (Fig. 3). They encode polypeptides with M_r values of 87,350/85,800, 23,070, and 30,789 (DmsA preprotein/mature, B, and C, respectively). Organization of the operon closely parallels that seen for fumarate reductase and formate dehydrogenase of E. coli. The 5'-proximal gene encodes the molybdenum cofactor-containing catalytic subunit followed by an electron transfer subunit that contains groups of cysteine residues for assembly of [4Fe-4S] clusters. The 3'-distal gene encodes the membrane anchor and quinol binding subunit DmsC.

Fig. 3Organization of the chromosomal regions encoding dmsABC, ynfEFGHdmsD, torCAD, torts, and torYZ operons.

DmsA—the Catalytic Subunit

The molybdenum cofactor.

The Mo-bisMGD cofactor in both TorA and DmsA (Fig. 4) is a tricyclic ring system comprising a pterin group with a fused pyran ring that is attached via a phosphodiester linkage to a guanine nucleotide (24, 99, 121, 122). The pyran ring of the pterin has two sulfur atoms attached in a dithiolene linkage. This cis-dithiolene provides bidentate coordination to the essential Mo atom at the active site of TMAO or DMSO reduction. The dithiolene sulfurs of the two pterins provide a total of four sulfur ligands to the active site Mo. Further coordination of the Mo atom is provided by the hydroxyl side chain of an amino acid and by an additional oxo group. Stable derivatives of the molybdopterin can be isolated, quantified, and characterized by acid hydrolysis of the protein preparation followed by controlled oxidation of the cofactor (65). The transition metals Mo and W are found associated with molybdopterin. W catalyzes reactions of very low potential (< −420 mV) (73), whereas Mo catalyzes reactions of high redox potential (in general, >0 mV). Molybdenum-containing enzymes have been divided into three subfamilies based on the structure of the molybdenum center at the active site (52): the xanthine oxidase family, the sulfite oxidase family, and the DMSO reductase family.

Fig. 4Opposite views of the Mo-bisMGD cofactor. The molybdenum is coordinated by two dithiolene sulfurs from each of the two molybdopterin guanine dinucleotides (MGDs).

Cofactor composition of DmsA.

DmsA is the 87.4-kDa catalytic subunit that binds the Mo-bisMGD that is essential for DMSO reduction (107) (Fig. 4). Exclusive supplementation with W (tungsten) in lieu of Mo during growth in minimal media renders the enzyme inactive and abolishes insertion of the metal-cofactor complex into DmsA despite proper targeting and assembly of DmsABC. However, overexpression of the enzyme or addition of both W and Mo to the minimal medium can rescue anaerobic growth on DMSO since small amounts of Mo can be salvaged and incorporated into approximately 10% of the DmsABC assembled (107). It is very likely that a high-spin [4Fe-4S] cluster is coordinated by the conserved Cys residues at positions 18, 22, 26, and 59 of mature DmsA (see below). (Note: Residue numbers used herein often differ from the numbers in earlier publications because the existence of the tat leader was not known prior to 2000.) This cluster would communicate between the [4Fe-4S] clusters in DmsB and the Mo-bisMGD and would parallel the recent discoveries of a spectroscopically elusive [4Fe-4S] cluster in nitrate reductase and formate dehydrogenase (14, 67, 105) (see below).

Crystal structures.

Although a three-dimensional structure of E. coli DmsA (EcDmsA) has not yet been determined, it has significant sequence similarity to the soluble, periplasmic DmsA of Rhodobacter sphaeroides and R. capsulatus (RhDmsA), which have been crystallized and their three-dimensional structures elucidated (83, 84, 121, 122). The RhDmsA protein possesses four distinct domains (Fig. 5A and 5B). Like EcDmsA, RhDmsA contains a Mo-bisMGD cofactor at the active site (see below). Domain IV forms a complex H-bonding network with the two pterin groups, while domains II and III interact with the guanine nucleotides of the P-MGD (proximal to the active site) and Q-MGD (distal to the active site) molecules, respectively (84, 121, 122) (Fig. 5A and 5B). The substrate cavity/funnel leading to the active Mo atom is encircled by domains I, II and III to form a channel that is approximately 10 Å wide and 8 Å deep (Fig. 6A and 6B). Additionally, a conserved loop of polypeptide serves as a lid during the reaction cycle to trap the substrate DMSO in the binding pocket. Domain IV sits at the "bottom" of the RhDmsA subunit and is where the Fe-S cluster in EcDmsA is predicted to be located and also the contact point with DmsB.

Fig. 5ASpace-filling structures (PDB ID 3DMR) of DMSO reductase from R. capsulatus. This panel shows a "side" view, whereas Fig. 5B shows a "top" view looking into the active-site funnel. Domain I, green; Domain II, purple; Domain III, orange; Domain IV, black. The Mo atom is yellow, and the MGD portions are shown as sticks. Images were generated using PyMol.

Fig. 5BSpace-filling structures (PDB ID 3DMR) of DMSO reductase from R. capsulatus. This panel B shows a "top" view, whereas Fig. 5A shows a "side" view looking into the active-site funnel. Domain I, green; Domain II, purple; Domain III, orange; Domain IV, black. The Mo atom is yellow, and the MGD portions are shown as sticks. Images were generated using PyMol.

Fig. 6AView of the active-site funnel in R. capsulatus DMSO reductase (PDB ID 3DMR). Shown in purple are the oxo ligands to the Mo atom. The sliced view shows the active-site funnel leading to the Mo-bisMGD cofactor.

Fig. 6BView of the active-site funnel in R. capsulatus DMSO reductase (PDB ID 3DMR). Shown in purple are the oxo ligands to the Mo atom. The view is directed straight into the active funnel to show the accessibility of the Mo-bisMGD.

The Mo-bisMGD cofactor of RhDmsA.

DmsA belongs in the DMSO reductase superfamily of molybdoenzymes that can be distinguished by the presence of two pyranopterin molecules per Mo atom and the covalent attachment of a GMP molecule to each pyranopterin via a phosphate bond (51, 52, 64). The coordination of the Mo atom in RhDmsA was somewhat controversial as the available crystal structures showed between five and seven ligands interacting with the Mo (83, 84, 121, 122). Early extended X-ray absorption fine structure (EXAFS) studies on the R. sphaeroides (36, 37) and the R. capsulatus (8) DMSO reductases also indicated differences between hexa- and heptacoordination, respectively. Currently, there is general consensus that the Mo is hexacoordinated with four ligands being provided by the dithiolenes of the bisMGD moieties, one provided by the hydroxymethyl side chain of Ser147 and the last one provided by DMSO (77) (Fig. 7A). This organization is also observed in the TMAO reductase enzyme (24) and these enzymes have been classified as the type III enzymes of the DMSO superfamily. This contrasts with other molybdoenzymes with Cys/SeCys or Asp side chains as a Mo ligand in the type I and type II clades, respectively (52). In a recent EXAFS study by George et al. (35) (Fig. 7B) of the active site of oxidized EcDmsA it was found that mono-oxo Mo^VI species was coordinated by two molybdopterin dithiolenes and a serine with four Mo–S ligands at 2.43 Å, one Mo=O at 1.71 Å, and a longer Mo–O at 1.90 Å. The active site of EcDmsA is very similar to that of the well-characterized RhDmsA, suggesting similar active-site structures for the two enzymes.

Fig. 7AChemical structure of the Mo-bisMGD cofactor from R. capsulatus DMSO reductase in the oxidized form. The Mo atom is coordinated by four dithiolene sulfurs from two molybdopterin molecules, one oxo ligand, and the oxygen atom from the side chain of Ser147. The bond lengths are taken from reference 84.

Fig. 7BEXAFS analysis of the active site Mo-bisMGD in E. coli DmsA (35).

The apparent multiple coordination states of Mo coordination is believed to stem from movement of the Ser147 side chain and the Mo atom as it cycles between the Mo(IV), Mo(V), and Mo(VI) states during catalysis (77). With reduction of DMSO and formation of the Mo-oxo intermediate, a Trp residue is responsible for H-bonding and stabilization of the oxo group (83). Residue Tyr114 has also been observed to interact with one of the oxo groups in the dioxo-pentacoordinated structure in which the dithiolenes of the Q-pterin do not serve as ligands (122). It has been proposed that the heptacoordinated structure is the result of the superimposition of the penta- and hexacoordinated states (84). The hexacoordinated mono-oxo form is believed to be correct as it is supported by EXAFS experiments (36, 37) as well as Raman spectroscopy (10, 29). The pentacoordinate state is thought to be a "damaged" state of the enzyme (19) whereby the dithiolene ligands of the Q-pyranopterin are lost.

Catalytic mechanism.

From studies on the R. capsulatus and R. sphaeroides DMSO reductases it appears that the active site undergoes complex structural transformations. The molybdenum site exists in multiple structural states, and rearrangement of atoms/groups at the active site is facilitated in the Mo(V) state. Johnson et al. (66) have suggested a catalytic cycle for RhDmsA, which likely is similar for the EcDmsABC (Fig. 8).

Fig. 8Suggested catalytic mechanism for the catalytic cycle of Rhodobacter DMSO reductase, as proposed by Johnson and Rajagopalan (66).

Source: Figure 8 is reprinted from Johnson and Rajagopalan (J. Biol. Chem.268:13178–13185, 2001), with permission.

Mutagenesis studies on Tyr114 and Trp116 have provided insights into the catalytic mechanism of DMSO reduction and oxo group transfer to the Mo atom. First, note that the equivalent residue to Tyr114 is missing in TMAO reductase (24). Second, a Tyr114→Phe mutation in the R. capsulatus enzyme causes the K _m to increase and the k_cat/K_m ratio to decrease for both DMSO and TMAO substrates (101). Third, insertion of an equivalent Tyr into the active site of TorA leads to an increased preference for utilizing DMSO over TMAO (66). Together, these experiments indicate the crucial role of Tyr114 to be breakage of the S–O bond during catalysis. Residue Trp116 has also been shown to play a crucial role. Site-directed mutagenesis of Trp116→Phe in the R. capsulatus enzyme results in alterations of its UV-Vis absorption spectra that become reminiscent of the protein in the "damaged" pentacoordinated state (19, 102). Furthermore, k_cat values were decreased with respect to both DMSO and TMAO, whereas K_m values were only moderately affected, suggesting that W116 plays more of a role in the catalytic cycle rather than substrate binding (102). In addition, the Trp116→Phe mutant also renders the Mo^V/IV transition to be pH dependent compared with a pH-independent oxidation exhibited by the wild-type enzyme. The data presented by Ridge et al. suggest that Trp116 plays a stabilization role in which it H-bonds to the Mo-oxo group to maintain the active site in a hexacoordinate state and prevent conversion to the inactive pentacoordinate Mo form (102). These residues are not conserved in EcDmsA, suggesting that other nearby residues fulfill the roles. Further research will be needed to elucidate the mechanism of EcDmsA.

Broad substrate specificity, S- and N-oxides.

In addition to its ability to reduce DMSO to DMS, the E. coli DMSO reductase can also reduce a variety of other S- and N-oxides (125). The promiscuity of the oxidant is believed to stem from the wide active-site funnel leading from the aqueous milieu into the active site (126). Although the structure of E. coli DmsA is unavailable, results from mutagenesis studies on conserved residues in DmsA inspired by the RhDmsA X-ray crystallographic structure agree with this hypothesis. In an extensive study Simala-Grant and Weiner (125) show that the major determinant of substrate reducibility lies in its compatibility with the substrate pocket as K_m values showed 20-fold-larger variance than k_cat values over a range of 22 substrates; K_m values for sulfoxides and pyridine N-oxide were lowest, followed by aliphatic N-oxides. Mutational analyses suggest that Thr132, Gly151, Ala162, Gln163, and Arg201 are part of the active-site funnel and are required for substrate binding and/or enzyme catalysis. In particular, large changes in the K_m value were observed in the Thr132Ser, Ala162Gln, and Arg201Gln mutants when DMSO was used as an electron acceptor. The Thr132Ser mutant did not exhibit a similar change in K_m value when pyridine N-oxide was used as the substrate. Finally, note that the Rhodobacter DMSO reductases selectively reduce the S-enantiomer of the sulfoxides, whereas the E. coli DMSO reductase is selective for the R-enantiomer (1, 48).

Catalytic optimum.

The midpoint potentials for the Mo^VI/V and Mo^V/IV transitions have been determined by potentiometric titration and electron paramagnetic resonance (EPR) spectroscopy to be −15 mV and −175 mV, respectively (141). Protein film voltammetry studies on DmsABC have shown that the enzyme activity is maximal within a narrow window of electrode potential that coincides with the appearance of the Mo(V) species (50). In fact, two switches at critical potentials were seen to control two one-electron processes separating three distinct states of the enzyme. Based on the elaborate series of protein film voltammetry studies, it was proposed that the most influential step during catalysis is the reduction of Mo(V) to Mo(IV)-H⁺ (50). This characteristic of DmsABC is called "tunnel-diode" behavior.

FS0—the predicted [4Fe-4S] cluster in DmsA.

Available X-ray crystal structures of E. coli anaerobic complex molybdoenzymes—namely, formate dehydrogenase-H, formate dehydrogenase-N, and nitrate reductase—show the assembly of a [4Fe-4S] cluster in the catalytic subunit (14, 18, 67). This cluster has not been observed by biophysical techniques such as EPR spectroscopy, possibly because of peptide flexibility and heterogeneity-induced spectral broadening. Mutational studies have shown that the cysteine group in DmsA can act as ligand to a [3Fe-4S] cluster in a Cys22Ser mutant (109, 142). This particular mutation may act in a similar manner to the DmsB-Cys102Ser mutant (see below), where the −50 mV [4Fe-4S] cluster is converted into a [3Fe-4S] cluster and electron transfer through this cluster is compromised (110, 111). In fact, mutation of Cys22 to Ser or Ala abolishes growth on DMSO and cuts the enzyme activity by 50% (142). EPR studies show alterations to the interaction between the reduced −120 mV cluster of DmsB and the Mo(V) atom in the Cys22Ser mutant suggesting that the engineered [3Fe-4S] cluster is in the path for electron tunneling from DmsB to DmsA (109). It is also plausible that the −120 mV signal corresponds to the FS0 species and is only observed through its interaction with the Mo. In an Arg61Ser DmsA mutant, this interaction is abolished, suggesting that this residue is directly on the path of electron flow and that its mutation prevents reduction of the Mo atom (109).

DmsB—the Electron Transfer Subunit

Cofactor composition of DmsB.

DmsB is an electron relay subunit bridging the Mo-bisMGD active site and the Q-site and is highly homologous to FdnH and NarH, the respective electron transfer subunits of formate dehydrogenase-N and nitrate reductase. The electron transfer subunits of DMSO reductase and formate dehydrogenase each contain four [4Fe-4S] clusters (21, 67) differing from the composition of NarH, which contains three [4Fe-4S] clusters and one [3Fe-4S] cluster (15, 45). The DmsB primary sequence contains 16 cysteine residues that serve as ligands to the four [4Fe-4S] clusters. The consensus sequences for assembly of Fe-S clusters in DmsB resemble those in typical bacterial ferredoxins and appear four times as a C-X₂-C-X_n-C-X_n-C motif, where the first three cysteines serve as ligands to one [4Fe-4S] cluster and the terminal cysteine interacts with a different [4Fe-4S] cluster (Fig. 9).

Fig. 9Cysteine coordination of the four [4Fe-4S] clusters in DmsB by four cysteine groups. The midpoint potential (mV) of each cluster is indicated.

DmsB model structure.

A computer-generated model of DmsB (22) based on the FdnH X-ray crystallography structure (67) is available (Fig. 10). The in silico DmsB structure is highly superimposable with the FdnH crystal structure with a root-mean-square deviation of only 0.64 Å through 151 common α-carbons. The four Fe-S clusters of DmsB are arranged in an almost linear fashion that facilitates rapid electron transfer between the Mo active site and the Q-site and are called FS1 (closest to catalytic subunit), FS2, FS3, and FS4 (closest to the Q-site).

Fig. 10A model for DmsB based on the structure of FdnH (PDB ID 1KQF). The four [4Fe-4S] clusters FS1 to FS4 are shown from top to bottom (22).

Electron transfer through DmsB.

Figure 11shows the E_m profile of electron transfer through DmsABC. The four [4Fe-4S] clusters in DmsB have midpoint potentials (E_m values) of −50 mV (FS4), −120 mV (FS3), −330 mV (FS2), and −240 mV (FS1) (21, 111). The relatively negative potentials of FS1 and FS2 compared with the E_m for MQH₂/MQ and DMSO/DMS (−100 mV and +160 mV, respectively) led to a debate about whether these clusters participated in the electron transfer relay or served a structural role. Elucidation of the X-ray crystallographic structures of FrdB (60), NarH (14), SdhB (155), and most importantly FdnH (67) clearly show that all Fe-S clusters are involved in the electron relay. As the in silico model of DmsB also has the Fe-S clusters positioned sequentially, one can infer that all Fe-S clusters in DmsB are also involved in electron transfer. The very negative clusters may provide a barrier for electron flow, slowing the rate of electron transfer and providing better coupling between quinol oxidation and DMSO reduction.

Fig. 11Redox profile of DmsABC (circles) showing the midpoint potential of each redox-active center. The altered profile due to mutation of DmsB Cys102Ser is shown by the blue solid (normal electron transfer) and dotted (electron transfer abolished) lines.

Mutagenesis studies.

Comparison of the cysteine motifs in DmsB and NarH shows that DmsB-Cys102-coordinating FS4 is a Trp residue in nitrate reductase, and the Fe-S cluster is assembled as a [3Fe-4S] cluster in NarH as opposed to the [4Fe-4S] cluster in DmsB. Replacement of Cys102 with Ser, Tyr, Trp, or Phe by site-directed mutagenesis allows conversion of FS4 from a [4Fe-4S] cluster to a [3Fe-4S] cluster with a concomitant increase in its midpoint potential from −50 mV to +275 mV (110). EPR studies on FS4 of the Cys102 mutants show that this cluster is magnetically and conformationally linked to the MQH₂ binding site (111). Traditional enzymatic assays and growth experiments demonstrate that the DmsB-Cys102Ser mutation renders the enzyme inactive as a DMSO reductase, but Q-pool coupling assays show that the mutant enzyme remains functional in the reverse direction to donate electrons to the MQ (111). It has been proposed that, with FS4 at +275 mV, the electron does not have enough energy to bypass the endergonic barrier of FS2 that is poised at −330 mV; incidentally, a similar barrier can be surmounted by electrons traveling through nitrate reductase, where two hemes are available to "push" the electrons past the energy barrier (22).

An extensive mutagenesis study on residues around FS4 by Cheng et al. yielded some interesting results (22). In the structure of formate dehydrogenase, a Tyr residue from the electron transfer subunit is positioned into the core of FdnI and interacts with the proximal heme_b (67). The equivalent Tyr residue in DmsB (Tyr104) was found to be critical for MQH₂ binding and oxidation. Mutation of Tyr to Glu or Asp also lowered the midpoint potential of FS4 in its [3Fe-4S] form, but both mutations were unable to render the enzyme back into a functional state. Two other residues, Pro80 and Ser81, are also predicted to be located at the DmsB/DmsC interface and were also demonstrated to be vital for MQH₂ binding and oxidation. Basic residues around FS4 appear to be responsible for fine-tuning the midpoint potential of FS4 but acidic residues play no apparent role. Redox titrations of Tyr mutants in the absence and presence of HOQNO in addition to pH studies also suggests that the maximum achievable midpoint potential for FS4 is approximately +275 mV (22).

DmsC—The Membrane Anchor Subunit

Topology.

DmsC is the 30.8-kDa membrane anchor subunit that tethers DmsAB to the cytoplasmic membrane. It is known that only the full-length DmsC can interact with DmsAB because truncated DmsC leads to accumulation of the soluble catalytic dimer in the cytoplasm (151). Hydropathy plots, such as those using the Kyte–Doolittle algorithm (76), of DmsC predict a subunit with eight distinct transmembrane α-helices. Application of the "positive-inside" rule by von Heijne (146) suggests that both the N- and C-terminal ends of DmsC are located on the periplasmic side of the membrane with three Arg+Lys residues on the periplasmic side and eight on the cytoplasmic side. Fusion experiments using alkaline phosphatase and β-lactamase (151) have confirmed this proposed arrangement of DmsC, which is shown in Fig. 12. The subunit has three relatively large periplasmic loops, P1 to P3, and four small cytoplasmic loops C1 to C4. Comparative analyses and site-directed mutagenesis studies of conserved residues also reaffirm the proposed topological arrangement of DmsC (30).

Fig. 12A topology model based on phoA and bla fusions of DmsC. Acidic residues are in red, and basic residues are in blue. His65 and Glu87 known to be required for quinol oxidation are enlarged.

Source: Figure 12 was adapted from Fig. 6 in Weiner et al. (J. Biol. Chem.268:3238–3244, 1993), with permission.

Quinol binding and oxidation.

The DmsC subunit plays an integral role in the binding of MQH₂ and its oxidation to MQ. Steady-state and rapid-reaction fluorescence quench titrations of DmsABC with the MQH₂ analogue 2-n-heptyl-4-hydroxyquinoline-N-oxide (HOQNO) indicate the presence of a single high-affinity quinol binding site with a dissociation constant of 5 nM (156). This differs from other anaerobic terminal oxidoreductases in E. coli, such as fumarate reductase (44, 81, 152, 153) and nitrate reductase (20, 148), which have been proposed to contain more than one Q-site and can receive reducing power from both ubiquinol and menaquinol. Although X-ray structural information of fumarate reductase FrdCD indicates the presence of both proximal and distal menaquinol binding sites, only the proximal menaquinol binding site is within electron-transferring distance to the [4Fe-4S] clusters (61); the role of the distal site remains to be determined. The crystal structure of nitrate reductase, NarI, indicates that there are two clefts where quinol can bind. However, binding studies with the HOQNO and crystallographic studies with a benzoquinol analogue, pentachlorophenol, suggest there is only one quinol binding site (13). The lack of an X-ray crystallographic structure for DmsABC means that the presence of a nondissociable Q-site that is undetectable by normal enzymology means cannot be discounted.

Other menaquinol-like molecules are oxidized by DmsABC. Two hydroxylated naphthoquinol analogues, reduced plumbagin (PBH₂; 5-hydroxy-2-methyl-1,4-naphthoquinol) and reduced lapachol (LPCH₂; 2-hydroxy-2-methyl-1,4-naphthoquinol), have been shown to be substrates. These compounds have optical, solubility, and redox properties that make them suitable for use in studies of the enzymology of menaquinol oxidation ( 106).

The lack of cytochromes or any other redox-active cofactor in the membrane anchor domain poses an interesting problem during MQH₂ oxidation: electrons from MQH₂ cannot dwell in the membrane anchor subunit as they do at the two b-hemes of NarI, but they must travel to the [4Fe-4S] clusters in the electron transfer subunit concomitant with MQH₂ oxidation. As such, the Q-site in DmsC has been demonstrated to be conformationally and functionally linked to the −50 mV [4Fe-4S] cluster (FS4) in DmsB (111). In addition, a Tyr side chain from the electron transfer subunit has also been suggested to point toward the Q-site and be involved in the binding and oxidation of MQH₂ (22). In comparison with nitrate reductase, the lack of hemes in DmsC has also been proposed to cause a diminished driving force for the electrons to surmount the energy barrier imposed by the low-potential Fe-S cluster (E_m = −330 mV) (22). For instance, when the midpoint potential of FS4 is raised from −50 mV to +275 mV via a [4Fe-4S] → [3Fe-4S] conversion in the DmsB-Cys102Ser mutant (Fig. 11), electrons reside at FS4 and are unable to progress toward DmsA (110, 111). The ΔE_m from FS4 to FS2 is approximately −600 mV in this case, which is comparable to the scenario observed in the native nitrate reductase enzyme (45, 46).

Site-directed mutagenesis studies.

Two residues in DmsC, His65 and Glu87, have been shown to exert negative effects on HOQNO binding and lapachol oxidation when mutated (30, 111, 156). EPR studies on a DmsB-Cys102Ser/DmsC-His65Arg mutant showed a clear functional link between the Q-site and the −50 mV Fe-S cluster (FS4) (111). Incidentally, both residues are predicted to be in the first periplasmic loop of DmsC, between transmembrane α-helices 2 and 3 (Fig. 12), placing the Q-site on the periplasmic side of the membrane. Given that MQH₂ must donate electrons directly to DmsB, this implies a periplasmic location for DmsAB (see section below for further discussion). Other residues, Leu66, Gly67, Arg71, Phe73, and Ser75, examined in DmsC are also located in this periplasmic loop, but they do not show enhanced or compromised activities associated with MQH₂ binding or oxidation when mutated (30).

Membrane Targeting and Translocation of DMSO Reductase

The DmsA subunit is synthesized as a pre-protein of 814 (90,400 Da) residues. A 45-amino-acid tat leader is removed to yield a mature protein of 769 (85,795 Da) amino acids. The DmsA polypeptide is translated and folded in the cytoplasm into a cofactor-containing protein together with DmsB before it is targeted to the cytoplasmic membrane as a catalytic dimer (116). The DmsA leader sequence contains two positively charged Lys near the N terminus, two Arg residues at positions 16 and 17 consistent with typical twin arginine consensus sequences (S/T)-R-R-X-F-L-K, and an arginine near the cleavage site (11):

¹MKTKIPDAVLAAEVSRRGLVKTAIGGLAMASSALTLPFSRIAHA⁴⁵

The leader is presumably cleaved by leader peptidase, but this has not been shown. Chromosomal deletion of the tatABC operon leads to accumulation of DmsAB in the cytoplasm and abolishment of growth on minimal media containing DMSO as terminal electron acceptor (114). Similarly, deletion of the leader sequence from DmsA does not allow full assembly and proper targeting of DmsABC, and in fact leads to degradation of DmsA in vivo (116). Thus the leader sequence also plays a protective role that prevents DmsA degradation in the cytoplasm. The DmsD protein encoded by the ynfEFGHIdmsD operon is a specific chaperone that binds to the DmsA leader prior to targeting to the tat translocon (see below) (100).

Topology of DmsABC

The membrane topology of the catalytic dimer, DmsAB, has been controversial for many years. The immature DmsA contains a leader sequence at its N terminus that is recognized by the tat translocation machinery (116) and, as such, DmsA along with the leaderless DmsB appears to be a standard cofactor-containing substrate for tat translocation to the periplasm. Additionally, chromosomal deletion of the tatABC operon leads to accumulation of DmsAB in the cytoplasm and abolishes growth on minimal media containing DMSO (114). Unlike other tat substrates such as TorA (62), insertion of the MGD cofactor is not required for interaction with the tat translocase and a cofactor-less DmsABC is assembled on the membrane. The structure of the apo-form of NarGHI has been determined by X-ray crystallography and it appears as though the cofactor-less protein assembles with almost superimposable precision compared with the wild-type enzyme (105).

An earlier comprehensive study by Sambasivarao et al. (115) used several biochemical and immunological approaches to examine the topology of DmsAB on the membrane. These studies concluded that DmsAB was located on the cytoplasmic side of the membrane. Traditional cold osmotic shock (92) and chloroform washing of whole cells (3), methods that in general release periplasmic proteins, did not release any DmsAB. While this might be expected if DmsAB is localized to the periplasm but bound tightly to DmsC, no DmsAB is released in a DmsC deletion (116) where DmsAB should be free in the periplasm. The protease sensitivity of DmsAB was tested in cells with permeabilized outer membranes. No protease degradation was observed until the membranes were solubilized with Triton X-100 to destroy any topological constraints. Everted membrane vesicles prepared by French pressure lysis yielded protease-sensitive DmsABC, but higher temperatures, higher pH, and the presence of solubilizing agents were required to achieve inactivation similar to the fumarate reductase control. Immunogold electron microscopy of thin sections by using anti-DmsA and anti-DmsB antibody probes also indicated that DmsAB is oriented toward the cytoplasmic surface of the inner membrane. Attempts to construct fusions of alkaline phosphatase (PhoA) and bla with either DmsA or DmsB did not yield any active PhoA or β-lactamase in the periplasm (151). Sambasivarao et al. (115) carried out many of the experiments with both chromosomal- and plasmid-encoded DmsABC to eliminate artifacts due to overexpression. An independent experiment using the spin probe dysprosium(III)-EDTA also suggested a cytoplasmic location for the Fe-S clusters in DmsB (112).

In contrast to the Sambasivarao study, a later study by Stanley et al. (132) argued that the above experimental designs were flawed because: (i) overexpression of DmsABC from a multicopy plasmid might cause overloading of the tat translocase, leading to a cytoplasmic buildup of DmsAB and (ii) the PhoA protein requires formation of two intramolecular disulfides (130) but the tat system transports proteins in their folded states (47, 104, 118, 119). Thus an inability to select for DmsA-PhoA and DmsB-PhoA fusions does not conclusively indicate a cytoplasmic location for the dimer. In fact, it was shown that PhoA and LacZ fusions are not ideal candidates to use when studying periplasmic proteins that are exported via the tat system, but Bla and Cat fusions remain active when transported through the tat translocon (132). A periplasmic location of DmsAB was first observed in a ΔdmsC strain by Weiner et al. (150), but this result could not be reproduced (114) although it was confirmed by Stanley et al. (132). Indeed, cellular subfractionation of a ΔdmsC strain wherein a His₆ tag was attached to the chromosomal dmsB allowed the majority of DMSO reductase activity as well as DmsB polypeptides to be detected in the periplasm by enzyme assays and immunoblotting, respectively (132).

A bioinformatic analysis by Rothery et al. (108) of the topology of membrane anchor subunits of multisubunit reductases and dehydrogenases suggested that DmsAB is bound to the periplasmic surface of DmsC. Additionally, the bioenergetic study by Bogachev et al. (17) showed that DmsABC does not generate a PMF but serves as an electron sink indicating that the quinol oxidase and DMSO reduction sites are on the same side of the membrane. Mutant studies of quinol binding in DmsC would place DmsAB on the periplasmic side. One additional result adds to the puzzle of DmsAB topology. A recent study by Lindebstrauss et al. (78) showed that the tat proteins of R. capsulatus could correctly target the E. coli TorA, AmiA, and FdnGH subunits but could not target DmsAB, suggesting that there is something unique about the assembly of this latter complex.

Overall, the arguments for a periplasmic location of DmsAB are more convincing than the case presented for a cytoplasmic topology. An alternative model for the assembly of DmsABC is a "ball-in-glove" model where DmsAB sits deeply embedded in the membrane but is surrounded by the eight helices of DmsC. This possibility could account for both cytoplasmic and periplasmic observations of DmsAB and can reinforce the theory that DmsC prevents DmsAB from entering the periplasm (150). In this case, only an X-ray crystallographic structure would provide the answer to this lasting puzzle.

Regulation of DmsABC Expression

The dmsABC operon is located at 20' on the E. coli chromosome. Transcription of the operon depends on the oxygenation state of the cell and the availability of nitrate and molybdenum but not S- or N-oxides. The transcriptional start site is under the control of two promoters, P1 and P2; both have A/T-rich segments at the −35 and −10 positions (87) (Fig. 13). Under anaerobic conditions, the global anaerobic transcription factor FNR binds to the sequence TTGATnnnnAACAA (FNR consensus is normally TTGATnnnnATCAA) located 48 bp upstream of P1 to initiate transcription (9, 23). Transcription of the dmsABC operon is also under the regulation of the two-component NarXL system. In the presence of nitrate, NarL is phosphorylated and binds to the P1 region such that FNR and RNA polymerase are unable to bind (9, 23, 133). The dominant repressive role of phosphorylated NarL over FNR ensures that nitrate is metabolized before DMSO to maximize PMF generation. A third level of control is provided by the ModE protein, which is required for both FNR activation and NarL repression of the dmsABC operon (23, 87). The ModE protein also regulates the modABCD and moaABCDE operons in response to the presence of molybdate (43, 88, 147). DmsABC expression does not depend on substrate availability of DMSO or TMAO (23).

Fig. 13Organization of the upstream region preceding dmsA. The known promoter P1 is marked, as is the putative promoter P2. The −10 and −35 promoter regions are shown in blue boxes. Three direct-repeat NarL boxes are noted by the orange arrows. The Fnr box centered at −41.5 is shown in green. The integration host factor consensus sequence is underlined in blue, and the ModE consensus is shown in red.

YnfEFGH/DmsD—a Dms Paralogue

The DMSO reductase enzyme has a paralogue in the E. coli genome: ynfEFGHdmsD (80). It is found at 1.65 kbp on the E. coli chromosome (Fig. 3). This is not surprising because both the torCAD (25, 41) and narGHJI operons (16, 56) have paralogues. The role of the ynf operon remains an enigma. While there is some evidence from transcriptome studies that ynfEFGH operons are expressed, there is no evidence that this paralogue has any functional role. Strains carrying a ΔdmsABC mutation are unable to grow using DMSO or other S-oxides, because a terminal electron acceptor and immunoblot analysis using polyclonal anti-DmsA and anti-DmsB antibodies fail to reveal any cross-reacting polypeptides (117) in cells grown anaerobically on minimal medium with fumarate or DMSO as terminal electron acceptor. However, it has been shown that YnfEFG produced by recombinant expression techniques does cross-react with anti-DmsA and anti-DmsB antibodies. The dmsD(ynfI) open reading frame in the operon encodes a specific chaperone that has an essential role in assembly of DmsABC (100) (see below) and it is possible that the ynf operon has survived to provide a source of DmsD protein.

The DNA region upstream of ynfE has been examined for regulatory and promoter sequences, and it appears that overlapping elements are used to control transcription of the ynf and dms operons. A putative FNR box upstream of the ynfE promoter has been identified (80), implying anaerobic expression of the paralogue. Two potential NarL binding sites have also been spotted (85) in the promoter region to allow nitrate repression of the ynf operon, preserving the substrate hierarchy of metabolizing nitrate before DMSO as a terminal electron acceptor. Response to molybdate availability is controlled by a potential ModE binding site (85) that was discovered by MatInspector (98). Two σ⁷⁰ promoters have been identified by a computational approach (113), one upstream of ynfE and one upstream of ynfF (80). The σ⁷⁰ promoter for ynfF is located in the coding sequence of ynfE and presumably regulates transcription of ynfFGHdmsD. Upstream of ynfE, two binding sites for the general stress response regulator transcription factor RpoS ( 103) are present to initiate mRNA synthesis. Shine-Dalgarno sequences are located immediately upstream of ynfE and ynfF and are ideal for initiation of translation.

The gene products YnfE and YnfF are tandem duplications and are highly homologous to DmsA, showing 66% identity and an additional 15% similarity to the catalytic subunit. Both begin with typical tat leader sequences responsible for membrane translocation and targeting. Both YnfE and YnfF also retain conserved amino acids that are thought to be critical in Mo-bisMGD binding and catalytic activity. The electron-transferring subunit YnfG is the most conserved and is 94% identical plus 3% similar to DmsB. YnfH is the membrane anchor and quinol binding subunit. It is only 57% identical with DmsC and is predicted to be a transmembrane protein comprising eight transmembranal α-helices. DmsD is unique to the ynf operon.

Lubitz and Weiner (80) attempted to characterize the Ynf open reading frames by constructing chimeric proteins between the Dms and Ynf polypeptides. The putative anchor subunit, YnfH, can interact with DmsAB to form a trimeric enzyme with MQH₂ oxidase and DMSO reductase activity. The YnfG paralogue can substitute for DmsB as an electron transfer subunit and can interact functionally with both DmsC/YnfH and DmsA/YnfF. The YnfF subunit is more restrictive as it preferentially interacts and accepts electrons from YnfGH but is unable to do so from DmsBC. Coexpression of YnfE and YnfF in conjunction with YnfG/DmsC could not complement growth on DMSO in a ΔdmsABC strain, whereas the YnfFG/DmsC chimera can do so effectively, indicating that YnfE interferes with assembly of the terminal reductase.

When YnfFGH is overexpressed with multicopy plasmids it can functionally replace DmsABC as a DMSO reductase, but growth on glycerol-DMSO minimal medium yielded approximately 50% cell mass compared with the latter enzyme when the former enzyme was utilized. Enzymological studies of YnfFGH and YnfFG/DmsC show that they can also use a broad array of N- and S-oxides as is the case with DmsABC, but the turnover numbers are relatively poor (80, 125). The best substrate for YnfFGH appears to be hydroxypyridine N-oxide, followed sequentially by isonicotinic acid N-oxide, pyridine N-oxide, 2-chloropyridine N-oxide, DMSO, and finally TMAO (80). The possibility that ynfEFGH encodes a hydroxypyridine N-oxide reductase was refuted given that this substrate does not upregulate expression of the enzyme nor does the enzyme augment growth on minimal media containing hydroxypyridine N-oxide. The physiological substrate and function for the ynfEFGH gene products remain unknown.

DmsD—a DMSO Reductase-Specific Chaperone

The DmsD protein, initially named YnfI, has been shown to be a chaperone necessary for DmsABC maturation (100). This gene is unexpectedly part of the ynfEFGHdmsD operon rather than the dmsABC operon. DmsD is a peripheral membrane protein localized to the E. coli inner membrane under anaerobic conditions in wild-type E. coli or a ΔdmsABC strain, but DmsD is retained in the cytoplasm under aerobic conditions or in a ΔtatABCDE strain (95). It was first purified as a protein that interacts with the twin-arginine leader of DmsA and subsequently shown to interact with premature forms of DmsA and TorA (93). This finding is quite surprising given that the DmsA and TorA leader sequences are not functionally interchangeable (116). A three-dimensional structure of the homologue from Salmonella enterica serovar Typhimurium LT2 has been deposited in the Protein Data Bank (Fig. 14).

Fig. 14A PyMol-generated ribbon model for the structure of serovar Typhimurium DmsD (PDB ID 1S9U). Two sulfate molecules (red/yellow) and five di(hydroxyethyl)ether molecules were cocrystallized. The dotted yellow line shows residues 117 to 122, which are disordered in the crystals.

Binding of DmsD to pre-DmsA has been studied extensively by using biophysical techniques. Fusion of the DmsA leader at the N terminus of glutathione S-transferase allowed interaction with DmsD, whereas a C-terminal fusion did not (154). Binding of DmsD to the DmsA leader occurs with 1:1 stoichiometry with relatively strong affinity; a K_d of 0.2 μM was derived from isothermal titration calorimetry (154). DmsD has been purified and can be separated into three different states: monomer, dimer, and a heterogeneously folded monomer containing multiple conformations (120). All three forms of DmsD can interact with the DmsA leader in vitro (154).

Although DmsD interacts with the leader sequence of DmsA, it does not appear that DmsD is necessary for proper translocation and targeting of periplasmic components since the DmsA leader can still target green fluorescent protein to the periplasm in a ΔdmsD strain (100). It has therefore been proposed that DmsD acts as a chaperone for cofactor insertion into DmsA (100), much like the TorD chaperone for TorA maturation and Mo-bisMGD insertion (53, 97). As noted above, DmsD is not associated with the membrane in a ΔtatABCDE strain, suggesting that it binds to the tat complex, and it has been shown that TatB and TatC are needed for the interactions (95). Thus, DmsD is seen as a DmsA chaperone necessary for cofactor insertion into DmsA and assists in its initial interaction with the tat machinery.

TRIMETHYLAMINE N-OXIDE REDUCTASE

TMAO in the Natural Environment

TMAO is found at low levels in seawater and phytoplankton. It is found in the tissues of saltwater organisms including fish, cephalopods, mollusks, crustaceans, and seaweed, where it can be as much as 1% wt/wt (7). Saltwater fish maintain a high intracellular osmolarity similar to the osmolarity of seawater by using TMAO in combination with urea as osmolytes (149). TMAO stabilizes proteins in solution by enhancing hydrogen bonding to water molecules and may play a role in the formation of disulfide bonds (143). This counteracts the denaturing effect of urea. TMAO is not found in freshwater fish that reside in an environment of low osmolarity. The product of TMAO reduction is trimethylamine (TMA), which is responsible for the pungent smell of rotting fish. Humans can detect TMA at concentrations of less the 1 ppb. Some species of bacteria grow anaerobically with TMAO as an alternate electron acceptor. TMAO reductase activity has been found in marine bacteria (Shewanella, Vibrio, and Photobacterium), in photosynthetic bacteria (Rhodobacter), and in enterobacteria such as E. coli and serovar Typhimurium (7, 26). TMAO is reduced to TMA by bacteria associated with or contaminating the marine organisms. E. coli or serovar Typhimurium does not further metabolize TMA and thus is not used as a carbon or nitrogen source. In methylotrophs the oxidation of TMA to TMAO is the initial step in the utilization of TMA as a carbon, oxygen, and energy source. In the early 1970s Kim and Chang (72) observed that TMAO gave increased cell yield of cells grown anaerobically on glucose supplemented with TMAO, and they proposed that reduction must be associated with anaerobic respiration that supported oxidative phosphorylation. The TMAO/TMA half-reaction is +0.130 V, which is less than the nitrate/nitrite couple (+0.42 V) or the DMSO/DMS couple (+0.16 V), but higher than the fumarate/succinate couple (+0.031 V) (85) and in line with the 3 to 4 H⁺/2e⁻ as measured by Takagi et al. (137). Because TMAO reduction to TMA consumes protons, anaerobic growth on TMAO results in alkalization of the growth medium (Fig. 15).

Fig. 15The reaction catalyzed by TMAO reductase.

The early data on the number of distinct TMAO reductases (one constitutive and three to four inducible) is confused (85). We now know that under most physiological anaerobic growth conditions there is one inducible, periplasmic TMAO reductase encoded by the torCAD operon, which uses only TMAO as reductant, and a membrane-bound enzyme complex encoded by dmsABC, which reduces both DMSO and TMAO, although it has been proposed that TMAO is the natural substrate for this enzyme (85).

The tor Locus

The tor (trimethylamine N-oxide reduction) locus of E. coli was originally identified by Takagi et al. (137) in the early 1980s by mutagenesis and the TorA structural gene was mapped to a region that is homologous with the location of tor on the serovar Typhimurium genome (89). The tor locus is located at approximately 1,053 and 1,061 mbp (22 min) on the E. coli MG1655 chromosome and consists of two parts: the torCAD structural operon and an adjoining region encoding the torRTS regulatory operon (89) (Fig. 3). torT is encoded on the same strand as torCAD, while torR and torS are divergently encoded.

TorA—the Mo-bisMGD Catalytic Subunit

TorA is the catalytic subunit responsible for converting TMAO to TMA. It is assembled with the Mo-bisMGD cofactor in the cytoplasm with the help of the TorD chaperone (see below) and translocated to the periplasm via the tat translocon (94, 150). TorA is active only with TMAO and 4-methylmorpholine-N-oxide and does not have any S-oxide activity. This differs from DMSO reductase, which is active with a wide range of S- and N-oxides (125).

TorA is a member of the DMSO reductase molybdenum cofactor family, and several structures have been determined for members of this family including formate dehydrogenase-H (18), two DMSO reductases (121, 122), formate dehydrogenase-N (67), and nitrate reductase (14). Although the structure of E. coli TorA has not been determined, the structure of the homologue from Shewanella massilia has been reported at 2.5 Å resolution (24). Like RhDmsA discussed above, TorA is organized into four domains around the bis-MGD cofactor, domains II and III are somewhat symmetrical. Domain IV at the C terminus is the only one with a continuous amino acid stretch.

In addition to the dithiolene sulfurs, further coordination of the Mo atom is provided by the hydroxyl side chain of Ser149 and by an additional oxo group. Substrates reach the Mo via a funnel from the surface. Czjzek et al. (24) tried to rationalize the substrate specificity of TorA compared with DMSO reductase as an active center issue, because TMAO reductase only reduces TMAO yet DMSO reductase has wide substrate specificity (125). One possibility focuses on Tyr114 of RhDmsA, which is missing in TorA; this may be crucial for reduction of S-oxide compounds but not N-oxides. Another important difference is the number and arrangement of charged groups lining the inner surface of the funnel-like entrance. Finally, a surface loop at the entrance to the funnel, which is well defined in the TorA sequences and is more flexible in the DmsA structure, may play a role in substrate accessibility (Fig. 16).

Fig. 16A PyMol-generated ribbon model for the structure of TorA from S. massilia (PDB ID 1TMO). Two oxo ligands (red), the Mo atom (blue), and the two MGD molecules can be seen when looking directly into the active site.

In general, attempts to grow E. coli with tungsten result in the accumulation of molybdoenzyme-lacking cofactor or enzyme with little or no activity. Buc and colleagues (20) were able to replace the transition metal Mo in E. coli TorA with W. TorA assembles with W and the W-substituted enzyme is active and now uses both TMAO and DMSO. This observation downplays the theory of a critical Tyr residue being responsible for the difference in the ability to reduce S- and N-oxides. The W-substituted enzyme is also more heat resistant.

TorC—the Pentaheme Cytochrome

The TorC protein is a pentaheme c-type cytochrome that is essential for TMAO respiration (39). The protein is synthesized as a 390-amino-acid precursor. TorC is translocated to the periplasm via the classical sec system (138). A 28-amino-acid N-terminal leader is cleaved from the mature enzyme yielding a 362-amino-acid (40.47-kDa) mature protein that is bound to the cytoplasmic membrane by a membrane anchor at the amino end. The bulk of TorC faces the periplasm where it can interact with TorA. TorC must interact with the menaquinol pool in the membrane and must have a MQH₂ binding site, but no information is available on the residues that make up this site.

The complexity of the TorA maturation and translocation machinery is matched by that needed to assemble the pentaheme cytochrome (TorC) onto the outer surface of the cytoplasmic membrane. TorC is largely matured by using the standard cytochrome c maturation machinery. After translocation via the sec system, the cysteines in the CXXCH are rapidly oxidized by DsbA/DsbB, which converts TorC to a prefolded form that precedes maturation (138). Final maturation involves reduction of the cysteine disulfides and requires the gene products of ccmABDEFGH and dsbD (75, 138).

TorC is composed of two domains. The N-terminal half is a member of the NapC/NirT family of c-type cytochromes and encodes four CXXCH motifs binding four c-type cytochromes (2). The C-terminal half has an additional CXXCH domain found only in TorC- and DorC-type proteins at residues 329 to 333 (124). Gon (39) cloned the amino (TorC_N) and carboxy (TorC_C) halves independently using the TorT leader to direct TorC_C to the periplasm. Neither domain alone supported growth, suggesting that the entire TorC protein was needed.

Attempts to overexpress TorC have been hampered by the limited capacity of E. coli to make c-type cytochromes encoding the c-type cytochrome biosynthetic activity (138) even when the ccm genes are added on another plasmid. The ccm genes were under the arabinose promoter so that expression could be exquisitely controlled. Very low levels of arabinose were used (0.0005%) to avoid overloading the c-type cytochrome biosynthetic system. TorC maturation is prevented by TorC overproduction, dithiothreitol (DTT) addition, or disruption of the ccm genes involved in holoTorC production.

TorC has a typical c-type cytochrome spectrum. In the oxidized state the Soret band at 411 nm is observed. In the reduced state bands were seen at 417 nm (Soret), 523 nm, and 552 nm. Redox potentiometry of TorC indicated E_m values of −177 mV corresponding to two hemes, −98 mV corresponding to two hemes, and +114 mV (one heme). The negative potential hemes were localized to the amino-terminal half and the positive heme to the carboxyl-terminal half of the protein. This suggests that electrons flow from the amino-terminal half to the carboxyl-terminal domain and then to TorA.

Gon et al. (39) examined the physical interaction of TorC with TorA. They found that TorC migration on native polyacrylamide gels was retarded by TorA, suggesting that the two proteins interact. These binding studies were extended by using histidine-tagged and -purified TorC_N and TorC_C halves of TorC by surface plasmon resonance. Surprisingly, they found that the TorC_C half does not interact with TorA but the TorC_N interacts with a K_d of 4.5 × 10⁻⁸ M. When intact TorC was used they noted that binding was in two steps with K_d1 = 1.7 × 10⁻⁸ M and K_d2 = 3.0 × 10⁻⁶ M. This suggests a model in which TorA folds around TorC such that TorC_N can bind TorA and TorC_C interacts with the molybdopterin cofactor in TorA (39). Based on the ease with which TorC is released by osmotic shock, this interaction must be relatively transient in the cell.

TorD—the Chaperone Protein

The third gene in the tor operon is not part of the functional complex. It is a 22-kDa cytoplasmic protein related to a large number of prokaryotic homologues with similarities ranging from 20 to 64%. The TorD family has been divided into four groups or clades by using an unrooted phylogenetic tree (54), and several of these clades are also related by molybdoenzyme classification (86). The clade that includes the E. coli TorD also contains S. massilia TorD (26), R. capsulatus DorC (123), and R. sphaeroides DmsB (91). It is related to the type III molybdoenzymes and is thus also defined as type III. Although these proteins are clearly related, the TorD of S. massilia could not replace E. coli TorD for either in vitro binding studies or in vivo complementation (54). Proteins of type II, which includes the DmsD and YcdY proteins, could not replace TorD for complementation, although they did compete for TorD binding to TorA.

The TorD protein of S. massilia has been the subject of structural investigation. This protein is 33% identical in sequence to E. coli TorD and exists in multiple oligomeric states. Interconversion between the monomeric and dimeric states required destabilization of the native fold of the protein (140). The 2.4-Å X-ray crystallographic structure of the dimeric form of S. massilia TorD has been solved (139) and it reveals a dumbbell-like structure with extreme domain swapping between the two subunits. TorD is an intertwined dimer with a globular domain of all α-helical composition.

During the past few years considerable information has become available on the chaperone function of TorD with TorA. Pommier et al. (97) originally showed that in E. coli the absence of TorD led to a loss of TorA activity. However, a significant amount of TorA was still present in the periplasm. These studies suggest that TorD alters the conformation of apoTorA (lacking the Mo-bisMGD cofactor) to allow for better cofactor insertion. It was found that, when the activation was carried out in vitro with purified components, apoTorA was activated only in presence of TorD and MobA (MobA converts Mo-MPT to Mo-MGD). In a strain deficient in TorD and the MGD cofactor, TorA was proteolytically digested; however, an excess of TorD prevented the proteolytic degradation. The proteolytic degradation was obvious in particular at elevated temperatures (42°C) (31). Using surface plasmon resonance and native polyacrylamide gel electrophoresis it could be shown that TorD physically binds to apo-TorA (53).

More recent experiments by Hatzixanthis et al. (49) and Iobbi-Nivol and colleagues (31, 32, 33, 53, 54, 139) indicate that TorD has a dual role: to act in the maturation of apo-TorA with MGD as noted above and to act as an escort for TorA to the tat translocation system (12, 150). These two roles are independent. The latter role is to prevent the export of immature TorA. Using a mutant of TorA lacking the Tat signal sequence it could be shown that the truncated protein still binds Mo-bisMGD and the binding was stimulated by TorD, as normal. It was shown (63) that TorD bound to the TorA signal sequence. Using a synthetic TorA leader it could be shown that TorD bound to the TorA leader (49), and this binding involved portions of the leader independent of the SRRxFLK motif. Hatzixanthis et al. (49) further showed that TorD is a GTP-binding protein. No GTPase activity could be measured, and the role of GTP is unclear although it may be regulatory. Addition of TorA to a GTP-TorC mixture results in increased GTP affinity.

The TorYZ Operon

The E. coli chromosome contains a paralogue of the torCAD operon at 42.1 minutes (1952 to 1956 kbp on the MG1655 chromosome) (Fig. 3) termed torYZ, originally termed yecKtorZ. Unlike the torCAD operon the paralogue lacks a gene related to the torD chaperone protein. The existence of a paralogue to the tor operon is similar to the nar and dms operons, which also contain paralogues on the chromosome (16, 80). In all cases the paralogue is expressed at relatively low levels.

torZwas originally characterized as BisZ, a second biotin sulfoxide reductase in E. coli (25) and it is 63% identical with BisC while only 44% identical with TorA. TorZ accounted for only 4% of the BSO reductase activity in E. coli. However, Gon et al. (41) have shown that unlike BSO reductase, TorZ has a typical tat leader similar to TorA and DmsA and the mature protein is located in the periplasm. Gon et al. (41) have shown that translocation depends on a functional tat system. TorZ has wide catalytic activity to S- and N-oxide compounds and, importantly, it displays higher activity with TMAO than BSO. Upstream of torZ is the 1-kbp open reading frame torY(yecK) that is cotranscribed with torZ. This open reading frame encodes a membrane-bound pentaheme cytochrome, TorY, that is homologous to TorC and the corresponding subunit of the Dor system of R. capsulatus. E. coli contains only six pentaheme cytochromes, two involved in the periplasmic nitrate reductase (Nap) (42), two involved in nitrite reduction (Nrf) (42), TorC and TorY. TorY can support anaerobic respiration with TorZ in a strain lacking torCAD and dmsABC when it is expressed from a ptac promoter in the presence of isopropyl-β-d-thiogalactopyranoside (IPTG). Thus TorYZ can form a respiratory system capable of supporting anaerobic growth on TMAO, DMSO, and BSO. Unlike TorCA, TorYZ is also able to support growth on BSO and DMSO in a strain lacking torCAD and dmsABC, indicating that TorZ has a much broader substrate specificity than TorA.

Like the expression of the ynfEFGHdmsD operon (41), expression of torYZ remains an enigma. Expression is very low and the operon under control of the native promoter is not induced by TMAO, DMSO, or BSO, and it is unable to support growth on S- or N-oxides when the torCAD or dmsABC operons are mutated. Because the activity profile of TorZ is not identical with TorA or DmsA, perhaps it has a unique activity and the inducer has not yet been discovered.

A major difference between the torCAD operon and torYZ is the absence of a TorD-like chaperone. As noted above TorD has multiple roles in the assembly of TorA. Gon et al. (41) have shown that TorD is not required for TorZ maturation, suggesting that TorZ maturation is chaperone independent or a chaperone is encoded elsewhere on the chromosome. A similar situation exists for DmsA where the DmsD chaperon is encoded in the ynf operon and not in the dmsABC operon. The lack of a chaperone may account for the very low level of TorZ expression.

Tor Regulation

The expression of the torCAD operon is controlled by anaerobiosis and TMAO (96). Surprisingly, although TorA will only reduce TMAO and methylmorpholine-N-oxide, the torCAD operon can be induced by a range of S- and N-oxide compounds including DMSO (55). Induction of the torCAD operon occurs in a somewhat unusual and complicated process (128). Anaerobiosis induces expression about 10-fold based on a torA'-lacZ reporter fusion but the torCAD operon is unusual in that the anaerobic induction is not under the control of the FNR (131) or ArcA (59) global regulatory proteins and no FNR or ArcA boxes can be found in the region upstream of the torC promoter (96, 129). torA expression is not affected by ModE, which mediates molybdenum regulation of dmsABC (87). Further, torA expression is not subject to the hierarchical control whereby substrates with higher redox potentials are preferentially induced (96, 134) and lacks the NarL consensus binding sequence (38). Regulation of torCAD is mediated by a group of three open reading frames located adjacent to the torCAD operon, torS, torT, and torR, which are divergently expressed (Fig. 3).

TorSR Two-Component Regulation

The regulation by TorA expression is strictly dependent on TMAO and is mediated by a two-component histidine/aspartate kinase phosphotransfer relay (136). torS was originally identified by insertion mutagenesis (69) and encodes a membrane-bound sensor protein for TMAO. Although TorA only utilizes TMAO and 4-methylmorpholine-N-oxide as substrates, a wide variety of S- and N-oxide compounds including DMSO can induce torCAD expression via TorS binding (55). The TorS sensor histidine kinase contains an N-terminal periplasmic domain that senses the signal. Jourlin et al. (69) have identified a three-amino-acid deletion mutation in this domain (presumably in the TMAO detector site) which results in full constitutive expression in the absence of TMAO, confirming that the periplasmic domain senses TMAO. The kinase domain (transmitter domain) of 240 amino acids is located in the cytoplasm. This domain contains four motifs including the active site where a histidine is autophosphorylated when TMAO is detected (135). A linker region that includes a transmembrane α-helix is located between the sensor and transmitter domain. The linker region plays an essential role in propagating conformational changes from the periplasm to the histidine kinase transmitter domain and mutations have been isolated in this domain, which gives some expression of torCAD in the absence of TMAO. This phosphoryl group is transferred to an aspartate on the response regulator, TorR (receiver domain), leading to transcription of torCAD.

Phosphotransfer by two-component regulatory systems occurs via one of two mechanisms (27). In the simple two-step mechanism the sensor kinase directly phosphorylates the response regulator. A more complex four-step mechanism utilizes a sensor kinase with two or three phosphorylation sites. TorS falls into the latter category because it involves three phosphorylation sites and is termed an "unorthodox kinase" (68). The additional phosphorylation sites are found in a histidine phosphotransferase domain that contains a receiver domain and an alternate transmitter domain at the C terminus. A similar architecture is seen in ArcB (57). ArcB is part of the ArcA/ArcB two-component regulatory system for anaerobic repression of certain respiratory enzymes (59). All three phosphorylation sites on TorS [His443 → Asp723 → His850 → Asp(TorR)] are required for TMAO induction of torCAD. TorS can rapidly dephosphorylate phospho-TorR when TMAO is removed, and this appears to work by a reverse phosphorelay since His850 and Asp723 are essential for dephosphorylation ( 4). The unorthodox kinase proteins often have an additional phosphatase that regulates the signal transduction at intermediate checkpoints. In the TorSR system, a novel protein TorI interacts with the effector domain of TorR to inhibit torCAD expression without affecting TorR binding to DNA (6). TorI is a very small protein of 66 amino acids. It appears that TorI prevents TorR from interacting with RNA polymerase, thus limiting recruitment of the polymerase to the torCAD promoter. TorI has no effect on signal transduction or the phosphorelay mechanism. Homologues of TorI bearing 100% identity include hkaC of the coliphage HK620 and gene 18 of Shigella phage Sf6, suggesting that TorI is of phage origin and belongs to the defective prophage KplE1 that is present on the E. coli genome (28). Elantak et al. have shown that TorI has excisionase activity and excises the cryptic prophage KplE1. The solution structure of TorI has been determined (28), and although it lacks sequence similarity, the three-dimensional structure is highly homologous to the ^λXis, Mu bacteriophage repressor and transposase.

An additional level of regulation of torCAD is mediated by apo-TorC (40). Immature TorC lacking the hemes cannot interact with TorA but can interact with the periplasmic domain of TorS inhibiting the TorS kinase and thus negatively regulating expression of torCAD. This regulation is highly specific to apo-TorC and cannot be replaced by other immature c-type cytochromes. The regulation depends on the mono-heme carboxy-terminal domain of TorC and not on the tetra-heme amino-terminal domain.

TorR is the 25-kDa response regulator protein of the two-component regulatory system that controls expression of torCAD in response to TMAO (127). TorR is a member of the OmpR family of response regulators (58, 74, 82). TorR binds to the short torR-torC intergenic regulatory region that contains four direct repeats of a decameric consensus sequence (tor box) (CTGTTCATAT, box 1, 2, and 3) (CCGTTCATCC, box 4) (Fig. 17) located in the untranslated nucleotide region between the divergently expressed tor operon and torR (Fig. 3). TorR protein binds to the torC regulatory region based on DNA footprinting and gel retardation assays. Three regions are protected by TorR binding (127) (Fig. 17): one of 24 nucleotides covers boxes 1 and 2, the second is upstream of the −35 region and covers box 3, and the third is downstream of the −35 sequence and covers the fourth tor box. Mutating any one box results in dramatic decrease of tor expression based on a torA'-lacZ reporter construct, indicating that all four boxes are required. The upstream boxes, 1 and 2, display high-affinity binding and form a nucleoprotein complex that covers the torR transcription start site. Simon et al. (127) have proposed that unphosphorylated TorR dimer binds to boxes 1 and 2. In the presence of TMAO, TorR is transphosphorylated by TorS leading to the cooperative formation of a tetramer attached to boxes 1 and 2. One subunit then binds to box 4, and subsequently another subunit contacts the very-low-affinity box 3. This complex positions the promoter for RNA polymerase binding and transcription of torCAD. All four boxes must be close to each other and on the same side of the DNA helix for induction of torCAD (5). High-level expression of TorR mimics the presence of the inducer TMAO, indicating that, if there is enough protein available, it can bypass the need for histidine phosphorylation (128). The anaerobic regulator of torCAD is unknown, and it does not appear that TorR plays a role in anaerobic regulation.

Fig. 17Organization of the divergently expressed intragenic region between torC and torR. The −10 and −35 promoter regions are shown in blue boxes. Four direct decameric tor boxes are noted by an orange bar. The thickness of the bar roughly approximates the affinity of TorR for the DNA.

The expression of torR is negatively autoregulated and is independent of TMAO and TorS (5), suggesting that both unphosphorylated and phosphorylated TorR can bind. It only involves TorR binding to the high-affinity sites boxes 1 and 2, thus preventing RNA polymerase from binding.

TorT

The torT gene encodes a periplasmic protein of 35.7 kDa. Transposon insertion within torT dramatically reduces expression of a torA'-lacZ fusion reporter (70). TorT bears some sequence similarity to the periplasmic ribose binding protein and it has been proposed that TorT serves as a ligand binding protein to deliver TMAO to the sensor kinase TorS in the cytoplasmic membrane. In support of this TorT has been found to be dispensable and excess TorR in a torT mutant results in partial constitutive expression of torA'-lacZ.

SUMMARY

The major attributes among DmsABC, TorCA, and their respective homologues are listed in Table 1. All four proteins are synthesized with a tat leader in the premature soluble domain and are thus targeted to the periplasm by the tat translocon; however, the arguments for DmsABC remain debatable and the case for YnfFGH has yet to be fully explored, although a periplasmic location could be conjectured based on its strong sequence and functional relation to DmsABC. Assembly of DmsABC and TorCA requires the chaperones DmsD and TorD, respectively. YnfFGH assembly has not been tested in the absence of DmsD, but its genetic codes are located in one operon. The assembly of TorYZ has been shown to be chaperone independent. The expression of DmsABC and its paralogue YnfFGH are positively regulated by the FNR protein under anaerobic conditions. TorCA is also expressed under anaerobic conditions, although via an unknown mechanism, and is additionally controlled by the presence of S- and N-oxides. In terms of substrate utilization, note that DmsABC, YnfFGH, and TorYZ can reduce a variety of S- and N-oxides; in contrast, the native TorCA can only reduce N-oxides, but the W-substituted TorCA can reduce S- and N-oxides.

Table 1Summary of attributes for the four enzymes discussed in this review

Although the two paralogues YnfFGH and TorYZ have enzyme activity toward S- and N-oxides, their physiological functions remain a mystery since their expressions are almost negligible and are negatively dominated by expressions of DmsABC and TorCA, respectively. It is possible that the ynfEFGH and torYZ operons are the evolutionary remains of inefficient enzyme systems that have been functionally replaced by the dmsABC and torCA gene loci. Even the DmsABC and TorCA enzymes themselves have overlapping abilities to reduce certain N-oxides. Such duplicate systems might be a compensatory mechanism established by bacterium such as E. coli to survive in environments that limit energy conservation.

ACKNOWLEDGMENTS

Research in our laboratory is funded by the Canadian Institutes of Health Research, the Alberta Heritage Foundation for Medical Research (AHFMR), and the National Institutes of Health. J.H.W. is a Canada Research Chair in Membrane Biochemistry, and V.W.T.C. holds an AHFMR Studentship.