Engineering a Novel Iron-Sulfur Cluster into the Catalytic Subunit of Escherichia coli Dimethyl-sulfoxide Reductase*

Dimethyl-sulfoxide reductase (DmsABC) is a complex [Fe-S] molybdoenzyme that contains four [4Fe-4S] clusters visible by electron paramagnetic resonance (EPR) spectroscopy. The enzyme contains four ferredoxin-like Cys groups in the electron transfer subunit, DmsB, and an additional group of Cys residues in the catalytic subunit, DmsA. Mutagenesis of the second Cys, Cys-38, in the DmsA group to either Ser or Ala promotes assembly of a fifth [Fe-S] cluster into the mutant enzyme. The EPR spectra, the temperature dependences, and the microwave power dependences demonstrate that the new clusters are [3Fe-4S] clusters. The [3Fe-4S] clusters in both of the C38S and C38A mutant enzymes are rela- tively unstable in redox titrations and have midpoint potentials of approximately 178 and 140 mV. Mutagene- sis of the DmsA Cys group to resemble a sequence capa-ble of binding an [4Fe-4S] cluster did not change the cluster type but reduced the amount of the cluster present in this mutant enzyme. This report demonstrates that all four EPR detectable [Fe-S] clusters in the wild-type enzyme are ligated by DmsB. Wild-type DmsA does not ligate an [Fe-S] cluster that is visible by EPR spectroscopy. Escherichia Me 2 SO as respi-ratory oxidant an electron transfer chain termi-nating with Me 2 SO reductase, DmsABC 1 (1).


Dimethyl-sulfoxide reductase (DmsABC) is a complex [Fe-S] molybdoenzyme that contains four [4Fe-4S] clusters visible by electron paramagnetic resonance (EPR) spectroscopy. The enzyme contains four ferredoxin-like
Escherichia coli grows anaerobically using Me 2 SO as respiratory oxidant by expressing an electron transfer chain terminating with Me 2 SO reductase, DmsABC 1 (1). DmsABC is a complex [Fe-S]-and molybdenum-containing enzyme located on the cytoplasmic surface of the inner membrane (2). DmsA is the largest subunit (87.4 kDa) and binds a molybdenum-molybdopterin guanine dinucleotide cofactor, Mo-MGD (3). DmsB (23.1 kDa) is an electron-transfer subunit containing four ferredoxin-like Cys groups proposed to ligate [4Fe-4S] clusters (1,4). DmsC (30.8 kDa) is a membrane-intrinsic subunit that anchors DmsAB to the membrane. DmsC accepts electrons from menaquinol, transferring them through the [4Fe-4S] clusters in DmsB to the active site in DmsA (1). The dmsABC operon has been cloned and sequenced, and the enzyme can be expressed to high levels in membranes (5,6).
DmsABC is a member of a family of molybdenum-containing oxidoreductases with highly conserved sequences (1, 6 -8). These are enzymes that reduce Me 2 SO, trimethylamine Noxide (TMAO) (9), nitrate (7, 10 -14), biotin sulfoxide (15,16), and polysulfide (17) or enzymes that oxidize formate (18 -24). Each enzyme contains a large catalytic subunit with a noncovalently bound molybdenum cofactor. The sequence identity is located in segments throughout the polypeptide. Many of these enzymes are similar to DmsABC in prosthetic groups and subunit composition.
Ferredoxins that contain [4Fe-4S] clusters usually ligate these clusters by Cys groups consisting of four Cys residues spaced such that the first two Cys residues are separated by two amino acids, while the spacing between the second, third, and fourth Cys residues is somewhat variable. A Cys group from the thermophilic methanogen Methanococcus thermolithotrophicus (25) has four amino acids separating the first two ligands, but we have not identified a Cys group with three intervening residues. The first three Cys residues and one distal Cys, often from a second Cys group elsewhere in the protein, provide the ligands to the cluster (26,27). Alignment of the amino-terminal regions of the large subunits of the molybdoenzymes ( Fig. 1) shows four conserved Cys residues arranged in a manner reminiscent of a [4Fe-4S] ferredoxin Cys group. The sequences can be divided into three types. Type I enzymes contain three Cys residues spaced similar to a bacterial ferredoxin Cys group and one other conserved Cys, which could provide the fourth ligand. The Type II enzymes also have four Cys residues, but the spacing is such that three amino acids instead of two separate the first and second Cys residues. DmsABC belongs to this group, as do the two membrane-bound E. coli nitrate reductases in which the first Cys is replaced by a His. His can be a ligand to a [4Fe-4S] cluster, as in the nickel-iron hydrogenase from Desulfovibrio gigas, but the first two ligands of this cluster, His and Cys, are separated by two amino acids (28). The Type III enzymes include biotin sulfoxide reductase (BisC) and TMAO reductase (TorA) which share sequence identity with the other molybdoenzymes but do not contain the Cys region.
The periplasmic nitrate reductase, NapAB, from Thiosphaera pantotropha has been shown to contain a [4Fe-4S] cluster (29). This is a Type I enzyme, and the Cys residues in NapA are the only candidates to ligate the [4Fe-4S] cluster (14,29). This raises the possibility that the Cys region may ligate a [4Fe-4S] cluster in other members of this family. The subunit of E. coli formate hydrogenlyase that contains the formate dehydrogenase activity, FdhF, may contain an [Fe-S] cluster based on iron analysis (30).
The [Fe-S] clusters of DmsABC have been characterized by electron paramagnetic resonance (EPR) spectroscopy. The enzyme contains four [4Fe-4S] clusters with midpoint potentials, E m,7 ϭ Ϫ50, Ϫ120, Ϫ240, and Ϫ330 mV (4). These clusters are believed to be ligated by the four ferredoxin-like Cys groups (I-IV) in DmsB (1,4), although the possibility exists that the Cys region in DmsA might be able to ligate a cluster. Sitedirected mutagenesis of DmsB groups III (31) and I 2 has demonstrated that these Cys groups provide ligands for two [4Fe-4S] clusters.
The DmsA Cys region has previously been examined through the use of site-directed mutagenesis (32). Cys-38, Cys-42, and Cys-75 were mutated to Ser, and only the C75S mutant enzyme was able to support growth on Me 2 SO. All three mutant enzymes retained some level of catalytic activity with an artificial electron donor, reduced benzyl viologen (BV . ϩ ) and with the quinol analog, 2,3-dimethyl-1,4-naphthoquinol. The C38S and C42S mutants were blocked in using electrons from the quinol pool to reduce substrate, although the [4Fe-4S] clusters responded to the redox state of the quinol pool. In this manuscript we show that the DmsA Cys group does not coordinate an EPR visible [Fe-S] cluster, but when the second Cys of this group is mutated, a [3Fe-4S] cluster assembles into the mutant enzyme.
Materials-Synthesis of oligonucleotides and DNA sequencing were carried out in the Department of Biochemistry DNA Core facility at the University of Alberta using an Applied Biosystems model 392 DNA synthesizer and a model 373A DNA sequencer (Perkin Elmer). Restriction endonucleases and modifying enzymes were obtained from Life Technologies Inc. All other materials were of reagent grade and were obtained from commercial sources.
Site-directed Mutagenesis-Manipulations of strains and plasmids were carried out as described in Sambrook et al. (34). The double mutant of dmsA with substitutions in residues Asn-37 and Cys-38 was generated through PCR mutagenesis of pC38S using the following mutagenic primers: 5Ј-GTACAGTTTGCTCTGGTAG-3Ј and 5Ј-CGAC-TACCAGAGCAAACTG-3Ј (35). The mutant product was cloned into pTZ18R and sequenced. The fragment bearing the double mutation was subcloned into the wild-type operon to construct the plasmid pC38S,N37C.
Harvesting of Cells and Preparation of Membrane Fractions-For whole cell EPR samples, cells were harvested, washed, and resuspended in degassed 100 mM MOPS, pH 7.0, 5 mM EDTA buffer containing 20 mM succinate (31). Glycerol-fumarate grown cells were harvested, washed, and resuspended in 50 mM, MOPS, pH 7.0, 5 mM EDTA. Phenylmethanesulfonyl fluoride (0.2 mM) was added, and cells were subjected to French pressure lysis and differential centrifugation to prepare crude membrane fractions (4). For EPR experiments, the membranes were washed and resuspended in 100 mM MOPS, pH 7.0, 5 mM EDTA. Membranes were stored at Ϫ70°C prior to use.
Protein Determination and Polyacrylamide Gel Electrophoresis-Protein concentrations were estimated by a modification of the Lowry procedure (38) using a Bio-Rad bovine serum albumin protein standard. Polyacrylamide gel electrophoresis was carried out using the Bio-Rad mini-gel system and a discontinuous SDS buffer system (39). Gels (12.5%) were stained with Coomassie Blue and destained, and the relative amount of DmsA protein was determined using a Joyce-Loebel Chromoscan 3 densitometer.
the individual figure legends. Spin quantitations were carried out by double integrations of spectra from redox titrations (microwave power, 2 milliwatts; temperature, 12 K). The presence of background signals (primarily fumarate reductase) in the membranes was partly compensated for by subtracting the base line intensity (at approximately 0 mV) from the fully oxidized and reduced intensities.

Growth Properties and Enzyme Activities of the Cys-38
Mutants-The ability of the mutant enzymes to support growth on Me 2 SO was assessed using the dmsABC deletion strain, DSS301 (Table I). Only the wild-type enzyme was able to support growth. DSS301 was not used for further characterization of the mutant enzymes for reasons previously mentioned (31,32). The specific activities of HB101 membrane preparations were assayed using TMAO as the electron acceptor and the artificial electron donor, BV . ϩ (Table I). Both mutant enzymes showed decreased specific activities when compared with the wild-type enzyme, as shown previously (32). The relative amounts of enzyme present in the F36 membrane preparations used for the EPR studies were determined by densitometry ( Table I). The relative percentage of DmsA in the F36/pC38A membranes is lower than in the wild-type or C38S preparations.
EPR Characteristics of Dithionite-reduced F36 Membranes Containing Overexpressed Wild-type and Mutant DmsABC-In F36 the Mo-MGD cofactor is not inserted into DmsABC and cannot interfere with the [Fe-S] signals in the redox titrations as occurs when DmsABC is expressed in HB101 (4). Although the enzyme produced in F36 is inactive, it assembles the [Fe-S] clusters normally. Fig. 2 shows the EPR spectra of reduced membranes at 12 K. F36/pBR322 membranes contained very little DmsABC, expressed from the chromosomal copy of the operon. The spectrum from these membranes shows a peak at g ϭ 2.02 and a peak-trough at about g ϭ 1.94. These features are characteristic of the reduced [2Fe-2S] cluster of fumarate reductase, FR1 (40,41). F36/pDMS160 membranes contained a high level of DmsABC, and the spectrum of the dithionitereduced samples (Fig. 2b) is very similar to spectra obtained from both purified and membrane-bound DmsABC (4,31). The spectrum contains peaks located at g ϭ 2.06, g ϭ 2.02, g ϭ 1.99, and g ϭ 1.95. Two troughs are observed at g ϭ 1.92 and g ϭ 1.87. The peak at g ϭ 2.02 is most likely due to fumarate reductase, which would also contribute to the peak/trough at g ϭ 1.95 to g ϭ 1.92 (31). The peaks at g ϭ 2.06 and g ϭ 1.99 and the peak/trough at g ϭ 1.95 to g ϭ 1.92 arise from DmsABC. Repeating the dithionite reduction at pH 9, to decrease the redox potential, did not change the spectrum observed. Fig. 2, c and d, shows the spectra of dithionite-reduced F36 membranes containing the C38A and C38S mutant enzymes, respectively. The features of these spectra are very similar to that of F36/pDMS160 membranes, and all of the DmsABC features are present. The slight difference in the size of some of the features in the reduced EPR spectrum of the C38A mutant enzyme is due to the lower amount of enzyme present. Reduction by dithionite at pH 9 did not highlight any difference between the spectra of the wild-type and mutant enzymes (data not shown). Fig. 3 shows EPR spectra recorded at 12 K of F36 membranes oxidized with ferricyanide. The spectrum of F36/pBR322 membranes (Fig.  3a) has a small peak at g ϭ 2.02 with a broad trough immediately upfield, characteristic of the oxidized [3Fe-4S] cluster of fumarate reductase, FR3 (40,41). The spectrum of F36/ pDMS160 membranes (Fig. 3b) shows similar features to that of F36/pBR322 membranes, indicating that FR3 is the major EPR visible species present. Spectra from the mutant enzymes show major new features. The spectrum of F36/pC38S membranes (Fig. 3d) is comprised of a sharp peak at g ϭ 2.03 and a peak/trough centered at g ϭ 2.00. The F36/pC38A spectrum (Fig. 3c) is similar, but the signal is broader than in F36/pC38S membranes. The EPR spectra of the oxidized mutant enzymes can be attributed to centers having axial symmetry with approximate g values of g z ϭ 2.03 and g xy ϭ 2.00. Oxidized spectra of HB101 membranes containing the C42S and C75S mutant enzymes were similar to that of FR3, indicating that these mutants do not assemble significant amounts of a [3Fe-4S] cluster ligated in DmsA (data not shown).

EPR Characteristics of Oxidized F36 Membranes Containing Amplified Wild-type and Mutant DmsABC-
To identify the nature of the paramagnetic species present in the oxidized Cys-38 mutants, we studied the microwave power saturation properties and the temperature dependences of the new signals. Fig. 4 shows the effect of increasing microwave power on the mutant center signals at 12 K. Microwave power  2. EPR spectra of reduced F36 membranes. a, F36/pBR322; b, F36/pDMS160; c, F36/pC38A; and d, F36/pC38S. Samples were incubated at 25°C under argon for 2 min. 5 mM dithionite was added, and the samples were incubated under argon an additional 10 min before freezing in liquid nitrogen. Spectra were recorded under the following conditions: temperature, 12 K; microwave power, 20 milliwatts; microwave frequency, 9.45 GHz; modulation amplitude, 10 G pp at 100 KHz.
[Fe-S] Cluster in DmsA Mutants saturation data obtained from F36/pC38A membranes were fitted to an empirical equation to obtain the microwave power required for half saturation of the signal, the P1 ⁄2 (42). A twocomponent model was required to fit the data giving P1 ⁄2 values of 1 (40%) and 185 milliwatts (60%). The presence of two components suggests that the protein conformation around the cluster is not homogeneous. Microwave power saturation data from the F36/pC38S membranes were fitted to one component with a P1 ⁄2 of 9 milliwatts. Fig. 5 shows the effect of temperature on the intensity of the new signals. The F36/pC38A and F36/ pC38S signals reached a maximum intensity at 9 K and 11 K that decreases until, at 30 K, they were hardly visible. The microwave power saturation and temperature dependences of the new centers in the C38A and C38S mutants are typical of the behavior of [3Fe-4S] clusters (43,44). We therefore assign these signals to new [3Fe-4S] clusters ligated by the DmsA Cys group in the mutant enzymes.
Redox Titrations of the C38A and C38S Mutant Enzymes-Redox titrations were carried out to determine the midpoint potentials for the [4Fe-4S] clusters and the new [3Fe-4S] clusters in the DmsA mutant enzymes (Table II). In the mutant enzymes, four [4Fe-4S] clusters were detected with midpoint potentials close to those of the wild-type enzyme. The amount of each cluster present was very similar in the wild-type and mutant enzymes. To determine the midpoint potential of the [3Fe-4S] cluster of C38A, F36/pC38A membranes were oxidized, reduced, and reoxidized (Fig. 6a). Upon reoxidation of the membranes, most of the [3Fe-4S] clusters were destroyed. Data obtained from the titration were fitted to a two-component model of the Nernst equation with E m,7 ϭ 140 (69%) and 75 mV (31%). The F36/pC38S redox titration data were fitted to one component, but the cluster routinely exhibited hysteresis (Fig. 6b). The E m,7 in the oxidizing direction was 190 mV, and the E m,7 in the reducing direction was 165 mV. The ratio of reduced [Fe-S] clusters to that of oxidized [Fe-S] clusters was determined from double integrations of reduced and oxidized samples. In the oxidized spectrum of membranes containing wild-type DmsABC, there is only a small amount of [3Fe-4S] cluster visible, so the ratio of reduced to oxidized [Fe-S] clusters is very large. The C38A and C38S mutants contain approximately four reduced clusters for each oxidized cluster, giving a total of five clusters.
Mutagenesis of the DmsA Cys Group to a Consensus Ferredoxin Cys Group-Sequences known to ligate [4Fe-4S] clusters usually contain four Cys residues (26,27). The first and second Cys residues are separated by two amino acids, an exception being M. thermolithotrophicus ferredoxin, which has four intervening residues (25). We altered the sequence of the DmsA

[Fe-S] Cluster in DmsA Mutants
Cys group so that the first two Cys residues would only be separated by two amino acids. The plasmid, pC38S, was further mutated to produce pC38S,N37C, which contains the sequence, CTVCSGSNC. This spacing of Cys residues occurs in Cys group II of DmsB and other electron transfer subunits of enzymes belonging to this family (1) and in Azotobacter vinelandii ferredoxin I (45,46). Expression and specific activity of the double mutant, C38S,N37C, are similar to that of C38S, and this enzyme is also unable to support growth on Me 2 SO in DSS301 (Table I). Fig. 7 shows spectra of the ferricyanide-oxidized membranes from the DmsA mutants. The double mutant ligated a [3Fe-4S] cluster, but the amount of cluster was reduced to approximately 25% of the amount of cluster ligated by C38S, estimated by double integration. The line shape is similar to C38S, but the signal is broader. Spectra of the mutants in whole cells are identical to that of the membrane preparations, indicating that the clusters in all three mutant enzymes are [3Fe-4S] clusters in vivo and are not [4Fe-4S] clusters altered upon oxidation during cell breakage (data not shown). The signal intensity of the double mutant appeared larger in whole cells than in the membrane samples. Redox titrations of the double mutant identify four [4Fe-4S] and one [3Fe-4S] cluster (Table II) with the ratio of reduced to oxidized [Fe-S] clusters being approximately 13:1. The high ratio is likely due to the reduced amount of [3Fe-4S] cluster present in these membrane preparations. DISCUSSION Wild-type DmsABC has a complex EPR spectrum (Fig. 2b) that has been analyzed as two pairs of interacting [4Fe-4S] clusters (1,4). Our research has been aimed at identifying which residues ligate the [Fe-S] clusters in DmsABC through the use of site-directed mutagenesis and EPR. Cys groups III (31) and I 2 of DmsB each ligate a [4Fe-4S] cluster. In this report, the Cys region of DmsA was mutated so that it is unlikely to bind a [4Fe-4S] cluster, but the EPR spectra of reduced membranes containing DmsA mutant enzymes was essentially identical to the wild-type enzyme (Fig. 2). All four previously characterized [4Fe-4S] clusters are present in the mutant enzymes in the correct amounts and with midpoint potentials similar to those of the wild-type enzyme (Table II). We conclude that the four EPR visible [4Fe-4S] clusters are all ligated by the Cys groups of DmsB.
A major new signal is observed in spectra of the oxidized a The ratio of reduced to oxidized [Fe-S] clusters was calculated from double integration of reduced and oxidized spectra from the redox titrations.
b Data from the redox titrations were fitted to a two-component model.
c Data from the redox titrations were fitted to two independent onecomponent models. (c) were oxidized with ferricyanide as described in Fig. 3. Instrument parameters were as described for Fig. 2.

[Fe-S] Cluster in DmsA Mutants
DmsA mutants with a peak at g ϭ 2.03 and a peak/trough at g ϭ 2.00 (Fig. 3). The line shapes of the new signals are distinct from the spectrum of fumarate reductase center FR3 (40,41 (31,47,48). The conversion of a [3Fe-4S] to a [4Fe-4S] cluster was generated in fumarate reductase (49).
It appears that the [3Fe-4S] cluster in DmsA mutants is not formed via cluster conversion but that mutagenesis of Cys-38 has altered the protein environment such that a cluster can assemble. This is supported by the lack of evidence to suggest that a fifth [4Fe-4S] cluster exists in wild-type DmsA. Reduction of the enzyme by dithionite gives only four clusters, and increasing the reduction potential of dithionite by increasing the pH did not reduce any additional clusters. Photoreduction with proflavin and EDTA did not reduce any additional centers (data not shown). No changes in the reduced EPR spectrum to indicate loss of a [4Fe-4S] cluster were visible in the Cys-38 mutant enzymes. Double integrations of the reduced and oxidized spectra of DmsA [3Fe-4S]-containing mutants gave ratios of four to one, indicating that these mutants gained an [Fe-S] cluster compared with the three to one ratio from DmsB C102 mutants, which altered one of the existing [Fe-S] clusters (31). Extensive EPR characterization of NarGHI (also a Type II enzyme) has identified four [Fe-S] clusters ligated by four Cys groups in the electron transfer subunit (NarH), but no fifth cluster was ligated by NarG (50 -52).
The spacing of the Cys residues in DmsA was altered so the first and second Cys residues were separated by only two amino acids, but the double mutant still ligated a [3Fe-4S] cluster with a line shape similar to that of the C38S cluster. The amount of this cluster was reduced, and multiple components were present in the analysis. [3Fe-4S] clusters can be generated from [4Fe-4S] clusters that are damaged by oxidation, but EPR analysis of whole cells expressing the double mutant demonstrated that the enzyme ligated a [3Fe-4S] cluster in vivo.
The natural function of the DmsA Cys group is unknown, but it could be involved in binding some factor, perhaps a metal ion, the loss of which may destroy function in the Cys-38 and Cys-42 mutant enzymes. The role of Cys-38 in DmsA is unique. Substitution of Ser or Ala for Cys should cause little perturbation of the protein structure, indicating that the sulfhydryl of the Cys-38 is important. In the Type I enzymes, there is an abundance of conserved Gly residues in this region that are not conserved in the Type II enzymes (Fig. 1). A cluster may be sterically hindered from assembling in this region until mutation of Cys-38 disrupts normal function and frees the Cys group to ligate an [3Fe-4S] cluster. Another possible reason for the loss of function in C38S is the presence of the [3Fe-4S] cluster. The C38S enzyme has a block in the electron pathway between the [4Fe-4S] clusters and Mo-MGD, where the substrate is reduced (32). The high E m,7 of the [3Fe-4S] cluster in C38S relative to the potentials of the Mo(VI)/(V) and Mo(V)/(IV) couples (Ϫ75 and Ϫ90 mV, (4)) suggests that the [3Fe-4S] cluster may act as an "insulating cluster" (28,53) between the [4Fe-4S] clusters and the Mo-MGD to decrease the rate of electron transfer.
We have divided the enzymes into three classes. The Type I enzymes such as NapAB, FdhF, and perhaps the other members of this type are likely to have a [4Fe-4S] cluster located in their amino terminus. The Type II enzymes, DmsABC, and the two E. coli nitrate reductases have a Cys region, but they are not likely to ligate an [Fe-S]. The Type III enzymes lack this region altogether, and neither TorA or BisC have been suggested to contain [Fe-S] clusters. This region in DmsA is likely a degenerate Cys group that has lost [Fe-S] binding capability upon evolution of the enzyme, although in DmsA the Cys group retains an essential role in electron transfer, perhaps interacting with the Mo-MGD.