Multiple pathways of electron transfer in dimethyl sulfoxide reductase of Escherichia coli.

The catalytic subunit of dimethyl sulfoxide (Me2SO) reductase, DmsA, contains six blocks of sequence that are homologous to other members of the superfamily of prokaryotic molybdoenzymes. The amino-terminal block contains 5 conserved residues (Cys38, Cys42, Cys75, Lys28, and Arg77). Site-directed mutagenesis of these residues did not alter membrane localization but in some cases less enzyme accumulated. The activity of Me2SO reductase was monitored by measuring Me2SO-dependent anaerobic growth, benzyl viologen, or dimethylnaphthoquinol oxidase activity, and using a quinol pool-coupling assay. Only Cys75 and Lys28 mutant enzymes were able to support anaerobic growth with Me2SO suggesting a critical role for Cys38, Cys42, and Arg77. Benzyl viologen oxidase activity was retained in the mutants although with reduced efficiency in Cys42-Ser. Electron transport with dimethylnaphthoquinol was reduced in Cys38-Ser, Cys42-Ser, and Cys75-Ser and almost totally eliminated in the Arg77-Ser mutant. Cys38-Ser, Cys42-Ser, and Arg77-Ser were unable to support quinol oxidation although electron transfer from the quinol pool to the [Fe-S] centers in DmsB was normal. These results indicate that the amino-terminal region is involved in functional electron transfer from the quinol pool to Me2SO and that electrons from benzyl viologen, dimethylnaphthoquinol, and menaquinol may follow different paths within the catalytic subunit.

Dimethyl sulfoxide (Me2SOI1 reductase, DmsABC, is a terminal electron transfer enzyme which allows Escherichia coli to grow anaerobically on Me2S0 as respiratory oxidant (Weiner et al., 1992). The enzyme is a complex iron-sulfur and molybdenum containing protein located on the cytoplasmic surface of the inner membrane (Sambasivarao et al., 1990). It consists of a molybdopterin cofactor containing subunit (DmsA, 87.4 kDa), an electron-transfer subunit with four [4Fe-4S] clusters (DmsB, 23.1 kDa) and a hydrophobic, membrane spanning anchor subunit (DmsC, 30.8 kDa). The enzyme accepts electrons from the menaquinol pool at the level of DmsC and passes them via DmsB to DmsA where the catalytic site is located (Weiner et al., 1992). The operon encoding the enzyme has been cloned *This work was funded in part by Medical Research Council of Canada Grant PG-11440. The costs of publication of this article were therefore be hereby marked "aduertisement" in accordance with 18 defrayed in part by the payment of page charges. This article must U.S.C. Section 1734 solely to indicate this fact.
Recently, we and others have found that DmsA is part of a superfamily of oxidoreductase enzymes with highly conserved organization and sequence Blasco et al., 1989;Wootton et al., 1991;Weiner et al., 1992). At present nine members of the family have been identified (Fig. 1). They are prokaryotic enzymes which reduce Me2S0, nitrate (Blasco et al., 1989;Blasco et al., 19901, biotin sulfoxide (Pierson and Campbell, 19901, and polysulfide (Krafft et al., 1992) or enzymes which oxidize formate (Berg et al., 1991;Bokranz et al., 1991;Sauter et al., 1992;Zinoni et al., 1986;White and Ferry, 1992;Schuber et al., 1986). Each contains a large catalytic subunit with a noncovalently bound molybdopterin cofactor. Eight of these enzymes are membrane-bound and contain a subunit with multiple [Fe-SI clusters and a membrane anchor subunit. An alignment of the catalytic subunits shows extensive blocks of homology throughout the length of the polypeptide but the amino-terminal region ( Fig. 1) is the most highly conserved. Three cysteine residues and 2 basic amino acids, Cys3', C Y S~~, C Y S~~, L Y S~~, and Arg77 of DmsA, are conserved in eight of the nine sequences. Only biotin sulfoxide reductase lacks this region. As biotin sulfoxide reductase is the only member of this group which does not accept electrons from an ironsulfur subunit (Pierson and Campbell, 1990) this suggested to us that these Cys and basic amino acids might participate in electron transfer from the iron-sulfur containing subunit to the catalytic subunit.
The activity of Me2S0 reductase can be monitored by: 1) measuring Me2S0 dependent anaerobic growth (Bilous and Weiner, 1985); 2) an artificial spectrophotometric assay using reduced benzyl viologen, BV.', as an electron donor ; 3) a more physiological spectrophotometric assay using the quinol analogue 2,3-dimethyl-1,4-napthoquinol, DMNH2, as the electron donor (Sambasivarao and Weiner, 1991a); and 4) the quinol-pool coupling electron paramagnetic resonance (EPR) assay (Rothery and Weiner, 1991). In this report, we have used site-directed mutagenesis to investigate the roles of residues Lys2ss C Y S~~, C Y S~~, C Y S~~, and Arg77 in Me2S0 reductase and to examine the growth, expression, localization, and electron transfer properties of the mutant enzymes.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The E. coli strains and plasmids used in this study are listed in Table I.
Materials-Deoxyoligonucleotides were synthesized on an Applied Biosystems 392 DNA Synthesizer in the DNA Core facility in the Dept. of Biochemistry, University of Alberta. DMN was a lund gift of Dr. A. Kroger, J. W. Goethe University, Frankfurt, Germany. Restriction endonucleases and modifying enzymes were from Life Technologies Inc. Sequenase and Sequenase reaction kits, Version 2.0, were from U. S. Biochemical Corp. The in uitro mutagenesis system was from Amer-  sham. Protein assay standard and low molecular weight polyacrylamide subcloned into the wild-type dms operon on a 1-kilobase Hind111 fraggel electrophoresis standards were obtained from Bio-Rad. All other ment to generate the mutant dmsA gene. The resultant plasmids were materials were reagent-grade and were obtained from commercial confirmed by restriction mapping and DNA sequencing. sources.
Preparation of Membrane Fractions-Membrane fractions were prepared by the method ofYamato et al. (1975) for use in expression studies and enzyme assays. This method gives a 2-fold enrichment of M e 8 0 specific activity over washed membranes.
For EPR analysis washed membrane fractions were prepared by French Pressure lysis and differential centrifugation as described by Cammack and Weiner (1990). Membranes were stored at -70 "C prior to use.
Enzyme Assays-Me2S0 reductase activity was determined by monitoring the Me,SO-or TMAO-dependent oxidation of reduced benzyl viologen (BV.+) as described by Cammack and Weiner (1990). The reduced DMN (DMNH,) and Me2SOor TMAO-dependent oxidation of dithionite was assayed a s described by Sambasivarao and Weiner (1991a). One unit of activity corresponds to 1 pmol of BV.' or dithionite oxidized min".
Protein Determination and Polyacrylamide Gel Electrophoresis-Protein concentrations were estimated by a modification of the Lowry procedure (Markwell et al., 1978) using a Bio-Rad protein standard. Polyacrylamide gel electrophoresis of 12.5% acrylamide gels was carried out using the Bio-Rad mini-gel system and a discontinuous SDS buffer system (Laemmli, 1970). Gels were stained with Coomassie Blue, destained, and the relative amount of DmsA protein was determined using a Joyce-Loebel Chromoscan 3 densitometer.
Electron Paramagnetic Spectroscopy-Samples were prepared as described by Cammack and Weiner (1990) from washed membranes of E.
coli DSS3Ol cells grown on glyceroVfumarate medium. Sample protein concentrations were 30 mg ml-l. Dithionite (5 mM) reduced samples were incubated under argon a t 23 "C for 2 min. 25 mM Me,SO or fumarate was added and the samples incubated for a further 2 min. Airoxidized samples were vigorously stirred with a coiled stainless steel wire. Spectra were recorded using a Bruker ESP300 EPR spectrometer equipped with an Oxford Instruments ESR-900 flowing helium cryostat under the following conditions: temperature, 10 K microwave power, 20 m W microwave frequency, 9.45 GHz; modulation amplitude, 10 Gpp; gain, 2 x lo4 (3.2 x lo4 for DSS301/pBR322).

RESULTS
Preparation of Mutants-Five conserved residues were selected for analysis. Mixed mutagenic oligonucleotides (Table 11) were used to replace C Y S~~, C Y S~~, C Y S~~, LysZ8, and Arg77 with serine or either alanine or glycine. Serine and alanine were chosen as replacements as they are unlikely to cause structural defects (Bordo and Argos, 1991). Glycine substitutions were obtained as a result of using mixed oligonucleotides.
Expression and Localization of Mutant Enzymes-In membrane preparations from E. coli HBlOl/pDMS159 (drnsABC) DmsA was recognized as a major 87-kDa polypeptide . DmsB and DmsC could also be identified in the membranes but these subunits were less obvious due to comigrating polypeptides and the diffuse staining of DmsC (Bilous . To compare the expression/ accumulation of DmsA from the mutant plasmids in HBlOl we used densitometric scanning of Coomassie Blue-stained SDSpolyacrylamide gel electrophoresis gels (Table 111). The amount of DmsA accumulated varied widely. In general, the serine substitutions accumulated DmsAto levels near the wild-type. With Percentage of DmsA relative to total protein was determined by densitometry scans of polyacrylamide gels. 45 pg of membrane protein was loaded per lane.
ND, not determined.
the exception of HB101/pK28A, the other alanine and glycine mutant plasmids accumulated lesser amounts of DmsA in the membrane. HB101/pC75A did not accumulate detectable DmsA. pC38S, pC42S, pC75S, pK28S, and pR77S were selected for detailed study as they presented the most consistent pattern of expression. In all but one case approximately 70% of the total MezSO reductase activity was localized to the cytoplasmic membrane with the remaining 30% found in the cytoplasm. This is similar to the distribution seen with wild-type enzyme activity (Sambasivarao et al., 1990). The exception was HB101/pR77G, which accumulated about 70% of the activity in the cytoplasmic fraction. Immunoblotting (not shown) confirmed a larger amount of DmsAB in the cytoplasmic fraction of HB101/pR77G than in HB101/pDMS160 suggesting that perturbation in the structure of DmsA had occurred altering membrane localization of the catalytic dimer.
We examined expression of the serine mutants in E. coli DSS301 which carries a total deletion of the dms operon (Sambasivarao and Weiner, 1991a). The expression levels of the pC38S, pC42S, and pR77S subunits correspond to only half the protein seen in HB101. DSS301/pK28S accumulated DmsA to levels comparable to the wild-type enzyme. The distribution of enzyme between membrane and cytoplasm was similar in DSS301 and HB101.
Growth Characteristics of E. coli DSSSOl Harboring Wildtype and Mutant Reductases-DSS301 is unable to grow anaerobically on Me2S0 because Me2S0 reductase is the only enzyme able to couple MezSO reduction to energy conservation (Sambasivarao and Weiner, 1991b). The deletion mutant can be complemented by the entire dms operon on a recombinant plasmid (Sambasivarao and Weiner, 1991a) and this provided a measure of the function of the mutant enzymes. DSS301 carrying the mutant plasmids were grown anaerobically on glycerol/MezSO medium producing the doubling times shown in Table IV. Only pDMS160, pK28A, pK28S, and pC75S were able to support growth of DSS301, although cells carrying pC75S grew much slower.
We examined growth of HBlOl harboring these dms plasmids (Table IV). Although HBlOl has a chromosomal copy of MezSO reductase, the mutants that did not support growth in DSS301 inhibited the growth of HB101. The overexpressed mutant subunits compete with the chromosomally encoded wild-type DmsA for assembly into the holoenzyme resulting in this inhibition. Such effects have previously been reported in studies of MezSO reductase (Rothery and Weiner, 1991) and fumarate reductase (Weiner et al., 1986;Westenberg et al., 1990).
Electron Dansfer from the Artificial Electron Donor Benzyl Viologen-Membrane fractions from HBlOl harboring the dms plasmids were assayed for MezSO reductase activity using BV.' as electron donor and Me2S0 or TMAO as oxidant ( Table  V). All of the mutant enzymes displayed a specific activity higher than HBlOVpBR322. HB101/pC42S had the lowest specific activity (35 units mg") which was only slightly enhanced over the specific activity of HBlOUpBR322 (29 units mg"). The HB101/pC42S activity may represent a combination of chromosomal dms expression and excess pC42S expression but the results suggest that this residue plays a role in BV.' oxidase NG, no growth on glycerol/Me,SO minimal medium. I, Doubling times were determined from measurements of cell densities in glycerol/Me,SO minimal medium using a Nett-Summerson spectrophotometer equiped with a number 66 filter. activity. When the activities were normalized to the amount of DmsA in the membrane (Table V), HB101/pC42S had the lowest activity again suggesting that this mutation had decreased the efficiency of MezSO reductase activity.
All of the plasmids expressed BV.+-dependent MezSO reductase activity in DSS301 (Sambasivarao and Weiner, 1991a). The specific activity of DSS301/pDMS160 was 79 units mg" (TMAO as oxidant), whereas DSS301/pC42S displayed the lowest specific activity (14 units mg-l). This activity must come from the mutant plasmid as DSS301 has a specific activity of 0.2 units mg" with TMAO. As in HB101, the other mutant plasmids demonstrated intermediate activities.
The ratio of TMAO to MezSO activity was generally between 7 and 9, in agreement with the reported activity ratio (Sambasivarao and Weiner, 1991a). Two of the mutants had substrate activity ratios outside this range (HBlOUpK28S had a ratio of 10.5 and HB101/pC42S had a ratio of 5) but it is unlikely that these variations represent a difference in substrate utilization. Together these results suggest that of these 5 conserved residues only Cys42 may play a role in BV.+ oxidase activity.
Electron D-ansfer from the Quinol Analogue Dimethylnaphthoquinol-DMNHz is a quinol analog previously used to analyze electron transfer within fumarate reductase (Weiner et al., 1986) and MezSO reductase (Sambasivarao and Weiner, 1991a). It is reactive only with the holoenzyme forms of these reductases, in contrast to benzyl viologen which will transfer electrons to either the catalytic dimer or holoenzyme. DMNH2dependent oxidation of TMAO and MezSO by HBlOl membranes is shown in Table V. All mutants, except A~-g~~-S e r , display specific activities greater or equal to HBlOUpBR322. HBlOUpC38S and HB101/pC42S have low levels of activity and it is unclear if this activity results from chromosomal or plasmid expression. HB101/pR77S has very low activity suggesting that this residue is essential for DMNH2 oxidase activity.
All mutant enzymes possessed DMNHz activity in DSS3Ol where the endogenous DMNHz:TMAO activity is only 0.025 units mg-l. DSS301/pC38S and DSS301/pC42S had specific activities of 0.3 and 0.2 units mg", respectively, suggesting that these residues are not essential for DMNHz oxidase activity but play a role in the efficiency of the reaction. Similarly, DSS301/pR77S had a very low specific activity of 0.05 units mg" suggesting that this residue, although not essential, has a major effect on DMNHz oxidase activity. Interestingly, the ratio of TMAO to MezSO activity with DMNHz as electron donor was near 2 compared to 5-10 for BV.' suggesting that the mechanisms of the benzyl viologen and DMN reactions differ.  ' BVTMAO activities were normalized for the amount of DmsAexpression determined by densitometry. One unit corresponds to 1 pmol of BV.+ Membranes were prepared from DSS301 cells grown on glycerollfumarate medium and the ability of Me,SO reductase to oxidize the menaquinol pool was examined.

Ability of Wild-type and Mutant Enzymes to Oxidize the Menaquinol Pool-
The membrane menaquinol pool is in equilibrium with several terminal reductases including fumarate and Me2S0 reductases. Functional MezSO reductase is able to draw electrons from fumarate reductase through the menaquinol pool to reduce MezSO and the reverse also is true. We can follow this reaction using EPR spectroscopy of the endogenous [Fe-SI clusters in these proteins (Rothery and Weiner, 1991). DSS301 membranes contain fumarate reductase which can be used to determine if plasmid encoded MezSO reductase is able to catalyze MezSO-dependent menaquinol oxidation. We have used this quinol-pool coupling (&-pool coupling) assay to determine whether mutants of Me2S0 reductase are able to accept electrons from the endogenous quinol pool (Table V). Fig.  2A shows the spectra obtained from E. coli DSS301/pBR322 membranes. When reduced with dithionite the prominent fea- tures are a peak at g = 2.03 and a peak trough at g = 1.94, which are characteristic of the reduced FR1 t2Fe-2SI center of fumarate reductase (Johnson et al., 1988). Addition of 25 mM MezSO or TMAO (data not shown) does not change the spectrum of these membranes but the addition of 25 mM fumarate causes the FR1 features to diminish and a sharp peak at g = 2.02 with a broad trough immediately upfield to appear. These new features are characteristic of the oxidized [3Fe-4S] center of fumarate reductase, FR3 (Johnson et al., 1988). In Fig. 2B the spectra of membranes from DSS301/pDMS160 are shown. In the dithionite-reduced sample the fumarate reductase signal is still present but a new peak at g = 1.99 and a trough at g = 1.88 are also visible. These features are part of the spectrum of reduced MezSO reductase (Cammack and Weiner, 1990 that MezSO reductase is able to interact with and accept electrons through the quinol pool and pass them on to the substrate. Fig. 3A shows EPR spectra of the DSS301/pK28S mutant membranes which have been reduced with dithionite. The dithionite-reduced spectrum is a composite of fumarate reductase and MezSO reductase which can be oxidized by either fumarate or MezSO. The K28S mutant enzyme is coupled t o the quinol pool as is expected from the wild-type growth properties of this enzyme. The C y~~~-S e r enzyme also enables DSS3Ol to grow on MezSO medium and it too shows coupling to the quinol pool. The C y~~~-S e r , C y~~~-S e r , and A~-g~~-Ser mutants were unable to support growth on MezSO and all have EPR spectra similar to that of DSS301/pR77S shown in Fig. 3B. There is no oxidation of the reduced spectra of either enzyme by the addition of MezSO. These mutant enzymes are not able to consume electrons by reducing MezSO to Me& Addition of fumarate causes both enzymes to be oxidized so the mutant MezSO reductases are still able to interact with the quinol pool and pass electrons to the level of the [Fe-SI clusters in DmsB, but they are not capable of completing the electron transfer to MezSO in the catalytic site of DmsA and are therefore unable to grow. DISCUSSION A newly identified superfamily of membrane-bound prokaryotic dehydrogenases and reductases includes E. coli MezSO reductase (DmsABC, Bilous et al. (1988)) nitrate reductases (NarGHJI, Blasco et al. (1989); NarZYWV, Blasco et al. (1990)), Wollinella succinogenes polysulfide reductase (PsrABC, Krafft et al. (199211, and the formate dehydrogenases from E. coli (FDNN-FdnGHI, Berg et al. (1991); FDNH-FdhF, Zinoni et al. (1986); and HycCD, Sauter et a l . (1992)), Methanobacterium formicicum (FdhAB, Schuber et al. (1986) and FdhC, White and Ferry (1992)) and W. succinogenes (FdhABCD, Bokranz et al. (1991)). Members typically contain three subunits: a large molybdopterin containing catalytic subunit (e.g. DmsA, NarG), a cysteine-rich electron transfer subunit ligating [Fe-SI clusters (e.g. DmsB, NarH), and a membrane anchor subunit (e.g. DmsC, NarI). Multiple amino acid sequence alignment clearly shows the relationship of the catalytic and electron transfer subunits while the anchor subunit appears to be far less conserved. The catalytic subunits contain several blocks of sequence homology which are present in all members of the fam-ily and presumably some of these regions participate in complexing the molybdopterin cofactor. The order of these blocks from amino to carboxyl terminus is constant but their relative spacing varies greatly (Weiner et al., 1992). Interestingly, biotin sulfoxide reductase (BisC, Pierson and Campbell (1990)), which receives its reducing equivalents from a thioredoxin-like soluble protein, lacks the amino-terminal block (Fig.  1). The other eight enzymes require the Cys-rich electron transfer subunit and this suggested to us that the first block may be necessary for electron transfer to the catalytic subunit. The results presented herein confirm this hypothesis.
The eight homologous proteins contain 3 Cys residues equivalent to C Y S~~, C Y S~~, and Cys75 of DmsA (Fig. 1). DmsA contains an additional conserved Cys residue at position 34 but this amino acid is replaced by a His in nitrate reductase. Changing Cys34 of DmsA to His does not appear to affect the properties of MezSO reductase.2 The proteins also contain conserved basic residues corresponding to LysZ8, Arg77, and Arggo of DmsA. Mutagenesis of k g s 0 t o Ser3 did not alter any measurable activity of MezSO reductase. Additional conserved residues including Gly78 and Pros4 were not mutated due to their potential structural role. The roles of LysZ8, C Y S~~, C Y S~~, C Y and Arg77 in MezSO reductase were examined by site-directed mutagenesis.
LysZ6-Mutation to the neutral amino acids Ser or Ala did not alter the growth, expression, or catalytic activities of MezSO reductase indicating that this conserved residue is not important for the activities we measured.
Cy~~~-Mutation to Ser produced an enzyme which could not support growth on MezSO although the membrane localization was wild-type and accumulation of mutant enzyme was not greatly impeded. The enzyme catalyzed BV.' oxidase activity at -50% of wild-type and had low DMNHz oxidase activity. It could not accept electrons from the menaquinol pool.
C y~~~--C y s~~-S e r could not support growth on MezSO. Again, the membrane localization was wild-type and accumulation of this enzyme was not greatly reduced. This enzyme had the poorest BV.' oxidase activity and low DMNHz activity suggesting that Cys42 was necessary for optimal activity. Cys4'-Ser could not accept electrons from the menaquinol pool. Cy~~~-Mutation to Ser produced a n enzyme which supported growth on MezSO. The membrane localization was normal. It displayed relatively high benzyl viologen and DMNHz activity and could transfer electrons from the menaquinol pool.
A~g~~-T h i s residue was essential for growth and both DMNHz and menaquinol oxidase activities. The A~-g~~-S e r mutant had near wild-type benzyl viologen activity, localization, and accumulation.
We have combined these results into a working model which is shown in Fig. 4. In this model the BV.' oxidase activity requires both the DmsA and B subunits, but not DmsC. It is not clear if any of the four [4Fe-4Sl centers are needed for this reaction but our preliminary data suggests that at least the clusters ligated primarily by Cys groups I and I11 of DmsB are not n e~e s s a r y .~ Of the conserved residues studied, only in DmsA plays a role in electron transfer from BV.'.
The DMNHz activity has been assumed to accurately reflect the physiological function of MezSO reductase as it required all three subunits Weiner, 1991a, 1991b). However, the data presented here indicates that it is possible to isolate mutants with reasonable levels of DMNHz activity, i.e. C y~~~-S e r and C y~~~-S e r which are not functional and unable to transfer electrons from the menaquinol pool. Interestingly, in HB101, these mutants had activity levels in excess of pBR322 which supported growth on MezSO. Thus, there is no correlation between DMNHz activity and growth. This suggests that electrons from DMNHz follow a different path to the active site. Functional electron transfer from the menaquinol pool requires Cys3', Cys4', and Arg77 but not LysZ8, C Y S~~, or ArggO. In all cases the enzyme assembled functional [4Fe-4S] clusters which could be reduced by dithionite and which could pass their electrons back to the menaquinol pool, but in C Y S~~, C Y S~~, and Arg77 the enzyme was unable to pass these electrons on through DmsA.