Interactions between the Molybdenum Cofactor and Iron-Sulfur Clusters of Escherichia coli Dimethylsulfoxide Reductase*

We have used site-directed mutagenesis to study the interactions between the molybdo-bis(molybdopterin guanine dinucleotide) cofactor (Mo-bisMGD) and the other prosthetic groups ofEscherichia coli Me2SO reductase (DmsABC). In redox-poised preparations, there is a significant spin-spin interaction between the reduced E m,7 = −120 mV [4Fe-4S] cluster of DmsB and the Mo(V) of the Mo-bisMGD of DmsA. This interaction is significantly modified in a DmsA-C38S mutant that contains a [3Fe-4S] cluster in DmsA, suggesting that the [3Fe-4S] cluster is in close juxtaposition to the vector connecting the Mo(V) and the E m,7 = −120 mV cluster of DmsB. In a DmsA-R77S mutant, the interaction is eliminated, indicating the importance of this residue in defining the interaction pathway. In ferricyanide-oxidized glycerol-inhibited DmsAC38SBC, there is no detectable interaction between the oxidized [3Fe-4S] cluster and the Mo-bisMGD, except for a minor broadening of the Mo(V) spectrum. In a double mutant, DmsAS176ABC102SC, which contains an engineered [3Fe-4S] cluster in DmsB, no significant paramagnetic interaction is detected between the oxidized [3Fe-4S] cluster and the Mo(V). These results have important implications for (i) understanding the magnetic interactions between the Mo(V) and other paramagnetic centers and (ii) delineating the electron transfer pathway from the [4Fe-4S] clusters of DmsB to the Mo-bisMGD of DmsA.

sequence comparisons (see below), the cofactor is almost certainly present in the form of Mo-bisMGD in the intact holoenzyme. DmsABC is a member of an emerging family of bacterial complex [Fe-S]-molybdoenzymes that includes the E. coli formate dehydrogenases (FdnGHI and FdoGHI) (10,11), E. coli nitrate reductases (NarGHI and NarZYV) (12,13), Salmonella typhimurium thiosulfate reductase (PhsABC) (14), and Wolinella succinogenes polysulfide reductase (PsrABC) (15). Each member of this family has a molybdenum cofactor-containing catalytic subunit (assumed to be Mo-bisMGD (16,17)), a four-[Fe-S] cluster-containing electron transfer subunit, and a hydrophobic membrane anchor subunit. The various prosthetic groups of these enzymes catalyze overall reactions that involve the transfer of two electrons through the electron transfer subunit to or from a quinol-binding site associated with the membrane anchor subunit (6,18). The use of EPR spectroscopy to study the interactions between the prosthetic groups in wild-type and appropriately mutagenized enzymes may yield important information on the electron transfer pathway through the [Fe-S] clusters of the electron transfer subunit to the Mo-bisMGD cofactor of the catalytic subunit.
In DmsABC, two of the [4Fe-4S] clusters of DmsB appear to be thermodynamically competent to transfer electrons from menaquinol (MQH 2 , reduced midpoint potential at pH 7 E m,7 ϭ Ϫ70 mV) to the Mo-bisMGD of DmsA (E m,7 values as follows: Mo(IV/V) ϭ Ϫ175 mV; Mo(V/VI) ϭ Ϫ15 mV (19)), and these have E m,7 values of Ϫ50 and Ϫ120 mV, respectively (2,6). It has been demonstrated that there is a significant conformational link between the E m,7 ϭ Ϫ50 mV [4Fe-4S] cluster and a single menaquinol (MQH 2 ) binding site associated with His-65 of DmsC (6,20). This was established by using the MQH 2 analog 2-n-heptyl-4-hydroxyquinoline-N-oxide to elicit an EPR line shape change on a genetically engineered [3Fe-4S] cluster in a DmsB-C102S mutant (4,6). In this mutant, the E m,7 ϭ Ϫ50 mV cluster is replaced by a high potential [3Fe-4S] cluster, and therefore a role for the E m,7 ϭ Ϫ50 mV [4Fe-4S] cluster of the wild-type enzyme can be envisioned in electron transfer from MQH 2 . The other cluster that appears to be thermodynamically competent to participate in catalytic electron transfer is the E m,7 ϭ Ϫ120 mV cluster. It has been demonstrated that there is a strong spin-spin interaction between this cluster and the E m,7 ϭ Ϫ50 mV cluster of the wild-type enzyme, consistent with both of them being part of an eight-iron (2[4Fe-4S]) ferredoxin motif (2,6). This pair may provide a conduit for electron flow through DmsB to the Mo-bisMGD of DmsA. The two remaining low potential clusters of DmsB (E m,7 ϭ Ϫ240 and Ϫ330 mV) may play a role in defining the overall structure of this subunit in a manner similar to that suggested for the low potential [4Fe-4S] clusters of E. coli nitrate reductase A (21) and fumarate reductase (22).
Evidence exists to suggest a pathway of electron transfer from DmsB to the Mo-bisMGD of DmsA. The microwave power saturation properties of the Mo-bisMGD Mo(V) EPR spectrum are sensitive to the redox state of the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster of DmsB (2). DmsA also contains a vestigial [4Fe-4S] cluster binding motif close to its N terminus, which, when appropriately mutagenized, can be made to bind an engineered [3Fe-4S] cluster with an E m,7 of approximately 178 mV (23). This motif has been shown to be involved in physiological electron transfer from the [4Fe-4S] clusters of DmsB to the Mo-bisMGD cofactor of DmsA (24). This suggests that the vestigial [4Fe-4S] cluster binding motif may be on the electron transfer pathway from the E m,7 ϭ Ϫ120 mV cluster of DmsB to the Mo-bisMGD of DmsA. It would therefore be interesting to determine if the presence of an engineered DmsA [3Fe-4S] cluster has any effect on the observed interaction between the Mo-bisMGD and the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster of DmsB.
The structures of three proteins that have significant sequence similarity to DmsA have been solved. These are the Me 2 SO reductases from Rhodobacter sphaeroides and Rhodobacter capsulatus (16,25) and the formate dehydrogenase H (FdhF) from E. coli (26). In each case, the molybdenum cofactor is a Mo-bisMGD. Given the sequence similarity between these proteins and DmsA (35-41% similar, 25-32% identical), it is almost certain that the latter contains a Mo-bisMGD cofactor as well. The structure of FdhF is particularly relevant, since it contains a [4Fe-4S] cluster coordinated by its N-terminal Cys group. This cluster has been proposed to have an important role in the electron transfer pathway from formate to the artificial electron acceptor benzyl viologen (26). Significant differences in the sequence of the N-terminal Cys groups have been identified between the bacterial molybdoenzyme catalytic subunits that contain a [4Fe-4S] cluster and those that do not (23). EPR evidence exists to place E. coli FdhF and Paracoccus denitrificans (formerly Thiosphaera pantotropha) periplasmic nitrate reductase (NapA) in the former group (27,28), and the catalytic subunits of the E. coli respiratory Me 2 SO and nitrate reductases (DmsA and NarG) in the latter group (23,29).
In the structure of FdhF (26), it is significant that there is a Lys residue located between the [4Fe-4S] cluster and one of the pterins of the Mo-bisMGD. This residue is highly conserved in the group of enzymes that contain a [4Fe-4S] cluster coordinated by the N-terminal Cys group of the catalytic subunit (23). In the group that has the Cys group but no [4Fe-4S] cluster, this residue is replaced by an Arg (Arg-77 of DmsA and Arg-98 of NarG). When DmsA-R77 is mutagenized to Ser, electron transfer from the [4Fe-4S] clusters of DmsB to the Mo-bisMGD of DmsA is blocked (24). It would therefore be interesting to test the effect of the DmsA-R77S mutation on the Mo(V)-[4Fe-4S] cluster interaction.
In this paper, we describe the interactions observed by EPR that involve the Mo-bisMGD cofactor, the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster of DmsB, and engineered [3Fe-4S] clusters of appropriately mutagenized DmsA and DmsB. In the absence of detailed crystallographic data, these studies provide important information on the interactions between the prosthetic groups of DmsABC as well as on the electron transfer pathway from MQH 2 to Me 2 SO.

MATERIALS AND METHODS
Bacterial Strain and Plasmids-The E. coli strain and plasmids used in this study are listed in Table I. pDMS170 bears the wild-type dmsABC operon behind an fnr-dependent promoter and was generated by ligating the 4.8-kilobase pair EcoRI-SalI fragment from pDMS223 (4) into pBR322 that had previously been cut with PvuII and NruI and self-ligated to destroy these sites. pDMS160-S176A-C102S encoding DmsA S176A B C102S C was constructed by subcloning a 3.37-kilobase pair EcoRV-SalI fragment bearing the dmsB C102S mutation from pDMS160-C102S into pDMS160-S176A. All routine cloning work was carried out essentially as described by Sambrook et al. (30). All plasmids were transformed into E. coli HB101. pDMS170 was used to express the wild-type DmsABC used in this study and is functionally equivalent to the plasmid pDMS160 used in previous studies (4,19,24). pDMS160-C38S, pDMS160-R77S, and pDMS160-S176A-C102S were used to express DmsA C38S BC, DmsA R77S BC, and DmsA S176A B C102S C, respectively.
Growth of Cells and Preparation of Membrane Vesicles-Cells were grown anaerobically in 20-liter batch cultures at 37°C for 48 h on a glycerol-fumarate minimal medium (4,31). Cells were harvested and washed in 100 mM MOPS and 5 mM EDTA (pH 7.0). Membranes were prepared by French pressure cell lysis and differential centrifugation as described previously (4,6).
Preparation of EPR Samples-Membrane vesicles were suspended at a protein concentration of approximately 30 mg ml Ϫ1 in 100 mM MOPS and 5 mM EDTA (pH 7.0). Dithionite-reduced (5 mM) samples were incubated under argon at 23°C for 5 min. Oxidized samples were prepared by incubating membranes in the presence of 0.2 mM potassium ferricyanide for 2 min. Ferricyanide-oxidized, glycerol-inhibited (32, 33) samples were prepared as described in the legend to Fig. 5.
EPR Spectroscopy-Spectra were recorded using a Bruker ESP300 EPR spectrometer equipped with an Oxford Instruments ESR-900 flowing helium cryostat. Instrument conditions and temperatures were as described in the individual figure legends. Microwave power saturation data were fitted to the equation, where S is the signal height, K is a proportionality factor, P is the microwave power, P1 ⁄2 is the microwave power for half-saturation, and b is the inhomogeneity parameter (35,36).
This study a pDMS170 was generated by ligating the 4.8-kilobase pair EcoRI-Sal1 fragment from pDMS223 (4) into pBR322 that had previously been cut with PvuII and NruI and self-ligated to destroy these sites.

Mo(V) EPR Spectra of the Wild-type, DmsA-C38S, and
DmsA-R77S Enzymes- Fig. 1 shows Mo(V) EPR spectra recorded at 75 K of membranes containing the wild-type ( Fig.  1A), DmsA-C38S (Fig. 1B), and DmsA-R77S ( Fig. 1C) mutant forms of DmsABC. All three spectra appear to be essentially identical, with g values of approximately 1.984, 1.980, and 1.960 (g 1 , g 2 , and g 3 ), suggesting that neither the DmsA-C38S nor the DmsA-R77S mutation has any significant effect on the immediate coordination sphere of the molybdenum of the Mo-bisMGD cofactor. Fig. 2A shows Mo(V) potentiometric titrations of wild-type and DmsA-C38S mutant DmsABC. For the wild-type enzyme, plots of the intensity of the g ϭ 1.980 peak-trough versus E h can be fitted to two E m,7 values of Ϫ3 Ϯ 3 mV (Mo(V/VI)) and Ϫ148 Ϯ 14 mV (Mo(IV/V)) (three determinations), in reasonable agreement with previously reported values (19). In the case of the DmsA-C38S mutant, plots of the g ϭ 1.980 peak-trough versus E h can be fitted to two E m,7 values of Ϫ5 Ϯ 4 mV (Mo(V/VI)) and Ϫ128 Ϯ 13 mV (Mo(IV/V)) (three determinations), indicating that the mutation and/or the presence of the DmsA [3Fe-4S] cluster has a minor effect on the Mo(IV/V) couple (a ⌬E m,7 of approximately ϩ20 mV). Fig. 2B shows the effect of the DmsA-R77S mutation on the E m,7 values of the Mo-bisMGD cofactor. The plots can be fitted to two E m,7 values of Ϫ27 Ϯ 3 mV (Mo(V/VI)) and Ϫ149 Ϯ 10 mV (Mo(IV/V)) (three determinations). In this case, the mutation has a minor effect on the Mo(V/VI) couple (a ⌬E m,7 of approximately Ϫ24 mV). Given their lack of effect on the Mo(V) EPR line shape and their minor electrochemical effects, it is unlikely that the two DmsA mutations cause gross changes in the overall structure of DmsABC.

Effect of the DmsA-C38S and DmsA-R77S Mutations on the Mo(IV/V) and Mo(V/VI) E m,7 Values-
The Spin-Spin Interaction between the E m,7 ϭ Ϫ120 mV [4Fe-4S] Cluster and the Mo(V) of Wild-type DmsABC-Cammack and Weiner (2) originally identified an enhancement of the spin relaxation of the Mo(V) species of wild-type DmsABC that was consistent with it being due to an interaction between the Mo(V) and the reduced E m,7 ϭ Ϫ120 mV cluster. Fig. 3A shows microwave power saturation curves recorded at various redox potentials for the g ϭ 1.98 peak-trough of the Mo(V) spectrum at 30 K. As the potential is reduced, the data can be fitted to two components, a saturable noninteracting component (dominant at high E h ) and a nonsaturable component (dominant at low E h ). The fraction of noninteracting Mo(V) decreases with decreasing E h with an apparent n ϭ 1 E m,7 of Ϫ140 mV (Fig. 3B). Given that the fraction of noninteracting Mo(V) at each E h is obtained from log-log plots of the microwave power saturation data, the observed E h of Ϫ140 mV is in reasonable agreement with the published E m,7 of Ϫ120 mV for one of the [4Fe-4S] clusters of DmsB (2, 6). These data are consistent with (i) the interacting [4Fe-4S] cluster being the E m,7 ϭ Ϫ120 mV cluster and (ii) the E m,7 ϭ Ϫ50 mV cluster not contributing independently to this interaction.
Effect of the DmsA-C38S and DmsA-R77S Mutations on the Paramagnetic Interaction between the E m,7 ϭ Ϫ120 mV Cluster of DmsB and the Mo(V) of DmsA- Fig. 4 shows microwave power saturation curves for wild-type, DmsA-C38S, and DmsA-R77S DmsABC at potentials of approximately Ϫ20 and Ϫ150 mV. In the DmsA-C38S mutant, the microwave power saturation curve of the E h ϭ Ϫ20 mV Mo(V) signal is essentially identical to that of wild-type DmsABC, saturating with a single P1 ⁄2 of 1.2 mW. In contrast to the wild type, at E h ϭ Ϫ155 mV the Mo(V) signal saturates as an apparent single component with a P1 ⁄2 of 2.0 mW and an inhomogeneity parameter (b) of 0.8. When microwave power saturation curves are recorded using redox poised samples (between E h ϭ Ϫ20 mV and E h ϭ Ϫ180 mV), b decreases to a lower limit of approximately 0.8 as the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster becomes reduced (data not shown). Potentiometric studies indicate that the E m,7 values of the DmsB [4Fe-4S] clusters (including the E m,7 ϭ Ϫ120 mV cluster) remain unaltered in the DmsA-C38S mutant (23). Thus, the presence of the [3Fe-4S] cluster appears to significantly modulate the interaction between the E m,7 ϭ Ϫ120 mV cluster of DmsB and the Mo(V) of the Mo-bisMGD cofactor of DmsA.
In the DmsA-R77S mutant, the microwave power saturation curve of the E h ϭ Ϫ18 mV sample is again very similar to that of the wild-type enzyme (Fig. 4). In this case, the P1 ⁄2 is estimated to be 1.4 mW. However, at E h ϭ Ϫ146 mV, the saturation curve is essentially identical to that observed at Ϫ18 mV, indicating that the DmsA-R77S mutation eliminates the magnetic interaction between the Mo(V) and E m,7 ϭ Ϫ120 mV cluster of DmsB. EPR spectra recorded at 12 K of redox-poised samples indicate that there is no detectable shift of the midpoint potential of the E m,7 ϭ Ϫ120 mV cluster in the DmsA-R77S mutant (data not shown). When the potential dependence of the microwave power saturation curves are investigated, greater than 95% of the Mo(V) saturates with a low P1 ⁄2 , even at potentials as low as Ϫ200 mV (data not shown). Because of the disappearance of the Mo(V) signal at low E h , it was not possible to investigate the effect of reduction of the two lowest potential [4Fe-4S] clusters (the E m,7 ϭ Ϫ240 and Ϫ330 mV clusters) on the microwave power saturation properties or line shape of the

Mo(V) spectrum. Interaction between Glycerol-inhibited Mo(V) and the Oxidized DmsA-C38S [3Fe-4S] Cluster-Me 2 SO reductase from
Rh. sphaeroides can be prepared in a glycerol-inhibited form by reduction followed by reoxidation in the presence of high concentrations of glycerol (32,33). The Mo-bisMGD of this form of the enzyme remains in the Mo(V) state under oxidizing conditions. In order to investigate potential interactions between the molybdenum and the DmsA [3Fe-4S] cluster in the DmsA-C38S mutant, we generated membrane preparations containing glycerol-inhibited wild-type and mutant DmsABC. Fig. 5A shows EPR spectra recorded at 75 K of the glycerol-inhibited forms of both wild-type (Fig. 5A, i) and DmsA-C38S mutant (Fig. 5A, ii) DmsABC. Glycerol-inhibited enzyme (33) was prepared by oxidation of dithionite-reduced enzyme in the presence of 50% (v/v) glycerol. In both cases, the EPR spectrum has g values (g 1 , g 2 , and g 3 ) of 1.986, 1.978, and 1.959. Thus, the spectrum of the glycerol-inhibited form of Mo(V) is slightly more rhombic than the noninhibited form (cf. Fig. 1), allowing resolution of the g 1 peak from the g 2 peak-trough. Except for a slight broadening of the g 2 peak-trough in the glycerol-inhibited DmsA-C38S mutant (Fig. 5A, ii), there appears to be little effect elicited on the glycerol-inhibited Mo(V) spectrum by the presence of a [3Fe-4S] cluster in DmsA.
The estimated P1 ⁄2 values for the Mo(V) signals of the glycerol-inhibited wild-type and DmsA-C38S DmsABC are 0.6 and 0.9 mW, respectively (Fig. 5B). These results suggest that there is no significant enhancement of the spin relaxation of the glycerol-inhibited Mo(V) caused by the presence of an engineered [3Fe-4S] cluster in DmsA. The redox state of the DmsA [3Fe-4S] cluster was verified by recording spectra at 12 K, and under the conditions used to generate the glycerol-inhibited samples it was found to be in the oxidized S ϭ 1 ⁄2 state. Interestingly, in the case of the redox-poised data presented in Fig.  4, it is clear that the reduced S ϭ 2 state of the DmsA [3Fe-4S] cluster also does not significantly enhance the spin relaxation of the Mo(V) spectrum.

Interaction between the Mo(V) of a DmsA-S176A Mutant and an Engineered [3Fe-4S] Cluster in DmsB-
We have previously demonstrated that mutagenesis of the protein-molybdenum ligand of DmsA (Ser-176) generates a form of the Mo-bisMGD in which the molybdenum remains in the paramagnetic Mo(V) redox state at high E h (19). We have also generated mutants of DmsB in which the [4Fe-4S] cluster coordinated primarily by Cys group III is replaced by a high potential [3Fe-4S] cluster (4,6). By straightforward subcloning (see "Materials and Methods"), it was possible to generate a double mutant, DmsA S176A B C102S C, which at high E h contains Mo(V) and the oxidized DmsB [3Fe-4S] cluster. Fig. 6A shows EPR spectra recorded at 30 K and 0.2-mW microwave power of ferricyanide-oxidized membranes containing overexpressed DmsA S176A BC (Fig. 6A, i) and DmsA S176A -B C102S C (Fig. 6A, ii). In both cases, a Mo(V) signal was observed with g values of approximately 2.018, 1.982, and 1.961 (g 1 , g 2 , and g 3 ), as previously reported (19). The minor features in the g ϭ 2.00 -2.03 region result from the incomplete broadening of the DmsB-C102S [3Fe-4S] cluster spectrum at the temperature and microwave power used to record the spectra of Fig. 6A. Fig.  6B shows microwave power saturation profiles of the Mo(V) spectra of DmsA S176A BC and DmsA S176A B C102S C. In both cases, the estimated P1 ⁄2 is 0.8 mW, indicating that the presence of the oxidized S ϭ 1 ⁄2 [3Fe-4S] cluster of DmsB has no detectable effect on the microwave power saturation properties of the Mo(V) of DmsA. DISCUSSION We have demonstrated herein the effects of mutations of residues along the electron transfer pathway of DmsA on the EPR properties and redox chemistry of the Mo-bisMGD cofactor of DmsABC. Based on their effects on the Mo(V) EPR spectrum, it is clear that the DmsA-C38S and DmsA-R77S mutants have little effect on the environment or coordination sphere of the molybdenum and only minor effects on its redox chemistry. Given these results, it is unlikely that the mutations result in any gross modification of the protein structure that may result in significant differences in intercofactor distances between the wild-type and mutant proteins.

. Interaction between the glycerol-inhibited Mo(V) and an engineered [3Fe-4S] cluster in DmsA.
A, EPR spectra of glycerolinhibited, ferricyanide-oxidized DmsABC (i) and DmsA C38S BC (ii). The glycerol-inhibited form was generated as follows. Membranes at approximately 40 mg ml Ϫ1 protein were reduced with a small excess of dithionite for 2 min under argon at 23°C. Glycerol was then added to 50% (v/v), and following incubation for a further 2 min, the membranes were oxidized with an excess of ferricyanide (approximately 0.4 mM). EPR spectra were recorded at 75 K with a modulation amplitude of 3 G pp and a microwave frequency of 9.47 GHz. B, microwave power saturation curves of glycerol-inhibited Mo(V) signals at 30 K with a modulation amplitude of 6 G pp . Triangles, DmsA C38S BC (P1 ⁄2 ϭ 0.9 mW, b ϭ 1.5); squares, DmsABC (P1 ⁄2 ϭ 0.6 mW, b ϭ 1.6). bisMGD cofactor is corroborated by comparison of DmsA with the structurally characterized FdhF (26). In FdhF, the residue equivalent to DmsA-C38 (FdhF-C11), provides a ligand to the [4Fe-4S] cluster found in this protein. The residue equivalent to DmsA Arg-77 (FdhF Lys-44) is located between the [4Fe-4S] cluster and one of the pterins of the Mo-bisMGD cofactor and may play an important role in the electron transfer mechanism (26). The effects of the DmsA-C38S and DmsA-R77S mutants on the EPR and redox properties of the molybdenum contrast with those of a mutation of a residue whose side chain contributes to the immediate coordination sphere of the molybdenum. For example, mutation of the protein-molybdenum ligand of DmsABC, DmsA Ser-176 to Ala has a profound effect on both the Mo(V) EPR spectrum and the overall redox chemistry of the Mo-bisMGD cofactor (19). In FdhF, mutation of the selenocysteine protein-molybdenum ligand also significantly modifies the Mo(V) EPR spectrum (37).
The DmsA-C38S and DmsA-R77S mutants appear to have minor, but opposite effects on the redox chemistry of the Mo-bisMGD. The shift elicited by the DmsA-C38S mutant is a ⌬E m,7 ϭ ϩ20 mV shift of the Mo(IV/V) couple, whereas the shift elicited by the DmsA-R77S mutant is a ⌬E m,7 ϭ Ϫ24 mV shift of the Mo(V/VI) couple. The [3Fe-4S] cluster of the DmsA-C38S mutant is reduced at E h values where the Mo(V) is visible (it has an E m,7 of approximately 178 mV (23)) and therefore does not carry a formal charge. Thus, it is likely that subtle changes in structure caused by the introduction of a [3Fe-4S] cluster are responsible for the small change in the Mo(IV/V) E m,7 . In the case of the DmsA-R77S mutant, the ⌬E m,7 ϭ Ϫ24 mV shift in the Mo(V/VI) E m,7 may be rationalized in terms of the loss of the electron-withdrawing effect of the positive charge of the Arg residue. This suggests that the residues that contact the pterin ring systems may have an important role in defining the Mo(IV/V) and Mo(V/VI) E m,7 values. It should be noted that in the case of both mutations, the observed effects are on only one of the two molybdenum E m,7 values. Further mutagenesis studies will be necessary to confirm the role of pterin contact residues in defining the electrochemistry of the Mo-bisMGD.
Spin-spin interactions between EPR-visible prosthetic groups can provide important information on the topographical arrangement of these centers in multicofactor enzymes. Of relevance to the data presented herein are the studies carried out on milk xanthine oxidase (36, 38 -41). In this enzyme, there is a strong spin-spin interaction between various accessible forms of the Mo(V) of the molybdo-molybdopterin cofactor and one of the [Fe-S] clusters. This interaction results in both splitting of the EPR spectrum of the Mo(V) and in an enhancement of its spin relaxation rate at low temperatures that manifests itself as a significant increase in the P1 ⁄2 . In the case of DmsABC, there is a strong spin-spin interaction between the Mo(V) species and the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster that manifests itself as an increase in the Mo(V) P1 ⁄2 (Fig. 3), but does not result in a line shape change that can readily be interpreted as arising from a dipolar interaction.
A structure has recently become available of a bacterial protein similar in structure and sequence to xanthine oxidase. This protein, Desulfovibrio gigas aldehyde oxidoreductase (42), contains a molybdo-molybdopterin cytosine dinucleotide cofactor, and two [2Fe-2S] clusters. This enzyme has EPR properties similar to those of xanthine oxidase (43). It has been suggested that the cluster equivalent to the one interacting with the molybdenum in xanthine oxidase may be located approximately 15 Å from the molybdenum (44). In the structure of aldehyde oxidoreductase (42), one of the Cys ligands of this cluster is also hydrogen-bonded to the pterin. These observations bear interesting comparison with distance estimates for xanthine oxidase based on EPR analyses that are in the 8 -25-Å range (41). However, it should be noted that the assignment of the interacting cluster in the structure of aldehyde oxidoreductase remains controversial (44). In another protein that is much more closely related to DmsA, E. coli FdhF, a [4Fe-4S] cluster is located approximately 13 Å from the molybdenum (26). No interaction between the [4Fe-4S] cluster and the molybdenum of FdhF has yet been reported.
Based on comparisons with xanthine oxidase and FdhF, a position for the E m,7 ϭ Ϫ120 mV cluster of DmsB that is consistent with the observed spin-spin interaction between this cluster and the molybdenum would be equivalent to that of the interacting [2Fe-2S] cluster of xanthine oxidase or the [4Fe-4S] cluster of FdhF. However, it is clear from analyses of mutants of the DmsA Cys group that there is no EPR-detectable cluster coordinated by DmsA in the wild-type enzyme (23,24). Also, mutants of DmsA Cys-38 and DmsA Arg-77 still contain a potentiometrically identifiable E m,7 ϭ Ϫ120 mV cluster, further supporting the assertion that there is no cluster coordi- FIG. 6. Interaction between the Mo(V) of DmsA and an engineered [3Fe-4S] cluster in DmsB. A, Mo(V) EPR spectra of ferricyanide-oxidized HB101 membranes containing overexpressed DmsA S176A BC (i) and DmsA S176A B C102S C (ii). EPR conditions were as described for Fig. 5A, except that the temperature was 30 K, the microwave power was 0.2 mW, and the modulation amplitude was 6 G pp . B, microwave power saturation curves of the Mo(V) signal from DmsA S176A BC (squares, P1 ⁄2 ϭ 0.8 mW, b ϭ 1.5) and DmsA C176A B C102S C (triangles, P1 ⁄2 ϭ 0.8 mW, b ϭ 1.5). nated by the DmsA Cys group. Based on qualitative comparisons with D. gigas aldehyde oxidoreductase, it is therefore likely that the portion of DmsB containing the E m,7 ϭ Ϫ120 mV cluster is located within approximately 15 Å of the molybdenum of DmsA.
One important distinction between the interaction reported herein and that observed in xanthine oxidase is the lack of apparent line shape change accompanying the enhancement of the spin relaxation of the Mo(V) signal. In xanthine oxidase, the line shape change manifests itself when the temperature of the sample is reduced to sufficiently slow down the spin relaxation rate of the interacting [2Fe-2S] cluster so that it is compatible with that of the Mo(V) species (39). The interacting [2Fe-2S] cluster of xanthine oxidase is a saturable (at 20 K) reduced [2Fe-2S] cluster (36) that displays readily interpretable behavior with increasing microwave power. In the case of DmsABC, the interacting species is the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster of DmsB, which comprises half of a 2[4Fe-4S] ferredoxin motif with the E m,7 ϭ Ϫ50 mV [4Fe-4S] cluster. At the potentials where the spin relaxation enhancement is observed, the E m,7 ϭ Ϫ120 mV cluster is itself undergoing a complex interaction with the E m,7 ϭ Ϫ50 mV cluster (2,6). This interaction is equivalent to that observed in the bacterial 2[4Fe-4S] ferredoxins and results in an unsaturable EPR spectrum indicative of a very rapid spin relaxation rate (40,(45)(46)(47). Thus, in the interaction between the Mo(V) of DmsA and the E m,7 ϭ Ϫ120 mV cluster of DmsB, the latter center is essentially unsaturable with increasing microwave power, and its relaxation rate at 30 K is very likely to be too high for the observation of significant splittings in the Mo(V) spectrum. A lack of quantifiable splitting of the DmsABC Mo(V) signal by the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster of DmsB precludes a quantitative estimate of the intercenter distance based on the data presented herein.
We have demonstrated that incorporation of a [3Fe-4S] cluster coordinated by the DmsA Cys group modifies the interaction between the E m,7 ϭ Ϫ120 mV cluster and the molybdenum (Fig. 4). This is consistent with (i) this cluster being located close to or on the vector joining the Mo(V) and the [4Fe-4S] cluster (for a dipolar interaction) and/or (ii) this cluster being on or close to the interaction pathway between the two centers (for an exchange interaction). The effect of the DmsA-R77S mutation (Fig. 4) clearly favors the second explanation, since presumably it has little effect on the distance between the two centers but is still able to eliminate the detectable interaction. Given that both the presence of a [3Fe-4S] cluster and the DmsA-R77S mutant both essentially eliminate electron transfer between DmsB and DmsA (23,24), the data presented herein are consistent with the interaction pathway being equivalent to the electron transfer pathway. In the case of the interaction in xanthine oxidase, it has also been proposed that it occurs through a specific pathway through the protein and that it is also primarily exchange in nature (39).
Given the significant increase in the observed P1 ⁄2 for the Mo(V) spectrum observed in the wild-type enzyme, we anticipated that it would be possible to detect a spin-spin interaction between the Mo(V) and either the reduced S ϭ 2 or oxidized S ϭ 1 ⁄2 [3Fe-4S] cluster of the DmsA-C38S mutant. In the case of the glycerol-inhibited, ferricyanide-oxidized DmsA-C38S enzyme, there appears to be no enhancement of the Mo(V) spin relaxation and only a minor broadening of its EPR spectrum (Fig. 5). This result is somewhat surprising, since the [3Fe-4S] cluster of the DmsA-C38S mutant enzyme is, based on comparison with the structurally characterized FdhF (26), very likely to be located approximately 13 Å from the molybdenum.
Given the lack of significant interaction between the glyc-erol-inhibited Mo(V) and the engineered [3Fe-4S] cluster in DmsA, it is not surprising that no interaction is observed between the Mo(V) of the DmsA-S176A mutant and the engineered [3Fe-4S] cluster of the DmsB-C102S mutant (Fig. 6). In this case, the lack of interaction may simply be due to the extra distance between the centers or perhaps due to the E m,7 ϭ Ϫ120 mV cluster acting as a "shield," preventing the observation of an interaction. In the DmsB-C102S mutant, the E m,7 ϭ Ϫ50 mV [4Fe-4S] cluster of DmsB is converted to a [3Fe-4S] cluster (4, 6). As described above, in the wild-type enzyme, this cluster appears to form half of an 2[4Fe-4S] ferredoxin motif. Based on the structurally characterized bacterial 2[4Fe-4S] ferredoxins (48), this cluster is potentially a maximum of approximately 12 Å further away from the molybdenum of DmsA than the E m,7 ϭ Ϫ120 mV cluster (this depends on the angle between the vector joining the E m,7 ϭ Ϫ120 and Ϫ50 mV clusters and that joining the E m,7 ϭ Ϫ120 mV cluster and the molybdenum center). Thus, an approximate maximum distance of 27 Å (12 plus 15 Å) is apparently too great for an interaction to be observed in DmsABC. The data presented herein and previously reported data on the topology and electron transfer pathway of DmsABC (4 -6, 9, 20, 23, 24, 49) allow a tentative model for electron transfer through the enzyme to be proposed. MQH 2 binding and oxidation occur at a single dissociable site in DmsC (20), which is conformationally linked to the E m,7 ϭ Ϫ50 mV [4Fe-4S] cluster of DmsB (6). This cluster forms half of a 2[4Fe-4S] ferredoxin pair (2,6) with the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster. Electrons pass from MQH 2 through these two clusters and continue to the Mo-bisMGD cofactor via a vestigial [4Fe-4S] cluster binding domain defined by the N-terminal Cys motif of DmsA (24). This segment of the pathway is sensitive both to the presence of an engineered [3Fe-4S] cluster in the DmsA-C38S mutant and to mutation of a residue (DmsA Arg-77) that is in close juxtaposition to one of the pterins of the Mo-bisMGD (23,24,26). We have presented evidence herein that is consistent with this segment of the pathway being equivalent to the pathway of the magnetic interaction between the E m,7 ϭ Ϫ120 mV [4Fe-4S] cluster and the molybdenum of the Mo-bisMGD cofactor. Given the potential position of DmsA Arg-77 in relation to one of the pterins, it is also very likely that one of the functions of this pterin is to provide a conduit for electron transfer to the molybdenum.