Mechanism of assembly of the Bis(Molybdopterin guanine dinucleotide)molybdenum cofactor in Rhodobacter sphaeroides dimethyl sulfoxide reductase.

A fully defined in vitro system has been developed for studying the mechanism of assembly of the bis(molybdopterin guanine dinucleotide)molybdenum cofactor in Rhodobacter sphaeroides dimethyl sulfoxide reductase (DMSOR). R. sphaeroides DMSOR expressed in a mobA(-) Escherichia coli strain lacks molybdopterin and molybdenum but contains a full complement of guanine in the form of GMP and GDP. Escherichia coli MobA, molybdopterin-Mo, GTP, and MgCl(2) are required and sufficient for the in vitro activation of purified DMSOR expressed in the absence of MobA. High levels of MobA inhibit the in vitro activation. A chaperone is not required for the in vitro activation process. The reconstituted DMSOR can exhibit up to 73% of the activity observed in recombinant DMSOR purified from a wild-type strain. The use of radiolabeled GTP has demonstrated incorporation of the guanine moiety from the GTP into the activated DMSOR. No role was observed for E. coli MobB in the in vitro activation of apo-DMSOR. This work also represents the first time that the MobA-mediated conversion of molybdopterin to molybdopterin guanine dinucleotide has been demonstrated directly without using the activation of a molybdoenzyme as an indicator for cofactor formation.

In all molybdenum-containing enzymes except nitrogenase, the metal is coordinated to the organic cofactor molybdopterin (MPT), 1 which also serves as the cofactor for tungsten-containing enzymes (1). The MPT molecule is a substituted pterin ring that coordinates the metal through a dithiolene linkage (Fig.  1A). Additional variability of the molybdenum cofactor is found in bacteria with the attachment of GMP, AMP, IMP, or CMP to the phosphate group of MPT. In Escherichia coli, the final stage in cofactor biosynthesis is the linkage of GMP to MPT to form molybdopterin guanine dinucleotide (MGD) (Fig. 1B). This form of the cofactor is found in a number of enzymes essential to the global nitrogen, carbon, and sulfur cycles including dimethyl sulfoxide reductase (DMSOR), nitrate reductase (NR), biotin sulfoxide reductase, and trimethylamine N-oxide reduc-tase (2). In several MGD-containing enzymes, including Rhodobacter DMSOR (3)(4)(5) and biotin sulfoxide reductase (6), Desulfovibrio desulfuricans periplasmic NR (7), and E. coli trimethylamine N-oxide reductase (8), the molybdenum atom is coordinated by two MGD molecules (Fig. 1C). This form of the cofactor is referred to as bis(molybdopterin guanine dinucleotide)molybdenum (bis(MGD)Mo). As first demonstrated in the crystal structure of Rhodobacter sphaeroides DMSOR (9), the bis(MGD)Mo cofactor is deeply buried in the interior of the protein such that only the molybdenum atom and three of the coordinating sulfur ligands are accessible from the solvent.
The conversion of MPT to MGD requires a gene product of the mob locus. While wild-type cells contain both MPT and MGD, mob-deficient cells contain elevated levels of MPT but no detectable MGD (10). In E. coli, the mob locus contains two genes, mobA and mobB, that encode the 22-kDa protein MobA (also referred to as protein FA) and the 19-kDa protein MobB (11,12). Both MobA and MobB have been cloned, expressed, and purified (13,14). The mobB gene is not essential for cofactor biosynthesis, since mobA alone can fully complement the deletion of both mob genes (15).
Studies of the conversion of MPT to MGD have been hampered by the extreme instability of both of these forms of the cofactor. In earlier studies, activation of E. coli NR expressed in a mob Ϫ strain served as an indirect assay for MGD formation. Using this technique, it was demonstrated that the MobA-dependent activation of NR from a mob Ϫ strain was enhanced by the addition of GTP and MgCl 2 to the soluble cell extract utilized for activation (16). Activation also required the presence of the NarJ protein (15), which appears to function as an NR-specific chaperone. (17). The MobB protein was not essential for this activation process, although its presence appeared to increase the extent of activation (15).
Despite the insight provided by these experiments regarding the role of MobA, activation of NR from a mob Ϫ strain using only purified, well defined components has not been achieved. In fact, most of the activation mixtures have included either a crude or soluble cell extract as an essential component. In the best defined activation mixtures, soluble cell extract was applied to a gel filtration column, and fractions from this column were added to NR purified from a mob Ϫ strain, purified MobA, GTP, and MgCl 2 (15,16). Additionally, as mentioned by Palmer et al. (15), the requirement for NarJ in the activation of NR from a mob Ϫ cell strain also complicated the use of this system for quantitative analysis of MGD formation. NarJ appears to play an NR-specific role in activation, since it does not influence MobA-dependent activation of trimethylamine N-oxide reductase (15). Blasco et al. (17) proposed that the main role of NarJ might be to maintain the NR complex in a suitable open conformation that is competent for molybdenum cofactor insertion. Although NarJ does not appear to be involved in forma-tion of the MGD cofactor, it is difficult to separate the NRspecific role of NarJ as a chaperone and the more general cofactor biosynthesis role of the mob proteins, since both functions are required to form the active NR enzyme in these experiments.
In view of the difficulties described above, the need was apparent for a new system for studying the conversion of MPT to MGD. The use of R. sphaeroides DMSOR eliminated many of the complications associated with NR. DMSOR is a single subunit, soluble enzyme that contains bis(MGD)Mo as the sole prosthetic group (9), whereas NR is a membrane-bound, heterotrimeric complex that contains [Fe-S] clusters and a heme group in addition to the molybdenum cofactor (18). Recombinant DMSOR can also be heterologously expressed in E. coli (19), and both the recombinant enzyme and that purified from Rhodobacter have been extensively characterized (4, 19 -21). One possible complication is the presence in the R. sphaeroides operon of a gene called dmsB (22) or dorB (23) that has been proposed to encode a chaperone for the DMSOR protein. However, heterologously expressed, recombinant DMSOR is active and incorporates the molybdenum cofactor in the absence of this proposed chaperone, indicating that a R. sphaeroides chaperone is not essential for successful expression of R. sphaeroides DMSOR in E. coli.
In this paper, we present a completely defined in vitro system for studying the mechanism of assembly of the bis(MG-D)Mo cofactor in R. sphaeroides DMSOR. The Rhodobacter enzyme has been expressed in E. coli in the absence of one or both of the mob gene products and activated using only purified components in a process that requires MobA, GTP, MgCl 2 , and MPT. After this in vitro activation, DMSOR from a mob Ϫ strain exhibited a maximum of 73% of the activity observed in recombinant DMSOR purified from a wild-type strain. The conversion of MPT to MGD by MobA has also been shown to occur in the absence of DMSOR.

EXPERIMENTAL PROCEDURES
Construction of Expression Plasmids-The DMSOR expression plasmid used in this work was created by digesting the pJH720 construct (19) (Table I) with NcoI and HindIII to release the complete mature R. sphaeroides DMSOR coding sequence including an N-terminal His 6 tag. This insert was subsequently ligated into similarly digested pTrc 99 A (Amersham Pharmacia Biotech) to form pJH820.
For creation of the MobA expression constructs, an oligonucleotide primer was designed that covered the translational start region of the mobA gene. This primer contained mismatches creating two restriction sites, an upstream BamHI site and an NcoI site that changed the start codon for mobA from GTG to ATG and the second codon from AAT to GCT, converting the second amino acid from Asn to Ala. A second oligonucleotide primer complementary to the translation stop site of the mobA gene with mismatches creating a downstream SphI site was also designed. These two oligonucleotides were used to prime PCRs with Elongase (Life Technologies) using E. coli MC4100 (Table I) genomic DNA as the template, and a fragment of the expected size (620 base pairs) was amplified. This PCR product was digested with BamHI and SphI and subsequently ligated with similarly digested pLysE to create pCT300A (Table I). To create a high yield expression vector for MobA, pCT300A was digested with NcoI and SalI to release the mobA gene, which was subsequently ligated into pTrc 99 A to form pCT800A.
A similar procedure was used for the mobB gene as well. The oligonucleotide covering the translation start site contained mismatches creating an upstream BglII site and a NcoI site at the ATG start codon. The primer covering the translation stop site contained mismatches creating a downstream SphI site. A PCR fragment of the expected size (569 base pairs) was amplified, digested with BglII and SphI, and ligated with pLysE digested with BamHI and SphI to create pCT300B (Table I). To create a high yield expression vector for MobB, pCT300B was digested with NcoI and SalI to release the mobB gene, which was subsequently ligated into pTrc 99 A to form pCT800B. Oligonucleotides were synthesized by Life Technologies Custom Primers, and automated sequencing was performed at the Duke University DNA Analysis Facility.
Protein Expression and Purification-Wild type DMSOR was purified from R. sphaeroides as described previously (3). Recombinant DMSOR was purified from MC4100 cells (Table I) containing pJH820. Cells were grown aerobically at 37°C overnight in LB supplemented with 100 g/ml ampicillin. This culture was then diluted 1:25 into LB supplemented with ampicillin and 0.5 mM Na 2 MoO 4 and subsequently grown at 30°C until A 600 ϭ 1. This culture was diluted 1:20 into M9ZB medium supplemented as described previously (19), except Na 2 MoO 4 was present at 1.0 mM, isopropyl-1-thio-␤-D-galactopyranoside was added to 5 M, and ampicillin replaced all other antibiotics. The cells were then grown anaerobically for 20 -24 h at room temperature before harvesting.
TP1000 cells (Table I) containing pJH820 and grown as described above were used to purify mobAB Ϫ DMSOR. TP1000 cells containing pJH820 and either pCT300A or pCT300B were grown as described above with the addition of 34 g/ml chloramphenicol and used to purify mobB Ϫ and mobA Ϫ DMSOR, respectively. All forms of recombinant DMSOR were purified as described previously (19) except that cell lysis was achieved using a Microfluidics M110L Microfluidizer Processor (6), and no heat step was performed.
MobA and MobB were purified from TP1000 cells containing pCT800A or pCT800B, respectively. A culture was grown overnight and then diluted 1:25 into LB containing 0.5 mM Na 2 MoO 4 and 100 g/ml ampicillin and grown at 30°C until A 600 ϭ 1. This culture was diluted 1:20 into LB supplemented with ampicillin, 1.0 mM Na 2 MoO 4 , and 1 mM isopropyl-1-thio-␤-D-galactopyranoside and grown overnight aerobically at 30°C. MobA and MobB were purified as described by Eaves et al. (13) except that cell lysis was achieved using a Microfluidics M110L Microfluidizer Processor, and a 0.5 M NaCl wash step was added to the Cibacron blue affinity column during the purification of MobA. After purification, MobA was dialyzed into 50 mM Hepes, 10 mM 2-mercaptoethanol, 0.1 M NaCl, pH 7.6.
Protein Analysis-Purified DMSOR was quantitated spectrophotometrically using an extinction coefficient at 280 nm of 200,000 M Ϫ1 cm Ϫ1 (3). Comparisons using the Pierce BCA assay indicated that the mob Ϫ forms of DMSOR have the same extinction coefficient at 280 nm as the wild-type protein. The concentrations of purified MobA and MobB were calculated using molar extinction coefficients at 280 nm of 26,000 and 18,220 M Ϫ1 cm Ϫ1 , respectively (13).
DMSOR activity and the total guanine and molybdenum content of the purified proteins were assayed as described previously (19). Form A analysis was performed on purified protein as described previously (3), using 10% methanol in the HPLC running buffer. Guanine nucleotides were extracted from purified DMSOR by boiling in an aqueous SDS solution (24) and subsequently identified by HPLC analysis using a method adapted from Nguyen and Sadee (25). Separation of the nucleotides was accomplished on a Partisil 10-SAX column using 0.02 M NH 4 H 2 PO 4 in 0.01 M NaCl at pH 3.6 (buffer A) and 1 M NH 4 H 2 PO 4 in 0.5 M NaCl at pH 3.6 (buffer B) at a flow rate of 2 ml/min. The column was initially equilibrated in 100% buffer A. After sample injection, a 4% buffer B solution was run for 5 min followed by a 15-min elution gradient from 4 to 100% buffer B.
Analysis of DMSOR Expressed in the Absence of Molybdate-Recombinant DMSOR was expressed using the plasmid pJH720 in BL21(DE3) pLysE cells as described previously (19) with minor modifications. The 80-ml overnight culture was used to inoculate a 1-liter culture of LB that was subsequently grown for 6 -8 h. The cells were then harvested by centrifugation and resuspended in M9ZB medium before addition to the 40-liter culture. No Na 2 MoO 4 was added to any of the growth media. Protein purification followed the previously published procedure (19) except the heat step was eliminated. The lysate was also fractionated with 40% saturated ammonium sulfate before precipitation of the DMSOR with 70% saturated ammonium sulfate and subsequent resuspension in 10 mM imidazole, 50 mM sodium phosphate, 300 mM NaCl, pH 7.5, for the Ni 2ϩ -nitrilotriacetic acid column. The purified protein was analyzed as described above.
Release of MPT from SO-MPT was released from SO by a procedure adapted from Schwarz et al. (26). Recombinant human SO was purified as described previously from TP1000 cells containing pTG918 (Table I) (27), dialyzed against 50 mM potassium phosphate, pH 7.8, and then concentrated to between 22 and 30 mg/ml using Centricon 30 or Centriprep 30 filtration devices (Amicon). In a Coy chamber containing a nitrogen, carbon dioxide, and hydrogen gas mix, the protein was gelfiltered using Nick columns (Amersham Pharmacia Biotech) equilibrated with 50 mM Tris/HCl, pH 7.5. The SO was then sealed in a Vacutainer (Fisher), removed from the Coy chamber, and denatured by heat treatment for 60 s at 90 -95°C to release MPT.
Activation of DMSOR Purified from a mob Ϫ Strain-In a standard assay, 10 -11 g of DMSOR was combined with MobA in a 50:1 molar ratio in a round cuvette (Fisher). Sufficient GTP and MgCl 2 were added to produce a final concentration of 1 mM each, and 50 mM Tris/HCl, pH 7.5, was added to bring the total volume to 150 l. This mixture was placed in the Coy chamber for a minimum of 15 min before 50 l of heat-denatured SO was added. The tube was then capped securely before removal from the Coy chamber and incubation at 37°C for 1.5 h. To monitor the extent of activation, the Me 2 SO reductase activity of the reconstituted mixture was assayed as described previously (19). Measurements were corrected for the activity observed under the same conditions in the absence of DMSOR, and at least two assays were performed for each data point.
For the time course experiment, 5.5 g of DMSOR was used per assay, and the incubation period varied from 0 to 675 min. To determine the effect of MobA concentration on the extent of activation, the amount of DMSOR was maintained at 10 g while the molar ratio of MobA to DMSOR was varied from 10:1 to 0.002:1. In both of these cases, the procedure described above was followed in all other aspects.
Activation of mob Ϫ DMSOR Using Radiolabeled GTP-For this activation, 0.11 mg of mobAB Ϫ DMSOR was combined with MgCl 2 and MobA in a 1:250 molar ratio of MobA to DMSOR in a total volume of 180 l. In two separate controls, MobA was left out of the mixture, and wild-type DMSOR was substituted for the protein purified from a mob Ϫ stain. An aliquot of 400 l of heat-denatured SO was added to each sample before it was sealed and removed from the Coy chamber. Using an airtight syringe, [␣-32 P]GTP (PerkinElmer Life Sciences) was added to a sealed, anaerobic solution of GTP, and 20 l of this solution was then injected into each reaction mixture. The final concentration of GTP and MgCl 2 was 0.1 mM, and the mixture contained 0.14 Ci/ml. Each sample was then incubated at 37°C for 3 h before the addition of 1.2 ml of 50 mM Tris/HCl, pH 7.5, to each sample to permit easy transfer to a Slide-A-Lyzer (Pierce). All three samples were then dialyzed two times against 2 liters of 50 mM Tris/HCl, pH 7.5, before measurement of bound radioactivity using a Beckman LS 1801 scintillation counter.
Form A-GMP Analysis-MobA (128 g) was mixed with a 1 mM concentration each of GTP and MgCl 2 and 200 l of heat-denatured SO in a total volume of 600 l under anaerobic conditions. The mixture was then incubated at 37°C for 3 h. Tubes containing no MobA or 0.5 mg of wild-type DMSOR substituted for MobA were treated in an identical manner. The volume of the samples was then increased to 9 ml by the addition of 8.4 ml of 10 mM sodium phosphate buffer, pH 7.
To produce form A-GMP, the pH was adjusted to 2.5, and 100 mg of SDS and 1 ml of 1% iodine, 2% KI solution were added (10). A 4-ml QAE-Sephadex column was used to separate form A and form A-GMP according to the procedure of Johnson et al. (10). The column was washed with 16 ml of 10 mM acetic acid, 136 ml of 10 mM HCl, and 40 ml of 50 mM HCl. After discarding the first 60 ml of the 10 mM HCl wash, 1.5-ml fractions were collected for the remainder of the column

RESULTS
Expression of DMSOR in mob Ϫ Cells-The previous construct used to express recombinant R. sphaeroides DMSOR, pJH720, required the presence of T7 RNA polymerase within the host cells (19), thus limiting the number of suitable expression cell types. To avoid this complication, the coding sequence for DMSOR including the N-terminal His 6 tag was transferred into the pTrc 99 A vector to form pJH820 (Table I). Using this new vector in E. coli MC4100 cells (Table I), an average of 3.4 mg of DMSOR were purified per liter of cell culture. This represents an 18-fold increase over the previous yield, while maintaining high levels of activity and molybdenum incorporation (Table II). This large increase in expression can probably be attributed to the use of the new expression strain, since problems have been observed with molybdoenzyme expression in the E. coli BL21(DE3) cells used previously (27). Recombinant enzyme purified from the wild-type parent strain, MC4100, is referred to herein as rDMSOR to differentiate it from enzyme purified from R. sphaeroides.
The mob locus of MC4100 was replaced by a kanamycinresistance cassette by Palmer et al. (15) to form the TP1000 strain. Transformation of pJH820 into this strain enabled purification of DMSOR expressed in the absence of both MobA and MobB, herein referred to as mobAB Ϫ DMSOR. Construction of the plasmids pCT300A and pCT300B (Table I) enabled coexpression of DMSOR with MobA or MobB, respectively. DMSOR purified from the strain TP1000 pCT300A was expressed in the absence of the MobB protein only and is referred to as mobB Ϫ DMSOR. Similarly, enzyme purified from the strain TP1000 pCT300B was expressed in the absence of the MobA protein only and is referred to as mobA Ϫ DMSOR.
Previous purification of rDMSOR included heating the soluble cell extract to 50°C and maintaining that temperature for 1 min (19); however, inclusion of this step in the purification of mobAB Ϫ or mobA Ϫ DMSOR resulted in precipitation of the recombinant enzyme. Therefore, the heat step was eliminated in the purification of all of the recombinant DMSOR used in this study. Even with the exclusion of the heat step, yield was considerably lower for both mobAB Ϫ and mobA Ϫ DMSOR, an average of only 0.51 mg of protein/liter of cell culture. No significant changes in yield were observed for mobB Ϫ DMSOR. Incidentally, the increased heat sensitivity observed in DM-SOR from a mobA Ϫ strain does demonstrate that the heat step is an effective method for separating cofactor-free enzyme from active, wild-type DMSOR.
Analysis of DMSOR Expressed in a mob Ϫ Strain-Purified DMSOR expressed in the absence of one or both of the gene products of the mob locus was compared with that from the wild-type cell strain. The total guanine and molybdenum content of each protein was quantitated, and the DMSOR activity was determined (Table II). The proteins were also examined for the presence of MPT, detected as form A (Fig. 2 and Table II). Both mobAB Ϫ and mobA Ϫ DMSOR were colorless, contained no detectable molybdenum or MPT, and exhibited no Me 2 SO reductase activity. Surprisingly, HPLC analysis showed that they both contained a full complement of guanine. In contrast,  the mobB Ϫ DMSOR had the typical color characteristic of rDM-SOR and was fully active. This is in accordance with work by Palmer et al. (15) demonstrating that MobB is not required for the expression of active NR, trimethylamine N-oxide reductase, or formate dehydrogenase lyase. The guanine moieties present in mobAB Ϫ and mobA Ϫ DM-SOR were extracted from the purified proteins by boiling in an aqueous SDS solution and subsequently identified by HPLC analysis. In both cases, approximately 70% of the guanine was present as GMP and 30% as GDP; no GTP was detected. There was no evidence of dGDP, but we were unable to separate GMP and dGMP under the conditions used.
Analysis of DMSOR Expressed in the Absence of Molybdate-Previously, it was observed that recombinant expression of active R. sphaeroides DMSOR in the BL21(DE3) E. coli strain required supplemental molybdate in the medium (19). To investigate the role of molybdate in bis(MGD)Mo biosynthesis, recombinant DMSOR was purified from BL21(DE3) cells grown in the absence of supplemental molybdate. While it had been previously shown that DMSOR expressed under the same conditions in the presence of supplemental molybdate is active and contains approximately 74% molybdenum in addition to MPT and a full complement of guanine (19), in the absence of exogenous molybdenum, the protein contained 2.1 mol of guanine/mol of protein, and no MPT or molybdenum was detected.
Activation of DMSOR Purified from an mobA Ϫ Strain-The in vitro activation of DMSOR from a mobA Ϫ strain presented itself as a sensitive and convenient assay for production of the bis(MGD)Mo cofactor. Since R. sphaeroides DMSOR has been successfully expressed in E. coli (19), no problems were anticipated with the use of the E. coli cofactor biosynthesis proteins in this in vitro assay. To assess the roles of E. coli MobA and MobB in the in vitro activation, the two proteins were expressed in TP1000 cells using pCT800A and pCT800B, respectively (Table I). Both proteins were purified using the method of Eaves et al. (13) with only minor modifications.
In view of the total lack of pterin in mobAB Ϫ and mobA Ϫ DMSOR, it was clear that a source of MPT had to be included in any in vitro activation attempt. Since sulfite oxidase (SO) contains the MPT form of the cofactor and can be purified in relatively large quantities (27), heat-denatured, recombinant human SO was chosen as a source of MPT. Any risk of MGD contamination was avoided by expressing the recombinant SO in the mobAB Ϫ TP1000 cells. Anaerobic conditions were used for the release of MPT from SO and subsequent activation of DMSOR due to the extreme sensitivity of the cofactor to oxygen (1).
The assay components were combined with mobAB Ϫ or mobA Ϫ DMSOR under anaerobic conditions and incubated at 37°C for 1.5 h. The extent of activation was determined by measuring the DMSOR activity generated in the sample. Using this method, it was determined that in vitro activation of mobAB Ϫ or mobA Ϫ DMSOR required the presence of MobA, GTP, MgCl 2 and heat-denatured SO as a source of MPT (Table  III). GTP and MgCl 2 were present at 1 mM each, and MobA was present in a 1:50 molar ratio with DMSOR. No activation was observed in the absence of any one of these components.
GMP and ATP were unable to substitute for GTP under the conditions used. The addition of 50 mM Na 2 MoO 4 also did not affect the extent of activation; nor did the addition of MobB in a 1:50 molar ratio with DMSOR (Table III). To examine the possible requirement for an additional component for the proper functioning of MobB, activation was carried out in the presence of 20 l of TP1000 lysate in addition to the standard assay components. Even under these conditions, there was no difference in the extent of activation seen with the addition of  MobB alone. Additionally, no significant differences were observed between the activation of mobAB Ϫ and mobA Ϫ DMSOR (Table III) The crystal structure of the E. coli MobA protein is presented in the accompanying article (28). In attempts at obtaining the structure of the MobA-GTP complex, it was found that the nucleotide was much better defined as a complex with Mn 2ϩ than when Mg 2ϩ was used as the cation. In the reconstitution procedure described here, no significant changes were observed in the level of activation when 1 mM MnCl 2 was substituted for 1 mM MgCl 2 , showing that the MobA-GTP-Mn 2ϩ complex is a catalytically relevant species.
Time Course for in Vitro Activation-To determine the highest level of in vitro activation possible, Me 2 SO reductase activity was measured as a function of incubation time (Fig. 3). Maximum activation required at least 7.5 h under the conditions used, at which time the reconstituted enzyme exhibited an average of 17.4 units of activity/mg of DMSOR. This represents about 73% of the activity observed in rDMSOR purified from the wild-type, MC4100 strain. Doubling the MPT concentration, increasing the molar ratio of MobA to DMSOR to 1:10, or doing both simultaneously did not significantly change the final specific activity. Preincubating MobA with GTP, MgCl 2 , and MPT for 2 h at 37°C prior to the addition of mobAB Ϫ DMSOR also did not change the extent of activation during the first 90 min.
Excess MobA Inhibits in Vitro Activation of mobA Ϫ DM-SOR-To determine the effect of MobA concentration on the extent of activation, the concentration of mobAB Ϫ DMSOR was maintained at 10 g, while the molar ratio of MobA to DMSOR was varied from 10:1 to 0.002:1. No significant differences were observed in the level of activation after a 1.5-h incubation period when MobA was present at or below an equimolar ratio with DMSOR (Fig. 4, top). However, when the molar concentration of MobA exceeded that of mobAB Ϫ DMSOR, the extent of activation was attenuated. In fact, almost no activation occurred when the molar ratio of MobA to DMSOR reached 4:1 (Fig. 4, bottom). The presence of MobB did not inhibit activation under the conditions used, even when the molar concentration of MobB was 5 times that of mobAB Ϫ DMSOR.
Activation of mob Ϫ DMSOR Using Radiolabeled GTP-The requirement for GTP in the activation of mobAB Ϫ DMSOR could indicate that the guanine moieties of the newly formed MGD were derived from the added GTP. Alternatively, it was possible that the guanine moieties present in the purified mobAB Ϫ DMSOR prior to activation were joined to the MPT to form MGD. To investigate the latter possibility, mobAB Ϫ DM-SOR was reconstituted in the presence of ␣-32 P-labeled GTP FIG. 4. Inhibition of in vitro activation by excess MobA. During in vitro activation, the amount of mob Ϫ DMSOR was maintained at 10 g while the molar ratio of MobA to DMSOR was varied from 10:1 to 0.002:1. Me 2 SO reductase activity was determined for each sample after incubation for 1.5 h at 37°C. followed by dialysis to remove any unincorporated nucleotides. For this experiment, the concentrations of GTP and MgCl 2 were decreased to 0.1 mM each, and the molar ratio of MobA to DMSOR was changed to 1:250. Neither of these changes significantly lowered the extent of activation from that observed under the standard conditions.
Following dialysis, the sample of mobAB Ϫ DMSOR activated in the presence of the radiolabeled GTP contained 8,590 cpm. In the presence of MobA, substitution of active DMSOR purified from R. sphaeroides for mobAB Ϫ DMSOR yielded only 688 cpm after dialysis. Similarly, when MobA was excluded from a sample containing mobAB Ϫ DMSOR, only 611 cpm were found after dialysis. These data indicate that the guanine moiety in the final form of the DMSOR cofactor is derived from the GTP added to the activation mix rather than the guanine moieties already present on the purified protein.
Direct Evidence for the in Vitro Formation of MGD-Up to this time, the formation of MGD by MobA has been demonstrated by indirect means such as the absence of MGD in a mob Ϫ cell strain (10) or the requirement for MobA to activate NR (14,15) or DMSOR purified from a mob Ϫ cell strain. The procedure developed here, however, offered the chance to directly demonstrate the production of MGD from MPT by MobA. To that affect, MobA was incubated for 3 h in the presence of MPT (supplied in the form of heat-denatured SO) and a 1 mM concentration each of GTP and MgCl 2 . This mixture was then incubated overnight with SDS and iodine according to the method of Johnson et al. (10) to convert MPT to the form A derivative and MGD to form A-GMP. A QAE-Sephadex column was then used to separate the two fluorescent derivatives.
As seen in the top of Fig. 5, in the absence of MobA, the form A peak generated from MPT was present, but there was no significant form A-GMP peak, indicating that no appreciable MGD was formed in the absence of MobA. When MobA was present in the incubation mixture, the form A peak was considerably smaller, and a form A-GMP peak was detected (Fig.  5, middle). To confirm the identity of form A-GMP, the fractions containing this derivative were treated with alkaline phosphatase and pyrophosphatase. After this treatment, the fluorescence of these fractions was measured again, and a large increase in the fluorescence was observed (Fig. 5, dashed lines). This fluorescence increase is associated with the scission of the pyrophosphate bond, thereby eliminating the quenching effects of the ribonucleotide on the inherent fluorescence of form A (10). The fluorescence spectrum of the species present after phosphatase treatment was identical to that of form A. The amount of form A-GMP produced under these conditions was comparable with that present in the control reaction containing 0.5 mg of R. sphaeroides DMSOR (Fig. 5, bottom). The addition of mobAB Ϫ DMSOR to the MobA reaction mixture did not appreciably alter the size of the form A-GMP peak produced. DISCUSSION We present here a fully defined, in vitro system for studying the mechanism of assembly of the bis(MGD)Mo cofactor. The in vitro generation of activity in R. sphaeroides DMSOR purified from a mob Ϫ strain serves as a sensitive and convenient assay of bis(MGD)Mo assembly and insertion into the apoprotein. Using this assay, it has been demonstrated that MobA, MPT, GTP, and MgCl 2 are required and sufficient for the activation of mobAB Ϫ and mobA Ϫ DMSOR. Clearly, activation of mobA Ϫ DMSOR, unlike that of E. coli NR, can proceed in the absence of a chaperone. Therefore, the specific roles of the mob proteins in the assembly of the DMSOR cofactor can be examined without the complication of the enzyme-specific role of a chaperone. In addition, the MobA-mediated conversion of MPT to MGD has been demonstrated directly for the first time without using the activation of a molybdoenzyme to indicate cofactor formation. This has also demonstrated that MGD formation can occur in the absence of a molybdoenzyme apoprotein.
The Rhodobacter mobAB Ϫ and mobA Ϫ DMSOR proteins used in this activation assay were devoid of any detectable MPT or molybdenum. It was demonstrated previously that E. coli NR (29) and DMSOR (30) also do not contain any MPT or molybdenum when expressed in a mob Ϫ strain. Therefore, the guanine moiety of the cofactors appears to play an essential role for binding of that cofactor in many MGD-containing proteins. Alternatively, the MPT-binding site could be latent in apo-DMSOR, although the binding sites for the two guanine moieties are obviously present as indicated by the presence of a full complement of guanine in mobA Ϫ DMSOR. Hä nzelmann et al. (24) have purified CO dehydrogenase, a molybdopterin cy- tosine dinucleotide (MCD)-containing enzyme, from Hydrogenophaga pseudoflava grown in the absence of molybdenum or in the presence of tungstate, and they found that it was devoid of MCD while still containing a stoichiometric amount of cytidine moieties, including CDP, dCDP, CMP, dCMP, CTP, and dCTP (listed in order of abundance). Since only GMP and GDP were found in mobA Ϫ DMSOR, this may indicate a greater specificity in the guanosine binding pocket of DMSOR. The presence or absence of guanine moieties in E. coli DMSOR and NR expressed in the absence of MobA was not investigated in the studies mentioned earlier (16,29,30).
It is unclear whether the guanine moieties present in mobA Ϫ DMSOR play any role in cofactor biosynthesis and insertion in vivo. However, the absence of MPT in apo-DMSOR shows that occupancy of the guanine nucleotide binding site does not by itself create the MPT binding site. The guanine moieties incorporated into the activated DMSOR originate from the GTP added to the activation mixture, as demonstrated by the use of radiolabeled GTP. Although the GMP and GDP are bound to apo-DMSOR strongly enough to be present stoichiometrically in the purified protein, the data presented here show that they can be replaced in the process of inserting the assembled bis(MGD)Mo cofactor into the protein. It may be that the guanine moieties found in mobAB Ϫ and mobA Ϫ DMSOR serve to stabilize the protein as a prelude to cofactor insertion, and one possible role for the proposed DMSOR chaperone is to catalyze the rapid exchange of the bound nucleotides for bis(MGD)Mo. It is also possible that, in vivo, the apoprotein interacts with a chaperone to prevent intermediate binding of GMP and GDP. Even with the guanine moieties present, heat sensitivity and overall low yield indicate that mobA Ϫ DMSOR is less stable than the holoenzyme.
As shown here, MPT is an essential component in the activation of DMSOR from a mobA Ϫ strain. MPT was not included in earlier studies on the activation of E. coli NR because inactive NR from a mob Ϫ strain was incorrectly believed to contain MPT (16). This has led to speculation on the source of the pterin component of the cofactor and to the suggestion that the lack of sufficient MPT may explain the low level of activation achieved in vitro with NR (29). Since cell lysate from a mob Ϫ strain was included in most NR activation mixtures, the lysate may have served as a source of trace amounts of MPT. This is especially possible given the increased level of MPT found in a mob Ϫ strain and the low level of activation attained, only 15% of that for the wild-type enzyme (16), even in the presence of lysate. The failure to include a source of MPT also accounts for the inability to achieve activation using only purified components.
It was surprising to find that high levels of MobA inhibit the activation of apo-DMSOR from a mob Ϫ strain. It is possible that a high concentration of MobA relative to that of available pterin makes it difficult for any two MGD moieties to assemble into a single bis(MGD)Mo cofactor. The MobA protein is able to form MGD in the absence of mobAB Ϫ DMSOR, as demonstrated by the formation of the form A-GMP derivative. However, this technique cannot distinguish between the synthesis of MGD and bis(MGD)Mo. Therefore, apoprotein may still be required for assembly of the complete bis(MGD)Mo cofactor. It is also unclear at this time whether MGD exists as an independent form of the cofactor or is only found as a protein-bound intermediate.
The highest level of activation described here, 73% of that observed in recombinant DMSOR purified from a wild-type strain, requires a long incubation period. Preincubation of MobA, MPT, GTP, and MgCl 2 did not change the rate of activation once mobAB Ϫ DMSOR was added. It is likely that the rate-limiting step is the replacement of bound guanine nucleotides with the bis(MGD)Mo cofactor. In view of the requirement for the NarJ chaperone in the activation of mob Ϫ NR (15,17), it would be interesting to determine if the addition of the proposed R. sphaeroides DMSOR chaperone to the reactivation mixture increases the rate of enzyme activation. The role of a chaperone in cofactor insertion is especially intriguing, since, in holo-DMSOR, the bis(MGD)Mo cofactor is deeply buried in the interior of the protein (9). Due to the larger size, multisubunit structure, and additional prosthetic groups present in NR, it is conceivable that such a chaperone is essential for cofactor insertion in NR while not required in the simpler DMSOR protein.
E. coli MobB appears to play no role in the in vivo or in vitro activation of R. sphaeroides DMSOR. In contrast, MobB appears to enhance NR activation, although it is not required for the process (15). Like the chaperone, MobB may not play as critical a role in cofactor production and insertion for the simpler molybdoenzymes. It may also be that E. coli MobB may not be able to substitute for the Rhodobacter counterpart in the reconstitution of R. sphaeroides DMSOR. Although the first 160 amino acids of R. sphaeroides MobB are highly similar to the entire sequence of E. coli MobB, the R. sphaeroides protein contains 292 additional amino acids not found in the E. coli protein (31). E. coli and R. sphaeroides MobA are much more closely related, with 39% identity in their amino acid sequences. Whereas E. coli MobB is obviously not required for activation of mobA Ϫ DMSOR, it would be interesting to see whether the addition of R. sphaeroides MobB to reconstitution assays would affect the rate or extent of DMSOR activation.
Although this study does not focus on metal chelation to the MPT cofactor, the lack of MPT and molybdenum in recombinant DMSOR expressed in BL21(DE3) cells in the absence of supplemental molybdenum does seem to imply that metal chelation precedes dinucleotide attachment. This is especially likely in view of the inability of H. pseudoflava to produce MCD in the absence of molybdenum despite the presence of MPT in the cells (24). Therefore, the MobA protein in E. coli, and the analogous MCD-synthesizing protein in H. pseudoflava are probably specific for Mo-MPT and are unable to attach the guanine or cytosine nucleotide to molybdenum-free MPT. Under the in vitro conditions described here, the pterin component acted upon by MobA is clearly Mo-MPT, since the cofactor is supplied by heat-denatured SO. This is supported by the inability of added molybdate to increase the extent of activation. A more complete model for cofactor biosynthesis and insertion can now be proposed for R. sphaeroides DMSOR. Molybdenum chelation to the pterin appears to precede dinucleotide formation. MobA then links the GMP moiety supplied by Mg-GTP to Mo-MPT to form Mo-MGD. Formation of Mo-MGD can proceed in the absence of an acceptor apoprotein. Subsequently, two Mo-MGD molecules are assembled to form bis(MGD)Mo, either on MobA or on the acceptor apoprotein. If both MGD moieties are already coordinated to molybdenum, the formation of bis(MGD)Mo would require elimination of one of these metal atoms. In the in vitro reconstitution procedure, neither MobB nor a DMSOR-specific chaperone is required in cofactor formation or insertion, although the presence of a chaperone may accelerate cofactor insertion.