In Vivo and in Vitro Function of theEscherichia coli Periplasmic Cysteine Oxidoreductase DsbG*

We have characterized in vivo andin vitro the recently identified DsbG fromEscherichia coli. In addition to sharing sequence homology with the thiol disulfide exchange protein DsbC, DsbG likewise was shown to form a stable periplasmic dimer, and it displays an equilibrium constant with glutathione comparable with DsbA and DsbC. DsbG was found to be expressed at approximately 25% the level of DsbC. In contrast to earlier results (Andersen, C. L., Matthey-Dupraz, A., Missiakas, D., and Raina, S. (1997) Mol. Microbiol. 26, 121–132), we showed that dsbG is not essential for growth and thatdsbG null mutants display no defect in folding of multiple disulfide-containing heterologous proteins. Overexpression of DsbG, however, was able to restore the ability ofdsbC mutants to express heterologous multidisulfide proteins, namely bovine pancreatic trypsin inhibitor, a protein with three disulfides, and to a lesser extent, mouse urokinase (12 disulfides). As in DsbC, the putative active site thiols in DsbG are completely reduced in vivo in adsbD-dependent fashion, as would be expected if DsbG is acting as a disulfide isomerase or reductase. However, the latter is not likely because DsbG could not catalyze insulin reductionin vitro. Overall, our results indicate that DsbG functions primarily as a periplasmic disulfide isomerase with a narrower substrate specificity than DsbC.

We have characterized in vivo and in vitro the recently identified DsbG from Escherichia coli. In addition to sharing sequence homology with the thiol disulfide exchange protein DsbC, DsbG likewise was shown to form a stable periplasmic dimer, and it displays an equilibrium constant with glutathione comparable with DsbA and DsbC. DsbG was found to be expressed at approximately 25% the level of DsbC. In contrast to earlier results (Andersen, C. L., Matthey-Dupraz, A., Missiakas, D., and Raina, S. (1997) Mol. Microbiol. 26, 121-132), we showed that dsbG is not essential for growth and that dsbG null mutants display no defect in folding of multiple disulfide-containing heterologous proteins. Overexpression of DsbG, however, was able to restore the ability of dsbC mutants to express heterologous multidisulfide proteins, namely bovine pancreatic trypsin inhibitor, a protein with three disulfides, and to a lesser extent, mouse urokinase (12 disulfides). As in DsbC, the putative active site thiols in DsbG are completely reduced in vivo in a dsbD-dependent fashion, as would be expected if DsbG is acting as a disulfide isomerase or reductase. However, the latter is not likely because DsbG could not catalyze insulin reduction in vitro. Overall, our results indicate that DsbG functions primarily as a periplasmic disulfide isomerase with a narrower substrate specificity than DsbC.
Disulfide bonds are essential for the correct folding and stability of many exocytoplasmic proteins (2). The oxidation of cysteine residues to form disulfide bonds can occur spontaneously in the presence of molecular oxygen. However, air oxidation is a slow, mechanistically complex reaction whose time scale is much longer than what is required for the folding of proteins under biosynthetic conditions, i.e. in the cell. As a result, both prokaryotic and eukaryotic cells have evolved elaborate enzymatic mechanisms for the catalysis of disulfide bond formation and for maintaining the proper thiol-disulfide redox balance in various cellular compartments.
In Gram-negative bacteria, disulfide bond formation normally occurs following export into the periplasmic space, which is topologically equivalent to the endoplasmic reticulum, albeit substantially more oxidizing (3,4). Genetic and biochemical studies have unequivocally defined four proteins (DsbA, DsbB, DsbC, and DsbD) that are involved in the formation of disulfide bonds in secreted proteins (5). All the known Dsb proteins contain a Cys-X-X-Cys sequence that is characteristic of the thioredoxin superfamily (5). DsbA is a soluble periplasmic enzyme that serves as a potent catalyst of protein and peptide cysteine oxidation (6 -8). It also exhibits some disulfide isomerization activity that may be important under some conditions in vivo (9 -11). Once it has transferred its disulfide bond to a substrate, DsbA is rapidly reoxidized by the membrane protein DsbB, which in turn transfers its electrons either to molecular oxygen or to the quinone system (12,13). The oxidation of protein cysteines by DsbA is very rapid but often results in the formation of incorrect disulfide bonds. The rearrangement of nonnative disulfides is catalyzed primarily by the dimeric periplasmic enzyme DsbC (8,14,15). For DsbC to be able to catalyze disulfide bond isomerization, its active site Cys-X-X-Cys sequence must be present in the dithiol form. Although the redox potentials of DsbA and DsbC are comparable, Ϫ89 mV (7) and Ϫ96 mV (calculated from Ref. 8), respectively, under steady state conditions in the periplasm DsbA is oxidized, whereas DsbC is almost exclusively reduced (10,16,17). Reduction of DsbC is mediated by cytoplasmic membrane protein DsbD and also depends on the cytoplasmic proteins thioredoxin (TrxA) and thioredoxin reductase (TrxB) (10,17).
In addition to dsbA, dsbB, dsbC, and dsbD, Missiakas and Raina (18) have isolated additional genes that affect sensitivity to dithiothreitol (DTT). 1 One of these genes was named dsbE and has been proposed to play a role in oxidative protein folding. However, dsbE is identical to ccmG, which has been shown to be an inner membrane protein involved in cytochrome biosynthesis in Escherichia coli and other bacteria (19 -21). It is not yet known whether DsbE affects the oxidation state of proteins other than cytochrome c. Recently, Raina and coworkers (1) isolated a gene that functioned as a multicopy suppressor of DTT sensitivity in a dsbB Ϫ background. The same gene (dsbG) was also isolated in a search for mutants conferring increased sensitivity to DTT and an increase in E -dependent periplasmic heat shock response. Interestingly, Andersen et al. (1) reported that dsbG is required for growth unless the cells are provided with exogenous oxidants. This is surprising, since none of the other dsb genes, including dsbA, which encodes the main catalyst of protein oxidation in the periplasm, is essential. In addition, Andersen et al. (1) reported * This work was supported by the National Science Foundation and National Institutes of Health (NIH) Grant GM-47520 (to G. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  that dsbG can catalyze insulin reduction in vitro and is partially responsible for the oxidation of alkaline phosphatase in the periplasm. We independently cloned and expressed DsbG and characterized its in vivo and in vitro function in detail. Contrary to the previous report, we show that dsbG is not an essential gene in E. coli; it is maintained by DsbD exclusively in reduced form, and it does not appear to affect protein oxidation in the periplasm. On the other hand, multicopy expression of dsbG suppressed the effect of a dsbC Ϫ mutation on the folding of heterologous multidisulfide substrates. DsbG was found to have a very unstable disulfide much like DsbA and DsbC, but unlike these two proteins it appears to have narrow substrate specificity. Our results indicate that DsbG functions predominantly either as a reductant or as a catalyst for disulfide isomerization.

EXPERIMENTAL PROCEDURES
Cloning and Expression-The strains and plasmids used in this work are listed in Table I. The 5-kbp Eco52I/SalI fragment containing the E. coli dsbG gene from Kohara clone 166 (22) was isolated and cloned into Eco52I/SalI-digested pBR322, generating plasmid pPBdsbG. The dsbG coding region was amplified from pPBdsbG by polymerase chain reaction (PCR) using the primers 5Ј-AGGAATTCAGGAGGTCTCTCATGT-TAAAAAAGATACTTTTAC-3Ј and 5Ј-CCATCCATGAGGATCCTTT-TATTTATTCCCCATAAT-3Ј. The PCR product was digested with BsaI and BamHI and ligated into NcoI/BamHI-digested pTrc99A (Amersham Pharmacia Biotech) or pET-11d (Novagen, Madison, WI), generating plasmids pTrcdsbG2 and pETdsbG2, respectively. Likewise, dsbG without its stop codon was amplified from pPBdsbG with the same forward primer and reverse primer 5Ј-CTAGAGGATCCTCGAGTTTAT-TCCCCATAATGATATT-3Ј, digested with BsaI and XhoI, and ligated into pET-28a (Novagen, Madison, WI) that had been cut with NcoI and XhoI. The resulting plasmid, designated pETdsbG2his, contains the coding sequence for DsbG fused to a 6ϫ histidine tag at its C terminus and under the control of the T7 promoter. Plasmid constructions were verified by automated DNA sequencing.
Expression and Purification of DsbG-Plasmid pETdsbG2his was transformed into strain BL21(DE3), and histidine-tagged DsbG was purified from the osmotic shockate using nickel-chelate chromatography, following standard protocols (Qiagen, Santa Clarita, CA). The purity of DsbG in the nickel-Sepharose™ eluant, was approximately 95% as judged by SDS-PAGE and Coomassie Brilliant Blue staining. This material was used to raise polyclonal antisera in mice using standard protocols (23). Rabbit polyclonal antisera against DsbA and DsbC were a gift of John Joly (Genentech, S. San Francisco, CA). Purified DsbA was purchased from Boehringer Mannheim. Purified DsbC was a kind gift of John Joly. All other chemicals were purchased from Sigma except as noted.
To purify DsbG expressed without a hexahistidine affinity tag, E. coli BL21(DE3) harboring the plasmid pETdsbG2 were grown in LB medium with 50 g/ml ampicillin at 37°C to midlog phase (A 600 ϳ 0.5). Cells were then transferred to a 25°C water bath and induced with 0.5 mM IPTG for 8 h to maximize the concentration of soluble, mature protein in the periplasm. Immediately after induction, the cells were collected by centrifugation, and periplasmic fractions were obtained by the cold osmotic shock procedure as modified by Thorstenson et al. (24). The periplasmic fraction was dialyzed against 30 mM Tris-HCl, pH 8.5, 0.5 mM EDTA, and 50 mM NaCl and applied to a DEAE anion exchange column (Bio-Rad), equilibrated with the same buffer. The column was developed with a linear gradient of NaCl from 50 to 300 mM. Analysis of the collected fractions by SDS-PAGE showed that DsbG eluted between 100 and 150 mM NaCl. The peak fractions were then pooled, concentrated, and dialyzed against 10 mM Tris-HCl, pH 8.0, 0.5 mM EDTA buffer and applied to a Sephadex TM -75 sizing column (Amersham Pharmacia Biotech) previously equilibrated with the same buffer. Densitometric analysis of a scanned SDS-PAGE gel revealed that following gel filtration the protein was at least 95% pure. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry was performed on a Perseptive Biosystems Voyager Biospectrometry Workstation with a mass resolution of greater than 95 and error of less than 0.07%.
Construction of dsbG Ϫ Mutant Strains-Plasmid pPBdsbG was digested with AflII and Sse8387I; the large fragment was isolated, and its ends blunted with T4 polymerase. This fragment was then ligated with either the 1.4-kbp BsaAI fragment of pACYC184 containing the chloramphenicol resistance gene to generate pPB⌬dsbG::Cm or the 1.3-kbp kanamycin resistance cassette from pUC-4K to generate pPB⌬dsbG::Kan. Plasmid pPB⌬dsbG::Cm was linearized and transformed into the recD Ϫ strain D301 (25), and Cm R , Amp S colonies were selected. Additionally, plasmid pPB⌬dsbG::Kan was digested with MluI and SphI, and the fragment containing the Kan R cassette with the dsbG chromosomal flanking regions was recovered and cloned into pBAD39, a vector containing a conditional IPTG-requiring replicon, an ampicillin resistance gene, and a wild type rpsL gene (Table I). MC4100 cells transformed with the resulting plasmid, pBAD⌬dsbG::Kan, were grown first in liquid medium without IPTG and then plated on LB agar containing kanamycin, and streptomycin for counterselection. Colonies that were Kan R , Strep R , and Amp S were selected, and correct constructs were confirmed both by immunoblotting to verify the absence of a DsbG band and by PCR using primers flanking the deletion region. The ⌬dsbG::Cm and ⌬dsbG::Kan mutations were transduced into different strains using P1vir following standard protocols (26).
Expression of BPTI and Urokinase-Bovine pancreatic trypsin inhibitor (BPTI) expression was monitored by enzyme-linked immunosorbent assay as described previously (27). Urokinase activity was detected by indirect chromogenic assay as follows. Cultures were grown at 37°C to A 600 ϭ 0.7, and protein synthesis was induced by adding IPTG to 1 mM. The cells were harvested 3 h later and lysed by French ® pressure cell. Following centrifugation to remove insolubles, the total soluble protein was quantified by the Bradford assay (Bio-Rad), using bovine serum albumin as a standard, and then diluted to 0.3 g/l in 50 mM Tris-HCl, pH 7.4, 0.01% Tween 80. In a microtiter plate, 50 l of whole cell lysate soluble fraction was mixed with 50 l of human plasminogen (Calbiochem), 0.1 g/l in 50 mM Tris-HCl, pH 7.4, 0.01% Tween 80, 6 mM 6-aminohexanoic acid. The plasminogen substrate, 50 l of 4 mM Spectrozyme ® PL (American Diagnostica, Greenwich, CT), was added immediately, and the plate was incubated at room temperature for 60 min. The absorbance at 405 nm was measured and, after subtracting the background activity of a strain not expressing urokinase, compared with a standard curve prepared using human urokinase (Calbiochem).
Redox Properties of DsbG-The in vivo redox state of DsbG was assayed by derivatization of free thiols by 4-acetamido-4Ј-maleimidylstilbene-2,2Ј-disulfonic acid (AMS) (Molecular Probes, Inc., Eugene, OR) under denaturing conditions essentially as described for DsbC (10). The oxidized standard was purified DsbG without AMS added. The reduced standard was generated by reducing purified DsbG in 40 mM DTT for 20 min at room temperature followed by removal of reducing agent using a gel filtration spin column (Bio-Rad) and AMS derivatization.
The redox equilibrium of DsbG with glutathione was assayed as described for DsbA (7). In this assay, the change in fluorescence intensity (excitation wavelength 280 nm) was measured at the wavelength of maximum emission (330 nm for DsbG and 324 nm for DsbA). Experiments were carried out in 100 mM sodium phosphate, pH 7.0, and 1.0 mM EDTA. Oxidized DsbG or DsbA (0.45 M) was incubated at 25°C in the presence of 0.1 mM GSSG and 0 -2.0 mM GSH for 12 h before recording the fluorescence emission in an SLM-Aminco Luminescence Spectrometer (series 2). The equilibrium concentrations of GSH and GSSG were calculated according to Equations 1-3, where [GSH] 0 and [GSSG] 0 are the initial concentrations of GSH and GSSG, R is the relative amount of reduced protein at equilibrium, [Dsb] 0 is the initial concentration of DsbG or DsbA in the oxidized form, F is the measured fluorescence intensity, and F ox and F red are the fluorescence intensities of completely oxidized and reduced protein. The equilibrium constant K eq was estimated from nonlinear regression analysis of the data according to Equation 4 (28).
Chemical Cross-linking-Chemical cross-linking was used to determine the oligomeric state of DsbG. The reactions were performed by adding 0.1 volumes of Me 2 SO containing various concentrations (0 -5 mM) of the amine cross-linker ethylene glycol-bis(succinimidylsuccinate) (EGS) (Pierce) to protein solutions containing 250 g/ml of purified DsbG or DsbA. Each reaction was incubated at 4°C for 30 min and then quenched by the addition of glycine to 75 mM. The samples were resolved by SDS-PAGE (10 -20% resolving gels), and the cross-linked products were visualized by Coomassie Brilliant Blue staining.
In Vitro Oxidoreductase Activity Assays-The ability of DsbA, DsbC, and DsbG to catalyze the reduction of human insulin (Sigma catalog no. I-5523) in the presence of DTT was tested as described previously by Holmgren (29). A stock solution of 5 mM insulin was freshly prepared in 0.1 M potassium phosphate buffer, pH 7.0, and 2 mM EDTA before each assay. The reaction mixtures were prepared directly in cuvettes using 0.1 M potassium phosphate buffer, pH 7.0, 2 mM EDTA, 131 M insulin, and various concentrations of Dsb proteins (typically between 5 and 15 M) in a final volume of 0.8 ml. The reactions were started by adding DTT to a final concentration of 0.35 mM. After thorough mixing, the cuvettes were placed in the spectrophotometer, and measurements were performed at 650 nm every 30 s. In all of the experiments, the uncatalyzed reduction of insulin by DTT was monitored in a control reaction without the addition of Dsb proteins.
The preparation of fully reduced, denatured RNase and refolding assays were performed as described previously (30). The recovery of ribonuclease activity resulting from oxidative renaturation was measured in a UV-visible spectrophotometer using cCMP as a substrate for the RNase. The reaction mixtures consisted of 50 mM Tris acetate buffer, pH 8.0; 4.5 mM cCMP; GSH and GSSG (at predetermined concentrations to provide a redox buffer); pure DsbG or DsbC at concentrations ranging from 1 to 20 M or 1.5 M bovine PDI as a positive control. After equilibrating the reaction at 25°C, the assay was initiated by the addition of reduced, denatured RNase to a final concentration of 8 M. Hydrolysis of cCMP by the refolded RNase was monitored every 30 s for 1 h as an increase in the absorbance at 296 nm. As a negative control, the uncatalyzed reaction was recorded in parallel under identical redox conditions. The concentration of active RNase at any time in each assay was calculated as described by Lyles and Gilbert (30). (31) search of the E. coli genome indicated the presence of an open reading frame, 2 having 49% similarity and 30% identity to 220 residues of the E. coli DsbC protein. Because of the high degree of similarity between DsbC and the identified hypothetical protein, we reasoned that it may have a role in the formation of disulfide bonds in the periplasmic space. While this work was in progress, Raina and co-workers (1) identified the same open reading frame in a genetic screen for resistance to dithiothreitol and named it dsbG (GenBank TM accession no. AF000956). We cloned the complete 248-amino acid coding region of dsbG into a T7 expression vector fused to a 6ϫ histidine tag and purified the expressed protein by immobilized metal affinity chromatography. The N-terminal sequence of the mature purified protein was verified by automated Edman degradation, and cleavage of the signal peptide was shown to occur after Ala 17 . Antibodies were raised against the resulting material and used to probe a Western blot of whole cell extracts of E. coli not carrying any plasmids. A band of the expected mobility could be detected in lysates of exponential phase cells growing This study aerobically in rich media, but the corresponding band was not detected in strains in which the dsbG coding region had been deleted (see below). These results demonstrate that dsbG is indeed normally expressed in wild type cells under these conditions. The relative amounts of DsbG, DsbC, and DsbA in exponentially grown cells were determined by quantitative Western blotting, using serial dilutions of the respective purified proteins as standards. Immunoreacting proteins were visualized by ECL TM and quantified by densitometry. It was found that in strain MC4100, DsbG is present at approximately one-fourth the level of DsbC, which, in turn, is approximately 6-fold less abundant than DsbA (not shown).

Cloning and Expression of dsbG-A BLAST
Construction and Characterization of dsbG Null Mutants-In earlier studies Andersen et al. (1) reported that a dsbG::⍀Tet mutation could be crossed onto the chromosome only when the cells were grown in the presence of low molecular weight oxidants such as cystine or oxidized DTT. Despite finding that a dsbG::⍀Kan mutation could be transduced without supplementation of oxidants, they nevertheless hypothesized that this observation resulted from the accumulation of second site suppressor mutations. On the basis of these results, they concluded that mutations in dsbG are conditionally lethal. To evaluate this hypothesis, we first constructed a large deletion in dsbG comprising 187 codons and marked it by the insertion of fragments of either 1.4 or 1.3 kbp containing, respectively, a Cm R or a Kan R gene. First, the Cm R -marked dsbG deletion was integrated into the chromosome by homologous recombination in a recD Ϫ strain. Hundreds of Cm R , Amp S colonies were obtained, as expected for a gene that is not essential for viability. Colonies were picked at random and proven to contain ⌬dsbG::Cm by PCR using the appropriate primers. Subsequently, the allele was transferred to different strains by P1 transduction.
In a second approach, the ⌬dsbG::Kan mutation was inserted into the suicide vector pBAD39 that can only replicate in cells grown in the presence of IPTG. The plasmid pBAD⌬dsbG::Kan was transformed into MC4100, and cells that had lost the plasmid and had recombined the ⌬dsbG::Kan allele on the chromosome were selected for resistance to kanamycin and streptomycin and sensitivity to ampicillin on plates that also lacked IPTG. Kan R , Strep R , and Amp S colonies were tested by PCR and were all found to carry the expected deletion in dsbG. In order to test whether suppressor mutations had arisen in our dsbG Ϫ strains, we transduced the ⌬dsbG::Kan mutation to either a wild type strain or the same strain carrying dsbGϩ on a multicopy plasmid (pTrcdsbG2). For comparison, the same procedure was repeated in order to recombine the degP41 mutation, a nonlethal Kan R -marked deletion in the gene encoding the periplasmic protease DegP (32), into the chromosome of the same two acceptor strains. The ratio of Kan R transductants in RI89 to RI89 (pTrcdsbG2) was the same (3:2) regardless of whether ⌬dsbG::Kan or the unrelated degP41 mutation was transduced. If the introduction of a dsbG gene disruption in the haploid strain were lethal, and growth could only occur via the accumulation of second site suppressor mutations, as suggested by Andersen et al. (1), then we would expect that transduction of the ⌬dsbG::Kan allele into RI89 to result in a significantly lower number of Kan R colonies relative to the number of colonies obtained by transducing the unrelated degP41 allele. This, however, was not the case. In summary, the above results clearly show that dsbG is not necessary for growth in RI89. Also, the frequency with which dsbG Ϫ transductants could be obtained in MC4100, RI90, and SR3324 strains suggests that dsbG is not essential in these genetic backgrounds either. The dsbG mutation was found to have no effect on alkaline phosphatase activity, growth at 42°C, or the growth rate in rich or minimal media (data not shown).
The dsbG deletion mutants were also tested for their effect on the folding of heterologous proteins containing multiple disulfide bonds, such as bovine pancreatic trypsin inhibitor (BPTI) and mouse urokinase. BPTI is a 6.5-kDa protein with three disulfides, whereas mouse urokinase is a 48-kDa protein containing 12 disulfides. The folding of both proteins had been shown to be strongly dependent on the presence of functional dsbA, dsbB, dsbC, and dsbD genes (14,17,33,34). However, the levels of folded BPTI and active urokinase in the dsbG Ϫ strain were identical to their wild type counterparts (data not shown).
We furthermore examined the effect of overexpressed DsbG on disulfide formation in heterologous proteins. While overexpression of dsbG from the trc promoter in a wild type strain could not improve the yield of BPTI or urokinase, it could complement the folding defect in a dsbC Ϫ strain. In the case of BPTI, the yield of native protein was restored to wild type levels in the dsbC Ϫ strain; dsbG overexpression could not, however, complement a null mutation in dsbD (Fig. 1A). With urokinase, the activity was only restored to approximately 15% of the wild type level (Fig. 1B).
In Vivo Redox State-Unlike DsbC, which contains a struc- tural disulfide in addition to its active site disulfide, mature DsbG has a single pair of cysteine residues. We examined the in vivo redox status of the active site disulfide in DsbG under normal growth conditions in wild type cells as well as in null mutants of dsb family genes. Briefly, exponentially grown cells were lysed in the presence of 10% trichloroacetic acid to denature DsbG while preventing the rearrangement of free thiols. Subsequently, free thiols were blocked with AMS, and the AMS-conjugated (reduced) and unconjugated (oxidized) proteins were resolved by SDS-PAGE under nonreducing conditions (10,12,35). We found that at steady state DsbG is present exclusively in reduced form in wild type cells and in dsbA Ϫ , dsbB Ϫ , and dsbC Ϫ mutants. In contrast, in dsbD Ϫ cells the protein was completely oxidized (Fig. 2A). Furthermore, there was no significant change in the expression level of DsbG in any of the dsb mutants. Upon introduction of a plasmid overexpressing DsbG from an inducible promoter, a small amount of DsbG is found to be oxidized. The reduced state, however, is overwhelmingly favored in all strains except, of course, the dsbD null (Fig. 2B).
In Vitro Characterization of DsbG-DsbG was purified from the periplasmic fraction of a culture harboring a plasmid containing the dsbG gene under the control of the T7 promoter. The overexpressed periplasmic protein was extracted by cold osmotic shock (Fig. 3A, lane 3) and partially purified by DEAE anion exchange chromatography, yielding a preparation that contained more than 90% pure DsbG protein (Fig. 3A, lane 4). To further remove other protein contaminants, peak fractions were loaded onto a Sephadex TM -75 size exclusion column. Fractions containing DsbG eluted slightly ahead of the ovalbumin (44 kDa) marker and yielded a preparation that was over 95% pure (Fig. 3A, lane 5). The molecular weight of the purified protein was determined by matrix-assisted laser desorption/ionization time-of-flight spectrometry and shown to agree with the calculated molecular weight within the error of measurement.
Upon gel filtration, DsbG eluted as a 45-50-kDa protein, raising the possibility that DsbG (subunit molecular mass 25.7 kDa) may form a homodimer as has been shown for DsbC (8).
As an independent means to establish the quaternary structure of DsbG, purified protein was subjected to cross-linking with EGS, and the migration of the complex was analyzed by SDS-PAGE and Coomassie Brilliant Blue staining (Fig. 3B). These data revealed that, indeed, with increasing concentration of EGS, the electrophoretic mobility of DsbG shifted to approximately twice the expected 25.7-kDa mass of the monomer (Fig.  3B, lanes 3-5). As a negative control, purified DsbA, known to be a monomer, did not shift in mobility when subjected to the same concentrations of cross-linking agent (data not shown). These results suggest that DsbG exists predominantly in the homodimer form in the periplasmic space of E. coli, which is consistent with its estimated molecular weight from gel filtration analysis.
It has been previously shown that the fluorescence emission spectra of thioredoxin and DsbA are strongly dependent on the redox states of the enzymes (7, 36 -38). The fluorescence intensity at 324 nm of reduced DsbA at pH 7.0 is 3-fold higher than that of the oxidized protein. Reduced DsbG likewise exhibits increased fluorescence emission at 330 nm at pH 7.0, compared with the oxidized form (Fig. 4A), but the difference is smaller FIG. 2. In vivo redox state of DsbG active site. Whole cells were precipitated with 10% trichloroacetic acid, and free thiols were derivatized with AMS, which adds 490 Da per thiol reacted, therefore reducing the electrophoretic mobility. Proteins were separated by 12% SDS-PAGE and visualized by Western blotting using DsbG antisera with ECL TM (Amersham Pharmacia Biotech) development. Strains used are as follows: RI89 (wt), RI90 (dsbA Ϫ ), RI317 (dsbB Ϫ ), RI179 (dsbC Ϫ ), RI242 (dsbD Ϫ ), PB303 (dsbG Ϫ ). A contains proteins from cells not carrying any plasmids; B contains proteins from cultures carrying the plasmid pTrcdsbG2, with dsbG induction by 1 mM IPTG for 30 min before trichloroacetic acid precipitation. Oxidized and reduced standards were generated from purified DsbG without AMS derivatization (ox.) or subjected to reduction by 40 mM DTT followed by AMS derivatization (red.). than that of DsbA. In contrast, there is no difference in the spectroscopic properties of active site reduced and oxidized DsbC (8).
The observed differences in fluorescence intensities of reduced or oxidized DsbA and DsbG were used to measure the equilibrium concentrations of the oxidized and reduced forms in the presence of different ratios of oxidized (GSSG) and reduced (GSH) glutathione. The DsbG/glutathione redox equilibrium and its equilibrium constant are given by Equations 5 and 6. DsbG red ϩ GSSG L | ; K eq DsbG ox ϩ 2GSH (Eq. 5) K eq ϭ ͓DsbG ox ͔͓GSH͔ 2 /͓DsbG red ͔͓GSSG͔ (Eq. 6) DsbA or DsbG was incubated in the presence of 100 M GSSG and increasing concentrations of GSH (0 -2 mM), and the relative amount of reduced enzyme at equilibrium (R) was measured over the range from fully oxidized to fully reduced protein (Fig. 4B). Nonlinear regression was used to fit the data to Equation 4. The equilibrium constant K eq for the reduction of DsbA by glutathione was determined to be 130 M. This result agrees well with the values of 120 and 80 M reported by Wunderlich and Glockshuber (7) and Zapun et al. (6), respec-tively, for DsbA. Under identical conditions, the equilibrium constant for DsbG was estimated to be 140 M. Thus, the DsbG disulfide bond is nearly as unstable as that of DsbA and slightly less so than that of DsbC (K eq ϭ 200 M) (8). The general behavior of the DsbG titration by glutathione indicates that its redox potential is comparable with those of DsbA and DsbC.
The ability of DsbG to catalyze the reduction of insulin in the presence of DTT was evaluated. Reduction of insulin leads to cleavage of the two interchain disulfide bonds, causing the ␤-chain of insulin, which is insoluble, to aggregate and form a precipitate whose formation can be recorded by measuring the turbidity in the sample at 650 nm. Assays were performed exactly as described by Holmgren and co-workers (29). As expected, the addition of 5.0 M DsbC or DsbA resulted in the rapid precipitation of insulin, which could be detected after 6 min for DsbC and 18 min for DsbA (Fig. 5A). At lower concentrations of either protein (2.5 M), the time at which the turbidity could initially be detected was increased, and the precipitation rate was slower (data not shown). Andersen et al. (1) had reported that DsbG has insulin reduction activity comparable with that of DsbA. However, in our hands DsbG was not able to reduce insulin above the background levels observed in the control reaction with DTT alone (Fig. 5A). No activity was observed with different preparations of DsbG expressed with or without a hexahistidine tail at 25 or 37°C (Fig. 5A) or with high concentrations of DsbG (up to 20 M). We also failed to see any activity with protein that had been prereduced with 15 mM DTT (data not shown). On the basis of these results, we concluded that DsbG is unable to catalyze the reduction of insulin, at least under the standard assay conditions.
We also examined whether DsbG could catalyze the oxidative renaturation of reduced, denatured RNase (30). The addition of reduced, denatured RNase to a glutathione redox buffer containing a catalytic protein and the RNase substrate cCMP results in an increase of RNase activity, which is monitored as an increase in the rate of cCMP hydrolysis measured at 296 nm. The concentration of active RNase can be calculated at any time from the instantaneous slope of the absorbance versus time trace after considering the change in cCMP concentration and the formation of the RNase inhibitor CMP (for calculations, see Lyles and Gilbert (30)). In a 60-min assay, there is no substantial gain in RNase activity with catalytic concentrations of enzyme in the absence of glutathione redox buffer, suggesting that oxygen-dependent activation of RNase does not provide a significant source of oxidizing equivalents under these conditions. Consistent with previous studies, bovine PDI was shown to catalyze the oxidative refolding of RNase (Fig. 5B). Similarly, DsbC was also efficient in catalyzing the reactivation of RNase. The k cat and K m values for this reaction were 0.24 Ϯ 0.01 min Ϫ1 and 16 Ϯ 3 M, which are 3-fold slower and 2-fold higher, respectively, than those reported for PDI (39). The presence of DsbG (5 M) resulted in a rate of RNase reactivation that was reproducibly slightly above background. Identical results were obtained with either native DsbG or DsbG⅐His-tagged protein (Fig. 5B). These results indicate that DsbG is a poor catalyst of oxidative protein refolding, at least using RNase as the substrate. Furthermore, DsbG (5 M) could not act synergistically with PDI or DsbC, failing to give a reactivation rate higher than PDI or DsbC alone. FIG. 5. Oxidoreductase activity of DsbG compared with other disulfide catalysts. A, the enzymatic reduction of insulin in the presence of DTT was assayed by turbidimetric assay as described previously (29). Reactions were performed in a final volume of 800 l containing 0.1 M potassium phosphate buffer, pH 7; 2 mM EDTA; 0.35 mM DTT; 0.13 mM insulin; and purified DsbA, DsbC, or DsbG at 5 M. The background reduction by DTT without any enzyme present was used as a control. B, analysis of the kinetics of oxidative refolding of RNase. The concentration of refolded RNase was determined by measuring the absorbance change at 296 nm produced by RNasecatalyzed hydrolysis of cCMP (initial concentration 4.5 mM) at pH 8.0 in 50 mM Tris acetate buffer (see "Experimental Procedures") in the presence of redox buffer (0.2 mM GSSG, 1.0 mM GSH). Concentrations of PDI, DsbC, DsbG, and DsbG⅐His are indicated in the key. Reactions were started with the addition of 8 M reduced, denatured RNase.

DISCUSSION
We first identified DsbG based on its homology with E. coli DsbC. A data base search also revealed that DsbG is highly similar to an unidentified open reading frame 3 from Pseudomonas aeruginosa (47% identity, 70% similarity, allowing for conservative substitutions), distinct from the putative DsbC (GenBank TM accession no. AF057031) of P. aeruginosa, with the two exhibiting similarity (25% identity, 43% similarity) comparable with that between the E. coli DsbC and DsbG.
The E. coli dsbG gene was shown to encode a dimeric periplasmic protein, which in exponentially grown cells is expressed at a level about 25% that of DsbC. The two proteins share a number of similarities; they are both dimeric, form an unstable disulfide bond, and are maintained in the reduced state in the periplasm via the action of DsbD (10,17). High level expression of DsbG could fully complement the defect in the folding of BPTI and partially complemented the formation of active urokinase in a dsbC Ϫ mutant background. In vitro, the folding pathway of BPTI has been shown to involve the formation and subsequent rearrangement of intermediates with nonnative disulfide bonds. There is evidence that DsbC catalyzes kinetically important disulfide bond isomerization steps in the folding of BPTI in the periplasm and most likely has a similar effect on urokinase (17,34). The ability of DsbG to fully restore the expression of BPTI and partially restore that of urokinase suggests that DsbG also is able to effect disulfide bond isomerization. Remaining to be determined is whether DsbG is a bona fide catalyst of disulfide isomerization or whether its overexpression acts indirectly, for example by altering the redox state of the periplasm.
DsbG contains a Cys-X-X-Cys sequence that is found in the active site of proteins belonging to the thioredoxin superfamily. This is consistent with DsbG's being a redox active protein that may play a role in disulfide bond isomerization. The Cys-Pro-Tyr-Cys sequence in DsbG forms an unstable disulfide bond that is readily reduced by glutathione. Interestingly, DsbA, DsbC, and DsbG all exhibit equilibrium constants with glutathione in the 80 -200 M range (6 -8), considerably lower than that of thioredoxin (10 M) (40). Despite the highly oxidizing nature of the periplasmic space, DsbG is found exclusively in the reduced state in wild type cells. Only under conditions of overexpression did we observe a small amount of oxidized protein (Fig. 2B). On the other hand, DsbG was completely oxidized in a dsbD Ϫ background, suggesting that, in the absence of a membrane protein reductant, DsbG becomes completely oxidized.
While this work was in progress, Andersen et al. isolated dsbG genetically and examined its function in vivo (1). We have found several discrepancies between our results and those of Andersen et al. First of all, they reported that a dsbG null linked to Tet R could not be crossed onto the chromosome unless the cells were supplemented with high concentrations of low molecular weight oxidants. Although a dsbG::⍀Kan null allele could be transduced without supplementation by oxidants, they attributed the appearance of transductants to suppressor mutations and went on to conclude that dsbG is essential. In contrast, we found the following: 1) a Kan R -or a Cm R -marked null allele could be transduced from a recD Ϫ strain into a variety of recipient strains at a normal frequency; 2) a null allele could be readily recombined into the chromosome from a suicide plasmid, again at a high frequency; and 3) finally, confirming the absence of second site suppressors, we found that the frequencies with which the ⌬dsbG allele can be transduced to a wild type background with or without dsbGϩ present in multicopy are comparable. These results prove unequivocally that dsbG is not required for growth.
In contrast to our results in Fig. 2A, Andersen et al. (1) reported that DsbG is present as a mixture of oxidized and reduced protein in vivo. There are two possible reasons for this discrepancy. First, in their experiments they expressed DsbG from a T7 promoter. As shown in Fig. 2B, under conditions of overexpression a small portion of DsbG is indeed oxidized. Second, they used a methodology for trapping free thiols that has been shown to be subject to artifacts (16).
Another discrepancy between our results and those of Andersen et al. (1) is that in our hands DsbG exhibited no insulin reduction activity, although different high purity protein preparations were tested. Furthermore, DsbG exhibits substantially lower activity than DsbC in catalyzing the in vitro refolding of RNase, indicating that it is a poor protein oxidant, at least for this substrate.
The inability of DsbG to catalyze the reduction of insulin or the oxidative refolding of RNase in vitro or to restore the defect in alkaline phosphatase formation in a dsbA mutant background (data not shown) argues that its biological function is to facilitate disulfide bond isomerization rather than the oxidation of protein thiols or net disulfide bond reduction. Consistent with this hypothesis, DsbG is maintained in a reduced state in the periplasm in a manner analogous to DsbC. Also, the fact that DsbG can partially rescue the defect in the formation of active multidisulfide proteins in a dsbC Ϫ background strongly suggests that it can play a role in protein folding in vivo. In agreement with this conclusion, Andersen et al. (1) reported that the synthesis of an htrA-lacZ fusion, which is dependent on the E factor, is restored to wild type levels by overexpression of DsbG in a dsbC Ϫ background. Induction of a E -dependent heat shock response has been shown to occur under conditions that lead to the accumulation of misfolded periplasmic proteins (41). In our hands, the deletion of dsbG had no negative effect on expression of disulfide-containing proteins in the periplasm, indicating that it may be redundant under the conditions tested or have a limited set of substrates for which it is necessary. Studies to identify the in vivo substrates of DsbG in the bacterial periplasm are under way.