Differential in vivo roles played by DsbA and DsbC in the formation of protein disulfide bonds.

Several Escherichia coli proteins participate in protein disulfide bond formation. Among them, DsbA is the primary factor that oxidizes target cysteines. Biochemical evidence indicates that DsbC has disulfide isomerization activity. To study intracellular functions of DsbA and DsbC, we used an alkaline phosphatase mutant, PhoA[SCCC], with the most amino-terminal cysteine replaced by serine. It was found that the remaining 3 cysteines in PhoA[SCCC] form a disulfide bond of incorrect as well as correct combinations. An aberrant disulfide bond was preferentially formed in wild-type cells, which was converted slowly to the normal disulfide bond. This conversion did not occur in the dsbC-disrupted cells. Overproduction of DsbC stimulated the formation of the correct disulfide bond. In contrast, the inefficiently formed disulfide bonds in the dsbA-disrupted cells, and the more efficiently formed disulfide bonds in the same strain in the presence of oxidized glutathione were mostly in the correct form. These results suggest that the DsbA-catalyzed reaction can be too rapid for some proteins. DsbA may simply oxidize available pairs of cysteines, which happen to be in an incorrect combination in the case of PhoA[SCCC]. In contrast, DsbC stimulates the formation of correct disulfide bonds and corrects previously introduced aberrant ones. Thus, DsbC acts to isomerize disulfide bonds in vivo.

Disulfide bonds are found in many extracytosolic proteins in all organisms and contribute to folding and stability of these proteins. While disulfide bond formation is a simple reaction of oxidation of cysteine residues, and it can be reproduced in vitro under appropriate conditions (1), recent studies established that it does not occur effectively in vivo without the aid of other proteins (2). In Escherichia coli, a periplasmic protein, DsbA, is required for disulfide bond formation in vivo (3,4). It directly oxidizes cysteines on the target proteins in vitro (5,6). It has a thioredoxin-like Cys 30 -X-X-Cys motif characteristically found in disulfide oxidoreductases (7).
DsbB, an integral membrane protein, is also required for the processes (8). The role of DsbB is to reoxidize DsbA to enable its catalytic turnover (8 -11). Genes dsbC and dsbD (dipZ) also encodes factors involved in disulfide bond metabolism (12)(13)(14)(15). DsbD is a membrane protein with a thioredoxin-like motif in the periplasmic domain, and it may have a regulatory role of conferring reducing power to the periplasm. DsbC is a periplasmic protein with 4 cysteines among which Cys 98 and Cys 101 forms a thioredoxin-like motif. Creighton and his colleagues (16) characterized the redox activity of DsbC using a model substrate. They showed that while DsbA merely oxidized cysteines on the substrate, DsbC efficiently isomerized preformed disulfide bonds. Bacterial alkaline phosphatase, a periplasmic protein, is a dimer of the phoA gene product (PhoA) with two intramolecular disulfide bonds (Cys 168 -Cys 178 and Cys 286 -Cys 336 ) (17). Disulfide bond formation is essential for the correct folding of this enzyme (3,4,18,19). We found that, of the two disulfide bonds in PhoA, the carboxyl-terminal one (Cys 286 -Cys 336 ) is required and sufficient for the active conformation of this enzyme (20). Thus, a mutant form of PhoA, termed PhoA[SSCC], with the two NH 2 -terminally located cysteines replaced by serine is as active as the wild-type enzyme, although it is no longer resistant to a protease. Interestingly, the presence of an additional cysteine at residue 178 lowered the enzymatic activity significantly (20). We show here that this mutant PhoA, termed PhoA[SCCC], forms an aberrant disulfide bond among Cys 178 , Cys 286 , and Cys 336 .
Using this unique experimental system, we investigated into the in vivo roles played by DsbA and DsbC. It was found that DsbA principally introduced an aberrant disulfide bond into PhoA[SCCC], whereas DsbC stimulated the eventual formation of the correct disulfide bond in vivo. Thus, DsbC functions, in concert with DsbA, as a disulfide isomerase in vivo.

EXPERIMENTAL PROCEDURES
E. coli Strains and Plasmids-Strain MS3 was a ⌬phoA strain, KS272 (21), into which FЈlacI Q lacPL8 LacZ ϩ Y ϩ A ϩ pro ϩ had been introduced (20). MS4 was a dsbA-33::Tn5 (4) transductant of MS3. As a dsbC deletion strain we used W3110 tonA ⌬dsbC, which was kindly provided by John Joly of Genentech. This strain had been constructed by integration and segregation of a plasmid carrying ⌬dsbC, and the dsbC deletion was confirmed by polymerase chain reaction analyses. 1 An isogenic dsbC ϩ strain, W3110 tonA, was also provided by J. Joly of Genentech.
PhoA and its Cys/Ser mutant forms were designated by the four letter notations in brackets, with C for Cys and S for Ser, for residues 168, 178, 286, and 336 in this order (20; numbering according to Ref. 17). They were expressed from the plasmids under the control of the lac operator/promoter (20); pMS002 for wild-type PhoA, pMS003 for PhoA [SCCC], pMS004 for PoA [SSCC], and pMS015 for PhoA [CCSS]. pMS022 was a DsbC-overproducing plasmid. For its construction, a 1.0-kilobase pair SacI-KpnI fragment, containing dsbC (including its own promoter), was excised from pDS30 (14) and cloned into the SmaI site of pSTV28 (a pACYC184-based lac promoter vector; Ref. 22).
Examination of Disulfide Bonds in PhoA-To examine PhoA molecules at steady states, cells were grown at 30°C to an exponential phase in L broth (10 g of tryptone, 5 g of yeast extract, 5 g of NaCl, and 1.7 ml of 1 NN NaOH/liter) supplemented with 1 mM IPTG 2 and appropriate antibiotics. A 200-l portion was mixed with an equal volume of 10% trichloroacetic acid. Protein precipitates were collected by centrifugation, washed with acetone, and dissolved in SDS/Tris-HCl solution * This work was supported by grants from the Ministry of Education, Science and Culture, Japan. 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 U.S.C. Section 1734 solely to indicate this fact.
To follow the biosynthesis and conversion of different isoforms, cells were grown at 30°C to an exponential phase in M9 medium (26) supplemented with 0.4% glycerol, appropriate antibiotics, and 20 g/ml each of amino acids (except methionine and cysteine). Fifteen minutes after addition of 1 mM IPTG, cells were pulse-labeled with 50 Ci/ml [ 35 S]methionine (1100 Ci/mmol, American Radiolabeled Chemicals) for 30 s, followed by chase with unlabeled L-methionine (200 g/ml) for indicated periods. Whole cell proteins were immediately precipitated with trichloroacetic acid and dissolved in the SDS/Tris-HCl/iodoacetamide solution as described above. Radioactive PhoA was immunoprecipitated (4), electrophoresed, and visualized using a Bioimaging Analyzer BAS2000 (Fuji Film). These results indicate that Cys 286 -Cys 336 disulfide bond mainly contributes to the increased electrophoretic mobility of the oxidized PhoA molecule. We designate this electrophoretic mobility "ox1" (Fig. 1). PhoA[SCCC] produced two bands when expressed in wild-type cells (Fig. 1, lane 4). The minor band was at the ox1 position, whereas the major band migrated even faster than ox1. The latter mobility is designated "ox2" (Fig. 1,  lane 4). Obviously, the former should represent Cys 286 -Cys 336 disulfide-bonded molecules. The latter species was not due to a proteolytic cleavage, since reduced PhoA[SCCC] migrated as a single band at the position ("red" in Fig. 1) identical to the reduced wild-type PhoA (Fig. 1, lanes 9 and 10). These results indicate that the ox2 form of PhoA[SCCC] contains an aberrant disulfide bond, between Cys 178 and either Cys 286 or Cys 336 .

DsbA-dependent Preferential Introduction of an Aberrant Disulfide Bond to PhoA[SCCC]-PhoA[SCCC]
was expressed in dsbA ϩ cells and dsbA-disrupted and dsbA-disrupted cells growing in broth medium, and its electrophoretic mobilities were examined by immunoblotting. Whereas the ox2 isoform was the major product in the wild-type cells, the ox1 isoform became the major accumulated product in the dsbA Ϫ strain (Fig. 1,  compare lanes 4 and 5). Apparently, the correct disulfide bond was preferred in the absence of DsbA. The enzymatic activity of PhoA[SCCC] was higher when produced in the dsbA-disrupted cells than in the dsbA ϩ cells. 3 We then studied the synthesis and conversion of the PhoA-[SCCC] isoforms by pulse-chase/immunoprecipitation experiments, using cells growing in minimal salt medium. In wildtype cells, the ox1 and ox2 forms of PhoA[SCCC] were initially labeled in about equal intensities. During chase with unlabeled methionine, intensity of the ox2 form decreased with concomitant increase in the ox1 form (Fig. 2, lanes 1-9). This ox2 to ox1 conversion took place over some 1 h.
When PhoA[SCCC] was expressed in the dsbA-disrupted cells, the majority of the newly synthesized molecules were now in the reduced form, which was slowly degraded (Fig. 2, lanes 10 -18). A small amount of oxidized form in this strain was in the ox1 form, and no aberrant form (ox2) was detected (Fig. 2,  lanes 10 -18). The experiments reported in Fig. 1 (lane 4) and Fig. 2 (lanes 10 -18) gave different proportions of PhoA [SCCC] isoforms for the dsbA Ϫ strain. This may be explained by the presence of some broth components, such as cystine, that may have acted as an oxidant in the former experiment (see below for oxidant effects). In addition, instability of the reduced form may have lowered the detection by immunoblotting.
We examined effects of oxidized glutathione (GSSG) added to the minimal medium (Fig. 3, E and G). In the presence of GSSG, PhoA [SCCC] was almost all in the ox2 form in the wild-type cells (Fig. 3E), whereas it was almost all in the ox1 form in the dsbA Ϫ mutant (Fig. 3G). Thus, DsbA introduces an abnormal disulfide bond into Pho[SCCC] molecule in vivo. DsbA-independent disulfide bond formation occurs mostly between the natural combination of cysteines; this was true even when disulfide bond formation was driven by a "nonspecific" oxidant, GSSG (Fig. 3G). Fig. 2 (lanes  1-9) demonstrated that the ox2 isoform of PhoA[SCCC] was gradually converted to the ox1 isoform in the wild-type cells. We found that this conversion did not occur in a dsbC deletion strain (Fig. 2, lanes 19 -27). The ox2 species in this strain was degraded over the time. These results suggest that the DsbC function is needed for the posttranslational conversion from the aberrant to the correct disulfide isoforms of PhoA[SCCC] in vivo.

DsbC Is Required for the Conversion of Aberrant to Normal Disulfide Isoforms of PhoA[SCCC]-The results in
DsbC Overproduction Enhances the Production of the Correctly Disulfide-bonded PhoA [SCCC] Molecules-In the presence of a DsbC-overproducing plasmid, even the dsbA ϩ strain produced preferentially the correctly disulfide-bonded PhoS-[SCCC] (Fig. 3B). Control cells with the vector produced mainly the ox2 isoform, which was later converted to the ox1 isoform (Fig. 3A). DsbC overproduction in the dsbA-disrupted strain resulted in almost exclusive production of the reduced form (Fig. 3D), whereas the control cells with the vector produced both the reduced and the ox1 forms (Fig. 3C) (lanes 19 -27), each carrying pMS003 (PhoA[SCCC]) were grown in minimal amino acids/glycerol medium at 30°C, induced for the synthesis of PhoA[SCCC] with IPTG, and pulse-labeled with [ 35 S]methionine for 30 s followed by chase with unlabeled methionine for 1-80 min as indicated. PhoA species were immunoprecipitated and subjected to nonreducing SDS-PAGE as described under "Experimental Procedures." Strain W3110 tonA, the isogenic dsb ϩ counterpart of W3110 tonA ⌬dsbC gave essentially the identical results as MS3 shown in lanes 1-9. WT, wild type.

Roles of DsbA and DsbC in Vivo 10350 disulfide bond formation.
DsbC Exhibits Different Properties in the Presence of an Excess Oxidant-We repeated these experiments in the presence of supplemented GSSG (Fig. 3, E-H). As already discussed, GSSG stimulated the correct disulfide bond formation in the absence of DsbA (Fig. 3G). Overproduction of DsbC in the presence of both GSSG and DsbA gave only a small stimulation of the formation of the ox1 isoform (Fig. 3, compare E and F). Overproduction of DsbC in the presence of GSSG and in the absence of DsbA resulted in the production of both ox1 and ox2 forms (Fig. 3H). Since GSSG alone (in the absence of DsbA) supported the formation of only the ox1 form (Fig. 3G), the results of Fig. 3H indicate that excess DsbC gained the ability to introduce the incorrect disulfide bond in the presence of GSSG. This is in marked contrast to the oxidation inhibition observed in the dsbA Ϫ cells in which DsbC was overproduced in the absence of added GSSG (Fig. 3D).
Thus, in the presence of GSSG, DsbC is transformed to have a DsbA-like ability to introduce disulfide bonds that are not necessarily in the correct combination. Under oxidative conditions, the isomerization activity of DsbC may be suppressed. DISCUSSION Artifacts in the determination of in vivo redox states of proteins can be minimized by examining them after acid denaturation (23,27), the method employed in this study. Of the two disulfide bonds of PhoA, the amino-terminal disulfide (Cys 168 -Cys 178 ) constrains a loop of only 9 amino acids, while the carboxyl-terminally located disulfide (Cys 286 -Cys 336 ) constrains a loop of 49 amino acids. The fast migration of the oxidized PhoA in SDS-PAGE can essentially be ascribed to the Cys 286 -Cys 336 disulfide bond (Fig. 1). Since any incorrect disulfide bonds that can be formed in PhoA should connect cysteines that flank at least 107 amino acids (in the case of Cys 178 -Cys 286 ), they are expected to confer even more increased mo-bility in gel electrophoresis.  (3,4) and in vitro (19), triggering the subsequent folding and dimerization reactions (19). The fact that PhoA[SSCC] has almost 100% enzymatic activity (20) indicates that the PhoA molecule without the NH 2 -terminal disulfide bond retains the ability to position the active site residues in proper geometry, but this event is incomplete until the carboxyl-terminal disulfide bond has been formed. In PhoA[SCCC], kinetic competition may occur between different combinations of the 3 cysteines for disulfide bond formation, and it should be modulated by ongoing folding reactions (28). Thus, extremely rapid disulfide bond formation may be more random than slower disulfide bond formation. Furthermore, DsbA may be so potent and efficient that it introduces or initiates to introduce a disulfide while a substrate polypeptide chain is still in the process of membrane translocation; the first translocating cysteine, Cys 178 , will be committed for disulfide bond formation when it reaches the periplasm and forms a transient disulfide with the reactive Cys 30 residue (29 -31) of DsbA. In contrast, "spontaneous" or glutathionedriven disulfide bond formation in the absence of DsbA will be less efficient and slower such that the correct combination is preferred due to the local folding properties of the polypeptide chain. DsbA-dependent formation of aberrant disulfide bonds was implicated previously (but not demonstrated) in a system where a foreign protein was expressed in E. coli (32,33).
We obtained two kinds of results that suggest that DsbC functions to stimulate the formation of the correct disulfide bond. First, slow conversion occurs from the aberrant to the correct disulfide bonds of pulse-labeled PhoA[SCCC] in wildtype cells, but not in the dsbC deletion strain. This observation clearly indicates that DsbC-dependent isomerization of disulfide bond occurs in the cell. Second, overproduction of DsbC in the dsbA ϩ cells enhanced the rapid formation of the correct disulfide bond. This observation could be interpreted in terms of either very rapid isomerization in the presence of excess DsbC or de novo introduction of the correct disulfide bond in the presence of excess DsbC. In view of the biochemical demonstration of isomerization activity of DsbC (16), we think it reasonable to assume the rapid isomerization. This is consistent with the observation (Fig. 3, D and H) that DsbA is required for DsbC to exhibit the correct disulfide-introducing function.
Recently, Rietsch et al. (34) reported in vivo results that a dsbC disruption had no effect on OmpA (normally with a single disulfide bond), that it resulted in accumulation of a small amount of reduced PhoA 4 , and that it resulted in greatly diminished production of active urokinase (normally with 12 disulfide bonds). Although these results are consistent with their interpretation that DsbC has a disulfide isomerizing role in vivo, the evidence for the aberrant disulfide bond formation in the absence of DsbC was inconclusive.
Our results provide some additional information about the DsbC functions in vivo. Excess DsbC was inhibitory against the

Roles of DsbA and DsbC in Vivo 10351
background disulfide bond formation that occurs inefficiently in the absence of DsbA. Probably, the reducing character of DsbC might dominate the air oxidation. It was also found, however, that excess DsbC can exhibit DsbA-like function when GSSG was supplemented to the medium. This is consistent with another finding of Rietsch et al. (34) that the suppression of dsbA null mutation by loss-of-function dsbD (dipZ) mutations requires the functional DsbC. They proposed that DsbD (DipZ) together with thioredoxin normally keeps DsbC in the partially reduced state, and the loss of reducing factor leads to the production of oxidized DsbC, which in turn substitutes for DsbA. This model is consistent with our results; excess DsbC is oxidized by GSSG and then it substitutes for DsbA.
We have now demonstrated that the principal role of DsbC is to facilitate disulfide bonds that are formed between correct pairs of cysteines. This demonstration was made possible by using a unique model protein, PhoA [SCCC]. The results of Rietsch et al. (34) that the formation of active urokinase, a natural (but still foreign to E. coli) protein with multiple disulfide bonds, depends absolutely on DsbC also supports this notion. These in vivo results establish that the cellular role of DsbC lies in the isomerization of disulfide bonds until the protein has been folded into a stable conformation. , consistent with its instability. Thus, the essential conclusions of the first publication hold. Although PhoA[SCCC] has a lower than normal enzyme activity, the abnormally migrating protein species reported in the second publication was not observed for PhoA [SCCC]. Thus, although the essential conclusion that DsbA introduces a disulfide bond of an aberrant combination (possibly involving residue 401 in PhoA[SC-CCC*]) and that DsbC isomerizes that bond holds, the interpretation that the abnormal disulfide bond observed was between intrinsic PhoA residues was incorrect. We regret that we overlooked the mutation that existed in one of the starting materials. We thank Dr. George Georgiou for communicating their sequencing results of pMS003 and Yuki Takahashi for eliminating the mutation and characterizing the new PhoA derivatives.

Additions and Corrections
Plasmid pMS002 and its derivatives used in the above two publications proved to contain an additional mutation for a Ser-401 3 Cys substitution within PhoA. We traced this mutation back to the phoA plasmid (provided by others) that was used to substitute the amplified segment, as described in the first publication (page 6174, "Experimental Procedures"). Given this fact, we eliminated this unwanted mutation from most of the plasmid constructions and repeated key experiments presented in both publications. Mutant forms of PhoA are shown by the abbreviations described in both articles; however, the previous constructions are indicated by C* attached to the end, and its absence indicates new constructions without the Ser-401 3 Cys mutation. . We also ob-We suggest that subscribers photocopy these corrections and insert the photocopies at the appropriate places where the article to be corrected originally appeared. Authors are urged to introduce these corrections into any reprints they distribute. Secondary (abstract) services are urged to carry notice of these corrections as prominently as they carried the original abstracts.