Engineered DsbC chimeras catalyze both protein oxidation and disulfide-bond isomerization in Escherichia coli: Reconciling two competing pathways
- Departments of *Chemical Engineering and ‡Biomedical Engineering and †Institute for Cell and Molecular Biology, University of Texas, Austin, TX 78712; and §Department of Biochemistry, Baylor College of Medicine, Houston, TX 77030
-
Communicated by Jonathan Beckwith, Harvard Medical School, Boston, MA, April 28, 2004 (received for review February 25, 2004)
Abstract
In the Escherichia coli periplasm, the formation of protein disulfide bonds is catalyzed by DsbA and DsbC. DsbA is a monomer that is maintained in a fully oxidized state by the membrane enzyme DsbB, whereas DsbC is a dimer that is kept reduced by a second membrane protein, DsbD. Although the catalytic regions of DsbA and DsbC are composed of structurally homologous thioredoxin motif domains, DsbA serves only as an oxidase in vivo, whereas DsbC catalyzes disulfide reduction and isomerization and also exhibits significant chaperone activity. To reconcile the distinct catalytic activities of DsbC and DsbA, we constructed a series of chimeras comprising of the dimerization domain of DsbC, with or without the adjacent α-helical linker region, fused either to the first, second, third, or fifth residue of intact DsbA or to thioredoxin. The chimeras fully substituted for DsbC in disulfide-bond rearrangement and also were able to restore protein oxidation in a dsbA background. Remarkably, the chimeras could serve as a single catalyst for both disulfide-bond formation and rearrangement, thus reconciling the kinetically competing DsbB–DsbA and DsbD–DsbC pathways. This property appeared to depend on the orientation of the DsbA active-site cysteines with respect to the DsbC dimerization domain. In vitro, the chimeras had high chaperone activity and significant reductase activity but only 15–22% of the disulfide-isomerization activity of DsbC, suggesting that rear-rangement of nonnative disulfides may be mediated primarily by cycles of random reduction and reoxidation.
Disulfide bonds play a key role in the folding process of many secretory and membrane proteins. A series of cysteine thiol:disulfide oxidoreductases have evolved in both eukaryotic and prokaryotic systems to catalyze this critical cellular process. In particular, the Dsb family of the Escherichia coli periplasm consists of two distinct pathways: DsbA–DsbB and DsbC/DsbG–DsbD. These pathways are involved in the formation of disulfides and the rearrangement of incorrectly formed bonds, respectively (1, 2). The extreme oxidizing nature of DsbA mediates rapid oxidation of substrate cysteines and, therefore, results in the formation of nonnative disulfides, which are in turn rearranged by DsbC and, to a lesser extent, DsbG. Despite the strong oxidizing environment of the periplasmic space, DsbC and DsbG have to be maintained in a reduced state to be able to catalyze the rearrangement of nonnative disulfide bonds (1, 2).
DsbA and DsbC are found in entirely oxidized and reduced states, respectively (3). DsbA is recycled by the membrane protein DsbB, which then passes its electrons to the quinones and the respiratory chain, whereas DsbC is maintained in the reduced state by another membrane protein, DsbD, by means of an electron-transfer relay that involves thioredoxin (TrxA) and TrxA reductase in the cytoplasm (4). Remarkably, the Dsb machinery has evolved to capitalize on a strong thiol oxidant (DsbA) and a strong thiol reductant (DsbC) that do not appear to exchange electrons with each other in the periplasmic space but instead act synergistically in oxidative protein folding. Studies have revealed that the transfer of electrons between the DsbA–DsbB and the DsbC/DsbG–DsbD pathways is strongly disfavored kinetically. Oxidation of DsbC by DsbA is very slow in vitro (5). Compelling evidence shows that the interaction of DsbB with DsbA is highly specific and favored kinetically over the oxidation of DsbC (6, 7). Also, Bader et al. (8) suggest that dimerization protects the DsbC active site from the interaction with DsbB, thus maintaining the segregation of the oxidative and reductive pathways. The recent characterization of the DsbC–DsbDα complex crystal structure revealed that the dimerization domain of DsbC is critical for the interaction with DsbD and is, therefore, required for reduction to take place (9). Nonetheless, interactions between enzymes within the same pathway are favored strongly over nonphysiological disulfide-exchange reactions between the two pathways by kinetic differences of 103- to 107-fold (10).
DsbC exhibits significantly higher in vitro isomerase and reductase activity compared with DsbA. The catalytic domains of DsbC and DsbA show a considerable degree of structural homology, and they both contain a CXXC TrxA active-site motif for the catalysis of disulfide-exchange reactions. These similarities in the catalytic domain of the two proteins raise the question whether the disulfide-isomerization activity of DsbC stems simply from the dimerization of a TrxA fold active site. Here, we report the construction, biochemical characterization, and in vivo function of a series of DsbC–DsbA chimeras and a DsbC–TrxA chimera in which the catalytic domains are linked to the dimerization domain of DsbC. Although these chimeras exhibit poor isomerase activity in vitro, they are capable of catalyzing both protein oxidation and disulfide-bond rearrangement in the periplasm of E. coli.
Materials and Methods
Strains and Plasmids. The bacterial strains and plasmids used in this study are listed in Table 2, which is published as supporting information on the PNAS web site. The DsbC chimeras were constructed by overlap extension PCR using the primers listed in Table 2 and cloned into pBAD33 (11). All of the chimeras contain a C-terminal hexahistidine tag. For protein purification, the gene fusions were digested with XbaI and HindIII, ligated into pET28(a), and transformed into E. coli BL21 cells.
In Vivo Assays. To determine alkaline phosphatase activity, overnight cultures were grown in low-phosphate minimal medium containing Mops salts, 0.2% glycerol, 0.2% casamino acids, and 0.5 μg/ml thiamine, supplemented with 50 μg/ml of kanamycin and 25 μg/ml of chloramphenicol, as needed. Cultures were diluted 1:100 in the same media, and arabinose at a final concentration of 0.2% (wt/vol) was added when the cell density reached OD600 = 0.4. The cells were collected 4 h later and mixed with a buffer containing 0.4 M iodoacetamide and lysis buffer (B-PER, bacterial protein extraction reagent; Pierce) in a 1:2 ratio. The activity of alkaline phosphatase was determined as described (12).
For cell-motility assays, overnight cultures were grown in M9 salts/0.1% casein amino acids/2 mM MgSO4/5 μg/ml thiamine/0.2% glycerol/0.2% arabinose, supplemented with 50 μg/ml of kanamycin and 25 μg/ml of cloramphenicol, as needed; diluted 1:100 in fresh media; and grown for an additional 6 h. We normalized 3-μl aliquots to the same optical density, spotted them in the center of plates containing the same media with 0.3% agar, and incubated them at 37°C for 24 h.
To determine the folding yield of a truncated version of human tissue plasminogen activator (vtPA) consisting of the kringle 2 and protease domains, E. coli PB351 (SF100 ΔdegP::kan, ΔdsbC), or PB401 (SF100 ΔdegP::kan, dsbA::kan) were cotransformed with pBAD33 derivatives encoding the DsbC chimeras and with pTrcStIIvtPA (13), a pTrc99 derivative encoding the vtPA gene fused to the stII leader peptide. Cultures were grown at 30°C in 96-well plates with LB media containing 50 μg/ml ampicillin and 25 μg/ml chloramphenicol, as needed. When the cells reached late stationary phase (OD595 ≈ 0.8), the expression of the DsbC chimeras and of vtPA was induced by the addition of 2% (wt/vol) arabinose and 1 mM isopropyl β-d-thiogalactoside, respectively. The culture absorbance (OD595) was recorded 3 h later, and the cells were lysed by transferring 30-μl aliquots into wells containing 20 μl of the lysis reagent BugBuster (Novagen), followed by incubation at room temperature for 30 min and freezing to -20°C. After thawing to room temperature, the amount of catalytically active vtPA was determined by using an indirect chromogenic assay for plasminogen activation. Briefly, 50-μl aliquots of cell lysates were mixed with 200 μl of 50 mM Tris·HCl,pH7.4/0.01% Tween 80/0.01 mg/ml human glu-plasminogen/0.1 mM Spectrozyme PL (American Diagnostica, Greenwich, CT) in microtiter plate wells. The reaction was allowed to proceed at 37°C, and the change in absorbance at 405 nm was recorded. ΔA 405 is a direct measure of the tissue plasminogen activator (tPA) activity.
In Vivo Redox State. The in vivo redox states of the Dsb chimeras was determined by derivatization of free thiols by 4-acetamido-4′-maleimidyl-stilbene-2,2′-disulfonic acid (AMS; Molecular Probes) in trichloroacetic acid-quenched samples, as described (14), followed by the detection of the reduced and oxidized protein bands by Western blotting using mouse anti-His tag polyclonal serum (Sigma). Oxidized protein standards were generated by omitting the alkylation treatment, whereas the reduced standard was generated by incubating 1 ml of cells with DTT to a final concentration of 100 mM for 20 min on ice, followed by alkylation, as described above.
Expression, Purification, and Biochemical Assays. The DsbC chimeras were expressed in E. coli BL21 transformed with the appropriate PET28(a) plasmids and grown in LB medium with 50 μg/ml kanamycin at 37°C. We added 0.01 mM isopropyl β-d-thiogalactoside at an OD600 of 0.6–0.8, and the cells were harvested by centrifugation 4 h later.Periplasmic proteins were isolated by the cold osmotic-shock procedure (15). The His-tagged DsbC mutant proteins were purified from the osmotic shockate by immobilized metal ion affinity chromatography using a 10 ml PolyPrep column (Bio-Rad) according to the manufacturer's instructions. The protein in the eluant was purified further by gel-filtration FPLC on a Superdex-200 size-exclusion column (Amersham Biosciences) equilibrated in PBS with 10% glycerol. The concentration of purified proteins was determined from the A 280 values by using estimated extinction coefficients (16) (Peptide Property Calculator, available at www.basic.nwu.edu/biotools/proteincalc.html).
Insulin reduction in the presence of DTT was determined as described (17). The rate of change in the A 650 due to the aggregation of reduced insulin was also measured. The activity is expressed as the ratio of the initial slope of the turbidity curve to the lag time (18). The renaturation of reduced, denatured RNase A was monitored according to the procedure in ref. 19. In this assay, the lag in the appearance of active RNase A is proportional to the oxidase activity and the initial rate in the RNase A vs. time plot is used to calculate the rate of disulfide isomerization as described in ref. 19. Activities are expressed as micromolars of native RNase A formed per minute per micromolars of enzyme.
The protection of citrate synthase (CS) from thermal inactivation was monitored according to ref. 20. The rate of thermal inactivation obtained with or without 4 μM chaperone was determined. Finally, the aggregation of guanidine hydrochloride-denatured GAPDH after dilution in the presence of chaperone ranging from 0- to 44.8-μM concentrations was studied according to Cai et al. (21).
Protein Structure Modeling. The secondary structure prediction was performed by using predictprotein. Manual model building of the chimeric proteins was done in xfit (22), based on the information obtained for the secondary structure of the region linking the DsbC dimerization domain to DsbA.
Results
Construction of DsbC Chimeras. The dimerization domain of DsbC comprises residues 1–59 and is joined to the C-terminal catalytic domain by a 12-aa-long α-helix linker (amino acids 60–72). We constructed a series of fusions encoding the dimerization domain of DsbC with or without the α-helix (DsbCdα and DsbCd, respectively). DsbCd was fused to the first, second, or third residue of mature DbsA [DsbA (1–189), DsbA (2–189), and DsbA (3–189), respectively] (Fig. 1A). DsbA (2–189) and DsbA (3–189) were similarly fused to DsbCdα. In addition, a fusion to the fifth residue of the mature DsbA (5–189) was constructed. In DsbC, the active-site cysteine pairs within each catalytic domain are oriented facing each other perpendicular to the axis of symmetry along the dimerization domain. Molecular modeling indicates a similar orientation of the active-site residues in DsbCdα–DsbA (2–198) and in DsbCdα–DsbA (5–189) (see Fig. 5, which is published as supporting information on the PNAS web site) but not for DsbCdα–DsbA (3–189), where the active site is predicted to be tilted 170° relative to the long axis of symmetry of the molecule. In other words, in DsbCdα–DsbA (3–189), the active-site cysteines are facing away from each other (Fig. 1B). Finally, a gene fusion encoding a chimera, consisting of the DsbC dimerization domain followed by the linker α-helix (amino acids 60–72 in DsbC), and the N-terminal residues of the catalytic domain fused to the complete sequence of TrxA, was constructed also. The redox potential of TrxA fold proteins involved in disulfide transfer is modulated by the identity of the two amino acids within the C—X—X—C catalytic motif (23). In DsbCdαN–TrxA, the G—P dipeptide in the active site of TrxA was substituted with the G—H sequence found in the active site of protein disulfide isomerase (PDI). TrxA (CGHC) protein exhibits a higher redox potential (E 0 = -235 mV), making it more similar to PDI and conferring higher isomerase activity (24, 25).
Schematic representation of the DsbC chimeras. (A) DsbC, DsbA, and TrxA “PDI-like” sequences, as well as domain composition of the DsbC fusion chimeras. (B) Orientation of the active sites relative to the symmetry axis of the dimerization domain. 1, DsbC; 2, DsbCdα–DsbA (2–189); 3, DsbCdα–DsbA (3–189); 4, DsbCdα–DsbA (5–189).
The dsbC fusions were placed downstream from the arabinose promoter in the medium copy number plasmid pBAD33 (11). After induction of protein expression with arabinose, the wild-type DsbC and all of the DsbC–DsbA chimeras accumulated to nearly identical levels, as determined by Western blotting with a polyclonal antibody that recognizes the C-terminal His tag. In contrast, DsbCdαN–TrxA“PDI-like” was expressed at a 4- to 5-fold higher level (Fig. 2A).
Disulfide-bond formation in vivo. (A) Yield of active vtPA in dsbC or dsbA cells and relative expression levels of the chimera proteins. PB351 (SF100 ΔdsbC), or PB401 (SF100 dsbA) transformed with pTrcStIIvtPA and pBAD derivatives encoding the respective fusion proteins were grown in LB media. Protein synthesis was induced as described in Materials and Methods, and the yield of active vtPA at 3 h after induction was determined. Relative activities were obtained by dividing the ΔA 405 (absorbance of each strain subtracted of the background consisting of a strain not expressing tPA) by the ΔA 405 of a strain expressing vtPA alone. Upper shows the fusion protein expression level, determined by Western blotting. Each lane in the gel shows the expression of the corresponding protein in the bar graph. (B) PhoA activity. Effect of the expression of the chimeric proteins on alkaline phosphatase activity in the periplasm of MC1000 dsbA (white bars) and MC1000 dsbB (black bar). The alkaline phosphatase activity of the parental isogenic strain MC1000 is shown by the gray bar. Cells were induced with 0.2% arabinose, harvested in mid-log phase, and lysed, and activity assays were conducted as described. (C) Cell-motility assays. Motility of MC1000 dsbA cells transformed with the following: pBADdsbCdα-dsbA (3–189) (a); pBADdsbCdα-dsbA (5–189) (b); pBADdsbCd-dsbA (1–189) (c); pBADdsbCd-dsbA (3–189) (d); pBADdsbC (e); pBADdsbCdαN-TrxA“PDI-like” (f); MC1000 dsbA (g); and MC1000 grown in low-phosphate media (h). Cultures were diluted with media to the same absorbance, and 3-μl aliquots were spotted on the center of each plate.
Disulfide-Bond Formation and Isomerization in Vivo. In E. coli, the folding yield of eukaryotic proteins with multiple disulfide bonds is limited by the isomerization activity afforded by DsbC (1, 2, 26, 27). In particular, the yield of proteloytically active vtPA, containing a total of nine disulfide bonds, depends on the DsbC expression level (28). In cells containing pBADdsbC and grown with 2% (wt/vol) arabinose, active vtPA accumulates at a 25-fold higher level (13) than in the absence of overexpressed DsbC. In contrast, expression of DsbA from a similar pBAD vector in cells grown under identical conditions confers essentially no increase in the yield of tPA relative to the control (26). However, expression of the DsbC–DsbA chimeras afforded vtPA yields comparable with those provided by pBADdsbC, irrespective of the nature of the fusion residue in DsbA or the presence or absence of the DsbC linker α-helix. DsbCdαN–TrxA also supported the folding of vtPA, but the yield of active vtPA in this case was ≈50% lower (Fig. 2A). For all of the chimeras, only background levels of active vtPA were detected in a dsbD strain background (data not shown). The finding that DsbC–DsbA and the DsbCdαN–TrxA support the folding of vtPA in a DsbD-dependent manner suggests that these chimeras function by facilitating disulfide-bond rearrangement.
DsbC cannot normally serve as an oxidant to complement the phenotypes of dsbA mutants, such as low PhoA activity or loss of cell motility. In contrast to DsbC, however, the DsbC–DsbA chimeras all were able to support protein oxidation to various degrees. In E. coli MC1000 dsbA grown in low-phosphate media, the PhoA activity is 30-fold lower than in its isogenic parent. Coexpression of the DsbC–DsbA chimeras restored PhoA activity to 45–100% of the values obtained in the parental strain MC1000, whereas DsbCdαN–TrxA was a somewhat weaker oxidant (Fig. 2B). As expected, neither the DsbC–DsbA chimeras nor DsbCdαN–TrxA could restore PhoA activity in dsbB cells (Fig. 2B).
The size of the motility halo in dsbA cells plated on soft agar plates represents an additional, more stringent measure of the ability of proteins to catalyze periplasmic oxidation (43). MC1000 dsbA cells expressing DsbC from pBAD33 were completely nonmotile, whereas the expression of the DsbC–DsbA chimeras restored cell motility to various degrees (Fig. 2C). Proteins containing the DsbC α-helix linker gave larger diameter motility halos compared with identical fusions lacking the 12-aa linker region [Fig. 2C; compare DdsbCdα–DsbA (3–189) and DdsbCdα–DsbA (5–189) with DdsbCd–DsbA (1–189) and DdsbCd–DsbA (3–189)]. In contrast, even though the DsbCdαN-TrxA fusion was capable of oxidizing PhoA partially, it could not restore cell motility, suggesting that it is a weaker catalyst of disulfide-bond formation, either in terms of its activity or with respect to its substrate specificity.
The data described above reveal that the DsbC chimeras can catalyze disulfide-bond rearrangement in dsbC cells and, separately, protein oxidation in dsbA mutants. These findings raised the possibility that the chimeras may be able to satisfy the role of a protein thiol oxidant and a disulfide-isomerization catalyst in dsbA cells simultaneously. The folding of vtPA, in addition to requiring a high level of DsbC activity as discussed above, also critically depends on the presence of DsbA. For this reason, in strain SF100 dsbA, overexpression of DsbC results in background levels of active vtPA (Fig. 2A). In contrast, expression of the DsbC–DsbA chimeras afforded a high yield of active vtPA, with the notable exception of the two fusions containing the DsbA (3–189) domain. Whereas fusions beginning with the first, second, or fifth residue of the DsbA domain were fully active in terms of their ability to support the formation of active vtPA, fusions beginning at the third amino acid in DsbA were completely inactive in this assay. Thus, DsbA (3–189) fusions exhibit the following properties: They (i) support the formation of active vtPA in wt cells; (ii) are capable of catalyzing protein oxidation in a dsbA background; and nonetheless, (iii) fail to allow folding of vtPA in the dsbA mutant. The surprising inability of the DsbA (3–189) fusions to support disulfide rearrangement in a dsbA background is analyzed in some detail in Discussion.
In order for the DsbC chimeras to catalyze cysteine thiol oxidation and disulfide rearrangement simultaneously, they have to be maintained in the periplasm as a mixture of oxidized and reduced species. The in vivo redox state of the chimeras was determined by harvesting cells into trichloroacetic acid to quench disulfide-bond rearrangements, alkylation of free cysteines with AMS, and detection of the oxidized and reduced forms of the proteins by Western blotting (29). All of the chimeras were maintained predominantly, but not exclusively, in the reduced form (Fig. 3). The fraction of the protein found in the oxidized form ranged from 15% for DsbCdα–DsbA (5–198) to ≈50% for DsbCd–DsbA (1–198). DsbC was present exclusively in the reduced state, as expected (3). In a dsbD strain background, the chimeras were present exclusively in the oxidized form (see Fig. 6, which is published as supporting information on the PNAS web site). The redox state of the DsbC–DsbA fusions should be contrasted to that of DsbA, which is always oxidized even when it is overexpressed (29).
In vivo redox state of the Dsb chimeras. PB401 (SF100 dsbA) were grown in LB media, protein synthesis was induced with 0.2% arabinose, and the redox state of the chimeras after AMS treatment was determined by Western blotting by using sera specific to the His tag. 1, DsbCdα–DsbA (3–189); 2, DsbCdα–DsbA (5–189); 3, DsbCd–DsbA (1–189); 4, DsbC–DsbA (3–189); 5, DsbC; 6, DsbCdαN–TrxA“PDI-like.” The first two lanes correspond to the reduced (red) and oxidized (ox) standards obtained by incubating the cells with DTT before trichloroacetic acid precipitation and by omitting AMS, respectively.
Biochemical Characterization. Four DsbC–DsbA chimeras, DsbCdαN–TrxA and DsbC as a control were purified by immobilized metal ion affinity chromatography, and oligomerization status was analyzed by gel-filtration FPLC. All of the proteins eluted exclusively as dimers (see Fig. 7, which is published as supporting information on the PNAS web site).
The insulin reduction activity (30) of the DsbC–DsbA chimeras was 20–62% of the insulin reduction activity of DsbC (Table 1). DsbCdαN–TrxA exhibited a slightly higher reductase activity. For comparison, DsbA has only ≈10% of the activity of DsbC. All of the chimeras exhibited low disulfide isomerase activity in the refolding of reduced RNase A. DsbCd–DsbA (1–189), DdsbCdα–DsbA (5–189), DdsbCdα–DsbA (3–189), DdsbCd–DsbA (3–189), and DsbCdαN–TrxA displayed 10–22% of the isomerase activity of DsbC, which was 8-fold less active than PDI. The low values of isomerase activity obtained were nonetheless higher (P < 0.05) relative to the background rate of RNase A refolding in the absence of catalyst or in the presence of DsbA.
The chaperone activity of the chimeras was evaluated based on the protection of CS from thermal inactivation and from the prevention of protein aggregation during the refolding of Gdn·HCl-denatured GAPDH (20). At a 2.7-fold stoichiometric excess, DsbCdαN–TrxA delayed the inactivation of CS to the same extent as DsbC. However, whereas DsbA did not have any effect on CS inactivation, dimerization by virtue of its fusion to DsbCd or to DdsbCdα gave rise to proteins that were 3-fold more efficient relative to DsbC (or 10-fold better, compared with DsbA) in this assay (Table 1). Fig. 4 shows the effect of protein activities on the suppression of GAPDH aggregation during refolding. Even when added at a 15-fold stoichiometric excess, TrxA had no effect on GAPDH aggregation, whereas a large excess of DsbA suppressed aggregation by <20%. In contrast, fusion to the DsbC dimerization domain markedly enhanced the chaperone activity of TrxA and DsbA. All of the tested chimeras could suppress the aggregation of GAPDH to a significant extent. Collectively, the data presented in Table 1 and in Fig. 4 reveal that fusion of DsbA or TrxA to the dimerization domain gives rise to proteins with appreciable chaperone activity, the exact magnitude of which depends on the substrate and the assay conditions.
Effect of foldases on the aggregation of chemically unfolded GAPDH. GAPDH (140 μM) was denaturated O/N in 3 M guanidine hydrochloride and 5 mM DTT at 4°C. After 1:50 dilution in phosphate buffer, with increasing concentration of catalysts, the A 488 was recorded continuously.
Discussion
Oxidative protein folding involves two complementary but also competing processes: cysteine thiol oxidation and isomerization of nonnative disulfide bonds. Throughout nature, enzymes that catalyze thiol:disulfide exchange reactions (with the exception of certain enzymes with narrow substrate specificity, such as DsbB) employ structurally homologous TrxA domains for catalysis (31). Relatively subtle changes, such as the presence of different amino acids within the dipetide sequence in the TrxA CXXC active site, the insertion of α-helical domains within the TrxA fold, and fusion to additional domains (23, 32–34), have been used during evolution to modulate the function of thiol:cysteine oxidoreductases. DsbA and DsbC provide an illustration of how two TrxA fold enzymes have evolved to perform different reactions in the cell. DsbA is a powerful oxidant (E 0 = -130 mV), but it exhibits marginal reductase, chaperone, or disulfide isomerase activity (Table 1) (3, 30). DsbC has an active-site cysteine pair with a redox potential almost as low as that of DsbA (35–37). However, DsbC not only catalyzes disulfide-bond oxidation in vivo, but, in contrast to DsbA, it also displays disulfide isomerase, reductase, and chaperone activities (5, 35).
In principle at least, DsbC should be sufficient to catalyze both disulfide-bond formation and rearrangement in the cell without the need for a specialized oxidant. PDI, the eukaryotic analog of DsbC, catalyzes protein oxidation in the endoplasmic reticulum in an Ero1p-dependent process (38), and recent data suggest that it also serves as an isomerase in the ER (H.F.G., unpublished data). However, this is not the case with DsbC; even though it has high oxidation activity per se, in the periplasmic space it is maintained exclusively in the reduced state and, therefore, can only catalyze disulfide-bond rearrangement. The presence of the DsbD–DsbC and the DsbB–DsbA systems enables the coexistence of a fully reduced and a fully oxidized catalyst in close proximity to each other, without the establishment of an energy-consuming futile cycle that would be draining and ultimately detrimental to the cell (8). The kinetic isolation of the cysteine thiol oxidation and disulfide-rearrangement pathways in prokaryotes partially stems from the fact that DsbC is not readily oxidized by either DsbB or DsbA (5, 7). Bader et al. (8) presented genetic and biochemical evidence indicating that the dimeric nature of DsbC represents the main barrier to its oxidation by DsbB. Also, Rozhkova et al. have measured the kinetics of disulfide-exchange reactions among periplasmic components of the DsbA–DsbB and the DsbC–DsbD pathways and compared them with nonfunctional reactions between redox active sites of periplasmic oxidoreductases. This analysis further highlights how prevention of nonphysiological interactions between the proteins involved in disulfide-bond formation guarantees the separation of the oxidative and reductive pathways in the same cellular environment in vivo (10).
We now show that DsbC chimeras, which consist of the dimerization domain fused to a TrxA catalytic module in close analogy to the architecture of the authentic DsbC, become oxidized by DsbB and are able to substitute for the lack of DsbA. The ability of the chimeras to interact with DsbB was independent of the presence or absence of the α-helix linker joining the two domains. In vitro, gel-filtration analysis showed that the DsbC chimeras are present exclusively as dimers (see Fig. 7), which is a conclusion that is consistent with the recent finding of Zhao et al. (30). Although we cannot rule out the possibility that the DsbC chimeras exist in equilibrium with a very small amount of monomer that interacts with DsbB and serves as an oxidant in vivo, we think that the presence of such monomeric species is unlikely. Our findings suggest that dimerization alone cannot account for the kinetic isolation of DsbC from DsbB, which must be dictated by additional structural features that presumably modulate the accessibility of the active-site cysteines within the catalytic domain.
Although the six DsbC–DsbA chimeras and DsbCdαN–TrxA can be oxidized by DsbB, in the periplasm these proteins are maintained mostly in the reduced form by the action of DsbD. In vitro, DsbD can readily reduce TrxA and DsbC but not the monomeric DsbA (39). Recent evidence shows that DsbA is reduced by DsbD extremely slowly in vitro (10); it is likely that the reduction of the DsbA chimeras by DsbD is due to the increase in the effective concentration of DsbA active sites that results from dimerization or by more favorable steric interactions originated by the fusion of DsbA with the DsbC dimerization domain.
In the absence of a kinetic barrier between DsbD and DsbB, how does the cell avoid the establishment of a draining futile cycle that would require energy to maintain and would ultimately be detrimental (8)? Such a futile cycle would ultimately result in the consumption of NADPH by TrxB, leading to the transfer of electrons from the TrxA–DsbD relay to DsbB–DsbA and, ultimately, to the respiratory chain. The finding that the chimeras are predominantly reduced in vivo suggests that oxidation by DsbB must be relatively slow and kinetically disfavored, compared with the transfer of the active-site disulfide to substrate proteins or to DsbD. Therefore, it appears that only a small fraction of the DsbC–DsbA and DsbC–TrxA chimeras ends up being shuttled between DsbB and DsbD, and hence, the energy expenditure due to the establishment of a futile cycle may be minimal. Also, we note that, as has been observed in other studies, the establishment of an artificial futile cycle affects the yield on the carbon source but not necessarily the growth rate (40).
The folding of vtPA in the periplasm normally requires the action of DsbA and also a high level of DsbC activity. However, expression of most, but not all, of the chimeric DsbC enzymes in a dsbA strain, afforded a high yield of vtPA, revealing that a single catalyst is perfectly capable of catalyzing both cysteine thiol oxidation and disulfide rearrangement in vivo. The genomes of several bacteria, including Helicobacter pylori (41) and Clostridium acetobutylicum, encode two DsbC homologues but have no DsbA homologue, indicating that a single catalyst for disulfide-bond formation may have been adapted during evolution. Interestingly, neither organism has a dsbB gene, which suggests that the reoxidation of DsbC in these organisms is accomplished by a mechanism that has yet to be determined. Why then do E. coli and other γ-proteobacteria employ DsbA–DsbB for protein oxidation and a separate system comprising of two (DsbC and DsbG) catalysts of disulfide rearrangement? A simple explanation is that the evolution of separate, kinetically isolated, catalysts allows the exocytoplasmic environment to be maintained in a highly oxidizing state and, as noted above, may provide an increase in the growth yield by avoiding the establishment of a futile cycle.
Structural and biochemical data indicate that the cleft formed by the dimerization domain is responsible for peptide binding and for the chaperone activity of DsbC (30, 42, 43). It is, therefore, not surprising that the fusion of TrxA domain proteins, such as DsbA and TrxA, which have little or no chaperone activity on their own, gives rise to chimeras that are able to prevent protein aggregation and inactivation. The dimerization of DsbA also conferred a 2- to 7-fold increase in the rate of insulin reduction. Dimerization of TrxA“PDI-like” decreased the reductase activity of the molecule compared with TrxA or TrxA“PDI-like” (24).
The isomerase and oxidase activity of the chimeras were determined by using RNase A as a substrate (19). In this assay, the reactivation of reduced, denatured RNase A is monitored as a function of time in the presence of a catalyst. The lag time for the appearance of active RNase A depends on the rate of protein oxidation by the catalyst, whereas the initial slope in the RNase A activity vs. time plot is proportional to the rate of disulfide isomerization. Analysis of the lag times in the refolding of RNase A indicated that the chimeras are much more effective in catalyzing disulfide-bond formation (oxidase activity, ≈50% of PDI), compared with DsbC, which is a very weak oxidant (oxidase activity, 4% of PDI, data not shown). However, the chimeras displayed very low isomerase activitym, which was only 10–22% of DsbCs, or 2–3% relative to PDI (Table 1).
On the basis of the analysis described above, we propose that it is the reduction of incorrect disulfide bonds, rather than the isomerization activity per se, that mediates the rearrangement of incorrect disulfides. TrxA and, to a much lesser extent, DsbA can catalyze the net reduction of disulfides in vitro, but both proteins become oxidized by DsbB in the periplasmic space (44) and are, therefore, unavailable for the reduction of nonnative disulfide bonds. Fusion to the DsbC dimerization domain allows the activesite cysteines in the chimeras to be partially maintained in the thiol state, as needed for the catalysis of disulfide-bond reduction. Recently, H.F.G. and coworkers (45) proposed that the isomerization of disulfides by PDI proceeds by means of trial-and-error cycles of timed reduction and reoxidation. The chimeras may be operating by similar reduction-oxidation cycles whereby the oxidant is either DsbA or, in dsbA mutant strains, the oxidized form of the chimera. The need for proper timing of the oxidation-reduction cycles (45) may be responsible for the inability of DdsbCd–DsbA (3–189) and DdsbCdα–DsbA (3–189) to support the folding of vtPA in the dsbA mutant. Secondary structure prediction reveals that in the chimeras the N-terminal residue of the DsbA domain extends a short (for DsbCd) or long (for DsbCdα) α-helix. The precise fusion amino acid determines the orientation of the catalytic domain with respect to the α-helix and also the dimerization domain (see Figs. 1B and 5). Fusion to the third N-terminal residue of DsbA results in positioning of the active site (CHPC) at the face opposite to the V-shaped cleft. This finding is in contrast to DsbCdα–DsbA (2–189), DsbCdα–DsbA (5–189), as well as in DSbCd–DsbA (1–189) and in DsbCd–DsbA (2–189), where the catalytic active sites of each monomer are positioned facing each other. It is possible that the localization of the active site on the opposite face to the V and away from the dimerization cleft incapacitates the ability of the protein to carry out timed oxidation and reduction cycles. In other words, although DdsbCd–DsbA (3–189) and DdsbCdα–DsbA (3–189) can form a proper redox environment when the oxidant is DsbA, the timing of the reduction and oxidation cycles is not optimal when they have to serve simultaneously as both oxidant and reductant. Therefore, even though these proteins posses both oxidation and reduction activities and are present as a mixture of reduced and oxidized form in vivo, they are incapable of assisting the productive rearrangement of nonnative disulfide bonds in vtPA. Consistent with this hypothesis, a DsbC mutant with a 2-aa deletion in the α-helical linker that causes the catalytic domain to be rotated with respect to the dimerization domain is devoid of isomerization activity in vivo but can fully complement dsbA cells (L.S. and G.G., unpublished data).
Zhao et al. (30) recently constructed a DsbC (1–77)–DsbA (1–189) and a DsbC (1–77)–TrxA(wt) fusion. Consistent with our findings, both of these proteins exhibited chaperone activity and were found to be dimers. However, Zhao et al. reported that DsbC (1–77)–DsbA (1–189) had only 10% of the reductase activity but 50% of the isomerase activity of DsbC. In contrast, we find that DsbCd–DsbA (1–189), which differs from the protein constructed by Zhao in that it has a slightly shorter DsbC domain (we used an 1–72 DsbC fragment vs. the 1–77 fragment used in Zhao et al.) exhibited much higher reductase activity and lower isomerase activity (60% and 11% of the DsbC activity, respectively). Analogous differences were observed in the biochemical properties of DsbCdαN–TrxA“PDI-like” reported here, compared with the DsbC (1–77)–TrxA fusion of Zhao et al. These rather large differences in biochemical properties are either due to rather drastic effects of the residues in the fusion junction or to the slightly different assay conditions that were used in the two studies.
Acknowledgments
We thank Lluis Masip for reading the manuscript, and Kandice Johnson and Lori Murphy for help with some of the experiments. This work was supported by National Intitutes of Health Grant GM55090.
Footnotes
-
↵ ¶ To whom correspondence should be addressed. E-mail: gg{at}che.utexas.edu.
-
Abbreviations: AMS, 4-acetamido-4′-maleimidyl-stilbene-2,2′-disulfonic acid; CS, citrate synthase; PDI, protein disulfide isomerase; TrxA, thioredoxin; tPA, tissue plasminogen activator; vtPA, truncated version of human tPA.
- Copyright © 2004, The National Academy of Sciences
