The disulfide isomerase DsbC cooperates with the oxidase DsbA in a DsbD-independent manner

Summary

In Escherichia coli, DsbA introduces disulfide bonds into secreted proteins. DsbA is recycled by DsbB, which generates disulfides from quinone reduction. DsbA is not known to have any proofreading activity and can form incorrect disulfides in proteins with multiple cysteines. These incorrect disulfides are thought to be corrected by a protein disulfide isomerase, DsbC, which is kept in the reduced and active configuration by DsbD. The DsbC/DsbD isomerization pathway is considered to be isolated from the DsbA/DsbB pathway. We show that the DsbC and DsbA pathways are more intimately connected than previously thought. dsbA⁻dsbC⁻ mutants have a number of phenotypes not exhibited by either dsbA⁻, dsbC⁻, or dsbA⁻dsbD⁻ mutations: they exhibit an increased permeability of the outer membrane, are resistant to the lambdoid phage Φ80, and are unable to assemble the maltoporin LamB. Using differential 2D-LC-MS/MS, we estimated the abundance of about 130 secreted proteins in various dsb⁻ strains. dsbA⁻dsbC⁻ mutants exhibit unique changes at the protein level that are not exhibited by dsbA⁻dsbD⁻ mutants. Our data indicate that DsbC can assist DsbA in a DsbD-independent manner to oxidatively fold envelope proteins. The view that DsbC’s function is limited to the disulfide isomerization pathway should therefore be reinterpreted.

Introduction

DsbA introduces disulfide bonds into secreted proteins in the Escherichia coli periplasm (Bardwell et al., 1991). DsbA has a CXXC catalytic site motif present within a thioredoxin fold. The cysteine residues of this motif are found oxidized in vivo. The disulfide bond of DsbA is very unstable and is rapidly transferred to secreted unfolded proteins. DsbA is then reoxidized by the inner-membrane protein DsbB (Bardwell et al., 1993). DsbA forms a mixed disulfide complex with DsbB. Disulfide bond transfer occurs after conformational changes within the DsbB protein (Inaba et al., 2006). DsbB has two pairs of cysteine residues and generates disulfide bonds de novo from quinone reduction. Electrons are then successively transferred from quinone to cytochrome oxidases and finally to molecular oxygen (Bader et al., 1999).

DsbA is likely to be involved in the oxidative folding of many periplasmic and outer membrane proteins. A search in the sequence databases reveals that about 40% of the 700 secreted proteins in E. coli have at least two cysteine residues and are therefore potential DsbA substrates (Dana Boyd, personal communication). It is therefore not surprising that dsbA⁻ strains have a pleiotropic phenotype; they show an attenuated virulence, lack motility, form mucoidal colonies on minimal media and in the presence of some antibiotics such as tetracycline, and are more sensitive to dithiothreitol and cadmium (reviewed in Collet and Bardwell, 2002). About 15 DsbA substrates have been identified so far; these were obtained by trapping of mixed disulfides with DsbA mutants (Kadokura et al., 2004), from differential thiol trapping experiments (Leichert and Jakob, 2004), and by 2D-gel analysis, which showed decreased levels of expression of several cysteine containing proteins (Hiniker and Bardwell, 2004).

Despite the significant number of proteins that are DsbA substrates, dsbA⁻ strains are surprisingly healthy, particularly when grown on rich media. This in part may be due to the fact that small molecule oxidants, like cysteine, are present in rich media (Bardwell et al., 1993). Consistent with this, dsbA⁻ strains grow very poorly on minimal media lacking cysteine, particularly in some strain backgrounds. Although the rate of disulfide bond formation is about a 100-fold decreased in dsbA⁻ strains (Bardwell et al., 1991), some slow residual disulfide bond formation does occur, and proteins that are stable in the absence of their disulfide accumulate in the oxidized form to near normal levels at steady state. This disulfide formation appears to be dependent on oxygen, as much lower levels of oxidized proteins are seen in dsbA⁻ strains when they are grown anaerobically.

DsbA is a powerful oxidant that apparently lacks proofreading activity. DsbA oxidizes cysteine residues on secreted proteins as they emerge into the periplasm. If the native disulfide bond pattern involves cysteine residues that are consecutive in the amino acid sequence, DsbA can form disulfides correctly. However, when secreted proteins have disulfides that need to be formed between nonconsecutive cysteines, DsbA can introduce non-native disulfides, leading to protein misfolding and degradation by proteases (Berkmen et al., 2005). Noteworthy, our recent work on RNase I, a periplasmic protein with one non-consecutive disulfide, showed that DsbA is more specific than generally assumed (Messens et al., 2007). The correction of non-native disulfides is thought to be the role of a disulfide isomerization system. This system is composed of two soluble periplasmic proteins, DsbC and DsbG, which are thought to function as disulfide isomerase proteins in vivo and in vitro (Zapun et al., 1995; Bessette et al., 1999). Like DsbA, DsbC and DsbG possess a thioredoxin fold and a CXXC catalytic site motif. In contrast to DsbA, whose CXXC active site motif is maintained in an oxidized form, the CXXC motif of DsbC and DsbG is kept reduced in the periplasm. This allows DsbC and DsbG to attack non-native disulfides, a necessary step in the isomerization reaction (reviewed in Messens and Collet, 2006). The protein that keeps DsbC and DsbG reduced is the inner-membrane protein, DsbD. DsbD transfers reducing equivalents from the cytoplasmic thioredoxin system to the periplasm via a succession of disulfide exchange reactions (Rietsch et al., 1996; Katzen and Beckwith, 2000; Collet et al., 2002; Rozhkova et al., 2004). Residues that are important for this electron cascade have recently been identified (Cho and Beckwith, 2006; Hiniker et al., 2006), but the precise mechanism used by DsbD to transport electrons from one side of the membrane to the other is still obscure.

In contrast to the central role played by DsbA in oxidizing a large number of periplasmic proteins, DsbC is thought to be required for the expression of a limited subset of proteins that contain nonconsecutive disulfides, including the penicillin insensitive endopeptidase MepA, the ribonuclease RNase I, and the acid phosphatase AppA (Hiniker and Bardwell, 2004; Berkmen et al., 2005). In agreement with the relatively small number of DsbC substrates, dsbC⁻ strains have a milder phenotype than dsbA⁻ mutants (reviewed in Collet and Bardwell, 2002). DsbC seems to be particularly important under some oxidative stress conditions. For instance, DsbC is required for growth in the presence of high concentrations of copper, a redox metal that catalyzes the formation of non-native disulfide bonds (Hiniker et al., 2005).

The role of DsbG is less clear. DsbG was originally reported to be an essential oxidase (Andersen et al., 1997). However, subsequent work showed it to be nonessential for the growth of E. coli (Bessette et al., 1999). DsbG null mutants have no defect in the folding of heterologous proteins containing multiple disulfide bonds, and are unable to catalyze disulfide bond rearrangement using either hirudin or the bovine pancreatic trypsin inhibitor intermediate as substrates (Bessette et al., 1999; Hiniker et al., 2007). However, DsbG overexpression is able to restore the ability of dsbC⁻ mutants to express some heterologous proteins containing multiple disulfide bonds (Bessette et al., 1999). It is also possible to select mutations in DsbG that complement DsbC, and these mutations show increased isomerase activity (Hiniker et al., 2007). These latter observations and DsbG’s homology to DsbC (Heras et al., 2004) have led to the conclusion that DsbG is a disulfide isomerase with restricted substrate specificity.

The situation in eukaryotic disulfide bond formation is more complex and controversial. Protein disulfide isomerase (PDI) is thought to function as both an oxidase and an isomerase in vivo. However, there has been quite a bit of controversy regarding the relative importance of the oxidative and isomerase activities to the cell (Sevier and Kaiser, 2006). In addition to PDI, yeast possesses several thioredoxin-like proteins that are localized to the ER. Those proteins are likely to play important roles in the isomerization and oxidation of proteins (reviewed in Gruber et al., 2006).

In E. coli, the current view is that the distinct disulfide catalytic pathways have well defined roles: the DsbA/DsbB system is important in oxidizing disulfide bonds, and the DsbC-G/DsbD system is important in isomerizing them. The results presented in this paper show that this conclusion needs to be reinterpreted. We show that the simultaneous absence of DsbA and DsbC has severe consequences on E. coli’s viability and outer membrane integrity. dsbA⁻dsbC⁻ mutants are resistant to the Φ80 bacteriophage and seem unable to fold the trimeric porin LamB. Using a 2D-LC-MS/MS proteomics approach, we show that the absence of DsbA and DsbC affects the global protein content of the periplasm and leads to a decreased abundance of several cysteine-containing proteins. Our data indicate that the function of DsbC goes beyond the correction of the non-native disulfides formed by DsbA. On the basis of our results, we propose a new model for the oxidative protein folding pathways in E. coli.

Results and Discussion

dsbA⁻dsbC⁻ double mutants have a more severe phenotype than dsbA⁻ mutants

To test whether DsbC can assist DsbA outside the framework of the isomerization pathway, we made dsbA⁻dsbC⁻ double mutants and compared their phenotype to that of dsbA null strains. The logic behind these experiments is as follows: if DsbC’s function is restricted to the correction of DsbA’s mistakes, then dsbA⁻ and dsbA⁻dsbC⁻ mutants should have a similar phenotype. In contrast, if DsbC has functions in addition to the correction of DsbA’s mistakes, then dsbA⁻dsbC⁻double mutants should have a more severe phenotype than dsbA⁻ mutants. We observed that a dsbA⁻dsbC⁻ mutant has a more severe growth defect than a dsbA⁻ strain when cells are grown in minimal media (Fig. 1A and Table 2A). Both the dsbA⁻ and dsbA⁻dsbC⁻ strains are more sensitive to antibiotics and detergents than wild-type strains. However, the sensitivity of the dsbA⁻dsbC⁻ double mutant is more severe than that of a dsbA⁻ mutant. The dsbA⁻dsbC⁻ mutant is more sensitive to SDS (Fig. 1B), and, unlike dsbA⁻ strains, is unable to grow in the presence of 4 μg/ml rifampin, a large hydrophobic antibiotic (Table 2A). These phenotypes suggest that the permeability of the outer membrane is increased in the double mutant. A single dsbC⁻ mutant does not exhibit any growth defect or sensitivity toward antibiotics and detergents compared to isogenic wild-type strains.

Fig. 1. The absence of dsbA and dsbC has phenotypical consequences.

A. Growth curves of wild-type (□), dsbC⁻ (▴), dsbA⁻ (*), dsbA⁻dsbD⁻ ( ) and dsbA⁻dsbC⁻ (●) strains in M63 minimal media at 37°C. Growth was monitored at A₆₀₀.

B. SDS sensitivity of wild-type (lane 1), dsbA⁻ (lane 2), dsbA⁻dsbC⁻ (lane 3) and dsbA⁻dsbD⁻ (lane 4) strains. Strains were grown in LB at 37°C to an A₆₀₀ of 0.5. The cultures were then serially diluted 10⁷-fold in 10-fold increments. 10 μl of each dilution were then spotted on LB plates containing 2.5% SDS and grown overnight.

C. Western blot showing protein expression levels. The upper bands correspond to the LamB protein. Outer membrane proteins prepared from wild-type (lane 1), dsbA⁻ (lane 2), dsbC⁻ (lane 3), dsbA⁻dsbC⁻ (lane 4), and LamB⁻ (lane 5) strains. The lower bands correspond to an unknown protein recognized by the anti-LamB antibody, which was used as an internal standard.

Table 2.

Phenotypic characterization of dsbA⁻dsbC⁻ mutants.

Table 2A.
Strain	Genotype	MacConkey Maltodextrin (1%)	Rifampin (4 μg/ml)	Growth rate (h−1) *
JFC209	Wild-type	Mal+ (red)	+++	1.15
MD1	dsbA−	Mal+ (red)	++	0.79
JFC383	dsbC−	Mal+ (red)	+++	1.15
MD3	dsbA−dsbC−	Mal+/− (pink)	−	0.69
AH396	dsbA−dsbD−	Mal+ (red)	++	0.87

Table 2B.
Strain	Genotype	Pfu with Φ80
JP114	Wild-type	5.5 x 1010
JP220	dsbA−	1.0 x 109
JP539	dsbC−	5.1 x 1010
JP557	dsbA−dsbC−	2.0 x 104
JP649	dsbA−dsbD−	7.0 x 109
C600	fhua-	0

^*Growth rate have been calculated using the following formula: μ=ln(A₆₀₀)/time

dsbA⁻dsbC⁻ mutants form pink colonies on maltodextrin MacConkey agar, in contrast to wild-type, dsbA⁻, and dsbC⁻ colonies, which are red (Table 2). This pink phenotype is often indicative of a decreased abundance of LamB, the outer membrane component of the maltose transport system in E. coli (Duguay and Silhavy, 2002). LamB is a trimeric protein that possesses one disulfide bond per subunit. By Western blot analysis, we confirmed that the expression level of LamB is strongly decreased in a dsbA⁻dsbC⁻ mutant compared to wild-type, dsbC⁻, and dsbA⁻ strains (Fig. 1C). The dsbA⁻dsbC⁻ mutants are also significantly more resistant to the lambdoid Φ80 phage (Table 2B). The E. coli receptor for this phage is the ferrichrome iron receptor protein FhuA. Interestingly, FhuA contains four cysteine residues that form two consecutive disulfide bonds. These data suggest that the function of DsbC is not restricted to the correct folding of proteins that have disulfides formed between nonconsecutive cysteines. As the disulfides of LamB and FhuA do not seem to be important for the function of these proteins (Ferenci and Stretton, 1989; Bos et al., 1998), it is tempting to speculate that in the absence of DsbA and DsbC intermolecular disulfides are formed, preventing these proteins from correctly folding in the outer membrane.

Our results clearly show that dsbA⁻dsbC⁻ mutants have a more severe phenotype than dsbA⁻ mutants and suggest that DsbC may be involved in the folding of LamB and FhuA, two proteins that have only consecutive disulfides. Our data thus support the hypothesis that DsbC’s function is not restricted to the correction of non-native disulfides. DsbC also has chaperone activity that is independent of its active site cysteine residues (Liu and Wang, 2001). Thus we needed to consider the possibility that the defects observed in dsbC⁻ strains is due to the lack of DsbC’s chaperone activity. We however observed that expression of a mutant of DsbC in which both catalytic site cysteines are replaced by serine failed to improve the growth rate of dsbA⁻dsbC⁻ strains, in contrast to expression of the wild-type DsbC protein. This mutant of DsbC is expected to lose its thiol-disulfide oxidoreductase activity, while retaining its chaperone activity (Liu and Wang, 2001). Thus the severe phenotype of the dsbA⁻dsbC⁻ mutant is unlikely to be due to a chaperone deficiency in dsbC⁻ strains, but rather a thiol-disulfide oxidoreductase defect.

DsbC is able to assist DsbA in a DsbD-independent manner

To test whether the activity of DsbC always depends on the presence of DsbD, we generated a dsbA⁻dsbD⁻ double mutant. This mutant is phenotypically similar to a dsbA⁻ strain in terms of detergent and antibiotics sensitivity. It is even slightly more resistant to detergent than a dsbA⁻strain. As such, its phenotype is less severe than that of a dsbA⁻dsbC⁻ strain. DsbD functions to keep DsbC reduced in the periplasm so that DsbC can react with non-native disulfides to correct them. The lack of equivalence between a dsbA⁻dsbC⁻ mutant and a dsbA⁻dsbD⁻ mutant indicates that DsbC does not always require DsbD to function in the periplasm.

DsbC is reduced in a dsbA⁻dsbD⁻ mutant

DsbC is found reduced in wild-type cells, but is oxidized in strains lacking DsbD. We determined the in vivo redox state of DsbC in a dsbA⁻dsbD⁻ mutant using AMS trapping. DsbC is found reduced in this genetic background, even when cells are grown in LB, a media that contains small molecule oxidants (Fig. 2). Addition of diamide, a disulfide generating compound, to growing dsbA⁻dsbD⁻ cells leads to immediate oxidation of DsbC (not shown). However, when this disulfide stress is over, DsbC goes back to its reduced state even though DsbD is absent. We propose that DsbC accumulates in the reduced state in the periplasm of dsbA⁻dsbD⁻ strains by donating its disulfide bond to folding proteins.

Fig. 2.

In vivo redox state of DsbC. Exponentially growing cells (in LB) were TCA-precipitated, free cysteines were modified by AMS, and DsbC was detected by Western blot analysis. Lanes: 1, wild-type; 2, dsbA⁻; 3, dsbA⁻dsbD⁻; 4, dsbD⁻.

The deletion of DsbG has no effect

The E. coli periplasm contains another protein disulfide isomerase, DsbG. DsbG has been proposed to function as an isomerase for essentially two reasons: first, like DsbC, DsbG is found reduced in the periplasm and second, DsbG can assist the folding of eukaryotic proteins with multiple cysteine residues when it is overexpressed (Bessette et al., 1999). However, no physiological substrate has been identified so far for DsbG, and the exact function of this protein remains unclear. To see whether DsbG is also able to function in the periplasm of dsbA⁻ strains in a manner similar to DsbC, we constructed a dsbA⁻dsbG⁻ double mutant.

We found that deletion of DsbG does not affect the phenotype of a dsbA⁻ mutant (data not shown). Similarly, a triple dsbA⁻dsbC⁻dsbG⁻ mutant is phenotypically similar to a dsbA⁻dsbC⁻double mutant. However, DsbG may be able to donate its disulfides in ways similar to DsbC as we found the protein to be mostly reduced in dsbA⁻dsbD⁻ double mutants.

2D-LC-MS/MS analysis of dsb⁻ strains

The data presented above suggest that DsbC cooperates with DsbA in a DsbD-independent manner. These observations prompted us to characterize the periplasmic proteome of various dsb⁻strains by two-dimensional liquid chromatographic mass-spec/mass spec analysis (2D-LC-MS/MS) to see whether deletion of both dsbA and dsbC has specific consequences on the protein content of the periplasm. 2D-LC-MS/MS allows a global and semi-quantitative analysis of protein expression ratios. The periplasmic proteomes of wild-type, dsbA⁻, dsbC⁻, dsbA⁻dsbC⁻, and dsbA⁻dsbD⁻ strains were compared.

Two proteins with multiple cysteine were not identified in the dsbC⁻ strains

To examine the consequences of the absence of DsbC at the protein level, wild-type and dsbC⁻strains were grown in minimal media, and periplasmic extracts were prepared. Periplasmic proteins were then digested by trypsin, and the generated peptides were analyzed by 2D-LC-MS/MS. The experiments were repeated three times for both strains. Each run allowed us to identify up to 175 secreted proteins, but only 115 proteins that could be reproducibly identified were kept for further analysis. To our knowledge, this is the first time that a proteomic approach allowed the identification of such a large number of secreted proteins, representing about 18% of all the proteins present in the cell envelope. A number of outer-membrane proteins, probably present in outer-membrane vesicles that did not pellet during the centrifugation, were reproducibly identified. Since they also represent potential targets for DsbA and DsbC, they were kept for further analysis.

The same proteins were identified as being present in both the wild-type and dsbC⁻ strains with the exception of three proteins that were absent in the latter strain. We discovered that these three proteins include DsbC itself and two proteins with multiple cysteine residues: a penicillin insensitive murine endopeptidase (MepA) and an endonuclease (End1). MepA had previously been found to depend on DsbC for expression (Hiniker and Bardwell, 2004), but End1 had not previously been reported to be a DsbC substrate. The structure of Vibrio cholera End1 shows four disulfide bonds, one of which is formed between nonconsecutive cysteines(Altermark et al., 2006). The eight cysteines that form these four disulfide bonds are conserved in the E. coli End1 protein. The E. coli and V. cholera proteins are 66% identical at the amino acid sequence level; this suggests that they have similar structures and makes it almost certain that the E. coli protein shares the V. cholera disulfide bond pattern. Since DsbC is required for the formation of nonconsecutive disulfides, the folding of End1 is likely to require the presence of DsbC.

We then searched for proteins that, although present in both strains, vary substantially in their abundance. For quantification of abundance, we used the number of spectral counts (SC) reported for every protein. The number of SC for a protein is the total number of MS/MS spectra taken on peptides from this protein in a given 2D-LC-MS/MS analysis. This value is linearly correlated with the protein abundance over a dynamic range of two orders of magnitude (Liu et al., 2004). Protein ratios determined by spectral counting agree well with those determined from peak area intensity measurements and are consistent with independent measurements based on gel staining intensities (Old et al., 2005). To validate this quantification method, we added varying amounts (2–60 pmoles) of two eukaryotic proteins, ovalbumin and carbonic anhydrase, to 300 μg of periplasmic proteins. Linear regression based on different sampling statistics was performed for each of the 2D-LC-MS/MS runs; the R² values obtained for the spectral counts were 0.94 and 0.92 for ovalbumin and carbonic anhydrase, respectively (Fig. 3). We concluded that this method is quantitative and that the ratio of spectral counts reliably reflects changes in protein expression levels.

Fig. 3.

The number of spectral counts correlates with the abundance of a protein. Varying amounts (2–60 pmoles) of two eukaryotic proteins, ovalbumin and carbonic anhydrase, were added to 300 μg of periplasmic proteins. The SC values obtained for these two proteins in the various samples were then plotted against the corresponding protein amounts. After linear regression, we found that the R² values obtained for the spectral counts were 0.94 and 0.92 for ovalbumin and carbonic anhydrase, respectively. This indicates that the number of spectral counts reliably reflects protein abundance in the sample.

We selected proteins whose abundance was decreased or increased by at least two-fold. To test the significance of the data, we used the unpaired Student's t test and defined significance as a P < 0.05 (2-tail 2-sample equal variance test). No protein was more abundant in the dsbC⁻ strain (Table S1), and only two were decreased. YebF, a small protein with an unknown function, was the protein most decreased by the absence of DsbC (6 SC instead of 36), whereas Ivy, an inhibitor of lysozyme, was about two-fold less abundant. Both of these proteins have two cysteine residues, and our results suggest that they may partially depend on DsbC for correct folding.

Deletion of both dsbA and dsbC affects the global protein content of the periplasm

The results of the analysis of the periplasmic proteome of dsbC⁻ and wild-type strains by 2D-LC-MS/MS agree well with those obtained using 2D gels (Hiniker and Bardwell, 2004). This indicates that our 2D-LC-MS/MS method is reliable and should allow for the detection of changes in the periplasmic protein content of other dsb⁻ strains. Periplasmic extracts were prepared from dsbA⁻, dsbA⁻dsbC⁻, and dsbA⁻dsbD⁻ mutants, proteins were digested, and peptides were separated by HPLC followed by LC-MS/MS analysis. The results from the MS/MS analysis were then compared to those obtained previously for the wild-type strain. First, the expression levels of several proteins were dramatically modified in all the strains that lack DsbA. These differences, which are described below, are consistent with the previously reported role of DsbA in oxidative protein folding. Second, we found that a dsbA⁻dsbC⁻ double mutant has a significantly altered periplasmic proteome when compared to dsbA⁻ and dsbA⁻dsbD⁻ strains. This is reflected by the graphs shown in Fig. 4. The SC values from the dsbA⁻ and dsbA⁻dsbD⁻ strains are linearly distributed, reflecting similar protein content. Comparison of the dsbA⁻ and dsbA⁻dsbD⁻ mutants shows indeed that there is only one protein (Spy) whose abundance is significantly different between these two strains (see supplementary table 2). In contrast, when the SC values reported for proteins from the dsbA⁻dsbC⁻ mutant are plotted against those from the dsbA⁻ strain, the distribution is much more dispersed. This indicates a distinct overall protein content of the dsbA⁻dsbC⁻ mutant and confirms the deleterious effect of the simultaneous absence of DsbA and DsbC on the periplasm. One possibility is that the broader dispersion of the SC values observed in the dsbA⁻dsbC⁻ mutant reflects an increased sensitivity of this strain to the osmotic shock procedure. However, this would probably be reflected by an increased overall yield of proteins in periplasmic extracts, not decreased amounts of specific proteins in these extracts. The distinct protein content of dsbA⁻dsbC⁻ strains is a new finding and suggests that DsbC and DsbA cooperate in the folding of proteins. Specifically our finding suggests that DsbC can take over the function of DsbA. This is not consistent with the assigned roles of DsbA as an oxidase and DsbC as an isomerase. It suggests that a revision of the current model of disulfide bond formation in E. coli is called for. We note that a more likely interpretation is that DsbC acts to compensate for the absence of DsbA, and DsbA and DsbC cooperate in helping to fold proteins.

Fig. 4. The overall protein content of a dsbA⁻dsbC⁻ mutant is different compared to dsbA⁻ and dsbA⁻dsbD⁻ strains.

A. The logarithms of the SC values reported for the dsbA⁻ strain were plotted against those reported for the dsbA⁻dsbD⁻ mutant. Most of the SC values are similar in both strains, which is reflected by a quasi-linear distribution.

B. The logarithms of the SC values reported for the dsbA⁻ strain were plotted against those reported for the dsbA⁻dsbC⁻ mutant. The distribution is more dispersed, which indicates that the overall protein content of this double mutant is different.

Several cysteine-containing proteins are less abundant in a dsbA⁻dsbC⁻ strain

We searched for proteins with a significantly decreased abundance in the dsbA⁻dsbC⁻ strain, compared to the dsbA⁻ and dsbA⁻dsbD⁻ strains. Ten proteins were not detected in the double dsbA⁻dsbC⁻ mutant (Cn16, DsbC, FepB, GltI, Slp, SubI, YcfS, YggN, YjhT, and YnjE), and 10 were at least two-fold less abundant in the dsbA⁻dsbC⁻ than in the dsbA⁻ and the dsbA⁻dsbD⁻mutants (P < 0.05) (Table 3). Interestingly, six of these proteins (Cn16, GltI, YggN, OppA, TreA, and YhjJ) have two cysteine residues. Noteworthy, GltI, YggN, and OppA are known DsbA substrates (Hiniker and Bardwell, 2004; Kadokura et al., 2004). Two other cysteine-containing DsbA substrates, PhoA and DppA, were also more than 2 fold less abundant in the dsbA⁻dsbC⁻ mutant than in the dsbA⁻ or the dsbA⁻dsbD⁻ mutant. However, the decrease was not statistically significant when compared to the dsbA⁻dsbD⁻ mutant (see Supplementary Table 2) and these proteins were not included in Table 3. Altogether, our data show that the simultaneous absence of both DsbA and DsbC decreases the level of several proteins, including several that contain cysteine residues. In contrast, the absence of DsbD has no effect on dsbA⁻ strains. Determination of RNA expression levels (Table 3) showed that, for most of the decreased proteins, their lower abundance is not due to a decreased transcription and is therefore likely to represent a direct consequence of the absence of DsbC. Altogether, our results further support the hypothesis that DsbC can assist DsbA in a DsbD-independent manner.

Table 3.

Proteins more than two-fold less abundant in a dsbA⁻dsbC⁻ mutant than in dsbA⁻ and dsbA⁻dsbD⁻strains (P < 0.05)¹.

The levels of expression of the genes coding for these proteins in the dsbA⁻dsbC⁻ double mutant relative to the dsbA⁻ mutant are shown in grey.

Protein	# Cysteines	Spectral counts			RNA2(dsbA−dsbC− vs dsbA−)
		dsbA−	dsbA−dsbC−	dsbA−dsbD−
Proteins with at least two cysteine residues
Cn16	2	2	0	3	NSD
DsbC	4	12	0	8	↓ (110)
GltI	2	5	0	6	NSD
OppA3	2	305	137	259	NSD
YggN	2	2	0	2	NSD
TreA3	2	8	1	11	nd
YhJJ	2	10	2	9	NSD
Proteins with one or no cysteine residue
FepB	0	5	0	3	NSD
FliY	0	151	65	132	NSD
SubI	0	9	0	6	NSD
YcfS	1	2	0	2	NSD
GgT	0	8	3	8	NSD
MalE	0	41	1	24	nd
MppA	0	14	3	15	NSD
PhnD	0	159	79	168	NSD
Slp	1	5	0	3	↓ (2.9)
YliB	1	10	4	11	NSD
YjhT	0	4	0	2	NSD
YnjE	1	14	0	6	NSD
YncE	0	15	4	14	NSD

¹All proteins that were identified in the dsbA⁻ and dsbA⁻dsbD⁻ strains but were absent in all three independent analysis of the dsbA⁻dsbC⁻ were also considered as significantly less abundant in the latter strain.

²Genes whose expression ratios were >2 fold different (P<0.05) in the dsbA⁻dsbC⁻ double mutant relative to the dsbA⁻ mutant were considered as induced (↑) or repressed (↓). All other genes were considered as not significantly different (NSD). The ratio indicated next to the arrows is the fold increase or decrease in transcript level in the dsbA⁻dsbC⁻ strain compared to the dsbA⁻ strain. “nd” means that no data were obtained for the corresponding gene.

³P=0.05 when SC values from the dsbA⁻dsbC⁻ are compared to those from the dsbA⁻dsbD⁻.

A complete list with all the identified proteins is available in Table S2. Microarray data for the genes corresponding to the identified proteins are available in Table S3.

Proteins are decreased in all strains impaired in disulfide bond formation

In all strains lacking DsbA, the relative abundance of several proteins is dramatically modified compared to wild-type (Table 4). In 125 proteins that were identified, we observed that the abundance of about 50 proteins was modified by at least two-fold.

Table 4. Proteins whose abundance is significantly changed in dsbA⁻ strains.

The levels of expression of the genes coding for these proteins in the dsbA⁻ mutant relative to the wild-type are shown in grey.

Proteins > two-fold less abundant in dsbA− strains (P < 0.05)1			RNA2dsbA− vs wild-type	Proteins > two-fold more abundant in dsbA− strains (P < 0.05)			RNA2 (dsbA− vs wild-type)
Protein	Spectral counts			Protein	Spectral counts
	wild-type	dsbA			Wild-type	dsbA−
ArgT (2 cys)a	54	9	NSD	DegP (2 cys)	1	27	↑ (2.0)
ArtJ (2 cys)a	301	150	↑ (3.5)	FkpA	24	62	NSD
DppA (4 cys)	99	34	NSD	GgT	0	8	↑ (2.3)
GltI (2 cys)	49	5	NSD	IvY (2 cys)	26	224	↑ (7.9)
HisJ (2 cys)	170	84	↓ (2.6)	OmpA (2 cys)	131	401	NSD
PhoA (4 cys)	212	29	↑ (19.4)	OmpX	27	152	NSD
ProX (2 cys)a	46	8	NSD	OpgG	8	32	NSD#
YfhM (2 cys)	1	0	NSD	OsmE	16	39	NSD#
YebF (2 cys)a	36	1	↑ (3.3)	OsmY	104	285	↑ (6.0)
YggN (2 cys)	5	2	NSD	PotF (2 cys)	9	31	↑ (9.6)
YtfQ (2 cys)a	8	0	↑ (27.7)	RseB	0	3	nd
CirA (2 cys)a	8	0	NSD	Spy	0	10	↑ (5.3)
CreA (2 cys)a	2	0	NSD	TreA (2 cys)	1	8	nd
End1 (8 cys)a	1	0	nd	WzA	0	8	↑ (9.3)
MepA (6 cys)	4	0	NSD	YbiS	8	30	↑ (4.5)
RNase I(8 cys)	4	0	NSD	YdeI	11	108	↑ (9.7)
YebY (2 cys)a	8	0	NSD	YehZ	7	41	↑ (2.8)
DsbA	131	0	↓ (3.4)	YggG (3 cys)	1	14	↑ (3.7)
FlgH	5	0	NSD	YgiW	6	20	NSD
FliC	219	0	↑ (92.5)	YhjJ (2 cys)	3	10	NSD
OmpF	136	13	↓ (10.7)	YjbG	0	6	↑ (52.0)
PhoE	86	9	↑ (3.9)	YjbH (3 cys)	0	5	↑ (5.9)
FlgG	5	0	↓ (*)	YjbF	0	2	nd
OmpN	2	0	NSD	YpfG (6 cys)	0	3	↑ (4.5)
				YraP	1	3	NSD

DsbA substrates are underlined.

^aNewly identified DsbA substrates.

^*flgG transcripts were not detected in the dsbA⁻

^#osmE and opgG transcription rates were increased by 1.82 (P=0.026) and 1.7 (P=0.0039), respectively.

¹All proteins that were identified in the wild-type strain but were absent in all three independent analysis of the dsbA⁻ were also considered as significantly less abundant in the latter strain.

²Genes whose expression ratios were >2 fold different (P<0.05) in the dsbA⁻ mutant relative to the wild-type were considered as induced (↑) or repressed (↓). All other genes were considered as not significantly different (NSD). The ratio indicated next to the arrows is the fold increase or decrease in transcript level in the dsbA⁻ strain compared to the wild-type strain. “nd” means that no data were obtained for the corresponding gene.

A complete list with all the identified proteins is available in Table S1. Microarray data for the genes corresponding to the identified proteins are available in Table S4.

In addition to DsbA, eight proteins with at least two cysteine residues were missing in dsbA⁻ strains. Two of these proteins (MepA and End1) were also missing in the dsbC⁻ mutant, which suggests they require the presence of both DsbA and DsbC. The other cysteine containing proteins are a periplasmic ribonuclease (RNase I), the outer membrane colicin 1 receptor protein (CirA), a protein that is exported to the periplasm according to Psort (CreA), and three proteins with an unknown function (YebY, YtfQ, and YfhM). We also found nine proteins with two or more cysteine residues whose abundance was decreased by at least two-fold (P < 0.05) (Table 4). HisJ, GltI, DppA, PhoA and YggN are known DsbA substrates (Hiniker and Bardwell, 2004; Kadokura et al., 2004), but the other proteins (ProX, ArgT, ArtJ, and YebF) had not previously been identified to depend on DsbA for correct folding. Sequence analysis revealed that the cysteine residues of these proteins are conserved in homologous sequences, which suggests that they are structurally important and probably form disulfide bonds. Our results allow us to add these proteins to the list of the potential DsbA substrates. The formation of a disulfide in YtfQ, YebF, and ArtJ was confirmed by differential thiol trapping (see below).

The absence of DsbA also leads to decreased levels of several proteins that do not contain cysteine residues (Table 4). This can either be a direct consequence of a misfolding problem caused by the absence of DsbA or can result from a decreased transcription of their genes. To discriminate between these two possibilities, we determined the RNA expression levels of the genes coding for these proteins (Table 4). Our data show that the absence of DsbA has no consequence on the transcription of most of these genes and that some genes, such as artJ, phoA, and yebF are even induced. This indicates that the decreased protein abundance is due to an impaired folding of these proteins. Because DsbA does not have a chaperone activity, we propose that envelope perturbations in dsbA⁻ strains prevent these proteins to correctly fold in the periplasm. In contrast, we found that the RNA expression levels of ompF and flgG were significantly decreased, which indicates that the diminution in the corresponding protein abundance is due to a decreased transcription of their genes. Regarding the outer membrane protein OmpF, our data agree with previous results (Pugsley, 1993). The other protein, FlgG, is a protein of the bacterial flagellum. Previous reports have shown that in a dsbA⁻ strain, the flagellar P-ring protein FlgI is not properly folded and is degraded (Dailey and Berg, 1993). The degradation of FlgI prevents the assembly of a functional flagellum, which leads to the repression of the transcription of other flagellum genes (Chilcott and Hughes, 2000). The misfolding of FlgI is therefore the reason why the transcription of flgG is repressed and the corresponding protein is not identified in the dsbA⁻ strain. Noteworthy, we found that the transcription of flgH, the gene coding for the other flagellum protein that was decreased in our proteomics analysis, was also diminished, but by less than 2 fold.

Several stress-related proteins are more abundant in strains lacking DsbA

The absence of DsbA leads to increased levels of 25 proteins, including nine proteins that were not identified in the wild-type. Similar data were obtained for the dsbA⁻dsbC⁻ and dsbA⁻dsbD⁻mutants (see supplementary tables). Determination of the RNA expression levels allowed us to show that the transcription rates of 14 of the genes coding for these proteins are increased (Table 3).

Seven of these 25 proteins are part of the SigmaE regulon. SigmaE is a transcriptional activator that controls the expression of a variety of genes involved in maintaining the integrity of the cell envelope (reviewed in Ruiz and Silhavy, 2005). SigmaE is induced under conditions of stress in the cell envelope, including accumulation of misfolded outer membrane proteins in the periplasm, aberrant lipopolysaccharides, and lack of periplasmic folding agent. Gross and coworkers already showed that the absence of DsbA leads to an increased transcription of the gene coding for SigmaE (Mecsas et al., 1993), but this is the first time that the induction of SigmaE was confirmed at the protein level. The SigmaE regulon members whose abundance is increased in the dsbA⁻ strain include the outer membrane proteins OmpA and OmpX, the periplasmic chaperone FkbA, the periplasmic proteases DegP and YhjJ, a protein involved in the biosynthesis of osmoregulated glycans (OpgG) and a negative regulator of SigmaE activity (RseB). The induction of these proteins further indicates that lack of disulfide bond formation leads to a global stress in the cell envelope.

In addition to the induction of SigmaE regulated proteins, we observed the induction of proteins that are known to be induced under high osmotic pressure: OsmE, an osmotically-inducible lipoprotein (Bordes et al., 2002), OsmY, a small protein of unknown function that has been proposed to interact with phospholipids on both sides of the periplasm (Lange et al., 1993), and the periplasmic trehalase TreA (Repoila and Gutierrez, 1991). Determination of the expression levels of the genes coding for these proteins allowed us to confirm that the increased protein abundance of OsmE and OsmY can be attributed to an increased RNA synthesis. Similar changes in the abundance of these three proteins, as well as a decreased transcription of ompF (see above), have been observed in strains grown under high osmotic pressure, suggesting that dsbA⁻ strains mimic the effects of increased osmotic pressure. The induction of Spy, a protein that is specifically induced in spheroplasts (Hagenmaier et al., 1997), suggests that the induction of these osmo-related proteins in dsbA⁻ strains may be the consequence of an altered peptidoglycan layer.

Determination of the in vivo redox state of periplasmic proteins by differential thiol trapping

To confirm the presence of disulfide bonds in the newly identified DsbA substrates, we adapted the differential thiol trapping technique developed by (Leichert and Jakob, 2004) to determine the redox state of the cysteine residues present in the periplasm of dsbA⁻ and wild-type strains.

As expected, the majority of the cysteine residues identified in peptides from the wild-type strain were oxidized (Table 5). In contrast, more reduced cysteine residues were found in proteins from the dsbA⁻ strain. In particular, cysteine residues from known DsbA substrates including OmpA, PhoA, DppA, and GltI were found oxidized in the wild-type and, when detected, reduced in the dsbA⁻ strain. The differential thiol trapping technique allowed us to confirm the formation of a disulfide bond in three newly identified DsbA substrates, ArtJ, YebF and YtfQ.

Table 5.

Redox state of cysteine-containing peptides.

Protein	cysteine-containing peptides found	wild-type (% red)	dsbA− (% red)	dsbA−dsbD− (% red)	dsbA−dsbC− (% red)
PhoA	ATYHGNIDKPAVTCTPNPQR	2	45	66	66
	CYGPSATSEK	0	-	-
	QDHAANPCGQIGETVDLDEAVQR	0	50	33	50
OmpA	AALIDCLAPDRR	0	47	32	52
	GMGESNPVTGNTCDNVK	2	85	83	90
ArtJ	QMQAECTFTNHAFDSLIPSLK	11	100	100	80
YebF	SADIHYQVSVDCK	15	-	-	-
	CEDLDAAGIAASVK	0	-	-	-
DppA	NECQVMPYPNPADIAR	0	-	-	-
GltI	PQSQEAYGCMLR	0	-	-	-
MepA	TPPPLPPSCQALLDEHVI	0	-	-	-
RNaseI	AVKLTCQGNPAYLTEIQISIK	0	-	-
YneA	VLTWDSDTKPECR	0	-	-	-
YtfQ	KPCNVVELQGTVGASVAIDR	0	-	-	-
EcoT	VSSPVSTMMACPDGK	-	100	100	100
	VELLIGQTLEVDCNLHR		100	100	100
YedD	VDRPTAECAAALDK	-	75	100	60
YggG	TLSDQACQEMDSK	-	66	66	50

We also determined the redox state of the cysteine residues in the periplasm of the dsbA⁻dsbC⁻ and dsbA⁻dsbD⁻ double mutants and we found that they are similar to those observed in the dsbA⁻ strain.

A revised model for the oxidative protein folding pathways in E. coli

In conclusion, our results show that the simultaneous absence of DsbA and DsbC leads to a decreased integrity of the cell envelope and affects the global protein content of the periplasm. In contrast, strains lacking both DsbA and DsbD, the protein that is responsible for keeping DsbC active as an isomerase, do not share these characteristics. Our results suggest therefore that DsbC cooperates with DsbA in a DsbD-independent manner to ensure the correct folding of E. coli envelope proteins.

Kinetic, structural and genetic data showed that DsbB is unable to oxidize DsbC at physiological rates, unless the dimerization domain is removed and DsbC is expressed as a monomeric protein (Bader et al., 2001). Similarly, DsbD is unable to reduce DsbA (Rozhkova et al., 2004). This led to the assumption that the DsbA/DsbB oxidation pathway was isolated from the DsbC/DsbD isomerization pathway. Our results show that one can open a door in the barrier separating the oxidative and isomerization pathways. Our results indicate that, in contrast to the current view, DsbC can function independently of DsbD and is therefore able to function in both the oxidation and isomerization pathways. When DsbC gets oxidized upon reduction of a non-native disulfide, it is either reduced by DsbD or by transferring its disulfide to a reduced protein. DsbC may possibly be acting as a stand-alone protein folding catalyst that is able to cycle from the reduced to the oxidized state upon substrate oxidation and substrate reduction, respectively. This activity of DsbC seems important to maintain the integrity of the cell envelope and is not restricted to the correction of non-consecutive disulfides. Our results extent those from Bader and co-workers who showed that monomeric mutants of DsbC are substrates for DsbB and can catalyze disulfide bond formation (Bader et al., 2001). On the basis of our results, we have adapted the model of disulfide bond formation in the E. coli periplasm (Fig. 5).

Fig. 5. A revised model for the formation of disulfide bonds in the E. coli periplasm.

Disulfide bonds are introduced by the DsbA/DsbB pathway. Non-native disulfides are corrected by DsbC, which is recycled by DsbD. Both pathways are kinetically isolated. Our results indicate that DsbC is also able to function on the other side of the barrier where it assists DsbA in a DsbD-independent manner. DsbC may be acting as a stand-alone protein folding catalyst that cycles from the reduced to the oxidized state upon substrate oxidation and substrate reduction, respectively. Although kinetics data showed that DsbC is not a good substrate for DsbB, we cannot exclude that a slow oxidation of DsbC by DsbB may play a more significant role in the absence of DsbA. The Western blot data presented in Figure 2 also suggest that in the absence of DsbD, DsbA may be responsible for the oxidation of DsbC. The redox potentials of DsbA and DsbC are −125 mV and −130 mV, respectively.

Experimental procedures

Bacterial strains and growth conditions

The bacterial strains used in this study are described in Table 1. Strains JP114, JP220, JP539, JP557, JP649 and C600 were used for the titration experiment with phage F80. All the other experiments were performed with strains AH50, JFC383, AH396, MD1 and MD3 in the MC1000 background. Strains JFC383, MD1, and MD3 were constructed by P1 transduction. Cells were grown aerobically in either LB or M63 minimal media, at 37°C. Unless otherwise indicated, M63 minimal medium was supplemented with 0.2% glucose, vitamins (Thiamine 10 μg/ml, Biotine 1 μg/ml, Riboflavine 10 μg/ml, and Nicotinamide 10 μg/ml), 1 mM MgSO₄, leucine (20 μg/ml ) and isoleucine (20 μg/ml). Sensitivity to antibiotics was assayed by streaking the strains on LB plates containing 4 μg/ml rifampin. To test the sensitivity to SDS, strains were grown in LB at 37°C to an A₆₀₀ of 0.5. The cultures were then serially diluted 10⁷-fold in 10-fold increments. 10 μl of each dilution were then spotted on LB plates containing 2.5% SDS and grown overnight. To study the ability of dsb⁻ strains to assemble a functional LamB protein, strains were streaked on MacConkey agar indicator plates containing 1% maltodextrin.

Table 1.

Strains used in this study and their relevant genotypes.

Strain	Relevant genotype	Source
AH50	MC1000 phoR Δara714leu+ phoA68	Hiniker et al., 2005
JFC383	AH50 dsbC::kan	This study
AH396	AH50 dsbD::cm, dsbA::kan1	Hiniker et al., 2005
MD1	AH50 dsbA::kan1	This study
MD3	AH50 dsbC::cm, dsbA::kan	This study
JP114	ER1821	New England Biolabs
JP220	JP114 ΔdsbA::kan	This study
JP539	JP114 ΔdsbC::cm	This study
JP557	JP220 ΔdsbC::cm	This study
JP649	JP114 ΔdsbA::kan dsbD−::cm	This study
C600	JP114 fhuA::kan	This study

Φ80 phage titration

Cells were grown overnight to late logarithmic phase at 30°C. 100 μL of cells was used to inoculate 3 mL of LB top agar (0.7% LB agar). The suspension was vortexed and plated onto a pre-warmed LB plate. Serial dilutions of F80 stock (> 10¹¹ pfu/mL) were made at 10⁻³, 10⁻⁶, and 10⁻⁹. 5 μL of these serial dilutions was spotted onto the LB agar plate containing cells, and allowed to incubate at 30°C overnight (16–18 h). The number of plaques and plaque sizes were tabulated.

Expression of a DsbC SXXS mutant

The catalytic site cysteine residues of DsbC were replaced by serine residues using the QuickChange Mutagenesis Protocol (Stratagene). Both the wild-type and mutated DNA sequence were then inserted in the pBAD33 expression plasmid. The plasmids were transferred into the MD3 strain and DsbC expression was induced by adding L-arabinose (0.2 %).

Periplasmic extracts preparation

Cells (100 ml) were grown aerobically at 37°C in M63 minimal media to an A₆₀₀ of 0.8, and periplasmic extracts were prepared as in (Hiniker and Bardwell, 2004). Protein concentration was determined using the Bradford assay.

Differential thiol trapping and digestion

300 μg of periplasmic proteins were precipitated by adding trichloroacetic acid (TCA) to a final concentration of 10% w/v, followed by incubation on ice for 30 min. Samples were then centrifuged at 14,000 rpm for 20 min and the resulting pellets washed with 5% ice cold TCA. The pellets were then resuspended in 100 μl denaturing buffer (6M urea, 200 mM Tris-HCl pH 8.5, 10 mM EDTA) supplemented with 100 mM iodoacetamide. At this stage, various amounts ranging from 2 to 60 pmoles of carbonic anhydrase and ovalbumin were added to the samples as internal standards. After a 20 min incubation at 25°C, the reaction was stopped by adding 10 μl of ice cold 100% TCA and left on ice for 20 min. The alkylated proteins were centrifuged and the pellet washed as described above. The proteins were then dissolved in 100 μl of 10 mM DTT in denaturing buffer. After a 1 h incubation at 25°C, 100 μl of denaturing buffer supplemented with 100 mM N-ethylmaleimide was added to titrate out the remaining DTT and alkylate all newly reduced cysteines. The reaction was stopped by addition of 10% TCA and the proteins collected by centrifugation. The resulting pellet was successively washed with TCA and ice cold acetone, dried in a Speedvac, resuspendend in 0.1 M NH₄HCO₃ pH 8.0 with 3 μg sequencing grade trypsin, and digested overnight at 30°C. Peptide samples were then acidified to pH 3.0 with formic acid and stored at −20°C.

Differential analysis of periplasmic proteins by label-free 2D-LC-MS/MS

Peptides were loaded onto a strong cation exchange column GROM-SIL 100 SCX (100 × 2 mm, GROM, Rottenburg, Germany) equilibrated with solvent A (5% acetonitrile v/v, 0.05% v/v formic acid pH 2.5 in water) and connected to an Agilent 1100 HPLC system. Peptides were separated using a 50 min elution gradient that consisted of 0%–50% solvent B (5% acetonitrile v/v, 1 M ammonium formate adjusted to pH 3.0 with formic acid in water) at a flow rate of 200 μl/min. Fractions were collected at 2 min intervals (20 in total) and dried using a Speedvac. Peptides were resuspended in 10 μl of solvent C (5% acetonitrile v/v, 0.01% v/v TFA in water) and analyzed by LC-MS/MS as described below.

The LC-MS/MS system consisted of an LCQ DECA XP Plus ion trap mass spectrometer (ThermoFinnigan, San José, CA, USA) equipped with a microflow electrospray ionization source and interfaced to an LCPackings Ultimate Plus Dual gradient pump, Switchos column switching device, and Famos Autosampler (Dionex, Amsterdam, Netherlands). Two reverse phase peptide traps C18 Pepmap 100 Dionex (300 μm × 5 mm) were used in parallel with two analytical BioBasic-C18 columns from ThermoElectron (0.18 mm × 150 mm). Samples were injected and desalted on the peptide trap equilibrated with solvent C at a flow rate of 30 μl/min. After valve switching, peptides were eluted in backflush mode from the trap onto the analytical column equilibrated in solvent D (5% acetonitrile v/v, 0.05% v/v formic acid in water) and separated using a 100 min gradient from 0% to 70% solvent E (80% acetontrile v/v, 0.05% formic acid in water) at a flow rate of 1.5 μl/min.

The mass spectrometer was set up to acquire one full MS scan in the mass range of 400–2000 m/z, followed by three MS/MS spectra of the three most intense peaks in the mass range 400-1500 m/z. The dynamic exclusion feature was enabled to obtain MS/MS spectra on co-eluting peptides, and the exclusion time was set at 2 min.

Protein identification

Raw data collection of approximately 54,000 MS/MS spectra per 2D-LC-MS/MS experiment was followed by protein identification using the TurboSequest algorithm in the Bioworks 3.2 software package (ThermoFinnigan) against an E. coli protein database (SwissProt) using the following constraints: only tryptic peptides up to one missed cleavage site were allowed; tolerances for MS and MS/MS fragment ions were set to 1.2 Da and 1.0 Da, respectively; and methionine oxidation (+ 16.0 Da), carboxamidomethyl cystein or N-ethylmaleimide cysteine (+ 57.0 Da or + 125.0 Da, respectively) were specified as variable modifications. The identified peptides were further evaluated using charge state versus cross correlation number (Xcorr). The criteria for positive identification of peptides were Xcorr > 1.5 for singly charged ions, Xcorr > 2.0 for doubly charged ions, and Xcorr > 2.5 for triply charged ions. Protein scores (Su, Xcorr), peak areas, and spectral counts were calculated within BioWorks 3.2. The data were converted into Microsoft Excel spreadsheets by the export function contained in BioWorks and the output files were compared and processed by an in house software program. Relative quantification of protein abundance was estimated by calculating the ratio of spectral counts determined within the BioWorks software package. This parameter was shown to follow a linear relationship over two orders of magnitude, as determined from spiked internal standard proteins, within a dynamic range of at least 10³.

Preparation of outer membrane proteins

Outer membrane proteins were prepared from strains grown in LB. Cultures were grown to an A₆₀₀ of 0.8; cells were harvested by centrifugation at 6000 rpm for 10 min then resuspended in 25 mM Tris pH 8.0, 0.5 M sucrose, 1 mM EDTA, and 0.25 mg/ml lysozyme. After 15 min at room temperature, 20 mM MgCl₂ was added. The extracts were then centrifuged for 5 min at 12,000 rpm. The pellets were discarded and supernatants were centrifuged at 45,000 rpm for 1 h at 4°C. The pellets were resuspended in Laemli buffer and loaded on SDS-PAGE.

In vivo redox state of DsbC

Cells were grown in LB at 37°C to an A₆₀₀ of 0.8, and 1 ml samples were taken. Proteins were precipitated with 5% ice cold TCA and centrifuged at 16,000 × g for 15 min. The pellets were washed with acetone, dried, and resuspended in 50 mM Tris-HCl, pH 7.5, 0.1% SDS, 10 mM EDTA, and 10 mM AMS. AMS is a reagent that covalently reacts with free thiol groups, adding a 490 Da group. This leads to a major mobility shift respective of modified protein in SDS-PAGE gels. Samples were analyzed by SDS-PAGE under nonreducing conditions.

Antibodies

Antibodies against LamB were kindly provided by Natividad Ruiz and Tom Silhavy (Princeton), and antibodies against DsbC were provided by Jon Beckwith (Harvard).

Microarray analysis

RNA from WT, dsbA- and dsbA-dsbC- strains was extracted using the Tripure reagent and the RNeasy purification kit (Qiagen). Microarray analysis were performed in triplicates by using “GeneChip® E. Coli Genome 2.0 Array” and the protocol provided by Affymetrix for prokaryotic expression analysis.