Introduction

The formation of structural disulfide bonds in Escherichia coli appears to be strictly segregated according to subcellular compartment. In the periplasm, disulfide bonds are actively formed in many proteins by the Dsb system (Rietsch and Beckwith, 1998). In the cytoplasm, the only disulfide bonds known to be present in proteins are formed in enzymes like ribonucleotide reductase during their catalytic cycles or in the oxidative response transcription factor OxyR during its regulatory cycle. The reduced forms of these proteins are regenerated via the action of the thioredoxin and glutathione/glutaredoxin pathways (Åberg et al., 1989; Zheng et al., 1998).

In the thioredoxin pathway, thioredoxin reductase (the product of the trxB gene) uses the reducing potential of NADPH to maintain thioredoxin 1 (the product of the trxA gene) in the reduced state, so that thioredoxin 1 in turn can reduce substrate proteins such as ribonucleotide reductase (Figure 1). The glutathione/glutaredoxin system also uses the reducing potential of NADPH in this case to reduce glutathione via the enzyme glutathione oxidoreductase. Glutathione is then able to reduce the three glutaredoxins (glutaredoxin 1, 2 and 3) (Holmgren, 1989). Only glutaredoxin 1 is able to reduce ribonucleotide reductase efficiently in vitro, whereas glutaredoxin 3 has a modest ability to reduce this enzyme (Åslund et al., 1994). Although glutaredoxins 2 and 3 are less efficient at reducing protein disulfides, they are active in reducing mixed disulfides of glutathione. Such mixed disulfides are generated and must be resolved during the catalytic cycle of enzymes such as arsenate reductase (Liu and Rosen, 1997). In E.coli, either the thioredoxin or the glutaredoxin pathway can be disabled by mutation without serious detriment to the cell. Nevertheless, if both pathways are disrupted, aerobic growth is almost completely eliminated, suggesting overlap between these two systems for the reduction of essential substrates such as ribonucleotide reductase (Prinz et al., 1997). However, for some substrates, such as methionine sulfoxide reductase, an apparent specificity towards thioredoxin 1 was deduced from the phenotype of trxA mutants (Russel and Model, 1986).

Structural disulfide bonds do not ordinarily form in cytoplasmic proteins (Pollitt and Zalkin, 1983; Derman and Beckwith, 1991). However, such bonds can form in cells that are defective for certain components of these reducing pathways (Derman et al., 1993a; Prinz et al., 1997). The formation of structural disulfide bonds in the cytoplasm of E.coli was assessed in strains in which the normally periplasmic enzyme alkaline phosphatase (AP) was localized to the cytoplasm by deleting its signal sequence (ssAP). AP is useful for this purpose since it contains two disulfide bonds that are required for it to fold into an active conformation (Sone et al., 1997).

We have found that strains mutant for the enzyme thioredoxin reductase (trxB mutants) express high levels of active cytoplasmic ssAP (Derman et al., 1993a). We considered two hypotheses to explain the oxidative properties of the trxB strain. According to the first hypothesis, the absence of reducing potential in the thioredoxin pathway (due to the loss of thioredoxin reductase) results in a failure to reduce ssAP that is spontaneously oxidized. According to the second hypothesis, oxidized thioredoxin accumulates in the trxB cytoplasm and can act to promote disulfide bond formation in ssAP. However, the properties of a trxA mutant and a double trxA, trxB mutant were not compatible with either hypothesis. These findings led us to suggest that there might be a hitherto undiscovered thioredoxin in E.coli (Derman et al., 1993a), a suggestion reinforced by subsequent studies in this laboratory (Prinz et al., 1997). Based on this reasoning, we sought and found an open reading frame in the E.coli genome sequence that could code for a new thioredoxin homolog, and initiated in vivo genetic studies on its function. During the course of this work, the purification of this gene (trxC) product, termed thioredoxin 2, and the demonstration of its function in vitro as a thioredoxin, was reported (Miranda-Vizuete et al., 1997).

In this paper, we utilize null mutations in the trxA and trxC genes to investigate the mechanism by which disulfide bond formation in ssAP occurs in the absence of thioredoxin reductase. We show that the two thioredoxins (thioredoxin 1 and 2) are both necessary for AP activity in the cytoplasm. We propose that the oxidized thioredoxins that accumulate in a trxB mutant can actively promote disulfide bond formation in appropriate substrate proteins. Further, we investigate the role of thioredoxin 2 as a disulfide-reducing protein in vivo, and show that it can fulfill many of the roles of thioredoxin 1. We also describe a strain that can be used to identify disulfide-reducing proteins from E.coli and other organisms.

Results

When AP is expressed in the cytoplasm of cells missing the enzyme thioredoxin reductase, substantial amounts (25–50%) of AP accumulate with its formed structural disulfide bonds. We have previously suggested that this phenomenon was due to the alteration of the oxidation state of thioredoxin and of a second, hitherto unknown, thioredoxin (Derman et al., 1993a; Prinz et al., 1997). Thus, we initiated genetic studies on the newly discovered open reading frame (now termed trxC) that appeared to encode a thioredoxin homolog.

Thioredoxins 1 and 2 are required for cytoplasmic disulfide bond formation in a trxB strain

To study the role of the second thioredoxin, thioredoxin 2, we generated a null mutation by deleting the coding region of trxC between the first and last four codons. This strategy was chosen due to the lower probability of affecting the expression of downstream genes compared with generating an insertion or replacement construct. The deletion strain was constructed in the presence of a complementing trxC plasmid in case the deletion was lethal. Cells deleted for trxC proved viable, and mutants containing this deletion were then used to investigate the mechanism of cytoplasmic disulfide bond formation. We again used the enzymatic activity of ssAP as a reporter of disulfide bond formation in the cytoplasm. As observed previously, a wild-type strain exhibits low AP activity, and a trxB strain yields high activity (Figure 2). Unlike the trxB strain, where nearly all the AP activity is in the cytoplasm, the low level of activity in a wild-type strain is due to the small amount of AP that gets exported to the periplasm despite the lack of a signal sequence (Derman et al., 1993a,b).

The first hypothesis outlined in the Introduction assumes that the AP activity in the trxB mutant is a result of the failure to reduce AP due to the absence of the reduced forms of the two thioredoxins. To test this hypothesis, we constructed the double mutant trxA, trxC strain which would eliminate the thioredoxins as reductants from the cytoplasm. However, the deletion of both trxA and trxC simultaneously gave no increase in AP activity relative to the wild-type strain. The second hypothesis proposes that the thioredoxins, which accumulate in the oxidized form in a trxB mutant, are capable of actively promoting disulfide bond formation in ssAP. This hypothesis predicts that we should be able to eliminate the activation of AP in the trxB mutant by generating the triple mutant trxA, trxB, trxC. This prediction proved correct. Deletion of both thioredoxins in a trxB strain expressing ssAP decreases the AP activity from the high level seen in a trxB strain to about the level of the control TrxB⁺ strain. The low AP activity of the trxA, trxB, trxC strain clearly shows that the thioredoxins must be present for disulfide bond formation in a trxB strain, suggesting an important role for them in the oxidation of AP. Furthermore, a mutation in either of the thioredoxins alone, combined with trxB, causes the resultant strain to show a decreased level of activity relative to a trxB strain, with the decrease in activity being more pronounced with the trxA mutation than with the trxC mutation.

To ensure that the differences in activity of AP are due to a failure of the protein to become active rather than to a change in the level of expression of AP, we pulse-labeled and immunoprecipitated AP. Quantitation showed that the level of AP produced, normalized to the amount of OmpA, is comparable between the strains, indicating that the differences in activity are due to the level of activity of the protein, rather than the level of expression (Figure 3).

To verify that the change in phosphatase activity was due to the deletion of the trxC gene, assays were performed in a trxA, trxB, trxC strain in the presence of a plasmid containing trxC and also in the presence of the equivalent vector as a control (Figure 2). The activity in the strain with the trxC plasmid increases to the level of a trxA, trxB strain, while the strain with the vector alone is not altered in phosphatase activity. Similarly, when these plasmids are introduced into a trxB, trxC strain, the strain with the trxC plasmid showed increased activity, while the strain with the vector alone did not (data not shown). These results indicate that the changes in activity are due to the deletion of the trxC gene.

The role of thioredoxin 2 in the reducing pathways of E.coli

The role of thioredoxin 2 as an oxidant in the trxB background is unlikely to reflect the function of this protein in the ordinarily reducing environment of a wild-type cytoplasm. We wished to determine whether thioredoxin 2 contributes to the important reducing processes normally taking place in the cytoplasm. To do this, we have constructed a number of multiple mutants lacking various combinations of the important disulfide-reducing components and determined the viability of these mutants.

Cells require either the thioredoxin or glutathione/glutaredoxin pathway for normal growth. For example, trxB, gshA double mutants lacking thioredoxin reductase and glutathione (the product of the gshA gene is essential for the synthesis of glutathione) are essentially inviable, and only grow in the presence of DTT (Prinz et al., 1997). This inviability is presumably due to the complete block of the glutathione/glutaredoxin and thioredoxin pathways. However, a trxA, gshA double mutant is viable (Prinz et al., 1997). Since this latter double mutant is blocked for the glutathione/glutaredoxin pathway and lacks thioredoxin 1, it appears likely that its viability is due to the presence of thioredoxin 2. To determine whether this was the case, we constructed a triple mutant, trxA, trxC, gshA. This strain is inviable and could only be constructed in the presence of DTT (Table I). This finding demonstrates that eliminating both thioredoxins completely disables the thioredoxin reductase/thioredoxin pathway, whereas eliminating only thioredoxin 1 does not.

Glutaredoxin 1, the product of the grxA gene, is the most efficient disulfide-reductant in the glutathione/glutaredoxin pathway (Åslund et al., 1997). It seemed possible that this protein along with the two thioredoxins are the important contributors to the thiol redox environment of the cytoplasm. To test this possibility, we attempted to construct a triple mutant trxA, trxC, grxA strain. We were only able to construct this mutant when one of the missing disulfide-reducing proteins, such as thioredoxin 2, was expressed from a plasmid, indicating that a trxA, trxC, grxA strain is inviable (Table I). These results demonstrate that any one of the three proteins, thioredoxin 1, thioredoxin 2 or glutaredoxin 1, is capable of supplying the disulfide-reducing capacity to the cell necessary for viability. We anticipate that this activity is required for cell growth because of the need to reduce enzymes such as ribonucleotide reductase.

Surprisingly, the trxA, trxC, grxA strain is not able to grow in the presence of DTT, even though other strains that knock out the reducing capacity of both pathways such as a trxB, gor strain can. This result shows that DTT can substitute for thioredoxin reductase or glutathione oxidoreductase, but not for the thioredoxins or glutaredoxins.

We examined the role of the thioredoxins under a variety of conditions including growth on minimal media, growth under anaerobic conditions, and growth on methionine sulfoxide, to determine which processes they participate in. A trxA, grxA strain is unable to grow on minimal media, despite the presence of the chromosomal trxC gene (Russel et al., 1990). Expression of trxC from the arabinose promoter on a plasmid restores the ability of this strain to grow on minimal media, presumably due to a higher level of trxC expression (Table I). Under anaerobic conditions, E.coli uses an alternative ribonucleotide reductase that is independent of disulfide reduction (Mulliez et al., 1995). A trxB, gor strain is able to grow anaerobically, suggesting that the disulfide-reducing pathways are not essential under these conditions (Prinz et al., 1997). We find that a trxA, trxC, grxA strain is also capable of anaerobic growth (data not shown), confirming that the essential aerobic substrates are either replaced, not needed or reduced in another way in the absence of oxygen. One of the few substrates that were thought to be specifically acted on by thioredoxin 1 is the enzyme methionine sulfoxide reductase (MsrA) (Russel and Model, 1986). To determine how stringent the requirement is for thioredoxin 1, we overproduced several disulfide-reducing proteins in a trxA strain and tested for the ability to use methionine sulfoxide as the sole source of methionine. Increasing the expression of glutaredoxin 1 allowed growth, indicating that at higher levels of expression, it can reduce methionine sulfoxide reductase (Table I). Expression of thioredoxin 2 allowed only very poor growth.

Given the apparent non-specificity of the thiol-disulfide oxidoreductases, we wondered if the trxA, trxC, grxA triple mutant strain could be used to address the function of other disulfide-reducing proteins. Indeed, we found that nrdH (a normally cryptic E.coli gene encoding a potent disulfide-reductant; Jordan et al., 1997), when expressed from a plasmid, was able to support the growth of the triple mutant. However, a plasmid expressing grxC, in addition to the chromosomal expression of glutaredoxin 3, was not able rescue this strain, despite the fact that glutaredoxin 3 has a low but significant activity reducing ribonucleotide reductase in vitro (Åslund et al., 1994). This suggests either that the reducing potential of glutaredoxin 3 is not low enough to reduce this enzyme in vivo, or that some other essential enzyme, which glutaredoxin 3 cannot act on, is responsible for the inviability.

Discussion

We initiated this series of studies to determine why stable structural disulfide bonds were absent from proteins in the cytoplasm of cells (Derman et al., 1993a). We began with the presumption that the reducing potential of cytoplasmic thiol redox components was required to prevent disulfide bonds from forming. To pursue this question, we have assessed a variety of mutant cells for their ability to alter the cytoplasmic environment so that such bonds can form. The results presented here suggest that our starting presumption was incorrect. Thus, it is the presence of an active thiol oxidant—a promoter of disulfide bond formation—that causes such bonds to form. This conclusion is based on the properties of mutations lacking components of the thioredoxin reductase/thioredoxin pathway.

Mutants lacking the enzyme thioredoxin reductase accumulate high levels of the disulfide-bonded active form of AP when this normally periplasmic protein is retained in the cytoplasm. Here we show that this effect is not due to the elimination of the reduced forms of thioredoxin 1 and 2, as a trxA, trxC double mutant, lacking both these proteins, exhibits little or no AP activity. Rather, we propose that the disulfide bond formation that occurs in the trxB mutant is due to the action of the oxidized forms of these thioredoxins. That is, the oxidized products of the trxA and trxC genes are catalyzing the formation of disulfide bonds in substrates such as AP.

It may appear surprising that thioredoxins, which are potent reductants, could also act as oxidants under certain conditions. However, such an activity of these proteins is not unexpected. While the redox potential of the disulfide bond in thioredoxin 1 is the lowest of the known thiol-disulfide oxidoreductases (-270 mV at pH 7.0), it is still considerably higher than that of the structural disulfide bonds in folded proteins such as bovine pancreatic trypsin inhibitor (BPTI; Creighton and Goldberg, 1984). This difference in redox potential makes the net transfer of the disulfide bond from thioredoxin to proteins such as alkaline phosphatase a favorable reaction. Such a process has been demonstrated in vitro; oxidized thioredoxin can transfer disulfide bonds to reduced ribonuclease (Pigiet and Schuster, 1986; Lundström et al., 1992). Thus, our explanation for the accumulation of the disulfide-bonded form of AP in the trxB mutant is consistent with the in vitro properties of thioredoxins.

These findings raise a number of interesting points: first, they further support the conclusion that a catalytic system is required for disulfide bond formation in proteins to occur with significant efficiency. For example, when the DsbA or DsbB proteins are absent from the periplasm, the oxidation of cysteines in exported proteins proceeds quite slowly (Bardwell et al., 1991). One of the potential roles considered for the two major disulfide-reducing pathways in the cytoplasm is that they reduce any incidental disulfide bonds that form in cytoplasmic proteins. However, as we have shown that in ssAP these bonds do not form in the cytoplasm of either a wild-type strain or a trxA, trxB, trxC strain, we conclude that disulfide bond formation in the cytoplasm does not occur very frequently unless it is catalyzed. This could mean that for most cytoplasmic proteins there is little need for the thioredoxin and glutathione/glutaredoxin pathways in the maintenance of reduced thiols under laboratory growth conditions. Secondly, folded and oxidized AP is not able to be reduced by the cytoplasmic reducing pathways, as disulfide bonds accumulate in AP in the cytoplasm of the trxB strain, despite the presence of the glutathione/glutaredoxin pathway. Similarly, in certain double mutants of the glutathione/glutaredoxin pathway (e.g. gor, grxA, assayed on NZ medium), high levels of cytoplasmic AP activity can be achieved, even though the thioredoxin pathway is intact (F.Åslund, unpublished data). Therefore, the disulfide-reductants do not appear to significantly affect the formation of structural disulfide bonds in cytoplasmic AP.

Our results also provide another example where the role of a thiol-disulfide oxidoreductase is altered by a change in the environment in which it is expressed. The protein disulfide bond isomerase, DsbC, acts as an oxidant in the E.coli periplasm, when the protein required for DsbC reduction, DsbD, is eliminated (Missiakas et al., 1995; Rietsch et al., 1996). More recently, we have shown that thioredoxin 1 can also act as an oxidant when it is exported to the oxidizing environment of the E.coli periplasm (Debarbieux and Beckwith, 1998). In this paper, we present evidence suggesting that both thioredoxins 1 and 2, whose normal role is to act as reductants, can act as oxidants in a trxB mutant.

In fact, it may be that under certain growth conditions or in some of the environments wild-type E.coli encounter, accumulation of oxidized thioredoxins does take place. Certain stress conditions, such as exposure to hydrogen peroxide, generate an increased oxidizing environment within the cytoplasm that results in the oxidation of thioredoxin 1 (F.Åslund, unpublished data). This accumulation of oxidized thioredoxins could be dangerous to the cell, as we have shown that they can induce disulfide bond formation in other proteins. Therefore, the cell responds to this challenge by activation of the regulatory protein OxyR, which increases the expression of some components of the reducing machinery (Zheng et al., 1998). This probably helps to reduce any aberrant disulfide bonds that may occur in cytoplasmic proteins. Another potential target of such oxidative stress could be the precursor forms of secreted proteins, before they are translocated across the membrane. However, we consider oxidation of these precursors to be unlikely, as the rate of secretion is much higher than that of thioredoxin-catalyzed disulfide bond formation. For example, secretion of a protein such as AP (with its signal sequence) is complete in <30 s (Michaelis et al., 1986), while cytoplasmic disulfide bond formation, even in a trxB mutant, takes as long as 10 min (Derman et al., 1993a).

In other organisms, there may be another mechanism at work to prevent disulfide bond formation by oxidized thioredoxins. Mammalian thioredoxins and glutaredoxins contain conserved cysteine residues other than those in the active site. These non-active site cysteines form disulfide bonds that inactivate thioredoxin upon oxidation (Ren et al., 1993). We propose that these bonds may play an autoregulatory role to prevent the disulfide-reducing proteins from promoting disulfide bond formation.

With the availability of a trxC null mutant, we have also been able to address a number of other questions concerning the disulfide-reducing components of the E.coli cytoplasm. We have obtained evidence for an in vivo role for the newly discovered thioredoxin 2, the product of the trxC gene. By examining the phenotype of trxC mutations in combination with other mutations in the thioredoxin and glutathione/glutaredoxin systems, we show that the loss of both thioredoxins 1 and 2 completely blocks the thioredoxin pathway. Since a trxA, trxC, grxA strain is not viable unless complemented by one of the three genes, we conclude that these three are the only (sufficiently expressed) proteins able to perform the disulfide-reductive functions essential for aerobic growth of E.coli. Furthermore, any one of these proteins expressed by itself is capable of supporting growth.

While the cytoplasmic disulfide-reducing proteins appear to overlap in their functions, there are some indications of specificity of function. In particular, genetic studies have suggested that, in vivo, thioredoxin 1 is essential for the reduction of methionine sulfoxide reductase. However, we have found that when glutaredoxin 1 is overproduced in a trxA mutant, it can substitute for thioredoxin 1 in the reduction of methionine sulfoxide reductase (Russel and Model, 1986). In another example, thioredoxin 2 is able to complement a trxA, grxA mutant for growth on minimal medium when the trxC gene is overexpressed from a plasmid, but not when expressed from the chromosome. These findings suggest that the specific requirements in these reactions may not be dependent only on a specific binding interaction with the substrate proteins. Rather, any observed specificity of disulfide-reducing proteins deduced from the phenotype of mutant strains may also be dependent on the relative concentration of that protein and its redox potential.

Given this lack of stringent specificity, it seems that E.coli could survive with a single reducing system. Organisms of the genus Mycoplasma appear to possess a single thioredoxin system, and no glutathione/glutaredoxin system (Fraser et al., 1995; Himmelreich et al., 1996). This raises the intriguing question of why E.coli maintains this plethora of disulfide-reductants. We point out that in the phage T4 genome, where it might be assumed that genetic space is at a premium, there are two glutaredoxins encoded (Gvakharia et al., 1996). This, despite the fact that when T4 infects E.coli, it is infecting a cell already providing three glutaredoxins and two thioredoxins. Escherichia coli is not alone in this regard, as a large number of organisms have more than one thioredoxin, and some substantially more (Wetterauer et al., 1992; Rivera-Madrid et al., 1995; Lim et al., 1996). Part of the answer may be that by co-evolving expression levels with substrate specificity, the cell can respond to different stress situations that are present in various environments. For example, hydrogen peroxide stimulates an OxyR response that results in the specific induction of grxA transcription. It could be that each of the thioredoxins and glutaredoxins has a particular role to play under conditions of stress, and that different sources of stress may require different responses. Organisms with less disulfide-reducing variety may not encounter the same sources of oxidative stress, or they may deal with these stresses in novel ways.

Materials and methods

Bacterial strains and growth conditions

Strains and plasmids used are listed in Table II. Cells were generally grown in NZ medium, described previously (Derman et al., 1993b), at 37°C. For phosphatase assays and pulse-labelings the cells were subcultured from overnight NZ cultures into minimal M63 medium [M63 salts with 0.2% glucose, 1 g/ml vitamin B1, 1 mM MgSO₄, 50 g/ml 18 amino acids (excluding methionine and cysteine) and supplemented with 2 g/ml nicotinic acid] at a 100-fold dilution, and then incubated at 37°C. Antibiotic selection was maintained for all markers on plasmids, at the following concentrations: ampicillin, 200 g/ml; spectinomycin, 100 g/ml; and chloramphenicol, 10 g/ml. Induction of alkaline phosphatase was accomplished by addition of IPTG to a final concentration of 5 mM at the time of subculturing. For testing growth on a minimal medium, M63 glucose (0.2%) plates were supplemented with leucine and isoleucine (50 g/ml each) and incubated for 2 days at 37°C. The ability to utilize methionine sulfoxide as the sole source of methionine was tested on M63 plates as above, with methionine sulfoxide added to 100 g/ml. The strain used (generously provided by M.Russel) is A313, a trxA metE mutant which is unable to synthesize methionine de novo. Nicotinic acid (2 g/ml) was added to the media when strains with an insertion in nadB were grown on M63. Constructs in pBAD plasmids were induced by addition of 0.1% L-arabinose.

Cloning of the disulfide-reducing proteins

A PCR fragment encoding the trxC gene was cloned into pBAD33. This resulted in the expression of trxC under the control of the araBAD operon promoter. A DNA fragment containing trxC and enough upstream DNA (214 bp) to include its putative promoter was also generated by PCR and cloned into the vector pAM238. This PCR fragment was ligated into the vector in the opposite orientation to the lac promoter on the plasmid, resulting in the expression of the trxC gene under its own promoter. All other plasmids created for this study were cloned using PCR. Plasmids pFÅ1, pEJS33, pEJS62 and pEJS80 use the Shine–Dalgarno ribosome binding site native to the gene cloned, while pFÅ12, pFÅ13, and pWP619 employed the optimized Shine–Dalgarno contained in the vector.

Deletion of trxC

As the sequence downstream of trxC indicates that it might be the first gene in an operon, we created a null mutation that is unlikely to affect expression of these downstream genes. DNA flanking the gene was amplified by PCR, and fragments consisting of 1.8 kb of upstream DNA and 1.4 kb of downstream DNA were generated using the primers (all are written 5' to 3', and were supplied by Genosys): upstream, F1-left-B, GGCCAGGATCCTTATCACGGACC; F1-right-P, ACAATGCTGCAGAACGGTATTCATAACTAACCT; downstream, F2-left-P, GATAGCCTGCAGAACGAATCTCTTTAATCTTAC; F2-right-S(Xho), GCG- CTCTCGAGACTGTCCCGGGCCAGATAGTCAAG. These fragments included the first four amino acids (upstream) and the last four amino acids (downstream) of trxC, but no other part of the coding sequence. They were cloned into pKO3, and following the published protocol (Link et al., 1997), this vector was used to sequentially select and then screen for replacement of the wild-type trxC allele with the deletion allele. Presence of the deletion allele was verified by a PCR screen across the chromosomal trxC gene. The final result removes the central 131 codons of the trxC gene, and replaces them with the six bases of a PstI site.

Transduction of trxC with a linked transposon

A P1 lysate of strain CAG18480, containing a Tn10 insertion in the nadB gene, was used to generate a marker linked to trxC. The tetracycline resistance marker co-transduces trxC with a frequency of 30%. The original deletion of trxC was moved by P1 transduction, using this marker to create the strains used in this study. The strain used to generate this lysate was PCR sequenced across the trxC region to verify the deletion.

Construction of multiple-mutant strains

Construction of the potentially lethal strains completely blocked for the disulfide-reducing pathways was accomplished as follows: a trxA, gshA strain was transduced with the lysate from the trxC strain described above, selecting for tetracycline resistance, on a plate containing a disk saturated with DTT. DTT-dependent colonies were then screened by PCR for the presence of the trxC deletion: a trxA, grxA mutant carrying a plasmid expressing trxC or another disulfide-reducing protein was also transduced with the trxC lysate, and selected for tetracycline resistance on plates containing arabinose, as described above. The resultant colonies were then screened for the trxC deletion.

Alkaline phosphatase assays

Cells were grown and induced as described above. At an optical density at 600 nm (OD₆₀₀) of 0.4–0.6, an aliquot of 1 ml was immediately added to 100 l of room-temperature 1 M iodoacetamide. The sample was then incubated on ice for at least 20 min. AP activity was then determined as described previously (Derman et al., 1993a). All cultures were grown, subcultured, and assayed in duplicate.

Pulse-labelings and immunoprecipitations

Cells were grown and induced as described above. At an OD₆₀₀ of 0.25–0.30, a 1 ml aliquot of culture was transferred to a Falcon 2059 culture tube in a shaking water bath at 37°C. After a 30 min incubation, radiolabel ([³⁵S]methionine, ICN) was added to a concentration of 40 Ci/ml. After a 1 min incubation, 100 l of unlabeled 1% methionine was added and 700 l of the culture was immediately removed to a tube containing 300 l of ice-cold, unlabeled 0.5% methionine.

These cells were then lysed and immunoprecipitations performed using antibody to AP (1 l/ml of culture; antibody purchased from Rockland) and antibody to OmpA protein (1 l/ml of culture; antibody from laboratory-generated stock). Immunoprecipitations were performed as described previously (Pogliano and Beckwith, 1993). A 12% SDS–polyacrylamide gel (Mini-ProteanII, Bio-Rad) was then loaded with 20% of each of the total sample volumes, and electrophoresed at a constant voltage of 120 V until the dye front reached, but did not leave, the bottom of the gel. The gel was dried under vacuum at 80°C for 1.5 h, then exposed to film (Kodak X-OMAT) for 14 h. The gel was then immediately used to expose a Molecular Imager cassette for 3 h. The cassette was scanned and quantified at a resolution of 100 m in a Bio-Rad GS-525 Molecular Imager System. The number of counts from the band corresponding to phosphatase was normalized to the number of counts from OmpA in each lane, to provide a control for the amount of protein loaded. All cultures were grown, labeled and quantitated in duplicate.

Acknowledgements

We would like to thank Dr W.Prinz for his help and insight into this work, and along with Dr S.Timmermann for their critical reading of the manuscript. We would also like to thank Dr L.Debarbieux for invaluable aid. In addition, we thank the other members of the laboratory for their help and support. This work was generously supported by the National Institutes of Health (NIH GM41883). J.B. is also supported by an American Cancer Society Research Professorship, and F.Å. is supported by a long-term fellowship from the European Molecular Biology Organization (EMBO).

References

Åberg A, Hahne S, Karlsson M, Larsson A, Ormö M, Åhgren A and Sjöberg BM (1989) Evidence for two different classes of redox-active cysteines in ribonucleotide reductase of Escherichia coli. J Biol Chem, 264, 12249–12252. | PubMed | ISI | ChemPort |

Åslund F, Ehn B, Miranda-Vizuete A, Pueyo C and Holmgren A (1994) Two additional glutaredoxins exist in Escherichia coli: Glutaredoxin 3 is a hydrogen donor for ribonucleotide reductase in a thioredoxin/glutaredoxin 1 double mutant. Proc Natl Acad Sci USA, 91, 9813–9817. | PubMed | ISI |

Åslund F, Berndt KD and Holmgren A (1997) Redox potentials of glutaredoxins and other thiol-disulfide oxidoreductases of the thioredoxin superfamily determined by direct protein–protein redox equilibria. J Biol Chem, 272, 30780–30786. | Article | PubMed | ISI | ChemPort |

Bardwell JCA, McGovern K and Beckwith J (1991) Identification of a protein required for disulfide bond formation in vivo. Cell, 67, 581–589. | Article | PubMed | ISI | ChemPort |

Boyd D, Manoil C and Beckwith J (1987) Determinants of membrane protein topology. Proc Natl Acad Sci USA, 84, 8525–8529. | Article | PubMed | ChemPort |

Creighton TE and Goldenberg DP (1984) Kinetic role of a meta-stable native-like two-disulphide species in the folding transition of bovine pancreatic trypsin inhibitor. J Mol Biol, 179, 497–526. | Article | PubMed | ISI | ChemPort |

Debarbieux L and Beckwith J (1998) The reductive enzyme thioredoxin 1 acts as an oxidant when it is exported to the Escherichia coli periplasm. Proc Natl Acad Sci USA, 95, 10751–10756. | Article | PubMed | ChemPort |

Derman AI and Beckwith J (1991) Escherichia coli alkaline phosphatase fails to acquire disulfide bonds when retained in the cytoplasm. J Bacteriol, 173, 7719–7722. | PubMed | ISI | ChemPort |

Derman AI, Prinz WA, Belin D and Beckwith J (1993a) Mutations that allow disulfide bond formation in the cytoplasm of Escherichia coli. Science, 262, 1744–1747. | Article | PubMed | ISI | ChemPort |

Derman AI, Puziss JW, Bassford PJ,Jr and Beckwith J (1993b) A signal sequence is not required for protein export in prlA mutants of Escherichia coli. EMBO J, 12, 879–888. | PubMed | ISI | ChemPort |

Fraser CM et al. (1995) The minimal gene complement of Mycoplasma genitalium. Science, 270, 397–403. | Article | PubMed | ISI | ChemPort |

Gil D and Bouché JP (1991) ColE1-type vectors with fully repressible replication. Gene, 105, 17–22. | Article | PubMed | ISI | ChemPort |

Guzman L-M, Belin D, Carson MJ and Beckwith J (1995) Tight regulation, modulation and high-level expression by vectors containing the arabinose p_BAD promoter. J Bacteriol, 177, 4121–4130. | PubMed | ISI | ChemPort |

Gvakharia BO, Hanson E, Koonin EK and Mathewes CK (1996) Identification of a second functional glutaredoxin encoded by the bacteriophage T4 genome. J Biol Chem, 271, 15307–15310. | Article | PubMed | ISI | ChemPort |

Himmelreich R, Hilbert H, Plagens H, Pirkl E, Li BC and Herrmann R (1996) Complete sequence analysis of the genome of the bacterium Mycoplasma pneumoniae. Nucleic Acids Res, 24, 4420–4449. | Article | PubMed | ISI | ChemPort |

Holmgren A (1989) Thioredoxin and glutaredoxin systems. J Biol Chem, 264, 13963–13966. | PubMed | ISI | ChemPort |

Jordan A, Åslund F, Pontis E, Reichard P and Holmgren A (1997) Characterization of Escherichia coli NrdH: A glutaredoxin-like protein with a thioredoxin-like activity profile. J Biol Chem, 272, 18044–18050. | Article | PubMed | ISI | ChemPort |

Lim CJ, Sa JH and Fuchs JA (1996) Identification of a third thioredoxin gene from Corynebacterium nephridii. Biochim Biophys Acta, 1307, 13–16. | PubMed | ISI | ChemPort |

Link AJ, Phillips D and Church GM (1997) Methods for generating precise deletions and insertions in the genome of wild-type Escherichia coli: application to open reading frame characterization. J Bacteriol, 179, 6228–6237. | PubMed | ISI | ChemPort |

Liu J and Rosen BP (1997) Ligand interactions of the ArsC arsenate reductase. J Biol Chem, 272, 21084–21089. | Article | PubMed | ISI | ChemPort |

Lundström J, Krause G and Holmgren A (1992) A Pro to His mutation in active site of thioredoxin increases its disulfide-isomerase activity 10-fold. New refolding systems for reduced or randomly oxidized ribonuclease. J Biol Chem, 267, 9047–9052. | PubMed | ISI | ChemPort |

Michaelis S, Hunt J and Beckwith J (1986) Effects of signal sequence mutations on the kinetics of alkaline phosphatase export to the periplasm in Escherichia coli. J Bacteriol, 167, 160–167. | PubMed | ISI | ChemPort |

Miranda-Vizuete A, Damdimopoulos AE, Gustafsson J-A and Spyrou G (1997) Cloning, expression and characterization of a novel Escherichia coli thioredoxin. J Biol Chem, 272, 30841–30847. | Article | PubMed | ISI | ChemPort |

Missiakas D, Schwager F and Raina S (1995) Identification and characterization of a new disulfide isomerase-like protein (DsbD) in Escherichia coli. EMBO J, 14, 3415–3424. | PubMed | ISI | ChemPort |

Mulliez E, Ollagnier S, Fontecave M, Eliasson R and Reichard P (1995) Formate is the hydrogen donor for the anaerobic ribonucleotide reductase from Escherichia coli. Proc Natl Acad Sci USA, 92, 8759–8762. | Article | PubMed | ChemPort |

Pigiet V and Schuster BJ (1986) Thioredoxin-catalyzed refolding of disulfide-containing proteins. Proc Natl Acad Sci USA, 83, 7643–7647. | Article | PubMed | ChemPort |

Pogliano KJ and Beckwith J (1993) The Cs sec mutants of Escherichia coli reflect the cold sensitivity of protein export itself. Genetics, 133, 763–773. | PubMed | ISI | ChemPort |

Pollitt S and Zalkin H (1983) Role of primary structure and disulfide bond formation in -lactamase secretion. J Bacteriol, 153, 27–32. | PubMed | ISI | ChemPort |

Prinz WA, Åslund F, Holmgren A and Beckwith J (1997) The role of the thioredoxin and glutaredoxin pathways in reducing protein disulfide bonds in the Escherichia coli cytoplasm. J Biol Chem, 272, 15661–15667. | Article | PubMed | ISI | ChemPort |

Ren X, Björnstedt M, Shen B, Ericson ML and Holmgren A (1993) Mutagenesis of structural half-cystine residues in human thioredoxin and effects on the regulation of activity by selenodiglutathione. Biochemistry, 32, 9701–9708. | Article | PubMed | ISI | ChemPort |

Rietsch A and Beckwith J (1998) The genetics of disulfide bond metabolism. Annu Rev Genet, 32, 163–184. | Article | PubMed | ISI | ChemPort |

Rietsch A, Belin D, Martin N and Beckwith J (1996) An in vivo pathway for disulfide bond isomerization in Escherichia coli. Proc Natl Acad Sci USA, 93, 13048–13053. | Article | PubMed | ChemPort |

Rivera-Madrid R, Mestres D, Marinho P, Jacquot J-P, Decottignies P, Miginiac-Maslow M and Meyer Y (1995) Evidence for five divergent thioredoxin h sequences in Arabidopsis thaliana. Proc Natl Acad Sci USA, 92, 5620–5624. | Article | PubMed | ChemPort |

Russel M and Model P (1986) The role of thioredoxin in filamentous phage assembly. Construction, isolation and characterization of mutant thioredoxins. J Biol Chem, 261, 14997–15005. | PubMed | ChemPort |

Russel M, Model P and Holmgren A (1990) Thioredoxin or glutaredoxin in Escherichia coli is essential for sulfate reduction but not for deoxyribonucleotide synthesis. J Bacteriol, 172, 1923–1929. | PubMed | ISI | ChemPort |

Singer M et al. (1989) A collection of strains containing genetically linked alternating antibiotic resistance elements for genetic mapping of Escherichia coli. Microbiol Rev, 53, 1–24. | PubMed | ISI | ChemPort |

Sone M, Kishigami S, Yoshihisa T and Ito K (1997) Roles of disulfide bonds in bacterial alkaline phosphatase. J Biol Chem, 272, 6174–6178. | Article | PubMed | ISI | ChemPort |

Wetterauer B, Jacquot JP and Veron M (1992) Thioredoxins from Dictyostelium discoideum are a developmentally regulated multigene family. J Biol Chem, 267, 9895–9904. | PubMed | ISI | ChemPort |

Zheng M, Åslund F and Storz G (1998) Activation of the OxyR transcription factor by reversible disulfide bond formation. Science, 279, 1718–1721. | Article | PubMed | ISI | ChemPort |