Importance of Redox Potential for the in VivoFunction of the Cytoplasmic Disulfide Reductant Thioredoxin fromEscherichia coli *

The thioredoxin superfamily consists of enzymes that catalyze the reduction, formation, and isomerization of disulfide bonds and exert their activity through a redox active disulfide in a Cys-Xaa1-Xaa2-Cys motif. The individual members of the family differ strongly in their intrinsic redox potentials. However, the role of the different redox potentials for thein vivo function of these enzymes is essentially unknown. To address the question of in vivo importance of redox potential for the most reducing member of the enzyme family, thioredoxin, we have employed a set of active site variants of thioredoxin with increased redox potentials (−270 to −195 mV) for functional studies in the cytoplasm of Escherichia coli. The variants proved to be efficient substrates of thioredoxin reductase, providing a basis for an in vivocharacterization of NADPH-dependent reductive processes catalyzed by the thioredoxin variants. The reduction of sulfate and methionine sulfoxide, as well as the isomerization of periplasmic disulfide bonds by DsbC, which all depend on thioredoxin as catalyst in the E. coli cytoplasm, proved to correlate well with the intrinsic redox potentials of the variants in complementation assays. The same correlation could be established in vitro by using the thioredoxin-catalyzed reduction of lipoic acid by NADPH as a model reaction. We propose that the rate of direct reduction of substrates by thioredoxin, which largely depends on the redox potential of thioredoxin, is the most important parameter for the in vivo function of thioredoxin, as recycling of reduced thioredoxin through NADPH and thioredoxin reductase is not rate-limiting for its catalytic cycle.

Thiol-disulfide oxidoreductases from Escherichia coli that belong to the thioredoxin superfamily can have either reducing properties, like thioredoxin with a redox potential of Ϫ270 mV (1,2), or oxidizing properties, like DsbA with a redox potential of Ϫ122 mV (3,4). In general, the bacterial cytosolic members of the family such as thioredoxin and glutaredoxin are efficient reductants of disulfide bonds and are believed to be important for the maintenance of the reducing redox status of the cytosol (5). In addition, thioredoxin catalyzes the NADPH-dependent reduction of a number of important metabolic enzymes such as ribonucleotide reductase, 3Ј-phosphoadenosine-5Ј-phosphosulfate reductase, and methionine-sulfoxide reductase, whose active sites become oxidized as part of their catalytic cycle (see Holmgren and Björnstedt (6)). In contrast, members with oxidizing redox properties like DsbA and DsbC are found in the bacterial periplasm. These enzymes catalyze the formation and isomerization of disulfide bonds during folding of newly translocated proteins (7,8). An exception to this correlation between cellular localization and the redox potential appears to be the periplasmically oriented thioredoxin-like proteins that are anchored to the inner bacterial membrane. An example is TlpA from Bradyrhizobium japonicum, which has a low redox potential (Ϫ259 mV) despite its periplasmic orientation and is required for cytochrome aa 3 maturation (9, 10).
Thiol-disulfide oxidoreductases with thioredoxin fold share a catalytic disulfide bond with the sequence Cys-Xaa 1 -Xaa 2 -Cys. An influence of the active site dipeptide Xaa 1 -Xaa 2 on the redox properties of these enzymes has been demonstrated, which results from a modulation of the pK a value of the more Nterminal, nucleophilic active site cysteine (11)(12)(13)(14). In DsbA, the extremely low pK a value (ϳ3.5) of the nucleophilic cysteine stabilizes the reduced state of the active site making the protein an excellent disulfide bond donor (15). Replacement of the Xaa 1 -Xaa 2 sequence of DsbA by that of thioredoxin decreased the redox potential by 90 mV making DsbA a 1000-fold better reductant in terms of equilibrium constants with substrate compounds (13). Conversely, a DsbA-like Xaa 1 -Xaa 2 variant of thioredoxin became a 200-fold better oxidant (14). Despite these dramatic changes in redox potentials, the corresponding mutant disulfide oxidoreductases did not show phenotypic effects in bacteria under physiological conditions (10,11,16,17). Furthermore, the most reducing member of the thioredoxin superfamily, thioredoxin, and variants thereof can, at least partially, substitute for the most oxidizing member DsbA when exported to the periplasm (18,19). This raises the question whether redox potential is at all an important property of these proteins under physiological growth conditions. A certain correlation between the redox potentials of DsbA variants and their ability to complement a DsbA null phenotype could only be demonstrated for the recovery of resistance toward the reductant dithiothreitol (DTT) 1 in the growth medium (11).
Thioredoxin and glutaredoxin-1 from E. coli serve as cyto-plasmic reductants of 3Ј-phosphoadenosine-5Ј-phosphosulfate reductase, an enzyme in the pathway of sulfate reduction. Absence of both trxA and grxA is accompanied by the inability to grow on sulfate minimal medium unless a source of reduced sulfur such as cysteine is provided (20). Thioredoxin is also the preferred reductant of methionine-sulfoxide reductase as shown by the inability of trxA mutants to use methionine sulfoxide as the sole source of methionine (21). Moreover, a pathway for periplasmic disulfide bond isomerization has recently been described (22) that funnels reducing equivalents from thioredoxin in the cytosol to the periplasmic disulfide isomerase DsbC via the membrane protein DsbD (23). To address the question whether the intrinsic redox potential of thiol-disulfide oxidoreductases is of any major importance in vivo, we have used these in vivo functions of thioredoxin to test a set of thioredoxin variants with increased redox potentials (ranging from Ϫ270 to Ϫ195 mV, cf. Table I) for complementing activity in E. coli thioredoxin null mutants. The selected mutant proteins are the three most oxidizing thioredoxin variants that have been reported so far and bear the Xaa 1 -Xaa 2 dipeptide sequences of the following members of the thioredoxin family: glutaredoxin (active site: Cys-Pro-Tyr-Cys), DsbA (active site: Cys-Pro-His-Cys), and eukaryotic protein-disulfide isomerase (PDI; active site: Cys-Gly-His-Cys). The basis for the in vivo studies presented in this paper is the finding that these thioredoxin variants are all substrates of thioredoxin reductase (TR) and show a clear in vitro correlation between their redox potentials and efficiency of catalyzing the reduction of lipoic acid by NADPH as a model reaction.
Purification of Thioredoxin-Thioredoxin wild type and the thioredoxin variants were purified from overproducing E. coli strains as described previously (14). The protein concentrations were measured by the specific absorbance at 280 nm (14).
Construction of Thioredoxin Expression Plasmids with Arabinose Promoter-Molecular cloning techniques were based on Sambrook et al. (24). The genes encoding the thioredoxin variants in the corresponding T7 expression plasmids (14) were amplified by the polymerase chain reaction using the following primers: trxA-SacI, 5Ј-AGG CGA GCT CCT GTG GAG TTA TAT ATG AGC GAT AAA ATT ATT CAC-3Ј; and trxA-HindIII, 5Ј-AAT AAG CTT ACG CCA GGT TAG CGT CG-3Ј. The polymerase chain reaction products were digested with SacI and Hind-III and cloned into the vector pBAD33 (25) cut with the same enzymes. The sequence of all thioredoxin genes was confirmed by dideoxy sequencing of the resulting plasmids.
Overexpression and Purification of Thioredoxin Reductase-The gene coding for TR, trxB (26), was amplified from the genome of the E. coli K12 wild-type strain W3110 (27) by polymerase chain reaction using the following oligonucleotide primers: N-terminal primer, 5Ј-CCA TTG TCT GTC TAG AAC TAT GGG GAT CTC ATG GGC ACG-3Ј; and C-terminal primer: 5Ј-CTG ATT TGT AAG CTT ATT TTG CGT CAG CTA AAC CAT C-3Ј. The amplified gene was cloned into a derivative of the expression vector pTrx (14) via the restriction sites XbaI and Hin-dIII. In the resulting expression plasmid pTR, the trxB gene is under control of the T7 promoter/lac operator (28). Successful cloning was verified by sequencing of the whole trxB gene. For overproduction of TR, E. coli BL21(DE3) (28) harboring pTR was grown at 25°C in 10 liters of LB medium containing ampicillin (100 g/ml). With an optical density at 550 nm of 1.0, isopropyl-␤-D-thiogalactopyranoside was added to a final concentration of 1 mM, and the cells were grown further for 18 h. Bacteria were harvested by centrifugation, suspended in 10 mM sodium phosphate, pH 7.0, and disrupted twice in a French Pressure cell (18.000 p.s.i.). TR was enriched by anion exchange chromatography on DE52-cellulose, fractional ammonium sulfate precipitation, and affinity chromatography on 2Ј,5Ј-ADP-agarose essentially as described (29). After gel filtration on Superdex 200 HR in 50 mM sodium phosphate, pH 7.0, and cation exchange chromatography on CM52-cellulose (pH 4.0, NaCl gradient), TR was more than 99% pure as judged by Coomassie Blue-stained SDS-polyacrylamide gels. 60 mg of purified TR were obtained from this procedure. A specific absorbance of A 280 nm, 1 mg/ml, 1 cm ϭ 1.39 was used for determination of the TR concentration (30).
Catalytic Parameters of TR for the Thioredoxin Variants-The assays were performed at 25°C in 100 mM Tris/HCl, pH 8.0, containing 0.2 mg/ml bovine serum albumin, 0.2 mM EDTA, 200 M NADPH, 500 M dithionitrobenzoic acid, 33 nM TR, and varying concentrations of thioredoxin (wild type or variant) or DsbA (0.04 -40 M). All reactants (without TR) were mixed and equilibrated for 5 min at 25°C before the reactions were started by the addition of TR. The reactions were followed by the formation of 2-nitro-5-thiobenzoate (⑀ 412 nm ϭ 13600 M Ϫ1 cm Ϫ1 ) (31).
Thioredoxin-catalyzed Reduction of Insulin by DTT-The test was performed in 100 mM potassium phosphate, pH 7.0, 2 mM EDTA (purged with nitrogen) containing insulin, DTT, and thioredoxin (wild type or variant) at concentrations of 130 M, 1.0 mM, and 1.0 M, respectively. After incubation of thioredoxin with DTT for 10 min at 25°C, the reaction was started by the addition of insulin. Precipitation of reduced insulin was monitored by the increase in optical density at 650 nm.
Thioredoxin-catalyzed Reduction of Insulin by NADPH-The reaction was performed at 25°C in 100 mM potassium phosphate, pH 7.0, 2 mM EDTA containing 500 M NADPH, 80 M insulin, 33 nM TR, and 20 M thioredoxin wild type (or the corresponding thioredoxin variants). The reaction was started by the addition of TR and followed by the increase in optical density at 650 nm. The same assay was also performed at pH 8.0 in 100 mM Tris/HCl, 0.2 mM EDTA using the same initial concentrations of reactants. In addition to the increase in optical density at 650 nm, the reaction at pH 8.0 was also followed by the consumption of NADPH (⑀ 340 nm ϭ 6200 M Ϫ1 cm Ϫ1 , 0.1-cm quartz cuvettes) and by HPLC analysis of the insulin reduction. For the latter analysis aliquots were removed from the reaction at different time points, the reactions were quenched with formic acid (20% (w/v) final concentration), and the insulin reaction products were separated by reversed-phase HPLC at 55°C on a Vydac 218TP54 column using a linear gradient from 30 to 60% acetonitrile in 0.1% trifluoroacetic acid. The decrease in the peak area of native insulin served to quantify the initial velocity of the reactions. Identical reactivities of the three disulfide bonds in insulin were assumed for evaluating the initial decrease in native insulin.
Reduction of Insulin, Lipoic Acid, and Oxidized DTT by Thioredoxin-Reduced thioredoxin was prepared by incubation of the oxidized protein with a 1000-fold molar excess of DTT (50 mM) at pH 7.0 for 1 h and subsequent removal of DTT on a PD10 gel filtration column (Amersham Pharmacia Biotech). Assays with thioredoxin as stoichiometric reductant were carried out at 25°C in 100 mM Tris/HCl, pH 8.0, 0.2 mM EDTA. The concentration of reduced thioredoxin was 1.0 M in all assays. Lipoic acid and oxidized DTT were used at concentrations between 0.1 and 10 mM guaranteeing pseudo first-order conditions. The stoichiometric reduction of insulin was performed in the same buffer as described above but at 15°C with identical initial concentrations (1.0 M) of reduced thioredoxin and insulin. All reactions were followed by the decrease in the specific tryptophan fluorescence of thioredoxin at 345 nm (excitation at 295 nm) on a Hitachi F-4500 fluorescence spectrometer. The fluorescence data were evaluated according to pseudo first-order kinetics (or quasi second-order kinetics in the case of insulin reduction).
Stoichiometric Reduction of Thioredoxin by DTT-Thioredoxin (6.0 M) was reduced at 25°C by equimolar amounts of DTT in 100 mM sodium phosphate, pH 7.0, 1 mM EDTA. The reaction was monitored by tryptophan fluorescence as described above and evaluated according to second-order kinetics.
Thioredoxin-catalyzed Reduction of Lipoic Acid-The assay was performed at 25°C in 100 mM Tris/HCl, pH 8.0, 0.2 mM EDTA containing 500 M NADPH, 33 nM TR, 20 M thioredoxin, and 1 mM lipoic acid. The reaction was started by the addition of thioredoxin reductase, and NADPH consumption was detected by the decrease in absorbance at 340 nm. Quartz cuvettes with a 0.1-cm path length were used.
In Vivo Complementation Assays-The corresponding trxA null strains were transformed with the pBAD33 derivatives containing the genes of thioredoxin wild type and the variants. All growth assays were performed at 37°C in the absence and in the presence of 0.2% (w/v) L-arabinose. Growth on minimal medium was assayed using E. coli FÅ47 (⌬trxA, grxA::kan, ara714, leu::tn10) (32) on M63-plates (33) containing 0.2% (w/v) glucose and supplemented with leucine and isoleucine (50 g/ml each). Methionine sulfoxide reduction assays were performed with the strain A313 (trxA::kan, metE::tn10) (21) on M63glucose plates supplemented with methionine sulfoxide (100 g/ml). After 72 h of growth, the sizes of at least 5 well separated colonies were measured and averaged.
Thioredoxin-and DsbC-dependent Folding of Urokinase-like Plasminogen Activator (uPA) in the Periplasm-Assays were performed with a wild-type (RI281) and trxA (RI363) strain (23) harboring pRDB8-A, a plasmid for constitutive periplasmic expression of mouse uPA (34). The pBAD33 expression plasmids for the thioredoxin variants were introduced into RI363. As a control, both RI363 and RI281 were cotransformed with the pBAD33 vector lacking the trxA gene. Bacteria were grown with ampicillin (100 g/ml) and chloramphenicol (10 g/ml) on NZ-rich medium (33) containing 0.2% glucose and 0.2% arabinose and harvested at mid-logarithmic phase. Whole cell lysates were prepared by five cycles of freezing/thawing in sample buffer followed by separation on a nonreducing SDS-polyacrylamide gel electrophoresis. uPA activity was demonstrated at 37°C by assaying plasminogen activation using plasminogen/casein agar as described (35).
Western Blot Analysis of Thioredoxin Expression-Aliquots of the same cell extracts that were used for the plasminogen activation assay were separated by reducing SDS-polyacrylamide gel electrophoresis; proteins were transferred to a nitrocellulose membrane and probed for thioredoxin by immunostaining with polyclonal rabbit antithioredoxin antibodies (gift from C. C. Richardson, Harvard Medical School, Boston, MA).

All Active Site Variants of Thioredoxin with Increased Redox
Potential Are Substrates of TR-NADPH constitutes the common donor of reducing equivalents for all thioredoxin-dependent reduction processes in E. coli. Thus, as a prerequisite for an in vivo characterization of the active site variants of E. coli thioredoxin listed in Table I, it had to be proven that the variants are still efficient substrates of TR. For this purpose, recombinant TR from E. coli was overexpressed and purified. The substrate properties of the variants toward E. coli TR were then tested using the TR-and thioredoxin-catalyzed reduction of Ellman's reagent (dithionitrobenzoic acid) by NADPH (31). In this assay, all thioredoxin variants showed, within a factor of two, the same substrate properties (k cat /K m values) as wildtype thioredoxin (Table I). Consequently, all variants principally have the potential of complementing thioredoxin deficiency in vivo. In contrast, DsbA wild type is not a substrate of TR (Table I).
In Vitro Model Reactions to Probe the Importance of Thioredoxin Redox Potential-To establish an in vitro model reaction that can probe the importance of thioredoxin redox potential in thioredoxin-catalyzed processes, we compared the different thioredoxin variants with respect to their ability to catalyze the reduction of insulin by DTT at pH 7.0. This reaction is a widely used standard assay for detecting thiol-disulfide oxidoreductase activity (36). Unexpectedly, this assay gave an inverse correlation between redox potential of the thioredoxin variants and the rate of insulin reduction, with the PDI-like, the glutaredoxin-like, and the DsbA-like variant being more efficient catalysts than wild-type thioredoxin (Fig. 1A). This result can however be rationalized by considering the fact that the ratelimiting step in the thioredoxin-catalyzed reduction of insulin by DTT is the re-reduction of thioredoxin by DTT (36) and that the more oxidizing variants, in accordance with their increased redox potentials, are indeed reduced 20 -40 times faster by DTT than the wild type (Table II, values of k 2 (DTT red )). A decreased rate of insulin reduction by the variants, which one might expect from their increased redox potentials, is obviously not a rate-limiting factor for this model reaction.
In a search for a model reaction that directly probes the redox potential of the thioredoxin variants, we thus had to choose another reaction where the reductive force of the thioredoxin variants and not their own reduction is rate-limiting for the catalytic cycle. First, we used NADPH and TR for rapid and efficient recycling of reduced thioredoxin (cf. Table I). As thioredoxin appears to be mainly in the reduced state in E. coli cells (37), it is likely that reduction of thioredoxin by NADPH is also not rate-limiting for the catalytic cycle of thioredoxin in vivo. Second, we looked for an alternative disulfide substrate that is reduced slower by thioredoxin than insulin. We compared insulin with the disulfide substrates lipoic acid and oxidized DTT with respect to the rate constants of their reduction by the reduced thioredoxin variants at pH 8.0. Unexpectedly, the rate constants of insulin reduction, measured by the in- The thioredoxin reductase assay was performed as described (31)  crease in thioredoxin fluorescence, were extremely high (10 4 -10 5 M Ϫ1 s Ϫ1 ). In contrast, oxidized DTT and lipoic acid were reduced 2-4 orders of magnitude slower (Table II). In particular, the rate constants decreased with increasing redox potential of the thioredoxin variants. Therefore, both oxidized DTT and lipoic acid appeared to be suitable substrates for a thioredoxin-catalyzed model reaction probing the redox potential of thioredoxin. As the apparent rate constants for the reduction of lipoic acid by the thioredoxin variants were more sensitive toward the redox potentials of the variants compared with the reduction of oxidized DTT (Table II), we chose lipoic acid as a model substrate for reactions with thioredoxin as reduction catalyst. Indeed, when the thioredoxin-catalyzed reduction of lipoic acid by NADPH was measured for all the variants at saturation of TR (20 M thioredoxin, cf. Table I), a good correlation between the initial velocities of lipoic acid reduction (measured by NADPH consumption) and the redox potentials of the variants was observed ( Fig. 1B and Table II).
Conversely, the reaction with insulin as substrate under the same conditions was entirely independent of the redox potentials of the variants with respect to initial NADPH consumption (Table II). HPLC analysis of insulin reduction also yielded very similar initial velocities for the reaction (Table II). The difference between insulin and lipoic acid reduction can be explained by calculating the actual velocities of each step catalyzed by thioredoxin. We found that recycling of thioredoxin by NADPH was still slower than the reduction of insulin under the conditions of our assay. With concentrations of 20 M for thioredoxin, 33 nM for TR, and 80 M for insulin and k cat values for thioredoxin reduction of 20 -30 s Ϫ1 (Table I), the calculated initial velocity of insulin reduction is at least ten times faster than recycling of reduced thioredoxin. Thus, insulin is such a good substrate of thioredoxin that at least 100-fold higher concentrations of TR would have been necessary to make insulin reduction the rate-limiting step.
In Vivo Characterization of Thioredoxin Variants with Different Redox Properties-Having established an in vitro assay for thioredoxin-catalyzed reduction of a disulfide substrate in which a clear correlation between redox potential and reactivity could be observed (Fig. 1B), we wanted to address the question whether the thioredoxin variants would show the same pattern of redox potential-dependent reactivity in vivo. The variants were cloned into the vector pBAD33 where protein expression is tightly regulated by the arabinose promoter (25). In the presence of the inducer arabinose (0.2% w/v) the expression levels in the trxA null mutant RI362 (23) were very similar for all thioredoxin variants as determined by Western blot analysis (Fig. 2).
Mutants of E. coli lacking thioredoxin have a number of readily assayed phenotypes. For instance, growth on minimal medium with sulfate as a sole source of sulfur requires either thioredoxin or glutaredoxin, because trxA, grxA double mutants cannot reduce sulfate (20). When the plasmids encoding the thioredoxin variants were introduced into a trxA, grxA double mutant, we found that they all complemented for growth on sulfate minimal medium, in contrast to the control plasmids lacking trxA or encoding the inactive thioredoxin variant C35A that lacks the buried active site cysteine (Table  III). The colony sizes correlated very well with the redox potentials of the complementing thioredoxin variants, with the most oxidizing glutaredoxin-like variant barely being able to sustain growth (Table III).
On minimal medium, growth in the presence of methionine sulfoxide should reflect the ability of the thioredoxin variants to complement a thioredoxin null mutant strain for the reduction of methionine sulfoxide to methionine in a methionine auxotroph strain (21). These experiments were carried out by introducing the set of thioredoxin expression plasmids to the trxA, metE strain A313 (21). Again, the same redox potentialdependent order of complementation as that for sulfate reduction was observed for the growth on methionine sulfoxide as the sole methionine source (Table III).
Periplasmic disulfide bond isomerization is dependent on the periplasmic disulfide isomerase DsbC. As only the reduced form of DsbC is catalytically active as an isomerase, it is maintained in the reduced state by reducing equivalents from NADPH in the cytosol, a process catalyzed by TR and thioredoxin in the cytoplasm, and by DsbD in the inner membrane of E. coli (22,23). The evidence for this pathway is that disruption of any of the corresponding genes leads to the accumulation of oxidized DsbC and a severe defect in disulfide bond isomerization during folding of proteins with multiple disulfide bonds in

Insulin reduction Lipoic acid reduction
Wild type (Gly-Pro) b Apparent second-order rate constant of the oxidation of thioredoxin by oxidized DTT at pH 8.0 and 25°C measured by the decrease in thioredoxin fluorescence.
c Apparent second-order rate constant of the reduction of insulin by thioredoxin at pH 8.0 and 15°C measured by the decrease in thioredoxin fluorescence.
d Apparent second-order rate constant of the reduction of lipoic acid by thioredoxin at pH 8.0 and 25°C measured by the decrease in thioredoxin fluorescence.
e Initial velocity of NADPH consumption during thioredoxin-catalyzed reduction of insulin by NADPH at pH 8.0 and 25°C, using concentrations of 500 M NADPH, 33 nM thioredoxin reductase, 80 M bovine insulin, and 20 M thioredoxin. The reaction was followed by the decrease in NADPH absorbance at 340 nm.
f Initial velocity of the decrease in native insulin during its thioredoxin-catalyzed reduction by NADPH, as judged by HPLC analysis. Same conditions as in Footnote e.
g Initial velocity of NADPH consumption during thioredoxin-catalyzed reduction of lipoic acid at pH 8.0 and 25°C, using concentrations of 500 M NADPH, 33 nM thioredoxin reductase, 1.0 mM lipoic acid, and 20 M thioredoxin. The reaction was followed by the decrease in NADPH absorbance at 340 nm.
the E. coli periplasm. Secretory expression of mouse uPA is a convenient model for monitoring periplasmic disulfide bond isomerase activity. uPA has six disulfide bonds in its protease domain and can only fold into a biologically active conformation when DsbC is functional in the periplasm. We assayed the ability of the thioredoxin active site variants to complement a chromosomal thioredoxin null mutation by monitoring their competence to restore DsbC activity and thus periplasmic uPA activity. Using this assay, complementation of the trxA null mutant RI362 (23) by the thioredoxin variants again decreased exactly with increasing redox potential of the variants (Fig. 2). As observed in the sulfate reduction assay, almost no complementation could be detected for the most oxidizing, glutaredoxin-like thioredoxin variant.

DISCUSSION
The question of the importance of the redox potential of thiol-disulfide oxidoreductases for their in vivo function has so far only been addressed for catalysts of disulfide bond formation and isomerization during protein folding in oxidizing cellular compartments. In the endoplasmic reticulum of yeast, PDI deficiency can only be complemented by secreted E. coli thioredoxin variants with increased redox potentials but not by wild-type thioredoxin (12), indicating that the oxidative force of this essential protein is crucial for its function in vivo. This view is supported by active site variants of yeast PDI that cause DTT sensitivity and a reduced rate of protein folding in the endoplasmic reticulum (38). In the periplasm of E. coli, active site variants of DsbA, which are more than 1000-fold weaker oxidants than the wild type, could still functionally replace DsbA under normal growth conditions (11,16,17). Efficient functional replacement of DsbA by thioredoxin could, however, only be achieved by secretion of the same three most oxidizing thioredoxin variants that were used in the present study (18,19). It thus appears that the redox potential of the final oxidant of polypeptides in the E. coli periplasm is also important and that a certain minimum of oxidative force must be retained for maintenance of disulfide bond formation in the periplasm.
The present investigation is the first study on the influence of the intrinsic redox potential of a cytoplasmic member of the thioredoxin family on its function in vivo. Complementation of three phenotypes of trxA strains, i.e. deficiency in reduction of sulfate, methionine sulfoxide, and periplasmic disulfide bond isomerization, by a series of thioredoxin variants with different redox potentials was investigated. In all cases, the degree of trxA complementation strictly increased with decreasing redox potential. Thus, the low redox potential of thioredoxin is clearly a very important factor for its in vivo function.
In general, a change in redox potential of a thiol-disulfide oxidoreductase can be caused by altered rate constants of its own reduction and/or its own oxidation. We found that the in vivo results on the different thioredoxin variants could only be reproduced in vitro under conditions where recycling of reduced thioredoxin is not rate-limiting for catalysis. We propose that the main requirement for the low redox potential of thioredoxin is the fact that reduction of thioredoxin by NADPH is not rate-limiting in vivo, as this reaction is very efficiently catalyzed by TR with a k cat /K m value of about 10 7 M Ϫ1 s Ϫ1 (Table I). Consequently, the reductive force of thioredoxin toward its protein substrates becomes the main determinant of its catalytic efficiency in vivo. This view is supported by the insulin reduction assays (Fig. 1A). Here the reduction of thioredoxin by DTT is rate-limiting for catalysis so that an inverse correlation between the redox potential of the thioredoxin variants and their catalytic efficiency was obtained, again as predicted from the order of their redox potentials. Consequently, an in vitro FIG. 2. Recovery of DsbC-dependent disulfide bond isomerization in the periplasm of the trxA strain RI363 by plasmid-encoded cytoplasmic expression of thioredoxin variants. A, thioredoxin-dependent disulfide bond isomerization in E. coli was assayed by the activity of plasmid-encoded mouse uPA in the periplasm, a protein with multiple disulfide bonds that requires DsbC and thioredoxin for efficient folding. Cell extracts were subjected to nonreducing SDSpolyacrylamide gel electrophoresis, and the gel was subsequently underlayed with agar containing plasminogen and casein. The dark zones result from proteolytic degradation of casein by activated plasmin, which is generated by functional uPA. uPA activity in the wild-type (WT) strain RI281 (23) cotransformed with the expression vector pBAD33 lacking trxA (lanes 1 and 2 (1/100 dilution of 1)) was compared with the uPA activity in the trxA strain RI363 (23) (32) was transformed with the pBAD33 derivatives containing the genes of the corresponding thioredoxin variants and then grown at 37°C on agar plates with M63 minimal medium (containing 10 g/liter ammonium sulfate as sole source of sulfur) in the presence of the inducer L-arabinose. The size of at least 5 colonies after 3 days of growth was measured for each complementation assay and averaged.
b The trxA, metE strain A313 (21) was transformed with the pBAD33 derivatives containing the genes of the corresponding thioredoxin variants and grown at 37°C on agar plates with M63 minimal medium supplemented with the inducer L-arabinose and methionine sulfoxide as sole methionine source. Colony sizes were determined after 3 days of growth as in Footnote a . assay for thiol-disulfide oxidoreductases that reflects the in vivo situation must be performed under conditions that guarantee that the rate-limiting catalytic step is the same as that in vivo. We believe that our newly established assay, making use of the catalyzed reduction of lipoic acid by NADPH, is a valuable tool to study the function of NADPH-dependent members of the thioredoxin family in vitro.
Another question that needs to be addressed in the context of our in vivo experiments is whether the intracellular redox potential in the cytoplasm determines a steady-state ratio of oxidized and reduced thioredoxin and thus possibly the available concentrations of reduced thioredoxin. One would indeed expect that the intracellular concentration of the reduced relative to the oxidized enzyme increases in the case of the more oxidizing thioredoxin variants compared with wild-type thioredoxin. However, it has been shown that wild-type thioredoxin is already mainly reduced in the cytoplasm (about 70% in freshly prepared cell extracts (37)). Therefore, assuming unchanged overall concentrations of thioredoxin, the concentration of reduced thioredoxin can only increase to a small extent in the case of the more oxidizing variants. Thus, the concentration of reduced thioredoxin cannot be affected significantly in the case of the more oxidizing variants. This further supports our conclusion that it is the rate by which substrates are reduced by thioredoxin that determines the in vivo function of this enzyme and that slower substrate reduction by the thioredoxin variants is a consequence of their increased redox potential. We have also demonstrated that the thioredoxin expression level in our in vivo complementation studies is practically the same as that in wild-type cells (Fig. 2B).
In the present study we have demonstrated the in vivo importance of redox potential for a single cytoplasmic member of the thioredoxin family. We would like to emphasize that a correlation between redox potential and in vivo function does not necessarily have to exist in general. On the one hand, it follows from the above consideration that the rate-limiting step in the catalytic cycle of a thiol-disulfide oxidoreductase is the main determinant to what extent and in which direction its intrinsic redox potential influences its in vivo function. On the other hand, the situation in vivo is more complex because there are often several enzymes that fulfill the same function but may recognize their substrates with different specificities. For instance, ribonucleotide reductase (NrdAB) has a lower K m value for glutaredoxin 1 than for thioredoxin, but glutaredoxin 1 has a higher redox potential (39,40). Similarly, only glutaredoxin 1 and the thiol-disulfide oxidoreductase NrdH are effective reductants of the second E. coli ribonucleotide reductase, NrdEF (41), whereas thioredoxin shows no reactivity despite the fact that it is the strongest reductant of the three enzymes. It will thus be the combined effect of redox potential, substrate specificity, and expression levels of related thiol-disulfide oxidoreductases that accounts for the different phenotypes that are observed for E. coli mutants lacking components of the thioredoxin-and glutaredoxin-dependent pathways. In the case of NADPH-dependent members of the thioredoxin family, the expression level of TR may determine the rate-limiting step of catalysis. Clearly, all of these factors have to be taken into account to get a complete view of the in vivo functions of each individual member of the thioredoxin family.