Catalysis of Thiol/Disulfide Exchange

Glutaredoxin (Grx) and protein-disulfide isomerase (PDI) are members of the thioredoxin superfamily of thiol/disulfide exchange catalysts. Thermodynamically, rat PDI is a 600-fold better oxidizing agent than Grx1 from Escherichia coli. Despite that, Grx1 is a surprisingly good protein oxidase. It catalyzes protein disulfide formation in a redox buffer with an initial velocity that is 30-fold faster than PDI. Catalysis of protein and peptide oxidation by the individual catalytic domains of PDI and by a Grx1-PDI chimera show that differences in active site chemistry are fundamental to their oxidase activity. Mutations in the active site cysteines reveal that Grx1 needs only one cysteine to catalyze rapid substrate oxidation, whereas PDI requires both cysteines. Grx1 is a good oxidase because of the high reactivity of a Grx1-glutathione mixed disulfide, and PDI is a good oxidase because of the high reactivity of the disulfide between the two active site cysteines. As a protein disulfide reductase, Grx1 is also superior to PDI. It catalyzes the reduction of nonnative disulfides in scrambled ribonuclease and protein-glutathione mixed disulfides 30–180 times faster than PDI. A multidomain structure is necessary for PDI to catalyze effective protein reduction; however, placing Grx1 into the PDI multidomain structure does not enhance its already high reductase activity. Grx1 and PDI have both found mechanisms to enhance active site reactivity toward proteins, particularly in the kinetically difficult direction: Grx1 by providing a reactive glutathione mixed disulfide to supplement its oxidase activity and PDI by utilizing its multidomain structure to supplement its reductase activity.

Members of the thioredoxin superfamily of redox proteins transfer reducing and oxidizing equivalents between proteins through thiol-disulfide exchange. With overall structural similarity, the classical thioredoxin and glutaredoxin members of the family all have an active site with the sequence CXXC (1). The two active site cysteines are able to cycle between dithiol and disulfide redox states. Despite the similarities in the active sites and a general mechanism involving thiol-disulfide exchange, thioredoxin family proteins serve a diverse set of bio-logical functions. The family patriarch is thioredoxin, the principal function of which is to provide reducing equivalents for deoxyribonucleotide synthesis by reducing the active site disulfide of ribonucleotide reductase (2). Glutaredoxins (Grx), 1 which have a GSH binding site (3), serve as reductants of protein-SG mixed disulfides and also provide reducing equivalents to ribonucleotide reductase apart from their many other functions (4). Protein-disulfide isomerase (PDI) and its prokaryotic counterpart, DsbA, function primarily to introduce disulfides into protein substrates during oxidative protein folding (5,6).
In thioredoxin and PDI, the more N-terminal cysteine, CXXC, of the active site is exposed to solution (7) and acts as a nucleophile to attack substrate disulfides (8). The second active cysteine, CXXC, is buried and constrained to react with only the N-terminal active site cysteine (7). The reduction potentials of the active site disulfides of the thioredoxin family span an extremely large range, depending, in part, on the identity of the two XX residues and the protein context in which the CXXC sequence occurs. The reduction potential of the best oxidant, DsbA (Ϫ124 mV) differs from the reduction potential of the worst oxidant, thioredoxin (Ϫ270 mV) by more than 146 mV, corresponding to a ratio of nearly 10 5 in thermodynamic stability during the disulfide/dithiol equilibrium (9). Within the family, there is a strong correlation between the identity of the two intervening residues and the redox potentials (10 -12). Both Raines and co-workers (12) and Creighton and co-workers (13) have argued that much of the difference in reduction potential can be attributed to factors that stabilize the thiolate anion of the reduced active site. The active site cysteines have been suggested to form a thiol-thiolate hydrogen bond network, which stabilizes the more N-terminal active site thiolate in the reduced form of the proteins (14 -16).
In general, the kinetic properties of the thioredoxin family proteins parallel their thermodynamic ones. In its reduced state, thioredoxin and glutaredoxin are exceptional reductants of both protein and nonprotein disulfides (2,4). By contrast, the poorest reductants thermodynamically, PDI and DsbA, are also poor reductants kinetically. When comparing the kinetics of substrate disulfide formation using the oxidized (disulfide) active site, DsbA and protein-disulfide isomerase are the best oxidants kinetically. A link between thermodynamics and kinetic reactivity is not unexpected, given the overall mechanism of thiol-disulfide exchange, where the fundamental chemical process is the same among all the family members. For thiol/ disulfide exchange, the transition state is symmetrical (18), suggesting that factors affecting the stabilities of either the thiolate or disulfide will also be reflected in the transition state for thiol-disulfide exchange.
Glutaredoxins, also known as thioltransferases, are predominately localized to the cytoplasm but also have been detected in the nucleus and mitochondria (19 -21). Glutaredoxins catalyze the reduction of protein-SG mixed disulfides, preventing their accumulation in the reducing environment of the cytoplasm (4). By contrast, PDI and DsbA are localized in the more oxidizing environment of the endoplasmic reticulum and periplasm, respectively, in keeping with their function as oxidases.
Escherichia coli Grx1 is a single domain protein, but mammalian PDI is a multidomain catalyst (22). PDI is composed of four structural domains, all of which adopt a thioredoxin fold. The two catalytic domains of PDI (a and aЈ) are separated by two intervening structural domains (b and bЈ). The multidomain structure is important in the catalysis of disulfide isomerization during protein folding, but it is not essential for PDI oxidase activity. PDI catalyzes disulfide isomerization in vivo, but its principal function in yeast is to serve as an oxidase for disulfide formation in the endoplasmic reticulum (6).
In light of the expected linkage of kinetics and thermodynamics in the thioredoxin family, we were struck by the observation that Grx1 is a much better oxidase for protein dithiols than PDI despite the fact that PDI is a 600-fold better oxidant, thermodynamically (9). To investigate the basis of the unusual versatility of glutaredoxin as a protein oxidase and reductase, we have compared the kinetics of disulfide formation and reduction catalyzed by Grx1 and PDI. We also constructed a chimerical protein with the Grx1 structure in place of one of the active site domains in PDI.
Mutation of the more C-terminal active site cysteines of Grx1 and PDI shows that Grx1, but not PDI, can utilize a single active site cysteine to catalyze protein oxidation through the formation of a Grx-SG mixed disulfide. As an oxidase, Grx1 provides oxidizing equivalents to its substrates through a reactive mixed disulfide with glutathione. As protein-disulfide reductases, both Grx1 and PDI require two active site cysteines. For PDI, its multidomain structure is needed to catalyze protein disulfide reduction, whereas multiple domains are not needed to catalyze substrate oxidation. It is suggested that both Grx1 and PDI have developed specialized mechanisms to enhance catalysis of reactions that would normally be difficult because of the thermodynamic stability of the active site thiols and disulfides.

EXPERIMENTAL PROCEDURES
Materials-Bovine pancreatic ribonuclease A (RNase), GSH, GSSG, NADPH, cCMP, and glutathione reductase were from Sigma. The peptide substrate used in this study was purchased from Genemed Synthesis (South San Francisco, CA). It has the sequence NRCSQGSCWN, with the N-and C-terminal groups acetylated and amidated, respectively. Peptide concentrations in solution were determined using the molar extinction coefficient at 280 nm of 5690 M Ϫ1 cm Ϫ1 calculated according to Gill and von Hippel (23).
Plasmids and Mutagenesis-PDI mutants with a single active site cysteine were produced previously (8). Cys 14 of E. coli Grx1 was mutated to Ser using the QuikChange® site-directed mutagenesis kit from Stratagene to create Grx1C14S. The method of Xiao et al. (24) was employed to create the Grx1-PDI chimera, Grx-b-bЈ-aЈ(SGHS)c. The vector combined an N-terminal Grx1 domain with the b-bЈ-aЈc domains of PDI using a GlySer linker. The cysteines in the C-terminal, aЈ, catalytic domain of PDI were both mutated to serine so that the only catalytic contribution must come from the Grx1 domain. For PDI, Grx1, and their mutants, proteins were expressed from pET15rx with an N-terminal His 6 tag (24). Proteins were purified using a HiTrap Chelating column (Amersham Biosciences) as described (24) and were greater than 90% pure by SDS-PAGE.
Preparation of RNase Variants-Fully reduced RNase was prepared by reduction of the native protein in urea as described (25). The mixed disulfide of RNase and glutathione (SG-RNase) was prepared as described in Ref. 26, but without HPLC purification. Scrambled RNase A was prepared as described in Ref. 27.
Peptide Oxidase Activity-Thiol-disulfide exchange reactions between peptide, Grx1, and PDI or its domains were carried out in 0.1 M Tris-HCl, pH 7.4, 0.2 M KCl, 1 mM EDTA, 0.5 mM GSSG, and 2 mM GSH at 25°C with a peptide concentration of 20 M and at an enzyme concentration of 1 M. Aliquots of 100 l were removed after different reaction times, and thiol-disulfide exchange was quenched by the addition of HCl to a final concentration of 0.2 M. Oxidized and reduced peptide species in the quenched reaction mixture were separated and analyzed on Sephasil C18 SC 2.1/1.0 column (SMART system; Amersham Biosciences) using a linear gradient of 5-25% (v/v) acetonitrile in 0.1% (v/v) trifluoroacetic acid in 25 min at a flow rate of 100 l min Ϫ1 . At the end of this gradient, the acetonitrile concentration was increased to 75% (v/v) to elute the absorbed enzyme. Peptides were detected and quantified by their absorbance at 215 nm. The initial velocity of the peptide oxidation was determined from the initial rate of formation of the oxidized peptide.
Oxidase Activity toward Reduced RNase-In 0.1 M Tris-HCl, pH 8.0, the renaturation of RNase activity was followed by the concurrent hydrolysis of cCMP due to the gain of RNase activity as described by Lyles and Gilbert (28). A glutathione redox buffer (1 mM GSH, 0.2 mM GSSG) was used to maintain the optimum redox state. The final concentration of reduced RNase was 8 M. Plots of the concentration of active RNase against time show a characteristic lag, during which time the reduced RNase is oxidized to a mixture of nearly random disulfides (25,29). The lag time was determined by extrapolating the linear gain in RNase activity after the lag back to zero active RNase formed. The initial velocity of dithiol oxidation was estimated from the lag time by assuming that lag represents approximately three half-lives for the oxidation of the four disulfides of RNase. Whereas this is a rough approximation of the initial velocity, the rate is proportional to the concentration of the oxidase (8), and the lag can be measured reproducibly. The background rate in the absence of catalyst was subtracted. For catalysts such as Grx1 that do not display any isomerase activity, a small amount of wild type PDI or an oxidase-deficient mutant with only a single cysteine active site (PDI N CGHS :C SGHS) was included at a concentration of 1 M. In this case, the background velocity was that observed in the absence of Grx1.
Reduction of Scrambled RNase and SG-RNase-The reduction of SG-RNase was followed by coupling the formation of GSSG to the oxidation of NADPH using glutathione reductase. Reaction mixtures contained 100 mM potassium phosphate buffer, pH 7.0, 1 mM EDTA, 1 mM GSH, 100 g/ml bovine serum albumin, 100 M NADPH, yeast glutathione reductase (0.14 units/ml), and 5 M SG-RNase. The reactions were started by adding PDI or Grx1. Rates were corrected for the slow uncatalyzed reduction that occurs in the absence of PDI or Grx1. The reduction of 25 M scrambled RNase was carried out in the presence of 3 mM GSH in 0.1 M Tris-HCl, pH 8.0, 0.15 mM NADPH, and 1 unit/ml glutathione reductase. The oxidation of NADPH was followed at 340 nm in the presence of 0 -10 M Grx1 or PDI, Grx1C14S, or PDI N CGHS :C CGHS . The specific activity was calculated using the molar extinction coefficient of 6300 M Ϫ1 cm Ϫ1 for NADPH.
Reduction of Grx1-The stoichiometric reduction of Grx1 by GSH was observed by coupling the formation of GSSG to the oxidation of NADPH with glutathione reductase. The reaction was monitored by the change in fluorescence of NADPH using a cut-off filter of 395 nm (excitation 340 nm). The reactions were initiated by mixing equal volumes of oxidized Grx1 (3 M final concentration) with a solution of NADPH (5 M final concentration), GSH (0 -1 mM), and glutathione reductase (10 units/ml), all in 0.2 M potassium phosphate buffer, pH 7.0, at 25°C. Data were acquired using a SF-2004 stopped-flow instrument (KinTek Corp.). At each GSH concentration, the first-order rate constant for the disappearance of NADPH fluorescence was determined by nonlinear least squares fitting to a monoexponential decrease followed by a slow, linear change in absorbance with time due to a slow oxidation of NADPH. This linear correction was significant only at concentrations of GSH below 0.2 mM, and, even then, the correction was small compared with the total change in fluorescence. To monitor the competence of the trapping reaction for GSSG, 3 M GSSG was reduced by NADPH and glutathione reductase at the same concentrations used to follow Grx1 reduction. The rate constant for the disappearance of GSSG (0.85 s Ϫ1 ) was at least 4 times the maximum observed rate constant for Grx1 reduction. The rate constants for Grx1 reduction as a function of GSH concentration were determined by nonlinear least squares fitting to Equation 2 as described under "Results" (30,31).

RESULTS
Catalysis of Peptide Oxidation-Protein substrates are complex, with multiple disulfides, both native and nonnative. To observe the intrinsic reactivity of the various active sites on a well defined substrate, and to determine whether protein substrates are unique substrates for PDI and/or Grx1, the catalyzed oxidation of a small peptide (NRCSQGSCWN) was examined. Typical data are shown in Fig. 1, and the results for a more extensive series of catalysts are summarized in Table I. With this peptide substrate, native PDI and Grx1 are both efficient oxidases. PDI requires both active site cysteines to catalyze peptide oxidation, consistent with a dithiol mechanism of disulfide formation as with protein substrates. The individual catalytic domains of PDI are also effective oxidases, comparable with the full-length molecule. The other domains of PDI (Table I) and the potential substrate interactions they may provide are not important to peptide oxidase activity. However, Grx1 is still a superior oxidase (5-fold based on active site concentrations) when compared with PDI. A CXXS mutation in Grx1(Grx1C14S) has no significant effect on the peptide oxidase activity, suggesting that Grx1 catalyzes the peptide oxidation through a monothiol mechanism that requires only the nucleophilic cysteine ( Fig. 1 and Table I).
Catalysis of Protein Oxidation-During the catalysis of the oxidative refolding of reduced RNase, the initial phase of disulfide formation rarely produces native RNase (32). Consequently, there is a significant lag time before the appearance of the enzymatically active, native RNase. This lag is associated with the formation of RNase disulfides that must be broken and reformed to produce native RNase, and it can be used as an indicator of the rate of protein oxidation (8). As can be seen in Fig. 2, the addition of the catalyst of oxidative folding, PDI, significantly decreases the lag compared with the uncatalyzed reaction as well as catalyzing the isomerization of incorrect disulfides leading to native RNase formation. A small amount of Grx1, in addition to the PDI that is present, greatly decreases the lag (accelerates the oxidation) without substantially contributing to the subsequent isomerase activity, consistent with the results reported by Lundström-Ljung and Holmgren (33). PDI is a 600-fold better oxidant than Grx1 thermodynamicaly (9), but enzymatically, the active site of Grx1 is a 30-fold better oxidase than PDI in a redox buffer consisting of 1 mM GSH and 0.2 mM GSSG (Table II).
During the oxidation of protein substrates, PDI catalyzes direct transfer of its active site disulfide into the substrate; both PDI active site cysteines are essential for effective catalysis of protein oxidation. PDI mutants with only the nucleophilic (N-terminal) cysteine (CGHS) are not active oxidases and do not decrease the lag during the oxidative folding of reduced RNase (Table II). However, mutation of the active site of Grx1 to remove the C-terminal cysteine has no significant effect on the ability of Grx1 to decrease the lag in RNase refolding. The observation that wild type Grx1 and Grx1C14S display comparable oxidase activities even at very low concentrations (50 nM) indicates that Grx1 can catalyze protein oxidation with only one of its active site cysteines and that the transfer of oxidizing equivalents from the redox buffer (GSSG) to the substrate most likely occurs through a Grx-SG mixed disulfide.
In dealing with protein substrates, significant interactions between the catalyst (Grx1 or PDI) and the substrate protein (RNase) might contribute to their effectiveness as an oxidase. However, the catalytic domains of PDI are, by themselves, just as effective an oxidant as the intact molecule (34). To determine whether Grx1 might utilize a protein interaction domain in making protein oxidation even more effective, the N-termi-   2. Oxidation of reduced RNase catalyzed by PDI and Grx1. The oxidative folding of reduced RNase is preceded by a significant lag period in which inactive, oxidized species of ribonuclease form from the reduced RNase. The length of the lag period is related to the ability of the catalyst to accelerate RNase oxidation (8). Following oxidation, the isomerase activity of PDI resolves misoxidized RNase producing the native RNase through disulfide isomerization. The activity of RNase was assayed continuously as it formed during the assay by including the substrate, cCMP, in the assay. The assay was performed at pH 8. nal catalytic domain of PDI was replaced by Grx1 to create a Grx1-PDI chimera, Grx-b-bЈ-aЈ(SGHS)c. The two cysteines of the C-terminal catalytic domain of PDI were also mutated to Ser so that the activity of Grx1 in the context of the rest of the PDI molecule might be observed. As with PDI, the addition of the protein interaction domains of PDI had no significant effect on the ability of Grx1 to function as an oxidant (Table II). Whereas we cannot be sure that the orientation of the catalytic site in the chimeric protein would allow Grx1 to take advantage of a peptide/protein biding site from the b-bЈ-aЈ-c domains of PDI, the ability of Grx1 and PDI to serve as oxidases relies almost exclusively on the high reactivity of an active site disulfide to introduce disulfides into protein substrates. In the case of Grx1, this active site disulfide is most likely a disulfide with glutathione, whereas for PDI, it represents an intramolecular disulfide between the two active site cysteines.
Catalysis of Reduction-Because Grx1 is such a superior oxidase compared with PDI, we were also interested in directly comparing the ability of the two proteins to catalyze reduction. In line with its very negative reduction potential, Grx1 is a more effective reductase than PDI (16-fold on a molar basis) in catalyzing the reduction of the protein disulfides in scrambled RNase. However, Grx1 now requires both active site cysteines to be an effective reductant. Mutation of the more C-terminal cysteine in the Grx1 active site decreases the reductase activity by 15-fold, indicating that like PDI, both active site cysteines make a significant contribution to catalyzing protein disulfide reduction (Table  III). Bushweller et al. (35) have found that Grx1 reduction of ribonucleotide reductase requires both active site cysteines, consistent with our observation that catalysis of protein disulfide reduction by Grx1, in general, needs both active site cysteines.
Given that the active site thermodynamics suggest that Grx1 is a 600-fold better reductant (redox potential 83 mV more negative) than PDI, the 15-fold difference observed in reductase activities (Table III) is surprisingly low. Darby et al. showed that the individual domains of PDI (a or aЈ) by themselves are very poor reductases (2-3% of wild type PDI activity) (22). Thus, the multidomain structure of PDI enhances the reductase activity considerably (30 -50-fold); the catalytic domains are very poor reductases in the absence of the catalystsubstrate interactions that are available from its multiple domains. If substrate interactions contribute 30 -50-fold to catalysis of reduction, and Grx1 is a 15-fold better reductant than PDI, we estimate that Grx1 is a 450 -750-fold better reductant than the catalytic domains of PDI, in line with the 600-fold (83 mV) difference in reduction potentials.
The effect of creating a multidomain structure with a Grx1 catalyst was examined in a chimera in which the catalytic PDIa domain was replaced by Grx1 (Grx-b-bЈ-aЈ(SGHS)-c). The activity of the chimeric protein was slightly lower than Grx1 (Table III). Thus, in contrast to our expectations, appending Grx1 to the peptide/protein interaction domains of PDI does not enhance the reductase activity of Grx1. The Grx1-PDI chimera is still an effective catalyst of protein oxidation (Table II) so that the noncatalytic PDI portion of the molecule does not affect the reactivity of the Grx1 active site as an oxidase. Thus, Grx1, a superior reductant to begin with, benefits little from  additional substrate-catalyst interactions that may be available from the noncatalytic domains of PDI, whereas the same interactions do enhance the reductase activity of the intrinsically poor reductant, PDI.
Lundström-Ljung et al. (26), found that the glutaredoxins, including Grx1, are good reductases for protein-glutathione mixed disulfides. Comparing Grx1 and PDI for their ability to catalyze reduction of the mixed disulfides of denatured RNase and glutathione (SG-RNase), Grx1 is 200-fold more active against this substrate than a single catalytic domain of PDI (Fig. 3, Table IV). For this substrate, mutating the second active site cysteine to serine decreases the reductase activity of Grx1 by ϳ3-fold. This indicates that there may be some contribution of a two-cysteine reduction mechanism for protein-SG mixed disulfides but that a single-cysteine dominates the reduction mechanism. The PDI catalytic domain (a domain; Table IV), on the other hand, loses its ability to reduce protein-SG mixed disulfides when the more C-terminal cysteine is mutated to serine, suggesting that both active site cysteines are needed, even for this substrate. Thus, for protein-SG mixed disulfides, the specificity of Grx1 for glutathione (35) provides an alternative mechanism of protein reduction that involves only one active site cysteine.
Reduction of the Grx1 Active Site-During catalysis of oxidation, the active site of Grx1 is reduced by the protein substrate and then recycled through oxidation by GSSG (Fig. 4). To observe the reactivity of the Grx1 disulfide and the Grx1-SG mixed disulfide, rapid reaction kinetics were employed. The oxidation of glutathione to GSSG by the oxidized active site of Grx1 requires two consecutive reactions, the formation of a Grx1-SG mixed disulfide and the subsequent reaction of the mixed disulfide with GSH (Scheme 1). These reactions must occur during the catalysis of oxidation and reduction by Grx1 or PDI in order to recycle the catalyst for another round of reduction (or oxidation). To determine the fundamental rate constants for the reaction of Grx1 with glutathione, oxidized Grx1 was rapidly mixed with various concentrations of GSH in a stopped-flow fluorescence spectrophotometer. The GSSG formed as a product of the reaction was trapped by the glutathione reductase-catalyzed oxidation of NADPH to NADP ϩ (31). Conditions were chosen to ensure that there would be sufficient NADPH to reduce all of the GSSG formed (5 M NADPH and 3 M Grx1) and that the trapping of GSSG was faster than the observed reactions (Fig. 5).
The observed pseudo-first-order rate constants for Grx1 reduction were determined at various GSH concentrations. At low concentrations of GSH, the reduction of Grx1 is clearly second-order in GSH concentration (Fig. 5). This second-order dependence requires that the reaction of GSH with the active site disulfide is at equilibrium (k 2 /k 3 Ͼ Ͼ [GSH]; Scheme 1), that the Grx-active site disulfide is not fully converted to the mixed disulfide, and that the rate-limiting step is the reaction of the Grx1-SG mixed disulfide with GSH (k 3 ). For PDI, at GSH SCHEME 1

FIG. 4. Mechanism of protein oxidation catalyzed by Grx1 and by PDI.
Protein oxidation by PDI proceeds predominantly through a mechanism that requires both active site cysteines and transfer of a disulfide into the substrate (A), whereas Grx1-catalyzed oxidation proceeds predominantly through pathways that do not employ the second active site cysteine (B).
concentrations of about 0.3 mM, there is a change in the ratedetermining step to a reaction that is first-order in GSH, as the reaction of the mixed disulfide of PDI with GSH becomes ratelimiting (31). In this case, the reaction changes from secondorder to first-order in GSH concentration. For Grx1, a fit of the data to the entire kinetic model, including the potential for a change in rate-limiting step (Equation 1), is not statistically better than a simple second-order dependence (Equation 2). Thus, at the GSH concentrations that can be used without the rate of Grx1 reduction becoming faster than the trapping reaction, there is no observable change in rate-determining step for Grx1 reduction. However, this places a lower limit on the GSH concentration where this change in rate-determining step could occur (k 2 /k 3 ratio). With a second-order dependence on GSH concentration, the apparent first-order rate constant for GSH reduction of Grx1 (k obs ), is given by Equation 2.
Under conditions where [GSH] Ͻ Ͻ k 2 /k 3 , the third-order rate constant for reduction of Grx1, k GSH (Table V), is equal to k 1 k 3 /k 2 . Because the overall equilibrium constant for the reaction is known (k 1 k 3 /k 2 k 4 ϭ 1/404 mM Ϫ1 ), the value of k 4 and k 3 can be calculated (Table V). The fact that there is no evidence for a change in rate-limiting step at the experimentally accessible concentrations of GSH sets a lower limit on k 2 /k 3 and places lower limits on k 1 and k 2 ( Table V). The rate constants for the GSH reduction of the isolated N-terminal thioredoxin (a) domain of PDI are also known under similar conditions (30,31). These results show that the reaction of the Grx1-SG mixed disulfide with GSH (k 3 ) is faster (ϳ3-fold) than the reaction of PDI-SG mixed disulfide with GSH. The Grx1-SG mixed disulfide (k 3 ) is also much more reactive with GSH than the intramolecular active site disulfide of PDI (k 1 ) by a factor of ϳ400-fold. It is possible that the Grx1-SG mixed disulfide is much more reactive with GSH (k 3 ) than the intramolecular active site disulfide of Grx1 (k 1 ); however, this suggestion is limited by our ability to set only a lower limit on k 1 . In addition, Grx1 and PDI reduction and oxidation by GSH and GSSG are sufficiently fast that the rate-limiting half-reaction during catalyzed protein and peptide oxidation and reduction must be the reaction of Grx1 and PDI with the peptide or protein substrates rather than the recycling of the active site redox state through reaction with GSH or GSSG. Consequently, the relative rates of peptide and protein oxidation and reduction by Grx1 and PDI reflect reactions of the catalysts with the peptide or protein substrates. DISCUSSION For most members of the thioredoxin family, the kinetic reactivities of the dithiol active sites generally parallel the thermodynamics of disulfide formation within the active site. The least stable disulfides are the best oxidases, whereas the least stable dithiols are the best reductases. Grx1 is an exception. As expected from its redox potential, Grx1 is a better reductase than an individual catalytic domain of PDI (30 -200fold ; Tables III and IV), but surprisingly, Grx1 is also a better oxidase than PDI (Tables I and II). One potential explanation for the superiority of Grx1 as an oxidase might lie in the proportion of the active site that is in the oxidized state under the conditions where its oxidase activity is measured. In the redox buffers used to measure the oxidase activity (Tables I  and II), only a fraction of the PDI is expected to be in the oxidized state (ϳ10%), whereas Grx1 is expected to be mostly oxidized (ϳ99%). In addition, the proportions of the Grx1 active site that are present as the intramolecular disulfide and the Grx1-SG mixed disulfide can be calculated from the micro-  scopic equilibrium constants (K mix and K intra ; Table V). At equilibrium with the redox buffer (1 mM GSH, 0.2 mM GSSG), Grx-SG would represent 1.4% of the total Grx1, whereas the intramolecular disulfide would represent 97%. Although a minor species at equilibrium, the high reactivity of the Grx1-SG mixed disulfide suggests that it can be a kinetically competent oxidant. For example, it reacts with GSH ϳ400-fold faster than the active site disulfide of PDI. The Grx1-SG mixed disulfide is the most reactive glutathione mixed disulfide described to date, and its increased thermodynamic stability allows it to accumulate to a significant extent in a redox buffer.
In contrast to Grx1, PDI requires both active site cysteines to accomplish effective disulfide formation. The fact that the velocity of substrate oxidation is less than 10% of a wild type active site (depending on the substrate) when the PDI active site has only the nucleophilic cysteine suggests that disulfide formation by PDI is dominated by a direct transfer of a disulfide to the substrate (Fig. 4A) and that glutathione involvement is not required. This is further supported by in vivo data from a genetic analysis in yeast that has defined a core pathway for protein disulfide bond formation in the ER, whereby PDI transfers oxidizing equivalents derived from Ero1p to secretory proteins by direct thiol-disulfide exchange reaction. In addition, depletion of glutathione from the ER has no effect on disulfide bond formation in secretory proteins (36,37) By contrast, the second active site cysteine in Grx1 contributes minimally to catalysis. Grx1 and the Grx1C14S mutant, have comparable activity in shortening the lag phase of RNase folding and introducing a disulfide bond into a peptide (Tables I and II). It has been previously shown that the synergistic effects of Grx1 together with PDI in refolding of RNase appear exclusively in the presence of glutathione (33). Thus, Grx1-catalyzed oxidation generally occurs by mono-thiol mechanisms (Fig. 4B), where glutathione involvement is necessary and only one active site cysteine is formally required.
The high oxidase activity of Grx1-SG suggests that in addition to its usual role as a cytoplasmic reducing agent, Grx1 may also participate in catalyzing protein oxidation during oxidative stress. The high reactivity of the disulfide of Grx1-SG may be due to the abnormally low pK a of the N-terminal Cys residue in Grx1 and the resulting ability to serve as a good leaving group when attacked by a protein thiol (17, 38 -40).
The oxidase activity of PDI does not utilize protein-protein interactions that might be available from the multidomain structure of PDI. A single catalytic domain is just as good an oxidase as wild type PDI. By contrast, PDI-catalylzed reduction does require the multidomain structure; the individual catalytic domains are ineffective reductases. Protein reduction is facilitated by interactions between the substrate and the other domains of PDI, substantially increasing the ability of PDI to serve as a protein reductant. The function of the additional substrate interactions may be to affect local unfolding of the substrate, or the enhanced binding interaction may decrease the dissociation rate of the substrate, making reduction more likely. Which mechanism dominates in the reduction of protein substrates by PDI is not clear. However, adding the proteininteraction domains of PDI to Grx1 does not enhance its oxidase or reductase activities, suggesting that the extraordinary reactivity of the Grx1 active site and the flexibility of providing a Grx1-SG mixed disulfide is responsible for enhancing its abnormal oxidase activity.
In summary, Grx1 and PDI are unusual members of the thioredoxin family. Grx1 provides mechanisms for protein oxidation and reduction that are not available to the other family members, mechanisms that involve a single-cysteine mecha-nism and the high reactivity of a Grx-SG mixed disulfide. This alternative mechanism compensates for the very negative reduction potential of the Grx1 active site and suggests that Grx1, despite its highly reducing nature, may also catalyze significant protein oxidation. The formation of protein-SG mixed disulfides by Grx1 through a monothiol mechanism may play an important role in protecting against more drastic, irreversible modifications of protein thiols, particularly when the redox state of the cytoplasm becomes more oxidizing, as under conditions of oxidative stress. Several monothiol glutaredoxins have been identified in yeast (yGrx3, yGrx4, and yGrx5) (41). The lethality of ⌬grx3⌬grx4⌬grx5 mutations suggests that monothiol glutaredoxins are very specific for their substrates, and their functions cannot be replaced by their dithiol counterparts. The biological activities of yGrx3 and yGrx4 are not known, but yGrx5 is involved in the maturation of Fe-S clustercontaining proteins in the mitochondria (21) and plays a central role in defense against protein oxidative damage (41). PDI is unusual because it is a more effective reductant than expected based on its redox potential. In this case, the efficiency of reduction is increased by involving PDI-substrate interactions in facilitating substrate reduction, an activity that is required to affect disulfide isomerization during oxidative protein folding (42).