In Vivo Control of Redox Potential during Protein Folding Catalyzed by Bacterial Protein Disulfide-isomerase (DsbA)*

The formation of disulfide bonds in Escherichia coli is catalyzed by periplasmic protein disulfide-isomerase (DsbA). When the a-amylaseltrypsin inhibitor from Ragi, a protein containing five intramolecular disulfide bridges, is secreted into the periplasm of E. coli, large amounts of misfolded inhibitor with incomplete or in- correct disulfides are accumulated. Folding of the inhibitor in the periplasm is not improved when DsbA is coexpressed and cosecreted. However, an up to 14-fold increase in correctly folded inhibitor is observed by co- expression of DsbA in conjugation with the addition of reduced glutathione to the growth medium. This pep- tide acts as a disulfide-shuffling reagent and can pass the outer membrane of E. coli. Since the influence of DsbA on the folding yield of the inhibitor is reduced in the presence of oxidized glutathione, the in vivo func- tion of DsbA appears to be dependent on the ratio between oxidizing and reducing thiol equivalents in the periplasm. The high stability of thiol reagents against air oxidation during growth of E. coli allows the investigation of oxidative protein folding in vivo under con- trolled, thiol-dependent redox conditions. by trypsin inhibition Trypsin was incubated with in- creasing amounts of the of E. coli JM83pFtBI or mM mM TridHCl, Triton to centration 405 nm. dissociation constant trypsin/RBI complex is below test: concentration of was calculated by linear extrapolation of to 100% inhibition assuming stoichiometry experiments identical of E. JM83 trypsin was not inhibited by other soluble E. coli proteins. by Coomassie-stained


The formation of disulfide bonds in
Escherichia coli is catalyzed by periplasmic protein disulfide-isomerase (DsbA). When the a-amylaseltrypsin inhibitor from Ragi, a protein containing five intramolecular disulfide bridges, is secreted into the periplasm of E. coli, large amounts of misfolded inhibitor with incomplete or incorrect disulfides are accumulated. Folding of the inhibitor in the periplasm is not improved when DsbA is coexpressed and cosecreted. However, an up to 14-fold increase in correctly folded inhibitor is observed by coexpression of DsbA in conjugation with the addition of reduced glutathione to the growth medium. This peptide acts as a disulfide-shuffling reagent and can pass the outer membrane of E. coli. Since the influence of DsbA on the folding yield of the inhibitor is reduced in the presence of oxidized glutathione, the in vivo function of DsbA appears to be dependent on the ratio between oxidizing and reducing thiol equivalents in the periplasm. The high stability of thiol reagents against air oxidation during growth of E. coli allows the investigation of oxidative protein folding in vivo under controlled, thiol-dependent redox conditions. Disulfide bonds are a typical feature of secretory proteins and are considered to contribute significantly to their overall stability (Goldenberg, 1993;Matsumura et al., 1989). The formation of a disulfide bridge is a posttranslational protein modification and involves a redox reaction that is catalyzed by protein disulfide-isomerase (PDI)' i n vivo Noiva and Lennarz, 1992). Disulfide formation in eukaryotic cells takes place in the lumen of the endoplasmic reticulum (ER). In this cellular compartment, the ratio between oxidized and reduced glutathione (GSSG/GSH) determines the thiol-dependent redox conditions. Since the GSSG/GSH ratio in the ER is about 100 times higher than in the cytoplasm, GSSG is likely to Grants G1159/1-1 and G11591-2 and the Fonds der Chemischen Indus-* This work was supported by Deutsche Forschungsgemeinschaft trie (grant to M. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. provide the oxidizing equivalents for the formation of protein disulfides in the ER (Hwang et al., 1992).
Numerous in vitro experiments have shown that the oxidation of a folding polypeptide chain in redox buffers containing GSH and GSSG occurs spontaneously via disulfide exchange reactions between GSWGSSG and the folding polypeptide (Saxena and Wetlaufer, 1970;Jaenicke and Rudolph, 1989). However, although oxidative protein folding does not necessarily have to be catalyzed by PDI, the enzyme has turned out to be essential for effective disulfide formation in the ER (Bulleid and Freedman, 1986). Eukaryotic PDI is a well characterized, multifunctional disulfide oxidoreductase and constitutes one of the most abundant proteins in the ER lumen. It is a strong oxidant with an intrinsic redox potential of -0.11 V Hawkins et al., 1991).
In contrast to eukaryotic cells, little is known about the process of disulfide formation and its necessity in bacteria. In Escherichia coli, disulfide formation takes place in the periplasmic space. Recently, the dsbA gene product was identified as the first protein required for effective disulfide bond formation in E. coli. The DsbA protein consists of 189 residues and contains an active disulfide (Cys-Pro-His-Cys) very similar to the catalytic disulfides present in eukaryotic PDI and other disulfide oxidoreductases like thioredoxins and glutaredoxins (Bardwell et al., 1991;Kamitani et al., 1992). This monomeric protein is a strong oxidant similar to eukaryotic PDI with an intrinsic redox potential of -0.089 V Zapun et al., 1993;.
Unlike in the ER lumen, the molecular species providing the oxidizing equivalents for disulfide formation in the periplasm of E. coli is unknown. However, it seems likely that different dithiols and molecular oxygen may be the real oxidants of cysteines in the periplasm, since the outer membrane of E. coli is permeable for molecules smaller than about 500 Da (Payne and Gilvarg, 1968;Decad and Nikaido, 1976). Consequently, the oxidant responsible for disulfide bond formation in E. coli may essentially depend on the growth conditions and the composition of the surrounding medium.
The permeability of the outer E. coli membrane for small peptides prompted us to use the periplasm of E. coli as "reaction vessel" and to mimic the situation in the ER by adding GSH and GSSG to the growth medium. To analyze the influence of DsbA on disulfide formation a t different concentrations of GSH and GSSG, DsbA was overexpressed in conjugation with the addition of GSWGSSG to the medium.
As a measure for the efficiency of disulfide formation, we analyzed the heterologous periplasmic expression of the bifunctional a-amylasdtrypsin-inhibitor (RBI) from Ragi (Eleusine coracana Gaertneri), since the yield of functionally expressed RBI in E. coli is strongly hampered by insufficient disulfide formation. RBI belongs to the family of plant a-amylasdtrypsin inhibitors (Laskowski, 1986) and contains 122 residues with five intramolecular disulfides that are essential for its activity (Shivaraj and Pattabiraman, 1981;Campos and Richardson, 1983). In this report, we show that the addition of reduced glutathione in combination with coexpression of DsbA leads to an up to 14-fold increase in the yield of native RBI, whereas the addition of oxidized glutathione and the overproduction of DsbA alone have no influence on the yield of correctly folded inhibitor in the periplasm. The results point out the importance of the reduction of incorrect disulfides during oxidative protein 24547 Oxidative Protein Folding in Vivo folding in vivo and are discussed in terms of a general strategy RESULTS AND DISCUSSION to improve the yield of recombinant secretory proteins ex-F~~ the functional expression ofthe ~~i a-amylase/trypsin pressed in E. coli.
inhibitor (RBI) in E. coli, a synthetic gene coding for the known peptide sequence (Campos and Richardson, 1983) was fused to the bacterial OmpA signal sequence (Inouye et al., 1982). This and oxidized glutathione (GSH and GSSG) were from Sigma. Benzoyl-

Materials-5,5'-dithiobis(2-nitrobenzoic acid) ( D m )
, and reduced allows the transport of the inhibitor into the periplasmic space L-arginine-p-nitroanilide was purchased from Boehringer (Mannheim, where disulfide bridges are formed. The OmpAmBI fusion was Laboratories (Detroit. MI). All other chemicals including trypsin were 1991). where it is controlled by the lac promoter/operator. Cells EXPERIMENTAL PROCEDURES Germany). Bacto-tryptone and B a h -y e a s t extract were from Difco cloned into the expression plasmid p a K 4 0 (Skerra et al.9 from Merck (Darmstadt) and of the highest purity available. Oligonucleotides were synthesized with an Applied Biosystems model 380A synthesizer using the phosphoramidite method. The plasmid pASK40 was a generous gift of Dr. A. Skerra (Max-Planck-Institut fur Biophysik, Frankfurt, Germany). General Methods-Molecular cloning techniques were based on the procedures described by Sambrook et al. (1989). SDS-PAGE was performed according to Fling and Gregerson (1986) using 15% (vh) acrylamide gels. For immunoblotting, the proteins were transferred onto nitrocellulose membranes after complete SDS-PAGE. RBI was specifically detected using anti-RBI rabbit antibodies and anti-rabbit swine antibodies conjugated with alkaline phosphatase according to Blake et al. (1984).
Construction of Expression Plasmids-The plasmid pRBI for functional periplasmic expression of RBI was constructed by cloning a synthetic gene coding for the OmpAlRBI fusion (Campos and Richardson, 1983;Inouye et al., 1982) into the plasmid pASK40 (Skerra et al., 1991) via the X6aI and HindIII restriction sites. The ds6A gene, including the natural ribosomal binding site and the natural signal sequence, was amplified from the genome of the E. coli K12 wild type strain W3110 (Bachmann, 1972) by the polymerase chain reaction using oligonucleotide primers based on the published ds6A sequence (Bardwell et al., 1991) as described . The amplified ds6A gene was cloned into pRBI directly at the 3' end of the RBI gene via the HindIII and BamHI sites. 1985) harboring the plasmids pRBI and pRBI-PDI were grown in LB medium containing ampicillin (100 pg/ml) a t 26 "C to a n optical density at 550 nm of 1.0 and were induced with IPTG (final concentration: 1 mM). At the time of induction, different amounts of GSWGSSG were added (see Table I). The cultures were shaken for 16 h, and the cells were harvested by centrifugation and suspended in lysis buffer (100 mM TridHC1, pH 7.5,20 mM EDTA) to an identical optical density. The cells were disrupted in a French pressure cell, and the lysate was centrifuged (48,000 x g, 30 min, 4 "C). The supernatant (soluble fraction) was removed, and the insoluble fraction was suspended in the original volume.
Stability of Reduced Glutathione-The stability of GSH against air oxidation in the growth medium was analyzed by determination of free thiol groups according to Ellman (1959). Assays were camed out at DTNB concentrations of 0.3 mM in 80 mM sodium phosphate, pH 8.0, 1 mM EDTA, 2% (wh) SDS. ARer incubation for 15 min at room temperature, the absorbance at 412 nm was recorded (e4lz " , , , = 13,600 M-' cm-Vfree thiol) (Ellman, 1959).
Determination of Inhibitor Concentration-The concentration of functional RBI in the soluble fractions of crude extracts was determined by trypsin inhibition assays. Trypsin (0.2 PM) was incubated with increasing amounts of the extracts of E. coli JM83pFtBI or JM83pRBI-PDI for 30 min in 100 m M NaCI, 50 mM TridHCl, pH 8.0,l mM CaCIz, 0.005% (wh) Triton X-100 at 25 "C. The activity of free trypsin was determined by adding benzoyl-barginine-p-nitroanilide to a final concentration of 0.2 mM and recording the increase in absorbance a t 405 nm. Since the dissociation constant of the trypsin/RBI complex is more than 2 orders of magnitude below than concentration of trypsin in the test: the concentration of RBI was calculated by linear extrapolation of the titration curves to 100% inhibition assuming a 1:l stoichiometry of the complex (Shivarqi and Pattabiraman, 1981;Bieth, 1974). Control experiments with identical extracts of E. coli JM83 proved that trypsin was not inhibited by other soluble E. coli proteins.
The amount of RBI present in the insoluble fractions was evaluated by densitomeric scans of Coomassie-stained SDS gels, where known amounts of purified RBI were applied simultaneously as a standard for the quantitative analysis. Gel scans were performed a t 546 nm using a Hirschmann Elscript 400 densitometer.
M. Wunderlich and R. Glockshuber, unpublished results. of E. coli JM83 harboring the resulting expression plasmid (pRBI) were induced with IPTG, and the expression of RBI was analyzed by SDS-PAGE. Densitometric scans of Coomassiestained gels revealed that about 95% of the expressed and correctly processed protein were accumulated in insoluble periplasmic aggregates, while only about 5% of processed RBI were obtained as soluble, native inhibitor. Immunoblot analysis using reducing and nonreducing SDS-PAGE and antibodies that preferably recognize the correct disulfide-bridged conformation of the inhibitor revealed that native RBI is only present in the soluble fraction (Fig. 1A ). When the aggregated material was solubilized by reducing agents and denaturanta, native RBI could be renatured in redox buffers containing GSH and GSSG (Jaenicke and Rudolph, 1989). Therefore, as a result of the complex cystine pattern of RBI (Fig. 1B 1, it is likely that the formation of native disulfides is rate-limiting for folding of RBI in vivo and competes with the aggregation of nonnative molecules (Kiefhaber et al., 1991).
Coexpression and cosecretion of DsbA in at least equimolar amounts using a dicistronic operon (Fig. 2, A and B ) did not increase the yield of functional RBI as determined by trypsin inhibition tests (Table I and Fig. 3A). Since the function of eukaryotic PDI is known to be dependent on the thiol-dependent redox potential (Lyles and Gilbert, 1991 The position of all cysteine residues in RBI is marked by boldface numbers.   lanes 1 and 2 ) and E.

Influence of GSH and GSSG on the functional expresawn of RBI and
coli JM831pRBI -PDI (lanes 3 and 4 ) grown in 0 and 5 mM GSH, respectively, were applied onto a 15% SDS-PAGE that was silver-stained. A molecular mass standard ( S t ) and the purified inhibitor from Ragi ( C ) were also applied. C, effect of GSH and coexpression of DsbA on the aggregation of processed RBI. Identical volumes of the insoluble parts of extracts of E. coli JM83/pRBI (lanes 1 and 2 ) and E. coli JM83/pRBI-PDI (lanes 3 and 4 ) grown in the absence (lanes 1 and 3) and in the presence of 5 mM GSH (lanes 2 and 4 ) were applied onto a 15% SDS-PAGE that was stained with Coomassie Blue. As a control, insolublefractions of induced E. coli JM83 cells (harboring no plasmid) and uninduced E. coli JM83/pRBI were applied. St, molecular mass standard; C, purified RBI from Ragi. GSSG were added to the culture medium assuming identical redox conditions in the medium and in the periplasm. Based on the observation that oxidative folding of RBI in vitro was strongly dependent on the GSWGSSG ratio (data not shown), we expected a significant effect of the added thiol reagents on the folding of RBI in vivo. The yield of native RBI in the periplasm could indeed be increased by adding GSH and GSSG in varying concentrations at the time of induction without coexpression of DsbA. A more than 5-fold increase in functional RBI was found at a GSH concentration of 5 mM in the medium (Table I). In contrast to coexpression of DsbA in the absence of GSH, coexpression of DsbA in the presence of 5 mM GSH further increased the yield of native inhibitor almost 3-fold leading to an overall, 14-fold improvement of periplasmic folding of RBI (Table I and Fig. 2 B ) . However, the analysis of the insoluble cellular protein by Coomassie-stained SDS gels revealed that the amount of insoluble RBI is practically independent of DsbA coexpression and presence of thiols in the medium (Fig. 2C). This demonstrates that improved folding of RBI in the periplasm mainly prevents the inhibitor from being proteolytically degraded, but does not decrease the overall amount of aggregated RBI significantly. Therefore, DsbA apparently does not act as an 'aggregation inhibitor" in a chaperone-like way (Buchner et al., 1990).
The catalytic activity of coexpressed DsbA was strongly in- Ffc. 3. Determination of the yield of functionally expressed RBI and the stability of CSH in the growth medium. A, trypsin inhibition assays were used to determine the concentration of native RBI in the soluble parts of extracts of E. coli JM83/pRBI (0) and E. coli JM83/pRBI-PDI ( 0 ) grown in the absence of GSH and E. coli JM83/ pRBI ( 0 ) and E. coli JM83/pRBI-PDI ( + ) grown in the presence of GSH ( 5 mM). All extracts were from cell suspensions with identical optical densities. B, stability of free thiol groups in the culture medium during growth of E. coli JM83IpRBI after the addition of GSH to a final concentration of 1 (0). 5 ( + ), and 10 mM (m).
fluenced by the GSWGSSG ratio. DsbA activity was reduced when the relative amount of GSSG was increased and was tightly correlated with the GSWGSSG-dependent yields of RBI without coexpression of DsbA (Table I).
The concentration of GSH added to the medium was found to be constant during the growth of the bacteria (Fig. 38). We conclude that the oxidation of GSH by air in the culture medium is kinetically hindered even under aerobic growth of E. coli and that a small molar excess of GSSG or other oxidizing molecules may be sufficient for disulfide formation. In a control experiment, misfolded periplasmic aggregates of RBI could not be rescued when induced cells ofE. coli JM83IpRBI were grown in the absence of disulfide shuffling reagents and were suspended in LB medium containing GSWGSSG after harvest. Therefore, GSH and GSSG added to the medium at the time of induction are directly involved in the oxidative folding of RBI in vivo.
The fact that the yield of native RBI in the periplasm of E.
coli is diminished by GSSG and increased with GSH up to concentrations of 10 mM with and without coexpression of DsbA suggests that the reduction of incorrect disulfides limits the rate and yield of folding of RBI in vivo, Due to the high cysteine content of RBI (10 out of 122 residues) it appears likely that folding intermediates with nonnative disulfides are significantly populated during oxidative folding of the inhibitor.
From the known equilibrium constant between DsbA and glutathione at pH 7 (1.2 . M; Wunderlich and Glockshuber, 19931, it becomes clear that less than 1% of all DsbA molecules were oxidized at the conditions providing maximal yields of functional inhibitor (5 mM GSH and 5 mM GSWl m~ GSSG). Therefore, the almost fully reduced state of DsbA at optimal conditions for folding of RBI further supports the view that DsbAis mainly involved in the breakage of nonnative disulfides during folding of RBI. Similar results were also reported for eukaryotic PDI during the oxidative folding of RNase A in vitro (Lyles and Gilbert, 1991). Thus, it seems likely that misfolded proteins with nonnative disulfides occur during protein folding in uiuo. However, the almost fully reduced state of DsbA at optimal conditions for folding of RBI does not necessarily exclude a significant contribution of DsbA in the formation of disulfides, since the oxidation of a folding polypeptide by DsbA is in the order of lo3 to lo4 times faster than its reduction . Therefore, high relative amounts of reduced DsbA may indeed be required to guarantee DsbA-catalyzed reduction of nonnative disulfides.
It appears likely that the optimal redox conditions for folding of other secretory proteins in the periplasm of E. co2i may be different from the outlined conditions for maximal folding yield of RBI. However, the best growth conditions for optimal yields in the periplasm (in conjugation with coexpression of DsbA) can easily be evaluated by varying the concentrations of GSH and GSSG in the medium.
Unlike oxidative folding in the ER (Bulleid and Freedman, 1988;Braakman et al., 1992), folding in the periplasm of E. coli is probably not assisted by ATP-dependent chaperones like BiP (Haas, 1991) due to the lack of an ATP-specific transport system and the presence of alkaline phosphatase that hydrolyzes ATP (Heppel et al., 1962). The absence of ATP-dependent chaperones that are known to prevent aggregation (Buchner et al., 1991) (for a review, see Gething and Sambrook (1992)) is consistent with the observation that periplasmic aggregates of RBI have no effect on the bacterial growth. Since DsbA and peptidylprolyl cis,trans-isomerase (Liu and Walsh, 1990) are the only known periplasmic components of E. coli promoting protein folding and may act synergistically (Schonbrunner and Schmid, 1992), oxidative folding in the bacterial periplasm appears to be less complex than in the ER lumen. Recently, another E. coli protein (termed DsbB) was found to be involved in disulfide formation in E. coli. It was suggested that DsbB, which is a component of the cytoplasmic membrane, may be required to transfer oxidizing equivalents from the cytoplasm to DsbA that in turn oxidizes folding polypeptide chains (Bardwell et al., 1993). However, our results indicate that folding of RBI and probably also the redox state of DsbA is determined by the thiols and disulfides present in the medium.
Thus, DsbB may become essential for disulfide formation in E. coli, when oxidants are rare in the medium (e.g. under anaerobic conditions), but may play a secondary role when oxidizing equivalents are abundant. This is consistent with the finding that the lack of DsbB (in dsb-strains) can be complemented by adding GSSG or cystine to the growth medium (Bardwell et al., 1993) and with our finding that the thiol-dependent redox conditions in the periplasm of E. coli can indeed be controlled by thiol reagents in the medium. We have found that other thiol reagents such as N-acetylcysteine exhibit improvements of the yield of functional RBI very similar to GSH when used a t concentrations of 5 m~ in the medium. Since N-acetylcysteine is less expensive than GSH, it may be more convenient for technological applications.
In conclusion, we believe that the possibility to control the thiol-dependent periplasmic redox potential and the demonstration of the thiol-dependent activity of DsbA in vivo may be useful for the investigation of oxidative protein folding in bacteria and may greatly facilitate the heterologous expression of secretory proteins in E. coli.