Human Protein Disulfide Isomerase Functionally Complements a dsbA Mutation and Enhances the Yield of Pectate Lyase C in Escherichia coli

Human PDI was expressed to the Escherichia coli periplasm, by using a plasmid encoded ompA-PDI fusion under the control of the trp promoter. Periplasmic ex- tracts were shown to contain active PDI using the scrambled ribonuclease assay. PDI activity was also demonstrated by complementation of two phenotypes associated with a dsbA mutation. Alkaline phosphatase activity, which is reduced in dsbA cells, was restored to wild type levels by PDI. PelC, a pectate lyase from Er- winia carotovora , was shown to be DsbA dependent in E. coli . PDI was able to restore its activity to that seen in wild type cells. Increased expression of PDI was found to increase the yield of active PelC above that seen in wild type cells. PDI also enhanced the yield of PelC in DsbA (cid:50) cells but only in the presence of exogenous oxidized glutathione. PDI is thus able to functionally sub- stitute for DsbA in the folding of disulfide-bonded proteins in the bacterial periplasm and to enhance the yield of highly expressed protein when the ability of the E. coli periplasm to fold protein may be saturated. How- ever, our results suggest that the activities of DsbA and PDI in vivo may be different. Protein folding in assisted by two general classes of proteins. Chaperones, such as GroEL and BiP, are thought to interact noncovalently with substrates preventing or reversing interactions that can to incorrect folding and aggregation. Folding

Protein folding in vivo is assisted by two general classes of proteins. Chaperones, such as GroEL and BiP, are thought to interact noncovalently with substrates preventing or reversing interactions that can lead to incorrect folding and aggregation. Folding enzymes, such as PDI 1 and peptidyl-prolyl-cis-transisomerases, catalyze rate-limiting steps in the folding of proteins (for a general review see Ref. 1). The effects of chaperones and folding enzymes on protein folding have been extensively studied in vitro (for reviews see Refs. 2 and 3). However, studying assisted folding of a protein in vivo is more difficult, partly because of its subcellular location. For example, the major site for folding, disulfide bonding, and modification of secretory polypeptides in eukaryotes is the endoplasmic reticulum lumen, the composition of which is not easily manipulable.
PDI is one of the major proteins of the endoplasmic reticulum lumen. It was first shown to oxidatively refold reduced RNase (4) and was subsequently shown to be the ␤-subunit of prolyl-4-hydroxylase (5). It is thought to be the main agent for forma-tion and isomerization of disulfide bonds in proteins folding in the endoplasmic reticulum lumen (for reviews see Refs. 6 and 7). However, this cellular function has largely been demonstrated by indirect methods; PDI has been shown to be physically associated with folding proteins in vivo (8,9), has restored defective co-translational disulfide bond formation in PDI-deficient microsomes (10), and has been shown to assist the folding of proteins in vitro (11)(12)(13). Here, we have studied PDI-mediated protein folding by expressing human PDI to the periplasm of Escherichia coli and assaying its effects on model proteins. PDI is amenable to study in the E. coli periplasm because it is highly soluble, is not glycosylated, and does not require ATP for its function.
The periplasm is an oxidizing compartment and is the site of disulfide bond formation in E. coli proteins. A number of proteins have been identified in E. coli that are involved in disulfide bond formation, including DsbA, DsbB, and DsbC (14 -19). These proteins possess the motif, NH 2 -CXXC-COOH, containing cysteines that are important for catalytic activity. PDI also possesses this motif. Purified DsbA acts as a disulfide oxidoreductase and also acts weakly as a disulfide isomerase (20,21). The relative importance of the two activities in vivo is not known. Strains lacking DsbA show a variety of defects in disulfide bond formation (see Ref. 22 for a review). We demonstrate here that human PDI can complement two dsbA-dependent phenotypes, confirming the idea that PDI and DsbA are functionally similar. We also show that when one particular disulfide-bonded protein (PelC) is overexpressed, the folding capacity of wild type E. coli cells can be saturated. The coexpression of PDI overcame this deficiency resulting in increased levels of active PelC. Increased levels of PelC could be produced from dsbA cells expressing PDI, but only when oxidized glutathione (GSSG) was added to the growth media. This suggests that although DsbA and PDI are functionally similar in complementation studies, their activities in vivo have some significant differences.

EXPERIMENTAL PROCEDURES
Materials-Polygalacturonic acid, indoleacrylic acid (IAA), dithiothreitol, and RNA (Torula yeast) were purchased from Sigma; paranitrophenyl phosphate was purchased from BDH Ltd. Vent polymerase was from New England Biolabs, and Sequenase was from U. S. Biochemical Corp. sRNase and anti-PDI antibodies were generous gifts of Prof. R. Freedman (Table I). The E. coli strain TG2 was used for all DNA manipulations.
DNA Manipulations-Standard methods were used (23). Growth Conditions-Cultures grown in Luria broth supplemented with antibiotics were shaken (200 r.p.m.) at 37°C. Expression of PDI from pDPH5 was either uninduced or induced by including IAA at concentrations from 1 to 120 g ml Ϫ1 . Expression of PelC from pDPH9 was induced by inclusion of IPTG at either 0.2 or 0.4 mM in cultures grown at 33°C.
Construction of PDI Expression Plasmids pDPH5 and pDPH13-DNA coding for the mature PDI peptide was generated as a HindIII-EcoRI polymerase chain reaction fragment using pUKC151 as a template and the primers 5Ј-GCGCAAGCTTACGCCCCCGAGGAG-3Ј and 5Ј-CGGAATTCTTACAGTTCATCTTTC-3Ј. Vent polymerase was used for the polymerase chain reaction. After cleavage with HindIII and EcoRI, this was ligated into pSKOMP in frame with the OmpA signal peptide codons. The HindIII site was mutagenized by the double primer method (24) to restore the correct post signal peptide cleavage codon using the primers 5Ј-CGTAGCGCAAGCGGACGCCCCCGA-3Ј and 5Ј-GTAAAACGACGGCCAGT-3Ј. Successful mutagenesis was confirmed by DNA sequence using Sequenase (data not shown). The OmpA-PDI fusion was then ligated into pUC118 as an XbaI fragment. To make pDPH5, the fusion was cut out as a BamHI fragment and ligated into the BclI site of the expression plasmid pCT54, allowing expression from the trp promoter. This also produced the control plasmid pDPH6, which had the insert in the reverse orientation. To make pDPH13, the fusion was ligated as a SacI-HindIII fragment into pSU18, thereby allowing expression of PDI from a chloramphenicol-resistant plasmid under the control of the lac promoter.
Construction of PelC Expression Plasmid pDPH9-Plasmid pJS6197 provided the coding region of PelC. The gene was ligated into pSU18 as a SacI/HindIII fragment allowing expression of PelC from the lac promoter.
Preparation of Cell Extracts-Periplasmic extracts were made by the cold osmotic shock method (28) or the lysozyme/EDTA method (29) and stored on ice before use. For the lysozyme/EDTA method, cultures were concentrated 5 or 10 ϫ by resuspending cell pellets in 1 ⁄5 or 1 ⁄10 the original culture volume of spheroplast buffer.
Protein Analysis-Protein concentrations were measured using the Bradford method using bovine serum albumin as a standard.
Assay of Alkaline Phosphatase Activity-The assay was based on methods described earlier (30). For the in vivo assay, 100 l of culture was added to 900 l of 1.5 M Tris, pH 8.0, at 30°C, containing 25 l of 0.1% SDS and 50 l of chloroform. The samples were vortex mixed for 5 s and then incubated at 30°C for 5 min. The assay was started by adding 200 l of 15 mM para-nitrophenyl phosphate and stopped after 10 min by adding 200 l of 1 M KH 2 PO 4 . After pelleting the cells, the A 420 of the supernatant was read against a blank. One unit of alkaline phosphatase activity is defined as ⌬A 420 of 1.0 min Ϫ1 A 600 Ϫ1 . Assay of PelC Activity-PelC was assayed essentially as described earlier (31) on periplasmic extracts prepared by the lysozyme/EDTA method. Small volumes (2-20 l) of 10 ϫ periplasmic extract were mixed with 1 ml of PelC assay media, and the change in A 235 measured at room temperature using a Unicam SP1800 spectrophotometer and chart recorder. One unit of PelC activity is defined as a ⌬A 235 of 1.0 min Ϫ1 mg Ϫ1 of periplasmic extract.
Assay of PDI Activity-PDI expressed to the periplasm was assayed at room temperature by the scrambled RNase method (32). 5ϫ periplasmic extracts made by the lysozyme/EDTA method were extensively dialyzed against TKM buffer (50 mM Tris, 25 mM KCl, 5 mM MgCl 2 , pH 7.5) at 4°C. Volumes of between 5 and 80 l were made up to 160 l with TKM. 20 l of 0.1 mM dithiothreitol was added, and the tube stood at room temperature for 2 min. 20 l of sRNase (at a concentration of 20 K m ) was added to start the reaction. 3.3-l samples were removed after 0.5, 3, 6, 9, and 12 min and mixed with 1 ml of 82 g ml Ϫ1 RNA in TKM. The rate of change of A 260 was measured for 30 s using a Hewlett Packard 8452A diode array spectrophotometer using HP 89532K multicell kinetics software. One unit of activity is defined as ⌬A 260 min Ϫ1 min Ϫ1 mg Ϫ1 periplasmic extract.

Expression of Human PDI in E. coli
Plasmid pDPH5 was shown to express PDI to the E. coli periplasm. Cells containing pDPH5 grown in the presence of the inducer IAA expressed a new protein in periplasmic fractions produced by osmotic shock with an apparent molecular mass of 57.4 kDa. Western blot analysis with a polyclonal rabbit anti-bovine PDI antibody showed that PDI was produced in the DsbA Ϯ strains used for further experimentation (Fig. 1). When pDPH5 was uninduced ( Fig. 1, lanes 1 and 5), PDI was still detectable in the periplasm. PDI is apparently more highly expressed in the dsbA strain JCB571 than dsbA ϩ JCB570 (  4). This may reflect a greater stability of PDI in DsbA Ϫ cells due to differences in the redox environment of the periplasm. Laser scanning densitometry of polyacrylamide gel electrophoresis Blue83 stained SDSpolyacrylamide gels showed that PDI constituted up to ϳ35% of periplasmic proteins greater than 45 kDa in the E. coli strain TG2. Accumulation of PDI was proportional to the concentration of IAA, and the protein was always in the soluble fraction (data not shown).

Assay of PDI Activity from E. coli Periplasmic Extracts
Crude periplasmic extracts were assayed by the sRNase method. Fig. 2 shows that large amounts of PDI activity could be measured in such extracts. This activity was proportional to the concentration of IAA and therefore to the final concentration of PDI. Periplasmic extracts from cells bearing pCT54 gave no measurable activity, showing that the activity was due to PDI alone.

Complementation of dsbA Phenotypes by PDI
Alkaline Phosphatase-This protein is known to be DsbA dependent in its folding (14,15) and was used as a model protein. We assayed the alkaline phosphatase activity from overnight cultures of the phoR strains JCB570 and JCB571, which express alkaline phosphatase constitutively. Fig. 3 shows that dsbA cells are deficient in alkaline phosphatase activity measured relative to the wild type (compare columns 1 and 2). Inclusion of the PDI expression plasmid, pDPH5, under noninduced conditions gave nearly full restoration of alkaline phosphatase activity. The control plasmid pDPH6 had no effect on alkaline phosphatase activity. This shows clearly that PDI can functionally substitute for DsbA. Further investigation showed that the level of alkaline phosphatase activity in the DsbA ϩ cells (JCB570) was not enhanced by overexpression of PDI from either pDPH5 or pDPH13 (data not shown).
PelC as a Model Folding Protein in Vivo-We wished to study the expression of an assayable protein that was not native to E. coli but that was likely to require the action of disulfide bond-forming proteins to reach its active form in vivo. PelC, a pectate lyase from Erwinia carotovora was found to be an ideal model protein. It is a secreted enzyme known to have two disulfide bonds (33). Erwinia has Dsb proteins (19), so it seemed possible that PelC folding would be DsbA-dependent in E. coli. PelC is assayable by a plate and turbidometric assay and can be expressed in E. coli, where it is retained in the periplasm (27). Use of a plasmid-encoded model protein would also facilitate investigation of overexpression phenomena.
Induction of PelC expression from pDPH9 with 0.2 mM IPTG resulted in a protein with a molecular mass of 39.8 kDa present in the periplasmic fraction, in agreement with the predicted mass (27). PelC activity was shown to be dsbA-dependent in E. coli by assaying crude periplasmic extracts in vitro by following the degradation of polygalacturonic acid spectrophotometrically. The results are shown in Fig. 4. There was a significant reduction in the PelC activity recovered from dsbA cells, as shown by comparing columns 1 and 2. When PDI was supplied by including the compatible plasmid pDPH5, the activity of PelC recovered was restored to near wild type levels.
This demonstrates clearly that PelC activity, and therefore presumably PelC folding, is dsbA-dependent in vivo in E. coli. PDI is able to act on PelC protein and restore its activity to near wild type levels in a dsbA background. pDPH5 was uninduced for PDI expression in this experiment. Under these conditions, there is only a very small amount of PDI in the periplasm, which is only easily visualized by immunoblotting of such periplasmic extracts separated on SDS-polyacrylamide gel electrophoresis gels (data not shown). However, the level of PelC activity recovered from the JCB571 ϩ pDPH9 ϩ pDPH5 cells appeared to be slightly greater than that from wild type cells. This led us to investigate whether increased PDI expression could give enhanced PelC yield.

Overexpression of PelC in the Presence of PDI
PelC expression was induced with 0.4 mM IPTG in cultures with increasing concentrations of IAA to give increased PDI expression. Fig. 5 shows that increasing co-expression of PDI gave increased recovery of active PelC relative to wild type cells. The activity recovered increased proportionally with IAA concentration. The increase was not due to the effect of IAA or pCT54 (expression vector) on the cells (Fig. 5, compare columns  7 and 8). The ability of E. coli to oxidatively fold PelC under these conditions appears to have been limited. This limitation can be overcome by additional PDI. We wondered if this overproduction of PelC could also be achieved in the dsbA cells JCB571. Fig. 6 shows that even with induction of PDI expression with 80 g ml Ϫ1 IAA, there is no increased recovery of PelC activity above that seen in wild type cells. This suggested that PDI on its own in the periplasm is unable to increase the yield of active PelC in this strain.
We tested if this limitation in DsbA Ϫ cells could be reversed by supplying GSSG in the media. It is thought that the relatively oxidizing environment of the endoplasmic reticulum lumen is maintained by a mixed glutathione buffer (34), and it has been shown that use of glutathione buffers in bacterial media affects the redox state of the periplasm (35-38). Fig. 7   FIG. 1. Western blot demonstration of the presence of PDI in E. coli periplasmic extracts. Periplasmic extracts prepared by osmotic shock were separated by 12.5% reducing SDS-polyacrylamide gel electrophoresis and then electroblotted to nitrocellulose. Lanes 1-4, periplasmic extracts from JCB570 containing pDPH5 (lanes 1-3) or pCT54 (lane 4). Lanes 5-8, periplasmic extracts from JCB571 containing pDPH5 (lanes 5-7) or pCT54 (lane 8). Cultures were grown overnight with 0 g ml Ϫ1 IAA (lanes 1 and 5), 20 g ml Ϫ1 IAA (lanes 2 and 6), and 80 gml Ϫ1 IAA (lanes 3, 4, 7, and 8). The antibody used was a rabbit polyclonal antibody raised against purified bovine PDI. Each well was loaded for an equal amount of total periplasmic protein. shows that in the presence of GSSG, PDI can enhance the yield of PelC above that seen in wild type cells even in a dsbA Ϫ background. The yield of PelC seen with 5 mM GSSG was approximately half that seen with an equivalent level of PDI induction in wild type cells as seen in Fig. 5 but is still approximately twice the yield seen in wild type cells without PDI. DISCUSSION In this paper we present direct evidence for the ability of human PDI to assist the oxidative folding of proteins in vivo in E. coli. We show for two different proteins that PDI can functionally substitute for DsbA. We also show that high levels of co-expression of PDI result in increased yields of active PelC above that achieved by wild type cells. This overproduction is itself DsbA-dependent, but the provision of GSSG in media overcame this dependence.
The rate of folding of alkaline phosphatase in dsbA strains is lower than that of wild type cells (14), resulting in lower overall alkaline phosphatase activity in dsbA cells. Co-expression of low levels of PDI restores the measured alkaline phosphatase activity to near wild type levels, suggesting that PDI is influencing the rate of folding of alkaline phosphatase. We used PelC as a model folding protein and by assaying crude periplasmic extracts in vitro showed that the amount of active PelC recovered from dsbA Ϫ cells is significantly reduced compared with that recovered from dsbA ϩ cells. We interpret this as demonstrating that the folding of PelC is DsbA-dependent in vivo. Because E. carotovora is known to contain Dsb proteins similar to those in E. coli, we would expect PelC folding to be DsbA-dependent in its native host also, as suggested previously (19). Co-expression of low levels of PDI from a plasmid compatible to the one carrying the PelC gene re- sulted in a restoration of the recovered PelC activity to near wild type levels, demonstrating again that PDI can functionally substitute for DsbA.
While doing PelC assay experiments on the phoR Ϫ strains JCB570 and JCB571, we observed that the total protein concentration of periplasmic extracts from dsbA Ϫ JCB571 was consistently about 30% lower than that from dsbA ϩ JCB570. Low levels of PDI expression restored protein concentrations in dsbA Ϫ cells to normal levels. 2 PhoR Ϫ strains constitutively express several periplasmic proteins (for example, alkaline phosphatase, phosphate-binding protein, and glycerol 3 phosphate-binding protein) (for reviews see Refs. 39 and 40). The lower rate of folding in the periplasm of dsbA cells probably results in a greater proteolytic susceptibility for these proteins, similar to that seen with alkaline phosphatase, ␤-lactamase, OmpA, and BPTI (14,36), and hence a lower final protein concentration. This suggests that PDI can act on several different E. coli proteins.
In the complementation experiments PDI was produced from uninduced pDPH5, so the level of expression of PDI is low; rich media are known to repress the trp promoter (41). This demonstrates that only small amounts of PDI are necessary to complement dsbA phenotypes, suggesting that PDI is acting catalytically and not just as an extra source of a periplasmic redox buffer. The fact that PDI appears to act on many periplasmic proteins suggests that PDI has a relaxed peptide binding specificity in vivo, which is consistent with experiments done in vitro (42,43).
We tested if increased production of PDI would result in increased yields of PelC. The results in Fig. 5 show that indeed higher levels of induction of PDI gave greatly increased yield of PelC activity above that seen in wild type cells. The yield of PelC activity increased with increasing levels of PDI expression, although it reached a plateau in the range of IAA concentrations tested. This suggests that the yield of overexpressed PelC in DsbA ϩ E. coli is limited by the cellular folding machinery. This may be due to saturation of the Dsb system, resulting in increased proteolytic susceptibility for some molecules of PelC in the periplasm.
We showed that higher levels of periplasmic PDI only achieved overproduction of PelC in the presence of DsbA. 1 and 5 mM GSSG partially restored the ability of PDI to increase the yield of active PelC in a dsbA background. Others have used redox buffers supplied in culture media while trying to influence folding of periplasmic proteins directly but with mixed success (36 -39). In our experiments, the addition of 1 and 5 mM GSSG had no effect on the yield of active PelC in DsbA Ϫ cells containing no PDI. Therefore, GSSG alone lacks the ability to increase the yield of oxidatively folding proteins in the periplasm, which is consistent with previous work. However, the same concentrations of GSSG in DsbA Ϫ cells in the presence of PDI has a marked positive effect.
The precise mechanism of complementation of dsbA by PDI is not known. It could simply be a result of PDIs dithiol oxidase activity substituting directly for that of DsbA. If this is true, an unknown oxidant(s) must be responsible for regenerating active PDI. Alternatively, PDI acting as an isomerase may increase the flux through a DsbA-independent pathway, for example by isomerization of disulfide-bonded intermediates that have spontaneously formed in the incorrect conformation. Such a pathway is probably normally catalyzed by DsbC, and it will be of great interest to investigate the ability of PDI to complement dsbC mutants. The fact that overproduction of PelC is only seen in wild type cells but not in dsbA cells suggests that a DsbA-dependent step is rate-limiting, at least for PelC. This can be overcome by provision of oxidized glutathione. GSSG may act either as an oxidant, thereby improving the regeneration of oxidized PDI, or by forming a mixed adducts with PelC and PDI, which may accelerate formation of the active protein, by the mechanisms demonstrated by Darby et al. (44).
Understanding the exact actions of PDI and DsbA on alkaline phosphatase and PelC folding are complicated in vivo by the dynamic nature of the periplasm, which may have changing needs for oxidation/isomerization activities during cell growth. Co-expressed PDI was able to effect increases in the yield of PelC in the absence of any supplied redox buffers. In contrast, co-expression of DsbA only gave increased yield of native ␣-amylase/trypsin inhibitor from Ragi (RBI) in the presence of media redox buffers (36). This suggests that PDI and DsbA may have different activities in vivo. We also note that overexpression of PelC in DsbA Ϫ cells in the presence of PDI and GSSG resulted in only approximately half the yield of PelC compared with that in the DsbA ϩ cells with PDI (compare column 7 in Fig. 7 with column 5 in Fig. 5). This suggests that the enhancement by GSSG (in DsbA Ϫ cells) was not due to direct replacement of the DsbA activity missing.
The phoR cells used for these experiments constitutively translocate at least six periplasmic proteins, alkaline phosphatase, phosphate-binding protein, glycerol 3 phosphate-binding protein, PelC, PDI, and ␤-lactamase (from pDPH5), without apparent difficulty. However, overexpression of PelC may have saturated the Dsb machinery. Perhaps PDI only has a beneficial role to play in E. coli under Dsb saturated conditions?
Bacterial proteins generally have fewer disulfide bonds than secreted eukaryotic proteins (20). The ability of PDI to catalyze the folding of more complex proteins in a range of genetic backgrounds is currently being evaluated. Co-expression systems such as this provide a powerful investigative tool. Individual proteins such as PDI can be investigated in an environment closer to that in the cell than can be generated using reconstituted systems in vitro, thereby enabling investigation of other factors that affect the function of PDI, such as redox conditions and PDI-protein interactions.