Catalytic activity and chaperone function of human protein-disulfide isomerase are required for the efficient refolding of proinsulin.

Protein-disulfide isomerase (PDI) catalyzes the formation, rearrangement, and breakage of disulfide bonds and is capable of binding peptides and unfolded proteins in a chaperone-like manner. In this study we examined which of these functions are required to facilitate efficient refolding of denatured and reduced proinsulin. In our model system, PDI and also a PDI mutant having only one active site increased the rate of oxidative folding when present in catalytic amounts. PDI variants that are completely devoid of isomerase activity are not able to accelerate proinsulin folding, but can increase the yield of refolding, indicating that they act as a chaperone. Maximum refolding yields, however, are only achieved with wild-type PDI. Using genistein, an inhibitor for the peptide-binding site, the ability of PDI to prevent aggregation of folding proinsulin was significantly suppressed. The present results suggest that PDI is acting both as an isomerase and as a chaperone during folding and disulfide bond formation of proinsulin.

Protein-isulfide isomerase (PDI) 1 is a protein of the endoplasmic reticulum involved in the oxidative folding of many disulfide-bonded proteins (1Ϫ3). PDI consists of four domains arranged in the order a-b-b-a with an acidic extension at the C terminus (c domain) that contains the KDEL retention sequence. The first two domains show significant homology to thioredoxin (4,5). Although there are no structural data available for the b and a domains, it can be infered from their sequence similarities to the b and a domain, respectively, that they also contain the thioredoxin fold. Both, the a and a domains of PDI, contain the local sequence -WCGHCK-that is essential for the catalytic activity of PDI. Both isolated domains can operate as thiol:disulfide oxidoreductases, but all other PDI domains are required to assist protein refolding and formation of native disulfide bonds with maximum catalytic activity (6). No function has been assigned to the b domain so far, however, the b domain provides the principle peptidebinding site (7). This domain alone is sufficient to bind peptides of 10 -15 residues but binding of larger peptides or non-native proteins requires the contribution of either the a and b domains or the a domain (7).
In vitro, PDI can assist folding of proteins that contain no disulfide bonds, demonstrating its function as a chaperone (8,9). PDI can influence folding of proteins with multiple disulfide bonds as a chaperone and also as an isomerase (10Ϫ17). Yao et al. (11) proposed that PDI can fulfil both activities on the refolding of acidic phospholipase A 2 . When using mitochondrial malate dehydrogenase (10) or rhodanese (8) as a substrate PDI influences refolding in a chaperone-like manner. PDI and also alkylated PDI displaying no isomerase activity can chaperone refolding of D-glyceraldehyde-3-phosphate dehydrogenase indicating that the active sites are not required for the chaperone activity of PDI (9,18). The isomerase activity is involved in refolding of lysozyme (19), an antibody fragment (13), insulin (16), and proinsulin (17). Furthermore, under certain redox conditions PDI can reduce protein substrates with native disulfide bonds (13,20). In addition, PDI can catalyze disulfide bond formation and rearrangements within kinetically trapped, structured folding intermediates (21). This demonstrates that the substrate specificity of PDI is not restricted to misfolded proteins containing no or non-native "scrambled" disulfide bonds.
It is possible to discriminate between isomerase and chaperone activity of PDI by using mutants or variants displaying only one of both activities. Active site mutants or alkylated PDI have been used to generate species that act as a chaperone only (11,13). On the other hand mutants lacking chaperone activity do not exist as it is not clear which amino acid residues in the peptide-binding domain are involved and essential for substrate binding. Tsibris et al. (22) could show that isomerization of scrambled RNase and reduction of insulin are inhibited by estrogens which might be due to the similarity of PDI parts with estrogen receptor segments. Additionally, compounds with estrogenic activity can inhibit peptide binding to PDIp, a pancreas-specific member of the protein-disulfide isomerase family (23).
To dissect the different activities of PDI we used human proinsulin as model substrate. Human proinsulin consists of a single polypeptide which, after trypsin and carboxypeptidase B cleavage, can be converted to the biologically active insulin and the c-peptide in vitro (24). Proinsulin contains three disulfide bonds (1) Cys 7 -Cys 72 , (2) Cys 19 -Cys 85 , and (3) Cys 71 -Cys 76 which are essential for its native structure. In addition to the efficient spontaneous oxidative folding of reduced proinsulin at * 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.
§ Supported by a grant from the Deutscher Akademischer Austauschdienst.
Here we investigated the mechanism of PDI function in proinsulin refolding. We show that PDI influences the refolding of denatured and reduced proinsulin both as an isomerase and as a chaperone. PDI in catalytic amounts is able to increase the refolding rate and the refolding yield while PDI variants devoid of isomerase activity influence the refolding yield only when present in stoichiometric amounts. The chaperone function is essential during the first seconds of refolding because aggregation of folding proinsulin is a major side reaction. Inhibition of the peptide-binding site of PDI leads to a suppression of the aggregation preventing role of PDI. Besides the chaperone function the isomerase activity is also required at the beginning of proinsulin folding, but the late refolding process does only depend on the isomerase activity.

EXPERIMENTAL PROCEDURES
Reagents-Recombinant native human proinsulin was provided by BIOBRÁ S, Montes Claros, Brazil. ELISA antibodies were obtained from Roche Molecular Biochemicals, Penzberg, Germany. The vectors pET23-PDI and pET23-PDI-bЈaЈc were received from Dr. L. W. Ruddock, University of Kent, Canterbury, UK. PDI and PDI variants were purified from Escherichia coli lysates as described below. Genistein and the homobifunctional cross-linking reagent disuccinimidyl glutarate were purchased from Sigma, iodoacetamide was obtained from ICN. X-ray films and Bolton-Hunter 125 I labeling reagent were obtained from Amersham Bioscience, Inc.
Unfolding and Refolding of Proinsulin-The human proinsulin, containing a N-terminal His 8 -Arg-tag was subjected to reductive unfolding by adding 20 mg of the native proinsulin to 1 ml of 6 M guanidinium chloride, 10 mM Tris/HCl, pH 8.5, 1 mM EDTA, 1.1 M dithiothreitol. The sample was incubated for 8 h at 37°C and then extensively dialyzed at 4°C against 6 M guanidinium chloride, 5 mM EDTA, pH 3. For crosslinking, the denatured and reduced proinsulin was desalted to a final concentration of 2 M guanidinium chloride using a NAP-5 column (Amersham Bioscience, Inc.).
Refolding of proinsulin was performed in 10 mM Tris, 10 mM glycine, pH 7.5, 1 mM EDTA, 1 mM GSH, 2 mM GSSG at 25°C by diluting the unfolded and reduced proinsulin to a final concentration of 100 g/ml, if not otherwise indicated. The sample was immediately mixed. At distinct time points aliquots were removed and acetonitrile (final concentration 20% (v/v)) and trifluoroacetic acid (final concentration 0.1% (v/v)) were added, if not otherwise indicated.
Refolding of proinsulin in the presence of genistein was performed with proinsulin at a final concentration of 27 g/ml. Genistein (10 mM solution in Me 2 SO) was added to a final concentration of 20 M. Aliquots were removed and digested with trypsin and carboxypeptidase B (see below).
Quantification of Proinsulin Folding-Refolding of denatured and reduced proinsulin was monitored by reversed phase HPLC (RP-HPLC) using a C 18 column (Macherey-Nagel) equilibrated with 100% solvent A (solvent A: 20% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid; solvent B: 80% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid). The protein was eluted at 20°C with a flow rate of 0.5 ml/min of a linear gradient from 20 to 40% solvent B within 20 min. For samples containing genistein and for samples after digestion, the column was equilibrated with 10% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid. The protein was eluted with a linear gradient from 10 to 38% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid with a flow rate of 0.5 ml/min within 40 min. Peaks were detected by absorbance at 214 nm and quantified according to a calibration curve.
Tryptic Digestion and ELISA-Refolding samples containing genistein were digested with trypsin and carboxypeptidase B (final concentration 1.35 g/ml each) for 20 min at ambient temperature. Afterward acetonitrile and trifluoroacetic acid were added to a final concentration of 10 and 0.1% (v/v), respectively, and the samples were analyzed by RP-HPLC as described above.
To verify the native structure of the refolded proinsulin, a digestion was performed with native and refolded proinsulin at refolding conditions (100 g/ml proinsulin). After incubation for 20 min at ambient temperature the digestion was stopped by addition of soybean trypsin inhibitor and EDTA. The final concentrations were 5 g/ml trypsin, 5 g/ml carboxypeptidase B, 25 g/ml trypsin inhibitor, and 125 mM EDTA. Samples were analyzed by RP-HPLC as described above and by an insulin-ELISA as described previously (25).
Light Scattering-Light scattering due to aggregation was measured at 500 nm in a stirrable 2-ml cuvette using a fluorescence spectrometer (FLUOROMAX, Spec Instruments). For proinsulin concentrations of 10 -100 g/ml both excitation and emission slits were adjusted to 1 nm band width, for lower concentrations to 3 nm. The temperature of the refolding buffer (see above) was adjusted to 25°C. The protein to be measured was added to the stirred refolding buffer and aggregation was recorded for 120 to 600 s. At the same conditions a calibration curve was generated for denatured and reduced proinsulin (from 0.2 to 10 M). According to this calibration curve, the light scattering signal caused by the aggregation of proinsulin could be calculated to micrograms/ml aggregating proinsulin. Related to the amount of proinsulin used one can calculate the corresponding amount of proinsulin protected from aggregation.
Generation of PDI Variants-Point mutations to replace the codons for cysteine in the active sites of PDI by serine were generated using the vector pET23-PDI for the quick change procedure (Qiagen). The wildtype sequences 5Ј-TGTGGCCACTGC-3Ј for the a domain (corresponding to -35 CGHC 38 -) and 5Ј-TGTGGTCACTGC-3Ј for the a domain (corresponding to -379 CGHC 382 -) were changed to 5Ј-TCCGGTCACTCT-3Ј encoding the amino acid sequence -SGHS-. The active site in either the a domain (PDI⌬C1) or a domain (PDI⌬C2) was mutated to generate a PDI-single mutant. The PDI-double mutant (PDI⌬C1,2) was produced by replacing the cysteine codons of both active sites by serine. The correct sequence was verified by DNA sequencing.
Expression and Purification of PDI and PDI Variants-For the production of wild-type PDI, PDI-bЈaЈc, and the mutants PDI⌬C1, PDI⌬C2, and PDI⌬C1,2, E. coli BL21(DE3)-pLysS was transformed with the corresponding plasmid-DNA. The cells were incubated at 37 or 30°C on LB medium containing 100 g/ml ampicillin and 25 g/ml chloramphenicol. Three hours after induction with 1 mM isopropyl-thio-␤-Dgalactopyranosid, the cells were harvested by centrifugation. The cell pellet was suspended in buffer A (20 mM Na-phosphate, pH 7.3) and DNase was added to a final concentration of 10 g/ml. After freezing and thawing the suspension twice, the lysate was centrifuged. After ultrafiltration (0.2 m), the supernatant was loaded onto a Ni-NTA column (volume 12 ml, Qiagen), which, after activation with NiSO 4 and washing with double-distilled water, was equilibrated with buffer A. Proteins bound nonspecifically onto the matrix were removed by washing the column with buffer A containing 500 mM NaCl and 50 mM imidazole, followed by buffer A. Recombinant protein was eluted with buffer A containing 10 mM EDTA. The elution fraction was loaded onto a ResourceQ column (6 ml, Amersham Bioscience, Inc.) equilibrated with buffer A and eluted with a linear gradient from 0 to 0.5 M NaCl in buffer A. Fractions containing homogenous PDI or PDI variants were pooled and dialyzed against 1 mM Tris, 1 mM glycine, pH 7.5, 0.1 mM EDTA. The concentration of purified protein was determined by UV spectroscopy (molar absorbance coefficient 45,380 M Ϫ1 cm Ϫ1 (280 nm)). The molecular mass of the purified proteins was determined by mass spectrometry. PDI⌬C1,2 was centrifuged (70,000 ϫ g, 4°C, 30 min) prior to use and afterward the correct concentration was determined by UV spectroscopy.
Alkylation of Cysteines-PDI was incubated with 5 mM dithiothreitol. After 20 min at 25°C, iodoacetamide was added to a final concentration of 50 mM and the sample was incubated for 45 min at 25°C. Excess dithiothreitol and iodoacetamide was removed by dialysis against 1 mM Tris, 1 mM glycine, pH 7.5, 0.1 mM EDTA and the alkylated PDI lyophilized and stored at Ϫ20°C. For refolding experiments, lyophilized alkylated PDI was dissolved in double distilled water to a concentration of ϳ5 mg/ml. After centrifugation (70,000 ϫ g, 4°C, 30 min), the correct concentration was determined by UV spectroscopy.
Cross-linking-Bolton-Hunter 125 I labeling of denatured and reduced proinsulin (5 mg/ml, in 2 M guanidinium chloride, pH 3.0) was carried out as recommended by the manufacturer. Radiolabeled proinsulin was incubated with the E. coli cell extract that expressed PDI (10 g/ml) on ice. After 10 min 0.1 volume of disuccinimidyl glutarate (5 mM in Me 2 SO) was added and the sample incubated for additional 60 min. The reaction was stopped by the addition of SDS sample buffer (26). Proteins were separated on 12.5 or 15% polyacrylamide gels and electrotransferred onto a polyvinylidene difluoride membrane and subsequently analyzed by autoradiography. 2 J. Winter, unpublished results.

Refolding of Proinsulin Is Catalyzed by PDI-PDI
catalyzed the refolding of denatured and reduced proinsulin with two effects (Fig. 1A). First, PDI increased the rate of oxidative folding in a concentration-dependent manner. In the absence of PDI the apparent rate constant of oxidative refolding of proinsulin was k app ϭ 0.0018 s Ϫ1 . With equimolar amounts of PDI present the rate constant of folding was increased drastically to approximately k app ϭ 0.033 s Ϫ1 . Second, PDI enhanced the refolding yield (Fig. 1C). The yield reached a plateau of about 50 -60% once PDI was present in a molar ratio of PDI/proinsulin of 0.033 and could not be increased further even in the presence of stoichiometric amounts of PDI. Under the same conditions bovine serum albumin did not affect proinsulin folding (data not shown). Refolding of proinsulin depended on the redox conditions; with 1 mM GSH and 2 mM GSSG the best yield of refolding was obtained for the spontaneous and the catalyzed reaction (data not shown).
A slow decrease in native proinsulin was observed after folding in the presence of high PDI concentrations (Fig. 1A). To analyze this effect native proinsulin was incubated under the same conditions in the absence and presence of PDI (data not shown). The amount of native proinsulin was unchanged under all conditions indicating that aggregation or proteolysis of native proinsulin did not occur under these conditions and that no isomerization of native disulfide bonds occurred, neither spontaneously nor catalyzed by PDI. In addition, native proinsulin at refolding conditions and refolded proinsulin were digested with trypsin and carboxypeptidase B and the digestion products were analyzed by an insulin-ELISA (25). At these conditions only correctly folded proinsulin can be converted to native insulin (24). The amount of insulin generated from native and refolded proinsulin, respectively, was identical indicating that the refolded proinsulin was indeed correctly folded and contained native disulfide bonds (data not shown).
PDI Variants Can Improve Refolding-PDI, PDI-bЈaЈc (a PDI variant containing only the last three domains), and mutants having one (PDI⌬C1 and PDI⌬C2) or both (PDI⌬C1,2) active sites inactivated by cysteine to serine substitutions were tested with respect to refolding of denatured and reduced proinsulin. As shown in Fig. 1B only wild-type PDI was able to increase refolding to the maximum rate and yield. PDI⌬C1 and PDI⌬C2 increased the folding rate and yield, but not to the same extent as wild-type PDI, even when they were present at a two times higher concentration which corresponded to the same amount of active sites present in wild-type PDI. This effect was independent of which active site, either in the a domain or a domain, was mutated. PDI-bЈaЈc had the same properties in proinsulin renaturation as PDI⌬C1 and PDI⌬C2 (data not shown). PDI⌬C1,2 was not capable of increasing the folding rate, however, PDI⌬C1,2 could improve the yield of refolding by about 30% (Fig. 1C). In contrast to wild-type PDI which increased the yield of refolding of proinsulin already at substoichiometric concentrations PDI⌬C1,2 had to be present at least at stoichiometric concentrations. These results indicate that PDI acts both as an isomerase and as a chaperone when present during refolding of proinsulin.
PDI Is Essential during the First Seconds of Refolding-Next, we asked at which stages of the renaturation process the isomerase and the chaperone activity of PDI are required to allow efficient refolding of proinsulin. Refolding was performed in the presence of wild-type PDI or PDI⌬C1,2 added to refolding proinsulin at different times of renaturation (Fig. 2). The chaperone activity was only effective when present during the first seconds of the refolding process. While PDI⌬C1,2 could chaperone renaturation of proinsulin when present at the be- ginning of the process, no increase in yield was observed if PDI⌬C1,2 was added to refolding proinsulin after 7 s of refolding. PDI inactivated by alkylation (13) was tested under the same conditions (data not shown) and its activity on refolding was identical to that of PDI⌬C1,2 thus excluding unspecific effects of possible structural changes caused by the point mutations in PDI⌬C1,2. Both, PDI⌬C1,2 and alkylated PDI, showed identical results in the timed addition experiment.
In contrast, wild-type PDI, even when added 1 min after initiation of folding could improve renaturation. This improvement, however, was not as efficient as when present from the beginning. This indicates that the significant decrease in the refolding yield due to the absence of the chaperone function in the first seconds of refolding can be partly compensated by the isomerase activity of PDI. The observed decrease in the yield of refolding of proinsulin when PDI was added to the refolding sample at different time points exhibits a biphasic kinetic. A very fast first phase with an apparent rate constant k app ϭ 0.075 s Ϫ1 was followed by a slow phase with a k app ϭ 0.0043 s Ϫ1 . Compared with the rate constant of formation of native proinsulin (k app ϭ 0.0018 s Ϫ1 , Fig. 1A) the second phase occurred in a similar time range but the first phase was much faster. From this we conclude that upon renaturation of proinsulin there is a kinetic competition between an overall slow folding process and an ϳ20 times faster off-pathway reaction. This off-pathway seems to depend on redox reactions subsequently leading to aggregation.
Chaperone Function of PDI Prevents Aggregation of Proinsulin-Furthermore, PDI and PDI⌬C1,2 were tested with respect to peptide binding. Chemical cross-linking with an amino-specific cross-linking reagent is a useful method to detect interaction partners for PDI (7,26). Both variants of PDI were capable of binding radiolabeled denatured and reduced proinsulin (Fig. 3). Furthermore, the binding was reversible, as it could be competed with other substrate peptides and proteins (data not shown). This clearly demonstrates that PDI and PDI⌬C1,2 were indistinguishable with respect to their interactions with substrates. PDI⌬C1,2 shows two bands in the cross-linking (Fig. 3). The signal at lower molecular weight corresponded to a degraded PDI⌬C1,2. Both, the full-length and the degraded PDI⌬C1,2 were able to bind denatured and reduced proinsulin. This degradation was probably due to proteases present in the E. coli cell extract used for the cross-linking assay. No degradation of PDI⌬C1,2 was observed with purified protein under refolding conditions as analyzed by SDS gels (data not shown). To demonstrate that the observed effects of the PDI⌬C1,2 were not due to misfolding of one or more of its domains its stability was compared with that of wild-type PDI. Incubation of PDI⌬C1,2 or the wild-type PDI with various concentrations of proteinase K showed that there was no significant difference in the protease sensitivity indicating that PDI⌬C1,2 can adopt as compact a structure as the wild-type and that there was no difference in the structural stability of both proteins (data not shown).
Aggregation of folding proinsulin as detected by light scattering was proportional to the proinsulin concentration in the refolding sample (Fig. 4B, inset). It occurred in the first seconds of refolding, reaching a maximum after 20 -30 s (Fig. 4A). This indicates that aggregation of proinsulin during renaturation occurred in the same time range as the fast phase of the unproductive reaction observed in the timed addition experiment of PDI (Fig. 2). PDI suppressed aggregation of refolding proinsulin substantially and was already effective in catalytic amounts (Fig. 4B) which is in agreement with the refolding data shown in Fig. 1A and the timed addition data from Fig. 2. PDI at a molar ratio of PDI/proinsulin of 0.05 could protect about 20% of the folding proinsulin from aggregation. In contrast, alkylated PDI in the same molar ratio could suppress only 10% of the aggregation (data not shown). Only when present in stoichiometric amounts could alkylated PDI prevent aggregation of folding proinsulin to the same extent as PDI. Under identical conditions bovine serum albumin did not suppress aggregation (data not shown).
This clearly indicates that PDI, present in catalytic amounts, possesses chaperone activity which alone is not sufficient to protect folding proinsulin from aggregation. Early intermediates in the proinsulin folding pathway, probably containing wrong disulfide bonds, seem to be highly susceptible to aggregation. In the presence of wild-type PDI these aggregationprone intermediates can isomerize to folding intermediates that are less susceptible to aggregation. This results in reduced aggregation and, as a consequence, in an enhanced refolding yield. When present in very high concentrations alkylated PDI was similarly effective, indicating that a noncatalyzed isomer-FIG. 2. Timed addition of PDI to refolding proinsulin. Reactivation was performed as described in the legend to Fig. 1, except that PDI (q) and PDI⌬C1,2 (E), respectively, were added after refolding start at the times indicated. The final concentration was 2 M for PDI and 28 M for PDI⌬C1,2. Refolding was analyzed after the renaturation was completed. The data (q) were fitted to a biphasic reaction with apparent rate constants of k app ϭ 0.075 min Ϫ1 and k app ϭ 0.0043 min Ϫ1 .

FIG. 3. Interaction of PDI and PDI⌬C1,2 with proinsulin.
Equal amounts of E. coli cell extracts that expressed wild-type PDI (PDI) and PDI⌬C1,2 (mutant), respectively, were incubated with radiolabeled denatured and reduced proinsulin (final concentration 33 M). For control, E. coli cell lysates that did not contain PDI or PDI⌬C1,2 (buffer) were incubated at the same conditions. Incubation was performed in a total volume of 10 l at 0°C for 10 min. Samples were subsequently incubated with disuccinimidyl glutarate (final concentration 0.5 mM) for 60 min at 0°C and analyzed on a 12.5% polyacrylamide gel with subsequent autoradiography. The cross-linking products are indicated by an arrow.
ization of disulfides during proinsulin folding can occur if aggregation is sufficiently suppressed by the chaperone activity. However, independent of the presence of PDI a significant part of proinsulin aggregated during the refolding. This correlates with the maximum refolding yield of about 50%.
Proinsulin Binding to PDI Is Suppressed by the Inhibitor Genistein-To analyze whether the interaction of PDI and PDI⌬C1,2 with denatured and reduced proinsulin can be suppressed by an inhibitor of the chaperone activity we used the small molecular weight substance genistein. It has been shown previously that genistein, a substance with estrogenic activity, can suppress the binding of ⌬-somatostatin to PDIp, a member of the protein-disulfide isomerase family (23), and also PDI. 3 Both, PDI and PDI⌬C1,2 showed the same genistein binding properties (Fig. 5A) indicating that the peptide-binding domain of PDI⌬C1,2 was not affected by the mutations in both active sites. By titration of genistein we found that a 3-fold molar excess of genistein over proinsulin was required to completely suppress binding of denatured and reduced proinsulin to PDI as well as to PDI⌬C1,2.
Genistein-inhibited PDI or alkylated PDI were significantly affected in their ability to prevent proinsulin from aggregation (Fig. 5B). In refolding samples analyzed by light scattering genistein-inhibited PDI and genistein-inhibited alkylated PDI could protect about 11% of the folding proinsulin from aggregation. This indicates that both PDIs still had some residual chaperone activity under the experimental conditions. Genistein inhibited 66% of the chaperone function of PDI. For comparison, the inhibitory effect of genistein on alkylated PDI was less pronounced (about 50% inhibition). The lower inhibition effect for alkylated PDI was due to its two times higher concentration in the refolding sample and therefore a two-times less excess of genistein over alkylated PDI compared with PDI.
In agreement with the results shown above, the inhibition of the chaperone function of PDI drastically changed the PDIassisted refolding behavior of proinsulin (data not shown). In PDI-catalyzed refolding a 10-fold molar excess of genistein over proinsulin decreased the refolding rate four times and the yield of proinsulin refolding was only 37% compared with 50% in the presence of non-inhibited PDI. For refolding in the presence of genistein-inhibited PDI⌬C1,2 or alkylated PDI the stimulating effect mediated by the variants with no isomerase activity decreased by 60% (corresponding to about 25% refolded proinsulin). The rate constant did not change compared with refolding without genistein. This shows that the inhibited PDIs still had some residual chaperone activity and that in the case of wild-type PDI the isomerase activity can facilitate a higher refolding yield. Similar results were obtained for refolding of proinsulin with a four times and a 100 times molar excess of genistein (data not shown). This is in agreement with the cross-linking data shown above. As a control we confirmed that the aggregation and refolding properties of proinsulin were not changed in the presence of genistein alone (see above). Additionally, Me 2 SO used to solubilize the genistein did not influence proinsulin refolding, neither in the absence nor the presence of PDI or PDI variants (data not shown). This proves that by binding of genistein to the peptide-binding domain of PDI binding of folding proinsulin is suppressed. This inhibition of the chaperone function of PDI leads to enhanced aggregation of folding intermediates and to a reduction of the yield of refolding.
The Chaperone and Isomerase Activity of PDI Are Not Required during the Entire Refolding Process-To analyze whether both the chaperone and isomerase function of PDI were only required during the first seconds of refolding or whether they were essential during the entire refolding process, the refolding of proinsulin was monitored in the presence of PDI and PDI⌬C1,2, and genistein was added at different times after initiation of refolding (Fig. 6A). Similar to these experiments we performed proinsulin refolding and added genisteininhibited PDI at different time points after refolding started (Fig. 6B).
As described above, refolding of proinsulin in the presence of genistein-inhibited PDI or PDI⌬C1,2 resulted in a significant decrease of the yield of refolding. In contrast, genistein added 25 s after initiation of refolding of proinsulin in the presence of PDI or PDI⌬C1,2 did not influence the refolding yield. For both, refolding of proinsulin with PDI or PDI⌬C1,2, the obtained refolding yield was similar to the yield of refolding without genistein. This indicates that at this time, 25 s after refolding started, the chaperone function was not longer required (Fig.  6A). These results are in agreement with the light scattering data (Fig. 4A) showing that aggregation occurred in the first few seconds of refolding. Timed addition of genistein-inhibited 3 L. W. Ruddock and P. Klappa, unpublished observations. PDI to refolding proinsulin exhibited a decrease in the refolding yield with a biphasic characteristic and an apparent rate constant for the fast, first phase of k app ϭ 0.029 s Ϫ1 . PDI with isomerase activity could increase the refolding yield significantly even when added 1 or 5 min after starting refolding. This catalytic effect was independent of the chaperone activity of PDI. The results of the above experiments clearly show that the isomerase activity could compensate for an impaired chaperone activity and that in this case the isomerase activity became essential to facilitate efficient proinsulin refolding. If the isomerase activity is available the chaperone activity seems to play an ancillary role.

DISCUSSION
Previous work on refolding activities of PDI in vitro showed that PDI has catalytic activity as a thiol:disulfide isomerase and can act as chaperone. In most cases PDI affects only one or the other of these functions (8,9,13). Yao et al. (11) proposed that for the reactivation of acidic phospholipase A 2 both functions are required. They stated that in their model system only a small part of PDI acts catalytically as an isomerase and the residual part functions as a molecular chaperone, which can be replaced by alkylated PDI. In our model system refolding of proinsulin is dependent on the redox conditions and influenced by aggregation and disulfide bond formation. A PDI variant with no isomerase activity can substantially suppress aggregation of refolding proinsulin. However, this PDI variant was not as efficient as wild-type PDI with respect to refolding yield and refolding rate, indicating that disulfide isomerization also plays an important role in the refolding of proinsulin. Maximum rate and yield of proinsulin refolding can only be achieved if both, the chaperone and the isomerase activity of PDI were present.
Although PDI exhibits isomerase and chaperone activity PDI seems to be more efficient as an isomerase than as a chaperone (13,27). PDI acts only as an isomerase on refolding of an antibody fragment (13). This effect is limited to the first seconds of refolding. However, if the chaperone BiP is simultaneously present BiP can efficiently bind and re-bind folding intermediates. Thus it keeps the cysteines of the folding antibody fragment accessible to PDI over a much longer time scale and PDI can sequentially act as an isomerase (27). The active site sequence -WCGHC-of PDI enables very efficient disulfide isomerization as compared with wild-type thioredoxin. A Pro to His mutation (ϪCGPCϪ to ϪCGHCϪ) in the active site of thioredoxin increases the isomerization activity (28). For thioredoxin a flat and hydrophobic molecular surface area on one side of the redox active disulfide bond has been described that is suggested to be the binding area for redox interactions with other proteins (29,30). On the basis of sequence homologies to thioredoxin, it was suggested that PDI also has protein binding capacity. The principle peptide-binding site is located in the b domain of PDI, however, for efficient binding of proteins the contribution of the a domain (bЈaЈc) or the a and b domains FIG. 5. Influence of the inhibitor genistein on peptide binding of PDI. A, inhibition of proinsulin binding to PDI and PDI⌬C1,2 by genistein analyzed by cross-linking. E. coli cell lysates that expressed PDI (PDI) and PDI⌬C1,2 (mutant), respectively, were incubated with 33 M radiolabeled denatured and reduced proinsulin in the presence of the indicated concentrations of genistein. The samples were cross-linked with disuccinimidyl glutarate and analyzed on a 15% polyacrylamide gel with subsequent autoradiography. The cross-linking products are indicated by an arrow; B, influence of genistein-inhibited PDI and genistein-inhibited alkylated PDI on proinsulin aggregation. Refolding was performed in 10 mM Tris, 10 mM glycine, pH 7.5, 1 mM EDTA, 1 mM GSH, 2 mM GSSG at 25°C in a stirred cuvette. The final concentration was 0.25 M for denatured and reduced proinsulin, 1 mM for guanidinium chloride, 0.025 M for PDI, 0.05 M for alkylated PDI, 10 M for genistein, and 0.009% (v/v) for Me 2 SO. PDI or alkylated PDI were added to the refolding buffer after a preincubation with genistein or Me 2 SO for 10 min. Afterward denatured and reduced proinsulin was added to the stirred sample. The amount of proinsulin protected from aggregation was determined according to a calibration curve. Aggregation of proinsulin was analyzed in the presence of PDI and Me 2 SO (bar 1) or genistein (bar 4), in the presence of alkylated PDI and Me 2 SO (bar 2) or genistein (bar 5), and in the absence of PDI but with Me 2 SO (bar 3) or genistein (bar 6).
(abbЈ) are required (7). In most cases, aggregation is a nonproductive, off-pathway reaction, which competes with the correct folding of proteins (31). Consequently, chaperones (e.g. GroEL/GroES, DnaK/DnaJ/GrpE, SecB, and ClpB from E. coli or the Hsp families of eukaryotes) interacting with folding intermediates and thereby preventing or minimizing aggregation can increase the refolding yield (32Ϫ35). When folding intermediates escape the protective function of chaperones, they can form stable protein aggregates. Few cases are known where chaperones (e.g. the DnaK system of E. coli, sometimes sequentially acting with ClpB) have the potential to disaggregate protein aggregates (36Ϫ38) and most chaperones and also PDI cannot actively dissolve protein aggregates.
Here, wild-type PDI, the PDI domain construct bЈaЈc, and the full-length PDI mutants with only one active site (PDI⌬C1 and PDI⌬C2) were used to catalyze refolding of proinsulin. Although all non-wild-type variants are less effective than PDI with respect to proinsulin refolding they can significantly increase the refolding rate and yield. We demonstrated that (i) the bЈaЈc domain construct of PDI can catalyze the refolding of proinsulin, (ii) the presence of one active site was sufficient to accelerate proinsulin refolding, however, (iii) full-length PDI was needed for maximum catalytic activity. This is in agreement with the data of Darby et al. (6) who demonstrated that the b domain of PDI has an especially important role in catalysis, and that maximum catalytic activity in disulfide bond rearrangements requires the involvement of all PDI domains. It was proposed that all PDI domains participate in substrate binding and especially the binding of non-native proteins might require all domains of PDI (7,39).
Using isomerase-inactive PDI variants and PDI with inhibited chaperone function, we now can clearly distinguish between both functions PDI provides. The mutation in the active site did change the enzyme characteristic but did not effect the chaperone activity. This was concluded as both, alkylated PDI and PDI⌬C1,2, were equally active regarding peptide binding as shown by cross-linking and refolding of denatured and reduced proinsulin. Genistein, which can suppress ⌬-somatostatin binding to the pancreatic protein-disulfide isomerase PDIp (23), is also able to bind to the peptide-binding site of PDI. Thus genistein can efficiently suppress binding of other substrates to PDI as shown by cross-linking experiments. By inhibition of the principal peptide-binding site the refolding yield of proinsulin was significantly decreased although not completely abolished indicating that other domains can contribute to the chaperone activity of PDI. However, in this case the interactions might be too weak to be detected in the cross-linking experiments. Genistein-inhibited PDI variants lacking isomerase activity that should not influence refolding were still able to increase proinsulin refolding, but not as efficient as the noninhibited variant. This might indicate that proinsulin binding occurs not only at the b domain but extends through all PDI domains or that PDI contains more than one peptide-binding site with only one site as a target for genistein. Furthermore, neither for genistein nor for proinsulin the binding constant to PDI is known. Hence we cannot exclude that the binding of refolding proinsulin to PDI is possible with different affinities for different folding intermediates, even in the presence of an excess of genistein.
In spontaneous refolding of proinsulin about 20% of the denatured and reduced proinsulin was folding competent. The remaining part was excluded from productive refolding by very fast aggregation of completely reduced proinsulin or folding intermediates. Providing optimum conditions (e.g. refolding at pH 10.5) 2 aggregation of folding proinsulin is reduced and about 60% native proinsulin can be formed. Similarly, IGF-I that belongs to the insulin superfamily and contains three motif-specific disulfide bonds yields under optimized conditions about 60% native protein within 5 h (40). Proinsulin and also porcine insulin precursor (41) form no stable intermediate as possible intermediates seem to be either highly susceptible to aggregation or can fold to the native protein (41). In contrast, IGF-I yields two isoforms that are stable under refolding conditions (42Ϫ44). The first one is native IGF-I with the disulfide bonds: 1) Cys 6 -Cys 48 , 2) Cys 18 -Cys 61 , and 3) Cys 47 -Cys 52 , which correspond to Cys 7 -Cys 72 , Cys 19 -Cys 85 , and Cys 71 -Cys 76 in proinsulin. The second isoform is non-native IGF-I with the disulfide bonds 1 and 3 mismatched. These species occur in a ratio of 60% native to 40% mismatched.
The apparent rate constant determined in timed addition experiments with PDI showed that in the first seconds of proinsulin refolding very fast off-pathway reactions play an important role. These unproductive reactions are reduced by both FIG. 6. Influence of the inhibitor genistein on PDI-catalyzed refolding of proinsulin. Refolding of denatured and reduced proinsulin was performed in 10 mM Tris, 10 mM glycine, pH 7.5, 1 mM EDTA, 1 mM GSH, 2 mM GSSG at 25°C. The final concentration was 2.5 M for proinsulin and 15 mM for guanidinium chloride. After 60 min, refolded proinsulin was digested with trypsin and carboxypeptidase B for 20 min. Afterward acetonitrile and trifluoroacetic acid were added to a final concentration of 10 and 0.1% (v/v), respectively, and the samples were analyzed by RP-HPLC. A, timed addition of genistein. Genistein (10 M) was added to refolding proinsulin in the absence of PDI (E), in the presence of 0.25 M PDI (ƒ) and 0.5 M PDI⌬C1,2 (Ⅺ) after renaturation start at the times indicated. The data were fitted single exponentially with apparent rate constants of k app ϭ 0.038 s Ϫ1 (for PDI), and k app ϭ 0.055 s Ϫ1 (for PDI⌬C1,2). B, timed addition of genistein-inhibited PDI to refolding proinsulin. PDI and genistein were preincubated for 10 min and added to refolding proinsulin after the renaturation was initiated at the times indicated. Final concentration was 0.25 M for PDI and 10 M for genistein. The data were fitted to a biphasic reaction with apparent rate constants of k app ϭ 0.029 min Ϫ1 and k app ϭ 0.00024 min Ϫ1 . the chaperone and isomerase activity. The chaperone function of PDI was important during the first seconds but not at later stages of refolding. In the absence of the chaperone activity, however, the isomerase function of PDI became essential to ensure an increased refolding yield. The isomerase function of PDI is acting during the whole refolding process although 10 min after initiation of refolding its effect was less apparent. From the kinetics of spontaneous proinsulin refolding we know that refolding was completed after about 30 -60 min. This shows that folding and possibly isomerization occurred even 10 -30 min after refolding started when aggregation was already completed. These late intermediates were not particularly susceptible to aggregation and PDI could catalyze these reactions but did not enable a significant increase in the yield of refolding of proinsulin. In the presence of PDI, aggregation as the major side reaction was suppressed although not completely even in stoichiometric concentrations. Under these conditions the proinsulin species protected from aggregation (about 40%) became folding competent and refolded completely as the refolding yield during catalyzed refolding was increased to about 50 -55%. This indicates that if side reactions can be efficiently suppressed, either by accelerated isomerization or reduced aggregation, denatured and reduced proinsulin can form the native molecule. However, off-pathway reactions occurred very fast and obviously they cannot be reversed by PDI. This might be the major reason why we cannot force the proinsulin to refold quantitatively.