Refolding Mechanism of Ovalbumin

Ovalbumin, a member of the serpin superfamily, contains one cystine disulfide (Cys73-Cys120) and four cysteine sulfhydryls (Cys11, Cys30, Cys367, and Cys382) in the native state. To investigate the folding mechanism of ovalbumin, a urea-denatured disulfide isomer with a mispaired disulfide Cys367-Cys382 (D[367–382]) and its derivative (D[367–382/CM-73]) in which a native cystine counterpart of Cys73 is blocked by carboxymethylation were produced. Both the denatured isomers refolded within an instrumental dead time of 4 ms into an initial burst intermediate IN with partially folded conformation. After the initial burst phase, most of the D[367–382] molecules further refolded into the native form. In contrast, upon dilution of D[367–382/CM-73] with the refolding buffer, the protein stayed in the IN state as a stable form, which displayed a partial regain of the native secondary structure and a compact conformation with a similar Stokes radius to the native form. The structural characteristics of IN were clearly differentiated from those of an equilibrium intermediate IA that was produced by dilution with an acidic buffer of urea-denatured ovalbumin; IA showed much more hydrophobic dye binding and a larger Stokes radius than the IN state, despite their indistinguishable far-UV circular dichroic spectra. The non-productive nature of IA highlighted the importance of a compact conformation of the IN state for subsequent native refolding. These observations were consistent with a refolding model of ovalbumin that includes the regain of the partial secondary structure and of the compactness of overall conformation in an initial burst phase before the subsequent native refolding.

The importance of protein folding studies has been increasing, because they give important information about the molecular mechanism for novel types of disease that involve a protein aggregation process (1)(2)(3)(4)(5). The serpins, a group of serine proteinase inhibitors, are an interesting target for investigation of the protein folding mechanism. They undergo unique conformational change upon exerting the inhibition activity; after receiving cleavage at the P1-P1Ј site by a target proteinase, the reactive center loop is inserted into the central ␤-sheet A (6). This dynamic conformational change accompanies a great structural stabilization and has been considered to be from a metastable state into a fully stabilized state (7). The folding analysis of serpins should therefore explain how a protein molecule folds into a metastable state with native protein nature. The folding process of serpins has, however, been poorly understood. This is largely related to the large protein size with a molecular mass of around 40 kDa, which makes it difficult for the application of powerful protein folding analysis, such as nuclear magnetic resonance.
Ovalbumin, despite its non-inhibitory nature (8), is a member of the serpin superfamily, because of a close similarity in the primary and tertiary structures (9 -11) and the specific loop cleavage at the canonical P1-P1Ј site by a serine proteinase (12). Furthermore, recent thermodynamic and crystallographic evidence from our laboratory has clearly demonstrated that ovalbumin exerts, by the replacement of a single hinge residue, the inherent metastable nature that undergoes structural transition into the loop-inserted, thermostabilized form following P1-P1Ј cleavage (13). Ovalbumin should be, therefore, a competent model for the analysis of the metastable folding mechanism. We have shown that intramolecular sulfhydryl/disulfide exchange reactions are a useful probe for the analysis of the unfolding and refolding process of ovalbumin (14 -19). Ovalbumin consists of a single polypeptide chain of 385 amino acid residues and contains 6 cysteine residues of Cys 11 , Cys 30 , Cys 73 , Cys 120 , Cys 367 , and Cys 382 ; only Cys 73 and Cys 120 form a disulfide bond in the native state (20,21). Under highly denaturing conditions, the egg white protein undergoes extensive sulfhydryl/disulfide exchanges producing all the 15 possible disulfide isomers with one disulfide and four sulfhydryls (14). Thus the disulfide isomers are distributed at equilibrium depending on loop length (the number of amino acid residues separating the two cysteines) to a power of about Ϫ2. From the complex denatured state, most of the ovalbumin molecules refold into the native form with Cys 73 -Cys 120 through intrachain sulfhydryl/disulfide exchanges (15). This enables us to investigate the refolding processes from different disulfide isomers with different conformational entropy. That the native cystine disulfide is regained without the help of an added chemical oxidant implies that the sulfhydryl accessibility problem (22,23), inherently inevitable in the oxidative refolding system, can be circumvented. As an alternative advantage, a possible folding intermediate can be trapped in a stable form under native folding conditions, as demonstrated here, when one of the cysteine residues participating in the native disulfide bond is blocked.
In a previous study using a disulfide isomer (16), we have demonstrated that the refolding process of ovalbumin includes a non-productive pathway as displayed in Scheme 1. When the urea-denatured isomer with the native disulfide Cys 73 -Cys 120 * 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.
(D[73-120]), 1 produced by protein incubation under acid/urea conditions, is diluted into a near neutral folding buffer, ovalbumin forms an initial burst intermediate I N . This intermediate can correctly refold into the native form N[73-120] with a first-order rate constant k f of about 0. 2 -73] was almost indistinguishable from disulfide-intact ovalbumin as evaluated by far-UV CD and intrinsic tryptophan fluorescence spectra and trypsin resistance analyses (data not shown).
Preparation of Urea-denatured, Mispaired Disulfide Isomers-The outline of preparation of the urea-denatured mispaired disulfide isomers D[367-382] and D[367-382/CM-73] is shown in Fig. 1. Step 1 was to produce a mixed-disulfide protein derivative; native ovalbumin (20 mg/ml) was incubated at 25°C for 1 h with 2 mM Pyr-S-S-Pyr in buffer A (50 mM potassium phosphate buffer, pH 2.2, containing 1 mM sodium EDTA), then excess Pyr-S-S-Pyr and Pyr-SH were removed by gel filtration using a Sephadex G-10 column (Amersham Biosciences, NAP-10) equilibrated with buffer B. In Step 2, the native cystine disulfide Cys 73 -Cys 120 of the derivative was reduced by incubation at 37°C for 1 h with 15 mM DTT in buffer B, then the sample was passed through a Sephadex G-10 column equilibrated with buffer A. The urea-denatured, mispaired protein D[367-382] was produced in Step 3 by incubation of the cystine-reduced, mixed-disulfide derivative at 37°C for 30 min with 6 M urea in buffer A; during this incubation, a mispaired disulfide Cys 367 -Cys 382 and Pyr-SH are produced by the attack of the Cys 382 sulfhydryl against the mixed-disulfide of Cys 367 (19).
Another urea-denatured mispaired disulfide isomer, D[367-382/CM-73] was prepared in the same way except that the Cys 73 sulfhydryl was carboxymethylated, prior to Step 3, by incubation of the cystine-reduced, mixed-disulfide derivative with 40 mM IAA at 37°C for 10 min. Refolding of the Urea-denatured Proteins-Refolding was initiated at 25°C by 20-fold dilution of the urea-denatured proteins with buffer C (50 mM Tris-HCl buffer, pH 8.6, containing 1 mM sodium EDTA) giving a final pH value of 8.2. The proteins were allowed to refold at 25°C. An equilibrium intermediate I A was produced by 20-fold dilution of the urea-denatured proteins with buffer A. The buffers were degassed at reduced pressure and equilibrated under N 2 atmosphere prior to the refolding and contained 0.5 mM DTT for the experiments for the disulfide-reduced proteins.
Analysis for Disulfide-involved Cysteines-Disulfide-involved cysteines were determined by a peptide mapping analysis as described previously (14). Briefly, the urea-denatured protein in acidic pH conditions was neutralized by addition of 1 M Tris solution and alkylated with 0.1 M iodoacetamide in the presence of 9 M urea. The alkylated protein was precipitated in a cold acetone-HCl solution, dissolved in buffer B containing 8 M urea, reduced with DTT, and then modified with a fluorescent alkylation reagent, N-iodoacetyl-NЈ-(5-sulfo-1-naphthyl)ethylenediamine. The modified protein was extensively proteolyzed with combinations of trypsin, chymotrypsin, and Achromobacter protease I. The resultant peptides were analyzed by reversed-phase high performance liquid chromatography, and the modified cysteine peptides were detected by fluorescence (excitation, 340 nm; emission, 520 nm).
For the refolding analysis, sulfhydryl/disulfide exchanges were quenched at various refolding times by mixing the protein samples with a 0.24 volume of 2 M HCl. Disulfide-involved cysteines were determined in the same way.
Intrinsic Tryptophan Fluorescence and Far-UV CD Spectrum-The fluorescence spectrum of ovalbumin was measured with a fluorescence spectrophotometer (Hitachi, model F-3000). The intrinsic tryptophan residues in ovalbumin were excited at 295 nm, and emission spectrum was recorded at a wavelength range from 300 to 420 nm. All measurements were carried out at a constant temperature of 25°C.
For the spectrum measurements at an early refolding time, a stopped-flow reaction analyzer (Applied Photophysics Ltd., UK) was employed. The time course of fluorescence intensity changes was monitored at various emission wavelengths (excitation wavelength, 295 nm), and the data at a refolding time of 5 ms were plotted. For the control experiments, the native ovalbumin and the urea-denatured proteins were diluted, respectively, by the refolding buffer (buffer C) containing 0.82 M urea and by 0.25 M HCl containing 9 M urea and 1 mM sodium EDTA (pH 2.2), and the fluorescence intensity at 338 nm was recorded in the same way. The averaged data of 10 -15 traces were obtained. The dead time for mixing was determined to be 4 ms by a model reaction between 2,6-dichlorophenolindophenol and L-ascorbate (26).
The far-UV CD spectrum was recorded at 25°C with a spectropolarimeter (JASCO, J-720). The CD data were expressed as mean residue ellipticity (degrees⅐cm 2 /dmol) by using 111 as the mean residue weight of ovalbumin. CD spectra at a short refolding time were determined by measuring the time-dependent increase in CD ellipticities at various wavelengths, and the values at the 10-s refolding time were plotted as a function of the wavelength. The averages of eight time traces were determined.
The time course of the refolding after the initial burst phase was monitored by the CD ellipticity at 222 nm and the intrinsic tryptophan fluorescence excited at 295 nm. The data were obtained as the averages of triplicate determinations. The fraction of the native form at the refolding time of t "F N (t)" was calculated by using the equation: , where X 0 and X t are the initial values and the values at the refolding time t, respectively. For the CD and fluorescence analyses, the values at 6-s and 5-ms refolding were taken as X 0 , respectively. X N is the value of the native form.
ANS Binding Experiments-Ovalbumin in the various states was mixed at 50 g/ml with 0.005 volume of 4.02 mM ANS solution, and the fluorescence emission spectra were measured with a fluorescence spectrophotometer (Hitachi, F-3000) after 30 s of the mixing. ANS was excited at 350 nm, and the emission spectra were recorded at 25°C in a wavelength range from 390 nm to 610 nm.
For the spectrum measurements at the refolding time of 30 s, the urea-denatured proteins (1.0 mg/ml) were diluted 20-fold with buffer C containing 21 M ANS, and the time course of fluorescence intensity change was monitored at various emission wavelengths (the excitation wavelength, 350 nm). The obtained data at a refolding time of 30 s were plotted as a function of the emission wavelength. The averages of eight time traces were determined. The time course of fluorescence intensity change was monitored at 470-nm emission, and the data were obtained as the averages of triplicate determinations.
Trypsin Resistance Assay-The regain of the native conformation during the refolding was examined by a trypsin resistance assay, because the protein in the native state, but not in the non-native state, is highly resistant against trypsin (17). At various refolding times, the protein was mixed with 0.01 volume of 12.5 mg/ml trypsin and digested at 25°C for 1 min. The digestion was terminated by addition of soybean trypsin inhibitor (17). The proteins were electrophoresed on an SDS-PAGE (10% polyacrylamide/0.27% bisacrylamide), according to the standard method of Laemmli (27), and then stained with Coomassie Brilliant Blue R-250. The amount of trypsin-resistant ovalbumin was determined from the band intensity with a densitometer (Shimadzu, CS-9000). The data were obtained as the averages of eight time determinations.
Differential Scanning Calorimetry-The denatured proteins were refolded for 20 h and concentrated about 15-fold using a concentrator (Amicon, Centriprep-10). The refolded proteins and native protein controls were analyzed with a differential scanning calorimeter (Micro Cal, MCS-DSC). The protein concentration was 0.5 mg/ml in buffer B, and the rate of temperature change was 1 K⅐min Ϫ1 .
Size-exclusion Chromatography-Analytical gel filtration chromatography of various states of ovalbumin was performed on a TSK gel column (G3000SW XL , 7.8 mm ϫ 30 cm) joined to a high performance liquid chromatography apparatus (Shimadzu, LC-10A). Proteins were chromatographed in buffer B (for the native and refolded proteins), in buffer A (for the I A state), or in 9 M urea-0.25 M HCl (for the ureadenatured proteins) at a flow rate of 0.5 ml/min at 25°C. Eluted protein was detected by absorbance at 220 nm with a UV absorbance monitor (Shimadzu, SPD-2AS). The Stokes radius was determined by the method of Corbett and Roche (28). The protein standards used were catalase, lactate dehydrogenase, human transferrin, bovine serum albumin, and bovine erythrocyte carbonic anhydrase. The void volume and total solvent-accessible volume of the column were determined from the elution volumes of blue dextran and sodium azide, respectively. The data (Table I) are the averages with standard deviations of quadruplicate determinations. Fig. 1. Ovalbumin takes a highly ordered molten-globule state at acidic pH, and Pyr-S-S-Pyr reacts specifically with the Cys 367 sulfhydryl at pH 2.2 (19). By use of the reaction, a mixed-disulfide protein derivative was produced in Step 1. As Step 2, the native cystine Cys 73 -Cys 120 in the derivative was reduced with DTT at near neutral pH, generating the cystine-reduced, mixed-disulfide derivative. In the presence of a high concentration of urea at pH 2.2, Pyr-SH is released from the mixed disulfide derivative by the nucleophilic attack of the nearest cysteine residue in the primary structure, Cys 382 (19). Thus the mispaired disulfide isomer D[367-382] was prepared by placing the cystine-reduced, mixed-disulfide derivative in acidic urea denaturing conditions as Step 3. The specific disulfide formation between Cys 367 and Cys 382 in the urea-denatured sample was confirmed by a peptide mapping analysis as shown in Fig. 2A; this figure clearly shows that almost all of the disulfide-forming cysteines consisted of Cys 367 and Cys 382 . By comparison with the overall labeling analysis that leads to the detection of all the six cysteines ( Fig. 2B), the yield of D[367-382] was estimated to be more than 97%.

Preparation of Mispaired Disulfide Isomers-Urea-denatured mispaired disulfide isomers D[367-382] and D[367-382/ CM-73] were prepared as summarized in
The other urea-denatured mispaired disulfide isomer D[367-382/CM-73] was prepared by inclusion of the carboxymethylation of the Cys 73 sulfhydryl prior to Step 3 ( Fig. 1). In the native conformation at near neutral pH, an alkylation reagent, such as IAA, is inaccessible to the sulfhydryls of Cys 11 , Cys 30 , Cys 367 , and Cys 382 , but Cys 73 and Cys 120 can be readily alkylated in their reduced state with IAA (17). In the present study, we found that the Cys 73 sulfhydryl is more reactive to IAA than the Cys 120 sulfhydryl and that, in an optimized condition, Cys 73 can be specifically carboxymethylated. As shown in Fig. 2D, when the cystine-reduced, mixed-disulfide derivative was carboxymethylated with 40 mM IAA at pH 8.2 (producing the carboxymethylated, mixed-disulfide derivative) and then subjected to the overall labeling analysis, the Cys 73 sulfhydryl was found to be almost completely blocked, but no other cysteine sulfhydryls were carboxymethylated with IAA. The carboxymethylated, mixed-disulfide derivative was then denatured during Step 3, and the sample was analyzed for the disulfideforming cysteines by the peptide mapping method. As shown in Fig. 2C, the disulfide-forming cysteines comprise almost totally Cys 367 and Cys 382 .
Ovalbumin can also refold correctly without any intrachain disulfide through an equivalent intermediate I N state (16); the absence of the interconversion of I N with I N [mis-SS] pool results in a more straightforward refolding of ovalbumin (16).
The refolding under disulfide-reduced conditions was therefore employed as a control experiment in the present refolding study. Refolding Detected by Optical Methods-The conformational change during the ovalbumin refolding was examined by intrinsic tryptophan fluorescence, far-UV CD spectrum and ANS binding analyses. As shown in Fig. 3 (A and B), all the ureadenatured proteins showed typical fluorescence and CD spectra for the unfolded protein; essentially no spectrum difference was detected among the different urea-denatured proteins. The native ovalbumin and the urea-denatured proteins displayed the maximum fluorescence at 338 and 352 nm, respectively, and the maximum fluorescence intensity of the urea-denatured proteins was about 33% of that of the native ovalbumin.
We have previously shown that two types of partially folded intermediates I N and I A are formed within an initial burst phase during the refolding of urea-denatured ovalbumin with the correct disulfide (D[73-120]) (16); the intermediate I N , produced by dilution of D[73-120] with a near neutral buffer, can refold to the native protein, but I A , produced by dilution of the same urea-denatured protein with an acidic buffer, stays in the partially folded state without being transferred to a near neutral pH condition in which the protein refolds correctly in the same way. In the present study, the equivalent intermediates I N and I A were found to be formed during the refolding of the urea-denatured, mispaired disulfide isomers and of their disulfide-reduced counterparts, as stated below.
Ovalbumin contains three tryptophan residues, Trp 148 in helix F, Trp 184 as the nearest neighbor residue of the carboxyl terminus of strand 3A, and Trp 267 in helix H (9 -11). Upon rapid stopped-flow dilution with a near neutral pH buffer, the urea-denatured, mispaired disulfide protein D[367-382] refolded within the mixing dead time of 4 ms into an initial burst state that showed at 338 nm an intermediate intrinsic tryptophan fluorescence intensity between the native and denatured forms (Fig. 3A, inset). The same experiment was done with the different urea-denatured proteins at various wavelengths of fluorescence emission. When the fluorescence intensity at 5 ms after the rapid mixing was plotted as a function of the wavelengths, the fluorescence spectrum obtained was almost exactly the same for all the intermediates produced from the different urea-denatured proteins; the spectrum of the initial burst intermediate was the one with the same peak wavelength (338 nm) but with much less intensity (56%) as compared with that of the native protein (Fig. 3A).
The conformational state of the early refolding intermediate was also analyzed by far-UV CD spectrum analysis. In this analysis, the CD ellipticities at the refolding time of 10 s after a manual mixing were determined at various wavelengths. As shown in Fig. 3B, the CD spectra at 10-s refolding were almost indistinguishable among the intermediates from the different urea-denatured proteins and displayed about 58% of the absolute value of the native CD ellipticity at 222 nm. When the same manual mixing experiment was carried out by the intrinsic tryptophan fluorescence analysis, the obtained fluorescence spectra at the refolding time of 10 s were almost indistinguishable from those of the initial burst intermediates at 5-ms refolding (data not displayed). This was reasonably accounted for by a slow rate of refolding after the initial burst intermediate formation; the most rapid refolding from the initial burst intermediate occurs for the disulfide-reduced forms (D 3.4% only at the refolding time of 10 s. The far-UV CD spectra obtained at the refolding time of 10 s (Fig. 3B) should be, therefore, essentially the same as the ones for the initial burst intermediates. The conformational characteristics of the initial burst intermediate, obtained by the fluorescence and far-UV CD spectra, were consistent with those of I N produced during the refolding from the urea-denatured, correct disulfide protein D[73-120] (16).
After the formation of the initial burst intermediate I N , all the urea-denatured proteins except for D[367-382/CM-73] slowly refolded as reflected in the increases of the fluorescence intensity and of the absolute value of CD ellipticity. At a prolonged refolding time of 20 h, the fluorescence intensity at 338 nm and the CD ellipticity at 222 nm reached 97 and 90% of the values for the native protein, respectively (Fig. 3, A and B). The refolded protein from D[367-382/CM-73], however, did not show any increase in the fluorescence intensity and CD ellipticity after the initial burst phase (Fig. 3, A and B).
When the urea-denatured proteins were refolded by dilution with an acidic buffer, they showed fluorescence spectra with 49% of the native intensity at the peak wavelength of 338 nm   (Fig. 3A). The CD spectrum of the acidic intermediate was essentially the same as that of I N , showing about 58% of the native CD ellipticity at 222 nm (Fig. 3B). The fluorescence intensity and the absolute value of the CD ellipticity did not increase after the initial burst phase in this acidic condition (data not shown). These spectral characteristics of the acidic intermediate were consistent with those of the previous equilibrium intermediate I A , produced by dilution of D[73-120] with the same acidic buffer (16).
The results from the intrinsic tryptophan fluorescence and far-UV CD spectra show that the initial burst intermediate I The conformational difference between the two intermediates of I N and I A was detected by ANS binding analyses. Fig. 3C shows the emission spectra of ANS in the presence of various states of ovalbumin. The ANS binding was almost undetectable for the native or urea-denatured ovalbumin. In contrast, greatly increased fluorescence emission with a peak at 472 nm was observed in the presence of the intermediate I A for any disulfide types. The I N state produced by refolding from D[367-382/CM-73] showed a much lower level of ANS binding. The results from the ANS binding analyses were consistent with the view that the intermediate I A is a more extended molecule with an exposed hydrophobic core than the intermediate I N .
Time Course of Refolding after the Initial Burst Phase-The time course of refolding after the initial burst phase was examined by trypsin resistance, far-UV CD, intrinsic tryptophan fluorescence, and ANS binding analyses (Fig. 4). The trypsin resistance assay is a sensitive and reliable probe for the refolding analysis, because ovalbumin is highly resistant against the protease in the native conformation (17), because of the nonbasic nature of the P1 residue (alanine residue) (7).  (Figs. 3 and 4). The integrity of native refolding was investigated more rigorously by differential scanning calorimetry. As shown in Fig. 5 and c, which are the same ones as in a previous report (16), are shown for direct comparison. The data were arbitrarily shifted on the ordinate scale for clarity. fide rearrangements by intrachain sulfhydryl/disulfide exchange reactions in the initial burst intermediate state I N as shown in Scheme 1 (16). This suggests the inclusion of intrachain sulfhydryl/disulfide exchanges in the I N state during the refolding from the urea-denatured, mispaired disulfide isomer D[367-382]. The disulfide-involved cysteines were determined at various refolding times by the peptide-mapping analysis. As shown in Fig. 6A, Cys 367 and Cys 382 , which were detected as only the disulfide-involved cysteines at the refolding time 0 decreased immediately and continuously after the initiation of the refolding; the decreases in these non-native cysteines accompanied the increases in the other four cysteines of Cys 11 , Cys 30 , Cys 73 , and Cys 120 . The disulfide-involved Cys 73 and Cys 120 both increased continuously, and their amounts were estimated to be about 80% at 20 h of the refolding. In contrast, the increases in Cys 11 and Cys 30 were only tentative, followed by gradual decreases with time of refolding. The data were consistent with the view that most, if not all, of the ureadenatured, mispaired disulfide isomer can refold to the native disulfide form through intrachain sulfhydryl/disulfide exchange reactions.
The intrachain sulfhydryl/disulfide exchanges also occurred when D[367-382/CM-73] was transferred into the refolding buffer. As shown in Fig. 6B, Cys 367 and Cys 382 detected as disulfide-involved cysteines decreased after the transfer into the refolding buffer, whereas Cys 11 , Cys 30 , and Cys 120 increased. It was confirmed that Cys 73 did not increase as a disulfide-involved cysteine because of its blocked sulfhydryl nature; this situation resulted in a limited level as disulfideinvolved cysteine for Cys 120 , which is the counterpart of the native disulfide pairing. The distribution of the various disulfide isomers reached equilibrium at 20 h; the amounts of disulfide-involved cysteines were about 47% for Cys 11 , Cys 30 , and Cys 120 and about 27% for Cys 367 and Cys 382 .
Size-exclusion Chromatography of Various States of Ovalbumin-The preceding data (Figs. [3][4][5] strongly suggest that ovalbumin refolded from D[367-382/CM-73] stays in the initial burst intermediate state with a partially folded conformation. As an alternative intermediate, ovalbumin assumes a partially folded conformation with the same far-UV CD spectrum as I N state upon dilution of the urea-denatured proteins with an acidic buffer (Fig. 3). The Stokes radii were determined by size-exclusion chromatography for various ovalbumin states, including I N and I A . As shown in Table I

DISCUSSION
The present study demonstrates the usefulness, for the analysis of a large protein, of a refolding system that includes intrachain disulfide rearrangements. A urea-denatured, mispaired disulfide isomer D[367-382] of ovalbumin was found to spontaneously refold via an initial burst intermediate I N into the native form N[73-120], as evidenced by a variety of conformational probes. More importantly, D[367-382] refolded into the state with the native thermostability, indicating the metastable refolding of ovalbumin in vitro. The obtained data, along with our previous ones (14 -18), demonstrate for the first time the serpin refolding mechanism from the fully denatured state and should provide important information about how the protein folds into a metastable state.

TABLE I Stokes radii of ovalbumin in various states
The ovalbumin samples were produced and analyzed by size-exclusion chromatography as described in the text. The times shown for I N in the parentheses represent the time periods from the 20-fold dilution of D[367-382/CM-73] to the chromatographic appearance of ovalbumin peak.  (Fig. 6) can be therefore accounted for by the occurrence of intrachain sulfhydryl/disulfide exchange reactions in the initial burst phase I N , which produces the folding competent intermediate I N  as the reverse reaction. The I N intermediate produced from D[367-382/CM-73] was found to lack the correct folding ability into the native form (Figs. 3 and 4) despite its potential to be converted into a variety of disulfide isomers. This reinforces that I N [73-120] is the only competent intermediate for the subsequent correct refolding under the disulfidebonded conditions (16).
The model shown in Fig. 7 also includes the refolding process under the disulfide-reduced conditions. In contrast to the disulfide-bonded refolding, disulfide-reduced ovalbumin can refold in a straightforward way to the native form N[SH] with essentially the same first order rate constant of 0.2 min Ϫ1 as the true folding rate constant from I N [73-120] to N[73-120] (16). This leads to the conclusion that the native disulfide Cys 73 -Cys 120 is not necessary for native refolding, but a mispaired disulfide prevents the intermediate I N from the correct refolding. In other words, Cys 73 -Cys 120 is just acceptable rather than crucial for the native conformation. In the native state, the redox conversion between N[73-120] and N[SH] is readily brought about without any conformational destruction in the presence of a reduced or oxidized form of a sulfhydryl reagent (17).
Refolding analyses under acidic conditions also provide useful information. When urea-denatured ovalbumin (D[73-120]) is diluted with an acidic buffer, pH 2, it is transformed into a partially folded conformational state I A that has the same far-UV CD spectrum as the I N state. Native ovalbumin retains its ordered conformational characteristics at this pH value (N A form in Fig. 7); this form displays the native far-UV CD spectrum, a highly resistant nature against pepsin, and a clear endothermic transition on differential scanning calorimetry (18,29). Under the acidic conditions, complex sulfhydryl/disulfide exchange reactions are completely blocked. The intermediate I A [73-120], however, cannot refold directly into N A [73-120] unless it is transferred to a near neutral pH buffer where the protein refolds correctly through the I N state (16). The absence of correct refolding under the acidic conditions is also found in the disulfide-reduced form of ovalbumin (16).
Conformational  (Figs. 3 and 4). The stable nature of this I N state enabled us to analyze the conformational state by a variety of probes. The intermediate I N displayed partial regaining (about 58% of the ␣-helix content) of the native far-UV CD spectrum (Fig. 3), an accessible nature against trypsin (Fig. 4), some ANS binding capacity (Figs. 3 and 4), and flexibility detected by the occurrence of the intrachain sulfhydryl/disulfide exchanges (Fig. 6). These characteristics were consistent with a non-native conformational state for the I N state. Nevertheless, the intermediate I N was a compact molecule, similar to the native form, as evaluated by size-exclusion chromatography (Table I).
Alternative stable intermediate states I A produced from D[73-120] and D [SH] have been shown to display a common far-UV CD spectrum to the I N states from the same ureadenatured ovalbumin forms (16). In the present study, the same I A state was confirmed to be produced by dilution of D displayed the same far-UV CD spectrum (Fig. 3). These indicate that the secondary structure contents in the I N and I A states are independent of the differences in the disulfide pairings and pH conditions. The secondary structure may be, therefore, formed in the initial burst intermediate, depending on the helix-forming propensity of local peptide segments, without the help of non-local native interactions. Indeed, the regaining of the native far-UV CD ellipticity in the I N state (58% of the CD ellipticity at 222 nm) is very similar to the percentage (55%) of the value, predicted on the basis of helix-forming propensities of individual amino acids (30), relative to the ␣-helix content in the crystal structure (11). The I A intermediate, however, cannot refold directly into the native-like conformer N A under the acidic conditions. This may be closely related to the findings that the I A state is a molecular state with a more extended Stokes radius and more exposed hydrophobic core than the I N intermediate (Table I and Fig. 3). These conformational properties of I A are similar to that of an intermediate called the "pre-molten globule state," which has partially folded secondary structures and somehow extended conformation with exposed hydrophobic core and disordered long-range interactions (31).
The native refolding after the initial burst phase should include the reorganization of the preformed secondary structure by which the formations of all the non-local, native interactions are attained. The main body of a serpin comprises the ␣-helix and ␤-structure domains. Unlike usual multidomain proteins, the serpin domains are closely packed with each other, and many of the ␣-helix and ␤-strand elements reside almost alternately on the primary structure. This unique domain structure might be related to the metastable nature of native serpin. That the intermediate I A with extended conformation cannot undergo the correct folding into the N A state strongly suggests that the partially formed secondary structure elements in the initial burst phase must be reorganized with strong mutual interactions in a compact conformational state of I N that enables subsequent structural reorganization into the metastable native state.
Comparison with the Refolding Mechanism of Other Proteins-The previous experimental results for the native refolding of small single-domain proteins have been consistent with the presence of a specific pathway, such as the two-or multistate transition pathways (32)(33)(34), although theoretical analyses have proposed a new mechanism that includes protein folding through undefined multiple pathways within a funnellike energy landscape (35,36). Unlike the extensive experimental evidence for small single-domain proteins, the folding mechanism has been poorly understood for large complex proteins. This is due, at least in part, to the difficulty for application of powerful methods for protein folding analysis, such as nuclear magnetic resonance.
As a general approach for a variety size of proteins, the oxidative refolding systems of disulfide proteins have been widely employed. The refolding processes of disulfide proteins may be categorized into three classes depending on the conformational state of disulfide-reduced forms. As first class proteins, most small single-domain proteins, such as bovine pancreatic trypsin inhibitor (37), lysozyme (38), and ribonuclease A (39), essentially assume unfolded conformation in the disulfidereduced state. In the most deeply investigated example of bovine pancreatic trypsin inhibitor, the refolding proceeds via a well defined pathway in which partially disulfide-bonded species with unique conformation are included as folding intermediates (37,40,41). A serious problem has, however, been pointed out for this oxidative refolding system. The first step for disulfide regeneration is the intermolecular attack of an oxidizing disulfide agent onto protein sulfhydryls; the differential accessibility of the protein sulfhydryls, due to the conformational situations of refolding intermediates, may perturb the true kinetic pathway of the refolding (22,23).
The second class of disulfide proteins, such as the constant fragment of immunoglobulin light chain (22) and ribonuclease T 1 in the presence of a high salt concentration (42), assumes a native-like conformation in their disulfide-reduced state. The third class of the proteins includes ovotransferrin (43)(44)(45) and serum albumin (46), which assume partially folded conformations in the disulfide-reduced forms. The urea-denatured forms of these two classes of proteins refold into the fully or partially folded conformation under disulfide-reduced conditions. Subsequent regeneration of the correct intrachain disulfide(s) stabilizes or improves the pre-formed conformations. It is therefore difficult to obtain information about the overall refolding pathway directly from the disulfide regeneration processes of the second and third classes of disulfide proteins.
In the present study, ovalbumin, which is categorized into the second class of disulfide proteins, was analyzed by a differential way. The refolding process involved regaining the native disulfide Cys 73 -Cys 120 from a non-native pairing of D[367-382] by intrachain sulfhydryl/disulfide exchange reactions without the help of any added chemical oxidant. The obtained refolding mechanism includes multiple intermediates that undergo extensive interconversion (Fig. 7). Such a refolding process is considered to be more adequately related to a funnel-like model than a pathway model with a small number of defined intermediates, although the quantitative description by an energy landscape cannot be done at present. The use of D[367-382/ CM-73] as the starting denatured state enabled us to analyze the structural characteristics of the initial burst intermediate as a stable form. The obtained results highlighted the unique structural features of I N that are characterized by a compact conformation and a partial regain of the ␣-helix content that is probably formed by the propensity of a local primary structure. These structural characteristics of I N are very similar to the partially folded intermediate of ␤-lactoglobulin, which is predominantly ␤-sheet protein. A refolding intermediate and equilibrium molten globule state of this protein assume a very compact conformation (47,48), and the polypeptide segment with a good helix propensity takes a non-native helix in an early refolding intermediate (49,50).