Structural and Functional Characteristics of Partially Disulfide-reduced Intermediates of Ovotransferrin N Lobe CYSTINE LOCALIZATION BY INDIRECT END-LABELING APPROACH AND IMPLICATIONS FOR THE REDUCTION PATHWAY*

Ovotransferrin N lobe contains six intrachain disul- fides (SS-I/Cys 10 –Cys 45 ; SS-II/Cys 20 –Cys 36 ; SS-III/Cys 115 – Cys 197 ; SS-IV/Cys 160 –Cys 174 ; SS-V/Cys 171 –Cys 182 ; SS-VI/ Cys 228 –Cys 242 ) in a single polypeptide chain of 332 amino acid residues. Upon the protein disulfide reduction with dithiothreitol under nondenaturing conditions, the intermediate species with four, three, and two disulfides were generated. The partially disulfide-reduced intermediates were isolated, and the localization of intact disulfides in the intermediates was determined by an indirect end-labeling method. This method included the S -cyanocysteine-specific protein fragmentation, fol-lowed by gel electrophoresis and the immunochemical visualization of the C terminus-intact fragments using antiserum raised against a non-cysteine C-terminal fragment (Ser 280 –Arg 332 ). Results clearly showed that first SS-IV and SS-V, second SS-III, and then SS-VI are cleaved. No reduction was observed for SS-I and SS-II under the employed reducing conditions. The conclu- sion was confirmed by peptide mapping analyses for the same disulfide intermediates using reverse phase high performance liquid chromatography. Transverse urea gradient gel electrophoresis and visible absorption spectra revealed that the four-disulfide intermediate, but not the three- or two-disulfide intermediate, retains essentially the same iron-binding function as the native protein. By far-UV CD analyses, the residual native con-

The redox folding/unfolding system of a disulfide protein is a useful model for the investigation of protein folding mechanism, because all disulfide intermediates may be trapped in stable forms (Creighton, 1986). In bovine pancreatic trypsin inhibitor of the most well documented example, the predominant folding/unfolding pathway includes intrachain disulfide rearrangements that inevitably require the formation of nonnative disulfide intermediates (Creighton, 1977;Creighton and Goldenberg, 1984;Weissman and Kim, 1992). A similar nonstraightforward regeneration/reduction pathway has been shown for many small single-domain proteins, such as lysozyme (Acharya and Taniuchi, 1980), ribonuclease A (Rothwarf and Scheraga, 1993), ribonuclease T1 (Pace and Creighton, 1986), ␣-lactalbumin (Ewbank and Creighton, 1993a), and hirudin (Chatrenet and Chang, 1993). To quantitatively elucidate the kinetic pathways, the structural characteristics of partially disulfide-bonded intermediates have been extensively studied with these small single-domain proteins (Pace et al., 1988;Radford et al., 1991;Darby et al., 1992;Staley and Kim, 1992;Ewbank and Creighton, 1993b;van Mierlo et al., 1994;Talluri et al., 1994). As a useful model for the multidomain protein, we have been interested in the folding/unfolding problem of ovotransferrin N lobe. Ovotransferrin consists of a single polypeptide chain with a molecular mass of about 80 kDa, which is folded to give N and C lobes; each of the lobes has one specific iron-binding site (Aisen and Listowsky, 1980). Either lobe can be isolated after mild proteolysis , and the x-ray crystallographic structure of the N lobe has recently been established (Dewan et al., 1993). As shown in Fig. 1, the N lobe of ovotransferrin consists of a single polypeptide chain of 332 amino acid residues, which is folded into two domains (domains 1 and 2); Domain 1 contains two disulfide bonds in the N-terminal ␤␣␤␣ motif (SS-I/Cys 10 -Cys 45 ; SS-II/Cys 20 -Cys 36 ), and domain 2 contains three disulfides in the kringle bridges (SS-III/Cys 115 -Cys 197 ; SS-IV/Cys 160 -Cys 174 ; SS-V/Cys 171 -Cys 182 ) and one disulfide in the C-terminal loop of this domain (SS-VI/Cys 228 -Cys 242 ). The structural and functional renaturation of the isolated N lobe has been shown to be attained by a two-step procedure, in which the denatured, reduced proteins are initially transformed into a partially folded, reduced state and are then reoxidized into the native, disulfide-bonded form (Hirose et al., 1989;Hirose and Yamashita, 1991;Yamashita and Hirose, 1993). The structural and functional characteristics of the disulfide intermediates involved in the redox folding/unfolding pathway, however, have still not been determined.
One of the major difficulties on studying the redox folding/ unfolding pathway of a complex protein is related to the analytical approach for the disulfide localization in the partially disulfide-bonded intermediates. The established method of the peptide mapping procedure that includes extensive protein digestion and subsequent identification of disulfide peptides is highly complex and time-consuming for the analysis of a large protein having many disulfides. A previous chemical fragmentation approach that includes specific polypeptide cleavages in S-cyanocysteine sites and subsequent SDS-PAGE 1 has been * This work was supported in part by a grant-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan. 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.
‡ To whom correspondence should be addressed. shown to be useful for the simple and sensitive analysis of the localization of half-cystines (Jacobson et al., 1973;Subramanian, 1980;Walser et al., 1981;Mikami and Morita, 1983;Nefsky and Bretscher, 1989). Practical application of the technique, however, has been limited to proteins with a small number of half-cystines.
In the present study, we developed an indirect end-labeling procedure that enables, in combination with the S-cyanocysteine-specific fragmentation technique, the localization of sulfhydryls and disulfides in a complex protein with many disulfides. The reliability of the indirect end-labeling procedure was confirmed in such a way that with respect to the partially disulfide-reduced intermediates of ovotransferrin N lobe, the sulfhydryl and disulfide localizations determined by this procedure were essentially the same as those determined by the established peptide mapping analysis. The structural and functional characteristics of the partially disulfide-reduced intermediates were also investigated using different analyses. The implications of the disulfide intermediates for the disulfidereduction pathway are discussed on the basis of their structural characteristics.

EXPERIMENTAL PROCEDURES
Materials-Ovotransferrin N lobe (the N-terminal half-molecule, Ala 1 -Arg 332 ) and the two fragments corresponding to Ala 1 -Lys 279 (N fragment) and Ser 280 -Arg 332 (C fragment) were prepared as described previously Kurokawa et al., 1994). Antiserum for the C fragment was raised in rabbits by injecting the protein emulsified with complete Freund's adjuvant subcutaneously. We confirmed that the antiserum specifically reacts with Ser 280 -Arg 332 segment, but not with any other part in the ovotransferrin N lobe; either the intact N lobe or the C fragment, but not the N fragment, was immunochemically detected using the antiserum after SDS-PAGE. Antiserum for whole ovotransferrin (Ala 1 -Lys 686 ) was prepared in the same way. Anti-IgG alkaline phosphatase conjugate, nitro blue tetrazolium, and 5-bromo-4-chloro-3-indolyl phosphate were obtained from Promega. SBD-F was obtained from Dojindo. Trypsin was purchased from Sigma (type XI, treated with diphenylcarbamyl chloride). Other chemicals were guaranteed grade from Nacalai Tesque.
Localization of Sulfhydryls and Disulfides by Indirect End Labeling-As shown in Fig. 2A, none of 12 half-cystines are included in the short C fragment. Thus, antiserum for the C fragment was employed in the present indirect end-labeling approach. FIG. 1. Schematic view of the native structure of ovotransferrin N lobe. The figure was drawn based on the x-ray crystallographic data of iron-loaded ovotransferrin N lobe (Dewan et al., 1993) using the program MolScript (Kraulis, 1991). The upper and lower halves correspond to domains 1 and 2, respectively. N and C represent the N-terminal and C-terminal ends, respectively.  In Panel A, the localizations of the six native disulfide bonds and the C fragment (Ser 280 -Arg 332 ) that was employed for the probe antiserum are shown. Panel B represents the strategy for determining the sulfhydryl and disulfide localizations in ovotransferrin N lobe.
Step 1 includes disulfide reduction with DTT under nondenaturing conditions, alkylation of the generated sulfhydryls with IAA yielding alkylated cysteines (A-), and then isolation of disulfide intermediates. The figure shows the case for the hypothetical two-disulfide intermediate having SS-III.
Step 2 is the cyanylation procedure, in which the intact disulfides are fully reduced by DTT in the presence of urea. The newly generated sulfhydryls are modified with DTNB and subsequently S-cyanylated (CN-). In step 3, the protein is fragmented by S-cyanocysteine-specific cleavage at pH 9.0, electrophoresed on a SDS-polyacrylamide gel, and visualized by the Western blotting technique using the antiserum specific for the C fragment. In the standard run, the native protein is analyzed in the same way except that step 1 is skipped. By the S-cyanocysteine specificfragmentation, 91 different polypeptide species should be generated. However, only the 13 species (12 fragments of F1-F12 and one intact molecule, Int) that retain the C-terminal segment of the intact protein should be detected by the Western blotting analysis. On the basis of the molecular size of the fragments, the half-cystine sites can be directly mapped from the C terminus of the intact N lobe. In the sample run, only the intact protein and the fragments that come from the cleavages at the disulfide-involved half-cystine sites in a disulfide intermediate should be detected. The fragments that are detected in the standard run, but not in the sample run, correspond to those coming from the cleavages at the free sulfhydryls in the disulfide intermediate. The term of "indirect end labeling" was employed after Wu (1980) who mapped the nuclease-sensitive sites in nuclear chromatin DNA as probed with a cloned radiolabeled DNA segment. in ovotransferrin N lobe by an indirect end-labeling approach, in combination with the S-cyanocysteine-specific fragmentation.
In step 1, protein disulfides were reduced under nondenaturing conditions by incubating the isolated ovotransferrin N lobe at 0.2 mg/ml and 6°C with 10 or 30 mM DTT for various times in buffer A (0.1 M Tris-HCl buffer, pH 8.2, 0.1 mM Na-EDTA). The buffer was degassed at reduced pressure and equilibrated under a N 2 atmosphere prior to the reduction experiments. Immediately after initiating the protein reduction, sample vials were placed under a stream of N 2 and then sealed with a screw cap. The protein was alkylated with 0.1 M IAA at 6°C for 5 min and then at 37°C for 5 min. The number of intact disulfide bonds was determined by the PAGE technique as described elsewhere . For the isolation of [3S-S] and [4S-S], the native protein was reduced with 10 mM DTT for 50 min in a total volume of 30 ml and alkylated with IAA in the same way. The sample was concentrated using a concentrator (Amicon, Centriprep-30), passed through a Sephadex column (NAP-10, Pharmacia Biotech Inc.) equilibrated with buffer B (50 mM sodium acetate buffer, pH 5.5, containing 8 M urea), and applied to an ion-exchange column (Mono S HR 5/5, Pharmacia) equilibrated with buffer B. Proteins were eluted by a linear 20-ml gradient of 0.05 to 0.9 M sodium acetate buffer, pH 5.5, in the presence of 8 M urea; peak fractions as monitored by absorbance at 280 nm were collected and analyzed for the number of disulfides by the same PAGE technique. Protein samples containing [3S-S] and [4S-S] were applied to a reverse-phase column (Cosmosil 5C18 AR-300, 4.6 ϫ 150 mm), eluted by a linear 20 ml gradient of 30 to 60% acetonitrile in 0.1% trifluoroacetic acid, lyophilized, and stored at Ϫ20°C until use for further analyses. [2S-S] was isolated in the same way after incubating the native half-molecule with 30 mM DTT for 5 h.
In step 2, the cysteine-specific cyanylation was carried out essentially as described by Nefsky and Bretscher (1989). Briefly, the isolated disulfide intermediates were fully reduced at 1.0 mg/ml by incubation with 3 mM DTT at 37°C for 20 min in buffer A containing 9 M urea in a total volume of 0.1 ml. The sample was mixed with 0.5 volume of 30 mM DTNB dissolved in 0.1 M sodium phosphate buffer, pH 7.5, incubated for 15 min at 25°C, and then passed through a Sephadex G-25 column (NAP-5, Pharmacia) equilibrated with 0.1 M Tris-HCl buffer, pH 7.5. The modified protein was applied to a reverse phase column (Cosmosil 5C18-AR-300) equilibrated with 0.1% trifluoroacetic acid, 30% acetonitrile, eluted by a linear 30 to 60% gradient of acetonitrile, and lyophilized. The protein modified with DTNB was dissolved in 50 l of cleavage buffer (0.1 M sodium borate, 0.5 M glycylglycine, 8 M urea, pH 9.0) and mixed with 0.5 l of 0.1 M KCN.
In step 3, the S-cyanylated protein was fragmented by incubation at 40°C. At different times, 10 l of the sample was withdrawn and the fragmentation was terminated by addition of one volume of 0.2 M DTT. The sample was analyzed by SDS-PAGE using the standard buffer system as described by Laemmli (1970). Polyacrylamide gels (13 ϫ 13 ϫ 0.075 cm) consisted of 15% acrylamide and 0.08% bisacrylamide. Protein fragments were electroblotted onto a polyvinylidene difluoride membrane and detected by an immunochemical procedure described by Sambrook et al. (1989), using the antiserum raised against the C fragment (Ser 280 -Arg 332 ), anti-IgG alkaline phosphatase conjugate, nitro blue tetrazolium, and 5-bromo-4-chloro-3-indolyl phosphate.
Analysis for the Localization of Disulfide Bonds by Peptide Mapping Analyses-The disulfide localization in the disulfide intermediates was also determined by analyzing the intact disulfide bonds by the peptide mapping procedure established in a previous report (Yamashita and Hirose, 1993). Briefly, [3S-S] and [2S-S] isolated as in the step 1 in Fig.  2B, along with the native protein control, were extensively proteolyzed with trypsin, and the trypsin digests were applied to a reverse phase HPLC column (YMC AP-302, ODS 4.6 ϫ 150 mm, pore size, 30 nm). Peptides were eluted with an acetonitrile linear gradient (0 -60%) in 0.1% trifluoroacetic acid. A part of each peak was assayed for disulfides using a fluorescent reagent, SBD-F. Disulfide-positive peaks were further purified by rechromatography on the same column but with a different buffer system (0 -60% acetonitrile gradient in 10 mM triethylamine, acetic acid buffer, pH 5.0). Purified disulfide peptides were analyzed for their amino acid compositions with an amino acid analyzer (Hitachi, model 835-30) and primary sequences with a gas phase protein sequenator (Applied Biosystems, model 477A/120A).
Transverse Urea Gradient Gel Electrophoresis-Transverse linear urea gradient polyacrylamide gels were prepared essentially as described by Goldenberg and Creighton (1984). The linear gradient was from 0 to 8 M urea, and the slab gels (8.0 ϫ 5.0 ϫ 0.1 cm) consisted of 6% acrylamide and 0.16% bisacrylamide. Both the gel and tray buffers consisted of 0.1 M Tris-0.03 M borate, pH 8.6, 15 mM oxalate, and 0.1 mM Na-EDTA as shown previously (Hirose et al., 1989).
One nmol (36 g) of the native protein or either of the isolated disulfide intermediates was dissolved in 35 l of the tray buffer, mixed with 2 l (0.5 nmol iron) of 0.25 mM iron-nitrilotriacetate (1:1 mixture of FeCl 3 and the chelating agent), so that half of protein molecule can be saturated with the metal. The sample was mixed with 0.33 volume of the tray buffer containing 50% glycerol and applied to the top of the gel. Electrophoresis was carried out at 4°C and a constant voltage of 100 V for 2 h. The gel was stained with Coomassie Blue.
CD Analysis-The native and isolated disulfide intermediates were dissolved at 60 g/ml in buffer A in the presence or absence of 30 M of the iron-nitrilotriacetate complex containing 0.7 mM oxalate. The CD spectra of the proteins were recorded at 6°C with a spectropolarimeter (Jasco J-501C) using a 0.2-cm cuvette.
Analyses for Iron Binding by Visible Absorption Spectrum-The visible absorption spectrum was determined essentially in the same way as described previously (Kurokawa et al., 1994). Briefly, the native protein, [3S-S], and [4S-S] were dissolved in buffer A containing 10 M of the iron-nitrilotriacetate complex and 25 mM sodium bicarbonate. The samples were loaded onto a Sephadex G-25 column (NAP-10, Pharmacia) equilibrated with buffer A containing 25 mM sodium bicarbonate to remove excess iron-nitrilotriacetate. The visible absorption spectra of the iron-loaded proteins were recorded at a protein concentration of 0.6 mg/ml with a Shimadzu UV-240 spectrometer using a cuvette with a 1-cm path length.

Disulfide Reduction under Nondenaturing Conditions and
Isolation of Disulfide Intermediates-As in step 1 in Fig. 2B, the native protein was incubated with 10 or 30 mM DTT for various times under nondenaturing conditions, alkylated with IAA, and then analyzed for the reduced number of disulfide bonds by the PAGE technique. As shown in Fig. 3, [4S-S] was the disulfide intermediate that was first generated in 30 mM DTT at an incubation time as short as 15 min; little fivedisulfide species was detected. At 30 min of incubation, [3S-S] was generated, and after 50 min of incubation the production of [2S-S] was observed. Fig. 3 also shows that two disulfides are intact at a prolonged incubation time of 24 h (lane 11). The two disulfides were still resistant against higher DTT concentrations of 75 and 150 mM for the same 24-h incubation (data not shown). Essentially the same profile for generations of the partially disulfide-reduced intermediates was observed at a lower DTT concentration of 10 mM, except that the reduction rate was slower; nor was any five-disulfide intermediate detected during the incubation with 10 mM DTT (data not shown).
To isolate the partially disulfide-reduced intermediates, the native protein was reduced with 10 mM DTT for 50 min, alkylated with iodoacetic acid, and then fractionated by ion-exchange column chromatography. As shown in Fig. 4, the protein was mainly separated into three peaks labeled 1, 2, and 3. Analyses for the number of disulfides by the PAGE technique revealed that the peaks 1, 2, and 3 correspond, respectively, to Localization of Disulfide Bonds in the Disulfide Intermediates by Indirect End-labeling Approach-As the standard run in Fig. 2B, it was examined using the native protein having six disulfides whether or not the polypeptide species that retain the C terminus of the intact protein can be specifically detected by the indirect end-labeling technique. The native protein was fully reduced with DTT, modified with DTNB, cyanylated, fragmented, electrophoresed, and then visualized immunochemically. Although 91 different polypeptide species should be produced from the protein having 12 S-cyanocysteine sites, only 13 bands should be detected, if the indirect end labeling works correctly. Fig. 5 demonstrates that this was indeed the case. The band number was restricted to 13 (one intact form plus 12 fragments of F1-F12) using the C-terminal antiserum (Fig. 5, A and C), while many other bands along with the 13 bands were detected using the whole ovotransferrin antiserum (data not shown). The molecular sizes of the 12 fragments estimated from their migrations were 35.0, 34.0, 31.0, 30.0, 25.0, 21.0, 19.0, 17.5, 17.0, 15.5, 11.7, and 9.8 kDa, respectively, for F1-F12. These values were consistent within experimental error with the theoretical ones calculated assuming that these fragments come from the specific cleavages at the S-cyanocysteine sites but retains the C terminus of the intact protein; the theoretical molecular weights are 35,442, 34,340, 32,456, 31,435, 24,055, 19,309, 18,170, 17,783, 16,923, 15,341, 11,918, and 10,342, for F1-F12, respectively. In addition, as shown in Fig. 5B, no fragmentation occurred when the fully reduced protein was alkylated with IAA prior to modification with DTNB, indicating that the fragmentation is specific to halfcystine sites and that the cleavage can be completely blocked by cysteine alkylation with IAA. We, therefore, concluded that F1-F12 fragments come, respectively, from the cleavages at the N-terminal sites of Cys 10 , Cys 20 , Cys 36 , Cys 45 , Cys 115 , Cys 160 , Cys 171 , Cys 174 , Cys 182 , Cys 197 , Cys 228 , and Cys 242 .  Fig. 2B. Fig. 6 clearly shows that, for either [4S-S], [3S-S], or [2S-S], the number of the detected fragments was exactly consistent with the one expected from the number of the intact disulfide bonds. The number of the fragment bands was eight for [4S-S]; the fragment bands corresponding to SS-I, SS-II, SS-III, and SS-VI were clearly detected, but no other fragments corresponding to SS-IV and SS-V were produced (Fig. 6C). For [3S-S], only the six fragments that corresponded to SS-I, SS-II, and SS-VI were detected (Fig. 6B). When [2S-S] was analyzed, only the four fragment bands corresponding to SS-I and SS-II were detected. These results were consistent with the idea that the disulfide cleavage occurs first at SS-IV and SS-V, second at SS-III, and then at SS-VI. Both SS-I and SS-II are resistant to DTT under the employed reducing conditions. Localization of Disulfide Bonds by Peptide Mapping Analyses-To investigate the reliability of the above conclusion FIG. 4. Isolation of partially disulfide-reduced intermediates. The native protein was reduced at 0.2 mg/ml with 10 mM DTT in buffer A at 6°C for 50 min in a total volume of 30 ml. Disulfide-reduced proteins were alkylated with IAA, concentrated, passed through a Sephadex column, applied to an ion-exchange column, and eluted as described in the text. The same PAGE analysis as in Fig. 3 revealed that the peaks 1, 2, and 3 correspond to [3S-S], [4S-S], and the six-disulfide native protein, respectively.

FIG. 5. Indirect end-labeling visualization for the six-disulfide protein.
In Panel A, the six intrachain disulfides in the native ovotransferrin N lobe were all reduced with DTT in the presence of 9 M urea. The protein was modified with DTNB and then cyanylated. In Panel B, the native protein was processed in the same way as in Panel A, except that disulfide-reduced protein was alkylated with IAA prior to the modification with DTNB. The protein was subjected to fragmentation at pH 9.  6). SDS-PAGE, electroblot onto a membrane, and immunochemical detection using the C fragment-specific antiserum were carried out as described in the text. Roman numerals on the right side represent the corresponding native disulfide bonds. drawn by the indirect end-labeling method, we analyzed the localization of intact disulfide bonds in the isolated disulfide intermediates by the peptide mapping procedure. In our previous report (Yamashita and Hirose, 1993), the peptide mapping for all six native disulfides in the ovotransferrin N lobe was established. In this report, we have obtained the same conclusion with regard to [4S-S]; this intermediate specifically lacks SS-IV and SS-V as determined by the peptide mapping analysis. In the present study, therefore, the peptide mapping analysis was carried out with the disulfide intermediates of [3S-S] and [2S-S].
According to our peptide mapping procedure (Yamashita and Hirose, 1993), the extensive tryptic digestion of the native half-molecule yields, on reverse-phase HPLC, nine disulfide peaks denoted A-I. By the rechromatography, peaks C and I, but not the other peaks, are separated into two peaks: peak C, into peaks C-1 and C-2, and peak I, into peaks I-1 and I-2. The peaks, A, B, C-1, C-2, D, E, F, G, H, I-1, and I-2, respectively, correspond to SS-V, SS-VI, SS-II, SS-III, SS-II, SS-III, SS-IV, SS-III, SS-VI, SS-VI, and SS-I.
The isolated disulfide intermediates of [3S-S] and [2S-S] along with the native protein control were analyzed by the same peptide mapping procedure. We reconfirmed by analyzing amino acid sequences and amino acid compositions that the 11 disulfide peptide peaks were all produced from the native protein control. For [3S-S], however, the disulfide peptides (A, C-2, E, F, and G) corresponding to SS-III, IV, and V were missing, while the occurrence of all the other disulfide peptides (B, C-1, D, H, I-1, and I-2) corresponding to SS-I, SS-II, and SS-VI was confirmed. For [2S-S], the occurrence of the peptides (C-1, D, and I-2) corresponding to SS-I and SS-II was confirmed, but no other disulfide peptides were detected. All disulfide peptides from [3S-S] and [2S-S] were analyzed for amino acid compositions, and peptide recoveries were compared between the partially disulfide-reduced proteins and the native protein control. As summarized in Table I, the recoveries of SS-I and SS-II from [3S-S] were more than 93% of those from the native protein, although the recovery of SS-VI showed a slightly lower value of 72.8%. The recoveries of SS-I and SS-II from [2S-S] were also greater than 94% of those of the native protein control. These data from the peptide mapping analyses along with previous results (Yamashita and Hirose, 1993) led us to the same conclusion as the one obtained by the indirect end-labeling approach: the disulfide reduction in ovotransferrin N lobe occurs first at SS-IV and SS-V, second at SS-III, and then at SS-VI. Both SS-I and SS-II are highly resistant against reduction with DTT.

Protein Stability and Iron Binding Capacity as Evaluated by
Urea Gradient PAGE-The structural and functional characteristics of the disulfide intermediates were investigated by transverse linear urea gradient PAGE , which enables the analyses both for the protein stability and the iron binding capacity of transferrins (Evans and Williams, 1980). As shown in Fig. 7A, the native protein that had been partially saturated with iron (iron to protein molar ratio of 0.5) migrated in a profile displaying the occurrence of two protein forms; one corresponded to the ironfree form showing urea-induced denaturation transition, and the other corresponded to the iron-bound form having marked stability against urea concentration as high as 8 M. The urea concentration giving the midpoint of the denaturation transition was 4.0 M for the iron-free form. An apparent single band in low urea concentrations may indicate that global native conformation is almost indistinguishable between the iron-free and loaded forms. When [4S-S] was analyzed in the same way, the protein was separated into two distinct bands (Fig. 7B); one band displayed urea-induced denaturation transition, and the other showed no denaturation in urea concentrations from 0 to 8 M. Protein species corresponding to the urea-resistant band can be again accounted for by the iron-bound form. The urea concentration for the midpoint of the denaturation was the same 4.0 M for the iron-free form.
In contrast to the native protein and [4S-S], either [3S-S] or [2S-S] that had been incubated with iron at metal to protein molar ratio of 0.5 showed a single band, suggesting the absence of iron-binding capacity. Both [3S-S] and [2S-S] showed some conformational transition by urea; the urea concentration giving the midpoint of the denaturation was 3.1 M for [3S-S] and 2.6 M for [2S-S] (Figs. 7, C and D).
CD Spectrum-The urea gradient PAGE data indicate that all disulfide intermediates retain some folded conformation. To examine this by a different approach, we determined far-UV CD spectra in the presence and absence of iron. As shown in Fig. 8A, [2S-S], [3S-S], and [4S-S] were found to contain less secondary structure than the native form in the absence of iron; the residual native conformation was decreased with the increased number of the reduced disulfides. All the partially disulfide-reduced proteins, however, appeared to retain some folded conformation, since their CD spectra were quite different from the spectrum of the urea-denatured form.  As shown in Fig. 8B, [3S-S] and [2S-S] appeared, also in the presence of iron, to contain significantly lower secondary structures than the native protein. When the spectrum data were compared between the presence and absence of iron, the CD profiles were almost indistinguishable for these protein species. This may be related to the absence of iron-binding capacity in either [3S-S] or [2S-S] (see Fig. 7, C and D). The disulfide intermediate, [4S-S], however, displayed significantly different CD spectra in the presence and absence of iron; the CD spectrum of this intermediate was very similar to that of the native protein in the presence of iron. This disulfide intermediate may, therefore, undergo some conformational change upon the iron binding in such a way that partially unfolded state is transformed into the native-like conformation.
Visible Absorption Spectrum-The preceding data of urea gradient PAGE and CD spectra indicate that [4S-S], but not [3S-S] or [2S-S], retains the iron-binding capacity. To investigate the iron-binding capacity more directly, we analyzed the visible absorption spectrum in the presence of iron. As shown in Fig. 9, the visible absorption spectrum of [4S-S] was very similar to that of the native protein; maximum absorption intensity was almost exactly the same for the two protein species, although the wavelength for the maximum absorption was slightly different (the native protein, 465 nm; [4S-S], 460 nm). The disulfide intermediate of [3S-S], however, showed no visible absorption peak at around 460 nm. We, therefore, concluded that [4S-S], but not [3S-S], retains essentially the same iron-binding capacity as the native protein.

DISCUSSION
The chemical fragmentation method that includes specific polypeptide cleavages in S-cyanocysteine sites and subsequent SDS-PAGE has been shown to be potentially useful for the simple and sensitive analysis of the localization of half-cystines in a protein (Jacobson et al., 1973;Subramanian, 1980;Walser et al., 1981;Mikami and Morita, 1983). The method has, however, yielded only indirect evidence about the localization without the use of a specific end-labeling procedure. Matsudaira et al. (1985) first attempted to localize the half-cysteine sites in villin by the immunoblot analysis of the S-cyanocysteine cleavage products using antiserum specific for the terminal sequence peptides. However, it has not been clear because of the use of a protein with unknown chemical structure whether their approach works as a general method for localizing half-cystine sites in a protein; the S-cyanocysteine cleavage data have shown the presence of 6 half-cystines in villin, while the data from the radiolabeled alkylation and from optical determinations using DTNB have shown the occurrence of 2.6 -3.9 half-cystines. Furthermore, their procedure has not included the way to differentiate cystine and cysteine sites. The data in our present report demonstrate that the indirect end labeling reliably works for the cystine and cysteine localizations in a complex protein, such as ovotransferrin N lobe with 12 halfcystines, providing a promising example for its general application for a variety of sulfhydryl and disulfide proteins.
The indirect end-labeling method has some advantages over previous "landmark mapping" (Nefsky and Bretscher, 1989), in which the carbon atom of 14 CN is introduced, upon the Scyanocysteine-specific fragmentation reaction, into the N termini of all the newly generated fragments. First, the number of the fragment bands detected in SDS-PAGE by a nonspecific staining should be (p ϩ 1)(p ϩ 2)/2 (91 bands for ovotransferrin N lobe) for a protein having p of the number of half-cystines, while that detected by autoradiography should be p(p ϩ 1)/2 (78 bands for the N lobe). The difference in these numbers, p ϩ 1, corresponds to the number of the bands that are assigned to the fragments retaining the N terminus of the original protein. This complex procedure may hamper the clear mapping of S-cyanocysteine sites for a protein having many disulfides. In the indirect end-labeling method, however, only (p ϩ 1) bands that retain the N or C terminus of intact protein should be directly detected. Second, cellular protein folding is one of the most important subjects in modern bioscience. A disulfide protein in which all disulfide intermediates may be trapped in stable forms (Creighton, 1986) is a useful model for investigating in vivo protein folding. The potential usefulness for such a nonpurified protein system appears to be an alternative advantage of the indirect end-labeling approach, since this technique includes specific protein visualization with high sensitivity after gel electrophoresis.
Data from the optical and PAGE analyses of the disulfide intermediates provide important information about structural and functional roles of the disulfide bonds in ovotransferrin N lobe. One of the most striking feature is that [4S-S] still retains the iron-binding function (Figs. 7 and 9). The observation unambiguously proves that SS-IV (Cys 160 -Cys 174 ) and SS-V (Cys 171 -Cys 182 ) that are both strictly conserved in N and C lobes of ovotransferrin, lactoferrin, and serum transferrin (Williams, 1982) are not essential for the iron-binding function. Another interesting observation for [4S-S] is the lower mobil- ity, in a low urea concentration range (from 0 to about 2.5 M urea) on the transverse urea gradient PAGE, for the iron-free form than for the iron-loaded form of [4S-S] (Fig. 7B). This may be explained as follows. As observed with the far-UV CD spectrum, [4S-S] may undergo some conformational change upon the iron binding in such a way that partially unfolded state is transformed into the native-like conformation (Fig. 8). The lower mobility of the iron-free form can, therefore, be accounted for, at least in part, by the partially unfolded conformation. In addition, the iron-dependent conformational change might be closely related to the amino acid residues that are involved in the disulfide loop of SS-III (Cys 115 -Cys 197 ); in the absence of iron, the loop might be more exposed to protein environment in [4S-S] than in the native protein, since [4S-S] lacks both SS-IV (Cys 160 -Cys 174 ) and SS-V (Cys 171 -Cys 182 ). The loop, however, should be less exposed to protein environment, since it includes an iron-coordinating ligand (Tyr 191 ) and the synergistic anionbinding ligand (Arg 121 ) that are both localized in the interior of the interdomain cleft in the iron-loaded form (Dewan et al., 1993). The disulfide loop contains 12 basic amino acid residues (Arg 4 , His 3 , and Lys 5 ) and 10 acidic residues (Glu 5 , Asp 1 , and carboxymethyl-Cys 4 ) (Jeltsch et al., 1987). These would confer increased positive charges on the accessible surface of the ironfree form of [4S-S], thereby reducing the electrophoretic mobility toward anode.
The results from the indirect end-labeling analysis clearly shows that the number of the detected fragments is exactly consistent with the one expected from the number of the intact disulfide bonds for either [4S-S], [3S-S], or [2S-S] (Fig. 6). Furthermore, the disulfide peptides are recovered in high yields (Table I)  A question arises as to the implication of these disulfide intermediates for the redox folding/unfolding pathway of ovotransferrin N lobe. In the oxidative folding pathway of bovine pancreatic trypsin inhibitor, the kinetically preferred mechanism involves intramolecular disulfide rearrangements (Creighton, 1977;Creighton and Goldenberg, 1984;Weissman and Kim, 1992), while the reductive unfolding mechanism includes the parallel pathways of the direct reduction and rearrangement pathway (Mendoza et al., 1994). As a technical problem on detecting the correct disulfide intermediates, it has been pointed out that the rate for sulfhydryl alkylation is sometimes too slow to quench rapid sulfhydryl/disulfide interconversions (Rothwarf and Scheraga, 1991;Weissman and Kim, 1991). The following observations, however, make it very unlikely that the partially disulfide-reduced intermediates of the ovotransferrin N lobe are artificial ones that are formed during slow alkylation with 0.1 M IAA employed in the present report. We observe by the peptide mapping analysis that [4S-S] and [3S-S], which are isolated from the disulfide intermediates quenched at an increased concentration of 0.9 M IAA, possess the same disulfide structures as the ones characterized in the present report. 2 All of the disulfide intermediates of [4S-S], [3S-S], and [2S-S] have some stable and partially folded conformations, and their residual native conformations are decreased with the increased number of the reduced disulfides ( Figs. 7 and 8). We therefore conclude that either [4S-S], [3S-S], or [2S-S] is implicated as a major stable disulfide-intermediate for the reductive unfolding pathway of ovotransferrin N lobe. The pathway for the disulfide reduction from the native form to [4S-S], however, is still not clear, since no five-disulfide intermediate is detected in the disulfide intermediates quenched at either 0.1 M IAA (Fig. 3) or 0.9 M IAA. 2 According to the x-ray crystallographic structure (Dewan et al., 1993), the sulfur atoms of SS-V are somehow exposed to solvent, while all sulfur atoms of the other disulfides are almost completely buried in the protein molecule. The SS-V, therefore, may be the disulfide that first cleaved by DTT. In an oxidatively refolded form of ovotransferrin N lobe, small amounts of non-native disulfides (Cys 174 -Cys 182 and Cys 171 -Cys 174 ) that correspond to the mispairings of SS-IV and SS-V have been detected (Yamashita and Hirose, 1993), probably reflecting the fact that the C ␤ -C ␤ distances between Cys 171 and Cys 174 and between Cys 174 and Cys 182 are only about 5 Å in the native structure (Dewan et al., 1993). It is, however, not clear whether or not such non-native disulfide intermediates are involved in the pathway for the second reduction of SS-IV. The question of direct reduction or rearrangement reduction remains until the kinetic pathway for the redox folding/unfolding of the ovotransferrin N lobe is established.