The Folding Pathway of α-Lactalbumin Elucidated by the Technique of Disulfide Scrambling

The technique of disulfide scrambling permits reversible conversion of the native and denatured (scrambled) proteins via shuffling and reshuffling of disulfide bonds. Under strong denaturing conditions (e.g. 6 m guanidinium chloride) and in the presence of a thiol initiator, α-lactalbumin (αLA) denatures by shuffling its four native disulfide bonds and converts to an assembly of 45 species of scrambled isomers. Among them, two predominant isomers, designated as X-αLA-a and X-αLA-d, account for about 50% of the total denatured structure of αLA. X-αLA-a and X-αLA-d, which adopt the disulfide patterns of (1–2,3–4,5–6,7–8) and (1–2,3–6,4–5,7–8), respectively, represent the most unfolded structures among the 104 possible scrambled isomers (Chang, J.-Y., and Li, L. (2001)J. Biol. Chem. 276, 9705–9712). In this study, X-αLA-a and X-αLA-d were purified and allowed to refold through disulfide scrambling to form the native αLA. Folding intermediates were trapped kinetically by acid quenching and analyzed quantitatively by reversed phase high pressure liquid chromatography. The results revealed two major on-pathway productive intermediates, two major off-pathway kinetic traps, and at least 30 additional minor transient intermediates. Of the two major on-pathway intermediates, one takes on a native-like α-helical domain, and the other comprises a structured β-sheet, calcium binding domain. The two major kinetic traps are apparently stabilized by locally formed non-native-like structures. Overall, the folding mechanism of αLA is essentially congruent with the model of “folding funnel” furnished with a rather intricate energy landscape.

Following the removal of denaturant (e.g. by gel filtration, dilution, or dialysis), pH jump, or temperature jump, denatured proteins usually refold spontaneously to form the native structure. The pathway of protein refolding is then monitored by the mechanism of restoration of some physicochemical signals that distinguish the native and unfolded states. The most commonly used signals are spectra of fluorescence, circular dichroism (15), infrared (16,17), UV, and NMR (18 -21), coupled with amide-proton exchange. This method is most versatile and can be practically applied to study folding behaviors of any protein. However, it does not permit, in most cases, isolation of folding intermediates. To date, the vast majority of our knowledge about protein folding has been acquired by the application of this approach.
Another established method is oxidative folding of disulfidecontaining proteins (22)(23)(24)(25)(26). Proteins are first reduced and denatured in the presence of reducing agent (e.g. dithiothreitol) and denaturant (e.g. 6 M GdmCl). After exclusion of the reductant and denaturant, the reduced and denatured protein is allowed to refold in the presence of redox buffer (27). The pathway of refolding is then tracked by the mechanism of formation of the native disulfide bonds. For example, a protein that contains three disulfide bonds can potentially assume 75 different disulfide isomers (15 one-disulfide, 45 two-disulfide, and 15 three-disulfide). The disulfide folding pathway is characterized by the heterogeneity and structures of disulfide isomers that accumulate along the process of oxidative folding that leads to the formation of the species containing the three native disulfide bonds. This technique, pioneered by Creighton (22,23), has been applied to the elucidation of folding pathways of several three-disulfide-containing proteins, including bovine pancreatic trypsin inhibitor (28,29), ribonuclease A (30), hirudin (31), tick anticoagulant peptide (32,33), potato carboxypeptidase inhibitor (34), epidermal growth factor (35)(36)(37), and insulin-like growth factor (38,39).
We describe here a third approach for analyzing the pathway of protein folding using the technique of disulfide scrambling, a process of shuffling and reshuffling of disulfide bonds accompanied by the reversible conformational change between the native and denatured proteins (40,41). In the presence of denaturant and a thiol initiator, a native protein denatures by shuffling its native disulfide bonds and converts to a mixture of heterogeneous scrambled isomers. Upon removal of the denaturant (but retaining the thiol initiator), all scrambled isomers refold spontaneously by reshuffling their non-native disulfide bonds to form the native structure.
The technique of disulfide scrambling was applied to analyze the structures of denatured ␣-lactalbumin (␣LA) (42). ␣LA contains four disulfide bonds and may adopt 104 possible denatured scrambled isomers. Each scrambled isomer assumes a specific disulfide pairings and adopts a varied extent of unfold-ing. At mild concentration of GdmCl (e.g. 1.5 M), denatured ␣LA was shown to consist of 45 fractions of scrambled species (42). None of them comprises more than 10% of the total structure of denatured ␣LA. However, under strong denaturing conditions, two predominant scrambled isomers, namely X-␣LA-a and X-␣LA-d, together account for about 50% of the total structure of denatured ␣LA (42). These two well populated isomers contain exclusively non-native disulfide bonds and adopt the disulfide patterns of (1-2,3-4,5-6,7-8) and (1-2,3-6,4 -5,7-8), respectively (see Fig. 1). They have the smallest sizes of disulfide loops of all possible scrambled isomers. Their disulfide structures, together with the fact that they are predominant only under strong denaturing conditions, indicate that X-␣LA-a and X-␣LA-d possess the highest free energy and represent the most unfolded structures among the 104 possible denatured isomers.
In this report, X-␣LA-a and X-␣LA-d were purified and allowed to refold through disulfide scrambling in the presence of thiol initiators. Folding intermediates were trapped by acid quenching and analyzed by HPLC. The studies have enabled us to identify and isolate productive folding intermediates, kinetic traps, and a large number of transient intermediates along the folding pathway of ␣LA.

EXPERIMENTAL PROCEDURES
Materials-Calcium-depleted bovine ␣LA (L-6010) was used throughout this study and was obtained from Sigma. The protein was further purified by HPLC and was shown to be more than 97% pure. Thermolysin (P-1512), dithiothreitol, reduced glutathione, cysteine, 2-mercaptoethanol, GdmCl, guanidinium thiocyanate, and urea were also purchased from Sigma with purity of greater than 99%. Protein disulfide isomerase (number 7318) was purchased from Takara, Kyoto, Japan.
Nomenclature of Scrambled Isomers of ␣LA-Scrambled species of ␣LA are designated by the following formula: X-␣LA-(species assigned on HPLC), where X stands for scrambled. For instance, X-␣LA-a represents species a of scrambled ␣LA.
Preparation of Scrambled Isomers of Denatured ␣LA as Starting Material of Folding-The native ␣LA (4 mg/ml) was dissolved in the Tris-HCl buffer (0.1 M, pH 8.4) containing 2-mercaptoethanol (0.2 mM) and GdmCl (6 M). Denaturation was carried out at 23°C for 24 h. Denatured ␣LA was then mixed with an equal volume of 4% aqueous trifluoroacetic acid and subjected directly to separation by HPLC. Under these conditions (42), denatured ␣LA consists of about 45 fractions of scrambled isomers, with two predominant species, namely X-␣LA-a and X-␣LA-d, each accounting for about 24 and 17% of the total denatured structure. X-␣LA-a and X-␣LA-d were purified by HPLC, freezedried, and used as starting materials of folding experiments. They were stored at Ϫ20°C.
Folding of Scrambled Isomers of ␣LA by Disulfide Scrambling-Folding of X-␣LA-a and X-␣LA-d was performed in the Tris-HCl buffer (0.1 M, pH 8.4) using 2-mercaptoethanol (0.1 mM) (control folding), cysteine (0.1 mM), or reduced glutathione (0.1 mM) as thiol initiators. Folding experiments of X-␣LA-a were also carried out in the presence of NaCl (1 M), CaCl 2 (5 mM), and protein disulfide isomerase (5 and 50 M) using 2-mercaptoethanol as the thiol initiator. Stop/go folding experiments of purified intermediates (X-␣LA-b, X-␣LA-c, X-␣LA-h, and X-␣LA-k) were similarly conducted in the Tris-HCl buffer containing 0.1 mM 2-mercaptoethanol. All experiments were carried out at 23°C with the protein concentration of 0.5 mg/ml. To monitor the kinetics of folding, the recovery of native structure, and the heterogeneity of folding intermediates, aliquots of the folding sample were removed at time intervals, quenched with an equal volume of 4% aqueous trifluoroacetic acid, and analyzed by HPLC. The samples were stored at Ϫ20°C.
Analysis of the Folding Intermediates by Reversed Phase HPLC-Analysis and isolation of folding intermediates of ␣LA were achieved by the reversed phase HPLC using the following conditions: Solvent A for HPLC was water containing 0.1% trifluoroacetic acid, and Solvent B was acetonitrile/water (9:1, by volume) containing 0.086% trifluoroacetic acid. The gradient was 22 to 37% Solvent B in 15 min, and 37 to 56% Solvent B from 15 to 45 min. The flow rate was 0.5 ml/min. The HPLC column was Zorbax 300SB C-18 for peptides and proteins, 4.6 mm, 5 m. Column temperature was 23°C.
Structural Analysis of Scrambled Isomers of ␣LA-Fractions of scrambled ␣LA (ϳ10 g) were isolated and treated with 1 g of ther-molysin in 30 l of N-ethylmorpholine/acetate buffer (50 mM, pH 6.4). Digestion was carried out at 37°C for 16 h. Peptides were then isolated by HPLC and analyzed by amino acid sequencing and mass spectrometry to identify the disulfide-containing peptides.
Amino Acid Sequencing and Mass Spectrometry-Amino acid sequence of disulfide-containing peptides were analyzed by automatic Edman degradation using a PerkinElmer Life Sciences Procise sequencer (model 494) equipped with an on-line phenylthiohydantoinderivative analyzer. The molecular masses of disulfide-containing peptides were determined by a matrix-assisted laser desorption/ ionization-time-of-flight mass spectrometer (PerkinElmer Life Sciences Voyager-DE STR).

RESULTS
Folding Mechanism of X-␣LA-a-A control folding experiment of X-␣LA-a was first performed at 23°C using 2-mercaptoethanol (0.1 mM) as a thiol initiator. The folding intermediates and the yield of native ␣LA, trapped by acid quenching and analyzed by HPLC, are presented in Fig. 2A. About 55% of the native ␣LA was recovered after 24 h of folding (see Fig. 3). Conversion of X-␣LA-a to the native ␣LA proceeds through a mixture of heterogeneous intermediates. Approximately 35 fractions of scrambled ␣LA were identified along the folding pathway. Among them, four major species have been structurally characterized ( Fig. 1) (42). X-␣LA-b contains two native disulfide bonds at the ␤-sheet domain, whereas X-␣LA-c forms two native disulfides within the ␣-helical domain. Both X-␣LA-h and X-␣LA-k undergo partial folding at the C-termi-  (42). The structure of isomer k is reported here for the first time. The disulfide structure of X-␣LA-k was derived from the Edman sequencing and matrix-assisted laser desorption/ionization mass analysis of disulfidecontaining peptides of thermolysin-digested samples (data not shown). nal and N-terminal regions of ␣LA. The relative concentration of these four major intermediates accumulated along the folding pathway is shown in Fig. 4A. Cysteine and reduced glutathione are useful thiol initiators, as well, but they are less effective than 2-mercaptoethanol. At the same concentration (0.1 mM), 2-mercaptoethanol is about 60 and 120% more efficient than cysteine and reduced glutathione in promoting the conversion of X-␣LA-a to native ␣LA (see Fig. 3). Regardless of the nature of thiol initiator, the patterns of folding intermediates remain indistinguishable.
The inclusion of a protein stabilizer (1 M NaCl) improves the rate of renaturation of native ␣LA by 2-fold (see Fig. 3). However, it is the presence of protein disulfide isomerase that exhibits the most remarkable effect on accelerating the folding of X-␣LA-a. As compared with the control folding experiment performed in the presence of 2-mercaptoethanol (0.1 mM) alone, addition of 5 and 50 M protein disulfide isomerase increases the rate of folding of X-␣LA-a by 6-and 60-fold, respectively (see Fig. 3). The concentrations of folding intermediates rise by an average of 75% (compare Fig. 4, A and B), but their heterogeneity and patterns remain essentially unchanged (compare Fig. 2, A and B). The rise of concentration of folding intermediates in the presence of protein disulfide isomerase reflects the significant increase of rate constant for the conversion of X-␣LA-a to the various intermediates.
␣LA is a calcium-binding protein. Not surprisingly, the presence of CaCl 2 also displays significant influence on both folding kinetic and folding intermediates of X-␣LA-a. Inclusion of CaCl 2 (5 mM) accelerates the folding of X-␣LA-a by about 2.5fold (Fig. 3)  Stop/Go Folding of X-␣LA-b and X-␣LA-c, Two Major Onpathway Productive Folding Intermediates-Of the four major folding intermediates of X-␣LA-a, two comprise native-like structural domains. X-␣LA-b forms two native disulfide bonds (Cys 61 -Cys 77 , Cys 73 -Cys 91 ) at the ␤-sheet region, whereas X-␣LA-c contains the other two native disulfide bonds (Cys 6 -Cys 120 , Cys 28 -Cys 111 ) within the ␣-helical domain (Fig. 1). X-␣LA-b and X-␣LA-c were purified by HPLC, freeze-dried, and allowed to continue the folding in the Tris-HCl buffer containing 2-mercaptoethanol (0.1 mM). Folding intermediates were trapped by acidification and analyzed by HPLC. Chromatograms are presented in Fig. 5.
Both X-␣LA-b and X-␣LA-c fold to form the native ␣LA without accumulation of any significant intermediate along the pathway (Fig. 5). These data suggest that existing native disulfide bonds of X-␣LA-b and X-␣LA-c remain stable and do not participate in further disulfide shuffling during the folding. Conversion of both X-␣LA-b and X-␣LA-c to the native ␣LA thus entail only disulfide scrambling of their remaining two non-native disulfide bonds. In the case of X-␣LA-b (Fig. 1), for instance, it requires reshuffling of Cys 6 -Cys 28 and Cys 111 -Cys 120 to form Cys 6 -Cys 120 and Cys 28 -Cys 111 to reach the native structure. In this setting, only one possible isomer (Cys 6 -Cys 111 , Cys 61 -Cys 77 , Cys 73 -Cys 91 , and Cys 28 -Cys 120 ) may serve as the folding intermediate of X-␣LA-b. Absence of this yet to be identified isomer indicates that it is unlikely to adopt a stable structure. A similar scenario is applied to the folding of X-␣LA-c.
The rates of renaturation of X-␣LA-b and X-␣LA-c are considerably greater than that of X-␣LA-a (Fig. 6). Between these two intermediates, X-␣LA-b is about 65% more competent than X-␣LA-c in converting to the native ␣LA, implying that folding (formation) of the ␣-helical domain is more efficient than that of the ␤-sheet domain.
Stop/Go Folding of X-␣LA-h and X-␣LA-k, Two Major Offpathway Kinetic Traps-Purified X-␣LA-h and X-␣LA-k were also permitted to carry on the folding under similar conditions (0.1 mM 2-mercaptoethanol). Their folding kinetics and the heterogeneity of folding intermediates were concluded from chromatograms presented in Fig. 7. Efficiencies of folding of X-␣LA-h and X-␣LA-k are slightly lower than that of X-␣LA-a and significantly lower than those of X-␣LA-b and X-␣LA-c The Folding Mechanism of ␣-Lactalbumin (Fig. 6). The rate of renaturation of X-␣LA-k is about 5-fold slower than that of X-␣LA-b. In addition, folding of both X-␣LA-h and X-␣LA-k undergo a mixture of heterogeneous intermediates that include X-␣LA-a, X-␣LA-b, and X-␣LA-c (Fig. 7). The low concentration of intermediates observed in the case of X-␣LA-k folding (Fig. 7, right panel) is consistent with its slow renaturation to form the native structure. In this case, disulfide reshuffling (reorganization) of X-␣LA-k and its conversion to other scrambled species represent a major ratelimiting step.
These data imply that X-␣LA-h and X-␣LA-k are stable kinetic traps along the folding pathway of X-␣LA-a. Their stability accounts for their accumulation and slow conversion during the folding. Because X-␣LA-h and X-␣LA-k both contain entirely non-native disulfide bonds, their conformational stability must be attributed to the non-native interactions formed at the N-terminal (X-␣LA-k) and C-terminal (X-␣LA-h) regions (Fig.  1). To continue the folding, such stable structural elements need first to be disrupted through disulfide shuffling. For example, one major route for X-␣LA-k to carry on folding is to scramble Cys 6 -Cys 61 and Cys 28 -Cys 73 at the N-terminal region (Fig. 1) and return back to X-␣LA-a (see Fig. 7 and Fig. 8).
Folding Mechanism of X-␣LA-d-X-␣LA-a and X-␣LA-d are two predominant isomers that constitute the structure of denatured ␣LA generated under strong denaturing conditions (42). According to the unfolding curves of ␣LA (42), the yield of X-␣LA-d as a fraction of the total denatured ␣LA increases from 20 to 23% as the concentration of guanidinium thiocyanate rises from 5 to 6 M. At the same time, the recovery of X-␣LA-a decreases from 28 to 26%. A similar phenomenon was observed with the GdmCl unfolding curves of ␣LA. These data indicate that X-␣LA-d possesses a higher free energy and possibly adopts a more unfolded structure than X-␣LA-a. To corroborate this finding, folding of X-␣LA-d was also examined here. Purified X-␣LA-d was allowed to refold in the Tris-HCl buffer using 2-mercaptoethanol (0.1 mM) as the thiol initiator, an experimental condition identical to that used in the control folding of X-␣LA-a ( Fig. 2A). Folding intermediates were similarly analyzed by HPLC. The results show that the patterns of folding intermediates of X-␣LA-d (Fig. 7, left panel) are largely comparable with those initiated with X-␣LA-a (Fig. 2A). These results, however, reveal two distinct properties of X-␣LA-d. 1) Folding of X-␣LA-d was found to proceed through X-␣LA-a as a major intermediate (Fig. 4D). By contrast, only a minute amount of X-␣LA-d appeared along the folding pathway of X-␣LA-a (Fig. 2, A and  B). This clearly demonstrates that conversion of X-␣LA-d to X-␣LA-a during the folding is essentially an irreversible one-way street, which in turn substantiates our previous finding (42) that X-␣LA-d is thermodynamically less stable than X-␣LA-a. 2) Under identical folding conditions, renaturation of X-␣LA-d is actually about 20% more efficient than that of X-␣LA-a (Fig. 6). This is most likely because of the fact that, aside from converting to X-␣LA-a, a substantial amount of X-␣LA-d may also refold directly via X-␣LA-b, which is an on-pathway productive intermediate that rapidly forms the native ␣LA (Fig. 5). DISCUSSION The Folding Pathway of ␣LA Is Congruent with the Model of a "Folding Funnel"-It is now commonly understood that for most proteins there is no single specific folding pathway. Instead, unfolded proteins must navigate their way through a maze of routes and an energy landscape (folding funnel) to reach the native structure (10,(43)(44)(45)(46)(47)(48). Complexity and smoothness of such a folding funnel are determined by the nature, size, and domain structures of proteins. This new model of protein folding has presented experimentalists a formidable challenge to monitor and describe simultaneously an exceedingly large number of transient intermediates with vastly diverse stability and local energy minima (7). Without separation and identification of these transient intermediates, the folding pathway at present can only be deduced and described from analysis of the average physicochemical properties of these collective transient isomers.
The technique of disulfide scrambling permits folding experiments to be initiated with a structurally defined isomer and also allows such heterogeneous folding intermediates to be trapped by covalent bonds (42). Qualitative and quantitative analysis of trapped folding intermediates can be performed with HPLC. We demonstrate here the application of this method to the folding of ␣LA. The results show that the folding mechanism of ␣LA is essentially congruent with the model of a folding funnel (43)(44)(45)(46)(47)(48). The energy landscape that constitutes the folding funnel of ␣LA is best illustrated by the intermediates presented in Fig. 2B. Along the folding pathway of ␣LA, there exists a large number of structural isoforms. A few of them represent productive intermediates (e.g. X-␣LA-b and X-␣LA-c), some are transient intermediates (local energy minima), whereas others account for major kinetic traps (e.g. X-␣LA-h and X-␣LA-k). All together, at least 35 fractions of major and minor folding intermediates of ␣LA are identified. In this study, only four predominant intermediates were characterized. A thorough structural analysis and stop/go folding experiments of these minor intermediates should reveal a more detailed energy landscape of the folding of ␣LA.
The Major Folding Pathway(s) of ␣LA-Despite the complexity of minor transient intermediates, analysis of the predominant on-pathway and off-pathway intermediates allows us to construct the major folding pathways of ␣LA (Fig. 8). In this scheme, folding of the unfolded protein (X-␣LA-a) undergoes two major routes, and each consists of two distinct pathways. One is to form on-pathway intermediates with native disulfide bonds (X-␣LA-b and X-␣LA-c). Such intermediates most likely comprise native-like structures and may directly proceed to form the native ␣LA. X-␣LA-b and X-␣LA-c do not exist in a state of equilibrium (Fig. 5). The extent of their accumulation along the folding pathway is kinetically partitioned by the rates of their formation and their subsequent conversion to the native ␣LA. Stop/go folding experiments with purified X-␣LA-b and X-␣LA-c have demonstrated that folding (formation) of the ␣-helical domain (conversion of X-␣LA-b to the native ␣LA) is about 65% more efficient than that of the ␤-sheet domain. The other route is to form off-pathway intermediates with nonnative disulfide bonds (X-␣LA-h and X-␣LA-k). Such interme-  diates act as kinetic traps that are stabilized by non-native interactions. For X-␣LA-h and X-␣LA-k to carry on the folding, they will have to overcome an energy barrier and unravel the structure that already formed. In some occasions, they may need to return to a state of higher free energy to continue the folding. This is evident in the case of X-␣LA-k. One major channel for X-␣LA-k to continue the folding is to convert back to X-␣LA-a (Fig. 7). This conclusion is based on the observation that X-␣LA-a represents the major intermediate along the stop/go folding pathway of X-␣LA-k. The overall low concentration of intermediates observed in the case of X-␣LA-k folding (Fig. 7, right panel) signals the stability of X-␣LA-k and indicates that disulfide reshuffling (reorganization) of X-␣LA-k and its conversion to other scrambled species represent a major rate-limiting step.
Folding can also be initiated with X-␣LA-d, an isomer that possesses a higher free energy than X-␣LA-a (42). In this case, folding pathways are kinetically segregated by converting to either X-␣LA-a or X-␣LA-b, which then keep folding as illustrated (Fig. 8).
It is relevant to mention that this outlined folding pathway of ␣LA, and hence the configuration of the energy landscape of ␣LA folding, appears to be independent of the rate of protein renaturation, at least under the folding conditions described here. For instance, in the presence of protein disulfide isomerase (50 M), the rate of folding of ␣LA increases by more than 60-fold (Fig. 4), yet the patterns of folding intermediates remain practically unaffected (compare Fig. 2, A and B). However, in the presence of ligand (Ca 2ϩ ) that binds to a specific structural domain of ␣LA, the folding pathways can be drastically altered. In this setting, folding of X-␣LA-a bypasses the majority of transient intermediates and kinetic traps and undergoes a single predominant intermediate that forms a structured ␤-sheet (calcium binding) domain (Fig. 2C). The energy landscape is greatly simplified and becomes much smoother.
The Folding Pathway of ␣LA Is Mostly Consistent with the Documented Findings-␣LA is among the most intensively investigated models for understanding the mechanism of protein folding (19,49). To date, the majority of existing data concerning the folding mechanism of ␣LA has been obtained from monitoring the refolding process of denatured ␣LA without disrupting its native disulfide bonds and without fractionation of folding intermediates (49 -59). However, these documented findings are largely consistent with the folding pathway of ␣LA elucidated here using the technique of disulfide scrambling. These consistencies are elaborated as follows. 1) The molten globule state of ␣LA has been widely observed by various spectroscopic and physicochemical techniques as a major kinetic folding intermediate of ␣LA (19, 57, 60 -65). Structural analysis by NMR revealed that the molten globule state of ␣LA comprises a structured ␣-helical domain and a disordered ␤-sheet domain (51,53,54). A similar intermediate was found along the folding pathway of ␣LA elucidated here (Fig. 8). The most predominant on-pathway folding intermediate, X-␣LA-c, bears the essential structural properties of the molten globule state of ␣LA (42,66). X-␣LA-c contains two native disulfide bonds (Cys 6 -Cys 120 , Cys 28 -Cys 111 ) within the ␣-helical domain and two scrambled disulfide bonds (Cys 61 -Cys 73 , Cys 77 -Cys 91 ) at the ␤-sheet region (Fig. 1). X-␣LA-c also retains a substantial amount of ␣-helical structure as exhibited by its far UV CD spectrum (66). 2) Transient non-native secondary structures (␤-sheet) were detected during the refolding of ␣LA by infrared spectroscopy (17). Here, the two major kinetic traps (X-␣LA-h and X-␣LA-k) and 30 minor transient intermediates are all potential candidates containing such non-native secondary structures.
3) The presence of calcium is known to accelerate the folding rate of ␣LA (17,(67)(68)(69). This effect has been suggested to be the consequence of calcium binding to the early intermediates (19,70). Here, we clearly demonstrate that the presence of CaCl 2 enhances the folding of calcium binding site of ␣LA and favors the formation of an on-pathway folding intermediate (X-␣LA-b) that adopts a structured, native-like ␤-sheet domain.