Pathway of oxidative folding of alpha-lactalbumin: a model for illustrating the diversity of disulfide folding pathways

The pathway of oxidative folding of alpha-lactalbumin (alpha LA) (four disulfide bonds) has been characterized by structural and kinetic analysis of the acid-trapped folding intermediates. In the absence of calcium, oxidative folding of alpha LA proceeds through highly heterogeneous species of one-, two-, three-, and four-disulfide (scrambled) intermediates to reach the native structure. In the presence of calcium, the folding intermediates of alpha LA comprise two predominant isomers (alpha LA-IIA and alpha LA-IIIA) adopting exclusively native disulfide bonds, including the two disulfide bonds (Cys(61)-Cys(77) and Cys(73)-Cys(91)) located within the beta-sheet calcium binding domain. alpha LA-IIA is a two-disulfide species consisting of Cys(61)-Cys(77) and Cys(73)-Cys(91) disulfide bonds. alpha LA-IIIA contains Cys(61)-Cys(77), Cys(73)-Cys(91), and Cys(28)-Cys(111) disulfide bonds. The underlying mechanism of the contrasting folding pathways of calcium-bound and calcium-depleted alpha LA is congruent with the cause of diversity of disulfide folding pathways observed among many well-characterized three-disulfide proteins, including bovine pancreatic trypsin inhibitor and hirudin. Our study also reveals novel aspects of the folding mechanism of alpha LA that have not been described previously.     

Evidence for the underlying cause of diversity of the disulfide folding pathway

The pathways of oxidative folding of disulfide proteins exhibit a high degree of diversity, which is illustrated by the varied extent of (a) the heterogeneity of folding intermediates, (b) the predominance of intermediates containing native disulfide bonds, and (c) the level of accumulation of fully oxidized scrambled isomers as intermediates. BPTI and hirudin exemplify two extreme cases of such divergent folding pathways. We previously proposed that the underlying cause of this diversity is associated with the degree of stability of protein subdomains. Here we present compelling evidence that substantiates this hypothesis by studying the folding pathway of alphaLA-IIA. alphaLA-IIA is a partially folded intermediate of alpha-lactalbumin (alphaLA). It comprises a structured beta-sheet (calcium-binding) domain linked by two native disulfide bonds (Cys(61)-Cys(77) and Cys(73)-Cys(91)) and a disordered alpha-helical domain with four free cysteines (Cys(6), Cys(28), Cys(111), and Cys(120)). Purified alphaLA-IIA was allowed to refold without and with stabilization of its structured beta-sheet domain by calcium. In the absence of calcium, the folding pathway of alphaLA-IIA resembles that of hirudin, displaying a highly heterogeneous population of folding intermediates, including fully oxidized scrambled species. Upon stabilization of its beta-sheet domain by bound calcium, oxidative folding of alphaLA-IIA undergoes a pathway conspicuously similar to that of BPTI, exhibiting limited species of folding intermediates containing mostly native disulfide bonds.     

Folding determinants of LDL receptor type A modules.

To investigate how three disulfide bonds and coordination of a calcium ion cooperate to specify the structure of an LDL-A module, we studied the interdependence of disulfide bond formation and calcium coordination in the folding of ligand-binding module 5 of the LDL receptor (LR5). In variants of LR5 containing only a single pair of cysteines normally disulfide-bonded in the native polypeptide, the addition of calcium does not alter the effective concentration of one cysteine for the other. LR5 only exhibits a calcium-dependent preference for formation of native disulfide bonds and detectable calcium-induced changes in structure when the two C-terminal disulfide bonds are present. Furthermore, when the conformation of this two-disulfide variant of LR5 is probed by NMR in the presence of calcium, only the C-terminal lobe of the module, which contains the calcium coordination site, acquires a near-native conformation; the N-terminal lobe appears to be disordered. These findings contrast with studies of other model proteins, like BPTI, in which formation of a single disulfide bond is sufficient to drive the entire domain to acquire a stable, nativelike fold.     

Folding and binding integrity of variants of a prototype ligand-binding module from the LDL receptor possessing multiple alanine substitutions

The LA repeats that comprise the ligand-binding domain of the LDL receptor are among the most common autonomously structured extracellular modules found in the nonredundant protein sequence database. Here, we investigate the information content of the amino acid sequence of a typical LA module by constructing sequences with alanine residues at nonconserved positions in the module. Starting with the sequence of the fifth ligand-binding repeat of the LDL receptor (LA5), we created generic LA modules with alanine substitutions of nonconserved residues in only the N-terminal lobe, only the C-terminal lobe, and throughout both lobes of the module. LA variants with alanine residues at as many as 18 of 37 positions fold to a preferred disulfide isomer in the presence of calcium. Indeed, the six cysteines, the C-terminal calcium coordinating residues, two hydrophobic residues involved in packing, two glycines, and five other residues that form side chain-intramodule hydrogen bonds are alone sufficient to specify the fold of an LA module when alanine residues are present at all other positions. The LA variants with multiple alanines in either the N- or C-terminal lobe were then exploited to identify residues of LA5 that contribute to the binding of apoE-containing ligands in LDL receptor-derived "minireceptors", implicating nonconserved residues of the N-terminal lobe of LA5 in recognition of apoE-DMPC. Our library of LA modules with multiple alanine substitutions should be generally useful for probing the roles of nonconserved side chains in ligand recognition by proteins of the LDL receptor family.     

Pathway of Oxidative Folding of a 3-Disulfide α-Lactalbumin May Resemble Either BPTI Model or Hirudin Model

Pathways of oxidative folding of disulfide proteins display a high degree of diversity and vary among two extreme models. The BPTI model is defined by limited species of folding intermediates adopting mainly native disulfide bonds. The hirudin model is characterized by highly heterogeneous folding intermediates containing mostly non-native disulfide bonds. αLA-IIIA is a 3-disulfide variant of α-lactalbumin (αLA) with a 3-D conformation essentially identical to that of intact αLA. αLA-IIIA contains 3 native disulfide bonds of αLA, two of them are located at the calcium binding β-subdomain (Cys⁶¹–Cys⁷⁷ and Cys⁷³–Cys⁹¹) and the third bridge is located within the α-helical domain of the molecule (Cys²⁸–Cys¹¹¹). We investigate here the pathway of oxidative folding of fully reduced αLA-IIIA with and without stabilization of its β-subdomain by calcium binding. In the absence of calcium, the folding pathway of αLA-IIIA was shown to resemble that of hirudin model. Upon stabilization of β-sheet domain by calcium binding, the folding pathway of αLA-IIIA exhibits a striking similarity to that of BPTI model. Three predominant folding intermediates of αLA-IIIA containing exclusively native disulfide bonds were isolated and structurally characterized. Our results further demonstrate that stabilization of subdomains in a protein may dictate its folding pathway and represent a major cause for the existing diversity in the folding pathways of the disulfide-containing proteins.     

The structure, dynamics, and binding of the LA45 module pair of the low-density lipoprotein receptor suggest an important role for LA4 in ligand release

The low-density lipoprotein receptor (LDLR), the primary receptor for cholesterol uptake, binds ligands through its seven LDL-A modules (LAs). We present nuclear magnetic resonance (NMR) and ligand binding measurements on the fourth and fifth modules of the LDLR (LA45), the modules critical for ApoE binding, at physiological pH. Unlike LA5 and all other modules in LDLR, LA4 has a very weak calcium affinity, which probably plays a critical role in endosomal ligand release. The NMR solution structure of each module in the LA45 pair only showed minor differences compared to the analogous domains in previously determined crystal structures. The 12-residue linker connecting the modules, though slightly structured through an interaction with LA4, is highly flexible. Although no intermodule nuclear Overhauser effects were detected, chemical shift perturbations and backbone dynamics suggest cross talk between the two modules. The ligand affinity of both modules is enhanced when the two are linked. LA4 is more flexible than LA5 and remains so even in the module pair, which likely is related to its weaker calcium binding affinity.     

Role of kinetic intermediates in the folding of leech carboxypeptidase inhibitor

The oxidative folding and reductive unfolding pathways of leech carboxypeptidase inhibitor (LCI; four disulfides) have been characterized in this work by structural and kinetic analysis of the acid-trapped folding intermediates. The oxidative folding of reduced and denatured LCI proceeds rapidly through a sequential flow of 1-, 2-, 3-, and 4-disulfide (scrambled) species to reach the native form. Folding intermediates of LCI comprise two predominant 3-disulfide species (designated as III-A and III-B) and a heterogeneous population of scrambled isomers that consecutively accumulate along the folding reaction. Our study reveals that forms III-A and III-B exclusively contain native disulfide bonds and correspond to stable and partially structured species that interconvert, reaching an equilibrium prior to the formation of the scrambled isomers. Given that these intermediates act as kinetic traps during the oxidative folding, their accumulation is prevented when they are destabilized, thus leading to a significant acceleration of the folding kinetics. III-A and III-B forms appear to have both native disulfides bonds and free thiols similarly protected from the solvent; major structural rearrangements through the formation of scrambled isomers are required to render native LCI. The reductive unfolding pathway of LCI undergoes an apparent all-or-none mechanism, although low amounts of intermediates III-A and III-B can be detected, suggesting differences in protection against reduction among the disulfide bonds. The characterization of III-A and III-B forms shows that the former intermediate structurally and functionally resembles native LCI, whereas the III-B form bears more resemblance to scrambled isomers.     

A major kinetic trap for the oxidative folding of human epidermal growth factor

The folding pathway of human epidermal growth factor (EGF) has been characterized by structural and kinetic analysis of the acid-trapped folding intermediates. Oxidative folding of the fully reduced EGF proceeds through 1-disulfide intermediates and accumulates rapidly as a single stable 2-disulfide intermediate (designated as EGF-II), which represents up to more than 85% of the total protein along the folding pathway. Among the five 1-disulfide intermediates that have been structurally characterized, only one is native, and nearly all of them are bridges by neighboring cysteines. Extensive accumulation of EGF-II indicates that it accounts for the major kinetic trap of EGF folding. EGF-II contains two of the three native disulfide bonds of EGF, Cys(14)-Cys(31) and Cys(33)-Cys(42). However, formation of the third native disulfide (Cys(6)-Cys(20)) for EGF-II is slow and does not occur directly. Kinetic analysis reveals that an important route for EGF-II to reach the native structure is via rearrangement pathway through 3-disulfide scrambled isomers. The pathway of EGF-II to attain the native structure differs from that of three major 2-disulfide intermediates of bovine pancreatic trypsin inhibitor (BPTI). The dissimilarities of folding mechanism(s) between EGF, BPTI, and hirudin are discussed in this paper.     

Conformational specificity of the chaperonin GroEL for the compact folding intermediates of alpha-lactalbumin.

The chaperonin GroEL binds unfolded polypeptides, preventing aggregation, and then mediates their folding in an ATP-dependent process. To understand the structural features in non-native polypeptides recognized by GroEL, we have used alpha-lactalbumin (alpha LA) as a model substrate. alpha LA (14.2 kDa) is stabilized by four disulfide bonds and a bound Ca2+ ion, offering the possibility of trapping partially folded disulfide intermediates between the native and the fully unfolded state. The conformers of alpha LA with high affinity for GroEL are compact, containing up to three disulfide bonds, and have significant secondary structure, but lack stable tertiary structure and expose hydrophobic surfaces. Complex formation requires almost the complete alpha LA sequence and is strongly dependent on salts that stabilize hydrophobic interactions. Unfolding of alpha LA to an extended state as well as the burial of hydrophobic surface upon formation of ordered tertiary structure prevent the binding to GroEL. Interestingly, GroEL interacts only with a specific subset of the many partially folded disulfide intermediates of alpha LA and thus may influence in vitro the kinetics of the folding pathways that lead to disulfide bonds with native combinations. We conclude that the chaperonin interacts with the hydrophobic surfaces exposed by proteins in a flexible compact intermediate or molten globule state.     

Major kinetic traps for the oxidative folding of leech carboxypeptidase inhibitor

The leech carboxypeptidase inhibitor (LCI) is a 66-amino acid protein, containing four disulfides that stabilize its structure. This polypeptide represents an excellent model for the study and understanding of the diversity of folding pathways in small, cysteine-rich proteins. The pathway of oxidative folding of LCI has been elucidated in this work, using structural and kinetic analysis of the folding intermediates trapped by acid quenching. Reduced and denatured LCI refolds through a rapid, sequential flow of one- and two-disulfide intermediates and reaches a rate-limiting step in which a mixture of three major three-disulfide species and a heterogeneous population of non-native four-disulfide (scrambled) isomers coexist. The three three-disulfide intermediates have been identified as major kinetic traps along the folding pathway of LCI, and their disulfide structures have been elucidated in this work. Two of them contain only native disulfide pairings, and one contains one native and two non-native disulfide bonds. The coexistence of three-disulfide kinetic traps adopting native disulfide bonds together with a significant proportion of fully oxidized scrambled isomers shows that the folding pathway of LCI features properties exhibited by both the bovine pancreatic trypsin inhibitor and hirudin, two diverse models with extreme folding characteristics. The results further demonstrate the large diversity of disulfide folding pathways.     

Characterizing the tick carboxypeptidase inhibitor: molecular basis for its two-domain nature

Tick carboxypeptidase inhibitor (TCI) is a small, disulfide-rich protein that selectively inhibits metallocarboxypeptidases and strongly accelerates the fibrinolysis of blood clots. TCI consists of two domains that are structurally very similar, each containing three disulfide bonds arranged in an almost identical fashion. The oxidative folding and reductive unfolding pathways of TCI and its separated domains have been characterized by kinetic and structural analysis of the acid-trapped folding intermediates. TCI folding proceeds through a sequential formation of 1-, 2-, 3-, 4-, 5-, and 6-disulfide species to reach the native form. Folding intermediates of TCI comprise two predominant 3-disulfide species (named IIIa and IIIb) and a major 6-disulfide scrambled isomer (Xa) that consecutively accumulate along the reaction and are strongly prevented by the presence of protein disulfide isomerase. This study demonstrates that IIIa and IIIb are 3-disulfide species containing the native disulfide pairings of the N- and C-terminal domains of TCI, respectively, and explains why the two domains of TCI fold sequentially and independently. Also, we show that the reductive unfolding of TCI undergoes two main independent unfolding events through the formation of IIIa and IIIb intermediates. Together, the comparison of the folding, stability, and inhibitory activity of TCI with those of the isolated domains reveals the reasons behind the two-domain nature of this protein: both domains contribute to the specificity and high affinity of its double-headed binding to carboxypeptidases. The results obtained herein provide valuable information for the design of more potent and selective TCI molecules.     

Identification of a major heparin and cell binding site in the LG4 module of the laminin alpha 5 chain

The G domain of the laminin alpha chains consists of five homologous G modules (LG1-5) and has been implicated in various biological functions. In this study, we identified an active site for cell and heparin binding within the laminin alpha5 G domain using recombinant proteins and synthetic peptides. Recombinant LG4, LG5, and LG4-5 modules were generated using a mammalian expression system. The LG4 and LG4-5 modules were highly active for cell binding, whereas the LG5 module alone showed only weak binding. Heparin inhibited cell binding to the LG4-5 module, whereas no inhibition was observed with EDTA or antibodies against the integrin beta(1) subunit. These results suggest that the LG4-5 module interacts with a cell surface receptor containing heparan sulfate but not with integrins. Solid-phase assays and surface plasmon resonance measurements demonstrated strong binding of the LG4 and LG4-5 modules to heparin with K(D) values in the nanomolar range, whereas a 16-fold lower value was determined for the LG5 module. Treatment with glycosidases demonstrated that N-linked carbohydrates on the LG5 module are complex-type oligosaccharides. The LG4-5 module, devoid of N-linked carbohydrates, exhibited similar binding kinetics toward heparin. Furthermore, cell binding was unaffected by removal of N-linked glycosylation. To localize active sites on the LG4 module, various synthetic peptides were used to compete with binding of the tandem module to heparin and cells. Peptide F4 (AGQWHRVSVRWG) inhibited binding, whereas a scrambled peptide of F4 failed to compete binding. Alanine replacements demonstrated that one arginine residue within F4 was important for cell and heparin binding. Our results suggest a critical role of the LG4 module for heparan sulfate-containing receptor binding within the laminin alpha5 chain.     

Localization of disulfide bonds in the frizzled module of Ror1 receptor tyrosine kinase

The frizzled (FRZ) module is a novel module type that was first identified in G-protein-coupled receptors of the frizzled and smoothened families and has since been shown to be present in several secreted frizzled-related proteins, in some modular proteases, in collagen XVIII, and in various receptor tyrosine kinases of the Ror family. The FRZ modules constitute the extracellular ligand-binding region of frizzled receptors and are known to mediate signals of WNT family members through these receptors. With an eye toward defining the structure of this important module family, we have expressed the FRZ domain of rat Ror1 receptor tyrosine kinase in Pichia pastoris. By proteolytic digestion and amino acid sequencing the disulfide bonds were found to connect the 10 conserved cysteines in a 1-5, 2-4, 3-8, 6-10, and 7-9 pattern. Circular dichroism and differential scanning calorimetry studies on the recombinant protein indicate that the disulfide-bonded FRZ module corresponds to a single, compact, and remarkably stable folding domain possessing both alpha-helices and beta-strands.     

Kinetic analysis of the folding and unfolding of a mutant form of bovine pancreatic trypsin inhibitor lacking the cysteine-14 and -38 thiols

The kinetics of the disulfide-coupled unfolding-refolding transition of a mutant form of bovine pancreatic trypsin inhibitor (BPTI) lacking Cys-14 and -38 were measured and compared to previous results for the wild-type protein and other modified forms. The altered cysteines, which were changed to serine in the mutant protein, are normally paired in a disulfide in the native protein but from disulfides with Cys-5 in two-disulfide kinetic intermediates during folding. Although the mutant protein could fold efficiently, the kinetics of both folding and unfolding were altered, reflecting the roles of these cysteines in the two-disulfide intermediates with "wrong" disulfides. The intramolecular rate constant for the formation of the second disulfide of the native mutant protein was more than 10(3)-fold lower than that for the formation of a second disulfide during the refolding of the wild-type protein. The observed rate of unfolding of the mutant protein was also lower than that of the wild-type protein, demonstrating that the altered cysteines are involved in the intramolecular rearrangements that are the rate-determining step in the unfolding of the wild-type protein. These results confirm the previous conclusion [Creighton, T.E. (1977) J. Mol. Biol. 113, 275-293] that the energetically preferred pathway for folding and unfolding of BPTI includes intramolecular rearrangements of intermediates in which Cys-14 and -38 are paired in disulfides not present in the native protein. The present results are also consistent with other, less detailed, studies with similar mutants lacking Cys-14 and -38 [Marks, C.B., Naderi, H., Kosen, P.A., Kuntz, I.D., & Anderson, S. (1987) Science (Washington, D.C.) 235, 1370-1371].     

Evidence that familial hypercholesterolemia mutations of the LDL receptor cause limited local misfolding in an LDL-A module pair

Mutations at conserved sites within the ligand-binding LDL-A modules of the LDL receptor cause the genetic disease familial hypercholesterolemia (FH), and several of these FH mutations in modules five and six prevent the isolated single modules from folding properly to a nativelike three-dimensional structure. Because LDL-A modules occur as a series of contiguous repeats in the LDLR and related proteins, we investigated the impact of two FH mutations in LDL-A module five (D203G and D206E) and two mutations in module six (E219K and D245E) in the context of the covalently connected module five-six pair. HPLC chromatography of the products formed under conditions that efficiently refold the native module five-six pair demonstrate that, for each mutation, a folding defect persists in the module pair. NMR spectroscopy and calcium affinity measurements of the ensemble of misfolded products demonstrate that the unaltered module of each pair can fold to its native structure regardless of the range of misfolded conformations adopted by its mutated neighbor. These findings lend additional support to a model in which individual LDL-A modules of the LDL receptor act as independent structural elements.     

The folding pathway of alpha-lactalbumin elucidated by the technique of disulfide scrambling. Isolation of on-pathway and off-pathway intermediates

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, alpha-lactalbumin (alphaLA) 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-alphaLA-a and X-alphaLA-d, account for about 50% of the total denatured structure of alphaLA. X-alphaLA-a and X-alphaLA-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-alphaLA-a and X-alphaLA-d were purified and allowed to refold through disulfide scrambling to form the native alphaLA. 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 alpha-helical domain, and the other comprises a structured beta-sheet, calcium binding domain. The two major kinetic traps are apparently stabilized by locally formed non-native-like structures. Overall, the folding mechanism of alphaLA is essentially congruent with the model of "folding funnel" furnished with a rather intricate energy landscape.     

Efficient folding of the FcepsilonRI alpha-chain membrane-proximal domain D2 depends on the presence of the N-terminal domain D1

Human high affinity receptor for IgE is a membrane glycoprotein multichain complex presenting two extracellular Ig modules in its alpha-chain (D1D2). The receptor IgE binding region is located within the membrane-proximal module D2, while the N-terminal module D1 appears to promote an optimal receptor conformation for IgE binding. To understand the structural relationship between the two modules, we dissected FcepsilonRI alpha-chain into its discrete Ig units and expressed them in mammalian cells. Unexpectedly, D2 was secreted as a disulphide-linked dimer, while D1 was monomeric. Active secretion and full glycosylation of dimeric D2 suggest a native-like conformation of the protein, justifying the escape from the endoplasmic reticulum/Golgi quality control systems. We then propose a domain-swapping model for D2, in which two interdigitated polypeptide chains assume the overall conformation of two Ig modules, as observed for rat CD2 N-terminal domain. Fusion of an unrelated Ig fold moiety at the N terminus of D2 did not interfere with its dimerisation. While D1D2 assumes a correct fold, co-expression of both isolated domains in the same cell did not restore monomeric folding of D2. Thus, D1 appears to assist the appropriate folding of FcepsilonRI alpha-chain, acting as an uncleavable intramolecular chaperone-like block towards D2.     

The precursor of beta-lactamase: purification, properties and folding kinetics.

The precursor of Escherichia coli RTEM beta-lactamase was purified to homogeneity on a milligram scale by a procedure independent of the binding properties of the protein and refolded to an active, reduced form. For comparing the folding kinetics, the wild-type enzyme was reduced and a mutant was constructed, in which the two cysteines that form a very stable disulfide bond in the RTEM enzyme were both changed into alanines. The rate of folding was determined by directly measuring the increase in enzymatic activity. The reduced precursor folds at least 15 times more slowly than either the reduced mature enzyme or the mature Cys----Ala double mutant under identical conditions. The wild-type enzyme, the Cys----Ala double mutant and the precursor protein all had similar KM values, demonstrating a very similar native state. The slow folding of the precursor compared with the mature form may be an essential and general feature to secure a transport competent conformation necessary for the translocation through a membrane in protein export. This folding assay of a precursor by directly following its enzymatic activity may facilitate the characterization of putative folding modulators in bacterial membrane transport.     

The NMR structures of the major intermediates of the two-domain tick carboxypeptidase inhibitor reveal symmetry in its folding and unfolding pathways

There is a lack of experimental structural information about folding intermediates of multidomain proteins. Tick carboxypeptidase inhibitor (TCI) is a small, disulfide-rich protein consisting of two domains that fold and unfold autonomously through the formation of two major intermediates, IIIa and IIIb. Each intermediate contains three native disulfide bonds in one domain and six free cysteines in the other domain. Here we have determined the NMR structures of these two intermediates trapped and isolated at acidic pH in which they are stable and compared their structures with that of the native protein analyzed under the same conditions. Both IIIa and IIIb were found to contain a folded region that corresponds to the N- and C-terminal domains of TCI, respectively, with structures very similar to the corresponding regions of the native protein. The remainder of the polypeptide chains of the intermediates was shown to be unfolded in a random coil conformation. Solvent exchange measurements further indicated that the two protein domains are not completely independent, but affect each other in terms of dynamics and stability, in agreement with reported inhibitory activity data. The derived results provide structural evidence for symmetric TCI folding and unfolding mechanisms that converge in IIIa and IIIb and reveal the structural basis that accounts for the strong and simultaneous accumulation of both intermediates. Altogether, this work has important implications for a better understanding of the folding mechanisms of multidomain, disulfide-rich proteins.