Chemistry and Enzymology of Disulfide Cross-linking in Proteins

Abstract

Cysteine thiols are among the most reactive functional groups in proteins, and their pairing in disulfide linkages is a common post-translational modification in proteins entering the secretory pathway. This modest amino acid alteration, the mere removal of a pair of hydrogen atoms from juxtaposed cysteine residues, contrasts with the substantial changes that characterize most other post-translational reactions. However, the wide variety of proteins that contain disulfides, the profound impact of cross-linking on the behavior of the protein polymer, the numerous and diverse players in intracellular pathways for disulfide formation, and the distinct biological settings in which disulfide bond formation can take place belie the simplicity of the process. Here we lay the groundwork for appreciating the mechanisms and consequences of disulfide bond formation in vivo by reviewing chemical principles underlying cysteine pairing and oxidation. We then show how enzymes tune redox-active cofactors and recruit oxidants to improve the specificity and efficiency of disulfide formation. Finally, we discuss disulfide bond formation in a cellular context and identify important principles that contribute to productive thiol oxidation in complex, crowded, dynamic environments.

Graphical abstract

1. INTRODUCTION

Cysteine thiols are among the most reactive functional groups in proteins, and their pairing in disulfide linkages is a common post-translational modification in proteins entering the secretory pathway. This modest amino acid alteration, the mere removal of a pair of hydrogen atoms from juxtaposed cysteine residues, contrasts with the substantial changes that characterize most other post-translational reactions. However, the wide variety of proteins that contain disulfides, the profound impact of cross-linking on the behavior of the protein polymer, the numerous and diverse players in intracellular pathways for disulfide formation, and the distinct biological settings in which disulfide bond formation can take place belie the simplicity of the process. Here we lay the groundwork for appreciating the mechanisms and consequences of disulfide bond formation in vivo by reviewing chemical principles underlying cysteine pairing and oxidation. We then show how enzymes tune redox-active cofactors and recruit oxidants to improve the specificity and efficiency of disulfide formation. Finally, we discuss disulfide bond formation in a cellular context and identify important principles that contribute to productive thiol oxidation in complex, crowded, dynamic environments.

2. WHICH PROTEINS CONTAIN DISULFIDES

Almost 40% of human protein-encoding genes are predicted to have either a signal sequence for targeting to the secretory pathway and/or at least one transmembrane segment (http://www.proteinatlas.org/humanproteome/secretome). Many of these proteins, which are either secreted or remain associated with the cell plasma membrane or the intracellular endomembrane system, are modified by disulfide bonding. Disulfide bonding is not restricted to proteins of a certain size, type, or function (Figure 1). Indeed, disulfides appear by the handful in small protein toxins containing less than 50 amino acids and by the hundreds, or even thousands, in the large proteins that make up certain biomaterials.

Figure 1.

Examples of proteins and protein complexes cross-linked by disulfide bonds. Protein backbones are represented as ribbon traces, and the Cβ and sulfur (yellow) atoms of cysteine side chains are shown as spheres. In protein complexes, different polypeptides are shown in distinct colors. PDB codes are as follows: conotoxin, 2LXG; theta-defensin, 5INZ; antifreeze protein, 1EZG; Izumo, 5JK9; ribonuclease A, 3MZR; laminin fragment, 4AQS; insulin, 2BN3; C4b-binding protein oligomerization domain, 4b0f; Kremen1-LRP6-Dickkopf complex, 5FWS.

Systems to introduce disulfide bonds into eukaryotic proteins have evolved in the endoplasmic reticulum (ER), Golgi apparatus, and mitochondria, so proteins that reside in or pass through these compartments have correspondingly evolved to contain cysteines in positions suitable for pairing. In bacteria, disulfide-generating catalysts, and thus disulfide bonded proteins, are found in the inner membrane and periplasm of Gram-negative bacteria¹ and on the membrane of certain Gram-positive bacteria.² Cytosolic proteins, whether in prokaryotes or eukaryotes, tend not to contain disulfide bonds that contribute crucially to structure or stability. However, certain thermophilic organisms provide exceptions to the generalization that cysteines in cytosolic proteins remain unpaired.^3–5 In addition, a mammalian protein that may be dual-localized to the cell surface and the cytosol has been shown to contain disulfides even in its cytosolic manifestation.⁶ Furthermore, some viral proteins produced in cytosolic “viral factories” are disulfide bonded.^7–9 Finally, cytosolic proteins may contain transient disulfide bonds, e.g., as catalytic intermediates,¹⁰ in response to chemical or thermal stress conditions,¹¹ or as a result of cell stimulation and changes in metabolic state.¹²

To provide an overview of the range of protein types modified by disulfide formation, a few examples are presented. Among the smallest are the conotoxins, a diverse set of peptides or tiny proteins (10 to 50 amino acids) produced by marine Conus snails.¹³ Most conotoxins have between two and four disulfides, and different conotoxin types have distinct cysteine pairings, leading to enormous diversity.¹⁴ Snake venoms such as disintegrins and fasciculins are also small disulfide-rich proteins.^15,16 Defensins, a class of antimicrobial peptides produced by animals and plants, have three-dimensional structures and activities that depend on disulfide formation.^17,18 The above protein families, as well as growth factors, cytokines, protease inhibitors, and other small, disulfide-rich proteins or protein domains, have been analyzed structurally and classified.¹⁹ Evidently, predation and defense are frequently based on an evolved armory of disulfide-packed bioactive peptides and proteins. Nature’s arsenal inspires investigations into engineered variants with novel pharmaceutical properties.^20–22

Disulfides are also a dominant structural and stabilizing element within larger proteins made up of tandem repeats of smaller, internally cross-linked units. Low density lipoprotein receptor-like repeats and epidermal growth factor-like (EGF-like) repeats are two examples of disulfide-rich modules that are linked in tandem to generate long, rod-shaped structures acting as cell-surface receptors or key components of the extracellular matrix (ECM). For example, laminin is an 800 kDa ECM protein that contains nearly 200 disulfide bonds; much of its architecture is based on distorted ladders of disulfides within tandem EGF-like repeats (Figure 1).^23,24 Interestingly, protein domains can be inserted into this ladder without disrupting its form,²⁵ suggesting that the packing of disulfides remains quite regular despite the sequencing variability among EGF-like repeats. Rigidity and regularity are supplied by disulfides in another scenario: antifreeze proteins. A few classes of highly active insect antifreeze protein consist of regular repeats of disulfide bonded loops, with the disulfides running through the protein cores (Figure 1). This orderly structural motif positions a precise array of threonine hydroxyl groups on the ice-binding face.^26,27

A number of fascinating, disulfide-rich biomaterials with as yet unknown structures show particularly regular placements of cysteines in their amino acid sequences, which are likely to form repeating structural elements based on disulfide bonds (Figure 2). The pairing pattern of these cysteines is not known in all cases, but disulfides clearly contribute to the functionality of the material. One such substance is hair. During the keratinization of developing fibers within the mammalian hair follicle, an intense oxidative transformation immobilizes keratin intermediate filaments within a matrix of disulfide-rich proteins.²⁸ The main contributors to this matrix are the high and ultra-high sulfur keratin-associated proteins (KRTAP), with cysteine contents up to 41%.^28–30 The amino acid sequence of a representative KRTAP, in which the (CCX₃)_n pattern is evident, is shown in Figure 2. Another cysteine-rich biomaterial is minicollagen, the protein that reinforces the walls of the venomous capsules of jellyfish, hydra, and corals.^31,32 The capsule walls require reinforcement because these nematocysts propel a harpoon into prey with enormous acceleration forces driven by pressures of >2000 psi.³³ Protein segments with the amino acid repeat pattern (CX₃)_n flank the Gly-X-Y repeats of minicollagens and contribute to formation of a resilient disulfide-bonded network. Surprisingly, the N- and C-terminal motifs have identical cysteine spacings (Figure 2) but show different disulfide connectivities.^34–36 A third cysteine-patterned structural material is the cysteine-rich eggshell membrane protein (CREMP), a major component of the proteinaceous fibers underlying the calcified shells of birds and oviparous reptiles.^37–39 The defining feature of CREMPs is repetition of a module with the pattern (CX₄CX₅CX₈CX_6/11)_n, which can extend for thousands of amino acids.³⁷

Figure 2.

Examples of diverse disulfide-rich proteins incorporated into biomaterials. The human keratin-associated protein has an ultra-high sulfur content, with cysteine constituting 37% of the residues in the amino acid sequence. In the cnidarian mini-collagens, the six-Cys motifs amino and carboxy terminal to the short collagen triple helical region (green) form disulfide-bridged networks to reinforce the pressurized nematocyst. CREMP proteins, with cysteine contents of ~11%, feature a highly repetitive and long sequence of disulfide-linked modules, eight of which are shown here.

Despite the many disulfide-rich proteins described above, disulfide cross-links are often found more sparingly in other proteins. In cases where disulfides are not the dominant structural element, they may occur within surface loops,⁴⁰ link surface loops to one another or to secondary structures,⁴¹ or bridge secondary structure elements that are part of the core fold.^42,43 Examples that illustrate how disulfides can enhance rather than dominate folds are urokinase-type plasminogen activator receptor (uPAR) and phospholipase A2 (Figure 3). In uPAR, disulfide bonds affix loops to the β-strands in the three copies of the β-sheet fold that constitute the receptor. In phospholipase A2, disulfides bridge core secondary structure elements within the fold. Most disulfides in uPAR and phospholipase A2 are at least partially surface-exposed and may therefore be accessible to disulfide catalysts or other oxidants even late in the folding process. In other structures, such as the influenza virus neuraminidase and many proteins containing immunoglobulin folds, disulfides linking core secondary structure elements are entirely buried.^42,43 This fact suggests that these disulfides are introduced prior to tertiary structure formation, while access by enzymes or small-molecule oxidants is still possible, or that the folded structure fluctuates sufficiently to permit exposure to oxidative agents.⁴⁴

Figure 3.

Two proteins demonstrating different ways in which disulfides can decorate a fold. Disulfides link surface loops to the termini of core β-strands in uPAR. Disulfides link core helices to one another and affix a β-hairpin to the helical core in phospholipase A2. PDB codes are UPAR, 2FD6; phospholipase A2, 1C74.

In addition to stabilizing domains within proteins, disulfides are also used to build larger, intermolecular assemblies. One example is the heptameric oligomerization domain of the C4b binding protein, an inhibitor of the complement system, in which neighboring subunits in the ring are disulfide bonded to one another (Figure 1).⁴⁵ Another, classic example of disulfide-mediated protein quaternary structure assembly is antibodies, in which disulfides bridge the subunits of the H₂L₂ heterotetramer and also link tetramers, and certain accessory proteins, into higher order structures such as IgA dimers and IgM pentamers.^46,47 Another physiologically important example of intermolecular disulfide-mediated cross-linking is network formation of mucins, the proteins that form the protective mucus layers coating exposed epithelia such as in the gastrointestinal tract. Mucins are giant, heavily O-glycosylated proteins that, when properly assembled into a disulfide-bridged covalent network,⁴⁸ form a physical barrier preventing the penetration of microorganisms into the epithelial cell layer.⁴⁹ Mucins contain more than 200 cysteines and undergo a complex, multi-stage assembly process. An important step in this process is disulfide-mediated trimerization in the Golgi apparatus, in which mucins arriving as dimers from the ER are recruited into a hexameric array.^50,51

3. HOW DISULFIDES STABILIZE PROTEINS

A number of issues must be considered before making generalizations regarding the effect of disulfide bonds on protein stability. Protein thermodynamic stability is the difference in free energy between the folded and unfolded states of a protein. Classically, disulfides are considered to stabilize proteins by decreasing the configurational entropy, and thus raising the free energy, of the unfolded state: two amino acids distant from one another in the primary structure of the protein but held in covalent association in three-dimensional space will dramatically lessen the number of conformations accessible to the unfolded protein. For a protein that has evolved to contain a disulfide between residues that are closely positioned in the folded state, there will be, to a first approximation, no accompanying decrease in entropy of the folded state. However, it has been observed that engineered disulfides are most stabilizing if they are introduced into regions of relatively high mobility in the native protein.^52,53 While this finding might seem to suggest that disulfides are paradoxically stabilizing when they decrease the configurational entropy of the folded state, the implication may be instead that introducing cysteines into regions that can relax to allow good disulfide stereochemistry may be less deleterious than mutation of regions that are tightly packed and thus less accommodating. In both cases, the greater effect on unfolded state entropy may be the actual stabilizing factor, but this stabilization may be undermined by disrupting favorable interactions in well-packed regions of the folded state. Understanding the effects of disulfide bonds on protein structures is important, as disulfides can be engineered into proteins to improve stability,⁵⁴ a desirable feature for protein pharmaceuticals. Engineered disulfides can also improve the functional properties of macromolecular reagents, such as the increased activity seen for certain luciferase mutants with added disulfides.⁵⁵ Disulfide engineering has further applications in the mechanistic study of protein conformational changes and their relationship to protein function.^56–58

Most natural disulfides have evolved to favor folded states of proteins, but disulfides can in principle also promote unfolded or partially folded states, or affect the kinetics of folding and assembly pathways. A disulfide has been proposed to stabilize an aggregation-prone mis-folded form of mutated γD-crystallin, contributing to cataract disease.⁵⁹ Remarkably, a study of non-synonymous codons in a related eye-lens protein, γB-crystallin, revealed that the effect of mRNA sequence on translation and folding rates can in turn influence cysteine oxidation states.⁶⁰ The order of cysteine pairing has been shown to contribute to formation of the active, metastable state of a serpin family protease inhibitor.⁶¹ Though claims for non-native disulfide formation during protein folding in vivo have been made,⁶² there is little evidence that non-native disulfides are required intermediates along most protein folding pathways. Instead, disulfides likely evolve primarily under the constraints of native-state functionality. The exception that proves the rule is the case of “pro regions” containing cysteine residues that form temporary, non-native disulfide bonds with other cysteines in the protein during folding.^63,64 Such disulfides are transient intermediates in isomerization reactions that promote escape from other, longer-lived non-native disulfide pairings that delay native folding. The evolution of cysteine-containing pro regions thus demonstrates not the importance of non-native disulfides but rather that minimizing the extent and duration of non-native disulfides is beneficial.

Interesting cases of specific disulfide rearrangements during the folding and assembly of proteins have been discovered in mechanisms to regulate the acquisition of function. One example is the E. coli lipopolysaccharide export complex, in which the major pore subunit undergoes a disulfide rearrangement and becomes functional only when triggered by the presence of a partner protein.⁶⁵ Disulfide rearrangements are also involved in the conversion from the receptor-binding form to the membrane-fusion active state of certain retrovirus surface proteins.^66,67 As a final example, a reduction or isomerization event is thought to activate integrins on the cell surface for ligand binding.^68,69

4. DETECTING DISULFIDES IN PROTEINS

A number of strategies have been used to demonstrate the presence of disulfide bonds within proteins. A change in the electrophoretic migration of a protein under reducing vs. non-reducing conditions is a common, straightforward, and high-throughput method,^62,70 but for certain proteins can be challenging or inconclusive.⁷¹ Typically, disulfide cross-links will decrease the hydrodynamic radii of polypeptides and thus increase the migration rates of proteins in gels,^72,73 but there are exceptions to this generalization, and the presence of disulfides may also decrease the migration rate in the gel.³⁷ Such exceptions may arise if disulfides decrease binding of the ionic detergents that provide the negative charge by which proteins are driven to the positive pole.⁷⁴ Disulfides may also enforce a highly elongated shape that can collapse to a faster-migrating form upon reduction.

If the electrophoretic mobility change upon reduction is insufficient or difficult to interpret, disulfides can be detected by reduction followed by modification using large alkylating groups.⁷⁵ For example, any unpaired cysteines in a protein are first alkylated with a small thiol-labeling reagent such as N-ethylmaleimide (NEM; 125 Da). Disulfides are then reduced and modified with a large thiol-labeling reagent such as polyethylene glycol (PEG) (typical range 2 to 5 kD) appended with a maleimide group. The shift in migration on a gel due to two copies of this large modification per disulfide reports on the number of disulfides that had been present prior to reduction. Alternatively, free thiols can be directly labeled by PEG, and a lack of modification at certain cysteines taken to indicate their protection in disulfide bonds, with the caveat that other explanations for lack of PEG modification must be considered. It should be noted that a PEG modification of 5 kD results in retardation on the gel comparable to about 15 kD of protein.⁷⁶ Another common reagent used to modify cysteines is 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS), which, with a molecular mass of 536 Da, causes a more modest retardation of the protein in the gel for each thiol group labeled.

Other methods are also available to detect disulfides. Disulfides can be identified from protein crystal structures⁷⁷ or using NMR-based methods.⁷⁸ Protein cysteines initially present in thiol form can be distinguished from cysteines that become reactive only upon reduction by differential labeling with small alkylating agents, which are then distinguished by mass spectrometry.⁷⁹ Liquid chromatography tandem mass spectrometry (LC-MS/MS) enables both disulfide detection and mapping of cysteine connectivity.⁷³ The presence and connectivity of disulfides can also be predicted based on the context of the cysteines in the amino acid sequence.^80,81 Lastly, disulfides can be inferred by bioinformatics studies examining the presence and conservation of encoded cysteines in genomic sequence data.^82–84

5. HOW DISULFIDES FORM

Disulfide bond formation requires oxidation of paired cysteine residues using either stand-alone, small-molecule oxidants or cofactors that are incorporated into enzyme active sites. These reagents will be described in turn below. We begin by presenting “disulfide exchange,” a common stratagem for forming a particular disulfide. In this process, no net disulfide formation occurs, but disulfides can be generated in new positions by sacrificing them at other loci.

5.1. Thiol-disulfide Exchange Reactions

A simple depiction of a thiol-disulfide exchange reaction is shown in equation 1:

$${\mathrm{R}}_{\mathrm{A}}-{\mathrm{S}}^{-}+{\mathrm{R}}_{\mathrm{B}}\text{-S-S-}{\mathrm{R}}_{\mathrm{C}}\iff {\mathrm{R}}_{\mathrm{A}}\text{-S-S-}{\mathrm{R}}_{\mathrm{B}}+{\mathrm{R}}_{\mathrm{C}}-{\mathrm{S}}^{-}$$ (1)

This reaction can be followed, according to the same principle, by a second thiol-disulfide exchange reaction as shown in equation 2.

$${\mathrm{R}}_{\mathrm{D}}-{\mathrm{S}}^{-}+{\mathrm{R}}_{\mathrm{A}}\text{-S-S-}{\mathrm{R}}_{\mathrm{B}}\iff {\mathrm{R}}_{\mathrm{D}}\text{-S-S-}{\mathrm{R}}_{\mathrm{A}}+{\mathrm{R}}_{\mathrm{B}}-{\mathrm{S}}^{-}$$ (2)

These two reactions occurring in series accomplish dithiol-disulfide exchange, in which the new disulfide (R_D-S-S-R_A) involves two cysteines (A and D) that did not participate in the original disulfide (R_B-S-S-R_C). A number of fundamental issues affect the kinetics and thermodynamics of such thiol-disulfide exchange reactions and are thus relevant to biological pathways that swap disulfide bonds, as described in subsequent sections.

Early model studies of thiol/disulfide exchange reactions demonstrated a first-order dependence on thiolate anion and disulfide^85,86 and were corroborated by a series of papers showing the S_N2 character of the reaction.^87–89 The attacking nucleophile is almost exclusively the thiolate anion; it is some 10¹⁰-fold more reactive than the protonated thiol.⁸⁷ For this reason, rapidly lowering the pH of the solution is a favored strategy for preserving the status of thiols and disulfides and preventing rearrangements prior to analysis of disulfide connectivity.⁸⁸ The pH dependence of the observed second order rate constant for the exchange reaction (k_obs) is related to the limiting rate constant at high pH values (k) as follows:

$${\mathrm{k}}_{\text{obs}}=\mathrm{k}/(1+{10}^{\text{pK-pH}})$$ (3)

While this relationship clearly identifies the importance of the pK of the nucleophile in dictating the extent to which the thiol is ionized (and therefore reactive in disulfide exchange), a series of Bronsted analyses using a range of thiols of comparable structure showed that the intrinsic nucleophilicity of the thiolates decreases with decreasing pK values.^89,90 Thus one cannot assume, for example, that an exceptionally low pK value translates to a correspondingly strong nucleophile at pH 7.5.

Theoretical studies using a range of models and solvent systems have considered whether a trisulfide (^∂-S-S-S^∂-) species represents a local minimum along the reaction coordinate for disulfide exchange.^91–95 While the outcome of these calculations depends on the model system chosen, these studies support the expectation that exchange is optimal with in-line attack of the thiolate and with rather modest change in the charge of the central sulfur atom along the reaction coordinate.^91–95

5.2. Oxidants and Oxidative Chemistry

5.2.1. Molecular Oxygen

For the net generation of disulfide bonds, other mechanisms besides thiol-disulfide exchange must come into play. Molecular oxygen is the ultimate oxidant in many pathways for disulfide bond generation. However, the direct reaction between triplet dioxygen and singlet thiol (equation 4)

$$2\text{R-SH}+{\mathrm{O}}_2->\text{R-S-S-R}+{\mathrm{H}}_2{\mathrm{O}}_2$$ (4)

is formally spin-forbidden and kinetically sluggish.⁹⁶ Autoxidation in aqueous solutions is greatly accelerated by traces of redox-active transition metal ions, most notably copper and iron, which are persistent contaminants of solution components and glass and plastic surfaces. Metal chelating reagents such as EDTA strongly suppress, but do not totally eliminate, the background thiol oxidation in aerobic solutions. The best understood system, the copper-catalyzed oxidation of cysteine, is of considerable kinetic complexity.^97–101 Oxidation proceeds via cysteine chelates and generates hydrogen peroxide at neutral pH values. Further thiol oxidation can then occur as described in section 5.2.2 below.

Given the precedents for such metal-catalyzed thiol oxidation in the chemical literature, and considering the many metal-dependent oxidoreductases in oxidative transformations, the dearth of well-documented metalloenzymes catalyzing disulfide formation is striking. The earlier literature contained a number of reports of both iron- and copper-dependent sulfhydryl oxidases, but none of them have been unequivocally confirmed.^102–104 Indeed, reinvestigation of sulfhydryl oxidases from both categories have established that their apparent metal content was adventitious, and that thiol oxidation is catalyzed instead by a flavin redox center.^102,103 Thus, while bona fide metal-dependent sulfhydryl oxidases capable of sustained turnovers would be expected to exist, they have yet to be definitively described.

5.2.2. Hydrogen Peroxide

The reaction between hydrogen peroxide and thiols has received sustained attention because of its importance in the context of oxidative stress and signaling in biology.¹⁰⁵ Oxidation by hydrogen peroxide comprises two spin-allowed steps:

$${\mathrm{H}}_2{\mathrm{O}}_2+\text{R-SH}->{\mathrm{H}}_2\mathrm{O}+\text{R-S-OH}$$ (5)

$$\text{R-S-OH}+\text{R-SH}->\text{R-S-S-R}+{\mathrm{H}}_2\mathrm{O}$$ (6)

Sulfenic acid formation (equation 5) involves nucleophilic attack by the thiolate and is slow for simple low molecular weight thiols (with second order rate constants of about 25 M⁻¹s⁻¹)^106,107 but can be greatly accelerated enzymatically. Subsequent disulfide bond generation (equation 6) is then inherently more rapid.

5.2.3. Dehydroascorbate

Dehydroascorbate (DHA) is reduced by reduced glutathione generating ascorbate (AA) and disulfides^108,109 with the stoichiometry:

$$2\text{G-SH}+\text{DHA}->\text{G-S-S-G}+\text{AA}$$ (7)

The reaction likely involves a thiohemiketal intermediate at C2 that can be resolved using a second nucleophilic attack (Figure 4).^109–111 DHA has been described as a direct oxidant of a range of thiols, including unfolded reduced proteins and protein disulfide isomerase.¹¹²

Figure 4.

Oxidation of thiols by dehydroascorbate. Generation of a thiohemiacetal intermediate is followed by capture of the adduct by a second thiolate species to yield ascorbate and the disulfide, RS-SR.

5.2.4. Quinones

Certain quinones serve as electron carriers in biological systems,¹¹³ whereas others are found as toxic components of environmental pollutants.^114,115 Quinones can form thiol adducts, catalyze the generation of reactive oxygen species, and oxidize thiols to disulfides (Figure 5).^116–118 The major systems for disulfide bond formation in the periplasm of bacteria rely on ubiquinone cofactors, and the chemistry of ubiquinone reactivity with cysteines has been described most thoroughly in these systems (see sections 6.3.1 and 6.3.2). Reduction of the quinone (form 1) is initiated by formation of a Michael adduct (form 2) followed by resolution using a second thiolate to yield the hydroquinone and the corresponding disulfide (form 4; Figure 5).^117,118

Figure 5.

Mechanism of disulfide bond formation by 1,4-quinones. Attack of a thiolate on quinone (1) forms the Michael adduct (2) which can be resolved as in (3) to generate the reduced cofactor and the corresponding disulfide, RS-SR.

5.2.5. Flavins

Free oxidized flavins are minimally reactive towards thiols under physiological conditions, but they contribute to catalysis of disulfide bond formation as tightly bound prosthetic groups in enzymes. While flavin-linked enzymes will be discussed more fully later, the key chemistry of flavin-mediated disulfide bond generation is summarized here (Figure 6). Precedents for these steps come from early mechanistic studies of the flavin-dependent thiol/disulfide oxidoreductases dihydrolipoamide dehydrogenase (DLD) and glutathione reductase (GR),^119–121 and parallels with the mechanism for quinone oxidation of thiols can be seen. Among the mechanistic commonalities are the thiolate-to-flavin charge-transfer complexes that precede the nucleophilic attack on the flavin depicted in Figure 6, form 1. Transitory C4a adducts (Figure 6, form 2) are intermediates in the net reduction of the flavin prosthetic group in other flavin-linked disulfide oxidoreductases^122–125 and are likely to form in sulfhydryl oxidases as well. The corresponding adduct is then resolved by attack from a second thiolate species (Figure 6, form 3) leading to the generation of the dihydroflavin species (Figure 6, form 4) and liberation of the corresponding disulfide (R-S-S-R).^122,123,126 In canonical sulfhydryl oxidases the reduced flavin is reoxidized by molecular oxygen via one-electron chemistry.^127,128 All sulfhydryl oxidases release substantial to stoichiometric levels of hydrogen peroxide;^129–131 however, in some cases a minor proportion of this peroxide comes from the superoxide ion that subsequently dismutes after release from the active site.¹³² While the nominal electron acceptor for the sulfhydryl oxidases is molecular oxygen, other electron acceptors may function in vitro or may substitute for oxygen in particular biological contexts.^131,133,134

Figure 6.

Mechanism of disulfide formation by flavin cofactors. The isoalloxazine ring system (1) can be attacked by a thiolate nucleophile at the C4a position to yield a thiol-flavin adduct (2). Resolution of this species with a second thiolate (3) leads to the generation of the disulfide. The resulting dihydroflavin (4) can be oxidized in two one-electron steps by molecular oxygen or by other oxidants.

6. HOW ENZYMES CONTRIBUTE TO THIOL OXIDATION REACTIONS

Here we revisit thiol-disulfide exchange reactions (section 5.1) and cofactors (sections 5.2.4 and 5.2.5) in the context of enzyme active sites. We will discuss how enzyme reactive groups are positioned to facilitate catalysis of disulfide formation and how other enzyme features promote substrate specificity.

6.1. Protein Disulfide Isomerase (PDI) and Other Thiol-disulfide Oxidoreductases

PDI family enzymes engage in thiol-disulfide rearrangements for disulfide introduction in proteins, and many principles can be gleaned from an examination of their active sites.

6.1.1. Geometry and Chemistry of Thiol-disulfide Exchange in the Context of PDI Active Sites

The PDI family is a large set of proteins (more than 20 in humans) that catalyze disulfide formation, isomerization, and in certain cases reduction of substrate proteins in the ER.^135–136 PDI family enzymes contain domains with the same fold as thioredoxins,^137,138 which are proteins with disulfide reductase activity.¹³⁹ In many enzymes of the PDI family, multiple thioredoxin-fold (trx) domains are linked in tandem, and some PDI enzymes contain other domain types as well. PDI itself contains four trx domains, the first and last of which contain redox-active CXXC motifs. The domain organization of PDI is referred to as a-b-b′-a′, with “a” designating redox-active domains and “b” redox-inactive domains. Structural differences between members of the PDI family that could lead to functional specialization have been reviewed.^138,140 As PDI enzymes use thiol/disulfide exchange for catalysis, they must fulfill the fundamental principles outlined above in the introduction to thiol/disulfide exchange reactions (section 5.1). Additional features may also be present to further enhance PDI-mediated reactions.

As in thioredoxin, the PDI active site contains a disulfide formed by the two cysteines in a CXXC motif, typically with the sequence CGHC. We will focus here on geometrical and chemical requirements for the catalysis of disulfide bond formation by PDI proteins using this motif. First, however, it is important to note that the presence of a CXXC motif in a thioredoxin-fold domain of a PDI family protein is not evidence of redox activity. S. cerevisiae Eps1p contains a domain with a CXXC motif in the appropriate primary structural context, but the structure of the enzyme shows the disulfide to be buried and unreactive, and the adjacent proline residue, while maintained, is found in the trans rather than the cis configuration.¹⁴¹

Most fundamentally, PDI and other enzymes that undergo thiol-disulfide exchange with substrate proteins should conform to the steric requirements of the reaction.¹⁴² In-line attack in the context of structured globular proteins would predict alignment of the catalytic disulfide axis as depicted schematically in Figure 7A. This arrangement allows an approaching nucleophilic thiolate from the substrate to form what is known as a “mixed” disulfide (e.g., a disulfide between a substrate and an enzyme or between two redox-active sites within a single protein or protein complex) without the need for substantial conformational rearrangement of the disulfide-containing peptide loop in the PDI protein.^104,143 In contrast, an enzyme disulfide with an axis parallel to the protein surface would not readily engage in mixed disulfide formation (Figure 7B). PDI catalytic domains, like other members of the thioredoxin superfamily, position the redox-active CXXC disulfide at the amino terminus of a long surface helix in an orientation consistent with geometrical predictions (Figure 8). The sulfur atom of the N-terminal cysteine of the CXXC motif (S1 in Figure 8), also known as the interchange cysteine, is solvent accessible, whereas its redox partner (S2) is typically buried within the protein.

Figure 7.

(A) Two proteins with cysteine sulfurs oriented as shown are able to engage in an in-line SN2 reaction, form a mixed disulfide, and complete a thiol-disulfide exchange reaction. (B) The orientation of cysteine sulfurs shown is incompatible with thiol-disulfide exchange.

Figure 8.

Representative active site of a PDI family protein. Amino acids commonly observed in PDIfamily active sites are shown in stick representation, with red indicating oxygen atoms and blue nitrogen. The Phe side chain is replaced by Tyr in some cases. Potential interactions of the largely buried S2 sulfur of the CXXC motif are indicated by dashed lines (black—hydrophobic; blue—hydrogen bonding; red— proton transfer). Green arrows indicate sites for hydrogen bonding interactions by the incoming substrate backbone. Red sphere is water. PDB code is 3ED3 (yeast Mpd1p).

Additional features contribute to the success of PDI family proteins in catalyzing thiol-disulfide exchange reactions. One such feature is the positioning of nearby functional groups to promote mixed disulfide formation. The second of the CXXC cysteines (S2) is released as the mixed disulfide between PDI and substrate forms (Figure 7A), and the resulting thiolate would be nestled 3 to 4 A from the three side-chain Cβ atoms of a conserved cis-proline on a neighboring loop and two nearby aromatic residues (Figure 8), an apparently unfavorable environment for development of negative charge. Even deeper within the protein, however, is a conserved glutamic acid in position to perform either a direct or a water-mediated proton transfer to the thiolate. An additional feature that may aid formation of a thiolate involving sulfur S2 is the proximity of a backbone N-H group at the carboxy terminus of the β-strand containing the buried acidic residue (Figure 8).

The PDI active site must not only be suitable for attack by a substrate thiolate to form the mixed disulfide between enzyme and substrate, it must also enable attack of this mixed disulfide by a second substrate thiolate and release of the substrate with a newly formed disulfide. Therefore, the mixed disulfide must be accessible for this second SN2 reaction. Structural information is starting to become available for intermediates in PDI thiol-disulfide exchange reactions. Stabilizing such intermediates typically requires mutation of cysteines that could potentially attack the mixed disulfides, so part of the chemical environment is sacrificed for the sake of illuminating the functional groups that remain. A few examples of mixed disulfides involving PDI family proteins stabilized by mutagenesis in this manner have been crystallized, including the PDI family protein ERp57 in association with tapasin,¹⁴⁴ which is a component of the peptide-loading system for the class I major histocompatibility complex. In addition, structures are available for the PDI proteins P5 and ERp46 linked to a peptide derived from peroxiredoxin Prx4 (Figure 9),¹⁴⁵ and for ERp44 linked to Prx4.¹⁴⁶ These structures demonstrate how substrate peptides engaging in mixed disulfides with PDI proteins form extended hydrogen bonding interactions with the enzyme loop containing the cis-proline. A basic side chain from the neighboring loop may also make interactions with a substrate backbone carbonyl (Figure 9). Similar principles are observed for analogous complexes involving thioredoxins, except that the backbone of the neighboring loop, rather than a side chain, participates in hydrogen bonding to the substrate backbone.^147,148 These interactions keep the substrate backbone low and tight against the enzyme surface. With the substrate main chain held to one side of the active site, the substrate cysteine is fully exposed for a second in-line attack (Figure 9).

Figure 9.

Model for substrate interaction with a PDI protein. The P5 protein (gene name PDIA6) is in gray, and the backbone (including a proline side chain) of a peptide from peroxiredoxin Prx4 is in magenta. The cysteine side chain atoms of Prx4 are labeled Cβ and S. Blue dashed lines are hydrogen bonds. Red in stick representations indicates oxygen atoms, blue nitrogens. The S1 sulfur of P5 is labeled. PDB code is 3W8J.

The final step in the electron transfer reaction that should be promoted by PDI is formation of the disulfide in the substrate and release of the reduced enzyme. The formation of a thiolate at S1 in the enzyme may be promoted by the location of the cysteine at the amino terminus of the active-site helix, i.e., at the positive end of the helix dipole. In addition, the histidine in the CGHC sequence, the presence of which raises the redox potential of the motif,¹⁴⁹ may help stabilize the S1 thiolate.

6.1.2. Oxidation vs. Isomerization

The above description of features that promote catalysis by PDI focused on substrate oxidation. Disulfide isomerization (Figure 10) is another activity catalyzed by PDI enzymes.¹⁵⁰ Like oxidation, isomerization occurs by formation of a mixed disulfide. The difference is that reduced PDI engages a substrate disulfide to initiate isomerization, rather than oxidized PDI acting on free substrate cysteines to pair them. The isomerization reaction may be promoted by the inherent reactivity of a disulfide between erroneously paired cysteines in a mis-folded substrate protein. The PDI-substrate mixed disulfide can then be resolved by nucleophilic attack from another substrate cysteine, forming a new substrate disulfide and releasing PDI again in the reduced state. It should be noted that the isomerization reaction is indistinguishable mechanistically from an oxidation reaction once the mixed disulfide has formed. Whether a PDI family enzyme will carry out oxidation or isomerization to promote native disulfide bond formation depends on the state of the cysteines in the substrate and the availability of the reduced form of the PDI enzyme. The presence of reduced PDI will depend on its redox potential and on various kinetic considerations, discussed in section 7.5 below.

Figure 10.

Oxidative protein folding comprises two conceptual steps: oxidation and isomerization of mispaired disulfides. These steps are schematically depicted here for the introduction and rearrangement of a single disulfide bond.

Interestingly, mutagenesis can be used to uncouple oxidation and isomerization in PDI proteins, as amino acid residues in the vicinity of the PDI active site selectively influence the isomerization vs. oxidation reactions.¹⁵¹ Using mutants that can oxidize but not isomerize, it was shown that the essential activity of PDI in yeast is oxidation and not isomerization.¹⁵²

6.1.3. Ligand Binding and Chaperone Activity of PDI

In addition to performing catalytic activities based on thiol-disulfide exchange, PDI and some of its family members also possess ligand binding capabilities that contribute to substrate specificity and promote interaction between the PDI redox-active sites and target cysteines. Both redox-active and -inactive domains of PDI proteins participate in substrate recognition and alignment.^153–155 Particularly notable is a large patch of hydrophobic surface on the b′ domain facing the interior of the U-shaped PDI structure (Figure 11).^156,157 This patch is a principal docking site for exposed hydrophobic residues in unfolded or mis-folded substrates.^153,156,158 Similar hydrophobic clefts or patches are found in other PDI family proteins with different domain composition.¹⁵⁹ Recently, sophisticated models have been put forth for redox-dependent substrate binding and for how substrate interaction with non-catalytic domains influences catalytic activity at the redox centers.^160–162 For example, the discovery that binding of certain small molecules to the PDI hydrophobic patch interferes with reduction of the protein substrate insulin but enhances reduction of a fluorescent GSSG analog suggested that engaging the hydrophobic site triggers an allosteric switch enhancing PDI reductase activity.¹⁶² Small-molecule binding allowed the two effects to be dissociated, but physiological protein substrates likely fulfill both roles, first binding the PDI hydrophobic patch and then benefitting from augmented catalytic activity at the PDI redox-active sites.

Figure 11.

Structure of PDI. The image shows S. cerevisiae PDI. Thioredoxin fold domains are labeled according to conventional nomenclature (a-b-b′-a′). Cysteine side chains are shown as spheres with yellow sulfurs. A set of exposed hydrophobic residues in the b′ domain that may constitute a binding site for folding or misfolded proteins is shown in space-filling representation (carbons in cyan). PDB code is 2B5E.

Attempts have been made to produce small-molecule catalysts that mimic the activity of PDI in promoting native disulfide bond formation and activity of substrate proteins.^163,164 Effective molecules of this type would make cheap and readily removable reagents for industrial production of disulfide bonded proteins, which include antibodies and hormones for clinical use. For basic science, success in matching the activity of PDI with a small molecule might indicate which features of PDI are key for its activity. A recent effort at producing such a reagent took into consideration not only the cysteine pK values and reduction potential of the disulfide in the designed molecules but also aimed to incorporate the capability of binding hydrophobic sites in target proteins.¹⁶⁵ A series of dithiols decorated with various hydrophobic groups was produced, and increasing ability to refold the model substrate RNase A was observed with increasing hydrophobic bulk, until the solubility of the reagent became limiting. However, the best small-molecule catalysts did not match the improvement in folding achieved by PDI itself, suggesting that the broad binding surfaces offered by protein catalysts and the cooperation between substrate binding and redox activities described above are important for catalysis of proper disulfide bond formation in substrate proteins. Given the highly evolved and coordinated activities of PDI, it is not surprising that small molecule oxidants do not achieve equal success in oxidative protein folding.

6.1.4. Mia40, a Thiol-Disulfide Oxidoreductase with a Different Scaffold

Mia40 is an enzyme that performs thiol-disulfide exchange reactions in the mitochondrial intermembrane space (IMS) to oxidize substrates and is a rare thiol thiol-disulfide oxidoreductase not based on the trx fold. Instead, Mia40 structures from yeast and human show a helix-turn-helix motif stapled by two disulfides and an amino-terminal flexible region housing the redox-active CPC motif (Figure 12).^166,167 Disulfides within CXC motifs have been shown to be relatively unstable,¹⁶⁸ but the CPC disulfide is the most stable of the CXC disulfides, perhaps because the restricted degrees of freedom in the backbone due to the presence of the proline limit the entropic gain upon reducing the disulfide. A redox potential of approximately −200 mV has been reported for the Mia40 CPC motif,¹⁶⁷ compared to a value of about −165 mV for PDI and −167 for a peptide with the sequence CGC.^142,163 Like the active site of PDI, the CPC motif of Mia40 is solvent exposed, and the disulfide is available for attack along the preferred axis (Figure 12). Reduced substrates of Mia40 bind to a shallow hydrophobic depression on the protein surface.^166,167,169 The rate of mixed disulfide formation then appears to depend on the presence of hydrophobic residues in the substrate immediately adjacent to the attacking cysteine thiolate.¹⁷⁰ The ultimate resolution of the mixed disulfide and formation of a new substrate disulfide depends on attack by a second substrate cysteine. It has been noted that Mia40 mixed disulfides with substrate proteins are longer-lived than the typically transient encounters made by PDI.¹⁷¹

Figure 12.

Structures of the Mia40 oxidoreductase. Sulfurs in the CPC motif are labeled S1 and S2. Gray arrow in the oxidized structure illustrates solvent accessibility of the redox-active disulfide. PDB codes are 2K3J (reduced; H. sapiens Mia40 solution NMR structure) and 2ZXT (oxidized; S. cerevisiae Mia40 crystallized as a fusion with maltose binding protein (not shown)).

6.2. Peroxiredoxins/Glutathione Peroxidases

Peroxiredoxins and glutathione peroxidases catalyze the reduction of peroxides to water, and a byproduct of this reaction is formation of a disulfide bond in the enzyme active site. Peroxidases located in cellular compartments in which oxidative protein folding occurs couple active-site regeneration to a second catalytic activity: the formation of disulfide bonds in substrate proteins.

6.2.1. Peroxiredoxin

Peroxiredoxins are a widespread protein family studied for their potential roles as antioxidants in cancer, neurodegeneration, and inflammatory disease. Many of these enzymes also function to sense and transduce redox signals borne by H₂O₂.¹⁷² Here we will focus on the role of peroxiredoxin as a catalyst of disulfide generation for oxidative protein folding. The peroxiredoxin relevant to disulfide formation is Prx4 of the ER,¹⁷³ as most other peroxiredoxins are found in locations, i.e., the cytosol, where disulfide bond formation is not integral to protein biosynthesis. Prx4 is a ring-shaped homodecamer, and active-sites are formed between subunit pairs. The peroxidatic cysteine of Prx4 reacts very rapidly (~ 2 × 10⁷ M⁻¹s⁻¹) with H₂O₂,¹⁷⁴ the end of the helix containing the newly sulfenylated cysteine unfolds, and another enzyme thiolate from the neighboring subunit attacks the sulfenylated cysteine to form a disulfide (Figure 13).¹⁷⁵ Reduction of this disulfide by thiol-disulfide exchange with a PDI family enzyme allows the oxidizing equivalents gained by reducing hydrogen peroxide to be funneled toward oxidative protein folding.^{145,176–178}

Figure 13.

Prx4 peroxidase cycle. Isolated segments of two neighboring Prx4 subunits are shown (purple and orange). Spheres are the side chains of cysteines participating in catalytic disulfide formation, with sulfurs colored yellow and Cβ atoms colored according to the parent chain. Based on PDB codes 3TJG and 3TJF.

6.2.2. Glutathione Peroxidase

Gpx7 and Gpx8 are the glutathione peroxidase family members located in the ER. Glutathione peroxidases are thioredoxin fold superfamily proteins,¹⁷⁹ and though some glutathione peroxidases contain selenocysteine as the active-site residue, Gpx7 and Gpx8 do not. These two enzymes are better described as PDI peroxidases, as their primary reductant appears to be PDI rather than glutathione.^180–181 Gpx8 has been reported as an important detoxifier of hydrogen peroxide produced by sulfhydryl oxidase activity in the endoplasmic reticulum.¹⁸² Studies are ongoing to identify the physiological functions and physiologically relevant mechanisms of these peroxidases,¹⁸³ as well as the respective roles of Gpx enzymes and Prx4.^175,182

6.3. Quinone-containing Enzymes

Seminal studies on oxidative protein folding in the bacterial periplasm led to the recognition of two classes of polytopic membrane protein that utilize quinones as cofactors of enzyme-mediated oxidation of thiols (Figure 14).^82,184–188 These classes are represented by the enzymes DsbB and homologs of vitamin K epoxide reductase (VKOR). Bacteria that form disulfide bonds in proteins exported from the cytosol are partitioned into those containing DsbB and those with a VKOR homolog.⁸² DsbB and VKOR share a similar arrangement of transmembrane helices and active-site chemistry, but their polypeptide chain termini are found at different positions in the helical bundle. These observations led to the suggestion, backed by a phylogenetic study, that DsbB and VKOR are related by circular permutation from a common ancestor.¹⁸⁹ Disulfide bond formation in both the DsbB and VKOR involves a stepwise reduction of an enzyme-bound 1,4-quinone species according to the mechanism shown in Figure 5.

Figure 14.

Active-site regions of quinone-containing disulfide catalysts. Quinone-proximal disulfides are shown as spheres with sulfur atoms in yellow and Cβ atoms in green. The quinone (orange sticks) is buried in the core of the helical bundle in each enzyme, and the amino-terminal cysteine of the quinone-proximal disulfide (S1) is exposed to in-line nucleophilic attack. PDB codes are 2ZUP and 3KP9.

6.3.1. DsbB

The DsbB protein from the periplasm of E. coli is the best understood mechanistically of the quinone-containing catalysts of disulfide bond formation. A series of disulfide exchange reactions transmit reducing equivalents to the quinone cofactor and ultimately to the respiratory chain (Figure 15A). The immediate reductant of DsbB is DsbA,^184,185 a thioredoxin-fold protein¹⁹⁰ that collects reducing equivalents from nascent chains undergoing oxidative protein folding in the periplasm. Mutagenesis, crystallographic and pre-steady state kinetic approaches have provided key mechanistic insight into the DsbA-DsbB system.^191–195 The first observable intermediate when reduced DsbA is mixed with oxidized DsbB is a thiolate-quinone charge-transfer intermediate.^194,196,197 This thiolate-to-quinone interaction is stabilized both by the charge-transfer interaction per se and via an ionic interaction with the side chain of a nearby arginine. This complex is resolved by disulfide exchange reactions that lead to rate-limiting release of DsbA and formation of the reduced quinone cofactor.¹⁹⁶

Figure 15.

(A) Schematic representation of the DsbA-DsbB system showing the key charge-transfer intermediate observed when reduced DsbA is mixed with oxidized DsbB. Reduction of the quinone is rate-limiting and is coupled to the release of oxidized DsbA. The quinone cofactor is shown as a hexagon, in which blue represents a charge-transfer state. (B) Thermodynamics along the reaction coordinate for the events depicted in panel A. The redox potential for DsbA would appear insufficiently reducing to efficiently transfer electrons to the first disulfide in DsbB. (C) While the net conversion of reactants to products is thermodynamically unfavorable, the mixed disulfide intermediate between them can be highly populated (black line) or destabilized (grey line). Selective stabilization of mixed disulfide intermediates may allow the formation of products if those products are in turn depleted by other favorable reactions, such as electron transfer to quinone or oxygen.

Analysis of the pathway whereby reducing equivalents are relayed from DsbA to the DsbB quinone highlights a striking (> −120 mV) thermodynamic mismatch between the reduction potential of the DsbA active-site di-cysteine motif and its target disulfide within DsbB (Figure 15B). However, while the net reduction of the first disulfide in DsbB would be unfavorable by a factor >10⁴, a mixed disulfide intermediate between them could be quite stable. This point is illustrated schematically in Figure 15C: the stability of mixed disulfides is not constrained by the energy differences between reactants and products.¹⁹⁸ Hence a series of such disulfide exchanges in DsbA-DsbB releases the cofactor-proximal thiolate to react with the strongly oxidizing quinone center.¹⁹⁹ This dynamic disulfide relay allows efficient redox coupling between a modest reductant, DsbA, and a distal electron-deficient center without having to traverse the formal redox barriers depicted in Figure 15B. Given the prevalence of disulfide relays in the enzymes of oxidative protein folding, this principle is likely to be exploited in other instances involving apparently mismatched redox partners.¹⁹⁸

6.3.2. VKOR

The VKOR enzyme family is named for its function of reducing vitamin K epoxide generated during enzymatic carboxylation of glutamic acid residues in factors involved in blood coagulation, bone development, and other physiological processes. In mammals, reduced PDI generated by oxidative protein folding can provide reducing equivalents to VKOR for reduction of vitamin K1 and K2 and their corresponding epoxides.^200–204 Notably, vitamin K-dependent enzymes are not restricted to mammals, suggesting that mechanisms for vitamin K reduction are needed in other organisms. For example, Conus snails contain a vitamin K-dependent carboxylase,²⁰⁵ which makes post-translational modifications to a class of disulfide-poor toxins,^206,207 a distinct set from the disulfide-rich toxins described in section 2 above. VKOR-like sequences can be identified in various marine invertebrates including Conus bullatus, and, though a diverse and rapidly evolving set of PDI proteins has been identified in Conus snails,²⁰⁸ whether the activities of PDIs and VKOR-like enzymes are linked in these organisms is not known. Notably, many VKOR homologs reduce only quinones (ubiquinone or menaquinone) and do not use vitamin K epoxide as a substrate. Despite the differences in electron acceptors, various VKOR family enzymes likely share structural and certain mechanistic properties.^209,210

The VKOR oxidoreductases are also mechanistically similar to the DsbB enzymes; like DsbB, they feature catalytic disulfides in redox communication with a quinone cofactor embedded in a transmembrane domain.^211,212 Some VKOR enzymes accept electrons from DsbA-like proteins.²¹³ In other VKOR-like enzymes, a trx domain is fused carboxy-terminally to the transmembrane domain (Figure 16),²¹⁴ from where it presumably substitutes for DsbA. Structures of the Synechococcus VKOR enzyme revealed the nature of the association of the transmembrane and trx domains and suggested how electrons accepted from protein clients are transferred to the quinone-proximal cysteines (Figure 15).^183,194 This interdomain transfer is mediated by a loop in the transmembrane domain bearing two cysteines and sandwiched between the main redox-active sites. A set of mutations were made in the VKOR electron transfer pathway to trap various mixed-disulfide intermediates and illuminate the conformational changes necessary to obtain them.¹⁹⁶ One major intramolecular electron-transfer intermediate was recently proposed as the binding target of the anticoagulant warfarin.²¹⁵

Figure 16.

Schematic representation of internal electron transfer events in a bacterial VKOR homolog. The quinone cofactor is shown as a hexagon, in which blue represents a charge-transfer state. It is not known to what extent the electron-transfer steps are concerted in VKOR enzymes, i.e., whether the electron-transfer loop is simultaneously disulfide bonded to both a trx domain active-site cysteine and the partner of the charge-transfer cysteine in the transmembrane domain. For comparison with Figure 14, the trx domain is shown to the left of the transmembrane domain, but it should be noted that the trx domain is fused carboxy terminally to the transmembrane domain.

6.4. Flavin-containing Enzymes

6.4.1. General Principles and Commonalities

All sulfhydryl oxidases described to date have been shown to be flavin-linked enzymes catalyzing the net reaction:

$$2\text{RSH}+{\mathrm{O}}_2->\text{R-S-S-R}+{\mathrm{H}}_2{\mathrm{O}}_2$$ (8)

Three broad categories have been identified: the Ero1 enzymes,^216–218 the Erv-fold proteins including members of the QSOX family, and enzymes based on the pyridine nucleotide disulfide oxidoreductase scaffold.^219–223 Within these categories, evolution of associated domains and quaternary structure assembly modes has produced further diversity in structure (Figure 17). Despite their many differences, sulfhydryl oxidases share the unifying feature of a catalytic disulfide bond formed from closely spaced cysteines (separated by two or four residues) in redox contact with the isoalloxazine ring of the flavin adenine dinucleotide (FAD) prosthetic group, similar to the positioning of a redox-active disulfide close to the quinone in DsbB and VKOR. Four representative structures of flavin-containing sulfhydryl oxidases, emphasizing this proximity, are shown in Figure 18. In addition to the funneling of reducing equivalents through a protein disulfide dedicated to direct communication with a bound cofactor, another similarity between flavin- and quinone-containing catalysts of disulfide bond formation is the presence of a second enzyme disulfide that mediates electron transfer between substrate thiols and the FAD-proximal disulfide. Such “shuttle” disulfides may help accommodate bulky incoming thiol-containing substrates. They also may allow substrate specificity to evolve partially independently of the strict geometric requirements of the active site. Lastly, shuttle disulfides may minimize undesirable changes in active-site polarity accompanying substrate docking.

Figure 17.

Gallery of sulfhydryl oxidase flavoenzyme structures. Protein subunits are purple and green, FAD is orange, disulfide bond sulfurs are yellow. Ero1, Saccharomyces cerevisiae Ero1; QSOX1, Rattus norvegicus Quiescin Sulfhydryl Oxidase 1; QSOX (T. brucei), Trypanosoma brucei Quiescin Sulfhydryl Oxidase; ALR, Homo sapiens Augmenter of Liver Regeneration; AcMNPV, Autographa californica multicapsid nucleopolyhedrovirus; ASFV, African swine fever virus; AfGliT, Aspergillus fumigatus Gliotoxin Sulfhydryl Oxidase; APMV, Acanthamoeba polyphaga mimivirus. With the exception of APMV R596, dimer structures are viewed down the two-fold axis. PDB codes are Ero1, 1RP4; QSOX1, 4P2L; QSOX (T. brucei), 3QCP; ALR, 1OQC; AcMNV Ac92, 3QZY; ASFV pB119L, 3GWL; AfGliT, 4NTC; APMV R596, 3GWN.

Figure 18.

Active-site regions of flavin-dependent sulfhydryl oxidases. Flavin-proximal disulfides are shown as spheres with sulfur atoms in yellow and Cβ atoms in gray. The sulfur in each enzyme targeted for nucleophilic attack by an incoming thiolate during catalysis (the interchange sulfur) is labeled S1. The relative solvent exposure of the S1 sulfurs and accessibility of the disulfide to in-line attack is evident in ALR, QSOX1, and GliT. Only in Ero1 is the FAD-proximal disulfide buried and the S1 cysteine apparently inaccessible. PDB files are ALR, 1OQC; QSOX1, 3LLI; Ero1, 1RP4; GliT, 4NTC.

6.4.2. Erv/ALR Enzymes

The designation Erv/ALR arises from co-identification of this enzyme family in yeast and in mammals. A yeast enzyme was named Essential for Respiration and Viability 1 (Erv1) due to the impact of mutations on mitochondrial function,^224,225 and a mammalian homolog was called Augmenter of Liver Regeneration for its activity in promoting hepatocyte proliferation during development or upon partial hepatectomy.²²⁶ These proteins were found to be sulfhydryl oxidase flavoenzymes.²²⁷ Fungi, animals, and plants have an Erv/ALR family enzyme localized to the mitochondrial IMS, where it contributes to disulfide bonding of IMS proteins with the aid of Mia40 (section 6.1.4). S. cerevisiae has an additional Erv/ALR paralog localized to the endoplasmic reticulum.^228–229

Erv/ALR enzymes conform to the principles common to flavin-based sulfhydryl oxidases by juxtaposing a CXXC motif with the flavin isoalloxazine (Figure 18, ALR). They are compact dimers consisting of little more than the few helices encasing the FAD (Figure 17, ALR). The two FAD-binding subunits are packed with C2 symmetry, and the active sites are exposed on opposite ends of the assembly. Electron delivery from substrate to the FAD-proximal cysteines is mediated by a shuttle disulfide tethered to the amino- or carboxy-terminus of the enzyme and interacting with the opposite subunit in the dimer (Figure 19).^230–232 On the other end of the internal electron-transport pathway, many Erv/ALR enzymes use oxygen as a preferred electron acceptor, but others reduce cytochrome c more rapidly than they reduce oxygen.²³³ A preference for cytochrome c may arise when a two-electron reduction of one of the FAD molecules bound in the Erv/ALR enzyme dimer is followed by comproportionation with the second FAD, leading to two one-electron reduced cofactors. Each active site can then perform a one-electron reduction of cytochrome c.^{231,233–237} In this reaction, the Erv/ALR enzyme is technically not functioning as a sulfhydryl oxidase, but because of the close evolutionary and structural relationship between Erv/ALR enzymes that utilize oxygen efficiently and those that do not, they are typically classified and discussed together.

Figure 19.

Schematic of an Erv/ALR dimeric enzyme with shuttle disulfides delivering electrons, derived from substrate oxidation, to the opposite subunit in the dimer.

In addition to the Erv/ALR orthologs localized to mitochondria or the ER, enzymes from this family are encoded by viruses. These enzymes appear to be expressed in the cytosol of virus-infected cells or in viral factories,²³⁸ which are organelle-like loci established within the cell cytosol to coordinate massive levels of DNA replication and virion self-assembly.²³⁹ Large, double-stranded DNA (dsDNA) viruses encode Erv/ALR family sulfhydryl oxidases as part of their core gene set, but the extent to which the oxidases have diversified in these viruses is extraordinary. Specifically, the tertiary and quaternary structural contexts in which the Erv/ALR module appears is highly variable.²⁴⁰ The canonical dimerization mode seen in cellular Erv/ALR enzymes is found in the mimivirus sulfhydryl oxidase R596, but the two subunits associate in completely different manners in the African swine fever virus and baculovirus enyzmes (Figure 20). The viral Erv/ALR enzymes thus provide a counterexample to the generalization that quaternary structures usually change by accretion of symmetries rather than by abandoning evolved interfaces and acquiring new packing modes.²⁴¹ Furthermore, virus Erv/ALR sulfhydryl oxidase modules can exist either as stand-alone domains or embedded within divergent sequence contexts. It remains to be determined whether the cysteines in the regions outside the Erv/ALR domain serve as electron shuttles to the FAD-proximal active-site cysteines, or whether other viral proteins partner with them in an extended electron-transfer pathway. The Erv/ALR sulfhydryl oxidase from poxviruses functions together with a small partner protein,²⁴² but the relationship of these two proteins in terms of electron transfer is not clear. Structural information is not yet available on the poxvirus disulfide formation enzymes, but sequence analysis suggests that they will demonstrate even further structural diversity than that already observed among the viral disulfide catalysts.²⁴³

Figure 20.

Incorporation of the Erv/ALR module into viral sulfhydryl oxidase dimers. Different dimerization modes are used in each case, and the module is embedded in different tertiary structural contexts. ASFV, African swine fever virus; AcMNPV, Autographa californica multicapsid nucleopolyhedrovirus; APMV, Acanthamoeba polyphaga mimivirus. PDB codes are ASFW pB119L, 3GWL; AcMNPV Ac92, 3QZY; APMV R596, 3GWN. For comparison, the structure of the mitochondrial enzyme ALR is also shown.

6.4.3. QSOX

The quiescin sulfhydryl oxidase, or QSOX, enzyme family is a cousin of the Erv/ALR enzymes. QSOX presumably arose from the canonical Erv/ALR dimer by gene duplication, fusion, and degradation of one of the copies of the fold. As a result of such a history, QSOX contains a single-chain pseudo-dimer with a catalytically active Erv-like sulfhydryl oxidase domain, supported by packing against a vestigial Erv domain lacking a catalytic center and flavin cofactor.²⁴⁴ This module was further fused to a segment resembling either the first (in plants and protists) or the first two (in animals) domains of PDI. QSOX has two redox-active CXXC motifs, one in the PDI-like module and the other adjacent to the FAD in the ERV/ALR module.^245,246

The QSOX PDI-like module assumes the function that PDI plays in multi-protein disulfide relays in the ER, such that QSOX inserts disulfide bonds directly into reduced proteins without the mediation of PDI itself.²⁴⁷ This arrangement resembles the fusion of a trx domain to the quinone-binding domain in VKOR enzymes, except that the interaction between the two redox-active CXXC motifs of QSOX is direct (Figure 21), rather than bridged by a shuttle disulfide (electron-transfer loop) as in VKOR (Figure 15). Although QSOX contains a third, conserved CXXC motif near its carboxy terminus, this motif does not appear to mediate internal electron-transfer events in the main catalytic cycle,^246,248 and the reason for its conservation is unknown. Mechanistic and structural studies support a minimal model for QSOX catalysis as depicted in Figure 22.^{198,248–252}

Figure 21.

Schematic representation of conformational flexibility and internal electron transfer events in QSOX. The FAD cofactor is represented by hexagons, with blue representing a charge-transfer state. The rightmost (i.e., most carboxy-terminal) CXXC disulfide does not appear to participate in electron transfer from substrate to the FAD.

Figure 22.

QSOX catalytic cycle. This model depicts turnover involving two-electron reduced forms of the enzyme for simplicity. Additional pathways involving four-electron reduced enzyme forms may contribute at high concentration of reducing substrates and/or low oxygen tensions. Curved lines represent motion. Step 1 is reduction of QSOX by substrate protein. Step 2 is the rapid formation of an interdomain disulfide. Step 3 is a rate limiting step involving reduction of the flavin cofactor. Step 4 regenerates the oxidized enzyme and occurs from either the closed or the open (shown here) state.

Since the QSOX PDI-like module accepts electrons from substrate and then transfers them directly to the ERV/ALR active site (Figure 21), large-scale conformational changes appear to be required. A structure of trypanosome QSOX was determined with the redox-active sites far apart from one another and the PDI-domain CXXC exposed at the protein surface, appropriate for substrate interaction.²⁴⁸ In addition, QSOX variants containing an inter-module disulfide were readily prepared, showing how a flexible linker between the modules enables a structural transition to juxtapose and bury the redox-active sites for interdomain electron transfer.²⁴⁸ Two observations indicated that the closed conformational state of QSOX is not an artifact of cysteine mutagenesis. First, the wild-type rat QSOX1 enzyme containing all native cysteines was crystallized in a closed conformation in which the four redoxactive cysteine residues were aligned for in-line attack (Figure 23).²⁵² Second, a single-molecule fluorescence resonance energy transfer (FRET) experiment on trypanosome QSOX showed that a fraction (about 5% under the conditions of the experiment) of the enzyme population assumes a closed conformation in the absence of substrate.²⁵⁰

Figure 23.

Structure of rat QSOX with juxtaposed redox-active disulfides. The protein is oriented to correspond roughly to Figure 20. The flexible linker is behind the domains. Disulfides are shown with yellow sulfur atoms. The three CXXC motifs and the FAD cofactor are labeled. Unpaired cysteines are not shown. Disulfides not in CXXC motifs are displayed but not labeled. The rightmost CXXC motif is distant from the redox-active centers. PDB code is 4P2L.

Notably, the amount of closed conformer increased substantially when substrate was added and FRET was monitored under steady-state turnover conditions. A population of charge-transfer intermediate was also seen in stopped-flow experiments monitoring FAD absorbance upon substrate addition.²⁴⁹ Mechanistic modeling of the single-molecule FRET results suggested that the charge-transfer species accumulating under steady-state turnover conditions does not contain the mixed disulfide. Instead, the charge-transfer species has a reduced pair of FAD-proximal cysteines, a disulfide in the PDI domain active site, and is in the closed conformation.²⁵⁰

An appraisal of substrate scope in vitro showed that QSOX is a facile oxidant of thiols located in areas of conformational flexibility in proteins irrespective of their amino acid sequence, isoelectric point, or overall protein size.^{130,247,253,254} In contrast, well-folded proteins with pairs of surface thiols are poor to negligible substrates of the enzyme.^253,254 The initial reduction of QSOX (step 1; Figure 22) by a model substrate, reduced RNase, is second-order with no evidence of kinetic saturation.^198,254 In addition, no obvious binding site for thiol substrates is evident from the crystal structures of QSOX family members.^248,252 Together, these experiments suggest an enzyme whose targets are likely governed by substrate availability and conformational flexibility. Interestingly, QSOX has evolved to serve in physiological environments different from those in which homologs of its component domains operate. PDI family proteins function in the ER, and Erv/ALR enzymes are known for their important role in the mitochondrial intermembrane space. QSOX, on the other hand, is localized to the Golgi apparatus and secreted from cells.^255,256 Physiological substrates in these contexts have yet to be discovered, but unlike proteins entering the ER or mitochondria, proteins in the late secretory pathway and extracellular environment are expected to be largely folded. Therefore, any required flexibility in natural QSOX substrates is likely to comprise only part of the protein. Alternatively, the preference for flexible model substrates may not reflect the capabilities of QSOX on its natural substrates. In any case, based on its localization, QSOX may be involved in protein assembly or re-organization rather than protein folding per se.

6.4.4. Ero1

Ero1 is a sulfhydryl oxidase found in the ER of animals, plants, fungi, and protists.^257–259 Ero1 enzymes consist of a single, large, helical domain decorated with disulfide-bonded loops.^77,260 Ero1 shows some underlying similarities with the Erv/ALR family, despite differences in size and quaternary structure (Figure 17). For example, the FAD cofactor is encased in a helical fold in both families and assumes a similar U-shaped structure. Like many Erv/ALR enzymes, Ero1 has a shuttle disulfide that delivers electrons to the FAD-proximal disulfide. However, some striking differences are also observed between Ero1 and Erv/ALR modules. One is that the FAD-proximal disulfide of Ero1 enzymes, as seen in crystal structures, is not exposed for in-line attack by either the shuttle motif or an exogenous cysteine (Figure 18). This situation, seen in yeast and mammalian Ero1 structures, ^77,260 differs from that of PDI as discussed above (section 6.1.1), as well as from Erv/ALR enzymes. In Ero1, the FAD is buried more deeply within the fold than in Erv/ALR, and surface loops and pieces of secondary structure obscure access to the FAD-proximal disulfide. Though structures of yeast Ero1 captured with different conformations of the loop containing the shuttle disulfide suggest mobility in this region (Figure 24),⁷⁷ neither of the observed conformations achieves the alignment of the relevant cysteine residues for an SN2 reaction to generate a mixed disulfide as seen for QSOX.²⁵² More extensive rearrangement of the Ero1 structure appears to be necessary to allow formation of an electron relay between shuttle cysteines and the FAD-proximal disulfide, but how exactly this rearrangement occurs and the nature of the resulting structure remain to be determined.

Figure 24.

Structure of yeast Ero1 displaying different conformations of the loop containing the shuttle disulfide. Only the flexible loop containing the shuttle disulfide is shown for the “out” conformation (magenta). The FAD is orange. The FAD-proximal cysteine is obscured by the S1 sulfur in this view, such that the thiolate nucleophile would be expected to approach S1 from the direction of the reader. PDB codes are 2RP4 and 1RQ1.

The shielding of the active-site disulfide in Ero1 suggests the potential for regulation of Ero1 by controlled conformational changes. Indeed, reduction of regulatory disulfides in Ero1 and the presumed conformational changes that ensue have been shown to increase the catalytic activity of the enzyme.^73,260–264 There is some distance between regulatory disulfides and the catalytic center, so changes in structure or flexibility upon reduction may propagate toward the active site. Reduced PDI, and in particular over-reduced PDI bearing a reduced CX₆C motif near the main, CXXC disulfide,²⁶⁵ are potent activators of yeast Ero1.²⁶⁶ Ero1 appears to mediate is own de-activation by re-oxidizing the regulatory disulfides.^261,266 No other sulfhydryl oxidase has been found to have such a complex and highly tuned regulatory mechanism.

6.4.5. Thioredoxin Reductase-like Enzymes

A number of secreted fungal flavin-linked sulfhydryl oxidases^267–270 are efficient catalysts of the oxidation of glutathione:^267,268

$$2\text{GSH}+{\mathrm{O}}_2->\text{GSSG}+{\mathrm{H}}_2{\mathrm{O}}_2$$ (9)

Cysteine, 2-mercaptoethanol, and dithiothreitol (DTT) are significantly poorer substrates of these enzymes.^267–269 In the cases examined, these fungal sulfhydryl oxidases have weak to negligible activity towards unfolded RNase thiols.^267,269,271 The physiological roles of these abundant secreted enzymes are currently unknown.

Sequence analyses revealed that these microbial glutathione oxidases are members of the pyridine nucleotide disulfide oxidoreductase superfamily, which includes glutathione and thioredoxin reductases.^{268,272–274} Thus the microbial sulfhydryl oxidases have appropriated the reductase framework, retaining the flavin and proximal disulfide, while eliminating a functional pyridine nucleotide binding site to allow them to use molecular oxygen as the terminal electron acceptor for substrate oxidation. For illustration, the flow of reducing equivalents in glutathione reductase and oxidase are compared here (equation 10a and 10b, respectively):

$$\text{NADPH}->\text{FAD}->\text{-S-S-}(\text{proximal})->\text{GS-SG}$$ (10a)

$$2\text{GSH}->\text{-S-S-}(\text{proximal})->\text{FAD}->{\mathrm{O}}_2$$ (10b)

Phylogenetic analyses identified additional clades of specialized bacterial and fungal flavin-linked sulfhydryl oxidases that generate disulfide bonds in a number of virulence factors, antibiotics, and anticancer compounds (Figure 25).^275–277 For example, the enzyme GliT catalyzes the last step in the biosynthesis of gliotoxin in the opportunistic human pathogen Aspergillus fumigatus by generating a disulfide bridge that is essential for toxicity.^276–278 Holomycin is one of a series of dithiolopyrrolone broad-spectrum antibiotics produced by Actinomycetes and proteobacteria.²⁷⁹ In Streptomyces clavuligerus, the intramolecular disulfide in this antibiotic is generated by the dedicated oxidase, HlmI.²⁷⁵ A third specialized disulfide-generating enzyme, DepH, from Chromobacterium violaceum, introduces the unusual transannular cross-link in FK228 (Romidepsin) (Figure 25). While DepH was originally thought to be an NADP⁺ dependent oxidoreductase,²⁸⁰ a more recent study showed that the enzyme does not contain a functional pyridine nucleotide binding site.²⁷⁷ Hence the catalytic mechanism of DepH is likely to be comparable to the other members of the thioredoxin reductase-fold sulfhydryl oxidases described above.

Figure 25.

Compounds with disulfide bonds introduced by fungal and bacterial sulfhydryl oxidases related to the thioredoxin reductase enzyme family. Gliotoxin is a epipolythiodioxopiperazine virulence factor in Aspergillus. Holomycin is a broad-spectrum antibiotic from Streptomyces. Romidepsin (FK228) is a depsipeptide anticancer agent from Chromobacterium species used against T-cell lymphomas.

7. OXIDATIVE FOLDING IN BIOLOGICAL SYSTEMS

Now that the activities and mechanisms of the various enzyme families contributing to oxidative folding in biological systems have been presented, the interplay between these enzymes and their integration into their cell biological context can be discussed.

7.1. General Principles

Classical questions that have been considered regarding the biosynthesis of disulfide bonded proteins include 1) when in the process of translation/folding does a given disulfide get introduced, 2) does the vectorial nature of protein biosynthesis affect the order and kinetics of cysteine pairing, 3) what is the enzyme that introduces the disulfide, and 4) what role does non-native cysteine pairing play in oxidative protein folding and how is it resolved? In relation to the first two questions, the order of folding vs. disulfide bond formation has been studied for model proteins.^281,282 A particularly creative approach was the use of single-molecule force measurements to distinguish the timing of protein folding from disulfide bond formation.²⁸³ These studies suggested that interactions of nascent proteins with PDI may delay disulfide formation until proper folding can occur, which is likely to promote native disulfide pairings. Nevertheless, due to the large number of substrates folding oxidatively and the potential parallelism of the contributing catalysts in vivo, obtaining general answers to questions regarding the order and mechanism of formation of native disulfide bonds is challenging. Apportioning the relative importance of each possible route to disulfide output in locales where multiple pathways are operating simultaneously may not be practical. Instead, it is valuable to outline general principles.

It is likely that the timing of folding vs. disulfide bonding and the requirement for isomerization following oxidation will be determined by the properties of the substrate protein to a greater extent than by the capabilities of the enzymes involved. In particular, if a protein folds sufficiently to bring correct cysteine pairs into proximity for preferred disulfide bond formation upon encounter with enzyme, or contains few competing cysteines, then pairing will likely be accurate. If, in contrast, the substrate protein arrives as a largely unfolded polymer with many cysteines available to resolve the mixed disulfide with an oxidizing enzyme, then the products will be a heterogeneous mixture of disulfide isomers. As the substrate set differs among the various organelles that promote cysteine oxidation, the balance and importance of oxidants vs. isomerases is also expected to be organelle-specific. Particularly interesting issues that are also organelle-specific are how electron flow into environments that support net protein oxidation is utilized and controlled and how electrons liberated by oxidation of folding proteins are fed into other cell metabolic pathways. In the subsequent sections we will therefore discuss key features of oxidative folding processes grouped by cellular locale.

7.2. Bacterial Periplasm

One conundrum that arose in the context of catalysts of disulfide formation in the bacterial periplasm is how reductive and oxidative pathways can co-exist in the same compartment without “short circuiting.” Disulfide bonds are produced in the E. coli periplasm by the pathway consisting of the soluble protein DsbA, the membrane-embedded DsbB, and quinones as electron acceptors. The periplasm is supplied with reducing equivalents by the soluble protein DsbC and the membrane-embedded DsbD, driven by thioredoxin, thioredoxin reductase, and NADPH in the cytosol. The activities of DsbA and DsbC, which, respectively, generate and isomerize disulfide bonds, would be undermined if DsbA could efficiently oxidize DsbC or DsbD, or if DsbC could reduce DsbA or DsbB. The solution to this potential short circuiting problem is kinetic isolation of the oxidative and reductive pathways. This isolation is accomplished in part by a simple stratagem: DsbB is prevented from oxidizing DsbC by steric seclusion of DsbC active sites through dimer formation.^284,285 In fact, with the solution of the DsbB structure, it was realized that DsbC fails to engage DsbB due to a steric clash of one subunit of the DsbC dimer with the membrane were the second subunit properly positioned to transfer electrons to DsbB.¹⁹³ DsbD, in turn, is prevented from reducing DsbA efficiently, most likely due to more minor steric incompatibility between regions surrounding the redox-active sites of DsbA and DsbD.²⁸⁶ Genetics studies investigating the phenotypes of null mutants of various Dsb pathway members, individually or in combination, led the investigators to argue that this kinetic isolation model is an over-simplification of actual events in the bacterial periplasm.²⁸⁷ Although the kinetic isolation may not be complete, or Dsb components may partially function off their canonical pathways, the elegant principles revealed by the initial studies of the E. coli Dsb system are nevertheless illustrative and likely also to be substantially correct.

Another principle that can be showcased with the Dsb system is how to use electrons derived from cysteine oxidation in a metabolically efficient manner. Whereas electrons accepted by sulfhydryl oxidases can be transferred, via bound cofactors, to molecular oxygen, this direct route appears to waste reducing potential. In contrast, reduction of mobile quinones by DsbB allows electrons to enter the respiratory chain where they can be additionally leveraged toward the generation of an electrochemical gradient and ATP synthesis before ultimately being delivered to oxygen.²⁸⁸

7.3 Mitochondrial Intermembrane Space

With the characterization of the disulfide bond formation pathway in the mitochondrial IMS, conceptual parallels were discovered with the Dsb system in the bacterial periplasm. Electrons received by DsbB are fed into the bacterial electron transport chain through lipid-soluble quinones, whereas electrons received by ALR enzymes in mammalian mitochondria are fed into the mitochondrial electron transport chain through cytochrome c.^233–235 In both cases, these pathways may minimize the accumulation of reactive oxygen species within the mitochondrial intermembrane space as well as using electrons productively.²³³

There are some significant differences, however, between reduction of quinones by DsbB and reduction of cytochrome c by ALR. As opposed to the two-electron transfer from DsbB to quinones, ALR performs successive one-electron transfer reactions to two molecules of cytochrome c per disulfide formed. Interestingly, clues to single electron transfer by ALR were detected in studies of the purified enzyme: the two electrons received by ALR during formation of one substrate disulfide were seen to produce large amounts of neutral semiquinone, suggesting comproportionation between the two FAD molecules bound by the ALR dimer.²³³ The isoalloxazine moieties are separated by about 17 A edge-to-edge, which is slightly longer than that expected for unbridged tunneling distances in proteins²⁸⁹ but is still expected to lead to electron transfer rates consistent with the relatively low overall turnover number of the enzyme.²³³ Though a detailed analysis of the features of Erv/ALR enzymes that support formation of the semiquinone and promote reduction of cytochrome c vs. molecular oxygen has not been done, the preference for cytochrome c is apparently not universal for the mitochondrial enzymes, since Arabidopsis thaliana Erv1 does not resemble mammalian ALR in this regard.²⁹⁰

In addition to electron acceptors, other major issues have arisen in the context of the mitochondrial disulfide formation pathway: protein substrate specificity, the role of glutathione, and the relationship between cytosolic and IMS redox potentials. In eukaryotes all proteins residing in the mitochondrial intermembrane space (IMS) are encoded in the nucleus, produced in the cytosol, and enter mitochondria in a reduced form through a translocase of the outer membrane (TOM) complex. Upon entry into the IMS, these proteins encounter Mia40 and may become oxidized.^291,292 Initially, Mia40 was thought to recognize substrates based on a 9-residue amphipathic helix containing a cysteine that engages in mixed disulfide formation with the redox active CPC motif in Mia40.²⁹³ However, in vitro and in vivo experiments showed that a number of kinetically-competent reducing substrates lack these sequence features (e.g., ATP23, Ccs1, DRE2, Erv1/ALR and SOD1).^294–298 Indeed, comprehensive studies using mutated IMS protein sequences,^171,299,300 and arrays of ARP23 peptides²⁹⁶ showed that Mia40 has a much broader specificity than originally envisaged. This conclusion is reinforced by the finding that the Mia40/ALR system is an efficient oxidant of three reduced proteins (RNase, lysozyme, and riboflavin binding protein) that lack the amphipathic helix motif mentioned earlier and are naturally oxidatively folded within the ER.³⁰¹ The catalytic promiscuity of the Mia40/ALR system suggests that proteins entering the IMS en route to their final mitochondrial destinations may be protected from extraneous disulfide bond formation by sequestration via translocases and chaperones.^302–305 Alternatively, misincorporated disulfides might be removed by IMS-resident reductases.^{170,294,306,307}

Although the Mia40/ERV1 pathway is widely distributed in eukaryotes, a number of protists, including the human pathogens Toxoplasma gondii, Plasmodium falciparum, and the Trypanosomatids lack a detectable Mia40.^237,308,309 There appears to be no compensating ability of the corresponding cognate protist Erv1 enzymes to oxidize reduced client proteins directly. Currently, the precise mode by which disulfides are introduced in the IMS of these organisms, and whether there is a factor that compensates for the missing Mia40, remains unclear.

The comparative lack of specificity in the initial oxidation step of IMS protein substrates, together with the significant complexity of disulfide connectivity in some of the products of the Mia40/ALR pathway, suggest the existence of an enzyme with isomerase activity to address mispairings that occur during oxidative protein folding in the IMS.^{296,297,299,310} It should be noted that Mia40 itself is an unlikely candidate for this isomerase. Mia40 showed weak to insignificant isomerase activity towards the model substrate RNase and very poor activity towards the IMS resident protein Cox17.^142,299 Although reduced glutathione can catalyze the resolution of mispaired disulfides in Cox19,¹⁷⁰ the ability of glutathione to correct mispairings in substrates with more complicated disulfide patterns in the IMS remains to be established. In any case, glutathione is certainly available to the IMS, as it readily traverses the mitochondrial outer membrane.³⁰⁶

7.4. Disulfide Bond Formation in Cytoplasmic Assembly of dsDNA Viruses

Among the most intriguing but poorly understood pathways involving sulfhydryl oxidases are those encoded by viruses. The genomes of large, dsDNA viruses, such as poxviruses (e.g., vaccinia virus) and mimivirus, contain hundreds or even thousands of open reading frames. These viruses assemble in viral factories established in the cytosol of infected cells, but they rely on processes considered to be characteristic of specialized, membrane-bounded compartments in eukaryotic cells. As an adaptation to their site of replication, the viruses encode their own versions of enzymes that are compartmentalized in cells. For example, mimivirus encodes a set of enzymes involved in glycan formation,^311–313 allowing for the biosynthesis of surface glycoproteins independently of the host secretory pathway. In addition, as noted above (section 6.4.2), large, dsDNA viruses encode sulfhydryl oxidases. These enzymes are presumably necessary because the cytosolic virus assembly process cannot exploit the activities of cellular sulfhydryl oxidases encapsulated in the ER, mitochondria, or Golgi apparatus. Furthermore, the viral sulfhydryl oxidases can evolve independently of their cellular counterparts and thus become well-suited to the specific needs of assembling virions. The substantial architectural diversity of the viral enzymes (Figure 17) supports the notion that they have evolved features unique to each virus and its structural and biochemical requirements for disulfide bond formation.

The vaccinia virus sulfhydryl oxidase (E10R) functions on a pathway together with a glutaredoxin (G4L) and an additional small protein not apparently belonging to any class of known redox-active proteins (A2.5L).^238,242,314 E10R and its partners are essential for virion morphogenesis,^315,316 are associated with virion membranes,³¹⁵ and have been detected by mass spectrometry proteomics in purified virions.^317–319 The targets of this pathway appear to be viral membrane-embedded proteins involved in viral maturation and membrane fusion during infection.^238,320 These proteins are topologically oriented such that their disulfide bonded ectodomains face the cytosol during virion assembly.²³⁸ Many glutathionylated proteins were found in a proteomics study of vaccinia virions,³²¹ but the significance of this finding is not yet clear.

A similar pathway as that described for the vaccinia sulfhydryl oxidase may function in African swine fever virus (ASFV), sole member of a large dsDNA virus family distinct from poxviruses.³²² In ASFV, the sulfhydryl oxidase was found to associate with another viral protein of unknown structure,³²² perhaps paralleling the function of A2.5L in poxviruses. Also analogous to poxviruses, viral membrane proteins are the targets of disulfide bond formation in ASFV.³²²

Sulfhydryl oxidases have been identified in other viruses,^323,324 but they have not been associated with particular pathways or viral targets. Mimiviruses encode two Erv/ALR family sulfhydryl oxidases and a number of proteins with thioredoxin-fold domains, as well as a probable glutaredoxin. A better understanding of the assembly and maturation of the virus, as well as the mechanisms by which it reprograms the cell cytosol, will require elucidating the roles of these enzymes.

In addition to the recognizable fold families corresponding to sulfhydryl oxidases and thioredoxin-like oxidoreductases, additional, novel proteins with roles in disulfide formation may be present in viruses. A protein with unknown structure but containing CXXC motifs may contribute to disulfide bond formation during herpesvirus assembly and maturation.³²⁵

7.5. Disulfide Bond Formation in the Endoplasmic Reticulum

The numerous pathways involved in generating disulfide bonds in the ER of higher eukaryotes distinguishes oxidative protein folding from other forms of posttranslational modification. The activity of PDI as a catalyst of oxidative protein folding in the ER has been recognized since the 1960s,³²⁶ but recent years have brought increasing appreciation for the complexity and multiplicity of players contributing to the redox environment of the early secretory pathway. Four enzyme classes that contribute disulfides to the ER folding environment have been identified: Ero1, VKOR, peroxiredoxin 4, and the PDI peroxidases gpx7 and gpx8. These enzymes partner with numerous PDI family enzymes, which can undergo thiol/disulfide exchange reactions among themselves in addition to oxidizing and isomerizing client proteins.^327,328 Together, the whole system is exposed to varying levels of glutathione, hydrogen peroxide, DHA, and perhaps other small-molecule redox-active factors. In addition, changes in ER luminal calcium concentration may affect the localization, partnerships, and activities of redox-active ER enymes.³²⁹

Though specific interactions and electron-transfer events can be studied in vitro and identified in vivo by, for example, trapping and identifying inter-protein mixed disulfides, oxidative protein folding in the ER may deviate from the model of a strict assembly line. A distinct perspective can be obtained by considering the flux of reducing equivalents within the ER (Figure 26). In terms of inputs, electrons flow into the ER principally with cysteines on nascent polypeptide chains and with reduced glutathione transported from the cytosol.³³⁰ Reduced protein clients are then co- or post-translationally oxidized, with the majority of the reducing equivalents passing through PDI-family proteins.³³¹ PDIs in turn reduce enzymes that catalyze the transfer of electrons from thiol groups to molecular oxygen or hydrogen peroxide, producing, respectively, either hydrogen peroxide or water. The rate of flux of electrons through this system is one crucial parameter that can be controlled, presumably by modulating the availability and activities of enzymes that oxidize PDIs. The levels and locales of hydrogen peroxide produced may be another important parameter. These levels may be controlled by the balance of, and possibly interactions between, Ero1 and PDI peroxidases,^176,182 or by as yet undetermined sources.³³²

Figure 26.

Schematic flow of reducing equivalents in the ER during oxidative protein folding. Electron input: reducing equivalents arrive from the cytosol as cysteine residues in translocating nascent chains and imported glutathione. Output: electrons ultimately reduce molecular oxygen. The major reductive flux in the ER pipeline is believed to involve multiple PDIs in redox communication with one another and with the glutathione redox system, and capable of transferring electrons to sulfhydryl oxidases and peroxidases, which in turn reduce terminal electron acceptors.

Initial insight into how the flux of reducing equivalents in the ER is controlled came from the finding that Ero1 has a mechanism to prevent over-oxidation of its surroundings,^260–263 a phenomenon not observed for other sulfhydryl oxidases. This mechanism is based on the PDI-mediated reduction of regulatory disulfide bonds for Ero1 activation and auto-oxidation for Ero1 inactivation.^73,261 More recently, evidence has been put forth that Ero1 participates in complexes that further control the input and output of electrons.^264,333 It has been observed that the pool of PDI family proteins in a variety of mammalian cell types is maintained largely in reduced form.^{142,145,330,334–337} For comparison, oxidative folding in vitro is most efficient under relatively reducing conditions.³³⁸ In one case study, the rate of oxidative refolding of riboflavin binding protein, with nine disulfides and >34 million possible isomers, progressively increased as the fraction of oxidized PDI decreased, and the most rapid folding occurred with an amount of oxidized PDI just sufficient to account stoichiometrically for the insertion of the disulfides.²⁴⁷ If in vitro conditions adequately represent those in the ER, it appears that the high intraluminal ratio of PDI_red/PDI_ox is tuned for efficient oxidative folding in vivo. Somewhat paradoxically, this large reduced PDI pool should also be efficient at activating Ero1, and thereby at promoting its own depletion. The observed steady state, with substantial reducing power available to nascent protein chains, may be achieved with aid from the cytosolic thioredoxin system.³³⁹ The concentrations of relevant ER enzymes and substrates and the reaction rates between them are also clearly central to this issue, and additional factors not yet discovered or considered may further impact ER redox potential.

Regarding measurements of PDI redox state in vivo, it is important to note that the glutathione redox system is in rapid equilibrium with the isomerases.

$$2\text{GSH}+\text{PDI}{(\text{CxxC})}_{\text{ox}}\iff \text{GSSG}+\text{PDI}{(\text{CxxC})}_{\text{red}}$$ (11)

In vitro studies following the reduction of both a and a′ domains of human PDI and those of human ERp57 and yeast PDI1p show that this equilibration is half complete in less than half a second using a physiologically realistic concentration of 5 mM GSH.^155,340,341 The similarities among the active sites of PDI proteins and the utility of glutathione redox buffers for measurements of isomerase redox potentials¹⁵⁵ suggests that facile redox communication with glutathione is likely to be a common feature among these oxidoreductases. Given this kinetic facility, stringent trapping procedures are needed to accurately assess the redox states of PDI within cells.¹⁴² As a flip side to this complication, equilibration of PDI proteins with glutathione has the benefit of allowing a comparison of PDI_red/PDI_ox ratios with the redox poise of the ER measured using roGFP probes, which also communicate with luminal glutathione. Independent investigations using such probes gave redox potentials for fibroblasts, HeLa, and pancreatic acinar cells lines of −231, −208 and −236 mV, respectively.^333,342,343 Applying the Nerst equation using an average intra-luminal redox potential of −225 mV predicts that the PDIs in Table 1 would be almost completely reduced if equilibrated with the ER glutathione pool. Hence the experimental observations of reduced PDI proteins in the ER and the calculations based on an equilibrium assumption are in substantial agreement. Furthermore, shifts in the balance of reduced and oxidized PDI proteins in perturbed cells are accompanied by parallel shifts in reduced and oxidized glutathione levels.³⁴⁴ It should be noted, however, that not all redox-active proteins in the ER need be equilibrated with the glutathione redox buffer. Protein sublocalization may allow specific pathways for oxidation to occur against the backdrop of the reducing environment. For example, the membrane-bound PDI family member Tmx1 appears to preferentially target substrates also anchored to the ER membrane.³⁴⁵ Reminiscent of the Dsb system in bacteria, various points of steric compatibility and incompatibility are likely to open and close routes by which reducing equivalents flow from nascent protein chains and glutathione, via PDI proteins, to sulfhydryl oxidases and PDI peroxidases.

Table 1.

Redox potentials and expected reduced/oxidized ratios of PDI proteins in the ER.

Protein	Redox Potential	Red/Ox at −225 mV (assuming equilibrium)
PDI	−160 to −176 mV	~ 90/1
ERp57	−155 to −167 mV	~ 130/1
ERp72	−155 mV	~ 195/1
P5	−146 mV	~ 360/1
ERp46	−161 mV	~ 120/1
TMX3	−157 mV	~ 160/1

To balance the emphasis on sulfhydryl oxidases, other potential oxidants should also be considered for their potential effects on ER redox balance. DHA (section 5.2.3) serves as an oxidant of the reduced catalytic centers of PDI^110,346, leading to the suggestion that this reaction would support oxidative folding in the ER of higher eukaryotes.³⁴⁷ A study of the kinetics of oxidation of reduced PDI and GSH by DHA yielded second-order rate constants of 12.5 ± 1.9 and 0.39 ± 0.01 M⁻¹s⁻¹ respectively at pH of 7.3.¹¹² In contrast, DHA was observed to oxidize reduced bovine trypsin pancreatic inhibitor (BPTI) and RNase A with maximal second-order rate constants of 163 ± 33 and 104 ± 9 M⁻¹s⁻¹ respectively.¹¹² DHA is thus a more efficient oxidant of unfolded reduced proteins compared to PDI; indeed, its utility was recognized decades ago by Anfinsen and Straub during their seminal work on the refolding of reduced RNase.^348,349 Though DHA could have pleiotropic effects within the ER by reacting with unfolded reduced proteins, the extent to which DHA contributes to the net oxidizing flux in the mammalian ER is not currently known.^350,351 An ascorbate oxidizing activity has been described on the cytosolic side of rat liver ER membranes, together with a transport system enabling dehydroascorbate to access the lumen, suggesting that a contribution is in principle possible.^351,352

7.6. Quality Control at the ER/Golgi Boundary

The error-prone process of oxidation, folding, and assembly of large, complex proteins suggests the need for quality control. Not covered in this review are the multiple branches of the ER stress response that bolster ER capabilities when excessive folding failures are perceived.³⁵³ Instead, we will focus on the PDI family protein ERp44, which cycles between the ER and Golgi apparatus retrieving proteins that can be engaged through exposed cysteines and that may therefore lack their full complement of disulfides. ERp44 gives wayward subunits of protein complexes repeated chances to meet their partners and is also responsible for ER localization of some proteins lacking classical retention signals.

In contrast to many other PDI proteins, which have a redox-active di-cysteine motif as an active site, ERp44 has only a single active-site cysteine. This cysteine can be sterically shielded by the carboxy-terminal tail of ERp44, which contains a sequence recognized by the KDEL receptor responsible for retrograde transport of proteins back to the ER. Both the redox-active cysteine and the carboxy-terminal tail become exposed in a pH-dependent manner. ERp44 thereby takes advantage of the pH gradient along the secretory pathway to toggle its activity between 1) a state that forms disulfide bonds with substrate and is retrotranslocated and 2) a state that is released from the KDEL receptor, liberated from its cargo, and shielded from redox reactions.

ERp44 was first identified as a factor associated with human Ero1α^354,355 and then shown to modulate the assembly of adiponectin complexes and IgM.^356,357 ERp44 retains certain other ER redox-active proteins, such as Ero1 and peroxiredoxin IV, in the ER as well.³⁵⁸ Structural studies have provided information both on the conformational changes that underlie the toggling of ERp44 activity,^359,360 as well as on the mechanism of engagement of peroxiredoxin IV.¹⁴⁶ Some remaining questions include the mechanism for disulfide formation between ERp44 and its clients undergoing oligomeric assembly. Does ERp44 displace glutathione from protein mixed disulfides? Or is the disulfide between ERp44 and its client formed from two free thiols by the action of a sulfhydryl oxidase or other enzyme?

7.7. Disulfide Bond Formation in the Late Secretory Pathway

As described in the previous sections, the ER is the standard site for introduction of disulfide bonds into proteins in the secretory pathway, and retrieval and quality control mechanisms acting at the ER/Golgi interface reinforce the importance of the ER in this regard. Nevertheless, a unique family of sulfhydryl oxidase is localized beyond this border. The QSOX enzymes are found in the Golgi apparatus or secreted from cells, suggesting either that they function on a class of substrates distinct from the clientele for ER oxidation or, perhaps reflecting a need to reoxidize disulfides originally formed by the canonical system in the ER.

A great deal is known biochemically and structurally about QSOX, but less is known about its physiological function. The enzyme is present in a variety of body fluids^{102,219,361–364} and to varying degrees in blood serum, depending on the organism,³⁶⁵ its developmental state,³⁶⁶ and its disease state.³⁶⁷ The QSOX enzymes are localized to the Golgi apparatus in many mammalian cell types.²⁵⁶ Specifically in confluent fibroblasts, QSOX expression is upregulated, and the enzyme is secreted from the cells.²⁵⁵ In cell culture, the activity of secreted QSOX is essential for assembly of a normal and functional extracellular matrix,²⁵⁶ though the precise targets of QSOX activity are not yet known.

One question that arises in considering a late-secretory pathway sulfhydryl oxidase is to what extent the enzyme contributes to oxidation of thiols in the ER in transit to the Golgi and extracellular environment.³⁶⁸ One possible mechanism to dampen QSOX activity in the ER would be pH-dependent activity optimized for the lower pH of the Golgi. However, purified QSOX exhibits a pH optimum for activity fully consistent with the ER environment.^246,249 A study of the QSOX conformational ensemble during turnover suggests another speculative but simple mechanism for controlling enzyme activity. A fraction (approximately 7%) of oxidized QSOX was observed to be in a conformation that buries the two redox-active sites against one another and secludes them from the surrounding solution.²⁵⁰ This observation indicates that the equilibrium between a “closed” and inactive state and an “open” and active state of QSOX only slightly favors the latter. Any hypothetical interaction made between an ER protein and QSOX as it transits that compartment, as long as the interaction is specific for the closed form, would be sufficient to tip the balance toward catalytic inactivity. Identifying such an interaction is likely to be challenging, as it would be expected to be weak and transient. Direct identification of the conformational state of QSOX in the ER is likely to be even more challenging, as undetectable levels of the enzyme accumulate there.²⁵⁶

Though QSOX is the only catalyst of de novo disulfide bond formation known to be efficiently secreted from mammalian cells, other enzymes typically localized to the ER have been reported to exist and function extracellularly under certain circumstances. For example, upon injury, platelets and endothelial cells release PDI.^369,370 This PDI population then activates factors that initiate thrombus formation.^371–373 Due to its role in strokes, PDI thus becomes a target for specific inhibition of its activity extracellularly.³⁷⁴ Additionally, the PDI family protein Agr2 has been found in mucus secreted from cells of the gastrointestinal tract and appears to be important for mucus structure and protective function.^375,376 In certain cases, extracellular localization of ER proteins has been associated with disease, such as the correlation of ECM-associated ERp57 with renal fibrosis.³⁷⁷ The secretion mechanisms by which ER proteins arrive in the extracellular environment are a topic of ongoing investigation.³⁷⁸

8. CONCLUSIONS AND PERSPECTIVES

While the oxidation of vicinal thiols to form a disulfide bond is the most chemically conservative of post-translational modifications, disulfide bonds can profoundly impact the stability, localization, oligomerization, and function of the recipient protein. Other cross-linking chemistries have been identified in proteins in nature, but disulfide bonding is by far the most widespread. The ready reversibility of the modification also contributes to its versatility. Indeed, facile thiol-disulfide exchange reactions underpin the mechanisms of enzymes that generate and rearrange disulfide bonds in client peptides and proteins. Importantly, disulfide bond formation in various intracellular compartments is integrated into cellular metabolism, since electrons derived from disulfide bonding are either delivered productively into other pathways or generate reactive oxygen species that are deactivated using dedicated cellular resources. Historically, the study of disulfide bond formation was a subfield of the investigation of ER function. With time, however, disulfide bond formation was identified and characterized in numerous organelles and physiological contexts. Though based on common underlying chemical principles, the formation of disulfides occurs through fascinating variations on the central themes and contributes to a diversity of wondrous biological structures and phenomena.

Biographies

Colin Thorpe

Colin Thorpe obtained a Chemistry Degree from the University of Cambridge in 1969. His Ph.D. work, at the University of Kent at Canterbury, investigated how enteric bacteria can stereoselectively reduce impermeant cobalt(III) complexes. These studies led to a postdoctoral position in redox enzymology at the University of Michigan with Charles H. Williams and to a sustained interest in the mechanism and physiological role of flavoproteins. Thorpe joined the Department of Chemistry and Biochemistry at the University of Delaware in 1978 where he is now a Willis F. Harrington professor. A chance observation in 1995 led his laboratory to discover a new family of flavin-linked disulfide-generating enzymes. The Thorpe group currently applies a range of chemical, spectroscopic and mechanistic tools to study disulfide bond formation and isomerization in diverse systems and cellular locales.

Deborah Fass

Deborah Fass studied biochemistry at Harvard University and obtained a Ph.D. degree in Structural Biology from the Massachusetts Institute of Technology in 1997. She became interested in disulfide bond formation through the study of viral envelope proteins and contemplation of the process by which they fold oxidatively in the endoplasmic reticulum. In 1998, Fass joined the department of Structural Biology at the Weizmann Institute of Science, where she now holds the Fred and Andrea Fallek professorial chair. The Fass laboratory has determined the structures of a number of sulfhydryl oxidases and contributed to the study of their mechanisms.