Main

The formation of biosynthetic disulphide bonds is an important step in the maturation of the extracellular domains of both membrane and secreted proteins in eukaryotic and prokaryotic cells. Not only are disulphide bridges often vital for the stability of a final protein structure, the incorrect pairing of cysteine residues (hereafter referred to as cysteines or Cys) usually prevents the folding of a protein into its native conformation. THIOL-REDOX REACTIONS are also an essential part of the catalytic activity of several metabolic enzymes. For example, the cytoplasmic enzyme ribonucleotide reductase becomes oxidized during its catalytic cycle, and it must be recycled to its reduced form to be reactivated (reviewed in Ref. 1). Protein activity can also be modulated by altering the redox state of cysteines. In plants, light-generated reducing equivalents are used to reduce the regulatory disulphide bonds in several photosynthetic enzymes, thereby inducing a switch from catabolic to anabolic respiration2,3. Under specific cellular conditions, several transcription factors, including the bacterial OxyR and Hsp33, also become activated by the oxidation of cysteines that form disulphide bonds4,5.

Important progress has been made towards understanding how disulphide bonds are formed in cellular proteins. Core pathways that promote disulphide-bond formation have been delineated in both prokaryotes and eukaryotes. Recently, a new appreciation of how these pathways operate at the mechanistic level has been achieved. It is now apparent that the prokaryotic and eukaryotic pathways of disulphide-bond formation have remarkable similarities, which include their orchestration at the mechanistic level.

The chemistry of disulphide-bond formation is well established. The formation of a disulphide bond from two thiols (-SH) is a two-electron reaction that requires an oxidant or electron acceptor. Disulphide bonds can be formed spontaneously in vitro by the loss of electrons from two cysteine thiols coupled with the gain of electrons by an available acceptor, such as molecular oxygen. When molecular oxygen is used as an electron acceptor, an intermediary, such as a transition metal or flavin, is required to overcome the kinetically sluggish, yet thermodynamically favourable, association of oxygen with protein thiols.

In vivo, however, the most common mechanism for the formation of protein disulphide bonds is a THIOL-DISULPHIDE EXCHANGE REACTION of free thiols with an already disulphide-bonded species. A thiol-disulphide exchange reaction can occur between a protein and any sulfhydryl-containing substrate, including small thiol-containing compounds, such as GLUTATHIONE, or a protein containing a disulphide bond. Thiol-disulphide exchange reactions provide the cornerstone of catalysed protein disulphide-bond formation in all living organisms, from prokaryotes to eukaryotes (for details, see Box 1).

A class of proteins commonly known as THIOL-DISULPHIDE OXIDOREDUCTASES catalyses thiol-disulphide exchange reactions in vivo. The activity of these proteins depends on a pair of cysteines that are often arranged in a Cys-X-X-Cys motif (where X is any amino acid). This motif is usually found embedded in a domain that shares structural homology with the small redox protein THIOREDOXIN. The presence of a Cys-X-X-Cys active-site motif has become the hallmark of proteins that are involved in forming or breaking disulphide bonds by an exchange reaction. PROTEIN DISULPHIDE ISOMERASE (PDI) was one of the first-identified thiol-disulphide oxidoreductases6 and, consequently, its activities have been well characterized (reviewed recently in Refs 7–9). PDI is a remarkably versatile enzyme. Depending on the redox environment and the characteristics of the substrate proteins, PDI can catalyse the formation, reduction or isomerization of disulphide bonds10. Analysis of the REDOX POTENTIAL of PDI indicates that it is a mild oxidant with a redox potential of between −110 and −190 mV (Refs 11,12; Box 2). However, it is not yet known whether the primary in vivo function of PDI is to reshuffle non-native disulphide bonds or to catalyse the formation of disulphide bonds at the outset.

In the cell, proteins that contain disulphide bonds are found primarily in relatively oxidizing environments. In eukaryotic cells, protein disulphide-bond formation proceeds predominantly in the lumen of the endoplasmic reticulum (ER), whereas in prokaryotic cells, most protein oxidation occurs in the periplasmic space. The oxidation of cysteines to form disulphide bonds is catalysed rapidly in both the ER and periplasmic space by several different thiol-disulphide oxidoreductases, including PDI. Thiol-disulphide oxidoreductases, such as thioredoxin, are also present in the relatively reducing environment of the cytoplasm, where they usually catalyse the reduction of protein disulphide bonds. However, the direction of the reaction catalysed by a thiol-disulphide oxidoreductase (reduction or oxidation) does not depend solely on the equilibrium redox potential of the compartment in which the enzyme resides. The redox potential of the thiol-disulphide oxidoreductase (Box 2), its propensity to interact with other redox-active proteins and substrates, and the concentration of the substrate and product proteins all contribute to the nature of the in vivo enzyme activity.

The past few years have seen notable advances in our understanding of the pathways of protein disulphide-bond formation. It is now clear that protein oxidation is a catalysed process that requires many cellular thiol-disulphide oxidoreductases, as well as enzymes that couple the activity of the thiol-disulphide oxidoreductases to the redox chemistry of the cell. This review focuses on the emerging similarities between the prokaryotic and eukaryotic systems that catalyse the formation of structural disulphide bonds, and the general principles of disulphide-bond formation that can be deduced from the genetic, biochemical and structural studies of these systems.

Eukaryotic pathways for protein oxidation

Genetic and biochemical analysis of Saccharomyces cerevisiae has defined an essential pathway for protein disulphide-bond formation that involves two ER proteins: Ero1 (ER oxidoreductin) and PDI (Fig. 1). Ero1 is a glycosylated lumenal ER protein that is tightly associated with the ER membrane13,14. Ero1 is a component essential for the introduction of OXIDIZING EQUIVALENTS into the ER lumen. A conditional ero1-1 mutant fails to provide the oxidizing equivalents that are necessary for protein disulphide-bond formation in the ER, and this results in the accumulation of misfolded proteins in the ER, the folding and transport of which are dependent on disulphide-bond formation13. The oxidizing capacity of the ER can be either increased or decreased by varying the cellular levels of active Ero1 (Refs 13,14).

Figure 1: Pathways for protein oxidation in the endoplasmic reticulum of Saccharomyces cerevisiae.
figure 1

Oxidizing equivalents for disulphide-bond formation are introduced into the endoplasmic reticulum (ER) by two parallel pathways. In the first pathway, oxidizing equivalents flow from Ero1 — or ER oxidoreductin — to the thiol-disulphide oxidoreductase protein disulphide isomerase (PDI), and from PDI to secretory proteins through a series of direct thiol-disulphide exchange reactions (thin arrows). Ero1 derives oxidizing equivalents from a flavin adenine dinucleotide (FAD) cofactor, but the ultimate source of oxidizing equivalents for Ero1 oxidation has not been identified. In the second pathway, the ER oxidase Erv2 transfers disulphide bonds to PDI before substrate oxidation. Erv2 obtains oxidizing equivalents directly from molecular oxygen through its flavin cofactor. In mammalian cells, two isoforms of Ero1 (Ero1-Lα and Ero1-Lβ) are present that transfer oxidizing equivalents to PDI. A mammalian, ER-localized Erv2-like protein has not been identified.

Ero1 influences protein oxidation by transferring oxidizing equivalents directly to PDI, which, in turn, oxidizes the substrate proteins15,16. The transmission of oxidizing equivalents in this pathway occurs through a series of direct thiol-disulphide exchange reactions between the proteins15. Mutational analysis of Ero1 identified four cysteines that are essential for the oxidative activity of this protein17. The positions of these four cysteines indicate that Ero1 contains two active-site cysteine pairs: Cys100-Cys105 and Cys352-Cys355 (Ref. 17). Mutation of any of these four essential cysteines disrupts the thiol-disulphide exchange between Ero1 and PDI, as assayed by the detection of Ero1–PDI MIXED-DISULPHIDES17.

An Ero1-dependent pathway for protein oxidation is also present in the mammalian ER. Two functional human homologues of yeast Ero1 have been identified, Ero1-Lα and Ero1-Lβ (Ero1-Like). Both Ero1-Lα and Ero1-Lα can complement the phenotypic defects associated with the yeast mutant ero1-1 strain18,19, although neither protein can complement the lethality that is associated with a complete disruption of yeast ERO1 (Ref. 20). The lack of complementation of an ERO1 deletion in yeast by human ERO1 has been attributed to the 127-residue carboxy-terminal domain of the yeast Ero1 protein, which is absent in the human proteins20. Ero1-Lα and Ero1-Lβ share a high degree of sequence similarity with each other and with yeast Ero1, but these mammalian proteins differ in their tissue distribution and transcriptional regulation19. Like yeast Ero1, both of the human Ero1 proteins facilitate disulphide-bond formation in substrate proteins, and Ero1–PDI mixed-disulphides have been isolated in mammalian cells21,22.

Recently, a second pathway of disulphide-bond formation in the yeast ER has been identified, which involves a small ER oxidase known as Erv2 (Fig. 1). ERV2 was identified as a gene that, when overexpressed, could restore viability to the mutant ero1-1 strain23. ERV2 encodes a 22-kDa membrane-associated ER protein with a lumenal domain that is non-covalently bound to flavin adenine dinucleotide (FAD)23,24. Using purified recombinant Erv2, it has been shown that this protein can catalyse the formation of disulphide bonds using molecular oxygen as an electron acceptor23. The in vivo activity of Erv2 is dependent on a pair of cysteines (Cys121 and Cys124) that are found in a Cys-X-X-Cys motif in a region of high sequence conservation among the Erv2 homologues, as well as on a second pair of cysteines (Cys176 and Cys178) that are present in a Cys-X-Cys arrangement in the carboxy-terminal portion of the protein23,25. Erv2 seems to drive the oxidation of substrate proteins in vivo as part of a cascade of disulphide-bond formation that involves PDI, as assayed by the capture of a mixed-disulphide intermediate of Erv2 and PDI23. However, in vitro, recombinant Erv2 is also able to oxidize substrate proteins directly24; the importance of this observation for Erv2 function in vivo is not clear. Erv2 is a member of a large family of thiol oxidases that are distributed widely among eukaryotic organisms and viruses (Box 3; Table 1). The first member of this family to be identified and biochemically characterized was a thiol oxidase known as SOX, which was purified from avian egg whites26,27,28,29.

Table 1 Characterized Erv-like proteins

The bacterial oxidation machinery

Many insights into the most fundamental aspects of protein disulphide-bond formation have come from studying the pathways of disulphide-bond formation in the periplasmic space of bacterial cells. In Gram-negative bacteria, the periplasmic space forms a compartment for the formation of disulphide bonds, which has similarities to the eukaryotic ER. Two proteins, the periplasmic thiol-disulphide oxidoreductase DsbA and the cytoplasmic inner-membrane protein DsbB, drive the formation of disulphide bonds in periplasmic proteins (reviewed recently in Refs 30–33) (Fig. 2). A disulphide bond that is formed between the active-site cysteines of DsbA is transferred directly to periplasmic substrate proteins, and the reduced form of DsbA is efficiently reoxidized by DsbB. The transfer of oxidizing equivalents between DsbB and DsbA occurs through direct protein-to-protein thiol-disulphide exchange, as shown by the capture of disulphide-linked complexes that contain DsbB and DsbA34,35.

Figure 2: Periplasmic pathways for protein oxidation and isomerization in Escherichia coli.
figure 2

Oxidizing equivalents are introduced into substrate proteins in the periplasmic space through a thiol-disulphide exchange reaction (thin arrows) with the soluble thiol-disulphide oxidoreductase DsbA. DsbA is reoxidized by the cytoplasmic membrane protein DsbB, which, in turn, obtains oxidizing equivalents from a quinone cofactor (Q). Electrons are ultimately transferred from quinone to oxygen under aerobic conditions, or to nitrate or fumarate under anaerobic conditions, by a cytochrome oxidase. Disulphide-bond rearrangement is catalysed by the thiol-disulphide oxidoreductase DsbC, which is maintained in a reduced state by the cytoplasmic membrane protein DsbD. DsbD is kept reduced by cytoplasmic thioredoxin, which ultimately obtains electrons from NADPH through thioredoxin reductase (not shown).

In addition to the DsbA–DsbB system for the formation of disulphide bonds, bacteria also contain a pathway that is dedicated to the isomerization of incorrectly paired cysteines. The two components of the isomerization pathway are the thiol-disulphide oxidoreductase DsbC and the cytoplasmic membrane protein DsbD (reviewed in Refs 30–33) (Fig. 2). DsbC catalyses disulphide reshuffling by reducing incorrectly paired disulphides. It is not known whether the complete reshuffling reaction can be carried out by DsbC alone, or whether a further oxidation step by DsbA is required. DsbC is maintained in a reduced active state by a continual flow of electrons from cytoplasmic thioredoxin to DsbC through the cytoplasmic membrane protein DsbD36,37. The transmission of a reducing potential across the cytoplasmic membrane is facilitated by a cascade of thiol-disulphide exchange reactions that take place between cysteines in the DsbD protein38,39,40.

Comparing prokaryotic and eukaryotic pathways

There are many similarities between the key components of the prokaryotic and eukaryotic pathways. Most notable are the phenotypic parallels between mutants in these systems. Disruption of either DsbA or PDI interferes with the oxidation of secretory proteins15,41. The functional overlap between DsbA and PDI is evident with the ability of dsbA mutants to be complemented by the introduction of PDI into the bacterial periplasm42. Loss of functional Ero1 results in the accumulation of reduced PDI15, and disrupting the function of DsbB causes defects in the reoxidation of DsbA43. Given the functional similarity between DsbB and Ero1, it is perhaps surprising that these proteins share no obvious sequence homology. The only apparent similarity between the primary sequence of these two proteins is the presence of two active-site cysteine pairs, which are essential for the function of either protein as a redox catalyst17,44. Alterations in the cysteines of DsbB result in phenotypes that are similar to those observed with Ero1 cysteine mutants17,45. It is interesting to note that, among the known bacterial DsbB homologues, the only amino acids that are strictly conserved are the two cysteine pairs and an arginine (Arg) at position 48 (Ref. 46). Arg48 seems to assist in the interaction of DsbB with a QUINONE cofactor, which is necessary for DsbB oxidation46.

A striking difference between the characterized prokaryotic and eukaryotic systems of disulphide-bond formation is the absence of an identified isomerization pathway in eukaryotes. At present, it is not known whether there is a reduction pathway that is analogous to the prokaryotic DsbC–DsbD system in eukaryotes.

Disulphide transfer in a single protein

All of the enzymatic pathways described above use a conserved thiol-disulphide exchange mechanism to transfer disulphide bonds between separate components of the cellular redox systems. In addition to these inter-protein transfer events, it seems that the eukaryotic and prokaryotic pathways share a similar mechanism for disulphide-bond transfer between several pairs of cysteines in a single protein. The best-characterized example of such an intra-protein transfer event is the passage of electrons from the cytoplasm to the periplasm by DsbD. The mechanism of electron transfer by DsbD involves a cascade of disulphide-bond reduction events that take place between the three pairs of essential cysteines that are present in the DsbD protein38,39,40,47,48,49,50.

It has been proposed that the activities of both DsbB and Ero1 rely on a similar disulphide shuttle between the two essential cysteine pairs in each protein17,34,45. For both DsbB and Ero1, one cysteine pair is thought to interact directly with its partner thiol-disulphide oxidoreductase (DsbA and PDI, respectively), and this thiol pair is thought to be reoxidized by the internal transfer of oxidizing equivalents from the second cysteine pair. A direct interaction between the two cysteine pairs of DsbB was confirmed recently by the detection of a transient disulphide bond between the two active-site cysteine pairs51. A similar transfer of oxidizing equivalents between cysteine pairs in a single molecule has also been indicated for Erv2 (Ref. 25). Here, a cysteine pair in a flexible tail region of the protein has been proposed to accept electrons from target proteins, and to shuttle these electrons to the FAD-proximal Cys-X-X-Cys cysteine pair25. The conservation of both pairs of cysteines in all of the Erv2 homologues (Table 1) lends support to such a transfer model. The only exception is the viral protein E10R, which contains a single Cys-X-X-Cys active site52. However, E10R associates with another viral protein, A2.5L, that contains a cysteine pair in a Cys-X-X-X-Cys motif53.

Intriguingly, the structural analysis of several of the cellular thiol-disulphide oxidoreductases indicates that the relay of oxidizing equivalents might follow an alternating pattern of transfer between thioredoxin-like domains and non-thioredoxin-like domains. Thioredoxin-like domains adopt a characteristic structure formed by α helices and β sheets with the overall fold βαβαβαββα (Ref. 54). The active-site Cys-X-X-Cys motif is found in an exposed turn that links β2 to α2 (Ref. 54). In the DsbC–DsbD system, electrons are transferred from the cytoplasmic thioredoxin, to cysteines in the non-thioredoxin-like transmembrane domain of DsbD, then to cysteines in a thioredoxin fold in the carboxy-terminal periplasmic domain of DsbD, on to the non-thioredoxin amino-terminal DsbD domain, and finally to the thioredoxin family member DsbC38,55,56. Although the FAD-proximal Cys-X-X-Cys cysteine pair of Erv2 is not in a thioredoxin fold, the structural environment of this Cys-X-X-Cys pair is similar to that of the DsbA and thioredoxin active sites25,57. The cysteine pair in the non-thioredoxin-like tail region of Erv2 might serve as a disulphide-bond shuttle between thioredoxin-like proteins, such as PDI, and the thioredoxin-like environment of the FAD-proximal cysteine pair. The alternation of disulphide-bond transfer between regions with different protein folds, which is observed with DsbD and Erv2, might reflect a fundamental regulatory mechanism that allows transfer only between thioredoxin-like and non-thioredoxin-like domains. These structural constraints might function to direct the flow of electrons along specific pathways.

Specificity of transfer

Several thioredoxin-like proteins have been identified in the ER and the periplasmic space (see Table 2 and Box 4 for a discussion of eukaryotic homologues). A central question that remains unanswered, however, is which functions each of the thioredoxin-like proteins have in disulphide-bond formation. In yeast, it has been speculated that Ero1 and Erv2 have different preferences for each of the PDI homologues, and perhaps act on different substrate proteins. Ero1 transfers oxidizing equivalents to PDI and Mpd2 (Ref. 15), and perhaps Mpd1 (Ref. 58), whereas, so far, Erv2 has only been shown to associate with PDI23. The interaction of mammalian Ero1-Lα with PDI and ERp44, but not with the homologous protein ERp57, might also reflect the presence of distinct oxidation pathways in the mammalian ER21,59. The idea of several oxidizing or reducing pathways that are designed for distinct substrates is borne out in the prokaryotic system in which DsbD promotes disulphide-bond isomerization (by DsbC) as well as cytochrome c maturation (by CcmG) (reviewed in Refs 31,32,60).

Table 2 Eukaryotic PDI-like proteins

The simultaneous operation of several protein-oxidation pathways within the ER and periplasmic space would require a way to ensure the specificity of disulphide transfer to appropriate substrates. So far, it is not clear how such specificity is achieved. Perhaps the flexible carboxy-terminal region of Erv2 (discussed above) is designed to interact specifically with a particular PDI-like partner molecule? It is interesting to note that all the ERV-LIKE FAMILY members contain a Cys-X-Cys or Cys-X-X-Cys cysteine pair in addition to the Cys-X-X-Cys pair found in the 100-residue conserved core domain, but the position of these cysteines relative to the core domain Cys-X-X-Cys pair varies between the family members (Table 1). The protein context of the second pair of cysteines might direct the interaction of each Erv-like protein with a unique thioredoxin-like partner. A DsbD homologue from Rhodobacter capsulatus — CdcA — contains a cysteine pair in a DsbD-like hydrophobic domain, but lacks the two extra active-site cysteine pairs that are found in the periplasmic domains of DsbD. Interestingly, a recent comparison of the functional domains of Escherichia coli DsbD and R. capsulatus CdcA indicates that the extra thiol-containing domains of DsbD might expand the substrate range of DsbD relative to CcdA61. Perhaps other redox-active proteins also contain cysteine pairs in unique structural motifs that promote interactions with specific substrate thiol-disulphide oxidoreductases?

The coincident operation of oxidation and isomerization pathways in the periplasmic space poses a similar problem for specific electron transfer. For proteins of the oxidizing pathway to be able to carry out their function, their active sites must be in an oxidized state. Similarly, proteins that function to reduce or isomerize substrates must achieve a reduced state to fulfil their roles (see, for example, Ref. 62). The accidental transfer of oxidizing equivalents into the isomerization pathway, or of reducing equivalents into the oxidizing pathway, would inactivate the enzymes in either pathway and result in the incapacitation of either system. Indeed, little cross-talk in the form of disulphide-bond transfer between the pathways for oxidation and reduction/isomerization is evident: in vitro, DsbB oxidizes only DsbA and not DsbC, despite the fact that both proteins have similar redox potentials63. A recent analysis of dsbC mutants that can complement a dsbA-null strain indicates that dimerization of the DsbC isomerase/reductase enzyme might normally block its active sites from recognition by DsbB, and prevent misoxidation of DsbC by DsbB64.

Coupling disulphide bonds and small molecules

Glutathione and the ER redox potential. To fully understand the process of cellular protein oxidation, it is necessary to determine the ultimate origin of the oxidative power for disulphide-bond formation. Several small thiol-containing molecules, such as cystamine65, vitamin K epoxide66 and glutathione67, have been proposed to contribute to the oxidation of proteins in the ER lumen. Of these molecules, oxidized glutathione has attracted the most attention and, until recently, it was widely considered to be the prime candidate for the source of the oxidizing equivalents that are necessary to generate protein disulphide bonds. The most compelling evidence in support of a role for glutathione in protein oxidation came from the observation that a higher ratio of oxidized to reduced glutathione is present in the ER relative to the cytosol67. The mixture of oxidized and reduced glutathione detected in the ER was similar to that found in redox buffers that afford optimal rates of protein oxidation in vitro68.

However, a direct experimental test in S. cerevisiae showed that, despite the abundance of oxidized glutathione in the ER lumen, glutathione is not required for oxidative protein folding in the eukaryotic ER13. Moreover, the in vitro oxidation of RNase A by purified Ero1 and PDI does not require oxidized or reduced glutathione16. Interestingly, the disulphide-bond formation that is driven by the in vitro Ero1–PDI system can proceed in the presence of a vast excess of reduced glutathione16. Likewise, the cytoplasmic vaccinia virus protein-oxidation pathway can operate in the presence of the excess of reduced glutathione that is found in the cytoplasm69,70,71. These observations have led to a revised view that protein-oxidation pathways proceed by the direct transfer of oxidizing equivalents between enzymes and do not rely on oxidizing equivalents provided by glutathione.

Further experiments have indicated that in vivo glutathione might compete with proteins for oxidizing equivalents. In ero1-1 mutants, the reduction of the intracellular glutathione level, by disruption of the GSH1 gene (which encodes the enzyme that catalyses the first and rate-limiting step in glutathione synthesis), actually restores disulphide-bond formation activity to the compromised protein-oxidation system72. So, glutathione acts as a net reductant in the ER that counteracts the oxidizing activity of the Ero1 pathway. The production of oxidized glutathione could result from the reduction of a protein disulphide bond in any component of the eukaryotic protein-oxidation pathway: Ero1, PDI or secretory proteins. Recent experiments have begun to narrow down the potential source of glutathione oxidation in the ER. The in vitro characterization of Ero1 and Erv2 has shown that neither protein directly oxidizes glutathione16,23. The complete reconstitution of the Ero1–PDI pathway for protein oxidation in vitro indicates that glutathione oxidation is driven by Ero1-derived disulphide bonds in PDI and/or substrate proteins16.

These observations raise the question of why the ER maintains two seemingly competing pathways: a glutathione-based pathway that introduces reducing equivalents and a protein-oxidation pathway that is driven by the enzymatic transfer of oxidizing equivalents. The importance of a proper ratio of reducing and oxidizing equivalents for in vitro refolding reactions has been shown repeatedly. An attractive possibility is that glutathione functions as a buffer for the ER redox environment. Under hyperoxidizing conditions, the reducing equivalents from glutathione might be used to reduce improperly paired cysteines, facilitating the correct folding of proteins. Instead of interacting directly with substrate proteins, glutathione could also reduce the normally oxidized PDI, shifting PDI activity from oxidation to isomerization. Glutathione might also counteract oxidative stress simply by consuming excess oxidizing equivalents during the conversion of reduced glutathione to oxidized glutathione. A role for glutathione in counteracting oxidative stress is supported by the observation that oxidative protein folding is more readily compromised by the addition of the oxidant diamide in a gsh1 mutant strain72.

Flavins and eukaryotic disulphide bonds. As glutathione seems to provide reducing, rather than oxidizing, equivalents in the ER, a renewed search has begun for the oxidative source for the ER. Recent experiments indicate that flavin moieties provide a source of oxidizing equivalents for both the Ero1 and Erv2 pathways of disulphide oxidation. In vivo, the depletion of riboflavin, and therefore its flavin derivatives, including FAD, inhibits disulphide-bond formation and results in the accumulation of reduced Ero116. The in vitro oxidative folding of reduced RNase A that is catalysed by purified Ero1 and PDI also seems to rely on the oxidizing equivalents that are provided by the addition of FAD16. However, the ultimate oxidizing source for FAD and Ero1 remains elusive. During a catalytic cycle of the Ero1 system, FAD will become reduced to FADH2 on the transfer of oxidizing equivalents to Ero1. In the in vitro pathway, the requirement for a stoichiometric excess of FAD indicates that Ero1 might exchange the reduced FADH2 for oxidized flavin, FAD, from solution. Although Ero1 might exchange FADH2 for free FAD in vivo, it seems unlikely that such an exchange is the normal physiological mechanism for Ero1 oxidation. Most flavoproteins tightly bind their cofactors, which would impede a catalytic exchange mechanism. In addition, the concentration of FAD in yeast cells is much lower than the levels required for the in vitro Ero1 oxidation reaction73.

Although the identity of the oxidant for Ero1 and its FAD cofactor remains elusive, physiological experiments give us some clues about the types of oxidation process that are possible. The ero1-1 mutant is not viable at high temperatures, either in the presence or the absence of oxygen, which indicates that Ero1 is an essential part of the oxidation pathway under aerobic and anaerobic conditions. Although Ero1 might use molecular oxygen as an electron acceptor during aerobic growth, the ability of Ero1 to operate under conditions in which oxygen is limited indicates that there must be a physiological electron acceptor for Ero1 that is not molecular oxygen and does not depend on oxygen for its generation. Conversely, molecular oxygen functions as the obligate electron acceptor for the second ER pathway that is driven by Erv2 (Ref. 23). In the Erv2 pathway, the flavin cofactor of Erv2 interacts directly with molecular oxygen to contribute the oxidizing equivalents that are necessary for disulphide-bond formation.

Quinones as prokaryotic electron carriers. In prokaryotes, a more complete understanding of how the oxidation of protein thiols is integrated into the redox chemistry of the cell has been achieved. Experiments in E. coli have shown that the RESPIRATORY ELECTRON-TRANSPORT CHAIN (Box 5) is necessary for the complete oxidation of DsbB. Disruption of the respiratory chain, by depletion of the intracellular pools of haem or ubiquinone and menaquinone, impedes the flow of oxidizing equivalents into the DsbA–DsbB system74. Under these depletion conditions, DsbA accumulates in its reduced form74. The recent reconstitution of the DsbA–DsbB system has established that DsbB uses a small electron carrier, a quinone cofactor, to transfer electrons to the terminal oxidases of the electron transport chain and then to either molecular oxygen or other electron acceptors63,75. Under conditions of aerobic growth, electrons flow from DsbB directly to ubiquinone that is associated with cytochrome bd or bo oxidase, and then to molecular oxygen. During anaerobic growth, DsbB uses menaquinone as an electron carrier that transfers electrons to alternative acceptors such as fumarate and nitrate, rather than oxygen75. Alleles of dsbB that encode single amino-acid substitutions for Arg48 show a greater defect in the use of menaquinone than of ubiquinone46. Consistent with the role of menaquinone as the anaerobic electron acceptor for DsbB, these mutants show the greatest defect in protein oxidation under anaerobic growth conditions.

Future directions and implications

The past few years have seen significant advances in our understanding of the pathways of protein disulphide-bond formation in the periplasm of bacteria and the ER of eukaryotic cells. This review has concentrated on the emerging similarities between the prokaryotic and eukaryotic systems. Both pathways include a conserved thiol-disulphide exchange mechanism that transfers disulphide bonds between the enzymatic components of the pathways of disulphide-bond formation. In addition, new mechanistic insights into the functions of several redox-active proteins show that cellular redox pathways often rely on the relay of electrons between pairs of cysteines in a single protein. The cellular oxidation pathways seem to be controlled by the specificity of intra-protein and inter-protein interactions. The work that was discussed here also introduced a new family of eukaryotic and viral thiol-oxidases, the Erv-like family, whose role in disulphide-bond formation was identified recently. Notably, the initial characterization of members of the Erv family shows that many of the same characteristics are shared between the more established ER and periplasmic pathways.

The studies reviewed here provide solid groundwork for future studies of protein disulphide-bond formation. It will be of interest to understand the biological significance and division of labour among the various homologues that are implicated in disulphide-bond formation in mammalian and yeast cells. Similarly, the diversity and ubiquity of the Erv family of proteins indicate that it might be possible to extend our understanding of how oxidizing equivalents can be transferred specifically from one protein to another, and to other compartments, such as the mitochondria, cytosol and extracellular space.

The early focus on the identification and initial characterization of the pathways of protein oxidation and/or reduction in prokaryotes and eukaryotes has clearly shifted during the past few years towards understanding the mechanistic and structural details of these pathways. Now, the structural data on the amino-terminal domain of DsbD56, the flavoprotein-oxidase Erv2 (Ref. 25) and the bacterial thioredoxin-like proteins DsbA and DsbC55,57, along with the recent ability to reconstitute the Ero1–PDI16, DsbB–DsbA63,76 and DsbD–DsbC40 pathways in vitro, have set the stage for a detailed structural picture of the pathways of disulphide-bond formation. The tools available allow rational mutagenesis, domain swapping and biochemical studies, to test the current models that are designed to explain the specificity observed in the electron transfer in and between proteins. Clearly, the structural analysis of Ero1, as well as of complexes between Ero1 and PDI, is a crucial goal for understanding the structural basis of selectivity in eukaryotic disulphide-bond formation.