2.2. Protein disulfide isomerase

2.2.1.1. Cloning, sequence analysis and cellular localization of PDI

The first full-length sequence information on PDI came from a rat liver cDNA which encoded for 508 amino acids, including an N-terminal ER signal sequence and C-terminal ER retrieval signal. The mature protein was seen to contain two segments with homology (47% identity) to each other, regions a (amino acids 9-90) and a’ (353-431). The midpart of the protein, regions b and b’, corresponding to amino acids 153-244 and 256-343 also showed some homology to each other, (28% identity) (Edman et al. 1985). Highly conserved sequences near the beginning of both a and a’ were seen to contain two cysteines separated by a glycine and histidine; these putative active sites were later identified as such (Hawkins & Freedman 1991, Vuori et al. 1992a, LaMantia & Lennarz 1993, Lyles & Gilbert 1994). Upon comparison, homology to thioredoxin was seen in regions a and a’, but no significant homology to other proteins was initially seen in regions b and b’ (Edman et al. 1985).

Despite the presence of the N-terminal ER signal sequence and the -KDEL ER retrieval sequence at the C-terminus of PDI, other cellular and extracellular localisations have been reported (for a review see Turano et al. 2002). PDI has been found to be secreted from hepatocytes (Terada et al. 1995), pancreatic exocrine cells (Yoshimori et al. 1990), endothelial cells (Hotchkiss et al. 1998) and platelets (Chen et al. 1992). It has also been found on the cell surface of a variety of different cell types, such as the retina in chicken embryo (Pariser et al. 2000). PDI has also been found in the cytosol of human and monkey liver cells (Wroblewksi et al. 1992) and also a nuclear localization has been reported (Gerner et al. 1999). It is unclear what the physiological relevance of these observations is.

2.2.1.2. Domain organization

The sequence analysis by Edman et al. 1985, led to further analysis and definition of the modular structure of PDI. While PDI probably consists of four structural domains, the work required to achieve a complete picture of the structure of the enzyme and of its individual domains is still continuing.

From the information gained from the DNA sequence of the enzyme, it was noted that PDI contains two segments with homology to thioredoxin, a small, ubiquitous protein which catalyses reduction of protein disulfides (Holmgren 1985). These homologous segments, designated a and a’, contain the active site motif -Cys-Gly-His-Cys-. Two other regions, b and b’ also have limited homology to each other. The initial analysis identifying these four segments only covered 337 amino acids, or 69% of the mature protein. Following limited proteolysis and homology modelling studies, a six-domain structure a-e-b-b’-a’-c was proposed (Freedman et al. 1994) The e domain was proposed to exist between a and b based on homology to the estrogen-binding domain of human estrogen receptor. Studies on recombinant fragments of PDI have since excluded the concept of the e domain (Kemmink et al. 1997, Freedman et al. 1998). A combination of studies based on limited proteolysis and recombinant protein production have confirmed the original predictions of a four-domain structure, a-b-b’-a’. However, the boundaries are not those originally defined, with all four domains being longer than first thought (Alanen et al. 2003a). In addition to these domains, PDI has a C-terminal extension of 29 amino acids, out of which 18 are either aspartate or glutamate (Pihlajaniemi et al. 1987). The significance of this very acidic tail of PDI is not known, but at the very C-terminus resides the endoplasmic retrieval sequence –KDEL which is necessary for retaining the enzyme in the lumen of the ER.

A schematic representation of the domain architecture of PDI is presented in Figure 1.

The three-dimensional structure of full-length PDI is not known. To date, the individual domains a (Kemmink et al. 1996) and b (Kemmink et al. 1999) have been structurally defined by NMR. The structures of domains a and b are represented as ribbon diagrams in figure 2. Both domains have a fold similar to that of thioredoxin (β αβ αβ β α). This is somewhat surprising, as there is no homology, between the b domain and either a or a’ or thioredoxin. Work on the structure of domain a’ is underway (Dijkstra et al. 1999) and sequence similarity to domain a strongly suggests that it also has the thioredoxin-like fold. The evolution of the four-domain structure of PDI probably arose from internal gene duplications, which during time lost homology but retained the fold. Phylogenetic studies show that the molecular evolution of PDI and the PDI family of proteins started from an ancestral enzyme with two thioredoxin-like domains (Kanai et al. 1998, McArthur et al. 2001). Given the homology between b and b’ it is probable that PDI consists of four thioredoxin-like modules, with two of these containing oxidoreductive active sites.

2.2.1.3. Thioredoxin fold

The thioredoxin fold was named after the first protein it was discovered in, E.coli thioredoxin, a structure which was solved in 1990 (Katti et al. 1990). Thioredoxin is a small cytosolic, nuclear and mitochondrial protein found in abundance in all species from archaebacteria to mammals. It is a general protein reductant, involved in a wide range of physiological functions from inhibition of apoptosis to being an antioxidant (Powis & Montfort 2001). Thioredoxin is a single-domain protein formed from a five-stranded β -sheet surrounded by four α-helices (Martin 1995). The thioredoxin fold, as found in all the members of the thioredoxin-like family, has an N-terminal β αβ motif, in which the beta strands run parallel to each other, and a C-terminal β β α motif, in which the β strands run antiparallel to each other. These are connected by a third α-helix which, in effect, forms a loop linking the two motifs (Martin 1995).

The thioredoxin fold has been found in a number of enzyme families involved in sulfur metabolism. Many of these enzymes share the Cys-X-Y-Cys active site motif and have a functional similarity, in that they are involved in sulfur-based redox reactions in the cell, e.g. thioredoxins, glutaredoxins, DsbAs and PDIs. The residues between the two active site cysteines vary between each protein family and they are important in determining the active site potential and hence the physiological function of the enzyme. The effects the residues between active site cysteines have on the active site potential have been studied extensively in the bacterial periplasmic thiol-disulfide oxidoreductase DsbA (for example Gane et al. 1995, Grauschopf et al. 1995, Chivers et al. 1997, Guddat et al. 1997, Jacobi et al. 1997, Huber-Wunderlich & Glockshuber 1998). The active site potential of the enzyme can be modulated by mutations in the -XY- residues, which in part explains why enzymes with overall structural similarity, can be reducing, such as thioredoxin (CGPC), or highly oxidizing, such as DsbA (CPHC) or acting as isomerase, such as PDI (CGHC).

Some of the proteins that have been identified as having a thioredoxin fold in their structure do not have a thioredoxin-like active site and their function is also different, but they are involved in sulfur metabolism. For example glutathione S-transferase and glutathione peroxidase both interact with the cysteine-containing substrate, glutathione (Reimener et al. 1991, Epp et al. 1983).

2.2.2.1. Redox-isomerase function

Disulfide bond formation as an essential part of protein folding has been studied for decades. The complexity of these processes involving reduction, oxidation and isomerization of disulfide bonds, the variety of catalysts known in eukaryotic cells and the recent discovery of multiple different pathways for disulfide bond formation have hindered the progression of a clear understanding of actual physiological events. In vitro studies have offered good models to study the catalysis and kinetics of PDI activities towards a variety of substrates (Creighton et al. 1980, Lyles & Gilbert 1991, Goldenberg 1992, Weissman & Kim 1993, Darby et al. 1994, Creighton et al. 1995, Darby & Creighton 1995a, Freedman et al. 1995, Ruoppolo & Freedman 1995, Ruoppolo et al. 1996). However, these in vitro studies are very dependent on the conditions used and do not necessarily explain how PDI functions in vivo in catalyzing these reactions. The use of mutant forms of PDI and individual domains or domain combinations in different assays have provided an insight into the mechanisms of PDI function (Vuori et al. 1992a, LaMantia & Lennarz 1993, Lyles & Gilbert 1994, Laboissiere et al. 1995, Darby & Creighton 1995b, Walker et al. 1996).

PDI is capable of catalyzing both oxidation and reduction of disulfides under physiological conditions as well as disulfide isomerization, and this adds considerably to the complexities of data interpretation. A schematic representation of the mechanisms of catalysis of reduction, oxidation and isomerization of disulfide bonds by PDI is presented in Figure 3.

The essential basis for the redox activity of PDI is the presence of two thioredoxin-like motifs -CGHC- in domains a and a’ (Edman et al. 1985), which by themselves account for the catalytic activity in thiol-disulfide exchange reactions (Hawkins & Freedman 1991, Lundström & Holmgren 1993, Darby & Creighton 1995a, 1995b, Kortemme et al. 1996). The thiol-disulfide redox state of the active sites determines the direction of catalysis: oxidation of a substrate requires the conversion of a disulfide -CGHC- site to a dithiol, while reduction of a substrate requires the conversion of a dithiol active site to a disulfide. In each active site, the N-terminal cysteine is surface-exposed and reactive towards the substrate, forming a mixed disulfide with a substrate cysteinyl residue (Hawkins & Freedman 1991). Isomerization by PDI occurs via two possible pathways: through direct isomerisation or through cycles of reduction followed by reoxidation (Schwaller et al. 2003).

The presence of two catalytic active sites in PDI probably, in part, accounts for the multiplicity of functions and broad specificity of the enzyme. By mutational studies both a and a’ active sites have been shown to account for 50% of the full activity of PDI, indicating that the two sites operate independently (Vuori et al. 1992a). What role each active site has in vivo is not known. It has been shown, however, that efficient catalysis of native disulfide bond formation requires domains other than just a and a’. By studying individual domains and their linear combinations Darby et al. (1998) showed that the presence of the b’ domain is essential for the catalysis of the full range of reactions tested. Specifically they demonstrated that simple oxidation reactions require only a and a’, simple isomerization reactions a linear combination of domains including b’ and either a or a’, while isomerisation reactions which require a significant conformational change in the substrate are only catalysed by full length PDI (excluding the c region). As the primary non-native protein binding site is located in domain b’ (Klappa et al. 1998a) this finding is very logical, as reactions such as the reorganisation of disulfide bonds in complex substrates could be predicted to require high affinity binding of the substrate.

2.2.2.2. Folding assistant/chaperone

In addition to its role in native disulfide bond formation PDI has been shown to catalyze reactivation of non-native proteins without disulfide bonds. This function, in the literature termed the chaperone or foldase function, is independent from the redox/isomerase function, as it has been shown to exist in vitro with several model substrates that do not contain disulfides in their native structure (Cai et al. 1994, Song & Wang 1995, Yao et al. 1997). In addition, PDI with its active sites inactivated is still capable of reactivating such proteins (Quan et al. 1995) which further indicates that the redox/isomerase and foldase functions are distinct from each other. The foldase function of PDI must be dependent on the ability to bind or interact with unfolded polypeptides with broad specificity. The reported interaction of PDI with its substrates is weak (K_d > 100 M) and is dependent on the length of the area of substrate backbone binding (Morjana & Gilbert 1991). However, in substrates that contain cysteine residues the interaction is stronger (Morjana & Gilbert 1991, Klappa et al. 1997), possibly due to the formation of a mixed disulfide between PDI and the substrate.

The first report on the site of peptide binding in PDI was in 1993 when Noiva et al. purified a tryptic fragment of PDI that had been photoaffinity-labelled with a tripeptide probe. They identified this site of interaction to be at the end of domain a’, near the highly acidic C-terminus of PDI (Noiva et al. 1993). However, the binding affinity of PDI to tripeptides had previously been shown to be very weak (Morjana & Gilbert 1991) and hence the significance of this observation is unclear.

A different substrate binding site was identified by Klappa et al. (1998a). They used longer substrates and chemical cross-linking with individual domains and combinations of domains of PDI expressed in E.coli. Domain b’ was found to be essential for the binding of all substrates and sufficient for the binding of short peptides. However, with longer substrates additional domains were needed, with the shortest construct able to bind a synthetic 28 amino acid fragment of BPTI being b’a’c or abb’. Binding of the non-native protein ‘scrambled’ RNase was found to require the presence of both of the b’ and a’ domains and efficient binding was only observed with full length PDI (excluding the c region).

The binding interaction of PDI with substrates was reported to be dependent on the redox state of PDI by Tsai et al. (2001), who demonstrated a glutathione-dependent dissociation of PDI from cholera toxin. However, the reduction or oxidation state of PDI does not affect the interaction of PDI with either C-propeptide of procollagen, which PDI is known to bind in vivo (Bottomley et al. 2001), or the interaction with collagen prolyl 4-hydroxylase α-subunit, which PDI is known to form a stable enzyme tetramer complex with (Pihlajaniemi et al. 1987). Furthermore, Lumb and Bulleid (2002) demonstrated that dissociation of PDI from cholera toxin was observed upon competition with the 14-amino acid mastoparan, which the authors say could explain the original observation of the effect of glutathione, a tripeptide which can act as a substrate to PDI binding, i.e. that substrate binding is independent of the redox state of PDI.

While the binding specificity of the pancreas specific PDI homologue PDIp has been reported to be a tyrosine or tryptophan with nonadjacent negative charge (Ruddock et al. 2000), the binding specificity of PDI is not known. Binding is reversible and primarily hydrophobic in nature (Klappa et al. 1998b), but no simple pattern is discernable for the full range of substrates bound by full length PDI.

2.2.2.3. The β subunit of collagen prolyl 4-hydroxylase

Collagen prolyl 4-hydroxylase (C-P4H) (EC 1.14.11.2) is an enzyme, which catalyzes the formation of 4-hydroxyprolines in collagens. Hydroxylation of prolines is an essential modification for the formation of the collagen triple helix at body temperature (see Kivirikko & Pihlajaniemi 1998 and Myllyharju 2003 for reviews). In addition to C-P4Hs, there exists another, only recently characterized family of cytoplasmic P4Hs, which regulate the hypoxia-inducible transcription factor HIFα (Bruick & McKnight 2001, Epstein et al. 2001). Other than their catalytic regions, C-P4Hs and HIF P4Hs do not share sequential homology (Myllyharju 2003).

In vertebrates, the C-P4H is an α₂β ₂ tetramer with molecular weight of around 240 kDa; the individual subunits being 63 kDa and 55kDa, respectively (Pänkäläinen et al. 1970, Berg & Prockop 1973, Tuderman et al. 1975). Several isoforms of the α subunit are known in human and mouse tissue, Caenorhabditis elegans and Drosophila melanogaster (Helaakoski et al. 1989, Veijola et al. 1994, Helaakoski et al. 1995, Annunen et al. 1997, Friedman et al. 2000, Hill et al. 2000, Winter & Page 2000, Abrams & Andrew 2002, Riihimaa et al. 2002). There are three isoforms of the α subunit in vertebrates, the original α(I) isoform and the more recently identified α(II) (Helaakoski et al. 1995, Annunen et al. 1997) and α(III) (Van Den Diepstraten et al. 2003). In addition, new human and mouse α subunit-like gene products are being characterized (Myllyharju 2003). While vertebrate C-P4Hs are only known to exist in the α₂β ₂ tetrameric form, in C. elegans some dimer formation is known to happen in vivo, although the main form is a mixed tetramer formed of two α isoforms PHY1 and PHY2 and two β subunits (Myllyharju et al. 2002). Human α subunits do not form a mixed α(I)/α(II) tetramer with β subunit (Annunen et al. 1997). In addition to the main enzyme α isoforms, other α subunit-like polypeptides are known to be encoded in the genome of C.elegans (Winter & Page 2000). Collagen prolyl 4-hydroxylase is a glycoprotein with all of the carbohydrate moieties being attached to α subunits (Kedersha et al. 1985). Glycosylation of the human and mouse α(I) subunit is not needed for the formation of a functional C-P4H tetramer (Lamberg et al. 1995)

The β subunit of collagen prolyl 4-hydroxylase was cloned 1987 and found to be identical to PDI (Koivu et al. 1987, Pihlajaniemi et al. 1987). The isolated subunits have no prolyl 4-hydroxylase activity, but the β /PDI subunit has approximately 50% of the wild-type PDI activity in rearranging “scrambled” ribonuclease disulfide bonds, an assay which is much-used in determining the isomerase activity of PDI, even when present in the α₂β ₂ enzyme tetramer. However, PDI activity is not needed for tetramer assembly or for prolyl 4-hydroxylase activity, as was shown by inactivating PDI active sites by mutating them to -SGHC-. Tetramer formation was not affected by these mutations in PDI and the tetramer formed was fully active in hydroxylation of prolines. Inactivating the N-terminal active site of PDI in domain a leads to about 50% isomerase activity of the PDI /β subunit in the α₂β ₂ tetramer. However, inactivating the C-terminal active site of PDI in domain a’ leads to a loss of isomerase activity of the PDI/β subunit in the α₂β ₂ tetramer, indicating that, in the tetramer structure, the N-terminal active site is unavailable for reaction (Vuori et al. 1992b). As with PDI, no high resolution structure exists for any C-P4H, so it is not known, in what orientation the subunits assemble into a tetramer.

Various studies have shown, that in the absence of the β /PDI subunit the α subunit aggregates and has no activity (Kivirikko et al. 1992, Vuori et al. 1992b, Veijola et al. 1994, Helaakoski et al. 1995) The primary function of PDI in collagen prolyl 4-hydroxylase therefore appears to be to keep the α subunit in soluble conformation. In addition, as the α subunit does not have an ER retention signal, PDI may be responsible for retaining C-P4H in the lumen of the ER (Vuori et al. 1992b). PDI probably also contributes significantly to the structure/function of the C-P4H, as coexpression of prolyl 4-hydroxylase α subunit with the molecular chaperone BiP results in the formation of soluble complexes but no prolyl 4-hydroxylase activity was generated (Veijola et al. 1996).

2.2.2.4. The smaller subunit of microsomal triglyceride-transfer protein

The microsomal triglyceride-transfer protein (MTP) has molecular weight of about 155 kDa. It catalyzes the transport of triglyceride, cholesteryl ester and phosphatidylcholine through membranes and is required for the assembly of apoB-containing lipoproteins in the liver and intestinal cells. It is an αβ heterodimer consisting of subunits with molecular weights of approximately 97 and 55 kDa. The smaller subunit has been identified as PDI (Wetterau et al. 1990). The isolated PDI subunit has no lipid-transfer activity, and the larger 97 kDa subunit is active only in the complex. PDI may have a role in the retention of MTP in the lumen of the ER, as the larger subunit lacks an ER retrieval signal. Also, as in the case of C-P4H, the α subunit aggregates after dissociation from the PDI subunit and therefore the role of PDI could be directly related to maintaining the complex soluble (Wetterau et al. 1991). This was studied further by Lamberg et al. (1996) who expressed the 97 kDa subunit alone in insect cells and found that it formed insoluble aggregates. When co-expressed with PDI, soluble dimers with MTP activity were formed. To study whether the redox activity of PDI is necessary for the activity of MTP dimer, the 97 kDa subunit was co-expressed with PDI in which both active sites had been inactivated by mutating them to -SGHC-. Soluble dimers with comparable levels of MTP activity were formed. This indicates that the role of PDI in MTP is to keep the 97 kDa subunit in a catalytically active, soluble conformation (Lamberg et al. 1996)

2.2.2.5. Other functions

A large number of other functions have been suggested for PDI in the literature. A major cellular tri-iodothyronine binding protein has been reported to be identical to PDI (Yamauchi et al. 1987). Since tri-iodothyronine is a thyroid hormone, which has nuclear receptors for induction and repression of transcription of selective genes, the physiological significance of this finding is unclear.

PDI has also been characterized as a glycosylation-site binding protein in the lumen of the ER, recognising multiple polypeptide sequences for oligosaccharyl transferase (Geetha-Habib et al. 1988), but this finding was subsequently shown to be due to an artefact (Noiva et al. 1991a). PDI has also been characterized as a retinal protein r-cognin (Krishna-Rao & Hausman 1993), a dehydroascorbate reductase (Wells et al. 1990) and as a major ER phosphoprotein undergoing ATP-dependent autophosphorylation stimulated by the presence of denatured polypeptides (Guthapfel et al. 1996).

In addition, calcium binding by PDI has been shown (Macer & Koch 1988). As with many other luminal proteins of the ER, the binding is weak, but high in capacity, as 19 Ca²⁺ ions per PDI molecule has been measured (Lebeche et al. 1994). The physiological significance of calcium binding by PDI is poorly understood.

A recent finding is the ability of PDI to covalently cross-link proteins, via transglutaminase-like activity. Transglutaminase-catalyzed cross-linking of glutamine and lysine through the formation of an isopeptide bond is important in creating tissue stability. This finding was made during the cloning of filarial nematode Dirofilaria immitis transglutaminase, which was found to contain some homology to PDI. Upon comparison, bovine PDI was shown to have transglutaminase activity (Chandrashekar et al. 1998).

Under certain conditions PDI has been reported to facilitate misfolding and aggregation of substrates; this has been termed anti-chaperone activity (Puig & Gilbert 1994). Denatured, reduced lysozyme aggregates strongly in the presence of low concentrations of PDI (Sideraki & Gilbert 2000). This behaviour is especially pronounced with aggregation-prone substrates. Whether this activity has function or significance in vivo is not known, but it has been suggested that the anti-chaperone activity is due to multivalent binding of partially aggregated substrates to PDI (Primm et al. 1996). Furthermore, since PDI is present in the lumen of the ER at millimolar concentrations, folding substrates in vivo are unlikely to experience the conditions under which PDI has significant anti-chaperone activity.

PDI located on the cell-surface (Hotchkiss et al. 1998) has been reported to act on the transfer of nitric oxide from extracellular protein into the cytosol by a mechanism which is not known (Ramachandran et al. 2001). Platelet-surface PDI may have a role in platelet activation via its oxido-reductase activity (Essex et al. 2001). A recent study reports that PDI binds estradiol and tri-iodotyronine through a distinct hormone-binding site and thus possibly acts as a hormone reservoir or mediates hormone-receptor binding (Primm & Gilbert 2001).

2.2.3.1. ERp57

After PDI, Erp57 (ER-60, ERp60, ERp61, GRP58, P58, HIP-70, Q2) is probably the best characterized protein of the PDI family. It was initially incorrectly identified as a phospholipase, when the enzyme was first cloned using screening of rat cDNA library with a guinea pig uterus phospholipase antibody (Bennett et al. 1988). The mature human polypeptide consists of 481 amino acids, excluding the 24 amino acid N-terminal signal sequence (Bourdi et al. 1995, Koivunen et al. 1996). Subsequent research using sequence comparison and recombinant protein production has shown that ERp57 has a similar domain organization to that of PDI (Ferrari & Söling 1999, Alanen et al. 2003a) and has two thioredoxin-like domains corresponding to domains a and a’ of PDI (Freedman et al. 1994) with -CGHC- active-sites in similar positions to those of PDI. ERp57 has an ER retrieval sequence variant -QEDL. The overall sequence identity between PDI and ERp57 is 29% and similarity 56% (Koivunen et al. 1996). ERp57 has been shown to have thiol-dependent reductase activity similar to that of PDI (Srivastava et al. 1993, Bourdi et al. 1995, Hirano et al. 1995). However, ERp57 does not substitute PDI as the β subunit of collagen prolyl 4-hydroxylase structurally or functionally (Koivunen et al. 1996).

ERp57 has been shown to have a distict physiological function in the folding of glycoproteins. Cross-linking studies identified an interaction between glycoproteins, ERp57 and either calnexin or calreticulin (Oliver et al. 1997). Calnexin and calreticulin are chaperones that interact specifically with newly synthesised glycoproteins taking part in both the folding process and the quality control. They recognize monoglucosylated proteins and bind N-glycan moieties of the form GlcNAc2Man9Glc1 (Rodan et al. 1996) i.e. they are lectins. After this recognition, the interaction of calnexin and calreticulin with ERp57 then allows for thiol-dependent protein-protein interaction of ERp57 with substrate (Helenius et al. 1997).

Immunohistochemical studies of murine tissues have shown a wide distribution of ERp57 in a variety of cell types (Kozaki et al. 1994). Erp57 is particularly abundant in plasma cells, mucus-secreting cells in various tissues, neuroendocrine cells including neurons and follicular epithelia of the thyroid gland (Iida et al. 1996). Furthermore, a high correlation has been observed between intracellular amounts of ERp57 and immunoglobulin production by hybridoma cells (Kozaki et al. 1994). This was confirmed by the finding of ERp57 in the assembly of MHC class I together with calnexin in complexes with MHC class I heavy chains (Lindquist et al. 1998).

Recently, a novel function for ERp57 was identified in Caenorhabditis elegans. The C.elegans homologue of ERp57 was shown to have in vitro transglutaminase activity, in addition to the PDI-like activity in the refolding of denatured RNase (Eschenlauer & Page 2003). Transglutaminase catalyzes the cross-linking of glutamine and lysine residues of cellular proteins, creating tissue stability for example in the cuticle of nematodes. In addition, nuclear localization and an interaction with DNA has been reported for ERp57 (Coppari et al. 2002).

Structural work on ERp57 is not as advanced as for PDI, but recently the a’ domain of ERp57 was shown to have the same overall fold as the a domain of PDI (Silvennoinen et al. 2001); the one significant difference being that the α3 is apparently missing in ERp57.

2.2.3.2. Other members of the PDI protein family

In addition to PDI and ERp57 the PDI family in higher eukaryotes includes PDIp, ERp72, P5, PDIr, ERp44, ERdj5/JPDI, ERp18 and ERp29/28.

PDIp is a pancreas specific isoform of the PDI family (Desilva et al. 1996), which is located in the acinar cells but not in the islet cells of the pancreas (Klappa et al. 1998b). It has a high sequence similarity to PDI (46% identity and 66% similarity (Desilva et al. 1996)). PDIp possesses two thioredoxin-like active sites -CGHC- and the unusual -CTHC- motif and an ER retrieval sequence variant -KEEL (Desilva et al. 1996). The domain organization of PDIp is the same as that of PDI, with two thioredoxin-active domains and two thioredoxin-inactive domains, abb’a’ (Ferrari & Söling 1999, Alanen et al. 2003a). The highly acidic C-terminal stretch of PDI is modified in PDIp, with a stretch of 32 predominantly hydrophobic residues but including a short predominantly acidic stretch. In addition, PDIp has a 21 amino acid N-terminal acidic stretch (Klappa et al. 2001). PDIp has been shown to have both reduction and oxidation activities (Desilva et al. 1996). It has also been shown to bind protein substrates and is therefore probably participating in the maturation processes of newly translocated proteins (Volkmer et al. 1997, Elliot et al. 1998, Klappa et al. 1998b). Recently the binding specificity of PDIp was determined by a cross-linking-based method. PDIp was found to recognize tyrosine or tryptophan residues within a peptide when there was no adjacent negatively charged residue (Ruddock et al. 2000). Substrate binding by PDIp is inhibited by certain oestrogens and xeno-estrogens (Klappa et al. 1998b), though the physiological significance of this is unclear.

ERp72 differs from archetypal PDI domain structure by having three active thioredoxin-like domains, with -CGHC- active site sequences, with the domain distribution of -a⁰abb’a’- and having a 42 amino acid acidic stretch of amino acids at the N-terminus (Ferrari & Söling 1999). ERp72 has both redox and disulfide-isomerase activity (VanNguyen et al. 1993, Rupp et al. 1994, Kramer et al. 2001) but has no observable peptide-binding ability (Kramer et al. 2001).

P5 was identified in 1992 (Chaudhuri et al. 1992) but little is still known about this protein. The domain structure includes two redox-active domains with -CGHC- active site sequences and in addition a C-terminal domain which has been reported to have similarity to b domain (Ferrari & Söling 1999), but which in a recent study was found to have no significant similarity to either b or b’ domains of PDI and to have a significantly different CD spectra from other purified domains indicating that it may not have a thioredoxin-like fold (Alanen et al. 2003a). P5 has been shown to have diminished PDI-like activity in the refolding of reduced RNase (Rupp et al. 1994, Kramer et al. 2001, Kikuchi et al. 2002).

Another already long known but virtually unstudied member of the PDI family is PDIr (Hayano & Kikuchi, 1995). PDIr has the domain architecture ba⁰aa’ and each thioredoxin-active domain has a different variant of active site sequence: -CSMC-, -CGHC- and -CPHC- (Ferrari & Söling 1999), with only one having the PDI consensus active site sequence.

Other recently reported new members of PDI family include as least ERp29 found in rat tissues (Demmer et al. 1997, Mkrtchian et al. 1998, Hubbard et al. 2000) and its human homologue ERp28 (Ferrari et al. 1998), ERp44 (Anelli et al. 2002), human protein ERdj5 (Cunnea et al. 2003) and its murine homologue JPDI (Hosoda et al. 2003), and smallest so far, ERp18 (Alanen et al. 2003b).

ERp28/29 is unique among the other members of the PDI family in that it has no -CGHC- active site in its sequence. ERp28/29 was first identified in rat (Demmer et al. 1997, Mkrtchian et al. 1998) with molecular weight of 29 kDa, and soon after in human (Ferrari et al. 1998) with the molecular weight of 28 kDa, both having a C-terminal retrieval sequence variant -KEEL. In rat hepatoma cells, ERp29 expression was found to be strongly induced following stress, indicating that the function of the protein could be related to the stress response in the ER (Mkrtchian et al. 1998). In another study the protein was found to be up-regulated during synthesis of secretory proteins and has highest expression in secretory tissue, suggesting a foldase-like role for ERp29/28 (Hubbard et al. 2000). The structure of ERp28 has been solved by NMR (Liepinsh et al. 2001). The protein comprises of two domains; the N-terminal domain having a thioredoxin fold similar to that of domain a of PDI and the C-terminal domain having a novel, all-helical structure. The protein has a tendency to dimerize both in vivo and in vitro, and the dimerization was found to occur through the N-terminal domain of ERp28 (Liepinsh et al. 2001).

ERp44 was identified as a protein associated in the human Ero1-mediated oxidative protein folding pathway (Anelli et al. 2002). The protein is approximately 44 kDa in size and has one thioredoxin-like domain with an unusual -CRFS- motif and a C-terminal ER retrieval sequence variant -RDEL. ERp44 is also found to be up-regulated during stress and forms mixed disulfides with Ero1 (Anelli et al. 2002). Otherwise, the physiological role of ERp44 is still unknown.

Two recent, independent reports announced the identification of a novel member of PDI family, ERjd5/JPDI. ERjd5 was cloned from human (Cunnea et al. 2003) and JPDI from mouse (Hosoda et al. 2003). ERjd5/JPDI has an N-terminal DnaJ-like domain followed by four thioredoxin-like sequences with a -CXYC- motif variants -CSHC-, -CPPC-, -CHPC- and CGPC-. In addition, it has a C-terminal -KDEL ER retrieval sequence (Cunnea et al. 2003, Hosada et al. 2003). DnaJs work in co-operation with molecular chaperones of the class Hsp70, such as BiP. The function of ERjd5/JPDI could therefore be related to the protein folding and quality control mechanisms of the ER.

Another very recently identified protein of the PDI family is ERp18 (Alanen et al. 2003b). ERp18 has an N-terminal signal sequence and a C-terminal ER retrieval variant -EDEL. The protein contains one thioredoxin-like -CXYC- motif, -CGAC-. ERp18 was shown to have some peptide oxidase activity, although lesser than that of the isolated a domain of PDI. This indicates that ERp18 may have a role in disulfide bond formation in the ER (Alanen et al. 2003b).