Proteins of the PDI family: Unpredicted non-ER locations and functions
Abstract
Protein disulfide isomerases (PDIs) constitute a family of structurally related enzymes which catalyze disulfide bonds formation, reduction, or isomerization of newly synthesized proteins in the lumen of the endoplasmic reticulum (ER). They act also as chaperones, and are, therefore, part of a quality-control system for the correct folding of the proteins in the same subcellular compartment. While their functions in the ER have been thoroughly studied, much less is known about their roles in non-ER locations, where, however, they have been shown to be involved in important biological processes. At least three proteins of this family from higher vertebrates have been found in unusual locations (i.e., the cell surface, the extracellular space, the cytosol, and the nucleus), reached through an export mechanism which has not yet been understood. In some cases their function in the non-ER location is clearly related to their redox properties, but in most cases their mechanism of action has still to be disclosed, although their propensity to associate with other proteins or even with DNA might be the main factor responsible for their activities. © 2002 Wiley-Liss, Inc.
The family of protein disulfide isomerases (PDIs) is composed of several well-characterized proteins formed by multiple domains, each presenting the typical fold of thioredoxin (reviewed by Freedman et al., 1994, and by Ferrari and Söling, 1999). Each protein contains two or more active sites similar to that of thioredoxin. Such proteins are present both in prokaryotes and eukaryotes, but only those from eukaryotes will be considered here, and specifically those found in mammals, as listed in Table 1. It can be noticed that the name PDI indicates not only the family but also that particular member of the family, which was the first to be characterized and also the one studied in more detail. Therefore, in the following discussion, PDI will indicate that particular protein. It should also be noticed that ERp57 has been indicated in the literature with a variety of synonyms.
Protein | Synonyms | Cellular localizations |
---|---|---|
Protein disulfide isomerase (PDI) | Endoplasmic reticulum, cell surface, nucleus, cytosol, secreted | |
ERp57 | ERp60a; ER-60; GRp58; ER-58; ERp58; ERp61; Q2 | Endoplasmic reticulum, nucleus, plasma membrane rafts, cell surface, cytosol, secreted |
ERp72 | CaBP2b | Endoplasmic reticulum, cell surface, secreted |
PDIR | Endoplasmic reticulum | |
PDIp | Endoplasmic reticulum | |
P5 | CaBP1b | Endoplasmic reticulum |
ERp44c | Endoplasmic reticulum | |
ERp29d | ERp31; ERp28 | Endoplasmic reticulum |
- a ERp60 can also designate calreticulin.
- b CaBP is an acronimus used also to indicate members of a family of calcium binding proteins.
- c This protein has only one thioredoxin-fold domain and has a CRFS sequence instead of the usual thioredoxin-like active site (Anelli et al., 2002).
- d This protein has only one thioredoxin-fold domain and no thioredoxin-like active site (Demmer et al., 1997).
All the proteins of the family are present in the endoplasmic reticulum (ER), in some cases in relatively high amounts, and are often considered to exert their activities exclusively in this location. Nowadays, however, the evidence from a vast literature supports the view that at least some members of the family also have a different subcellular distribution and have activities which may differ from those displayed in the lumen of the ER. The activities in the latter location have been extensively studied, and particularly those of classical PDI and of ERp57. They are involved in the proper folding and in the formation and reshuffling of the disulfide bridges of the proteins synthesized in the rough ER, imported in the lumen of this structure and destined to be secreted or incorporated in the cell membrane. Therefore their function, which is essential for cell viability, is that of chaperones and redox-catalysts, this latter activity being due to their thioredoxin-like active sites. Considering these functions, it is not surprising that the expression of these proteins is usually upregulated under stress conditions. They have a tendency to associate with peptides or other proteins, this property obviously being related to their activity as chaperones. This feature has been studied in some detail for PDI, and can explain the finding that PDI is part of at least two heteromeric enzymatic complexes found in the ER, i.e., the prolyl hydroxylase (Pihlajaniemi et al., 1987) and the microsomal triglyceride transfer protein (Wetterau et al., 1990). The other protein studied in detail, i.e., the ERp57, has been shown to bind to calnexin or calreticulin, which have lectin-like activity, and to intervene in the proper folding of glycoproteins (Elliott et al., 1997; Zapun et al., 1998; Oliver et al., 1999). By exploiting this activity, ERp57 participates in the assembly of the major histocompatibility complex I (MHC I) (Lindquist et al., 1998).
At least some of these proteins also have other properties, such as as proteolytic activity or the capability of binding calcium, ATP and other small ligands. Although the relevance of these properties for their biological roles is not clear, they can assume a great importance in the ER where these proteins are present in high concentrations. In non-ER locations, where they are present in much lower amounts, their functions might be quite different. Also the difference in redox potential between ER and non-ER compartments must be taken into account. The highly oxidative environment of ER directs the catalytic action of the PDI-related proteins mainly towards the formation of disulfide bonds of proteins, the oxidized form of PDI probably being regenerated by the Ero1 protein (Frand and Kaiser, 1999; Cabibbo et al., 2000). In non-ER locations, with their more reducing environments, the same PDI-related proteins catalyze more often the reduction of disulfides, the reduced form of the enzymes probably being regenerated by the NADPH-thioredoxin reductase system (Lundstrom and Holmgren, 1990).
STRUCTURAL AND FUNCTIONAL FEATURES
The proteins of the PDI-family, identified in mammals and other higher vertebrates, are characterized by a multidomain structure. Each domain has the typical thioredoxin fold, formed by this sequence of α-helices and β-strands: β–α–β–α–β–α–β–β–α (Kemmink et al., 1997; Ferrari et al., 1998). The structural aspects of this family have been extensively reviewed by Ferrari and Söling (1999). In two or three domains, but never in all, the particular thioredoxin-like active site sequence is present, formed by two cysteines separated by two amino acids. This is the redox-active center, with a mechanism of action similar to that of thioredoxin. Thioredoxin, a small-molecular mass protein (12 kDa), is present in many cellular compartments, and has a variety of functions, based mainly on its reducing properties (Holmgren, 1989). The PDI-family proteins differ from thioredoxin in two main respects: their redox potentials are usually higher and the presence of the domains devoid of the active site confers additional properties to them, the most important being the capacity to bind peptides or proteins, and consequently to display chaperone and isomerase activities. Although thioredoxin has some activity in reshuffling the disulfide bonds of ribonuclease, PDI is much more efficient (Hawkins et al., 1991). Furthermore, it should be kept in mind that also the simple reduction of disulfide bonds of a particular protein is not determined uniquely by the thermodynamic parameter, i.e., the redox potential of the reductant, but also, or rather mainly, by kinetic factors (Danon, 2002). Thus, although thioredoxin has an ubiquitous distribution in the cell, some reductive reactions occur more efficiently with members of the PDI-family, which in spite of the similarity of the active sites, have the advantage of a better enzyme–substrate interaction for the reduction of a particular disulfide bond. This is clearly shown, as discussed later, by some specific reductions catalyzed on the cell surface by PDI, and not by thioredoxin, which is also present in the same location.
The schematic structures of the three proteins of the family that have been detected in a non-ER location are shown in Figure 1. The a and a′ domains are those containing the thioredoxin-like sites, having the same sequences in all three proteins, i.e., WCGHCK. The mechanism of action of these sites is still open to discussion (reviewed by Ferrari and Söling, 1999), but the transient formation of a mixed disulfide between one of these cysteines and one from another protein appears highly probable. The difference of redox potential of these sites between PDI (E′0=−175/−190 mV; Lundstrom-Ljung et al., 1995) and thioredoxin (E′0=−270 mV; Joelson et al., 1990) has been explained by the slightly different amino acid sequence in the latter (WCGPCK (Huber-Wunderlich and Glockshuber, 1998)). However, additional factors must intervene to determine the value of the redox potential, since ERp57, which has the same active sites as PDI, appears to be significantly more reducing than PDI (Hirano et al., 1995; Ferraro et al., 1999).
The b and b′ domains do not have these active sites, but fold essentially in the same way (Kemmink et al., 1999). Highly negatively charged sequences exist both in ERp72 and PDI, in specific domains called c. These sequences should be those responsible for the low affinity, high capacity calcium-binding properties of PDI and ERp72 (Macer and Koch, 1988; Van Nguyen et al., 1989).
The c and the b′ domains of PDI have been shown to be responsible for the interaction with other proteins or peptides (Noiva et al., 1993; Klappa et al., 1998). No proof exists that the same property can be attributed to a particular domain of ERp57 or ERp72.
At the C-terminal the three proteins have an ER-retention sequence, although this is present in its canonical form (lys-asp-glu-leu) only in the PDI.
Activities other than the well-ascertained chaperone, isomerase, and redox-activities have also been attributed to these proteins. When ERp57 was first characterized, it was believed to be a phospholipase C (Bennett et al., 1988). This claim was later shown to be wrong (Mazzarella et al., 1994). Instead some degree of protease activity was found to be displayed by ERp57 (Urade et al., 1992), and by PDI and ERp72 as well (Urade et al., 1993; Urade et al., 1999). The biological relevance of this activity, which anyway is modest and in the case of ERp72 has been questioned (Ferrari and Söling, 1999), is still uncertain. However, ERp57 has been thoroughly investigated in this respect, and the involvement of a cysteine in its thioredoxin-like active sites has been demonstrated (Urade et al., 1997). ERp57 has been shown to be identical to one of the glucose-regulated proteins (GRPs), which are overexpressed in response to glucose starvation. A microsomal protein displaying carnitine-palmitoyl transferase activity has been described as homologous to ERp57 (Murthy and Pande, 1994).
An exclusive characteristic of ERp57 is a nuclear localization signal near the C-terminal end of the polypeptide chain. Although this sequence is not the most common bipartite one, it is very similar to the SV40 signal. This binds with high affinity to a specific site of the protein importin, which is responsible for the nuclear import process (Dingwall and Laskey, 1998).
PDI is also able to bind small ligands, like estradiol (Tsibris et al., 1989), 3,3′,5-triiodo-l-thyronine (Yamauchi et al., 1987), and ATP (Nigam et al., 1994; Quemeneur et al., 1994). The latter may lead to an autophosphorylation of the protein. PDI is also a substrate of sphingosine-dependent kinases (SDKs) (Megidish et al., 1999).
ERp72 can be phosphorylated on serine(s) by casein kinase II (Chen et al., 1996; Janson et al., 1997), while ERp57 has been shown to be phosphorylated on tyrosines 444, 453, and 466 by the Lyn kinase (Donella-Deana et al., 1996).
Some of these features may be relevant only in the ER, where these proteins are highly concentrated. Here PDI has been estimated to reach a concentration of 0.2 mM (Lyles and Gilbert, 1991). Thus, the estradiol- and 3,3′,5-triiodo-l-thyronine-binding properties of PDI have been proposed to contribute significantly to the storage of these hormones in the cell, given the considerable amount of the protein present in the ER (Primm and Gilbert, 2001). Clearly, a similar function cannot be performed in non-ER locations, where the members of the PDI-family are present in smaller, or even minute, amounts. Instead, the redox-properties and the protein-interacting capability can often explain, or are compatible with, the extra-ER locations and functions of these proteins, but still a lot has to be learned to establish a satisfactory structure/function relationship in these locations.
NON-ER LOCATIONS
The mechanism by which the proteins of the PDI family can escape from ER is unknown, and in fact the reports of their extra-ER locations have often been questioned. However, this mechanism should operate not only for these proteins, but also for other typical ER proteins, like calreticulin, which have been demonstrated as being not only secreted, but also located on the cell surface, in the cytoplasm and in the nucleus (Roderick et al., 1997; Holaska et al., 2001; Johnson et al., 2001).
Even secretion is not easily explained since calreticulin, and the proteins of the PDI-family as well, have the amino acid sequence acting as an ER retention signal, which is KDEL or similar sequences. Many tentative mechanisms have been proposed, as, for example, saturation of the ER-retention machinery, removal of the KDEL sequence or secretion from the cell by complex formation with other macromolecules (Johnson et al., 2001). A facultative translocation to ER has been proposed to explain the presence of proteins provided with the N-terminal signal sequence in other intracellular compartments (Belin et al., 1996). According to this proposal, the protein molecules being synthesized are partitioned between the cytosolic and the ER compartments. Alternatively, a retrotranslocation could be hypothesized (Johnson and van Waes, 1999).
In any case, since no safe explanation can at present be provided, it is obvious that claims of finding these proteins in sites other than ER must be accepted with caution, since in preparations of any subcellular compartment the danger of a contamination by proteins abundant in the ER is quite real. This is especially true for a possible nuclear localization of PDI, which has been found abundantly present in the ER region surrounding the nuclear membrane. However, in many cases, as described below, the non-ER location has been substantiated by experimental approaches capable of ruling out the possibility of contamination.
Another source of error may originate from the fact that proteins that are not members of the PDI-family display some degree of protein-disulfide isomerase activity. Among these are integrins (O'Neill et al., 2000), a human spliceosomal protein (Reuter et al., 1999), thyroglobulin (Klein et al., 2000), fibronectin (Langenbach and Sottile, 1999), and the elongation factor EF-Tu (Richarme, 1998). Therefore, claims of finding proteins of the PDI-family in unusual locations cannot be made only on the basis of their catalytic activity but should always be confirmed by the use of specific antibodies, amino acid sequencing, or other safe identification criteria.
SECRETION AND LOCATION ON CELL SURFACE
Proteins of the PDI-family have been found outside the cell, either as secreted proteins or as proteins located on the cell surface. In these cases their escape from ER can be ascertained with confidence, since no disruption of the cells is required for their detection, and the proteins can be identified by the use of antibodies or of specific, membrane impermeant ligands.
PDI has been found to be secreted from a variety of cell types, among which hepatocytes (Terada et al., 1995), pancreatic exocrine cells (Yoshimori et al., 1990), endothelial cells (Hotchkiss et al., 1998), activated platelets (Chen et al., 1992). Secretion of overexpressed ERp72 (Dorner et al., 1990) and ERp57 (Hirano et al., 1995) from cultured cells has also been observed. While the biological importance of these secreted proteins remains in most cases obscure, a function of PDI secreted by thyrocytes into the lumen of the thyroid follicles has been identified (Delom et al., 2001). It has been shown that the enzyme is involved, together with the BiP protein (the immunoglobulin heavy chain-binding protein) in the control of thyroglobulin folding and multimerization, probably by reducing the intermolecular disulfide bridges and thus limiting the extent of multimers formation.
PDI has been found also on the surface of a variety of cell types. It has been suggested that, after secretion, it binds on the cell surface through electrostatic interactions (Terada et al., 1995). In any case, an interaction with other surface-located proteins can be envisaged, since PDI, as other members of the family, are soluble proteins, which are not likely to become inserted into the membrane.
Some interesting roles have been assigned to the PDI located on the cell surface. In general, the thioredoxin sites of the protein appear to be involved in the reducing activity of the cell exterior (Mandel et al., 1993), so that disulfide bridges of macromolecules coming in contact with, or present on the cell surface, can be reduced or reshuffled. Thus, the level of cell surface thiols in lymphocytes and in fibrosarcoma cells was shown to be positively correlated to the amount of cell surface PDI (Lawrence et al., 1996; Jiang et al., 1999). Surprisingly, an inverse correlation was instead found in B cell chronic lymphocytic leukemia, in which a high increment in thiols of the cell membrane proteins was detected upon inhibition of surface-bound PDI with bacitracin or anti-PDI antibodies (Tager et al., 1997). Although the role of these external thiols and their relationships with PDI are still unclear, their importance is underlined by the fact that the amount of surface thiols is significantly increased in conditions like lymphocyte activation, leukemia, or HIV-infection (Lawrence et al., 1996; Tager et al., 1997). Moreover, this amount seems to be correlated with resistance to cytostatic drugs (Tager et al., 1997).
Interestingly, thiol groups of surface proteins are involved in cellular adhesion with various mechanisms. In some cases the direct intervention of PDI has been demonstrated, while more often the participation, even if indirect, of this protein can only be considered a possibility.
A clear-cut example of intervention of PDI in cell adhesion has been found in retina cells from chicken embryo, where a protein of the cell surface called R-cognin, responsible for adhesion (Hausman and Moscona, 1975), was later found to be identical to PDI (Krishna Rao and Hausman, 1993; Pariser et al., 2000). Moreover, its adhesion property required the integrity of the thioredoxin-like active site.
Compelling evidence has also been presented for a role of cell-surface PDI in the regulation of leukocytes adhesion (Bennett et al., 2000). The adhesion protein l-selectin is present normally on the cell membrane of leukocytes in a particular conformation, which makes it unaccessible to a proteolytic activity. The mantainance of this conformation requires the reducing activity of PDI. Upon activation of leukocytes a new conformation of l-selectin takes place, with a consequent cleavage of the extracellular moiety of the protein and a loss of the adhesive properties of the cell. Bennett et al. (2000) showed that this conformational change and l-selectin cleavage are brought up even in the absence of cell activation by inhibitors of PDI, i.e., bacitracin, phenylarsine oxide (which blocks vicinal thiols like those present in the thioredoxin-like active site), and anti-PDI antibodies.
An extensive literature has illustrated the presence and function of PDI on the surface of platelets. PDI has been detected in this location even in non-activated platelets (Essex et al., 1995). Platelets activation is accompanied by an increase of surface thiol groups and, in particular, by an increased amount of the reduced form of the PDI active sites (Burgess et al., 2000). Thiols on platelet surface have been clearly shown to be involved in the integrin-mediated platelet adhesion. In fact, the activated integrin conformation requires the reduction and/or reshuffling of disulfide bonds. The use of specific inhibitors demonstrated the requirement of PDI at least in some steps of the adhesion process (Essex and Li, 1999a; Lahav et al., 2000; Essex et al., 2001). However, some unexpected results point out the need for caution in the interpretation of the effects of inhibitors. Thus, while Fab anti-PDI fragments elicited an anti-aggregation response (Essex and Li, 1999a), a polyclonal antibody induced aggregation (Essex and Li, 1999b); this was shown to be an effect of the Fc moiety of the antibody. Likewise, bacitracin, usually considered a specific inhibitor of PDI and widely used as such, inhibited adhesion of lymphocytes through a PDI-unrelated mechanism (Mou et al., 1998). This might be explained by the fact that platelet integrin αIIbβ3 displays disulfide isomerase activity, that is inhibited by bacitracin (O'Neill et al., 2000). An additional problem with the use of bacitracin is represented by the frequent contamination of commercial preparations of this antibiotic with protease activity, which could be the real source of any inhibitory activity apparently displayed by bacitracin (Rogelj et al., 2000).
It has also been suggested that PDI participates in the modification of structure and function of thrombospondin, a large protein secreted by platelets and other cells, involved in cell–cell and cell–matrix interactions and capable of various specific interactions with other proteins. Speziale and Detwiler (1990) demonstrated that thrombospondin is subjected to disulfide isomerization, which has later been shown to be relevant for its multiple functions (Huang et al., 1997). In particular, isomerization can unmask the RGD sequence which is bound by integrins of platelets or endothelial cells (Hotchkiss et al., 1998). The presence of PDI on the surface of these cells would suggest that PDI is involved in vivo in the modulation of thrombospondin activity. The importance of the latter protein is such that search of a definite evidence for the involvement of PDI is highly desirable.
The reducing activity of cell-surface PDI has also been found to act on the thyroid stimulating hormone receptor of thyrocytes. Cleavage of disulfide bonds between the transmembrane β subunit of the receptor and the external α subunit releases the latter in the extracellular fluid or in the bloodstream (Couet et al., 1996).
Recent observations have linked the surface-bound PDI to the nitric oxide signaling pathways. It has been shown that extracellular S-nitrosylated proteins can transfer NO to the cytosolic compartment with a process exploiting the reducing activity of cell-surface PDI (Zai et al., 1999; Ramachandran et al., 2001). The NO formed by such reduction might enter into the membrane, where presumably it becomes oxidized by molecular oxygen, and hence able to nitrosylate cytosolic thiols. While some stages of this process are still hypothetical, the intervention of PDI has been well documented.
While the full biological importance of the protein disulfide isomerase activity on the cell surface must still be understood, some interesting examples of involvement of PDI in pathological events have been demonstrated. Thus, cell-surface PDI has been shown to be involved in the reduction of disulfide bridges of the diphtheria toxin, which in this way acquires the capability of entering the cell and displaying its cytotoxic activity (Mandel et al., 1993). Also the infectivity of viruses, such as the Sindbis virus (Abell and Brown, 1993) and HIV (Ryser et al., 1994), appears to depend on the reducing activity of the cell surface. In the case of HIV, a detailed study has ascertained that the reducing agent is, in fact, PDI. Inhibitors of this enzyme, i.e., the membrane impermeant 5,5′-dithiobis (2-nitrobenzoic acid), bacitracin, and an anti-PDI antibody do not allow entry of the virus in human lymphoid cells, although the binding to the CD4 receptor still occurs. The entry of HIV is strictly dependent on the reduction of disulfide bridges of the viral envelope glycoprotein gp120 (Ryser et al., 1994; Fenouillet et al., 2001).
PDI is not the only member of its family to be located on the cell surface. ERp72 has been found on the membrane of neutrophils, where apparently is involved in the process of priming of neutrophils, with a concomitant burst of oxidative reactions (Weisbart, 1992). ERp57 has been found among the proteins of sperm membrane (Bohring et al., 2001).
CYTOSOL
Reports of the presence of PDI-family proteins in the cytosol have mainly, but not exclusively, concerned ERp57. Soon after its characterization it was found that ERp57 is present in significant amounts not only in the ER but also in the cytosol (Lewis et al., 1986). Since this initial evidence was based on physical separation of subcellular compartments, it could be argued that a contamination from ER might have occurred. Recently, however, the cytosolic location was securely demonstrated by detecting ERp57 associated to proteins, which are well known to be cytosolic.
Thus activated STAT3, which is a member of the family of STAT proteins, responsible both for signal transduction and transcription regulation, has been shown to become complexed in the cytosol of Hep3B cells with other proteins among which is ERp57 (Ndubuisi et al., 1999). Activation of STAT3 requires tyrosine and serine phosphorylation, and the protein dimerizes through its own SH2 domain. It should be remembered that also ERp57 can be phosphorylated on tyrosine residues, but the relevance of this phosphorylation to its association with STAT3 has not yet been investigated. STAT proteins have also been detected in plasma membrane rafts, which are microdomains of the membrane involved in signal transduction (Sehgal et al., 2002) and ERp57 has been found associated with STAT proteins also in these rafts.
The role of these associations has not been understood, although it is tempting to hypothesize that it can be important for the nuclear import of STAT3, as discussed below.
Wyse et al. (2002) have recently demonstrated that grp58 (which is identical to ERp57, see Table 1) binds to the cytosolic moiety of the thiazide-sensitive sodium chloride cotransporter (NCC) present in the cells of the distal convoluted tubule of mammalian kidney. This interaction, which takes place on the COOH terminus of the cotransporter, increases the activity of NCC and plays a fundamental role in maintaining the NCC function. Immunofluorescence and immunoprecipitation experiments confirmed that this interaction occurs in vivo.
In both the afore-mentioned complexes the mechanism of action or of the complex formation of ERp57 is unknown, and no evidence has been presented on the possible role of its redox properties.
PDI has also been found in cytosol, and precisely in the cytosolic fraction of human and monkey liver. PDI has been proposed to act in concert with insulin-degrading enzime (IDE) in the cellular metabolism of insulin, catalyzing the cleavage of disulfide bonds with reduced glutathione as cosubstrate. Thus the degradation of insulin should occur by a sequential pathway involving initial scission of its disulfide bonds followed by proteolysis (Wroblewski et al., 1992).
MITOCHONDRIA
Protein disulfide isomerase activity has been detected in mitochondria, and a protein responsible for this activity has been purified (Rigobello et al., 2000). Although many features of the catalytic activity were the same as those of real PDI, the molecular weight of the purified protein was lower and there was no recognition by an anti-PDI antibody. No biochemical characterization was performed, so that the identification of this protein as a member of the PDI-family is doubtful. It should be considered that, as mentioned before, many proteins not belonging to the PDI-family display some degree of protein disulfide isomerase activity.
NUCLEI
A great number of observations have suggested that at least two proteins of the PDI-family, i.e., PDI and ERp57, are present in the nuclei. These suggestions are based on two types of evidence: either a direct detection of these proteins in the nuclear compartment or the demonstration in vitro that these proteins can influence processes that occur at the DNA level, i.e., the DNA binding of transcription factors or of structural nuclear proteins. It is quite obvious that both types of evidence must be received with caution. The PDI-like proteins have been shown to be present in low amounts in nuclei, so that the danger of contamination from the ER is high and real. On the other hand, in vitro evidence that these proteins are involved in processes at the DNA level is no proof that a similar situation exists in vivo. In some cases, however, as discussed below, a safe demonstration of the nuclear location has been provided.
ERp57 has been isolated from the internal nuclear matrix of nuclei purified from chicken liver cells (Altieri et al., 1993). PDI and ERp57 have been detected among the proteins from the nuclear matrix of human lymphocytes and monocytes (Gerner et al., 1999). ERp57 was found also among the nucleolar proteins from HeLa cells, as shown in the Swiss-2DPAGE database (Scherl et al., unpublished results); it should be noted that the nucleolus is usually copurified together with the nuclear matrix. Notwithstanding the purity of nuclear or subnuclear preparations, these results are not immune from criticism with regard to a possible contamination from the cytosolic or the ER compartments. It should be remembered that PDI, in particular, is concentrated on the ER compartment lining the nuclear periphery. Moreover, nuclei can easily be damaged and made more permeable during purification procedures.
The first safe demonstrations of nuclear locations of ERp57 were derived by immunohistochemical experiments on whole cells. ERp57 was found by this method in nuclei of rat spermatids and spermatozoa (Ohtani et al., 1993), in nuclei of chicken embryo fibroblasts (Altieri et al., 1993), and in nuclei of HeLa cells (Coppari et al., 2002).
The immunohistochemical detection, however, gave positive results only with certain cell types, so that a widespread presence of these proteins in the nucleus could not be ascertained. It is conceivable that the detection by immunofluorescence of ERp57 or PDI present in low amounts in a particular subcellular compartment is easily obscured by a much higher amount of the protein in the ER.
A safe criterium for the demonstration of the nuclear location would be one analogous to that used for the cytosolic location, i.e., finding the protein interacting with a specific nuclear component. The ideal one is DNA, and in fact by using DNA-protein cross-linking reagents on intact, viable cultured cells, ERp57 has been found complexed to DNA in all cell types examined. By this approach ERp57 has been demonstrated to be present in HeLa, HL-60, F9 (differentiated and non-differentiated), 3T3 (Coppari et al., 2002), and M14 melanoma cells (M. Eufemi, unpublished results). These results, obtained both with UV-radiation and cis-diamminedichloroplatinum as cross-linking agents, demonstrate not only the nuclear location of ERp57 but also its interaction with DNA. Recently, VanderWaal et al. (2002) showed that also PDI could be cross-linked to DNA in nuclei isolated from HeLa cells.
The demonstration of the presence in nuclei of ERp57 and PDI strongly supports the view that a variety of effects displayed in vitro by ERp57 or PDI on other protein–DNA interactions do indeed occur also in vivo.
Johnson et al. (1992) first demonstrated by gel shift experiments that ERp57 alters the formation of complexes between nuclear proteins and the regulatory domain of interferon-inducible genes. According to Clive and Greene (1996) PDI activates the binding of transcription factors NF-κB and AP-1 to DNA, although more recently the activation of AP-1 has been attributed to thioredoxin (Hirota et al., 1997). Markus and Benezra (1999) demonstrated that both PDI and ERp57 have a regulatory effect on the activity of the transcription factor E2A. PDI has been described as contributing to the DNA-binding of the estrogen receptor (Landel et al., 1995). It has also been shown that ERp57 by itself binds DNA with moderate affinity (Kd of the order of 10−7 M), with a preferential recognition towards A/T rich and MAR-like sequences (Ferraro et al., 1999). The latter are the matrix-associated-regions, that are at the base of the chromatin loops and are believed to provide an anchorage of the DNA to the nuclear matrix or nuclear scaffold (Berezney, 1991). An interesting feature of this ERp57-DNA binding is that it is strongly dependent on the redox state of the protein, since only the oxidized form is able to bind while the reduced one completely loses its affinity for DNA. Recently it has been found that the DNA binding takes place through the C-terminal moiety of the protein (Grillo et al., 2002). An isolated a′ domain binds efficiently DNA, but only when its thioredoxin-like active site is in the disulfide form.
Although it is not possible to assign with any certainty a function to the nuclear PDI and ERp57 on the basis of these results, a few hypotheses can be put forward.
It is a well-ascertained fact that many transcription factors depend for their DNA-binding activity on their redox-state, usually the reduced form being in this case the active one. Since, as mentioned before, a regulatory effect of PDI and/or ERp57 has been shown in vitro in a number of cases, the hypothesis of their intervention also in vivo in the nuclear compartment is legitimate. The presence in nuclei of the NADPH-thioredoxin reductase system (Rozell et al., 1988) is compatible with this hypothesis. In fact, this system has been shown to act not only on thioredoxin but also on PDI (Lundstrom and Holmgren, 1990), and thus could regenerate the reduced form of the PDIs.
It has also been suggested that ERp57 and PDI contribute to the anchoring of the DNA loops to the nuclear matrix. If this is the case, the capacity of ERp57 to bind DNA only in the oxidized form would provide a redox-modulated mechanism for the interaction between DNA and the nuclear matrix, which might have consequences for transcriptional regulation (Stein et al., 1995). This possibility received experimental support from a recent study by VanderWaal et al. (2002), who showed that in isolated nuclei from HeLa cells PDI could be cross-linked to DNA by cis-diamminedichloroplatinum, but that after treatment of nuclei with dithiothreitol the DNA was partially detached from the nuclear matrix, and PDI could no more become cross-linked. It could be argued that both ERp57 and PDI are present in nuclei in very small amounts, so that their possible contribution to the DNA-nuclear matrix anchoring is unlikely to be a major one. However, the implications of this redox-dependent anchorage of DNA are of such importance that this hypothesis deserves to be investigated further.
The chaperone and the isomerase activities of the two proteins should also be kept in mind for their function in the nuclear context. The nuclear matrix, where these proteins are located, is an ill-defined structure formed by protein complexes, in which disulfide bridges are believed to play an important, stabilizing role. The propensity of PDI and ERp57 to associate with other proteins should be relevant to their participation in such structure, and perhaps for the isomerization and/or the formation of disulfides. Moreover, Roti Roti et al. (1998) have shown that the proteins of the nuclear matrix are among the most labile and heat-sensitive of all cellular proteins. This might suggest that PDI and ERp57 act as chaperones in conditions of heat-shock or other protein-damaging events, also considering that their expression does indeed increase under various stress conditions. However, a strong objection to this role for nuclear PDI is that the chaperone activity of this protein is strongly dependent on its concentration. At the low concentration existing in the nucleus, PDI is expected to behave as an anti-chaperone (Puig and Gilbert, 1994).
Finally, considering that activated STAT3 in the cytosol forms a complex that contains, besides other components, also ERp57, it might be possible that the latter, by means of its nuclear localization signal, is responsible for the nuclear import of STAT3, which then acts as a transcription factor. However, up to now, no evidence of a ERp57–STAT3 complex in the nucleus has been provided.
It should be noticed that PDI, contrary to ERp57, lacks a known nuclear localization signal, so that the mechanism of its nuclear import remains obscure. It is possible that PDI enters into the nucleus by exploiting its well-known tendency of association with other proteins, that might act as transporters.
PERSPECTIVES
The non-ER localization of some proteins has been demonstrated convincingly in many cases, particularly regarding PDI, ERp57 and, more rarely, ERp72. In these locations, however, a satisfactory knowledge of their function or of their mechanism of action is often missing. Only the isomerase and disulfide-reducing activity of the PDI on the cell surface appears to be well correlated with the known properties of the protein. A different situation, instead, arises in the case of the complexes containing ERp57. These are the complexes formed in the cytosol with the NaCl-cotransporter and with the activated STAT3, and in the nucleus, with DNA (together with PDI). The situation is in a way analogous to that existing for the complexes formed in the ER by PDI with the α subunit of prolyl-hydroxylase or with the large subunit of microsomal triglyceride transfer protein: here too the PDI is an essential moiety of the fully functional enzyme but its exact role must still be disclosed. The interactions with DNA allow some reasonable hypotheses to be formulated, but these require confirmation. Undoubtedly the mechanism of action of these PDI-family proteins cannot be considered as depending only on their thioredoxin-like active sites, but is to a large extent dependent on the protein-binding and the recently found DNA-binding properties.
The presence in nuclei of ERp57 and PDI should encourage a further investigation of their regulatory activity on transcription factors.
An extensive search for the presence on the cell surface of all members of the PDI-family could also be very fruitful, taking into account the interesting activities already identified and attributed to PDI. If PDI, provided with the classical ER-retention signal, is found on the cell surface, the other members of the family, which have different, probably less efficient signals, are even more likely to be exported. Furthermore, the search for the function of the proteins in this location can be conveniently aided by the use of membrane-impermeant inhibitors or specific antibodies.