Relationship between calnexin and BiP in suppressing aggregation and promoting refolding of protein and glycoprotein substrates.

Calnexin (CNX) is a membrane protein of the endoplasmic reticulum that has been defined primarily as a lectin, yet is capable of functioning as a molecular chaperone with non-glycosylated proteins in vitro. Here, we assess the relative contributions of the oligosaccharide- and polypeptide-binding sites of CNX to its in vitro chaperone functions by comparing it with the Hsp70 chaperone of the endoplasmic reticulum, BiP. Both proteins were equally effective in preventing the aggregation of non-glycosylated citrate synthase, indicating that the polypeptide-binding site of CNX is capable of functioning at a level similar to that of Hsp70. However, when confronted with glycoprotein substrates, the lectin site of CNX provided a significant advantage over BiP in suppressing aggregation. CNX also cooperated with BiP and the J domain of Sec63p in the ATP-dependent refolding of glycoprotein and non-glycosylated substrates. The lectin site of CNX was essential for refolding of the glycoprotein. These findings reinforce the function of CNX as a bona fide chaperone and illustrate how its lectin site confers advantages relative to other chaperones when confronted with glycoprotein substrates.

Calnexin (CNX) 1 is a type I membrane protein of the endoplasmic reticulum (ER) that interacts transiently with newly synthesized Asn-linked glycoproteins. This preference for glycoproteins is due to the fact that CNX is a lectin that binds to monoglucosylated oligosaccharides of the form Glc 1 Man 5-9 -GlcNAc 2 (1,2). These oligosaccharides are formed transiently in the ER following attachment of the precursor oligosaccharide, Glc 3 Man 9 GlcNAc 2 , to nascent polypeptide chains. In vivo studies have shown that CNX binds primarily to folding intermediates rather than to native conformers (3)(4)(5), and it can enhance the efficiency of glycoprotein folding and subunit assembly (6 -9). Furthermore, CNX plays a role in the ER quality control system, preventing the export of incompletely folded or misfolded glycoproteins (10,11). Based on these observations, it is generally thought that CNX functions as a molecular chaperone for Asn-linked glycoproteins.
How CNX effects its molecular chaperone and quality control functions is the subject of considerable debate. In one view, CNX functions solely as a lectin, binding only to glycoproteins bearing Glc 1 Man 5-9 GlcNAc 2 oligosaccharides. Dissociation of CNX from the glycan is caused by the action of glucosidase II, which removes the terminal glucose residue of the oligosaccharide. Rebinding of CNX occurs upon re-addition of the glucose, an action carried out by UDP-Glc:glycoprotein glucosyltransferase (1,12). In this model, CNX does not function as a classical molecular chaperone that prevents the aggregation of folding intermediates by binding to exposed hydrophobic polypeptide segments. Instead, CNX is thought to retain glycoprotein folding intermediates and act as a scaffold for the binding of other ER chaperones and folding catalysts. For example, the thiol oxidoreductase ERp57 binds to CNX, and as shown in vitro, this interaction promotes disulfide formation/ isomerization in glycoproteins associated with CNX (13,14).
An alternative view is that CNX does indeed function as a molecular chaperone by virtue of a polypeptide-binding site that it possesses in addition to its lectin site (2,15). This dual binding model is supported by a variety of in vivo studies in which CNX has been shown to associate with proteins that lack the appropriate monoglucosylated oligosaccharides (11, 16 -23). Furthermore, we have demonstrated that a soluble form of CNX, consisting of its entire ER luminal domain (S-CNX), is capable of binding to unfolded but not native forms of nonglycosylated proteins in vitro and that this interaction potently suppresses the thermally induced aggregation of the unfolded conformers (15). The aggregation-suppressing function of S-CNX is enhanced in the presence of ATP and is attenuated when its lectin site is occupied by a synthetic monoglucosylated oligosaccharide. Consequently, we proposed that S-CNX possesses a polypeptide-binding site that recognizes some feature of non-native proteins and that it is differentially regulated by ATP and monoglucosylated oligosaccharides. In addition, S-CNX was found to collaborate with ATP-dependent factors in rabbit reticulocyte lysate (RRL) to promote the refolding of thermally inactivated citrate synthase (CS), a non-glycosylated protein. However, the active components of RRL that collaborate with S-CNX were not identified, and no corresponding ER factors were identified that might substitute for the active components of RRL. Collectively, these results strongly support the view that CNX functions as a molecular chaperone in vitro in addition to its lectin functions. That CNX probably engages in similar functions in vivo has recently been suggested by the finding that when the formation of monoglucosylated oligosaccharides is completely blocked, CNX is still capable of specific associations via polypeptide-based interactions with a large variety of newly synthesized proteins within the ER (24).
The in vitro studies on the molecular chaperone functions of CNX have raised a number of questions. How do these functions compare with those of other "classical" molecular chaperones such as those of the Hsp70 family? Does the existence of both lectin and polypeptide-binding sites confer an advantage to CNX over other molecular chaperones that bind solely through polypeptide-based interactions, e.g. either by increasing the avidity of CNX for glycoprotein substrates or by increasing the number of potential substrates that are able to bind to CNX? Finally, is it possible to identify other molecular chaperones of the ER that are capable of replacing RRL and collaborating with CNX in the refolding of denatured proteins?
To address these issues, we examined the in vitro chaperone properties of the ER Hsp70 protein, BiP, and compared them with those of S-CNX. Remarkably, despite the availability of purified preparations of BiP for the past decade, there has been a paucity of studies focused on its chaperone functions in vitro.
Only very recently has it been shown that BiP can act synergistically with protein disulfide isomerase in the oxidative folding of an antibody Fab fragment (25). However, its ability to suppress protein aggregation or to promote ATP-dependent folding in conjunction with Hsp40 co-chaperones has not been investigated. We found that BiP was indeed capable of suppressing the aggregation of a non-glycosylated protein when present in stoichiometric amounts. Furthermore, S-CNX was found to suppress the aggregation of the same protein as efficiently as BiP. Of particular interest, S-CNX proved to be considerably more effective than BiP in preventing the aggregation of glycoproteins that carry monoglucosylated oligosaccharides, suggesting an important role for the lectin site of S-CNX in this process. Finally, we found that factors present in RRL that cooperate with S-CNX in the refolding of glycoprotein or non-glycoprotein substrates could be replaced with purified ER components, viz. BiP and its Hsp40 co-chaperone, the J domain of Sec63p. These latter findings suggest that the CNX and BiP chaperone systems have the potential to collaborate in the folding of nascent polypeptides within the lumen of the ER.
Purification of Soluble CNX, BiP, the Sec63p J Domain, and IgY-A soluble form of canine CNX (S-CNX) consisting of its entire ER luminal domain was expressed in Escherichia coli as a fusion protein with glutathione S-transferase, excised from glutathione S-transferase with Factor Xa, and purified as described previously (15). Approximately 2 mg of purified S-CNX was obtained from 2 liters of bacterial culture.
DH5␣ cells containing the pQE9 plasmid (QIAGEN Inc.) encoding Saccharomyces cerevisiae BiP lacking its signal sequence but with a C-terminal His 6 tag were provided by Dr. Tom Rapoport (Harvard Medical School). Protein expression was induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside, followed by incubation at room temperature for 16 h. Cells were lysed by three passes through a French press at 1700 p.s.i. and centrifuged at 100,000 ϫ g for 1 h to remove cell debris. The supernatant fraction was applied to a nickel-nitrilotriacetic acidagarose column (QIAGEN Inc.), and BiP was eluted with 225 mM imidazole. Approximately 16 mg of purified BiP was obtained from 2 liters of bacterial culture.
The J domain of S. cerevisiae Sec63p was produced using the New England Biolabs intein system. ER2566 cells transformed with the TYB4 plasmid encoding amino acids 121-198 of Sec63p were obtained from Dr. Tom Rapoport. Induction of protein expression and cell lysis were performed as described for BiP. The J domain was purified by intein-chitin chromatography (26). Approximately 6 mg of purified J domain was obtained from 2 liters of culture.
IgY was purified essentially as described previously (27). Approximately 250 mg of IgY was obtained from five chicken eggs.
All proteins contained Ͻ5% contaminants as assessed by Coomassie Blue staining of SDS-polyacrylamide gels.
Endoglycosidase H Treatment-Jack bean ␣-mannosidase (Sigma) was desalted using a NAP-25 gel filtration column (Amersham Pharmacia Biotech) equilibrated with 10 mM Tris-HCl (pH 8.0). The lyophilized protein was then dissolved at 5 mg/ml in 20 mM sodium phosphate buffer (pH 5.5) containing 0.1% SDS. IgY was dissolved at 5 mg/ml in the same buffer with the addition of 40 mM dithiothreitol. ␣-Mannosidase and IgY were boiled for 5 and 20 min, respectively; diluted to 0.05% SDS; and subjected to digestion with 50 units/ml endo-␤-Nacetylglucosaminidase H (Endo H; New England Biolabs Inc.) according to the manufacturer's protocol. The proteins were precipitated with 95% ethanol and then evaporated to dryness. Aggregation Assays-IgY or Endo H-deglycosylated IgY was dissolved at 33.3 M in 0.1 M Tris-HCl (pH 8.0), 6 M GdnHCl, and 40 mM dithiothreitol. After denaturation for 2 h at room temperature, samples were diluted out of the denaturant to 0.67 M in 0.5 ml of 10 mM Tris-HCl (pH 7.2), 150 mM NaCl, and 5 mM CaCl 2 (TSC buffer) plus 3 mM ATP and various concentrations of S-CNX, BiP, or mouse IgG. The temperature was raised to 45°C, and aggregation was monitored over a period of 60 min by measuring light scattering at 360 nm in a Shimadzu 1610 spectrophotometer equipped with a temperature-controlled cell holder.
Desalted jack bean ␣-mannosidase (see above) or Endo H-treated ␣-mannosidase was dissolved at 21.7 M in 0.1 M Tris-HCl (pH 8.0) and 6 M GdnHCl. After denaturation for 60 min at room temperature, samples were diluted to 0.3 M in 0.5 ml of TSC buffer containing 3 mM ATP and various concentrations of S-CNX, BiP, or IgG. Protein aggregation was monitored over 5 min at room temperature.
The non-glycosylated protein CS (Roche Molecular Biochemicals) aggregates upon incubation at 45°C without a need for prior chemical denaturation. CS (1 M) was incubated in 0.5 ml of TSC buffer containing 3 mM ATP and various concentrations of S-CNX, BiP, or IgG. The samples were shifted to 45°C, and protein aggregation was monitored over 60 min.
Reactivation of Thermally Inactivated CS-CS (1 M) was inactivated in 50 l of 10 mM Tris (pH 7.2), 50 mM NaCl, and 5 mM CaCl 2 at 43°C in the presence of 1 M S-CNX or mouse IgG. Reactivation was initiated after 60 min by shifting the temperature to 23°C and diluting the sample with 2 volumes of 10 mM Tris (pH 7.2), 50 mM NaCl, 5 mM CaCl 2 , 37.5 mM KCl, and 15 mM MgCl 2 containing an ATP-regenerating system consisting of 1 mM ATP, 10 mM phosphocreatine, and 50 g/ml creatine phosphokinase. The reaction mixture was supplemented with 13.3% RRL (Promega), 1 M Grp94 (Stressgen Biotech Corp.), or various combinations of BiP, the Sec63p J domain, and 40 M G1M3 oligosaccharide. Aliquots were taken at various times and assayed for CS activity (28).
Reactivation of Chemically Denatured ␣-Mannosidase-␣-Mannosidase was denatured by dissolving desalted lyophilized samples in 6 M GdnHCl as described above. Reactivation was initiated by diluting the sample to a final concentration of 0.1 M in 10 mM Tris (pH 7.2), 0.15 M NaCl, 10 mM MgCl 2 , 5 mM CaCl 2 , and 50 M ZnCl 2 containing 1 mM ATP and an ATP-regenerating system. The reactivation solution was supplemented with 0.8 M S-CNX or mouse IgG and 10% RRL with or without 40 M G1M3 oligosaccharide and incubated at 4°C for 90 min. Aliquots were taken at various times and assayed for ␣-mannosidase activity (29).
Reactivation of Thermally Denatured ␣-Mannosidase-␣-Mannosidase (0.25 M) was thermally inactivated in 50 l of 10 mM Tris (pH 7.2), 50 mM NaCl, and 5 mM CaCl 2 for 90 min at 45°C in the presence of 0.25 M S-CNX or IgG. Reactivation was initiated by shifting the temperature to 23°C and adding 2 volumes of 10 mM Tris (pH 7.2), 50 mM NaCl, 5 mM CaCl 2 , 10 mM MgCl 2 , 1 mM ATP, 10 M ZnCl 2 , and an ATP-regenerating system. The reaction mixture was supplemented with either 13.3% RRL or various combinations of BiP, the Sec63p J domain, and 40 M G1M3 oligosaccharide. Aliquots were taken at various times and assayed for ␣-mannosidase activity (29).

S-CNX and BiP Are Equally Effective in Suppressing the
Aggregation of a Non-glycosylated Protein-We recently demonstrated that the soluble ER luminal domain of CNX (S-CNX) is capable of suppressing the aggregation of non-glycosylated proteins in vitro, suggesting the presence of a polypeptidebinding site that allows S-CNX to function as a bona fide molecular chaperone in addition to its functions as a lectin (15). However, the extent to which this polypeptide-based interaction compares with that of other molecular chaperones is not known. To address this issue, we compared the in vitro aggre-gation-suppressing function of canine S-CNX with that of another ER chaperone, the S. cerevisiae Hsp70 protein, BiP (Kar2p). Although the chaperone functions of BiP are well established in vivo, its ability to suppress the aggregation of non-native proteins in vitro has not been studied. Canine S-CNX was used since we have characterized it extensively in terms of its lectin (30) and in vitro chaperone (15) functions, and S. cerevisiae BiP was chosen because its co-chaperone Sec63p has been identified and thoroughly studied (reviewed in Ref. 31). Co-chaperones for BiP proteins from other species have not been unequivocally identified.
CS is a mitochondrial protein that exists as a homodimer of 50-kDa subunits. It has been used extensively as a substrate in aggregation and refolding assays with molecular chaperones of the Hsp90, Hsp60, and small heat shock protein families. As shown in Fig. 1, when CS was heated alone at 45°C, it formed large aggregates detectable by light scattering at 360 nm. Consistent with our previous studies (15), S-CNX completely suppressed the aggregation of CS at a 1:1 to 2:1 molar ratio (S-CNX monomer/CS dimer). Similar experiments performed with BiP demonstrated that BiP was also capable of suppressing the aggregation of CS in a concentration-dependent manner, with maximal suppression also occurring at a 1:1 to 2:1 molar ratio. In contrast, an equivalent amount of mouse IgG added as a control protein had no effect on the aggregation of CS. Additional control proteins (lysozyme and catalase) were also without effect (data not shown). Thus, BiP, like other Hsp70 proteins, is capable of suppressing the aggregation of a denatured protein substrate in vitro. Furthermore, these experiments demonstrate that the polypeptide-binding site of S-CNX is capable of functioning just as effectively as that of BiP in suppressing the aggregation of a non-glycosylated protein, providing additional support for the notion that S-CNX acts as a true molecular chaperone in vitro.
The Lectin Site of S-CNX Confers Enhanced Potency Relative to BiP in Suppressing the Aggregation of Glycoproteins Con-taining Glc 1 Man 7-9 GlcNAc 2 Oligosaccharides-The dual binding model for the interaction of CNX with glycoproteins predicts that engagement of both the lectin and polypeptidebinding sites of CNX should increase its binding avidity for monoglucosylated glycoproteins relative to binding through either site alone (15). This may confer an advantage over other molecular chaperones that bind to glycoproteins solely through polypeptide-based interactions. To test this hypothesis, the relative abilities of S-CNX and BiP to prevent the aggregation of denatured glycoproteins were compared. In these experiments, jack bean ␣-mannosidase and chicken IgY were used as substrates since they contain the monoglucosylated oligosaccharides recognized by the lectin site of S-CNX. ␣-Mannosidase is a heterodimer of 44-and 64-kDa subunits and contains at least one oligosaccharide of the form Glc 1 Man 9 GlcNAc 2 on its 64-kDa subunit (32). The oligosaccharides of the IgY heavy chain are more heterogeneous, with 27.1% being monoglucosylated (Glc 1 Man 7-9 GlcNAc 2 ) (33).
As shown in Fig. 2A, when ␣-mannosidase was denatured in 6 M GdnHCl and then diluted out of the denaturant, it aggregated rapidly at room temperature. S-CNX was capable of preventing the aggregation of ␣-mannosidase, with essentially complete suppression occurring at an ϳ2-fold molar excess of S-CNX over ␣-mannosidase dimer ( Fig. 2A, upper panel). By comparison, a 32-fold molar excess of BiP was needed to suppress the aggregation of ␣-mannosidase to the same extent ( Fig. 2A, lower panel). Thus, S-CNX is ϳ16 times more effective than BiP in suppressing the aggregation of this glycoprotein substrate.
A similar but less dramatic difference between S-CNX and BiP was observed with the IgY substrate. When reduced and chemically denatured IgY was diluted rapidly into solutions containing various concentrations of S-CNX or BiP and heated to 45°C, S-CNX was consistently 2-4-fold more potent than BiP in suppressing IgY aggregation (Fig. 2B). For example, whereas a stoichiometric concentration of S-CNX was capable of suppressing IgY aggregation completely, a 4-fold excess of BiP was required to achieve the same effect. Therefore, when presented with monoglucosylated glycoprotein substrates, S-CNX appears to have a significant advantage over a molecular chaperone that is restricted solely to polypeptide-based interactions.
To confirm that the lectin-carbohydrate interaction is responsible for the enhanced potency of S-CNX over BiP in preventing aggregation, we repeated the experiments under identical conditions, except that deglycosylated ␣-mannosidase and IgY were used instead of the glycosylated proteins. Glycans were removed by treatment with Endo H, and complete digestion was verified by the increased mobility of the treated proteins following SDS-polyacrylamide gel electrophoresis analysis and by their loss of reactivity when blotted with concanavalin A (data not shown). Deglycosylation of ␣-mannosidase profoundly impaired the ability of S-CNX to suppress aggregation (Fig. 3A, upper panel). A 32-fold molar excess of S-CNX suppressed the aggregation of deglycosylated ␣-mannosidase by ϳ50%, whereas a 1:2 molar ratio (S-CNX/␣-mannosidase) was able to sustain this level of aggregation suppression for the glycosylated form (compare Figs. 2A and 3A, upper  panels). The reduced efficacy of S-CNX was unlikely to be a consequence of deglycosylation creating a much more aggregation-prone substrate since BiP was still capable of completely suppressing the aggregation of deglycosylated ␣-mannosidase at a 32-fold excess, the same molar ratio as with glycosylated ␣-mannosidase (Fig. 3A, lower panel). When a similar experiment was performed with deglycosylated IgY (Fig. 3B), S-CNX again became less effective than BiP. A 4-fold molar excess of coproteins resides in its ability to bind to such substrates via lectin-oligosaccharide interactions in addition to its polypeptide-based associations.
The Aggregation-suppressing Functions of S-CNX and BiP Are Regulated Differently by ATP and Monoglucosylated Oligosaccharides-Previously, we demonstrated that the ability of S-CNX to suppress the aggregation of non-glycosylated proteins is enhanced by the addition of ATP and attenuated when its lectin site is engaged with the tetrasaccharide Glc␣1-3Man␣1-2Man␣1-2Man (G1M3) (15). The addition of other nucleotides or non-glucosylated oligosaccharides has no effect. However, no ATPase activity could be detected in highly purified preparations of S-CNX (15). In contrast, Hsp70 family members, stimulated by their Hsp40 co-chaperones, hydrolyze ATP in a cycle that regulates their binding to and release from unfolded proteins (34). Interestingly, different Hsp70 proteins seem to vary with respect to the effect that ATP has on their abilities to suppress protein aggregation (35)(36)(37)(38). Consequently, to characterize further the functional differences between S-CNX and BiP, we compared the effects of ATP and G1M3 on their abilities to suppress the thermal aggregation of citrate synthase.
Consistent with our previous findings, Fig. 4 demonstrates that the presence of 3.0 mM ATP significantly enhanced the ability of S-CNX to suppress the aggregation of CS, whereas the addition of 40 M G1M3 attenuated this function. In contrast, the presence of ATP had no significant effect on the ability of BiP to suppress CS aggregation. Even in the presence of the J domain of Sec63p, no effect on the aggregation of CS by BiP was observed. This result was not due to the inability of purified BiP to hydrolyze ATP since the BiP preparation possessed ATPase activity that was stimulated 2-4-fold by the addition of a 2-4-fold molar excess of purified J domain (data not shown). Not surprisingly, the G1M3 oligosaccharide did not affect aggregation suppression by BiP since this chaperone lacks a binding site for monoglucosylated oligosaccharides. Thus, BiP and S-CNX differ substantially in the manner by which ATP and monoglucosylated oligosaccharides affect their abilities to suppress the aggregation of unfolded proteins.
S-CNX, BiP, and the J Domain of Sec63p Cooperate in the Refolding of Citrate Synthase-Previous experiments in our laboratory demonstrated that S-CNX is capable of maintaining heat-inactivated CS in a folding-competent state and acts in conjunction with additional ATP-dependent chaperones in RRL to refold the denatured substrate (15). However, the individual factors in RRL participating in the refolding of CS were not identified, and it was unclear whether ER chaperones could replace RRL in cooperating with S-CNX. To address these issues, CS was incubated at 43°C for 60 min in the presence of equimolar concentrations of either S-CNX or mouse IgG. Under these conditions ϳ50% of the initial CS activity was lost. The temperature was then reduced to 23°C to allow renaturation to occur either alone or in the presence of various additional chaperone systems. When CS was thermally inactivated in the presence of IgG, only a minimal recovery of activity was observed, even when RRL was present in the reactivation mixture (Fig. 5, upper panel). This indicates that once denatured, chaperones present in RRL are unable to refold CS. In contrast, when S-CNX was present during inactivation, ϳ40% of the lost CS activity could be recovered in the presence of RRL and ATP, similar to our previous results (15). S-CNX present on its own during inactivation and reactivation failed to promote refolding, indicating that a cooperation with other chaperone systems is required. Furthermore, as described previously, the critical role of S-CNX is during the thermal inactivation stage since its inclusion along with RRL only during reactivation failed to enhance productive refolding (data not shown).
We then investigated whether various ER chaperone systems could replace RRL during the reactivation stage. The addition of canine Grp94, the Hsp90 family member within the ER, along with ATP was unable to substitute for RRL (Fig. 5, upper panel). However, when RRL was replaced with purified yeast BiP and the Sec63p J domain, 30% of the lost CS activity could be recovered (Fig. 5, middle panel). This degree of reactivation was dependent on both ATP and the J domain since no reactivation was observed in the absence of ATP, and reactivation reached only 18% when the J domain was omitted. The J domain alone or in combination with ATP had no effect (data not shown). The dependence of BiP function on both the Sec63p J domain and ATP is consistent with many other studies on refolding of substrates by Hsp70 and Hsp40 proteins, wherein the J domain of Hsp40 promotes binding and release cycles of Hsp70 through stimulation of its ATPase activity (34). Since the addition of the G1M3 oligosaccharide diminished the ability of S-CNX to prevent the aggregation of thermally denatured CS, we tested the effect of the presence of G1M3 during thermal inactivation of CS or during BiP/Sec63p J domain-mediated reactivation. The addition of G1M3 to S-CNX during thermal inactivation reduced the level of CS reactivation by ϳ50% (Fig. 5, lower panel), whereas there was no inhibitory effect observed when it was included in the reactivation mixture only (middle panel).
Collectively, these results indicate that S-CNX and the BiP/ Sec63p chaperone systems can cooperate in the refolding of a non-glycosylated protein. Furthermore, the role of S-CNX in this process is to maintain CS in a folding-competent state during thermal inactivation, whereas BiP and the J domain participate directly in refolding.

S-CNX Cooperates with Other Chaperones in Refolding Glycosylated ␣-Mannosidase, an Effect Dependent on Its Lectin
Site-Since the lectin site of S-CNX participated significantly in its ability to suppress the aggregation of monoglucosylated glycoproteins (Fig. 2, A and B), it was of considerable interest to determine if lectin-carbohydrate interactions are also involved when S-CNX is included during refolding of a glycoprotein substrate. Initially, we examined whether S-CNX was capable of participating in the reactivation of chemically denatured ␣-mannosidase in conjunction with RRL. ␣-Mannosidase was denatured with 6 M GdnHCl and then diluted out of the denaturant into a reactivation mixture containing S-CNX or mouse IgG, RRL, and ATP. Fig. 6 (upper panel) shows that Ͼ30% of the ␣-mannosidase activity was recovered upon incubation with a mixture containing S-CNX, RRL, and ATP. Removal of any one of these three components resulted in minimal recovery of activity. Therefore, neither S-CNX nor RRL was capable of reactivating ␣-mannosidase on its own; rather cooperation between the two chaperone systems was required. To determine whether the participation of the lectin site of S-CNX was involved in its interaction with the glycoprotein, we repeated the experiment with ␣-mannosidase that had been deglycosylated with Endo H. The complete system of S-CNX, RRL, and ATP failed to recover any activity, consistent with a role for the lectin site of S-CNX in this process. However, this experiment was limited by the possibility that deglycosylated ␣-mannosidase may simply be incapable of refolding to a native state. As an alternative approach, glycosylated ␣-mannosidase was chemically denatured, and the G1M3 oligosaccharide was added to the reactivation mixture. Although G1M3 partially attenuates the polypeptide-based binding of S-CNX as observed with CS (see Fig. 4, upper panel; and Fig. 5, lower panel), it will also compete for lectin-carbohydrate interactions when a monoglucosylated substrate such as ␣-mannosidase is used. Indeed, the presence of G1M3 completely abolished the reactivation of ␣-mannosidase (Fig. 6, upper panel), in contrast to the partial inhibition of CS refolding that was observed when G1M3 attenuated polypeptide binding alone (Fig. 5, lower panel). This result is consistent with the view that the lectin site of S-CNX is crucial for its ability to participate in the refolding of a glycoprotein substrate.
In the preceding experiment, we were unable to replace RRL with BiP and the Sec63p J domain (data not shown). Consequently, to investigate the involvement of the lectin site of S-CNX in glycoprotein folding using entirely defined components, we subjected ␣-mannosidase to a thermal inactivationreactivation protocol similar to that employed for CS reactivation (Fig. 5). ␣-Mannosidase was heated at 45°C for 90 min in the presence of S-CNX or mouse IgG, after which time ϳ50% of its activity remained. Reactivation was initiated by shifting the temperature to 23°C and adding BiP, the Sec63p J domain, and ATP. As shown in Fig. 6 (lower panel), ϳ30% of the ␣-mannosidase activity that was lost during heating was recovered when S-CNX was present during the inactivation stage. As was the case for CS, only minimal recovery of activity was observed when IgG replaced S-CNX during the inactivation (Fig. 6, lower  panel) or when S-CNX was present on its own during inactivation and reactivation (data not shown), indicating that cooperation is required between the S-CNX and BiP/Sec63p chaper- one systems. The addition of the G1M3 tetrasaccharide during thermal inactivation in the presence of S-CNX resulted in a complete block of ␣-mannosidase reactivation. This result confirms the participation of the lectin site of S-CNX in reactivation of a denatured glycoprotein substrate by defined ER chaperone systems. DISCUSSION Although the functions of CNX as a lectin and as a factor involved in protein folding and quality control in the ER are widely accepted, its proposed role as a classical molecular chaperone that associates with non-native substrates through polypeptide-based interactions is controversial (15, 39 -41). In this study, we compared the ability of S-CNX to suppress the aggregation and to promote the folding of non-native substrates in vitro with that of a bona fide molecular chaperone of the ER, the Hsp70 family member BiP. BiP has been shown to recognize non-native folding intermediates through a binding site that can accommodate 9-mer peptides with large hydrophobic residues at alternate positions (42). However, despite intensive analysis of the functions of other Hsp70 family members, the ability of BiP to suppress protein aggregation in vitro has not been examined previously. Our experiments show that, like other Hsp70 proteins, BiP potently suppresses the thermally induced aggregation of a non-glycosylated substrate (citrate synthase) when present in stoichiometric amounts. Most importantly, S-CNX was found to be just as effective as BiP in preventing the aggregation of citrate synthase. This result supports the notion that S-CNX functions as a true molecular chaperone since its ability to bind to polypeptide segments of a non-native protein can be just as potent as that of a member of the well characterized Hsp70 chaperone family.
We also compared the regulatory effects of ATP on the ag-gregation-suppressing functions of BiP and S-CNX. Hsp70 proteins cycle between an ATP-bound state that has low affinity for unfolded polypeptide and a high affinity ADP-bound state. Conversion between these states is regulated by the intrinsic ATPase of Hsp70 and by co-chaperones that accelerate the ATPase activity (Hsp40/DnaJ) or facilitate ADP-to-ATP exchange (GrpE/Bag1) (43). However, conflicting data have been reported on the effects of ATP on Hsp70 proteins during in vitro aggregation experiments. Whereas the addition of ATP reduced the ability of E. coli DnaK or mammalian Hsc70 to suppress the aggregation of certain denatured substrates (36,38,44), either no effect or the opposite effect was observed in other studies using either Hsc70 or S. cerevisiae Ssa2p (35,37). More consistent results were observed upon addition of both ATP and the Hsp40/DnaJ co-chaperone, wherein an increase in aggregation suppression was observed (35)(36)(37). This was most likely a result of enhanced ATPase activity shifting the equilibrium of the various Hsp70 proteins to the higher affinity ADP state. In the present case, ATP did not significantly alter the aggregation-suppressing ability of BiP for either thermally denatured citrate synthase or chemically denatured ␣-mannosidase (data not shown). Furthermore, no change in aggregation suppression was observed upon addition of both ATP and the J domain of Sec63p, despite the fact that the J domain was capable of stimulating the ATPase activity of BiP. This atypical behavior of BiP was also reflected in recent experiments by others. It was demonstrated that the Sec63p J domain did not stimulate the binding of BiP to a peptide substrate in solution in the presence or absence of ATP (26). It was postulated that the J domain-activated state of BiP might be too short-lived to be detectable in solution. If this is the case, it would provide an explanation for the apparent lack of effect of the J domain on BiP in our aggregation suppression assays. It should be noted, however, that in our experiments focused on refolding of thermally denatured substrates, both ATP and the Sec63p J domain were essential for optimal BiP function, as expected for an Hsp70 protein.
In contrast to BiP, ATP enhanced the aggregation-suppressing ability of S-CNX for both CS and ␣-mannosidase. ATP (but not ADP) is known to cause a conformational change in CNX, potentially increasing the affinity of its peptide-binding site for substrates (15). Therefore, S-CNX and BiP differ fundamentally in the regulation of their polypeptide binding by adenosine nucleotides.
Given that CNX has a demonstrated capacity to function as a molecular chaperone in vitro, the question arises as to its role within the ER relative to other molecular chaperones. Are its functions simply redundant compared with those of other polypeptide-binding chaperones such as BiP or Grp94, or does the dual oligosaccharide-and polypeptide-binding properties of CNX provide it with unique capabilities? Support for a unique role was provided by experiments in which S-CNX and BiP were equally potent in suppressing the aggregation of a nonglycosylated substrate, but S-CNX performed much more efficiently than BiP when presented with glycoprotein substrates possessing monoglucosylated oligosaccharides. This advantage could be attributed to lectin-oligosaccharide interactions because removal of the monoglucosylated glycans from either the ␣-mannosidase or IgY substrate resulted in a substantial attenuation of the aggregation-suppressing ability of CNX, while leaving the ability of BiP to suppress aggregation largely unaffected. A similar contribution from the lectin site of CNX was observed in the refolding of chemically denatured ␣-mannosidase. Molecular chaperones present in RRL were unable to refold this monoglucosylated substrate on their own. However, the addition of S-CNX permitted a substantial degree of refolding, an effect that was highly dependent on the lectin site of S-CNX. Collectively, these findings suggest that the functions of CNX are not merely redundant with those of other ER chaperones, but rather the unique lectin site of CNX provides it with a substantial binding advantage for glycoprotein substrates compared with other chaperones that are restricted solely to polypeptide-based interactions.
It is noteworthy that, upon deglycosylation of IgY and ␣-mannosidase, the advantage of S-CNX over BiP was not only lost, but that S-CNX became ϳ3-4-fold less effective than BiP in suppressing the aggregation of these substrates. Given their comparable potencies with the non-glycosylated substrate (citrate synthase), one might have expected that CNX would be as effective as BiP with the deglycosylated proteins. Presumably this reflects inherent differences in the nature of each chaperone's polypeptide-binding site as well as the types of non-native determinants that are exposed in the various denatured substrates. Indeed, we observed significant variability in the ability of either BiP or S-CNX to suppress aggregation when presented with different non-glycosylated substrates. For example, to suppress the aggregation of the CS, non-glycosylated IgY, and non-glycosylated ␣-mannosidase substrates by 50%, BiP/substrate ratios of 1:2, ϳ2:1, and ϳ12:1 were required, respectively. Comparable values for S-CNX were ϳ1:3, ϳ6:1, and Ͼ32:1, respectively. From this comparison, it is apparent that the aggregation of some substrates is inherently more difficult to suppress than that of others and that the polypeptide-binding site of BiP seems to be more versatile and able to function more effectively than that of S-CNX when presented with diverse non-glycosylated substrates. The latter notion may help to explain the many conflicting reports in the literature of CNX being capable of stable binding to some substrates that lack monoglucosylated oligosaccharides, but not others (reviewed in Ref. 24).
A distinguishing feature of many molecular chaperones is their ability to cooperate with other chaperone systems in the folding of non-native proteins. Examples include the sequential interaction of nascent polypeptides with the Hsp70 and Hsp60 chaperone systems (43) and the involvement of both Hsp90 and Hsp70 in the conformational regulation of steroid hormone receptors (45). In this study, we show that CNX is also capable of cooperating with other chaperones to effect folding of nonglycosylated or glycosylated substrates. This was observed in the refolding of chemically denatured ␣-mannosidase, wherein the presence of S-CNX along with ATP-dependent chaperones of RRL was essential for recovery of enzyme activity. Furthermore, using completely defined ER components, S-CNX was found to cooperate with BiP in conjunction with the Sec63p J domain and ATP in the refolding of either non-glycosylated CS or monoglucosylated ␣-mannosidase. S-CNX was required to maintain substrates in a folding-competent state during thermal denaturation, whereas BiP and the Sec63p J domain were responsible for promoting subsequent ATP-dependent folding. This effect of BiP could not be replaced by Grp94, the Hsp90 homolog within the ER. As was observed in the aggregation suppression assays, the lectin site of S-CNX was crucial for refolding of the ␣-mannosidase substrate, which again underscores the flexibility of its dual mode of binding compared with exclusively polypeptide-binding chaperones. However, it is clear from the aggregation suppression assays that ␣-mannosidase is a particularly challenging substrate for the polypeptide-binding sites of either BiP or S-CNX. It will be of interest to examine the folding of other monoglucosylated proteins that interact more avidly with the polypeptide-binding site of S-CNX such as IgY and soybean agglutinin (15). In such cases, one might expect to observe a less strict dependence on the S-CNX lectin site for refolding.
The demonstration that the CNX and BiP chaperone systems have the capacity to cooperate in the folding of denatured substrates in vitro suggests the potential for such collaboration in vivo. Indeed, several studies examining chaperone interactions with nascent glycoproteins in cultured cells have documented sequential or simultaneous associations of BiP and CNX with the human major histocompatibility complex class I molecule (3); thyroglobulin (19); acid phosphatase (17); and the viral glycoproteins vesicular stomatitis virus G (7), influenza hemagglutinin (5), and Semliki Forest virus E1 (46). Furthermore, chemical cross-linking of intact cells revealed the presence of a large network of associated ER chaperones, including CNX, BiP, calreticulin, and Grp94 (5). The in vitro experiments described in this study provide a starting point for understanding the roles these various chaperones play in the complex process of nascent glycoprotein folding and quality control within the ER. Our findings suggest that CNX and undoubtedly calreticulin as well occupy a unique niche within the ER folding and quality control machinery as a consequence of their dual mode of substrate recognition.