Contributions of the Lectin and Polypeptide Binding Sites of Calreticulin to Its Chaperone Functions in Vitro and in Cells *

Calreticulin is a lectin chaperone of the endoplasmic reticulum that interacts with newly synthesized glycoproteins by binding to Glc1Man9GlcNAc2 oligosaccharides as well as to the polypeptide chain. In vitro, the latter interaction potently suppresses the aggregation of various non-glycosylated proteins. Although the lectin-oligosaccharide association is well understood, the polypeptide-based interaction is more controversial because the binding site on calreticulin has not been identified, and its significance in the biogenesis of glycoproteins in cells remains unknown. In this study, we identified the polypeptide binding site responsible for the in vitro aggregation suppression function by mutating four candidate hydrophobic surface patches. Mutations in only one patch, P19K/I21E and Y22K/F84E, impaired the ability of calreticulin to suppress the thermally induced aggregation of non-glycosylated firefly luciferase. These mutants also failed to bind several hydrophobic peptides that act as substrate mimetics and compete in the luciferase aggregation suppression assay. To assess the relative contributions of the glycan-dependent and -independent interactions in living cells, we expressed lectin-deficient, polypeptide binding-deficient, and doubly deficient calreticulin constructs in calreticulin-negative cells and monitored the effects on the biogenesis of MHC class I molecules, the solubility of mutant forms of α1-antitrypsin, and interactions with newly synthesized glycoproteins. In all cases, we observed a profound impairment in calreticulin function when its lectin site was inactivated. Remarkably, inactivation of the polypeptide binding site had little impact. These findings indicate that the lectin-based mode of client interaction is the predominant contributor to the chaperone functions of calreticulin within the endoplasmic reticulum.

Membrane-bound calnexin (Cnx) 3 and its soluble paralog calreticulin (Crt) are glycoprotein-specific chaperones of the endoplasmic reticulum (ER). As components of the ER quality control system, they retain glycoprotein folding intermediates within this organelle and assist folding by preventing aggregation and by recruiting folding catalysts such as the thiol oxidoreductase ERp57 and peptidyl-prolyl isomerase cyclophilin B (for reviews, see Refs. 1 and 2). Both chaperones consist of a globular lectin domain and an extended arm or P domain (3)(4)(5) with the tip of the arm domain comprising the binding site for ERp57 (6) and cyclophilin B (7). The specificity for glycoproteins resides within their lectin domains, which recognize a monoglucosylated oligosaccharide-processing intermediate of composition Glc 1 Man 9 GlcNAc 2 (8 -10). Cycles of chaperone binding and release are regulated by the availability of the terminal glucose residue on this oligosaccharide with glucose removal catalyzed by glucosidase II and its readdition by UDPglucose:glycoprotein glucosyltransferase I (11). UDP-glucose: glycoprotein glucosyltransferase I acts as the folding sensor in the cycle, only reglucosylating non-native glycoprotein conformers (12).
The importance of lectin-based interactions in mediating associations of Cnx and Crt with folding glycoproteins has been well established. Many studies have examined this issue using glucosidase inhibitors or glucosidase-deficient cells to prevent the formation of the monoglucosylated oligosaccharide from its triglucosylated precursor. In most cases, glycoprotein interactions with the chaperones are lost or greatly diminished, and in several instances this is accompanied by reduced folding efficiency and altered export rates from the ER (Refs. 1, 11, and 13-19 and references therein). Further detailed studies using class I histocompatibility molecules have shown that interactions with Crt are dependent on a functional lectin site and the presence of monoglucosylated oligosaccharide on the substrate (20,21). The ability of these chaperones to suppress client protein aggregation appears to be due in part to their sequestration of glycoprotein folding intermediates between the globular lectin domain and the flexible extended arm domain. This is suggested on the basis of FRET experiments using probes within the two domains (22) as well as arm domain truncation mutants that exhibited reduced abilities to suppress aggregation in vitro (23,24).
There is also substantial evidence supporting the existence of polypeptide-based interactions between Cnx or Crt and their client proteins, interactions that may also contribute to the suppression of aggregation. This is based on the finding that binding of these chaperones to a number of clients is unaltered or only modestly diminished in glucosidase-deficient cells, cells treated with glucosidase inhibitors, or cells either expressing non-glycosylated clients or mutated to lack Asn-linked oligosaccharides (Refs. 2,13,19, and 25 and references therein). Furthermore, both Cnx and Crt bind non-glycosylated hydrophobic peptides in vitro (23,24,26) and are capable of suppressing the in vitro aggregation of non-glycosylated proteins under physiological conditions of the ER (24,27,28). This aggregation suppression function resides at a single site within the globular domain and can be blocked by the addition of hydrophobic peptides (23,24). Efforts have been ongoing to localize more precisely the site responsible for aggregation suppression with most emphasis placed on Crt. Several protein binding sites have been identified on Crt such as those specific for GABARAP (29) and thrombospondin (30), but interacting peptides derived from these proteins fail to compete in aggregation suppression assays, indicative of distinct binding sites (24). Additional binding sites have been proposed in proximity to the lectin site (22,31,32), but these either do not exhibit the expected specificity for hydrophobic substrates (22) or would be expected to be blocked upon addition of monoglucosylated oligosaccharide (31), a treatment that does not affect the binding of hydrophobic peptides to Crt (24).
In this study, we tested four candidate hydrophobic sites on the globular domain of Crt for their involvement in suppressing client aggregation in vitro. One site, when mutated, disrupted the ability of Crt to bind hydrophobic peptides and resulted in an impaired ability of Crt to suppress the in vitro aggregation of both non-glycosylated and monoglucosylated protein clients.
We then expressed this mutant in Crt-deficient cells and compared it with a previously described lectin-deficient mutant to assess the relative contribution of each site to chaperone function in living cells. Using three distinct assays, we show that, despite the clear involvement of the newly identified polypeptide binding site in vitro, it is lectin-based interactions that are critical for the functions of Crt with glycoprotein clients within the ER.

Identification of the Polypeptide Binding Site on Calreticulin That Mediates Aggregation Suppression of Non-glycosylated
Substrates-Previously, we showed that Crt suppresses the aggregation of non-glycosylated client proteins through a polypeptide binding site that resides within its globular lectin domain but at a site distinct from the site of oligosaccharide binding (24). To localize the polypeptide binding site more precisely, we turned to our crystal structure of the mouse Crt lectin domain (4) and identified four surface-exposed hydrophobic patches at locations removed from the lectin site that could potentially mediate aggregation suppression (Fig. 1A). Nonconservative double mutants were created at each site (Fig. 1A) and tested for their abilities to suppress the thermally induced aggregation of non-glycosylated firefly luciferase (FFL) at 37°C.
As shown in Fig. 1B, three of the mutants (sites 1, 3, and 4) actually enhanced the ability of Crt to suppress aggregation. This was unexpected and presumably reflects some enhanced exposure of hydrophobic residues as a consequence of the  [1][2][3][4] were identified on the basis of surface exposure (yellow), and within each site two solvent-accessible residues were chosen for mutagenesis to polar or charged amino acids (mutated residues depicted in green). For orientation purposes, bound Glc 1 Man 3 oligosaccharide is shown in red/blue, and residues at the beginning and end of the arm domain are in dark gray. B, aggregation suppression potency of Crt with mutations in sites 1, 3, and 4. Wild type or mutant Crt (0.5 M) was equilibrated at 37°C, then FFL was diluted into the reactions to a final concentration of 2 M, and light scattering was recorded every 60 s at 370 nm. mutations, although there was no change in melting temperature detected for any of these mutants (Table 1). In contrast, the P19K/I21E mutant (site 2) substantially impeded the chaperone function of Crt ( Fig. 2A). Consequently, we constructed an adjacent site 2 double mutant, Y22K/F84E, and obtained similar results ( Fig. 2A). Doubling the concentration of each mutant in the aggregation suppression assay resulted in potency similar to wild type Crt ( Fig. 2A), suggesting that the two mutants were about 50% impaired in their aggregation suppression function. The P19K/I21E mutant was properly folded as evidenced by an unchanged circular dichroism (CD) spectrum (Fig. 2B), melting temperature (Table 1), and retention of lectin function (stabilization by Glc 1 Man 3 tetrasaccharide; Table 1). The Y22K/F84E mutant exhibited some structural alterations by these criteria (Fig. 2B and Table 1). We attempted to generate several triple mutants in an effort to make variants that completely lacked aggregation suppression function, but these attempts were unsuccessful due to instability of the mutants.
To confirm that site 2 is indeed the site responsible for the aggregation suppression function of Crt, we examined site 2 mutants for their abilities to bind to the hydrophobic peptides KKAFAF and KHPYAYLAAAIAAEVAGTTALKLSK (herein referred to as KHP). We showed previously that these peptides compete in the FFL aggregation suppression assay by binding to a single site within the lectin domain of Crt (24). As shown in Fig. 2C, although wild type Crt bound to the KKAFAF and KHP peptides with dissociation constants of 1.2 and 1.7 M, respectively, neither of the site 2 mutants was capable of binding these peptides, thereby confirming that the correct site had been identified.
Both the Lectin and Polypeptide Binding Sites of Crt Participate in Suppressing the Aggregation of a Monoglucosylated Glycoprotein Substrate-To investigate the relationship between the polypeptide binding site and the lectin site of Crt in suppressing client protein aggregation, we compared the behavior of Crt constructs with mutations in either or both sites. We showed previously that the D317A mutation ablates the lectin function of Crt while maintaining native tertiary structure (33). This mutation could be combined with the polypeptide binding-deficient mutant (P19K/I21E/D317A) with no effect on its CD spectrum (Fig. 2B) and only a modest decrease in melting temperature ( Table 1). The triple mutant exhibited no stabilization upon addition of Glc 1 Man 3 oligosaccharide in keeping with its defect in lectin function (Table 1).
When the lectin-deficient, polypeptide binding-deficient, and combined mutants were tested for their abilities to sup-press FFL aggregation, those mutants with impaired polypeptide binding exhibited reduced aggregation suppression potency (Fig. 3A). In contrast, loss of lectin function in the D317A mutant had no impact on the ability of Crt to suppress the aggregation of this non-glycosylated substrate. To determine whether both lectin and polypeptide binding sites are utilized in suppressing glycoprotein aggregation, we used jack bean ␣-mannosidase as substrate. This glycoprotein possesses a monoglucosylated oligosaccharide that is recognized by the lectin site of Crt and, when diluted from denaturant, aggregates very rapidly (34,35). As shown in Fig. 3B, mutation of either the lectin site or the polypeptide binding site of Crt resulted in an impaired ability to suppress ␣-mannosidase aggregation. When both sites were mutated simultaneously, an even greater impairment in aggregation suppression function was observed. When the concentration of each mutant was adjusted to achieve the same degree of aggregation suppression as wild type Crt, it was observed that the polypeptide binding-deficient mutant was about 2.5-fold less potent than wild type Crt, the lectin-deficient mutant was 10-fold less potent, and the combined mutant was 20-fold less potent (Fig. 3C). These findings indicate that both lectin and polypeptide-based interactions are used by Crt to suppress the aggregation of a monoglucosylated glycoprotein in vitro.
The Polypeptide Binding Site of Crt Is Dispensable in Supporting the Biogenesis of Glycoproteins in Living Cells-We used three approaches to assess the relative roles of the lectin site and newly identified polypeptide binding site of Crt in the biogenesis of glycoproteins in cells. The first approach was based on the observation that Crt plays an important role in the biogenesis of class I molecules of the major histocompatibility complex. Crt is present in a peptide loading complex (PLC) consisting of the class I heavy chain, ␤ 2 -microglobulin, Crt, ERp57, tapasin, and the transporter associated with antigen processing (TAP). The PLC ensures that class I molecules are loaded with a spectrum of high affinity peptides prior to their delivery to the cell surface. Once at the cell surface, they are examined for the presence of foreign peptide antigens by cytotoxic T cells. In Crt-deficient K42 mouse fibroblasts, class I molecules undergo inefficient peptide loading and exhibit markedly reduced expression at the cell surface (36).
We expressed wild type Crt in K42 fibroblasts as well as the lectin-deficient, polypeptide binding-deficient, and combined mutants and then examined whether the mutants were incorporated into the peptide loading complex and whether they were capable of supporting high level expression of class I molecules at the cell surface. All constructs possessed an influenza hemagglutinin (HA) epitope tag just prior to the C-terminal KDEL ER localization sequence. As shown in Fig. 4A, the constructs were expressed at similar levels when transiently transfected into K42 cells. Consistent with previous work, wild type Crt increased the cell surface levels of the class I molecules H-2K b and H-2D b by 4 -5-fold as assessed by flow cytometry (Fig. 4B). By comparison, the lectin-deficient Crt mutant (D317A; L Ϫ ) was ineffective in supporting increased surface expression of either class I molecule, whereas the polypeptide binding-deficient mutant (P19K/I21E; P Ϫ ) was just as effective as wild type Crt (Fig. 4B). The combined mutant (L Ϫ P Ϫ ) behaved in the same manner as the lectin-deficient mutant. Consistent with these findings, when the PLC was immunoisolated with anti-tapasin antibody and immunoblotted for Crt, mutants lacking a functional lectin site were not present in the complex, whereas wild type Crt and polypeptide binding-deficient Crt were readily detected (Fig. 4C). These findings indicate that a functional lectin site is critical for incorporation of Crt into the PLC and for supporting high level expression of class I molecules at the cell surface. The polypeptide binding site does not appear to participate in these processes. Given the important contribution of the polypeptide binding site of Crt in suppressing protein aggregation in vitro, we speculated that its apparent lack of involvement in class I biogenesis might be due to Crt functioning more to stabilize the PLC rather than suppressing the aggregation of newly synthesized class I molecules (20,21). Consequently, we turned to another assay where Crt has been shown to play a clearer role in suppressing client protein aggregation. The null Hong Kong (NHK) and Z variants (ATZ) of ␣ 1 -antitrypsin are prone to misfolding and are largely retained in the ER for disposal by ER-associated degradation (37,38). Kaufman and co-workers (39) recently demonstrated that the NHK variant interacts with Crt and that the presence of Crt increases NHK solubility.
We co-expressed the lectin and polypeptide binding-deficient mutants of Crt along with the NHK or ATZ variants in K42 cells to assess the relative roles of these sites in promoting the solubility of non-native glycoprotein substrates. Cell lysates were prepared using CHAPS detergent and separated into soluble and insoluble fractions that were subsequently immunoblotted to detect NHK (Fig. 5A) or ATZ (Fig. 5B). A substantial portion of both variants was detected in the insoluble fraction in the absence of Crt (Control lanes) with a marked increase in solubility upon expression of wild type Crt (quantified in Fig.  5C). Remarkably, when wild type Crt was replaced with the lectin-deficient, polypeptide binding-deficient, or combined mutants, only those mutants with a defective lectin site were compromised in their abilities to suppress NHK or ATZ aggregation and insolubility. These findings indicate that the poly- peptide binding site of Crt is dispensable for enhancing the solubility of the non-native NHK and ATZ variants in cells.
As a final test of the involvement of the polypeptide binding site in the functions of Crt within cells, we examined the ability of the various Crt mutants to bind to newly synthesized glycoproteins in a pulse radiolabeling experiment. Mouse L cells transiently expressing HA-tagged wild type and mutant forms of Crt were radiolabeled with [ 35 S]Met for 20 min and lysed in digitonin lysis buffer to preserve weak Crt-substrate interactions (13), and then Crt and associated client proteins were immunoisolated with anti-HA monoclonal antibody (mAb). Many newly synthesized proteins were co-isolated with wild type Crt (Fig. 6, compare untransfected control and wild type (WT) Crt lanes). Remarkably, mutation of the Crt polypeptide binding site had little if any effect on the pattern or intensity of interacting client proteins (Fig. 6, lane P Ϫ ). In contrast, Crt lacking a functional lectin site exhibited profoundly reduced association with the vast majority of clients (Fig. 6, lane L Ϫ ). Of particular note, there was no additional impairment in client protein interaction observed when both the lectin and polypeptide binding sites were mutated (Fig. 6, lane L Ϫ P Ϫ ). In an effort to preserve and detect potentially weak polypeptide-based interactions (22), we repeated the radiolabeling experiment, cross-linked intact cells with 1 mM dithiobis(succinimidyl propionate), and then lysed cells in Nonidet P-40 lysis buffer followed by anti-HA immunoisolation and analysis by reducing SDS-PAGE. However, the outcome of the experiment was the same (data not shown). We conclude that the predominant interaction of Crt with newly synthesized clients occurs through lectin-based binding with little or no detectable contribution from the newly identified polypeptide binding site.

Discussion
Since the initial description of glycan-independent binding of Crt or Cnx to non-native protein conformers (10,40) and their capacity to suppress the aggregation of non-glycosylated proteins in vitro (41,42), there has been considerable interest in defining the binding site responsible and assessing the relative contributions of glycan-dependent and -independent interactions to chaperone function in vivo. Most effort has been directed toward Crt because it is a soluble protein, and it has been implicated in a remarkable array of protein interactions not only in the ER but also the cytosol, nucleus, cell surface, and extracellular matrix. Many of these interactions are unlikely to be lectin-mediated based on their subcellular location or the lack of monoglucosylated oligosaccharide on the Crt binding partner. Examples include binding in the cytosol to glucocorticoid receptors and the ubiquitin-like protein GABARAP as well as to the cytosolic tail of integrins. At the cell surface, there are well documented associations with thrombospondin, CD91, CD69, collectin, ficolin, the complement protein C1q, and others (for reviews, see Refs. 43 and 44). In some instances, the binding site on Crt has been identified such as residues 36 -53 for thrombospondin (30) and residues 195-205 for GABARAP (29) (numbering includes signal sequence), and these sites have been considered as potential candidates responsible for the ability of Crt to suppress client protein aggregation. However, these sites could be excluded because neither GABARAP nor the thrombospondin-derived hep1 peptide could compete with the ability of Crt to suppress protein aggregation (24).
Additional efforts to identify the glycan-independent binding site have involved mutagenesis of suspected regions. These have included His-170 (45) as well as a variety of residues predicted to be hydrophobic and surface-exposed based on a homology model of the Crt structure (28). However, either these mutants were destabilizing (45), or they did not impair aggregation suppression function under physiological conditions of the ER (28). More recently, interest has focused around the lectin binding site because a crystal structure of the Crt globular domain detected an N-terminal affinity tag interacting with the edge of the lectin site (31). Indeed, a mutation at Trp-319 within the lectin site has been reported to attenuate aggregation suppression in vitro (32). This region was also implicated by the finding that occupancy of the lectin site with a monoglucosylated tetrasaccharide reduced the binding of several glycosy- FIGURE 5. The lectin function of Crt is required to increase the solubility of aggregation-prone NHK and Z mutants of ␣ 1 -antitrypsin. Crt-deficient K42 cells were co-transfected with empty vector (Control) or plasmids encoding WT Crt, lectin-deficient Crt (D317A; L Ϫ ), polypeptide binding-deficient Crt (P19K/I21E; P Ϫ ), or combined lectin and polypeptide binding-deficient Crt (P19K/I21E/D317A; L Ϫ P Ϫ ) along with pcDNA encoding the NHK or Z variants of ␣ 1 -antitrypsin. After 48 h, cells were lysed in CHAPS buffer and separated into soluble (S) and insoluble (I) fractions. Proteins in each fraction were separated by SDS-PAGE and immunoblotted with ␣ 1 -antitrypsin or anti-Crt antiserum. A and B depict the solubilities of the NHK and ATZ variants, respectively, in the presence of the various Crt constructs. Asterisks denote crossreactive bands that were also detected in cells transfected with empty plasmids. C, quantification of NHK and ATZ protein bands in soluble and insoluble fractions after normalizing with Ponceau S-stained loading controls. The soluble to insoluble ratio drops in the absence of a functional lectin site. Error bars represent standard error of the mean from three independent experiments. Con, control. lated and non-glycosylated proteins and that a fluorophore immobilized in the vicinity of the lectin site was quenched upon binding these proteins (22). Although these findings collectively suggest the presence of a protein binding site in proximity to the lectin site, this site does not exhibit the expected specificity for non-native conformers. Previous studies have established that the Crt site responsible for suppressing aggregation binds hydrophobic peptides selectively and that such peptides compete in the in vitro aggregation suppression assay (24). By contrast, the lectin-proximal binding site does not exhibit a requirement for substrate denaturation as it is capable of binding several proteins in their native states (22).
Our current efforts to identify the site responsible for aggregation suppression used two assays that report on this activity: aggregation suppression of denatured protein substrates under physiological conditions of the ER and direct binding of hydrophobic peptides. Using structure-guided mutagenesis of four candidate surface-exposed hydrophobic sites, we were able to identify a single site that, when mutated, resulted in both impaired aggregation suppression function and an inability to bind hydrophobic peptides. The mutations did not alter the overall structure of Crt as assessed by CD and melting temperature. Additional independent mutations within the same site impaired function to a similar extent, leading us to conclude that we had indeed identified the site on Crt responsible for its in vitro aggregation suppression function. Surprisingly, this site is located near the mature N terminus of Crt on the opposite face of the globular domain from the lectin site (Fig. 1A) and, as expected, is well removed from the sites of GABARAP and thrombospondin binding. Of interest, both the lectin site and this hydrophobic polypeptide binding site participated in preventing the aggregation of the monoglucosylated glycoprotein client jack bean ␣-mannosidase. We showed previously that these two modes of substrate interaction resulted in Crt and a soluble form of Cnx being much more potent in suppressing glycoprotein aggregation in vitro than a non-lectin chaperone such as the ER Hsp70 BiP (33,35). Given the lack of proximity between the two binding sites, it seems likely that they contribute independently to aggregation suppression rather than engaging a client glycoprotein simultaneously to increase binding avidity.
Three distinct assays were used to assess the relative contributions of the lectin site and hydrophobic polypeptide binding site to the chaperone functions of Crt within the ER of living cells. Disabling mutations in each site were examined for their effects on the ability of Crt to support the biogenesis and surface expression of MHC class I molecules, promote the solubility of mutant forms of ␣ 1 -antitrypsin, and associate with diverse newly synthesized glycoproteins. Given the potency of the polypeptide binding site of Crt in suppressing the aggregation of both glycosylated and non-glycosylated clients in vitro, it was surprising that in all three systems there was little if any impact of mutating this site on Crt function in cells. By contrast, disabling the lectin site profoundly impacted Crt function to an extent similar to that observed in Crt-deficient cells. We considered the possibility that the mutations in the polypeptide binding site, which impaired aggregation suppression by roughly 50% in vitro, might retain greater function within the ER luminal environment. However, if this was the case, one would have expected to see partial Crt function upon mutating the lectin site or in the double site mutant. Rather, we conclude that lectin-oligosaccharide binding is the primary means whereby Crt interacts with newly synthesized glycoproteins, maintains mutant ␣ 1 -antitrypsin solubility, and stabilizes the MHC class I peptide loading complex.
This conclusion is consistent with recent studies, particularly those focused on MHC class I molecules. Del Cid et al. (20) showed that when the lectin-deficient Y109A mutant of Crt (numbered Y92A without signal sequence) was expressed in Crt-deficient cells, neither it nor MHC class I molecules were stably incorporated into the class I peptide loading complex. The expression of class I molecules at the cell surface was also impaired (20). Similarly, the Y109A mutant was unable to support peptide loading onto class I molecules when the PLC was assembled from purified components in vitro (21). Lectin-deficient Crt has also been reported to be strongly impaired in its ability to bind to newly synthesized monoglucosylated glycoproteins in chemically cross-linked cell lysates (22).
Despite the predominant role of lectin-glycan interactions in the binding of Crt (and Cnx) to client glycoproteins, it is clear that lectin-independent associations do exist in cells. This has been observed in many studies using glucosidase-deficient or castanospermine-treated cells (1,2,13,22,46). Another recent example comes from the experiment of Wijeyesakere et al. (22) noted above. Although the lectin-deficient Y109A Crt was impaired in its ability to bind newly synthesized glycoproteins, significant residual interactions were readily detected (22). How are these interactions mediated given our inability to detect any involvement of the hydrophobic aggregation suppression site? The previously described glycan-independent binding site proximal to the lectin site (22) as well as contributions from the arm or P domain seem to be likely candidates (22,24). The lectin-proximal site binds several non-glycosylated proteins with affinities similar to those observed with monoglucosylated glycoproteins albeit with different kinetics (22,24), and mutants of Crt with arm domain truncations exhibit significantly impaired binding to purified non-glycosylated proteins in vitro (22,24) as well as to newly synthesized proteins in lysates of castanospermine-treated cells (22). Binding contributions from both domains could occur for clients entering the cleft between the arm and lectin domains. Determining the functional relevance of these interactions in vivo, either in the biogenesis of non-glycoproteins or in providing additional binding contributions to monoglucosylated glycoproteins, must await the generation of the appropriate disabling mutations.

Experimental Procedures
Materials-The KKAFAF peptide and the KHP peptide, corresponding to the first transmembrane helix (residues 2-24) of the Halobacterium salinarum protein Hsmr, were synthesized on an Intavis Multipep (Intavis AG, Germany) following the manufacturer's directions using Tentagel-S-RAM resin. Peptide purity was confirmed by reverse-phase HPLC to be greater than 90%. Peptides were dissolved in 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 , aliquoted, and stored at Ϫ20°C.
Cell Lines-Wild type (K41) and Crt-deficient (K42) mouse embryonic fibroblasts, a gift from Dr. M. Michalak (University of Alberta), were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 4.5 g/liter D-glucose and supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, 60 g/ml penicillin, and 100 g/ml streptomycin. Mouse L cells were maintained in RPMI 1640 medium with 10% FBS and antibiotics. All cells were grown at 37°C in the presence of 5% CO 2 .
Mutagenesis of Calreticulin cDNA-Amino acid numbering refers to the mouse Crt sequence with residue 1 corresponding to the first residue of the signal sequence. Constructs used for bacterial expression and purification were based on the pET15b-TEV-mCRT-CS plasmid described previously (24). It encodes an N-terminal His 6 tag followed by a TEV protease cleavage site fused to mouse Crt (mCrt) lacking its signal sequence and with an unpaired Cys at residue 163 mutated to Ser. pET15b-TEV-mCRT-CS was used as a template for sitedirected mutagenesis of potential polypeptide binding sites using the primers listed in Table 2. Primer pairs that encoded only a single amino acid change were used in multiple rounds of site-directed mutagenesis to generate double mutants. The lectin-deficient D317A mutant and combined lectin and polypeptide binding-deficient mutant P19K/I21E/D317A were constructed by subcloning the NsiI/BamHI fragment from pQCXIH mCRT D317A into pET15b TEV mCRT or pET15b TEV mCRT P19K/I21E, respectively. Correct introduction of all mutations was confirmed by DNA sequencing.
Purification of Crt and Mutants-Crt or its various mutants were expressed in Escherichia coli BL21-DE3 cells and purified as described previously with some modifications (23). Briefly, cells at an A 600 nm of 0.6 were induced for 3 h at 37°C with 1 mM isopropyl ␤-D-1-thiogalactopyranoside. Cells were harvested and washed once with ice-cold water. Cell pellets were resuspended in lysis buffer (50 mM Tris, pH 8, 300 mM NaCl, 3 mM CaCl 2 , and protease inhibitors) and were lysed using a French press. The lysate was centrifuged at 16,000 ϫ g at 4°C for 60 min and loaded onto a 20-ml nickel-nitrilotriacetic acid column (ThermoFisher, Burlington, Canada). The column was washed extensively, and the protein was eluted in lysis buffer containing 300 mM imidazole. Peak fractions were pooled and dialyzed against 20 mM Tris, pH 8, 150 mM NaCl, 0.4 mM CaCl 2 , and 10% glycerol. Proteins were filtered using a 0.22-m filter, aliquoted, and stored at Ϫ70°C.

Client Binding Sites of Calreticulin
Purification of Firefly Luciferase-FFL was expressed and purified from a pPROEX-FFL vector (gift from Dr. J. Glover, University of Toronto) as described previously (50).
Luciferase Aggregation Assay-Purified Crt and Crt mutants were dialyzed overnight against 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 . Prior to the aggregation assay, Crt proteins at the concentrations specified in the figures were equilibrated at 37°C for 10 min, and then FFL was added to a final concentration of 2 M in a total volume of 200 l. Aggregation was measured every 60 s over a period of 2.5 h at 37°C by monitoring light scattering at 370 nm using a SpectraMax Peptide Binding-To measure peptide binding to Crt and its mutants, fluorescence was measured at 346 nm using a Spex Fluorolog-3 with an excitation wavelength of 280 nm and a 5-nm band pass. Crt (1 M) was equilibrated in 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 for 10 min at 37°C prior to the addition of peptides. Peptides were titrated from a 100 M stock solution. Following each addition, samples were equilibrated for 10 min before reading. Data were fitted to a single site saturation equation for binding using Sigmaplot. All fluorescence measurements were corrected by subtracting the contribution of the respective peptide to fluorescence intensity.
Circular Dichroism Measurements-Reaction mixtures for CD measurements contained Crt (5 M) in 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 . Samples were allowed to equilibrate for 1 h at 37°C prior to measuring spectra between 200 and 260 nm. For thermal denaturation experiments, Crt (5 M) was equilibrated in 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 for 1 h at room temperature. Thermal denaturation curves were recorded from 20 to 70°C at 228 nm with a scan rate of 2°C/min. Assuming a two-state transition process, the thermal denaturation data were fit to a standard equation by nonlinear least square regression using SigmaPlot. The calculated T m value represents the transition midpoint temperature of the thermal unfolding. All CD experiments were measured on a Jasco J-810 spectropolarimeter equipped with a Jasco PTC-423S temperature controlling unit in a 1-mm-path length cuvette. The results are expressed as mean residue molar ellipticity [⌰] with units of degrees cm 2 dmol Ϫ1 . To measure stabi-lization of the Crt structure by the tetrasaccharide Glc 1 Man 3 , the ligand was added to 100 M, and the reaction was equilibrated for 1 h at room temperature before thermal denaturation was initiated.
Flow Cytometry-K42 cells transiently transfected with plasmids encoding various HA-tagged Crt constructs were washed with fresh medium after 24 h and then harvested by trypsinization after 48 h. Following a wash in phosphate-buffered saline (PBS) containing 3% FBS (PBS-FBS), cells were incubated on ice for 20 min in 100 l of ice-cold PBS-FBS containing 2 g of mAb Y3 or B22-249.R1. Cells were washed and then incubated for 30 min on ice in the dark with 2 g of Alexa Fluor 647conjugated goat anti-mouse IgG (Life Technologies) in 100 l of PBS-FBS. Cells were washed twice and fixed in 300 l of PBS containing 0.5% paraformaldehyde. Samples were analyzed using a BD FACSCalibur flow cytometer (BD Biosciences). The data were analyzed using FlowJo software (version 8.8.6). Gating was done to select for GFP-positive and -negative cells, and the difference in the median values of Alexa Fluor 647 fluorescence between GFP-positive and GFP-negative cells was compared to analyze surface expression of MHC class I molecules.
Immunoisolation-To isolate Crt-containing MHC class I peptide loading complexes, transiently transfected K42 cells were lysed at 4°C for 30 min with 1% digitonin lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 1% digitonin, 10 mM iodoacetamide, and protease inhibitors). Lysates were centrifuged at 10,000 rpm for 10 min, and the supernatant fractions were assayed for protein content using the Pierce BCA Protein Assay kit (Thermo Scientific). Lysates (750 g of protein) were incubated with 20 l of anti-tapasin antibody for 1 h at 4°C followed by the addition of 30 l of protein A-agarose beads (Thermo Scientific) for 2 h with rocking at 4°C. The beads were washed with 20 mM HEPES, pH 7.4, 150 mM NaCl, and 1% Nonidet P-40 followed by two washes with 20 mM HEPES, pH 7.4, 150 mM NaCl, and 0.2% digitonin. The peptide loading complex was eluted by incubating for 1 h at 4°C with 300 l of 100 M tapasin C-terminal peptide (SKEKATAASLTIPRNSKKSQ) used to raise the antibody. Any antibody in the eluate was cleared by incubation for 30 min at 4°C with 30 l of protein A-agarose beads. The beads were removed, and the supernatant was dried in a SpeedVac concentrator (Savant Instruments). Samples were dissolved in SDS-PAGE sample buffer, subjected to SDS-PAGE (10% gel) separation, and analyzed for the presence of Crt and TAP1 by immunoblotting.
Mutant ␣ 1 -Antitrypsin Solubility Assay-The effect of calreticulin mutants on the solubility of the misfolded NHK and ATZ variants of ␣ 1 -antitrypsin was investigated using protocols modified from Ferris et al. (39). K42 cells were transiently cotransfected with a pLVX IRES GFP plasmid containing wild type or mutant Crt and a pcDNA3.1 plasmid containing either the NHK or Z variant of ␣ 1 -antitrypsin (2 g of total DNA; 2:1 ratio of Crt:␣ 1 -antitrypsin plasmids) using Lipofectamine 2000. The medium was replaced after 24 h with fresh DMEM without FBS and antibiotics. After 48 h, cells were lysed in 650 l of 2% CHAPS lysis buffer (50 mM HEPES, pH 6.8, 200 mM NaCl, 20 mM N-ethylmaleimide, 2% CHAPS, and protease inhibitors) for 40 min at 4°C. Lysates were centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant was designated the soluble fraction.
The pellet was washed with 150 l of lysis buffer and then solubilized in 100 l of 1% SDS by vigorous vortexing for 30 s, sonication for 20 min, and then boiling until the pellet was completely dissolved. This was designated the insoluble fraction. The soluble and insoluble fractions were subjected to SDS-PAGE, then transferred to PVDF membrane, and stained with 0.1% Ponceau S to visualize proteins for use as loading controls. The membranes were subsequently immunoblotted to detect ␣ 1 -antitrypsin variants and Crt (anti-Crt antiserum). The NHK and ATZ band intensities, obtained from scanned films, were quantified using ImageQuant software and normalized to the Ponceau S loading control. After normalization, the soluble fraction band intensity was divided by the insoluble fraction band intensity to obtain the soluble to insoluble ratio.
Metabolic Radiolabeling-L cells in p60 dishes were transfected with 3 g of plasmid encoding HA-tagged wild type or mutant Crt 24 h prior to radiolabeling. Cells were starved for 15 min in Met-deficient ␣-minimum Eagle's medium and then radiolabeled for 20 min with 0.2 mCi/ml [ 35 S]Met (Ͼ1000 Ci/mmol; PerkinElmer Life Sciences). Following lysis in digitonin lysis buffer (1% digitonin, 20 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM iodoacetamide, and protease inhibitors), Crt and associated proteins were immunoisolated by incubation for 1.5 h with 21 g of anti-HA 12CA5 antibody. Immune complexes were collected for 1 h with 30 l of protein A-agarose, and the beads were washed using 0.25% digitonin, 20 mM HEPES, pH 7.5, and 150 mM NaCl. Proteins were separated by SDS-PAGE (10% gel), and radioactive proteins were detected by fluorography using x-ray film.  Fig.  4B, provided technical advice, and analyzed data. G. K. analyzed surface hydrophobic patches on calreticulin, provided advice, and suggested sites for mutation. D. B. W coordinated the study, helped conceive and design experiments, analyzed data, prepared figures, and wrote the final manuscript. All authors approved the final version of the manuscript.