Interactions of Substrate with Calreticulin, an Endoplasmic Reticulum Chaperone* □ S

Calreticulin is a molecular chaperone found in the endoplasmic reticulum in eukaryotes, and its interaction with N -glycosylated polypeptides is mediated by the glycan Glc 1 Man GlcNAc present on the target gly- coproteins. Here, we report the thermodynamic parameters of its interaction with di-, tri-, and tetrasaccharide, which are truncated versions of the glucosylated arm of Glc 1 Man 7–9 GlcNAc 2 , determined by the quantitative technique of isothermal titration calorimetry. This method provides a direct estimate of the binding con- stants ( K b ) and changes in enthalpy of binding ( (cid:1) H b °) as well as the stoichiometry of the reaction. Unlike past speculations, these studies demonstrate unambiguously that calreticulin has only one site per molecule for binding its complementary glucosylated ligands. Although the binding of glucose by itself is not detectable, a binding constant of 4.19 (cid:2) 10 4 M (cid:3) 1 at 279 K is obtained when glucose occurs in (cid:4) -1,3

Calreticulin (CRT), 1 along with calnexin, serves as a molecular chaperone in the endoplasmic reticulum (ER) of eukaryotic cells. Although calreticulin is a soluble, luminal protein, cal-nexin is a type I membrane protein (1,2). Segments of these proteins share amino acid identity ranging from 42 to 78% (3). Calreticulin is a highly conserved ubiquitous protein (M r 46,000) and has been implicated in Ca 2ϩ storage and intracellular Ca 2ϩ signaling in the sarcoplasmic and endoplasmic reticula (4,5). CRT has been divided into three regions: the N-terminal, the C-terminal, and the central P-domain, which consists of short sequence motifs repeated three times in tandem. The N-terminal domain is highly conserved among CRTs from different species and potentially mediates interactions between CRT and the ER folding catalysts, protein disulfide isomerase and ERp57 (6,7). Recent studies show that the P-domain, previously thought to be involved in oligosaccharide binding, interacts directly with ERp57 (8 -10). The C-domain is characterized by a high content of acidic residues (4,11), which is consistent with the location of a low affinity (K d ϭ ϳ1-2 mM), high capacity (ϳ25-50 mol) calcium-binding site (12) and contains the ER retrieval sequence.
The ER plays an essential role in the folding and maturation of newly synthesized proteins in the secretory pathway. ER quality control operates at various levels; one of the most common modifications of proteins translocated into the ER is the addition of N-linked glycans (13). CRT has been found to bind to partially trimmed, monoglucosylated glycans, and it has been shown that the binding of calreticulin involves a direct oligosaccharide-protein interaction specific for monoglucosylated oligosaccharides (14 -17). Recently the crystal structure of the luminal domain of calnexin has been published (18), which shows a single carbohydrate-binding site in its lectin domain. Given the sequence similarity between calnexin and calreticulin, the carbohydrate-binding site in the latter should reside in the corresponding homologous domain.
The binding of CRT is exquisitely specific for the monoglucosylated N-linked oligosaccharides, which appear as transient intermediates in vivo. The chaperone function of calreticulin is assisted by two enzymes of contrasting catalytic activities: UDP-glucose:glycoprotein glucosyltransferase and glucosidase. The sequential action of glucosidase I and II removes two glucose residues from Glc 3 Man 5-9 GlcNAc 2 from N-linked glycoproteins in the ER (19), whereas UDP-glucose:glycoprotein glucosyltransferase catalyzes the addition of glucose on the nonreducing end mannose of the ␣1-3 arm of Man 5-9 GlcNAc 2 chains of unfolded glycoproteins, thereby acting as a folding sensor (20,21). Repeated cycles of reglucosylation by the glucosyltransferase lead to prolonged association of calreticulin with the unfolded protein until its appropriate folding occurs. Should this process fail, the protein is retrotranslocated to the cytosol and subsequently degraded by the proteasome (22).
Substrate studies have identified the single-terminal glucose residue as a critical determinant recognized by calreticulin * This work was supported by grants from the Department of Biotechnology and Department of Science and Technology, Government of India (to A. S.), the Swiss National Science Foundation (to A. H.), and the European Union (Contract No. HPRN-CT-2000-00001 to S. O.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. because oligosaccharides containing 0, 2, or 3 glucose residues fail to bind (14,23). Oligosaccharide binding is critical for the formation of complexes between glycoproteins and calreticulin. If N-linked glycosylation is blocked with tunicamycin or if production of the Glc 1 Man 9 GlcNAc 2 species is prevented by treatment with the glucosidase inhibitors castenospermine or deoxynojirimycin, the binding of calreticulin to a vast majority of proteins is inhibited (15,24).
Thus, because of the importance of the calreticulin-glycoprotein interaction in various biological processes, the study of calreticulin-oligosaccharide interactions at the molecular level has become imperative. The oligosaccharide binding studies on calreticulin and calnexin have been limited because of the nonavailability of N-linked sugar Glc 1 Man 9 GlcNAc 2 in sufficient amounts from natural sources as well as the difficulty in synthesizing such large oligosaccharide structures. Here, we report the interaction of di-, tri-, and tetrasaccharide with calreticulin using the quantitative technique of isothermal titration calorimetry. We report the stoichiometry, binding constant, and various thermodynamic parameters for these interactions. Calreticulin-sugar binding is enthalpically driven, the binding enthalpies of tetrasaccharide being larger than those of the trisaccharide and the enthalpies of trisaccharide binding being larger than those observed for disaccharide. These thermodynamic data thus define in quantitative terms the extended combining site of calreticulin for the glucosylated arm of the oligosaccharide. Moreover, these data are consistent with the model of the complexes of calreticulin with these sugars.
Synthesis of Oligosaccharides-All three of the oligosaccharides have been synthesized previously (25)(26)(27). However, in our approach the glucose moiety was introduced last, using thioglycoside donors in both tetrasaccharide and trisaccharide synthesis. This is the route followed by Matta and colleagues (25)  Expression and Purification of Recombinant Calreticulin-The GST-CRT full-length fusion protein was purified essentially as described by Peterson and Helenius (29). Briefly, the construct was transformed into Escherichia coli DH5␣ cells. An overnight plateau phase culture was used to inoculate fresh Luria broth containing 50 g/ml ampicillin, 2 mM calcium chloride at a dilution of 1:100. Cells were grown at 30°C to an A 600 of 0.4 -0.6 and were induced by 0.2 mM isopropyl-1-thio-␤-Dgalactopyranoside followed by incubation for an additional 6 h. The cells were harvested and resuspended in a 1:10 culture volume of lysis buffer (500 mM Tris-HCl, 100 mM NaCl, pH 7.5) and sonicated on ice using a Branson tip sonicator. Lysates thus obtained were loaded onto GSTagarose (equilibrated with 2 bed volumes of lysis buffer). The specifically bound protein was eluted by 5 mM reduced glutathione in Tris buffer. The eluate was concentrated using a 10-kDa Amicon concentrator and dialyzed against a 10 mM MOPS, 5 mM CaCl 2 , 150 mM NaCl, pH 7.5, buffer.
Protein Estimation-The E 280 of calreticulin was estimated as 1.94 using the method of Gill and von Hippel (30). For a 1-mg ml Ϫ1 GST-CRT solution, A 280 was determined as 1.06.
Isothermal Titration Calorimetry-ITC experiments were performed using a VP-ITC calorimeter from Microcal Inc. (Northampton, MA) as described previously by Wiseman et al. (31). For titrations involving triand tetrasaccharide, typically 90 M protein (in 10 mM MOPS, 5 mM CaCl 2 , 150 mM NaCl, pH 7.4) in a sample cell, volume 1.4181 ml, was titrated with 10ϫ excess sugar (present in same buffer) from the 300-l stirrer syringe. For the titration of disaccharide, for example, 100 M protein was used, and the sugar (1.3 mM) was added as 30 -40 injections of 6 l into the sample cell. The titrations were performed while samples were being stirred at 400 rpm at the required temperatures. An interval of 3 min between each injection was given for the base line to stabilize. The heat-of-dilution values of the sugar were subtracted from the titration data. The data so obtained were fitted via the nonlinear least-squares minimization method to determine binding stoichiometry (n), the binding constant (K b ), and the change in enthalpy (⌬H b°) using Origin software (Microcal) as described previously (31,32). The experimental conditions ensured that the c value ranged between 2 and 130, where c ϭ K b ϫ M t (0) and M t (0) is the initial macromolecule concentration for all of the titrations.
Modeling Studies-Because the crystal structure of calreticulin has not been determined, the substrate specificity of this protein to various sugars was investigated by computer modeling using the MODELLER suite of programs (33,34 ture of calreticulin. The modeled structure of calreticulin was then subjected to simulated annealing and positional refinement using the CNS suite of programs (35,36). All of the oligosaccharide structures used to examine the binding mode to calreticulin were obtained from SWEET dB (37). The mode of binding of the following oligosaccharides was examined: The sugar-binding site in the modeled calreticulin was deduced from analogy to the crystal structure of calnexin. The monosaccharide was docked into the binding site using the docking package SYBYL (SYBYL® 6.7.1 Tripos Inc., St. Louis, MO). The docking geometry was improved by optimizing the overlap between the functional description of the binding site and the ligand, generating multiple solutions. The least energy solution was taken as the best conformation of the glucose to interact with the protein. This solution was then subjected to positional refinement using CNS. Retaining the position and orientation of the glucose ring with respect to the protein, the disaccharide was superposed on the monosaccharide using the program ALIGN. The protein and sugar coordinates were further examined using INSIGHT II. All possible conformations relating the two sugar moieties were explored at intervals of 10°. The best conformation was chosen as the one making the maximum number of hydrogen bonds and minimum short contacts with the protein. Following a similar procedure, the trisaccharide was superposed on the best conformer of the disaccharide. Retaining the torsion angles along the ␣-1,3 linkage of the disaccharide, the dihedral angle of the 1-2 linkage was allowed to vary from 0 to 360°i n intervals of 10°. A similar procedure was followed for building the tetrasaccharide. The conformation along the ␣-1,2 linkage between the third and fourth mannose sugars was varied through 0 to 360°at intervals of 10°. The best conformer at every stage was subjected to positional refinement using CNS.

RESULTS
We studied the oligosaccharide binding properties of calreticulin by isothermal titration calorimetry as it provides a direct  estimate of the binding constants (K b ) and changes in enthalpy of binding (⌬H b°) as well as the stoichiometry of the reaction (n). Additionally, the results obtained were rationalized on the model of calreticulin built by its homology with the recently solved structure of calnexin (18). Thermodynamics of Calreticulin-Sugar Interactions-The results of a representative titration of the disaccharide, trisaccharide, and tetrasaccharide to calreticulin at 293 K are shown in Fig. 1. Each of these titrations exhibits a monotonic decrease in the exothermic heat of binding with each successive injection until saturation is achieved. A nonlinear least-squares fit of the ITC data to the identical site model is also shown. The close fit of the data to the identical site model shows that calreticulin has only one type of site for its complementary sugar ligand. The thermodynamic parameters, ⌬G b°, ⌬H b°, and ⌬S b at 279 and 293 K, determined from the titration calorimetry measurements are presented in Table I. The values of K b increase almost by a factor of 2 from tri-to tetrasaccharide. Also, there is a greater than 25-fold increase in the binding constant from di-to trisaccharide. These findings demonstrate that calreticulin recognizes all four of the residues on the glucosylated branch of the Glc 1 Man 9 GlcNAc 2 oligosaccharide. The disaccharide Man␣1-2Man␣Me and Man␣1-3Man␣Me did not show any binding to the protein, thus demonstrating the importance of the terminal glucose residue for the interaction. These studies are qualitatively consistent with those of Williams and colleagues (38), who tested a variety of mono-, di-, and oligosaccharides for their ability to inhibit the binding of [ 3 H]Glc 1 Man 9 GlcNAc 2 to GST-CRT immobilized on glutathione-agarose. The smallest inhibitory compounds were the disaccharides, Glc␣1-3Man and Glc␣1-3Glc. Also, the trisaccharide, Glc␣1-3Man␣1-2Man, and the tetrasaccharide, Glc␣1-3Man␣1-2Man␣1-2Man, were ϳ100and ϳ500-fold, respectively, more potent inhibitors than the disaccharides.
The thermodynamics of the binding reaction presented in Table I shows that the binding reaction is entropically driven for Glc␣1-3Man␣Me and enthalpically driven for the tetrasaccharide, whereas for the trisaccharide, Glc␣1-3Man␣1-2Man␣Me, the switch between these two modes occur. The changes in enthalpy observed upon the binding of tetrasaccharide are the highest. Also from the data in Table I, it appears that the changes in enthalpy and entropy upon binding are compensatory (Fig. 2). Similar enthalpy-entropy compensation has been observed earlier for many protein-ligand interactions (39).
Calreticulin Modeling Studies-A model of calreticulin based upon the structure of calnexin is shown in Fig. 3B. The modeling studies provided a rationale for the sugar specificity of calreticulin. The binding site residues in calreticulin for glucose corresponding to those in calnexin are Tyr-109, Lys-111, Tyr-128, Met-131, Asp-135, Gly-124, and Asp-317 ( Fig. 4 and Ref. 18). The monosaccharide, i.e. Glc␣Me, was hydrogenbonded to the side chains of Tyr-109, Tyr-128, Asp-135, and Asp-317. The best conformer of the disaccharide interacted with the protein through six hydrogen bonds made by the second sugar moiety along with the five hydrogen bonds contributed by the first sugar. The addition of the third sugar to the disaccharide introduced three more hydrogen bonds with the protein. The fourth sugar of the tetrasaccharide contributed three additional hydrogen bonds to the interactions with the protein. Likewise, the nonpolar interactions also increased with the size of the oligosaccharide chain. The contacts with the residues of the protein made by the four sugars are listed in Table II. DISCUSSION The Glc 1 Man 9 GlcNAc 2 glycan intermediate formed after cleavage of two glucose residues by ␣-glucosidase I and II from the precursor N-glycan Glc 3 Man 9 GlcNAc 2 in the ER mediates the interaction of glycoproteins with calreticulin and calnexin. These two lectin chaperones promote the correct folding of their glycoprotein substrates. FIG. 3. A, the sequence alignment of calnexin and calreticulin. The structure of calreticulin was obtained by homology modeling using the crystal structure of calnexin (PDB code 1JHN). The figure shows the sequence alignment of those regions of calreticulin that were modeled against calnexin (CNX). Residues 21-349 of calreticulin share similarity with residues 70 -458 of calnexin. SSHM, secondary structure of the homology modeled structure and E-␤-strand, H-helix, L-loop. This figure was made using CHROMA. B, modeled structure of calreticulin using calnexin as the template. The structure contains six ␤-strands in the concave sheet (red), seven ␤-strands in the convex sheet (green), and two additional ␤-strands forming the roof sheet (blue). The ␣-helix, which shields the hydrophobic regions of the convex ␤-sheet, thus preventing the oligomerization of monomers, is shown in yellow. As the structure of the P-domain was not of importance to the present study, it is not shown in the model.
The recently solved crystal structure of calnexin shows that its sugar-binding domain has a striking similarity with that of a "legume lectin" fold (40 -42). Thus, a wealth of data on the molecular features of carbohydrate recognition by legume lectins helps us to explain in molecular terms the mode of interaction of the saccharides to calreticulin (43)(44)(45). In CRT, as in legume lectins, the concave and convex sheets are made of six and seven strands, respectively, whereas the roof region consists of only two strands as compared with five in legume lectins (Fig. 3B). An important finding that comes from the modeling studies is that the presence of a helix behind the convex ␤-sheet perhaps shields the hydrophobic regions of the sheet, thus preventing the oligomerization through interactions between the "convex" sheets of the monomers. This in turn explains why CRT, in contrast to the legume lectins, exists as a monomer.
Elucidation of the stoichiometry of interaction of calreticulin with its substrate has been a matter of speculations and debate. The ITC data presented here establish unambiguously that calreticulin-sugar interactions have a stoichiometry of 1 (Fig. 1). Indeed, the recent structural solution of calnexin also reports one site for glucose binding in this molecule, although the low affinity of this ligand and the consequent low occupancy of the chaperone with glucose cannot be ignored at this moment. The extremely low affinity of glucose for calreticulin is not surprising (K b ϳ 60 M Ϫ1 ), 2 as the modeling studies do not show any aromatic residue within the striking distance for stacking with the monosaccharide. The importance of the stacking interactions between the hydrophobic ␤-face of the pyranose ring of the monosaccharide and the lectins is well known (46). For example, the studies of Sharma and Surolia (47) explicitly noted that the failure of Phaseolus vulgaris lectin to bind to monosaccharides is related to the absence of such a stacking interaction. Indeed, the introduction of an aromatic residue in a P. vulgaris mutant that can stack with the monosaccharide makes up for this deficiency of the native protein (48). The inability of manno-oligosaccharides by themselves to bind shows that glucose at the nonreducing end is necessary for saccharide substrates to bind to calreticulin. An essential requirement of glucose in ␣-1,3 linkage to mannose is supported by modeling studies showing that the equatorially oriented hydroxyl group at C-2 of glucose in the saccharide substrates is making a favorable contribution to the binding process by hydrogen bonding with Asp-317 and Tyr-109 of calreticulin (Table II). The lack of such an interaction between oligosaccharides devoid of ␣-1,3-linked nonreducing end glucose, i.e. as in Man␣1-3Man, thus explains the obligatory requirement of glucose in ␣-1,3 linkage for binding to CRT.
The affinity of glucose per se or methyl ␣-glucopyranoside to calreticulin is very poor. However, its extension by mannose in ␣-1,3 linkage increases the binding constant in a striking manner. The extension of this structure by mannose in ␣-1,2 linkage, as in Gluc␣1-3Man␣1-2Man␣Me and Gluc␣1-3Man␣1-2Man␣1-2Man␣Me, enhances the binding further by 25-and 50-fold, respectively, indicating that the combining site of the chaperone consists of a number of subsites each of which can accommodate a hexapyranosyl residue. As the Ϫ⌬G for any of the saccharides equals the sum of the free energies from each subsite filled, the contribution of Ϫ⌬G from each subsite of calreticulin is obtained by comparing a pair of saccharides. These values together with the corresponding subsite occupied by the constituent sugar units of the saccharides are shown in Table III. An increase in Ϫ⌬G as the successive units of mannose are added supports our suggestion that subsites A-D are contiguous, favorable subsites. Of these subsites, B, C, and D are the strong binding subsites, with subsite C being the strongest, whereas subsite A is a weakly interacting subsite. These interpretations are supported by favorable changes in enthalpies as the size of the oligosaccharides is increased. Thus, subsites C and D contribute Ϫ5.0 and Ϫ3.14 kcal/mol, respectively, to the enthalpy of binding.
The decrease in the enthalpy of binding of saccharides with the decrease in their sizes implies more extensive hydrogen bonding and van der Waals interactions for the larger saccharides. This is also borne out by modeling studies, which show a progressive increase in the number of hydrogen bonds and van der Waals interactions from glucose to the tetrasaccharide. As the maximum increment in ⌬H occurs for subsite C, it is not 2 M. Kapoor and A. Surolia, unpublished results. surprising to find that the mannosyl residue occupying this subsite is involved in a maximum number of hydrogen bonds. Generally, the hydrogen bond between the charged atom and a hydroxyl group is stronger than that involving a neutral atom (49). A perusal of the model of calreticulin-sugar interaction indeed shows many hydrogen bonds between sugar hydroxyls and the charged atoms of CRT. This explains the greater affinities of CRT-oligosaccharide interactions as compared with other lectin-sugar interactions. In addition to hydrogen bonding, residues Arg-73, Tyr-109, Asp-125, Met-131, Leu-318, and Tyr-319 make hydrophobic interactions with the sugar. Interestingly, the phenyl ring of Trp-319 may possibly make an OH-interaction with the third sugar (mannose of Glc␣1-3Man␣1-2Man␣1-2Man␣Me). The O-4 hydroxyl group of the mannose moiety makes contacts with all of the carbon atoms of the phenyl ring of Trp-319 within a distance ranging from 2.9 to 3.6 Å. OH-interactions have been observed only recently in protein structures (50,51). In the case of protein-sugar interactions, there is only one recent example from our own laboratory (52). Interestingly, calreticulin-saccharide interactions, unlike most other protein-carbohydrate reactions, show that for the binding of Glc␣1-3Man␣Me and Glc␣1-3Man␣1-2Man␣Me, the values of free energies are more negative than those of the enthalpies (44). This indicates the favorable entropic effects for binding, especially for Glc␣1-3Man␣Me. This favorable entropic contribution could reflect the hydrophobic nature of the reaction. This is also borne out by greater burial of the apolar surface as the size of the oligosaccharide increases and is more pronounced for the disaccharide. The burial of the hydrophobic surface area was calculated by subtracting the accessible hydrophobic surface area (calreticulin and sugar) from that of the complex. A value of 122 Å was obtained for the monosaccharide, and it increased to 220 Å for the disaccharide. For the trisaccharide and tetrasaccharide, this value was 233 and 260 Å, respectively. However, as the plot of enthalpy-entropy compensation has a slope of 1.25, enthalpic factors, viz. hydrogen bonding and van der Waals interactions, appear to dominate calreticulin-saccharide interactions as is also attested to by modeling studies. The observation of enthalpy-entropy compensation in calreticulin-saccharide recognition can be explained in two ways (Fig. 2). In one, the release or uptake of water has been explained as the primary source of this compensation (53). Another and perhaps more likely explanation is that a progressively greater portion of the hydrophobic surface gets buried with an increase in the size of ligand. However, this favorable term is countered by the effect of the restriction of the ligand in the combining site; the larger the ligand, the tighter the binding (i.e. more number of interactions), and consequently, the greater the extent of restriction of degrees of freedom and the larger the change in the unfavorable entropy term with increasing ligand size, as observed experimentally.
In conclusion, our calorimetric studies demonstrate that calreticulin has a single binding site for its respective glycan moiety and provide evidence for the interaction of various oligosaccharides with calreticulin, showing that the entire Glc␣1-3Man␣1-2Man␣1-2Man structure of the oligosaccharide is recognized by calreticulin. Our modeling studies demonstrate that an increased number of hydrogen bonds and van der Waals interactions occurs as the size of the oligosaccharide is increased, which supports the observed thermodynamic data.