Distinct Contributions of the Lectin and Arm Domains of Calnexin to Its Molecular Chaperone Function*

Calnexin is a Ca2+-binding transmembrane chaperone of the endoplasmic reticulum that recognizes Glc1Man5–9GlcNAc2 oligosaccharides on folding glycoproteins as well as non-native elements of the polypeptide backbone. This latter mode of recognition enables calnexin to suppress the aggregation of both glycosylated and nonglycosylated substrates. The luminal portion of calnexin (S-Cnx) consists of two domains, a globular lectin domain and an extended arm domain. To understand the function of these domains during the interaction of calnexin with non-native protein conformers, we tested deletion mutants of S-Cnx for their abilities to suppress the aggregation of nonglycosylated firefly luciferase. The arm domain alone exhibited no capacity to suppress aggregation. However, stepwise truncation of the arm domain in S-Cnx resulted in a progressive reduction in aggregation suppression potency to the point where the globular domain alone exhibited 25% potency. To characterize the polypeptide-binding site, we used hydrophobic peptides that were competitors of the ability of S-Cnx to suppress luciferase aggregation. Direct binding experiments revealed a single site of peptide binding in the globular domain (Kd = 0.9 μm) at a location distinct from the lectin site. Progressive truncation of the arm domain in S-Cnx had no effect on the binding of small peptides but reduced the binding affinity of S-Cnx for large, non-native protein substrates. Because protein substrates exhibited no binding to the isolated arm domain, our findings support a model in which calnexin suppresses aggregation through a polypeptide-binding site in its globular domain, with the arm domain enhancing aggregation suppression by sterically constraining large substrates.

As nascent proteins are translocated into the lumen of the endoplasmic reticulum (ER), 2 an ensemble of enzymes catalyzes a variety of co-and post-translational modifications, whereas molecular chaperones assist in protein folding as part of the ER quality control system (1)(2)(3). The quality control system also detects nonnative protein conformers, retaining them until folding is com-pleted or, if proper folding cannot be achieved, marking them for ER-associated degradation (4 -6). One of the major chaperone systems in the ER consists of the type I membrane protein calnexin (Cnx) and its soluble paralog calreticulin (Crt). The ER luminal segment of Cnx consists of two domains, a globular lectin domain and an elongated arm domain (7). The lectin domain confers specific binding to glycoproteins bearing Asn-linked oligosaccharides of the form Glc 1 Man 5-9 GlcNAc 2 (8,9). A 140-Å hairpin loop forms the arm domain, the tip of which includes the binding site for a thiol oxidoreductase termed ERp57 (8,10). In vitro studies have shown that the recruitment of a reduced glycoprotein to the Cnx-ERp57 complex greatly enhances oxidative folding relative to ERp57 alone (11). Both the globular and arm domains have been shown to bind Ca 2ϩ by overlay experiments (12) but the crystal structure revealed only a single bound Ca 2ϩ within the globular domain (7). Although the complete structure of Crt has not been solved, an ϳ39% overall sequence identity to Cnx, combined with conserved oligosaccharide binding specificity (13), a shorter but similar arm domain structure as determined by NMR (14), and conserved ERp57 association properties (15) all suggest a similar overall structure for the two chaperones. Ca 2ϩ also binds to the globular and arm domains of Crt (16), and this ion is important for stabilizing both chaperones (17,18).
Cnx and Crt interact with folding glycoproteins via lectinoligosaccharide interactions as well as through polypeptidebased recognition of non-native conformers (9, 19 -21). The lectin-oligosaccharide interaction is regulated by the availability of the terminal glucose on the Glc 1 Man 5-9 Glc-NAc 2 oligosaccharides. Such monoglucosylated oligosaccharides are formed by the sequential action of glucosidases I and II on the Glc 3 Man 9 GlcNAc 2 oligosaccharide that is initially added to nascent chains. Release of the oligosaccharide from Cnx or Crt occurs in conjunction with removal of the terminal glucose residue by the further action of glucosidase II. If folding of the glycoprotein does not occur promptly, the folding sensor UDP-glucose:glycoprotein glucosytransferase recognizes non-native conformers and reglucosylates N-glycans, thereby allowing re-entry into the chaperone cycle (22)(23)(24). Terminally folding-defective glycoproteins are further processed by demannosylation that diverts them from the cycle and into the ER-associated degradation disposal pathway (25).
The polypeptide-binding site on Cnx and Crt permits them to function as classical chaperones capable of recognizing non-native features of protein folding intermediates and suppressing their aggregation. This function was initially uncovered through in vitro experiments that demonstrated that both Cnx and Crt can suppress the aggregation not only of glycoproteins bearing monoglycosylated oligosaccharides but that of nonglycosylated proteins as well (21, 26 -29). Aggregation suppression ability was enhanced in the presence of physiological ER Ca 2ϩ concentrations as well as millimolar ATP, the latter causing an increased hydrophobic surface on the chaperones (17). Recent studies have validated the existence of functional polypeptide-based interactions between either Cnx or Crt and folding glycoproteins in living cells. Lectindeficient mutants of Cnx were shown to interact with heavy chains of major histocompatibility complex (MHC) class I molecules in insect cells and to prevent their rapid degradation (30). Similarly, lectin-deficient Crt was found not only to interact with a broad spectrum of newly synthesized proteins and dissociate with normal kinetics, but it was also able to complement all MHC class I biosynthetic defects associated with Crt deficiency (31).
Collectively, these studies are consistent with a model wherein Cnx and Crt associate with folding glycoproteins through both lectin-and polypeptide-based interactions thereby increasing the avidity of the association relative to either interaction alone (32). Binding of Cnx or Crt serves to prevent premature release of folding intermediates from the ER and promotes proper folding by suppressing off-pathway aggregation and by providing a privileged environment in which associated ERp57 promotes thiol oxidation and isomerization reactions (11,33).
Although the lectin sites of Cnx and Crt have been well defined through structural and mutagenesis studies (7,34,35), less is known about the location and substrate specificity of the polypeptide-binding sites. Deletion mutagenesis of rabbit Crt as well as dog or yeast Cnx suggested that their abilities to suppress the aggregation of nonglycosylated proteins reside primarily within their globular domains (8,36). Furthermore, in vitro binding experiments with nonglycosylated proteins such as citrate synthase and malate dehydrogenase have indicated that both chaperones interact preferentially with non-native conformers, suggesting that they act as folding sensors in addition to the role provided by UDP-glucose:glycoprotein glucosytransferase (21,28). To characterize the specificity of the polypeptide-binding site of Crt, Houen and co-workers (37,38) examined an extensive panel of peptides for their binding to Crt using a competitive enzyme-linked immunosorbent assay. Peptide binding required a minimum peptide length of five residues that were hydrophobic in character. In another study, a hydrophobic Crt-binding peptide was shown to compete with the ability of Crt to suppress the thermally induced aggregation of a soluble MHC class I molecule (27). Collectively, these findings are consistent with the presence of a site on Cnx and Crt that recognizes non-native protein conformers and that, in the case of Crt, exhibits specificity for hydrophobic peptide segments. However, the preceding experiments were performed either in the absence of Ca 2ϩ or at temperatures ranging from 45 to 50°C. Given the role for Ca 2ϩ in stabilizing these chaperones as well as their relatively low melting temperature, Crt T m ϭ 46.4°C (18) and Cnx T m ϭ 49.5°C (17), it is possible that the observed polypeptide-based interactions were influenced by partial unfolding of the chaperones.
Given these limitations and the fact that there have been no reports examining the nature of the polypeptide-binding site of Cnx, we decided to investigate the location and characteristics of the polypeptide-binding function of Cnx under physiological conditions of the ER lumen. Using a recently developed assay in which the soluble ER luminal domain of Cnx (S-Cnx) suppresses the aggregation of nonglycosylated firefly luciferase at 37°C and 0.4 mM Ca 2ϩ (17), we show that this aggregation suppression function resides within the globular lectin domain but is enhanced by the presence of the full-length arm domain. The site in the globular domain responsible for aggregation suppression is distinct from the lectin site and is capable of binding hydrophobic peptides with micromolar affinity. Furthermore, binding studies with peptides and non-native proteins of increasing size revealed that the arm domain contributes to the aggregation suppression function of S-Cnx not through direct substrate binding but rather by sterically constraining large polypeptide chains.
Mutagenesis of Calnexin cDNA-Amino acid numbering refers to the canine Cnx sequence with residue 1 corresponding to the first residue following signal cleavage. Oligonucleotides used in this study are listed in Table 1. The soluble ER luminal domain of canine calnexin (S-Cnx, residues 1-461) and its globular domain (residues 1-255/390 -461) were each amplified in standard PCRs with primer pair S-Cnx-forward/S-Cnxreverse using as templates the previously generated glutathione S-transferase fusion constructs, pGEX-3X S-Cnx-His 6 (21) and pGEX-3X CNX 1-255/390 -461 (8), respectively. The latter construct contains a GSGSG linker between residues 255 and 390. The arm domain (arm, residues 256 -389) was synthesized in a standard PCR with the primer pair arm-forward/arm-reverse using pET15b-Tev-S-Cnx as the template (see below). To generate two variants of S-Cnx with a truncated arm domain, ⌬arm1 (⌬amino acids 315-334) and ⌬arm2 (⌬amino acids 309 -347), an overlap extension PCR (39) was performed. In both cases, plasmid pET15b-Tev-S-Cnx was used as the template in the first round of PCRs. For construct ⌬arm1, a 975and a 420-bp DNA fragment were amplified with primer pairs S-Cnx-forward/⌬arm1-reverse and ⌬arm1-forward/S-Cnx-reverse, respectively. For construct ⌬arm2, a 958-and a 381-bp DNA fragment were amplified with primer pairs S-Cnx-forward/ ⌬arm2-reverse and ⌬arm2-forward/S-Cnx-reverse, respectively. After purification, the synthesized DNA fragments were used in equimolar amounts as the template in a second PCR to amplify the complete constructs using the primer pair S-Cnx-forward/S-Cnx-reverse. The mutagenic PCR primers for both the ⌬arm1 and ⌬arm2 constructs introduced a GSG linker at the site of truncation (Table 1). To generate a lectin-deficient mutant of S-Cnx (LD-S-Cnx), an overlap extension PCR was performed to introduce the amino acid exchanges Y166A, M169A, and I184A into the lectin site of S-Cnx. In the first round of PCRs, plasmid pET15b-Tev-S-Cnx was used as the template. A 577-and a 919-bp DNA fragment were amplified with primer pairs S-Cnx-forward/LD-S-Cnx-reverse and LD-S-Cnx-forward/S-Cnx-reverse, respectively. After purification, the synthesized DNA fragments were used in equimolar amounts as the template in a second PCR to amplify the complete constructs using the primer pair S-Cnx-forward/S-Cnxreverse. All generated PCR products were digested with NdeI and BamHI and subsequently ligated into the expression vector pET15b-Tev, a modified pET15b vector (Novagen) carrying a tobacco etch virus protease cleavage site after the N-terminal His 6 tag. The integrity of each construct was confirmed by DNA sequencing.
Purification of S-Cnx and Deletion Mutants-The soluble ER luminal portion of Cnx or its various deletion mutants were expressed in Escherichia coli BL21-CodonPlus cells (Stratagene) following induction with 1 mM isopropyl thio-␤-D-galactopyranoside for 4 h at 30°C. Bacteria were lysed by French press at 4°C in 50 mM Tris, pH 8, 300 mM NaCl, and 3 mM CaCl 2 containing protease inhibitors. The lysate was centrifuged at 30,000 ϫ g for 60 min at 4°C, and the supernatant containing the hexahistidine-tagged S-Cnx construct was loaded onto a nickel-NTA-agarose column (Qiagen). Subsequent washing steps and the final elution of the recombinant protein with 250 mM imidazole were performed as described in the manufacturer's protocol (Qiagen). The eluate was dialyzed against 20 mM Tris, pH 8, 50 mM NaCl, and 3 mM CaCl 2 and subjected to further purification by Mono Q anion exchange chromatography (GE Healthcare) applying a linear 200-ml NaCl gradient (0.05-1 M) in 20 mM Tris, pH 8, and 3 mM CaCl 2 . S-Cnx and its deletion mutants eluted between 0.35 and 0.5 M NaCl and were judged to be greater than 95% pure by SDS-PAGE and Coomassie Blue staining. After a final dialysis against 20 mM Hepes, pH 7.4, 150 mM NaCl, and 1 mM CaCl 2 , purified proteins were aliquoted and stored at Ϫ70°C. The following average yields were obtained for the various S-Cnx constructs from 2 liters of bacterial culture: S-Cnx, 20 mg; ⌬arm1, 22 mg; ⌬arm2, 19 mg; globular domain, 5 mg; arm, 34 mg; LD-S-Cnx, 20 mg.
Aggregation Assay-Aliquots of recombinant firefly luciferase (FL) (Promega) were stored at Ϫ70°C at a concentration of 226 M in 20 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM CaCl 2 , and 0.5% glycerol and were thawed only once directly before use. The various S-Cnx constructs were equilibrated in 20 mM Hepes, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 with or without peptides (20 M) for 1 h at 37°C. FL was then added to a final concentration of 3 M in a total volume of 150 l. Aggregation was measured over a period of 1 h at 37°C by monitoring light scattering at 360 nm in a temperature-controlled cuvette holder. Absorbance readings were recorded every 6 s using a Shimadzu 1601 spectrophotometer.
Fluorescence Experiments-Intrinsic fluorescence of S-Cnx constructs (0.8 M) was measured with an excitation wavelength of 280 nm in 20 mM Hepes, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 in the presence or absence of various peptides (20 M) after an equilibration period of 1 h at 37°C. For the titration of the various peptides against S-Cnx or the ⌬arm2 variant, samples were allowed to equilibrate for 10 min between the respective titration steps. Fluorescence spectra were recorded from 290 to 390 nm at 37°C using a Photon Technology International QM-1 fluorescence spectrofluorometer with excitation and emission slit widths set to 2 and 5 nm, respectively. 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 S-Cnx or S-Cnx deletion mutants (1.81 M) in 2.5 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 1.4 mM CaCl 2 in the presence or absence of various peptides (20 M). Samples were allowed to equilibrate for 1 h at 37°C prior to measuring spectra between 200 and 260 nm. For thermal denaturation experiments, S-Cnx or its deletion mutants (7.24 M) were equilibrated in 20 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA and either 1.0 or 1.4 mM CaCl 2 for 1 h at 37°C. Thermal denaturation curves were recorded from 20 to 70°C at 231 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 squares regression using SigmaPlot 2004 version 9.0. 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 . In the case of samples containing both S-Cnx and peptide, the obtained spectra were corrected by subtracting the contribution of the peptide alone.

TABLE 1 Oligonucleotides used in this study
Restriction sites are underlined, stop codons are marked in italics and introduced codons are highlighted in bold letters.

Primer Sequence 5 3 3 Modifications
Gly-Ser-Gly linker ⌬arm1-reverse ggcgatctgtccactaccatcaggtacatattcgggttc Gly-Ser-Gly linker ⌬arm2-forward gatgatgaaggtagtggaggtgtctggcagcgacctatg Gly-Ser-Gly linker ⌬arm2-reverse ccagacacctccactaccttcatcatctaaccagccatc Gly-Ser-Gly linker LD-S-Cnx-forward caagacccctgccacgattgcctttggtccagataaatgtggagaagactataagcttcac ttcgccttccgccacaaaaacccc The sample was mixed for 15 s at room temperature, and the reaction was stopped by the addition of 110 l of 36 mM Na 2 S 2 O 5 and 16.4 mM KI in 0.2 M sodium phosphate buffer, pH 7.4. Thin layer chromatography and ␥ counting of separated peptide revealed 61% of input radioactivity incorporated into the peptide resulting in a specific activity of 39.7 mCi/mol peptide. The radioiodination mixture was loaded onto a C 18 Sep-Pak column (Millipore) and was washed with 25 ml of 0.05% trifluoroacetic acid and subsequently with 10 ml of 0.05% trifluoroacetic acid and 5% acetonitrile to remove unbound Na 123 I. The radioiodinated peptide was eluted with 2 ml of 0.05% trifluoroacetic acid and 50% acetonitrile, vacuum dried, and resuspended in 20 mM Hepes, pH 7.4, 150 mM NaCl, and 0.4 mM CaCl 2 .
To analyze the interaction between S-Cnx and KHP peptide, the various His 6 -tagged S-Cnx constructs ( and 250 mM imidazole. The amount of radioiodinated KHP peptide that remained bound to the various S-Cnx constructs in the eluates was quantified using a PerkinElmer Life Sciences 1470 ␥ counter. To control for nonspecific peptide binding to nickel-agarose beads, separate incubations lacking S-Cnx were performed, and radioactive peptide eluted from the beads was subtracted from the values obtained with the complete incubations. Typically, ϳ130,000 cpm of peptide was specifically bound to S-Cnx. Reduction and Carboxymethylation of ␣-Lactalbumin-Bovine ␣-lactalbumin (500 M, Sigma) was reduced and carboxymethylated (R-CMLA) as described previously except that guanidinium chloride was not included (40). R-CMLA was dialyzed exhaustively against 20 mM Hepes, pH 7.4, 150 mM NaCl, and 1 mM CaCl 2 , aliquoted, and stored at Ϫ70°C.
Surface Plasmon Resonance Assays-Real time binding between S-Cnx and non-native protein conformers was analyzed by surface plasmon resonance spectroscopy using a Biacore X instrument (Biacore AB Corp., Uppsala, Sweden). Recombinant firefly luciferase or reduced, carboxymethylated ␣-lactalbumin were covalently immobilized to the activated dextran surface of a CM5 sensor chip (Biacore Inc., Piscataway, NJ) in 10 mM sodium acetate buffer at 1 pH unit below the pI of the respective protein according to the standard amine coupling procedure recommended by the manufacturer. Binding of S-Cnx and deletion mutants to FL and R-CMLA were conducted at 30 and at 25°C, respectively, at a flow rate of 20 l/min. Prior to the experiment, S-Cnx constructs were equilibrated in 20 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1.4 mM CaCl 2 , and 0.005% surfactant P20 (Biacore) for 1 h at the respective temperature. Complex association was monitored for 1 min followed by a 2-min dissociation phase in the same buffer before the surface was regenerated with 50% ethylene glycol, pH 8.5.

RESULTS
Domain Contributions to the Aggregation Suppression Function of Calnexin-To localize the polypeptide-binding site of S-Cnx within the context of its arm and globular domains, we generated the various deletion mutants depicted in Fig. 1 and compared their abilities to suppress the aggregation of FL under physiological conditions of the ER lumen.
Before undertaking the aggregation suppression assays, the structural integrity of the various constructs was assessed by  (7) and theoretical depiction of deletion mutants as rendered by Cn3D Viewer (www.ncbi.nlm.nih.gov). Note that deleted residues were replaced by a GSG linker in the case of the ⌬arm1 and ⌬arm2 constructs and by a GSGSG linker in the case of the globular domain construct (glob). B, illustration of the linear sequences of S-Cnx constructs. ⌬arm1 represents S-Cnx with an arm domain lacking amino acids 315-334. ⌬arm2 is shortened further by removal of amino acids 309 -347 and, for the globular domain mutant, the complete arm domain was removed. Numbers 1 and 2 within the arm domain represent two proline-rich, tandemly repeated sequence motifs. Removed amino acids are shown as dashed lines.
far-UV CD and by measuring their thermal stabilities. As described previously (17), the CD spectrum of S-Cnx shown in Fig. 2A exhibited a negative band at ϳ226 nm and a strong increase in signal intensity below 213 nm. Although the secondary structure of S-Cnx consists mainly of ␤-sheet with a minor ␣-helical component (Fig. 1A), it has a high tryptophan content (2.9% versus 1.7% for an average protein (41)), which results in a CD spectrum that is dominated by the influence of tryptophan residues rather than expressing typical secondary structure characteristics (42). Aromatic side chains likely contribute as well to the distinct far-UV spectrum of the arm domain ( Fig.  2A), with a negative band at 230 nm, intercepting the base line with a pronounced positive maximum at 222 nm. This closely resembles the reported far-UV CD spectrum of the arm domain of Crt (43). The CD spectra of the ⌬arm1 and ⌬arm2 constructs were essentially superimposable with that of S-Cnx except for a minor decrease in intensity in the 200 -209 nm region, correlating with the extent of the truncation. By contrast, the CD spectrum of the globular domain had a comparatively low intensity, a characteristic also reported for the globular domain of Crt (18,43). In addition a negative band at 220 nm and a modest increase in signal intensity below 210 nm were observed. The individual globular and arm domains of S-Cnx have previously been shown to exhibit native functionality as assessed by retention of oligosaccharide and ERp57 binding, respectively (8).
To compare the stabilities of the various deletion constructs, their thermal denaturation curves were examined between 20 and 70°C. A T m value of 48.6°C was calculated from the well defined thermal transition curve of S-Cnx ( Fig. 2B and Table 2). Partial or complete truncation of the arm domain did not result in a significant destabilization of the proteins; only the arm domain appeared to be less stable with a T m value of 44.2°C (Table 2). Importantly, no evidence of denaturation was observed for any of the mutants at the 37°C temperature used in subsequent aggregation suppression assays (Fig. 2B). We previously demonstrated destabilization of S-Cnx upon reducing the Ca 2ϩ concentration below the 0.4 mM level of the resting ER (17). To assess the Ca 2ϩ binding properties of the various mutants, the thermal denaturation experiments were repeated in the absence of Ca 2ϩ . As shown in Table 2, the removal of Ca 2ϩ resulted in decreased thermal stability for S-Cnx and all mutants, with the exception of the arm domain, for which the T m value remained the same. These findings demonstrate that all mutants possessing the globular domain retain the ability to bind Ca 2ϩ and, furthermore, that the observed Ca 2ϩ -dependent changes in S-Cnx structure are mediated by the globular domain rather than by the arm domain.
The various deletion mutants were next compared for their capacities to suppress the aggregation of FL. As demonstrated previously (17), FL is an ideal client protein for these assays because, being unglycosylated, aggregation suppression is mediated solely through polypeptide-based interactions. Furthermore, because FL unfolds and aggregates at 37°C in the presence of 0.4 mM Ca 2ϩ , suppression of aggregation by S-Cnx can be examined under physiological conditions of the ER lumen. As shown in Fig. 2C, we compared the concentrations of S-Cnx and the various deletion mutants required to suppress the aggregation of 3 M FL to ϳ40% of the level observed with FL alone. This is a more sensitive assay than the complete suppression of FL aggregation, and in the case of S-Cnx, a 4 M concentration suppressed FL aggregation to 40%. Progressive truncation of the arm domain resulted in a corresponding increase in chaperone concentration required to suppress aggregation to the same level. By comparing these concentrations, ⌬arm1, ⌬arm2, and the globular domain were 2.5-, 3.25-, and 3.75-fold less potent than S-Cnx, respectively (Fig. 2C). Although this suggested that a polypeptide-based interaction site may reside within the arm domain, the arm domain itself did not reveal any capacity to suppress FL aggregation at a 10-fold molar excess. Thus, the polypeptide-binding site of S-Cnx appears to reside within its globular domain, but the arm domain is required to effect maximal aggregation suppression.
Hydrophobic Peptides Compete with S-Cnx in the Suppression of FL Aggregation-Small peptides have been used as substrate mimetics to report on the location and characteristics of substrate-binding sites on a variety of molecular chaperones (44 -51). In most cases, effective peptides exhibit substantial hydrophobicity. To determine whether this approach could be applied to S-Cnx, we first tested whether peptides with hydrophobic character could compete with S-Cnx to suppress FL aggregation. Again, the S-Cnx:FL ratio was chosen to provide partial aggregation suppression as a sensitive readout of the effects of additives (Fig. 3A). The three peptides tested did not affect the aggregation of FL by themselves (data not shown). However, when S-Cnx was preincubated for 1 h with peptide 6KAAW (KKKKKKAAWAAWAAWAA), the ability of S-Cnx to subsequently suppress FL aggregation was lost. To exclude the possibility that the inhibitory effect of the 6KAAW peptide was because of the positively charged hexa-lysine segment, we tested the more hydrophilic peptide 6KSGG (KKKKKKSGGS-GGSGGSC) and found that it had no significant effect on the aggregation suppression function of S-Cnx (Fig. 3A). We also tested another largely hydrophobic peptide representing the first transmembrane segment of the Hsmr protein from Halobacterium salinarum, for which just two lysines flanking the C and N termini were sufficient to confer water solubility. This KHP peptide (KHPYAYLAAAIAAEVAGTTALKLSK) also blocked the ability of S-Cnx to suppress FL aggregation, although somewhat less potently than 6KAAW. The experiments were repeated with the arm domain truncation mutants, ⌬arm1 and ⌬arm2, as well as with the globular domain with similar results (Fig. 3, B-D).
Hydrophobic Peptides Bind to the Globular Domain of S-Cnx at a Location Distinct from the Lectin Site-To confirm that the ability of hydrophobic peptides to compete in the aggregation suppression assay was because of their binding to one or more sites on S-Cnx, we examined peptide interaction with S-Cnx and its mutants using two direct binding assays. As mentioned previously, the far-UV CD spectrum of a protein is influenced by structural elements other than secondary structure. Disulfide bonds (52,53), the length and regularity of structural elements (54), as well as aromatic amino acid side chains (especially tryptophans) in combination with a low content of ␣-helix contribute to far-UV CD spectra (42,55). Consequently, the far-UV CD spectrum can detect changes in an asymmetric environment upon ligand binding. Thus, in the first assay, we compared the far-UV spectra of S-Cnx with and without the largely hydrophobic 6KAAW peptide and observed a decrease in the magnitude of the mean residue ellipticity values below 237 nm, indicative of interaction (Fig. 4). These changes were also observed upon incubation of S-Cnx with the hydrophobic KHP peptide. By contrast, the more hydrophilic 6KSGG peptide had no detectable effect on the CD spectrum of S-Cnx. Similar  intensity shifts in mean residue ellipticity below 237 nm were measured for ⌬arm1, ⌬arm2, and the globular domain in the presence of the hydrophobic peptides but not with the 6KSGG peptide (Fig. 4). Notably, none of the peptides caused any significant change in the far-UV spectra of the arm domain. The data are consistent with peptide binding only to the globular domain of S-Cnx because observed intensity changes in mean residue ellipticity were independent of the length or presence of the arm domain, and no spectral changes were detected upon incubation of peptide with the arm domain. Furthermore, these findings are in good agreement with the aggregation suppres-sion experiments, where only the hydrophobic 6KAAW and KHP peptides were able to compete with the various S-Cnx constructs to suppress FL aggregation.
To further localize the site of peptide binding to S-Cnx, the KHP peptide was radioiodinated and incubated with the various His 6 -tagged S-Cnx constructs. Following collection of 125 I-peptide-S-Cnx complexes on nickel-agarose beads, complexes were eluted and radioactivity quantified by ␥ counting. As shown in Fig. 5, all deletion mutants with the exception of the arm domain exhibited similar ability to bind radioiodinated KHP peptide. Binding was specific as evidenced by the lack of arm domain binding as well as by the ability of a 20-fold excess of unlabeled KHP peptide to compete with radiolabeled peptide in the assay. These results confirm the suggestion from the CD-based binding assay that the peptide-binding site resides solely within the globular domain of S-Cnx.
The lectin site of S-Cnx possesses significant hydrophobic character because two tyrosine residues and one methionine residue interact with the terminal glucose of Glc 1 Man 5-9 -GlcNAc 2 oligosaccharides (7). As such, it was a candidate for the site of hydrophobic peptide binding. We evaluated this possibility in two ways. In the first approach, the tetrasaccharide G 1 M 3 , which binds to S-Cnx or Crt with similar affinity as fulllength oligosaccharide (K d ϭ 1-2 M (56, 57)), was tested as a potential competitor of the S-Cnx-KHP peptide interaction. As shown in Fig. 5, binding of this oligosaccharide to the lectin site of S-Cnx at a concentration 20-fold greater than that of the KHP peptide had no effect on the formation of the S-Cnx-125 I-KHP complex. In the second approach, we mutated three hydrophobic residues within the lectin site to alanine (Y166A, M169A, and I184A). This triple mutant lacked the ability to bind oligosaccharide as demonstrated by its unaltered thermal stability in the absence and presence of oligosaccharide (compared with wild type S-Cnx that exhibits a 3°C increase in T m ), yet it retained the ability to suppress FL aggregation (data not shown). As depicted in Fig. 5, this lectin-deficient mutant (LD-S-Cnx) was fully capable of binding to 125 I-KHP peptide. Collectively, these findings demonstrate that the peptide-binding site of S-Cnx is distinct from the lectin site.
Characteristics of Hydrophobic Peptide Binding to S-Cnx-As a means to investigate the affinity and stoichiometry of peptide binding to S-Cnx, we compared the intrinsic fluorescence spectra of S-Cnx upon incubation with KHP peptide, YSN peptide (YSNENMETM) and 6KSGG peptide. The YSN peptide was included as a control because it contains a single tyrosine and exhibits weak intrinsic fluorescence, as does the KHP peptide that possesses two tyrosine residues. Furthermore, like   FEBRUARY 6, 2009 • VOLUME 284 • NUMBER 6 6KSGG, the YSN peptide did not compete with S-Cnx in the aggregation suppression assay with FL and was not expected to bind to S-Cnx (data not shown). Fig. 6 shows that, following subtraction of the weak peptide emission spectrum, neither the YSN nor the 6KSGG peptide altered the fluorescent emission spectra of S-Cnx. By contrast, S-Cnx fluorescence was enhanced significantly in the presence of the KHP peptide, indicative of a conformational change associated with complex formation.

Chaperone Functions of Calnexin Domains
We then used the change in fluorescent emission at the peak wavelength of 333 nm to produce titration profiles with increasing peptide concentrations as shown in Fig. 6B. For the KHP peptide, saturable binding to S-Cnx was observed. By contrast, the nonhydrophobic 6KSGG and YSN peptides did not exhibit any binding to S-Cnx over the entire concentration range, confirming the specificity of the assay. Drawing two lines of best fit through the initial linear and saturation portions of the KHP peptide binding isotherm allowed the determination of the equivalence point (58,59), resulting in an approximate stoichiometry of 1.6 mol of KHP peptide bound per mol of S-Cnx. Furthermore, the data could be readily fit to an equation describing single site binding with a K d of 0.9 M for KHP peptide binding to S-Cnx. The measured stoichiometry of 1.6 most likely reflects peptide binding to a single site given the error associated with determining the equivalence point and the finding that the binding isotherm could most readily be fit to an equation describing single site binding. Collectively, these binding experiments with hydrophobic peptides, which inhibit the aggregation suppression function of S-Cnx, strongly suggest that the capacity of S-Cnx to suppress the aggregation of the nonglycosylated client proteins resides within its globular domain at a probable single site distinct from the site of oligosaccharide binding.
Arm Domain of S-Cnx Influences Interactions with Large Substrates-Given that the polypeptide-binding site of S-Cnx resides within its globular domain, we speculated that the observed influence of the extended arm domain on the suppression of FL aggregation (Fig. 2C) might be due to steric effects on the substrate. To investigate this possibility, we examined the influence of the arm domain on the binding affinity of S-Cnx for substrates of differing size. For a small substrate, we used the 2.6-kDa KHP peptide. As a substrate of intermediate size, we employed the 14-kDa R-CMLA, which remains soluble despite assuming a non-native, extended conformation with limited secondary structure (60). It competes with S-Cnx in suppressing the aggregation of FL and thus was considered likely to interact with the chaperone (data not shown). For a large substrate, we used thermally unfolded FL (61 kDa).
Initially, we compared the binding affinity of S-Cnx and the truncated ⌬arm2 mutant for the small KHP peptide. The ⌬arm2 construct was used rather than the globular domain because it was easier to purify in large quantities from E. coli, and it exhibited almost the same reduction in potency, relative to S-Cnx, as the globular domain in the suppression of FL aggregation (Fig. 2C). As shown in Fig. 6B, the peptide binding curves were nearly superimposable, with the calculated K d value for binding to the ⌬arm2 mutant (0.7 M) closely resembling the K d value for binding to S-Cnx (0.9 M). This is consistent with the radioactive peptide binding assay in which the 125 I-KHP peptide bound equally well to all globular domaincontaining S-Cnx variants regardless of the length of the arm domain (Fig. 5).
To examine the binding of R-CMLA and FL to S-Cnx and its truncated arm variants, we employed surface plasmon resonance using sensor chips derivatized with the substrate proteins and S-Cnx mutants as the injected analytes. In the case of FL, experiments were performed at 30°C, a temperature where it undergoes slow unfolding as assessed by light scattering measurements (data not shown). Fig. 7A shows sensorgrams obtained when various concentrations of S-Cnx were injected over immobilized R-CMLA, the intermediate-sized substrate. Binding was readily detected and could be fit to an equation describing single site binding with a K d ϭ 20.2 M (Fig. 7B). Interestingly, when the experiment was repeated with the truncated ⌬arm2 mutant of S-Cnx, a 1.7-fold reduction in binding affinity was observed (K d ϭ 33.5 M, see Fig. 7B and Table 3).
No binding of the isolated arm domain to R-CMLA could be detected (Fig. 7B). The reduction in binding affinity accompanying truncation of the arm domain became even more pronounced when the large FL substrate was examined. For these experiments, the complete set of S-Cnx mutants was tested. As shown in Fig. 7, C and D, binding of S-Cnx to FL was observed with a K d of 1.6 M. However, as depicted in Fig. 7D and summarized in Table 3, a progressive loss of binding affinity occurred as the arm domain was increasingly truncated from 15% (⌬arm1; K d ϭ 3.4 M), to 30% (⌬arm2; K d ϭ 4.7 M), to 100% (globular domain; K d ϭ 6.1 M). This trend closely mirrored the progressive reduction in potency of the mutants in the aggregation suppression assay (Fig. 2C). Again, no specific binding to the isolated arm domain was detected (Fig.  7D); the low signal observed was nonspecific, exhibiting no evidence of saturation up to 160 M (not shown).
Importantly, the magnitude of the reduction in binding affinity when comparing S-Cnx with the ⌬arm2 mutant correlated with the size of the substrate tested (Table 3). There was no loss in binding affinity with the small KHP peptide, a 1.7fold reduction with the intermediate-sized R-CMLA substrate and a 3-fold reduction with the large FL substrate. Because there was no specific binding of any of the substrates to the isolated arm domain, the results are consistent with the arm domain contributing to S-Cnx binding affinity through steric constraint of the larger substrates. Such a mechanism should also be detectable by changes in substrate dissociation rate constants obtained from the surface plasmon resonance data. However, we were unable to fit the dissociation component of the binding curves to obtain these rate constants, presumably because of the multiple conformational states of the non-native substrates immobilized on the sensor chips.

DISCUSSION
In this study, we created mutants of Cnx that consist of its arm or globular domains alone as well as mutants that lack the distal 15% (⌬arm1) or 30% (⌬arm2) of the arm domain to assess the role of these domains in the suppression of substrate aggregation. The far-UV CD spectra and melting temperatures for  the ⌬arm1 and ⌬arm2 mutants were similar to those of S-Cnx, indicating that truncation of as much as 30% of the arm domain does not significantly affect the secondary structure or stability of the chaperone. Not surprisingly, the individual arm and globular domains exhibited far-UV CD spectra different from that of S-Cnx, with the arm domain closely resembling that of the isolated arm domain of Crt (43). The thermal stability of the globular domain was not significantly different from that of S-Cnx, whereas the arm domain exhibited a melting temperature ϳ4°C less than that of S-Cnx. Notably, none of the mutants exhibited any evidence of thermal denaturation at 37°C, the temperature employed in subsequent aggregation suppression and peptide binding assays. These findings contrast with those described previously for Crt where the isolated globular and arm domains exhibited a 7.5°C drop and an 11.6°C drop, respectively, in melting temperature relative to the intact molecule (18,43). This suggests that excision of individual domains from Crt has a greater destabilizing effect than for Cnx. Coupled with previous demonstrations that the individual arm and globular domains of S-Cnx retain ERp57 and oligosaccharide binding function, respectively (8), we conclude that the various deletion mutants examined in this study retain a high degree of structural and functional integrity. These biophysical studies also provided the opportunity to examine which domain of S-Cnx is responsible for the structural stabilization that accompanies Ca 2ϩ binding, previously detected as an increase in thermal stability, increased resistance to exogenous proteases, and reduced binding of the hydrophobic probe 1,1Ј-bis(4-anilino)naphthalene-5,5Ј-disulfonic acid (17). There is some uncertainty concerning the site(s) of Ca 2ϩ binding in Cnx with an early report documenting high affinity binding to the arm domain with lower affinity binding within the globular domain (12). By contrast, the crystal structure revealed only a single putative binding site within the globular domain (7). We compared the thermal melting curves of the various deletion mutants of S-Cnx in the presence and absence of Ca 2ϩ , and we observed that all constructs containing the globular domain exhibited a marked decrease in thermal stability upon complete removal of Ca 2ϩ (Table 2). However, the thermal stability of the arm domain was unaffected. These findings cannot rule out Ca 2ϩ binding to the arm domain, but they do demonstrate that the structural stabilization that accompanies Ca 2ϩ binding is an exclusive property of the globular domain. These findings are consistent with the identification of a putative Ca 2ϩ -binding site within the globular domain of the S-Cnx crystal structure (7) and also with the findings of Bouvier and co-workers (43) who localized the Ca 2ϩ -responsive region of Crt to its globular domain.
Examination of the various deletion mutants for their abilities to suppress the aggregation of nonglycosylated FL under physiological ER conditions revealed that the arm domain itself possessed no aggregation suppression capacity. However, shortening or complete removal of the arm domain resulted in a progressive impairment of chaperone function such that the globular domain alone exhibited about 25% aggregation suppression potency relative to intact S-Cnx. These findings indicate that the region primarily responsible for the chaperone function of S-Cnx resides within its globular domain with the extended arm domain somehow contributing to optimal aggregation suppression. To clarify the role of the arm domain in the polypeptide-based chaperone function of S-Cnx, hydrophobic peptides were used as substrate mimetics to localize and characterize the unfolded polypeptide-binding site(s) on S-Cnx. These peptides were effective competitors in the aggregation suppression assay indicating that they were reporting on the relevant function of S-Cnx. Subsequent peptide binding experiments, using assays based on CD spectral changes or the recovery of radioiodinated peptide-S-Cnx complexes, revealed that peptides bound equally well to S-Cnx constructs regardless of the length or even the presence of the arm domain. Indeed, no binding to the isolated arm domain could be detected. This demonstrates that the site through which S-Cnx exerts its aggregation suppression function resides exclusively within the globular domain and, because peptide binding could not be competed with mono-glucosylated oligosaccharide and was unaffected by lectin-inactivating mutations, that this site is at a location distinct from the lectin site. Further analysis of the peptide binding isotherm revealed a probable single binding site with a K d of 0.9 M for peptide KHPYAYLAAAIAAE-VAGTTALKLSK.
If the arm domain does not possess a binding site for nonnative polypeptides, how does it enhance by as much as 4-fold the aggregation suppression potency of S-Cnx? Using direct binding assays with substrates ranging in size from a 2.3-kDa peptide to 14-kDa R-CMLA to 61-kDa FL, we show that the arm domain increasingly contributes to S-Cnx binding affinity as a function of substrate size. Furthermore, with the large FL substrate, binding affinity also increased as a function of arm length. However, the isolated arm domain itself exhibited no specific binding to any of these substrates suggesting that its contribution to binding affinity occurs through steric constraint of the larger substrates. We envision that as a glycoprotein folding intermediate interacts with both lectin and polypeptide-based binding sites on the globular domain of S-Cnx, it enters the cavity between the arm and globular domains. Electron microscopic studies have revealed that the arm domain appears to be highly flexible, adopting a variety of curved shapes (43). Furthermore, in the S-Cnx crystal structure, the arm domain of each molecule within the crystal lattice is wrapped around the globular domain of an adjacent molecule (7). Consequently, the arm domain could enhance the aggregation suppression function of S-Cnx by physically sequestering a folding glycoprotein. This may have the combined effect of an increase in binding affinity as well as the removal of the folding glycoprotein from the vicinity of other aggregation-prone folding intermediates. Thus, aggregation suppression of client proteins would occur by a combination of direct binding of S-Cnx to exposed hydrophobic patches as well as physical sequestration. That such physical sequestration may be important for substrate interaction in living cells is supported by our recent finding that partial truncation of the arm domain of Crt prevented its interaction with MHC class I molecules despite the continued presence of its lectin and polypeptide-binding sites. 3