Scavenger Receptor C-type Lectin Binds to the Leukocyte Cell Surface Glycan Lewisx by a Novel Mechanism*

The scavenger receptor C-type lectin (SRCL) is unique in the family of class A scavenger receptors, because in addition to binding sites for oxidized lipoproteins it also contains a C-type carbohydrate-recognition domain (CRD) that interacts with specific glycans. Both human and mouse SRCL are highly specific for the Lewisx trisaccharide, which is commonly found on the surfaces of leukocytes and some tumor cells. Structural analysis of the CRD of mouse SRCL in complex with Lewisx and mutagenesis show the basis for this specificity. The interaction between mouse SRCL and Lewisx is analogous to the way that selectins and DC-SIGN bind to related fucosylated glycans, but the mechanism of the interaction is novel, because it is based on a primary galactose-binding site similar to the binding site in the asialoglycoprotein receptor. Crystals of the human receptor lacking bound calcium ions reveal an alternative conformation in which a glycan ligand would be released during receptor-mediated endocytosis.

The scavenger receptor C-type lectin (SRCL) 2 is an unusual endothelial cell scavenger receptor. It contains a C-terminal Ca 2ϩ -dependent C-type carbohydrate recognition domain (CRD) that is projected from the cell surface by collagenous and coiled-coil domains that are characteristic of the class A scavenger receptors (1,2). SRCL binds modified low density lipoproteins through these common domains, but the CRD additionally confers a glycan binding function not found in any other scavenger receptors. Recent studies have revealed that the CRD of human SRCL shows remarkably selective binding to glycans containing the Lewis x trisaccharide Gal␤1-4(Fuc␣1-3)GlcNAc, along with weaker binding to the closely related Lewis a trisaccharide Gal␤1-3(Fuc␣1-4)GlcNAc (3,4). Among receptors containing C-type CRDs, only the selectins show specific binding to such a limited set of sugar structures, primarily sialylated and sulfated derivatives of Lewis x and Lewis a (5).
The endothelial localization of SRCL and its ability to interact selectively with a sugar epitope that is commonly displayed on adhesion molecules on the surface of various types of leukocytes and tumor cells suggest further parallels with the selectins. For example, recognition of Lewis x -containing glycoproteins on a breast cancer cell line by SRCL suggests that it might mediate interactions between tumor cells and endothelia during metastasis (6). SRCL also shares several characteristics with the dendritic cell surface receptor DC-SIGN, which binds to Lewis x and Lewis a -containing glycans as well as to high mannose oligosaccharides (7). Like DC-SIGN, SRCL has the ability to serve as a cell adhesion molecule as well as being an endocytic receptor.
Despite these parallels, the structure of the CRD of SRCL suggests that it must bind Lewis x in a fundamentally different way from the way that the selectins and DC-SIGN bind such fucosylated ligands. In the CRDs of the latter receptors, the disposition of amino acid residues around the conserved Ca 2ϩ generates a primary binding site that is configured to bind monosaccharides in which the 3 and 4 hydroxyl groups have the stereochemistry found in mannose or fucose (8,9). Selective binding of Lewis x and related structures results from interaction of the terminal fucose with this primary binding site and additional interactions of the other terminal residues, such as galactose and sialic acid, with adjacent secondary binding sites on the surface of the CRD. In contrast, the amino acid sequence around the conserved Ca 2ϩ in SRCL is characteristic of galactose-binding C-type CRDs, and it would not be expected to accommodate fucose.
In the present studies, human and mouse SRCL are shown to have a similar narrow binding selectivity for Lewis X -containing glycans. The structural basis for such selective binding in a galactose-type CRD has been elucidated by x-ray crystallography and site-directed mutagenesis. In addition, the molecular basis for ligand release at endosomal pH, required for endocytic function of the receptor, has been determined.

EXPERIMENTAL PROCEDURES
Cloning and Expression of Mouse SRCL-The cDNA coding for mouse SRCL was amplified from a mouse lung cDNA library (Clontech). The portion of the DNA coding for the CRD, from residue 603 to the C terminus, was cloned into the pINII-IompA2 expression vector for expression in Escherichia coli as * This work was supported by Grant GM50565 from the National Institutes of Health (to W. I. W.), Grant 075565 from the Wellcome Trust (to M. E. T.), and Grant GM62116 from the National Institutes of Health to the Consortium for Functional Glycomics. 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.  (3), except that in some cases, cell lysis was achieved by passing the washed cell suspension 2-3 times through an EmulsiFlex-C3 homogenizer (Avestin) at a pressure of 10 -15,000 psi. For crystallization, the isolated protein was dialyzed against low salt buffer (25 mM NaCl, 10 mM Tris, pH 7.8, 10 mM CaCl 2 ), applied to an anion exchange column (MonoQ; G.E. Healthcare), and eluted with a linear NaCl gradient from 25 to 1000 mM NaCl. Protein which eluted at ϳ180 mM NaCl was exchanged back to the low salt buffer and concentrated to ϳ15 mg/ml using a spin concentrator. Analysis of Ligand Binding-Fluorescein-labeled extracellular domain of mouse SRCL prepared as described for human SRCL (3) was used to probe the glycan array following the standard procedure of Core H of the Consortium for Functional Glycomics. The specificity of wild-type and mutant CRDs for Lewis x and galactose was determined using a solidphase binding assay with CRDs immobilized to polystyrene wells (3).
Crystallization-Crystals of the CRD from human SRCL were grown at 21°C, using the hanging drop method (1 l of protein to 0.5 l of reservoir in a drop). The protein solution contained 10 mg/ml protein, 8 mM CaCl 2 , 8 mM Tris, pH 7.8, 20 mM NaCl, and 10 mM Lewis x (V-labs, Inc. and Toronto Research Chemicals). The reservoir solution contained 8% polyethylene glycol 8000, 0.2 M Zn(CH 3 COO) 2 , and 0.1 M Tris-Cl, pH 7.0. Crystals were transferred to synthetic mother liquor consisting of all the salts and buffers that were present in the drop, as well as 10 mM Lewis x and 15% ethylene glycol, for 5 min and were then frozen in liquid nitrogen for data collection. Crystals used for the low resolution data set of this protein were grown at 21°C (2 l of protein to 1 l of reservoir in a drop).
The protein solution contained 13 mg/ml protein, 9 mM CaCl 2 , 9 mM Tris-Cl pH 7.8, 22.5 mM NaCl, 5 mM Lewis x . The reservoir solution contained 9% polyethylene glycol 8K, 0.1 M sodium cacodylate, pH 6.5, and 0.2 M Zn(CH 3 COO) 2 . Crystals were transferred to a fresh reservoir solution containing 5 mM Lewis x and 15% methyl pentane diol and then frozen in liquid nitrogen for data collection.
Crystals of the CRD from mouse SRCL were grown at 21°C (1 l of protein to 1 l of reservoir in a drop). The protein solution contained 6 mg/ml protein, 9 mM CaCl 2, 9 mM Tris, pH 7.8, 22.5 mM NaCl, and 5 mM Lewis x . The reservoir solution contained 30% polyethylene glycol 8K, 0.2 M NaCl, and 0.1 M imidazole, pH 8.5. Crystals were transferred to a solution containing all the salts and buffers that are present in the drop, including 5 mM Lewis x , and then frozen in liquid nitrogen for data collection.
Data Collection-Diffraction data were measured at 100 K on ADSC Q315 CCD detectors, at the Advanced Light Source beam line 8.2.1 (high and low resolution CRD from human SRCL) and the Stanford Synchrotron Radiation Laboratory beam line 11-1 (CRD from mouse SRCL). Data were processed with MOSFLM and SCALA (10), and are summarized in Table 1.
Structure Determination-A lower resolution (2.8 Å) data set was measured for the human CRD. These data scaled with P6 symmetry and gave a molecular replacement solution in space group P6 5 using the program Amore (11), with the CRD of DC-SIGNR, Protein Data Bank (PDB) ID 1k9j, as a search model. The best solution gave a correlation coefficient of 42% and an R value of 49% (resolution range 15-3 Å). A partial model for SRCL CRD was built into the electron density map, and although the electron density maps were unambiguous, refinement did not lower the R free below 34%. The original data were incomplete along the 00l axis. However, the high resolution data set showed systematic absences along this axis, with significant intensities only for 00l ϭ 3n. This observation is incompatible with space group P6 5 , implying a lower symmetry trigonal space group (P6 2 and P6 4 , the only hexagonal space groups consistent with these absences, did not give translation function solutions). Molecular replacement for the higher res- is the observed intensity, and ϽI(h)Ͼ is the mean intensity obtained from multiple measurements where ͉F o ͉ is the observed structure factor amplitude and ͉F c ͉ is the calculated structure factor amplitude for the working and test sets, respectively. JUNE 8, 2007 • VOLUME 282 • NUMBER 23

JOURNAL OF BIOLOGICAL CHEMISTRY 17251
olution data set was performed with the program COMO (12) using the partially refined model from the lower resolution data set as a search model. The best solution had two mono-mers in space group P3 2 , with a correlation coefficient of 42% and R value of 41% in the resolution range 12-3.5 Å. Maximum likelihood amplitude refinement was performed using the program CNS (13), with bulk solvent and anisotropic temperature factor corrections applied at all stages. Missing loops were built in gradually, and the resolution was increased to 2.5 Å. After several rounds of positional and isotropic temperature factor refinement alternating with manual model adjustment, most of the residues in the two monomers, designated A and B, could be added to the model. Given the presence of 200 mM ZnCl 2 in the crystallization medium, several strong difference electron density peaks were modeled as Zn 2ϩ , based on the geometry of surrounding ligands and the fact that their refined temperature factors were comparable in magnitude to the surrounding ligands. The human CRDs showed binding to 5 Zn 2ϩ per monomer, but did not show density for Ca 2ϩ or the Lewis x trisaccharide in the expected binding site. Each monomer is cross-linked to its crystallographic symmetry equivalent copy by a Zn 2ϩ (Fig. 1, A and B): His 610 from one monomer A and His 641 from a symmetry-related monomer A bind to the same Zn 2ϩ , and the same holds for monomer B and its symmetry equivalent. Monomers A and B are related to each other by a Ϫ60°rotation and a translation of 1/6 along the z-axis. These two monomers are cross-linked to each other by another Zn 2ϩ , with His 700 from one monomer and Asp 616 and Asp 733 from another providing the coordination ligands. The six monomers (three A and three B) in the unit cell are related to each other by a 6 5 screw axis, to form a hexameric "barrel" that surrounds a large central space (Fig. 1, A and C).
Electron density outside of monomers A and B was seen in the center of the hexameric barrel, with four large peaks (Ͼ6) present in an F o Ϫ F c map of the asymmetric unit. After fixing monomers A and B, a search for an additional CRD was per-FIGURE 1. Arrangement of the CRDs from human SRCL in crystals. A, diagram of multiple P3 2 unit cells. The four monomers A, B, C, and D are shown in red, blue, orange, and green. The relative heights of the molecules along the z-axis are indicated for A and B, and for C and D. Note that monomers C and D are translated along the z-axis relative to A and B by ϳ5.5 Å, such that the local 2-fold axis relating A to C or B to D is not at z ϭ 0. A given unit cell can only have a copy of C or D, which would otherwise overlap. Thus, copies C and D are randomly distributed throughout the crystal, each with a net occupancy of 50%. Dashed lines indicate cross-links mediated by Zn 2ϩ . B, cross-linking of h-SR-CRD monomer A and its symmetry-related molecules. His 610 from molecule A and His 641 from a symmetry-related molecule A bind the same Zn 2ϩ ion. The same cross-linking occurs for the other monomers B, C, and D. C, cross-linking of molecules A and B forms a 6 5 -symmetric hexamer in the unit cell. A Zn 2ϩ binds to His 700 of one molecule and Asp 616 and Asp 733 from the other. D, monomer C and its symmetry mates (yellow), form a chain of monomers with 3 2 symmetry in the center of the hexamer formed by molecules A and B. A similar arrangement occurs for monomer D.
formed in COMO, using a CRD model with the Zn 2ϩ and some loops removed. The search yielded two equivalent solutions related by a 6 5 screw axis, but which overlap each other (monomers C and D, Fig. 1A). Surprisingly, the four large peaks seen in the F o Ϫ F c map calculated only with monomers A and B fit two Zn 2ϩ positions for both monomers C and D (Zn 2ϩ number 1 and 2, Fig. 2A). One of the Zn 2ϩ cross-links each monomer with its symmetry-related copy in the same manner observed for monomers A and B (Fig. 1B), supporting the validity of the solution for monomers C and D. A 2-fold rotation axis relates monomers A or B to monomers C or D, causing the filament formed by cross-linking C or D to run in the opposite direction from monomers A and B along the z-axis. Note that although monomers C and D are related by a 6 5 screw axis, their packing is not compatible with space group P6 5 , as application of a Ϫ60°r otation and a 1/6 translation along z to either C or D results in overlap with its symmetry mate. Instead, the crystal has the lower symmetry of P3 2 , with the unit cell containing monomers A, B, and either C or D. Presumably, C and D are randomly distributed through crystal to give a statistical mixture that effectively makes them present at 50% occupancy (Fig. 1).
To refine the arrangement of molecules in P3 2 , monomers C and D were treated as alternative conformations each with 50% occupancy, i.e. there are three independent copies in the asymmetric unit. Because C and D overlap, the maps around them are not as clear as for monomers A and B, but it appears that they have loop conformations and Ca 2ϩ in similar positions as in the mouse SRCL CRD (see below). Water molecules were added to peaks Ͼ3 in F o Ϫ F c maps and were within hydrogen bond distance to monomers A and B or to other water molecules. Because the quality of the maps around monomers C and D is not as high as for monomers A and B, the only water mol-ecules that were added in the vicinity of monomers C and D are ligands bound to the Zn 2ϩ or Ca 2ϩ . Temperature factor refinement suggested that in some cases Cl Ϫ serves as a Zn 2ϩ ligand instead of water. The final human CRD model contains residues 606 -734 for all protein monomers, 16 Zn 2ϩ , 6 Ca 2ϩ , 12 Cl Ϫ , and 33 water molecules.
Molecular replacement for the mouse SRCL CRD data set, using the program COMO and the partially refined model of the human SRCL CRD as a search model, gave a solution for four monomers in the P1 unit cell. The best solution had a correlation coefficient of 31% and an R value of 43% for the resolution range 12-3.5 Å. The rotation between monomers A and C, and between B and D, is almost 180°, whereas the rotation between the A-C pair and the B-D pair is about 80°. The structure was refined in CNS using a maximum likelihood amplitude target, and bulk solvent and anisotropic temperature factor corrections were applied throughout. Test set reflections for calculating R free were chosen in thin shells. Strict non-crystallographic symmetry was initially applied, but was released later in the refinement as some side chains showed different conformations among the four independent copies. These loops were built in gradually and the resolution was increased to 1.95 Å. In each of the four monomers, four Ca 2ϩ and one Lewis x molecule were visible. The final model contains residues 606 -735 for monomers A and D, 607-698 and 704 -738 for monomer B, 607-737 for monomer C, 16 Ca 2ϩ , 4 Lewis x trisaccharides, and 357 water molecules.

RESULTS
Glycan Ligands for SRCL-To facilitate structural and functional analysis of SRCL, both the human and mouse proteins were investigated. The sequences of human SRCL and mouse SRCL are 91% identical overall, with no insertions or deletions, indicating that this protein is highly conserved between the two species. Soluble fragments of human SRCL consisting of just the CRD or the whole extracellular domain containing the coiled-coil region, the collagen-like region and the CRD have been characterized previously (3). For this study, the equivalent fragments of mouse SRCL were produced.
The binding specificity of the human receptor was previously characterized by probing a glycan array consisting of biotinylated oligosaccharides immobilized on streptavidin in polystyrene wells (3). For comparison, the trimeric extracellular domain of the mouse receptor expressed in Chinese hamster ovary cells was tested against a second generation glycan array, in which oligosaccharides are covalently immobilized on a glass surface (14). Despite the difference in the assay format, the results reveal that, like human SRCL, mouse SRCL is highly specific for Lewis x -and Lewis a -containing oligosaccharides and shows some preference for Lewis x compared with Lewis a (Fig. 3A). The mouse receptor also binds to forms of these ligands in which the 6 position of GlcNAc or glucose is sulfated (glycans 274 and 275), but as expected it does not bind forms in which the 3 position of galactose bears sulfate (glycans 28 and 259 -262). Thus, this receptor shows partial similarity in specificity to the selectins, which can also bind sulfated ligands (5). The fact that the mouse and human receptors show the same restricted specificity for Lewis x and Lewis a is not surprising given that the amino acid sequences of the CRDs of the two proteins are very similar. The CRD sequences are 86% identical overall and in the region shown to form the sugar binding site in other C-type CRDs there is only one amino acid difference between the mouse and human proteins (Fig. 3B).

Recognition of Lewis x by SRCL by a Novel
Mechanism-With the goal of elucidating the mechanism of SRCL binding to Lewis x , attempts were made to crystallize the carbohydrate recognition domain of human SRCL with bound ligand. These efforts proved unsuccessful, but parallel studies on the mouse CRD resulted in determination of the structure in the presence of Ca 2ϩ and Lewis x trisaccharide. The crystals contain four independent copies, each of which reveals four Ca 2ϩ and a Lewis x molecule.
The CRD adopts the typical long form C-type lectin fold (Fig.  2B), including a third ␤-strand at the bottom of the domain (␤0) and a disulfide bond that connects the loops preceding ␤0 and ␤1. As predicted from the amino acid sequence of SRCL, the galactose residue in the Lewis x oligosaccharide interacts with the conserved Ca 2ϩ site in the CRD: the equatorial 3-and axial 4-hydroxyl groups form coordination and hydrogen bonds similar to those seen in other galactose-binding C-type CRDs (Figs. 4 and 5). Carbonyl oxygen atoms from the side chains of Gln 694 and Asn 718 act as Ca 2ϩ ligands, and the amide groups of these side chains serve as hydrogen bond donors to the 3 and 4 hydroxyl groups of galactose. The side chains of Asp 696 and Glu 706 also serve as Ca 2ϩ ligands and act as hydrogen bond acceptors from the same sugar hydroxyl groups. Interactions of the apolar face of galactose with an aromatic side chain are a hallmark of galactose-binding lectins (15). In this case, both C4 and the exocyclic C6 pack against Trp 698 .
The interactions of galactose at the principal Ca 2ϩ site orient Lewis x so that the central GlcNAc residue points away from the protein, while the terminal fucose residue contacts the protein in a secondary binding site, providing specificity for Lewis x over other galactose-containing ligands. In the secondary site, Lys 691 forms hydrogen bonds with the 4-hydroxyl group and the ring oxygen of fucose, and there are van der Waals contacts between the exocyclic methyl group of fucose and C␦1 of Ile 712 . Changing Ile 712 to valine results in a 3-fold loss in selectivity for Lewis x compared with galactose, confirming the importance of this interaction ( Table 2). Mutation of Ile 712 to Ala results in a reduction in sugar-binding activity. Although this mutant still bound weakly to galactose-Sepharose so that some protein could be purified, binding to the LNFPIII-BSA reporter ligand was too weak to allow quantification of binding in solid phase assays. In addition to contacting the fucose residue, Ile 712 also makes contact with Asn 718 , so reducing the size of the side chain at position 712 probably allows Asn 718 to move out of position, disrupting the primary binding site. In previous studies, a mutant CRD in which Lys 691 was changed to alanine still showed preferential binding to Lewis x (3). Thus, in the absence of Lys 691 , hydrogen bonds between the fucose oxygens and water are energetically equivalent to the bonds with the amino group of the lysine residue in the wild-type CRD, probably because of the high solvent accessibility of these hydrogen bonds. Finally, Phe 720 appears to play a critical role in organizing both the primary and secondary binding sites as it packs against Ile 712 as well as main chain and side chain atoms of residues that form the conserved Ca 2ϩ -binding site. Mutation of this residue to alanine results in complete loss of sugar binding activity, as shown by the inability of the mutant to bind to galactose-Sepharose.
The interaction of SRCL with Lewis x is fundamentally different from the way that DC-SIGN and the selectins bind to related glycans, although the conformation of the Lewis x trisaccharide is similar in the DC-SIGN and SRCL complexes (Fig. 4,  A-D). When bound to SRCL, the trisaccharide is oriented with the central GlcNAc residue tipped away from the protein so that the terminal fucose residue contacts the protein in the secondary binding site. In contrast, with the fucose residue in the primary binding site of DC-SIGN, the internal GlcNAc residue points away from the protein in the opposite direction and galactose makes secondary contact with the protein surface (7).
The conformation of Lewis x bound to SRCL explains the ability of this protein to interact with oligosaccharides bearing the Lewis a epitope. When the structure of Lewis a is superimposed onto the Lewis x structure observed here, it is clear that the Gal and Fuc moieties of both trisaccharides can form the same contacts with SRCL (Fig. 4E). This arises from the local 2-fold symmetry that relates the 3-and 4-OH groups of Glc-NAc; the superposition simply results in a reversal of the 2-and 6-substitutents of the GlcNAc pyranose ring. The GlcNAc does not interact directly with the protein in either orientation.
Plasticity in Galactose-binding Sites in C-type CRDs-No other crystal structures for ligand-bound forms of natural galactose-type binding sites in mammalian receptors containing C-type CRDs have been determined. However, the binding site of serum mannose-binding protein has been engineered to resemble very closely the binding site of the asialoglycoprotein receptor and the structure of this CRD in complex with GalNAc has been determined (16). Comparison of the contacts in this binding site with the interactions between galactose and SRCL reveals that the hydrogen and coordination bond networks to the sugar hydroxyl groups are almost identical, but the packing interactions with tryptophan are different because the side chain of the binding site tryptophan has been rotated by nearly 180° (Fig. 5, A and B). More distantly related galactose-binding proteins diverge even further in structure. A tyrosine rather than a tryptophan residue is present in the binding site of rattlesnake venom lectin, although the remainder of the binding site is relatively conserved (17) (Fig. 5C). In a galactose-binding tunicate C-type lectin, the galactose-binding site is reversed because the locations of hydrogen bond donors and acceptors around the conserved Ca 2ϩ have been switched, causing the positions of the 3-and 4-hydroxyl groups of galactose to be swapped (18) (Fig. 5D). Although there is still a packing interaction with a tryptophan residue, it comes from a different portion of the polypeptide than in the vertebrate CRDs.
Ca 2ϩ -dependent Changes in Conformation of the Sugarbinding Site-The mouse SRCL crystals contain four Ca 2ϩ that are found at sites observed in other C-type CRDs (Fig. 2B). In addition to the conserved Ca 2ϩ (site 2), an auxiliary Ca 2ϩ (site 1) is also bound to loops in the upper part of the protein near the carbohydratebinding site and is found in many other C-type CRDs, including mannose-binding proteins and DC-SIGN (8,19,20). The side chains of conserved residues Asp 670 , Glu 674 , Asn 697 , Asp 707 , and the main chain oxygen of Glu 706 , form the auxiliary site. Ca 2ϩ site 3 has been observed in some other crystal structures of C-type lectins where the Ca 2ϩ concentration is high. This Ca 2ϩ shares protein ligands with the auxiliary site and is also bound to several water molecules. The fourth site is in the lower part of the CRD. The coordination ligands for this Ca 2ϩ are the side chain of Glu 731 from the last ␤-strand in the C-terminal part of the protein, the side chain of Asn 646 and the main chain oxygen of Phe 644 , both of which are in the loop connecting the two ␣-helices, the side chain of Glu 650 in the second ␣-helix, and two water molecules. Glu 731 also forms a salt bridge to Lys 617 , a residue from the central ␤-strand of the lower sheet, to stabilize this region further. This Ca 2ϩ site is found in other C-type CRDs, for example in the human asialoglycoprotein receptor (21), whereas in other C-type CRDs, including mannose-binding protein A and DC-SIGN, a salt bridge stabilizes this region (8,19). The presence of this site in this structure may be a result of the high Ca 2ϩ concentration used for crystallization. It is not clear whether the absence of this Ca 2ϩ would significantly affect the structure of the protein, given that the side chains that form the salt bridges in this region in other C-type CRDs are present in SRCL.
Although ligand-containing crystals of the CRD from human SRCL were not obtained, crystals grown at somewhat reduced pH (7.0 versus 8.0) were analyzed. In the asymmetric unit, two copies, designated A and B, are present at full occupancy whereas a third molecule is present at 50% occupancy in one of two overlapping positions (modeled as copies C and D). Neither Ca 2ϩ nor Lewis x is observed in copies A and B, whereas Ca 2ϩ can be discerned in the electron density maps of C and D. However, the overlapping electron density of C and D makes definitive assessment impossible. The absence of bound Ca 2ϩ under these crystallization conditions probably reflects both the pHdependent loss of binding activity that allows SRCL to function as a recycling endocytic receptor and the presence of Zn 2ϩ , which binds to side chains present on exposed loops, resulting in cross-linking of monomers. It is possible that these interac-  tions stabilize loop conformations associated with loss of Ca 2ϩ binding.
The absence of Ca 2ϩ in monomers A and B, as well as the high quality of the electron density maps, provides the opportunity to compare the structures of the SRCL CRDs in the presence and absence of Ca 2ϩ . Monomers A and B each bind five Zn 2ϩ ( Fig. 2A) The first Zn 2ϩ cross-links a given monomer to its symmetry mate along the crystallographic 3 2 axis by binding to His 610 from one monomer and His 641 from a symmetryrelated monomer (Fig. 1). The second Zn 2ϩ sits in the lower part of the CRD in the position of the fourth Ca 2ϩ ion in the mouse SRCL CRD model. The third Zn 2ϩ links the bottom part of one CRD to the top part of another CRD (monomer A to B and B to A, Fig. 1). The fourth Zn 2ϩ occupies a position similar to that of the auxiliary Ca 2ϩ , but binds to the side chain of His 702 rather than the side chain of Asn 697 . This mode of binding is possible because the loop containing these residues adopts different conformations in the absence and presence of Ca 2ϩ (Fig. 6). The fifth Zn 2ϩ binds to the side chains of residues Glu 662 and His 664 . In contrast to monomers A and B, C and D appear to bind both Ca 2ϩ and Zn 2ϩ . The three Ca 2ϩ at the upper part of the CRD observed in the mouse SRCL CRD are present in C and D. The first and second Zn 2ϩ seen in monomers A and B of the human SRCL CRD are also present in monomers C and D; the first Zn 2ϩ forms cross-links between C or D and their symmetry mates in the same way seen for monomers A and B. There is a third Zn 2ϩ that cross-links Asp 696 from monomer A or B to His 700 from monomer C or D (marked 3b in Fig. 1A).
The most dramatic change between the mouse CRD with Ca 2ϩ bound and the human CRD is in a loop formed by residues 696 -707 in the vicinity of the conserved Ca 2ϩ site, designated loop 2. In different C-type lectins, the release of sugar is coupled to rearrangements in this Ca 2ϩ site (16,22) (Fig. 6). Because loop 2 contains residues of both the primary and auxiliary Ca 2ϩ sites, as well as Trp 698 , which interacts with the bound galactose, altering its conformation would be expected to lead to changes in Ca 2ϩ affinity, and therefore sugar binding, as a function of pH. Endocytic activity of SRCL requires the receptor to release ligand at endosomal pH. At physiological Ca 2ϩ concentrations of 1 mM, the midpoint of ligand binding to SRCL as a function of pH occurs at pH 6.5, which might suggest that a histidine residue serves as a sensor for the binding-to-nonbinding transition (3). His 702 in loop 2 would be a candidate sensor (Figs. 2 and 6). There are three other histidine residues conserved in the CRDs of mouse and human SRCL, but they are positioned farther away from the binding site and their positions do not change significantly between the sugar-bound and ligand-free structures.

DISCUSSION
The structural studies help to explain the preferential binding of SRCL to Lewis x -related glycans. Lewis x glycans are commonly represented on the surface of subpopulations of leukocytes, suggesting a potential mode of interaction of this endothelial receptor with cells in the circulation. Thus, parallels can be drawn between SRCL and the E-and P-selectin cell adhesion molecules, which mediate interactions between endothelial cells and leukocytes by binding to specific glycoprotein ligands on the leukocyte surface. There are also similarities between SRCL and DC-SIGN, which is expressed on dendritic cells rather than endothelial cells, but also mediates cell-cell interactions by binding to Lewis x and related oligosaccharides on leukocytes. Clearly, there are also important differences among SRCL, the selectins, and DC-SIGN. The restricted specificity of SRCL for a narrow class of oligosaccharide ligands is unusual for receptors that utilize C-type CRDs. DC-SIGN binds to a range of glycans that bear terminal fucose residues as part of a branched terminal structure and in addition binds to a distinct class of high mannose ligands (7). Even the selectins, which bind primarily to sialylated and/or sulfated forms of Lewis trisaccharides, have been shown to bind to additional classes of charged oligosaccharides (23). There are also clear mechanistic differences between the ligand binding activities of SRCL and other cell adhesion receptors such as the selectins and DC-SIGN, because the interaction with the trisaccharide core of the Lewis x -type glycans by SRCL is based primarily on recognition of galactose rather than fucose. Thus, although the oligosaccharide-binding characteristics of SRCL and DC-SIGN overlap, the fact that they have some common ligands represents a convergence of binding specificity from the two general categories of galactose and mannose/fucose binding groups of C-type CRDs.
It is intriguing that in addition to having the potential to function in cell adhesion, both SRCL and DC-SIGN are able to mediate endocytosis. The structural studies suggest that the pH-dependent release of ligands needed to allow receptor recycling during endocytosis results from conformational changes that cause loss of Ca 2ϩ binding. However, comparison of multiple pH-sensitive C-type CRDs suggests that the pH-sensing mechanisms in different CRDs are likely to be different.