Mechanism of Ligand Binding to E- and P-selectin Analyzed Using Selectin/Mannose-binding Protein Chimeras*

The mechanism of oligosaccharide binding to the selectin cell adhesion molecules has been analyzed by transferring regions of the carbohydrate-recognition domains of E- and P-selectin into corresponding sites in the homologous rat serum mannose-binding protein. Insertion of two basic regions and an adjacent glutamic acid residue leads to efficient binding of HL-60 cells and sialyl-Lewisx-conjugated serum albumin. Substitution of glycine for a histidine residue known to stabilize mannose in the binding site of wild type mannose-binding protein results in dramatic loss of affinity for mannose without decreasing binding to sialyl-Lewisx. The accumulated effect of these changes is to alter the ligand binding selectivity of the domain so that it resembles E- or P-selectin more closely than it resembles the parental mannose-binding domain. Affinity labeling using sialyl-Lewisx in which the sialic acid has been mildly oxidized has been used to verify this switch in specificity and to show that the sialic acid-containing portion of the ligand interacts near the sequence Lys-Lys-Lys corresponding to residues 111–113 of E-selectin. The binding of sialyl-Lewisx-serum albumin is inhibited dramatically at physiological and higher salt concentrations, consistent with a significant electrostatic component to the binding interaction. The binding characteristics of these gain-of-function chimeras suggest that they contain many of the selectin residues responsible for selective ligand binding.

The selectin cell adhesion molecules mediate initial binding interactions between moving leukocytes and stationary endothelia (1). The interaction of E-and P-selectins on endothelial cells with saccharide ligands on leukocytes represents a well characterized example of a carbohydrate-based cell-cell recognition mechanism. Among the best counter-receptors for the selectins that have been identified to date are derivatives of either the Lewis x trisaccharide, Gal␤134(Fuc␣133)GlcNAc, or of the Lewis a trisaccharide, Gal␤133(Fuc␣134)GlcNAc, which are sialylated and/or sulfated on the 3 position of the terminal galactose residue (2).
The sugar binding activity of each selectin has been mapped to an NH 2 -terminal C-type carbohydrate-recognition domain (CRD) 1 that is homologous to carbohydrate-binding domains in other calcium-dependent animal lectins (3). The structures of this module and the adjacent epidermal growth factor-like domain have been determined by x-ray crystallography (4). No complex with a saccharide ligand has been reported. However, complexes of sugar ligands with CRDs from two other members of the C-type animal lectin family have been deduced by x-ray crystallography (5,6). These studies of rat serum and liver mannose-binding proteins (MBPs) reveal that terminal mannose, N-acetylglucosamine, or fucose residues interact directly with the protein as well as indirectly through a bound Ca 2ϩ .
Modification of a cluster of three lysine residues near the COOH terminus of the E-selectin CRD has been shown to decrease or abolish binding to leukocytic cells as well as to 3Ј-sialyl-Lewis x (sLe x ) test ligand (4,(7)(8)(9)(10). In previous studies, a gain-of-function mutation in the CRD of MBP was created by incorporation of this cluster of basic residues (11). The resulting chimeric CRD binds to both mannose and to sLe x on HL-60 cells and conjugated to BSA. Subsequent structural analysis (12) revealed that the fucose residue of sLe x binds to this mutant CRD in much the same way as mannose binds to the wild type domain, forming direct coordination bonds to Ca 2ϩ and hydrogen bonds to amino acid side chains that serve as additional coordination ligands for this Ca 2ϩ . The Lewis x portion of the ligand has the same conformation as observed in solution by NMR studies and in crystals of the free oligosaccharide (for review, see Ref. 13), but the Gal and GlcNAc residues make only limited contact with the protein. The appended NeuAc residue is rotated approximately 180°from the orientation observed in solution, as predicted from previous transferred NOE analysis of ligand bound to the native Eselectin CRD (13). This residue does not make direct contact with the protein but is located in the vicinity of the incorporated cluster of lysine residues.
To probe further the mechanism of ligand binding to the selectins, additional gain-of-function mutations of MBPs have been created and analyzed. Binding and affinity labeling experiments demonstrate that a switch of preferred ligand can be achieved so that the mutants are closer to the selectins than to the parental MBPs in ligand binding specificity. The effect of changes in the lysine cluster and the sensitivity of binding to salt concentration suggests that electrostatic attractions may play an important role in initiating or stabilizing the binding interaction.

EXPERIMENTAL PROCEDURES
Mutagenesis-Mutant cDNAs were generated by substitution of synthetic double-stranded oligonucleotides, purchased from GenoSys or Applied Biosystems, for restriction fragments excised from the wild type cDNA for MBP. The mutations were created in an SacI to BamHI * This work was supported by a grant from the Wellcome Trust. 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  fragment of the cDNA in the vector pGEM-3 (14) and were then transferred into the cDNA in the expression vector pINIIIompA2 containing the wild type CRD (15). All mutations were confirmed by DNA sequence analysis using a Sequenase II kit from Amersham. Other reagents for molecular biology were purchased from New England BioLabs.
Protein Expression and Purification-Proteins were expressed in Escherichia coli strain JA221 grown in the presence of Ca 2ϩ to allow isolation of correctly folded and disulfide-bonded CRD directly from the periplasm. LB medium (500 ml) containing 50 g/ml ampicillin was innoculated with 10 ml of an overnight culture and grown at 23°C shaking at 200 rpm until the A 550 nm reached 0.8 (6 -8 h). After the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 40 M and CaCl 2 to a concentration of 100 mM, incubation was continued at 23°C for a further 16 h. Cells were harvested by centrifugation at 3,600 ϫ g for 10 min in a Beckman JS-4.3 rotor and resuspended in 40 ml of loading buffer (1.25 M NaCl, 25 mM CaCl 2 , 25 mM Tris-Cl, pH 7.8). Sonication was performed with the microprobe of a Branson model 250 sonifier for five bursts of 1 min interspersed with cooling on ice. The extract was centrifuged for 10 min at 11,000 ϫ g in a Beckman JA-21 rotor and then 1 h at 100,000 ϫ g in a Ti55.2 rotor. The final supernatant was passed over a 1-ml column of Man-Sepharose or invertase-Sepharose, which was washed with 5 ml of loading buffer and eluted with 5 aliquots of 0.5 ml of eluting buffer (1.25 M NaCl, 2.5 mM EDTA, 25 mM Tris-Cl, pH 7.8). Man-Sepharose was prepared by the method of Fornstedt and Porath (16), and invertase-Sepharose was prepared following exactly the same protocol but substituting a 20% (w/v) solution of yeast invertase, obtained from Sigma Chemical Company, for the mannose.
Further purification was performed on a 4.6 ϫ 75-mm C4 reverse phase column (Atlantis series from Phenomenex). Protein in the eluting buffer was loaded directly onto the column in multiple aliquots of 100 l under starting conditions of 10% acetonitrile and 0.1% trifluoroacetic acid. Protein was eluted with a gradient of acetonitrile increasing at a rate of 1.25%/min. Fractions were dried for 30 min in a Savant Speed-Vac concentrator to remove acetonitrile and lyophilized.
Binding Assays-Protein was adjusted to a concentration of approximately 0.1 mg/ml in loading buffer before coating into Immulon 4 Removawell strips from Dynex Technologies. Wells were coated with CRD and blocked with BSA, and HL-60 cell binding assays were conducted as described previously (11). 125 I-sLe x -BSA was prepared by the chloramine-T method (17) using 50 g of sLe x -BSA (containing 12.6 mol of conjugated oligosaccharide/mol of BSA; from Oxford GlycoSciences) and 1 mCi of Na 125 I from Amersham. For standard binding assays, ligand was diluted to a concentration of approximately 1 g/ml in HEPES-buffered saline (136 mM NaCl, 2.7 mM KCl, 0.5 mM MgCl 2 , 0.9 mM CaCl 2 , 19 mM Na-HEPES, pH 7.5) containing 0.1 mg/ml BSA. Ligand solution (50 l) was incubated in CRD-coated wells for 2 h at room temperature and then removed by aspiration followed by three washes with 200 l/well HEPES-buffered saline containing 0.1 mg/ml BSA. Separated wells were counted in a Wallac Wizard ␥ counter. Results reported for both HL-60 cell and S-Le x -BSA binding assays are the average Ϯ S.D. for 3-28 experiments, each performed in triplicate. Background values obtained in the presence of 10 mM EDTA were always less than 1% of values in the absence of EDTA.
Affinity Labeling-sLe x obtained from Oxford GlycoSciences was labeled with [ 3 H]NaBH 4 (NEN Life Science Products) and purified by washing after spotting on Whatman 3MM chromatography paper by published procedures (18). An aliquot (60 nmol) of lyophilized oligosaccharide was dissolved in 100 l of 100 mM sodium acetate, pH 5.6, containing 5 mM NaIO 4 and reacted for 10 min on ice (19). The reaction was quenched by the addition of 100 l of 10 mM ethylene glycol followed by further incubation on ice for 10 min. The oxidized oligosaccharide was separated from reagent by chromatography on a 2.5-ml column (0.8 ϫ 5 cm) of Sephadex G-10 eluted with water. Fractions of 100 l were collected, and eluted oligosaccharide was detected by scintillation counting of 2-l aliquots. Aliquots of pooled oligosaccharide in approximately 300 l were mixed immediately with protein (at a concentration of approximately 50 g/ml) in the presence of 25 mM Na-HEPES, pH 7.8, containing 10 mM CaCl 2 . NaCNBH 4 was added to a final concentration of 10 mM. After incubation for 10 min at room temperature, the protein was precipitated by the addition of 0.5 volume of 30% (w/v) trichloroacetic acid and incubation on ice for 10 min. The precipitated protein was collected by centrifugation for 5 min at 16,000 ϫ g and washed twice with 100-l aliquots of 50:50 ethanol: ether with intermediate spins of 2 min at 16,000 ϫ g. After drying under vacuum, the protein was dissolved directly in SDS-polyacrylamide gel sample buffer containing 1% 2-mercaptoethanol, incubated 5 min at 100°C, and resolved on a 17.5% polyacrylamide gel (20). The gel was stained with Coomassie Blue and then treated by Amplify solution (Amersham) for fluorography. After exposure to Kodak XAR-5 film for 1-8 days, protein on the stained gel and the radioactive bands were quantified using a Bio-Rad model 620 densitometer. Data were imported into the Origin data analysis program (Microcal) for integration.
For hydroxylamine treatment, protein at a concentration of approximately 0.4 mg/ml was labeled using a reagent concentration of 60 M and precipitated as described above. The protein was dissolved in 0.5 ml of 6 M guanidine hydrochloride and 100 mM Tris-Cl, pH 8.5, containing 100 mM 2-mercaptoethanol and reacted for 30 min at 37°C. Iodoacetamide was added to a final concentration of 70 mM, and alkylation was allowed to proceed for 30 min. After dialysis against two changes of 400 volumes of water, the alkylated protein was lyophilized. Hydoxylamine reagent was prepared at a concentration of 2 M in the presence of 2 M guanidine hydrochloride and 0.2 M potassium carbonate and adjusted to pH 10 with NaOH. Reagent was added directly to the dried protein and incubated for 6 h at 45°C (21). Controls were performed by incubation for 6 h at 0°C and 45°C in 2 M guanidine hydrochloride and 0.1 M potassium carbonate adjusted to pH 10. After incubation, all reactions were dialyzed against two changes of 1,000 volumes of water and lyophilized. Aliquots were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography as described above.

Incorporation of Additional Regions of the Selectin CRDs-
The effects of incorporating three different regions of E-selectin into the CRD of MBP have been investigated previously (11). The regions, two of which are shown in Fig. 1, were selected based on earlier mutagenesis studies conducted with E-selectin because changes in these regions lead to loss of binding activity (4,7,22). Only region 5, when tested by itself, is able to confer HL-60 cell binding activity onto MBP. In an attempt to improve the binding of MBP to these cells, further chimeras have been constructed in which both regions 4 and 5 are included. Because E-and P-selectin differ substantially in this region, sequences corresponding to both were tested. The results (Table  I) indicate that when the three lysine residues of region 5 are present, E-selectin region 4 improves the HL-60 cell binding activity, whereas region 4 of P-selectin actually causes a loss of binding activity.
These results were confirmed using 125 I-sLe x -BSA instead of HL-60 cells as the test ligand. Analysis of the region 5 mutant indicated saturable binding to this ligand (11). The limited availability of the test ligand prevented analysis of binding constants for the multiple mutants created in the current study. Instead, binding was assessed at a single, relatively low concentration of ligand in the linear region of the binding curve. Expression of the resulting binding activity as a fraction of the input ligand makes the value independent of fluctuations in test ligand concentration and gives values analogous to those obtained in the HL-60 cell assay. The relative activities of mutants containing the three different sequences in region 4 correspond reasonably well in the two assays (Table I).
Structural analysis of the original MBP mutant containing just region 5 revealed that the ␤-carbon of His 189 contacts the fucose that is bound near Ca 2ϩ site 2 much as it interacts with terminal mannose bound to wild type MBP and that the imidazole ring makes contact with the GlcNAc portion of the ligand (12). The position of the His 189 side chain in this mutant is illustrated in Fig. 2. The conformation of MBP in the loop containing this residue differs substantially from the corresponding segment of E-selectin such that sugar bound to Eselectin in the observed orientation would not contact this portion of the protein at all. To eliminate the possibility that this contact affects the binding of sLe x , the histidine side chain was removed by mutation to glycine. In previous studies with MBP, this change was found to decrease affinity for Man-BSA and ␣-methyl mannoside (14). Changing this residue in the mutants containing region 5 also results in loss of mannose binding activity as shown by loss of binding to Man-Sepharose. This loss of activity makes it difficult to purify mutations containing His 189 3 Gly mutations on Man-Sepharose. However, yeast invertase immobilized on Sepharose retains these mutants effectively. A comparison of wild type and mutant behavior on Man-Sepharose and invertase-Sepharose columns is shown in Fig. 3.
CRDs prepared on invertase-Sepharose bind HL-60 cells poorly when these CRDs are coated onto polystyrene. The possibility that some immobilized ligand might be released from the affinity columns 2 suggested that such released material could inhibit binding. Therefore, further purification of the CRDs was undertaken by reverse phase chromatography. This approach has been used previously as a final purification step to prepare active forms of other C-type CRDs (24). The repurified material does exhibit HL-60 cell binding activity. Comparison of the His 189 and Gly 189 proteins containing the wild type MBP sequence in region 4 shows that removal of the histidine side chain results in a roughly 2-fold increase in binding of both HL-60 cells and sLe x -BSA (Table I). This finding suggests that some portion of the histidine side chain at position 189 might hinder sLe x binding.
In the presence of Gly 189 , region 4 of E-selectin does not provide significant enhancement of HL-60 cell binding and actually decreases sLe x -BSA binding (Table I). Region 4 of P-selectin supports less binding than the wild type MBP sequence in both assays, although the presence of Gly 189 improves binding compared with His 189 . These results suggest that the ability of E-and P-selectin regions 4 to assume an optimal sLe x binding conformation in the MBP background is influenced differentially by the residue at position 189. Additional changes that appear to enhance formation of the correct conformations of the selectin-like parts of the CRD in the presence of Gly 189 are discussed below.
Affinity Labeling-The failure of the mutants containing glycine at position 189 to bind to Man-Sepharose suggested that the combination of this change with the insertion of lysine residues in region 5 results in a change in the preferred ligand for the CRD of MBP from mannose to sLe x . The limited quantities of sLe x and related ligands available make competition binding assays in the usual format impractical, so an alternative approach using affinity labeling to compare affinities was developed. Previous studies have demonstrated that the interaction between L-selectin and its ligand can be stabilized by mild oxidation of the ligand resulting in removal of C-8 and C-9 of NeuAc and generation an of aldehyde group at C-7 (25,26). Upon binding, this group forms a Schiff's base with lysine residues in the selectins, which can then be stabilized by reduction with cyanoborohydride. To create an affinity label, this approach was applied to sLe x oligosaccharide labeled reductively in the GlcNAc moiety with [ 3 H]NaBH 4 .
The results of incubating oxidized 3 H-sLe x with mutant CRD containing Gly 189 plus regions 4 and 5 of E-selectin are shown in Fig. 4. Concentration-dependent labeling can be detected after gel electrophoresis and fluorography. Controls in the 2 K. Drickamer, unpublished observations.  presence of EDTA at a concentration of 25 mM show no labeling, confirming that the binding is Ca 2ϩ -dependent (data not shown).
The location of the covalently attached affinity label was determined by treatment of the labeled protein with hydroxylamine to cleave at the single Asn-Gly sequence located at positions 201-202 of MBP (Fig. 1), thus removing the COOHterminal 20 residues of the protein. The presence of the expected predominant cleavage product of 20 kDa is observed below the remaining intact 22-kDa starting polypeptide on the Coomassie Blue-stained gel (Fig. 4). A minor cleavage product, at 18 kDa, probably results from double cleavage at positions 201-202 and at the Asn-Met sequence 22 amino acids from the NH 2 terminus of the expressed fragment of MBP. Low efficiency cleavage of such bonds with hydroxylamine has been reported (21). Fluorography reveals that the NH 2 -terminal 20-kDa fragment visible on the gel is not radioactively labeled, indicating that the label must be attached in the COOH-terminal 20-residue segment, which migrates off the end of the gel. Stability of the label under cleavage conditions is evident from the mock cleavages and the fact that remaining intact polypeptide in the hydroxylamine reaction retains bound radioactivity. The only lysine residues in the COOH-terminal segment are the three lysine residues inserted in region 5, so one or more of these residues must be labeled.
The COOH-terminal 20-residue fragment of the peptide was isolated by a combination of gel filtration and reverse phase chromatography. The NH 2 -terminal sequence of this fragment confirmed that cleavage occurs at the predicted Asn-Gly bond. Attempts to localize the labeled residue(s) more specifically using Edman degradation were unsuccessful. However, the labeling observed is consistent with binding of sLe x in the orientation seen in co-crystals with the region 5 chimera. Localization of the NeuAc residue in proximity to the region 5 lysine residues would lead to enhanced and selective reactivity with the affinity reagent. The structure of the sLe x -region 5 complex suggests that the reactive aldehyde group derived from the glycerol side chain of NeuAc would be oriented away from the region 5 lysine residues in the final bound conformation (Fig. 2). The reaction observed may reflect the presence of a portion of the ligand in the solution phase conformation, which would place the reactive group closer to these lysine residues. Such ligand might represent a kinetic intermediate in the association or dissociation process. Alternatively, selective reactivity might result from a local concentration effect as ligand binds and dissociates from the binding site.
The affinity labeling approach was used to obtain semiquantitative information about the relative affinity of this mutant for different saccharide ligands. Radioactive labeling in the presence of competing sugars was quantified and normalized to the protein recovered after gel electrophoresis. The small volumes used for the labeling reaction made it practical to test sLe x and 3Ј-sulfo-Lewis x oligosaccharides and compare them with mannose (Fig. 5A). The results confirm that the Lewis x derivatives are substantially better inhibitors of binding than is mannose. The results shown in Fig. 5B demonstrate that the absence of either the terminal Fuc or NeuAc acid residues of sLe x results in loss in inhibitory potency. These experiments provide further evidence that the modified CRD mimics the ligand binding behavior of authentic selectins.
Region 5 Interactions-The importance of the three lysine residues in region 5 was explored further in two ways. First, the roles of individual lysine residues in this region were examined. Mutations with single lysine residues in the absence of region 4 did not display any Ca 2ϩ -dependent binding activity (data not shown). Therefore, mutations containing one, two, or

FIG. 3. Gel analysis of wild type and selectin-like mutant of MBP purified on Man-and invertase-Sepharose.
Purification of wild type MBP (left) and mutants containing region 5 of E-selectin in the presence of Gly 189 (right) are shown. Starting material for each preparation was extract prepared from a 500-ml bacterial culture that was applied to 1-ml columns. Columns were eluted with 5 aliquots (0.5 ml) of eluting buffer. Aliquots (20 l) of fractions 2, 3, and 4 from each column were combined with an equal volume of 2 ϫ sample buffer (80 mM Tris-Cl, pH 6.8, 2% SDS, 15% glycerol, 2% 2-mercaptoethanol, and 0.006% bromphenol blue), heated to 100°C, and resolved on a 17.5% SDS-polyacrylamide gel (20) that was visualized with Coomassie Blue.  (Table II). The results indicate a cumulative effect of the lysine residues, with particularly the first and third residues capable of supporting some binding. The very limited effect of the middle residue may correlate with the disposition of this side chain furthest from the sialic acid portion of the bound ligand (Fig. 2). The HL-60 cell binding data suggest that the only combination of two residues which supports maximal binding is the second and third. However, the results using 125 I-sLe x -BSA as test ligand, which has relatively narrower error margins, suggest that any combination of two lysines supports full binding. The data indicate that no one residue is essential for binding activity and reveal a partial correlation of binding with total positive charge in this region. These findings are consistent with the structural results showing that none of the lysine side chains makes a specific, direct contact with the NeuAc portion of the ligand. Moreover, the effective binding achieved in the absence of Lys 211 suggests that the hydrogen bond between this residue and the galactose moiety is not an essential component of the binding interaction.
A second approach to understanding the role of the region 5 lysine residues was based on the observation that two of the corresponding side chains in E-selectin form hydrogen bonds with glutamic acid residues. A glutamic acid side chain at the position corresponding to Ile 207 in MBP hydrogen bonds to the lysine residue in region 5 corresponding to Lys 211 , whereas a glutamic acid side chain at the position of Thr 197 hydrogen bonds to the lysine residue corresponding to Lys 213 (Fig. 2). Because these bonds might stabilize the conformation of the region, the two glutamic acid residues were incorporated into the CRD in the presence of the region 5 lysine residues, glycine at position 189, and MBP, E-selectin, or P-selectin sequences in region 4. Incorporation of glutamic acid at position 197 results in a 2-3-fold increase in the binding of both HL-60 cells and sLe x -BSA (comparing the results in Tables I and III). Further incorporation of glutamic acid at position 207 actually has a deleterious effect. This last effect suggests distortion of the binding site by virtue of introduction of a negative charge into a position of MBP normally occupied by a hydrophobic isoleucine side chain.
In the presence of a glutamic acid residue at position 197, incorporation of region 4 of E-selectin (but not P-selectin) en-hances binding to both HL-60 cells and sLe x -BSA in the presence of His 189 but not Gly 189 . In the presence of glutamic acid residues at both positions 197 and 207, a similar enhancement by E-selectin region 4 is observed, although overall binding is much less in the Glu 207 -containing proteins. These effects suggest that region 4 of selectin can make a positive contribution to ligand binding but that this contribution is only evident when the binding site is in other respects suboptimal. Structural analysis will be necessary to determine if residues in   Effect of Salt on Binding-The fact that the NeuAc residue in sLe x does not make direct contact with lysine residues in region 5 and the apparent role of overall charge in this region of the CRD surface suggest that there may be a strong electrostatic component to the interaction between the selectin-like binding site and its ligand. To assess this possibility, the effect of salt on binding activity was evaluated. As shown in Fig. 6, the interaction of wild type CRD with Man-BSA is relatively insensitive to salt, whereas the interaction of a selectin-like mutant with sLe x -BSA is weakened drastically as the ionic strength is increased. At physiological salt concentration, roughly 40% of the activity in the absence of added NaCl is observed. Previous studies with L-selectin on cell surfaces revealed a very similar inhibition of binding, with roughly 40% of activity retained under physiological conditions (27). Similar effects, over a more limited concentration range, have been observed for the other selectins (28). Thus, in addition to providing evidence for the electrostatic contribution to binding of the selectin-like mutant, these studies also demonstrate further parallels between the activity of the mutant and natural selectins.
Role of Loop 3-Studies using the extracellular portion of E-selectin have suggested that residues in loop 3 of the CRD might affect ligand binding activity. Substitution of a lysine residue from this region of MBP into the E-selectin CRD has been proposed to introduce mannose binding activity into the domain (22). Therefore, it was of interest to examine the effect of changing the residues in this loop to resemble selectin sequences to see if they affect the relative affinity for Man-BSA and sLe x -BSA. In the case of mutants designed to resemble E-selectin either at the single position tested previously (residue 182 in MBP, corresponding to position 77 in E-selectin) or for all three residues in this loop which differ between MBP and E-selectin (positions 182-184), there was no detectable effect on relative binding to these two ligands (Table IV). It should be noted that the absolute value of the binding ratio has no direct meaning, as the ligands are not used at precisely the same concentration, and, more importantly, the degree of substitution with sugars is substantially different. Nevertheless, the results do not support a role for loop 3 residues in ligand discrimination.
A similar set of mutations designed to mimic P-selectin revealed an increase in relative affinity for Man-BSA when the entire loop was changed from the MBP to the P-selectin sequence (Table IV). This effect is the opposite of what would have been predicted from the previous E-selectin studies and is associated with the changes at positions 183 and 184, not 182. Although there is no simple explanation for these observations, they again do not support a role for the residues in this loop in determining ligand binding selectivity for P-selectin. DISCUSSION Mutational analysis of the selectins has revealed that changes at roughly 15 positions affect their ligand binding activity (4, 7-9, 22, 29). Several of the critical residues are Ca 2ϩ ligands, whereas many of the rest have been included in the current studies, which extend the analysis of loss-of-function mutations in the selectins. The changes introduced into the MBP framework have been analyzed for increased binding to selectin ligands and thus represent gain-of-function mutations. It is often easier to interpret such effects rather than loss of function because they are less likely to result from indirect alterations in CRD structure and stability. Similarly, the use of purified CRDs rather than proteins expressed on cell surfaces or as Fc chimeras, combined with use of sLe x -BSA ligand as well as HL-60 cells as target ligands, eliminates possible effects on cell surface expression and other confounding factors.
The fact that all of the mutants analyzed show Ca 2ϩ -dependent binding to either Man-Sepharose or invertase-Sepharose argues that they have achieved the correct overall conformation. Differential binding to invertase, sLe x -BSA, and HL-60 cells by the various mutants investigated therefore presumably reflects conformational differences near the modified binding sites. The lack of any detectable retardation of the glycinecontaining mutants of MBP on Man-Sepharose suggests that the affinity for invertase-Sepharose is not just a result of higher concentrations of mannose residues on these columns. Previous studies indicate that mannose phosphate residues in yeast polysaccharides are ligands for L-selectin (30), so similar structures in the invertase preparation might provide binding sites for the selectin-like mutants of MBP.
The phenotypes of mutant CRDs containing various portions of the E-and P-selectin CRDs suggest specific roles for several of these regions in ligand discrimination and stabilization of ligand-CRD complexes. Affinity for sLe x seems to be dependent primarily upon the presence of a sufficient number of lysine   (4,(7)(8)(9)29). These results are consistent with the findings reported in Table II.
In the crystal structure of the initial MBP mutant containing only the three lysines of region 5, the lysine side chains align closely with the arrangement seen in the natural E-selectin domain (4,12). Thus, although the addition of a glutamic acid residue that hydrogen bonds to one of these lysine residues results in increased binding activity, its role is likely to be in stabilization of the binding site rather than orientation of the lysine side chains. Mutagenesis studies with E-and P-selectin indicating that alanine substitutions for either of these residues results in at least partial loss of binding activity (4,9) are consistent with this suggestion.
The effects of region 4 are more subtle because the E-selectin sequence inserted in this region supports increased binding to HL-60 cells and sLe x -BSA only in the presence of a histidine residue at position 189. When present, this side chain has a potential role in the binding site because the imidazole ring makes contact with the N-acetylglucosamine portion of sLe x in the crystal structure of the region 5 mutant (31). Thus, the effect of the E-selectin sequence in the presence of His 189 may be caused by a slight reorientation of the ligand resulting from interactions with either the ␤-carbon or the imidazole ring, or it may reflect compensating interactions with the ligand which are only evident when the binding site has an initially weaker affinity.
In addition to providing insight into common aspects of ligand binding to the selectins, these studies provide some initial suggestion of why different selectins show preferential binding to different ligands. In particular, the differing behavior of mutants containing E-and P-selectin sequences in region 4 implicates this segment of the CRD in differential ligand binding. In the most selectin-like constructs, the effect can be viewed as a relative loss of affinity for the simple test ligand, sLe x -BSA, in the presence of the P-selectin rather than the E-selectin sequence (Table III). High affinity P-selectin ligands contain a tyrosine-sulfated region in addition to a specific set of sugar structures (32)(33)(34). Additional interactions with such a region, either through the CRD or the adjacent EGF-like domain, might compensate for lower affinity for sLe x reflected in the behavior of the mutants analyzed here.
Mutagenesis studies with E-selectin identified one additional portion of the protein, region 3, which might be involved in sLe x -BSA binding, although the phenotype resulting from changes in this segment were less severe than changes in regions 4 and 5 (4,7,22). Previous studies showed that incorporation of region 3 alone does not confer sLe x -BSA binding activity on MBP (11). Further studies in which regions 3, 4, and 5 of E-selectin were all incorporated into MBP failed to show any enhancement of sLe x -BSA binding compared with regions 4 and 5 alone. 2 These results and the relatively large distance between region 5 and sLe x in the crystal structure of the selectin-like mutant of MBP (Fig. 2) suggest that loss of binding caused by changes in region 3 of E-selectin represents an indirect effect on stability or on the arrangement of residues that are directly in the binding site.
The current results do not support a role for loop 3 of the CRD (region 6) in ligand binding selectivity suggested by previous studies with E-selectin (22). The earlier conclusions were based in part on the effect of mutations in this region on binding to yeast and to yeast invertase. The results reported here indicate that binding activity for commercial preparations of invertase is actually correlated with sLe x binding activity, a result that probably confounds the previous use of such material as a test for mannose binding activity.
The oligomeric state of the selectin molecules has not been documented clearly. However, many previous studies have utilized chimeric constructs containing immunoglobulin Fc domain fused to the COOH terminus of CRD and epidermal growth factor domains of the selectins. Such chimeras form either dimers, when the Fc from IgG is employed, or higher oligomers in the case of the Fc domain of IgM. All mutants employed in the present studies contain the neck region from MBP and are thus almost certainly trimers stabilized by the formation of a coiled coil of ␣-helices as observed in the crystal structures of both wild type MBP and the initial region 5 selectin-like mutant (12,35). Neither the Fc chimeras nor the MBP mutants are likely to reflect the relative arrangement of multiple CRDs in intact selectins at the cell surface. Therefore, the fact that the binding activities of Fc fusions and MBP chimeras closely parallel the activity of the membrane-bound molecules argues against the need for a specific geometry of interaction between an array of ligands and the multiple binding sites in oligomeric selectins. However, the way CRDs are presented may have more influence on transient interactions with ligands observed in cell rolling assays in contrast to static, equilibrium assays employed in the present studies.
Several modified versions of the CRD from MBP have been crystallized successfully in forms that are isomorphous with the wild type CRD, and the spacing in these crystals accommodates introduction of ligands by soaking (12,36). Thus, structural analysis of some of the mutants described here may prove feasible. The evidence presented showing that the binding activity of some of these mutants closely parallels the activity of the natural selectins suggests that such studies may provide a viable route to a detailed molecular understanding of ligand binding to the selectins.