In Vivo Ligand Specificity of E-selectin Binding to Multivalent Sialyl Lewisx N-linked Oligosaccharides*

The in vivo specificity for E-selectin binding to a panel of N-linked oligosaccharides containing a clustered array of one to four sialyl Lewisx (SLex; NeuAcα2–3Gal[Fucα1–3]β1–4GlcNAc) determinants was studied in mice. Following intraperitoneal dosing with lipopolysaccharide, radioiodinated tyrosinamide N-linked oligosaccharides were dosed i.v. and analyzed for their pharmacokinetics and biodistribution. Specific targeting was determined from the degree of SLexoligosaccharide targeting relative to a sialyl oligosaccharide control. Oligosaccharides targeted the kidney with the greatest selectivity after a 4-h induction period following lipopolysaccharide dosing. Unique pharmacokinetic profiles were identified for SLexbiantennary and triantennary oligosaccharides but not for monovalent and tetraantennary SLex oligosaccharides or sialyl oligosaccharide controls. Biodistribution studies established that both SLex biantennary and triantennary oligosaccharides distributed to the kidney with 2–3-fold selectivity over sialyl oligosaccharide controls, whereas monovalent and tetraantennary SLex oligosaccharides failed to mediate specific kidney targeting. Simultaneous dosing of SLex biantennary or triantennary oligosaccharide with a mouse anti-E-selectin monoclonal antibody blocked kidney targeting, whereas co-administration with anti-P-selectin monoclonal antibody did not significantly block kidney targeting. The results suggest that SLex biantennary and triantennary are N-linked oligosaccharide ligands for E-selectin and implicate E-selectin as a bivalent receptor in the murine kidney endothelium.

The regulation of leukocyte trafficking in response to a microbial insult or injury is controlled by a sequential process of cell to cell interactions and signaling events (1,2). The first critical step of this process is regulated by the selectins, which are a family of cell adhesion molecules composed of three members, E-(expressed on the endothelium), P-(endothelium and platelet), and L-selectin (leukocyte) (3,4). Each selectin is composed of a type I monomeric glycoprotein with a functional C-type lectin domain at the amino terminus followed by a single EGF-type domain, a variable number of short consensus repeats, a transmembrane domain, and a cytoplasmic tail (4).
The selectins function in mediating the initial cell-to-cell contact between leukocytes and the vascular endothelium during an inflammatory response through binding of their lectin domain to glycoconjugate counterligands (1,4). The selectin counterligands are proposed to be glycoproteins expressing sialylated, fucosylated, and/or sulfated oligosaccharides (5)(6)(7). The sialyl Lewis x (SLe x ) 1 tetrasaccharide acts as a ligand for all three selectins. However, numerous studies have established that binding of monovalent SLe x to E-, P-, and L-selectin is very weak (K d , mM) relative to the affinity (K d , nM) required to mediate leukocyte rolling (8,9). To account for this discrepancy in binding affinity, it has been hypothesized that the selectins, like other C-type lectins (10,11), bind with high affinity to natural ligands possessing a clustered array of sugar determinants (12).
In support of this hypothesis, the natural ligand of P-selectin (PSGL-1) contains a clustered array of SLe x O-linked oligosaccharides necessary for high affinity binding (13)(14)(15). Detergentsolubilized P-selectin binds PSGL-1 on HL-60 cells with a K d of 70 nM (8). L-selectin binds a tetravalent SLe x O-linked oligosaccharide with greater affinity than monovalent or divalent oligosaccharides as determined by a Stamper-Woodruff assay (16).
Likewise, a body of evidence suggests that E-selectin binds to a clustered array of SLe x determinants presented on N-linked oligosaccharides. Confocal laser scanning microscopy determined that E-selectin clusters on activated human umbilical venule endothelial cells in response to binding to HL-60 cells (17). A proposed natural ligand for E-selectin (ESL-1) requires its N-glycosylation for high affinity binding to the lectin (18,19). Pooled SLe x N-linked glycopeptides have been used to block E-selectin-mediated inflammation in vivo (20). Also, a bivalent SLe x N-linked oligosaccharide was shown to have a higher affinity for binding E-selectin in vitro relative to monovalent ligands (21,22).
In the present study we used a panel of multivalent SLe x N-linked oligosaccharide ligands to examine the ligand specificity of E-selectin induced in LPS mice. Using radiolabeled monoclonal antibodies, earlier studies determined LPS induces the expression of E-and P-selectin in all major tissues in mice (23,24). Using radiolabeled SLe x N-linked oligosaccharides as ligand probes, we report the kidney as the major target site with 2-3-fold selectivity over controls and an unexpected ligand cluster effect that implicates E-selectin as a divalent lectin that accepts a SLe x biantennary oligosaccharide as its minimal N-linked oligosaccharide ligand. MATERIALS (30 -35 g) and BALB/c mice (22-25 g) were purchased from Harlan, Indianapolis, IN and housed in cages located in a limited access area maintaining a 12 h light-dark cycle and controlled temperature (26 -28°C). Rat anti-mouse E-selectin mAb (IgG2a 10E9.6) was purchased from PharMigen, Los Angeles, CA. Hamster anti-mouse P-selectin mAb (C104) was a generous gift from Dr. Peter Ward, University of Michigan.
Radiolabeling of N-linked Oligosaccharides-SLe x and sialylated Nlinked oligosaccharides were synthesized as their t-butoxycarbonyltyrosinamide derivative as described previously (25). Oligosaccharides were radioiodinated at the tyrosine aglycone using a modification of the chloramine-T method as described previously (26). The purity of each iodinated oligosaccharide was analyzed by spotting 1 l (10 nCi) at the origin of a TLC plate developed with ethyl acetate/acetic acid/pyridine/ water at a ratio optimized for each oligosaccharide (see Fig. 1 for structures; I and I, 2:1:1:1; II and II, 3:3:2:2; III and III, 2.5:2.8:2:2; IV and IV, 2:2:2:2) and detected by autoradiography on a Phosphor Imager (Molecular Dynamics, Sunnyvale, CA).
Biodistribution Analysis of SLe x Oligosaccharides-ICR mice were dosed intraperitoneal with 0 -200 g of LPS in 200 l of phosphatebuffered saline. At times ranging from 1-6 h, mice were anesthetized by intraperitoneal injection of ketamine hydrochloride (100 mg/kg) and xylazine hydrochloride (10 mg/kg) followed by insertion of a single cannula into the right jugular vein. SLe x or sialyl oligosaccharides (15 l, 2 Ci in saline) were dosed i.v. and allowed to biodistribute for 6 -45 min after which mice were sacrificed by cervical dislocation. The major organs (liver, lung, spleen, stomach, kidney, heart, large intestine, and small intestine) were harvested, rinsed with saline, and measured by direct ␥-counting for total radioactivity (26). Biodistribution was also studied in LPS mice dosed i.v. with 7 Ci (50 l) of oligosaccharide. After a 15 min biodistribution period, mice were euthanized by a lethal injection of phenobarbital (100 mg/kg) and were subjected to wholebody autoradiography analysis as described previously (26).
Pharmacokinetic Analysis of Oligosaccharides-The pharmacokinetic analysis of SLe x and sialyl oligosaccharides was performed in triplicate in both normal and LPS mice. Mice were dosed intraperitoneal with LPS (20 g in 200 l of phosphate-buffered saline), and after 4 h a dual jugular vein cannulation was performed. Oligosaccharides were dosed in the left vein while blood time points were taken at 1, 3, 6, 10, 15, 20, 30, 40, and 60 min from the right vein. Serial blood time points were analyzed by direct ␥-counting, after which the oligosaccharide was extracted from blood by adding 60 l of water and 200 l of acetonitrile. Proteins were precipitated by centrifugation for 10 min (13,000 ϫ g), and the pellet was washed twice with 50 l of 80% v/v acetonitrile, resulting in an 80% recovery of the radioactivity. Extracts were combined and evaporated to dryness on a Centra-vap under reduced pressure and reconstituted in 3 l of water. Each time point was analyzed by spotting 1 l onto a TLC plate that was developed and autoradiographed as described above.
Pharmacokinetic parameters were derived from direct blood counts versus time for triplicate data sets of each oligosaccharide and averaged to obtain the mean and standard deviation (26). Iterative nonlinear least squares fits for individual data sets were obtained with PCNON-LIN (SCI Software, Lexington, KY) using a two-compartment open model described by the integrated Equation 1: where C b is the concentration of oligosaccharide in blood. A and B are constants, and ␣ and ␤ are hybrid first-order rate constants that characterize the slopes of the fast and slow phases of decline in a plasma concentration versus time profile (27). The mean residence time (MRT) was calculated according to Equation 2: which is the average time that the oligosaccharide was in the mouse (28). The total body clearance (Cl tb ) was calculated using Equation 3: and the volume of distribution at steady-state (Vd ss ) was calculated according to Equation 4 (29).
Inhibition of Oligosaccharide Targeting-Inhibition experiments were performed in Balb/c mice dosed i.p. with 100 g of LPS. After 4 h, mice were cannulated and dosed i.v. with 200 l of phosphate-buffered saline containing 2 Ci of oligosaccharide and 40 g of anti-E-selectin or 40 g of anti-P-selectin mAb. Following a 15 min biodistribution period, mice were euthanized by cervical dislocation; the organs were harvested and ␥ counted as described above.

RESULTS
Selectin targeting was studied using a panel of N-linked oligosaccharides that contained a clustered array of one to four SLe x determinants (Fig. 1, structures I-IV). Sialyl oligosaccharides I-IV were chosen to control for specific targeting because they possess a similar size and charge but lacked the full SLe x tetrasaccharide required for selectin binding (Fig. 1). Each oligosaccharide was radioiodinated at the t-butoxycarbonyltyrosine aglycone, resulting in probes with a specific activity of 125 Ci/nmol. All doses of oligosaccharide were 16 pmol (2 Ci), thereby avoiding the potential saturation of the receptor activity.
LPS was chosen as an immunostimulant because it was previously shown to systemically induce E-selectin expression (30). The optimum selectivity in SLe x oligosaccharide III to sialyl oligosaccharide III tissue targeting ratio was determined by varying the LPS dose, induction period, and oligosaccharide biodistribution time.
Initially, the biodistribution of oligosaccharide III and III was determined at 30 min following an LPS dose of 0 -200 g and a 4 h induction period. The kidney was the major target site for both III and III, which demonstrated an LPS dose-dependent targeting efficiency (% of dose in the target organ) that ranged from 2 to 16% (Fig. 2). The oligosaccharide targeting to other organs was unaffected by the LPS dose, whereas the kidney targeting was highest at an LPS dose of 200 g but lacked selectivity for III over III. The greatest selectivity in kidney targeting occurred at an LPS dose of 20 g, which produced a targeting efficiency of 12% for III compared with 6% for III (Fig. 2). Therefore, a 20-g dose of LPS was used throughout the rest of the study in ICR mice.
At an LPS dose of 20 g and induction time of 4 h, oligosaccharide biodistribution time was analyzed at times ranging from 6 to 45 min. The kidney targeting efficiency for III was 10% at 6 min, which increased to 18% at 15 min, followed by a gradual decline to 3% at 45 min (Fig. 3A, dashed line). In contrast, the kidney targeting efficiency for III in normal mice was 10% at 6 min and then declined to 4% or lower over 45 min (Fig. 3A, solid line). Likewise, control oligosaccharide III produced a maximal targeting efficiency of either 10 (LPS) or 7% (normal) at 6 min, which slowly declined to 5% over 45 min (Fig. 3B). The liver was the second most active targeted site for both III and III with 3% targeting efficiency at 6 min, which declined to 1% at 45 min (Fig. 3, C and D). Consequently, a biodistribution time of 15 min was used throughout the rest of the study because it afforded the greatest kidney targeting selectivity between III and III.
The kidney targeting efficiency for III and III was determined at an LPS dose of 20 g and a biodistribution time of 15 min while varying the induction time from 0 to 6 h. Oligosaccharide III targeted the kidney at approximately 4 -7% during the first 3 h, after which the targeting efficiency increased to 18% at 4 -5 h and then declined to 12% at 6 h (Fig. 4). In contrast, oligosaccharide III targeted the kidney with 4 -7% efficiency during the first 3 h, but failed to increase at induction times of 4 h or longer (Fig. 4). Based on these results, an induction period of 4 h was used throughout the rest of the study.
Using the parameters derived above, the pharmacokinetic profiles for oligosaccharides I-IV were compared with I-IV in normal and LPS-dosed mice. Pharmacokinetic analysis revealed that each oligosaccharide showed a biexponential decline in blood after i.v. dosing, which was fit by an open twocompartment model with elimination from the central compartment. In normal mice, all oligosaccharides rapidly distributed and showed a MRT under 126 min (Table I).
With the exception of II and III, the pharmacokinetic halflife of each SLe x and sialyl oligosaccharide was significantly increased in LPS mice versus normal mice (Table I). Oligosaccharide II and III demonstrated a significantly shorter half-life than sialyl oligosaccharides II and III in LPS mice (Fig. 5), consistent with the shorter half-life of oligosaccharides that target the asialoglycoprotein receptor (26,31).
Comparison of the kidney targeting efficiency for I-IV in normal and LPS mice provided evidence of receptor binding specificity for II and III (Fig. 6). Analysis of I and I in normal mice established that approximately 5% of the dose was recovered in the kidney after a 15 min biodistribution period (Fig.  6A). In LPS mice, I displayed a slight increase in the kidney targeting (6%) compared with the negligible change observed for I. These results were supported by whole-body autoradiography analysis of I in LPS mice, which indicated only nonspecific targeting to the kidney (Fig. 6E).
By comparison, the kidney targeting efficiency of II was dramatically increased in LPS mice. The whole-body autoradiography predicted a very high kidney targeting efficiency (Fig. 6F), which was supported by the finding that 20% of II targeted the kidney in LPS mice compared with only 4% in the normal mice (Fig. 6B). In contrast, the kidney targeting of control oligosaccharide II was only moderately increased in LPS mice (9%) relative to normal mice (4%) establishing a significance (p Ͻ 0.05) for II/II kidney targeting in LPS mice.
Whole-body autoradiographic analysis established a similar biodistribution profile for III in LPS mice (Fig. 6G), confirming the kidney as the major target site. The kidney targeting efficiency was 18% for III in LPS mice compared with 4% in normal mice. The targeting efficiency for III increased only slightly from 4 up to 7% upon dosing with LPS, which established a significance (p Ͻ 0.05) for III/III kidney targeting selectivity (Fig. 6C).
In contrast to the results presented above, biodistribution analysis of oligosaccharides IV and IV failed to establish specific targeting to the kidney. In LPS mice, 10% of IV targeted the kidney relative to 4% determined in normal mice (Fig. 6D). However, control experiments with IV also demonstrated kidney targeting of 12% in LPS mice and 4% in normal mice, indicating no selectivity in the kidney targeting for oligosaccharides IV and IV in LPS mice (Fig. 6D).
To establish if the specific kidney targeting determined for II and III was E-or P-selectin-dependent, blocking mAbs were co-administered with II and II, III and III, and IV and IV to  inhibit kidney targeting. Balb/c mice were used during these experiments because they reportedly produce higher levels of E-selectin when stimulated with LPS (32). However, we found it necessary to dose with 100 g of LPS to stimulate specific targeting in Balb/c mice compared with 20 g in ICR mice. Under these conditions, a kidney targeting efficiency of 17% was determined for II in LPS mice compared with 5% for II (Fig. 7A). Simultaneous dosing of II with 40 g of anti-Eselectin mAb blocked the kidney targeting to 8% (p Ͻ 0.005), whereas co-administration of II with 40 g of anti-P-selectin mAb only reduced the kidney targeting to 15%. Co-administration of 40 g of anti-E-or anti-P-selectin mAb with II failed to inhibit the kidney targeting.
Analysis of III in LPS Balb/c mice established a targeting efficiency of 21% relative to 10% for III (Fig. 7B). Co-administration of III with 40 g of anti-E-selectin mAb inhibited kidney targeting to 6% (p Ͻ 0.001), whereas the same dose of anti-P-selectin mAb only caused inhibition to 15%. Likewise, co-administration of either antibody with III failed to inhibit kidney targeting (Fig. 7B).
Comparison of oligosaccharide IV and IV in Balb/c mice established that each targeted the kidney at 3% in normal mice, which increased to 11% in LPS mice (Fig. 7C). In contrast to the inhibition results presented for oligosaccharide II and III, co-administration of 40 g of anti-E-selectin or anti-Pselectin mAb failed to inhibit the kidney targeting of either IV or IV (Fig. 7C). DISCUSSION Since the discovery of the selectins, rapid progress has been made toward unraveling their role in leukocyte trafficking and identifying their glycoprotein ligands. In the case of P-and L-selectin, glycoprotein ligands have been elucidated that possess multiple O-linked oligosaccharides terminating in SLe x or sulfo-Le x glycans (33)(34)(35). Alternatively, a proposed ligand for E-selectin (ESL-1) loses its binding activity upon treatment with N-glycosidase F, implicating the involvement of N-linked oligosaccharides as ligands for E-selectin (18). However, it remains unclear how the ligand valency for E-, P-, and L-selectin relates to the oligomeric state of the receptors on the cell  Table I. B and C illustrate a shorter half-life for II and III relative to II and III, whereas A and D demonstrate no distinction in the halflives of oligosaccharides I and IV relative to I and IV controls. surface. The present study attempts to further elucidate the N-linked oligosaccharide ligand specificity for E-selectin and to shed light on the oligomeric state of the receptor on the cell surface.
The data presented indicate the kidney is the major target site for both SLe x and sialyl oligosaccharide in LPS and normal mice (Fig. 2). This result can be rationalized because i.v. dosed N-linked oligosaccharides are rapidly cleared by renal filtration and recovered from the kidney in the range of 3-5% (26,31). LPS dosing has been shown to slow the glomerular filtration rate (36,37), which also results in an elevation in the oligosaccharide concentration recovered from the kidney. However, the finding that a 20-g dose of LPS produced a greater kidney targeting for SLe x triantennary III relative to sialyl triantennary III (Fig. 2) provided early evidence of selective receptor binding of III. This selectivity was then further enhanced at an optimized biodistribution time of 15 min (Fig. 3).
Several lines of experimental evidence established that the 17-20% kidney targeting efficiency determined for SLe x oligosaccharides II and III in LPS mice was due to specific receptor binding (Fig. 6, B and C). Sialyl oligosaccharide II and III only target the kidney with 9% or lower efficiency demonstrating the requirement for subterminal fucose (Fig. 6, B and C). Only oligosaccharides II and III possessed a shorter pharmacokinetic half-life (Fig. 5, B and C), a result analogous to that reported for N-linked oligosaccharides that target the asialoglycoprotein receptor (26,31). Likewise, the lack of specific targeting determined for monovalent SLe x (I) and SLe x tetraantennary oligosaccharide (IV) demonstrates a requirement for either bi-or trivalent SLe x N-linked oligosaccharides (Figs. 6D and 7C).
Two experimental results suggest that the targeting activity is primarily related to E-selectin binding. First, antibody inhibition experiments demonstrated that the majority of the kidney targeting of oligosaccharides II and III can be blocked with anti-E-selectin mAb but not with anti-P-selectin mAb (Fig. 7, A  and B). Both antibodies failed to block the kidney targeting of sialyl oligosaccharide controls as well as the SLe x tetraantennary oligosaccharide. Second, the kidney targeting induction profile of 4 -5 h closely matches the profile of E-selectin mRNA induction reported previously in LPS mice (Fig. 4) (30). Together these data suggest that SLe x biantennary and triantennary target E-selectin on the kidney endothelium.
The major structural difference between oligosaccharides II and III relative to IV is the valency of their terminal SLe x . Other C-type lectins recognize clustered arrays of oligosaccharides through a multivalent interaction leading to an enormous increase in binding affinity with increasing valency (38,39). In the case of the asialoglycoprotein receptor, a triantennary Nlinked oligosaccharide mediates a 57% liver targeting efficiency, whereas a biantennary oligosaccharide only targets at 6% efficiency because of suboccupancy of the three available galactose binding pockets (26,31,40).
A similar interpretation of the kidney targeting data presented here suggests that E-selectin binds optimally to a biantennary oligosaccharide because nearly equivalent targeting efficiencies were determined for II and III, suggesting that the third antenna possessing SLe x on III fails to contribute to the binding affinity. The lack of specific E-selectin targeting observed for IV indicates this is not a simple multivalency effect because one would expect an increased, not a decreased, affinity if the receptor could oligomerize to accept all four SLe x determinants. Rather, the inability of E-selectin to bind IV may be due to steric hinderance that prevents even two of the four SLe x epitopes from binding.
This unexpected multivalent ligand effect leads to the hy-pothesis that E-selectin is dimeric, possessing two carbohydrate recognition domains that must position binding sites to simultaneously accept two SLe x ligands. These are apparently closely spaced by no more than 25-30 Å on the cell surface, because this is the maximum distance spanned by the antennae on an N-linked oligosaccharide (40). Further studies are needed to support this hypothesis by attempting to cross-linking E-selectin subunits on the cell surface and by determining the exact chemical nature of the N-linked oligosaccharides on ESL-1. But it is intriguing to consider that the oligomeric state of E-, P-, and L-selectin may each be unique, providing different selectivity and binding affinity for ligands possessing multivalent SLe x .