Affinity and Kinetic Analysis of L-selectin (CD62L) Binding to Glycosylation-dependent Cell-adhesion Molecule-1*

The selectin family of cell adhesion molecules mediates the tethering and rolling of leukocytes on blood vessel endothelium. It has been postulated that the molecular basis of this highly dynamic adhesion is the low affinity and rapid kinetics of selectin interactions. However, affinity and kinetic analyses of monomeric selectins binding their natural ligands have not previously been reported. Leukocyte selectin (L-selectin, CD62L) binds preferentially to O-linked carbohydrates present on a small number of mucin-like glycoproteins, such as glycosylation-dependent cell adhesion molecule-1 (GlyCAM-1), expressed in high endothelial venules. GlyCAM-1 is a soluble secreted protein which, following binding to CD62L, stimulates β2-integrin-mediated adhesion of lymphocytes. Using surface plasmon resonance, we show that a soluble monomeric form of CD62L binds to purified immobilized GlyCAM-1 with a dissociation constant (K d ) of 108 μm. CD62L dissociates from GlyCAM-1 with a very fast dissociation rate constant (≥10 s−1) which agrees well with the reported dissociation rate constant of CD62L-mediated leukocyte tethers. The calculated association rate constant is ≥105 m −1 s−1. At concentrations just above its mean serum level (∼1.5 μg/ml or ∼30 nm), GlyCAM-1 binds multivalently to immobilized CD62L. It follows that soluble GlyCAM-1 may cross-link CD62L when it binds to cells, suggesting a mechanism for signal transduction.

It has been postulated that selectins are able to mediate tethering and rolling on vascular endothelium because they bind their ligands with very fast association and dissociation rate constants (10). However, the affinity and kinetics of selectin interactions with their physiological ligands remain poorly characterized. Selectins have been shown to bind synthetic oligosaccharides related to sialylated and/or sulfated Lewis x (Le x , galactose ␤134(fucose ␣133)(N-acetyl)glucosamine) or its stereoisomer Lewis a (Le a , galactose ␤133(fucose ␣134)(Nacetyl)glucosamine) with very low affinities (K d 0.1-5 mM (11)(12)(13)(14)(15)(16)(17)(18)(19)). It is possible that these studies underestimated the affinities because: (i) with few exceptions (13,18,19), they were based on inhibition by synthetic oligosaccharides of multivalent selectin-ligand interactions, and (ii) these oligosaccharides may differ in structure from physiological selectin ligands (7). Indeed soluble, recombinant forms of CD62P (20) and CD62E (21) have been reported to bind leukocytes with much higher affinities (K d Յ 1 M). However, the accuracy of the latter affinity measurements is also in doubt because the CD62E used was oligomeric (as assessed by size exclusion chromatography (21)), and the CD62P may have been contaminated by small amounts of multivalent CD62P (4).
In the present study we expressed a soluble, monomeric form of rat CD62L and used surface plasmon resonance to measure the monovalent affinity and kinetics of its interaction with native GlyCAM-1 purified from mouse serum. We show that CD62L binds immobilized GlyCAM-1 with a very low affinity (K d ϳ 105 M) and a very fast dissociation rate constant (k off Ն 10 s Ϫ1 ). In contrast, the binding of soluble GlyCAM-1 to immobilized CD62L is detectable at concentrations just above 1.5 g/ml (ϳ30 nM) and dissociates with a k off Ͻ 0.001 s Ϫ1 . Our results establish that CD62L binds its physiological ligand with a very low affinity and very fast kinetics. In addition, our data suggest that soluble GlyCAM-1 binds multivalently to, and therefore cross-links, cell surface CD62L, suggesting a mechanism for signal transduction.
GlyCAM-1 was purified from mouse serum as described (27). Mouse CD62L Ig was expressed in human kidney 293 cells and purified as described (36). Rat CD62L Ig was expressed in silkworm cells and was purified by protein A-Sepharose (Pharmacia Biotech AB, Uppsala, Sweden) affinity chromatography from silkworm hemolymph provided by Dr. M. Miyasaka (33).
Production of Recombinant Soluble CD62L-CD4 -DNA encoding the extracellular portion of rat CD62L was amplified by polymerase chain reaction from rat spleen cDNA. The 5Ј primer (5Ј-GCCCGCTCTA-GAACTTACAGAAGAGACC) was complementary to the 5Ј-untranslated region and added an XbaI site (underlined). The 3Ј-primer (5Ј-GAGAAAGTCGACTTTGTCTTTTGACATATTGG) was designed to include CD62L up to Lys-282 (numbered as the mature protein) and add a SalI site (underlined). To facilitate cloning, a silent mutation was introduced into the CD62L sequence (TCTAGA 3 TCTCGA) to remove an internal XbaI site. The CD62L fragment was ligated into the XbaI/ SalI sites of a previously described pBluescript vector containing an insert encoding domains 3 and 4 of rat CD4 (CD4d3ϩ4) (37). The resulting cDNA encoded the leader and most of the extracellular portion of CD62L fused at its carboxyl-terminal end to CD4d3ϩ4 (CD62L-CD4, Fig. 1A). The intervening SalI site introduced a Ser (underlined) at the junction between CD62L and CD4 (..QKTKSTSITA..). The DNA encoding CD62L-CD4 was excised with XbaI and BamHI, subcloned into expression vector pEE14 (38) using its XbaI and BclI restriction sites, and then checked by dideoxy sequencing. Chinese hamster ovary-K1 cells were transfected with the CD62L-CD4/pEE14 plasmid using calcium phosphate as described (38,39). Clones expressing high levels of CD62L-CD4 were identified as described (37) by inhibition enzymelinked immunosorbent assay, using the anti-CD4 mAb OX68 (40). The best clone (secreting 40 -60 mg/liter) was grown up to confluence in bulk culture before switching to serum-free medium supplemented with 2 mM sodium butyrate. The cultures were then left for a further 3-4 weeks prior to harvesting. CD62L-CD4 was purified from the spent tissue culture supernatant by affinity chromatography using OX68 coupled to Sepharose CL-4B (37), followed by size exclusion chromatography on a Superdex S200 HR10/30 column (Fig. 1C). The extinction coefficient (at 280 nm) of purified CD62L-CD4 was determined by amino acid analysis to be 1.87 cm 2 mg Ϫ1 . Briefly, the duplicate 20-l samples of CD62L-CD4 at an A 280 (path length 1 cm) of 0.46 were subjected to acid hydrolysis and the following amino acids were quantitated: Asp ϩ Asn, Glu ϩ Gln, His, Arg, Ala, Pro, Val, Leu, Phe, Lys. The amino acid composition was as expected from the primary sequence (data not shown). The extinction coefficient was calculated based on a CD62L-CD4 protein M r (excluding carbohydrate) of 51,999.
The proportion of purified CD62L-CD4 that bound OX85 was estimated as follows. Protein A-Sepharose beads (packed volume 100 l) were incubated with 1 mg of OX85 or the control antibody W3/25 (IgG 1 , binds domain 1 of rat CD4 (40,41)) in 200 l of Tris saline (150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM Tris (pH 7.5)) for 1 h at 4°C, with rotation, and then washed three times with 1.5 ml of Tris saline. OX85 or W3/25 beads (40 l) were added to 20 l of CD62L-CD4 (1 g/l in Tris saline) and incubated for 4 h at 4°C with rotation. The beads were then pelleted and the supernatants (8 l) analyzed by SDS-PAGE.
BIAcore Experiments-All binding experiments were performed at 25°C (unless otherwise indicated) on a BIAcore TM (BIAcore AB, Stevenage, Herts, United Kingdom) with Hepes-buffered saline as running buffer. Hepes-buffered saline comprised (in mM): NaCl 150, MgCl 2 1, CaCl 2 1, Hepes 10 (pH 7.4), and 0.005% surfactant P20. All directly immobilized proteins were covalently coupled to research grade CM5 sensor chips (BIAcore) via primary amine groups using the Amine Coupling Kit (BIAcore) as described (42) except that a flow rate of 10 l/min was used throughout. The purified mAbs OX68 and R10Z8E9 were coupled and regenerated as described previously (43,44). The polyclonal antibodies CAMO2 and CAMO5 were coupled by injecting them for 7 min at 60 g/ml in 10 mM sodium acetate (pH 5.0). Immobilized CAMO2 was regenerated with minimal loss of GlyCAM-1 binding activity by injecting 100 mM HCl over the surface for 3 min (data not shown).
Because GlyCAM-1 is not covalently coupled to the surface, it dissociates continuously, with the result that the amount of immobilized GlyCAM-1 decreases by 10 -20% between the first and last CD62L-CD4 injections (Fig. 3A). If this is not taken into account the K d determined for the CD62L-GlyCAM-1 interaction is artificially increased (K d ϳ 140 M) when proceeding from high to low CD62L-CD4 concentrations and decreased (K d ϳ 90 M) when proceeding from low to high concentrations (data not shown). Therefore the binding (CD62L bound ) at each CD62L-CD4 concentration was adjusted (CD62L adjusted ) for the level of GlyCAM-1 (GlyCAM injection ) on the surface immediately preceding that injection, using the formula, where GlyCAM initial is the level of immobilized GlyCAM-1 immediately preceding the first CD62L-CD4 injection. When this adjustment is made the same K d values are obtained irrespective of the order of CD62L-CD4 injections (Fig. 3D). Several lines of evidence suggest that the CD62L-CD4 is correctly folded. First, it bound to 3 previously described CD62L mAbs (HRL1, HRL2, and HRL3) (Fig. 2), two of which (HRL1 and HRL3) block binding of CD62L to its natural ligands (32,33). Second, CD62L-CD4 was used to raise a new mAb (OX85, see "Experimental Procedures"). OX85, in addition to binding CD62L-CD4 (Fig. 2), binds lymphocyte populations known to express CD62L (34,35) and to a well characterized (33) chimeric protein comprising rat CD62L fused to the Fc portion of human IgG 1 (CD62L Ig, data not shown). Third, CD62L-CD4-coated fluorescent microspheres bind selectively to high-endothelial venules (HEV) in lymph node sections (35). Fourth, CD62L-CD4 binds both porcine peripheral node addressin (PNAd, purified using the MECA-79 mAb) (35) and mouse GlyCAM-1 (see below). And finally, as expected for interactions involving C-type lectins, binding of CD62L-CD4 to HEV, PNAd, and GlyCAM-1 was inhibited by EDTA (see Ref. 35 and Table I).

Expression and Analysis of
Accurate affinity measurements require knowledge of the proportion of the CD62L-CD4 that is active with respect to ligand binding. It was not possible to measure directly ligand binding activity and so, as a surrogate, we used binding to mAbs, including two mAbs (HRL1 and HRL3) which block ligand binding (32,33). If one assumes bivalent binding, the mAbs HRL1, HRL3, and OX85 bind to Ն60% of CD62L-CD4 immobilized on the sensor surface (Fig. 2). Furthermore, protein A-Sepharose beads coated with OX85 depleted ϳ90% of CD62L-CD4 (Fig. 1B), despite the fact that OX85 dissociates rapidly from CD62L (Fig. 2). Since these data show that most of the recombinant CD62L-CD4 is correctly folded, the affinity measurements below assume ligand binding activity of 100%.
To determine whether CD62L-CD4 was monomeric it was analyzed by size exclusion chromatography (Fig. 1C). Using globular, unglycosylated proteins as calibration markers, CD62L-CD4 eluted at molecular mass ϳ 140,000 (Fig. 1C), which is higher than the molecular mass measured by SDS-PAGE (ϳ76 kDa, Fig. 1B). However, asymmetric glycosylated proteins such as CD62L-CD4 typically elute much earlier in size exclusion chromatography than predicted by their M r . For example, the asymmetric, glycosylated proteins sCD2, sCD80, and sCD48-CD4 (M r ϳ 30,000, 35,000, and 50,000 on SDS-PAGE), which are known to be monomeric in solution, elute at M r ϳ 52,000, 63,000, and 84,000 on the same column (45). Taken together, these data suggest that CD62L-CD4 exists as a monomer in solution. The monomeric peak of CD62L-CD4 (Fig. 1C) was used for affinity and kinetic measurements which were performed within 48 h of size exclusion chromatography to minimize the accumulation of multivalent aggregates (46).
Affinity of CD62L-CD4 Binding to GlyCAM-1-GlyCAM-1 purified from mouse serum was immobilized on the sensor surface indirectly using the rabbit polyclonal antibody CAMO2, which was raised against a peptide from the middle (nonmucin) region of GlyCAM-1 (see "Experimental Procedures"). When GlyCAM-1 is injected over a sensor surface to which CAMO2 had been covalently coupled, there is an increase in response, which indicates binding (Fig. 3A). Following the injection, while the GlyCAM-1 remains bound, a range of CD62L-CD4 concentrations are then injected briefly over this surface (Fig. 3A) and simultaneously injected over a control sensor surface with only CAMO2 (not shown). An expanded view of the response during injection of three concentrations of CD62L-CD4 over GlyCAM-1 reveals that the response attains equilibrium within seconds of the start of each injection and returns to baseline within seconds of the end of the injection (Fig. 3B). Because the BIAcore detects changes in refractive index, the high protein concentrations injected (up to 26 mg/ml or 0.5 mM) give a large background signal. This is evident when the response trace from the control surface is overlaid (Fig. 3B). The difference between the response seen with injection over the GlyCAM-1 surface compared with the response seen with injection over the control surface represents the actual binding of CD62L-CD4 to GlyCAM-1 (Fig. 3, B and C). Measured in this way, no binding is seen when CD62L-CD4 is injected in the presence of EDTA, or when the control CD4 chimera sCD48- CL, E, and C refer to C-type lectin, epidermal growth factor, and complement control protein superfamily domains, respectively (67). V and C2 refer to V-set and C2-set immunoglobulin superfamily domains (67). Predicted N-linked glycosylation sites are represented by filled circles. B, top: CD62L-CD4 (3 g) was analyzed by SDS-PAGE on a 12% acrylamide gel under reducing (ϩ␤-mercaptoethanol) and nonreducing conditions. Bottom: protein A-Sepharose beads coated either with OX85 or a control mAb (W3/25) were incubated with CD62L-CD4, pelleted, and the supernatants analyzed for the presence of CD62L-CD4 by reducing SDS-PAGE on 12% acrylamide. C, purification of CD62L-CD4 by size-exclusion chromatography. CD62L-CD4 (3 mg in 0.5 ml) was run on a Superdex S200 HR10/30 column (Pharmacia) at 0.5 ml/min in Hepes-buffered saline. The calibration markers shown (Sigma) were alcohol dehydrogenase (M r 150,000) and bovine serum albumin (M r 66,000). The indicated fractions (*) were combined, concentrated 10 -20-fold, and used within 48 h with storage at 4°C.

FIG. 2.
Binding of mAbs to CD62L-CD4. CD62L-CD4 was immobilized to the sensor surface by injecting it at 64 g/ml for 700 s (thick bar) over a sensor surface to which ϳ12,100 response units of the anti-CD4 mAb OX68 had been covalently coupled. The increase in the response during the CD62L-CD4 injection reflects binding of CD62L-CD4 to the sensor surface. MAbs were injected at the indicated concentration for 700 s (thin bars) both before and after the immobilization of CD62L-CD4 to the surface. The traces for each mAb are overlaid. No mAbs bound when injected before immobilization of CD62L-CD4, whereas the CD62L mAbs (HRL1, HRL2, HRL3, and OX85), but not the control mAb OX55 (IgG 1 , anti-rat CD2 (68)), bound to the immobilized CD62L-CD4. This experiment was performed at a flow rate of 3 l/min. CD4 (43) is injected, indicating that the binding involves the CD62L portion of CD62L-CD4 (Table I). Direct fitting of a standard Langmuir binding isotherm to the data indicates that the binding is saturable, with a K d of 105 M (Fig. 3C, inset). A Scatchard plot of the same data is linear and also gives a K d value of 105 M (Fig. 3D, closed circles). Provided that binding is adjusted to compensate for the slow dissociation of Gly-CAM-1 from the surface (see "Experimental Procedures"), the same K d is obtained when the order of CD62L-CD4 injections is reversed (Fig. 3D, open circles). These affinity measurements were highly reproducible (Table II). Interestingly, CD62L-CD4 bound with the same affinity at 25 and 37°C (Table II), consistent with a small enthalpic and a large entropic contribution to the binding energy over this temperature range. All subsequent measurements were performed at 25°C.
Recent evidence suggests that the binding of CD62L to PSGL-1 may involve the protein backbone as well as O-linked carbohydrates (9). A polyclonal antibody directed at an Nterminal PSGL-1 peptide inhibits CD62L binding (9). This raises the question as to whether the immobilization of Gly-CAM-1 via CAMO2 (which was raised to a peptide from the middle region of GlyCAM-1) (22) somehow diminishes CD62L-CD4 binding. To address this we studied CD62L-CD4 binding to GlyCAM-1 immobilized via the antibody CAMO5 (anti-peptide 3 antibody in Ref. 22), which was raised against a peptide from the carboxyl terminus of GlyCAM-1. CD62L-CD4 bound with the same affinity to CAMO5-and CAMO2-immobilized GlyCAM-1 (Table II), arguing strongly against any effect of GlyCAM-1 immobilization on CD62L-CD4 binding.
Kinetics of CD62L-CD4 Binding to GlyCAM-1-Following the injection of CD62L-CD4, the response dropped with a halftime of ϳ0.07 s (Fig. 4), which is similar to the time it takes to wash the sample out of the flow-cell at the flow-rate used (100 l/min) (47). This is confirmed by the observation that the background response (when CD62L-CD4 is injected through a control flow-cell) falls at the same rate (Fig. 4). Thus the rate at which the response falls represents the washing time rather than the intrinsic dissociation rate constant. Although the washing time can be decreased further by increasing the flowrate (up to a maximum of 500 l/min) (47), it would still not be possible to measure directly the dissociation rate constant because data cannot be collected on the current BIAcore at intervals shorter than 0.1 s. Nevertheless, it is possible to conclude from the available data that CD62L-CD4 dissociates from Gly-CAM-1 with a k off of at least 10 s Ϫ1 . Direct measurement of the association rate constant was not possible because equilibrium was reached within 1 s (Fig. 3B). However, with the k off Ն 10 s Ϫ1 and the K d ϳ 100 M the association rate constant (k on ) can be calculated to be Ն100,000 M Ϫ1 s Ϫ1 .

DISCUSSION
Accuracy of the Measurements-Accurate affinity measurements require that the recombinant CD62L-CD4 chimeric protein possesses the same ligand binding properties as native rat CD62L. We believe this to be the case for the following reasons. First, the chimera contains almost the entire extracellular portion of CD62L. Second, all four CD62L mAbs tested bound CD62L-CD4. Third, CD62L-CD4 bound selectively to HEV in lymph node sections (35). Fourth, the binding of CD62L-CD4 to GlyCAM-1, porcine PNAd, and lymph node HEV was inhibited by EDTA (35), and finally, GlyCAM-1 bound with a similar avidity to CD62L-CD4 as it bound to other well characterized and independently made CD62L proteins such as rat (33) and mouse (36) CD62L Ig (Fig. 5).
The affinity we obtained for the CD62L-CD4-GlyCAM-1 interaction could represent an underestimate if only a small proportion of the soluble CD62L-CD4 is correctly folded and able to bind GlyCAM-1. Our demonstration that Ն60% of the CD62L-CD4 retains mAb binding activity suggest that this possibility is very unlikely. One caveat is that we measured the affinity and kinetics of rat CD62L binding to mouse GlyCAM-1. However, because mouse GlyCAM-1 binds with a similar avidity to mouse and rat CD62L Ig (Fig. 5), we believe we are justified in assuming that mouse and rat CD62L bind mouse GlyCAM-1 with similar properties. This is not unexpected considering the high degree of conservation between mouse and rat CD62L (93% identity between C-type lectin domains (48)).
While migration of CD62L-CD4 on size exclusion chromatography was consistent with an asymmetric monomer, we could not rule out the possibility that it existed as a dimer. If CD62L-CD4 does indeed exists as a dimer, and binds divalently, our measurements would represent an underestimate of the K D and the k off . However, this would not alter the main conclusions of this study, which are that CD62L-CD4 binds to a physiological glycoprotein ligand with an exceptionally low affinity and fast kinetics, that this affinity is in agreement with the affinity measurements obtained for CD62L binding to sulfated forms of sialyl Lewis x, and that GlyCAM-1 is likely to bind multivalently to cell surface CD62L.
Comparison with Previous Studies on CD62L-To our knowledge this is the first affinity and kinetic analysis carried out in a cell-free system of the interaction between a selectin molecule and a defined physiological glycoprotein ligand (Table III). CD62L binding to GlyCAM-1 involves O-glycans which carry sialic acid, sulfate, and fucose groups (49 -51), consistent with the involvement of sulfated and sialylated derivatives of Le x or its stereoisomer Le a . The affinity of CD62L for synthetic forms of these oligosaccharides has been estimated by measuring the concentrations of soluble oligosaccharide required to inhibit by 50% (IC 50 ) multivalent CD62L-ligand interactions (Table III) (12-14, 19, 52). It is noteworthy that the IC 50 values for 6Јsulfo-sLe x and 6-sulfo-sLe x (Table III), two major capping groups present in GlyCAM-1 O-linked oligosaccharides (53,54), are only slightly higher (250 -800 M) than the K d meas-ured in the present study for CD62L binding to GlyCAM-1 (ϳ108 M). Because these inhibition studies relied on inhibition of multivalent interactions by monomeric oligosaccharides (12-14, 19, 52), the IC 50 values obtained are likely to underestimate the actual affinity. Thus, our results are consistent with the main CD62L ligands carried by GlyCAM-1 being 6Ј-sulfo-sLe x and 6-sulfo-sLe x (14,(52)(53)(54), or the branched and extended O-glycans in which these capping structures occur (54). The kinetics of CD62L interactions have been studied indirectly by analysis, in laminar flow, of transient leukocyte binding events (tethers) to planar surfaces coated with PNAd, a heterogenous mixture of CD62L ligands including CD34 (55). Since flow subjects these leukocytes to a shear force which increases the k off , the k off in the absence of an applied force ("intrinsic" k off ) was estimated by extrapolating to zero flow rate (55). Using this approach, Alon and colleagues (55) showed that these tethers are mediated by one or a few CD62L/PNAd bonds and detach with an intrinsic k off of ϳ7 s Ϫ1 , which agrees well with the solution k off for the CD62L/GlyCAM-1 interaction obtained in the present study (Ն10 s Ϫ1 ).
Comparison with CD62E and CD62P-Attempts have been made to measure the affinity of both CD62P and CD62E for   8,250 -8,450 RUs) of CAMO5 were covalently coupled to sensor surfaces (see "Experimental Procedures") and 350 -460 RUs of GlyCAM-1 were bound to these surfaces via these Abs.
b The affinity was measured by equilibrium binding as described in the legend to Fig. 3. physiological ligands present on leukocytes (20,21). Radiolabeled soluble recombinant monomeric CD62P has been reported to bind neutrophils and HL60 cells with an affinity (K d 70 nM) at least 3 orders of magnitude higher than the affinity we report for CD62L binding GlyCAM-1 (20). However, Ushiyama et al. (20) did not exclude the possibility that the CD62P preparation contained small amounts of multimeric material (4), and so it is possible that they overestimated the affinity (43, 46,56). Similarly, a soluble recombinant form of CD62E inhibited the binding of HL60 cells to immobilized CD62E with an IC 50 of ϳ1 M (21). However, size exclusion chromatography showed that the CD62E existed as a multimer in solution, suggesting that this study may also have overestimated the true affinity (21).
There are many published measurements of the affinity of CD62E and CD62P binding to sulfated and/or sialylated derivatives of Le x or Le a (Table III). The best of these oligosaccharide ligands have been reported to bind CD62E and CD62P with affinities of K d ϳ 107 and ϳ 220 M, respectively (Table III). Because of the discrepancy between the high affinities reported for CD62E and CD62P binding to cells and their low affinity for these ubiquitous oligosaccharides (Table III), it has been suggested that these selectins might bind to carbohydrate (7) and, perhaps, protein structures restricted to these physiological ligands. There is evidence that CD62E binds preferentially to tetraantenary N-linked carbohydrates with an unusual sialylated di-Le x on the one arm (57). This is consistent with the finding that the binding of CD62E to E-selectin ligand-1, a major glycoprotein ligand purified from myeloid cells, requires sialylated, fucosylated N-linked carbohydrates (58 -60). Optimal binding of CD62P to its ligand PSGL-1 (CD162) requires sulfation of tyrosine groups near the NH 2 terminus of PSGL-1, in addition to sialylated and fucosylated O-linked oligosaccharides (8).
The kinetics of CD62P-and CD62E-ligand interaction have been studied indirectly by analysis of transient leukocyte tethers to CD62E and CD62P immobilized onto planar surfaces (55,61). The intrinsic k off for CD62P-and CD62E-mediated tethers was ϳ1 s Ϫ1 and ϳ0.7 s Ϫ1 , respectively (55,61). These values are ϳ10-fold slower than the intrinsic k off of CD62L-mediated tethers (55) and also Ն10-fold slower than the k off reported in the present study for the CD62L-GlyCAM-1 interaction (Table  III). Taken together, these data suggest that CD62E and CD62P interact with their respective physiological ligands with higher affinities and slower dissociation rate constants than CD62L (Table III). These differences may contribute to the slower kinetics of CD62L-versus CD62P-/CD62E-mediated leukocyte tethering and rolling (see below).
Implications for Adhesion-Since GlyCAM-1 is a soluble secreted molecule (25,26), it could be argued that affinity and kinetic data for the CD62L/GlyCAM-1 interaction do not have direct implications for understanding leukocyte-endothelium interactions. However, it has recently been shown that lymphocytes and neutrophils can tether and roll on surfaces coated with GlyCAM-1. 2 Furthermore, it seems likely that the carbohydrate structures on GlyCAM-1, MAdCAM-1, and CD34 to which CD62L binds are very similar, if not identical. First, the CD62L-binding glycoforms of all three of these mucin-like molecules are expressed by the same cell type, namely high-endothelial cells (22)(23)(24). Second, the O-glycans on both CD34 and GlyCAM-1 contain sulfate, sialic acid, and fucose (49). Finally, the binding of CD62L to both CD34 and GlyCAM-1 has been shown to require sialylation and sulfation (49 -51, 62).
Since selectins seem to have evolved to mediate highly dynamic leukocyte-endothelial interactions such as tethering and rolling, there has been speculation as to what properties of selectins facilitate these interactions. One suggestion has been that selectins are effective because they bind their carbohydrate ligands with exceptionally fast association and dissociation rate constants (10). Consistent with this hypothesis, we show that the k on and k off values for the CD62L/GlyCAM-1 interaction are Ն10 5 M Ϫ1 s Ϫ1 and Ն10 s Ϫ1 , respectively. However, kinetic studies of other cell-cell recognition molecules, which are not known to mediate tethering and/or rolling, have revealed that rapid binding kinetics may be a general feature of the molecular interactions mediating cell-cell recognition (43, 45,47). For example, the ligand/receptor pairs CD2/CD58 (47) and CD28/CD80 (45) have k on values of Ն4 ϫ 10 5 and Ն6 ϫ 10 5 M Ϫ1 s Ϫ1 and k off values of Ն4 and Ն1.5 s Ϫ1 . This suggests that fast binding constants, although perhaps necessary, are not sufficient for tethering and rolling. It should be emphasized, however, that fast association rates can be achieved both by fast association rate constants (k on ) and by high surface densi- FIG. 5. Binding of soluble GlyCAM-1 to immobilized rat and mouse CD62L chimeras. The indicated GlyCAM-1 concentrations were injected (at 1 l/min) simultaneously (in multichannel mode) over sensor surfaces to which CTLA-4 Ig (1540 response units), mouse CD62L Ig (1550 response units), rat CD62L Ig (1629 response units), or rat CD2L-CD4 (1633 response units) had been immobilized. The Ig chimeras were immobilized via the anti-Ig mAb R10Z8E9 (ϳ4,200 response units), whereas CD62L-CD4 was immobilized via the anti-CD4 mAb OX68 (ϳ8,800 response units). The gradual decrease in the CD62L-CD4 baseline during the experiment is the result of dissociation of CD62L-CD4 from the sensor surface. There is a small background response during each injection but, in the absence of binding, the response returns to baseline at the end of each injection (see CTLA-4 Ig flow cell). The dotted lines show the baseline responses expected after injection of 1.7 g/ml GlyCAM-1, illustrating that there is binding of GlyCAM-1 to rat CD62L-CD4 and mouse CD62L Ig. ties of one or both interacting molecules. Williams (63) proposed that selectins might achieve fast association rates because their oligosaccharide ligands are presented on mucin-like molecules at very high densities. Subsequently all selectin ligands identified, with the exception of E-selectin ligand-1, have been mucin-like molecules (8). One property clearly important for selectin-mediated tethering and rolling is the localization of selectins (e.g. CD62L) or their ligands (e.g. PSGL-1) to the tips of microvilli on leukocytes (64,65). Interestingly the ␣4 integrins, which have recently been shown to be capable of mediating tethering and rolling of leukocytes on endothelium, are also apparently localized to the tips of microvilli (2).
The good agreement between the intrinsic k off of CD62Lmediated leukocyte tethers (55) and the solution k off of the CD62L/GlyCAM-1 interaction (Table III), suggests that the duration of leukocyte tethers is dominated by the k off of the underlying molecular interaction. Furthermore, there is an excellent correlation between the k off of CD62L-, CD62P-, and CD62E-mediated tethers and the velocity of CD62L-, CD62P-, and CD62E-mediated leukocyte rolling (55,66). Taken together these results are consistent with the hypothesis that the k off of selectin-ligand interactions has a major influence on the duration of leukocyte tethers and the velocity of leukocyte rolling. Analysis of the solution kinetics of CD62E-and CD62P-ligand interactions will provide a critical test of this hypothesis.
One or more CD62L-binding protein(s) present in normal mouse serum can partially inhibit adhesion of lymphocytes to HEV in a Stamper-Woodruff assay (25). Our finding that Gly-CAM-1 does indeed bind to CD62L at concentrations just above its mean serum level suggests that it may inhibit CD62Lmediated adhesion in vivo (25,28), but direct evidence for such a role is lacking.
Implications for Signaling-GlyCAM-1 is present in mouse serum at a concentration of ϳ1.5 g/ml (30 nM) (27). It has recently been reported that murine GlyCAM-1 at concentrations as low as 2.5 g/ml can stimulate ␤ 2 -integrin-mediated adhesion of naive human peripheral blood lymphocytes to ICAM-1 (CD54), and that antibodies to human CD62L block this effect, suggesting that the GlyCAM-1 acts by binding to CD62L (31). Consistent with this, we report here that Gly-CAM-1 binds to purified immobilized CD62L at concentrations as low as 1.7 g/ml (ϳ34 nM). Since monovalent CD62L-Gly-CAM-1 interaction has an affinity of 108 M, this result shows that at these low concentrations GlyCAM-1 must bind multivalently to the immobilized CD62L. It follows that GlyCAM-1 binds to preclustered CD62L and/or that it induces clustering of CD62L when it binds to the cell surface. Taken together with the observation that antibody-induced cross-linking of CD62L activates ␤ 2 -integrin-mediated adhesion (29,31), these results suggest that GlyCAM-1 activates lymphocyte adhesion by cross-linking CD62L.
In principle, GlyCAM-1 may bind multivalently either because it self-associates to form multimers or because each GlyCAM-1 molecule carries multiple copies of the CD62L binding carbohydrate structure(s). However, the elution position of purified GlyCAM-1 on size exclusion chromatography (M r 45,000 -66,000) agrees with its M r determined by SDS-PAGE, 3 arguing strongly against a multimeric form of GlyCAM-1. Instead we favor the explanation that each molecule of GlyCAM-1 carries multiple CD62L-binding carbohydrate structures, although there is no direct evidence to support this.
In conclusion, in the first affinity and kinetic study of the interaction between a selectin and a defined physiological ligand, we have shown that CD62L binds to GlyCAM-1 with a very low affinity (K d 108 M) and very fast kinetics (k off Ն 10 s Ϫ1 ). We have also provided evidence that, at concentrations just above the level at which it is present in serum, soluble GlyCAM-1 is able to bind multivalently to immobilized CD62L, suggesting a potential mechanism for signaling through CD62L.