Abstract
The leukocyte integrin α4β1 is a receptor for both cell surface ligands (VCAM-1 and mucosal addressin cell adhesion molecule-1 (MAdCAM-1)) and extracellular matrix components (fibronectin). Through regulated interactions with these molecules, α4β1 mediates leukocyte migration from the vasculature at sites of inflammation. Regulation of integrin activity plays a key role in controlling leukocyte-adhesive events and appears to be partly determined by changes in integrin conformation. Several mAbs that recognize ligand-induced binding site epitopes on integrins have been characterized, and a subset of these mAbs are capable of stimulating integrin-ligand binding. Conversely, some mAbs recognize epitopes that are attenuated by ligand engagement and allosterically inhibit ligand binding. To gain insight into ligand-specific effects on integrin conformation, we have examined the ability of different ligands to modulate the binding of four distinct classes (A, B1, B2, and C) of anti-α4 Abs to α4β1. VCAM-1 attenuated B (antifunctional) class epitopes via an allosteric mechanism and also allosterically inhibited the binding of the function-blocking anti-β1 mAb 13. Additional α4β1 ligands (fibronectin fragments, MAdCAM-1, and the CS1 peptide) also inhibited mAb 13-integrin binding; however, the epitopes of the class B anti-α4 mAbs were attenuated by the fibronectin fragments, but not by MAdCAM-1 or the CS1 peptide. Of the two anti-α4 class A mAbs examined, one recognized an epitope that was induced uniquely by VCAM-1. Taken together, these data suggest that overlapping but distinct binding mechanisms exist for different α4β1 ligands and that distinct conformational changes are induced upon integrin engagement by different ligands.
The ability of cells to bind to, interact with, and migrate relative to their immediate environment is of central importance for processes such as development, wound healing, and immune responses. These events are mediated by several classes of cell surface adhesion molecules, but are dominated by members of the integrin family (1, 2, 3, 4, 5).
The integrin α4β1 (VLA-4, CD49d/CD29) is expressed by a restricted set of cell types, including neural crest cells, lymphocytes, monocytes, eosinophils, and myoblasts and has also been reported to be present at low levels on polymorphonuclear cells (6, 7, 8, 9, 10). α4β1 is a key receptor for the extracellular matrix macromolecule fibronectin and the cell surface Ig superfamily member VCAM-1 (CD106); it has also been reported to bind mucosal addressin cell adhesion molecule-1 (MAdCAM-1)4 (11, 12, 13, 14). In vivo, α4β1 has been implicated in a diverse range of functions, including leukocyte interactions with activated endothelia, anchorage of hematopoietic progenitor cells in the bone marrow, and myoblast fusion during myogenesis (8, 15, 16).
The recognition of short peptide motifs by integrins is a characteristic feature of integrin-ligand interactions (17). The main α4β1 binding site in fibronectin is located in the type III connecting segment (IIICS), a region distinct from that recognized by other integrins (11). Analysis of the IIICS region of fibronectin has identified a key adhesive motif, LDV, encoded in the first alternatively spliced segment (represented by the CS1 peptide) (18). Interestingly, mutagenesis studies performed on VCAM-1 and MAdCAM-1 identified two homologous sequences (IDSPL and LDTSL, respectively) that were responsible for α4β1 interactions (14, 19, 20, 21, 22). Furthermore, all α4β1-ligand interactions can be blocked by the CS1 peptide, suggesting common or overlapping ligand binding sites on α4β1, and importantly, that the LDV motif of fibronectin represents the prototype of a common α4β1-binding motif found in all of its ligands (23).
Despite advances in our understanding of ligand active sites, the binding sites for ligands on integrins are less well characterized. A number of approaches have been used to localize regions involved in the ligand binding process. Peptide cross-linking, epitope mapping of antifunctional mAbs, expression of recombinant integrin fragments, and site-directed mutagenesis have localized active sites to several regions in the N-terminal portions of both the α and β subunits (see Refs. 1 and 3–5 for reviews). Putative ligand-binding sites in the α subunit map to a region that is predicted to be an array of seven homologous modules organized into a β-propeller structure (24). A module homologous to the von Willebrand factor A domain, found in a subset of integrin α subunits, also represents a major ligand-binding site, but this is absent in α4β1 (25, 26, 27, 28, 29). However, there is also strong evidence to suggest that a region resembling a von Willebrand factor A domain is present in all β subunits and that this domain is a key module for ligand binding (Refs. 30 and 31; reviewed in Refs. 3 and 4).
Integrins can exist in both active (ligand-competent) and inactive (ligand-incompetent) states. Physical data obtained from fluorescence resonance energy transfer experiments, the identification of enhanced ligand binding induced by mAbs, and the detection of neo epitopes expressed upon integrin-ligand engagement strongly suggest that integrins undergo conformational changes in response to cation-binding, ligand-binding, and mAb-binding events (reviewed in 4 . Several mAbs recognize ligand-induced binding sites (LIBS) on integrins, and a subset of these mAbs are also capable of activating integrins to bind ligand (32, 33, 34, 35, 36). One interpretation of these data is that integrins exist in an equilibrium between conformationally distinct inactive and active states. According to this model, Ab-induced integrin activation would be achieved when anti-LIBS mAbs bind to and stabilize integrin in an active ligand-competent or ligand-bound conformation. Recently, the mode of action of an inhibitory anti-β1 mAb, mAb 13, has been described (37). Intriguingly, ligand-occupied integrin demonstrated a reduced affinity for mAb 13, implying that the Ab epitope is attenuated by ligand (i.e., mAb 13 recognizes a ligand-attenuated binding site, or LABS). Surprisingly, kinetic analysis of the competition between mAb 13 and ligand for integrin binding revealed an allosteric mechanism of inhibition in contrast to the expected direct competitive, steric mechanism (37). Thus, anti-integrin mAbs can either activate or perturb ligand recognition by stabilizing or destabilizing integrin conformers by an allosteric mode of action.
A large number of anti-α4 mAbs have been described, and these can be classified into several groups depending on epitope location and the functional properties of the Ab (38, 39, 40, 41). Group A mAbs induce leukocyte homotypic aggregation and have been reported to be weak inhibitors of α4β1-fibronectin interactions but to have no effect on α4β1-VCAM-1 binding. Group B mAbs are potent inhibitors of all α4β1-ligand interactions, but these can be further subdivided into B1 and B2 mAbs, where B2 classes stimulate homotypic aggregation and B1 mAbs do not. Group C mAbs neither stimulate homotypic aggregation nor possess antifunctional activity; however, they block homotypic aggregation induced by either A or B2 mAbs. Interestingly, B1 and B2 mAbs block aggregation induced by type A mAbs, whereas group A and B1 mAbs block aggregation induced by type B2 mAbs.
Recently, we have shown for α5β1 that inhibitory anti-integrin mAbs can be used as probes to map the integrin-ligand binding interface and that different regions of ligand can differentially affect the binding of anti-α and anti-β mAbs (42). Here, we have examined the integrin-binding characteristics and epitope expression pattern of a panel of anti-α4 mAbs representing the different epitope classes and the anti-β1 mAb 13 in response to specific α4β1 ligands. The data suggest that different α4β1 ligands have overlapping but distinct binding mechanisms. Moreover, the effects on mAb epitope expression observed suggest that distinct conformational changes may be induced upon integrin-ligand engagement, implying that α4β1 may be able to transduce ligand-specific signals into the cell.
Materials and Methods
Monoclonal antibodies
Mouse anti-human α4 mAb HP2/1 (directed against epitope B1) was purchased from Serotec (Oxford, U.K.). Mouse anti-α4 mAbs HP1/1 (anti-epitope A), HP1/2 (anti-epitope B1), and HP1/3 (anti-epitope A) were purified from culture supernatants (generous gifts of Dr. Francisco Sánchez-Madrid, Universidad Autónoma de Madrid, Madrid, Spain) using protein G-Sepharose affinity chromatography. Mouse anti-α4 mAb 8F2 (epitope C) was a gift of Dr. Chikao Morimoto (Dana-Farber Cancer Institute, Boston, MA). Mouse anti-α4 mAbs GG5/3 and TY21/6 (both vs epitope B2) were gifts of Dr. Ted Yednock (Athena Neurosciences, San Francisco, CA). Rat anti-human β1 mAb 13 and rat anti-human α5 mAb 16 were gifts of Dr. Steven Akiyama (National Institute for Environmental Health Sciences, Research Triangle Park, NC).
Recombinant ligands and peptides
Truncated CAM-Fc chimeras consisting of Ig domains 1 and 2 of human VCAM-1 or murine MAdCAM-1 fused to the hinge and Fc region of human IgG1 were constructed and purified as described previously (14). Control protein MUC18-Fc was a gift of Dr. David Simmons (University of Oxford, Oxford, U.K.). Recombinant fragments of fibronectin HepII/IIICS splice variants were expressed and purified as described previously (43). Variant H/120 comprises fibronectin type III repeats 12 to 14, full-length IIICS, and fibronectin type III repeat 15. Variant H/89 comprises fibronectin type III repeats 12 to 14, IIICS region (minus the final 31 amino acids) and fibronectin type III repeat 15. The synthetic peptide CS1 (DELPQLVTLPHPNLHGPEILDVPST) was synthesized using Fastmoc chemistry on an Applied Biosystems (Foster City, CA) model 431A peptide synthesizer and purified as previously described (44, 45).
Purification of α4β1
MOLT4 cells (ECACC, Porton Down, U.K.) were cultured in RPMI 1640 supplemented with 2 mM glutamine and 10% (v/v) fetal bovine serum with antibiotics (all from Life Technologies, Paisley, U.K.). Confluent cultures (20 liters) were centrifuged at 2,000 × g for 5 min, and the cell pellet was washed in PBS (Life Technologies). The cell pellet (approximately 10–15 ml) was then extracted for 30 min at room temperature in 2% (v/v) Triton X-100, 150 mM NaCl, 2 mM PMSF, 10 μg/ml leupeptin, 5 mg/ml BSA, and 25 mM Tris-HCl, pH 7.4. Lysed cells were then centrifuged at 1,600 × g for 5 min, and the supernatant was recentrifuged at 30,000 × g for 30 min. The supernatant was absorbed first by filtration through Sepharose 4B and then by mixing with rat IgG-Sepharose (2 mg IgG/ml beads) for 1 h at room temperature. IgG-Sepharose was removed by filtration, and a 2-ml wash with 0.1% (v/v) Triton X-100, 150 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, and 25 mM Tris-HCl, pH 7.4 (wash buffer) was pooled with the cleared filtrate. The supernatant was then mixed with 3 ml mAb 13-Sepharose (2 mg IgG/ml beads) for 2 h at room temperature. The suspension was then packed into a 1.6-cm diameter column (Pharmacia) and washed with wash buffer at 4 to 8 ml/min for 16 h at 4°C. Material retained on the column was eluted with 1 mM CaCl2, 1 mM MgCl2, 0.1%(v/v) Triton X-100, and 10 mM sodium acetate, pH 3.5, at 0.5 ml/min at 4°C, and 0.5-ml fractions were collected into a neutralizing volume of 1 M Tris-HCl, pH 8.2. Aliquots of fractions were then analyzed by electrophoresis on a 6% nonreducing SDS-polyacrylamide gel. After Coomassie blue staining, bands corresponding to the expected positions for α4, α5, β1, and α4 cleavage products were detected. Peak integrin fractions were pooled and mixed for 2 h at room temperature with 50 μl of mAb 16 (anti-α5)-Sepharose (5 mg IgG/ml beads) to remove α5β1 from the preparation. A second α5β1 depletion was performed and mAb 16-Sepharose removed by centrifugation. α4 and β1 were the only integrin subunits detected in the cleared fractions by enhanced chemiluminescence ELISA.
Effect of ligand on mAb-α4β1 binding
Purified α4β1 (∼100 μg/ml) was diluted 1:100 in PBS (containing divalent cations; Life Technologies), and 100-μl aliquots were added to a 96-well ELISA plate (Dynatech Immunulon-3; Dynatech Laboratories, Chantilly, VA). Protein was allowed to coat wells for 16 h at 4°C, and then nonspecific sites on the plate were blocked for 2 h at room temperature with 200 μl 5% (w/v) BSA, 150 mM NaCl, 0.05% (w/v) NaN3, and 25 mM Tris-HCl, pH 7.4. Wells were then washed three times with 200-μl aliquots of binding buffer (0.1% (w/v) BSA, 150 mM NaCl, 1 mM MnCl2, and 25 mM Tris-HCl, pH 7.4), before adding aliquots of mAb diluted to appropriate concentrations in binding buffer (with or without varying concentrations of ligand) to triplicate wells. HP1/1, HP1/3 (approximate concentration, 50 μg/ml), and HP1/2 were diluted 1:300, 1:50, and 1:1000 respectively; HP2/1, TY21/6, GG5/3, 8F2, and mAb 13 were used at 0.3 μg/ml. Plates were then incubated for 2 to 2.5 h at 37°C. Unbound material was removed with three 200-μl washes with binding buffer and bound Ab detected with species-specific secondary Ab (IgG-specific, peroxidase conjugate; Dako, High Wycombe, Bucks, U.K.) diluted 1:1000 in binding buffer. After a 25-min incubation at room temperature, wells were washed three times with 200-μl aliquots of binding buffer, and color was developed using 0.55 mg/ml 2,2′-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid), 0.0026% (v/v) H2O2, 0.1 M sodium acetate, and 50 mM NaH2PO4, pH 5. The effect of ligand on mAb-α4β1 binding affinity was analyzed as above, except that the level of mAb binding over a range of Ab concentrations in the presence or absence of ligand (25 μg/ml) was detected. The mode of ligand-mAb competition was determined by monitoring the effect on mAb binding to α4β1 at three different concentrations (10-fold range) over a dose range of ligand.
VCAM-Fc-α4β1 binding
Binding experiments were performed (with or without mAb at ∼3 μg/ml) as before, except that VCAM-Fc (1 μg/ml) was detected using an alkaline-phosphatase-conjugated goat anti-human, IgG Fc-specific secondary Ab (Sigma, Poole, U.K.). Color was developed using soluble phosphatase substrate (p-nitrophenyl phosphate; Sigma) in 0.1 M glycine, 1 mM MgCl2, and 1 mM ZnCl2, pH 10.4. Maximum α4β1-VCAM-Fc binding was observed at 10 μg/ml, with half-maximal binding of VCAM-Fc at 1 μg/ml.
Results
VCAM-1 modulates mAb-α4β1 binding
Recent work from this laboratory has investigated the mechanism underlying the antifunctional activity of mAbs that block adhesion mediated by the integrin α5β1. These studies showed that both anti-α5 and anti-β1 Abs are allosteric inhibitors that work by preventing conformational changes within the integrin subunits that are necessary for ligand engagement. This finding implies that anti-integrin Abs can be used as probes for dissecting the sites within integrin ligands that contact specific receptor active sites. Because the integrin α4β1 has the interesting property of recognizing both the alternatively spliced extracellular matrix protein fibronectin and the cell surface Igs VCAM-1 and MAdCAM-1, we decided to investigate the mechanism of action of anti-α4 Abs. We hypothesized that these studies would inform us about the molecular basis of differential ligand recognition by α4β1 and may ultimately feed into the process of rational design of α4β1 antagonists. To determine the nature of such protein-protein interactions, it is necessary to accurately measure the apparent binding affinity over a range of conditions; to do this we employed a cell-free solid-phase binding assay using purified integrin and recombinant ligands.
Anti-α4 mAbs can be grouped into four classes according to activity, and for this study we used the following representative panel: HP1/1 and HP1/3 (class A); HP1/2, HP2/1, and TY21/6 (class B1); GG5/3 (class B2); and 8F2 (class C). We first monitored the effect of the mAb panel on α4β1 binding to VCAM-Fc. The results confirm previous reports (40), in that only mAbs directed against either B1 or B2 class epitopes inhibited integrin-ligand binding (Fig. 1⇓).
Effect of anti-α4 mAbs on α4β1-VCAM-Fc binding. The y-axis shows percentage VCAM-Fc binding: data are normalized as percentage of maximum α4β1-VCAM-Fc binding detected in the absence of mAb. Solid bar, control α4β1-VCAM-Fc binding in the absence of mAb. Hatched bars, α4β1-VCAM-Fc binding in the presence of mAb indicated. Suboptimal levels of VCAM-Fc (2 μg/ml) were incubated in the presence of mAb (2–5 μg/ml), and the level of ligand binding was quantitated after 2 h. Error bars represent SD of mean. The statistical significance of the difference in binding between control and mAb conditions was evaluated by Student’s t test; for HP1/2, HP2/1, TY21/6, and GG5/3, inhibition was significant at p < 0.0001.
Next, the ability of VCAM-Fc to perturb the binding of this panel of α4-specific was investigated. To maximize ligand binding, experiments were performed at saturating ligand concentrations (50 μg/ml VCAM-Fc) and in the presence of 2 mM Mn2+. The effect of VCAM-Fc occupancy of α4β1 produced contrasting effects on mAb-α4β1 binding (Fig. 2⇓). The antifunctional mAbs, HP1/2, HP2/1, TY21/6 (class B1), GG5/3 (class B2), and mAb 13 (anti-β1) were all substantially inhibited by VCAM-Fc, whereas the epitope A mAb HP1/1 and the epitope C mAb 8F2 were unperturbed. Intriguingly, binding of one epitope A mAb (HP1/3) to α4β1 was markedly enhanced by VCAM-Fc (192%). The control protein MUC18-Fc had no significant effect on any of the mAbs tested.
Effect of VCAM-Fc on anti-α4 mAb-α4β1 binding. The y-axis shows mAb binding: data normalized as percentage of mAb-α4β1 binding detected in the absence of ligand. Solid bars, binding of mAb-α4β1 in the presence of VCAM-Fc (50 μg/ml). Hatched bars, binding of mAb-α4β1 in the presence of control protein, MUC18-Fc (50 μg/ml). Error bars represent SD of mean. The statistical significance of the difference in binding between control and ligand-treated conditions was evaluated by Student’s t test; for HP1/3 stimulation and HP1/2, HP2/1, TY21/6, GG5/3, and mAb 13 inhibition, the data were significant at p < 0.001.
VCAM-1 decreases the apparent binding affinity of antifunctional anti-α4 and anti-β1 mAbs
Analysis of the mode of competition between mAb 13 and fibronectin for α5β1 integrin binding revealed an allosteric mode of inhibition in contrast to a direct competitive steric mechanism (37). To determine the type of inhibition observed with the anti-α4 B1 class, B2 class, and anti-β1-mAbs recognizing α4β1, we examined the effect of a constant concentration of VCAM-Fc (25 μg/ml) on mAb binding to α4β1 over a range of mAb concentrations. The results obtained with HP2/1 (B1 class), and GG5/3 (B2 class) were essentially identical (Fig. 3⇓, A and B). Although the maximum level of mAb-α4β1 binding appeared to be unchanged in the presence of VCAM-Fc, a marked shift in the apparent affinity of mAb for integrin was observed. Double-reciprocal plots of these data revealed that the apparent affinities of each anti-α4 mAb for α4β1 were similar (HP2/1, 0.29 nM; GG5/3, 0.23 nM; Fig. 3⇓, D and E). VCAM-Fc, however, had marked effects on their apparent affinities for α4β1 (HP2/1, 0.9 nM; GG5/3, 1.0 nM). Thus, occupancy of integrin by ligand resulted in a more than threefold decrease in the apparent affinity of both mAbs.
Analysis of the Inhibition of mAb-α4β1 Binding by VCAM-Fc. A–C, Effect of VCAM-Fc on indicated mAb-α4β1 binding: HP2/1 (A), GG5/3 (B), and mAb 13 (C). The • signifies mAb-α4β1 binding in the absence of ligand; and the ○ signifies mAb-α4β1 binding in the presence of VCAM-Fc (25 μg/ml). Error bars represent SD. D–F, double-reciprocal plot of data shown in A–C, respectively. Lines were extrapolated by linear regression. A molecular mass of 150 kDa was assumed for IgG. The y-axis intercept = 1/level of binding; and the x-axis intercept = −1/apparent KD of mAb binding to α4β1. Apparent mAb-α4β1 binding affinities: HP2/1, 0.29 nM (no ligand), 0.9 nM (+VCAM-Fc); GG5/3, 0.23 nM (no ligand), 1.0 nM (+VCAM-Fc); mAb 13, 0.8 nM (no ligand). Note in F, in the presence of VCAM-Fc, a biphasic mAb 13-α4β1 binding curve is obtained, indicative of two classes of binding sites: first population (solid line), 0.8 nM; second population (broken line), 28 nM (see text for further explanation).
VCAM-Fc also decreased the binding of the anti-β1 mAb 13 to α4β1 (Fig. 3⇑C). A double-reciprocal plot of the mAb 13 data revealed that Ab binding was biphasic (Fig. 3⇑F), in contrast to the monophasic binding obtained with both of the anti-α4 mAbs. By separating the mAb 13 data into two populations and performing linear regression analysis, the apparent affinity of mAb 13 for both of the integrin populations could be determined. This analysis revealed that 20% of the total integrin could bind mAb 13 with identical affinity with that in the absence of ligand (0.8 nM), suggesting that this population of integrin represented integrin not occupied by ligand. However, the second population of integrin was found to bind mAb 13 with much lower affinity (28 nM) and could only be detected with high concentrations of mAb 13 (Fig. 3⇑F, dashed line). Thus, these results suggest that mAb 13 has a 35-fold lower affinity for integrin occupied by VCAM-1.
VCAM-1 is an allosteric inhibitor of antifunctional anti-α4 and anti-β1 mAb-α4β1 binding
The above data indicate that VCAM-Fc is capable of inhibiting the binding of all classes of antifunctional α4 mAbs and the antifunctional anti-β1 mAb, mAb 13 to α4β1. The shifts in apparent integrin binding affinities indicate a competitive mode of inhibition. To determine whether ligand was acting as a direct competitor of mAbs or an allosteric inhibitor, we performed binding experiments with several different mAb concentrations and a dose range of VCAM-Fc (Fig. 4⇓, A–C). All three mAb-α4β1 interactions were inhibited by VCAM-Fc in a dose-dependent fashion. Importantly, the maximum extent of inhibition decreased with increasing mAb concentration. These data are inconsistent with the behavior of a direct-competitive inhibitor, in which the maximum level of inhibition observed would be unchanged with increasing mAb concentration. Single-reciprocal plots of these data reveal that in all cases, curves of mAb-α4β1 binding with increasing dose ligand were hyperbolic, indicative of an allosteric mechanism of inhibition (46) (Fig. 4⇓, D–F).
Analysis of VCAM-Fc/mAb competition mechanism. A–C, Effect of VCAM-Fc concentration on mAb-α4β1 binding over a 10-fold range of mAb concentrations: HP2/1 (A), GG5/3 (B), mAb 13 (C). mAb concentrations: A and B, • (1 μg/ml), ▪ (0.3 μg/ml), and ▴ (0.1 μg/ml); C, • (3 μg/ml), ▪ (1 μg/ml), and ▴ (0.3 μg/ml). D–F, single-reciprocal plot of data shown in A–C, respectively. Reciprocal data are nonlinear, indicative of an allosteric mode of action. Error bars represent SD of mean.
VCAM-1 enhances HP1/3-α4β1 binding, yet HP1/3 weakly inhibits VCAM-Fc-α4β1 binding
We next further analyzed the increase in HP1/3-α4β1 binding detected in the presence of VCAM-Fc (Fig. 1⇑). Inclusion of VCAM-Fc in the binding assay did not change the apparent affinity of the mAb for integrin; however, the maximum level of HP1/3 binding was markedly enhanced, suggesting induction of HP1/3 epitope expression by VCAM-Fc (Fig. 5⇓A). HP1/3 has previously been reported to be a weak inhibitor of α4β1-fibronectin binding but to have no effect on α4β1-VCAM-1 interactions (40). Because VCAM-Fc altered HP1/3-α4β1 binding characteristics, we examined the effect of HP1/3 on α4β1-VCAM-Fc binding. The results showed that in contrast to the effect of VCAM-Fc on α4β1-HP1/3 binding, HP1/3 increased the concentration for half-maximal VCAM-Fc binding and slightly reduced the maximal level of α4β1-VCAM-Fc binding (Fig. 5⇓B). Although integrin was activated in these experiments by the inclusion of 2 mM Mn2+, VCAM-Fc binding could be increased when the activating mAb TS2/16 was coincubated with ligand (Fig. 5⇓C). Thus, these results show that despite recognizing a LIBS, HP1/3 is not an activator of α4β1-VCAM-Fc binding.
Analysis of HP1/3-α4β1 binding. A, Effect of VCAM-Fc on HP1/3-α4β1 binding. •, HP1/3-α4β1 binding in absence of ligand; ○, HP1/3-α4β1 binding in presence of 25 μg/ml VCAM-Fc. B, Effect of HP1/3 on VCAM-Fc-α4β1 binding; ○, VCAM-Fc-α4β1 binding in absence of mAb; •, VCAM-Fc-α4β1 binding in presence of HP1/3 (∼1 μg/ml). C, Effect of activating mAb TS2/16 (1 μg/ml) on VCAM-Fc (0.1 μg/ml)-α4β1 binding. Error bars represent SD of mean. In C, the statistical significance of the difference in binding between control and TS2/16 treatment was evaluated by Student’s t test and was significant at p < 0.005.
Differential effects on mAb panel-α4β1 binding observed with MAdCAM-Fc and fibronectin variants
Because different biologic properties and different binding affinities are associated with each α4β1 ligand, they might be expected to trigger different conformational changes when they occupy α4β1. We therefore examined the effect of an array of α4β1 ligands including recombinant fibronectin (IIICS region) splice variants and MAdCAM-Fc on the binding of the anti-α4 mAbs. Two fibronectin variants were tested: H/120, encompassing fibronectin type III repeats 12 to 14, full-length IIICS (120 amino acids), and fibronectin type III repeat 15, and H/89, comprising fibronectin type III repeats 12 to 14, IIICS region (minus the CS5 and CS6 peptides; 89 amino acids), and fibronectin type III repeat 15. The control protein, MUC18-Fc, had a negligible effect on mAb-α4β1 binding, whereas the effects of MAdCAM-Fc and the fibronectin variants differed from VCAM-Fc in a number of respects (Fig. 6⇓). MAdCAM-Fc did not attenuate any of the anti-α4 mAbs, whereas fibronectin variants H/120 and H/89, which both possess a potent LDV motif, attenuated the binding of all antifunctional anti-α4 mAbs. Fibronectin variants that lacked the LDV motif did not affect any mAb-integrin interactions at the concentrations tested (data not shown). In addition, MAdCAM-Fc as well as H/120 and H/89 markedly reduced the level of mAb 13-α4β1 binding. However, none of the alternative α4β1 ligands were capable of enhancing the binding of HP1/3 to α4β1.
Analysis of MAdCAM-Fc and fibronectin variants (H/120 and H/89) on binding of anti-α4 mAb panel to α4β1. The y-axis shows mAb binding: data normalized as percentage of mAb-α4β1 binding detected in the absence of ligand. Binding of mAb-α4β1 is shown in the presence of MAdCAM-Fc (solid bars), fibronectin variant H/120 (heavy hatch bars), fibronectin variant H/89 (light hatch), and MUC18-Fc (open bars). All ligands were used at 50 μg/ml. Error bars represent SD of mean. The statistical significance of the difference in binding between control and ligand treatments was evaluated by Student’s t test: for HP1/2, HP2/1, TY21/6, and GG5/3 effects on H/120 and H/89 and for mAb 13 effects on MadCAM-1, H/120, and H/89, inhibition was significant at p < 0.01.
MAdCAM-1, H/120, and CS1 attenuate mAb13-α4β1 binding
Despite variations in their effects on the anti-α4 mAbs, all of the ligands tested here were capable of perturbing mAb 13-α4β1 binding. Because all physiologic ligands for α4β1 possess an adhesion motif closely related to LDVP, we compared the abilities of the LDV-peptide CS1 (derived from the IIICS of fibronectin), H/120, and MAdCAM-Fc to attenuate mAb13-α4β1 binding. The results obtained indicate that MAdCAM-Fc, H/120, and CS1 are dose-dependent inhibitors of mAb 13-α4β1 binding (Fig. 7⇓, A–C). Although CS1 is a weaker inhibitor than either MAdCAM-Fc or H/120, a significant level of mAb attenuation was observed. However, CS1 did not perturb the binding of any of the anti-α4 mAbs to α4β1 (data not shown).
Analysis of the inhibition of mAb13-α4β1 binding by MAdCAM-Fc, H/120, and CS1 peptide. A, Effect of MAdCAM-Fc; B, effect of H/120; C, effect of CS1 on mAb13-α4β1 binding (mAb 13 at 0.3 μg/ml). The y-axis shows mAb binding; data are normalized as percentage of mAb13-α4β1 binding detected in the absence of ligand. Error bars represent SD of mean.
Discussion
In this report, we have characterized the mode of action of a panel of anti-α4 mAbs and identified differential responses of the epitopes of these mAbs to occupancy of the integrin by a range of α4β1 ligands. Our novel findings are that 1) the binding of function-blocking mAbs, directed against either the α4 or β1 subunit, are attenuated by ligand occupancy and that ligand-dependent mAb attenuation is via an allosteric mechanism in all cases; 2) the degenerate LDV motif, a version of which is present in all α4β1 ligands, is the minimal sequence required for attenuation of the mAb 13 binding site on the β1 subunit (MAdCAM-1 or CS1 peptide did not perturb any α4 epitopes); and 3) a non-antifunctional class A α4 mAb recognizes a LIBS, uniquely induced by VCAM-1. Taken together, these data show that engagement of different α4β1 ligands elicits different conformational responses in the integrin.
Four distinct activities have been attributed to anti-α4 mAbs depending on their ability to perturb integrin-ligand binding or induce homotypic cell aggregation and have been classified as either A, B1, B2, or C mAbs (40). The epitopes of these mAbs have since been localized on the α4 subunit using human/mouse chimeras and site-directed mutagenesis. Class A mAbs map to the extreme N-terminus of the α-chain, whereas class B mAbs locate to the third repeat; class C mAbs map to residues 422 to 606 (38, 39, 41). The mechanism of both antifunctional and homotypic adhesion-inducing activities is presently poorly understood.
We first demonstrated that only mAbs of the B1 and B2 subsets blocked the binding of VCAM-Fc to α4β1 in a cell-free solid-phase assay, confirming previous data (40). A recent report has demonstrated that ligand (fibronectin-fragments) could attenuate the antifunctional anti-β1 mAb 13 binding site on α5β1 (37). Thus, we examined the effect of VCAM-Fc on the ability of the anti-α4 mAb panel to bind α4β1. We found that in the presence of VCAM-Fc, the binding of representatives of both B1 and B2 (antifunctional class) mAbs was reduced by approximately 70%, indicating competition between ligand and mAb for integrin binding. In addition, mAb 13 binding was also reduced by VCAM-Fc. In contrast to these observations, the binding of HP1/3 (class A) was dramatically enhanced (twofold increase) by VCAM-Fc, suggesting that HP1/3 recognizes an epitope induced by ligand. Thus, HP1/3 represents a member of the anti-LIBS family.
We further investigated the mechanism of ligand attenuation of mAb-α4β1 binding and found that in the presence of VCAM-Fc, the affinity of class B1 (HP2/1), class B2 (GG5/3), and anti-β1 (mAb 13) mAbs was reduced. These data closely resemble those obtained for mAb 13-α5β1 interactions in the presence of fibronectin central cell-binding domain region fragments (37). Thus, it would appear that mAb 13, HP2/1, and GG5/3 have high affinities for unliganded integrin but low affinities for ligand-occupied integrin.
Further data indicated that the mechanism of ligand-dependent mAb epitope attenuation was not directly competitive but was allosteric in nature. Thus, HP2/1 and GG5/3 can be grouped with mAb 13 as anti-LABS Abs. These findings have important implications for our understanding of integrin function and the mode of action of antifunctional integrin mAbs. First, the concept that integrins are allosteric proteins and undergo conformational changes when ligands or mAbs engage is further reinforced. Second, the use of anti-functional mAbs to identify ligand binding sites should be interpreted cautiously, because mAb-dependent perturbation of integrin function could potentially be distal from the ligand binding site. Third, the fact that ligand-induced conformational changes, as detected by modulation of mAb epitope recognition, indicates that both the α and β integrin subunits are involved in integrin-ligand binding events. These data also provide an insight into the mode of action of A, B1, B2, and C class mAbs. Although the precise molecular basis of anti-α4 mAb (class A and class B2)-induced homotypic cell aggregation is unclear, a possible explanation for mAb effects is apparent. Inhibition of class B2 mAb-induced aggregation by class A mAbs, the inhibition of class A mAb-induced aggregation by class B2 mAbs, and the blockade of both by B1 and C class mAbs have been observed; however, the epitopes of class A, class B, and class C mAbs are distinct according to competition assay and epitope mapping experiments (38, 39, 41). Furthermore, the class C mAb, 8F2, which did not block α4β1-VCAM-Fc binding and was not affected by VCAM-Fc occupancy of integrin, is capable of blocking some ligand-dependent responses (47). A likely explanation for these effects might be that integrin activity is modulated by conformational changes induced when mAbs engage their epitopes.
In addition to these observations, we show that a class A mAb, HP1/3, recognizes a site induced by VCAM-Fc, further reinforcing the perception that ligand engagement results in a conformational response by ligand. Analysis of HP1/3 binding data in the presence of VCAM-Fc reveals an increase in the maximal level of HP1/3-α4β1 binding (but not an increase in binding affinity), indicating that HP1/3 represents an anti-LIBS mAb as opposed to the anti-LABS mAbs described above. Several anti-LIBS mAbs have been previously described, including a number that map to the β1 subunit (e.g., 15/7, 9EG7, and 12G10); however, HP1/3 represents the first anti-α4 anti-LIBS mAb (35, 36, 37, 48). Interestingly, data obtained here with VCAM-Fc and elsewhere with fibronectin fragments demonstrate that HP1/3 is capable of inhibiting integrin-ligand interactions. Published findings have suggested that HP1/3 is an inhibitor of α4β1-fibronectin binding but not α4β1-VCAM-1 binding (40). Here, we demonstrate that HP1/3 is a weak inhibitor of α4β1-VCAM-1 interactions. A possible explanation to reconcile these data may reside in the higher apparent affinity of α4β1 for VCAM-1 as opposed to fibronectin and the more sensitive cell-free assay system used here. In the cases examined so far, all anti-LIBS mAbs are also activators of integrin-ligand binding activity; however, for reasons that are unclear, HP1/3 does not fall into this category.
Integrin α4β1 is also a weak receptor for the HepII/IIICS region of fibronectin and MAdCAM-1 (13, 14). When we analyzed the effect of fibronectin splice variants and MAdCAM-Fc on the binding of the anti-α4 mAb panel to α4β1, a set of epitope expression profiles were obtained that were ligand-specific. Only fibronectin variants and VCAM-Fc were capable of perturbing B1 class, B2 class, and mAb 13 epitopes, whereas MAdCAM-Fc only affected mAb 13-α4β1 binding. In addition, only VCAM-Fc was capable of enhancing the HP1/3 epitope. Thus, ligand occupancy of integrin appears to result in the modulation of integrin epitopes, with an epitope profile unique to each ligand. These data therefore imply that α4β1 undergoes distinct conformational changes, depending upon the ligand engaged. Interestingly, mutagenesis of the α4 subunit class A epitope region has been shown to abolish homotypic cell aggregation and α4β1-fibronectin binding but not α4β1-VCAM-1 binding, further reinforcing the unique effects of mAb epitope expression induced by VCAM-Fc (49). On the basis of these findings, α4β1 might be expected to respond differently to distinct α4β1 ligands, with important implications for integrin signaling and cell phenotypic changes.
α4β1 represents an attractive pharmacologic target, given that it plays a key role in a variety of adhesion-dependent disease states, including atherosclerosis, asthma, and rheumatoid arthritis (5, 50). The data obtained with the anti-α4 mAb panel suggest that careful consideration of α4β1 antagonists needs to be made, because potent anti-α4β1 agents may be capable of inducing cellular responses that result in unwanted side effects. Further characterization of the cellular responses associated with anti-α4 mAb binding activity might be useful in identifying potent antifunctional compounds with true antagonistic characteristics.
A propeller model for the N-terminus of the α4 subunit has been proposed (24). In this model, the seven 60-amino acid repeats found in all integrin α subunits are organized into a large domain with the overall appearance of a seven-bladed propeller with pseudo-sevenfold symmetry. In this model, each repeat folds into a four-stranded β-sheet and constitutes a single blade of the propeller. When mapped onto the model, the class A mAbs locate to blade 1, whereas B1 and B2 mAbs map to blades 3 and 4. Class C mAbs fall outside the propeller structure. The data described here suggest that the propeller structure would be capable of conformational changes in response to ligand; indeed, HP1/3 binding is sensitive to mutations at apparently distal sites (38, 39, 41). Moreover, ligands appear to differentially affect the conformation of blades independently of other parts of the propeller.
A related, but degenerate adhesion motif is a common feature of all α4β1 ligands. A consensus sequence, based upon amino acid homology, can be detected within α4β1 ligands (LDV (fibronectin IIICS region), IDS (VCAM-1), and LDT (MAdCAM-1)), with the first aliphatic and aspartate residue of each sequence being crucial for biologic activity (14). Interestingly, the peptide CS1, derived from the IIICS of fibronectin and encompassing the LDV motif, is capable of inhibiting α4β1-fibronectin and α4β1-VCAM-1 interactions via a competitive mechanism, indicating that the ligand binding site(s) on α4β1 for fibronectin and VCAM-1 are closely overlapping or identical (23). When we analyzed the effect of CS1 on the binding of the anti-α4 mAb panel to α4β1, we found that no attenuation or enhancement of epitopes was observed, a response identical with that of MAdCAM-1. Interestingly, like MAdCAM-1, CS1 was able to perturb mAb 13-α4β1 binding. These data suggest that the active LDV motif (present in CS1) interacts primarily with the β1 subunit of α4β1, with no detectable effect on the α4 subunit. These data are comparable with the recent finding that the FnIII(10) of fibronectin interacts with the β1 subunit of α5β1 via its RGD adhesion motif (42). Both RGD and LDV represent important integrin adhesion motifs in a variety of ligands, and it is intriguing that both perturb the same β1 epitope (mAb 13) in separate integrins. The implication of these findings is that key adhesion sequences within integrin ligands appear to engage integrin-active sites via similar mechanisms with similar conformational responses and therefore that RGD and LDV perform equivalent functions in the integrin-ligand binding mechanism.
Further analogy between α5β1-fibronectin interactions and α4β1-ligand interactions is also apparent; in both cases, additional sequences outside the RGD/LDV motif are required for further regulation of anti-α mAb epitopes. In the case of α5β1, anti-α5 mAbs were sensitive to the presence of the synergy sequence of FnIII(9), indicating that this region of fibronectin contacts the α5 subunit (42). Recently, additional sites important for integrin-binding activity in VCAM-1 Ig domain 2 (adjacent to the IDS motif domain) have been identified; however, none of the anti-α4 mAbs tested here demonstrated altered binding to α4β1 in the presence of VCAM-Fc when these sites were mutated (P. Newham and M. J. Humphries, unpublished observations; 14 . Nevertheless, for two β1 integrins at least, there appears to be the requirement for multiple, subunit-specific contact sites on ligands. It is noteworthy that MAdCAM-1 is a relatively poor ligand for α4β1 and is only capable of attenuating mAb 13-α4β1 binding and not the binding of anti-α4 mAbs. The lack of activity associated with MAdCAM-1 may therefore be attributable to poor or no contact with the α4 subunit. However, it should be noted that α4β1 and α4β7-MAdCAM-1 interactions can be perturbed by antifunctional anti-α4 mAbs. Thus, although MAdCAM-1 contact sites on α4β1 are distinct from other α4β1 ligands, they can be perturbed by antifunctional mAbs, perhaps the result of a close functional relationship between both of the integrin subunits.
Acknowledgments
We are grateful to Drs. Chikao Morimoto, Francisco Sánchez-Madrid, David Simmons, and Ted Yednock for their generous gifts of reagents; Janet Askari for technical assistance; and Drs. Allan Lowe and Danny Tuckwell for helpful advice and discussions.
Footnotes
-
↵1 These studies were supported by grants from the Biotechnology and Biologic Sciences Research Council, the Medical Research Council, and the Wellcome Trust. K.C. was supported by a Medical Research Council postgraduate studentship.
-
↵2 Current address: Zeneca Pharmaceuticals, Alderley Park, Macclesfield, Cheshire, SK10 4TG, U.K.
-
↵3 Address correspondence and reprint requests to Professor Martin J. Humphries, Wellcome Trust Centre for Cell-Matrix Research, School of Biological Sciences, University of Manchester, 2.205 Stopford Building, Oxford Road, Manchester, M13 9PT U.K. E-mail address: martin.humphries{at}man.ac.uk
-
↵4 Abbreviations used in this paper: MAdCAM-1, mucosal addressin cell adhesion molecule-1; IIICS, type III connecting segment; LIBS, ligand-induced binding site; LABS, ligand-attenuated binding site.
- Received September 8, 1997.
- Accepted December 29, 1997.
- Copyright © 1998 by The American Association of Immunologists
References
In this issue
