The trafficking of leukocytes through tissues is supported by an interaction between the β2 (CD18) integrins CD11a/CD18 (LFA-1) and CD11b/CD18 (Mac-1) and their ligand ICAM-1. The most recently identified and fourth member of the β2 integrins, αDβ2, selectively binds ICAM-3 and does not appear to bind ICAM-1. We have reported recently that αDβ2 can support eosinophil adhesion to VCAM-1. Here we demonstrate that expression of αDβ2 in a lymphoid cell that does not express α4 integrins confers efficient binding to VCAM-1. In addition, a soluble form of αDβ2 binds VCAM-1 with greater efficiency relative to ICAM-3. The I domain of αD contains a binding site for VCAM-1 since recombinant αD I domain binds specifically to VCAM-1. In addition, αD mAb that block cellular binding to VCAM-1 bind the αD I domain. Using VCAM-1 mutants we have determined that the binding site on VCAM-1 for αDβ2 overlaps with that of α4 integrins. Substitution of VCAM-1 aspartate at position 40, D40, within the conserved integrin binding site, diminishes binding to αDβ2 and abrogates binding to the αD I domain. The corresponding integrin binding site residue in ICAM-3 is also essential to αDβ2 binding. Finally, we demonstrate that αDβ2 can support lymphoid cell adhesion to VCAM-1 under flow conditions at levels equivalent to those mediated by α4β1. These results indicate that VCAM-1 can bind to an I domain and that the binding of αDβ2 to VCAM-1 may contribute to the trafficking of a subpopulation of leukocytes that express αDβ2.
Leukocyte trafficking is regulated by several families of cell adhesion molecules that function concomitantly or sequentially to mediate leukocyte adhesion to endothelium (1). In some vascular beds the initial contact of circulating leukocytes with postcapillary endothelium can be mediated by selectins interacting with their carbohydrate ligands on the opposing cell surface as well as the binding of leukocyte α4 integrins to endothelial VCAM-1 (2, 3, 4, 5). These interactions increase the duration of leukocyte contact with endothelium and can support rolling adhesion. β2 and α4 integrins may then be activated by stimuli such as chemokines presented on endothelium to support avid binding to endothelial ICAM-1 and VCAM-1, respectively. The more avid binding of β2 and α4 integrins may result in leukocyte arrest and extravasation.
Integrins are heterodimers that may possess two contact sites for CAMs. The α subunit of each of the β2 integrins CD11a (LFA-1), CD11b (Mac-1), CD11c, and αD contains an approximate 200-residue A or I domain. The I domain is homologous to the A domains in von Willebrand factor and repeats in cartilage matrix protein and collagen (6). The crystal structures of CD11a and CD11b I domains and von Willebrand factor A domains are very similar (7, 8). The I domains of CD11a and CD11b possess an essential binding site for ICAM-1. I domain-specific Abs can block binding to ICAM-1, and soluble recombinant I domain has been shown to bind directly to ICAM-1 and to compete ICAM-1 binding (9, 10). A cation binding site in the I domain, the metal-dependent adhesion site (MIDAS),2 may directly bind ICAM-1 (7, 11). Amino acid substitutions in the MIDAS abrogate ICAM-1 binding (12, 13, 14). A conserved integrin binding site or motif (IBS) in the amino-terminal ICAM-1 Ig-like domain is essential to ICAM-1-dependent adhesion and may form the CD11a I domain contact site (11, 15, 16). In contrast, the α4 subunit of α4β1 and α4β7 does not contain an I domain and does not appear to bind ICAM-1. Binding to VCAM-1 appears to use residues within the seven amino-terminal repeats of α4 (17). It is interesting that although VCAM-1 binds to α4 integrins, its binding is dependent on an IBS highly homologous to that in ICAM-1 (18). The β subunits also possess a cation binding site that appears to interact directly with CAMs (19). Substitution of residues in this site also abrogates CAM binding. Thus, both α and β subunits contribute to a CAM binding site.
The α4 integrins can mediate adhesion under flow and thus support early leukocyte contact with endothelium before extravasation. Binding of α4 integrins to endothelial VCAM-1 can support adhesion within the range of shear stresses reported for postcapillary venules of 1.5–30 dynes/mm2 (20, 21). However, the β2 integrins CD11a/CD18 (LFA-1) and CD11b/CD18 (Mac-1) do not appear to support leukocyte adhesion to purified ICAM-1 under flow conditions with shear stresses >0.5 dynes/cm2 (22).
Previously we isolated a fourth member of the human β2 subfamily, αDβ2 (23). αDβ2 is expressed on monocytes, macrophages, NK cells, a subpopulation of T and B cells, basophils, eosinophils, and neutrophils (24). Overall αD is more closely related to CD11b and CD11c (50–60% amino acid identity) than to CD11a (35% identity). The I domain of αD is also most closely related to the CD11b and CD11c I domains. We have demonstrated that αDβ2 binds to ICAM-3, but demonstrates weak or undetectable binding to ICAM-1 (23). Recently we reported that static adhesion of eosinophils to VCAM-1 can be supported by αDβ2 (25). Here we show that both cell surface and soluble recombinant αDβ2 bind to VCAM-1, and this binding is blocked with αD and VCAM-1 mAb. The αD I domain appears to provide a critical binding site for VCAM-1. We also demonstrate that αD binding to VCAM-1 supports adhesion under both static and flow conditions. In addition, VCAM-1 mutants were used to identify a VCAM-1 binding site for αD that overlaps with that for α4.
Materials and Methods
Soluble αD/CD18 and CAMs
A soluble form of αDβ2 was generated by subcloning cDNA encoding residues encompassing the extracellular domains into the expression vectors pcDNA.3 (Invitrogen, San Diego, CA) and pDC1 (ICOS, Bothell, WA). The resulting plasmids were electroporated into DG44 CHO cells (23). Transfectants were selected in nucleoside-deficient medium with dialyzed serum and 800 μg/ml G418 for 2 wk before screening of supernatants for αD/CD18 in a sandwich ELISA using mAb that bind α- and β-chains. Clonal populations were obtained by limiting dilution culture in selective medium. Soluble protein was characterized by Western blot analysis and SDS-PAGE analysis of affinity purified material.
To generate a soluble recombinant VCAM-1 domain 1 and 2/Fc chimera (VCAM-2D), we inserted the first two domains of VCAM-1 in-frame with the Fc region of human IgG1 (hinge-CH2-CH3) in the expression plasmid pDC37 (ICOS). A mutation in domain 1 of VCAM-1 that replaced the aspartic acid, D40, of the VLA-4 binding site, with an alanine was engineered by PCR. Then the mutated domain 1 and the domain 2 of VCAM-1 were ligated to the IgG1 Fc (VCAM-2D, D40A) sequence and pDC37 as described above. VCAM-2D and VCAM-2D, D40A, were expressed in COS cells and purified from culture media.
Expression and purification of the αD I domain
The boundaries of the αD I domain were defined based on alignment with previously determined sequences of CD11a and CD11b I domains. This fragment was isolated from the αD cDNA by PCR amplification using the primers GGAGGAGACATATGAAGAAGATGGACATCGTCTTCCT (5′) and CTCGAGGGTTCCCTCAACTGCATAGATC (3′). The resulting 580-nt fragment was cloned into pCR2.1 (Invitrogen) for sequence verification and then isolated by digestion with NdeI and XhoI and subcloned into pET20b+. The insertion of the I domain fragment at the NdeI site resulted in utilization of the N-terminal methionine of pelB, which was removed in the NdeI/XhoI digestion of the vector. Insertion into the XhoI site allowed fusion of the C-terminal amino acids of the I domain with the 6× His tag encoded by the plasmid.
Escherichia coli BL21 (DE3) cells were transformed with pET20b+/αDI and grown overnight in a volume of 2 ml of Luria Bertani (LB) broth containing 100 μg/ml ampicillin. One milliliter of this culture was expanded in 50 ml of fresh broth for growth to an OD 600 of 1.0 (4–6 h at 25 or 30°C. Expression occurred independently of the presence of isopropyl β-d-thiogalactopyranoside. Cells were lysed by sonication or use of B-PER reagent (Pierce, Rockford, IL), and soluble and insoluble fractions were obtained by centrifugation of the lysate at 15,000 rpm. SDS-PAGE analysis revealed that ∼30% of the αD I domain was available in soluble form.
The soluble fraction was loaded on a Q-Sepharose column pre-equilibrated with 10 mM NaCl in 10 mM Tris, pH 7.5. Fractions were obtained by elution with a linear gradient of 10–500 mM NaCl in 10 mM Tris (pH 7.5). After analysis by SDS-PAGE, fractions eluting between 30–70 mM NaCl were pooled and applied to a 250-μl nickel column (Qiagen, Chatsworth, CA). The column was washed with 5 ml of 20 mM imidazole in 300 mM NaCl and 20 mM Tris (pH 7.5) before elution of the αD I domain with buffer containing 200 mM imidazole. A PD-10 gel filtration column (Pharmacia Biotech, Piscataway, NJ) equilibrated with PBS was used to remove imidazole. The αD I domain was determined to be >95% pure by SDS-PAGE.
Generation of αD Abs
BALB/c mice were immunized with the affinity-purified, soluble form αD/CD18 described above (26). The α chain specificity of IgG-positive hybridomas was determined in a sandwich ELISA using soluble αD/CD18 and LFA-1. Briefly, soluble integrins were immobilized in 96-well microtiter plates (Immulon IV, Corning, Medfield, MA) with the F(ab)2 of a nonblocking anti-CD18 Ab (mAb 195N, ICOS). Following incubation of hybridoma supernatants with immobilized integrin, bound IgG was detected with a goat anti-mouse Fc-specific mAb. Hybridomas with reactivity to αD/CD18 only were cloned. The specificities of αD mAb 217I, 217K, 240I, and 212D, all IgG1, were confirmed by flow cytofluorometric and Western blot analyses and immunoprecipitation.
Recombinant protein binding assays
Soluble αD/CD18 from culture supernatants was immobilized on Immulon IV microtiter plates (Corning) previously coated with 5 μg/ml 212D mAb in bicarbonate buffer (pH 9.5) and blocked with 3% fish skin gelatin. After 16 h at 4°C, plates were washed and incubated for 1 h with 1–5 μg/ml Ab as indicated. Biotinylated CAM/Fc was diluted in Tris, 150 mM NaCl buffer (1 mM MgCl2, 1 mM CaCl2, and 0.5 mM MnCl2) and added to plates at 100 μl/well at the concentrations indicated in Fig. 2. Following a 90- to 120-min incubation at room temperature, bound CAMs were detected with streptavidin-HRP complex and o-phenyldiamine substrate.
To determine whether blocking αD mAb bind to the I domain, the recombinant αD and CD11b I domains were plated separately in 96-well microtiter plates. Wells were blocked with 2% fish skin gelatin (Sigma, St. Louis, MO) followed by incubation with 1 μg/ml αD-specific (217I and 240I, ICOS) or CD11b-specific (44AACB and OKM1, American Type Culture Collection, Manassas, VA) mAb. An irrelevant isotype-matched control mAb (IgG1) was also used. Bound mAb was detected with goat anti-mouse Fc HRP conjugate and o-phenyldiamine substrate. Binding of αD mAb to the αD I domain and not the CD11b I domain indicates that these adhesion-blocking αD mAb bind specifically to the αD I domain.
To determine whether the αD I domain binds VCAM-1, the I domain was immobilized in nickel-chelated flow cells for surface plasmon resonance (27). An NTA chip (BIAcore, Uppsala, Sweden) was charged with 20 μl of 500 μM NiCl2 in eluent buffer (10 mM HEPES, 0.15 M NaCl, and 0.005% surfactant P20, pH 3.0) at a flow rate of 20 μl/min. The αD I domain (200 nM) in eluent buffer was bound at a flow rate of 10 μl/min. Binding resulted in 200 response units. CAM binding was determined in eluent buffer with 1 mM MnCl2.
Generation of αDβ2-expressing Jurkat cells
Jurkat 77 (J77) cells were irradiated at 300 rad with a 60Co gamma source. Following 5 days of culture, J77α4β1-expressing cells were depleted by panning on polystyrene plates coated with α4 mAb (A4.1, ICOS Corp.) and β1 mAb (K20 and 3S3). After panning, nonadherent J77 cells were selected for further expansion. Three sequential cycles of panning and expansion were performed to obtain an enriched population of J77α4β1 null cells. To obtain J77 clones lacking surface expression of α4 and β1, cells from the enriched nonadherent population were stained with the β1 mAb TS2/16, followed by FACS sorting of single J77α4β1− cells into a 96-well plate. After expansion, these clones were subsequently analyzed for surface expression of α4β1 by staining with the aforementioned panel of anti-α4 and anti-β1 mAbs. Those clones lacking detectable surface expression of α4β1 were then tested in static adhesion assays to VCAM-1 to confirm the functional loss of VLA-4 (data not shown). Stably transfected cells used in flow and adhesion assays were obtained by transfection with full-length αD cDNA subcloned into the HindIII and XhoI sites of pCDNA3 (Invitrogen). Ten million J77/α4− cells were transfected by electroporation with 30 μg of plasmid DNA, followed by selection in 1 mg/ml G418 (Life Technologies). The bulk population was screened by FACS analysis using the Ab 212D. Clones were derived from this population by one round of limiting dilution cloning. Immunoprecipitation experiments were performed to confirm that the αD-chain was associated with CD18. Abs to either chain precipitated the same two proteins (∼95 and 150 kDa) as visualized by SDS-PAGE and Coomassie stain under reducing conditions, and preclearance with Abs to either chain prevented immunoprecipitation of these proteins by Abs to the other chain (data not shown).
Static cell adhesion assay
Static cell adhesion assays were performed as described previously (16). Microtiter plates (Immulon IV, Corning) were coated overnight with 5 μg/ml purified CAM/Fc or capture mAb in bicarbonate coating buffer (pH 9.5). Plates were blocked with 3% fish skin gelatin (Sigma) in Dulbecco’s PBS (Life Technologies, Grand Island, NY). Where relevant, Abs were incubated at 1 μg/ml for 1 h before cell addition. Cells were harvested by centrifugation and were resuspended in binding buffer (RPMI with 0.1% BSA) before addition to plates at 100,000/well (three wells per group). Binding was allowed to occur at 37°C for 30 min (Jurkat) before addition of 1% glutaraldehyde. After fixation for 1 h, plates were washed copiously with distilled water. Crystal violet solution (1% crystal violet in 10% absolute ethanol and 90% water) at 50 μl/well was added for 5 min. After copious washing, 100 μl/well destaining buffer (33% absolute ethanol in 66% sterile dH20) was added. After 12 h, signal was detected at 570 nm on a Dynatech MR500 plate reader (Chantilly, VA). Percent cell binding was calculated as A570 of all binding to CAM divided by A570 of cell binding to capture mAb × 100.
Flow adhesion assay
Borasilicate capillary assay tubes (Drummond Scientific, Broomall, MA) were cut into 2.5-cm lengths and soaked overnight in 95% ethanol. Tubes were then rinsed with sterile water. VCAM-1/Fc and ICAM-1/Fc were diluted to a concentration of 100 μg/ml in HBSS without Ca2+ and Mg2+ (Cellgro, Mediatech, Herndon, VA) buffered to pH 7 with 20 mM HEPES (Cellgro, Mediatech), and instilled into tubes. Tubes were incubated at 37°C in moist petri dishes for 1.5 h and then refrigerated for 30 min before use. Just before experiments, excess VCAM-1/Fc and/or ICAM-1/Fc was rinsed out of the tubes and replaced with the assay medium (HBSS with Ca2+ and Mg2+, 20 mM HEPES (pH 7), and 2% human serum). A silicone tubing loop and roller pump (Masterflex, Cole-Palmer, Vernon Hills, IL) was used to circulate J77 (T cell line Jurkat 77) cells through the CAM-coated assay tubes mounted on an inverted microscope stage. Just before assay J77 cells were centrifuged and resuspended in assay medium for a final loop concentration of 1 × 106 cells/ml.
The J77 cell lines were infused into the closed loop in independent experiments. The J77 αD cells were pretreated for 10 min with an isotype type-matched control mAb (40 μg/ml) or an αD mAb mixture of 217I and 240I (20 μg/ml each), and then infused into the loop system. Alternatively, a VCAM-1/Fc tube was pretreated for 10 min at 37°C with VCAM-1 mAb (130K and 130P, 20 μg/ml each of 130K and P, plus a 1/100 dilution of the anti-VCAM-1 mAb 1G11 (Caltag)), and then untreated J77αD cells were infused into the loop system.
Interactions between the J77 cell lines and bound VCAM-1 were monitored by videomicroscopy and recorded on video tape for analysis. Interactions were monitored for 15 min; the shear rate for the first 5 min was 2 dynes/cm2; that for the last 10 min was1.2 dynes/cm2. The number of cells attached to the VCAM-1/Fc-coated substrate was determined at 1-min intervals by individual frame computer analysis of the recorded experiments, using NIH Image Software, Montana ImmunoTech Macros (Bozeman, MT), and an Apple Computer PowerMac 7100-66 (Cupertino, CA).
Binding of αDβ2-expressing cells to VCAM-1
Previously we had demonstrated that αDβ2 can mediate static adhesion of eosinophils to VCAM-1 (25). We further characterize αDβ2 binding to VCAM-1 here in both cell and recombinant protein systems. For static and flow adhesion assays we used the T cell line Jurkat 77 (J77). J77 do not express α4β7 (data not presented). To eliminate the contribution of α4β1 to VCAM-1 binding, we generated an α4 null variant of Jurkat 77 (J77α4−). J77α4− cells did not stain with α4 mAb, whereas parental J77 clearly express α4 (Fig. 1 A). J77α4− did not stain with several α4 mAb (data not shown) indicating that loss of α4 expression was not due to loss of a specific α4 mAb epitope. J77α4− also did not stain with β1 or β7 mAb (data not shown). αD was expressed in J77α4− to generate an α4− αD+ line (J77αD). Expression of αD in J77αD was ∼2-fold less than that of α4 in parental J77.
J77 cell lines were tested for binding to immobilized VCAM-1 in static adhesion assays. J77αD was found to bind efficiently to VCAM-1 (Fig. 1 B). As expected J77α4− did not bind to VCAM-1, but retained the ability to bind to ICAM-1 via LFA-1. The binding of J77αD to VCAM-1 was blocked partially by the αDβ2-specific mAb 217I and 240I. Together these αD mAb completely blocked binding. J77αD binding to VCAM-1 was also blocked with CD18 and VCAM-1 mAb, but not by a VCAM-1 mAb that is nonblocking for α4 integrin binding (92C4D). Thus, expression of αDβ2 in J77α4− cells confers binding to VCAM-1. The identical pattern of results were obtained using CHO cells expressing αDβ2 (25) (data not presented).
Recombinant αDβ2 binds to VCAM-1
To further demonstrate binding between αDβ2 and VCAM-1, we used a cell-free assay. In this assay a secreted form of αDβ2, containing the entire extracellular region, was captured with a nonblocking αD mAb, and the binding of biotinylated VCAM-1 was determined using strepavidin-HRP in a standard ELISA format. VCAM-1 bound to recombinant immobilized αDβ2 in a saturable manner (Fig. 2,A). The binding of VCAM-1 was ∼6-fold greater than that of background binding to ICAM-1 and 2-fold greater than the binding of ICAM-3. In this assay low levels of αDβ2 binding to ICAM-1 were detected at the highest CAM concentration, 10 μg/ml. The binding of αDβ2 to VCAM-1 was blocked with either αD or CD18 mAb (Fig. 2,B). VCAM-1 possesses two binding sites for α4 integrins and thus may require two distinct mAb for complete blocking (18). VCAM-1 binding to αDβ2 was also partially blocked by one set of VCAM-1 mAb, 130K and 130P, and more completely by another VCAM-1 mAb, 1G11B1 (Fig. 2,C). The ability of the different VCAM-1 mAb to block in this cell-free assay is similar to that observed in cellular binding assays (Fig. 1 B). In contrast to αDβ2, immobilized recombinant LFA-1 does not bind to VCAM-1 (data not shown). Thus, the specificity of recombinant αDβ2 binding to CAMs is consistent with cell-based binding assays. In both cellular and recombinant protein assays αDβ2 binds specifically to VCAM-1 with greater efficiency than that of ICAM-3.
The binding site on VCAM-1 for αD overlaps with that of α4
To localize the binding site on VCAM-1 for αDβ2, we generated two mutants based on the binding sites for α4 integrins. VCAM-1 possesses two binding sites for α4: one in domain 1 and the other in domain 4. Both sites involve a critical aspartate residue in an identical IBS motif, IDSPL. To remove the domain 4 IBS we deleted domains 3, 4, and 5 from the 5D VCAM-1 chimera resulting in a 2 domain VCAM-1 (VCAM-2D). To disrupt the α4 binding site in domain 1, the critical aspartic residue at position 40 was substituted with alanine in VCAM-2D (VCAM-2D, D40A). The binding of J77αD to VCAM-2D was concentration dependent and equivalent to that of VCAM-1 (Fig. 3). The binding of J77αD to VCAM-2D, D40A was decreased by ∼50% relative to VCAM-2D. J77αD binding to VCAM-1 was also equivalent to that of the parental J77 cells mediated by α4β1. Neither J77αD nor J77 bound to ICAM-3 possessing a substitution in the acidic residue of its IBS, E37A. Thus, the VCAM-1 binding site for αDβ2, when expressed in J77, does not require domain 4 and appears to overlap with the site in domain 1 that interacts with α4β1. Identical results were obtained with CHO cells expressing αDβ2 (data not presented).
αD I domain binds blocking mAb and VCAM-1
The I domain of CD11a contains a critical binding site for ICAM-1. To determine whether αD I domain contributes to αDβ2 binding to VCAM-1, a soluble form of αD I domain was produced (Fig. 4,A). The binding of the αD I domain to both αD blocking mAb and VCAM-1 was determined. The αD blocking mAb 217I and 240I bound to αD I domain as determined by ELISA (Fig. 4,B). Binding was specific as αD mAb did not bind to a similar recombinant CD11b I domain and CD11b mAb did not bind αD I domain. Binding of 217I and 240I was noncompetitive (data not presented), indicating that the I domain possesses two distinct epitopes present also on native αDβ2 and thus appears to possess a native conformation. We have demonstrated that mAb 217I and 240I block αD binding to VCAM-1 (Fig. 1 B). This suggests that αD I domain is a critical binding site for VCAM-1 interaction.
Binding of αD I domain directly to VCAM-1 was demonstrated by surface plasmon resonance (Fig. 4 C). VCAM-1 did not bind to αD I domain in the presence of 1 mM EDTA. Binding was specific, as there was no detectable interaction with ICAM-1. LFA-1 I domain bound ICAM-1 Fc, but exhibited negligible interaction with VCAM-1 Fc in the same system (data not presented). VCAM-2D also interacted with αD I domain at a level similar to the 5 domain form of VCAM-1. In contrast, there was no detectable binding of VCAM-2D, D40A to αD I domain. Thus, αD I domain appears to contain a binding site for VCAM-1, and similar to other integrin-CAM interactions, binding is cation dependent.
The binding of αDβ2 to VCAM-1 mediates adhesion under flow conditions
VCAM-1 supports adhesion of leukocytes under flow conditions through an interaction with α4 integrins. To determine whether αDβ2 can also support adhesion to VCAM-1 under flow conditions, J77αD was tested for binding in the presence of shear stress (Fig. 5). J77αD and J77 bound to VCAM-1 at equivalent levels at 2 dynes/mm2 (0–5 min) and at 1.2 dynes/mm2 (5–15 min). Binding of J77αD was blocked with αD and VCAM-1 mAb, but not by a nonblocking CD18 mAb. Neither J77αD nor J77 bound to ICAM-1 under flow conditions. Thus, αDβ2 expressed in J77αD binds specifically to VCAM-1 under flow conditions. The level of binding was similar to that mediated by α4β1 expressed on J77.
We have extended our previous work (25) by determining the types of cellular interactions supported by αDβ2 binding to VCAM-1 and by identifying critical structural components. Expression of αDβ2 in Jurkat 77α4− cells conferred efficient adhesion to VCAM-1-coated substrates. αDβ2 supported adhesion to VCAM-1 under both static and flow conditions and was blocked by αD and VCAM-1 mAb. In addition, recombinant soluble αDβ2 binds VCAM-1 with greater efficiency relative to ICAM-3. The binding of recombinant proteins was also specifically blocked by αD and VCAM-1 mAb. These mAb bind the αD I domain, implicating this domain in the VCAM-1 interaction. We demonstrate that he αD I domain indeed binds directly to VCAM-1 and that the metal and structural requirements for binding are similar to those for cellular adhesion. Thus, VCAM-1 appears to serve as a ligand for αDβ2 and αD I domain functions as an essential binding site. VCAM-1 appears to be the highest affinity αDβ2 ligand described to date. Previously we have indicated that VCAM-1 in solution did not bind to CHO cells expressing αDβ2 (23). This may have been due to differences in assays used. Previously we had used VCAM-1 in solution. Here we used VCAM-1 that was immobilized, and thus a substrate that can support a more avid interaction.
The binding of αDβ2 to VCAM-1 is unexpected and indicates that a CAM can interact with different integrins by binding distinct α-chain domains. The binding of β2 integrins to ICAMs has been well characterized. These studies indicate that the A or I domain present in the α subunits of β2 integrins is a critical site for binding of ICAMs (9, 10). Substitution of I domain amino acids in or proximal to those that coordinate cations, the MIDAS, disrupts ICAM binding (12, 13, 14). ICAMs have not been reported to bind non-I domain integrins. Conversely, VCAM-1 has not previously been reported to bind to an I domain. VCAM-1 instead has been shown to bind to the amino-terminal repeats in α4 that lacks an I domain (17). The alignment of αD and α4 amino-terminal repeats demonstrates conservation of some residues that have been implicated in VCAM-1 binding (data not presented). Thus, although we demonstrate here that VCAM-1 can interact with the αD I domain, this does not preclude the possibility that VCAM-1 might also interact with αD amino-terminal repeats.
The conservation of IBS on CAMs suggests a potential for VCAM-1 binding to an I domain in a manner similar to ICAM-1. CAM Ig-like domains are composed of two β-sheets containing β-strands abed or cfg. Essential residues for integrin binding, the IBS, localize primarily to the mostly flat face composed of β-strands gfc (15, 28). This binding face possesses two components. First, there is the conserved IBS, a motif of approximately five amino acids, IETPL in ICAM-1 and IDSPL in VCAM-1 (16, 18). This sequence, located between β-strands C–D, contains the invariable critical acidic residues E or D that protrude from the binding face. The acidic residue in the IBS has been suggested to form an additional coordination site for cations in the I domain MIDAS. Replacing VCAM-1 IBS with that of ICAM-1 (29) or VCAM-1 with that of MadCAM-1 IBS (30) does not disrupt integrin-specific binding. Thus, this conserved motif that is essential to binding is not key to integrin specificity. Secondly, there are other less conserved CAM sequences on the binding face of VCAM-1 (29, 30), ICAM-1 (15), and ICAM-3 (16, 31) that have been shown to function in binding, and these have been proposed to determine integrin specificity. Thus, the integrin binding face on VCAM-1 closely resembles that on ICAM-1, which appears to bind to I domain (11). One exception may be that D40 of VCAM-1 protrudes somewhat further from the binding face relative to the corresponding functional residue E34 on ICAM-1 (28). Because we have demonstrated that VCAM-1 D40 functions in binding to the αD I domain, the αD I domain must accommodate the somewhat extended presentation of D40. The substitution D40A results in a 50% decrease in binding to αD and a complete loss in α4β1-specific binding. This indicates that additional residues in VCAM-1 and αD contribute to binding.
We find that αDβ2 binding to VCAM-1 can support cell adhesion under flow conditions. Interactions between leukocytes and endothelium play a critical role in leukocyte trafficking and occur in a broad range of shear stresses, 2–30 dynes/mm2 or less, depending on the particular vascular bed. At very low shear stresses under 0.5 dynes/mm2, β2 integrins may support adhesion to ICAM-1 (32, 33). In contrast, we find that at 2 and 1.2 dynes/mm2, αDβ2 mediates adhesion to VCAM-1 at levels equivalent to that of α4β1 (Fig. 5). Thus, αDβ2 may contribute to the tethering of leukocytes at different in vivo sites where VCAM-1 is expressed. Spleen and liver constituitively express VCAM-1, consistent with the localization of αD-expressing macrophages in normal individuals (34). Thus, αD binding to VCAM-1 may contribute to both the homing and the retention of leukocytes in certain tissues. Recently, we have demonstrated that αDβ2 can mediate adhesion of eosinophils to VCAM-1 and that this interaction might function in chronic inflammation (25).
We thank Rick Jasman and Janelle Taylor for αD mAb production, and Alice Dersham for manuscript preparation.
Abbreviations used in this paper: MIDAS, metal-dependent adhesion site; IBS, integrin binding site; CAM, cellular adhesion molecule.