Differential Effects of Integrin α Chain Mutations on Invasin and Natural Ligand Interaction*

To determine if recognition of the Yersinia pseudotuberculosis invasin protein and natural substrates requires identical integrin residues, a region of the human α3 integrin chain predicted to be involved in substrate adhesion was targeted for mutation. One point mutation located in a region of the third N-terminal repeat of the α3 chain, α3-W220A, failed to promote adhesion to the natural α3β1 substrate epiligrin but maintained near wild type levels of adhesion to invasin. A second nearby mutation, α3-Y218A, which showed no detectable adhesion to epiligrin, was only partially attenuated for invasin binding as well as invasin-mediated bacterial uptake. A third substitution, α3-D154A, predicted to be in the second N-terminal repeat not known to be implicated in cell adhesion, was competent for invasin-promoted adhesion events and appeared to encode a receptor of increased activity, as it had a higher efficiency than wild type receptor for adhesion to epiligrin. Cell lines expressing this derivative were not recognized by a function blocking anti-α3 antibody, indicating that the second and third repeats of the α3 chain are either closely linked in space or the second repeat can modulate activity of the third. Differential effects on substrate adhesion do not appear to be associated with all integrin α chain mutations, as α4chain mutations affecting the divalent cation binding domains depressed adhesion to invasin to a significant extent.

Invasin is a 986-amino acid outer membrane protein encoded by the Gram-negative pathogen Yersinia pseudotuberculosis (1,2) that binds at least five different integrins containing the ␤ 1 chain, including ␣ 3 ␤ 1 and ␣ 4 ␤ 1 (3). 1 Attachment to integrins by invasin leads to internalization of the bacterium (1,2). Crossinhibition studies with human fibronectin indicate that invasin recognizes a site on the integrin identical or nearby to that utilized by natural substrates (4).
Substrate recognition of natural substrates appears to involve the extracellular domains of both the ␣ and ␤ integrin chains (5)(6)(7)(8). Each ␣ chain has a series of 7 nonidentical Nterminal repeats, approximately 60 amino acids in length ( Fig.  1; see Refs. 9 and 10). Analyses of function-blocking monoclonal antibodies and site-directed mutagenesis indicate that the third ␣ chain N-terminal repeat is required for substrate recognition (11)(12)(13)(14)(15)(16). One striking feature of the identified critical ␣ chain residues found within this repeat is that aromatic side chains are important for ligand recognition (Fig. 1), reminiscent of previous observations regarding human growth hormone interaction with its receptor (17). Also found in the ␣ chain N-terminal region are three predicted divalent cationbinding motifs within repeats V-VII. Each putative cationbinding N-terminal repeat lacks a potential coordinating residue at the most C-terminal position of the cation-binding loop (26,27). One possible model is that a residue from a ligand may complete metal coordination within one of these domains to stabilize ligand binding (28), although data regarding substitution mutations in this region are inconsistent with this proposition (29).
Recent molecular modeling of the 7-repeat N-terminal domain of the integrin ␣ chain has led to the proposal that it is arranged in a 7-bladed "propeller" structure (18). This model uses intradomain contacts to bury hydrophobic residues and provides an explanation for why ligand-derived peptides can cross-link to multiple ␣ chain repeats. This is based on the hypothesis that the ligand contact surface of the integrin chain may involve several neighboring repeats binding to a single substrate (19 -23).
To investigate the model that identical integrin residues are involved in invasin and natural substrate adhesion, the following work examines three derivatives harboring mutations in a region of ␣ 3 homologous to the regions shown to be important for substrate recognition (13,14). Two mutations (␣ 3 -Y218A and ␣ 3 -W220A) lie in the third N-terminal repeat in a region predicted to adopt a ␤-turn structure on the surface of the ␣ chain integrin subunit (13,16), whereas the third mutation (␣ 3 -D154A) lies in the second N-terminal repeat. The results indicate that loss of side chains previously implicated in natural substrate recognition of the ␣ subunit is not sufficient to cause loss of recognition of invasin by integrin receptors. Invasin and natural substrates may recognize an overlapping, but nonidentical, set of residues on the integrin receptor during adhesion.
Construction of ␣ 3 Substitutions-A plasmid harboring the wild type complete cDNA of integrin ␣ 3 , pFneo␣3, was described (34). Plasmid pFneo␣3, containing the entire ␣ 3 cDNA, was digested with XbaI and the liberated ϳ3500-base pair full-length ␣ 3 -encoding fragment was inserted into XbaI-cut pSelect vector (Promega). Five ␣ 3 substitutions were constructed using the pSelect system as described (see Ref. 31; Promega Biotech). Mutations were incorporated into the gene using the following mutagenic oligonucleotides: 5ЈGGGTCAGAAGCCCAGCGGC-GCATG (D154A), 5ЈGCCCCCGGTGCCGCCAACTGGAAAGG (Y218A), 5ЈCCCGGTGCCTACGCCTGGAAAGGAAA C (N219A), 5ЈGGTGCCTA-CAACGCGAAAGGAAACAGC (W220A), 5ЈTACAACTGGAAAGCAAA-CAGCTACATG (G222A). Substitutions were confirmed by sequencing, and the altered ␣ 3 genes were returned to plasmid pFneo␣3 as XbaI fragments. Sequencing confirmed proper gene replacement within the pFneo␣3 vector. pFneo␣3 containing each of the ␣ 3 inserts was transfected into K562 cells by electroporation using 20 g of plasmid DNA and 5 ϫ 10 6 cells at 230 V, 960 microfarads (Bio-Rad Gene Pulser). Cells were grown 48 h in RPMI 1640 containing 10% FCS and penicillin/ streptomycin. After 48 h, cells were supplemented with 400 g/ml G418, and stable transfectants were selected. Following 2 weeks of selection, ␣ 3 -expressing cells were enriched by sorting with magnetic anti-mouse IgG-coated beads (Dynal Co. Oslo, Norway) coated with a 1:10 dilution of mouse anti-␣ 3 antibody A3-X8 hybridoma supernatant. ␣ 4 and ␣ 3 Microtiter Binding Assay-K562 cells transfected with wild type or mutant human ␣ 4 clones were sorted with magnetic anti-mouse IgG-coated beads (Dynal Co.) coated with a 1:1000 dilution of the purified anti-␣ 4 antibody B-5G10 (ϳ1 g/ml), and expression was analyzed by FACS analysis after 2 weeks grow out as described previously (29). 96-Well microtiter plates (ICN Biochemicals) were coated with a series of dilutions of MBP-INV497 at a starting concentration of 100 g/ml (33,35). Cells were pelleted, incubated with PBS containing 1 mM EDTA for 5 min at 37°C, recharged with 2 mM MgCl 2 , and repelleted. Cells were resuspended in RPMI containing 20 mM HEPES (pH 7.0) and 0.4% bovine serum albumin (Binding Buffer), counted, pelleted and resuspended in Binding Buffer at 5 ϫ 10 6 cells/ml. 25 l of cells were transferred to wells in a 96-well plate and mixed with 25 l of an undiluted supernatant of anti-␣ 5 ␤ 1 antibody VD1 or 25 l of Binding Buffer for 1 h at room temperature to block ␣ 5 ␤ 1 binding to invasin. The 50-l mixtures were transferred to wells of an invasin-coated 96-well microtiter plate and allowed to bind for 50 min at 37°C, 5% CO 2 . Unbound cells were removed by washing three times with PBS, and bound cells were fixed with 100 l of acetone/methanol (1:1). Cells were stained with 0.01% crystal violet in H 2 O for 30 min at room temperature, washed 5-10 times with H 2 O, and lysed with 100 l of 2% deoxycholate. Cell binding was determined at A 595 using an enzyme-linked immunosorbent assay plate reader (Bio-Rad).
Cells transfected with human ␣ 3 derivatives were enriched for those with the highest ␣ 3 expression using magnetic bead selection as described above. Adhesion of ␣ 3 transfectants to epiligrin was determined by incubating cells in 96-well tissue culture-treated microtiter plates (Falcon) in which a series of 1:2 dilutions of A431 cells starting at a highest density of 1.5 ϫ 10 4 cells had been allowed to deposit extracellular matrix for 48 h as described (36,39). Epiligrin concentration in wells was determined using anti-laminin 5 mAb 6F12 (see Ref. 37; generous gift of Dr. R. E. Burgeson) and a colorimetric enzyme-linked immunosorbent assay. To measure adhesion, 25 l of cells, resuspended at 2 ϫ 10 6 cells/ml in Binding Buffer, were mixed with 25 l of either undiluted supernatant of the blocking anti-␣ 3 antibody A3-IVA5 or a 1:100 dilution of the blocking anti-␣ 3 antibody P1B5 ascites in Binding Buffer. Cells were incubated with antibody for 1 h at room temperature, and the 50-l mixture was added to matrix-coated wells for 45 min at 37°C. Unbound cells were removed by washing three times with PBS containing 2 mM MgCl 2 , and bound cells were fixed and stained as described above for ␣ 4 transfectants. Binding to invasin was determined similarly, except that 96-well microtiter dishes (ICN Biochemicals) were coated with 100 l of 1 or 5 g/ml MBP-INV497, and cells were preincubated with 25 l of a 1:2 dilution of culture supernatants of hybridomas producing VD1 or 25 l of an equal mixture of the hybridoma supernatant and mAB A3-IVA5. The concentration of VD1 was sufficient to eliminate detectable binding to ␣ 5 ␤ 1 (see Fig. 5).
Assay of Bacterial Uptake via ␣ 4 ␤ 1 and ␣ 3 ␤ 1 Integrins-K562 transfectants expressing wild type or mutant ␣ 4 chains were pelleted, washed twice with PBS to remove G418 and penicillin/streptomycin antibiotics, grown overnight in RPMI 1640 containing 10% FCS, washed once with PBS, and grown overnight again in RPMI 1640, 10% FCS. To assay uptake, mammalian cells were pelleted and resuspended in Binding Buffer at 1 ϫ 10 6 cells/ml, and 500-l cells were added for 60 min at room temperature to a 24-well tissue culture-treated plate in the presence or absence of 50 l anti-␣ 5 ␤1 VD1 antibody. A 10-l aliquot of 1 ϫ 10 8 bacteria/ml of MC4100 harboring plasmid pJL277 (inv ϩ (31)) or pT7-12 (vector without insert) was added to cells preincubated with or without mAb VD1 and allowed to internalize bacteria for 90 min at 37°C. 50 g/ml gentamicin was then added for 90 min at 37°C; cells were washed three times by pelleting at 2000 rpm in microcentrifuge tubes, and the eukaryotic cells were lysed with 200 l of 0.1% Triton X-100. Liberated bacteria were diluted and plated on L agar containing 100 g/ml ampicillin. Bacterial uptake was calculated as number of bacteria internalized/bacteria added.
K562 transfectants expressing ␣ 3 were tested for bacterial uptake in a similar fashion except that the inv ϩ strain used was MC4100/pRI203 (2), and the inv Ϫ strain was MC4100. In addition, the effects of anti-␣ 3 antibodies P1B5 and A3-IVA5 were determined by including them in the VD1 preincubation at a dilution of 1:200 or 1:10, respectively.

Site-directed ␣ 3 Mutants Fail to Bind
Epiligrin-Analyses of blocking antibodies as well as site-specific mutagenesis indicate that in the ␣ 4 , ␣ 5 , and ␣ IIb chains, there exists a region within the third "FG-GAP" repeat ( Fig. 1) that is critical for adhesion (11)(12)(13)(14)(15)(16). To begin to analyze the role of the ␣ 3 chain in invasin-mediated uptake, and to determine if residues recognized by invasin are identical to those recognized by natural substrates, site-directed mutagenesis was used to generate five substitutions of the human ␣ 3 chain. The substitutions changed residues of the third N-terminal repeat shown to be important in substrate recognition in other ␣ chains, as well as a residue located in the second N-terminal repeat. The derivatives ␣ 3 -Y218A, ␣ 3 -W220A (third repeat), and ␣ 3 -D154A (second repeat) were stably expressed in K562 cells as determined by FACS analysis using the anti-␣ 3 monoclonal antibody P1B5 (Fig. 2), whereas ␣ 3 -N219A and ␣ 3 -G222A were not expressed (data not shown).
Stable cell lines expressing ␣ 3 substitutions were tested for adhesion to invasin and to an extracellular matrix containing epiligrin, a natural ␣ 3 ␤ 1 integrin ligand comprised of a mixture of laminins 5 and 6 (38). A dose-response curve of cell adhesion to increasing amounts of epiligrin was performed by incubating transfectants in microtiter wells that had been coated by ECM deposited by A431 cells (see "Materials and Methods"). The concentration of laminin 5 deposited in the wells was dependent on the density of A431 cells (Fig. 3A). K562 cells harboring substitutions ␣ 3 -Y218A and ␣ 3 -W220A showed no detectable adhesion to the laminin 5-containing matrix in the absence of blocking anti-␣ 3 antibody, even at the highest concentrations of laminin 5 (Fig. 3B). In contrast, the derivative ␣ 3 -D154A containing a substitution in the second repeat adhered considerably better than the wild type ␣ 3 (␣ 3 -wt) transfectant (Fig. 3B). This high adhesion efficiency was clearly dependent on the presence of the ␣ 3 chain, as both wild type receptor and the D154A derivative were unable to promote adhesion to matrix if anti-␣ 3 mAb P1B5 was included in the adhesion mix (Fig. 3C).
These results indicate that residues present in repeat III of ␣ 3 that are predicted to contribute to adhesion to natural substrate are required for adhesion by ␣ 3 ␤ 1 . Furthermore, there appears to be a region within repeat II that modulates the binding efficiency of the integrin to natural substrate (Fig. 3B). 1. A, schematic of the structure of the integrin ␣ 3 chain. Shown are the seven N-terminal repeats, the putative metal-binding sites (Me 2ϩ ), and the presumed substrate adhesion repeat III. B, sequence alignment of 5 of the N-terminal repeats from integrin ␣ 3 , ␣ 4 , ␣ 5 , and ␣ IIb chains. Sequences were manually aligned according to recent molecular modeling (18). The FG and GAP sequences are aligned separately, and repeat domains are terminated with the GXXYX consensus in which the 4th position is typically aromatic and the final position is hydrophobic (10). The cation binding domains in repeats V-VII are in bold. Integrin ␣ IIb peptides that inhibit ligand binding correspond to the divalent cation-binding region of repeat V (24,25). Boldface residues in the C terminus of the third repeat of each ␣ chain represent sites in which mutagenesis results in defective adhesion (see Refs. 13, 14, and this study). Asp-154 in the second repeat of ␣ 3 (bold) is the site that was changed as a control in this study. A second result that implicated a role for repeat II in adhesion to substrate was the fact that transfectants expressing ␣ 3 -D154A were not blocked for adhesion to matrix by the anti-␣ 3 mAb A3-IVA5 (Fig. 3B) because they were not recognized by this reagent. Antibody A3-IVA5 failed to elicit homotypic aggregation of transfectants harboring this derivative (data not shown; see Ref. 34). FACS analysis showed that this derivative failed to be recognized by A3-IVA5, although the other blocking mAb, P1B5, recognized ␣ 3 -D154A as well as the other ␣ 3 derivatives (Fig. 2, C and H). As derivatives that failed to bind epiligrin could be recognized by A3-IVA5 and P1B5, this indicates that recognition of substrate and of blocking mAbs does not necessarily involve interaction with identical residues. Furthermore, a region not previously demonstrated to be involved in recognition of natural substrates is recognized by a blocking mAb (Fig. 2, C and H) and appears to modulate binding to epiligrin.
Differential Effects of ␣ 3 Substitutions on Adhesion to Inva-sin-Adhesion to invasin by K562 cells expressing the ␣ 3 derivatives was determined in the presence of the blocking anti-␣ 5 ␤ 1 antibody VD1, which inhibits binding to the sole endogenous invasin receptor, the ␣5␤1 integrin (40). The ␣ 3 -Y218A derivative showed a small defect for cell adhesion to invasin, relative to wild type, when invasin was coated at either 1 or 5 g/ml (Fig. 4, A and B). This compares to the total loss of adhesion to epiligrin by this derivative (Fig. 3). The ␣ 3 -W220A and ␣ 3 -D154A derivatives, on the other hand, adhered to invasin as well as or better than transfectants harboring the wild type ␣ 3 chain (Fig. 4, A and B). That adhesion was dependent on ␣ 3 ␤ 1 was shown by the addition of the blocking anti-␣ 3 antibody A3-IVA5, which reduced adhesion of K562 cells expressing ␣ 3 -Y218A to background levels if wells were coated with 1 g/ml invasin or to 50% when coated at 5 g/ml invasin (relative to "inv ϩ VD1") (Fig. 4, A and B). Antibody A3-IVA5 decreased adhesion of cells expressing ␣ 3 -W220A or ␣ 3 -wt to 1 g/ml invasin to ϳ50% levels. A3-IVA5 had no effect on ␣ 3 -D154A-mediated adhesion to invasin consistent with results of the previous experiment indicating that this substitution derivative lacks the A3-IVA5 recognition epitope (Figs. 2 and 4, A and B). Bacterial Uptake via ␣ 3 Substitution Derivatives-As bacterial uptake into cultured cells is a more stringent assay than adhesion (33), the ability of ␣ 3 derivatives to internalize E. coli expressing invasin was determined. K562 transfectants were challenged with bacteria in the presence or absence of anti-␣ 5 and anti-␣ 3 antibodies, and uptake was assayed (see "Materials and Methods"). Transfectants harboring the ␣ 3 -wt, ␣ 3 -Y218A, ␣ 3 -W220A, and ␣ 3 -D154A derivatives were all capable of internalizing bacteria expressing invasin, although uptake promoted by ␣ 3 -Y218A was consistently 50 -75% of that seen with ␣ 3 -wt (Fig. 5). Interestingly, in the absence of anti-␣5 antibody, both ␣ 3 -Y218A and ␣ 3 -W220A displayed reduced uptake levels relative to wild type ␣ 3 transfectants (Fig. 5), indicating that binding to ␣ 5 ␤ 1 may be less active in promoting bacterial internalization than ␣ 3 ␤ 1 , and binding of bacteria to ␣ 5 ␤ 1 may interfere with access to ␣ 3 ␤ 1 . The antibody A3-IVA5 was able to inhibit uptake by all ␣3 derivatives, except ␣ 3 -D154A, indicating uptake was dependent on ␣ 3 ␤ 1 (Fig. 5).
Ligand-binding Defects of Three ␣ 4 Derivatives Harboring Mutations in Their Divalent Cation Binding Domains-As the effects of ␣ 3 chain mutations that eliminate integrin adhesion to natural substrate had little or no effect on cellular adhesion to invasin or on invasin-promoted uptake, we decided to compare these results to the analysis of other previously characterized mutants within human ␣ chains. Potentially, the higher affinity of invasin for integrin receptors relative to natural ligands means that single amino acid substitutions that affect natural substrate adhesion are not sufficiently drastic to eliminate adhesion to invasin. Masumoto and Hemler (29) isolated three integrin ␣ 4 substitutions containing the conservative substitutions N283E, D346E, or D408E at position 3, respectively, of the three proposed divalent cation-binding loops residing within N-terminal repeats 5-7 (Fig. 1). All three derivatives showed strong defects in their adhesion to VCAM-1 (10 -50% binding) and CS-1 (alternatively spliced fibronectin, no binding) relative to wild type ␣ 4 . Binding to invasin shows a moderate defect for each mutant (65-70% binding).
The invasin-binding phenotype of these K562 mutants was investigated further. Transfected cells were sorted using magnetic beads coated with the anti-␣ 4 antibody B-5G10 (see "Materials and Methods"), and surface expression was confirmed by FACS analysis (Fig. 6). When tested for binding to invasincoated surfaces in the presence of anti-␣ 5 antibody, the coating concentration of invasin necessary to support half-maximal binding of K562 transfectants was only slightly greater for the mutants than for wild type ␣ 4 (Fig. 7, A and B).
The ability of these transfectants harboring ␣ 4 integrin substitutions to internalize an E. coli inv ϩ strain was analyzed to determine if the small effects on adhesion had strong effects on uptake (see "Materials and Methods"). In contrast to the subtle defect of these mutants in adhesion to invasin-coated microtiter wells, the three ␣ 4 mutants displayed strong defects in ␣ 4 ␤ 1 -dependent, invasin-mediated bacterial uptake (Fig. 8). The strong defects observed support the model that moderate differences in adhesion can result in more pronounced defects in bacterial uptake (40). These results also indicate that mutations that cause defects in cellular adhesion to natural substrates can cause drastic defects in invasin-mediated bacterial uptake, in contrast to the results with the mutations in repeat III of the ␣ 3 chain. DISCUSSION The work presented here analyzed the relative effects of substitutions in the integrin ␣ chain on substrate adhesion. Whereas mutations in the third N-terminal repeat of ␣ 3 abolished adhesion to the natural integrin ligand epiligrin (laminin 5; Fig. 3), integrins harboring these changes showed only small defects in adhesion to invasin (Fig. 4) even when tested in the more stringent assay of bacterial uptake (Fig. 5). A control mutation in the analogous region of the second N-terminal repeat, ␣ 3 -D154A, gave the surprising result of enhanced adhesion to epiligrin. In contrast, integrins harboring substitutions in any of the three putative divalent cation binding domains of integrin ␣ 4 previously shown to be defective for natural ligand binding (29) were severely defective for invasinmediated bacterial uptake (Fig. 8). These results suggest that although ligand recognition may be mediated by the integrin ␣ chain third N-terminal repeat (11)(12)(13)(14)(15)(16), mutational change of any single residue in this region results in a receptor with considerable residual invasin binding activity. This is very different from results with natural ligands, in which substitutions of single critical residues eliminate adhesion. Furthermore, sequences in repeat II (Fig. 1) may play an important role in modulating the adhesion activity of the integrin, as a mutation in this repeat simultaneously stimulates adhesion to epiligrin and results in failure to bind a function blocking mAb.
Although the ␣ 3 substitution mutants were much more defective for adhesion to epiligrin than invasin, one derivative, ␣ 3 -Y218A, was more susceptible to antibody A3-IVA5 inhibition of invasin adhesion than was wild type receptor, consistent with this derivative being reduced in its affinity for invasin. Incomplete antibody inhibition of ␣ 3 -W220A and ␣ 3 -wt adhesion to invasin is most likely a reflection of the high affinity interaction of invasin with integrins ( Fig. 4; see Ref. 4). Similar results were found with the anti-␣ 3 antibody P1B5 (data not shown).
The unusually strong affinity of integrin binding to invasin probably plays a role in the observed differences in recognition of altered integrin derivatives by invasin and epiligrin. Presumably, an overlapping set of residues on the ␣ 3 chain is required for adhesion to both invasin and natural substrates. Consistent with this hypothesis, cell lines expressing ␣ 3 -Y218A showed defective adhesion to both a natural ligand and invasin, and mAb A3-IVA5 blocked interaction with both ligands. The high affinity of invasin for integrin receptors has the presumed consequence that single amino acid substitutions eliminating natural substrate adhesion are not sufficiently drastic to greatly affect adhesion to invasin. There are two likely explanations for why adhesion to invasin is more tolerant of such changes. First, during adhesion, invasin may engage more residues on the integrin receptor than do natural ligands. The overall binding energy contributed by these novel contacts may be sufficient to support adhesion to invasin in the absence of residues necessary for adhesion to natural substrates. Second, the residues engaged by invasin and natural substrates may be FIG. 5. Bacterial uptake into K562 cells harboring different ␣ 3 derivatives. 1 ϫ 10 6 cells were preincubated with the blocking anti-␣ 5 antibody VD1 in the presence or absence of anti-␣ 3 antibodies P1B5 or A3-IVA5 and challenged with ϳ1-5 ϫ 10 6 bacteria prior to determining uptake (n ϭ 3). Bacterial uptake is expressed as the percentage of bacteria added to the cells that were protected from the antibiotic, and error bars represent S.D. of four independent experiments done in triplicate.
identical, but the binding energy contributed by the individual unaltered residues may be greater for adhesion to invasin than to other substrates. As a result, the relative contributions of the mutated residues to invasin adhesion would be less than that observed for adhesion to a natural substrate.
The results presented here suggest that the ␣ 3 ␤ 1 integrin is more active than ␣ 5 ␤ 1 at promoting bacterial uptake. Substitution derivative ␣ 3 -W220A showed no defect in the invasin adhesion assay, yet it appeared to be attenuated, compared with wild type ␣ 3 , in its competition with endogenous ␣ 5 ␤ 1 for promoting bacterial uptake (Fig. 5). This is consistent with the result that an anti-␣ 5 antibody stimulated bacterial uptake in cell lines expressing ␣ 3 -W220A or ␣ 3 -Y218A, whereas the wild type derivative showed no stimulation of bacterial uptake by this treatment (Fig. 5). The reduced bacterial uptake by ␣ 3 -Y218A and ␣ 3 -W220A in absence of anti-␣ 5 ␤ 1 antibody is apparently a by-product of two phenomena. When ␣ 3 ␤ 1 competes with ␣ 5 ␤ 1 for adhesion to invasin, derivatives ␣ 3 -Y218A and ␣ 3 -W220A are presumably less efficient competitors of ␣ 5 ␤ 1 than is ␣ 3 -wt. In addition, bacterial adhesion to ␣ 3 ␤ 1 would be predicted to result in more efficient phagocytosis than adhesion to ␣ 5 ␤ 1 . ␣ 3 ␤ 1 has recently been shown to be specifically associated with tetraspan membrane proteins that are involved in a variety of signaling activities (41), whereas other integrins are apparently not associated with these proteins. Perhaps these integrin-associated proteins may provide important signals for invasin-mediated uptake.
All three ␣ 4 derivatives tested were dramatically defective for invasin-mediated bacterial uptake relative to wild type receptor, indicating that substitution in three different N-terminal repeats affected invasin binding. As the bacterial uptake defect of each of the ␣ 4 mutants is approximately equal (Fig. 8), this suggests that substitution within any one cation-binding loop is no more detrimental to ligand interaction than another. These three cation binding domains may regulate the structure of the integrin ␣ subunit, affecting the ligand-binding site at a distance.
It has recently been suggested that the integrin ␣ chain N-terminal repeats adopt a 7-bladed propeller structure resembling G-protein ␤ subunits (18,42,43). According to this model, ␣ 3 integrin chain residues Tyr-218 and Trp-220 reside on a solvent-exposed surface of the third blade (repeat) of the propeller. The finding that the ␣ 3 integrin substitution D154A in the second repeat is not recognized by the function-blocking antibody A3-IVA5 (Fig. 2) and stimulates ligand binding is surprising (Fig. 3) but consistent with this model. There are several explanations for the results regarding repeat II. The   FIG. 6. Sorting of K562 cells expressing integrin ␣ 4 derivatives. G418-resistant K562 cell transfectants were incubated with the anti-␣ 4 antibody B-5G10 and sorted with a secondary anti-mouse IgG antibody coupled to magnetic beads (see "Materials and Methods"). Following a 2-week grow out, cells were again labeled with B-5G10 and analyzed for ␣ 4 ␤ 1 surface expression by FACS. Histograms were generated to reflect fluorescence intensity due to cell labeling via fluorescein isothiocyanate-conjugated secondary antibody. second and third repeats may be in sufficiently close proximity to allow a single antibody to bind to the repeats or allow steric interference of ligand binding to repeat III by an antibody bound to the neighboring repeat II. Antibody binding to one domain may also disrupt the structure of its neighbor by causing a conformational change. Additionally, sequences that are proximal in the linear sequence to Asp-154 may be required for ligand interaction. Finally, sequences in repeat II could modulate binding of substrate to repeat III. The D154A mutation may lock the integrin into a high affinity state that allows high efficiency substrate adhesion. In the propeller model, ␣ 3 residues Tyr-218, Trp-220, and Asp-154 are in close proximity on a solvent-exposed surface of the molecule, consistent with each of these scenarios (18). Physical studies on the interaction of this region with substrates will directly test the validity of these models and provide an explanation for the apparent interaction between neighboring repeats.