Asp-698 and Asp-811 of the integrin alpha4-subunit are critical for the formation of a functional heterodimer.

The amino acid motif LDV is the principal binding site for alpha4 integrins in fibronectin, and homologous motifs are recognized in vascular cell adhesion molecule-1 and MAdCAM-1. Three conserved LDV motifs (LDV-1 to 3) occur in the ectodomain of the human and mouse alpha4-subunit, the functions of which are unknown. We demonstrate here that alpha4-transfected fibroblasts with mutation in LDV-1 (D489N) behaved like alpha4-wild type but that LDV-2 (D698N) and LDV-3 (D811N) mutants were impaired in binding and spreading on alpha4-specific substrates. On the RGD-containing fibronectin fragment FN-120 there was an inverse behavior; now the alpha4-wild type and the LDV-1 mutant could not adhere whereas the two other mutants could. The beta1 chain was critical for the differential integrin response. Biochemical analysis demonstrated that the LDV-2 and -3 mutations reduced the strength of the alpha4beta1 association, favored the formation of alpha5beta1, and prevented the expression of alpha4beta7 on the cell surface. Our results indicate that LDV-2 and LDV-3 are critical for the formation of a functional heterodimer. The presence of similar amino acid motifs in ligands and the alpha4-subunit suggest that metal coordination plays an important role in integrin-ligand binding as well as for heterodimer formation.

Integrins are a class of heterodimeric cell-surface molecules that are important mediators of cell function. In addition to providing the necessary mechanical stability for cell-cell or cell-matrix contact, integrins can transduce signals out of cells and into the cell from the microenvironment. The cytoplasmic portion of integrin chains can interact with signal-and cytoskeleton-associated proteins and thereby influence cellular properties like shape, adhesion, motility, growth, and differentiation (for review see Refs. 1 and 2).
An essential feature of integrins is their complex regulation of function which despite their biological importance as adhesive structures is not fully understood. Both subunits form a non-covalently linked heterodimer that appears to be important for transport to the cell surface as well as for ligand binding. It is believed that integrins contain multiple ligand contact sites, and several regions and residues have been iden-tified in both ␣ and ␤ chains. All of the identified contact sites carry amino acid sequences with oxygenated side chains that can potentially bind divalent cations. It has been estimated that integrin heterodimers can bind up to five divalent cations. The observation that all integrin ligands possess a critical aspartate or glutamate residue in their contact site to the integrin has led to the proposal that a metal ion provides a bridge between ligand and receptor. In addition to ligand binding, divalent cations are believed to contribute to the regulation of receptor conformation (for review see Refs. [3][4][5]. The ␣ 4 integrins, ␣ 4 ␤ 1 and ␣ 4 ␤ 7 , are adhesion molecules that play an important role in hematopoiesis (6,7), lymphocyte migration (8,9), mouse skeletal muscle formation (10), placental or cardiac development (11), and possibly tumor metastasis (12,13). VLA-4 (␣ 4 ␤ 1 ) mediates the adhesion of cells to fibronectin (14 -16) or the cytokine inducible endothelial cell ligand VCAM-1 1 (17,18). The heterodimer ␣ 4 ␤ 7 is a homing receptor that mediates the lymphocyte entry into the gut-associated lymphoid tissues (8,19). The ligand for ␣ 4 ␤ 7 is the addressin MAdCAM-1 (20). Only ␣ 4 ␤ 7 -positive cells can bind to MAdCAM-1, whereas both ␣ 4 ␤ 1 -and ␣ 4 ␤ 7 -positive cells can bind to fibronectin and VCAM-1 (21,22). Molecular studies have identified amino acid sequences in each ligand that are recognized by ␣ 4 integrins. A dominant binding site in fibronectin involves the LDV motif in the HepII/IIICS region, and the peptide surrounding and encompassing these residues has been termed CS-1 peptide (23,24). A homologous sequence IDS, present in domains 1 and 4 of VCAM-1, is essential for the binding to ␣ 4 integrins under static conditions (25)(26)(27) or flow (28), and an LDT sequence in the first domain of MAdCAM-1 is important for ␣ 4 ␤ 7 binding (29). Three conserved LDV motifs occur in the extracellular sequence of the ␣ 4 -subunit in mouse and man which have been termed LDV-1-3 (30).
We previously reported that the purified ␣ 4 -subunit as well as an LDV-containing peptide derived from the 80-kDa Nterminal portion could support the binding of lymphocytes via ␣ 4 ␤ 1 or ␣ 4 ␤ 7 (30) which could be important for ␣ 4 -mediated homotypic cell aggregation (22,31,32). In another study Ma et al. (33) have suggested that the LDV sites, in particular LDV sites 1 and 2, may serve as additional cation binding motifs that are required for cell adhesion. To understand further the functional role of the LDV motifs in the ␣ 4 -subunit, we have mutated the central aspartic acid to asparagine in each site in a consecutive fashion and have examined the effect on cell adhesion and spreading. Our results suggest a role for LDV-2 and -3 but not LDV-1 in the formation of a functional ␣␤ heterodimer. It is therefore possible that both ligand binding and the dimerization of ␣and ␤-subunits follow the same principal rule of metal ion coordination.
Peptides-The RGDS peptide was purchased from Sigma (Taufkirchen, Germany). The CS-1 peptide CEILDVPST was synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry and purified by preparative high pressure liquid chromatography. It was characterized further by analytical high pressure liquid chromatography and mass spectroscopy.
Cytofluorography-The cell-surface staining of cells with saturating amounts of mAbs, either hybridoma supernatants or purified antibodies, and phycoerythrin-conjugated goat antibodies to rat Ig (SERVA, Heidelberg, FRG) or hamster Ig (Dianova, Hamburg, FRG) has been described elsewhere (39). For the detection of cytoplasmic staining cells were first incubated on ice with 0.5% formaldehyde in PBS. After washing in PBS cells were permeabilized with 0.05% saponin (Sigma) for 10 min at room temperature. The cells were washed again and stained as above.
Stained cells were analyzed with a FACScan fluorescence-activated cell analyzer (Becton Dickinson, Heidelberg, FRG). For enrichment of ␣ 4 -transfectants the cells were stained under sterile conditions with mAb 5/3 and sorted on a FACS Vantage using counter staining with propidium iodide to exclude dead cells.
For the activation of the 9EG7 epitope, the cells were pretreated with peptide at the indicated concentration for 15 min followed by staining with the mAb and phycoerythrin-conjugated antibodies to rat IgG in the presence of the activating peptides.
Site-directed Mutagenesis and DNA Transfection-The mouse ␣ 4 -cDNA clone in Bluescript SKϩ was obtained from Dr. B. Holzmann (University of Munich, Germany) and was used as template for sitedirected mutagenesis with the Stratagene Chameleon kit (Stratagene, Heidelberg, Germany). In successive rounds of mutagenesis the Asp in positions Asp-489, Asp-698, and Asp-811 were changed to Asn as indicated in Fig. 1. Mutations were confirmed by DNA sequence analysis. Wild-type and mutated ␣ 4 -cDNAs were subsequently cloned in the pcDNA3 expression vector (Invitrogen) and transfected into Balb/3T3 fibroblasts using LipofectAMINE (Life Technologies, Inc., Eggenstein, Germany). Transfectants were selected for similar expression levels using FACS sorting (see above). The sorting of transfectants was repeated when the expression level of ␣ 4 integrin dropped by more than 20 mean fluorescent units.
Cell Adhesion and Spreading-Fibronectin or its fragments FN-120 or FN-40 (Life Technologies, Inc.) were coated to LABTEK glass chamber slides (Nunc, Wiesbaden, Germany) at a concentration of 10 g/ml or as otherwise indicated for 16 h at 4°C. Wells were blocked with 1% bovine serum albumin in PBS for 1 h at room temperature, washed with HBSS containing 10 mM HEPES, 2 mM Ca 2ϩ , and 2 mM Mg 2ϩ (binding buffer) and used for the assay. For adhesion, cells (5-10 ϫ 10 6 /ml) were suspended in the same buffer, and 0.2-ml aliquots were added to the coated slides. The binding assay was performed for 10 min at 37°C without shaking, and the slides were fixed in 2% glutaraldehyde/PBS after briefly dipping into PBS. For antibody or peptide blocking studies, cells were preincubated with purified antibody (10 g/ml final concentration, or as otherwise indicated) or peptides (500 g/ml or otherwise indicated) for 10 min at room temperature and then transferred to the chamber slides. For Mn 2ϩ activation, the Ca 2ϩ and Mg 2ϩ ions in the buffer were substituted with the indicated concentrations of Mn 2ϩ . Cell binding was measured by counting six independent 10 ϫ fields by video microscopy using IMAGE 1.55 software. To analyze cell spreading, the plated cells were incubated at 37°C at a microscope stage, and pictures were taken for 2 h every 10 min and stored on an optical disc. Pictures were analyzed, and the percentage of spread versus non-spread cells in each frame was determined. Clearly visible pseudopodia formation served as criterium for spread cells.
Binding of cells to human VCAM-1-transfected CHO cell monolayers was done using Eu 3ϩ -labeled cells (40). Briefly, cells were loaded with EuCl 3 for 1 h at 4°C, washed with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, and resuspended in Hanks' balanced salt solution plus 2 mM Ca 2ϩ and Mg 2ϩ . 100 l of the cell suspension (5 ϫ 10 5 cells/ml) were allowed to bind to the monolayer of VCAM-1-CHO cells in a 96-well plate for 30 min at 37°C. Following the binding assay bound and non-bound cells were separated on the basis of buoyant density using Percoll as described (44). For detection of the bound Eu 3ϩ -labeled cells, the plates were inverted to remove the Percoll and fixative, then washed once in PBS, and refilled with 50 l of PBS, and 50 l of Europium enhancement solution (LKB Wallac, Turku, Finland) was added. The fluorescence was measured in a time-resolved fluorometer (Arcus 1230, LKB-Wallac, Turku, Finland) after 30 min.
Isolation of RNA and Reverse Transcriptase-PCR Analysis-The isolation of total RNA from cells has been described in detail elsewhere (41). Total RNA (6 g) was transcribed into cDNA using Moloney murine leukemia virus reverse transcriptase (Promega, Heidelberg, Germany) and oligo(dT) 20 for priming. The RNA/DNA hybrid was treated with RNase H and used as template for PCR using an annealing temperature of 60°C and 35 cycles of 80 s. The following primers for the ␤ 7 integrin subunit were used: forward, ATAGGTTTTGGCTCCT-TCGTG; reverse, AGTGGAGAGTGCTCAAGAGTCACAGT. PCR products were separated on a 1% agarose gel containing 0.5 g/ml ethidium bromide. The mouse ␤ 7 cDNA clone was obtained from G. Krissansen, University of Auckland, New Zealand.
Biochemical Analysis-Lactoperoxidase-catalyzed iodination of intact cells was carried out as described (39). Following the labeling reaction, the cells were washed three times in PBS and lysed at 4°C for 15 min in Tris-buffered saline containing either 0.3% CHAPS (Sigma), 1% IGEPAL CA-630 (Sigma) or 1% IGEPAL CA-630 in the presence of 2 mM Ca 2ϩ and Mg 2ϩ ions. Lysates were prepared by centrifugation in an Eppendorf centrifuge at 4°C for 15 min and precleared with 30 l of packed rat Ig coupled to Sepharose. Immunoprecipitations were carried out using mAb 5/3 coupled to Sepharose or mAbs HM␤-1 preadsorbed to Protein G-Sepharose for 1 h at 4°C. The precipitates were washed 4 times in the respective lysis buffers and eluted from the Sepharose by boiling for 2 min in non-reducing SDS sample buffer. SDS-polyacrylamide gel electrophoresis was performed on 7,5% slab gels. Gels were dried and exposed to x-ray-sensitive films (Kodak Biomax-MS) using the Biomax MS intensifying screen. Structures of mutant ␣ 4 -subunits used in this study are shown. Aspartic acid to asparagine point mutations were introduced into the mouse ␣ 4 -cDNA consecutively rendering the LDV motifs to LNV. Known divalent cation binding sites in the ␣ 4 -subunit are also indicated. ␣ 4 -subunits (referred to as mut 1, mut 12, and mut 123) were transfected into ␣ 4 -negative Balb/3T3 fibroblasts and selected for equal cell-surface expression by FACS sorting. As shown in Fig. 2 the level of expression as detected by fluorescent staining with an ␣ 4 -specific mAb was comparable for all transfected cells. As revealed by staining with the respective mAbs, the expression levels of VLA-5, ␣v-integrins, as well as the ␤ 1 integrin subunit were not changed significantly following transfection. There was, however, induction of ␤ 7 surface expression in ␣ 4 -wt and mut 1 cells and little or undetectable expression in mut 12 and mut 123 cells, respectively. Induction of new ␤ chains upon transfection was also observed in another study (42).

Expression of Wild-type and Mutant
␣ 4 -Transfectants Are Altered in Adhesion-We investigated the adhesion of the different cells to immobilized fibronectin. As shown in Fig. 3A all cells could readily bind. The binding of non-transfected 3T3 cells is due to the expression of other integrins than ␣ 4 that can support adhesion to fibronectin (␣ 5 ␤ 1 , ␣v integrins see below). The binding site for ␣ 4 integrins is located on the FN-40 chymotryptic cleavage fragment of fibronectin. As shown in Fig. 3B the binding of ␣ 4 -wt and mut 1 cells to FN-40 was of similar magnitude and blocked in the presence of the mAb to the ␣ 4 -subunit. A partial blocking was obtained with mAbs against the ␤ 1 -or the ␤ 7 -subunit suggesting that the binding was mediated by ␣ 4 ␤ 1 and ␣ 4 ␤ 7 integrins. The mut 123 cells, and to a lesser extend mut 12 cells, showed a significant reduction in binding ability. Untransfected 3T3 cells did not bind.
To analyze whether a differential binding could be detected on other ␣ 4 -ligands, we tested the ability of the cells to bind to VCAM-1 stably expressed in CHO cells. As shown in Fig. 4A, the ␣ 4 -transfectants were indistinguishable in their adhesion to the full-length 7-domain VCAM-1 that contains two binding sites in domains 1 and 4, respectively. In contrast, a 1-3 domain deletion mutant of VCAM-1 (VCAM ⌬1-3) having only one binding site showed a clearly reduced binding of mut 12 and mut 123 cells (Fig. 4B). As expected, the 3T3 cells gave only background binding to VCAM-1-CHO.
Additional analysis of mut 123 cells were carried out on FN-40 by increasing the coating density of immobilized protein. Fig. 5A shows that although the adhesion of mut 123 cells was enhanced under these conditions, the level of ␣ 4 -wt cells was not reached. The adhesion of mut 123 cells was also improved in the presence of Mn 2ϩ ions that can activate ␣ 4 integrins. As FIG. 2. Analysis of ␣ 4 and ␤ 1 expression on transfected fibroblasts. Indirect immunofluorescence staining of the cells used in the present study is shown. Cells were stained by indirect immunofluorescence using mAb 5/3 (␣ 4 integrin subunit), HM␤-1 (␤ 1 integrin-subunit), HM␣5-1 (␣ 5 integrin subunit), Fib 30 (␤ 7 integrin subunit), and RMV-7 (␣v integrin subunit) followed by phycoerythrin-conjugated goat anti-rat IgG or anti-hamster IgG, respectively. For negative control the first antibody was omitted, and the peak fluorescence is indicated with an arrow. cient than ␣ 4 -wt cells. This could not be attributed to a lower ␣ 4 expression level in mutant cells since the phenomenon was also seen when ␣ 4 expression as detected by fluorescent staining was higher than on ␣ 4 -wt cells (not shown). Thus, the mutant ␣ 4 integrin receptors could be activated by Mn 2ϩ ions yet, due to an intrinsic defect, were less avid than the wild-type ␣ 4 integrin.
Presence of the ␣ 4 -Subunit Affects Spreading of Cells on FN-120 -We next analyzed the behavior of the cells on FN-120, which contains the RGD-binding site in fibronectin. Fig. 8 shows that 3T3 fibroblasts were able to spread on FN-120. Unexpectedly, the ␣ 4 -wt cells and mut 1 cells could not or only poorly spread on FN-120. In sharp contrast, in mut 12 and mut 123 cells spreading was again observed.
As indicated in Fig. 9, similar findings were made when the cells were analyzed for adhesion to FN-120. The adhesion of 3T3, mut 12, and mut 123 cells could be blocked in the presence of an RGDS peptide (Fig. 9A), by mAbs against the ␤ 1 integrin chain or by a mAb against ␣ 5 integrins, respectively (Fig. 9B). mAb RMV-7 against ␣v integrins had no or only marginal effects (data not shown). This observation suggested that on FN-120 the cells used mainly the ␣ 5 ␤ 1 integrin for adhesion and also indicated that mut 12 and mut 123 cells used similar integrin receptors for adhesion to FN-120 than 3T3 fibroblasts.
The differential behavior of the cells on FN-120 could not be explained by differences in cell-surface expression of the relevant cell surface receptors since the density of ␣ 5 and ␤ 1 was comparable on all cells (see Fig. 2). To demonstrate further that the effects observed were caused by the ␣ 4 -subunit and not due to an intrinsic failure of the fibroblasts, we sorted revertant cells that had lost ␣ 4 expression. These ␣ 4 -negative or low expressing cells were fully restored in their ability to spread on FN-120 (data not shown). We concluded, that the ␣ 4 -wt or the mut 1 subunit altered the function of the RGD-binding integrins in transfected cells. The removal of Asp-698 and Asp-811 by mutagenesis released these RGD-binding integrins from the block of function.
Modulation of the 9EG7 Epitope by Soluble Ligand-To demonstrate a differential activity of either the ␣ 4 ␤ 1 or RGD-binding integrins also in response to soluble ligands, the expression of the 9EG7 epitope, which is a conformation-dependent epitope of the ␤ 1 chain (38), was investigated. The expression of this epitope can be enhanced by Mn 2ϩ ions or soluble ligand (43). The binding site for mAb 9EG7 has been mapped to the cysteine-rich membrane proximal site of the ␤ 1 chain (43). As shown in Fig. 10A, the ␣ 4 -wt transfected cells and mut 1 cells showed staining with mAb 9EG7 which was of similar magnitude as 3T3 cells. In contrast, both mut 12 and mut 123 cells showed a consistently higher staining suggesting that the epitope was more accessible. These differences were not seen with mAb HM␤-1 which binds to a site distinct from the 9EG7 epitope on the ␤ 1 chain.
The modulation of the 9EG7 epitope expression following addition of soluble ligand in the form of the CS-1 peptide for ␣ 4 integrins or the RGDS peptide for RGD-binding integrins, respectively, are shown in Fig. 10, B and C. An increase in the concentration of CS-1 peptide led to a drastic up-regulation of 9EG7 epitope expression in ␣ 4 -wt cells. The mut 123 cells increased the expression of the 9EG7 epitope only marginally, whereas 3T3 fibroblasts showed no increase. Addition of RGDS peptide to the cells showed an inverse behavior (Fig. 10C). Now the 3T3 cells gave an appreciable dose-dependent response in the expression of the 9EG7 epitope. Also the mut 123 cells showed a dose-response curve comparable to 3T3 fibroblasts, although these cells started at a much higher expression level (see above). In contrast, the ␣ 4 -wt cells gave a poor response to the RGDS peptide consistent with the previous notion that RGD-binding integrins were impaired in these cells.
LDV-2 and -3 Mutations Affect the Stability of the ␣ 4 ␤ 1 Heterodimer-The functional data had indicated that mutations in the ␣ 4 -subunit were able to regulate the integrin response in the transfected cells and that the common ␤ 1 chain might be of crucial importance. To address this question further biochemical analyses were carried out. Immunoprecipitation with the ␣ 4 -specific mAb showed the typical pattern of ␣ 4 integrins on SDS-polyacrylamide gel electrophoresis consisting of the 150-kDa band representing the intact ␣ 4 chain and the smaller bands of 80 and 70 kDa that are proteolytic cleavage fragments of the ␣ 4 chain (see Fig. 11A). The mutations slightly affected the electrophoretic mobility of these fragments.
We analyzed the stability of the ␣ 4 ␤ 1 heterodimer from each cell type in the presence of different detergents. Following cell-surface iodination, aliquots of each transfectant were solubilized in 0.3% CHAPS, 1% IGEPAL CA-630, or 1% IGEPAL CA-630 in the presence of 2 mM Ca 2ϩ and Mg 2ϩ ions and the lysates were subjected to precipitation using mAbs to the ␣ 4 -or ␤ 1 -subunit. As shown in Fig. 11, A and B, the different solubilization conditions preserved the heterodimer in ␣ 4 -wt and mut 1 cells since the 80-and 70-kDa subunits were present in both ␣ 4 and ␤ 1 precipitates in similar amounts. Mut 12 and mut 123 cells showed a different behavior. Here the 70-and 80-kDa subunits were equally present only in the ␣ 4 -specific precipitates, whereas in the anti-␤ 1 immunoprecipitates the heterodimer was only detectable in the presence of Ca 2ϩ and Mg 2ϩ (compare mut 12 or mut 123 A versus B). IGEPAL CA-630 alone almost completely disrupted the heterodimer, whereas 0.3% CHAPS showed a partial dissociation of the complex. This differential sensitivity toward detergents indicated that in mut 12 and mut 123 cells the ␣ 4 ␤ 1 heterodimer was significantly weaker or present in much smaller amounts than in ␣ 4 -wt or mut 1 cells. There was also a clearly reduced presence of ␤ chains in the ␣ 4 -specific precipitates of mut 12 and mut 123 cells (see Fig. 11A).
The additional analysis of the ␤ 1 -associated proteins by shorter exposure of the gels presented in Fig. 11B revealed another important finding. As shown in Fig. 11C the ␣ 5 integrin chain was found to be complexed to ␤ 1 only in mut 12 and mut 123 cells and was barely detectable in ␣ 4 -wt and mut 1 cells.

LDV-2 and -3 Mutations Prevent Surface
Expression of ␣ 4 ␤ 7 -Finally we reinvestigated the failure of mut 12 and mut 123 cells to express ␤ 7 at the cell surface following transfection. We used mAb DATK32 that recognizes a conformational epitope of ␣ 4 ␤ 7 when it is associated (34). Fig. 12A shows that ␣ 4 -wt and mut 1 cells showed cell-surface staining with this mAb, whereas mut 12 and mut 123 cells were negative. This was not due to the lack of ␤ 7 message that was detectable by PCR analysis of reverse-transcribed cDNA in all transfected cell lines as well as in 3T3 cells (Fig. 12B). Indeed, cytoplasmic staining of permeabilized cells with a ␤ 7 mAb revealed the presence of ␤ 7 protein in all transfectants (not shown). We concluded that the mutations in the ␣ 4 -subunit most likely affected the formation of the ␣ 4 ␤ 7 heterodimer or its transport to the cell surface. DISCUSSION The goal of the present study was to assess the functional role of the three LDV sites in the mouse ␣ 4 -subunit by site- FIG. 6. Spreading of 3T3 and ␣ 4transfectants on FN-40. Spreading of 3T3 fibroblasts and ␣ 4 -transfectants on the FN-40 fragment coated at 10 g/ml is shown. The spreading of cells was monitored by video microscopy, and pictures were taken at 20-min intervals. Data shown were taken at 40 min after plating of the cells, and the percentage of cells spread was evaluated. Blocking of cell spreading is shown for ␣ 4 -wt cells using mAb 5/3 at 10 g/ml. The mAb HM␣5-1 was used as control.
directed mutagenesis. We find that Asp 3 Asn substitution in LDV-2 and -3 but not in LDV-1 affects adhesion and spreading of transfected fibroblasts on ␣ 4 -specific substrates. The ␣ 4 mutant cells were differentially affected in adhesion and spreading on FN-40. Whereas the mut 12 cells were mainly compromised in spreading, the mut 123 cells were affected in both adhesion and spreading suggesting that the mutation of the LDV-3 site was giving an additional effect. In the presence of Mn 2ϩ ions the defects of mut 12 and mut 123 cells could be largely restored, and the lack of spreading on FN-40 was overcome by prolonged assay times. This suggested that the receptor was not completely non-functional and that LDV-2 and -3 were probably the most important in events early after ␣ 4 ␤ 1 receptor engagement. Ma et al. (33) have performed similar mutagenesis studies using K562 cells transfected with human ␣ 4 -subunits in which the LDV sites were mutated to LEV. These authors demonstrated impaired adhesion of D489E and D698E mutants (equivalent to LDV-1 and -2) on FN-40 and a 1-3 domain form of VCAM-1. There was, however, no effect in D811E mutants (33). The reasons for the discrepancies between the previous study and our data are not entirely clear but could be due, at least in part, to the choice of amino acid exchanges. Whereas Asp 3 Glu mutations retain the negative charge in the amino acid side chain, our Asp 3 Asn mutations abolished it. Also the type of mutagenesis (single versus accumulating mutations) is different, and cell spreading was not analyzed by Ma et al. (33). Based on sequence comparison it was proposed that Asp-698 and possibly -489 (identical with LDV-2 and -1) may be putative metal-binding sites that might be required to stabilize a protein-protein interaction. The reason why these sites might be important for ␣ 4 integrin function remained unclear.
Our results provide suggestive evidence that the LDV-2 and -3 sites are important for a functional interaction with the ␤ chain. We postulate that in the presence of a transfected ␣ 4subunit the formation of a functional ␣ 4 ␤ 1 heterodimer is dominant. The formation of other heterodimers, i.e. ␣ 5 ␤ 1 , is suppressed presumably due to the lack of available ␤ 1 chains, and these ␣-subunits reach the cell surface by other means. The mutations of LDV-2 and -3 impair the ability of ␣ 4 to interact with ␤ 1 and favor again the formation of ␣ 5 ␤ 1 at the cell surface.
Evidence for this hypothesis came from several independent approaches. First we studied the binding of ␣ 4 -transfectants to FN-120. We found that in the presence of a transfected ␣ 4 -wt or mut 1 subunit the fibroblasts were unable to adhere and spread via ␣ 5 ␤ 1 on FN-120. In contrast, mut 12 and mut 123 cells could again spread and adhere and behaved similar to the parental 3T3 cells. We could exclude the possibility that the observed differences in the binding were the results of phenotypic changes in the transfectants since the expression levels of ␣ 5 , ␣v, ␣ 4 , and ␤ 1 were comparable. Furthermore, back selection of transfectants for ␣ 4 loss or low expressing variants fully restored the ability to spread on FN-120, supporting the view that the suppression was due to the presence of the ␣ 4 -subunit and not an intrinsic failure of the cells. Importantly, the dependence of the phenomenon on amino acid substitutions in the extracellular part of the molecule argued against an involvement of cytoplasmic proteins that may be required for integrin function. It rather appeared that cis-type of interactions with other proteins at the membrane surface was the reason.
An important observation was that in the binding of cells both to FN-40 and to FN-120 the ␤ 1 chain was involved. In particular, the inverse behavior of the transfectants on both types of substrates was striking. This led us to consider that the ␤ 1 chain might be a decisive factor in regulating the integrin response in the mutants. That indeed the ␤ 1 chain was crucial was supported by studies on the 9EG7 epitope that has been characterized as a ligand-induced binding site of the ␤ 1 chain (43). Interestingly, the epitope for this mAb was located in the cysteine-rich site of the ␤ 1 -subunit (43) which is juxtaposed to the site on the ␣ 4 -subunit where Asp-698 and Asp-811 are located. The level of staining for this epitope in the absence of any ligand was consistently higher on mut 12 and mut 123 cells than on ␣ 4 -wt or mut 1 cells. Thus, the LDV-2 and -3 mutations in the ␣ 4 chain induced a change in the accessibility of this particular ␤ 1 chain epitope that is usually up-regulated in the presence of bound ligand or activating divalent cations but also by the addition of EDTA (41). To study the effect of ligand binding on the 9EG7 epitope, we exposed the ␣ 4 ␤ 1 heterodimer to CS-1 peptide in solution. A clear-cut up-regulation was seen in ␣ 4 -wt cells. In contrast, there was only little change in mut 123 cells indicating that the epitope was already fully exposed in the first place and could not be up-regulated much further. These data demonstrated that the phenotype of the ␣ 4 -subunit, either mutated or not, was mirrored in the 9EG7 epitope presumably indicating a conformational change of the ␤ 1 -subunits imposed by the respective ␣ 4 chain. Thus, the 9EG7 epitope acted as an indicator of a change in the ␣␤ conformation.
More direct proof that the LDV-2 and -3 mutation affected the conformation of the ␣␤ heterodimer came from biochemical studies. By using different detergents for the solubilization of the cells, we noticed a much decreased stability of the ␣ 4 ␤ 1 heterodimer in mut 12 and mut 123 cells. In the absence of divalent cations the heterodimer could not resist the detergent Note that in mutant cells the 70-and 80-kDa ␣ 4 -subunits migrate slightly differently from wild-type ␣ 4 -transfectants. B, lysates were incubated with mAb HM␤-1 adsorbed to protein G-Sepharose. All samples were analyzed by SDS-polyacrylamide gel electrophoresis under nonreducing conditions. C, shorter exposure of B. The bracket shows the area of the gel. The position of the ␣ 5 -subunit is indicated and was confirmed by precipitation with mAb HM␣5-1 in parallel experiments. milieu as evidenced by the reduced presence of ␤ 1 chains in the ␣ 4 -specific precipitates and the failure to detect the ␣ 4 80 and ␣ 4 70-kDa fragments in anti-␤ 1 -specific precipitates of mut 12 and mut 123 cells. This could mean that the complex was only very poorly associated or even free ␣ 4 chain was present at the surface of these cells. Due to the lack of heterodimer-specific mAbs in the mouse, at the present we cannot distinguish between these two possibilities. Precipitation analysis of ␤ 1 -associated integrin subunits also revealed that the ␣ 5 chain was detectable only in mut 12 and mut 123 cells and not present or only weakly present in ␣ 4 -wt and mut 1 cells. Thus, the precipitation analysis reflected the results from the functional analysis of mutant cells.
Further evidence for a role of LDV-2 and -3 in heterodimer formation came from studies on the expression of ␣ 4 ␤ 7 in mutant cells. At the cell surface a ␣ 4 ␤ 7 heterodimer was only seen in ␣ 4 -wt and mut 1 cells, whereas in mut 12 and mut 123 cells it was not detectable. Despite this the ␤ 7 -subunit was available in all mutant cells as detected by reverse transcriptase-PCR and cytoplasmic staining. It is likely that the mutant ␣ 4 -subunits had a decreased ability to interact with the ␤ 7 chain thus preventing the assembly of the heterodimer or its transport to the cell surface.
The ␣ 4 and other integrins can physically interact with transmembrane-4 superfamily proteins like CD81 (TAPA-1) and others in the cell membrane of different cell lines (44). The binding site for TAPA-1 in the ␣ 4 chain is not entirely clear but has been mapped outside the ␣ 4 cytoplasmic tail (44). Transmembrane-4 superfamily proteins can associate with several integrins, but a direct role in the regulation of ␣ 4 integrin function has so far not convincingly been demonstrated. Although a possible interaction of these molecules with ␣ 4 in our transfected fibroblasts has to be considered, we do not regard this as a reasonable mechanism to explain our results.
Sanchez-Aparicio et al. (45) have reported that stimulation of human cells with the ␤ 1 -specific mAb TS2/16 could not only activate ␣ 4 ␤ 1 but also led to recognition of the RGDS sequence in fibronectin. Thus, the conformational change induced in ␣ 4 ␤ 1 by the mAb resulted in the ability to recognize the RGD sequence. Could the mutated ␣ 4 -subunit in our transfectants induce a similar change? The data presented in Fig. 9 argue against this possibility. It is evident that the adhesion of mut 12 and mut 123 cells to FN-120 was dependent on ␣ 5 ␤ 1 since it was blocked by the respective mAb but not by the mAb against the ␣ 4 integrin. Divalent cations regulate integrins in a complex way. All integrin-ligand interactions are divalent cation-dependent, and putative metal ion-binding sites have been identified in ␣and ␤-subunits (see Refs. [3][4][5]. The ␣-subunit of all integrins contain 3-4 divalent cation binding modules with homology to the EF-hand Ca 2ϩ -binding motifs (46). In the I domain which is present in many but not all ␣-subunits a metal ion-dependent adhesion site is present that coordinates Mg 2ϩ (47), and a metal ion-dependent adhesion site-like motif is also present in the most conserved region of ␤ integrin subunits (47). All these regions are involved in ligand binding implying that the ligandbinding pocket is complex and involves both subunits. Since the ligand epitopes recognized by integrins are often short acidic peptide motifs with central oxygenated amino acids, it has been suggested that receptor-bound cation might act as an integrinligand bridge. Recently a model was proposed for the ␣ chain 7-fold repeats that represent about 40% of the extracellular portion of the ␣ chains (48). This model predicts a ␤-propeller domain composed of seven blades made out of ␤-sheets around a central axis (48). The I domain is inserted into the ␤-propeller between the second and the third blade and is predicted to sit on the upper rim ot the ␤-propeller domain (49). The ␤-propeller model and the crystal structure of the ␣-subunit I domain from CD11a and CD11b (47,49) have allowed for the first time to propose a dynamic quaternary structure model of integrinligand interaction sites (5).
Outside the region of ligand contact the structure and conformation of integrin subunits is less well defined. It is known that the ␣and ␤-subunits form a heterodimer also in the absence of ligand and that withdrawal of divalent cations by chelating agents destabilizes the ␣␤ heterodimer (50 -52). It is quite possible that the conformation of the ␣␤ heterodimer also depends on multiple interactions that, at least to some part, uses metal ion bridging for stabilization. The metal ion-dependent adhesion site-like domain of several integrin ␤-subunits was suggested to be important for the association with the ␣-subunit (53)(54)(55). Interestingly, the ␤-propeller model predicts that the EF-hand Ca 2ϩ -binding motifs are located near one another on the lower surface of the ␤-propeller domain and might be involved in interactions with the ␤-subunit rather than in ligand binding (49). The LDV-2 and -3 sites are located on the stalk region of the ␣ 4 -subunit. Due to this localization they are probably not directly involved in ligand binding but may be necessary to establish a proper association with the ␤ chain. The lack of this putative association affects the affinity of the receptor and leads to defects that seem to be important in the early phase of ligand binding. These defects in receptor function of mutant cells were restored, although not completely, in the presence of Mn 2ϩ ions and by prolonged assay times. It is possible that the presence of Mn 2ϩ enforced additional binding sites. Collectively, our data suggest that LDV-2 and -3 sites represent important contact sites between ␣ 4 and its ␤-subunits. The presence of similar amino acid motifs in ligands and the ␣ 4 -subunit suggests that metal coordination plays an important role in integrin-ligand binding as well as for heterodimer formation.