Regulation of Integrin α5β1-Fibronectin Interactions by Divalent Cations

Integrin-ligand interactions are known to be dependent on divalent cations, although the precise role of cations in ligand binding is still unclear. Using the interaction between α5β1 and fibronectin as a model system, we have performed a comprehensive analysis of the effects of Mn2+, Mg2+, and Ca2+ on ligand binding. Each cation had distinct effects on the ligand-binding capacity of α5β1: Mn2+ promoted high levels of ligand binding, Mg2+ promoted low levels of binding, and Ca2+ failed to support binding. Studies of the effects of different combinations of cations on ligand binding indicated that the cation-binding sites within α5β1 are not all identical, or of broad specificity, but instead each site shows a distinct preference for one or more cations. Ca2+ strongly inhibited Mn2+-supported ligand binding, but this inhibition was noncompetitive, suggesting that Ca2+ recognizes different cation-binding sites to Mn2+. In contrast, Ca2+ acted as a direct competitive inhibitor of Mg2+-supported ligand binding, implying that Ca2+ can displace Mg2+ from the integrin. However, low concentrations of Ca2+ greatly increased the apparent affinity of Mg2+ for its binding site, suggesting the existence of a distinct high affinity Ca2+-binding site. Taken together, our results imply that the ligand-binding capacity of α5β1 can be regulated in a complex manner through separate classes of binding sites for Mn2+, Mg2+, and Ca2+.

has not yet been elucidated. The N-terminal portion of the integrin ␣ subunits comprises seven homologous, tandemlyrepeated domains of ϳ50 amino acid residues. Domains 4 -7 (or in some subunits, 5-7) contain sequence motifs similar to the Ca 2ϩ -binding EF-hands in proteins such as calmodulin. However, the integrin divalent cation-binding motif differs from classical EF-hand sequences in that it lacks an essential oxygenated residue at the Ϫz coordination position. Hence, it has been proposed that integrin ligands, such as RGD, may supply a crucial aspartate residue to complete the coordination geometry of the divalent cation (Corbi et al., 1987;Humphries, 1990). This hypothesis therefore suggests that divalent cations may act as bridge between ligand and receptor.
Chemical cross-linking experiments have provided direct evidence for the role of integrin EF-hand-like sites in ligand binding. Specifically, the binding site of a sequence from the fibrinogen ␥ chain has been mapped to the fifth repeat of ␣IIb (D'Souza et al., 1990), and a peptide corresponding to the EFhand-like sequence in this repeat bound to fibrinogen in a divalent-cation-dependent manner (D'Souza et al., 1991). Similarly, cross-linking of an RGD peptide to ␣V␤3 localized the ligand-binding site between the second and sixth domains of the ␣V subunit .
The importance of cation coordination in ligand binding by integrins has been demonstrated directly by the covalent coupling of Co(III) to ␣v␤3 (Smith and Cheresh, 1991). In addition, a recombinant fragment of ␣IIb that spans the EF-hand-like domains has been shown to contain multiple Ca 2ϩ binding sites (Gulino et al., 1992), and a modeling study of hybrid integrincalmodulin EF-hands predicts the integrin loop to support divalent cation chelation (Tuckwell et al., 1992).
A highly conserved region is found toward the N terminus of integrin ␤ subunits suggesting that this sequence may be functionally important. Cross-linking studies have shown that this region in the ␤3 subunit is proximal to the ligand-binding site (D'Souza et al., 1988;Smith and Cheresh, 1988). In addition, mutation of oxygenated residues in this region (Loftus et al., 1990;Bajt and Loftus, 1994) results in a receptor that is deficient in binding both cation and ligand. The sequence containing these residues shows some homology to a cation-binding sequence found in integrin I domains (Michishita et al., 1993;Bajt and Loftus, 1994;Lee et al., 1995). In an important recent advance, a synthetic peptide comprising this region of ␤3 (residues 118 -131) was shown to bind both the Ca 2ϩ analogue Tb 3ϩ and RGD peptides (D'Souza et al. 1994). Ligand binding caused displacement of cation from this peptide, and a similar displacement of cations from ␣IIb␤3 by ligands was also observed. Based on these results, a mechanism of integrin-ligand binding has been proposed, termed the "cation displacement hypothesis" (D' Souza et al., 1994). In this mechanism, cation, ligand, and receptor initially form a ternary complex in which ligand is bridged to the integrin through the cation, and cation is subsequently displaced from the ligand-binding site.
Cation binding by integrins has also been shown to be associated with conformational changes. For example, expression of the epitope recognized by monoclonal antibody (mAb) 1 24 on ␣L␤2 is dependent on Mg 2ϩ (or Mn 2ϩ ), and mAb 24 epitope expression correlates with the ability to bind ligand (Dransfield and Hogg, 1989;Dransfield et al., 1992aDransfield et al., , 1992b. Hence, an alternative hypothesis for the role of divalent ions in integrin function is that cation binding is required to cause a conformational change in the integrin that renders it competent to bind ligand. A number of important questions concerning the cation-binding sites on integrins are currently unresolved. First, do the cation-binding sites bind only one type of divalent cation, or do they all have a broad specificity? Second, is there only a single cation-binding site involved in ligand recognition or can occupancy of more than one site support ligand binding? Third, why does Mn 2ϩ confer a much higher ligand-binding affinity on many integrins than Ca 2ϩ or Mg 2ϩ (Gailit and Ruoslahti, 1988;Altieri, 1991;Elices et al., 1991;Dransfield et al., 1992b;Kern et al., 1993;Sanchez-Aparicio et al., 1993)?
The extracellular matrix glycoprotein fibronectin has served as a prototype substrate for the study of integrin-ligand interactions, and several regions of the molecule have been shown to be responsible for its adhesive activity. One domain that is recognized by a wide variety of cell types lies close to the center of the fibronectin subunit and contains the tripeptide RGD as a key active site (Pierschbacher and Ruoslahti, 1984;Yamada and Kennedy, 1984). The integrin ␣5␤1 is the major receptor for this central cell-binding domain (CCBD) and is expressed on many cell types. Here we have studied the role of Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ in modulating ␣5␤1-fibronectin interactions. We show that either Mn 2ϩ or Mg 2ϩ , but not Ca 2ϩ , can support ligand binding. However, Ca 2ϩ strongly modulates ligand binding supported by Mn 2ϩ or Mg 2ϩ and acts as a direct competitive inhibitor of Mg 2ϩ -supported binding but not of Mn 2ϩ -supported binding. Our results suggest the existence of distinct classes of cation-binding sites for each divalent ion.

EXPERIMENTAL PROCEDURES
Materials-Rat mAbs 16 and 13 recognizing the human ␣5 and ␤1 integrin subunits, respectively, were produced and purified as described previously (Akiyama et al., 1989). An 80-kDa fragment of fibronectin containing the CCBD was purified from a trypsin digest of plasma fibronectin as described by Garcia-Pardo et al. (1989). The synthetic peptide GRGDS was synthesized using Fmoc (N-9-fluoroenyl)methoxycarbonyl) chemistry on an Applied Biosystems 431A peptide synthesizer and purified as described previously (Humphries et al., 1986(Humphries et al., , 1987. Cell Attachment Assay-K562 erythroleukemia cells were obtained from the European Collection of Animal Cell Cultures (Porton Down, UK) and were grown in RPMI 1640 medium containing 10% (v/v) fetal calf serum, and 2 mM glutamine (all from Life Technologies, Inc., Paisley, Scotland, UK). Cell attachment assays were performed in 96well microtiter plates (Costar, High Wycombe, Bucks, UK). Wells were coated for 60 min at room temperature with 100-l aliquots of 80-kDa CCBD fragment (1 g/ml) diluted with Dulbecco's phosphate-buffered saline, and then sites on the plastic for nonspecific cell adhesion were blocked for 30 min at room temperature with 100 l of 10 mg/ml heat-denatured BSA. Cells were resuspended to 2 ϫ 10 6 /ml in 150 mM NaCl, 25 mM Hepes, 2 mM EDTA, pH 7.4, and incubated at 37°C for 30 min. Cells were then washed twice in Hepes-buffered saline (HBS; 150 mM NaCl, 25 mM Hepes, pH 7.4) and resuspended in the same buffer. Aliquots of cells (50 l) were then added to the microtiter wells and incubated with 50-l aliquots of Hepes-buffered saline containing 2ϫ the final concentration of divalent cations (MnCl 2 , MgCl 2 , or CaCl 2 ) for 20 min at 37°C in a humidified atmosphere of 6% (v/v) CO 2 . To estimate the reference value for 100% attachment, cells in quadruplicate wells coated with polylysine (500 g/ml) were fixed immediately by direct addition of 100 l of 5% (w/v) glutaraldehyde for 30 min at room temperature. Unbound and loosely bound cells in experimental wells were removed by shaking, and the remaining cells were then fixed as described above for reference wells. The fixative was aspirated, the wells were washed 3 times with 200 l of H 2 O, and attached cells were stained with Crystal Violet (Sigma) by a modification of the method of Kueng et al. (1989). 100 l of 0.1% (w/v) Crystal Violet in 200 mM MES, pH 6.0, was added to each well and incubated at room temperature for 1 h. Excess dye was removed by three washes of 200 l of H 2 O, and bound dye was solubilized with 100 l of 10% (v/v) acetic acid. The absorbance of each well at 570 nm was then measured using a multiscan ELISA reader (Dynatech, Billingshurst, UK). Each sample was assayed in quadruplicate, and background attachment to BSA was subtracted from all measurements.
Purification of a5␤1 Integrin from Human Placenta-Term placenta was obtained from Dr. J. Aplin, St. Mary's Hospital, Manchester, UK. Placenta (ϳ500 g) was cut into small chunks with scissors and homogenized in a blender (Philips) with 400 ml of buffer A (150 mM NaCl, 25 mM Tris-HCl, pH 7.4, 0.005% digitonin). The homogenate was stored at Ϫ70°C. Homogenate was thawed at room temperature and centrifuged at 5,000 ϫ g for 10 min. The pelleted material was then mixed with 600 ml of buffer A on ice for 10 min and centrifuged as above. The pellet was extracted on ice for 1.5 h with 400 ml of buffer B (150 mM NaCl, 25 mM Tris-HCl, pH 7.4, 2% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 2 mg/ml BSA). The extract was centrifuged at 6,000 ϫ g for 10 min and then at 40,000 ϫ g for 30 min. The supernatant was preadsorbed by passing it through a column of Sepharose 4B (30 ml), and then by mixing it with 10 ml of rat IgG-Sepharose (2 mg IgG/ml of beads) for 2 h on ice. IgG-Sepharose was then removed by centrifugation (5 min at 180 ϫ g) and column filtration, and the eluate was mixed with 8 ml of mAb 13-Sepharose (2 mg of IgG/ml of beads) for 2 h on ice. The suspension was then packed into a 1.6 cm diameter column (Pharmacia Biotech Inc.) and washed overnight (16 h) at 15 ml/h with buffer C (150 mM NaCl, 25 mM Tris-HCl, pH 7.4, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.1% Triton X-100). Bound material was eluted with buffer D (10 mM NaOAc, pH 3.5, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.1% Triton X-100) at 45 ml/h. 1.5-ml fractions were collected into 0.5 ml of 1 M Tris-HCl, pH 8.2. Aliquots of the fractions (25 l) were analyzed by SDS-polyacrylamide gel electrophoresis using a 6% nonreducing resolving gel and Coomassie Blue staining and found to contain ␤1 integrins of Ն90% purity.
Pooled fractions were then mixed with 2 ml of mAb 16-Sepharose (5 mg IgG/ml Sepharose) for 2 h on ice. The suspension was then packed into a 0.8-cm diameter column and washed with 12 ml of buffer C. Bound material was eluted with buffer D, and 0.5-ml fractions were collected and neutralized with 0.1 ml of 1 M Tris-HCl, pH 8.2. Aliquots of the fractions (25 l) were analyzed by SDS-polyacrylamide gel electrophoresis using a 6% nonreducing resolving gel. The only bands detected by Coomassie Blue staining were those corresponding to expected positions of the ␣5 and ␤1 subunits. ␣5 and ␤1 were the only integrin subunits detected in the eluted fractions by ELISA.
Purification of ␤1 Integrins from Peripheral Blood Mononuclear Cells-Mononuclear cells were purified from ϳ100 leukocyte concentrates (provided by the Blood Transfusion Service, Lancaster, UK) by centrifugation on Histopaque 1077 (Sigma, Poole, Dorset, UK). Cells were then washed with phosphate-buffered saline, centrifuged at 400 ϫ g for 10 min, and extracted with 100 ml of buffer B on ice for 30 min. The extract was clarified by centrifugation at 40,000 ϫ g for 30 min, and the resulting supernatant was precleared by rotary mixing with 4 ml of rat IgG-Sepharose beads for 1 h at room temperature. IgG-Sepharose was prepared by coupling at a ratio of 2 mg of IgG to 1 ml of CNBr-activated Sepharose (Pharmacia) according to the manufacturer's instructions. The IgG-Sepharose was removed by column filtration, and the filtrate was rotary mixed with 5 ml of mAb 13-Sepharose (2 mg IgG/ml Sepharose) for 2 h at room temperature. The suspension was then packed into a 1.6-cm diameter column and washed overnight at 4°C with buffer C. Bound material was eluted with buffer D, and 1.5-ml fractions were collected and neutralized with 0.5 ml of 1 M Tris-HCl, pH 8.2. Aliquots of the fractions (50 l) were analyzed by SDS-polyacrylamide gel electrophoresis using a 7.5% nonreducing resolving gel and Coomassie Blue staining and were shown to contain ␤1 integrins of Ն90% purity. ␣5␤1 was found to be a major component of this mixture by ELISA, although ␣2␤1, ␣4␤1, and ␣6␤1 were also present. Only trace amounts of other ␤1 integrins (␣3␤1 and ␣V␤1) could be detected by ELISA.
Solid Phase Receptor-Ligand Binding Assay-80-kDa CCBD fragment of fibronectin (500 g/ml in phosphate-buffered saline) was mixed with an equal mass of sulfo-N-hydroxysuccinimido biotin (Pierce, Chester, UK) and rotary mixed for 30 -40 min at room temperature. The mixture was then dialyzed against several changes of 150 mM NaCl, 25 mM Tris-HCl, pH 7.4, to remove excess biotin. Solid-phase receptorligand binding was performed by a modification of the method of Charo et al. (1990). Purified ␣5␤1 from placenta or ␤1 integrins from mononuclear cells (both at a concentration of ϳ0.5 mg/ml) were diluted 1:500 or 1:200, respectively, with phosphate-buffered saline containing divalent cations, and 100-l aliquots were added to the wells of a 96-well ELISA plate (Immulon 3, Dynatech). Plates were incubated overnight at room temperature, and wells were blocked for 1-3 h with 200 l of 5% (w/v) BSA, 150 mM NaCl, 0.05% (w/v) NaN 3 , 25 mM Tris-HCl, pH 7.4. Wells were then washed 3 times with 200 l of 150 mM NaCl, 25 mM Tris-HCl, pH 7.4, with 1 mg/ml BSA (buffer E). In experiments examining the role of cations in ligand binding, 100-l aliquots of biotinylated CCBD fragment (typically 0.1 g/ml) were diluted in buffer E with varying concentrations of MnCl 2 , MgCl 2 , or CaCl 2 (alone or in combination). For experiments examining the effect of antibodies or RGD peptides on ligand binding, biotinylated ligand diluted in buffer E with 1 mM MnCl 2 was added to the wells with or without antibodies or peptides. The plate was then incubated at 30°C for 3 h. Biotinylated ligand was aspirated, and the wells were washed 3 times with buffer A. Bound ligand was quantitated by the addition of 1:200 ExtrAvidinperoxidase conjugate (Sigma) in buffer E with 1 mM MnCl 2 for 10 min. Wells were then washed 4 times with buffer E, and color was developed using 2,2Ј-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (Sigma). Measurements obtained were the mean Ϯ S.D. of six replicate wells. The amount of nonspecific binding was measured by determining the level of ligand binding to wells coated with BSA alone; these values were subtracted from the corresponding values for receptor-coated wells. Results shown are representative of at least three separate experiments. Data shown in Figs. 1-3 were obtained using placental ␣5␤1, data shown in Figs. 4 -6 were obtained using ␤1 integrins from mononuclear cells, although both integrin preparations gave essentially identical results. Curve fitting (nonlinear regression analysis) to estimate apparent dissociation constants was performed as described previously (Mould et al., 1994). For double-reciprocal plots, data were normalized to zero ligand binding in the absence of the supporting cation (Mn 2ϩ or Mg 2ϩ ). Linear regression analysis of double-reciprocal plots was performed using SigmaPlot version 6; for clarity, not all of the data points are shown in some of these analyses. Under the conditions used in these assays (␣5␤1 coating concentration ϳ4 nM) the ELISA signal appeared to be directly proportional to the amount of bound ligand because plots of 1/absorbance versus 1/[free ligand] (not shown) did not deviate from linearity at high ligand concentrations.

RESULTS
Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ Have Distinct Effects on ␣5␤1-Fibronectin Interactions-To examine the role of divalent cations in regulating the activity of ␣5␤1, we compared the ability of Mn 2ϩ , Ca 2ϩ and Mg 2ϩ to modulate integrin function in a solid phase receptor-ligand binding assay and in a K562 cell attachment assay. We chose to use an 80-kDa fragment from the CCBD of fibronectin as the ligand in these experiments because it contains only one integrin recognition domain. Native dimeric fibronectin, which contains an ␣5␤1 recognition site in both subunits, could show cooperative binding to immobilized integrins. For the solid phase assays described in this report, we used either an affinity-purified preparation of ␣5␤1 from human placenta or a partially pure preparation of ␣5␤1 from human peripheral blood mononuclear cells; essentially identical results were obtained from these two preparations. In the initial characterization of this assay, binding of the CCBD fragment was found to be inhibited Ͼ90% by an antifunctional antibody to the ␣5 subunit and to be completely inhibited by either EDTA or GRGDS peptide (results not shown). K562 cells were chosen for the attachment assays because ␣5␤1 is the only ␤1 integrin expressed by these cells (Hemler et al., 1987). In agreement with previous studies Faull et al., 1993), we found that the attachment of K562 cells to the CCBD fragment was mediated solely by ␣5␤1 (data not shown).
Similar results were obtained in the two assay systems (Fig.   1, A and B). These data suggest that cell-surface ␣5␤1 and ␣5␤1 in solid phase assays behave in a similar manner with respect to the divalent cation dependence of receptor-ligand interactions. Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ had markedly different effects on ␣5␤1-fibronectin interactions. Both Mn 2ϩ and Mg 2ϩ promoted ligand binding, although Mn 2ϩ supported higher levels of binding than Mg 2ϩ . In contrast, Ca 2ϩ supported little or no binding. In the cell attachment assay, low levels of attachment were observed in the absence of cations, and this level was decreased with increasing concentrations of Ca 2ϩ . Comparison of the concentration of Mn 2ϩ and Mg 2ϩ to give halfmaximal ligand binding in the solid phase assay (Table I) suggested that the affinity of Mn 2ϩ for its binding site(s) on ␣5␤1 was ϳ40-fold higher than that of Mg 2ϩ . Scatchard-type analysis of the binding curves (not shown) indicated that there was only a single site (or a single class of sites) for Mn 2ϩ and Mg 2ϩ on ␣5␤1 for which cation occupancy supports ligand binding. Such sites have been termed "ligand-competent" sites Binding of biotinylated CCBD fragment in A or cell attachment in B was measured for a range of concentrations of each individual cation. q, Mn 2ϩ ; f, Mg 2ϩ ; , Ca 2ϩ . .
Since the solid phase assay was found to be highly reproducible, and also avoided possible artifacts from cation effects on cellular components other than integrins, we chose to use this assay for a detailed study of the effects of combinations of cations on ligand binding.
Ca 2ϩ Is a Noncompetitive Inhibitor of Ligand Binding Supported by Mn 2ϩ -Since Ca 2ϩ failed to support ligand binding, we first investigated whether Ca 2ϩ could reduce the high levels of ligand binding supported by Mn 2ϩ . Fig. 2 shows the result of an experiment in which the concentration of Mn 2ϩ was kept constant at 100 M and the concentration of Ca 2ϩ was varied. Ca 2ϩ strongly inhibited Mn 2ϩ -supported ligand binding, although even at a very high Ca 2ϩ concentration (8 mM) ligand binding was not abrogated. In contrast, Mg 2ϩ did not significantly inhibit Mn 2ϩ -supported binding even at 8 mM (results not shown).
To further analyze the effects of Ca 2ϩ on Mn 2ϩ -supported ligand-binding, we examined the effects of varying the concentration of Mn 2ϩ at constant Ca 2ϩ . Fig. 3A shows the inhibition of ligand binding by 8 mM Ca 2ϩ . Ca 2ϩ greatly reduced the maximal level of ligand binding but did not significantly alter the concentration of Mn 2ϩ required for half-maximal ligand binding. A double-reciprocal plot of these data (Fig. 3B) indicated that the inhibition observed at high Ca 2ϩ concentrations is noncompetitive in nature. A detailed analysis of the effects of lower Ca 2ϩ concentrations on Mn 2ϩ -supported ligand binding (not shown) suggested that Ca 2ϩ binding at multiple sites on the integrin was responsible for its inhibitory effects; however, Ca 2ϩ binding to any of these sites did not decrease the apparent affinity of Mn 2ϩ for its ligand-competent site. An important inference from these studies is therefore that, Ca 2ϩ does not compete with Mn 2ϩ for binding to the Mn 2ϩ ligand-competent site on ␣5␤1, and therefore appears to bind to different sites. Ca 2ϩ Can Stimulate or Inhibit Ligand Binding Supported by Mg 2ϩ -To investigate whether Ca 2ϩ could act as an inhibitor of ligand binding promoted by Mg 2ϩ , we initially investigated the effect of Ca 2ϩ on ligand binding supported by different Mg 2ϩ concentrations. At high Mg 2ϩ concentrations, only a partial inhibition of ligand binding was observed (result not shown), whereas for low Mg 2ϩ concentrations ligand binding was significantly increased at low Ca 2ϩ concentrations. Fig. 4 shows an experiment in which the concentration of Mg 2ϩ was kept constant at 50 M and the concentration of Ca 2ϩ varied between 0 and 4 mM. Two phases were apparent. In the first phase there was a marked stimulation in ligand binding, with a concentration of Ca 2ϩ for half-maximal increase of ϳ30 M and which reached a maximum at ϳ0.2-0.5 mM Ca 2ϩ . In the second phase, ligand binding was inhibited at high concentrations of Ca 2ϩ ; at very high Ca 2ϩ concentrations ligand binding approached the low levels observed in Ca 2ϩ alone (not shown). These results suggest that there are at least two Ca 2ϩ -binding sites on ␣5␤1 and that these sites have opposing influences on ligand binding supported by low concentrations of Mg 2ϩ . Binding of Ca 2ϩ to a high affinity site (the first phase) stimulates ligand binding, whereas binding of Ca 2ϩ to a low affinity site (the second phase) inhibits ligand binding. To analyze the first of these two phases, we tested the effect of a low concentration of Ca 2ϩ (0.25 mM) on ligand binding supported by different Mg 2ϩ concentrations (Fig. 5A). The results showed that low Ca 2ϩ concentrations have two effects on ligand binding; at low Mg 2ϩ concentrations ligand binding is increased, whereas at high Mg 2ϩ concentrations the maximal amount of bound ligand is reduced. The concentration of Mg 2ϩ required for half-maximal ligand binding in these assays (ϳ40 M) was decreased ϳ30-fold compared with that for Mg 2ϩ alone (see Table I). This increase in affinity of Mg 2ϩ for its ligand-competent site appears to be the mechanism by which Ca 2ϩ stimulates ligand binding at low Mg 2ϩ concentrations. A double-reciprocal plot of these data (Fig. 5B) showed that, at low concentrations, Ca 2ϩ is a mixed-type inhibitor of ligand binding supported by Mg 2ϩ ; it both increases the affinity of Mg 2ϩ for its site and reduces the maximal level of ligand binding.
To analyze the inhibition of Mg 2ϩ -supported ligand-binding by high Ca 2ϩ concentrations (the second phase in Fig. 4) we compared the effect of 0.2 mM Ca 2ϩ (a concentration that caused maximal stimulation of binding supported by low Mg 2ϩ concentrations) with that of 8 mM (Fig. 6A). High concentrations of Ca 2ϩ increased the concentration of Mg 2ϩ required for half-maximal ligand binding but did not affect the maximal amount of ligand bound at high Mg 2ϩ concentrations. The reciprocal plot (Fig. 6B) shows that the inhibition of ligand binding at high Ca 2ϩ concentrations is competitive in nature. By nonlinear regression analysis, the K I value was calculated as ϳ2 mM. Further analysis of the inhibition of Mg 2ϩ -supported ligand binding at different Ca 2ϩ concentrations (not shown) indicated that this inhibition is directly competitive, i.e. Ca 2ϩ is able to compete with Mg 2ϩ for binding to the Mg 2ϩ ligandcompetent site. However, when this site is occupied by Ca 2ϩ , the integrin fails to bind ligand.
Taken together, these data suggest: (a) that there is a Ca 2ϩbinding site of high affinity, the occupancy of which converts the Mg 2ϩ ligand-competent site into a high-affinity binding site The ability of either Mn 2ϩ , Mg 2ϩ , or Ca 2ϩ to support ligand binding was measured across a wide range of cation concentrations. The concentration of the ion required to support half-maximal ligand binding reflects the apparent dissociation constant (K D ) of the reaction for the ion . for Mg 2ϩ and (b) that Ca 2ϩ is, however, also able to competitively inhibit the binding of Mg 2ϩ to its ligand-competent site. However, since Ca 2ϩ binds this latter site with only low affinity, high concentrations of Ca 2ϩ are required to oppose the high affinity binding of Mg 2ϩ . Importantly, the observation that Ca 2ϩ can act as a direct competitive inhibitor of Mg 2ϩ -supported binding but not of Mn 2ϩ -supported binding also suggests that the ligand-competent sites for these two ions may be distinct.

DISCUSSION
In this report, we have performed a comprehensive analysis of the effects of Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ ions on the ligandbinding capacity of the integrin ␣5␤1. Our data show the following. (a) Only Mn 2ϩ and Mg 2ϩ support ligand binding. Although Ca 2ϩ does not support ligand binding, it strongly modulates ligand binding supported by Mn 2ϩ or Mg 2ϩ . (b) Ca 2ϩ is a noncompetitive inhibitor of Mn 2ϩ -supported ligand binding, suggesting that it does not compete with Mn 2ϩ for binding to the Mn 2ϩ ligand-competent site. (c) Ca 2ϩ can either enhance or inhibit Mg 2ϩ -supported binding, depending on the concentrations of each ion. The results suggest that Ca 2ϩ can compete directly with Mg 2ϩ for binding to the Mg 2ϩ ligand-competent site, but Ca 2ϩ binding to a separate high affinity site also greatly increases the affinity of Mg 2ϩ for its ligand-competent site. Taken together, these findings indicate that ␣5␤1 possesses several distinct cation-binding sites, each of which has a different specificity for Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ .
Our studies of the effects of Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ on fibronectin binding to ␣5␤1 showed that this interaction was strongly promoted by Mn 2ϩ , and to a lesser extent by Mg 2ϩ . This pattern has been observed for many other integrins including ␣1␤1 (Luque et al., 1994), ␣2␤1 (Staatz et al., 1989;Kern et al., 1993), ␣3␤1 (Weitzman et al., 1993), ␣6␤1 (Sonnenberg et al., 1988), ␣V␤1 (Kirchhofer et al., 1991), and ␣L␤2 (Dransfield et al., 1992b). Ca 2ϩ does, however, support ligand binding by a small number of integrins, including those of the ␤3 family (Kirchhofer et al., 1991;Smith et al., 1994). In a previous study (Gailit and Ruoslahti, 1988), Mn 2ϩ and Mg 2ϩ were found to support ligand binding by purified ␣5␤1 in liposomes, with similar values for the cation concentrations required for half-maximal ligand binding as reported here. However, Ca 2ϩ was also found to support ligand binding in the above study. Based on our findings that Ca 2ϩ can synergize with low concentrations of Mg 2ϩ , this result may have been due to contamination of the Ca 2ϩ samples with low concentrations of Mg 2ϩ . Similarly, an explanation for the low levels of K562 cell attachment we observed in the absence of exogenous cations or in the presence of low Ca 2ϩ concentrations is probably that small amounts of Mg 2ϩ are released from the cells during the time course of the experiment, high concentrations of Ca 2ϩ were observed to inhibit this effect. A recent study of myeloid cell adhesion to fibronectin confirms our finding that Ca 2ϩ alone does not support ␣5␤1-ligand interactions and that Ca 2ϩ also inhibits Mn 2ϩ -and Mg 2ϩ -supported adhesion (Davis and Camarillo, 1993).
We found that the affinity of ␣5␤1 for Mn 2ϩ was ϳ40-fold greater than that for Mg 2ϩ . It is a common feature of integrinligand interactions that typically Ͼ10-fold lower concentra- tions of Mn 2ϩ are required to support ligand binding than Mg 2ϩ (Altieri, 1991;Dransfield et al., 1992b;Kern et al., 1993;Michishita et al., 1993;Luque et al., 1994;Smith et al., 1994), indicating that many other integrins also contain one or more high affinity Mn 2ϩ -binding sites. An important implication from our observation that Ca 2ϩ could competitively inhibit Mg 2ϩ -supported ligand binding but not Mn 2ϩ -supported binding is that there may be separate ligand-competent sites on ␣5␤1 for Mn 2ϩ and Mg 2ϩ . This may shed light on why Mn 2ϩsupported ligand binding is of much higher affinity than that supported by Mg 2ϩ . We have also found that Mn 2ϩ causes a larger increase than Mg 2ϩ in the expression of an activation epitope on ␣5␤1 recognized by the mAb 12G10 (Mould et al., 1995), 2 suggesting that Mn 2ϩ is better than Mg 2ϩ at stabilizing a conformational change in the integrin required for ligand recognition.
The existence of a separate high affinity Ca 2ϩ -binding site, distinct from the Mg 2ϩ ligand-competent site, was suggested by the observation that low concentrations of Ca 2ϩ greatly increased the apparent affinity of Mg 2ϩ for its ligand-competent site. In summary, our studies suggest that at least three distinct cation-binding sites on ␣5␤1 are involved in the regulation of integrin activity; a tentative model of these sites is shown in Fig. 7. Site 1 binds Mn 2ϩ with high affinity; Ca 2ϩ does not appear to compete with Mn 2ϩ for binding to this site. Site 2 binds Mg 2ϩ with low affinity. Both sites 1 and 2 are ligandcompetent sites . Since, at high concentrations, Ca 2ϩ acts as a direct competitive inhibitor of Mg 2ϩ binding, Ca 2ϩ can also bind to site 2, although with low affinity. Site 3 is a Ca 2ϩ -binding site of high affinity with characteristics of the "effector" site proposed by Smith et al., (1994). Ca 2ϩ binding to this site dramatically increases the affinity of Mg 2ϩ for site 2. Ca 2ϩ binding at sites 2 and 3 (or possibly at additional sites) may be responsible for its ability to noncompetitively inhibit ligand binding supported by Mn 2ϩ . Our results clearly indicate that the cation-binding sites on ␣5␤1 are not all equivalent, neither are they of broad specificity, but instead each site shows a distinct selectivity for one or more cations. Binding of one cation to its site(s) can also affect the affinity of a second cation for its site(s); a similar cooperativity in cation binding has been observed for proteins such as calmodulin that contain multiple EF-hands (Strynadka and James, 1989). How the proposed cation-binding sites in Fig. 7 correspond to the putative divalent cation-binding sites in the ␣5 and ␤1 subunits and the molecular basis of their specificity will be the subject of future investigations. Such studies have the prospect of maping key sites involved in modulating integrin function.
The model shown in Fig. 7 is similar in several respects to that proposed to explain the effects of Ca 2ϩ and Mn 2ϩ on ligand binding to ␣V␤3 . In this study, Ca 2ϩ was found to be a mixed-type inhibitor of Mn 2ϩ -supported fibrinogen binding to ␣V␤3: low concentrations of Ca 2ϩ slightly increased the affinity of Mn 2ϩ binding to ␣V␤3 but decreased the amount of ligand bound at saturating Mn 2ϩ concentrations. Since high concentrations of Ca 2ϩ completely inhibited ligand binding, it was suggested that Ca 2ϩ could also competitively inhibit the binding of Mn 2ϩ to its site. Based on these results, a two-site model for ␣V␤3 was proposed in which Ca 2ϩ binds to an effector site and Mn 2ϩ (or Ca 2ϩ ) to a ligand-competent site . However, in a recent study of osteopontin binding to ␣V␤3 (Hu et al., 1995), it was found that Ca 2ϩ only decreased the association-rate of Mn 2ϩ -supported ligand binding but had no effect on the dissociation-rate, suggesting that Mn 2ϩ and Ca 2ϩ must bind to different sites on the integrin, with one cation influencing the on-rate and the other the offrate of ligand binding. Hence this latter report endorses the concept that there are cation-binding sites on integrins that selectively bind only one type of divalent ion.
Our model of the cation-binding sites in ␣5␤1 may be broadly applicable to the other integrins that show similar divalentcation requirements for ligand binding. For example, it has been shown for the ␣L␤2-ICAM-1 interaction that high concentrations of Ca 2ϩ can compete with Mg 2ϩ , but not with Mn 2ϩ , for binding to the integrin (Jackson et al., 1994). Low concentrations of Ca 2ϩ can, however, synergize with low concentrations of Mg 2ϩ to increase ligand binding (Marlin and Springer, 1987). Hence ␣L␤2 may have a similar arrangement of cation-binding sites to ␣5␤1.
One intriguing implication from our results is that there may be one cation-binding site on the integrin that selectively binds Mn 2ϩ . Whether or not Mn 2ϩ (or other transition metals) have a role in the regulation of integrin activity in vivo has been a matter of some speculation. It has been estimated that the concentration of Mn 2ϩ in tissues is in the range 1-14 M . Significant binding of fibronectin to ␣5␤1 was induced by these concentrations of Mn 2ϩ in vitro, hence Mn 2ϩ could potentially act as a physiological effector of ␣5␤1. Since the majority of Mn 2ϩ in the body is sequestered in bone, Mn 2ϩ might also be an important regulator of integrins during bone resorption. Alternatively, the effects of Mn 2ϩ observed in vitro could be fortuitous and not relevant in vivo. If it is possible to identify the Mn 2ϩ -binding site in ␣5␤1 (e.g. through site-directed mutagenesis), then several approaches could be adopted to resolve this question. This will be the subject of future work.
In conclusion, we have shown that Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ appear to recognize different sites on ␣5␤1 and that each ion has distinct effects on the capacity of ␣5␤1 to bind ligand. Since these cations have the ability to differentially regulate ␣5␤1 function, it will be important in the future to examine the in vivo consequences of fluctuations in divalent cation concentrations on ␣5␤1-mediated cell adhesion and migration, for example, during wound healing (Banai et al., 1990;Sank et al., 1989). FIG. 7. Model of the type and specificity of cation-binding sites in ␣5␤1. Occupancy of site 1 by Mn 2ϩ , or occupancy of site 2 by Mg 2ϩ , renders the integrin competent to bind ligand. Although Ca 2ϩ can compete with Mg 2ϩ for binding to site 2, Ca 2ϩ occupancy of this site does not permit ligand binding. Site 3 is a Ca 2ϩ -binding site of high affinity; occupancy of this site by Ca 2ϩ increases the affinity of Mg 2ϩ for site 2 (K D ϳ 40 M). The term "site" in the above context could refer either to individual cation-binding sites or to classes of site; our data do not allow us to distinguish between these possibilities.