Integrin α3β1 Binding to Fibronectin Is Dependent on the Ninth Type III Repeat*

Background: The fibronectin (Fn) ninth type III repeat can modulate integrin binding and resulting cell spreading. Results: Mutations within the Fn integrin binding domains affect integrin α3β1 binding. Conclusion: Integrin α3β1-fibronectin binding depends on the presence and spacing of the RGD and synergy sites within Fn. Significance: α3β1-fibronectin binding may modulate epithelial cell wound healing responses. Fibronectin (Fn) is a promiscuous ligand for numerous cell adhesion receptors or integrins. The vast majority of Fn-integrin interactions are mediated through the Fn Arg-Gly-Asp (RGD) motif located within the tenth type III repeat. In the case of integrins αIIbβ3 and α5β1, the integrin binds RGD and the synergy site (PHSRN) located within the adjacent ninth type III repeat. Prior work has shown that these synergy-dependent integrins are exquisitely sensitive to perturbations in the Fn integrin binding domain conformation. Our own prior studies of epithelial cell responses to recombinant fragments of the Fn integrin binding domain led us to hypothesize that integrin α3β1 binding may also be modulated by the synergy site. To explore this hypothesis, we created a variety of recombinant variants of the Fn integrin binding domain: (i) a previously reported (Leu → Pro) stabilizing mutant (FnIII9′10), (ii) an Arg to Ala synergy site mutation (FnIII9R→A10), (iii) a two-Gly (FnIII92G10) insertion, and (iv) a four-Gly (FNIII94G10) insertion in the interdomain linker region and used surface plasmon resonance to determine binding kinetics of integrin α3β1 to the Fn fragments. Integrin α3β1 had the highest affinity for FnIII9′10 and FnIII92G10. Mutation within the synergy site decreased integrin α3β1 binding 17-fold, and the four-Gly insertion decreased binding 39-fold compared with FnIII9′10. Cell attachment studies demonstrate that α3β1-mediated epithelial cell binding is greater on FnIII9′10 compared with the other fragments. These studies suggest that the presence and spacing of the RGD and synergy sites modulate integrin α3β1 binding to Fn.


Fibronectin (Fn) is a promiscuous ligand for numerous cell adhesion receptors or integrins. The vast majority of Fn-integrin inter-
actions are mediated through the Fn Arg-Gly-Asp (RGD) motif located within the tenth type III repeat. In the case of integrins ␣IIb␤3 and ␣5␤1, the integrin binds RGD and the synergy site (PHSRN) located within the adjacent ninth type III repeat. Prior work has shown that these synergy-dependent integrins are exquisitely sensitive to perturbations in the Fn integrin binding domain conformation. Our own prior studies of epithelial cell responses to recombinant fragments of the Fn integrin binding domain led us to hypothesize that integrin ␣3␤1 binding may also be modulated by the synergy site. To explore this hypothesis, we created a variety of recombinant variants of the Fn integrin binding domain: (i) a previously reported (Leu 3 Pro) stabilizing mutant (FnIII910), (ii) an Arg to Ala synergy site mutation (FnIII9 R 3 A 10), (iii) a two-Gly (FnIII9 2G 10) insertion, and (iv) a four-Gly (FNIII9 4G 10) insertion in the interdomain linker region and used surface plasmon resonance to determine binding kinetics of integrin ␣3␤1 to the Fn fragments. Integrin ␣3␤1 had the highest affinity for FnIII910 and FnIII9 2G 10. Mutation within the synergy site decreased integrin ␣3␤1 binding 17-fold, and the four-Gly insertion decreased binding 39-fold compared with FnIII910. Cell attachment studies demonstrate that ␣3␤1-mediated epithelial cell binding is greater on FnIII910 compared with the other fragments. These studies suggest that the presence and spacing of the RGD and synergy sites modulate integrin ␣3␤1 binding to Fn.
Cells interact with their surrounding extracellular matrix (ECM) 2 via transmembrane cell surface receptors, known as integrins. Integrins are heterodimeric proteins consisting of one ␣ and one ␤ subunit, which are known to form at least 24 unique heterodimers (1). Integrin interactions with their ECM ligands facilitate a host of cellular responses, including cell spreading, migration, proliferation, and differentiation, and can contribute to more orchestrated cellular events, such as angiogenesis and epithelial to mesenchymal transitions, among others (2)(3)(4). Integrin binding to ECM ligands occurs through specific binding sequences, the most notable of these sequences being Arg-Gly-Asp (RGD), which is found on a large number of ECM proteins, including fibronectin, vitronectin, osteoponin, laminin, thrombospondin, and several others (5). Furthermore, integrin heterodimers can interact with multiple ECM ligands and bind to multiple binding sequences (6,7).
Fibronectin (Fn) is a widely expressed ECM protein and is known to bind at least 16 integrins. Biochemically, Fn exists as a soluble dimeric glycoprotein composed of two nearly identical 230 -270-kDa monomers linked covalently near their C termini by a pair of disulfide bonds (8,9). Each monomeric subunit consists of three types of repeating modules: types I, II, and III. These modules comprise functional domains that mediate interactions with other ECM components, cell surface receptors, and Fn itself (9). Whereas type I and II repeats are structurally stabilized with two intrachain disulfide bonds in each repeat, type III repeats have no disulfide bonds and therefore are highly sensitive to external stimuli, including application of force, resulting in alterations of conformation of the molecule (8, 10 -12). Interestingly, a large number of Fn-integrin interactions occur through the RGD site located on the tenth type III repeat. The recognition of this simple tripeptide sequence can be quite complex and greatly depends on flanking residues, its three-dimensional presentation, and individual features of the integrin-binding pockets. This dependence is most well characterized in ␣5␤1 integrin binding to Fn, where RGD in concert with a second recognition sequence (PHSRN), the "synergy" site, in the adjacent ninth type III repeat is believed to promote the specific interaction of ␣5␤1 integrin binding to Fn through interactions with the ␣5 subunit (13-15). The synergy site is located ϳ32 Å from the RGD loop on the tenth type III repeat. The type III repeats show great elasticity in the loops between their F-and G-␤ strands, known as the FG loop, which allows the ninth and tenth type III repeats to present multiple conformations. Under small applied forces (on the order of 10 pN), the Fn tenth type III repeat is susceptible to partial unfolding. Computational models of force application to the Fn tenth type III repeat suggest that the RGD loop within the tenth type III repeat translocates away from the ninth type III repeat, resulting in an increase in the distance between the RGD and synergy sites from ϳ32 to ϳ55 Å (16). This capacity to present multiple spatial orientations of the ninth and tenth type III repeats has implications on cell binding, because the relative positioning of these two domains has been shown to influence binding of integrins such as ␣IIb␤3 and ␣5␤1 (17)(18)(19)(20)(21). Further evidence suggesting that alternate conformational states induce integrin "switching" comes from studies in which conformational stability was conferred to the ninth or tenth type III repeats, resulting in modulation of integrin affinity. In these studies, the ninth type III repeat was stabilized via a Leu-Pro mutation at amino acid 1408 (22,23), or the tenth type III repeat was stabilized through facilitation of greater hydrogen bonding within the repeat (24). In both cases, stabilization of the relative positions of the two repeats resulted in increased affinity for integrin ␣5␤1 over integrin ␣v␤3. In addition, studies in which the linker region between the ninth and tenth type III repeats was increased in length showed reduction in ␣5␤1 binding (17).
Previous studies in our lab using recombinant Fn fragments displaying the RGD and synergy sites with a stabilizing (L1408P) point mutation (FnIII9Ј10) or RGD alone (FnIII10) suggest that integrin ␣3␤1 binding to Fn may also be promoted by the ninth type III repeat (26). Although the classical ligand for integrin ␣3␤1 is laminin (Ln) (27,28), it has been reported to bind collagen and Fn (9,29,30) and facilitate cell-cell interactions through both homophilic binding and binding to E-cadherin (31)(32)(33). Furthermore, integrin ␣3␤1 binding to Fn has been shown to occur in an RGD-dependent fashion; however, ␣3␤1 binding to Ln and collagen does not occur through RGD sites (29). Integrin ␣3␤1 is highly expressed by many epithelial cells, including alveolar epithelial cells, renal epithelial cells, and keratinocytes, and has been shown to be critical for regulation of epithelial phenotype (34,35). Integrin ␣3␤1 plays an important role in maintaining epithelial integrity and facilitating wound repair responses, and if not bound to its ECM ligands, it can contribute to pathologies through induction of epithelial to mesenchymal transitions (34,36). During wound healing responses, the ECM rapidly undergoes changes in composition from mostly Ln and elastin to higher concentrations of provisional matrix components, such as Fn. Understanding the ability of integrin ␣3␤1 to bind to both its classical ligand Ln as well as Fn is important to understand cell interactions with these dynamic matrices observed in wound healing and may elucidate mechanisms involved in normal versus pathological wound healing responses.
To explore the hypothesis that ␣3␤1 binding to Fn is enhanced by the presence of the ninth type III repeat and its spacing relative to the tenth type III repeat, we created (i) a dominant negative Arg-Ala mutation in the synergy site (FnIII9 R-A 10), (ii) a two-Gly insertion in the linker region (FnIII9 2G 10), or (iii) a four-Gly insertion in the linker region (FnIII9 4G 10). Surface plasmon resonance (SPR) was then utilized to determine the binding kinetics of integrin ␣3␤1 to the Fn fragments, and cell attachment and spreading assays were then utilized to validate the binding data.

Experimental Procedures
Construction of Mutant pGEX4T1-FnIII9Ј10 Clones-Cloning of the Leu 1408 to Pro mutation (FnIII9Ј10) was made using a parent pGEX4T-1-FN III9 -10 encoding plasmid as previously described (22,23,26). FnIII9Ј10 variants displaying (i) a dominant negative Arg-Ala mutation in the synergy site (FnIII9 R-A 10), (ii) a 2ϫ Gly insertion in the linker region (FnIII9 2G 10), or (iii) a 4ϫ Gly insertion in the linker region (FnIII9 4G 10) were then created using the QuikChange II-E site-directed mutagenesis kit (Stratagene, La Jolla, CA). Sequences are presented in Table 1. All plasmids were introduced into and maintained in the electrocompetent XL-1 Blue Escherichia coli strain and cultured on agarose with LB and ampicillin (0.1 mg/ml) at 37°C. Plasmids were extracted from cultures using the QIAquick spin miniprep kit (Qiagen) and verified via sequencing (Johns Hopkins Synthesis & Sequencing Facility, Baltimore, MD).
Expression and Purification of Recombinant FnIII9Ј10 Proteins-Recombinant Fn fragments were produced as previously described (23,26). Briefly, the expression vectors described above were transformed into BL21 E. coli, and cells were grown to the exponential growth phase and treated with isopropyl ␤-D-thiogalactopyranoside for 3 h. Cells were then lysed by the addition of 10 mg/ml lysozyme and sonication, followed by incubation with 1% Triton X-100 and 10 units/ml of DNase I. Fn fragments were purified by GST affinity chromatography (AKTA Purifier; GE Healthcare). GST tags were removed using bovine thrombin (Sigma-Aldrich). A second round of purification was performed using GST and serine pro- tease affinity chromatography to remove cleaved GST tags and thrombin. Proteins were verified as Ͼ98% pure by SDS-PAGE (see Fig. 1).
Circular Dichroism-CD spectra were collected on a JASCO-J810 spectropolarimeter. Proteins were desalted and concentrated in DI water using PALL centrifugation tubes (3-kDa molecular mass cutoff). Aliquots (0.25 mg/ml) were placed in a quartz cuvette (1-mm path length) and scanned at 50 nm/min. The temperature was maintained at 20°C.
Surface Plasmon Resonance Studies-A Biacore 2000 (Biacore Lifesciences, GE Healthcare) was used to investigate kinetic binding constants (k a and k d ) of Fn fragments variants for integrin ␣3␤1 and control integrins ␣5␤1 and ␣v␤3. Briefly, integrins (R & D Systems, Minneapolis, MN) were covalently immobilized to gold-coated SPR sensor chips via self-assembled monolayer surface chemistry to generate a nonfouling surface with a controlled density of reactive carboxylic acid groups. Mixed self-assembled monolayers were generated on goldcoated chips as previously described (37,38) by incubating with a 10:1 mixture of 1 mM of tri(ethylene glycol)-terminated alkanethiols (HS-(CH2)11-(OCH2CH2)3-OH (ProChimia, Gdansk, Poland) and carboxylic acid-terminated alkanethiols (HS-(CH2)11-(OCH2CH2)6-OCH2COOH) overnight. The senor chip was then loaded into the Biacore 2000, and the carboxylic acid-terminated alkanethiol surface was activated by flowing 200 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (Sigma-Aldrich) and 50 mM N-hydroxysuccinimide (Sigma-Aldrich; 5 l/min for 10 min). Immediately after activation, integrins (100 g/ml) were flowed through the device and allowed to react with the functionalized surfaces at a flow rate of 5 l/min for 10 min to achieve ϳ1500 resonance units (1 resonance unit, ϳ1 pg/mm 2 ). An additional channel was immobilized with BSA to serve as a reference channel; BSA (100 g/ml) was likewise immobilized to achieve ϳ1500 resonance units. Immobilization of integrin and BSA were performed in 0.1 M sodium acetate, pH 4.5. Unreacted N-hydroxysuccinimide groups were quenched in all flow cells with 1 M ethanolamine, pH 8.5 (10 l/min for 10 min). Upon stabilization of the baseline signal, kinetic binding experiments were run with Fn fragments, full-length Fn, or Ln as the flow analytes. Various concentrations for each Fn fragment (10 M to 1 nM) were flowed at 30 l/min for 5 min immediately followed by a 10-min dissociation phase. Between each injection, the surface was regenerated with two 30-s pulses (10 l/min) of 20 mM EDTA and 1 M NaCl (pH 6.0). ␣3␤1 binding experiments were performed in 10 mM HEPES, 150 mM NaCl, 0.0001% Triton X-100, and 2 mM each MgCl 2 and MnCl 2, pH 7.4; ␣5␤1 and ␣v␤3 experiments were performed in this same buffer without Triton X-100. ␣3␤1 integrin interactions were also analyzed in the presence of free 100 g/ml RGD peptides. To determine the role of metal ions in Fn fragment-␣3␤1integrin interactions, experiments were performed in 10 mM HEPES, 150 mM NaCl, 0.0001% Triton X-100, and 2 mM MgCl 2 /2 mM CaCl 2 . The sensor chips in different experiments were freshly immobilized with integrin (␣3␤1, ␣5␤1, or ␣v␤3). A single experiment consisted of flowing various concentrations of FnIII9Ј10 variants, Fn, or Ln over the immobilized integrin and control channel. Between each injection, the surface was regenerated. At least three inde-pendent experiments were performed per integrin. For replicate experiments, FnIII9Ј10 variant, Fn, and Ln injection order was varied to rule out binding trends associated with injection sequence, such as potential decreased activity of the immobilized ligand following regeneration. To characterize binding to integrin ␣3␤1 in the presence of CaCl 2 and RGD, independent experiments were performed with freshly immobilized integrin.
SPR Analysis and Evaluation-SPR sensorgrams were analyzed with Scrubber 2 and ClampXP software (Center for Biomolecular Interactions Analysis, University of Utah) (39 -41). Reference cell responses were subtracted from corresponding active response curves. The resulting curves were then analyzed and fit to the kinetic models. Kinetic modeling and simulations were performed with ClampXP software with the heterogeneous surface model; globally fitted parameters were determined for each kinetic data set per Fn fragment. Equilibrium binding constants (K d1 and K d2 ) were calculated from fit kinetic constants by dividing k d /k a for each experiment; mean K d1/2 values are presented in data tables. Goodness of fit for each model was evaluated by analyzing the residual plots and residual sum of squares.
Cell Attachment Assays-Wells of a 96-well plate were coated with FnIII9Ј10 To determine 100% attachment, RLE-6TN cells were plated on poly-L-lysine-coated tissue culture plastic wells. Cells with or without the addition of antibodies were then plated on each of the Fn fragment-, Fn-, or Ln-coated wells for 30 min, and then all control and treated wells were washed and fixed with 5% gluteraldehyde and stained with 0.1% crystal violet stain. The dye was solubilized in 10% acetic acid, and absorbance was measured on a Biotek Synergy H4 multi-mode plate reader. The results are pooled from three independent triplicate experiments and presented as a percentage of attachment with respect to the poly-L-lysine signal.
Epithelial To determine the effect of various integrins on cell spreading, cells were incubated with ␣3, ␣5, or ␣v blocking antibodies at 37°C for 30 min prior to plating. Following a 3-h incubation period, the cells were washed with PBS, fixed with 4% formaldehyde, permeabilized with 0.2% Triton X-100, and then blocked with 10% goat serum. Polymerized actin was stained with Alexa Flour 546 phalloidin (Invitrogen), and nuclei were stained with Hoescht stain (Invitrogen). Coverslips were mounted, and images were examined using confocal microscopy (40ϫ oil immersion objective; Zeiss 700 -405). Repre-sentative images are presented. The area and perimeter of individual cells were determined for each condition using ImageJ (NIH Freeware) image processing software, and circularity was determined using the equation circularity ϭ 4(area/perimeter 2 ). At least 50 cells were analyzed per condition. The cumulative frequency distribution of area and circularity of cells was determined for each data set. Statistical Analysis-All statistical analyses were performed with Prism software program (GraphPad, San Diego CA). The data were analyzed using a one-way analysis of variance using the Tukey test at a 95% confidence interval.

Results
Analysis of FnIII9Ј10 Variants-FnIII9Ј10 variants were analyzed by SDS-PAGE and CD (Fig. 1). Variants were verified as Ͼ98% pure by SDS-PAGE. CD spectra of all FnIII9Ј10 variants are in good agreement with that of full-length fibronectin (42) (i.e. minimum at 215 nm and maxima at 226 and 202 nm), except for fragment FNIII9 4G 10, which has a minimum of ϳ216 nm and a maximum of ϳ194 nm. These bands are suggestive of ␤-sheets. Similar changes in CD spectra for FNIII9 -10 have been observed upon partial unfolding (43,44). The additional glycine residues within the loop region may have resulted in a more open conformation of the FNIII9 4G 10 fragment, analogous to partial unfolding of the domain. However, the secondary structures of the fragments clearly remain intact, because spectra for unfolded, random coils are not present.
Fitted ␣3␤1 Binding Affinity Parameters: Heterogeneous Surface Model-SPR with soluble recombinant integrin immobilized to the surface of the sensor chip was utilized to determine the role of FnIII9Ј10 in integrin ␣3␤1 interactions with Fn ( Fig.  1). A representative sensorgram obtained from integrin ␣3␤1 immobilization is shown in Fig. 1C. The large spike in RU seen following injection of ethanolamine is due to differences in refractive indices between the immobilization buffer and ethanolamine. Experimental data were collected and then fit to a two-site heterogeneous surface model to account for variable orientation and conformational state of immobilized integrin. Integrins inherently can present both high and low affinity states, which can result in a heterogeneous surface displaying complex binding kinetics. Furthermore, the amine coupling procedure utilized for immobilization of integrin onto the surface can result in a random receptor orientation (Fig. 1D) and a heterogeneous surface, also presenting both high and low affinity states of receptor. To account for the possibility of both high and low affinity states of the integrin, a heterogeneous surface model was utilized to fit the response curves. The fitted parameters k a1 , k d1 , k a2 , k d2 , calculated K d1 and K d2 , and residual sum of squares of the fit of each FnIII9Ј10 variant, Fn, and Ln using a heterogeneous surface model are displayed in Table 2. Experimental response and fit simulation curves are presented for FnIII9Ј10 variants binding to integrin ␣3␤1 in Fig. 2A. Residual plots for FnIII9Ј10 variants fit to the heterogeneous surface model are shown in Fig. 2B.
As expected, examination of the fitted parameters obtained from a heterogeneous surface model demonstrates a high affinity binding event and a secondary low affinity binding event. High affinity states of integrin result in high affinity binding events, described collectively by k a1 , k d1 , and the resulting calculated K d1 . Likewise, lower affinity states of the integrin result in a series of lower affinity binding events, described collectively by k a2 , k d2 , and the resulting calculated K d2 .
It was found that FnIII9Ј10 had the lowest k d1 of all FnIII9Ј10 variants analyzed, with a value of 136 ϫ 10 Ϫ4 s Ϫ1 . Increasing the linker region between the ninth and tenth type III repeats with either two Gly residues or four Gly residues, resulted in an increase in the k d to 492 ϫ 10 Ϫ4 and 247 ϫ 10 Ϫ4 s Ϫ1 , respectively, suggesting that increasing the space between the PHSRN and RGD sites disrupts the interactions between FnIII9 and FnIII10 domains and integrin ␣3␤1. Furthermore, the disrupting Arg to Ala point mutation within the PHSRN site resulted in an increased k d of 471 ϫ 10 Ϫ4 s Ϫ1 . Equilibrium dissociation constants, which takes k d1 and k a1 values into account, were calculated for each analyte for the high affinity binding state (K d1 ). FnIII9Ј10 and FnIII9 2G 10 were found to have similar K d1 values: 4.1 and 3.2 nM, respectively. Direct disruption in the PHSRN site resulted in a 17-fold increase in K d to 67.2 nM (compared with FnIII9Ј10), whereas increasing the spacing between the ninth and tenth type III repeats by four Gly residues resulted in a 39-fold increase to 156 nM. Ln and Fn analytes were found to have a high affinity for the integrin as well, with a K d of 9.4 and 1.6 nM, respectively. Experimental response and fit simulation curves for Fn and Ln binding to integrin ␣3␤1 are shown in Figs. 3 and 4, respectively. Focusing on the high affinity binding parameters obtained from the heterogeneous surface model demonstrates that stabilization of the ninth and tenth type III domains of Fn results in affinities for integrin ␣3␤1 similar to those observed for the classical ␣3␤1 ligand Ln.
These results indicate that Fn does bind integrin ␣3␤1 with a high affinity, and this binding is dependent on the spacing of the PHSRN and RGD sites. Integrin binding is typically dependent on the presence of divalent cations (29,45), and so all SPR experimental samples were performed in the presence of 2 mM of MgCl 2 and MnCl 2 . To confirm that integrin binding to FnIII9Ј10 variants is dependent on the presence of divalent cations, binding events were characterized in the presence of 2 mM MgCl 2 and CaCl 2 , because manganese has been known to acti- Kinetic association (k a ) and dissociation (k d ) rates obtained from the fits are presented. k a1 and k d1 are associated with the high affinity binding state, and k a2 and k d2 are associated with the low affinity binding state. Equilibrium dissociation constants, which take k d and k a values into account, were calculated for each analyte for the high (K d1 ) and low (K d2 ) affinity binding state. RSS are also presented. vate integrin by occupying the metal ion binding sites of integrin. In the presence of CaCl 2 instead of MnCl 2 , minimal binding to integrin ␣3␤1 was observed (Fig. 5A), indicating as expected, that ␣3␤1 binding is dependent on the presence of divalent cations. To determine whether the RGD site of FN is critical for the observed integrin interactions, binding of FnIII9Ј10 variants to integrin ␣3␤1 was analyzed in standard running buffer (2 mM of MgCl 2 and MnCl 2 ) in the presence of 100 g/ml of RGD peptide (Fig. 5B). Binding was inhibited in the presence of RGD, indicating that RGD is critical for integrin ␣3␤1 binding to FnIII9Ј10 variants. Control Integrins: Fitted ␣5␤1 and ␣v␤3 Binding Affinity Parameters-The role of the PHSRN site, as well as the relative spacing with respect to the RGD site, is critical for ␣5␤1 binding to Fn, whereas binding to ␣v␤3 only requires the RGD site (46). To confirm these previous reports and provide validation of our methods, binding of FnIII9Ј10, FnIII9 R-A 10, FnIII9 2G 10, FnIII9 4G 10, Fn, and Ln to immobilized integrin ␣5␤1 and ␣v␤3 was investigated. As with ␣3␤1, curves were fit to a heterogeneous surface model. Representative response curves and heterogeneous surface model fits for binding of one concentration (2 M) of FnIII9Ј10 variants to integrin ␣5␤1 and ␣v␤3 are shown in Fig. 6A, and residual plots are shown in Fig. 6B. Calculated K d values for both fits are displayed in Tables 3 and 4. As expected, ␣5␤1 binding was found to depend on the presence of the PHSRN site. Analysis of binding parameters using the het-erogeneous surface model demonstrates that the calculated K d1 for the FnIII9Ј10 variants follow similar trends as those observed for integrin ␣3␤1; FnIII9Ј10 had the lowest K d1 of all the variants, with a value of 41 nM. FnIII9 R-A 10, FnIII9 2G 10, and FnIII9 4G 10 were found to have K d1 values of 735, 523, and 830   nM, respectively. Fn and Ln were found to have K d1 values of 86 and 3509 nM, respectively. Although Ln is not typically regarded as a ligand for ␣5␤1, there have been previous reports demonstrating interactions between these molecules (47). K d2 values were higher than the calculated K d1 values for all analytes, ranging from 531 to 162,333 nM.
FnIII9Ј10 binding to integrin ␣v␤3 was not dependent on the presence of the PHSRN site, which was expected because ␣v␤3-Fn interactions are dominated by the RGD loop. Using the heterogeneous surface model, K d1 values for FnIII9Ј10, FnIII9 R-A 10, FnIII9 2G 10, FnIII9 4G 10, and Fn were 65, 57, 47, 30, and 82 nM, respectively, whereas minimal binding was observed to Ln. K d2 values were considerably higher than the calculated K d1 values for all analytes, ranging from 219 to 41,269 nM.
Cell Attachment Assays-To characterize how epithelial cells attach to different FnIII9Ј10 variants, RLE-6TN alveolar epithelial cells were incubated with ␣3, ␣5, or ␣v antibodies or combinations of the three for 30 min, then plated on each of the FnIII9Ј10 variants, Fn, or Ln and allowed to attach for 30 min.
This cell type was chosen because it is known to express ␣3, ␣5, and ␣v integrin subunits (26,49); expression of ␣3 is typically approximately 10 times higher than ␣5 or ␣v expression. We have confirmed this integrin expression pattern by flow cytometry analysis (data not shown). We have previously demonstrated that attachment of RLE-6TN cells to FnIII9Ј10 is dependent on both the RGD and PHSRN sites of FN. 1 g/ml of RGD peptide inhibits RLE-6TN attachment to FnIII9Ј10 by 70%, and in the presence of both 1 g/ml of RGD and 1 g/ml PHSRN peptides, RLE-6TN attachment to FnIII9Ј10 is completely abrogated (26). Following the 30-min incubation period, attached cells were fixed to the surface and stained, and absorbance was measured. The results shown are presented as a percentage of attachment compared with attachment on poly-Llysine. These data coincide closely with the SPR results, indicating that cell attachment to Ln and FnIII9Ј10 surfaces is highly dependent on ␣3 integrin, whereas cell attachment to FnIII9 2G 10, FnIII9 4G 10, and Fn was predominantly mediated by ␣v integrin (Fig. 7). Interestingly, cells cultured on the FIGURE 6. Attachment to control integrins. Response curves obtained from FnIII9Ј10 variants binding to immobilized integrin ␣5␤1 or ␣v␤3 were fit to a heterogeneous surface model. A range of concentrations was tested for each variant. A, representative experimental response curves and resulting fits are shown for 2 M FnIII9Ј10, FnIII9 R-A 10, FnIII9 2G 10, and FnIII9 4G 10. Solid lines indicate experimental SPR response curves, and dashed curves indicate fitted models. B, representative residuals are shown for FnIII9Ј10, FnIII9 R-A 10, FnIII9 2G 10, and FnIII9 4G 10.

TABLE 3 K d values: ␣5␤1-heterogeneous surface model
Kinetic association (k a ) and dissociation (k d ) rates obtained from the fits are presented. k a1 and k d1 are associated with the high affinity binding state, and k a2 and k d2 are associated with the low affinity binding state. Equilibrium dissociation constants, which take k d and k a values into account, were calculated for each analyte for the high (K d1 ) and low (K d2 ) affinity binding state. RSS are also presented. FnIII9 R-A 10 surface appear to mediate attachment through both ␣3 and ␣5 integrins. Previous studies have suggested that the RGD and PHSRN sites could serve as an on/off switch for ␣5␤1 binding to Fn such that decoupling of the PHSRN and RGD sites may turn the switch "off" for ␣5␤1 binding and "on" for ␣v integrin binding. Although this is supported by our results with the two-Gly and four-Gly insertion variants, it appears that a mutation in the synergy site is not sufficient to drive epithe-  Kinetic association (k a ) and dissociation (k d ) rates obtained from the fits are presented. k a1 and k d1 are associated with the high affinity binding state, and k a2 and k d2 are associated with the low affinity binding state. Equilibrium dissociation constants, which take k d and k a values into account, were calculated for each analyte for the high (K d1 ) and low (K d2 ) affinity binding state. RSS are also presented.  (Fig. 8). Cell spread area was analyzed on the various substrates. Spreading was found to be greatest on Ln and Fn. On FnIII9Ј10 variants, cell spread area was greatest on FnIII9Ј10 and least spread on FnIII9 2G 10; however, these differences in spread area on the variants were not significantly different. Furthermore, inhibition of integrin with antibodies did not significantly change cell spreading on FnIII9Ј10 variants (Fig. 9). Differences in cell shape were additionally analyzed by calculating cell circularity, with values closer to 1 indicating a more rounded cell. Cell circularity has previously been utilized to characterize epithelial to mesenchymal responses of alveolar epithelial type II cells (2,3,26). In these previous studies, cells that engage ␣3 integrins have been shown to display a rounder, more cuboidal morphology, indicative of a more epithelial phenotype. Conversely, cells that engage ␣v integrins typically display a more elongated morphology, accompanied by aligned actin filaments, which is indicative of a more mesenchymal phenotype. Here we aimed to determine whether differential integrin engagement by RLE-6TN cells on the FNIII9Ј10 variants would elicit similar responses. Based on SPR and cell attachment results, cells should strongly engage ␣3 integrins on Ln and FNIII9Ј10 substrates, and therefore, these substrates would be expected to support a predominantly round, epithelial morphology. Indeed, cell circularity was highest on these substrates compared with all other conditions. Furthermore, inhibition of ␣3 integrin with antibodies resulted in a decrease in cell circularity on both Ln and FNIII9Ј10 (Fig. 10). Based on SPR and cell attachment results, cells should strongly engage ␣v integrin on FNIII9 R-A 10 and FnIII9 4G 10, and therefore, these substrates would be expected to support a more elongated, mesenchymal phenotype. As expected, cell circularity was lowest on these substrates. Inhibition of ␣v integrin with antibodies resulted in an increase in cell circularity on both FNIII9 R-A 10 and FnIII9 4G 10, whereas inhibition of ␣3 and ␣5 integrins had a minimal effect on cell circularity (Fig. 10).

Discussion
These studies demonstrate that Fn binds integrin ␣3␤1 in a cell-free system, and this binding is dependent on the presence of the ninth type III repeat and its spacing relative to the tenth type III repeat, presumably mediated through the RGD and synergy sites. These results provide significant insight into the role of the central cell binding domain of Fn, i.e. the ninth and tenth type III repeats, and what is classically considered a Ln receptor, integrin ␣3␤1.
The role of the PHSRN site in ␣5␤1 integrin binding to Fn is well characterized, and in addition, this site has been shown to enhance binding of additional integrins, as illustrated by its role in facilitating strong interactions between Fn and the platelet integrin ␣IIb␤3 (50). Studies suggest that the PHSRN site acts to stabilize the high affinity conformation of the RGD site required for ␣5␤1 integrin binding. Here we demonstrate that ␣3␤1 integrin binding to Fn is also dependent on the relative spacing of the RGD and PHSRN sites. Previous investigations to elucidate potential ligands for integrin ␣3␤1 have been conflicting regarding interactions with Fn. Many of these studies involve cell-based attachment assays, which are at times difficult to interpret, especially in the case of integrin ␣3␤1, in which binding affinity to ECM ligands is influenced by the presence of additional integrins (51). Studies utilizing soluble integrins in cell free systems are also conflicting. For example, studies by Eble et al. (28) reported minimal ␣3␤1 binding to Fn. Interestingly, these studies showing minimal binding were performed on tissue culture plastic, whereas several studies confirming Fn/␣3␤1 interactions utilized ECM affinity columns (29,52). Our SPR studies possibly bridge these discrepancies. Fn has been shown to undergo unfolding upon adsorption as a consequence of the hydrophobic effect and the protein adopting its lowest energy state (53)(54)(55)(56). Such unfolding events likely disrupt the relative positioning of the RGD and PHSRN sites and, according to the SPR studies presented here, would disrupt ␣3␤1 binding. This conclusion is further supported by data demonstrating that fulllength Fn in solution binds ␣3␤1 exceptionally well, as indicated by the presented SPR results.
Because ␣3␤1 and ␣5␤1 integrin are involved in epithelial wound repair, these findings that ␣3␤1 integrin, like ␣5␤1 integrin, binds to Fn in a synergy-dependent manner have implications in cell interactions during epithelial wound heal- ing. Alveolar epithelial type II cells increase expression of ␣5␤1 integrin in response to injury (22); therefore, it is likely that integrin ␣3␤1 binds Fn with great affinity in this cell type in early wound repair before ␣5␤1 integrin is highly expressed. Although ␣5␤1 appears to be the main Fn receptor during alveolar wound repair, integrin blocking experiments with airway epithelial cells showed that in addition to ␣5, ␤1, and Fn blocking antibodies, ␣3 blocking antibodies also resulted in a significant decrease in wound repair (22). Furthermore, it has been shown that integrin ␣3␤1 binding to Fn can be affected by the presence of ␣5␤1 integrin such that ␣3␤1 binding is low in cells expressing ␣3␤1 and ␣5␤1 at comparable levels, but ␣3␤1 binding to Fn is greatly enhanced in H69 cells that highly express ␣3␤1 compared with other integrins (29). These reports, taken along with our findings here, suggest that ␣3␤1 binding to Fn could potentially modulate wound repair by facilitating binding to ECM in a wound healing environment as the ECM composition transitions from predominantly Ln to high levels of Fn.
Control SPR experiments demonstrate similar affinities of all FnIII9Ј10 variants for integrin ␣v␤3, with all FnIII9Ј10 variants showing affinities in the range of 30 -65 nM. The RGD and PHSRN sites have previously been described as a possible on/off switch for ␣5␤1 binding to Fn (25). Our findings show that in the context of epithelial cells, destabilizing the PHSRN and RGD sites may additionally serve as an off switch for ␣3␤1/ ␣5␤1 binding and on for ␣v integrin binding. Highlighting the significance of such a possible integrin switch, binding of ␣3␤1/ ␣5␤1 integrins have markedly different effects on epithelial phenotype compared with ␣v integrin binding. Integrin ␣3␤1 epithelial cell interactions are associated with maintenance of epithelial phenotype and wound healing, whereas interactions with ␣v integrins have been associated with aberrant wound healing, scar tissue formation, epithelial to mesenchymal transitions, and tumor metastasis (22,27,34,36). The extracellular microenvironment of disease states associated with increased ␣v integrin binding, scar tissue in particular, is associated with contractile fibroblasts (3). Evidence suggests that Fn is capable of undergoing force-mediated unfolding in response to cellular forces (48,50), which has been hypothesized to lead to an increase in spacing between the RGD and PHSRN sites within the integrin binding domain. Our data suggest that if such a speculative force-activated conformational change does occur, this would lead to diminished integrin binding to Fn while not affecting ␣v binding, which could further contribute to the nonhomeostatic condition. These studies provide insights into how Fn interactions could potentially contribute to normal versus pathological wound healing.