αVβ6 Is a Novel Receptor for Human Fibrillin-1

Human fibrillin-1, the major structural protein of connective tissue 10-12 nm microfibrils, contains multiple calcium binding epidermal growth factor-like domains interspersed with transforming growth factor β-binding protein-like (TB) domains. TB4 contains a flexible RGD loop that mediates cell adhesion via αVβ3 and α5β1 integrins. This study identifies integrin αVβ6 as a novel cellular receptor for fibrillin-1 with a Kd of ∼0.45 μm. Analyses of this interaction by surface plasmon resonance and immunocytochemistry reveal different module requirements for αVβ6 activation compared with those of αVβ3, suggesting that a covalent linkage of an N-terminal calcium binding epidermal growth factor-like domain to TB4 can modulate αV integrin binding specificity. Furthermore, our data suggest α5β1 is a low affinity fibrillin-1 receptor (Kd > 1 μm), thus providing a molecular explanation for the different α5β1 distribution patterns seen when human keratinocytes and fibroblasts are plated on recombinant fibrillin fragments versus those derived from the physiological ligand fibronectin. Non-focal contact distribution of α5β1 suggests that its engagement by fibrillin-1 may elicit a lesser degree and/or different type of intracellular signaling compared with that seen with a high affinity ligand.

Human fibrillin-1, the major structural protein of connective tissue 10 -12 nm microfibrils, contains multiple calcium binding epidermal growth factor-like domains interspersed with transforming growth factor ␤-binding protein-like (TB) domains. TB4 contains a flexible RGD loop that mediates cell adhesion via ␣V␤3 and ␣5␤1 integrins. This study identifies integrin ␣V␤6 as a novel cellular receptor for fibrillin-1 with a K d of ϳ0.45 M. Analyses of this interaction by surface plasmon resonance and immunocytochemistry reveal different module requirements for ␣V␤6 activation compared with those of ␣V␤3, suggesting that a covalent linkage of an N-terminal calcium binding epidermal growth factor-like domain to TB4 can modulate ␣V integrin binding specificity. Furthermore, our data suggest ␣5␤1 is a low affinity fibrillin-1 receptor (K d > 1 M), thus providing a molecular explanation for the different ␣5␤1 distribution patterns seen when human keratinocytes and fibroblasts are plated on recombinant fibrillin fragments versus those derived from the physiological ligand fibronectin. Nonfocal contact distribution of ␣5␤1 suggests that its engagement by fibrillin-1 may elicit a lesser degree and/or different type of intracellular signaling compared with that seen with a high affinity ligand.
Human fibrillin-1, a 350-kDa extracellular matrix glycoprotein, is the major structural component of the 10 -12-nm connective tissue microfibrils (1). It has a modular structure dominated by 43 calcium binding epidermal growth factor-like (cbEGF) 3 domains, tandem repeats of which are separated by transforming growth factor ␤-binding protein-like (TB) domains. Mutations in the fibrillin-1 gene (FBN1) give rise to the connective tissue disease Marfan syndrome and related disorders.
In elastic tissues such as arteries, lung, and elastic ligaments, fibrillin microfibrils appear attached to cell membranes. These areas of attachment resemble focal contacts with an abundance of actin microfilaments on the cytoplasmic side of the membrane. Clustering of microfilaments at the cell surface at points of contact with the microfibrils suggest that microfibrillar components interact with cell-surface receptors which in turn serve as a dynamic link between cells and their microenvironment (2,3).
Fibrillin-1 is one of the microfibrillar proteins shown to mediate cell adhesion through binding to heterodimeric cell surface receptors of the integrin family (4,5). This binding is at least in part mediated by the TB4 domain that contains an RGD (Arg-Gly-Asp) motif, a minimal integrin binding motif found in a number of extracellular matrix proteins. Fibrillin-integrin interactions are likely to be particularly important in tissues where microfibrils are in close proximity to cells, such as in the elastic lamellae, and may play a role in the assembly of the microfibril network, as has been shown in the case of fibronectin fibrillogenesis (6). Furthermore, the loss of cell-matrix interactions is likely to underlie some of the pleiotropic manifestations of Marfan syndrome.
Until recently integrin ␣V␤3 was thought to be the major fibrillin-1 receptor in several cell lines, and this interaction has been shown to influence cell spreading, focal contact assembly, cytoskeletal rearrangements, and extracellular matrix deposition (7,8). Our quantitative studies on ␣V␤3-mediated interactions with fibrillin-1 identified a high affinity interaction (K d ϳ 0.04 M) and suggested a previously unrecognized requirement for N-terminal linkage of cbEGF22 to RGD-containing TB4 for activation of ␣V␤3 (8). Integrin ␣5␤1 has also been shown to bind fibrillin-1 RGD based on inhibition of fibrillin-mediated cell adhesion by ␣5␤1-specific function-blocking antibodies (7). However, the biophysical properties and cellular consequences of this interaction have not been investigated.
In this study we sought to further define the integrin binding selectivity of fibrillin-1 using recombinant TB4-containing fragments. Our cellular and biophysical data identify ␣V␤6 as a novel moderate affinity receptor for fibrillin-1 RGD-containing fragments (K d ϳ 0.45 M) and demonstrate that molecular determinants of this interaction are contained within TB4. Thus ␣V␤3 and ␣V␤6 each have unique structural requirements for binding to fibrillin-1. In addition, we demonstrate that ␣5␤1 binding to fibrillin-1 gives rise to a low affinity interaction, which contrasts sharply with the high affinity observed for the physiological ligand fibronectin. This may explain the different cellular responses observed when ␣5␤1-expressing cells are plated onto these two substrates. Our data, therefore, provide new insights into the affinity and specificity of fibrillinintegrin interactions and suggest a broader and more diverse involvement of the integrin superfamily in fibrillin-1 function.
Expression and Purification of Recombinant Fibrillin-1 Constructs-DNA fragments encoding the wild type sequences of the domain pairs cbEGF22-TB4 (nucleotides 4589 -4948, numbering according to Pereira et al. (11)), TB4-cbEGF23 (nucleotides 4712-5074), and the triple construct cbEGF22-TB4-cbEGF23 (nucleotides 4589 -5074) were amplified from fibrillin-1 cDNA. Amplified DNA was inserted into pQE30 expression vector (Qiagen) and transformed into Escherichia coli NM554 [pREP4]. After Ni 2ϩ affinity chromatography, the recombinant His 6 fusion proteins were purified and refolded using a previously described redox-shuffling system (12). The D1543A substitution in cbEGF22-TB4-cbEGF23 was introduced by PCR mutagenesis according to standard procedures. For BIAcore experiments, wild type domain pairs and D1543A triple construct were modified to contain a C-terminal BirA sequence, which was used for biotinylation and immobilization onto streptavidin sensor chips. The identity of purified products was confirmed by mass spectrometry. All proteins were estimated to be Ͼ90% pure, as judged by SDS-PAGE. Onedimensional and two-dimensional NMR analyses were performed to confirm correct folding of wild type fragments. Ca 2ϩ binding to the fragments, an important criterion for proper folding, was measured by fluorometric titration (13) or titration with chromophoric chelator (14).
Cell Adhesion Assays-The cell attachment assay was conducted essentially as published previously (15). Briefly, 96-well plates (MaxiSorp c ) were coated with doubling dilutions of recombinant fibrillin-1 fusion proteins and fibronectin controls in Tris-buffered saline containing 2 mM Ca 2ϩ and incubated overnight at 4°C. VB6 cells were harvested with 0.5% trypsin-EDTA in phosphate-buffered saline (Sigma), neutralized with 1 mg/ml trypsin-inhibitor (Sigma), washed, and resuspended in serum-free Dulbecco's minimum essential medium. 50 l of 2 ϫ 10 5 cell/ml suspension was added to each well and incubated for 30 min at 37°C, 5% CO 2 . Adherent cells were washed once with phosphate-buffered saline (PBS) and fixed in 4% (v/v) glutaraldehyde, 4% (v/v) formaldehyde in PBS. Adherent cells were scored for spread morphology using a phase contrast microscope according to the criteria of Mardon and Grant (15). Results were confirmed by measuring the average surface area of 100 cells per well using a Leica (Milton Keynes, UK) DM IRB inverted microscope equipped with the Openlab imaging software (Improvision, Coventry, UK). Cell attachment was quantified by staining the nuclei with 0.1% crystal violet in 10% ethanol. The dye was solubilized with methanol, and optical density read at 595 nm. For antibody inhibition assay, cells were added to the wells together with serial dilutions of antibodies in 25 l of phosphate-buffered saline.
Immunocytochemistry-Glass coverslips were coated with 100 g/ml fibrillin-1 or fibronectin fragments overnight at 4°C. VB6 cells were prepared in the same way as for cell adhesion assays and replated on fibrillin-1 fragment-coated coverslips for 30 min. After fixation with 3% (w/v) paraformaldehyde, actin was visualized using fluorescein-conjugated phalloidin. Human integrins ␣V␤6 and ␣5␤1 were visualized using mAbs CS␤6 and JBS5, respectively, except for the double staining of the two integrins when ␣5 was labeled with the polyclonal rabbit anti-␣5. Assembly of focal adhesions was demonstrated by staining with an antibody to vinculin. Mouse and/or rabbit IgG 1 substituted for the primary antibody at 10 g/ml was used as a negative control. Images were captured using the objective with PL FLUOTAR 100ϫ oil immersion lens with fixed numerical aperture of 1.3 on a Leica DMRBE microscope (Leica) equipped with a Hamamatsu Orca C4742-95 digital camera and analyzed with the OpenLab software (both Improvision).
Flow Cytometry-Cells were prepared as for cell adhesion assays, except that accutase (PAA Laboratories) was used for harvesting cells. Washed and pelleted cells were resuspended in 100 l of primary antibody diluted in Hanks' balanced salt solution, 5% normal human serum and incubated for 1 h at 4°C. The labeling procedure was repeated with secondary antibody (fluorescein isothiocyanate-conjugated goat anti mouse (Fab) 2 , Dako). After washing, cells were analyzed using a FACscan flow cytometer (Beckman Coulter Epics Altra). Mouse IgG 1 was used as a negative control.
Surface Plasmon Resonance (SPR) Studies-Real time biomolecular interaction analysis was performed using a BIACORE 2000 instrument (BIAcore, Uppsala, Sweden). All experiments were performed at 25°C using Tris-buffered saline (25 mM Tris, 150 mM NaCl, pH 7.4) containing CaCl 2 , MgCl 2 , and MnCl 2 at 2 mM each as running buffer to ensure that integrins are in the extended conformation optimal for ligand binding (16). Fibrillin-1 fragments were coupled to the surface of CM5 sensor chips (BIAcore) either directly via amine coupling (between 1000 and 2500 RU) or indirectly via streptavidin using a C-terminal biotin tag (ϳ500 RU) according to the manufacturer's instructions (BIAapplications Handbook, BIAcore). Recombinant ␣V␤6 or ␣5␤1 integrins were used as the analyte. Dilutions of integrins (766-23.9 nM for ␣V␤6; 512-8 nM for ␣5␤1) were injected at a constant flow rate of 10 l/min for 5 min. All measurements were base-line-corrected by subtracting the sensorgram obtained with the control flow cell containing immobilized RGA mutant or a nonspecific fibrillin-1 fragment (TB6-cbEGF23). Data were analyzed using the BIAevaluation 3.0 software package and fitting the data to a simple 1:1 Langmuir binding model. At least three independent experiments were performed for each fibrillin-integrin couple. Mass transport effects were excluded by showing that changing the level of immobilized ligand or analyte flow rate had no effect on the measured rate constants.
Ligand Binding Assay-Binding of soluble fibrillin-1 and fibronectin fragments to full-length recombinant ␣5␤1 integrin was measured using an enzyme-linked immunosorbent assay. Purified integrin (150 ng) was coated onto microtiter wells followed by blocking of nonspecific binding sites with 1% bovine serum albumin. Biotinylated ligands were added to wells in 50 mM Tris, pH 7.4, 150 mM NaCl containing CaCl 2 and MgCl 2 at 1 mM each and incubated for 3 h at room temperature. After washing three times with buffer, the bound ligands were quantitated by peroxidase-conjugated streptavidin. Apparent K d values were calculated from each curve using BIAevaluation 3.0 software steady state fitting mode.

RESULTS
␣V␤6 Integrin Is a Novel Receptor for Fibrillin-1-Epithelial cells have been shown to synthesize and deposit fibrillin-1 into the extracellular matrix of human skin (17). We investigated the possibility that epithelial cell-specific ␣V␤6 integrin may have a role in cell adhesion to fibrillin-1 by testing the capacity of fibrillin-1 fragments to mediate adhesion of the high ␣V␤6-expressing keratinocyte cell line (VB6) (18). Initially the integrin expression profile of the VB6 cells was determined by flow cytometry (Fig. 1A). This analysis confirmed that a large proportion of VB6 cells (Ն80%) express high levels of ␣V␤6 and ␣5␤1 integrins. Approximately 60% of cells were also found to be ␣V␤5-positive, whereas very low levels of ␣V␤3 could be detected on less than 6% of cells.
To determine the contribution of each of these receptors to the overall keratinocyte adhesion to fibrillin-1, VB6 cells were allowed to adhere to plates coated with fibrillin-1 cbEGF22-TB4 domain pair in the presence of function-  blocking antibodies (Fig. 1B). cbEGF22-TB4 was chosen as the substrate since preliminary data showed that it contained the molecular determinants necessary for cell attachment and spreading. The antibody against ␣V␤3 had no detectable effect on VB6 spreading, consistent with its low abundance on these cells. Similarly, the antibody against ␣V␤5 did not cause any spreading inhibition, suggesting that this integrin may not be able to bind fibrillin-1. At concentrations Ն10 g/ml, the antibodies against ␣V␤6 and ␣5␤1 caused ϳ80 and 20% inhibition of the overall VB6 cell spreading, respectively. These data indicate that ␣V␤6 integrin is the major receptor for fibrillin-1 on this keratinocyte cell line. Furthermore, the role of this integrin in fibrillin-1-mediated adhesion extends to other ␣V␤6-expressing cells such as lung-derived H441 adenocarcinoma cells (data not shown).
Module Requirements for Fibrillin-mediated ␣V␤6 Activation-Because ␣V␤6 is the most abundant integrin on VB6 cells and the main mediator of their adhesion to fibrillin-1 fragments, this cell line was deemed suitable for investigating the structural requirements of fibrillin-mediated ␣V␤6 activation. Importantly, given that less than 6% of VB6 cells express low levels of ␣V␤3, the molecular determinants of ␣V␤6 activation could be studied in this cell line without any interference from ␣V␤3 receptor.
To establish whether the ␣V␤6fibrillin interface extends beyond TB4 to involve regions in flanking cbEGFs, three partially overlapping RGD-containing fragments cbEGF22-TB4, TB4-cbEGF23, and cbEGF22-TB4-cbEGF23 (Fig. 2) were tested for their ability to induce VB6 cell attachment, spreading, and focal contact formation. Quantification of VB6 adhesion to protein-coated surfaces (Fig. 3) revealed that structureactivity relationships were somewhat different from those observed previously for ␣V␤3-mediated adhesion of baby hamster kidney and human endometrial stromal fibroblasts (8). TB4-cbEGF23 and cbEGF22-TB4 were found to be equally potent not only in inducing cell attachment but also in triggering cell spreading, with a dose-dependent spreading activity equal to that of cbEGF22-TB4-cbEGF23 (Fig. 3, A and B). A D1543A substitution in the triple domain construct (RGA mutant) abolished the attachment and spreading activity as expected. Phase contrast micrographs in Fig. 3C show the morphology of VB6 keratinocytes on fibrillin-1 and illustrate that there are no significant differences in VB6 attachment or spreading between fibrillin-1 domain pairs. This result suggested that the RGD-containing TB4 domain might be sufficient not only for the initial ␣V␤6-mediated cell attachment to fibrillin-1 but also for the post-attachment intracellular signaling leading to cell spreading.
To gain further insight into domain requirements for the establishment of intracellular architecture characteristic of well spread cells, VB6 cells adhering to cbEGF22-TB4, TB4-cbEGF23, and cbEGF22-TB4-cbEGF23 were immunofluorescently labeled for ␣V␤6 integrin and F-actin (Fig. 4). Coverslips coated with the fibronectin FIII9-10 domain pair, a known ligand for ␣V␤6, were used for comparison (10,19). In VB6 cells cultured on cbEGF22-TB4 and cbEGF22-TB4-cbEGF23, immunofluorescent labeling revealed clusters of ␣V␤6 integrin colocalized with actin fibers within integrin-rich focal contacts (Fig. 4). Furthermore, adhesion to TB4-cbEGF23 also resulted in ␣V␤6 recruitment to focal contacts of approximately the same size and number as on cbEGF22-TB4 (Fig. 4). Because previous experiments with baby hamster kidney and human endometrial stromal fibroblast cells indicated a requirement for cbEGF22 in addition to TB4 for ␣V␤3 recruitment into focal contacts, this result suggests different molecular determinants for ␣V␤6 activation (8).
Kinetic Analyses of ␣V␤6 Binding to Fibrillin-1-The molecular mechanism underlying fibrillin-1 recognition by ␣V␤6 integrin was further investigated by monitoring fibrillinintegrin complex formation using SPR. The RGD-containing fragments cbEGF22-TB4, TB4-cbEGF23, and cbEGF-TB4-cbEGF23 (Fig. 2) were immobilized at equimolar levels on the surface of a CM5 sensor chip via primary amine coupling (Fig.  5). The rates of association (k on ) and dissociation (k off ), obtained by fitting the binding data onto a monoexponential binding model, were almost identical for all three ␣V␤6-fibrillin-1 complexes with K d values of ϳ0.9 M (Table 1). These SPR results are in good agreement with the observed cellular data since all three fibrillin-1 fragments were able to support recruitment of ␣V␤6 into focal contacts and the ensuing cell spreading (Figs. 3 and 4).
Because amine coupling of ligands to the sensor surface often leads to their partial inactivation and reduction of the apparent affinity for analyte (20), the kinetics of ␣V␤6 binding to fibrillin was remeasured with ligands immobilized via streptavidin affinity capture. cbEGF22-TB4, TB4-cbEGF23, and the RGA mutant were expressed as C-terminal BirA fusion proteins and subjected to site-specific biotinylation. Biotinylated FIII9-10 (9) was used as a positive control. Ligands were immobilized at the same density, and binding of soluble ␣V␤6 was monitored under the same conditions as before. Analysis of sensorgrams confirmed that the k on and the k off of ␣V␤6 binding to cbEGF22-TB4 and TB4-cbEGF23 are essentially the same, suggesting an equivalent integrin-ligand interface is formed in both cases ( Table 1). The overall affinity of the interaction was calculated to be 2-fold higher than that obtained with aminecoupled ligands (K d ϳ 0.45 M versus 0.9 M). This difference is most likely due to the partial inactivation of amine-coupled fibrillin fragments and the more native-like presentation of the streptavidin-captured ligand.
Together, these SPR results show that the two RGD-containing domain pairs and triple domain fragment form equally stable complexes with ␣V␤6 integrin. Furthermore, the data demonstrate that the molecular determinants of the ␣V␤6/ fibrillin-1 interaction are contained within the TB4 domain alone and are independent of the flanking cbEGF domains. To the best of our knowledge, this is also the first report of the K d value for ␣V␤6/fibronectin FIII9-10 interaction (ϳ0.18 M). Given that K d values for integrin binding to physiological ligands vary from a few nanomolar to high micromolar, the binding of ␣V␤6 to either fibronectin FIII9-10 or fibrillin fragments can be considered as a medium affinity interaction that is not as high affinity as that observed for ␣V␤3 (8).
Integrin ␣5␤1 Is Differentially Distributed in Cells Adhering to Fibrillin-1 and Fibronectin-The VB6 cell adhesion assay with function blocking antibodies (Fig. 1B) suggested integrin ␣5␤1 in addition to ␣V␤6 mediates some adhesion to fibrillin-1 and contributes to the intracellular signaling leading to cell spreading. This observation is consistent with an earlier study in which dermal fibroblasts were shown to adhere to recombinant fibrillin fragments and purified microfibrils predominantly via ␣5␤1 (7). However, the signaling events triggered by ␣5␤1 binding to fibrillin- 1 have not yet been investigated. To ascertain whether ␣5␤1 assembles into focal contacts upon fibrillin-1 engagement, VB6 cells were immunofluorescently labeled for ␣5␤1 integrin and F-actin after short term culturing on fibrillin-1 fragments and control fibronectin domain pair. Interestingly, the subcellular distribution  of the integrin on these two substrates was found to be quite different (Fig. 6A). In VB6 cells spread on cbEGF22-TB4, ␣5␤1 showed limited colocalization with actin fibers and was mostly distributed homogeneously over the entire cell surface. A similar diffuse staining pattern was observed on TB4-cbEGF23 and the triple domain construct. In contrast, on FIII9-10, the majority of the integrin was found to colocalize with actin in sharply defined focal contacts at the ends of stress fibers. To confirm the observed distribution patterns, VB6 cells were simultaneously immunolabeled for ␣5␤1 and vinculin, a protein marker of an assembled focal complex. As shown in Fig. 6B on full-length fibronectin and the FIII9-10 domain pair, the integrin is almost completely colocalized with vinculin in prominent focal contacts. However, in cells adherent on fibrillin ligands, integrin remains diffuse and shows minimal colocalization with vinculin. Finally, VB6 cells were double-labeled for ␣5 and ␤6 integrins to directly compare their distribution on fibrillin-1 substrates within the same cell (Fig. 6C). In contrast with fibronectin-associated focal contacts, which were found to be both ␣5and ␤6-rich, fibrillin ligands selectively recruited ␤6 into well defined focal contacts, whereas ␣5 retained its uniform and indistinct distribution over much of the ventral cell surface.
The non-focal distribution of ␣5␤1 on fibrillin-1 constructs was recapitulated in stromal fibroblasts, which like dermal fibroblasts characterized by Bax et al. (7), adhere to fibrillin predominantly via ␣5␤1 with a contribution from ␣V␤3 (supplemental Fig. 1). This indicates that the fibrillin-induced nonfocal distribution of ␣5␤1 is an intrinsic feature of this ligandreceptor complex that does not depend on the degree of ␣5␤1 involvement in the overall adhesion. Taken together, the variation in ␣5␤1 staining pattern and the notable difference in the stromal fibroblast morphology upon adhesion to fibrillin and fibronectin substrates (supplemental Fig. 1C), suggest that these ligands might elicit distinct degrees or different types of intracellular signaling upon ␣5␤1 engagement.
Fibrillin-1 cbEGF22-TB4 Is a Low Affinity Ligand for ␣5␤1 Integrin-The inability of any of the fibrillin-1 RGD-containing fragments to trigger efficient ␣5␤1 clustering in focal contacts suggested that its affinity for fibrillin-1 might be substantially lower than that of ␣V␤3 or ␣V␤6. To test this hypothesis, we analyzed ␣5␤1 binding to cbEGF22-TB4 and TB4-cbEGF23 by SPR using ␣V␤6 as a control analyte. Recombinant full-length ␣5␤1 integrin was injected over fibrillin ligands or FIII9-10 immobilized at equimolar levels via a C-terminal biotin tag. The binding of ␣5␤1 to fibrillin ligands was only marginal compared with the FIII9-10 control, indicating that fibrillin-1 is a weak ligand for this integrin (Fig. 7A). Plotting the near equilibrium binding responses obtained with different concentrations of the integrin shows that at saturation there is an ϳ30-fold difference between the levels of ␣5␤1 binding to FIII9-10 and cbEGF22-TB4 domain pairs (Fig. 7B). Kinetic analysis of sensorgrams yielded a K d of 10.8 Ϯ 2.3 nM for ␣5␤1/FIII9-10 interaction, which is in agreement with the published data (21,22). However, the kinetics of ␣5␤1-fibrillin-1 complex formation could not be reliably measured by SPR due to its low affinity and the consequent requirement for the large amounts of analyte. We, therefore, sought to characterize the ␣5␤1/fibrillin-1 interaction in reverse orientation by enzyme-linked immunosorbent assay. Integrin was immobilized on the surface of a microtiter plate, and biotinylated fibrilllin-1 fragments were added as soluble ligands. Titration curves demonstrated that binding of cbEGF22-TB4 or TB4-cbEGF23 to ␣5␤1 was negligible compared with that of equimolar FIII9-10 (Fig. 7C). An EC 50 of 10.7 Ϯ 4.4 nM was obtained for ␣5␤1/FIII9-10 interaction. At a concentration of 1000 nM cbEGF22-TB4, half-maximum binding to ␣5␤1 was still not reached, indicating that the K d of ␣5␤1/cbEGF22-TB4 interaction was Ͼ1 M if one assumes simple 1:1 Langmuir binding. These results suggest fibrillin-1 is a low affinity ligand for ␣5␤1 integrin, with the K d of this interaction Ͼ100-fold higher than that of the ␣5␤1/ FIII9-10 interaction.

DISCUSSION
This study identifies the epithelial integrin ␣V␤6 as a novel medium affinity receptor for human fibrillin-1. ␣V␤6 is normally expressed only at low levels in adult tissue but is rapidly up-regulated in response to injury, inflammation, tumorigenesis, and wound repair and during development. Because fibrillin is known to be secreted by epithelial as well as mesenchymal cells, it is present in the extracellular matrix surrounding cells in which ␣V␤6 becomes up-regulated (17). The fibrillin-1/ ␣V␤6 interaction is, therefore, likely to play a physiological role in epidermal cell adhesion, and fibrillin-1 joins the list of other known extracellular matrix ligands for ␣V␤6, including fibronectin, vitronectin, and tenascin (10,31,32,33).
Despite sharing the RGD binding ability, many integrins can nevertheless discriminate between their RGD-containing ligands. In the absence of high resolution structures of integrins in complex with macromolecular ligands, the structural basis for the binding specificity between integrins and their RGD ligands remains poorly understood. In this study we applied an interdisciplinary approach to elucidate the molecular determinants of fibrillin-integrin interactions beyond the known requirement for a solvent-exposed RGD motif in TB4. The dissection of structure-activity relationships using partially overlapping TB4-containing domain pairs as ligands revealed that molecular determinants for ␣V␤6 binding to fibrillin-1 reside within the RGD-containing TB4. The stability of the ␣V␤6fibrillin-1 complex is unaffected by the removal of flanking cbEGF domains, clearly showing that they do not contribute to the ␣V␤6 binding interface either directly or indirectly. In con-  (8) shows that, in addition to the proposed role of cbEGF22 in ␣V␤3 activation, residues just downstream of RGD are likely to interact with the ␣1 helix and the specificity-determining loop of the ␤3 I-like domain. These structural elements show very little sequence FIGURE 6. Subcellular localization of ␣5␤1 integrin in human VB6 keratinocytes upon adhesion to fibrillin-1. VB6 cells were incubated for 30 min on coverslips coated with cbEGF22-TB4, TB4-cbEGF23, or cbEGF22-TB4-cbEGF23 fragment. Coverslips coated with FIII9-10 were used as a positive control. After 30 min cells were fixed, permeabilized, and fluorescently labeled. A, F-actin was stained green with phalloidin, and nuclei were stained blue with 4Ј,6-diamidino-2-phenylindole, and ␣5␤1 integrin is red by indirect fluorescence with the JBS5 mAb. In VB6 cells adhering to FIII9-10, the majority of ␣5␤1 integrin colocalizes with actin in focal contacts at the ends of stress fibers. However, in cells adherent on fibrillin-1 fragments, the integrin has a mostly diffuse distribution over the entire cell surface, showing little colocalization with actin fibers. B, double immunofluorescent labeling of VB6 cells for ␣5␤1 integrin (green) and focal contact marker vinculin (red). Colocalization of these two proteins within focal contacts on full-length fibronectin (Fn) and FIII9-10 produces an orange-yellow color. On fibrillin-1 ␣5␤1 integrin remains diffusely distributed over the entire cell surface, leaving the focal contacts at the cell periphery colored red due to their high vinculin content. C, simultaneous immunofluorescent detection of ␤6 (green) and ␣5 (red) integrins with the corresponding antibodies (CS␤6 and AB1928, respectively) in VB6 cells shows that fibronectin-associated focal contacts have an orange-yellow color due to colocalization of integrins within these assemblies. However, a homogenous ␣5 distribution pattern and selective focal contact recruitment of ␤6 make these structures predominantly green on fibrillin-1 substrates. Scale bar, 10 m.
The role of ␣5␤1 integrin in fibrillin-1 function has been controversial for some time. Earlier cell adhesion assays with function-blocking antibodies, performed on purified microfibrils and recombinant fibrillin-1 fragments, suggested that ␣5␤1 integrin functions as a fibrillin-1 receptor in certain cell lines (7,26). However, several attempts to demonstrate this interaction in a cell-free system have been unsuccessful (4,8).
Because specific interference of one integrin has been known to influence the behavior of other integrins on the cell (a phenomenon called "integrin cross-talk" (27)), functionblocking studies that use integrinspecific antibodies cannot by themselves be considered as conclusive proof of integrin-ligand interactions. Using a biochemical approach and recombinant integrin, we show for the first time that ␣5␤1 is a low affinity receptor for fibrillin-1. Immunolocalization of ␣5␤1 integrin in cells spread on fibrillin-1 cbEGF22-TB4 or cbEGF22-TB4-cbEGF23 fragments revealed very limited recruitment of the integrin into focal contacts, giving rise to diffuse staining. Because assembly of integrins into focal contacts depends on the affinity of the integrin-ligand pair (28), the inability of fibrillin ligands to induce efficient ␣5␤1 clustering confirms the low affinity of this interaction at the cell surface. Furthermore, the distinct ␣5␤1 distribution observed in cells adherent on cbEGF22-TB4 and FIII9-10 domain pairs (Figs. 6, A-C, and supplemental Fig. 1B) raises the possibility that engagement of ␣5␤1 integrin by fibrillin-1 and fibronectin may have distinct cellular consequences. The initial event in outside-in signaling is integrin clustering in the plane of the membrane upon ligation, leading to the assembly of focal complexes and their subsequent maturation into focal contacts (30). The rate of integrin clustering depends on a number of factors including ligand and receptor density, rate of integrin diffusion in membrane, and the lifetime of the ligand-receptor complex (28). All other factors being equal, the lifetime of the receptor-ligand complex, determined by the affinity of the interaction, will be the crucial factor in the initiation and growth of focal contacts. With the affinity of ␣5␤1/cbEGF22-TB4 interaction at least 100-fold lower than that of ␣5␤1/FIII9-10, the dissociation rate of ␣5␤1/cbEGF22-TB4 complex might be too fast to ensure efficient nucleation and growth of focal contacts, giving rise to small and transient integrin clusters. These, although contributing to the overall cell attachment, might trigger a lesser degree or different type of signaling resulting in the gain of  A and B) and enzyme-linked immunosorbent assay (C). A, sensorgrams shown were obtained when recombinant ␣5␤1 integrin was injected over sensor surfaces with equimolar amounts of cbEGF22-TB4 or FIII9-10 immobilized via a C-terminal biotin tag (ϳ1000 RU). The bulk response, obtained in the flow cell with an RGA mutant of cbEGF22-TB4-cbEGF23, was subtracted from all sensorgrams. Sensorgrams obtained with recombinant ␣V␤6 integrin are included to demonstrate that low levels of ␣5␤1 binding are not due to inactivation of fibrillin upon immobilization. B, the near-equilibrium SPR binding response was plotted versus analyte concentration to obtain the dose-dependent binding curve. The error bars represent S.D. C, solid phase binding assay of biotinylated fibrillin-1 and fibronectin fragments to immobilized ␣5␤1. Soluble cbEGF22-TB4, TB4-cbEGF23, and FIII9-10 (3-1000 nM) were added to wells coated with 150 ng of the integrin. Binding was detected with peroxidase-conjugated streptavidin. The figure shows a representative result from five independent experiments. specific cellular function such as motility. This would account for the differences in ␣5␤1 distribution (Fig. 6, A-C, and supplemental Fig. 1B) and fibroblast morphology on the two substrates, which was also noted by Bax et al. (7). Although it remains possible that long range effects may modulate ␣5␤1 integrin affinity in full-length fibrillin-1, the similarity of our data with that of Bax et al. (7), where larger fibrillin fragments and microfibrils were utilized, suggests that these effects would be moderate.
It would be interesting to use site-directed mutagenesis to endow cbEGF22-TB4 with higher ␣5␤1 affinity. A model of the ␣5␤1 headpiece predicts an extensive stretch of acidic residues on the top of the ␣/␤ interface, where the ligand binding site is located (29). cbEGF22 has five acidic residues (Asp-1512, Asp-1516, Glu-1518, Glu-1552, and Glu-1584) projecting from the same face of the molecule as the RGD loop that are not counterbalanced by neighboring basic residues and might, therefore, cause an electrostatic repulsion from the integrin surface.
The findings described in this study suggest a broader and more diverse involvement of the integrin superfamily in fibrillin-1 function than previously anticipated. Our data show that the RGD motif of fibrillin-1 is recognized by at least three different integrin receptors with a large spectrum of binding affinities and provide novel insights into structural requirements and cellular consequences of these interactions. These results may find application in the understanding of the mouse Tsk phenotype, which is caused by secretion of a fibrillin-1 polypeptide with a duplication of RGD-containing TB4 domain, and also in the advancement of vascular tissue grafts, where the current challenge is to develop a cell-adhesive matrix that supports selective cell adhesion and activity without the risk of thrombosis (23)(24)(25).