INTRODUCTION

Interaction between cells and the extracellular matrix (ECM)¹ is critical for diverse biological processes. Binding of cells to ECM proteins is attributed primarily to the interaction between integrin receptors, heterodimeric transmembrane proteins involved in adhesion and communication, and specific peptide sequences in each ECM molecule (1, 2). For example, certain v- and 1-containing integrins recognize the RGD sequence displayed by many ECM proteins. 41 integrin binds the sequence LDV in the CS-1 region of fibronectin as well as the sequence QIDS in VCAM. Accordingly, much research has focused on the identification of sequence motifs responsible for integrin-ligand interaction.

Changes in integrin activity, which affect functions as diverse as strength of adhesion, natural ligand specificity, and matrix assembly (3, 4), can be induced by several agents in vitro and are associated with changes in integrin conformation. Mn²⁺ and activating antibodies such as TS2/16 and 8A2 bind to the receptor, independent of activation state, and induce an active conformation (5-7). Less specific cellular activators such as PMA can also change the integrin activation state (8). Changes in receptor conformation have been documented by anti-integrin antibodies that specifically bind receptors in the activated state (9, 10).

Many cell-surface and ECM proteins are, in part, composed of multiple repeating domains of ~90 amino acids known as fibronectin type III (FNIII) repeats. Cell-surface receptors containing these repeats include the human growth hormone receptor, the erythropoietin receptor, and multiple interleukin receptors; cell-surface adhesion molecules include chicken L1 and Drosophila neuroglian; and ECM proteins including fibronectin, tenascin, and certain collagens (11, 12). In fibronectin and tenascin, the integrin-binding RGD sequence is displayed on specific FNIII repeats. NMR and x-ray analyses (13, 14) demonstrate that FNIII repeats, although only weakly homologous (~20% identity) at the protein sequence level (12), have very similar tertiary structures. The dominant feature of all FNIII repeats is a sandwich formed by two anti-parallel -sheets enclosing a hydrophobic core. FNIII repeats adjacent to the RGD-containing FNIII₁₀ contribute to cell adhesion mediated by fibronectin (15-17). With the exception of the synergy sequence (PHSRN) in FNIII₉ of fibronectin, no systematic effort has been made to understand this contribution or to identify potential cell-surface receptors for FNIII repeats.

We now report that cells adhere and spread on FNIII repeats lacking RGD after experimental activation of 1 integrins. FNIII repeats derived from both extracellular matrix proteins and cytokine receptors mediate adhesion of multiple cell types. Interaction between individual FNIII domains and 1 integrins mediates focal adhesion kinase phosphorylation and subsequent stress fiber and focal contact formation. These data suggest that, in vivo, many FNIII-containing proteins may bind and signal through activated 1 integrins, dramatically expanding the potential for integrin-dependent intercellular and cell-matrix communication.

EXPERIMENTAL PROCEDURES

Cell adhesion assays were performed and quantified as described (18). Cell adhesion buffer contained 10 mM Hepes, pH 7.4, 150 mM NaCl, 0.25% bovine serum albumin, and 2 mM glucose with varying concentrations of Mn²⁺. The relative amount of each recombinant FNIII repeat bound to plastic was quantitated by an enzyme-linked immunosorbent assay-based assay using a mAb to the histidine tag (diaNovo). The absorbance of FNIII repeats to plastic was linear between 0.5 and 5.0 µg/ml and reached saturation at ~8.0 µg/ml. FNIII repeats were plated at 10 µg/ml to standardize the relative amount of each repeat bound to plastic. Native FNIII domains (a gift of Drs. Sergei V. Litvinovitch and Kenneth C. Ingham, American Red Cross, Rockville, MD) were coated at 10 µg/ml, human plasma fibronectin at 2.5 µg/ml, and poly-L-lysine (Sigma) at 10 µg/ml. Anti-RGD antibody 16G3 (19) was used at 50 µg/ml. Isolated peripheral blood mononuclear cells were cultured for up to 2 weeks in the wells of a 24-well dish coated previously with anti-CD3 antibody OKT-3. T blasts were recovered and tested for adhesion in the presence of 2.5 µM PMA (Sigma).

Fibronectin contains 15-17 FNIII repeats numbered sequentially from the most proximal repeat to the amino terminus (20). DNA encoding individual FNIII repeats was amplified by polymerase chain reaction from a full-length rat fibronectin cDNA. Purified amplification products were cloned into the expression vector pQE30 (QIAGEN Inc.) and sequenced. Recombinant FNIII repeats, which include an additional 18 amino acids (MRGSH₆GSACELGT) at the amino terminus and 3 additional amino acids (KLN) at the carboxyl terminus, were expressed in Escherichia coli JM109 (Stratagene). All FNIII domains were soluble and purified on Ni²⁺-nitrilotriacetic acid-agarose (QIAGEN Inc.) according to the manufacturer's instructions. Native FNIII repeats as well as FNIII repeats expressed as glutathione S-transferase fusion proteins mediate cell adhesion. Polyhistidine did not support cell adhesion, indicating that cells did not adhere to FNIII repeats through the histidine tag.

The soluble extracellular domains of the human IL-2 receptor  chain, the IL-4 receptor  chain, and the IL-2 receptor c chain were expressed by cloning the corresponding polymerase chain reaction fragments into pBlueBac II (Invitrogen) or pFASTBAC I (Life Technologies, Inc.) baculovirus expression vectors. Recombinant proteins were expressed in Hi-5 insect cells (Invitrogen) and purified by mAb 18741D (Pharmingen) or mAb 230 (R&D Systems) affinity chromotography for the IL-2 receptor  and IL-4 receptor  chains, respectively, or by nickel chelate (Ni²⁺-nitrilotriacetic acid) affinity chromatography for the IL-2 receptor c chain.

PAC1 cells, in 10 mM Hepes, pH 7.4, 150 mM NaCl, 0.25% bovine serum albumin, 2 mM glucose, 100 µM MnCl₂, and 1% fetal calf serum, were plated on 100-mm plastic dishes (Corning Inc.) coated previously with either 20 µg/ml FNIII repeats or 10 µg/ml poly-L-lysine. At 20, 40, or 60 min, cells were lysed in 1 ml of 1% Triton X-100, 0.1% SDS, 0.5% deoxycholate, 150 mM NaCl, 10 mM Tris, pH 7.4, 1 mM sodium vanadate, and 0.2 mM phenylmethylsulfonyl fluoride, and cellular debris was removed by centrifugation. To each 1 ml of lysate were 5 µg of anti-FAK mAb (Signal Transduction Laboratories) and/or 1 µg of anti-phosphotyrosine polyclonal antibody (Signal Transduction Laboratories). After a 1-h incubation at 4 °C, 60 µl of a 50% slurry of protein A-Sepharose beads were added, followed by an additional 1-h incubation at 4 °C. Immunoprecipitated proteins were subjected to Western blot analysis as described (18).

PAC1 cells in 10 mM Hepes, pH 7.4, 150 mM NaCl, 0.25% bovine serum albumin, 2 mM glucose, 100 µM MnCl₂, and 1% fetal calf serum were plated in each well of two-well chamber slides (Nunc) coated previously with 20 µg/ml FNIII repeats or 10 µg/ml poly-L-lysine and allowed to adhere for 2 h. Cells were fixed; permeabilized; and incubated with rhodamine-conjugated phalloidin (Molecular Probes, Inc.), anti-phosphotyrosine polyclonal antibody, rabbit polyclonal antibody directed against the cytoplasmic domain of chicken 1 integrin (21), or anti-vinculin mAb (Sigma). Primary antibodies were detected with either an anti-mouse or anti-rabbit biotinylated IgG (Amersham Corp.) and streptavidin coupled to either Texas Red or fluorescein (Amersham Corp.). Cells were mounted in Movoil (Hoechst Celanese) and viewed under fluorescence at 630× magnification. Note that secreted or serum-derived proteins are not required for focal contact formation. Focal contacts form in cyclohexamide-treated cells and under conditions in which serum is replaced with lysophosphatidic acid.

RESULTS

While analyzing pulmonary artery smooth muscle cell (PAC1) (22) adhesion to fibronectin, we observed that, in the presence of Mn²⁺, recombinant or native FNIII domains derived from fibronectin support cell adhesion nearly equal in extent compared with the RGD-containing central binding domain, FNIII_8-10 (Fig. 1A). Single FNIII repeats support adhesion equally as well as multiple repeats. Furthermore, an antibody that blocks RGD-dependent adhesion inhibits PAC1 cell adhesion to intact fibronectin in the presence of Ca²⁺ and Mg²⁺, but not in the presence of Mn²⁺, suggesting that cells adhere through FNIII repeats other than FNIII₁₀, which contains RGD. Recombinant FNIII repeats derived from tenascin also support Mn²⁺-dependent adhesion.²

[View Larger Version of this Image (29K GIF file)]

To investigate systematically FNIII-mediated adhesion, we studied recombinant fibronectin repeats FNIII₁, FNIII₃, FNIII₄, and FNIII₁₀. FNIII₁₀, which contains the integrin-binding RGD sequence, serves as a positive control and comparison standard for studying adhesive events. Pairwise amino acid sequence comparison of FNIII₁, FNIII₃, FNIII₄, and FNIII₁₀ reveals a range of identity from 12% (FNIII₁ versus FNIII₃) to 30% (FNIII₃ versus FNIII₄). Simultaneous comparison of all four sequences reveals 21% identity at positions in at least three of the sequences, and there is no obvious region of sequence identity that might specifically mediate cell adhesion. A pairwise analysis of 26 animal FNIII sequences showed 20% identity (12). Thus, FNIII₁, FNIII₃, FNIII₄, and FNIII₁₀ resemble each other no more than they resemble FNIII domains from proteins other than fibronectin. We therefore believe that FNIII₁, FNIII₃, FNIII₄, and FNIII₁₀ constitute a representative spectrum of FNIII repeat sequences.

Cell adhesion to FNIII₁₀ does not require Mn²⁺, but does require Ca²⁺ or Mg²⁺. Cell adhesion to FNIII₁, FNIII₃, and FNIII₄ is Mn²⁺-dependent (Fig. 1B) and is inhibited by the addition of Ca²⁺ (Fig. 1C). Different FNIII repeats require various concentrations of Mn²⁺ to support specific levels of adhesion. FNIII₃ requires 30 µM Mn²⁺ to support 50% cell adhesion, whereas FNIII₁ and FNIII₄ require 50 and >100 µM, respectively (Fig. 1B). Conversely, a higher concentration of Ca²⁺ is required to completely inhibit FNIII₃-mediated adhesion relative to FNIII₁-mediated adhesion (Fig. 1C). Relative levels of the FNIII repeats bound to tissue culture plastic were equal as measured by an immunoassay. Thus, specificity of adhesion apparently reflects sequence variation among FNIII repeats.

Soluble FNIII repeats block FNIII-mediated adhesion (Fig. 1D). FNIII₄- and FNIII₁-mediated cell adhesion is competed by FNIII₃, suggesting that all FNIII repeats are using the same receptor(s). IgG domains are very closely related to FNIII repeats in structure (12), but cells do not adhere to immunoglobulin, which contains multiple IgG domains, or to recombinant CD2, which is composed of a pair of IgG domains (Figs. 1B and 3). FNIII-dependent cell adhesion is not competed by soluble VCAM (23), which contains seven IgG domains, or by fibronectin CS-1, which contains the 41 integrin-binding site, demonstrating apparent specificity for the FNIII repeat structure (Fig. 1D). Furthermore, neither ovalbumin nor recombinant CD40, a member of the tumor necrosis factor superfamily, supports cell adhesion (data not shown).

Adhesion of embryonic kidney 293 cells to the recombinant IL-2 receptor  chain (IL-2R), IL-4 receptor  chain (IL-4R), and IL-2 receptor c chain (IL-2Rc) and to the soluble extracellular domain of CD2 (composed of two IgG domains) in the presence of 1 mM Ca²⁺ and 1 mM Mg²⁺ (white bars), antibody TS2/16 (stippled bars), 100 µM Mn²⁺ (black bars), or 100 µM Mn²⁺ and 10 µg/ml blocking anti-1 integrin antibody Ha2/11 (hatched bars) (37).

[View Larger Version of this Image (18K GIF file)]

Cell Adhesion to FNIII Repeats Is Mediated by Activated 1 Integrins

Numerous studies have demonstrated that integrin-dependent adhesion to certain ligands is enhanced by the presence of Mn²⁺ and blocked by the addition of Ca²⁺. For instance, Mn²⁺ stimulates and Ca²⁺ abrogates the 11 integrin-dependent adhesion of NB100 cells to collagen (24). Similarly, 41 integrin has multiple affinity states, the highest of which is induced by Mn²⁺ as well as other integrin-activating reagents (25). We therefore investigated the involvement of integrins in FNIII-mediated adhesion.

Adhesion to all tested repeats can be inhibited by blocking antibodies to 1 integrin (Fig. 2, A and B). Blocking cell adhesion to FNIII₃ required a higher concentration of antibody than blocking adhesion to FNIII₁ or FNIII₄. In the presence of 100 µM Mn²⁺, cell adhesion to FNIII₁, FNIII₃, and FNIII₄ was blocked completely with 5 µg/ml anti-1 integrin antibody, whereas adhesion to FNIII₁₀ was blocked by 80% (Fig. 2A). In the presence of 200 µM Mn²⁺, however, only adhesion to FNIII₄ was completely blocked by antibody at 5 µg/ml (Fig. 2B), whereas there was little, if any, blocking of FNIII₃- and FNIII₁₀-dependent adhesion at up to 10 µg/ml antibody. These data are consistent with results demonstrating that different FNIII repeats require different amounts of Mn²⁺ to permit equivalent levels of cell adhesion.

Activated 1 integrin-dependent cell adhesion to FNIII repeats.

[View Larger Version of this Image (24K GIF file)]

To take advantage of integrin-activating reagents other than Mn²⁺, we investigated human kidney epithelial 293 cell adhesion to FNIII repeats. This cell line, along with others (see below and Fig. 4), adheres to FNIII repeats in the presence of Mn²⁺. Activating anti-1 integrin antibody TS2/16 (6) stimulates adhesion of 293 cells to FNIII₁, FNIII₃, and FNIII₄ at concentrations of Mn²⁺ that do not support adhesion (Fig. 2C) as well as under conditions in which Mn²⁺ is replaced with Ca²⁺ and Mg²⁺ (Fig. 2D). Thus, Mn²⁺ can be replaced with an activating anti-1 integrin antibody to support FNIII-mediated adhesion. We conclude that activated 1 integrins mediate cell adhesion to FNIII repeats.

Adhesion of umbilical vein endothelial cells (UVEC; Clonetics Corp.), fibroblasts (FIB.; Clonetics Corp.), K562 cells transfected with 4 (K562(4)) (38), embryonic kidney 293 cells, and rat aortic smooth muscle cells (SMC) (39) to FNIII₃ and of CD3-stimulated T cells to FNIII₁ in the presence of 1 mM Ca²⁺ and 1 mM Mg²⁺ (white bars), 100 µM Mn²⁺ (black bars), 2.5 µM PMA (stippled bar), 100 µM Mn²⁺ and 10 µg/ml blocking anti-1 integrin antibody Ha2/11 (hatched bars) (37), or 2.5 µM PMA and 10 µg/ml blocking anti-1 integrin antibody Ha2/11 (striped bars).

[View Larger Version of this Image (18K GIF file)]

Most 1 integrin ligands interact with a restricted subset of integrin heterodimers. For instance, collagen interacts with 11, 21, and 31 integrins, but not with other members of the 1 integrin family. However, no tested blocking monoclonal antibodies to  integrin subunits blocked FNIII-mediated adhesion (data not shown). Furthermore, a screen designed to identify monoclonal antibodies that block FNIII-mediated cell adhesion identified multiple anti-1 integrin antibodies, but no anti- integrin antibodies, consistent with results reported here.² We speculate that FNIII-mediated adhesion is independent of specific  integrin subunits.

The extracellular domains of many cytokine receptors are composed of a tandem repeat, the repeating unit of which bears evolutionary resemblance to FNIII repeats (11, 12, 26). We therefore investigated the potential adhesion of cells to FNIII repeats derived from various cytokine receptors. 293 cells adhere to the recombinant extracellular domains of the interleukin-2 receptor  chain, the interleukin-4 receptor  chain, and the interleukin-2 receptor c chain in the presence of Mn²⁺ or antibody TS2/16. Cells do not adhere to recombinant CD2, which is composed of a pair of IgG domains, and adhesion is completely abrogated by anti-1 integrin-neutralizing antibody (Fig. 3). These data show that FNIII repeats derived from molecules other than extracellular domain proteins interact with activated 1 integrins and suggest the potential for direct interaction of integrins and cytokine receptors.

FNIII-mediated adhesion is not specific to the PAC1 smooth muscle cell line. Primary human dermal fibroblasts, endothelial cells, and 293 epithelial cells also adhere to FNIII repeats in an activated 1 integrin-dependent manner (Fig. 4). Nonadherent T lymphocytes, after activation by an anti-CD3 antibody, also bound FNIII repeats when stimulated further by PMA or Mn²⁺. As the effects of PMA on integrin activation are well documented, these data further support the observation that adhesion to FNIII repeats is mediated by activated integrins.

Certain cell lines did not support FNIII-mediated adhesion. The extent to which 1 integrins are "activated" may not be sufficient to promote adhesion. K562 cells are particularly sensitive to activation. For example, K562 cells expressing 11 integrin will bind laminin in an 1-dependent manner only in the presence of integrin-stimulating reagents (27). Alternatively, certain cultured cells may not express the intracellular components required for strong adhesion to the FNIII repeats. For instance, binding of the integrin LFA-1(L2) to its ligand ICAM-1 (intercellular adhesion molecule 1) can be induced by expression of the intracellular protein cytohesin-1, which interacts with the intracellular domain of 2 integrin (28). It is important to note that all primary cell lines we have tested (T lymphocytes, fibroblasts, and endothelial cells) adhere to FNIII repeats.

Interaction between integrins and ligands mediates intracellular signaling cascades that influence many physiological processes, including changes in intracellular Ca²⁺, pH, tyrosine phosphorylation, gene expression, and rearrangement of the actin cytoskeleton (29). Phosphorylation of FAK is associated with integrin-mediated signal transduction (29). To investigate whether the interaction between cells and FNIII repeats supports signaling, we tested for adhesion-dependent FAK phosphorylation. Within 40 min, FAK was phosphorylated in response to cell adhesion on all tested FNIII domains as well as intact fibronectin (Fig. 5). At 40 min after plating, very few cells have begun to spread on FNIII₁, FNIII₃, and FNIII₄ (data not shown), consistent with published data suggesting that phosphorylation of FAK precedes cell spreading (30). Furthermore, immunostaining of cells spread on the different FNIII domains with an anti-phosphotyrosine antibody localized sites of phosphorylation to focal contacts (Fig. 5B; see below). These data demonstrate that the interaction of cells with FNIII domains can precipitate the initial events associated with integrin-mediated intercellular signaling.

[View Larger Version of this Image (69K GIF file)]

Signaling via integrins can result in the rearrangement of the actin cytoskeleton and the formation of focal contacts, sites where transmembrane integrins link the extracellular matrix to the intracellular cytoskeleton. Adhesion of cells to FNIII domains results in the formation of stress fibers (Fig. 6), indicative of actin cytoskeleton rearrangement, as well as in the formation of focal contacts as demonstrated by double immunofluorescence using anti-vinculin and anti-1 integrin antibodies (Fig. 6). Thus, in the presence of integrin-activating reagents, FNIII repeats mediate integrin-dependent architectural changes within the cell.

1

[View Larger Version of this Image (81K GIF file)]

DISCUSSION

Our data demonstrate that FNIII repeats mediate adhesion and signaling through experimentally activated 1 integrins. Although a comparison of aligned sequences reveals no obvious region common to FNIII repeats that might participate in integrin binding, 1 integrins may interact through particular charged residues in a specific loop. D'Souza et al. (31) have proposed that integrins bind RGD-containing ligands through a cation displacement mechanism. In this scenario, the Asp residue provides a transitional cation coordination site during cation displacement and ligand binding (31). All FNIII domains contain solvent-accessible Asp or Glu residues in loop E-F, on which RGD is displayed in FNIII₁₀. The activated carboxyl residues in either of these amino acids could provide an analogous transitional cation coordination site. Alternatively, 1 integrin subunits may bind FNIII domains through a set of amino acids that occupy homologous positions in the tertiary structure, but are separated in primary sequence and therefore undetectable by simple sequence alignment.

Although the physiological activation of integrins is not clearly understood, integrin activity can be modulated in vitro by a variety of agents, including divalent cations, phorbol esters, and activating antibodies. We have demonstrated that 1 integrins, activated by three independent reagents (Mn²⁺, PMA, and mAb TS2/16), will adhere to FNIII repeats. PMA and antibody TS2/16 are clearly not physiological activators of 1 integrins, although the former implicates the protein kinase C pathway in activation. Mn²⁺ may, however, be a physiological activator. The concentration of Mn²⁺ in tissue is estimated at 1-14 µM, and estimates as high as 50 µM in bone and 30 µM in liver have been reported (32, 33). These latter figures are consistent with the concentration of Mn²⁺ required for 1 integrin-mediated adhesion to FNIII repeats.

The increase in cell adhesivity associated with activated integrins is attributed either to an increase in affinity for ligand or to post-occupancy-mediated events such as cytoskeleton assembly (4). Both mechanisms assume an interaction between the integrin and a defined sequence within the ligand (e.g. RGD). We have now demonstrated a third mechanism by which activated integrins can increase cell adhesivity, namely, through promiscuous interactions with multiple FNIII repeats. Activated integrins are associated with numerous physiological processes in multiple cell types. Vascular smooth muscle cells, endothelial cells, and activated T lymphocytes express activated integrins in vivo (34, 35). Assembly of a fibronectin matrix, critical to diverse biological processes, requires both activated integrins and reorganization of an actin cytoskeleton (4). Interactions between activated integrins and FNIII repeats could stabilize the interaction of cells with fibronectin and enhance cytoskeleton rearrangement to help modulate matrix assembly. Interactions between RGD and 51 integrin or between CS-1 and 41 integrin are thought to account for fibronectin-mediated T cell functions such as migration, signal transduction, and differentiation (36). FNIII repeats found in fibronectin and other matrix proteins may augment RGD- and CS-1-dependent interactions. Numerous extracellular domains of interleukin receptors contain FNIII repeats (11, 12). It is speculative, but reasonable to suggest, that activated 1 integrins may bind directly to cytokine receptors via FNIII repeats to induce or otherwise modulate both integrin- and cytokine-mediated signaling. This suggestion is supported by our observation that FNIII repeats derived from cytokine receptors support 1 integrin-dependent adhesion.

In conclusion, we have demonstrated that non-RGD-containing FNIII repeats with diverse sequences mediate cell adhesion through activated 1 integrins. This interaction is apparently specific to FNIII repeat structure as we have not consistently observed activated integrin-dependent adhesion to other proteins. The specific adhesivity of individual repeats apparently depends on variations in FNIII sequence. Furthermore, engagement of activated 1 integrins by FNIII repeats results in physiological responses by cells including tyrosine phosphorylation and cytoskeleton rearrangement. Given the number of molecules that contain FNIII domains, the potential for integrin-dependent intercellular and cell-matrix communication is dramatically increased and may significantly enhance sequence-specific integrin-ligand interactions.