Mechanically Activated Integrin Switch Controls α5β1 Function

See allHide authors and affiliations

Science  30 Jan 2009:
Vol. 323, Issue 5914, pp. 642-644
DOI: 10.1126/science.1168441


The cytoskeleton, integrin-mediated adhesion, and substrate stiffness control a common set of cell functions required for development and homeostasis that are often deranged in cancer. The connection between these mechanical elements and chemical signaling processes is not known. Here, we show that α5β1 integrin switches between relaxed and tensioned states in response to myosin II–generated cytoskeletal force. Force combines with extracellular matrix stiffness to generate tension that triggers the integrin switch. This switch directly controls the α5β1-fibronectin bond strength through engaging the synergy site in fibronectin and is required to generate signals through phosphorylation of focal adhesion kinase. In the context of tissues, this integrin switch connects cytoskeleton and extracellular matrix mechanics to adhesion-dependent motility and signaling pathways.

Integrins are expressed on the surface of most cells in the body, and their primary function is to maintain a dynamic adhesion between the cell and its microenvironment (1). Integrin-mediated cell adhesion controls critical intracellular signals that regulate cell differentiation, proliferation, and survival (24). Recent analyses have found that tissue stiffness and the ability of cells to generate tension also control the same signaling pathways (5, 6). How the mechanical properties of the cell's microenvironment are transmitted to intracellular biochemical pathways is not known. However, several intracellular signaling molecules, including src, Cas, and vinculin, show tension-dependent conformational changes that affect kinase activity, availability of phosphorylation sites, or intracellular localization (79). Binding proteins to the extracellular and cytoplasmic domains of integrins forges a link between the extracellular matrix and the actin cytoskeleton that is tensioned by myosin II motors acting on actin filaments (7). This complex contains the major mechanical elements that have been associated with the mechano-sensing capabilities of the cell (10). Integrins are necessary elements for most mechano-sensing models and lie at the beginning of the sensing pathway. Here, we investigate the mechano-responsive properties of α5β1 integrin bound to surface-attached fibronectin.

HT1080 cells plated on a stiff fibronectin-coated substrate attach through α5β1 integrin, and these adhesive bonds are then tensioned by actin and myosin II (11). To analyze the role of tension for α5β1-fibronectin adhesive bonds, we used pharmacological inhibitors of myosin II function and activation (12). The effects of these inhibitors on adhesive α5β1-fibronectin bonds were measured by chemical cross-linking. In previous analyses, the fraction of total α5β1 integrin that cross-linked was proportional to the number of adhesive bonds (11). This assay demonstrated a dose-dependent decrease in cross-linked α5β1 integrin for cells treated with the myosin light-chain kinase inhibitor, ML-7 (Fig. 1A), and the myosin II inhibitors 2,3-butanedione monoxime (BDM) and blebbistatin (fig. S1, A to C) [supporting online text (SOM text), note S1]. Similar decreases were observed with the actin assembly inhibitors cytochalasin D and latrunculin A (fig. S1, D and E). Thus, blocking the ability of the cell to apply tension reduced the number of cross-linkable α5β1-fibronectin bonds. In contrast, these inhibitors had no effect on the chemical binding of soluble fibronectin to α5β1 integrin on cells in suspension (fig. S2, A and B). Another approach to measuring adhesive bonds is to measure the force required to detach the cell. We developed a spinning disc device for this purpose and demonstrated that the cell detachment force was proportional to the number of adhesive bonds (11, 13, 14). This measure of adhesive bonds was not affected by inhibiting actin-myosin–induced cell tension (Fig. 1B). Hence, we hypothesized that there were two distinct forms of the α5β1-fibronectin adhesive bonds: one that formed in the absence of cell tension (relaxed bonds) that could not be cross-linked and one that formed in the presence of cell tension (tensioned bonds) that could be cross-linked. Tension involves holding one end of the bond and pulling on the other. This can have two outcomes; either the bond separates or the bond switches to a new, stronger state that resists more tension.

Fig. 1.

Analysis of adhesive α5β1-fibronectin bonds by chemical cross-linking before and after application of external force. (A) Proportion of cross-linked α5 and β1 integrin in ML-7–treated cells. (B) Tension-inhibited cells were analyzed by using the spinning disc: [ML-7] = 15 μM; [BDM] = 10 mM. (C) Distribution of cross-linked α5β1 integrin 60 min after plating, before and after spinning ±10 mM BDM. Nonspecific aggregates of secondary antibody provide an internal focus or exposure control. Scale bar, 25 μm. (D) Prespin represents tensioned bonds, and post spin represents total bonds; the difference between the two represents relaxed bonds. Data show percentage of total α5 and β1 cross-linked as a function of time. (E) BDM-treated cells, the same as (D). Means ± SEM; n = 3.

To determine whether a state shift occurred when tension was applied, cells were allowed to adhere for 60 min (prespin) or incubated for 60 min and spun 5 min, which would apply tension to adhesive bonds by pulling on the cell (post spin) (Fig. 1C). The control showed considerable cross-linked α5β1 integrin prespin and only a slight increase post spin, particularly toward the centers of the cells. The BDM-inhibited cells showed no detectable cross-linking before spinning, in agreement with the earlier result, but post spin, the levels of cross-linking were similar to those of the control. Thus, external force could substitute for intracellular force. The resulting tension switched the relaxed (non–cross-linkable) α5β1-fibronectin bonds to the tensioned (cross-linkable) state. To expand this result, a quantitative analysis of cross-linkable α5β1 was performed at 15, 30, and 60 min after plating, both before and after a 5-min spin (Fig. 1, D and E, and fig. S2C). The prespin values represent tensioned bonds; the spinning would convert all bonds to tensioned bonds, and hence, the post spin values represent total bonds; and the difference represents the relaxed bonds. Control cells showed a progressive increase in the level of tensioned bonds with time. In the BDM-treated cells, there was very little increase in tensioned bonds, although the total number of bonds increased with similar kinetics and extent as the control. Thus, in the absence of tension, adhesive α5β1-fibronectin bonds accumulated at a rate similar to that in the presence of tension, but the bonds remained in a relaxed state. Because it is necessary to generate the linkage from substrate through α5β1-fibronectin bonds to actin and myosin before tension can be exerted, the relaxed bond is a necessary, logical precursor to the tensioned bond. It is unlikely that the increase in bonds observed after spinning could be caused by the spinning itself because: (i) the extent of the increase depended on the time of the prespin incubation and not on the spinning time, and (ii) α5β1-fibronectin bonds accumulate relatively slowly because of issues of diffusion and the requirements for complex formation; 5 min is not sufficient time for the observed increases (fig. S2D and Fig. 1C).

If tension converts α5β1 integrin–fibronectin bonds from the relaxed to the tensioned state rather than causing the bond to dissociate, the intermolecular contacts that form the two states must be distinct (15). The relaxed state can be approximated by the chemical binding between fibronectin [the fragment containing the 7th to 10th type II repeats (Fn7-10), including both the Arg-Gly-Asp (RGD) and synergy sites (16) (Fig. 2A)] and the purified extracellular domain of α5β1 integrin, because these reactions were done with purified components, the bonds could not be tensioned (17). A combination of electron microscopy and surface plasmon resonance analyses demonstrated that the binding interface includes the RGD but not the synergy site. To determine the intermolecular contacts for tensioned bonds, the spinning disc was used to analyze fibronectin mutants. Bonds were allowed to form for 60 min to approximate steady-state adhesion (fig. S2D). When RGD was deleted, the relaxed bonds could not form and no cells attached. Individual synergy site mutations in which Ala replaced Arg at residue 1374 and 1379 (R1374A and R1379A, respectively) produced weaker bonds that required less force to rupture, and this effect was roughly additive in the double R1374/1379A mutant (Fig. 2B). Thus, the relaxed bonds involved only the RGD site, but the tensioned bonds required also R1374 and R1379 in the synergy site. The reduction of ∼90% in adhesion strength for the double R1374/1379A mutant implies that most of the bond strength depended on synergy-site contacts. Both molecular docking and the presence of epitopes for inhibitory monoclonal antibodies suggest that the synergy site would contact the β-propeller in α5 (18). The differential cross-linking observed for the relaxed and tensioned states can be explained by the resulting change in intermolecular distance.

Fig. 2.

Tensioned α5β1-fibronectin bonds engage the synergy site of fibronectin. (A) Crystal structure of fibronectin type III repeats 9 and 10. α5β1 integrin binds from the top overlapping Fn9 and Fn10 surfaces. The RGD extended loop in Fn10 acts as the recognition sequence. Synergy-site R1374 and R1379 (white) binds near the edge of the β-propeller of α5. Charged residues (black). (B) Spinning disc analysis of mutant Fn7-10. Means ± SEM; n = 3.

The generation of tension involves pulling against a resistance, which is provided by the substrate to which the fibronectin is attached. Up to this point, the experiments have used very stiff substrates that provide strong resistance to forces applied to the cell or cytoskeleton. On these substrates, most α5β1 integrin–fibronectin bonds were tensioned, but softer substrates in the range of that found in tissues (0.2 to 20 kPa) provide less resistance, and the proportion of adhesive bonds in the relaxed state should increase. Fibronectin was attached to polyacrylamide gel substrates with different stiffness and α5β1 integrin–fibronectin bonds were measured by using both spinning disc and chemical cross-linking methods (Fig. 3, A and B). Note that corrections were made for cell spreading and shape differences (see fig. S3, A and B, and SOM text, note S2). Spinning disc analysis showed that the total number of adhesive bonds that formed was independent of substrate stiffness. In contrast, the cross-linking assay showed a strong dependence of the number of tensioned bonds on substrate stiffness. Thus, as the substrate stiffness increased, the proportion of α5β1-fibronectin adhesive bonds in the tensioned state increased. If only tensioned bonds generated downstream signals, this would provide a mechanism for cells to sense microenvironmental stiffness. Inhibitors of myosin II and actin that blocked the conversion of relaxed to tensioned bonds exhibited a parallel dose-dependent reduction in phosphorylation of the focal adhesion kinase FAK on Y397 (Fig. 3C and fig. S4, A and B). The R1379A mutant that had the largest effect on synergy-site engagement was sufficient to block FAK Y397 phosphorylation, but the weaker R1374A mutant was below the tension threshold (Fig. 3D). Finally, FAK Y397 exhibited a stiffness-dependent increase in phosphorylation, whereas FAK Y861, which is tethering-independent (11), showed no stiffness-dependent change in phosphorylation in the same cells (Fig. 3E).

Fig. 3.

Substrate stiffness controls the mechanical activation of α5β1 integrin and signaling to FAK. (A) Mean cell detachment force per α5β1 integrin in the cell-gel interface. See fig. S2 calculations. (B) Proportion of β1 integrin cross-linked to fibronectin-polyacrylamide gels 4 hours after plating. Background was subtracted and normalized to 400 Pa gels and for proportion of α5β1 integrin in the interface. (C) Effect of ML-7 on FAK Y397 phosphorylation. Sus denotes cells in suspension; data were normalized to untreated controls. (D) Effect of Fn7-10 mutants on FAK Y397 phosphorylation normalized to wild type. T, Thr. (E) Effect of gel stiffness on FAK Y397 (◯) and Y861 (▲) phosphorylation. Means ± SEM; n = 3.

α5β1 integrin functions both to adhere cells to a fibronectin substrate and to mediate downstream signals that control cell physiology and cell fates (1, 3). Previous analyses have associated those functions with integrin activation and clustering (19, 20). The current models for integrin activation are based on analyses in which tension was not applied (17, 21). This activation switches the integrin into a state that can bind ligand and is necessary for the clustering of integrins and the formation of focal complexes (22). Although these processes were required to form adhesion complexes, they were insufficient to generate either strong adhesion or downstream signals (fig. S5). The initial α5β1-fibronectin bond is equivalent to the previously described activated-bound state (17, 19). Application of force switches the relaxed state to a new tensioned state, which has increased bond strength. Bonds that undergo tension-strengthening have been called “catch-bonds” and have been previously reported for leukocyte selectin and Escherichia coli FimH adhesion receptors (23, 24). A catch-bond can function as a molecular clutch that is engaged under tension and that will release when tension is released. Thus, α5β1-fibronectin bonds provide mechanical clutches for cell migration (25). For α5β1 integrin, the catch-bond mechanism depended on the synergy-site engagement, which raises the question of whether synergy sites exist more broadly in integrin ligands.

Mechano-responsive proteins undergo functional conformational transitions, in response to tension, that alter binding properties and/or enzymatic functions (8, 10). Intracellular proteins, like src, that have this property require cytoskeletal structure and integrin-mediated adhesion for mechanical activation (9). Thus, the mechanical activation of src would depend on the mechanical activation of integrins, which suggests a model that contains multiple intra- and extracellular mechano-sensitive nodes to generate intracellular signaling responses. Control of intracellular signals by mechanical triggers provides a mechanism for spatial distribution of signals within cells. This is important for the cells' capability to sense and respond to differences in microenvironmental stiffness and generates a need for mechanisms to control the cell's “tensional” homeostasis (5).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5


References and Notes

View Abstract

Navigate This Article