Vinculin forms a directionally asymmetric catch bond with F-actin

See allHide authors and affiliations

Science  18 Aug 2017:
Vol. 357, Issue 6352, pp. 703-706
DOI: 10.1126/science.aan2556

Making the right catch

Tension reveals cryptic vinculin-binding sites on α-catenin and talin at cadherin-based cell-cell and integrin-based cell-matrix adhesions, respectively. The enrichment of vinculin at cellular adhesions is thus an indicator of load-induced reinforcement of the cytoskeletal linkage. Huang et al. used a single-molecule optical trap assay to measure the binding lifetimes of vinculin to single actin filaments under load. The vinculin-F-actin interaction formed a directional catch bond—one that is very weak at low force but that greatly increases in lifetime with increasing force. This explains vinculin's role as a reinforcing linker at both cell-cell and cell-matrix adhesions.

Science, this issue p. 703


Vinculin is an actin-binding protein thought to reinforce cell-cell and cell-matrix adhesions. However, how mechanical load affects the vinculin–F-actin bond is unclear. Using a single-molecule optical trap assay, we found that vinculin forms a force-dependent catch bond with F-actin through its tail domain, but with lifetimes that depend strongly on the direction of the applied force. Force toward the pointed (–) end of the actin filament resulted in a bond that was maximally stable at 8 piconewtons, with a mean lifetime (12 seconds) 10 times as long as the mean lifetime when force was applied toward the barbed (+) end. A computational model of lamellipodial actin dynamics suggests that the directionality of the vinculin–F-actin bond could establish long-range order in the actin cytoskeleton. The directional and force-stabilized binding of vinculin to F-actin may be a mechanism by which adhesion complexes maintain front-rear asymmetry in migrating cells.

Cadherin- and integrin-based protein assemblies link cells to each other and to the extracellular matrix (ECM), respectively, and together provide the physical basis for the organization of multicellular tissues (1). Both classes of adhesion complexes are highly sensitive to mechanical load and change rapidly in size and composition to maintain the physical integrity of living tissues (2). These adhesions are also essential in defining the physical asymmetries that underlie both individual and collective cell migration in the context of embryonic development (3), wound healing (4), and cancer metastasis (5). However, the molecular basis of how cadherin- and integrin-based adhesions respond to mechanical stimuli to regulate tissue cohesion and directional cell migration has thus far remained poorly understood.

The protein vinculin is a component of both cadherin- and integrin-based adhesion complexes and is rapidly recruited to both types of adhesions in response to mechanical load through its interactions with α-catenin and talin, respectively (68). Vinculin plays a key role in maintaining tissue integrity (9, 10); for example, loss of vinculin in mice results in the death of the developing embryo owing to defects in neural tube closure and heart development (11). Importantly, vinculin is required for persistent directional cell migration, suggestive of a role in generating a polarized connection between adhesions and the actin cytoskeleton (12, 13). Although vinculin is also recruited to cadherin-based adhesions in a force-dependent manner (6, 14), comparatively little is known about how it might regulate actin organization and dynamics at those sites.

Vinculin binds directly to filamentous actin (F-actin) through its actin-binding tail domain (Vt) (15, 16), but how and whether this bond may be regulated by mechanical load is not known. Defining this mechanism is critical to understanding the role of vinculin as a reinforcing link between adhesion complexes and the actin cytoskeleton. We modified a previously developed optical trap (OT)–based assay (17) to define the load dependence of the binding interaction between vinculin and F-actin (Fig. 1A). Actin binding to full-length, wild-type vinculin is blocked by autoinhibitory intramolecular interactions between the N-terminal head domain and Vt (15). We thus used a well-characterized vinculin construct, termed T12 vinculin, bearing mutations that disrupt the autoinhibitory head-Vt interaction (18). In the OT assay, the displacement of the microsphere (1 μm in diameter) from the trap center typically decreased in several distinct steps (Fig. 1, A and B), indicative of the sequential release of multiple vinculin molecules from the actin filament (17). The last stair step in this series likely reflects a state in which actin is bound by a single vinculin molecule (19), allowing us to determine the effect of mechanical load on the kinetics of vinculin unbinding.

Fig. 1 The optical trap (OT) assay measures vinculin–F-actin bond lifetimes under load.

(A) (Top) An actin filament attached to two microspheres is held taut by two OTs near a platform bead with vinculin on its surface. A motorized stage moves the platform back and forth. Vinculin binding results in the displacement of one of the optically trapped microspheres. F, force. (Bottom) Both OTs exert force on the actin filament. Here we plot the summed force [the total force transmitted from both traps to the vinculin molecule(s) (19)] versus time, decimated from 40 to 4 kHz (light blue) and median-filtered with a 100-point moving window (overlaid in black). If force surpasses a defined threshold, stage motion halts until detachment of the bound vinculin molecule(s). (B) Representative traces produced when force on T12 vinculin is directed toward the pointed (–) end (blue) or the barbed (+) end (green) of the actin filament. (C) Binding events in which the force on T12 vinculin was directed toward the filament’s pointed (–) end (blue circles; n = 102; mean lifetime of 5.6 s) or barbed (+) end (green circles; n = 65; mean lifetime of 0.54 s).

The duration of vinculin-actin binding events depended on the level of mechanical load: Events were short (<1 s) at low loads (0 to 3 pN), markedly longer (~10 s) at intermediate loads (7 to 10 pN), and again short (<1 s) at loads >15 pN (Fig. 1C). This counterintuitive behavior, in which increasing load results in an increase in bond lifetime over a given force range, is termed a catch bond and differs from the more usual behavior of molecular complexes, in which increasing load leads to an exponential increase in the dissociation rate (termed a slip bond) (20, 21). Owing to the nature of the loaded OT assay, few binding events were observed at loads less than 2 pN, where catch and slip bond models differ greatly in their predicted lifetimes. To measure lifetimes in this force range, we developed an alternative assay in which an actin filament was positioned near a platform oscillating at 200 Hz over 5 nm. Binding of vinculin to the actin filament was detected by a decrease in the amplitude of microsphere fluctuations (fig. S1). This alternative assay could not resolve sequential unbinding steps, which means that the mean binding lifetimes measured in this way may be overestimated. Nonetheless, the binding events observed near zero force were markedly shorter than those measured at loads higher than 2 pN [P < 0.01 in a one-tailed two-sample Kolmogorov-Smirnov (KS) test], consistent with the idea that mechanical load increases the lifetime of the vinculin-actin bond.

A vinculin binding event can displace a trapped actin filament in one of two directions, depending on which direction the platform is moving when binding occurs (Fig. 1B). We noted that mechanical load applied in one direction resulted in longer mean bond lifetimes than when load was applied in the opposite direction. This result is unlikely to have been caused by a systematic error in the OT assay, because the direction yielding longer lifetimes switched when we physically rotated the filament by 180°.

F-actin is polar, with a barbed (+) and pointed (–) end, so we tested whether the observed directionality could arise from the asymmetric binding interface between vinculin and the actin filament. To determine the polarity of a trapped actin filament, we used platforms that were coated with myosin VI (fig. S2). Myosin VI moves in a processive manner toward the pointed (–) end of the actin filament (22), so the displacement of one of the microspheres from its OT allowed us to assign the polarity of the trapped actin filament (fig. S3). The same filament, now with known polarity, was moved to a vinculin-coated platform and tested as before. Over the range of loads tested (1 to 30 pN), the vinculin–actin filament bond exhibited a mean lifetime that was ~10 times as long when the force experienced by vinculin was directed toward the pointed (–) as opposed to the barbed (+) end of the actin filament (Fig. 1C and fig. S3).

To explain the observed distribution of bond lifetimes, we tested several kinetic models that have been used to describe how bond lifetimes depend on mechanical load (19). Catch bond models that feature a single bound state (23) predict a single exponential distribution of lifetimes at any given force and thus do not accurately describe the mixture of short- and long-lived events that we observed (fig. S4). We therefore chose a two-bound-state model (24, 25) in which the bond can transition between unbound (0), weakly bound (1), and strongly bound (2) states with load-dependent rates (Fig. 2A, inset) given by a modified Bell modelEmbedded Imagewhere the transition rate kij(F) between two states depends on xij, which reflects both the physical distance between the initial state i and the peak of the energy barrier (the transition state) between states i and j, as well as the extent to which the applied force F is aligned with this direction (kB, Boltzmann constant; T, temperature) (fig. S5A). Using maximum likelihood estimation, we found that this two-bound-state model best fit the data for displacements in both the pointed (–) and barbed (+) end directions (Fig. 2). To obtain accurate parameter estimates, we augmented the force-lifetime data set in which actin filament polarity was explicitly determined (Fig. 1 and fig. S3B) with a data set in which the actin polarity was inferred from a statistical test (fig. S3, C and D, and supplemental text). In the model fit, vinculin bound in the weak-binding state 90% of the time and equilibrated with the strong-binding state. Importantly, mechanical load lowered the rate at which the strong-binding state transitioned back to the weak-binding state, thereby stabilizing the vinculin–actin filament bond (fig. S6). The direction of applied force determined the extent to which it stabilized the strongly bound state, which is encoded in the distance parameters xij (Table 1). Thus, the two-bound-state catch bond model provides a kinetic basis for the directionality of vinculin–actin filament binding.

Fig. 2 A two-bound-state directional catch bond model of T12 vinculin and vinculin tail (Vt) binding lifetimes.

(A) Mean actin binding lifetimes and model of best fit for T12 vinculin (n = 728) and (B) Vt (n = 702). A subset of these events is collected from filaments of known polarity, and the rest are assigned a polarity using a 2D KS test (19). In each 2-pN bin, the area of each blue circle is proportional to the number of events. Red curves show mean lifetimes predicted by the two-bound-state catch bond model (inset), and purple envelopes indicate 95% confidence intervals (CIs) for the fit, obtained by parametric bootstrapping. The model was constrained to possess a mean lifetime at zero force that was less than or equal to the lifetime measured using the low-force binding assay (fig. S1). Survival plots of the data and models are shown in fig. S8 for T12 vinculin and fig. S9 for Vt.

Fig. 3 Computational model of actin dynamics in a cell protrusion.

(A) Long-range order of the actin cytoskeleton (orange) induced by adhesion proteins, including vinculin, may reinforce polarized cell migration. (B) In our computational model, actin filaments (orange) undergo 2D Brownian diffusion with a drift velocity. The effect of vinculin on filament motion is modeled as a simple spring that dampens diffusion and resists the drift velocity (19). Directionality in the model depends on the angle θ between the filament’s long axis (u) and the force vector (F). (C) Sample trace of the 2D position of a filament’s center of mass over time. Blue dots indicate times at which an additional vinculin binds. (D) Filaments start at x = 0 μm with a random initial angle. Red squares show the fraction of filaments with their barbed (+) end facing forward after 60 s. Gray circles show a control simulation in which vinculin’s actin-binding kinetics do not depend on actin polarity. Filaments oriented with their barbed (+) end to the right are defined as forward-facing (angles –90° to 90°). The top-left inset shows angles of filaments within x = –12 ± 1 μm, and the bottom-right inset shows angles of filaments within x = –1 ± 1 μm. For each condition, n = 224,000 filaments.

We next examined which domain(s) of vinculin were required for the formation of a directional catch bond to F-actin. We tested a protein construct containing only the Vt (residues 879 to 1066), which binds to F-actin. Similarly to the case of T12 vinculin, the Vt bond to F-actin was stabilized by mechanical load in either direction, and load directed toward the pointed (–) end resulted in higher stability (fig. S3, B and D). Thus, Vt alone is sufficient to form a directional catch bond with the actin filament (Fig. 2B). The less pronounced directionality in Vt binding was reflected in differences in the distance parameter x21 for the strong-to-weak transition (Table 1). This parameter is sensitive to the angle of applied load, and the presence of the vinculin head might change the direction of force experienced by the Vt (fig. S5). Alternatively, the vinculin head may contain cryptic actin-binding site(s) that are revealed upon application of load. In any case, formation of a directionally asymmetric catch bond appears to be intrinsic to the interaction of the Vt with the actin filament.

Table 1 Kinetic parameters for the two-bound-state catch bond model for T12 vinculin (top) and Vt (bottom).

State 0 is the unbound state. States 1 and 2 are the weak and strong binding states, respectively. Each parameter is assigned a 95% confidence interval (CI; in parentheses) obtained through parametric bootstrapping (17, 19). The transition rate from state i to state j in the absence of load is indicated by kij0. The application of load alters the equilibrium between state i and state j according to a distance parameter xij, which can be positive or negative, indicating that the transition rate is increased or decreased by force, respectively. These distance parameters vary depending on whether force is oriented toward the barbed (+) end or pointed (–) end of the actin filament.

View this table:

In migrating cells, actin filaments near the leading edge are almost all oriented with their barbed (+) ends facing the membrane, whereas filament orientation becomes increasingly isotropic toward the cell center (2629). We used computational modeling to investigate whether the directional interaction between vinculin and actin could potentially generate long-range order in the actin cytoskeleton (Fig. 3A). In our model, actin filaments underwent two-dimensional (2D) Brownian diffusion with a drift velocity that represented retrograde actin flow in a migrating cell (Fig. 3B). Vinculin binding was modeled as the attachment of a simple spring that resisted retrograde actin filament motion. The unbinding kinetics of the vinculin-actin bond were those obtained from the two-bound-state model of the OT data and thus captured both directionality and load-induced stabilization. An ensemble of randomly oriented actin filaments was subjected to a drift velocity, thermal fluctuations, and stochastic vinculin binding (Fig. 3C). As the simulation progressed, these conditions resulted in the development of orientational asymmetry, with the majority of filaments closer to the starting position (x = 0 μm) oriented with their barbed (+) ends facing the leading edge (Fig. 3D). Thus, directional asymmetry of the vinculin-actin bond can potentially establish a spatial asymmetry in F-actin polarity (fig. S7), an effect that may underlie the prior observation that vinculin is essential for generating persistent, directional cell migration (12, 13, 30). The directionally asymmetric linkage of vinculin to actin may complement other mechanisms by which cells generate persistent migration, such as the signaling cascade that activates the barbed (+) end nucleators Arp2/3 (31) and formins (32), which reinforce cell polarity by promoting actin polymerization at the cell’s leading edge.

We have shown that vinculin forms a directionally asymmetric and force-stabilized linkage to F-actin. Directional asymmetry of this sort is not unknown; the binding of kinesin with microtubules is destabilized by load, characteristic of a slip bond, but to a different extent depending on loading direction (33), and the binding of the myosin V head with actin, and dynein with microtubules, is asymmetrically stabilized by load, which may promote motor processivity (34, 35). The directional interaction between vinculin and actin may play a critical role in tissue patterning. Consistent with this idea, vinculin-null mouse embryos exhibit phenotypes suggestive of failures not only in cell adhesion, but also in tissue patterning driven by cell migration, notably axon outgrowth (11). Vinculin-deficient cells likewise exhibit defects in the formation of stable lamellipodia and filopodia (36) and are more motile (37) but less directionally persistent (13, 30). Thus, vinculin does not appear to be required for adhesion or motility per se, but instead for stable protrusion and polarized migration. The observation that vinculin forms a catch bond with actin in both directions of applied load (Fig. 2) likely reflects its biological function as a reinforcing linker at adhesions, including those in which actin may not be polarized. A comparison between actin binding by vinculin (Fig. 2A) and the E-cadherin/β-catenin/αE-catenin (cadherin-catenin) complex (17)—which vinculin is believed to mechanically reinforce—suggests that under barbed (+) end–directed load, vinculin binds with stability comparable to that of the cadherin-catenin complex, but that under pointed (–) end–directed load, vinculin stability is ~10 times as high. Within the limits of our assay, the binding of the cadherin-catenin complex to actin did not exhibit a directional response to force. We suggest that the asymmetric catch bond behavior of vinculin, in addition to enhancing overall resistance to mechanical load, could contribute to organizing the polarity of the actin cytoskeleton in response to internally and externally applied loads at cell-cell contacts.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Tables S1 to S4

References (3876)

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: The authors thank T. Omabegho and P.V. Ruijgrok of the Bryant Lab (Stanford University) for providing the myosin VI protein construct and W. J. Nelson for advice and discussion. The data reported in this paper are further detailed in the supplementary materials. Research reported in this publication was supported by grant R01GM11462 from the National Institutes of Health (NIH). The research of A.R.D. was supported in part by a Faculty Scholar grant from the Howard Hughes Medical Institute. D.L.H. and N.A.B. were supported by training grant T32 GM007276 from the NIH. D.L.H. was supported by a Graduate Research Fellowship from the National Science Foundation. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
View Abstract

Navigate This Article