Research Article

The minimal cadherin-catenin complex binds to actin filaments under force

See allHide authors and affiliations

Science  31 Oct 2014:
Vol. 346, Issue 6209, 1254211
DOI: 10.1126/science.1254211

Structured Abstract


Cadherins are an ancient class of transmembrane proteins that are essential for the formation of multicellular tissues in metazoans. Cadherins link intercellular adhesions to the cellular cytoskeleton, but how they are connected specifically to actin filaments is a hotly debated issue. Genetic and cell culture experiments indicate that E-cadherin, β-catenin, and the actin filament binding protein αE-catenin form a minimal cadherin-catenin complex that binds to the actin cytoskeleton directly in epithelial tissues. However, experiments with purified proteins showed that a stable cadherin-catenin complex can be reconstituted, but it does not bind strongly to actin filaments in solution. Nevertheless, cell culture experiments indicated that the cadherin-catenin complex is under constitutive actomyosin-generated tension and that this connection is required for mechanotransduction at cadherin-based adhesions. Here, we tested the hypothesis that tension is required to stabilize a linkage between the cadherin-catenin complex and actin filaments, and clarify how the cadherin-catenin complex could interact directly with the actin cytoskeleton in cells.

Embedded Image

Two-state catch bond model of cadherin-catenin/F-actin interactions. Force stabilizes the cadherin-catenin/F-actin bond by shifting it from a weakly to a strongly bound state. The force dependence of the connection between the cadherin-catenin complex and actin f laments may explain the mechanosensitivity of cadherin-mediated intercellular adhesions.


We developed an optical trap–based assay to measure the lifetime of cadherin-catenin complex/actin filament bonds under tension. An actin filament was attached to two optically trapped beads and suspended above purified cadherin-catenin complexes immobilized on a glass coverslip surface that was precoated with glass microspheres. The coverslip was mounted on a motorized stage. This spatial arrangement was informed by electron tomography of cell-cell junctions, which showed actin filaments parallel to the plasma membrane. Tension was applied to cadherin-catenin complex/actin bonds by moving the sample stage back and forth parallel to the actin filament; if the immobilized cadherin-catenin complexes bound the actin filament, the attached beads were displaced from the optical trap. The lifetime of the bond was measured from the time series of the force exerted on the trapped beads. Kinetic models were fit to bond lifetime distributions with respect to applied force.


We observed robust, reproducible cadherin-catenin complex/actin filament binding under force in optical trap–based experiments. Bond lifetime distributions had a biphasic dependence on force. The mean lifetimes increased from ~60 ms at low force to ~1.2 s at ~10 pN, after which they decreased. A two-state catch bond model is consistent with the biphasic mean lifetime distribution and the presence of two distinct lifetime subpopulations. In this model, bonds between a cadherin-catenin complex and an actin filament form in a weakly bound state and quickly dissociate, but rapidly transition to a strongly bound state as applied force increases. Long lifetimes are achieved in this state until higher forces accelerate dissociation from the strongly bound state (see the figure).


Our data and kinetic model reconcile previous in vitro and in vivo work by demonstrating that the cadherin-catenin complex binds robustly to actin filaments under force. Thus, it seems that direct cadherin-catenin complex/actin filament binding was not detected in previous solution-based assays because bonds were not strengthened by tension. The two bound states in our model may correspond to different conformational states of αE-catenin, consistent with previous observations that αE-catenin may undergo changes in conformation in response to actomyosin-generated cytoskeletal tension. Our model of cadherin-catenin complex/ actin filament bond dissociation, combined with previous evidence of cooperative binding of αE-catenin to actin filaments, indicates that the linkage is self-reinforcing and that its stability is dynamically regulated by mechanical force during tissue development and maintenance.

Pulling me apart only makes me stronger

Tension transmitted between neighboring cells can exert profound effects on cell proliferation, differentiation, and tissue organization. Exactly how intercellular mechanical tension is sensed at the molecular level is unknown. One attractive hypothesis is that a linkage between the cell-cell adhesion molecule E-cadherin, its binding partners α- and β-catenin, and actin filaments may act as a tension sensor. However, how this linkage is established at the molecular level is not known. Buckley et al. used optical tweezers to determine how mechanical load influences interactions of the cadherin/catenin complex with single actin filaments. The data support a model in which force shifts the interaction from a force-independent, weakly bound state to a highly force-sensitive, strongly bound state. The findings may explain how cells maintain tissue integrity while still being able to move and change shape.

Science, this issue p. 10.1126/science.1254211


Linkage between the adherens junction (AJ) and the actin cytoskeleton is required for tissue development and homeostasis. In vivo findings indicated that the AJ proteins E-cadherin, β-catenin, and the filamentous (F)–actin binding protein αE-catenin form a minimal cadherin-catenin complex that binds directly to F-actin. Biochemical studies challenged this model because the purified cadherin-catenin complex does not bind F-actin in solution. Here, we reconciled this difference. Using an optical trap–based assay, we showed that the minimal cadherin-catenin complex formed stable bonds with an actin filament under force. Bond dissociation kinetics can be explained by a catch-bond model in which force shifts the bond from a weakly to a strongly bound state. These results may explain how the cadherin-catenin complex transduces mechanical forces at cell-cell junctions.

Epithelia serve as barriers between the organism and its environment. A defining feature of these tissues is adhesion between cells at specialized intercellular junctions. The mechanical connection at these junctions imparts shape, organization, and structural integrity to the tissue and enables morphogenetic changes such as the movement of epithelial sheets and the formation of tubes during development (1, 2). Dysregulation of cell-cell junctions is common in cancer metastasis, which is characterized by loss of contact inhibition, epithelial-to-mesenchymal transitions, and abnormal cell invasiveness (3).

Classical cadherins and their cytoplasmic binding partners play a central role in intercellular adhesion in many tissues (4). In epithelial tissues, the extracellular domain of cadherin forms adhesive contacts between neighboring cells, and its cytoplasmic domain binds β-catenin, which in turn binds the F-actin binding protein αE-catenin (5), the most widely expressed of the three α-catenin family members (6). αE-catenin binds strongly to the E-cadherin/β-catenin complex [dissociation constant (Kd) ~1 nM] (7, 8) but more weakly to F-actin (Kd ~1 μM) (911). Cell biological studies led to the hypothesis that αE-catenin directly links the E-cadherin/β-catenin complex to F-actin, consistent with its role in force transmission between cadherins and the actin cytoskeleton (12). However, in vitro binding of αE-catenin to the cadherin cytoplasmic domain/β-catenin complex further weakens the affinity of αE-catenin for F-actin by at least a factor of 20, to a level that would not be useful for transmitting force between E-cadherin and the actin cytoskeleton (8, 10, 13).

These biochemical studies, performed with proteins from Mus musculus (Mm; mouse) and Danio rerio (Dr; zebrafish), cast doubt on the simple model that αE-catenin directly links the cadherin/catenin complex to the actin cytoskeleton. Other proteins, including vinculin (1418), epithelial protein lost in neoplasm (EPLIN) (19), α-actinin (20), and afadin (21, 22), bind to both αE-catenin and F-actin and could be a link between the cadherin-catenin complex and F-actin. Notably, force exerted on the cadherin-catenin complex appears to recruit vinculin to cell junctions (15, 2326). However, it is unclear how the changes in αE-catenin conformation required for vinculin binding (14, 27) can be induced by force if αE-catenin does not link the E-cadherin/β-catenin complex to F-actin.

Given the importance of actomyosin-generated tension in cell-cell adhesion (12, 15, 25), we posited that tension stabilizes a direct link between the minimal cadherin-catenin complex and F-actin. Therefore, we developed a single-molecule, optical trap–based assay that replicated the geometry of the adherens junction (AJ) and mechanical forces between cadherin-catenin complexes and actin filaments. We found that the application of physiological, pN-level forces increased the lifetime of normally transient bonds between cadherin-catenin complexes and an actin filament. This behavior is indicative of a catch bond, in which the dissociation rate decreases with applied force (28, 29), rather than the more typical slip bond in which the dissociation rate increases exponentially with increasing applied force. We show that a two-state catch-bond model (30) fits the distribution of lifetimes of cadherin-catenin complex/F-actin bonds under force. In this model, the cadherin-catenin complex and an actin filament interact in a short-lived, weakly bound state under low forces and transition to a stable, strongly bound state at higher forces. Thus, our data reveal that the cadherin-catenin complex is a force-sensitive, direct linker to the actin cytoskeleton, and our model offers a kinetic basis for understanding mechanotransduction at AJs.

Experimental approach

To replicate in vitro the spatial organization of the cadherin-catenin complex and F-actin, we examined electron tomographic reconstructions of cell-cell contacts in Caco-2 cells (31), which are derived from human intestinal epithelia. These images showed dense arrays of actin filaments parallel to the plasma membrane (Fig. 1, A and B, and fig. S1). Although the centerline distance between cell-cell contacts and the F-actin arrays averaged between 0.5 and 1.5 μm, there were regions where the edges of the parallel F-actin arrays were in close proximity to the cell-cell contacts (see the supplementary text), and in other areas a less dense organization of actin filaments was present between the actin arrays and plasma membrane. Cell-cell contacts are dynamic (8, 32), and therefore these images represent only a temporal snapshot of the proximity of junctions and the underlying actin filament bundles. Parallel actin arrays tended to appear at cell-cell contacts tens of nm above the extracellular matrix (ECM) interface, whereas nonparallel actin networks were spatially correlated with the basal membrane near the ECM contact. These observations are consistent with previous super-resolution microscopy data, which revealed actin filaments parallel to intercellular junctions in simple epithelia (9).

Fig. 1 Electron microscopy of cell-cell junctions and optical trap–based assay setup.

(A) Transmission electron microscopy (TEM) image of a cell-cell junction between Caco-2 epithelial cells cultured on EM-amenable substrates. Scale bar, 1 μm. Brackets label actin filament arrays parallel to the junction, and the yellow arrow marks where the actin arrays were in close proximity to cell-cell contacts. (B) Three-dimensional electron tomography reconstruction of the same region shown in (A) rotated 90° clockwise and then tilted 45° around the horizontal axis. The cell-cell junction is highlighted in red. Yellow arrow marks the same region as in (A). Scale bar, 1 μm. (C and D) Fluorescence time-lapse micrographs of tetramethylrhodamine (TMR)–labeled F-actin (white lines) at the surface of a coverslip coated with either 1 μM Mm cadherin-catenin complex (C) or 2 μM Mm αE-catenin homodimer (D). Insets in solid yellow lines show the corresponding images for regions bounded in the dashed yellow lines after 30 s had elapsed; actin filaments remained stably bound to coverslips coated with Mm αE-catenin homodimer, whereas actin filaments diffused away from coverslips coated with the Mm cadherin-catenin complex within seconds. Out-of-focus features are glass microsphere platforms. Scale bars, 10 μm. (E) Illustration of a cadherin-catenin complex and actin filament reconstituted in the optical trap assay (not to scale). GFP–E-cadherin cytoplasmic tail (green), β-catenin (yellow) and αE-catenin (blue) are immobilized on a coverslip. Glass microspheres (1.5-μm diameter) on the coverslip act as platforms such that cadherin-catenin complexes can contact the actin filament. A single biotinylated, TMR-phalloidin-coated actin filament (red) extends between two 1-μm diameter streptavidin-coated beads held in optical traps (pink). (F) Top view of the assay in bright field. Beads attached to the actin filament (not visible) are held in traps 1 and 2, and the platform bead is positioned between the traps and below the filament. Scale bar, 2 μm.

Without applied tension, cadherin-catenin complex/F-actin bonds are much weaker than αE-catenin homodimer/F-actin bonds (8, 10, 13). This difference was apparent in time-lapse movies of tetramethylrhodamine-labeled actin filaments diffusing in solution near the surface of coverslips coated with either 1 μM Mm green fluorescent protein (GFP)–E-cadherin/β-catenin/αE-catenin or 2 μM Mm GFP–αE-catenin homodimer (Kd for F-actin is ~1 μM) (Fig. 1, C and D). Actin filaments bound stably to Mm GFP–αE-catenin homodimers for at least 30 seconds, whereas actin filaments did not bind stably to the reconstituted Mm cadherin-catenin complex. Thus, these results replicate previous bulk sedimentation binding assays (13, 19).

To apply tension on transient cadherin-catenin complex/F-actin bonds, we replicated the orientation of actin filaments and cadherin-catenin complexes observed at intercellular junctions in cells (Fig. 1). A single actin filament was optically trapped and extended above Mm or Dr cadherin-catenin complexes immobilized on a coverslip surface precoated with glass microspheres (31) (Fig. 1, E and F). The glass microspheres acted as spacers to prevent the surface of the coverslip from interfering with force measurements. The coverslip was mounted on a stage that moved back and forth parallel to the actin filament. Upon formation of cadherin-catenin complex/F-actin bonds, the motion of the stage was transmitted to the trapped beads, and the displacement of beads within the optical trap caused a restoring force that was applied to the cadherin-catenin complex/F-actin bonds (Fig. 2A). The amount of applied force and the rate of its application were controlled by adjusting the frequency and amplitude of the stage oscillation. Because an optical trap works like a simple spring, the magnitude of this force was calculated from the stiffness of the trap and the displacement of the trapped beads caused by this force (31).

Fig. 2 Mm cadherin-catenin complexes bound actin filaments in oscillating-stage experiments.

(A) The illustrations show that upon binding, the motion of the stage was transmitted to the trapped beads, and the force exerted on them was correlated with the motion of the stage, as shown in the time series in (B). The trapped beads stopped moving along with the stage when the surface-bound cadherin-catenin complex detached from the actin filament. (B) Part of a representative time series of force exerted on one of the two optically trapped beads attached to an actin filament (blue). About 1 μM Mm cadherin-catenin complex was purified and added to the flow cell. The stage was driven by a 150-nm amplitude, 50-Hz frequency sine wave (gray). (C) Full force time series from which the binding events in (B) came (shaded in teal). Peaks in the series are individual binding events, most of which lasted for about one half-period of the sine wave used to drive the stage. (D) Force time series from a negative control experiment in which ~1 μM Mm E-cadherin/β-catenin complex was purified without Mm αE-catenin and added to the flow cell. Oscillation amplitude and frequency of the sine wave used to drive the stage were the same as in (B). (E) Frequency of observed Mm cadherin-catenin complex/F-actin binding as a function of maximum stage speed (amplitude × angular frequency); n = 297 binding events; bin width is 104 nm/s. Event frequency is the number of binding events divided by the total time sampled. (F) Sine histogram of the number of Mm cadherin-catenin complex/F-actin binding events that started in each angle bin (legend shows counts; n = 235 binding events; bin width is 36°; binding events from trap 1 are in blue, and those from trap 2 are in green). All events were from the 4 × 104 nm/s bin shown in (E). The glass microsphere platform (A) was farthest from trap 1 when it was at +1, and from trap 2 when the stage position was at –1. Stage position was normalized by the maximum amplitude of the wave.

The cadherin-catenin complex binds to F-actin under tension

Mm cadherin-catenin complexes, added at a concentration of 1 μM to the flow cell, bound an actin filament when the stage was driven back and forth by sine waves with 150-nm amplitudes and frequencies of up to 150 Hz (6.7-ms period) (Fig. 2B). The stage motion was transmitted to the optically trapped beads whenever cadherin-catenin complexes bound to the suspended actin filament (Fig. 2A). We observed many changes in the force experienced by the trapped beads as a function of time, indicating the formation of robust Mm cadherin-catenin complex/F-actin bonds (Fig. 2, B and C).

Binding of the cadherin-catenin complex to F-actin required αE-catenin (Fig. 2D and supplementary text). The observed unbinding events were most likely caused by the dissociation of the intact cadherin-catenin complex from the suspended actin filament, rather than dissociation of αE-catenin from the E-cadherin/β-catenin heterodimer. In solution (without applied tension), the αE-catenin monomer binds strongly to E-cadherin/β-catenin (Kd = 1 nM) but binds more weakly by a factor of at least 1000 to F-actin (79, 13). At the concentrations in our experiments, any αE-catenin molecules that detached from the surface-bound cadherin-catenin complex would occupy a negligible number of binding sites on the actin filament (31). Thus, it is unlikely that αE-catenin molecules attached to the actin filament could generate reversible binding events with surface-bound E-cadherin/β-catenin heterodimers. In contrast, the actin filament provided a very large number of binding sites for αE-catenin in the platform-bound cadherin-catenin complexes. Moreover, if αE-catenin separated from the E-cadherin/β-catenin complex during every actin filament-binding event, αE-catenin would most likely dissociate quickly from the actin filament and be irreversibly lost (this assumes a reasonably high off-rate, consistent with the weak affinity of monomeric αE-catenin for F-actin). This scenario would rapidly depopulate a platform of active cadherin-catenin complexes, whereas we observe tens to hundreds of binding events per glass microsphere platform (see, for example, Fig. 2C). In any case, the key finding remains that the cadherin-catenin complex bound the actin filament under tension.

Cadherin-catenin complex/F-actin binding was observed more frequently at intermediate stage oscillation speeds (150-nm amplitude, 50-Hz frequency) than at either lower or higher speeds (Fig. 2E). At these intermediate oscillation speeds, the majority of binding events started shortly after the stage had changed direction and was in the slowest part of the sine cycle (Fig. 2F). This indicates that some minimum contact time was necessary to establish Mm cadherin-catenin complex/F-actin bonds before they were subjected to increasing tension. The observation that more events occurred at intermediate rather than low loading rates further indicated that the load stabilized the transient initial bonds. These results motivated us to examine the mechanism by which force might modulate the lifetime of the cadherin-catenin complex bond to F-actin.

Force regulates cadherin-catenin complex dissociation from F-actin

To investigate how force modulated the dissociation kinetics of individual cadherin-catenin complex/F-actin bonds, we modified our optical trap–based assay to observe the detachment of cadherin-catenin complexes from the actin filament under constant force. In these experiments, a signal drove the stage 100 nm back and forth at a constant rate of 1 × 104 nm/s (1.5 × 103 pN/s). Before reversing direction, the stage paused for 150 ms and the forces exerted on the trapped beads were measured. If these forces surpassed a user-defined threshold that indicated binding of cadherin-catenin complexes to the trapped actin filament, then the stage paused until complete detachment of all cadherin-catenin complex/F-actin bonds returned the trapped beads to their zero-force baselines. In these experiments, we used αE-catenin and β-catenin from Dr rather than Mm. Mm αE-catenin forms homodimers (13) whose potential presence during the preparation of the flow cell or during the assay could complicate the interpretation of cadherin-catenin complex/F-actin binding events. Importantly, Dr αE-catenin is a monomer and, like Mm αE-catenin, its affinity for F-actin decreases by a factor of 20 upon binding Dr β-catenin (10).

When we reconstituted cadherin-catenin complexes with 10 nM Dr αE-catenin [in these experiments, Dr β-catenin and E-cadherin were preabsorbed onto the coverslip and glass microspheres (31)], we observed stepwise changes in the forces exerted on the trapped beads (Fig. 3, A and B), indicating that several cadherin-catenin complexes were initially bound to the actin filament and that they unbound sequentially. The number of stepwise changes scaled with the concentration of Dr αE-catenin used to reconstitute the complexes, but the binding frequency decreased abruptly when less than 10 nM Dr αE-catenin was added to the flow cell (table S1). The lowest concentration of Dr αE-catenin that still resulted in binding activity was ~5 nM. Even at this minimal concentration, however, unbinding events still comprised a few stepwise changes (similar to Fig. 3A), indicating that binding by several complexes was favored over binding by a single complex (table S1).

Fig. 3 Measurement of Dr cadherin-catenin complex/F-actin bond lifetimes in constant-stage-position experiments.

(A) An actin filament binding event representative of those observed in the presence of cadherin-catenin complexes reconstituted with 10 nM of added Dr αE-catenin. The stage was driven using a wave in the shape of a trapezoid with a 100-nm height and 1 × 104 nm/s slope (top, force time series in blue; bottom, stage position time series in black). Piecewise-constant fit of the force signal is shown in red, and the black arrow points to the segment whose duration and magnitude are the lifetime of the last Dr cadherin-catenin/F-actin bond and the force exerted on it, respectively. The black bar underneath the trace represents the total bound time of the entire event. (B) A representative force time series, with the event in (A) shaded in teal. (C) Representative single-molecule force time series showing binding between a surface-bound Dr cadherin-catenin complex and an actin filament in the presence of 100 nM Mm αE-catenin ABD. Flow cell was prepared using 1 nM of added Dr αE-catenin. (D) Force time series showing an actin-binding event measured for multiple surface-bound Dr cadherin-catenin complexes and 100 nM Mm ABD. The black bar above the trace represents the total bound time of the entire event. Flow cell was prepared using 5 nM of added Dr αE-catenin. (E) Survival frequency of total bound times, as marked by the black bars in (A) and (D), measured in experiments using flow cells prepared with 5 nM of added Dr cadherin-catenin complex (red, no Mm ABD present, and n = 412 binding events; blue, 100 nM Mm ABD, and n = 107 binding events). The survival frequency at time t is the fraction of complexes that remain bound for durations greater than t.

To investigate whether multiple cadherin-catenin complexes might bind F-actin more readily than a single, isolated complex, we introduced the actin-binding domain (ABD) of αE-catenin into the reaction buffer. Because ABD binds cooperatively to F-actin but does not bind to β-catenin (9), we reasoned that the presence of ABD would mimic the effect of having many αE-catenin molecules bound to F-actin. When we prepared flow cells with 1 nM Dr αE-catenin to reconstitute the cadherin-catenin complex, none of the microsphere platforms interacted with the actin filament. However, when we included 100 nM of ABD in the assay buffer, we observed many binding interactions that dissociated in a single step (Fig. 3C), indicative of the interaction of a single cadherin-catenin complex with the actin filament. Under these conditions, approximately 1 in 10 platforms interacted with the actin filament, providing further evidence that the large majority of the platforms contained at most one active cadherin-catenin complex (supplementary text). These observations indicate that addition of ABD was sufficient to replicate the presence of multiple cadherin-catenin complexes interacting with the actin filament. Additionally, because we could observe unbinding of single cadherin-catenin complexes reliably only in the presence of ABD, we conclude that multiple cadherin-catenin complexes may be required for actin filament binding to occur at an observable rate. The increased on-rates for individual complexes in the presence of ABD may be due to changes in the actin filament induced by cooperative binding of ABD as reported previously (9), although further experiments are needed to show this unequivocally.

We next asked how the presence of ABD might alter the interaction of the actin filament with many surface-bound cadherin-catenin complexes. Remarkably, in experiments using 5 nM of Dr αE-catenin to reconstitute the cadherin-catenin complex, addition of ABD greatly increased the total bound times of the cadherin-catenin complex/F-actin bonds (Fig. 3, D and E). This observation indicates that cooperative interactions between neighboring cadherin-catenin complexes and the actin filament enhanced the load-bearing capacity of cadherin-catenin/F-actin bonds by substantially extending their total bound time (Fig. 3D).

Two bound states in force-lifetime distributions

The duration of the last segment of stepwise unbinding events (black arrow in Fig. 3A) represented the lifetime of a single cadherin-catenin complex/F-actin bond, and the displacement from baseline represented the load experienced by the bond. The distribution of lifetimes of the last segment in multistep unbinding events revealed the existence of at least two bound states: a subpopulation of short-lived events at all forces and a subpopulation of long-lived events (up to 25 s) at forces between 5 pN and 10 pN (Fig. 4A).

Fig. 4 Force-lifetime distribution of Dr cadherin-catenin complex/F-actin bonds and lifetime survival analysis.

(A) Two-dimensional histogram of Dr cadherin-catenin complex/F-actin bond force-lifetime values measured from the last piecewise constant segment in multistep unbinding events, as shown in Fig. 3A. Tick labels on the color bar are bin counts (17 force bins, 32 lifetime bins, and n = 803 observations). (B) Kinetic schemes representing dissociation from either one bound state (blue) or two bound states (red). (C) Survival frequencies of Dr cadherin-catenin complex/F-actin bond lifetimes from three force bins indicated in Fig. 4A (black arrows, n = 188 observations in bin F1, 185 in F2, and 129 in F3; errors are SEM). Red lines are least-square fits of a biexponential function (two bound states), and blue lines are those of a single exponential function (one bound state). R2 > 0.90 for the biexponential function in all force bins, and the additional parameters of the biexponential function are justified (P ~ 0 in F test). Additional force bins are shown in fig. S2A.

These results formed the basis for testing several models to explain the distribution of cadherin-catenin complex/F-actin bond lifetimes. Models in which dissociation occurs from a single bound state (state 1) result in a bond survival probability that is described by a single exponential function. In contrast, dissociation from two distinct bound states (states 1 and 2) results in a survival probability that is described by a biexponential function, with a separate exponential decay rate corresponding to each bound state (Fig. 4B). In both of these models, state 0 represents the unbound state.

Bond survival probabilities over a broad range of forces were better fit by a biexponential function than a single exponential function; the improved fit was not due to chance (P ~ 0 in F test) (Fig. 4C and fig. S2). Furthermore, the biexponential function fits identified 24% of the lifetimes in the 4-pN bin as long-lived, 43% in the 6.8-pN bin, and 45% in the 13.7-pN bin. Based on this analysis, we conclude that (i) a model with a single bound state did not explain the distribution of cadherin-catenin complex/F-actin bond lifetimes, and (ii) models with two-bound states must recapitulate how the ratio of short-lived to long-lived lifetimes depended on force.

Two-state catch-bond model

Several quantitative models have been developed to account for the effect of force on how fast a bond dissociates. Most models are based on the Bell equation (3335):Embedded ImageIn this equation, Embedded Image is the rate of transition from state Embedded Image to state Embedded Image at force Embedded Image, Embedded Image the rate constant, Embedded Image the distance to the transition state, Embedded Image Boltzmann's constant, and Embedded Image the absolute temperature. As formulated, the Bell equation fits the behavior of slip bonds, in which dissociation rates increase exponentially with respect to applied tension. On the other hand, the equation with a negative exponential argument has been used in catch-bond models that include transitions in which rates decrease exponentially with respect to applied tension (29).

Based on the requirements determined by survival analysis of cadherin-catenin complex/F-actin bond lifetimes, we considered a two-state catch-bond model, which has been proposed to explain FimH adhesion (30, 36), and a two independently bound states model (Fig. 5A). In the two-state catch-bond model, bonds at low force prefer a weakly bound state (state 1) and dissociate quickly (k10) to an unbound state (state 0). As force increases to an intermediate value, bonds transition (k12) into a strongly bound state (state 2) and remain in that state because force opposes transitions (k21) back to the weakly bound state (state 1). In this intermediate force regime, dissociation from the strongly bound state (k20) is not appreciably accelerated, resulting in long bond lifetimes. As force increases further, unbinding occurs directly from the strongly bound state. In a two independently bound states model (Fig. 5A), transitions between states 1 and 2 are not allowed, and bonds remain in the weakly (state 1) or strongly (state 2) bound state in which they originally formed until the cadherin-catenin complex dissociates (k10 or k20) from the actin filament to an unbound state (state 0).

Fig. 5 Two-state catch-bond model.

(A) Kinetic schematics of the two-bound state models that were considered. In a two-state catch-bond model (left), states 1 and 2 are weakly and strongly bound states, respectively. State 0 is the unbound state. Unbinding rates k10 and k20 increase exponentially with respect to force. Transitions between states 1 and 2 occur at rates k12 and k21. The transition rate k12 increases exponentially with force, whereas k21 does the opposite. These transitions do not occur in the independently bound states model (right). (B) Averages of Dr cadherin-catenin complex/F-actin bond lifetimes binned by force (black dots, error bars are SEM, and n = 803 bond force–lifetime measurements). The red curves show the mean lifetime distributions predicted by the two-state catch-bond model (solid red) and a two independently bound states model (dashed red). Model parameters were computed using maximum likelihood estimation. (C) Survival frequencies of Dr cadherin-catenin complex/F-actin bond lifetimes from Fig. 4C, compared with survival probability curves predicted by the two-state catch-bond model (red lines, R2 > 0.90 for all bins except for the 4 pN bin, R2 = 0.67). (D) Two-state catch-bond model dissociation rates as functions of force. (E) Table of maximum likelihood–estimated parameters of the two-state catch-bond model and their 95% confidence intervals determined by parametric bootstrapping.

We found that the two-state catch-bond model was consistent with mean lifetime distributions and bond survival frequencies from our data (Fig. 5, B and C). We used maximum likelihood estimation to calculate the parameters of the two-state catch-bond model most likely to generate our unbinned, bond force-lifetime distributions (n = 803 bond force–lifetime measurements) (supplementary text; Fig. 5, D and E; fig. S3; and table S2). The estimated parameters predicted a mean bond lifetime curve that agreed well with the mean lifetime distribution (R2 = 0.86) (Fig. 5B). Predicted survival probabilities also agreed well with survival frequencies of bond lifetimes in all force bins (R2 > 0.90), except for the lowest force bin (F = 4 pN, R2 = 0.61) in which the model underestimated the fraction of short-lived bonds (Fig. 5C). The most likely explanation for this discrepancy is that bond lifetimes at low forces were undersampled, because our assay cannot detect events that were both short lived (less than ~5 ms) and low force (less than ~1 pN). The two independently bound states model, in which the bond is not allowed to transition between bound states, did not fit the mean lifetime distribution (Fig. 5B), emphasizing the importance of force-dependent transitions in the bound states in the two-state catch-bond model.


Our data and kinetic model reconcile previous in vitro and in vivo studies of the interaction between the cadherin-catenin complex and the actin cytoskeleton. The cadherin-catenin complex/F-actin linkage could not be reconstituted in solution using purified proteins (8, 13, 19) because, as shown here, force needed to be applied to the αE-catenin/F-actin interface to form a stable bond between F-actin and the cadherin-catenin complex.

In our experiments, the duration of direct bonds between the cadherin-catenin complex and an actin filament was sensitive to load, such that at moderate loads (~8 pN) the lifetime of the bonds dramatically increased relative to those observed at lower forces. These data were explained better by a two-state catch-bond model than any of the alternative models that we tested (supplementary text). In the two-state catch-bond model, force shifts the equilibrium from a weakly to a strongly bound state. Our model predicts that cadherin-catenin complex/F-actin bonds formed in the weakly bound state ~90% of the time. This probability was derived from the assumption that bound state transitions reached equilibrium before the application of force, which resulted in a model that was more consistent with our results than those that assumed different initial conditions, including those that incorporated force-loading history (30, 3739) (supplementary text and figs. S4 to S6).

In the two-state catch-bond model, cadherin-catenin complexes in the weakly bound state rapidly dissociate from actin filaments (k10). Force accelerates the transition into the strongly bound state (k12), from which cadherin-catenin complexes dissociate from actin filaments at a rate of ~0.1 s−1 at zero force (k20), a reduction by a factor of 70 compared with the dissociation rate from the weakly bound state (k10). Importantly, the transition from the strongly bound to weakly bound state is greatly decreased by tension (k21): In effect, tension locks the complex in the strongly bound state. Finally, forces greater than 10 pN are sufficient to accelerate the dissociation of cadherin-catenin complexes from the strongly bound state (k20), leading to a decrease in bond lifetimes at high forces.

The force-dependent distance parameters obtained from the model may be related to changes in protein structure that accompany the transition between the weakly and strongly bound states. In particular, the large value of x21, 4 nm, suggests that the αE-catenin/F-actin linkage undergoes a large decrease in length between the strongly and weakly bound states. This putative conformational change underpins catch-bond behavior, because force that opposes this change maintains the bond in the strongly bound state. Previous work indicated that αE-catenin has multiple flexible domains that could be capable of mediating such a transition (14, 16, 21, 27, 4043). At present, however, the precise structural changes in αE-catenin that accompany the transition between the weakly and strongly F-actin–bound states are unknown.

Catch-bond models have been used to describe the bonds formed by integrins, selectins, FimH, and myosin (28, 30, 4447). Interestingly, the homophilic contacts between cadherin extracellular domains have also been shown to exhibit catch-bond behavior (48, 49), indicating that the cadherin complex may be regulated by force on multiple levels. Although a two-state catch-bond model is a kinetic model that fits our data quantitatively, alternative models considering molecular details, such as the sliding-rebinding, deformation, and hydrogen bond network models that have been applied previously to selectins (5052), may also describe the interaction between the cadherin-catenin complex and F-actin. Structural information about how αE-catenin, in a complex with β-catenin, binds F-actin will help establish the molecular basis of the kinetics observed in our experiments.

In summary, our study demonstrates that a strong interaction between the reconstituted cadherin-catenin complex and F-actin requires force and is best described by a two-state catch-bond model. Catch-bond behavior provides a possible explanation for how cells transduce mechanical signals through cadherin-based adhesions, as observed in cell culture and in vivo studies (12, 15, 2326, 53). Given the evidence for force-dependent conformational changes in αE-catenin (15, 16), the two-state catch-bond model may also correspond to distinct conformational states of αE-catenin. In addition, changes in the structure of actin protomers within filaments observed in the presence of αE-catenin ABD (9) indicate that structural changes in the actin filament may also contribute to enhancing αE-catenin/F-actin interactions. Tension-stabilized states have been shown to regulate αE-catenin binding to vinculin, another actin-binding protein (15, 16). Thus, tension may not only shift the cadherin-catenin complex into a strongly bound state but also promote binding of vinculin, thereby creating a self-reinforcing system for strong linkage of the complex to the actin cytoskeleton. The experiments described here provide an approach to address this possibility directly.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 and S2

References (5472)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The authors thank S. Pokutta, J. Bianchini, and P. Miller for help with protein purification and biochemistry. C.D.B. thanks J. Sung and S. Nag for assistance with optical trapping instrumentation. D.H. and N.V. thank W. Ochoa, M. Rodriguez Lee, and C. Page for their contribution to the study, and H. Freeze and F. Levine for advice and discussion. The data reported in this paper are further detailed in the supplementary materials. This work was supported by NSF predoctoral fellowships (C.D.B. and J.T.), Stanford Bio-X predoctoral fellowships (C.D.B. and J.T.), a Burroughs Wellcome Career Award at the Scientific Interface (A.R.D.), the NSF (EFRI 1136790 to W.I.W., W.J.N., and A.R.D.), and the NIH (New Innovator Award 1DP2OD007078 to A.R.D., GM35527 to W.J.N., GM56169 to W.I.W., P01GM098412 to D.H. and N.V., and UO1GM094663 to W.J.N., W.I.W., D.H., and N.V.).
View Abstract

Stay Connected to Science

Navigate This Article