Kindlin stabilizes the talin·integrin bond under mechanical load by generating an ideal bond

Significance

Integrin-mediated cell adhesion is essential for metazoan life. Integrins mediate the mechanical communication between extracellular matrix and cytoskeleton. How a stable mechanical connection between integrins and the actomyosin cytoskeleton can be sustained is far from being fully understood. Key players in the mechanical connection between integrin and actomyosin are the adaptor proteins, talin and kindlin. Using single-molecule mechanical measurements, we found that the bond between talin and integrin alone is weak. However, in the presence of kindlin, the integrin·talin bond is stabilized and rendered force independent, making it a so-called ideal bond. Our findings suggest a unique stabilization mechanism that could help in forming stable integrin adhesion sites.

Abstract

Integrin-mediated adhesion is essential for metazoan life. Integrin binding to ligand requires an activation step prior to binding ligand that depends on direct binding of talin and kindlin to the β-integrin cytoplasmic tail and the transmission of force from the actomyosin via talin to the integrin–ligand bonds. However, the affinity of talin for integrin tails is low. It is therefore still unclear how such low-affinity bonds are reinforced to transmit forces up to 10 to 40 pN. In this study, we use single-molecule force spectroscopy by optical tweezers to investigate the mechanical stability of the talin•integrin bond in the presence and absence of kindlin. While talin and integrin alone form a weak and highly dynamic slip bond, the addition of kindlin-2 induces a force-independent, ideal talin•integrin bond, which relies on the steric proximity of and the intervening amino acid sequences between the talin- and kindlin-binding sites in the β-integrin tail. Our findings show how kindlin cooperates with talin to enable transmission of high forces required to stabilize cell adhesion.

Integrins are a large family of heterodimeric transmembrane receptors that bind to the extracellular matrix (ECM) proteins such as collagens, fibronectin, and laminin, and mediate cell adhesion, spreading, migration, proliferation, and survival (1, 2). In mammals, there are 18 α and 8 β subunits, which assemble into 24 distinct integrin heterodimers. The heterodimers consist of a large ECM-binding ectodomain, single-span transmembrane domains, and short, largely unstructured cytoplasmic tails (Fig. 1A) that mediate interaction with the contractile actomyosin cytoskeleton (3, 4). The β1-integrin subunit forms the largest integrin subfamily, is found on almost all cells, and can be alternatively spliced to produce either the widely expressed β1A-integrin or the β1D-integrin isoform that is highly enriched in the heart and skeletal muscle (5). A hallmark of integrins is that they can regulate their affinity for ligand. During affinity regulation, the unbound form of the integrin ectodomain switches from a bent, low-affinity conformation (inactive integrin, Fig. 1 A, Top Left) into an extended, high-affinity conformation (active integrin, Fig. 1 A, Top Right). Activation results in ECM binding and is followed by integrin clustering and the assembly of hundreds of different molecules which eventually form adhesion complexes (4).

Fig. 1.

The talin•integrin bond is weak and dynamic. (A) Illustration showing the reversible switch from an inactive to an active integrin conformation/state, which requires binding of talin and kindlin to the integrin β-cytoplasmic tail and the propagation of force (red arrows) from actomyosin via talin to the integrin–ligand bond. Once in the active state, integrins bind to the extracellular matrix proteins (i.e., fibronectin) with a substantially increased affinity. (B) Force versus extension trace of a THD2•β1D fusion protein (Inset), in which the β1D-integrin. cytoplasmic tail (gray) is connected via a flexible 75 aa linker (green) to the N-terminal THD2 (purple). The points of force application (pulling geometry) are indicated by red arrows: N-terminus of the talin F0-domain and N-terminus of the integrin β-tail. Once the stretching is initiated and the force load increases, a dynamic/weak flipping signal around 4 to 5 pN is observed. This represents the association and dissociation of the talin•integrin bond and the amino acid linker is stretched. After bond dissociation and linker stretching, unfolding of the talin-2 FERM domains (F0-3) is observed. (C) Equilibrium traces of THD2•β1D bond recorded at increasing (Top to Bottom) biasing forces. Note that with increasing load, the equilibrium of the bond switches from the closed (purple) toward the open (red) state. (D) Binding/closing (red) and unbinding/opening (purple) rates as a function of force for the THD2•β1D bond. Solid lines are extrapolations of the rates to zero load (for details, see SI Appendix, Zero-Load Extrapolation of Binding/Unbinding Rates). (E) Equilibrium traces of THD1•β1D (Top), THD2•β1A (Middle), and THD1•β1A (Bottom) bonds. While a binding signal was observed for THD1•β1D and THD2•β1A, the THD1•β1A bond is too weak to be distinguished from thermal noise and cannot be resolved by our optical-tweezer setup. (F and G) Binding/closing (red) and unbinding/opening (purple) rates as a function of force for the THD1•β1D and THD2•β1A bonds, respectively (compare D).

Two molecules that play a key role for integrin activation are talin and kindlin. They bind to distinct NxxY motifs in the β-integrin cytoplasmic tails and link the integrin tails to the contractile F-actin cytoskeleton (3). Talin and kindlin binding to the β-integrin tail fulfills two major functions. The first is the induction and/or stabilization of the active integrin conformation, which allows ligand binding, however, with only a quite short bond lifetime (<2 s) (6). The second function is to link the β-integrin tail to F-actin, which leads to the transmission of high actomyosin-generated forces followed by the formation of a catch bond between integrin and ECM ligand and a substantial increase of the bond lifetime (6, 7).

Talin and kindlin are FERM (protein 4.1, ezrin, radixin, moesin) domain–containing proteins that consist of the canonical F1-F3 domains and an additional, N-terminal ubiquitin–like F0 domain. The FERM domains of talin, termed talin head domain (THD, F0-F3 in Fig. 1A), are connected by an unstructured linker to a long rod domain, which contains multiple vinculin- and F-actin-binding sites (8, 9). Mammals have three kindlin (KIND1, KIND2, KIND3) and two talin (TLN1, TLN2) isoforms that are expressed in a cell type– and tissue-specific manner (10, 11). KIND1 is primarily expressed in epithelial cells, while KIND2 is expressed ubiquitously except for hematopoietic cells that exclusively express KIND3. TLN1 is widely expressed, and TLN2 is highly enriched together with KIND2 and the β1D-integrin isoform in cardiomyocytes and striated muscle (12, 13).

Although the extracellular catch bond behavior between integrin and the ECM ligand fibronectin has been well studied, it is still unknown how talin with an affinity for β-tails in the high micromolar range (14) can form force-bearing connecting structures to the cytoskeleton capable of resisting forces of 10 to 40 pN (6, 15, 16). To address this enigma, we used single-molecule force spectroscopy to measure the mechanical strength of the THD•integrin tail bond (Fig. 1B) in the absence and presence of KIND2. Our findings revealed that kindlin converts the weak talin•integrin slip bond into an ideal, force-insensitive talin•integrin bond. The ramifications of our findings for integrin function are discussed.

Results

The Talin•Integrin Bond Is Mechanically Weak.

To investigate the talin•integrin binding strength, we fused the β1A- or β1D-tails to THD1 or THD2, respectively, using a peptide linker containing 75 unstructured amino acid residues (cf. Fig. 1B, for details, see SI Appendix, Materials and Methods). The fusion construct was designed such that force is applied at the membrane-proximal part of the integrin tail, which mimics the force transmission in adherent cells. The recombinant fusion protein was expressed in bacteria, purified, and mounted in a dual-beam optical trap using two 170 nm-long dsDNA molecules that were attached via ybbR-CoA conjugation to the N-termini of the integrin β1-tails and the THD isoforms (SI Appendix, Fig. S1 and Material and Methods).

A force versus extension trace for the recombinant THD2•β1D fusion protein is shown in Fig. 1B. At extensions of around 330 nm when the DNA is fully stretched, the force rises and, at around 5 pN, rapid fluctuations between two states separated by 25.4 ± 0.3 nm (from N = 34 Worm-like-chain fits) in contour length can be observed (Fig. 1 B, Inset). The contour length gain of this transition is consistent with the length gain expected when the β1D-tail detaches from the talin-F3 domain and the 75 amino acid residue linker becomes fully stretched. Interestingly, the talin•integrin bond is weak, is dynamic, and requires only small forces of around 4.5 pN to break. At forces exceeding 10 pN, four consecutive unfolding peaks become visible which match the expected increase in length for the fully unfolded FERM subdomains of THD (SI Appendix, Table S1).

Next, we investigated the force-dependent binding and unbinding kinetics of the THD2•β1D bond in passive mode experiments, where the trap centers were held at a constant distance, thus subjecting the fluctuating THD2•β1D fusion protein to a constant average force (Fig. 1C). As expected, at lower forces (around 4.5 pN, upper trace), the THD2•β1D fusion protein populates more often the bound/closed state (purple) than the unbound/open state (red). With increasing force, this equilibrium shifts toward the unbound/open state (5 pN, middle; 5.5 pN, lower trace). From the lifetimes of the bound (purple) and the unbound (red) states, we can determine the on- and off-rates (using SI Appendix, Eq. S3) for THD2•β1D binding as a function of force (Fig. 1D). We find that the on-rates (red data points) drop with increasing force, while the off-rates rise, indicating that the THD2•β1D bond behaves as a slip bond under the influence of load. A fit of the data to the accessible force range using the elastic parameters of the polypeptide (17) yields off-rates and on-rates at zero force of 5.4 ± 0.6/s and 2135 ± 432/s, respectively. Hence, even in the absence of load, the THD2•β1D bond is dynamic with an average lifetime of just 0.19 s. The nonlinear fit model we used for extrapolation considers the force-dependent shift of the transition state position owed to the elasticity of unfolded polypeptide (see SI Appendix for details). Errors represent the random error obtained from a Levenberg–Marquardt fitting algorithm where the individual data points are weighted by their SEM (see SI Appendix for details).

Experiments with combinations of other THD1/2•β1A/D isoforms revealed that THD1•β1D fusion protein produced rapid unbinding events (Fig. 1 E, Top trace), with higher off-rates (Fig. 1F) and lower binding free energies (SI Appendix, Fig. S2F and Table S2) than those of THD2•β1D (SI Appendix, Fig. S2D). The THD2•β1A fusion protein bond is weaker than that of THD2•β1D, with an even lower lifetime of the bound state (Fig. 1E, Middle trace). Extrapolation of the force-independent rate plots to zero load shows that the unloaded off-rates (38.6 ± 20.1 /s) are faster by a factor of 7, while the on-rates (1239 ± 151 /s) are similar to THD2•β1D. The THD1•β1A fusion protein did not produce any binding signal (Fig. 1 E, Top) but only the open conformation could be observed. This finding was corroborated by force–extension curves, which revealed only the unfolding events of the THD1-FERM subdomains but no transition at low force indicative of binding (SI Appendix, Fig. S2A). All talin•integrin isoform pairs that showed detectable binding signals behaved as weak and dynamic slip bonds.

Kindlin Stabilizes the THD•β1D Bond against Force by Creating an Ideal Bond.

Since kindlin cooperates with talin to activate integrins, we investigated how addition of kindlin affects the mechanical strength of the talin•integrin bond. To this end, we performed measurements of various talin•integrin fusion constructs in the absence (Fig. 2A, 1st row, average preload of 5.5 pN) and presence of recombinant KIND2. To increase the solubility of KIND2, we removed the PH domain and the flexible loop in the KIND2-FERM-F1 subdomain (SI Appendix, Materials and Methods). For THD2•β1D, we find that with increasing concentrations of KIND2, the lifetime of the states where THD2 is bound to the β1D tail increases (purple levels in Fig. 2A, 2nd and 3rd row). However, the distribution of this lifetime has become double exponential (SI Appendix, Fig. S3A). To guide the eye, we split (SI Appendix, Fig. S3B) the bound lifetimes in the sample trace where 8 μM KIND2 was added into two classes: short-lived populations also observed in the absence of KIND2 are colored in purple and long-lived populations induced by KIND2 are colored in green (Fig. 2A, 4th row). Splitting was done by applying a double-exponential fit to the data and introducing a cutoff to produce two single-exponential lifetime distributions (SI Appendix, Fig. S3 B and C). Lifetimes of the two separated states were obtained from the double-exponential fit (SI Appendix, Eq. S4) of the normalized cumulative probability distribution. To confirm that the prolonged THD2•β1D bound lifetimes are indeed a consequence of KIND2 binding to the β1D-tail, we measured the THD2•β1D–Y795A fusion protein that carries the kindlin binding–deficient Y795A mutation (SI Appendix, Fig. S3D) (18, 19). With this mutant, THD2-bound lifetimes in the presence of KIND2 are short and indistinguishable from experiments in the absence of KIND2 (SI Appendix, Fig. S3 E and F). Our findings indicate so far that the lifetimes of the bound THD2•β1D states change in the presence of KIND2 by inducing long-lived bound events that we assign to kindlin association with the β1D-tail and hence, the formation of the THD2•β1D•KIND2 ternary complex. Note that bound states occurring from either a binary THD2•β1D (Fig. 2A, purple events) or a ternary THD2•β1D•KIND2 (Fig. 2A, green events) complex formation appear at identical force levels since they involve the same length of stretched linker.

Fig. 2.

Kindlin-2 mechanically strengthens the talin•β1D-integrin bond. (A) Equilibrium traces of THD2•β1D bond measured at an average preload of 5.5 pN. Addition of KIND2 (2nd row to 8 μM, 3rd row to 32 μM) in solution induces bound states with longer lifetimes. For clarity, the long-lifetime states are colored in green (Bottom row) by introducing a lifetime cutoff according to a double-exponential fit (as detailed in SI Appendix, Fig. S3). (B) Binding/closing (red) and unbinding/opening (purple) rates as a function of force for the THD2•β1D bond. Measurements performed in the absence of KIND2 are indicated by open symbols, while those in the presence of KIND2 are indicated by filled symbols. The green symbols represent the rates obtained from the KIND2-induced slow phase of the double-exponential fit to the lifetimes of the bound state (SI Appendix, Fig. S3). (C) Equilibrium traces of THD1•β1D (Top), THD2•β1A (Middle), and THD1•β1A (Bottom) bonds measured in the presence of KIND2 (5 μM, 10 μM, and 10 μM, respectively). Bound/closed states with a prolonged lifetime (green) are observed only for THD1•β1D. (D) Binding/closing (red) and unbinding/opening (purple, green) rates as a function of force for the THD1•β1D bond (see description in B).

The force-dependent rate plots shown in Fig. 2B revealed similar on-rates for THD2•β1D binding in the presence (red filled symbols) or absence (red open symbols) of KIND2. Furthermore, the short-lived THD2•β1D bound events in the presence of KIND2 (filled purple symbols) show the same off-rate and force dependence as THD2•β1D without addition of KIND2 (open symbols), indicating that those events reflect unbinding of a binary THD2•β1D complex. The long-lived (green) population, however, shows no force dependence and a slow unbinding rate of 6.1 ± 0.5/s at all forces measured, indicating that KIND2 stabilizes the THD2•β1D bond under load by making the bond insensitive to force. Extrapolation of the off-rates to zero force shows that the unbinding rates in the absence of load are very similar for the short-lived (purple) and the long-lived (green) population, strongly indicating that the kindlin-induced THD2•β1D bond stabilization arises under load and not at zero force conditions. The extrapolated zero-force rates and the average values of the kindlin-stabilized rates are summarized in SI Appendix, Table S2.

We find that the stabilization of the talin•integrin bond mediated by kindlin only occurs for combinations involving the β1D-tail isoform (Fig. 2 A and C, Top trace). Rate plots for the combination of THD1•β1D show that the off-rates of the green states are significantly slower than those of the purple data points, however, faster than in the case of THD2•β1D (Fig. 2 B and D). Furthermore, the extrapolation to zero load of the fast THD1•β1D unbinding rates agrees with the slow THD1•β1D•KIND2 unbinding rates (SI Appendix, Table S2), indicating that binding of KIND2 prolongs the THD1•β1D bond lifetime in the presence of cytoskeletal forces in the low piconewton regime. In contrast, however, addition of kindlin to combinations with the β1A-tail isoform affects neither off-rate nor force dependence (Fig. 2C, Middle and Bottom traces).

To extend the narrow force range accessible in passive-mode measurements, we employed a jump protocol (20) where we rapidly switch the force applied to the talin•integrin bond between high and low forces (Fig. 3A). Thus, we allowed the bond to form during the low-force phases and measured its lifetime during the high-force phases. This protocol allowed us to investigate the force sensitivity of the KIND2-stabilized THD2•β1D bond up to 18 pN of force. The extended off-rate versus force plot in Fig. 3B corroborates our finding that kindlin-stabilized THD2•β1D bond states (green) remain strictly force independent up to the highest loads applied. The THD2•β1D bond states without addition of KIND2 (purple) continue to be force dependent. However, above 6 pN of force, the slope changes, possibly indicating a force-induced shift of the transition state position (21). Toan and Thirumalai (22) have provided an elegant explanation for such a change in slope based on a free energy model that changes the transition state position abruptly with increasing force. To test the sensitivity of the KIND2-mediated stabilization effect on the pulling direction, we tested our results using two recombinant fusion constructs, in which the pulling point was shifted from FERM subdomain F0 to FERM subdomain F3 in either the full-length THD2 or only the THD2-F3 (SI Appendix, Fig. S4). We decided to also investigate this pulling direction because it may be the preferred direction of pulling in cells (Fig. 1A). Importantly, we found that both constructs behave similarly and respond equally to the stabilization effect of KIND2 as the recombinant fusion protein with the F0 subdomain pulling point used in all experiments above. These findings indicate that the talin•β1-integrin tail bond stabilization by kindlin is independent of the pulling geometry and the talin-F0, -F1, and -F2 subdomains.

Fig. 3.

Kindlin-2 renders the talin•integrin bond force insensitive. (A) Jump assay to study the THD2•β1D bond in the presence of KIND2 at high loads. The bond formation was induced by jumping to a low-force level (3.0 pN) and its unbinding/opening time was measured after quickly jumping back at high forces (18.1 pN)–see upper trace. Even at the much higher load, the lifetimes of the long phases are independent of force. The lifetimes of the short purple phases (zooms at the bottom) show the very rapid dissociation (purple) of the bond when it is not stabilized by kindlin. A further drop (16.5 pN to 13.6 pN) in force is observed after the talin•integrin bond dissociates. This is caused by the additional unfolding of a THD FERM domain. (B) Extended rate plot for the THD2•β1D bond covering the full force range measured in the absence (open symbols) and presence (filled symbols) of KIND2 in solution. Jump data are identified by triangular symbols, while normal equilibrium measurements are identified by round symbols. The presence of kindlin in solution stabilizes the talin•integrin bond and makes it force independent (green data points).

The Proximity of the Kindlin and Talin Binding Sites Is Required for the Ideal Bond Behavior.

A bond with an off-rate that is strictly independent of load has been termed ideal bond (23). It means that the transition state for unbinding under load coincides with the bound state, which implies that the bond will not slide into a different conformation before it breaks. In the absence of kindlin, we find from fitting (using SI Appendix, Eq. S13) the slope of the force-dependent rate plot in the low-force region in Fig. 3B, that the transition state for force-dependent bond rupture is at 7.7 nm of peptide contour length (SI Appendix, Table S3), indicating that there will be some mechanical motion between talin and β1-integrin tail along the pulling direction before the bond ruptures (Fig. 4A) (24). Since the binding sites of talin and kindlin on the β1-integrin tail are very close with some amino acid residues (KSPIN motif) potentially even overlapping, we hypothesized that the physical proximity of the kindlin and talin molecules might block sliding of talin along the β1-integrin tail prior to bond rupture (Fig. 4B), making the kindlin-bound conformation of the talin•β1-integrin tail bond independent of force. We tested this hypothesis and separated the talin- and kindlin-binding sites on the integrin tail by introducing an unstructured spacer consisting of 10 amino acid residues (see sequence in Fig. 4C), which should allow sliding between talin and integrin in the presence of kindlin and abrogate the ideal bond effect. To ensure an unaltered binding strength of the β1D-tail to both kindlin and talin, we duplicated the KSPIN motif at the C-terminal end of the spacer (25) and confirmed in microscale thermophoresis (MST) measurements that the affinities to talin and kindlin remain unchanged (SI Appendix, Fig. S5 B–D). The equilibrium traces recorded for the recombinant THD2•β1D-spacer-containing fusion protein indeed revealed that the addition of KIND2 failed to stabilize the talin•β1-integrin tail bond (Fig. 4 D and E and SI Appendix, Fig. S5A).

Fig. 4.

Possible mechanism of talin•integrin ideal bond formation initiated by the binding of kindlin. (A) Schematics of talin•integrin bond rupture when force (red arrows) is applied. Bond-forming contacts are indicated by lines. Force application induces sliding of talin (purple) along the integrin tail (gray), thus breaking the slip bond. (B) Talin•integrin ideal bond formation initiated by the binding of kindlin (green). Kindlin sterically blocks the sliding of talin across integrin when force (red arrows) is applied to the bond. This renders the bond insensitive to force. (C) Design of an insert mutant breaking the steric proximity of kindlin and talin. Compared to the native sequence (Top), the mutant construct (Bottom) is designed such that an unstructured insert of amino acids (red) between the talin and kindlin binding sites breaks their close proximity. The KSPIN motif is duplicated to ensure unaltered affinity to kindlin and talin. The two NPxY motifs required for talin and kindlin binding are indicated in purple and green, respectively. (D) Equilibrium traces of the THD2•β1D insert mutant recorded in the absence (Top) and presence (Bottom) of KIND2. The kindlin-induced stabilization effect is lost. (E) Model for abrogation of the kindlin-induced ideal bond effect in the insert mutant. Due to the separation of the talin- and kindlin-binding sites, kindlin binding is unable to block the sliding of talin along the integrin insert mutant.

It has been suggested in molecular dynamics simulations that the membrane-proximal helix of the integrin β3-tail might play an important role during the integrin inside-out activation process (26). According to the model proposed by Haydari et al. (27), kindlin triggers the upward sliding of talin to the membrane-proximal region of β3-integrin tail, thereby enhancing the talin•integrin interaction. To test such a potential effect, we performed experiments with a THD2 fused to the integrin β1D-tail lacking the membrane proximal α-helix (SI Appendix, Fig. S6) and found that the ability of KIND2 to stabilize the THD2 bond with the mutant β1D-tail is unchanged. This finding rules out a contribution of the membrane-proximal α-helix to the kindlin-induced stabilization of the talin•β1-integrin tail bond.

A Single Proline Residue Mediates the Ideal Bond Behavior.

What are the molecular determinants within the β1A- and β1D-integrin isoform sequences that cause the different molecular bond behavior? Anthis et al. (14) identified two major differences in the talin-binding region between integrin β1D- and β1A-tail: a glutamine versus a glycine residue in position 778 and a proline versus an alanine residue in position 786. We tested, whether changing the residues at either position can induce an ideal bond behavior in integrin β1A. We found that the G778Q substitution slightly stabilized the THD2 bond with the mutant β1A-tail (Fig. 5F) but did not induce an ideal bond behavior. In sharp contrast, the A786P substitution induced an ideal bond behavior (Fig. 5G) with clearly visible long-lived and force-independent THD2•β1A-A786P•KIND2 bonds (green) (Fig. 5D), highlighting the importance of residue A786 in β1-tail isoforms for the mechanics of talin•kindlin•integrin crosstalk.

Fig. 5.

Proline-786 is key for the ideal bond formation. (A) Sequences of the integrin β1D and β1A cytoplasmic tails. Residues that are not conserved between the talin-binding sites of β1A and β1D are indicated by arrows. (B–D) Equilibrium traces of (B) THD2•β1D-Q778G, (C) THD2•β1A-G778Q, and (D) THD2•β1A-A786P fusion proteins in the absence (Left) and presence of 5 μM KIND2 (Right). (E–G) Binding (red circles) and unbinding (purple circles) rates as a function of force for (E) THD2•β1D-Q778G, (F) THD2•β1A-G778Q, and (G) THD2•β1A-A786P fusion proteins. Upon KIND2 addition, slower/stabilized unbinding rates were only observed for THD2•β1A-G778Q and THD2•β1A-A786P fusion proteins (green; E and G). The unbinding rates of the THD2•β1D (gray triangles) and THD2•β1A bond (gray squares) are shown for comparison.

Discussion

Strength of the Talin•Integrin Bond.

Force transmission across integrins occurs via mechano-regulated linkages that connect ECM-bound integrins via talin to actomyosin. These linkages generate ligated integrins exposed to low and high forces (28) and include the stretching of the talin rod by actomyosin leading to exposure of cryptic vinculin-binding sites and force-induced allosteric change(s) of the integrin ectodomain leading to the formation of integrin catch bonds with ECM ligands and adhesion strengthening (6, 29–32). However, these studies also raised the question how the high pulling forces of actomyosin (16) and the forces required to induce an integrin catch bond (6, 33, 34) can be achieved with the low talin affinity for β-integrin tails. To solve this conundrum, we developed several THD•β1-integrin tail fusion constructs to probe the force response of the talin•integrin bond in the absence and presence of kindlin, which tightly cooperates with talin to maintain integrins in a ligated state (35, 36).

Our results on the breaking forces of the bond between talin and integrin clearly show that, in isolation, this bond is highly dynamic and mechanically weak and behaves as a slip bond under force. The strongest combination of talin•integrin isoforms, THD2•β1D, breaks at forces of around 4 to 4.5 pN while binding of the combination of THD1•β1A was too weak to be detected in our single-molecule assay. The higher strength of the THD2•β1D bond is consistent with our and published (14) affinity measurements that revealed highest affinity of talin-2 for β1D integrin among the talin and β-integrin tails tested. Importantly, however, neither affinity suffices to sustain the forces up to 10 pN reported to act on talin ex vivo (16), let alone the 40 pN reported as peak loads on the bond between integrin and fibronectin (33). Although the talin rod domains may act as force buffers (15) by unfolding at forces between 5 and 10 pN and thus, increasing the lifetime of the bonds involved in the force-bearing chain between integrins and the cytoskeleton, it is a reasonable assumption that more factors support the mechanics of the talin•integrin bond required to maintain integrin in an active state. Interestingly, Jiang et al. (37) reported a weak slip bond breaking at forces of only 2 pN located somewhere between integrin and actin on the intracellular side, showing that extremely weak bonds are involved in the force-bearing chain. However, it remains unclear how the ECM–integrin–actomyosin linkages can bear the high forces in an integrin adhesion site.

Kindlin Converts the Talin•Integrin Slip Bond into an Ideal Bond.

Since kindlin cooperates with talin to stabilize the integrin–ligand interaction, we tested whether kindlin can modulate the strength of the weak talin•integrin bond. The results of our measurements revealed that kindlin efficiently stabilizes the THD2•β1D bond by making the bond force insensitive up to the highest forces measured (18 pN) in our experiments.

A strictly force-independent bond has been termed an ideal bond (23) and so far reported for very few force-bearing interactions. Most often, bimolecular bonds exhibit slip bond behavior which is characterized by an unbinding rate that increases with the load applied to the bond. In a one-dimensional energy landscape model for bonds, all bonds with a finite transition state position will behave as slip bonds. In contrast to slip bonds, catch bonds become stabilized against breaking under mechanical load (6, 38–41). A macroscopic analog of a catch bond would be a finger trap that traps stronger the harder one pulls, which makes their bond lifetimes insensitive to applied mechanical loads. Ideal bonds have been reported for the dynein–microtubule (42) and cadherin–cadherin interactions (38). Ideal bond behavior can arise if the pulling process does not affect the barrier height of the bond. In a one-dimensional energy landscape picture, the native state and the transition state would lie at the same position. In this case, application of force would affect the energy of the native as well as the transition state in exactly the same way and therefore, the energy barrier separating them would not depend on force. It is, however, well possible that like in the case of catch bonds (43), 2-dimensional energy landscapes provide better explanations for the ideal bond effect (44).

Even though high-resolution structural information about the ternary complex of kindlin, talin, and integrin is not yet available, kindlin binds very close to the talin-binding site on the integrin tail upon ternary complex formation (eye-guided fill model, see ref. 45). Consistent with this view, the mutation Y795A, which prevents kindlin binding to the β-integrin tail (18, 19), also abrogates the ideal bond effect (SI Appendix, Fig. S3 D–G). A binding scenario schematically outlined in Fig. 4 A and B can explain how binding of kindlin in close proximity to talin on the integrin tail, namely acting as a roadblock preventing sliding of talin along the integrin tail when force is applied. Hence, in the presence of kindlin, force cannot actively move talin and integrin apart from each other and unbinding would have to occur along a coordinate perpendicular to the direction of force applied to the THD2•β1D•KIND2 complex. Indeed, control experiments using a β1D-tail with the talin- and kindlin-binding sites separated by the insertion of a spacer sequence restored the slip bond behavior. Furthermore, we identified the proline residue-786 in the β1D-tail as key residue for the ideal bond behavior. Substitution of alanine-786 for proline in β1A converted the apparent slip bond in a readily visible ideal bond. Since proline residues generally restrict the orientational flexibility of the amino acid backbone, it might help keeping kindlin and talin in a fixed orientation to each other when force is applied and thus, preventing slippage of the talin•integrin tail bond. While the roadblock model we propose is consistent with our results, other mechanisms involving formation of a tripartite complex of integrin, talin, and kindlin with direct interactions among the three players may also lead or contribute to the observed ideal bond behavior.

Possible Role for Kindlin-Induced Stabilization during Adhesion Maturation.

It is striking that the ideal bond effect is readily visible for the β1D isoform that is highly enriched in the heart and striated muscle. A study comparing the dynamics of β1D and β1A variants in cells (46) reported significantly increased exchange dynamics of β1A in focal adhesions as compared to the β1D variant and that the exchange dynamics of β1A was slowed down by introducing the β1A-A786P substitution. Although the different dynamics was explained merely by the increased affinity of talin-2 to β1D as compared to β1A, our findings indicate that the differential mechanical stability and force sensitivity of the tripartite complexes with kindlin provide a further and probably decisive contribution to maintain integrin ligation. It is obvious that the stabilization effect of the talin•integrin bond induced by kindlin is significant but its effect is transient because even the stabilized bond will break on average within 200 ms. Particularly in the high-force environment of striated muscle cells, the stability reinforcement of the talin•integrin bond by kindlin is probably providing the necessary time frame to induce and/or maintain integrin adhesions to tendon cells (47). On the contrary, outside the striated muscle system, forced expression of the β1D integrin isoform leads to embryonic lethality with a plethora of severe defects (47), indicating that talin•β1-integrin tail bond dynamics must be fine-tuned in different cell types.

While the adaptation of the extracellular domains of integrins to sustain large mechanical forces through a catch-bond effect between ECM proteins such as fibronectin and integrin ectodomains has been established long time ago (6), our study sheds light on the interplay of important downstream components of the force-signaling chain. While the ideal-bond effect we report in this study is an attractive mechanism for increasing bond lifetimes under load, it is obvious that understanding the mechanics of a dynamic system like the focal adhesion involving several hundreds of components (48–50) is far from complete. In particular, the role of membrane-associated phospholipids in enhancing binding affinities and mechanical attachment strength will be important to include in future studies. Furthermore, quantitative understanding of a cellular adhesion will critically rely on quantitative single-molecule studies of the many components involved.

Materials and Methods

All protein constructs were prepared using standard recombinant techniques as described in SI Appendix. The single-molecule force-spectroscopy measurements were carried out using a custom-built dual-beam optical tweezers, while the binding affinity solution measurements were done on a commercial (Nanotemper, Munich, Germany) MST Monolith NT115 setup (see SI Appendix for details). Fitting of force–extension cycles and analysis of passive mode traces are explained in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.