Molecular mechanisms of mechanotransduction in integrin-mediated cell-matrix adhesion

Abstract

Cell-matrix adhesion complexes are multi-protein structures linking the extracellular matrix (ECM) to the cytoskeleton. They are essential to both cell motility and function by bidirectionally sensing and transmitting mechanical and biochemical stimulations. Several types of cell-matrix adhesions have been identified and they share many key molecular components, such as integrins and actin-integrin linkers. Mechanochemical coupling between ECM molecules and the actin cytoskeleton has been observed from the single cell to the single molecule level and from immune cells to neuronal cells. However, the mechanisms underlying force regulation of integrin-mediated mechanotransduction still need to be elucidated. In this review article, we focus on integrin-mediated adhesions and discuss force regulation of cell-matrix adhesions and key adaptor molecules, three different force-dependent behaviors, and molecular mechanisms for mechanochemical coupling in force regulation.

1. Introduction

Cell–matrix adhesions are discretely distributed on the cell surface and mediate cell interactions with the ECM. They are essential for cellular functions, such as cell rolling, migration, differentiation, tissue remodeling and homeostasis [1]. The adhesions can either be quickly disassembled or progressively matured from focal complexes to fibrillar adhesions in changing the shapes and molecular compositions of the adhesion complexes. Mechanical force plays a key role in adhesion maturation [2]. During the past two decades researchers have focused on force regulation of molecular interactions including integrins and adaptor molecules involved in cell-matrix interactions. Along with the development of sophisticated single molecule experimental techniques, molecular and structural insights into the mechanical regulation of cell-matrix adhesions have been provided [3].

Before we explore the details of force regulation and transmission at the adhesion sites, we first describe characteristics of different integrin-mediated adhesions. Four different cell-matrix adhesion structures have been defined in cultured cells: focal complexes (nascent adhesions), focal adhesions (FA), fibrillar adhesions, and three-dimensional (3D) matrix adhesions [4, 5]. Nascent adhesions are force-independent and are transient adhesion structures, which are formed at the periphery of spreading or migrating cells in early stages of cell attachment. They can mature into a force-dependent FA when proper linkages are established with the actin cytoskeleton, and the typical size of nascent adhesions is less than 0.2 μm² [6]. Compared to nascent adhesions, FAs are relatively stable and highly regulated structures by modulating association and dissociation of focal adhesion proteins. FAs can further develop into fibrillar adhesions, which are more stable structures. Recent studies have shown that cells in a 3D matrix interact with the environment with distinguished properties and behaviors compared to that of the adhesions formed in the two-dimensional (2D) matrix. This type of cell-matrix adhesion is called 3D matrix adhesion [4, 7].

In different stages of cell-matrix interactions, the composition of adhesions and adaptor molecules varies (Fig. 1). Focal complexes are multi-molecular structures that include integrin, talin, paxillin, vinculin, and actin [8–12]. Focal complexes develop into FA by recruiting zyxin [9, 13, 14]; and tyrosine phosphorylation is also required for the recruitment and further maturation of FAs [15, 16]. FAs typically disassemble within 10–20 minutes, however when α₅β₁-integrin adheres to a fibronectin matrix, FAs further mature into fibrillar adhesions as they elongate the shape of adhesion sites and change the molecular compositions. Stress-fibers are anchored on adhesion sites and the sites move toward to the center of the cell. Fibrillar adhesions specifically contain α₅β₁ integrin and tensin to form a very stable α₅β₁-tensin complex, which is scarcely observed in the migrating cell [17, 18]. Throughout the maturation of cell-matrix adhesions, integrins are key players. Integrins are a family of transmembrane proteins: the ectodomain of integrin binds to its ECM ligands and the cytoplasmic tail links to the cytoskeleton via adapter molecules including talin, α-actinin, vinculin, and tensin, thereby connecting the intracellular cytoskeletons to the ECM through the adhesion sites and transmitting bidirectional signals (Table 1).

Figure 1. Schematic of the integrated mechanotransduction at the integrin-mediated adhesion site.

(A) Interaction of cell with the ECM, which is mediated by a multi-protein complex, FA. The red rectangle is enlarged in B. (B) A widely accepted model for force-mediated FA maturation with key adaptor molecules. Integrin interacts with its ECM ligands (e.g. fibronectin) and it triggers signal transduction of followings: (1) Integrin in inactivated state. Talin is recruited to the FA site and starts interacting with integrin β3-tail. Vinculin and other adaptor molecules are also recruited and concentration of free G-actin is accumulated at the adhesion site. (2) Integrin engages with its ligand, and transition to intermediate state. F-actin polymerization is accelerated and force is loaded on the linkage of integrin-adaptor molecule to the actin cytoskeleton. Force-activatable molecules (e.g. vinculin and talin) in the adhesion complex sense the externally applied force and modulate their activation and affinity states for their ligands to transduce the mechanical signaling. (3) Conformational activation of integrin, an extended-open high-affinity state, is fully induced by talin-β-integrin tail binding, integrin ligand binding to integrin head piece, or force applied on integrin. The activation results in modulation of interaction kinetics for its ECM ligands to further develop FAs. FA is matured and stabilized at this stage.

Table 1. Adaptor molecules involved in integrin-mediated mechanotransduction at FAs.

Integrins transmit mechanical outside-in and inside-out signaling via adaptor molecules, which are recruited to the adhesion sites under mechanical force loading conditions.

Protein	Functions	Force-dependent behavior	Ref
Integrins	Transmembrane receptor Linkage of the actin cytoskeleton to the ECM	Clustering at FAs Conformational activation with high-affinity for its ligands
Zyxin	LIM protein Localized at FAs and along the actin cytoskeleton	Recruiting Ena/VASP and α-actinin Enhancing actin polymerization in cooperation with Ena/VASP	1*,3*
Tensin	Linkage of integrins to the actin cytoskeleton during FA maturation	Maturing FAs into fibrillar adhesion by inducing a5b1 integrin translocation
Talin	Linkage of integrins to F-actin at FAs	Involving in integrin activation by binding to the integrin β-tail Conformational activation with high-affinity for its ligands (e.g. exposing cryptic VBSs)	39, 40, 41, 2*
Paxillin	Interacting with tyrosine kinases (e.g. Src and FAK) and adaptor molecules (e.g. vinculin and talin)	Stabilizing FAs as force increases Recruiting FAK	4*
Vinculin	Linkage of integrin-talin complex to F-actin Binding alternately to tlain or α-actinin	Conformational activation with regulation of affinity for the ligands (e.g. talin, α-actinin and F-actin)
α-actinin	Linkage of integrin to the actin cytoskeleton via interaction with vinculin, zyxin and F-actin	Stabilizing FAs as tensile force increases	3*, 4*
filamin	Crosslinking two F-actins into an orthogonal structure Interacting with integrin	Conformational activation with regulation of affinity for the ligands (e.g. integrin)	5*

¹H. Hirata, Zyxin emerges as a key player in the mechanotransduction at cell adhesive structures, 2008, 1(2): 192–5

²M. Yao, Mechanical activation of vinculin binding to talin locks talin in an unfolded conformation, Scientific reports, 2014, 4:4610

³Reinhard M, An alpha-actinin binding site of zyxin is essential for subcellular zyxin localization and alpha-actinin recruitment, JBC, 1999, 274(19): 13410-

⁴Ye N, Direct observation of α-actinin tension and recruitment at focal adhesions during contact growth, 2014, Exp Cell Res, 327(1):57-

⁵L. Rognoni, Dynamic force sensing of filamin revealed in single-molecule experiments, PNSA, 2012, 109(48):19679–19684.

There are other types of integrin-mediated structure: podosome and invadopodia. Briefly, these are actin-rich dynamic protrusions on the periphery of the plasma membrane [19]. They are found in many cell types, including invasive cancer cells, endothelial cells and immune cells and act as sites of ECM degradation and attachment for cell migration. These two structures share many characteristics in structure and molecular composition (e.g. biomarkers such as WASP, Tks4, Tks5 and MT MMP) [20]. Typically podosomes imply normal cell and invadopodia is cancer cells [21]. Podosomes are highly regulated by many actin regulators, so actin turnover occurs within seconds and the structure lasts only minutes. Compare to that, FAs are more stabilized structures and can last hours.

In this review article, we have focused on molecular mechanisms underlying force regulation of the integrin-mediated adhesion complex. The first part of this review is about the key molecules involved in mechanotransduction and the second part covers characteristics and mechanisms of force regulation.

2. Key adaptor molecules in the integrin-mediated adhesion

1) Integrin

Integrins are the most intensively studied and best characterized among cell-matrix adhesion proteins. In 1986, Hynes and colleagues had revealed structural features of a transmembrane glycoprotein and named the complex ‘integrin’ to denote the role in transmembrane association between ECM and the cytoskeleton [22]. At the same year, Ruoslahti lab also had independently discovered a family of RGD-specific adhesion receptors, a integrin family [23].

Integrins are a heterodimeric transmembrane protein of non-covalently associated α and β subunits. There are 18 α- and 8 β-subunits known, forming at least 24 distinct αβ combinations. Integrins are classified according to their ligands. Collagen-binding integrins include α₁β₁, α₂β₁, α₁₀β₁, and α₁₁β₁ [24–27]; laminin-binding integrins include α₃β₁, α₆β₁, α₆β₄ and α₇β₁ [28, 29]; and fibronectin-binding integrins include α₅β₁, α₈β₁, α_IIbβ₃ and the α_vβ₃ [30, 31]. Integrins consist of a relatively large ectodomain, membrane-spanning helixes, and relatively short cytoplasmic tails. The molecular weight of integrins varies from 90 to 160 kDa, and the molecular weight of the ectodomain is between 80 to 150 kDa. Integrins can be classified into two main groups, according to the presence or absence of an αI-domain in the α-subunit. In αI-less integrins, the binding sites to the ligands reside at the interface between the β-propeller domain of α subunits and βI domain of β subunits. In αI integrins, the binding sites are located in the αI domain [32].

The integrin ectodomains bind to ECM glycoproteins (e.g. fibronectins, collagens, and laminins), other cellular receptors (e.g. vascular cell adhesion molecule-1) and the intercellular cell adhesion molecule family (e.g. ICAM-1 and ICAM-3) [24, 33]. The cytoplasmic tails interact with adaptor molecules, such as integrin activators (e.g. talin and kindlin) [34–37] and inhibitors (e.g. filamin) [38, 39]. Competitive binding between these molecules regulates integrin-mediated signal transduction [38, 39]. Focal adhesion kinase (FAK), a cytoplasmic tyrosine kinase, is thought to interact with integrin through a FERM domain [40, 41], which is also found in the talin head and interacts with the integrin β leg, or interact indirectly through vinculin or paxillin [41]. The direct and indirect integrin interactions with FAK and other tyrosine kinases including ILK and TAP20 may trigger downstream signal transduction [42].

Electron microscopy studies suggest that the ectodomains are constrained in either a bent conformation with a height of ~11 nm (Fig. 1B-1) or an extended conformation with a closed headpiece with a height of ~20 nm (Fig. 1B-2). The bent conformation is found to be in the physiological low affinity state for its ligands [43]. The extended-closed headpiece integrin may undergo further changes to a conformation with an open headpiece by swinging the hybrid domain away from the α-subunit during signal transduction (Fig. 1B-3) [43–46]. The extended-open conformation exhibits a high affinity for its ligands compared to the bent conformation [47, 48]. This conformation change is termed integrin activation. The intermediate conformation, the extended- closed headpiece (Fig. 1B-2), is predicted to possess an intermediate affinity to its ligands [44, 45]. The conformational change from a bent to an extended-open conformation with separation of the α and β leg domains consists of three major domain motions: extracellular integrin extension; headpiece opening and hybrid domain swing-out; and separation of the cytoplasmic α and β legs. The order of these motions may depend on the stimulations applied on the integrin [44].

Dynamic regulation of integrin activation is induced by ECM ligand to the ectodomain as well as association of intracellular adaptor molecules (e.g. kindlin and talin) to the cytoplasmic tail. The bidirectional activation processes are so-called outside-in and inside-out signaling: (outside-in) ligand binding to the integrin stimulates the head piece opening followed by integrin extension and the leg separation; (inside-out) the integrin cytoplasmic tail binding to cytoskeleton proteins leads to leg separation and the conformational change propagates to the headpiece.

The inside-out and outside-in signaling processes are not independent or mutually exclusive. In physiological conditions, a small fraction of integrins reside in the extended-open conformation. Integrin binding to specific ECM ligands (e.g. α_IIbβ₃/α_Vβ₃ to fibronectin) induces and stabilizes the extended-open high-affinity conformation [43]. This conformational change recruits integrin-binding adaptor molecules to the integrin tail, and subsequently triggers further outside-in downstream signaling, such as FAK phosphorylation and recruitment of other signaling proteins. These recruited adaptor proteins can lead inside-out signaling, and vice versa [46, 53, 62, 63].

In the perspective of conformation, both experimental [53] and computational studies [54] showed that the binding of a cytoskeleton protein, talin, to the integrin β cytoplasmic tail is able to induce separation of the α and β legs. Steered molecular dynamics (SMD) stimulation suggested that ligand binding also may induce integrin activation. Soaking αIIbβ3 crystals with linear RGD peptide [57] and soaking α5β1 crystals with cyclic RGD peptide [56] leads to headpiece opening and hybrid domain swing-out. Integrin conformation is also regulated by different divalent cation conditions. Mn²⁺ activates integrin by inducing the extended conformation, while Ca²⁺ stabilizes an inactive, bent conformation of integrin [64]. In addition, mechanical force is a key player of integrin activation. The first observed outside-in force transmission by integrins was integrin β1-induced FA formation and cytoskeletal stiffening [49]. External forces enhance integrin activity and clustering, and induce cell-matrix adhesion and cytoskeletal rearrangement. The inside-out force signaling has been observed as actomyosin contraction regulates FA structure and dynamics by strengthening integrin-ECM binding [50–52]. SMD study suggested that mechanical force may induce hybrid domain swing-out and integrin extension [58, 59]. This force-induced extension has been observed in single-molecule experiments as well [60, 61]. More details about force-mediated activation will be discussed in section-3.

Integrin clustering and spacing between the molecules are another important in integrin adhesion and activation. Ligand binding recruits cytoplasmic proteins, such as actin-binding, signaling and adaptor proteins linking integrin to cytoskeletons [65, 66]. The recruited proteins, like talin and kindlin, promote integrin clustering followed by additional recruitment of signaling proteins to FAs [67–72], and leads to reinforcement actin cytoskeleton link to the integrin-mediated adhesion site [73, 74]. As many integrin ligands are multivalent, integrin clustering significantly increases adhesion through avidity modulation [75, 76]. How close is clustered integrin? On the patterned substrates adhesion reinforcement was observed when the distance between integrins was <70nm [77]. Difference integrins play different roles in establishing adhesion. Integrin α5β1 clustered within 40 nm reinforced adhesion strength to the fibronectin under force; however α_vβ₃ or talin connect to the cytoskeleton and reinforce the connection in response to the force [78].

To measure forces on a single cell or molecule level, different techniques from traction force microscopy to molecular tension sensors have been developed. Forces applied at single adhesion sites were first measured using elastic micropatterned substrates and the stress is found to be 5.5 ± 2 nN/μm² [79]. Forces generated by single integrin has measured by FRET-based molecular tension sensor [80] and DNA-based sensor in living cells [81] in the range of 1 – 40 pN. Mostly forces exerted at single molecules have read in a range of force, 2 – 40 pN [82–84]. Therefore, single molecular experiments using force microscopies are typically conducted in the range of force.

2) Talin

Talin is the most highlighted among adaptor proteins due to the direct interaction with integrins. Talin is the critical linking component between integrin receptors and the actin cytoskeleton. It was discovered in the focal adhesion in 1983 [85] and in the following year it was found that talin interacts with vinculin [86], another linking component between the cell membrane and actin cytoskeleton. Talin consists of an N-terminal globular head region (~47 kDa) and a large flexible C- terminal rod domain (~220 kDa). The head region is composed of four subdomains: the F1, F2 and F3 domains, which form a FERM (protein 4.1/ezrin/radixin/moesin) domain, and an F0 domain. The F3 subdomain resembles a phosphotyrosine binding-like domain, which binds to the membrane-proximal NPxY motif of the β-integrin cytoplasmic tails [35, 87, 88]. The F1 and F2 domains anchor to the cell membrane. The talin rod (TR) is composed of 62 periodic amphipathic α-helices (H1–H62), which form into a series of 13 helical bundles (R1–R13), followed by a helical dimerization domain (DD) at the C-terminal forming an anti-parallel dimer. In the default physiological native state, talin forms an autoinhibited conformation by intracellular head-to-rod binding. In the inactive autoinhibited conformation, many binding sites, including (at least) three actin, several vinculin, and two integrin binding sites, are cryptic [89–91]. Cell-matrix adhesion stabilization requires talin conformational activation, however the detailed process still remains poorly understood.

Recently, molecular dynamics (MD) simulations have been proposed force-induced activation of talin and its binding sites [92, 93]. The TR contains 11 vinculin binding sites (VBSs) that bind to a hydrophobic pocket in the vinculin head (Vh) [94]. Crystal structure studies revealed that most TR VBSs are buried within the 5-helix bundles stably packed by hydrophobic interactions and result in low binding affinity for vinculin in the inactive state. Magnetic tweezers experiments have demonstrated that mechanical activation by stretching a single talin rod promoted vinculin binding, suggesting that force loading exposes the cryptic VBSs [95]. The R1–R3 domains unfolded under tensile force and vinculin binding to VBSs in those regions inhibited talin refolding after force was removed [58]. These findings suggest that talin acts as a mechanosensor.

Talin also plays a critical role in integrin-mediated inside-out signaling. Talin-induced integrin activation was observed for the first time in CHO cells [87]. Talin-integrin binding is the final step of integrin activation [96] and thought to induce integrin activation by disrupting the association between two α- and β-integrin tails [97]. A structural study showed that the F3 subdomain of the talin head forms a salt bridge (i.e. a strong ionic interaction) with the β-integrin tail. A positively charged patch in the F2 subdomain binds to the negatively charged cell membrane anchoring talin to the membrane [98]. It was found that talin binding alters the integrin β-transmembrane domain topology and disrupts the salt bridge interactions between the α- and β-tail in the transmembrane domain, resulting in conformational activation of the extracellular domain of integrin [53].

3) Vinculin

Vinculin (117 kDa) was first extracted from chicken gizzards and localized at the ends of microfilaments bundles coinciding with cell-matrix adhesion sites in 1979 [99]. Later, vinculin was found to bind to actin filaments [100] and Arp2/3 with high affinity, recruiting Arp2/3 to integrin clusters [101]. In fibroblasts, vinculin localization was correlated with fibronectin distribution on the cell surface [102]. These findings suggest that vinculin is a key player in linking the cytoskeleton to the cell-matrix adhesion site. Electron microscopy studies revealed that vinculin comprises a N-terminal globular head (Vh), a C-terminal rod tail (Vt), and a short flexible proline-rich linker in between them [103–106]. Crystal structure studies revealed that eight antiparallel α-helical bundles organized into five domains in vinculin: the first three domains form the globular head and the last two form the rod-like tail [107, 108].

Vinculin interacts with other structural adaptor molecules: the head domain contains binding sites for talin, α-catenin, α-actinin, and IpaA [109–112]; and the tail domain contains binding sites for actin, paxillin, and PIP2 [113–116]. In physiological conditions, vinculin adopts an autoinhibited conformation [117, 118] formed by the Vh-Vt intracellular interactions through oppositely charged binding sites [119]. The autoinhibition masks the talin and F-actin binding sites on the Vh and Vt respectively and promotes binding to α-actinin and VASP [120, 121]. Vinculin only gets activated at the adhesion site, dependent on actin and talin recruitment [107, 122, 123]. Crystal structure studies have shown that talin-Vh interactions induce a conformational change in the Vh resulting in Vt displacement from the head to open the structure [124]. The open conformation activates vinculin and recruits actin bundles. As mentioned in the talin section, vinculin recruitment to the adhesion site is regulated by force-induced talin activation.

Recently, the development of molecule-based tension sensors allows researchers to measure intracellular molecular tension in vivo. For example, a vinculin-tension sensor was prepared by inserting an elastic protein domain and intermolecular FRET to visualize tension-induced conformational changes. In vivo experiments using the tension sensor have demonstrated that vinculin was required to stabilize FA under force, and tension across vinculin was about 2.5 pN in stable FAs [82].

4) Actin

Actin is a highly conserved protein and found in most eukaryotic cells as an essential element of the cytoskeleton. Actin exits in a freely monomeric state (G-actin; ~42 kDa) or linear filamentous state (F-actin) within the cytoplasm and nucleus [125]. Vertebrates express three main actin isoforms, α, β and ν isoforms. G-actin consists of 375 amino acids and is organized into four subdomains, subdomain 1–4, with an ATP binding site in a cleft between subdomains 2 and 4. Variations between isoforms are differences in few amino acids localized to the proximal N-terminal [126]. The molecules of G-actin spontaneously polymerize to form F-actin, a double-stranded right-hand helical filament. F-actin is a kinetically polarized structure with a fast-growing barbed end and a slow-growing pointed end. The transition between these two states, G-actin and F-actin, is regulated by local concentration of G-actin, different ionic conditions, and a large number of actin-binding proteins. Recently, it has been shown that mechanical forces also affect depolymerization kinetics of G-actin dimer and F-actin [127].

F-actin polymerization is usually powered by ATP hydrolysis, and it occurs over three sequential phases: nucleation, elongation, and steady static phase. During the nucleation phase, few actin subunits assemble spontaneously. Due to a kinetic barrier, the initial aggregate is very slow and unstable, therefore actin nucleators (formin, Arp2/3 and spire) are often required to promote the formation of a stable nucleus [128]. In the elongation phase, the G-actin monomer rapidly associates with the barbed end of F-actin. Filament polymerization reaches an equilibrium state such that association and dissociation of G-actin monomer at both the barbed and the pointed end are balanced in the steady static phase. A group of actin-binding proteins (profilin, formin, Arp2/3, and cofilin) are directly involved in nucleation and polymerization/depolymerization of F-actin and regulate the equilibrium [129, 130]. By regulating actin-binding proteins and adaptor molecules, F-actins form a larger-scale network known as the actin cytoskeleton, which plays critical roles in maintaining the cell shape, providing mechanical support, enabling cell motility, and serving as tracks for transportation. For example, fast polymerization at the barbed end of F-actin pushes the cell membrane and mechanically links it to the adhesion site to protrude forward.

Actin acts as both a force-bearing and force-sensing structure. Tension changes mechanical properties. MD studies have demonstrated that tension on F-actin decreased its twist angle and the conformational change caused an increase of mechanical stiffness [131]. Force loading using optical tweezers exhibited force-dependent distortion of single F-actins [36]. The tension-induced structural distortion affects the affinities for actin-binding proteins, such as cofilin and myosin: F-actin severing was delayed compared to unloaded F-actin, indicating cofilin binding to F-actin was inhibited [37]; and the stretched F-actin showed high affinity for the myosin II domain [34]. These data suggest that mechano-chemical coupling is essential in regulating the actin cytoskeleton.

3. Force regulation of cell-matrix adhesions

Intuitively force would break molecular interactions. However, interestingly, cell-matrix adhesions are strengthened and matured under mechanical force-loading conditions. Two explanations have been proposed for interpretation of the force-induced strengthening and maturation: force-dependent recruitment of adaptor molecules and force-induced stabilization/activation of molecular interactions. Intensive studies have been done in the past decade and provided significant evidence supporting these hypotheses. In many cases, the two hypotheses come together rather than mutually exclude each other. In this section, we will review how force regulates cell-matrix adhesion.

1) Force recruits cell-matrix adhesion molecules

a. Force-dependent recruitment

Studies have shown that force exerted on the cell leads to recruitment of adhesion-related molecules. When force was applied on cells using collagen-coated magnetic beads, filamin A, one of the actin-binding proteins, was recruited to the cell membrane and F-actin polymerization was accelerated [132]. Force loading by an optical trap promoted vinculin and paxillin recruitment in a talin1-dependent manner [133, 134]. Also, internal force generated by actomyosin contraction regulates recruitment of adaptor molecules. Adding myosin kinase inhibitor to block contractile forces resulted in the significant decrease of vinculin accumulation around the adhesion site [133].

Two molecular mechanisms have been proposed to illustrate the force-dependent recruitment: thermodynamics-based and structural mechanism. In thermodynamics-based mechanism, mechanical force leads to a chemical potential shift or molecular density variation in the complex of anchored proteins and results in molecule recruitment to the adhesion site. In structural mechanism, mechanical force induces conformational changes of molecules in the adhesion complex, and results in exposing cryptic binding sites for their ligands, thereby recruiting adaptor molecules to the adhesion site.

i. Thermodynamics-based mechanism

Two thermodynamic mechanisms for molecular recruitment under pulling forces have been proposed [135–137] by Shemesh et al. and Safran et al. In these models, one or two-dimensional aggregate of identical molecules or molecule clusters is considered. The recruited molecules are anchored to the substrate (e.g. cell membrane) and pulling force is applied to the aggregates. The molecules in the anchored aggregate can be exchanged with those freely in the surrounding medium. Shemesh et al. proposed the exchange rate is governed by their chemical potential. If the chemical potential in the anchored aggregate is lower than that in the surrounding medium, the free molecule will be added into the anchored aggregate. In an equilibrium condition, the chemical potentials of the anchored and free molecules are equal and thereby, there is no recruitment. When the anchored molecule is pulled, the chemical potential decreases and leads to recruitment of free molecules from the medium to the anchored site (e.g. adhesion site). This model is largely consistent with experimental observations of FA behaviors under internal and/or external force loading [138, 139]. Safran and his co-workers assumed that the presence of mechanosensors in FA, which cause molecular interactions and are modified by the local density of the protein. Mechanical force deforms FA and changes local density, which in turn affects the molecular interaction. It further induces the downstream biochemical cascade, and recruitment of new molecules to FA [136, 137]. These two different mechanisms could lead to different protein recruitment patterns. A carefully designed fluorescence-recovery-after-photobleaching experiment was suggested by Safran et al. to test these two models [137].

ii. Structural model

The interaction of talin and vinculin is a typical example of force-dependent molecular interactions induced by conformational changes. The talin tail contains at least one cryptic VBS in its native state and when mechanical force is applied to cells, vinculin is recruited to the adhesion site [140]. Therefore, it was proposed that pulling force would induce the conformational change of the talin tail and exposes cryptic VBSs based on MD simulation [92]. In the simulation, VBS1 in the TR rotated ~62.0 degrees and released buried VBSs from the hydrophobic core. This hypothesis has been experimentally validated and it demonstrated that there are more than one cryptic VBS in the tail using magnetic tweezers [95]. As pulling force increased, the number of Vhs bound to the TR increased. Without applying force, only one Vh was observed to bind to the TR; with 2 pN of pulling force, the second and even the third Vh bound to the TR; and when pulling force increased to 12 pN, most TRs were bound with three Vhs. Both the MD simulation and single molecule experiment revealed the possibility that force-induced conformational change, such as exposure of cryptic binding sites, could be one of the molecular mechanisms underlying force-induced recruitment of adaptor molecules.

In thermodynamics-based model, the recruitment of molecules is a thermodynamically favorable consequence of the chemical potential shift or local density variation of FAs, without considering individual molecules in FAs. The structural model focuses on the specific molecules and their binding partners in FAs. However these mechanisms are not exclusive to each other. Both mechanisms may take place concurrently or independently in the recruitment of molecules. The structural mechanism proposed by MD simulation [92] has been validated by single-molecule experiments [95]. The thermodynamics-based mechanisms were proposed based on theoretical considerations. The possible experiments have been proposed to validate the mechanism [135, 137] and further studies are needed to investigate these models.

b. Force induces stabilization of molecular interactions

Another possible explanation for force-induced strengthening of cell-matrix adhesions is that a specific molecular interaction in the adhesion complex is strengthened by the loaded force. There are two possible molecular mechanisms: 1) in case of autoinhibited molecules, mechanical force disrupts the autoinhibited conformation and it prevents competition between autoinhibitory domains and its ligands thereby stabilizing the interaction; and 2) mechanical force directly enhances the bond lifetime of the molecular interaction.

Besides cryptic binding sites, autoinhibition is another mechanism to keep molecules inactive. NMR spectroscopy of talin showed that the C-terminal domain of the talin tail bound to the N-terminal domain of the talin head in its native state, thereby masking the integrin β-tail binding sites on the talin head. Therefore, the integrin β-tail has to compete with the talin tail to bind the talin head for further activation [141]. It suggests that modulation of the talin head-to-tail interaction affects talin affinity for the ligands and talin recruitment to the adhesion site. Similarly, vinculin also consists of head and tail domains. Study with truncation mutants of the Vh and Vt showed that the head and tail directly bind to each other. Pre-incubation of the Vh with talin significantly reduced the intramolecular head-to-tail interaction [117]. This intramolecular interaction also inhibited binding to actin [123]. These studies suggest that talin and the Vt compete to bind to the Vh, and actin and the Vh compete to bind to the Vt. Therefore, vinculin-ligand interactions and vinculin recruitment to the adhesion site can be modulated by affinity changes of the head-to-tail binding. Mechanical force is one of the critical factors which modulates the intramolecular interactions of vinculin and talin. Force applied on the serial molecular bonds in the integrin-mediated adhesion complex induces the open-conformation of talin and vinculin by disrupting the head to tail binding, thereby enhancing the affinity of talin and vinculin for their ligands, such as integrin β-tail, vinculin and actin.

In addition, force directly affects molecular interaction kinetics.

2) Force regulates molecular interaction kinetics

a. Slip, ideal, and catch bond

Force directly affects individual molecular interaction kinetics. In 1978, Bell proposed a theoretical model coined in his name. The bell model predicts that force on the bond exponentially increases off-rates of the bond, meaning force weakens molecular bonds, a phenomenon called slip bond (Fig. 2A). A slip bond behavior is intuitive and most commonly observed in biomolecular interactions, such as CD2 and CD48 [48], streptavidin and biotin [55], E-selectin and its ligands [57], integrins and their ligands [142], and antibodies and their antigens [59]. Besides a slip bond, force-independent interaction has been observed, such as the interaction between E-cadherins under certain conditions [47]. The force-independent behavior was theoretically predicted and termed an ideal bond by Dembo et al [143]. They proposed another force-dependent behavior, a catch bond (Fig. 2B). Catch bonds exhibit force-induced bond strengthening distinguished from slip bonds.

Figure 2. Force-dependent interaction kinetics under dynamic tensile force loading conditions: slip and catch bond and cyclic mechanical reinforcement.

The left column is a representative force-time trace loaded on the bond of each force-dependent behavior; the middle column is the curve of mean bond lifetime (MBL) [sec] versus loaded tensile force [pN] in A and B, MBL versus value peak force of single-cycled loading [pN] in C and MBL versus number of cycles of multi-cycled loading in D; and the right is the distribution of bond lifetimes in the middle column. The notation of I, II and III in the middle and right column represents different force ranges, relatively low, intermediate and high, respectively. A typical force range for biomolecular bonds is between 0 and 100 pN. (A) In a slip bond, when tensile force is linearly ramped and clamped at a given force (left), intuitively, mean bond lifetime decreases as force increases (middle). Distributions of bond lifetimes typically exhibit single-state dissociation (right). The lifetime distribution of longer mean lifetime shifts toward to the right, meaning lower off-rate. The off-rate is calculated from the distribution and a reciprocal of the off-rate is a theoretical mean lifetime of the interaction. (B) Typically, a catch bond follows by a slip bond, exhibiting a biphasic behavior, in which counterintuitively mean bond lifetime increases as force increases until to reach a peak and deceases as force further increases (middle) under the same path of force loading with A (left). A typical lifetime distribution is single decay dissociation (right). The distribution of bond lifetime in II force range exhibits the lowest off-rate. (C) CMR induced by mechanical preconditioning with single-cycled tensile force loading. A relatively high peak force for the molecular bond is loaded, reduced to a low force and then clamped to measure bond lifetimes (left). Interestingly, after releasing the peak force, bond lifetime does not immediately revert to the low state, which is measured under linear force-clamping (e.g. A and B), but is significantly enhanced. The degree of reinforcement is much more effective than that by a catch bond behavior. As the peak force increases, mean lifetime increases and reaches a plateau as force further increases, meaning force-history dependence (middle). Under the dynamic force loading condition, a multi-state dissociation is often observed (right). The fraction of bond population in each state changes with different cyclic mechanical preconditionings. (D) CMR induced by multi-cycled tensile loading. A combination of linear ramping to a given peak force and unloading to zero force is repetitively applied (left). As a number of cycle increases, mean lifetime increases and reaches a plateau (middle). The lifetime distribution exhibits the similar feature to the single-cycled reinforcement.

b. Possible mechanisms for catch bonds

A catch bond behavior is widely accepted as a mechanism of mechanical strengthening in biomolecular interactions, in which force prolongs bond lifetimes (i.e. decrease of off-rates). Before any experimental observations of catch bonds, Dembo et al. had proposed theoretical models of catch bonds and predicted its existence in biomolecular interactions [143]. Development of sophisticated single molecule experimental techniques, such as atomic force microscopy (AFM), biomembrane force probe (BFP), an optical and magnetic traps, allow researchers to investigate biomolecular interactions under force loading conditions at the single molecule level. Integrins, such as α₅β₁, α_Lβ₂, α₄β₁, and α_Mβ₂, exhibit catch bond behaviors in interaction with their ligands [142, 144–146]. In the last decade, more cell membrane proteins, such as cadherins, selectins, and T-cell receptors, were also found to form a catch bond with their ligands [147–152]. Recently, it was observed that the intracellular protein actin also exhibits a catch bond behavior [127]. Catch bonds are observed only over certain force and time ranges, and followed by a slip bond phase beyond that range. These catch-bond observations attract great attention due to their counterintuitive behavior of possible physiological importance. Theoreticians have proposed many models from different perspectives to explain mechanisms underlying catch bonds. In this review, we will focus on molecular and conformational mechanisms: sliding and rebinding, allosteric regulation, and deformation of energy landscape.

i. Sliding and rebinding

Selectin and PSGL-1 interaction was the first experimentally observed catch bond by single molecule experiment [153, 154]. MD simulation, Monte Carlo simulation and adhesion studies in a flow chamber system showed that force increases the flexibility of L-selectin by breaking a hydrogen bond in an interdomain hinge. The forced opening of the hinge may cause ligands to slide on the binding interface and increase the opportunity of rebinding, thereby enhancing bond lifetimes. G-actin/G-actin and G-actin/F-actin interactions also have shown catch bond behaviors. Steered MD simulation showed that pulling the end unit of F-actin or G-actin dimer resulted in formation of salt bridges, which were induced by sliding and rebinding between the interaction surface of actin subunits [127].

ii. Allosteric regulation

Springer and his co-workers proposed an allosteric regulation model to illustrate catch bond of selectin-ligand interactions. Tensile force exerted on a P-selectin-ligand complex favored an extended conformation and this allosteric change was transmitted to the binding pocket in the lectin-EGF hinge domain, thereby forming high affinity conformation and increasing bond lifetimes [155, 156]. Allosteric and sliding-rebinding models are not necessarily mutually exclusive. Another example of allosteric regulation is interactions between integrins and their ligands. As a major protein in the adhesion complex, characteristics and mechanisms of the integrin catch bond have been extensively studied. During the interaction of integrin and their ligands, force-induced conformational changes, bending and unbending, were observed by BFP experiments [60]. Also, integrin-ligand interactions exhibited multi-state dissociations consisting of short-, intermediate-, and long-lived states, which have fast, intermediate, and slow off-rates, respectively. Tensile force loading switched the short-lived state to the intermediate- or long-lived state, thereby prolonging average bond lifetimes [142].

iii. Deformation of energy landscapes

Besides the structural models, researchers have proposed energy landscape-based models of catch bonds. Evans and his colleagues employed a two-state model based on force spectrum measurements of P-selectin-PSGL-1 interaction [157]. In this model, switching between two dissociation pathways is regulated by force. The first pathway follows Bell’s rule, meaning that the off-rate of this pathway increases with increasing force, like a slip bond. The off-rate of the second pathway is force-independent and slower than that of the first. The first pathway has a deeper energy well under force-free condition and force loading tilts the energy landscape of the bond to shift from the first to the second pathway, resulting in increase of average bond lifetimes. The switching between two dissociation pathways provides a mechanism for the catch bond behavior. In a following study, a simpler, one-state two-pathway model was proposed [158]. In this model, all bonds reside in a single state, but can dissociate along one of two pathways, which have a low and high off-rate respectively. Force tilts the energy landscape and favors bond dissociation from the pathway with a low off-rate, resulting in longer bond lifetime.

c. Effects of force-loading history: cyclic mechanical reinforcement

Most single molecular studies using force microscopies have been conducted under linearly ramped, constant force conditions. In physiological conditions, cells are subject to dynamic force loading rather than static loading, and dynamic force loading over time might act as mechanical preconditioning.

Recent research has demonstrated that mechanical preconditioning by loading dynamic forces on biomolecular bonds significantly strengthened bond interactions. This mechanical reinforcement has been observed in different scales of cellular structure: at the single cell level, cyclic stretching and compression on fibroblasts using AFM induced mechanical stiffening [38]; cyclic shearing on actin-α-actinin meshes mechanically hardened the structure by accelerating α-actinin-mediated actin bundling [39]; and at the single molecule level, applying cyclic tensile forces yielded bond strengthening of both integrin/fibronectin and FimH/mannose interactions using BFP and AFM respectively [61, 159].

The left column of figure 2C and 2D are representative force-time traces of mechanical preconditioning by single-cycle or multi-cycle regimens respectively. In single molecule mechanical preconditioning, the experimental setup typically consists of a combination of bond loading and unloading with a peak force in intermediate to high range, followed by clamping at a given force for lifetime measurement. The middle and right column of figure 2C and 2D represent typical reinforcement after cyclic preconditioning: the stability of tensile force-induced high-affinity states is not only retained, but reinforced; multi-state dissociation is often observed; and different force-loading paths yield different dissociation kinetics.

Interactions between FimH and mannose formed catch bonds in linear force-clamping mode, and the bond strengths were enhanced upon addition of preloaded cycles. All bonds ruptured at significantly higher force compared to those without preloading. Allosterically stabilizing a high affinity conformation by mutations or antibody treatment exhibited the same strengthening effect, suggesting the mechanical reinforcement is due to allosteric regulation [61]. Integrin α₅β₁ formed catch bonds binding to fibronectin, when the bonds were clamped at a given force. Mechanical preconditioning by cyclic tensile loading further prolonged the bond lifetimes by switching the bond states from a short-lived to long-lived. The fraction of the long-lived population increased as peak force or number of loading cycles increased [159]. Actin also exhibited significant enhancement of bond lifetimes after cyclic mechanical preconditioning compared to catch bond lifetimes induced by constant clamping (data not published). This phenomenon is termed cyclic mechanical reinforcement (CMR) or mechano-memory.

CMR is shown to be significantly more effective in strengthening molecular interactions than the traditional catch bond. In addition to slip and catch bonds, CMR and force-history dependence are possible critical mechanotransduction mechanisms underlying force regulation of integrin-mediated adhesion. [40] [41] [71] [72]

4. Conclusions and future perspectives

We have discussed the key molecules involved in the integrin-mediated cell-matrix interaction and different molecular mechanisms underlying force regulation of integrin and key adaptor molecules. In addition to the traditional force-dependent behaviors, such as a catch, slip and ideal bond, we highlighted CMR, which can mimic physiological conditions more closely. CMR has shown significant effective bond strengthening in some systems and independent of catch or slip behaviors. Therefore, other adhesion molecules also need to be investigated under dynamic force loading to elucidate mechanisms underlying mechanochemical coupling of CMR.

As briefly mentioned in the introduction, recently, cell-matrix adhesions studies have often been conducted on 3D substrates. This has been accompanied by the development of microfluidic-based cell culture systems [70] and 3D ECM-mimicking scaffolds [69]. Although 3D matrix adhesions are comprised of many of the same molecules of 2D cell-matrix adhesions, cell adhesions and behaviors are distinguished from the interactions on the 2D substrates due to inherent complexity of 3D matrices. 2D cell-matrix interactions mostly depend on substrate stiffness and coated ECM molecules, yet in 3D scaffolds, cells sense porosity (pore size and density), topography of ECM ligands or scaffold matrix materials, and isotropicity of material compositions and mechanical properties [67, 68, 160, 161]. Therefore, mechanosensing and mechanotransduction of 3D adhesion differ from traditional 2D adhesion, and in vitro and in vivo studies to elucidate the different characteristics and mechanisms are just beginning.

Force regulation of transmembrane receptors and their ligands has been intensively studied at the single-cell and single–molecule level using AFM, BFP, optical tweezers, and flow chamber systems. However, studies of intracellular force regulation are relatively limited due to the technical hurdles of live cell experiments. A full understanding of the bidirectional signaling cascades requires more attention to intracellular adaptor proteins and cytoskeletons. Single molecule force spectroscopy combined with fluorescence imaging technique or optical microscopy has shown great promise in investigating individual biomolecules both in vivo and in vitro. Cutting edge imaging techniques (e.g. super-resolution-PALM and TIRF-SIM) [162, 163], molecule-based force sensors (e.g. FRET tension sensors) [82, 164] and combinations of single-molecule force spectroscopy and imaging systems [165] are now providing opportunities to further explore the challenging questions regarding mechanosensing and mechanotransduction.

Highlights.

• General review of integrin-mediated cell-matrix adhesion

• Force regulation of integrin activation and key adaptor proteins

• Molecular mechanisms for mechanochemical coupling in force regulation