Cellular force‐sensing through actin filaments

The actin cytoskeleton orchestrates cell mechanics and facilitates the physical integration of cells into tissues, while tissue‐scale forces and extracellular rigidity in turn govern cell behaviour. Here, we discuss recent evidence that actin filaments (F‐actin), the core building blocks of the actin cytoskeleton, also serve as molecular force sensors. We delineate two classes of proteins, which interpret forces applied to F‐actin through enhanced binding interactions: ‘mechanically tuned’ canonical actin‐binding proteins, whose constitutive F‐actin affinity is increased by force, and ‘mechanically switched’ proteins, which bind F‐actin only in the presence of force. We speculate mechanically tuned and mechanically switched actin‐binding proteins are biophysically suitable for coordinating cytoskeletal force‐feedback and mechanical signalling processes, respectively. Finally, we discuss potential mechanisms mediating force‐activated actin binding, which likely occurs both through the structural remodelling of F‐actin itself and geometric rearrangements of higher‐order actin networks. Understanding the interplay of these mechanisms will enable the dissection of force‐activated actin binding's specific biological functions.


Introduction
The ability of cells to perceive and respond to mechanical cues and forces in their environments ('mechanosensing') plays a critical role in numerous physiological processes [1,2], and the dysfunction of cellular force-sensing has correspondingly been implicated in disease states [3], including cancer [4] and developmental disorders [3]. Forces are converted into biochemical pathways that modulate cellular physiology and behaviour (mechanical signal transduction or 'mechanotransduction') through mechanisms that largely remain poorly understood, limiting dissection of mechanosensation in vivo for fundamental biological discovery and the development of therapeutics targeting mechanotransduction. The actin cytoskeleton, a cell-spanning physical network of dynamic actin filaments (F-actin), myosin motor proteins and dozens of associated actin-binding proteins (ABPs), governs cell mechanics. Recent studies have highlighted cytoskeletal force-feedback mechanisms mediating cellular mechanosensation [5][6][7][8] and cytoskeletonassociated proteins which serve as core components of mechanotransduction pathways [9][10][11]. In this Viewpoint, we discuss the emerging concept that F-actin itself can serve as an important mechanical sensor element ('mechanosensor') through direct force-regulated binding interactions [12][13][14][15], primarily focusing on cases where forces applied to F-actin itself enhances the binding of ABPs (which we term 'force-activated binding').
In recent years, force-activated F-actin-binding interactions are being discovered at an accelerating pace, likely driven by the development of sensitive biophysical assays for identifying them. We view these interactions as falling into two biophysically distinct categories that we speculate are intrinsically suited for different biological functions. Force has been reported to moderately enhance the F-actin binding of several canonical ABPs which display a constitutive affinity for F-actin, a phenomenon we term 'mechanical tuning' (Fig. 1A). We propose mechanical tuning is wellsuited for cellular mechanosensing functions requiring force-feedback to the cytoskeleton itself, such as modulation of cell adhesion or cell migration. Very recently, force applied to F-actin has also been shown to activate actin binding by certain proteins whose binding is undetectable in the absence of force, which we term 'mechanical switching' (Fig. 1B). Mechanical switching, which has thus far only been identified in proteins from the LIN-11, Isl1 and MEC-3 (LIM) domain superfamily, could be broadly suitable for initiating mechanotransduction, with potential downstream outputs beyond cytoskeletal regulation, such as controlling gene expression.
Finally, we discuss potential biophysical and structural mechanisms of force-activated F-actin binding (Fig. 2). F-actin has been reported to exhibit structural polymorphism [16,17], which could be modulated by forces to mediate force-activated actin binding at the level of individual actin filaments ( Fig. 2B-C). However, at the cellular scale, forces will also impact the geometric properties of actin networks (such as the spacing between filaments and their orientations), which could modulate engagement by the many ABPs that contact multiple filaments simultaneously, notably crosslinking proteins (Fig. 2D). Dissecting the interplay between these nonexclusive single-filament and network geometry models is of outstanding importance, and we discuss the development of structural approaches that may render this feasible.
Mechanically tuned ABP binding facilitates forcefeedback to cell mechanics Many of the more than 150 identified ABPs function in the assembly, disassembly, and regulation of cytoskeletal networks [18]. While the fundamental actin-binding interactions of these proteins have been biochemically well-studied for decades, recent efforts using functional biophysical approaches have also frequently uncovered mechanical tuning. This highlights the capacity for mechanosensation to occur at the level of cytoskeletal dynamics via fine adjustments to the binding affinities and kinetics of ABPs that already have the capacity to bind F-actin in the absence of force, providing force-feedback to biochemical interactions that determine cytoskeletal architectures controlling cell mechanics.
Cells form mechanical interfaces between their intracellular actin cytoskeleton and their local extracellular environments through plasma membranelocalized, integrin-mediated cell-extracellular-matrix focal adhesions [19] and cadherin-mediated cell-cell junctions (adherens junctions) [20]. Adhesions are linked to contractile actin-myosin cables known as stress fibres through organized layers of bridging molecules, including many ABPs, which undergo conformational [21][22][23][24][25][26][27][28][29][30] and compositional [9][10][11]31,32] remodelling in response to force. This facilitates mechanically enhanced recruitment of ABPs to adhesions and the reinforcement of cytoskeletonmembrane receptor linkages. In addition to mechanically integrating with tissues through the actinmyosin-adhesion system, cells generate force and movement through the polymerization dynamics of branched actin networks, which power cell movement [33], propel the internal dynamics of organelles [34] and facilitate membrane remodelling processes including endocytosis [35]. Recent studies have revealed that the assembly dynamics, force output capacity, and nanoscale network architecture of branched actin networks are modulated by resistive loads in vitro [36] and in cells [5]. This may contribute to mechanosensing behaviours such as polarity maintenance [37] and navigation along rigidity gradients ('durotaxis') [38].
Several essential adhesion proteins that bind F-actin through conserved 'talin-HIP1/R/Sla2p actin-tethering C-terminal homology' (THATCH) domains, notably talin, a-catenin and vinculin, feature mechanical tuning. Talin [21,22] and a-catenin [23] serve as bridges between transmembrane receptors and stress fibres in focal adhesions and adherens junctions, respectively. These connections are reinforced by vinculin, which can bind either talin or a-catenin while simultaneously binding F-actin, thereby strengthening both types of adhesions. Optical tweezer experiments have revealed all three of these proteins form catch bonds (defined as a bond whose lifetime is enhanced when force is applied across the binding interface after bond formation) with F-actin in the presence of forces on the order of 10 pN [39][40][41][42]. Interestingly, these catch bonds all feature a directional asymmetry defined by the polarity of the actin filament, wherein forces directed towards the minus ('pointed') end produce substantially greater stabilization. This could feasibly play a role in cell polarization by orienting adhesion and stress fibre assembly [39]. Directional catch bonding may be encoded by the structural topology of THATCH domains, as cryogenic electron microscopy (cryo-EM) structures of vinculin and a-catenin bound to F-actin revealed similar partial unfolding transitions when they engage the filament that could be stabilized by force [27][28][29][30].
Mechanically tuned force-activated F-actin binding, where tension solely across F-actin promotes ABP engagement, is a mechanistically distinct phenomenon from catch bonding. Using a combined optical tweezer and fluorescence microscopy assay, we recently reported that a-catenin also displays force-activated actin binding in the presence of lower forces, approximately 1 pN across the actin filament [29], the force regime produced by single myosin motors [43]. Consistently, forces exerted by myosin motors in a reconstituted system could also enhance a-catenin binding. Interestingly, vinculin did not display force-activated F-actin binding, suggesting this property varies between THATCH domains, likely mediated by differential flexible elements that engage distinct binding surfaces on F-actin [29]. We speculate that low force (1 pN)-activated binding promotes initial attachment between a-catenin and stress fibres, while high force (10 pN) catch bonding [42] subsequently reinforces adhesion (Fig. 1A). As vinculin also serves a reinforcing function, its lack of force-activated actin binding is consistent with this model. Since talin fulfils a similar function to a-catenin, forming initial F-actin attachments in focal adhesions, it will be informative to probe whether it also possesses force-activated actinbinding activity. Key ABPs mediating the assembly and disassembly of actin networks have also been shown to feature mechanical tuning. The first ABP reported to feature mechanical regulation was actin-depolymerizing factor (ADF)/cofilin, where tensile forces applied to single actin filaments by optical tweezers inhibited F-actin binding and slowed filament severing [44]. However, two recent studies using microfluidics to apply fluid flow forces found that tension has no effect on cofilin binding or filament severing [45,46]. Instead, these studies reported that filament bending and torque accelerate cofilin-mediated filament severing at the boundary between bare and cofilin-decorated F-actin segments, despite these mechanical perturbations also having no impact on the kinetics of cofilin binding [45,46]. These more recent findings are consistent with a theoretical study proposing that stretching filaments disperses strain uniformly over subunit interfaces with no detectable partial rupture, whereas bending or  twisting filaments imposes nonuniform interface strain and partial rupture, accelerating filament fragmentation [47]. Formins, a family of actin polymerases that mediate nucleation and enhance the elongation rate of individual actin filaments by processively associating with growing plus ('barbed') ends, are responsive to both tension (0-10 pN) and torque as demonstrated by a variety of biophysical methods (flow-based systems, optical traps and magnetic tweezers) [48][49][50][51][52][53]. Distinct modes of mechanical regulation have been reported for different formin family members, with tension either accelerating or slowing formin-mediated actin elongation, depending on the formin's identity and the presence of the soluble globular actin (G-actin) binding co-factor profilin [48][49][50][51][52][53]. Additionally, singlemolecule fluorescence polarization imaging of labelled F-actin elongating from surface-immobilized formin revealed periodic oscillations, suggesting a relative rotation between formin and the polymerizing actin filament [54]. If this rotation were to be constrained, torsion would accumulate in the filament and in turn tune the activity of formin, providing force-feedback to actin polymerization.
The Arp2/3 complex, the sole branched F-actin nucleator, nucleates daughter filaments at a stereotypic 70°angle from the side of pre-existing filaments, remaining associated with both filaments to form a branch junction. A flow-based study revealed a weak bias for Arp2/3 to form branches from the convex side of mechanically bent filaments [55]. A recent microfluidics-based study further demonstrated a substantial increase in the rate of Arp2/3 branch dissociation ('debranching') in the presence of shear forces [56]. The ATPase activity of the Arp2/3 complex produces two distinct biochemical states with different sensitivity to force. The 'old/weak' ADP-bound Arp2/ 3 complex is 20 times more sensitive to force than its 'young/strong' ADP-P i -bound counterpart, leading to accelerated debranching of aged branches under shear flow [56]. Since branches are initiated in close proximity to the plasma membrane by nucleation-promoting factors, the higher debranching force threshold of young branches facilitates their pushing against the plasma membrane more efficiently, while aged branches are prone to dissociate, facilitating replenishment of the G-actin pool [56]. This coordination between biochemical and mechanical regulation of Arp2/3 is thus likely to function as a timer that governs actin network dynamics, a function, which has also been ascribed to F-actin nucleotide states traversed during ATP hydrolysis and phosphate release by actin protomers upon polymerization [57].
As an ever-expanding 'parts list' of mechanically tuned ABPs continues to be identified, the precise functions of mechanical tuning generally remain speculative. We expect probing how the collective action of these nanoscale-binding interactions mediates cellular mechanosensing through actin networks at the micron scale will be a fruitful topic in the coming years.

Mechanically switched ABPs as initiators of mechanotransduction
Recently, we and Winkelman et al. reported a categorically distinct form of force-activated actin binding. Tandem LIM domain-containing proteins from the FHL, paxillin and zyxin families do not detectably bind F-actin in the absence of force, yet they directly engage filaments in the presence of myosin-generated forces in reconstitution assays [7,8], binding in 'patches' along local regions within individual filaments (Fig. 1B). Within the cytoplasm, proteomics studies have revealed many LIM proteins to be enriched in stress fibres and focal adhesions in the presence of myosin activity [9][10][11]58]. Nuclear localization of several superfamily members has also independently been reported, where they have been suggested to regulate gene expression as transcriptional coactivators [59]. The dual activities of LIM proteins in force-dependent cytoskeletal localization via mechanically switched actin binding and in gene expression regulation suggest they could function as relay factors between the cytoskeleton and nucleus to mediate mechanotransduction (Fig. 1B) [59] in a manner conceptually analogous to the well-established mechanotransduction effectors YAP/TAZ [60]. Furthermore, subfamilies contain multiple classes of effector domains fused to tandem LIM domains, suggesting that LIM proteins could serve as modular force transducers in diverse signalling networks. The switch-like binding of LIM proteins to F-actin solely in the presence of force is reminiscent of the initiating processes in canonical signal transduction, e.g. the binding of extracellular ligands to activate cell surface receptors, and it is biophysically well-suited for initiating mechanotransduction.
Zyxin, the first LIM protein identified to feature force-dependent cytoskeletal localization, repairs mechanically damaged stress fibres to maintain cytoskeletal integrity and cellular rigidity [6,61,62], a function that is prominent in mechanically active tissues [63,64]. Through its three C-terminal tandem LIM domains, zyxin localizes to stress fibre strain sites, mechanically evoked tears which spontaneously emerge in highly contractile cells, within seconds of their appearance [6,65]. There it initiates stress fibre repair through the rapid recruitment and downstream spatiotemporal coordination of multiple effector proteins, a function similar to scaffolding proteins in canonical signalling cascades. Zyxin's N-terminus contains one binding site for the dimeric actin crosslinker a-actinin [66,67] and four binding sites for the tetrameric actin polymerization factor Ena/VASP [68,69]. While direct binding interactions between these proteins have been observed biochemically, the precise mechanism for coordinating SF repair remains unknown [58,[66][67][68][69].
A recent study reported that VASP promotes actin assembly by forming multivalent clusters on actin filaments with lamellipodin, a partner which features tandem VASP-binding sites similar to zyxin's [70]. Given the potential for multivalent interaction between zyxin and its actin regulating partners, zyxin may employ a biomolecular condensation mechanism to mediate actin repair, similar to other actin regulatory pathways coupled to canonical upstream biochemical signalling [71]. Consistently, the zyxin paralog LIMD1 was reported to form condensates in vitro and in cells featuring active contractility [72]. Whether zyxin's force-activated actin binding initiates the formation of effector condensates to orchestrate stress fibre repair and how the activities of zyxin, VASP and a-actinin are coordinated are important open questions. More recently, evidence implicating LIM proteins in mechanically regulated gene expression has begun to appear. Four and a half LIM domain 2 (FHL2), a transcriptional co-activator dysregulated in many cancers [73][74][75][76], was shown to translocate into the nucleus in soft microenvironments, mechanically upregulating the expression of the cell-cycle inhibitor p21 (Fig. 1B) [77]. We found that FHL2's binding to tensed F-actin retains it in the cytoplasm in stiff microenvironments, whereas an FHL2 mutant deficient in tensed F-actin binding constitutively localizes to the nucleus, regardless of the rigidity of the microenvironment [7]. The rigiditydependent nuclear shuttling of FHL2 suggests the potential for a direct link between upstream forceactivated actin binding by LIM proteins and downstream gene regulation (Fig. 1B). Other LIM proteins featuring mechanically switched actin binding, including zyxin [78,79], paxillin [80] and HIC5 [81,82], have been reported to translocate into the nucleus under conditions not explicitly linked to microenvironment rigidity, suggesting context-dependent nuclear localization may be a conserved feature of the superfamily. Zyxin was reported to bind the transcriptional regulator cell division cycle and apoptosis regulator protein 1 (CCAR1, also known as CARP-1/DIS) through its LIM domains in response to UV irradiation, thereby promoting apoptosis [83]. An emerging theme in LIM proteinregulated gene expression reported to date is inhibition of cell division and promotion of cell death, the inverse of YAP's promotion of cell growth and survival [84], consistent with the anticorrelated rigidity-dependent nuclear localization of YAP and FHL2.
The functional significance of mechanically regulated gene expression by LIM proteins remains a largely unexplored vista with many open questions. FHL2 lacks canonical NLS and NES sequences, and standard pharmacological inhibitors of nuclear transport do not impact its nuclear localization [77]. While FHL2's small size (32.2 kDa) could potentially allow it to passively diffuse through the nuclear pore, other larger LIM proteins, such as zyxin (61.3 kDa), which also lack NLS, are likely to employ noncanonical nuclear import mechanisms to enter the nucleus. Defining the molecular mechanisms for LIM proteins' nuclear translocation is therefore of outstanding interest. Furthermore, although multiple genes, including IL6, IL8, TGF-b1 and cyclin D1 [74,[85][86][87][88], have been reported as FHL2 targets, p21 is the only gene identified thus far to be mechanically regulated by FHL2 [77]. Moreover, most LIM proteins, including FHL2, have multiple paralogs, often featuring partially overlapping tissue-specific expression patterns, with the potential for partial functional redundancy [59]. Future efforts to systematically identify the mechanically regulated target genes of LIM proteins, guided by dissection of the mechanisms governing their nuclear localization, will pave the way towards assessing their precise functions in mechanotransduction in vivo.

Potential structural mechanisms for forceactivated actin binding
Despite recent progress in identifying force-activated ABPs, the structural mechanisms mediating their detection of mechanical loads along F-actin remain unknown. Ideal actin filaments are helices composed of two identical strands ( Fig. 2A). Along the filament axis, actin protomers are periodically arranged around a tight left-handed helix (the 'short-pitch helix'), which alternates between the strands. The angular twist of the short-pitch helix is slightly offset from 180°, producing a precession of the strands around one another along a right-handed helical path (the 'long-pitch helix') that switches sides at 'crossovers' approximately every 13 subunits (Fig. 2A). The structure of the actin protomer itself consists of four subdomains ( Fig. 2A), which are known to undergo rearrangements coupled to actin's ATP nucleotide binding and hydrolysis, polymerization dynamics and ABP binding [89]. We and others have speculated that forces also modulate the conformation of F-actin, which can be 'read' through direct binding interactions, the most parsimonious explanation for force-activated binding to single filaments [7,8,12]. In principle, this could occur through rearrangements of actin protomers relative to one another in the helical lattice (an 'architectural remodelling' model, Fig. 2B), distortion of the structure of individual protomers (a 'subunit deformation' model, Fig. 2C) or both, as these phenomena are likely to be coupled due to steric considerations. Most ABPs engage interfaces on F-actin spanning at least two protomers, a binding mode that is particularly well-suited for detecting architectural remodelling [89].
At the steady state, F-actin has been observed to feature structural polymorphism, with reports of filaments featuring varying short-pitch twists [90] and crossover lengths [91] in negative stain EM, spontaneous transitions of individual actin protomers between low and high single-molecule FRET states [17], and co-existing protomer subdomain arrangements in frozen-hydrated actin filaments observed by cryo-EM in thick ice films [12,16]. This structural polymorphism likely reflects an energy landscape featuring multiple wells representing F-actin conformational states, some of which are separated by low enough energy barriers that filaments may spontaneously shift between them through thermal fluctuations. Mechanical forces are expected to tilt this energy landscape, potentially stabilizing F-actin conformations that are low occupancy in the absence of force [92] and/or evoking conformations that are otherwise inaccessible, which could be detected by ABPs. Consistent with this hypothesis, ABPs featuring tandem calponin homology domains have been reported to preferentially bind F-actin in the presence of stabilizing drugs and other ABPs, suggesting that subtle alterations of F-actin conformation can modulate ABP engagement [93]. Additionally, 'mechanical allostery' [94], wherein the force applied to one region of an actin subunit induces structural changes in another region of the same subunit or propagates into structural rearrangements of neighbouring subunits, could impact ABP binding in a manner similar to traditional allostery mediated through binding interactions. Supporting the feasibility of this mechanism, several ABPs have been suggested to be allosteric regulators of F-actin, including myosin, drebrin, VASP, a-actinin, and filamin, as their sparse binding alters the engagement of other ABPs, actin polymerization kinetics, and filament mechanics [93,95]. Cofilin provides the most striking example of this phenomenon, as its binding evokes substantial alteration of F-actin helical twist and actin subunit flattening [96,97], changes that are transmitted to neighbouring actin protomers at boundaries between bare and cofilin-decorated filament segments, leading to cooperative cofilin binding and filament severing at these boundaries [96,97]. Whether mechanical allostery indeed plays a role in mechanical tuning or switching of ABP engagement is an important topic for future studies.
Coarse-grained and mesoscopic molecular modelling efforts have predicted that forces applied to F-actin induce strain [98] and plastic deformation [47] at the nanoscale, producing architectural remodelling, which can control mechanically tuned ABP binding. Bendtwist [99] and stretch-twist [100] coupling have been observed in simulations, suggesting that intersubunit rotation (i.e. short-pitch helical twist) is a feature of the filament lattice particularly susceptible to mechanical regulation. Cryo-EM structures of actin filaments saturated with cofilin reveal substantially altered twists [101], presumably stabilized by binding energy. This suggests that cofilin engagement could potentially be upregulated and downregulated by different force regimes, depending on whether they favour or disfavour its preferred F-actin twist. A recent theoretical study proposes that filament twisting enhances the dissociation of cofilin [98], with overtwisting displaying a more pronounced effect than under twisting. This twisting-accelerated cofilin dissociation is consistent with the observation that formin-generated torsional forces protect F-actin from severing by cofilin [102]. Conversely bend-twist coupling could explain cofilin's accelerated severing of bent filaments [45], as the anticipated local asymmetry in twist at the bend site could provoke nonuniform cofilin binding across strands, thereby destabilizing the filament. Bend-twist coupling is also a feasible mechanism for explaining Arp2/3's preference for the convex side of curved filaments [55].
Evidence for subunit deformations in force-activated actin binding, on the other hand, remains circumstantial. Our group recently reported that the flexible Cterminal extension (CTE) of a-catenin engages a distinct site on the surface of F-actin from that of vinculin, and chimeric swaps implicated a-catenin's CTE as the key determinant of mechanically tuned binding [29]. As the actin interface engaged by this extended peptide is confined to a single protomer, it is possible that it preferentially binds a force-evoked subunit conformation. However, the CTE is also positioned at the interface between longitudinally adjacent actin-binding domains in our cryo-EM structure, which was determined in the presence of saturating a-catenin. Thus, it may also mediate communication between a-catenin binding sites along the filament that are mechanically regulated through architectural remodelling, a model which is not mutually exclusive with the detection of subunit deformations. Architectural remodelling of F-actin is also likely to be a feature of mechanically switched binding by LIM proteins, although, in this case, the existence of Factin conformations specifically evoked by myosin forces is supported by available indirect evidence. All mechanically switched LIM proteins that we and Winkelman et al. identified contain at least three tandem LIM domains [7,8], each featuring a conserved phenylalanine residue at the same position. Combinatorial mutagenesis of these residues to alanine in FHL2 results in a graded reduction in stress fibre localization in cells, and mutating all four in FHL3 completely disrupts force-activated F-actin binding in vitro, suggesting an avidity-based mechanism [7]. Correspondingly, Winkelman et al. found that altering the lengths of inter-LIM domain linkers disrupted mechanically switched F-actin binding, indicating that these linkers could serve as rulers to measure forceinduced architectural rearrangements [8]. We furthermore observed that LIM proteins bind by accumulating in elongated patches, which can grow along individual filaments to more than a micron in length [7]. This suggests a specific conformation that licences LIM protein binding can be evoked by myosingenerated forces in continuous domains spanning hundreds of protomers, co-existing with regions featuring a nonlicencing conformation within the same filament.
Experimentally establishing the structural mechanisms of force-activated actin binding will require technical barriers to be overcome. Fluid forces exerted by blotting and surface tension in thin films associated with the preparation of cryo-EM specimens have been suggested to impact F-actin structure [12], and it may be possible to productively harness these forces experimentally. Indeed, our lab has very recently used singleparticle cryo-EM to interrogate F-actin structural transitions evoked by bending forces under such standard conditions [103]. By developing a machine learningbased particle picking approach, we detected and characterized bent filament segments featuring a continuous range of curvatures up to 8 lm À1 [103]. Structural analysis revealed helical lattice architectural remodelling through bend-twist coupling as anticipated theoretically [99], coupled with protomer deformations characterized by shearing around actin's nucleotide cleft that was modulated by actin nucleotide state [103]. It is likely that studies of ABP-F-actin interactions using this approach will be feasible. Myosin motors are also potentially suitable as force-generators compatible with cryo-EM experiments, although their stochasticity will inevitably introduce heterogeneity which is refractory to high-resolution structure determination. Nevertheless, continuous progress in classification algorithms, particularly the recent introduction of approaches for handling continuous structural variability [104,105], which were successfully employed to characterize F-actin bending [103], may render this approach tractable.
Linking force-evoked conformational transitions in F-actin to functional biophysical studies of forceactivated ABP binding will further require associating particular conformations with specific force thresholds. This is of particular importance for dissecting the mechanisms of ABPs, which are sensitive to low force regimes (e.g. a-catenin [29]), as it is unclear whether 1 pN of force can evoke unique F-actin conformations versus modulating the steady-state landscape of thermal fluctuations. Theoretically, the magnitude of force experienced by a bending filament can be estimated from its curvature, as thermal fluctuations observed with fluorescence microscopy at the micron scale have been well-fit by semiflexible polymer models to calculate F-actin's flexural rigidity (7.3 9 10 À26 Nm 2 for phalloidin-stabilized F-actin [106]). This approach has been used to interpret in situ cryo-electron tomography (cryo-ET) studies of the actin cytoskeleton at podosomes [107] and clathrin-mediated endocytic vesicles [108], revealing bent actin filaments storing elastic energy estimated to be approximately 150 pNÁnm on average in podosomes [107]. However, it is unclear whether such polymer models of F-actin are applicable at the 10s of nanometres length scale (the span of a few subunits) relevant to the mechanical regulation of filament conformation. In the longer term, introducing molecular force reporters compatible with structural experiments are likely to be necessary for decisively dissecting force-structure relationships, a substantial technical challenge.
While experimental approaches for associating particular force regimes with specific conformational transitions remain to be developed, molecular dynamics (MD) simulations are a valuable tool for probing plausible mechanisms [97,[109][110][111][112][113][114][115][116]. Coarse-grained simulations of filaments composed of 13 subunits have successfully reproduced F-actin's persistence length, torsional stiffness, and the broad distribution of filament twist angles observed in experiments [111]. Another coarse-grained model has further suggested that compressive strains lower F-actin's ATP hydrolysis rate and enhance its P i release rate, while tensile strains have the opposite effects [113], predictions that are experimentally testable. Recently, all-atom MD simulations have proven feasible for actin filaments composed of tens of subunits and interacting ABPs, albeit at the limited time (< 100 ns) and length scales (< 50 nm). Simulations of bare F-actin / cofilin-bound F-actin boundaries revealed actin subunits adopting structures intermediate between those of bare and cofilin-bound states, producing compromised intersubunit contacts that lead to filament fragmentation [97]. This study also predicted the structure of the mechanically labile bare F-actin / cofilin-bound F-actin interface, which has yet to be experimentally resolved [97]. A major strength of MD simulations is their quantitative power, as relative force-structure relationships can be scrutinized by systematically varying the force applied in silico [117], although caveats remain about interpreting absolute force magnitudes as discussed above. As computation continues to develop rapidly, studies that combine anticipated advances in experimental structure determination in the presence of forces with MD simulations are likely to provide detailed insights into the mechanochemical mechanisms of force-activated ABP binding.
Beyond the conformation of individual filaments, an additional, nonexclusive layer of mechanical regulation occurs at the level of actin networks (Fig. 2D). In the cell, actin filaments are essentially always embedded in many filament assemblies whose geometry is controlled by crosslinking proteins. These proteins contain multiple actin-binding domains within the same molecule, which they employ to bridge multiple filaments, thereby controlling their relative orientations and spacings. In turn, the structural properties of crosslinkers render their F-actin-binding interactions intrinsically sensitive to network geometry (Fig. 2D). Electron microscopy and atomic force microscopy (AFM) studies have revealed that branched networks formed in the presence of nN mechanical loads feature a broadened range of angles, increased filament density and tighter filament packing, features that could alter the profile of ABP engagement throughout the network [5,36]. Several crosslinkers and filament-forming nonmuscle myosin II motor proteins were furthermore found to 'mechanoaccumulate' in cells in response to micropipette aspiration [118], a phenomenon that could be explained by network geometry modulation. LIM proteins also display more robust localization to stress fibres in cells than individual actin filaments in vitro [6,7], suggesting that tandem LIM domains may also be able to engage multiple filaments simultaneously and thereby detect stress fibre geometry. However, truncated myosin II motor domains unable to oligomerize also showed similar enhanced localization in mechanically stimulated cells [119], which cannot be explained by the simultaneous engagement of multiple filaments. Nevertheless, geometric control over the local concentration of available binding sites could result in force-activated actin binding by monomeric ABPs at the network level.
In the longer term, it will be necessary to dissect the structural interplay between mechanical regulation of the conformation of single filaments and the networks in which they are embedded. Recent cryo-ET studies of muscle fibres have proven highly successful in visualizing the domain-level structural organization of crosslinkers in situ [120,121], an approach that could be applied both in cells and reconstituted in vitro preparations where forces can be experimentally controlled. Very recently, we have also introduced a machine learning-based pipeline for reconstructing crosslinkers bridging filaments using single-particle cryo-EM [122], which could be adapted to mechanically active networks. We thus believe the coming years are likely to see major progress in uncovering the structural basis of force-activated actin binding.

Conclusion
The continued identification of mechanically regulated actin-binding proteins suggests F-actin is likely to serve as a physiologically relevant mechanosensor. While the biophysical phenomenon of force-activated actin binding has now widely been reported, the biological functions of both mechanical tuning and mechanical switching in ABPs remain largely speculative. Beyond providing fundamental insights, establishing the structural mechanisms of force-activated actin binding will be key for discriminating force-detecting elements in ABPs that can be experimentally manipulated. This will enable precision dissection of forceactivated actin-binding activity in cells and model organisms, a first step towards assessing its relevance in health and disease and long-term potential as a therapeutic target.