Differential Effect of Actomyosin Relaxation on the Dynamic Properties of Focal Adhesion Proteins

Treatment of cultured cells with inhibitors of actomyosin contractility induces rapid deterioration of stress fibers, and disassembly of the associated focal adhesions (FAs). In this study, we show that treatment with the Rho kinase inhibitor Y-27632, which blocks actomyosin contractility, induces disarray in the FA-associated actin bundles, followed by the differential dissociation of eight FA components from the adhesion sites. Live-cell microscopy indicated that the drug triggers rapid dissociation of VASP and zyxin from FAs (τ values of 7-8 min), followed by talin, paxillin and ILK (τ ~16 min), and then by FAK, vinculin and kindlin-2 (τ = 25-28 min). Examination of the molecular kinetics of the various FA constituents, using Fluorescence Recovery After Photobleaching (FRAP), in the absence of or following short-term treatment with the drug, revealed major changes in the kon and koff values of the different proteins tested, which are in close agreement with their differential dissociation rates from the adhesion sites. These findings indicate that mechanical, actomyosin-generated forces differentially regulate the molecular kinetics of individual FA-associated molecules, and thereby modulate FA composition and stability.


Introduction
Integrin-mediated cell-extracellular matrix (ECM) adhesions play key roles in tissue formation and morphogenesis, and in the generation and transmission of adhesion-dependent signals [1][2][3]. Recent studies indicate that the integrin family of matrix adhesions is highly heterogeneous, displaying conspicuous variations in overall structure, subcellular distribution, and specific molecular composition [1,4,5]. Consequently, different adhesions display diverse functional properties, including selective binding to the ECM, and a differential capacity to sense its mechanical properties and to actively remodel it [6][7][8].
Live-cell microscopy of cells tagged with specific focal adhesion (FA) components demonstrated that integrin adhesions are dynamic structures that undergo major morphological transformation during their formation and maturation, initially forming nascent adhesions, mainly along the leading lamellae, and later expanding into large focal adhesions, typically several square micrometers in size, that are associated with actomyosin-rich stress fibers [9][10][11][12]. Depending on the cell type and ECM properties, these FAs can induce ECM fibrillogenesis and transform into fibrillar adhesions [13,14]. These transformations were shown to be highly mechanosensitive processes; thus, the formation and stability of FAs depend on contractile forces generated by the associated actin cytoskeleton. Inhibition of these contractions (e.g., by Rho-kinase or specific actomyosin inhibitors) leads to FA dissociation, and to disruption of the associated stress fibers [15][16][17][18][19]. At the same time, it was shown that myosin IIindependent integrin adhesions also exist, and their properties were characterized [20,21].
The molecular composition and nano-architecture of FAs are believed to play key roles in regulating the diverse scaffolding and signaling activities of cells; yet the molecular mechanisms underlying these processes are still largely unclear. Attempts to characterize the molecular composition of integrin adhesions revealed a rich variety of "adhesome" molecules (over 200 components known at present) that collectively perform and regulate the various scaffolding and signaling functions of these adhesions [10,22]. Among them are membrane receptors, adaptor molecules, and cytoskeleton-associated proteins, which collectively bridge between the ECM and the Factin cytoskeleton. Additional regulatory molecules, including diverse kinases, phosphatases, and G-protein regulators, participate in both modulation of the adhesions, and in integrinmediated signaling processes that affect cell behavior and fate [10,22].
In this study, we tested the hypothesis that variations in the mechanical force applied to FAs by means of the cellular contractile machinery, differentially affect the binding and dissociation of various adhesome components and, hence, modulate FA composition, molecular architecture and, eventually, function. Specifically, we examined how inhibition of actomyosin contractility affects the association of different FA components with the adhesion sites, by quantifying temporal changes in the levels and organization of eight different adhesome proteins, following treatment with the Rho-kinase inhibitor Y-27632. We demonstrate here that the components tested dissociate from FAs at differing rates, accompanied by major structural changes in the FA-associated cytoskeleton, as revealed by cryo-electron tomography. We further show that the differential dissociation of the tested proteins can be attributed to specific changes in their k on and k off , values, induced by the drug. Calculation of the expected dissociation rate of each molecule from FAs in treated cells, based on these kinetic changes, accurately fit the dissociation values measured by Fluorescence Recovery After Photobleaching (FRAP) microscopy. Furthermore, we show that Y-27632treated cells can still form and maintain modified integrin adhesions with the underlying substrate. Notably, such adhesions, formed around the cell center, predominantly contain ILK, kindlin-2 and paxillin, but little or no zyxin, VASP, talin, vinculin and FAK, while those formed at the cell periphery are similar to those of untreated ones, except that they are largely devoid of VASP and zyxin.

Immunofluorescence staining, immunofluorescence microscopy and image analysis
For immunostaining, cells cultured on glass coverslips were fixed/permeabilized for 2 min in phosphate-buffered saline (PBS) containing 0.5% Triton X-100 and 3% paraformaldehyde, and post-fixed with 3% paraformaldehyde in PBS for 30 min. The cells were stained with primary antibodies for 1 h at room temperature, washed in PBS, incubated for 1 h with fluorophore-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA), washed again, and mounted in Elvanol (Moviol 4-88; Hoechst, Frankfurt, Germany). Images were acquired using the DeltaVision RT microscopy system (Applied Precision Inc., Issaquah, WA, USA). Primary antibodies used in this study included: mouse anti-paxillin and mouse anti-FAK (BD Transduction Laboratories, San Jose, CA, USA); mouse anti-ILK (Santa Cruz Biotechnology, Santa Cruz, CA, USA); mouse anti-vinculin (Sigma-Aldrich); and mouse anti-VASP, anti-talin and rabbit anti-kindlin 2 (Abcam, Cambridge, UK). Phalloidin labeled by FITC, TRITC or coumarin, for F-actin staining, was obtained from Invitrogen (Carlsbad, CA, USA). Other antibodies were prepared by the Antibody Production Laboratory of the Department of Biological Services, Weizmann Institute of Science: (http:// bioservices.weizmann.ac.il/antibody/about.html).
TIRF images were acquired using the DeltaVision Elite microscopy system equipped with a multi-line TIRF module (Applied Precision, Inc.), and time-lapse movies were taken at 3-min intervals, unless otherwise indicated, over a period of 2 h, using the same system. Quantification of fluorescence intensity and FA area was performed using image analysis software developed in-house, within the UCSF Priism environment (http://msg.ucsf.edu/IVE/), after correcting for fluorescence bleaching. To segment FAs, local background was subtracted and a watershed algorithm was applied, to separate neighboring adhesions. Quantification enabled the definition of ~30 geometric and intensity features for each adhesion site, including dimension, shape and intensity. The intensity and lifespan of individual FAs as a function of time were analyzed, following tracking of individual FAs. For fluorescence ratio imaging, cells were double-labeled for zyxin and vinculin, and the intensity ratio was computed per pixel, as previously described [24].

Fluorescence Recovery After Photobleaching (FRAP) assay and FRAP data analysis
mM HEPES, pH 7.2. An argon ion laser beam (Innova 70C; Coherent, Palo Alto, CA, USA) was focused through a fluorescence microscope (AxioImager D.1; Carl Zeiss MicroImaging, Oberkochen, Germany) to a Gaussian spot of 0.77±0.01 µm (with a plan-apochromat 63x/1.4 NA oilimmersion objective). After a brief measurement at monitoring intensity (528 nm, 1 µW), a 5 mW pulse (~10 ms) was used to bleach 50-75% of the fluorescence in the spot, after which the monitoring intensity beam was used to track the recovery of fluorescence over time.
To reduce the noise present in single FRAP curves, each curve was normalized by setting the pre-bleach intensity to 1, and 20-30 curves were then averaged, starting from the bleach point for synchronization. The resulting curve was fitted to the mathematical expression describing the full diffusion-exchange model, to extract the k on and k off values for each protein [27]. To increase the stability of the fitting procedure, some of the parameters were introduced as pre-determined, fixed inputs: D C (the diffusion coefficient in the juxtamembrane cytoplasm surrounding the FAs [26]), N FA (number of molecules in the FA), and N C (number of molecules in the juxtamembrane cytoplasm surrounding the FA). To determine D C for each protein, we performed FRAP measurements on FAs at a short timescale (2-3 s; data not shown), where the contribution of slow exchange is negligible, and fit the data as in [25][26][27]. For all proteins, these measurements yielded D C values of 1-1.5 µm 2 /s. For N FA and N C , we first estimated their values by measuring the fluorescence in a similarly sized region at an FA, as well as immediately adjacent to it, and then verified these estimates by fitting the FRAP data to the full model, with N FA and N C set as fitting parameters. Finally, we fitted the data to the full model, in which the D C , N FA , and N C values that were extracted as described above, were entered as fixed input. In all cases, the values of D M (the diffusion coefficient inside the FA) were negligible, relative to the fluorescence recovery time (i.e., <0.01 µm 2 /s).
Since both k on and k off were extracted in this manner, we could use these values to calculate, for each protein, the theoretical dissociation rate from the adhesion. For an adhesion to disassemble, the outward flux of proteins must be greater than the inward flux. The turnover of proteins in adhesions, at steady state, is a first-order process, i.e., for a protein to enter the adhesion, a bound protein must first detach from its binding site. We propose that during adhesion disassembly, the available binding sites for proteins from the cytoplasm are gradually reduced as a result of loss in tension; namely, after a specific protein molecule exits its binding site [25], the conformation of this site changes, thereby blocking rebinding of the cytoplasmic molecule. Since the "on rate" depends on the number of available binding sites, which changes over time at an unknown rate, the measured "on rate" constant is an apparent k on . As a first approximation, and for the purpose of comparing the different plaque proteins on both molecular and structural scales during adhesion disassembly, we defined: K dis = k off -k on . We then generated a theoretical disassembly curve for each protein, using: y=e −K dis *t , assuming that FA disassembly behaves as a mono-exponential process [28].

Cryo-electron tomography, correlative micros y=e −K dis *t copy, and image processing
Carbon-coated 200-mesh gold grids (Quantifoil, Micro Tools GmbH, Jena, Germany) were rinsed in PBS and overlaid on a drop of 50 µg/ml FN for 45 minutes. REF52 cells expressing YFP-paxillin were applied to the grids in concentrations of 100 cells per grid, and cultured for 24 h. As previously shown [29], these cells form FAs that are compatible with cryo-electron tomography (cryo-ET). A 5 µl drop of BSA-coated 15 nm gold colloids in PBS was added to the grids before plunging them into liquid nitrogen-cooled ethane, as described in [30]. Specimens were then transferred into a 300 FEG Polara microscope (FEI, Eindhoven, The Netherlands) equipped with a Gatan post-column GIF 2002 energy filter, and tilt series were collected, typically covering an angular range of -60° to 66°, sampled in 2° tilt increments, and at a 14 µm underfocus. Pixel size was 0.82 nm at the specimen level.
Projection images (2048 x 2048) were aligned to a common origin, using gold fiducial markers, and reconstructed by means of weighted back-projection, as implemented by means of a TOM toolbox software package [31]. All tomograms were reconstructed with a binning factor, to yield a 1.64 nm voxel size. For visualization purposes, the reconstructed volumes were processed by an anisotropic de-noising algorithm [32]. Individual adhesion complexes were identified and extracted, in silico. Two-dimensional images of these elements were calculated by projecting the volumes (32 x 32 x 11 voxels) along the Z axis, using the EM software package [33]. The particle stack was then masked and normalized prior to PCA, and followed by K-means classification (SPIDER). Class averages containing many features were chosen to be the first references for a multi-reference alignment of the data set. This strategy was iterated for five rounds until no major changes occurred in the classes, and in the alignment of single images.

Differential dissociation of adhesome components from FAs upon Y-27632 treatment
To explore the differential molecular sensitivity of adhesome molecules to relaxation of actomyosin-generated forces, we expressed, in HeLa JW cells, eight distinct, GFP-tagged FAassociated proteins, and examined the temporal effects of treatment with the Rho-kinase inhibitor Y-27632 on the levels of each, in association with FAs. Specifically, in this study we Differential Molecular Reorganization of FAs PLOS ONE | www.plosone.org used quantitative live-cell imaging to examine the drug's effect on zyxin, VASP, vinculin, paxillin, talin, ILK, kindlin-2, and FAK. In Figure 1A, representative images from time-lapse movies, corresponding to pre-treatment (control), and to 10, 30 and 90 minutes of treatment, are shown.
To exclude the possibility that the apparent changes in protein distribution are affected by the expression of the exogenous, tagged protein, we validated all live-cell data by immunofluorescence labeling of the corresponding endogenous proteins in non-transfected cells at several time points, following the same treatment (data not shown). Upon longer exposure to Y-27632, de novo formation of matrix adhesion sites occurred throughout the entire ventral cell membrane ( Figure 1B), despite the complete inhibition of MLC phosphorylation ( Figure 1C). Notably, all these effects were reversible, as cells regained normal morphology within 2 hours of Y-27632 withdrawal ( Figure 2).
Closer inspection of the treated cells indicated that FAs located within the lamellipodium or the lamella responded quite differently than those located around the cell center. In view of these differences (for details, see below), we quantified the responses of FAs located in these two subcellular regions (namely, "peripheral FAs" and "central FAs") separately. Examination of the differential effect of Y-27632 on the FA components tested, in both peripheral and central adhesions, indicated that the different proteins dissociate from FAs at distinct rates ( Figure 3A), as previously noted for some of these molecules [25,34,35]; see also Discussion.
To further quantify FA turnover, we then calculated FA lifespan (from "birth" to "death") for each of the studied proteins, as well as the exponential time-constant (τ) and fluorescence intensity of individual adhesion sites, during Y-27632 treatment ( Figure 3A for FA number and intensity, and 3B for FA lifespan, sorted by protein). The decay profiles are well described by a single-exponential curve. The timeconstant (τ) extracted from curve fitting the data ( Figure 3A), provides a natural physical time scale, as the system relaxes to 1/e (~37%) of its initial value. It is noteworthy that in untreated HeLa JW cells, the apparent τ of FAs labeled by all of the adhesome components tested varies, usually ranging between a few tens of minutes to a few hours (data not shown).
Quantification of the decline in the number and intensity of the initial peripheral and central adhesions, following Y-27632 treatment ( Figure 3A), indicates that Y-27632 dramatically reduced the lifespan of all existing FAs (both "peripheral" and "central"), though different proteins dissociate from FAs at varying rates. The first molecules to leave FAs were zyxin and VASP, with τ values of 7-8 min; talin, paxillin and ILK dissociate from FAs at a slower rate (τ ~16 min); while FAK, vinculin and kindlin-2 showed the slowest dissociation dynamics, with τ values of 25-28 min (see inserts in Figure 3A, values for the lifespans of peripheral FAs). To further emphasize the vast differences between the lifespans of FAs containing the different FA proteins, we constructed a histogram ( Figure 3B), depicting the lifespans of Y-27632-treated peripheral adhesions. Comparison of peripheral and central adhesions indicated that the latter dissociated at a 1.3-twofold faster rate; yet the order of exit of the different molecules was essentially the same.

Effect of Y-27632 treatment on FA nano-architecture
Examination of FAs of Y-27632-treated REF52 cells by means of cryo-ET revealed substantial changes in the molecular architecture of the cell-matrix adhesions ( Figure 4, and Figure S1). REF52 cells were chosen for these experiments since they are thin, form FAs that are suitable for cryo-ET [29], and the spatio-temporal dynamics of their FAs in the presence or absence of Y-27632 are similar to those of HeLa cells ( Figure S2). Image reconstructions of FAs after 3, 10 and 30 minutes of incubation with Y-27632 ( Figure 4B, C and D, respectively) show major structural alterations in the adhesion machinery, compared to untreated cells ( Figure 4A). These alterations manifest themselves in disorganization of the FA-associated actin network (see also 29) with a major loss of aligned filaments, already detected after 3 min of incubation with the drug ( Figure 4B). Upon longer treatment (10-30 minutes), the filaments become increasingly disorganized and sparse, as shown in Figure 4D. These early alterations of FAassociated actin filamentsoccur before a major decline in the levels of plaque proteins is evident, suggesting that loss of actin alignment in FAs constitutes an early response to Y-27632 treatment.

The nature of the adhesion structures formed in the presence of Y-27632
Long-term treatment of cells with Y-27632 enables the de novo formation and maintenance of integrin adhesions at the cell periphery and at the cell center, both of which differ from those formed by untreated cells. The new peripheral adhesions are insensitive to Y-27632, smaller than those of untreated cells, but display a protein molecular composition similar to those of untreated cells, yet with reduced levels of zyxin and kindlin-2 ( Figure 5A and C). Two-color TIRF microscopy revealed that ILK density, used as a marker for these adhesion structures, correlated with F-actin density at the same locations, suggesting that these structures are associated with actin filaments ( Figure 5B). The lifespan calculated for each protein ( Figure 5D) is much shorter than that found in untreated FAs, and varies from a few minutes for zyxin and VASP, to half an hour for kindlin-2 (see Figure 3), compared to over 2 hours in untreated cells.
The new central adhesions formed in the presence of the drug display a dot-like morphology, and consist of only a subset of the components present in classical FAs: ILK and α v integrin are prominent components, as well as kindlin-2 and low levels of paxillin. Notably, zyxin, VASP, vinculin, talin, FAK and the α v β 1 integrin appear to be absent from these adhesion structures ( Figure 6A and B). Furthermore, the ILK-rich patches were clearly visualized by two-color TIRF microscopy, confirming their association with the ventral cell membrane, while F-actin was hardly detectable in the same structures ( Figure 6C).

The molecular mechanism underlying ROCKdependent modulation of FAs
To explore the molecular basis for the differential turnover of adhesome components, we performed FRAP assays for the different FA-associated proteins studied herein, both prior to or within a few minutes of Y-27632 treatment. The results, shown in Figure 7A and 7B, demonstrate that the various proteins tested responded differently to the drug. The FRAP rates and the mobile fractions (R f ) of VASP, FAK and talin were not affected by Y-27632. Paxillin and zyxin, on the other hand, displayed reduced recovery rates and R f values, while vinculin and kindlin-2 exhibited faster turnover in the presence of the drug. Interestingly, a very limited effect of Y-27632 on recovery rate was initially seen in ILK; yet the mobile fraction was apparently reduced.
Calculation of the kinetic parameters (k on and k off ), performed for each protein in the presence and absence of Y-27632 (see Materials and Methods), showed distinct changes in the ratio between k on and k off values, for all proteins tested. Kindlin-2 and vinculin showed major increases in both k on and k off values; paxillin, talin and FAK showed minor decreases in both values; zyxin and VASP showed a small decrease in k on only, and ILK showed a slight increase in k off only (see Figure 7B). Notably, in untreated cells, the k on values for all proteins tested were higher than the k off values, whereas following Y-27632 treatment, k off exceeded k on for all these proteins. Addition of blebbistatin, instead of Y-27632, induced similar changes in the kinetic parameters of FA components (data not shown), suggesting that it is the relaxation of actomyosin contractility that triggers the unbinding of proteins from FAs, exceeding their binding rate, and consequently leading to a gradual disassembly of the FA structure.
Based on the differences between the k on and k off values induced by Y-27632, we calculated the expected dissociation rate of each of the tested proteins from the adhesion site (assuming that FA disassembly behaves as a monoexponential process [28]), and compared the predicted dissociation curves with those directly measured by live-cell microscopy ( Figure 7C and D). This comparison revealed a remarkable similarity between the calculated and measured values, suggesting that the basis for the differential dissociation of specific adhesion molecules upon suppression of actomyosin contractility is a protein-specific switch in the relative k on and k off values, induced by the drug.
FRAP measurements of the central adhesions formed following inhibition of actomyosin contractility for 30 minutes were also taken, specifically to quantify the recovery parameters of EGFP-ILK ( Figure S3). These measurements showed that ILK in these newly formed adhesions displayed k on and k off values that were essentially identical, though an order of magnitude lower, than those measured on FAs of untreated cells. Comparison of k on and k off values confirmed that in untreated FAs, ILK is in an "assembly state" (k on > k off ), but switches to a "disassembly state" (k on < k off ) following a few minutes of Y-27632 treatment; while in the novel central adhesions, ILK exists in a "stable state" (k on = k off ), even in the presence of the inhibitor.

Discussion
In this study, we addressed a basic aspect of adhesion biology; namely, the molecular, structural and functional heterogeneity of integrin adhesions, and the mechanisms whereby their molecular variability is regulated by mechanical force. The diversity of integrin adhesions has been extensively  [4,5,15,36]. It was further demonstrated that different regions within individual adhesion sites can display distinct molecular compositions [37]. This molecular heterogeneity is not surprising, given the extraordinary complexity and diversity of the ECM itself, displaying complex and irregular molecular composition and mechanical properties [5,[38][39][40]. Furthermore, the molecular machinery of the adhesion sites (known, collectively, as the "integrin adhesome"), comprising a large variety of receptors, adaptor molecules, cytoskeletal components and signaling proteins, can assemble into diverse complexes with distinct structural and functional properties [22]. In this study, we explored the The general notion that FA assembly depends on local mechanical forces is well-established (e.g. [17,[41][42][43]); yet the nature of the molecular cues whereby changes in local forces ("physical cues") affect FA composition ("chemical response"), is still largely unclear. For example, it is not known whether contractile forces are needed to keep the FAs structure intact, and upon reduction in these forces, the entire structure falls apart, or whether the turnover of different FA components is individually regulated. Our main objective in this study was to dynamically track the fate of multiple adhesome components after shifting live cells from their normal, contractile state to a "relaxed" state, induced by Rho-kinase inhibition. Progressing 27632, were analyzed by correlative (fluorescence-cryoelectron tomography) microscopy. Individual FAs, identified by light microscopy, were subjected to cryo-ET and 3D image reconstruction. The major axis of each FA is indicated by a white arrow; some of the more prominent filaments are highlighted by a thin red line. Ten nm-thick slices through the cryo-tomograms of representative FAs indicated a major reduction in actin alignment, already observed following 3 min treatment with the drug (compare B to A: The black arrow, in B, points to an example of misaligned filaments, commonly seen after Y-27632 treatment). Upon longer treatment with Y-27632, progressive misalignment of actin filaments is apparent, accompanied by a marked reduction in filament density. The scale bar in B represents 500 nm. stepwise from the structural-phenomenological level to the molecular-mechanistic level, we made the following observations: 1 Each of the eight proteins tested in this work dissociated from FAs at a different, characteristic rate, with zyxin and VASP being the fastest to exit, and kindlin-2 the slowest. 2 The subcellular location of the FAs had a marked effect on the rate of dissociation of all components, with "central adhesions" dissociating faster than those located at the cell periphery. 3 Structural analysis of FAs, based on cryo-ET, revealed major changes in the nanostructure of these adhesion sites, characterized by rapid disorganization of actin filaments shortly (<3 minutes) after addition of the inhibitory drug, suggesting that the loss of stress fiber-mediated tension precedes the dissociation of the adhesome molecules. 4 De novo formation of integrin adhesions with distinct subcellular distributions and molecular compositions takes place during and after the Y-27632-induced dissociation of the pre-existing FAs. 5 Rho-kinase inhibition had a differential effect on the molecular turnover kinetics (measured by FRAP) of the tested adhesome components. The apparent changes in k on and k off values and, hence, the predicted dissociation rates calculated for the different proteins following Y-27632 treatment, fit their measured dissociation rates.
A critical observation reported here is the distinct dynamic behavior of the different adhesome components tested, either in unperturbed cells, or in cells treated with Y-27632. These differences manifest themselves in: (1) different rates of dissociation from FAs following Rho-kinase inhibition; and (2) different turnover rates of the respective proteins in the presence and absence of the drug. These two features appear to be affected by mechanical stress: the major challenge of this work was to decipher the mechanistic interrelationships between the macroscopic observation (FA integrity) and the nanoscopic processes (e.g., turnover of specific proteins, or the It is noteworthy that the differential exit and entry of distinct FA proteins under different force regimes has been extensively documented in the past. Zamir et al. (2000) demonstrated that tensin and α 5 β 1 integrin exit FAs to form fibrillar adhesions, leaving behind paxillin, vinculin, α v β 3 integrin, and most of the tyrosine phosphorylated components of FAs. Rottner et al. (2001) reported on the recruitment of zyxin to new adhesion sites, yet its absence from lamellipodial and filopodial tips. They also noted that zyxin is an early target for signals leading to adhesion disassembly, in as much as it exits FAs before vinculin and paxillin, in line with the data reported here, and VASP dissociation occurs at an intermediate rate, between  zyxin and vinculin/ paxillin. Small differences between their results and the findings reported here can be attributed to the choice of other cell types, and use of combined expression of constitutively active Rac1 and Y-27632, in contrast with Y-27632 alone, in our experiments. In a more recent study, by Pasapera et al (2010), differential recruitment and dissociation of FA proteins is described, including the involvement of myosin II-dependent contractility in these processes. The results reported here extend the scope of these studies by examining a considerably larger array of FA components, and go one important step further, towards understanding the mechanism underlying the differential kinetic behavior of the different proteins, by calculating the specific changes in k on and k off values that drive the turnover of each of the proteins, and their dissociation from FAs, upon Y-27632 treatment.
Mechanical force is a critical factor in the regulation of FA formation and stability. This notion is based on the welldocumented dissociation of FAs upon treatment with Rhokinase or actomyosin inhibitors [16,18,19,34,35,44]. However, the temporal order of the reduction in stress and loss of adhesome components has not been established. The cryo-ET images acquired following treatment with Y-27632 support the notion that Rho-kinase inhibition induces marked structural changes in the adhesion sites shortly after addition of the inhibitor, but prior to the apparent protein dissociation from FAs. This evidence corroborates our previous finding that reduction in traction force upon myosin II inhibition, deduced by observing wrinkles produced by HeLa cells plated onto a flexible silicone rubber substrate, is essentially instantaneous, occurring in less than 20 seconds [25]. Y-27632-mediated effects manifest themselves in early loss of alignment of actin filaments (Figure 4 and [29]), and appear to be related to a reduction in the level of mechanical stress applied by the actomyosin system.
Once the forces applied to the adhesion sites are reduced, a general dissociation of FAs takes place, characterized by differential exit rates of various adhesome molecules. Attempts to understand the mechanism underlying this phenomenon indicated that the relaxation of Rho kinase-dependent actomyosin contractility radically and differentially affects the k on and k off values (calculated based on FRAP analyses) of the different proteins, in the presence and absence of Y-27632. In general, an increase in k off over k on values was seen, inevitably inducing net dissociation of each of the tested proteins from the adhesion sites. We further demonstrated that the effects of Y-27632 on the k off and k on values were highly protein-specific, leading to different predicted dissociation rates of the various proteins tested. Importantly, these predicted dissociation values were in excellent agreement with those actually measured by quantitative live-cell microscopy, supporting the notion that the mechanosensitivity of FAs may be attributed to differential, force-dependent regulation of the assemblydisassembly kinetics of individual FA components. Notably, although there is a similar trend among the calculated and measured disassembly curves for the different proteins, the calculated disassembly rates were faster than those actually measured. The reason could be that the disassembly process depends both on diffusion and exchange between membrane-bound and cytoplasmic proteins. Therefore, describing the process as dependent only on the exchange rate constants is an approximation. Nevertheless, the remarkable similarity between the measured and calculated curves suggests that exchange is the more dominant process in adhesion disassembly. Incorporating all dynamic processes into a model that describes FA disassembly is a focus for future efforts.
Another observation reported here is the apparent difference between integrin adhesions formed following Y-27632 treatment, and FAs of untreated cells. In general, the assembly of adhesion sites was shown to be highly dependent on actin organization, even in structures devoid of actomyosin contractility, e.g., small adhesions near the leading edge were reported to transmit strong traction forces in an actin polymerization-dependent but myosin II-independent manner [44][45][46][47]. Apparently, peripheral FAs, formed in the presence of Y-27632, display markedly reduced levels of zyxin and kindlin-2 (see Figure 5). It is noteworthy that zyxin accumulation in adhesion sites in a Rho-kinase-and myosin IImediated, force-dependent manner was previously reported [25,35,[45][46][47], yet mechanosensitive recruitment of kindlin-2 has not been previously described.
The integrin adhesions formed in the central regions of Y-27632-treated HeLa cells display limited association with Factin (in contrast to α V integrin-based adhesions reported by [48]), as well as a unique molecular composition, with prominent levels of α V integrin, ILK, kindlin-2 and paxillin, the adhesome components known to strongly interact with one another [49][50][51][52][53]; and little or no vinculin, talin, FAK, zyxin and VASP. It is interesting to note that FRAP experiments involving ILK in these central adhesions indicated that the k on and k off values were essentially the same, which would explain their stability in the presence of the drug.
In conclusion, this study highlights the molecular and structural diversity of integrin adhesions, and demonstrates the central role of mechanical forces in regulating this diversity. Quantifying eight different adhesome molecules, we show that integrin adhesions display diverse "molecular signatures", and that mechanical forces can play a key role in regulating FA composition. We further propose a novel mechanism underlying this mechanosensitivity, in which forces applied to the adhesion site selectively modulate the on-and off-rates of different FA constituents, thereby regulating the molecular composition of FAs, and thus, their stability and fate. Figure S1.

Supporting Information
Correlative fluorescence and cryo-ET microscopy. Correlated microscopy, combining fluorescence microscopy and cryo-electron tomography, was used to study the effect of Y-27632 on actin organization in REF-52 cells expressing YFP-tagged paxillin. Cells growing on EM grids were first examined by fluorescence light microscopy, then transferred to the cryo-EM, and specific FAs were identified (white border). Cryo-electron tomograms of the adhesion sites were collected. The inset shows the adhesion site (black arrow) at low magnification cryo-EM. (TIF) Figure S2.
Effect of Y-27632 treatment on FA reorganization in REF52 cells. Untreated cells (control) and cells treated with 10 µM Y-27632 for 3,10 and 30 min were stained for actin, paxillin and vinculin, demonstrating that the spatio-temporal dynamics of FAs in response to Y-27632 are similar to those of HeLa JW cells (Figure 1). (TIF) Figure S3. k on vs. k off values of GFP-ILK at different adhesion sites. FRAP measurements were performed either on untreated HeLa cells, within 10 minutes of their treatment with Y-27632 (initial FAs), or after 30 minutes of such treatment (newly formed central adhesions). The resulting k on and k off values ± SEM are presented. Note that in the new adhesions, k on > k off , indicating a new steady-state, though with slower kinetics than that of ILK in untreated cells. (TIF)