Myosin forces elicit an F-actin structural landscape that mediates mechanosensitive protein recognition

Cells mechanically interface with their surroundings through cytoskeleton-linked adhesions, allowing them to sense physical cues that instruct development and drive diseases such as cancer. Contractile forces generated by myosin motor proteins mediate these mechanical signal transduction processes through unclear protein structural mechanisms. Here, we show that myosin forces elicit structural changes in actin filaments (F-actin) that modulate binding by the mechanosensitive adhesion protein α-catenin. Using correlative cryo-fluorescence microscopy and cryo-electron tomography, we identify F-actin featuring domains of nanoscale oscillating curvature at cytoskeleton-adhesion interfaces enriched in zyxin, a marker of actin-myosin generated traction forces. We next introduce a reconstitution system for visualizing F-actin in the presence of myosin forces with cryo-electron microscopy, which reveals morphologically similar superhelical F-actin spirals. In simulations, transient forces mimicking tugging and release of filaments by motors produce spirals, supporting a mechanistic link to myosin’s ATPase mechanochemical cycle. Three-dimensional reconstruction of spirals uncovers extensive asymmetric remodeling of F-actin’s helical lattice. This is recognized by α-catenin, which cooperatively binds along individual strands, preferentially engaging interfaces featuring extended inter-subunit distances while simultaneously suppressing rotational deviations to regularize the lattice. Collectively, we find that myosin forces can deform F-actin, generating a conformational landscape that is detected and reciprocally modulated by a mechanosensitive protein, providing a direct structural glimpse at active force transduction through the cytoskeleton.


Fig. S1 :
Fig. S1: Correlative cryo-fluorescence and cryo-electron tomography.a, Schematic of sample preparation and data collection workflow.b and c, Additional examples of oscillatory domains imaged in zyxin-enriched adhesion sites from N = 2 independent experiments.The site in b features parallel bundled F-actin and was visualized in a cell not treated with Rho Activator II, while the site in c features more disorganized F-actin and was visualized in a cell treated with Rho Activator II.Correlative LM / EM and tomograms / segmentations are presented as in Fig. 1a,b.

Fig. S2 :
Fig. S2: Analysis of F-actin oscillatory domain morphology.a, Quantification of oscillatory domain wavelengths in the dual motor condition.n = 251 from N = 3 independent experiments.b and c, False-colored cryo-EM images of the dual motor condition in the presence of ATP, displaying oscillatory domains in a hole-spanning filament (b) and in a pair of broken filaments (c).Oscillatory domains, magenta; canonical F-actin, blue; carbon film, orange; ice contamination, yellow.d, Polar arrow plot of oscillatory domain major axis orientation relative to the ice plane from n = 16 observations, where 0° corresponds to the major axis being parallel to the ice plane.e, Serial Z slices through an oscillatory domain tomogram (barbed-end directed force condition) featuring protruding densities consistent with subunit dislocations (red arrows).f, Calibration of the F-actin forceextension curve to determine harmonic bond stiffness in coarse-grained molecular dynamics simulations.g, Calibration of cumulative twist variance to determine the bending constants of the harmonic angle potential and the dihedral potential used in coarse-grained molecular dynamics simulations.

Fig. S3 :
Fig. S3: Adaptation of neural network picker and filament curvature analysis.a, Workflow for network training and subsequent picking.A synthetic particle dataset is generated by projecting PDB models of filaments featuring different computationally-generated curvatures, which are modulated by the CTF followed by the addition of a pink noise box (top).The network is then trained to denoise these noisy particles, which is then used as an input for semantic segmentation.After training with synthetic data, the network can be used to denoise and segment real data.b, Sample micrograph containing both straight filaments and filaments featuring oscillating curvature.Filaments are assigned estimated signed curvature values and categorized for subsequent selection.Traces are offset by one Factin width for visualization.c, Curvature distributions of filaments identified in the dual myosin-motor evoked force condition (pink) or the -ATP control (blue).Measured curvature distributions of all picked filament segments (left), non-superhelical segments (middle), and superhelical segments (right) are shown as histograms, and modeled thermal bending fluctuation distributions are overlaid as blue, pink, and grey curves.The gray curve corresponds to a simple bending model described by equations 5-7 (Methods) using actin's experimentally determined persistence length of 9 μm (ref.63), while the pink and blue curves correspond to a model fitted with a multiplicative adjustment factor as described in detail in the Methods.

Fig. S4 :
Fig. S4: Superhelix structure determination workflow and variability analysis.a, Initial single particle cryo-EM data processing workflow for superhelical F-actin, using a single dataset collected for the dual motor condition.Transparent red and green boxes indicate rejected and accepted classes, respectively.b, Final processing workflow, incorporating data from two additional datasets to boost particle number and enhance the quality of the final map.Additionally, 3DVA variability analysis is displayed, highlighting the presence of continuous structural variability despite extensive classification.

Fig. S6 :
Fig. S6: Subdomain displacements in superhelical F-actin.Superimposed subdomain displacement vectors from all protomers after MDFF analysis of +ATP myosin force-evoked superhelical F-actin (top) and -ATP control reconstructions (bottom).Subdomains 1 and 4 versus 2 and 3 are displayed separately for clarity, and vectors are scaled 15X for visualization.The averages of these vectors are displayed in Fig.3c.

Fig. S7 :
Fig.S7: Cryo-EM processing workflow for the myosin force-activated ⍺-catenin-F-actin complex.Top: Cryo-EM processing workflow for visualizing the force-activated ⍺-catenin-F-actin complex, from a specimen prepared in the dual motor condition.Green and red boxes represent 2D class averages which were selected and rejected for additional processing, respectively.Bottom: a magnified 2D class average is displayed, highlighting preferential binding of ⍺-catenin along one side of the filament (arrowheads) on alternating F-actin strands.

Fig. S9 :
Fig.S9: Additional analysis of the force-activated ⍺-catenin-F-actin complex structure.a, Instantaneous helical twist of selected 3DVA frames from Fig.4b,c.Vertical dashed lines indicate canonical F-actin twist.b, Orthogonal views of the post-3DVA α-catenin-F-actin complex map, colored as in Fig.4a.c, Quantification of α-catenin density intensity in post-3DVA map.Yellow corresponds to light blue strand from a, and orange corresponds to dark blue strand.d, Instantaneous helical parameters of consensus map, colored as in b.Shaded regions represent 95% CI from 3 independent analyses.Vertical dashed lines indicate parameters of canonical F-actin.e, Superimposed subdomain displacement vectors versus a canonical F-actin subunit from all protomers after MDFF analysis of the consensus α-catenin-Factin complex map, displayed as in Fig.S6.The averages of these vectors are displayed in Fig.4f, top.f, Quantification of actin subdomain 2 reorientation versus α-catenin intensity.Dashed line indicates 50%.