Fascin structural plasticity mediates flexible actin bundle construction

Summary: Fascin crosslinks actin filaments (F-actin) into bundles that support tubular membrane protrusions including filopodia and stereocilia. Fascin dysregulation drives aberrant cell migration during metastasis, and fascin inhibitors are under development as cancer therapeutics. Here, we use cryo-electron microscopy, cryo-electron tomography coupled with custom denoising, and computational modeling to probe fascin’s F-actin crosslinking mechanisms across spatial scales. Our fascin crossbridge structure reveals an asymmetric F-actin binding conformation that is allosterically blocked by the inhibitor G2. Reconstructions of seven-filament hexagonal bundle elements, variability analysis, and simulations show how structural plasticity enables fascin to bridge varied inter-filament orientations, accommodating mismatches between F-actin’s helical symmetry and bundle hexagonal packing. Tomography of many-filament bundles and modeling uncovers geometric rules underlying emergent fascin binding patterns, as well as the accumulation of unfavorable crosslinks that limit bundle size. Collectively, this work shows how fascin harnesses fine-tuned nanoscale structural dynamics to build and regulate micron-scale F-actin bundles.


Figure S2 .
Figure S2.Analyses of fascin conformation and bundle element architecture (A) Superposition of fascin's four individual β-trefoil domains extracted from the consensus atomic model (Figure 1F).(B) Cryo-EM data processing workflow for reconstructing a hexagonal bundle element.(C-D) FSC curves for the bundle element reconstructed with two different box sizes.(E-F) Side views of actin planes 1-0-4 (E) and 5-0-2 (F) from the 8.7 Å reconstruction docking model.(G-I) Side view of each actin plane of models from 5 bundle element classes aligned on the central filament.(J) Distributions of poses of the 6 central-filament associated fascins from the four additional bundle element classes.

Figure S3 .
Figure S3.Analyses of multi-body derived reconstructions (A) Cryo-EM density maps of eigen_left, eigen_middle and eigen_right reconstructions.(B-C) FSC curves (B) and local resolution assessment (C) of the three reconstructions.(D) Superposition of all six Factin models from the three reconstructions.All F-actin 1 models are colored light blue, while F-actin 2 models are colored as in Figure 5B.(E) Comparison of the fascin-F-actin 1 interface across the three snapshots, superimposed on F-actin 1. (F-H) Detail views of contacts indicated in E. (I) Comparison of the fascin-F-actin 2 interface across the three snapshots, superimposed on F-actin 2. (J-M) Detail views of contacts indicated in I.

Figure S4 .
Figure S4.Computational model parametrization and analyses (A) Flowchart of the computational model.Filaments are added iteratively, and the crosslinking probabilities for all angular rotations, θB, of the new filament B are evaluated.The angular rotation corresponding to the maximum probability of crosslinking is selected, as long as this probability is greater than the minimum cross-linking threshold, .(B) Atomic model of fascin crosslinked F-actin in ribbon representation.Filament A is depicted in shades of blue, while filament B is depicted in shades of green.Interfacial residues used as fiducials to calculate characteristic distances are indicated with spheres of varying colors.Grey cylinders represent F-actin axes.(C) Fascin binding probabilities as a function of

Figure S5 .
Figure S5.Neural network architecture and performance (A) Neural network architecture used for pretraining on synthetic data and training on both synthetic and experimental subtomograms.(B) Neural network architecture and performance used for inference on experimental subtomogram.(C) Representative denoising and semantic segmentation performance on synthetic, noisy subtomogram.(D) Neural network denoising performance on cross-sectional slab of experimental tomogram.

Figure S6 .
Figure S6.Additional structural analysis of fascin crosslinked bundles (A) Top view of semantically segmented F-actin highlighting interconnected bundles.(B) End-on view of bundles in A, highlighting supertwist along their respective longitudinal axes.(C) End-on view schematics of filament rotational phase shifts and fascin poses of four additional bundles, analyzed as in Figure 6B.Dark grey and gold filaments correspond to actin planes displayed in D. (D) Side views of rigid-body docking models of two additional actin planes.Filament rotational phase offsets and fascin poses are indicated.

Figure S7 .
Figure S7.Subtomogram averaging workflow and analyses (A) Subtomogram averaging data processing workflow.(B) Local resolution map of subtomogram average.(C) Extreme frames (1 representing 5 th percentile, and 10 representing 95 th percentile) of interpolation along the first principal component of multibody refinement.Similar inter-filament rotation is present as observed in single particle analysis.(D) FSC curves of the consensus (gray) and multi-body refinement reconstructions (yellow, blue).(E) Distribution of amplitudes along eigenvector 1 of all subtomograms.(F) Distribution of Mahalanobis distances of all subtomograms (n = 129,948).Fit represents gamma distribution (R 2 = 0.9951).