Extended Data Fig. 6: Workflows for MIP analysis and stress-fiber subtomogram averaging.

a) Processing workflow for the MIP analysis. i) Zoomed-in view of a MT in a 2D slice of a tomogram, scale bar: 50 nm. ii) The centerline is traced by Amira. iii) The lumen is segmented by applying distance transform on the centerline. iv) PySeg analysis extracting the local minima and their connections in the MT lumen and MT shell. The color bar represents the topological persistence, the metric used to rank the minima relevance. The information of the shell is used to determine the threshold for the lumen. v) Thresholded local minima by persistence in the MT lumen. vi) MIP obtained by applying mean shift clustering to v. b) Processing workflow for stress-fiber subtomogram averaging. Tilt-series alignment and tomographic reconstructions were performed within RELION 5.0. A total of 1555 segments of filaments were manually selected from the area of interest using IMOD, and the corresponding coordinates were imported into RELION for subtomogram extraction in the form of a 2D stack. For subtomogram averaging, a cylinder reference was used, followed by 3D classification to create an initial model, and further refinement was performed using a combination of classification and refinement schemes without applying the actin helical symmetry. The nominal resolution was 27.7 Å. CTF correction was applied only during subtomographic analysis, following 3D reconstruction using RELION 5.0.