(A, D, G) Map cross-sections following application of the three masks used during these calculations, corresponding to the whole filament, core, and sheath regions, respectively (see Methods). (B, E, H) Non-directional Fourier Shell Correlation estimates for the three masked regions, calculated with the Relion software suite. (C, F, I) Directional Fourier Shell Correlation estimates for the three masked regions using the 3D FSC program (Tan et al., 2017) (see Materials and methods). The histogram (blue bars) reports the distribution of estimated resolutions for different angular directions (conical regions in Fourier space). Due to strongly preferred filament orientations in the sample (see Figure 1—figure supplement 1), a cone of data is missing from the map Fourier transform. As a result, a significant drop-off is observed in the reported resolution for angular directions that coincide with the missing cone, as reflected in the histogram: reported resolutions are distributed widely, from ~35 Å in the worst directions (those close to the vertical axis in (A, D, G) to ~8 Å in the best directions. Although this resolution anisotropy leads to ‘smearing’ of density features (vertical direction in panels (A), (D), (G), the asymmetric groove is large enough to be readily resolved in the wild-type map. While features on the left side in panels (A), (D), (G) (where the groove is located) diverge from the right side (where there is no groove), the average map resolution is comparable at these two locations (Figure 2—figure supplement 4), supporting the observed asymmetry between the two sides of the filament. See Figure 3—figure supplement 4.