Skip to main content

Structural heterogeneity of cellular K5/K14 filaments as revealed by cryo-electron microscopy

Cite this dataset

Weber, Miriam Sarah et al. (2021). Structural heterogeneity of cellular K5/K14 filaments as revealed by cryo-electron microscopy [Dataset]. Dryad.


Keratin intermediate filaments are an essential and major component of the cytoskeleton in epithelial cells. They form a stable yet dynamic filamentous network extending from the nucleus to the cell periphery, which provides resistance to mechanical stresses. Mutations in keratin genes are related to a variety of epithelial tissue diseases. Despite their importance, the molecular structure of keratin filaments remains largely unknown. In this study, we analyzed the structure of keratin 5/keratin 14 filaments within ghost keratinocytes by cryo-electron microscopy and cryo-electron tomography. By averaging a large number of keratin segments, we have gained insights into the helical architecture of the filaments. Two-dimensional classification revealed profound variations in the diameter of keratin filaments and their subunit organization. Reconstitution of filaments of substantial length uncovered a high degree of internal heterogeneity along single filaments, which can contain regions of helical symmetry, regions with less symmetry and regions with significant diameter fluctuations. Cross section views of filaments revealed that keratins form hollow cylinders consisting of multiple protofilaments, with an electron dense core located in the center of the filament. These findings shed light on the complex and remarkable heterogenic architecture of keratin filaments, suggesting that they are highly flexible, dynamic cytoskeletal structures.


K5/K14_1 cells expressing only K5/K14 intermediate filaments were seeded on glow-discharged Collagen I coated holey carbon gold EM grids (Au R2/1, 200 mesh, Quantifoil) and incubated overnight at 32 °C and in 5% CO2. The grids were rinsed in washing buffer (1x PBS, 2 mM MgCl2), cells were permeabilized for 15 – 20 s in permeabilization buffer (1x PBS, 0.1% Triton X-100, 600 mM KCl, 10 mM MgCl2 and protease inhibitors), and rinsed again in washing buffer. Next, the grids were incubated with 2.5 units/µl benzonase (Merck, 71206-3) in washing buffer for 30 min and washed again before vitrification in liquid ethane using a manual plunge freezing device.

The grids were analyzed using a 300 kV Titan Krios electron microscope (Thermo Fisher) equipped with a K2 Summit direct electron detector (Gatan) mounted on a post-column energy filter (Gatan). Cryo-EM micrographs were acquired in zero-loss energy mode using a 20 eV slit. Data were recorded with SerialEM 3.5.8 in low dose mode (74). Micrographs were acquired at a nominal magnification of 22,665 x with a pixel size of 2.206 Å. A defocus range between -0.5 and -3.5 μm was chosen. Dose-fractionation was used with a frame exposure of 0.2 s with a total exposure time of 10 s (50 frames in total). This corresponds to a total electron dose of ~20 e/Å2.

1,860 cryo-EM micrographs were processed with RELION 2.1 and RELION 3.0 using the helical toolbox. Frame-based motion correction and dose-weighting were performed using MotionCor2. The contrast transfer function was estimated using CTFFIND4. Low-quality micrographs showing high defocus, high astigmatism or low resolution were excluded, resulting in 1,763 micrographs used for further processing steps. Keratin filaments were either picked manually or automatically using the RELION helical toolbox. To generate a template for autopicking, 55,073 keratin particles were picked manually as start-to-end helices, extracted with a box size of 250 pixels (~55 nm) and 2D classified twice to create classes with straight keratin segments. These classes served as a reference for automated picking of 505,211 particles. For manual picking, 298,056 particles were selected as start-to-end helices. Particles were extracted in boxes of 250 pixels, corresponding to ~55 nm, or 164 pix, corresponding to ~36 nm, with an inter-box distance of 50 Å. Iterative 2D classification procedures were performed, using a spherical mask of 500 Å or 356 Å, respectively.

Usage notes

55 nm long keratin segments were finally classified into 50 representative 2D classes containing 305,495 particles. 36 nm long keratin segments were finally classified into 100 representative 2D classes containing 197,541 particles.


Swiss National Science Foundation, Award: 31003A, 179418

University of Zurich, Award: FK-18-041