Cryo-EM analysis and image processing

Cryo-EM data were collected on a Titan Krios II transmission electron microscope (FEI) operated at 300 keV. A K3 summit direct detector (GATAN) with a GIF Quantum Filter with a slit width of 20 eV was used for the data acquisition. Data acquisition was performed using SerialEM at a nominal magnification of 81,000×, corresponding to a pixel size of 1.05 Å per pixel in nanoprobe EF-TEM mode. For the Pol II−Spt4/5−nucleosome−Chd1 dataset, image stacks of 40 frames were collected in counting mode over 2.2 s. The dose rate was ~18.12 e⁻ per Ų per s for a total dose of 39.87 e⁻ per Ų. For the Pol II−Spt4/5−nucleosome−FACT dataset, image stacks of 40 frames were collected in counting mode over 2.2 s. The dose rate was ~18.30 e− per Ų per s for a total dose of 40.25 e⁻ per Ų.

Micrographs were stacked and processed using Warp⁴⁸. CTF estimation and motion correction was performed using Warp⁴⁸. Particles were picked using an in-house trained instance of the neural network BoxNet2 as implemented in Warp, yielding 3,755,390 particles for the Pol II−Spt4/5−nucleosome−Chd1 dataset and 5,227,093 particles for the Pol II−Spt4/5−nucleosome−FACT dataset. Particles were extracted with a box size of 400 pixels and normalized. Further image processing was performed with cryoSPARC⁴⁹ and RELION 3.0.7.

For the Pol II−Spt4/5−nucleosome−Chd1 dataset, particles were separated into two subsets and subsequently 3D classified using cryoSPARC⁴⁹. Particles were selected for the presence of a nucleosome-like density. Selected particles were imported into RELION⁵⁰. Two subsequent rounds of 3D classification resulted in 247,604 particles with clear nucleosomal density. A 3D refinement of these particles resulted in an overall model of 2.6 Å. The particles were CTF refined, and Bayesian polishing was conducted. A mask encompassing the nucleosome and additional density at SHL +2 was applied to two rounds of 3D classification to select for particles that contain Chd1. This resulted in 30,876 particles with clear density for Chd1 bound to the partially unwrapped nucleosome. Particles were then subsequently 3D refined resulting in map A (EMD-12666) with a resolution of 2.9 Å (FSC threshold 0.143 criterion). The map showed excellent density for Pol II, but the nucleosome−Chd1 part of the map showed flexibility. Therefore, Pol II with Spt4/5 and the nucleosome−Chd1 parts of the maps were individually refined using a mask for Pol II−Spt4/5 or nucleosome−Chd1, respectively. Signal subtraction was applied. This resulted in two masked refinements (Pol II−Spt4/5, map B, EMD-12667; nucleosome−Chd1 with signal subtraction, map C, EMD-12668) with overall improved density. These two masked refinements were combined using the Frankenmap and Noise2map tool set of Warp, resulting in the final composite map (map D, EMD-12449).

For the Pol II−Spt4/5−nucleosome−FACT dataset, particles were separated into two subsets and subsequently 3D classified using cryoSPARC and RELION. Particles were selected for the presence of nucleosome-like densities. The selection resulted in 603,550 particles with clear nucleosome-like density. The selected particles were 3D refined using an angular sampling rate of 7.5° and subjected to further classifications to select for particles with bound FACT. This ultimately resulted in 48,718 particles. These particles were again subjected to 3D classification. After 3D refinement, CTF refinement and Bayesian polishing, the remaining 47,138 particles resulted in a final refinement (map 1, EMD-12669) with an overall resolution of 3.1 Å (FSC threshold 0.143 criterion). To improve densities for the Pol II−Spt4/5 and nucleosome−FACT parts of the map, the particles were subjected to masked refinements (maps 2, EMD-12670 and map 3, EMD-12671). Similar to the Chd1 dataset, the masked refinements were combined using the Frankenmap and Noise2map tools included in Warp, yielding the final map (composite map 4, EMD-12450). Local resolutions of the composite maps were estimated using the RELION built-in tool for the determination of local resolutions.