Effect of multiple scattering and the ice thickness on the information transfer in cryo transmission electron microscopy

Abstract text (incl. figure legends and references)

In the field of electron cryo-microscopy it remains a challenge to improve the resolution and to decrease the number of micrographs needed for challenging protein samples. To get a better understanding of the scattering inside a biological specimen and the surrounding glass-like ice, we performed multislice simulations of the tobacco mosaic virus (TMV) (PDBe: 6sag [1]) in ice. The molecular structure of ice was modelled by molecular dynamics (LAMMPS [2]) and the atomic model of TMV was added [3]. First, we varied the number of slices used for the multislice simulations. Secondly, we added additional ice above and below the TMV.

The contrast transfer, as seen in Fig.1 by the Fourier ring correlations (FRC) of the exit waves decreases when the number of slices is reduced. The effect is strongest at high spatial frequencies. However, even for a simulation of a single slice that accounts for the propagation to the same optical plane as the multislice simulation, the FRC values are larger than 0.75 up to a resolution of 2.3 Å. So, multiple scattering has only a small effect up to this resolution, which suggests the principal applicability of dose-efficient scanning TEM (STEM) techniques that are based on single scattering, such as integrated differential phase contrast (iDPC) [4] and the single side band (SSB) ptychographic reconstruction. Compared to CTEM simulations, iDPC and the phase of SSB in Fig. 1 are easier to interpret because both contrast transfer functions do not show contrast inversions.

To simulate the averaging of single particle reconstructions, we also calculated the average of 600 phase-corrected simulated CTEM images (defocus range: [-0.1,-1.0] μm, spherical aberration: 1.5 mm) for different ice thickness and compared the results to the ground truth, using the FRC curves (Fig. 2). The contrast transfer decreases with increasing ice thickness, which is reflected in a lower contrast in real space for TMVs and in lower FRC values. The decrease of the FRC values is stronger at higher spatial frequencies, with a plateau above a resolution of 3 Å (Fig. 2). The FRC decrease is most dominant when adding the first ice layers, whereas depositing further ice to form rather thick layer of up to 90nm affect the fidelity of the exit wave and the CTEM simulation as to the TMV less.

Our results will be discussed in the context of experimental applicability of momentum-resolved STEM techniques to biological specimen under low-dose cryo-conditions, and to which extent such new acquisition schemes might improve the structural characterisations of proteins.

Figure 1) (a): FRC curves of the exit waves for multislice simulations of TMV with different number of slices, the number is given in the legend. (b): iDPC reconstruction. (c): Angle of the SSB reconstruction.

Figure 2) (a) and (b): Phase of the exit wave of simulations with minimal and maximal ice thickness (thickness in the label). (c) and (d): CTEM simulations of the exit waves in (a) and (b) with a defocus of -500 nm and a spherical aberration of 1.5 mm. (e) and (f): FRC curves of the exit wave and the average of 600 phase corrected CTEM simulations (defocus: -0.1 μm to -1.0 μm, spherical aberration: 1.5 mm).

[1] Weis, et al. EMBO reports 20.11 (2019)

[2] Thompson, et al. Computer Physics Communications 271 (2022)

[3] Shang, et al. Journal of structural biology 180.1 (2012)

[4] Lazic, et al. (accepted for publication).