Accurate magnification determination for cryoEM using gold

Abstract

Determining the correct magnified pixel size of single-particle cryoEM micrographs is necessary to maximize resolution and enable accurate model building. Here we describe a simple and rapid procedure for determining the absolute magnification in an electron cryomicroscope to a precision of <0.5%. We show how to use the atomic lattice spacings of crystals of thin and readily available test specimens, such as gold, as an absolute reference to determine magnification for both room temperature and cryogenic imaging. We compare this method to other commonly used methods, and show that it provides comparable accuracy in spite of its simplicity. This magnification calibration method provides a definitive reference quantity for data analysis and processing, simplifies the combination of multiple datasets from different microscopes and detectors, and improves the accuracy with which the contrast transfer function of the microscope can be determined. We also provide an open source program, magCalEM, which can be used to accurately estimate the magnified pixel size of a cryoEM dataset ex post facto.

Keywords: CryoEM; Electron cryomicroscopy; Gold; Magnification calibration; Pixel size; Protein structure; Single-particle reconstruction.

Fig. 1

(A) shows two models of DPS, built using ModelAngelo and refined using REFMAC5 , with the correct pixel size of 0.648 Å (green) and incorrect pixel size of 0.680 Å (purple). (B) and (C) show the error in the CTF as a function of spatial frequency and pixel size error for data collected at an accelerating voltage of 300 kV, a defocus of $-1.5\mathrm{\mu }\mathrm{m}$, and a $C_s$ of 2.7 mm. In (B), the defocus is kept at $-1.5\mathrm{\mu }\mathrm{m}$, whereas in (C) the defocus has been adjusted to minimize the error in the CTF.

Fig. 2

Demonstration of how to calculate the magnified pixel size from the Fourier transform of gold. In (A), a micrograph of polycrystalline gold foil from HexAuFoil grids gives a complete (111) ring in the Fourier transform shown in (B). The radius of the (111) ring can be used to estimate the magnified pixel size, given a gold lattice resolution of $\approx$2.35 Å.

Fig. 3

Issues in data processing of lattice spacing data that must be correctly handled. (A) The gold foil on UltrAuFoil grids displays two clear peaks at 2.35 Å and 2.48 Å. The inner peak at 2.48 Å is likely from hcp gold and selection of this results in a 5% error in magnified pixel size. (B) Although by eye the ring in the Fourier transform looks round, magCalEM measured significant anisotropic magnification. (C) Plots of the radial profile, centered on the true peak position. The peak of a weak signal is shifted towards the center when noise whitening is not employed (left). Noise whitening the power spectrum (right) prevents this.

Fig. 4

Cross correlation between maps generated from high resolution X-ray models of DPS and the cryoEM map. The cross correlation was measured as described in Section 2.3. The average is marked with the red vertical dashed line. The magnified pixel size estimated from gold HexAuFoil grids is marked by the blue dotted vertical line. The standard deviations are shown by the red and blue regions for the cross correlations and HexAuFoil measurements respectively.

Fig. 5

The effect of heating to 693 K on the gold foil on HexAuFoil gold grids. The summed Fourier transform from several foil regions on HexAuFoil grids (A) exhibits a strong peak at 2.35 Å resolution (111) and at 2.00 Å (200). After heat treatment (B), the grain size is noticeably larger and a peak at 2.48 Å is now the dominant peak in the Fourier transform.

Fig. 6

Selected area diffraction patterns from a large gold grain on the heat treated HexAuFoil grids with no tilt (A) and tilted to 30° (B). After tilting to 30°, a reflection at 2.23 Å resolution is now visible.