An electron counting algorithm improves imaging of proteins with low-acceleration-voltage cryo-electron microscope

ARTICLE https://doi.org/10.1038/s42003-022-03284-1 OPEN An electron counting algorithm improves imaging of proteins with low-acceleration-voltage cryoelectron microscope 1234567890():,; Dongjie Zhu1,2, Huigang Shi 2,3, Chunling Wu2,3 & Xinzheng Zhang 2,3 ✉ Relative to the 300-kV accelerating field, electrons accelerated under lower voltages are potentially scattered more strongly. Lowering the accelerate voltage has been suggested to enhance the signal-to-noise ratio (SNR) of cryo-electron microscopy (cryo-EM) images of small-molecular-weight proteins (<100 kD). However, the detection efficient of current Direct Detection Devices (DDDs) and temporal coherence of cryo-EM decrease at lower voltage, leading to loss of SNR. Here, we present an electron counting algorithm to improve the detection of low-energy electrons. The counting algorithm increased the SNR of 120-kV and 200-kV cryo-EM image from a Falcon III camera by 8%, 20% at half the Nyquist frequency and 21%, 80% at Nyquist frequency, respectively, resulting in a considerable improvement in resolution of 3D reconstructions. Our results indicate that with further improved temporal coherence and a dedicated designed camera, a 120-kV cryo-electron microscope has potential to match the 300-kV microscope at imaging small proteins. 1 School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China, 230026 Hefei, China. 2 National Laboratory of Biomacromolecules, CAS Center for Excellence in Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, 100101 Beijing, China. 3 University of Chinese Academy of Sciences, 100049 Beijing, China. ✉email: COMMUNICATIONS BIOLOGY | (2022)5:321 | https://doi.org/10.1038/s42003-022-03284-1 | www.nature.com/commsbio 1 ARTICLE D COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-022-03284-1 uring imaging, the exposure of proteins to accelerating electrons leads to irreversible and accumulative radiation damage1–4. Consequently, the total electron dose used in imaging proteins is limited, resulting in elevated shot noise, which yields a low signal-to-noise ratio (SNR) in cryo-electron microscopy (cryo-EM) images. A reconstruction of the 3D structure of a protein employs two steps: alignment and averaging5. A high SNR of the cryo-EM image is crucial for alignment accuracy, which then permits an effective averaging that increases the SNR in the 3D reconstruction, equivalent to improving the resolution of the reconstruction in cryo-EM. The signal from proteins in cryo-EM images is correlated with its size. Small proteins scatter less electrons, thereby producing weak signals, leading to low SNR images and therefore preventing accurate alignments6,7. From the perspective of SNR, the difficulty in reconstructions increases with decreasing protein size. Therefore, improving the SNR of cryo-EM images is extremely helpful in the reconstruction of small proteins. The power of cryo-EM in resolving the 3D structure of proteins increased rapidly since 2013 through the application of direct detection devices8,9 (DDDs). The DDD has high frame rate and improved detective quantum efficiency (DQE) compared with those of charge-coupled devices (CCDs) and films, allowing the recording of cryo-EM images with better SNR10, helping to set off the ‘resolution revolution’11. Currently, the state-of-the-art cryoEM technique enables the reconstruction of proteins with sizes as low as 40 kD12 but requires an ideal buffer condition and a thin layer of ice to minimise background noise apart from shot noise. However, in reality, an ideal cryo-EM specimen is difficult for many proteins. Consequently, any method that further improves the SNR of cryo-EM images benefits the structural determination of small proteins. The power of scattered electrons by an atom is negatively correlated with the energy of the incident electron13. Therefore, low-acceleration-voltage electrons produce more signal upon their interaction with the cryo-EM specimen. However, lowenergy electrons create severer radiation damage14, allowing fewer electrons for imaging, which may compromise the overall SNR of a cryo-EM image. To measure the dependence of the radiation damage rate on the electron energy, Peet et al.15 measured the inelastic scattering cross-section of electrons from graphene and the radiation damage ratio of electrons with energies between 100 kV and 300 kV from 2D-crystals of bacteriorhodopsin and paraffin. Instead of using the measured inelastic scattering section, they used Estar database16 to calculate the information coefficient (IFC). Their results indicate an optimal voltage of 100 kV for 30-nm-thick samples, with the signal improving by 5% compared with that at 300 kV. For 10 nm-thick samples, their result indicates a 25% improvement in signal by lowering the acceleration voltage to 30 kV. However, the radiation damage estimated from the critical dose was notably different between the crystal sample and the vitrified non-crystal sample17. As a result, the radiation damage measured from the 2D-crystals may be inaccurate and leads to a different optimal voltage. Therefore, an accurate electron energy-dependent radiation damage rate for cryo-EM specimens remains undetermined. The currently available DDD cameras for cryo-EM images were designed for detecting 300-kV electrons18. Their DQE drops with decreasing acceleration voltage19. Counting algorithms20,21 have been developed to determine the position of the electron entering the camera and to normalise its intensity. Some DDDs implement a counting algorithm that calculates the position of the centre of mass/gravity for each signal as the position of the incident electron19. The accuracy of this position has been shown sometimes at a sub-pixel level22. If the column charge from the incident electron is isotropic, the centre-of-mass position is an 2 effective approach in determining the incident position. However, if the incident electron is back-scattered or stimulates numerous of pixels, the centre-of-mass position may diverge from the true incident position. In contrast, a hybrid pixel camera23–26 (HPC) has a higher DQE for lower energy electrons and has the potential to replace the DDD in low-acceleration-voltage cryo-EM. However, current HPCs have significantly larger physical pixel size than DDD cameras, meaning fewer pixels on a single chip. Because of the large physical pixel size of the HPC and low DQE at 300 kV27, the low data throughput limits the usage of the HPC in cryo-EM. For instance, one HPC26 that was used for recording 100-kV cryo-EM images had only approximately 3% of the number of pixels in a current 4k-DDD because its physical pixel size is 75 µm, producing only 3% of the throughput during data collection with the same magnification. The advantage of throughput in DDDs suggests that improvements in the detection efficiency of lowenergy electrons for existing DDD cameras is valuable and re (...truncated)