X-ray crystallography (MX) and electron cryomicroscopy (cryoEM) are the dominant methods for determining the structures of biological macromolecules. Both are fundamentally limited in the same way. The radiation that produces the data (X-ray photons and electrons, respectively) damages biological material. This limits the amount of data that can be obtained before a sample is destroyed. Two recent papers show that a judicious choice of energy maximizes the amount of data obtained per unit damage.
Joshua Dickerson and Elspeth Garman from the University of Oxford have run simulations to show that the diffraction power of a crystal grows faster than the absorbed dose as the photon energy increases to about 30 keV (Reference 1). This is quite a bit higher than the 12.4 keV most commonly used at synchrotrons. In the past, collecting data near 30 keV would not have been sensible because of poor detector properties.
The arrival of large-area Hybrid Photon Counting with CdTe sensor detectors changes everything. Experimenters can choose an energy just below the absorption edge of cadmium (26.7 keV) to take advantage of the detector's near-perfect quantum efficiency and the high-energy photons' low relative damage. Dickerson and Garman estimate 26 keV photons are scattered 1.6 times better than 12.4 keV photons relative to radiation damage. The advantage can reach five times for crystals only a few micrometers across where photoelectrons can escape the confines of the crystal before causing damage (Figure 1).
Mathew Peet, Richard Henderson and Chris Russo from the Laboratory of Molecular Biology in Cambridge have shown that questions of energy are pertinent to radiation damage in cryoEM as well (Reference 2). High-resolution data are normally collected at 300 keV. At 100 keV, the radiation damage per electron is 1.6 times higher than at 300 keV but elastic scattering – the basis of the measured data – increases twofold. Put the two together and you get a quarter more elastic scattering per unit damage at 100 keV compared to 300 keV.
Anything lower than a factor of two is probably not a robust strategy to mitigate radiation damage, but this is not the only point here. A microscope operated at 100 keV is much cheaper to manufacture and operate than a 300 keV instrument. Getting data of the same or even slightly higher quality from a much simpler instrument is thus an attractive proposition for any laboratory interested in structural biology.
In a follow-up paper, the groups of Russo and Henderson have recently described such a microscope and presented the kind of data that can be obtained with it (Reference 3). The keys to high performance at 100 keV are thin ice, a low vacuum, a field emission gun for high beam coherence, and an optimized detector.
The back-thinned CMOS detectors that have been so powerful for data collection at 300 keV are poor options at 100 keV because lower-voltage electrons will scatter widely even in the thin sensor. Instead, Russo and Henderson used an EIGER X 500K with a thick silicon sensor and relatively large pixels. Data acquisition at low dose and high frame rates (up to 9 kHz) allowed them to count individual electrons and get a high quantum efficiency. With this setup, they solved the structure of DPS, a 220 kDa protein, to 3.4 Å resolution (Figure 3).
A similar follow-up to the Dickerson and Garman paper is still outstanding. The case for collecting MX data with a CdTe detector and 26 keV photons is compelling but has not been made experimentally yet. At synchrotron beamlines where such high-energy radiation is available, these experiments are currently being performed. The upcoming diffraction-limited storage rings with their higher flux at higher energy will be ideally suited to exploit this opportunity further.
A couple of papers published only a few months apart illuminate the virtues of unconventional energies for MX and cryoEM data collection. It is curious that both approaches strictly depend on a suitable detector. At DECTRIS, we offer you the right detector for the right energy. For structural biologists, this means better data whatever the choice of method.
Figure 1. The best energy for MX is 26 keV. The number of elastically scattered photons increases faster than the observed damage with the photon energy increasing to around 30 keV. The effect is largest for microcrystals ≤ 2 µm across. The data are normalized for the quantum efficiency of 750 µm CdTe (dotted line, right axis). See reference 1 for more details.
Figure 2. Thin samples benefit from data collection at 100 keV. The graph shows the electron energy to maximize the information content depending on sample thickness. See reference 2 for more details.
Figure 3. Structure of DPS determined at 100 keV. (a) Typical micrograph of DPS after motion correction. (b) 2D class averages. (c) 3D reconstruction of DPS. (d) Gold-standard FSC plot showing a resolution limit of 3.4 Å. (e) Mollweide projection showing nearly uniform sampling of particle views. See reference 3 for more details.