Date Published: February 01, 2019
Publisher: International Union of Crystallography
Author(s): Karol Nass.
Research in the area of protein crystallography at X-ray free-electron laser sources is summarized from the perspective of radiation damage, including its mechanisms and effects, ways to minimize it and a comparison with the damage observable at synchrotrons.
Macromolecular X-ray crystallography (MX) has been the most powerful approach for obtaining three-dimensional structural information on biological species such as proteins, nucleic acids or viruses at up to atomic resolution which, together with functional studies, is crucial for understanding the mechanism underlying the given biological process (Shi, 2014 ▸). MX requires the use of radiation with wavelengths similar to or shorter than the length scale of atoms in order to yield high-resolution structures. However, this electromagnetic radiation causes damage to the sample as it carries sufficient energy to overcome the binding energies of electrons in atoms and molecules, which results in the ionization and excitation of the atoms in the specimen (Als-Nielsen & McMorrow, 2011 ▸).
Photoabsorption is the main initiator of radiation damage in MX. It leads to the ionization of atoms and, in the case of XFELs, the high pulse intensity and ultrashort pulse duration cause excessive ionization of atoms in the sample that develops with time. As predicted by simulations, very high charge states of ions can be reached in the sample at the end of the XFEL pulse, which ultimately destroys the sample (Caleman, Bergh et al., 2011 ▸; Hau-Riege, 2013 ▸). The X-ray pulse duration from an XFEL source is typically of the order of a few femtoseconds to a few tens of femtoseconds. Each pulse can contain up to 1012 photons. When a typical ∼30 fs pulse with 1012 photons of 12.4 keV energy (1 Å wavelength) is focused to an area of 1 µm2, the resulting surface power density at the sample is ∼6.6 × 1018 W cm−2, which corresponds to a photon density of approximately 104 photons Å−2. Under such extreme irradiation, the consequence of exposure to a non-attenuated and tightly focused XFEL pulse is the rapid ionization of the atoms in the sample to high charge states via (multiple) photoabsorption(s) and electron-impact ionization (Caleman, Bergh et al., 2011 ▸; Young et al., 2010 ▸; Caleman, Huldt et al., 2011 ▸). The ionization rate and the final level of ionization at the end of the pulse depend on the specific cross-sections, which depend on the atomic number, the photon energy and the current ionization state of an atom. The rapidly increasing ionization of atoms within the duration of an XFEL pulse will ultimately lead to destruction of the sample by Coulomb explosion. Therefore, the timescales of radiation damage in macromolecular crystallography at XFELs and synchrotrons are very different owing to differences in the exposure time and the photon flux density. Consequently, damage caused by chemically reactive species, which is one of the main sources of damage in MX at synchrotrons, is avoided at XFELs as the XFEL pulse terminates before these molecules are created; thus, the reactions that they cause in contact with other atoms are not present (Chapman et al., 2014 ▸).
When the ionization level in the crystal is high enough for the ions to feel each other’s electrostatic potential, they start moving away from each other owing to repulsive forces (Barty et al., 2012 ▸). This movement initiates a gradual increase of the disorder parameter of the crystal structure during the pulse duration, which results in a decay of Bragg peak intensities at high scattering angles. Consequently, Bragg peaks at low scattering angles will be observed for longer and thus accumulate more scattered photons than those at high angles. The outcome of this effect is similar to that of the global damage effects observed in macromolecular crystallography at synchrotrons, where the high-resolution diffraction spots disappear from consecutive diffraction images as the dose absorbed by the crystals increases. The difference is that the temporal evolution of this damage effect is recorded in a single frame at an XFEL, whereas it can be spread over multiple frames at a synchrotron. It has been predicted that if the damage was distributed uniformly within the asymmetric units of the crystals, then it could be possible to correct for it by scaling (Barty et al., 2012 ▸). However, another study found that such scaling of data is difficult to accomplish, and attributed this complexity to possible non-uniformity of the radiation-damage dynamics across the sample and argued for the existence of ‘hot spots’ of damage (Lomb et al., 2011 ▸). In addition to induced disorder, changes in the atomic scattering form factors during the pulse can occur owing to the complex ionization dynamics of different atom types during the pulse, which can modify the scattered intensities (Hau-Riege, 2007 ▸). For example, at one ionization per atom on average, the scattering power of atoms is reduced, leading to a decrease of up to 30% in the scattered intensity (Caleman, Huldt et al., 2011 ▸).
It has been estimated that to ionize every atom in a protein crystal of an average composition at the end of a typical XFEL pulse once, an absorbed dose of 400 MGy is required (Chapman et al., 2014 ▸). Since scattering and ionization occur during the entire pulse duration, it is assumed that the scattering signal is obtained from atoms mostly in their intact (pristine) state and that this dose can be used as a threshold marker to obtain signal before any modification of the electronic structure of the sample has occurred (Kern et al., 2015 ▸). Nevertheless, on timescales of several to tens of femtoseconds photoionization of atoms cannot be avoided. Elements with higher atomic numbers such as iron or sulfur are more susceptible to X-ray-induced damage by photoelectron- and electron-impact ionization because they are characterized by higher atomic cross-sections for this type of interaction than lighter elements such as carbon, nitrogen and oxygen, which are the main components of proteins (Henke et al., 1993 ▸). Therefore, it has been predicted that heavy atoms and atoms in their vicinity could form areas of increased localized structural and electronic damage compared with the rest of the protein (Jurek & Faigel, 2009 ▸; Hau-Riege et al., 2004 ▸) and create effects of charge migration from lighter elements to rapidly ionizing heavier elements (Erk et al., 2013 ▸, 2014 ▸). Consequently, these regions are more likely to be structurally damaged than the rest of the protein. Importantly, metalloproteins containing a metal cofactor are involved in many essential processes (e.g. photosynthesis and respiration). Indications of such localized structural and electronic damage have been obtained experimentally (Nass et al., 2015 ▸) and have been predicted by simulations (Hau-Riege & Bennion, 2015 ▸) when using specific conditions. Pulses with a photon energy above the absorption edge of the metal atom in the active site, iron in this case, were focused to submicrometre dimensions and the pulse duration was slightly longer (80 fs) than that typically used in SFX experiments. The ionized S atoms of the 4Fe–4S iron–sulfur clusters in ferredoxin crystals moved in specific directions owing to repulsion forces, as favoured by their geometrical arrangement (Hau-Riege & Bennion, 2015 ▸; Nass et al., 2015 ▸). In the experimental results, one of the two 4Fe–4S clusters in the structure appeared to be more damaged than the other, indicating that the local protein environment plays a role in damage dynamics (Fig. 3 ▸). The increased sensitivity to damage selectivity of heavy elements in protein crystal structures has been used to propose a new phasing method, in which high-intensity XFEL pulses selectively modify the electronic structure of heavy atoms in the protein. This results in shifts of element-specific X-ray absorption edges; therefore, the peak and remote data sets typical for a MAD experiment could be recorded at the same wavelength with high and low pulse intensity (Son, Chapman et al., 2011 ▸). Recently, a theoretical study focused on exploring the possibility of mapping the non-uniform ion distribution in a protein as it undergoes the Coulomb explosion following intense ionization for applications in orientation recovery in single-particle imaging experiments at XFELs was published (Östlin et al., 2018 ▸). It showed the existence of localized hot and cold ‘spots’ of ion density in a protein exposed to an XFEL pulse of high intensity and that the predicted reproducibility of trajectories of carbon and sulfur ions in lysozyme exposed to XFEL radiation varied.
It may be possible to reduce the number of multiple photoabsorption events per atom and consequently the level of ionization in the sample by using pulses shorter than the lifetime of the Auger decay processes. After the initial photoionization and Auger relaxation is complete, two electrons are removed from the outer shells of an atom, leading to a doubly charged ion with all inner-shell electrons filled that is ready for the next photoionization event. However, when the pulses are shorter than the Auger decay lifetime, the generation of high-charge ions is suppressed. For lighter elements, the relatively slow Auger decay process is more favourable than the faster fluorescence relaxation pathway; therefore, when using pulses shorter than Auger processes the ionization level of lighter elements at the end of the pulse can be reduced. This phenomenon has been called frustrated X-ray absorption or intensity-induced X-ray transparency (Young et al., 2010 ▸; Hoener et al., 2010 ▸). In this phenomenon the production of high charge states of atoms via multiple photoabsorptions is suppressed in comparison to longer pulse durations because the core-hole that is left after the first absorption event is not filled for as long as the Auger decay lifetime lasts, reducing the number of inner-shell electrons available for X-ray absorption. For example, the measured lifetime of the Auger decay for iron is 0.55 fs, that for sulfur is 1.3 fs and that for carbon is 10 fs (Campbell & Papp, 2001 ▸). In contrast, when the XFEL pulse is longer than the Auger lifetime, core-excited states have sufficient time to decay, which results in the refilling of inner-shell holes with electrons from outer shells, and sequential multi-photon ionization can occur, possibly removing all electrons from an atom if the pulse is sufficiently intense and long (Young et al., 2010 ▸). Using X-ray pulses shorter than the Auger lifetime of atoms has another advantage for reducing radiation damage by outrunning the development of secondary electron-impact cascades. It has been estimated that one 6 keV photoelectron will lead to the creation of ∼300 secondary electrons via impact ionization cascades before the secondary electrons thermalize (Caleman, Huldt et al., 2011 ▸). Most impact ionization cascades will be completed after tens of femtoseconds; therefore, electron-impact ionization cascades caused by highly energetic photoelectrons released after the initial X-ray absorption would not have enough time to fully develop if sufficiently short pulses were used. This would result in a reduction of the ionization level in the sample and in the reduction of the radiation damage observable during the X-ray pulse. In order to completely outrun the creation of electron-impact ionization cascades created by photoelectrons, the pulse needs to be shorter than the time it takes for the first collision to occur, which depends on the energy of the photoelectron and is typically much less than 1 fs (Son, Young et al., 2011 ▸).
In this review, an overview of radiation-damage processes occurring on ultrafast timescales and a summary of published research articles that have investigated the radiation-damage processes occurring in samples exposed to high-intensity XFEL pulses have been presented. In contrast to decades of research in the field of radiation damage in macromolecular crystallography at conventional X-ray sources, only a handful of articles have investigated this phenomenon at X-ray free-electron lasers. It appears that in the case of SFX most studies have not observed damage effects in electron-density maps or in the X-ray emission spectra when using modest photon flux densities. The degree to which radiation damage will modify protein structures obtained from experiments that use higher flux densities (>1019 W cm−2) with pulses focused to submicrometre dimensions and shorter pulse durations on the few femtoseconds and subfemtosecond timescales remains to be explored. As the number of available XFEL sources and the user community grows, it is expected that this research field will also advance, allowing us to better understand the nature of radiation damage at XFELs and to aid the development of methods to overcome it.