Date Published: March 01, 2016
Publisher: International Union of Crystallography
Author(s): Robin L. Owen, Darren A. Sherrell.
Both site-specific radiation damage and heavy-atom derivatization result in small changes to the intensity of reflections. The size of the change owing to each is calculated and compared for individual reflections.
Radiation damage in macromolecular crystallography (MX) is an inevitable and, for the most part, unwelcome aspect of data collection. In many cases the presence of radiation damage is made clear by changes in metrics such as the diffracting power: a common observation during data collection at synchrotron sources is a decrease in the diffracting power of a crystal as the collection of a data set progresses. The decay in diffracting power is particularly striking at higher resolution, with spot patterns in outer resolution shells often fading to become invisible by the end of data collection. Changes in the crystal mosaicity, unit-cell volume or B factor may be less immediately visible, but nonetheless retrospective analysis can also reveal the heavy footprint left by X-rays.
This article is not intended to be a comprehensive review of radiation damage in MX: for this, the reader is referred to recent excellent reviews such as Holton (2009 ▸) and Garman (2010 ▸). A fundamental observation is that at cryotemperatures damage is proportional to the energy per unit mass, or dose, absorbed by a crystal. At temperatures above ∼200 K, experiments have shown that damage can be, at least partially, outrun (Warkentin et al., 2013 ▸; Owen et al., 2014 ▸). Absorbed dose is still however the dominating factor: all crystals have a finite life dose in the X-ray beam. Below, we will primarily consider the dose scales and limits associated with cryo-crystallography.
One of the most familiar symptoms of radiation damage is the fading of diffraction patterns, especially at high resolution. Global effects such as these occur on the order of tens of megagrays, with Howells et al. (2009 ▸) suggesting a criterion of 10 MGy per Å resolution based on a number of studies in the literature. Thus, when collecting a data set to 3 Å resolution a dose of ∼30 MGy can be tolerated, while for a 1 Å resolution data set this reduces to ∼10 MGy. When compared with the doses in Tables 1 ▸ and 2 ▸, these dose limits appear to be rather generous, with total exposure times of several (tens) of seconds possible before they are reached. As outlined in §1, the dose scales associated with site-specific damage are significantly less than this however, with damage observed on dose scales of tens of kilograys in some cases.
In the examples described here, the soluble protein FcγRIII is used as a test case (Zhang et al., 2000 ▸). The structure of FcγRIII has been deposited in the PDB with accession code 1fnl, with unit-cell parameters a = 67, b = 86, c = 36 Å in space group P21212. Each FcγRIII monomer consists of 175 amino acids, with no methionines and four cysteines. The four cysteine residues form two disulfide bridges. The unit cell contains approximately 8000 atoms, of which 16 are sulfur.
Certain amino acids such as glutamate, aspartate and tyrosine are particularly susceptible to localized damage, but one of the most well recognized calling cards that X-rays leave in an electron-density map is changes at disufide bonds (Burmeister, 2000 ▸; Weik et al., 2000 ▸; Ravelli & McSweeney, 2000 ▸). The rapid formation of large numbers of solvated electrons upon X-ray irradiation, and the ability of these electrons to travel through a crystal, provides a means for the rapid cleavage of disulfide bonds even at 100 K (Sutton et al., 2013 ▸).
It is well known that the addition of a small number of heavy atoms to a protein results in changes in the intensity of reflections. A simple rule of thumb for the size of the change 〈ΔI〉/I induced iswhere Ne is number of heavy atoms, Np is the number of non-H protein atoms, fh is the atomic number of the heavy atom introduced and feff is the mean atomic number of protein atoms (∼6.7; Crick & Magdoff, 1956 ▸). Applying this formulation to FcγRIII, the expected change 〈ΔI〉/I for both a mercury and a platinum derivative is ∼0.45 (assuming two heavy atoms per monomer in each case).
Non-isomorphism between crystals can arise for a number of reasons. These may originate from variations in crystallization conditions, (de)hydration or the cryocooling process in addition to the main topics of this manuscript: radiation damage and derivatization. There are two dominant sources of non-isomorphism: changes in unit-cell parameters or changes in the relative positions of protein molecules. Crick & Magdoff (1956 ▸) showed that a relatively minor change in unit-cell parameters of 0.5%, or a shift of a protein molecule by 0.1 Å, will cause an average change in intensity of >15%. If non-isomorphism is introduced by derivatization, the expected intensity change owing to anomalous scattering (2) must therefore be somewhat greater than this if experimental phasing is to succeed. Non-isomorphism must also be considered for a single crystal type or derivative, as it is frequently not possible to obtain sufficient data from a single crystal, and small changes in the derivatization process, crystal handling or cryocooling may induce crystal-to-crystal variation. Radiation damage can also cause non-isomorphism to be introduced during the experiment, although the resulting size and rate of change of unit-cell parameters can be unpredictable, even for a particular crystal form (Murray & Garman, 2002 ▸; Ravelli et al., 2002 ▸).
§5, §6 and §7 illustrate how radiation damage, derivatization and non-isomorphism can affect structure factors. Radiation damage in macromolecular crystallography results in global effects, such as a decrease in diffracting power, on dose scales of the order of tens of megagrays and site-specific damage, which occurs much faster, on dose scales of tens of kilograys. Both of these dose scales can easily be, and routinely are, surpassed during data collection at synchrotron sources, with the result that both types of damage can be considered to be an inevitable part of structure determination. Site-specific damage results in different rates of decay for individual reflections and we have shown that the size of these changes in intensity is comparable to those introduced by derivatization. The primary way in which radiation damage can be reduced is through reduction of the dose absorbed by a crystal during data collection. This can be achieved most simply by decreasing the exposure time/increasing the attenuation of the X-ray beam, but also by increasing the beam size to match that of the crystal (if applicable), translating the crystal through the beam during data collection or through the merging of low-dose partial data sets from multiple crystals. Prior to data collection, efficient backsoaking of derivatives removes heavy atoms which contribute greatly to the absorbed dose but do not add anomalous signal. Non-isomorphism can be minimized through consistent crystal handling and treatment during, for example, the mounting or cryocooling processes. If multiple crystals are used then software such as BLEND (Foadi et al., 2013 ▸) can be used to group crystals into clusters with similar unit-cell parameters.