Origin and temperature dependence of radiation damage in biological samples at cryogenic temperatures

Results

In this study global radiation damage was analyzed with respect to various crystallographic parameters such as the decay of the recorded Bragg intensities, changes of the unit cell volume, and changes of the crystal mosaicity of cubic insulin crystals as a function of the absorbed x-ray dose at six different temperatures ranging from 5 to 160 K and of elastase crystals at five temperatures between 5 and 100 K. Cubic porcine insulin and orthorhombic elastase crystals were chosen as model systems for different reasons. Cubic insulin crystals are of reproducibly good quality and diffract to high resolution. Elastase is a larger protein with a molecular weight of 26 kDa compared to cubic insulin with a weight of 6 kDa and was chosen to verify the results obtained for the insulin crystals. Data were analyzed as previously reported (24). Metric parameters strongly depend on the experimental conditions and thus should be regarded with care (9, 30). Therefore, refinements of atomic coordinates against all recorded datasets were carried out. The increase of the R_free factors from these refinements with dose can be regarded as a measure for the integrity of the structure and was taken as an additional radiation damage parameter that is much less sensitive to the experimental conditions than the changes of the mosaicity and the unit cell volume.

For both cubic insulin and elastase the decay of the absolute intensities in the 1.5–2.5 Å shell is most prominent at temperatures of 160 and 100 K, respectively, and is reduced at lower temperatures (Fig. 1A). Unexpectedly, for both systems a local minimum of the decay is observed at 50 K with a 23% reduced decay for insulin and an 18% reduced decay for elastase compared to 100 K.

In both systems the increase of the R-factors (R_free) with dose show a minimum at 50 K (Figs. 1B and 2). For cubic insulin the R_free increase is reduced by 74% and for elastase by 62% at 50 K compared to 100 K. Reducing the temperature further does not lead to an additional reduction of the R_free increase.

In a similar way to the intensity decay, the unit cell volume (Fig. 1C) and the crystal mosaicity (Fig. 1D) change with increasing absorbed dose. For insulin a larger increase of the unit cell volume was observed at lower temperatures indicating a greater distortion of the crystal lattice at these temperatures. However, this does not give rise to a faster decay of the intensities (Fig. 1A). In the case of elastase, the unit cell volume increase is slightly reduced while going from 100 to 50 K and rises again between 50 and 30 K, a trend that is also observed for insulin. The mosaicity increase of insulin shows a local maximum at 30 K (Fig. 1D). Mosaicity changes of elastase crystals show a similar behavior with dose but are generally smaller when compared to those of insulin.

In an attempt to directly observe radiation induced lattice distortions of the samples as a function of temperature, small-angle x-ray scattering (SAXS) measurements were performed on cubic insulin crystals. SAXS is sensitive to electron density variations. Contrast in crystalline materials arises from lattice imperfections like grain boundaries, where solid crystalline material is in contact with gas, liquid, or vacuum. In the SAXS measurements on cubic insulin crystals, an increase of the diffuse scattering signal as a function of dose is observed. Fig. 3A and B show the change of the integrated SAXS signal for q-ranges from 0.02 to 0.27 Å^-1 as function of dose for different data collection temperatures. The signal at smaller q-values than 0.02 Å^-1 could not be recorded due to geometrical restrictions of the experimental setup. With the exception of 160 K, where an immediate steep incline of the SAXS signal is observed, an almost linear and relatively moderate increase with dose is observed. The slopes in this region reflect a similar behavior with temperature as found independently for the crystal mosaicity increase derived from the diffraction data (Fig. 3C). Again, the steepest slope is found for the 30 K data, whereas the increase is significantly less pronounced at 5 K. At 50 and 100 K the increase with dose is moderate. The stronger linear increase at 5 and 30 K arises mainly from a strong intensity increase at q-values smaller than 0.03 Å^-1, which is indicative for particle sizes in the range of more than 20 nm or larger than 3 unit cells (Fig. 3C). This is likely caused by the misalignment of very small mosaic blocks with regards to the crystal lattice, as a result of a space requiring process. After a certain dose is accumulated, and depending on the temperature, the signal begins to increase rapidly indicating a very fast breakdown of the crystalline order after reaching this critical dose (Fig. 3B). This critical dose linearly decreases with temperature.

The increase of the unit cell volume in combination with the mosaicity changes indicate the generation of a radiolysis product that negatively affects the crystalline order. The fact that the unit cell volume increases with decreasing temperature further suggests that the observed temperature dependence is either caused by a more extended radiation chemistry at lower temperatures, or is related to diffusion processes. Warming macromolecular crystals after x-ray data collection above 160 K, which presumably is their glass transition temperature (12), results in a sudden release of a gas of so far unknown composition (Fig. 4). Similar observations have been made in cryo-EM, where the formation of molecular hydrogen bubbles in irradiated organic samples has been described (31, 32). To investigate if the generation of this gas is the main cause for crystal lattice distortions, the existence and composition of gaseous radiolysis products was verified by irradiating a set of liquid test substances, including pure water, hexane, acetone, methanol, ethanol, ethylene glycol, an aqueous 0.1 M sodium acetate solution, and a 10% (wt/vol) suspension of yeast in water with x-rays at room temperature (Table S1). In all cases H₂ was the main component of the gaseous phase (> 80%), except for pure water, where the produced gas volume was not large enough for subsequent analysis. Although this experiment was carried out at room temperature, our findings are in agreement with results previously published from cryoelectron energy loss spectroscopy, where the formation of hydrogen bubbles in a glycerol solution as a result of electron beam irradiation at cryogenic temperatures was reported (31, 32).

The gas volumes released after irradiation of water, ethylene glycol, and a 10% (wt/vol) aqueous lysozyme solution at 5, 100, and 150 K were also determined quantitatively (Fig. S1). The amount of gas formed is, with the exception of water and by correcting the water contribution for the aqueous lysozyme solution, temperature independent. The results strongly suggest that the majority of the molecular hydrogen originates from organic compounds present in the irradiated sample and not from the water itself (Table S1). By far the largest gas volumes are found for ethylene glycol, where about 400 molecules of H₂ are formed from every 12.5 keV photon absorbed.

Specific radiation damage at the disulfide bonds of insulin was analyzed by inspection of electron density maps, which were calculated after refinement of the atomic coordinates against diffraction data collected at temperatures of 5, 50, and 100 K. Fig. 5 depicts electron densities around one of the three disulfide bridges with accumulated x-ray doses of 9, 34, and 60 MGy. At 5 and 50 K the 2Fo-Fc maps are continuous up to an accumulated x-ray dose of 60 MGy (Fig. 5Top and Middle rows). At 100 K, the same disulfide bridge is much more susceptible to specific radiation damage. Already at 34 MGy, the 2Fo-Fc electron density around the sulfur atom of the cysteine 7B sulfur atom starts to vanish, whereas the negative peak in the negative Fo-Fc map appears at the same site. After an accumulated dose of 60 MGy, the integrity of the disulfide bridge is severely affected, as judged by a loss of 2Fo-Fc density and the presence of negative Fo-Fc density around the cysteine residue 7B (Fig. 5, Bottom row). The occupancy decay at 50 K is reduced by a factor of about 4 compared to 100 K. Cooling down to 5 K did not yield any further improvements. This observation follows the trend reported in recently published XANES data (25, 27), where photoreduction of metal binding sites in enzymes was found to be reduced at temperatures below 100 K.

In the case of insulin, only the solvent exposed disulfide bridge (residues A7 and B7) displays significant dose-dependent changes in electron density maps. The other two disulfide bridges, which are buried inside the molecule, remain almost intact, even at accumulated doses of 60 MGy, and exhibit only a very small temperature-dependent susceptibility to radiation damage. Electron density maps at 130 and 160 K follow a similar trend as mentioned above: The quality of the 2Fo-Fc difference maps is gradually decreased at 130 and 160 K. At 30 K the electron density maps look similar to those calculated for the 50 K data.

Discussion

Data collection at temperatures below 100 K drastically reduces specific radiation damage to disulfide residues by a factor of about 4, and structural details are much better preserved at these temperatures. However, such a strong reduction was not observed for global radiation damage, which is only reduced by about 18% in the case of elastase and 23% in the case of insulin across a temperature range from 100 to 50 or 5 K, as indicated by the results shown in Fig. 1. These values agree well with a previous study, where a 7% reduction of radiation damage at 15 K compared to 100 K was observed (24).

In our x-ray radiolysis experiments, hydrogen gas was identified as the main reaction product. As discussed in the literature, radiolysis of aliphatic organic molecules leads to C-H rather than C-C bond breakage (33, 34), which was also observed indirectly in the experiments presented in this study. Direct breakage of C-C bonds caused by radiolysis of hexane or ethylene glycol would yield smaller fragments such as ethane and methane. These fragments, however, were only minor constituents of the analyzed products (Table S1). Moreover, the results of the gas volume measurements are in agreement with previous results from electron microscopy, reporting that the gas formation takes place at the location of the protein and not in the solvent (22).

We conclude that the hydrogen formed during irradiation is responsible for the observed gap between the large reduction of specific damage and the better preservation of the structure (as observed by the reduced increase of the R_free values) and the relatively small reduction of global damage (as observed in the decay of the mean intensities) with decreasing temperature. The hydrogen exerts a disruptive force on the internal order of macromolecular crystals at cryogenic temperatures. At 160 K, hydrogen can easily diffuse inside the crystal and accumulate at vacancies or lattice imperfections as illustrated in Fig. 6 (Left). This leads to an increase of the crystal mosaicity (Fig. 1D). By lowering the temperature to 50 K, the mobility of molecules and radicals is reduced, which is reflected in a lower mean intensity decay at lower temperatures (Fig. 1A). Only hydrogen can still diffuse within the crystal, but diffusion rates are lower and more hydrogen is trapped locally inside the crystal (35). This results in a stronger unit cell volume expansion with decreasing temperature as observed in the case of insulin (Fig. 1C). At the same time, the destructive effect of the hydrogen gas is reduced with decreasing temperature. Assuming that H₂ at grain boundaries and lattice imperfections acts as an ideal gas, its molecular gas volume should become smaller with decreasing temperature and therefore exert weaker forces on the crystalline lattice. This in turn results in a lower mosaicity increase at lower temperatures as independently observed in x-ray diffraction and SAXS experiments for insulin crystals (Figs. 1D and 3A). Such an ideal gas behavior is indeed observed in the linear dependence of the critical dose in the SAXS experiments with temperature (Fig. 3B).

At 30 K, hydrogen becomes immobile and remains trapped locally inside the crystal (35). This results in an increased deterioration of the crystalline lattice as observed in higher unit cell expansion, a faster decay of the mean Bragg intensity at 30 K compared to 50 K, and a faster mosaicity increase (Fig. 1). This effect is illustrated on the Right side of Fig. 6.

The results presented in this paper show that hydrogen radical abstraction and the subsequent formation of molecular hydrogen is the major cause for global radiation damage in macromolecular crystallography at cryogenic temperatures. The hydrogen gas formed is distorting the sample, whereas intramolecular bonds between non-hydrogen atoms and hence the main geometrical features of the molecules themselves remain more or less intact. In the case of macromolecular crystallography, the formation of hydrogen gas results in a loss of crystallinity, which is expressed in a decay of the Bragg intensities and an increase of the unit cell volume and the crystal mosaicity. In the case of nonperiodic biological objects, like cells or large molecular assemblies, this unavoidable process will result in unfavorable sample deformations. Although our results are mainly based on x-ray diffraction experiments, we assume that they are of general validity for other methods employing high doses of ionizing radiation that create low energy electrons inside the sample, such as x-ray imaging techniques and transmission electron microscopy, because of the same underlying physical and chemical processes. In electron microscopy much thinner samples are used, and most of the hydrogen gas formed can evade the sample at liquid nitrogen and higher temperatures. At helium temperatures diffusion rates are close to zero, leading to the local formation of hydrogen bubbles (22).

The current standard data collection temperature of 100 K has been proven not to be the optimal choice. Global radiation damage is reduced by about 18–23% and specific damage up to a factor of 4 when experiments are conducted at temperatures around 50 K.

The reduction of specific damage and the better preservation of the structure should help to better understand the fundamental processes in radiation sensitive metalloproteins and increase the integrity of anomalous substructures used for phasing experiments. The reduction of global damage will allow higher resolution information to be obtained from a given sample in x-ray and electron diffraction or imaging experiments.

Methods

Supplementary Material