Research Article: Temperature-dependent macromolecular X-ray crystallography

Date Published: April 01, 2010

Publisher: International Union of Crystallography

Author(s): Martin Weik, Jacques-Philippe Colletier.


The dynamical behaviour of crystalline macromolecules and their surrounding solvent as a function of cryo-temperature is reviewed.

Partial Text

Macromolecular X-ray crystallography has greatly benefited from several innovations at the end of the last century, including the implementation of cryo-methods (Hope, 1990 ▶; Teng, 1990 ▶; Garman & Schneider, 1997 ▶; Garman & Owen, 2006 ▶) and the availability of brilliant X-ray beams from third-generation synchrotron sources. The timing was fortuitous, since the widespread use of the latter would not have been possible without the former. Indeed, the deleterious effects of X-ray irradiation on crystalline proteins were recognized early on (Blake & Philips, 1962 ▶) and make room-temperature experiments with modern synchrotron-based X-ray beams difficult, if not impossible. By flash-cooling a macromolecular crystal to 100 K its lifetime in the beam is increased by about 100-fold (Nave & Garman, 2005 ▶; Southworth-Davies et al., 2007 ▶) because the diffusion of the radicals created by X-ray irradiation is limited in the glassy matrix of the crystal solvent. Another beneficial effect of cryocooling originates from reduced dynamic disorder. For example, about twice as many water molecules are detected at cryo-temperature compared with room temperature in protein structures determined using X-ray (Nakasako, 1999 ▶) or neutron crystallography (Blakeley et al., 2004 ▶). Today, more than 90% of all macromolecular X-­ray crystal structures are determined from data collected at 100 K (Garman & Owen, 2006 ▶).

Over the past decade(s), intensive effort has been invested in the exploration of protein structures and rightly so. Because the delicate balance between both structural and dynamical aspects forms the basis of biomolecular function, efforts are now multiplying to uncover the ‘dynamic personalities of proteins’ (Henzler-Wildman & Kern, 2007 ▶). Studying protein motions at subzero (°C) temperatures is a valuable approach that permits the slowing down and teasing apart of the multitude of motions that otherwise occur simultaneously under physiological conditions (Parak, 2003 ▶). This dynamical complexity stems from the multidimensional energy landscape formed by the conformational substates accessible to a protein and its surrounding solvent (Frauenfelder et al., 1991 ▶). Macromolecular motions lead to interconversions between substates and hence ‘bring a protein to life’. When a hydrated biological macromolecule is cooled to cryo-temperatures, anharmonic macromolecular motions cease at the so-called dynamical transition, which occurs at a temperature of between about 180 and 220 K. The dynamical transition occurs in solution as well as in powder and crystalline samples of proteins, RNA and DNA. It was first discovered by researchers using Mössbauer spectroscopy to probe haem-iron movements in myoglobin (Parak et al., 1982 ▶) and has subsequently been studied by other experimental techniques including neutron scattering (Doster et al., 1989 ▶; Ferrand et al., 1993 ▶) and X-ray crystallography (see §4 below). In addition to its importance as a prominent feature in the low-temperature physics of biological macromolecules, the dynamical transition has been linked to the onset of biological activity (Rasmussen et al., 1992 ▶; Lichtenegger et al., 1999 ▶; Ostermann et al., 2000 ▶). However, certain enzymes are active below the dynamical transition (Daniel et al., 1998 ▶), or are at least able to undergo part of their catalytic cycle (Heyes et al., 2002 ▶; Durin et al., 2009 ▶).

Cryocrystallographic experiments require that macromolecular crystals are flash-cooled in a cryogen such as liquid (63–77 K) or gaseous (typically 100 K) nitrogen, liquid propane (83–231 K) or liquid ethane (90–185 K). The goal of the rapid temperature decrease allowed by the flash-cooling is the avoidance of crystalline ice formation in the water fraction of the crystal solvent. The change in density associated with water crystallization disturbs the crystal packing and results in a deterioration in the diffraction quality. In order to avoid crystalline ice formation, the solvent needs to be vitrified to an amorphous state before the water molecules have had the time to reorient and diffuse to form a crystalline arrangement. The higher the viscosity of the solvent and the more pronounced the solvent confinement by the macromolecules, the higher the temperature of the solvent glass transition and the easier it is to avoid crystalline ice formation by vitrification. In most cases the solvent viscosity has to be raised above that of the mother liquor in which the crystal grew by the addition of penetrating cryoprotectants such as glycerol, low-molecular-weight polyethylene glycol or salts (Garman & Schneider, 1997 ▶). For some crystalline proteins, crystalline ice formation is absent during flash-cooling even without the addition of penetrating cryoprotectants. In those cases, the viscosity of the mother liquor confined in the crystal is already sufficiently high to allow vitrification by flash-cooling. Recently, it has been reported that crystalline ice formation does not occur in thaumatin crystals without penetrating cryoprotectants when the temperature is decreased from 300 to 100 K at a very slow rate (0.1 K s−1; Warkentin & Thorne, 2009 ▶). The potential interest of this new slow-cooling procedure in kinetic crystallography is discussed in §4. Crystalline ice formation in the absence of penetrating cryoprotectants can also be pre­vented by flash-cooling protein crystals under high pressure (200 MPa; Kim et al., 2005 ▶). Once the solvent has been rendered amorphous, the protein crystal is in a metastable state at cryo-temperatures; it has ‘fallen out of thermodynamic equilibrium’. What happens when a flash-cooled protein crystal is warmed above 100 K, viz. the tem­perature at which most cryocrystallographic data are collected? A short summary of the behaviour of flash-cooled pure water is a prerequisite for understanding the more complex case of protein crystals.

The structure of myoglobin determined at various temp­eratures between 220 and 300 K provided the first evidence that dynamical information could be obtained from protein crystallography (Frauenfelder et al., 1979 ▶). Subsequently, several other protein structures have been studied at more temperature points and in a broader temperature range from 80 to 300 K (Singh et al., 1980 ▶; Hartmann et al., 1982 ▶; Parak et al., 1987 ▶; Tilton et al., 1992 ▶; Kurinov & Harrison, 1995 ▶; Nagata et al., 1996 ▶; Teeter et al., 2001 ▶; Joti et al., 2002 ▶; Edayathuman­galam & Luger, 2005 ▶; Schmidt et al., 2009 ▶; Kim et al., 2009 ▶). Crystallographic B factors (Debye–Waller factors) can also provide some insights into protein dynamics. Indeed, atomic mean square displacements 〈x2〉 extracted from B factors (〈x2〉 = B/8π2) stem from both dynamic and static disorder. Extrapolating the temperature-dependence of B factors to 0 K provides an estimate of the static contribution. In the case of the crystalline haemprotein nitrophorin 4, the contribution of static disorder to the B factor averaged over all non-H main-chain atoms was 40% and 65% at room temperature and 100 K, respectively (Schmidt et al., 2009 ▶). The temperature-dependence of averaged B factors, however, differs from protein to protein. A linear behaviour has been observed for nitrophorin 4 (Schmidt et al., 2009 ▶) and myoglobin (Parak et al., 1987 ▶; Chong et al., 2001 ▶), whereas a biphasic behaviour of the temperature-dependence of B factors with a kink at a temperature between 150 and 200 K has been reported for ribonuclease A (Tilton et al., 1992 ▶), crambin (Teeter et al., 2001 ▶) and lysozyme (Joti et al., 2002 ▶). The kink has been interpreted as a manifestation of the protein dynamical transition from harmonic to anharmonic motions and in the case of ribonuclease A it has been shown that the substrate binds to the active site above but not below the transition temperature (220 K; Rasmussen et al., 1992 ▶). Likewise, the water structure at the surface of crystalline crambin decreased above the transition (200 K). Does the linear temperature-dependence of B factors in nitrophorin and myoglobin indicate the absence of a dynamical transition in these proteins? Joti and coworkers offered an explanation for the apparent difference in the temperature-dependence of B factors in different protein crystals by arguing that a dynamical transition can take place despite linearity of the B factors at temperatures around 200 K (Joti et al., 2002 ▶). If the same set of conformational substates in the energy landscape of the crystalline protein is occupied throughout the entire temperature range studied, a dynamical transition cannot be observed by examining crystallographic B factors. In contrast, a transition can be observed when certain substates are depleted at lower temperature. Also, inspecting the B factors of individual amino acids might reveal non­linearity despite there being a linear behaviour of B factors averaged over the entire protein, indicating a local change in populated substates at the dynamical transition. Indeed, a reduction in the number of alternate side-chain conformations is often observed in protein structures determined at cryo-temperatures compared with structures determined at room temperature (Parak et al., 1987 ▶; Dunlop et al., 2005 ▶).

X-ray irradiation of macromolecular crystals during crystallo­graphic data collection leads to a decrease in diffraction quality and to specific damage to the macromolecules {see Proceedings of the Second to the Fifth International Workshops on X-ray Damage to Crystalline Biological Samples published in special issues of the Journal of Synchrotron Radiation [Vol. 9, Part 6 (2002), Vol. 12, Part 3 (2005), Vol. 14, Part 1 (2007) and Vol. 16, Part 2 (2009)]; for a review, see Ravelli & Garman, 2006 ▶}. Two types of damage are distinguished: primary and secondary. The former results from the inter­action of an X-ray photon with atoms in the sample, leading to the ejection of a highly energetic electron as a result of the photoelectric effect, which is the dominant inelastic event at the photon energies used in macromolecular crystallo­graphy (Murray et al., 2005 ▶). Primary radiation damage is temperature-independent (Teng & Moffat, 2002 ▶). Secondary damage arises from the many secondary radicals created by the primary photoelectron. Radiolysis of water plays a prominent role among secondary events and leads to a variety of radicals, including hydrated electrons (e−aq), hydroxyl radicals (), atomic hydrogen and protons.

Dynamical transitions at cryotemperatures in proteins and in their surrounding solvent have been proposed to be linked to biological function, as outlined in §2. Temperature-controlled protein crystallography can thus be exploited to generate, trap and structurally characterize macromolecular intermediate states (Ringe & Petsko, 2003 ▶) by combining reaction triggering with appropriate temperature profiles. Together with real-time Laue diffraction close to room temperature, temperature-controlled crystallography is part of the kinetic crystallography toolbox that provides structural biologists with means to address macromolecular function via crystallography. Temperature-controlled kinetic crystallography either follows a trigger–cool or a cool–trigger sequence. In the former, reaction initiation is achieved at room temperature, followed by trapping of the generated intermediate state by rapidly lowering the temperature to 200 K or below. In the latter, the crystalline macromolecule is first flash-cooled and the reaction is then initiated. A reaction initiated at low temperatures, e.g. at 100 K, can only proceed when the protein flexibility is enhanced by raising the tem­perature, typically to above the dynamical transitions of the solvent and protein (Weik, Ravelli et al., 2001 ▶; Kim et al., 2009 ▶). Several ways exist of triggering a reaction, including the irradiation of endogenous or exogenous photosensitive macromolecules with UV–visible light, the diffusion of small molecules such as substrates or products and X-ray irradiation creating radicals and specific bond cleavage. Kinetic crystallo­graphy greatly benefits from complementary spectroscopy techniques, such as offline (Bourgeois et al., 2002 ▶) and online (McGeehan et al., 2009 ▶) UV–visible fluorescence and absorption (Pearson et al., 2004 ▶; De la Mora-Rey & Wilmot, 2007 ▶), Raman (Carpentier et al., 2007 ▶), EPR (Utschig et al., 2008 ▶) and X-ray absorption spectroscopies (Hough et al., 2008 ▶). Extensive recent reviews of kinetic crystallography exist (Petsko & Ringe, 2000 ▶; Bourgeois & Royant, 2005 ▶; Bourgeois & Weik, 2009 ▶; Hirai et al., 2009 ▶) and we focus here on temperature-controlled crystallography using X-ray irradiation as a reaction trigger.

Macromolecular crystallography as a function of temperature currently comprises a small niche of experiments that can be enlarged. The possibility of performing slow-cooling experiments (Warkentin & Thorne, 2009 ▶) has already been discussed in §4. Comparing protein structures determined during slow cooling and during slow heating after flash-cooling might teach us more about the ensemble of conformational substates trapped in a flash-cooled crystalline protein. Temperature-controlled crystallographic experiments could also be carried out with neutrons instead of X-rays. Neutron crystallography allows the visualization of protons, which can be of interest for the interpretation of enzymatic intermediate states. Temperature-controlled kinetic neutron crystallography will be more accessible when more open-flow cooling systems are available on neutron diffractometers. Another perspective is to multiply temperature-controlled X-ray crystallographic experiments on membrane proteins (Hirai et al., 2009 ▶). To this end, carefully characterizing the temperature-dependent X-ray diffraction of membrane-protein crystals and their lipid and/or detergent matrix will be beneficial. Neutron spectroscopy experiments have indeed shown that lipid rather than water dynamics control the dynamics of membrane proteins (Wood et al., 2007 ▶). There is certainly a need to further explore the temperature-dependence of X-ray radiation damage to macromolecular crystals and their components. In crystallo spectroscopic techniques (UV–vis, Raman, EPR, XAS etc.) are a valuable complement to crystallography in this context.