Sean Parkin's Crystallographic Interests

There is a continuum of complexity in crystalline matter from the relatively simple (e.g. elemental metals and simple salts) to the extremely complex (e.g. proteins, nucleic acids and even viruses and the ribosome.)  The structures of all these entities are accessible to varying degrees by X–ray diffraction, but an arbitrary distinction has often split the discipline into ‘small’ and ‘macro’ molecule factions with surprisingly little inter-communication.  Advances in instrumentation, computation and experimental methods constantly serve to erode this division.  While the number of protein structures known at better than 1.2Å resolution grows rapidly, many synthetic chemists still encounter crystals deemed unsatisfactory for one reason or another. Perhaps the most obvious difference between small and macromolecular crystals is the attainable resolution. Nevertheless, some protein crystals do diffract to atomic resolution and can thus be refined like small–molecule structures. My interests lie in methods development and in extracting information from biological samples at a level of detail approaching that of normal ‘small’ molecules. The term “ultra–high resolution” has recently been coined to describe sub-ångstrom resolution protein structures. Unfortunately, this is still somewhat worse than the majority of true small–molecule crystals. In order to keep definitions as precise as possible, I try to limit the use of “ultra–high resolution” to those structures that actually are ultra–high resolution, such as charge–density analyses.

Small-Molecule Crystallography

As Director of the X–ray laboratory in Chemistry at the University of Kentucky, I provide education, training and structural data to organic, inorganic, materials and physical chemists in Chemistry, Pharmacy, Physics and Toxicology at UK as well as advice and ad–hoc training to nearby undergraduate universities and colleges. Much of this work is collaborative in nature and is detailed on the web pages of individual research groups involved. The service is free for academic institutions in the region (or rather we try very hard to absorb the costs!) Crystals from further afield and from industry are charged on a per job basis and is very competitive with other facilities. Data are collected at as low a temperature as possible, usually 90K, but occasionally higher if a destructive phase transition happens to shatter the crystal. A typical structure determination will result in a thermal ellipsoid plot, packing diagrams, and several files (including CIF) that provide all necessary details for publication. For non–routine structures, extra materials will be produced.

Ellipsoid plot	View down a	View down b	View down c
Typical graphics from a routine small-molecule structure determination (Anthony laboratory)

Large-Molecule Crystallography

Low–Temperature Atomic–Resolution Protein Structure: The limiting resolution usually quoted in descriptions of protein crystal structures is a measure of the fineness of detail that can be interpreted in electron density maps. Most small–molecule crystals diffract to better than about 0.8Å and some to well beyond 0.5Å. Crystals of biological macromolecules are different, however. Most diffract to worse than 2.0Å or 3.0Å while only a select few diffract to better than 1.2Å, which is generally considered to be the onset of true atomic resolution.

1.00 Å	1.25 Å	1.50 Å	2.00 Å	2.50 Å	3.00 Å
Detail in electron density maps of actinoxanthin (6) as a function of resolution. The situation is usually somewhat worse than shown here because these maps were simply truncated rather than being limited by the crystal itself. This particular dataset is missing a non-trivial wedge of data, so it is not quite ready for formal publication. Somewhat weaker data to 1.3Å that partially fills the gap exists, and time permitting this will enable the refinement to be finished.

One of the major problems is that high–resolution reflections, which are weak in any case, tend to diminish in intensity due to radiation damage much faster than low or medium resolution reflections. By cooling the crystal to cryogenic temperatures, much of this radiation damage can be suppressed, at least until the end of data collection.

Low temperature is the key to collecting data that preserves the high–resolution limit, but the act of cooling a crystal often causes significant damage in itself. In mid 1991 when I first stepped over from small to macromolecules, there were only a dozen or so entries in the PDB that were 1.2Å resolution or better. Four of these were DNA oligomer structures, a couple were short peptides, and of the protein entries, two were repeats. Some years earlier, Håkon Hope, had adapted his small–molecule cooling methods to biocrystals (1) and had succeeded in collecting data on two small proteins, crambin and BPTI. The beauty and detail present in those first low–temperature atomic–resolution electron density maps of BPTI (3) enabled the whole protein, including several disordered residues and many more water molecules to be modelled and refined.

BPTI N - terminus	BPTI C - terminus	BPTI Arginine 42	BPTI Waters
Representative electron density maps from the 1.1 Å resolution structure of BPTI (3). At low temperature, much of the disorder has been resolved, including the C-terminal residues, which had been hopelessly disordered at room temperature. Well-organized networks of water are also evident.

Given suitable data and a few computational tricks it is even possible to find many hydrogen atoms. Intense X–rays are a necessity in borderline cases, but synchrotron facilities are inconvenient and not always required (2-7). Naturally, increased detail from atomic resolution allows the observation and study of previously unforeseen phenomena.

Experimental Methods and Tools

Crystal manipulation and data collection at low temperature pose problems of their own. To overcome these problems, the appropriate tools and techniques (8) are required. Nowadays, there are several hundred atomic resolution entries in the PDB and nearly all protein data collections are done at low temperature. Recent advances in high–throughput structure determination as part of the structural genomics initiatives have resulted in robotic crystal manipulation devices which rely on Håkon Hope's high thermal mass tong–block concept (8).

Mounting tongs	Tongs schematic	Crystal cooling	Crystal mounting
Tools and techniques for reproducible, reliable crystal mounting and retrieval (8).

Protein–Water Structure Rearrangements and Crystal Annealing

Irreversible water reconstructions (2,9) can occur during protein crystal annealing (10-12). Studies on a small number of proteins (2,13) have shown that the behaviour of solvent regions is dictated by their size and shape and is intimately linked to macroscopic changes in cell parameters, mosaicity and limiting resolution. Nevertheless, the exact relationship between water reconstruction and the various meta–stable substructures (e.g. cubic and hexagonal ice, high and low density amorphous ice, supercooled and glassy water etc.) known from the phase diagram of water (14) remains obscure. Proof–of–principle experiments (2,4) using diffraction data from before and after the transition point in concanavalin A (16) showed abrupt expansion of the b and c axes, but not a.

Con A cell volume	Con A a axis	Con A b axis	Con A c axis
Variation of concanavalin A cell parameters with temperature (9). The crystal was flash-cooled to 120 K, cooled further to 95 K, warmed slowly to 180 K and re-cooled. Note the abrupt change between 160 and 165 K.

This expansion is correlated with movement of water molecules within oriented channels that run along the a axis. Such changes in cell parameters are in marked contrast to the movement of individual water molecules, which typically shift by about 0.1Å in the bc plane and by as much as 1Å or more parallel to the a axis.

Con A view down c	Con A view down b	Con A view down a
Water channels along the a axis provide the link between crystal structure and changes in cell parameters. There is no apparent expansion along a because of movement along the channels. Water molecules close to the protein undergo small shifts while those on the edge of the channels can undergo large shifts. Within the channels, movement is even less restricted.

To gain insight into the physics behind water substructure reconstruction, carefully chosen protein crystals that diffract to atomic resolution and that possess solvent regions of varying size and geometry are targetted. Intense X–rays, high data quality, reproducible crystal handling techniques (8), careful phase bias removal (9,15) and the highest attainable resolution are essential factors governing the success of this work.

References