Protein–drug interaction

Growing single crystals of proteins is an arduous process, requiring careful selection of conditions to achieve success. A protein structure study would be completely halted by the inability to produce a single crystal. Furthermore, single-crystal growth of a protein–ligand complex often presents completely new challenges compared with growing crystals of the protein alone. NMR, meanwhile, is limited by spectrometer resolution to proteins on the order of 25 kDa. The difficulties of growing and analyzing single crystals of proteins are best seen through a cursory examination of the Protein Data Base, which shows that only about 10% of the entries involve protein– ligand complexes.

Ways and means
An alternative to these techniques is suggested by the fact that the most readily available form for solid materials is a polycrystalline powder consisting of a complex mélange of small crystals. These can be formed over a wide range of conditions and timescales, unlike the restricted circumstances required for producing large single crystals. In many cases, polycrystalline powders can be readily made, but large, single crystals prove impossible to grow. Furthermore, powder-diffraction patterns can display considerable sensitivity to subtle structural changes, as typified by shifts in diffraction peak positions and changes in intensity. This has long been recognized by materials scientists, and over the past 30 years considerable progress has been made in extracting structural information from powders by Rietveld refinement (see box below) (1, 2).

Rietveld refinement

Because a polycrystalline powder consists of many small crystallites, a powder-diffraction experiment superimposes the diffractions from all the crystals in the sample, smearing all the single-crystal diffraction spots into rings. For a material of even moderate complexity, many of these rings partially or completely superimpose, and so the resulting profile becomes a curve of considerable complexity. In Rietveld refinement, a model is constructed to represent the entire powder-diffraction profile. This model includes a complete description of the crystal structure (e.g., atom positions, thermal motions, and scattering factors), the shapes and positions of the individual contributing peaks that make up the profile, and a contribution from any smooth background scattering. The refinement is a least-squares process that minimizes the differences between the calculated and measured experimental profiles by adjusting the suite of parameters needed to describe the model. In essence, this method is a complex curve-fitting exercise that may involve hundreds of parameters. Although the full three-dimensional character of a single-crystal diffraction pattern is lost, the technique of Rietveld refinement allows one to extract the maximum available information contained in the powder-diffraction profile.

For example, virtually all our structural knowledge about high-temperature superconductors comes from X-ray and neutron powder-diffraction experiments. These materials readily form powders but are not amenable to growth of large single crystals. Many of them are also subject to changes in structural phase that would render single crystals useless for diffraction experiments. Powder-diffraction experiments have elucidated the nature of these phase changes and the subtle structural changes that accompany changes in their superconducting properties with, for example, composition.

Until recently, protein crystal structures were considered far too complex for powder-diffraction experiments to give any useful information. However, our recent work at the Brookhaven National Laboratory’s National Synchrotron Light Source has shown that proteins give extremely sharp X-ray powder-diffraction patterns that can be analyzed by a combined Rietveld and stereochemical restraint refinement to give structures of moderate resolution (~3 Å) and, in one case, has led to the first solution of a protein structure from powder-diffraction data (3, 4). Protein lattice parameters determined from this powder data are perhaps 2 orders of magnitude more precise than those obtained from typical single-crystal experiments.

For the study of protein–ligand complexes, powder diffraction offers a distinct advantage over single-crystal work in its complete immunity to crystal fracture and to any phase change that may accompany complex formation. The extreme sensitivity of diffraction patterns to changes in lattice parameters makes powder diffraction sensitive to complex formation. Furthermore, rapid formation of a polycrystalline precipitate allows possible exploration of initial complex formation under a wide variety of conditions not accessible in slow soaking or single-crystal growth experiments.

Figure 1. Before and after. A small segment of high-resolution, X-ray powder-diffraction patterns of lysozyme (red) and N-acetylglucosamine–lysozyme complex (blue). The two patterns have been offset for clarity.

Figure 2. Binding in profile. High-resolution X-ray powder-diffraction profile from the final Rietveld refinement of the N-acetylglucosamine–lysozyme complex. Observed intensities are shown as red (+), calculated and difference curves as green and purple lines, and reflection positions as black (|). The background intensity found in the refinement has been subtracted from the observed and calculated intensities for clarity.

For example, the change in the powder-diffraction pattern of chicken egg lysozyme upon binding with the inhibitor N-acetylglucosamine is easily seen by changes in the diffraction peak positions and intensities (see Figure 1 at right), even though the changes in the lattice parameters are less than 0.5%. Moreover, these data can be used to locate the position of the ligand, and the protein– ligand structure can then be subjected to a combined Rietveld and stereochemical restraint refinement (Figure 2 at right) to determine the details of complex formation.

A bright future
High-resolution powder diffraction of proteins is still in its infancy, and the current molecular weight limit of perhaps 50 kDa is largely due to the density of reflection overlaps in the diffraction pattern. These limits are stricter than those of single-crystal diffraction, and there is no present way of solving protein structures ab initio from powder data; but model building and molecular replacement work quite well. Nonetheless, we can easily see future developments of the method that will allow examination of protein structures that exceed 100 kDa.

In particular, current data-collection technology scans the powder-diffraction pattern a few points at a time over a narrow field of view. Consequently, data collection times at a synchrotron source are on the order of half a day. The use of high-resolution imaging technology and X-ray focusing optics should improve this 1000-fold or more, making it possible to use powder diffraction on a laboratory X-ray source to screen for the formation of protein–drug complexes and to determine their structures.

References

1. Rietveld, H. M. J. Appl. Crystallogr. 1969, 2, 65–71.
2. The Rietveld Method; Young, R. A., Ed.; Oxford University Press: New York, 1993.
3. Von Dreele, R. B. J. Appl. Crystallogr. 1999, 32, 1084–1089.
4. Von Dreele, R. B., et al. Acta Crystallogr. 2000, D56, 1549–1553.