Section 23.4.5. The classic models: small proteins with high-resolution crystal structures

Crambin is a plant-seed hydrophobic protein of unknown function. It contains 46 amino-acid residues and was reported to form crystals that diffract to 0.88 Å resolution (Teeter & Hendrickson, 1979). The crystal structure of crambin was determined to 0.945 Å resolution directly from anomalous scattering by the six sulfur atoms involved in three disulfide bonds (Hendrickson & Teeter, 1981). Crambin is an amphipathic molecule in that the hydrophilic components (including six charged groups) are segregated from a mainly hydrophobic surface.

A total of 64 water molecules and two ethanol molecules were located in the electron-density map, despite the fact that the structure was determined in 60% ethanol. The overwhelming number of water molecules compared to ethanol is consistent with the results of the multiple-solvent crystal structures experiments described above for elastase (Mattos & Ringe, 1996).

Most of the 64 water molecules found in crambin interact with polar side chains in the typical manner described previously. The unusual information about solvent structure offered by the crambin model is that the arrangement of water molecules around hydrophobic residues is similar to that observed for clathrate hydrate structures (Teeter, 1991). Pentagonal water rings are observed to cap the Cδ2 atom of Leu18 as well as the hydrophobic methylene groups of Arg17 (Teeter, 1984, 1991). The set of five connected water rings is shown in Fig. 23.4.5.1. This ring cluster extends toward the protein, forming heterocyclic rings that are described in detail in the original article (Teeter, 1984).

Figure 23.4.5.1| top | pdf | van der Waals surface diagram of the water pentagons A, C, D and E in crambin viewed in the negative a direction. Rings A, C and E form a cap around leucine 18. Hydrophobic atoms are shown as dark circles, and water oxygens are shown as light circles. The methyl group of leucine 18 can be seen through the C ring. Adjacent translationally related molecules are shaded. The van der Waals radii used for the protein C, N and O atoms are 1.7, 1.4 and 1.4 Å, respectively, and for water oxygen, 1.8 Å. The larger radius is used for the water oxygens because hydrogen atoms have been omitted. Reprinted with the permission of the author from Teeter (1984).

Although crambin provides the clearest example of pentagonal water rings on a hydrophobic protein surface, it is not the only one. Other high-resolution crystal structures (better than 1.4 Å), such as insulin and cytochrome c, have also revealed pentagonal rings, but never to the extent seen in crambin (Teeter, 1984). This is very likely to be a general mode of interaction between water and hydrophobic moieties, be it in inorganic, organic, or biological molecules. The fact that it is not observed in protein structures in general may be related to the lower resolution of most X-ray structures, where it is not possible to model the more disordered areas where these patterns are likely to be found.

Bovine pancreatic trypsin inhibitor (BPTI) is a protein of 58 amino-acid residues whose X-ray crystal structure was obtained in the original crystal form to 1.5 Å resolution (Deisenhofer & Steigemann, 1975). Subsequently, 1 Å X-ray data were obtained from a different crystal form, and the new model was jointly refined with 1.8 Å neutron diffraction data (Wlodawer et al., 1984). Minor differences in structure between the two crystal forms of BPTI were observed (Wlodawer et al., 1984). The interesting contribution of the 1 Å model to the understanding of solvent structure resulted from the ability to refine occupancy at this resolution. A total of 63 water molecules were placed in the model, 20 of them within 1 Å of a water molecule found in the structure solved in the original crystal form. During refinement against the 1 Å data set, full occupancy was assigned to all protein atoms, and water occupancy was allowed to refine. Of the 63 water-molecule positions, 29 were found to be fully occupied. The remaining 34 had partial occupancies, with 0.4 being the minimum occupancy found. Given that there are very few contacts between protein molecules in the crystal (Wlodawer et al., 1984), it is reasonable to assume that this observation is representative of water occupancies on protein surfaces in general. It is likely that well over half of the water positions found on protein surfaces are less than fully occupied, although there is no definitive proof that this is true.

In general, small proteins serve as important models where results of X-ray crystallography, NMR and molecular-dynamics calculations can be easily compared and cross-validated, since larger proteins are more difficult to study by the latter two techniques. Small proteins are also more likely to form relatively ordered crystals, which are able to diffract X-rays to atomic resolution (of the order of 1 Å). With respect to understanding solvent structure, the two major contributions of these very high resolution protein models are the observation of solvent structure around hydrophobic residues, where at lower resolution the water molecules `look' disordered, and a glimpse at the pattern of water occupancy likely to occur on protein surfaces.