Tackling the bottleneck of protein crystallization in the post-genomic era.

Current genome projects are expected to reveal numerous new targets for therapy of human disease. However, it is usually not the genes themselves that are the targets but the proteins encoded by them. The function of these proteins is determined by their 3D structure and drugs can influence the function by precise atomic interactions with the protein. The most powerful technique for determining protein structure is X-ray crystallography, which requires highly ordered crystals.Protein crystallization has thus gained a strategic and commercial relevance in the post-genomic era in which X-ray crystallography will have a major role. There have been considerable advances in the automation of protein preparation [1xDesign of high-throughput methods of protein production for structural biology. Stevens, R.C. Structure. 2000; 8: R177–R185Abstract | Full Text | Full Text PDF | PubMed | Scopus (156)See all References[1], X-ray analysis [2xAutomation of X-ray crystallography. Abola, E. et al. Nat. Struct. Biol. 2000; 7: 973–977Crossref | PubMed | Scopus (145)See all References[2], modelling and bioinformatics [3xProtein structure modelling for structural genomics. Sanchez, R. et al. Nature Struct. Biol. 2000; 7: 986–990Crossref | PubMed | Scopus (155)See all References[3] but these advances have not yet been matched by improvements of the crystallization process. In the field of crystallization, the main effort and resources are being invested into the automation of screening procedures. Screening involves search procedures in which the molecule to be crystallized is exposed to numerous crystallization agents to identify conditions that might produce crystals. However, despite the ability to generate numerous trials and the high expectations, so far only a small percentage of the proteins produced have led to structure determinations.Pilot structural-genomics projects show that the success rate of getting from cloned protein to structure determination is ∼10%. For example, figures taken from a Human Proteome Structural Genomics pilot project (Brookhaven National Laboratory, The Rockefeller University and Albert Einstein College of Medicine, New York, USA: http://proteome.bnl.gov/progress.html) show that out of 124 proteins cloned, 62 were purified and 33 yielded crystals of some sort. Of these, only 16 yielded crystals suitable for structure determination (Fig. 1Fig. 1). Similar success rates by other pilot projects from around the world have been reported at relevant conferences and meetings. This highlights a general problem where, even when proteins can be cloned, expressed, solubilized and purified, and even if crystallization trials yield some crystals, this does not guarantee that the crystals will be good enough for structure determination. For structural genomics approaches to be productive, it is essential that this problem is addressed.Fig. 1The histogram shows the different stages involved from clone to X-ray structure determination and their relative success rates.View Large Image | Download PowerPoint SlideGeneration of high-throughput screening crystallization trials using robotics is well underway [4xHigh-throughput protein crystallization. Stevens, R.C. Curr. Opin. Struct. Biol. 2000; 10: 558–563Crossref | PubMed | Scopus (166)See all References, 5xMacromolecular crystallization in a high throughput laboratory – the search phase. Luft, J.R. et al. J. Cryst. Growth. 2001; 232: 591–595Crossref | Scopus (59)See all References, 6xDevelopment of a technology for automation and miniaturisation of protein crystallisation. Mueller, U. et al. J. Biotech. 2001; 85: 7–14Crossref | PubMed | Scopus (68)See all References]. Some proteins will crystallize during this initial screening but because screening is a hit-and-miss process most trials are likely to yield microcrystals or low-ordered crystals. The conversion of such crystals into useful ones requires intellectual input and individualized optimization techniques (for examples see [7xThe role of oil in macromolecular crystallisation. Chayen, N.E. Structure. 1997; 5: 1269–1274Abstract | Full Text | Full Text PDF | PubMedSee all References[7] and [8xImproving protein crystal quality by de-coupling nucleation and growth in vapour diffusion. Saridakis, E. and Chayen, N.E. Protein Sci. 2000; 9: 755–757Crossref | PubMedSee all References[8]). Such techniques do not lend themselves readily to automation and have yet to be adapted to cope with the huge volume of experiments required by genome projects. Consequently, the subject of optimization has been somewhat neglected, apart from the obvious first step of merely changing the concentrations, pH and/or temperature around the conditions that were found (by the screening trials) to produce small crystals or an indication of crystalline precipitate. Numerous recent articles and special issues of scientific journals have highlighted the importance of structural genomics in the post-genome era (for example, see [9xStructures by numbers. Abbott, A. Nature. 2000; 408: 130–132Crossref | PubMed | Scopus (19)See all References, 10xFocussed issue: Structural Genomics. Noble, D. et al. Prog. Biophys. Mol. Biol. 2000; 73: 289–362Crossref | PubMed | Scopus (38)See all References, 11xGlobal Efforts in Structural Genomics. Stevens, R.C. et al. Science. 2001; 294: 89–92Crossref | PubMed | Scopus (160)See all References, 12xCompleteness in structural genomics. Vitcup, D. et al. Nat. Struct. Biol. 2001; 8: 559–565Crossref | PubMed | Scopus (259)See all References]). However, surprisingly little attention has been given to improving methods of protein crystallization and optimization. The statistics shown in Fig. 1Fig. 1 and the enormous number of expressed proteins that will soon need to be dealt with, indicate that, like screening, optimization must be adapted to high-throughput. Without this we will soon run out of resources and be left with a backlog of useless microcrystals. The combination of automated screening and further development of automated crystal optimization methods will remove the main bottleneck in structural genomics and equip the genome project to deal with its awesome second-phase task.