Accelerating X-ray Crystallography

Courtesy of Astex Technology For years, the process of X-ray crystallography has moved at a tortoise's pace. "When I started in the field, it would typically take 20 person-years to produce a complete atomic model of one single protein. It was like a traffic jam in New York City. Every single part of the process was slow," says Stephen Burley, chief scientific officer and senior vice president of research at San Diego-based Structural GenomiX. The biggest holdup: obtaining a suitable crystal. Researchers can spend weeks or even months searching for optimal crystallization conditions. "You need a crystal and without a crystal you have nothing to do," explains Tom Hurley, a professor of biochemistry and molecular biology at Indiana University. Other bottlenecks include data collection and data analysis. Now the biotech industry and some academics are trying to kick the process up to hare speed. New high-throughput (HT) methods being developed will dramatically improve the pace at which researchers collect and analyze data on larger and more poorly diffracting proteins, macromolecular complexes, and nucleic acids. This improvement will transform the labor-intensive and time-consuming process of molecular structure determination by X-ray crystallography into a powerful and viable method for advancing functional genomics, and ultimately, for translating raw genomic information into new pharmaceuticals. "The whole notion is to do more, faster, using automation as much as possible," explains Mark Rould, director of the University of Vermont's Center for X-ray Crystallography and assistant professor of molecular physiology and biophysics at the school's college of medicine. "Crystallography has gone from solving a structure once every decade to solving 75% of the projects in the lab within a week of collecting the data," explains Hurley. SCHOLARLY DEBATE Not everyone agrees about the feasibility and benefits of HT X-ray crystallography; some consider the goal of Structural GenomiX--5,000 proteins in five years--overly ambitious. Crystallographer Gongyi Zhang, National Jewish Medical and Research Center and University of Colorado Health Sciences Center, Denver, says that crystallization was, is, and will continue to be a rate-limiting step. Eventually, he says, biotech companies will fail to keep the pace they have been bragging about. Hurley, who believes HT techniques "will lead to a huge explosion of understanding in biochemistry," says that considerable development is still needed before leaping from structural determination to drug design. Nevertheless, a great deal of money is being spent to prove otherwise. Companies such as Cambridge, UK-based Astex Technology are experimenting with additives to enhance crystal formation. Others are deploying high-tech robots with the capability to perform standard molecular biology manipulations and reactions as well as protein-expression experiments. Add in automated storage-retrieval devices and robotics that analyze millions of crystallization trials, and the rate-limiting steps of crystallography become "hands-off" processes. This surge in automation has addressed the worrisome dearth of protein-structure researchers: Projects that once required a full-time expert can now be performed by a novice--with an expert's guidance, of course. Rould's lab, for example, uses HT X-ray crystallography to determine empirically what kinds of molecules bind to a given protein target. His group's commercial crystallization robot, the Cyberlab C200 from Gilson of Middleton, Wis., is helping them in this endeavor. It "eliminates much of the repetitive work," he says. "Without the robot, setting up the massive number of crystallization trials can cripple the hands of a researcher or technician." Robotic assistants can come with a hefty price tag, however, putting them out of reach of all but the most well-financed investigators. The beneficiaries of such technology are more likely to be the big competitors in the private sector, such as Structural GenomiX; its HT approach to sample preparation relies heavily on robotic enhancements and illustrates the potential power of robotics. "We have an automated liquid-handling system for mixing crystallization solutions; automated setup of 96-well-format sitting-drop crystallization trials with bar coding; automated storage, retrieval, and digital imaging of the trials; and finally, automated scoring of crystallization trials based on those digital images," says Burley. Courtesy of Structural GenomiX HIGH BEAMS: Structural GenomiX's dedicated beam line facility at the Advanced Photon Source (APS) at the Argonne National Laboratory in Illinois. The APS is the most advanced source of synchrotron radiation in the United States and is widely regarded as the premier location in the country for the collection of X-ray crystallographic data. BETTER BEAMS Structural GenomiX is also one of the lucky few in this burgeoning field to have unrestricted access to a state-of-the-art facility for X-ray data collection. Beam time at advanced light sources is a must for HT macromolecular crystallography, but limited facilities exist. Construction of the company's dedicated synchrotron beam line at the Advanced Photon Source at the Argonne National Laboratory--currently the most powerful and precise source of radiation in the United States--was completed in 2001. The extremely intense, focused X-ray radiation generated by a synchrotron permits the use of smaller crystals, which are more easily attainable. Synchrotrons also allow researchers to tune or select the wavelengths at which data is collected. Multiwavelength anomalous dispersion techniques help identify the optimal wavelengths for obtaining phase information, which is key to reconstructing an image. Such information is impossible to collect with conventional X-ray crystallography methods and is simply inferred. Academic researchers also need beam time; as with their corporate counterparts, they are beginning to realize that sometimes the best solution is to buy a resource instead of sharing it. Indiana University is part of an academic consortium that is currently building a dedicated X-ray beam line at Lawrence Berkeley National Laboratory's Advanced Light Source. "Universities are now getting into the idea that their macromolecular researchers need beam time," says Hurley, who uses HT techniques to examine the cellular process of glycogen synthesis. SOPHISTICATED SOFTWARE The success of HT X-ray crystallography relies greatly on sophisticated software that helps massage the mammoth datasets obtained during synchrotron experiments into three-dimensional structural models. The process is fundamentally the same as when the brain reconstructs an image from light entering the eyes. Just as the brain computes what the real-world images must look like based on the multitude of light waves colliding with the retina, the computer attempts to calculate backwards from the X-ray scatter pattern to determine where in the crystal the reflecting atoms reside. At the University at Buffalo, scientists working on protein-structure determination were recently awarded a $2 million (US) grant to improve SnB, a software package used worldwide that enables researchers to solve protein structures with virtually no user input. SnB can solve critical substructures that allow for the determination of proteins with tens of thousands of atoms, a task that would have been considered impossible just a few years ago. The university project will explore new technologies, such as grid computing, data mining, and collaborative environments, to refine SnB and ultimately advance structure determination. Astex Technology, a structure-based drug-discovery company, is using HT X-ray crystallography for the rapid identification of novel drug candidates. The company has developed its own software, AutoSolve®, a Web browser-based program that automates and accelerates data analysis and interpretation. According to Harren Jhoti, Astex's founder and chief scientific officer, "The time taken to produce protein-ligand docked structures is reduced from days to minutes." Astex's HT technologies have enabled the company to solve the 3-D crystalline structure of human cytochrome P450 3A4, a key drug-metabolizing enzyme. It is estimated that as many as half of all known drugs interact with this form of cytochrome P450. Since cytochrome P450s were first linked to drug metabolism in the 1970s, numerous drugs have been pulled from the market or clinical trials due to adverse effects from their interactions with these enzymes. Courtesy of Structural GenomiX PROTEIN STRUCTURE: A bacterial protein with an entirely new three-dimensional fold. The protein supports the uptake of essential nutrients and can be used as a target to discover new antimicrobial substances. SUPER STUFF Improvements in X-ray crystallography will aid researchers in their colossal mission to map out the shape of the entire proteome. As that long-term goal is currently untenable, scientists have come up with a more reasonable short-term goal of determining the structure of at least one representative from each protein superfamily, or group of proteins with similar shapes. "The number of fundamental protein shapes that exist is on the order of 2,000 to 5,000," explains Burley. Experts say determining the structure of a few examples in each family would arm biologists with the information needed to model virtually any other protein. Over the next decade, the National Institutes of Health's Protein Structure Initiative will use HT technologies to determine the structure of between 10,000 and 20,000 proteins and to pinpoint at least one structure for each protein superfamily. According to the NIH, recent and anticipated technology developments in protein structural determinations make this formidable task feasible. The drug industry is also spending big bucks on structural genomics initiatives. The challenge lies in converting all this information into new drugs. "Pharmaceuticals act on proteins at the molecular level," observes Rould. "If we can control the interaction between proteins and the molecules they bind and act on, we ought to be able to solve most diseases." There are plenty of data: Since proteomics was thrust into the research spotlight less than a decade ago, more than 13,000 protein structures have been deposited into the federally funded Protein Data Bank (www.rcsb.org/pdb/). Today's drugs act only on a mere 500 known proteins, providing plenty of opportunity for new research. Yet nearly half of all drug-discovery targets are membrane proteins, which are almost impossible to crystallize and, therefore, not yet amenable to HT technologies. "The medications that do exist were invented blind," says Burley. "We can work without knowing the structure, but we think it will work even better if we can solve the structures." Burley's team is currently working on solving the structure of the cystic fibrosis transmembrane regulator (CFTR), the membrane protein responsible for the thick lung mucus indicative of cystic fibrosis. During a five-year collaboration with the Cystic Fibrosis Foundation, Structural GenomiX will use CFTR as its test bed for developing methods to determine the structure of membrane proteins. Likewise, in December 2002, Structural GenomiX and the Hereditary Foundation announced a large-scale effort to investigate the structure of the protein responsible for Huntington's disease. Burley, a former principal investgator for the New York Structural Genomics Research Consortium, who left a prestigious position at Rockefeller University for HT proteomics work in the private sector, has great expectations. "I am convinced that all the technology is in place to change the way in which a protein structure guides drug discovery," he says. And, he predicts, "We are going to be able to make drugs with higher selectivity and higher potency and a lot less side effects." Kelli Miller (science@mindspring.com) is a freelance science and medical writer in Atlanta. THE PROTEIN PROFIT SEEKERS Accelrys www.accelrys.com Formed in 2001, Accelrys represents the collaboration of Molecular Simulations, Synopsys Scientific Systems, Oxford Molecular, and the Genetics Computer Group. The company develops and commercializes molecular modeling software. Astex Technology www.astex-technology.com Founded in 1999, the UK-based drug discovery company uses high-throughput (HT) X-ray crystallography technology for the rapid identification of novel drug candidates. It recently signed a collaboration agreement with Mitsubishi Pharma. GeneFormatics www.geneformatics.com Founded in 2000, the company uses sophisticated algorithm-predictive software to determine the structures of proteins based on the structures of other proteins. It recently joined hands with De Novo Pharmaceuticals to elucidate inhibitors of matrix metalloproteinases, specifically the M10 family. IBM http://researchweb.watson.ibm.com/bluegene/ IBM's Blue Gene supercomputer will aid in the understanding of protein shapes and superfamilies. The Protein Structure Initiative www.nigms.nih.gov Launched in 1999 by the National Institutes of Health, the initiative is designed to organize a cooperative, large-scale effort in the emergent field of structural genomics. It supports researchers who study structural genomics, including protein-family classification, target selection, sample preparation, structure determination by X-ray crystallography and nuclear magnetic resonance spectroscopy, and data analysis. Structural Bioinformatics www.strubix.com A computational proteomics company founded in 1996, its ProMaxä drug-target database contains thousands of proprietary drug target and "antitarget" structures in more than 470 structural families. The company has partnerships with IBM and Quest Diagnostics. Structural GenomiX www.stromix.com Founded in 1998, this San Diego-based drug discovery company utilizes a genomics-driven, HT structure-based platform to increase the efficiency of the drug discovery process. The company has partnerships with Millennium Pharmaceuticals, Aventis, and the Cystic Fibrosis Foundation. Syrrx www.syrrx.com Founded in 1999 as a spin-off from the Novartis Research Foundation, the San Diego-based company intends to leverage HT structure determination and HT screening with computational methods to discover novel small-molecule drug candidates.

Accelerating X-ray Crystallography

Interested in reading more?

Become a Member of