In 1958, Max Perutz and John Cowdery Kendrew determined the first protein crystal structure, for sperm whale myoglobin, using X-ray crystallography. The technique, which also played a key role in James Watson and Francis Crick's work on the double-helix structure of DNA, has also been integral to the drug discovery and development process. Because X-ray crystallography can be used to determine molecular structure, it plays a key role in rational drug design.
But growing crystals can be a difficult, tedious process, particularly for proteins and DNA complexes. Over the last several years, vendors have responded by introducing tools to automate the process and make it more high throughput, and software has been developed to manage data and analyze proteins. Automation has continued to evolve, miniaturization has begun to play a role, and a new technology has the potential to bring a synchrotron-grade light source into the lab.
''The state of the art for protein crystallography has advanced to the point that you still have to spend a substantial amount of time trying to get a crystal of a protein that is appropriate,'' says John McAllister III, PhD, CEO, Tripos Inc., St. Louis, a provider of drug discovery chemistry and informatics products. ''But people now are using robots to set up the crystallization experiments, and they're doing a very systematic exploration of crystallization parameters'' by varying solvents and protein and salt concentrations.
High-powered beams from neutron reactors have also played a role, McAllister says. ''This allows you to collect data very rapidly and collect whole spheres of data simultaneously. That speeds up the data collection enormously, which means that you get a full data set from one crystal instead of multiple crystals, something which had a lot of scaling problems associated with it.'' McAllister says this enables much more refined analysis of how potentially therapeutic compounds bind into active sites of proteins, allowing for a much more designed approach to finding alternative ligands or new drugs. ''I think crystallography will play an increasing role in that area. I believe that's the final step in a process and not necessarily a routine discovery process, but it's a refinement process that is applied in the drug discovery activity, and it will be increasingly important in the pharmaceutical discovery activity.''
Michael Hennig, PhD, vice director of molecular structure and design at Hoffmann-La Roche AG, Basel, Switzerland, believes more refined crystallographic techniques have already had an effect in their role as a prerequisite for structure-based drug design. X-ray structures can now be determined in a timeframe that fits the drug discovery process, he says. ''Ten years ago, the problem was that the X-ray crystallography was quite slow and usually structural information was available very late in the project. Today, it's possible to have structural information much earlier in the lead generation process. . . Of course, structural information is just one aspect. Then you have to translate this information into chemistry, and for this I think computational chemistry is key so people can translate the structural information into molecules that the chemist can synthesize.''
Although new tools and techniques will help advance the field, Hennig says challenges remain. While more molecular targets are accessible for X-ray analysis, the structure determination of membrane proteins, especially G-protein-coupled receptors (GPCRs), are still a challenge. ''GPCRs are still not feasible, especially in industry.''
No Crystals Needed
In an effort to make crystallographic techniques easier to use, researchers at Argonne National Laboratory (ANL), Argonne, Ill., have developed a method that provides the same information as X-ray crystallography without the laborious step of growing a high-quality crystal. In a study published recently in Chemistry and Biology, researchers reported on the use of wide angle X-ray scattering (WAXS), a diffraction technique that has been used to determine the crystalline structure of polymers. They adapted it to study ligand-induced structural changes in proteins.
|Initiative Elucidates Protein Structures
Crystallization has been one of the limiting steps in determining 3D macromolecular structures, a fact that has sparked efforts to develop methods for high-throughput crystallization. These methods have played a key role in the Protein Structure Initiative (PSI), a 10-year $600 million project funded largely by the National Institute of General Medical Sciences at the National Institutes of Health. The program seeks to determine the 3D shape of proteins in order to elucidate how they function in many life processes. Although the program was only launched in 2000, more than 1,000 different structures have already been determined, many with the potential to lead to targets for the development of new drugs.
''The goal is to try to get a crystal structure of every type of protein domain that's out there. For every protein fold that exists, we'd like a crystal structure,'' says Lee Makowski, PhD, director of the biosciences division at the Argonne National Laboratory, Argonne, Ill. ''The whole idea of the protein structure initiative is to make it easier and easier, even with proteins that we don't know how to crystallize, to predict their 3D structure. From there, it would be great to be able to predict what small-molecule ligands they will interact with.''
Raymond Stevens, PhD, professor of molecular biology at The Scripps Research Institute, La Jolla, Calif., heads the crystallomics core at the Joint Center for Structural Genomics (JCSG). The JCSG is a collaborative effort involving more than 60 researchers from Scripps and a number of other public and private institutions. It is one of nine pilot centers participating in the first phase of the PSI. Stevens' group will use a high-throughput robotic production line that can produce thousands of samples of purified proteins per year and perform more than 100,000 crystallization screens a day.
Makowski adds that another advantage of the technique is that it allows scientists to analyze molecules that X-ray crystallography can't handle because many small molecules induce structural changes in proteins that are too big to be accommodated within a crystal lattice. Crystal lattices are disrupted by ligands that cause the protein to change structure. ''The crystal shatters, it falls apart when you add ligand in many cases, so this is a simple way of trying to detect structural changes that you just can't study easily in a crystal.''
Makowski says the research team chose four proteins whose structure is both known in the presence and absence of a particular ligand. They collected solution scattering patterns from the proteins, plus and minus the ligand, and compared the scattering changes that were detected by the addition of the ligand. They then compared those changes to what would have been predicted on the basis of the crystal structure. Makowski says that while there are many easy screening methods to determine if a small molecule binds to a protein, it is much more difficult to perform a high-throughput screen to tell whether a small molecule alters the function of a protein. ''The nice thing is that this is a moderate throughput system for telling whether or not an interaction is changing the structure, and since structural change is almost always correlated with functional change, that gets you one step closer to knowing if you've got a functional molecule.''
Makowski says WAXS could have a significant impact on drug development. ''The bar is much lower . . . While we can't tell you the details of the structural change, we can tell you if one has been induced by the ligand.'' Because it is sensitive enough to discern the difference between a ligand that is just sticking to a protein's surface and a ligand that is actually changing the protein's structure, it could identify drugs that will bind to target proteins. It could also be used to determine how effective drugs are at binding to and modifying targeted proteins. Data collection using WAXS takes only a few minutes, as opposed to functional cell-based assays that can take weeks or months.
Makowski says future research will focus on, among other things, automating the process. For example, a researcher could take a very large library of molecules, screen them for binding to a particular protein, and find 50 hits. But many of those hits are going to bind without altering the function of the protein. ''Now, I think we can go through and collect 50 data sets and tell you which of those molecules that bind to the protein actually alter the structure of the protein. Those are the ones that you'll have to suspect first as leading to changes in the function of the protein. The other thing, of course, is that with a protein that has a very flexible active site, this is a way of not having to worry about the fact that you have to get a crystal. It is very hard to get a crystal of a protein that has a large distorted region. It's done, but it's difficult.''
In addition to work such as Makowski's, which finds new applications for existing technologies, other scientists are developing new technologies. Researchers at The Scripps Research Institute (TSRI), La Jolla, Calif., have focused on miniaturization, automation and integration over the last several years, says Raymond Stevens, PhD, a professor of molecular biology at TSRI. He says the biggest contribution of his research group has been the miniaturization of the crystallization process to the point where they only need microgram amounts of protein as opposed to milligrams of protein, and where they can work as low as 20 nL as opposed to 2 mL. Researchers at the institute have also miniaturized nuclear magnetic resonance spectroscopy, Stevens says. ''You can miniaturize pretty much all of the biophysical techniques. There just hasn't been a pressure to do it in 20 years. Because of all the developments in microfluidics, we can copy what they've done with microfluidics and high-throughput screening and use those same lessons and apply them to structural biology.''
Five years ago, Stevens helped found Syrrx Inc., San Diego. Syrrx uses high-throughput X-ray crystallography to determine the 3D structures of drug targets to aid in drug design. It recently merged with Takeda Pharmaceutical Company Ltd., Osaka, Japan, in an effort to beef up Takeda's research and development pipeline. At about the same time that Syrrx was being founded, Stevens was also involved with the launch of the Joint Center for Structural Genomics, a high-throughput structural biology effort funded by the National Institutes of Health as part of its Protein Structure Initiative, a national program to determine the 3D shapes of a wide range of proteins. ''The reason for starting up these two entities was we knew that we needed to develop better tools for high-throughput structural biology. We just felt like we could use it in two different ways, either private-sector drug discovery or in the public sector where there would be more structural information for the biologist.''
But despite the efforts of private and public entities, much work remains. ''One of my biggest concerns is that right now the success rate for these high-throughput structural biology efforts is actually pretty low,'' Stevens says. For structural genomics programs, the success rates range from 4% to 10%; for companies such as Syrrx that focus on one target, the success rates can be greater than 70%. Attaining those rates, however, is expensive. ''My worry is that in this next phase of high-throughput structural biology, we don't have the solution to fix these problems. We still need a fundamental breakthrough in structural biology, protein crystallography in particular, that will increase our success rate while lowering the cost.''
A Lab Synchrotron
Some of the areas which could produce breakthroughs include crystal screening and data collection, Stevens says. Researchers in his group performed an analysis of where they spend their time during the crystallography process and found that while growing crystals was time consuming, screening them also took a long time. ''Finding the right ones that are good for data collection [and] collecting the data, that feedback loop is very slow.'' Stevens says a new tool that is the marriage of two mature technologies will help eliminate that bottleneck. The Compact Light Source (CLS) from Lyncean Technologies Inc., Palo Alto, Calif., is a miniature synchrotron that uses particle accelerator technology and solid-state laser technology. The CLS was miniaturized by reducing the electron beam energy and replacing conventional undulator magnets with a laser. Despite its size, CLS has an average flux comparable to beamlines at large synchrotrons and allows for multiwavelength anomalous dispersion data collection.
''It's sort of like bringing the desktop computer into the lab instead of the mainframe. If you start bringing [CLS] into the laboratories, people are going to get creative and they'll start developing more technologies around it,'' Stevens says. He adds that Lyncean, which has been funded by more than $7 million in grants from the Protein Structure Initiative, expects to have the first diffraction from protein crystals in May using their first prototype.
Stevens says CLS could be combined with microfluidic crystallization tools being developed at Scripps. These tools would allow crystallization trials to be performed on a glass capillary, which could then be run through the X-ray beam to read it. Stevens is also interested in exploring the possibility of getting rid of imaging and integrating crystallization robots with the CLS. ''That will allow us to screen crystals as they're growing to know which ones are going to be the ones to collect data on.''