There are an estimated one million different protein molecules expressed by the human genome, representing a plethora of potential new drug targets. Crystallography provides a way to "look" at the arrangement of atoms inside of proteins, determining their structure and revealing chemical and biological functions.
Advances are being made to automate methods to obtain high-quality protein crystals, process and enhance x-ray diffraction data into useable structural information, and to use three-dimensional protein structures as a basis for drug discovery and lead optimization.
Automated Crystallization Lab
In order to rapidly provide sufficient amounts of high-quality crystals, scientists at Exelixis (www.exelixis.com) developed their own automated crystallization laboratory, Crystalbot. Using off-the-shelf hardware, the company customized equipment to meet the needs of their drug discovery group.
"We don't need to screen huge numbers of proteins, but every protein we do address is a high-priority target. Our goal is to solve what is bound to those protein structures to give us information on how to improve those molecules and build better drugs," says David Matthews, Ph.D., director of structural biology.
The system can screen from 500 to 800 different mixtures of protein precipitates in order to find one that will crystallize. "It's unlikely you'll get something right off the bat that's a really excellent quality diffraction crystal," explains Dr. Matthews.
However, he says, "There have been a few cases where we've been able to go from an initial, not very attractive crystal, to something of high quality in a matter of days, where a more traditional approach would have taken at least weeks, if not months."
Exelixis plans to increase capacity and streamline methods for solving compound protein-complexes in a high-throughput manner. But, the system is strictly an in-house tool. "Our main focus is drug-lead optimization," summarizes Dr. Matthews.
Whole Gene Synthesis
Protein modifications (deletions and mutations for example) can have profound effects on expression, solubility, and crystallization. These are often difficult to identify, requiring testing of various protein constructs, which can be difficult to make.
Researchers at DeCode Biostructures (www.decode.com) developed a whole gene synthesis technology that allows rapid generation of multiple gene variants for expression and crystallization trials. "Once you decide gene synthesis is a good idea, one of the problems is how to actually design the gene," explains Alex Burgin, Ph.D., director of molecular biology.
GeneBuilder, a database and software package, helps scientists decide what is the best gene to synthesize for a given protein sequence. It also allows the user to define protein constructs based on crystal structures of homologous proteins, phylogenetic analyses, and protein prediction algorithms.
"With whole gene synthesis, you're not tied to what nature provides, you can synthesize any amino acid sequence. This provides huge flexibility and enables us to make proteins that are difficult to make by conventional cloning methods," says Dr. Burgin.
The software scans the web for a given gene, finds homologous structures and sequences, analyzes the structure, and comes up with an amino acid sequence to synthesize. This can all be done in approximately two weeks; it used to take weeks just to find the clone.
Although originally designed for crystallographers, Dr. Burgin says that GeneBuilder is useful for anyone who wants to do protein design. "It's all about making a stable structurefiguring out what's the best sequence to make. If you can build a gene, you can build a protein."
Protein Database Provides 3-D Structures
StructureBank, developed by Cengent Therapeutics (www.cengent.com), is a scalable, relational database that stores protein sequences, 3-D coordinates, annotations, and other data.
"We took the sequence data from the human genome and converted it to 3-D structures, which can be adapted to in silico screening methods for developing compounds," explains Kal Ramnarayan, Ph.D., vp and CSO.
The database currently has protein models for 550 protein families (as defined by PROSITE) and more than 5,500 unique, proprietary structures. Average time from sequence to protein structure ranges from one week up to four weeks. This is much faster than crystallography, which can take up to six months, depending on the protein.
Instead of depending on the expertise of a computational chemist, says Dr. Ramnarayan, this database enables the user to look at sequence and structure data, and perform searches and comparisons across families of proteins.
Searches can be done using textual, sequential, or structural parameters. Proteins can be viewed as skeleton, ribbon, surface-view, or space-filled. Other customizable views include electrostatic potential, flexibility, amino acid type, strain energy, and others. Images can be magnified and rotated, and structures can be superposed or separated to highlight similarities and differences.
Protein information can be obtained from multiple sources including in-house research, public domain sources, and the company's proprietary database. In order to transfer data across disparate platforms in different fields, the company incorporated XML technology. This enables users to write just a few lines of code to read protein sequences or structure-based information for their own analysis.
"People now realize the total utility of protein structure information in drug discovery. Our platform will help people access protein structures readily if they have limited resources to access novel structure information by other experimental methodologies," says Dr. Ramnarayan.
Enhancing Protein Activity
Historically, formation of inclusion bodies during protein expression creates a bottleneck in the drug discovery process. However, ProteomTech (www.proteomtech-inc.com) developed a proprietary method to obtain pure, active proteins expressed in E.coli. Pt-Fold produces research quantities of recombinant human proteins by refolding the proteins.
Developed by Xinli Lin, Ph.D., CSO and company co-founder, the process starts with inclusion bodies or the clone for expression in E.coli. These are isolated to high purity then solubilized in a proprietary buffer solution, adjusted to a high pH (10), so the protein doesn't form aggregates and remains soluble.
The pH is shifted over time (24 to 72 hours) to a lower pH (7 or 8). This causes the protein to refold and form a secondary, active structure. When refolding is complete, the protein is further purified and tested for activity.
More than 250 recombinant proteins have been refolded to date, including beta-secretase (an Alzheimer's disease-related protein), pro-urokinase, and other types of proteins, such as cell receptors, proteinases, and signal transduction-related kinases.
"One of our leading drug candidates is a vascular endothelial growth inhibitor, an anti-angiogenesis cancer drug," says Dr. Lin. "Other companies tried to refold this protein to make it active, but we were the only one to successfully refold it from E.coli into an active form."
The company is also developing an automated refolding machine that will accommodate 96 different formulations, automatically adjust the pH, and control environmental conditions. Projected availability is late 2005.
Bringing the Synchrotron Home
Synchrotron radiation sources operating at x-ray wavelengths allow atomic protein data to be collected. By using a laser pulse to bend electrons and produce x-rays, Lyncean Technologies (www.lynceantech.com) has developed a tabletop synchrotron. The Compact Light Source (CLS) is a miniaturized version of conventional, large synchrotrons (minus the big magnetic rings) that provide comparable x-ray beams, but with some noted advantages.
"Our device is naturally monochromatic versus the large synchrotrons, which have a broad spectrum of radiation," explains Ronald Ruth, Ph.D., company president and chief scientist.
"Broad spectrum means high power on the x-ray optics, which have to be cooled. Since our energy spectrum is so narrow, the power is low and no cooling is needed, making the x-ray optics very inexpensive and simple."
Another main feature is that the electron beam is tunable; the x-ray energy can be varied from about 6 keV (kilo electron volts) to about 16 kEV. This is key in multi-wavelength anomalous dispersion (MAD) studies. The sudden increase in the attenuation coefficient of photons occurs at a photon energy just above the binding energy of the K shell electron of the atoms interacting with the photons.
Crystallographers use this property to determine the phase in a diffraction pattern, and it also permits the solution of the structure. The "tunability" of the system allows the adjustment of x-ray energy by small amounts near the K edge, so that different diffraction patterns can be taken.
A major reason for developing this system, Dr. Ruth explains, is the increasing demand for high throughput protein crystallography and tremendous interest by the crystallography community. "There is no question that was the driving force behind developing a practical device that could go into anyone's home laboratory." The company will begin testing it this spring.
Compact Protein Diffractometer
A high intensity x-ray source for protein crystallography is now available without rotating anode generators. The Xcalibur PX Ultra, a compact protein diffractometer, is Oxford-Diffraction's (www. oxford-diffraction.com) first protein crystallography product.
Although the company's focus has been on the small molecule market, Leigh Rees, Ph.D., product manager, says it was a natural progression to move into the protein market because both areas use the same core technologies. "We were just extending our existing product lines and fine tuning them for protein crystallography. You need a slightly different wavelength (copper) and a bigger detector for proteins."
The unit incorporates the company's high intensity x-ray source, based on a sealed x-ray tube. It produces 0.3-mm high intensity x-ray beams at copper wavelengths. Custom optics (Osmic Confocal Max-Flux optics) are used to increase the intensity of the x-ray tube.
"This actually throws out a huge amount of x-rays and, typically, less than one percent of that is used. The optics increase the efficiency of how much of these x-rays are harvestedthis is the whole principal behind it," says Dr. Rees. He says comparative studies have shown that this system can provide the same intensity as a 5-kilowatt rotating anode generator, but without the expense and required regular maintenance.
Other features include a large area 165-mm Onyx CCD detector, the kappa 4-circle Xcalibur platform, and custom CrysALis protein software. "For companies just starting out, this can be used to quickly kick-start their protein crystallography programs, and for well-established firms, it will enable them to now spend time on the science rather than maintaining the machine," summarizes Dr. Rees.
The companies covered in this article will provide additional information on their technologies at the upcoming Select Conferences "Protein Crystallography in Drug Discovery" meeting on January 1718 in San Francisco.