The idea for the initiative, NIGMS officials explain, is both to advance the technology for high-throughput protein structure determination and to reach the goal of determining thousands of protein structures with industrial speed and efficiency. Nine centers have been established for this work, the latest two centers announced earlier this month. A list of the centers and their principal investigators is available at http://www.nigms.nih.gov/
funding/psi/psi_research_
centers.html.

The multidisciplinary PSI generally draws high praise from the life sciences research community. Each center represents a consortium of universities, private research institutions, government agencies--most of the Department of Energy multiprogram national laboratories are consortium members--and industry partners. Together, they are providing the personnel, facilities, and equipment that will make each center a veritable factory for protein structure determination.

RESEARCHERS PARTICIPATING in PSI--including many chemists--will first organize all known proteins into structural families. Then, using X-ray crystallography, nuclear magnetic resonance spectroscopy, and computer modeling, they will determine the structure of one or more proteins from each family for a total of 10,000 to 15,000 protein structures to be determined. By determining these structures, PSI will more than triple the number of unique structures available to researchers and provide more thorough coverage of structural families.

	Norvell COURTESY OF NIGMS
	Lattman JOHNS HOPKINS UNIVERSITY
	Terwilliger LOS ALAMOS NATIONAL LABORATORY
	Steitz PHOTO BY HAROLD SHAPIRO
	Berman COURTESY OF RUTGERS UNIVERSITY

The initiative is viewed as a natural successor to NIH's Human Genome Project (HGP), in part because many of the protein structures to be determined will be identified from HGP sequence data. And like HGP, much of the work will be automated. That is, researchers at all of the centers are working to develop robotics and other technical innovations that will speed structure determination.

The initiative "has many of the trappings of big science more commonly seen in physics, not biology," says Eaton E. Lattman, chairman of the department of biophysics at Johns Hopkins University.

He points out that, along with praise, concerns about PSI are being heard in the scientific community--equitable funding of work in structural biology, the scientific value of the information scientists can obtain from PSI research, publishing and intellectual property issues, and so-called pipeline issues that deal with the education and training of scientists in this new era.

Lattman and others point out that PSI is a very different way of doing science, especially biology. It is a process that begins with less emphasis on theory and hypothesis and that is focused more on the technical ability to gather huge data sets of a particular type, in this case, of course, structure data.

But the initiative is moving ahead while scientists grapple with the issues. Whether it is automated protein purification and crystallization, desktop NMR spectroscopy, or the rapid exposure of many different protein crystals to synchrotron light sources in a single session, there is a drive shared by many PSI researchers to keep finding and keep pushing the technical boundaries.

"We're working now at a couple of orders of magnitude faster than we used to," says Duncan E. McRee, director of structure determination at Syrrx, a San Diego biotechnology company that uses structure determination for rational drug discovery. The company is also a partner in the PSI Joint Center for Structural Genomics (JCSG) at Scripps Research Institute.

With robotics and other automated processes, "We can do several thousand trials on crystals of a protein in one week," McRee says. "People are talking about solving crystal structures in terms of hours, not days. Crystallography will become a tool. This is the next wave."

Current X-ray crystallography of very large, complex protein structures is a process that can take several months to years and can cost in excess of $200,000 per structure. PSI investigators typically focus on smaller molecules. They say they would like to bring down the price tag of structure determination to about $20,000.

With knowledge of structure, say PSI participants, researchers can enlarge scientific understanding of protein structure as well as protein folding and function. In turn, that information can be used to pursue rational drug design or other breakthroughs in medical therapies. As was true for efforts to complete the human genome sequence, they expect the mass of protein structure data to yield scientific insights that cannot be predicted in advance.

"We [established PSI]" says PSI Director John C. Norvell, "because there was a sense from the scientific community that it should be done, and we held three workshops and several advisory meetings before we proceeded. We included scientists with wide-ranging backgrounds and opinions." However, he adds, "some people are not as convinced as others."

Indeed, PSI has its share of critics. Scientists who work in more traditional areas of structural biology, for example, see PSI as a possible threat to their funding, especially if future NIH budgets are trimmed or if the quantity of work completed by the PSI centers somehow comes to be valued more than long-term structure determination studies. They charge that decisions to proceed with PSI were not peer reviewed by scientific panels but were decided upon by NIH administrators.

"THE THRUST of PSI is to solve structures that are easy to solve," says Peter B. Moore, a professor of chemistry at Yale University. Structures that are not easily determined, he says, will be left to researchers working in more traditional areas of structural biology--if they have adequate funding.

"There's only one pot [of federal money] for funding science," Moore says. He fears that blind enthusiasm for PSI might outweigh more scientific considerations that would favor a traditional structural biology approach.

"What's driving PSI?" Moore asks. "People who think it's a good thing to do."

For NIH, Moore continues, "it's top-down science [in which] administrators are deciding what's to be done rather than scientists."

"Times change," replies NIGMS Director Marvin Cassman. "One of the things that's changing is the way science is done."

Cassman views some of the criticism as a "reflex response" to anything other than investigator-initiated research. He says that rather than a top-down decision, NIGMS scientific advisory groups held at least a year's worth of meetings in order to hammer out the PSI strategy.

"The level of enthusiasm has increased in the general scientific community for certain structure studies and for the high-throughput approach," Cassman explains. "We're a basic research institution. One goal is to get a compendium of protein families in nature.

"Everybody always gets nervous about the funding," Cassman adds. "Let them prove it to me. Our success rate [for NIGMS grant proposals] is currently in the mid 30% range. Nobody is going to tell me that we are giving short shrift to [individual] investigators."

Cassman points out that approximately one-third of all known human proteins are membrane proteins, which are very difficult to crystallize. The structures are too large and too complex to be determined with the high-throughput approach. "There's a lot of room [in the NIGMS budget] for doing other things," he says.

"The bottom line is, there has been an influx of funding invested by a rigorous peer review process," says Keith O. Hodgson, a professor of chemistry at Stanford University and director of the Stanford Synchrotron Radiation Laboratory, a DOE user facility that is part of the PSI center network. "These consortia bring together a large number of talented people in a multidisciplinary setting."

At Stanford, Hodgson continues, "synchrotron structural biology--enhanced in both capability and capacity by the PSI-funded efforts--will be available to anyone who wants to use it. So the investment [in PSI] will accrue to a larger group of people than those in the consortia."

Nonetheless, PSI critics fear that tremendous resources are being poured into a big science effort that will churn out structure data devoid of meaning. They contend that knowledge of structure alone often will not allow scientists to accurately determine protein folding and function.

Hodgson also notes DOE-operated synchrotrons are one of the essential ingredients in the success of the PSI. This federal partnership with NIH, he says, "is a very good example of interagency cooperation in enabling large-scale science in biology."

"I am all for structure determination," states Thomas A. Steitz, a professor of molecular biophysics, biochemistry, and chemistry at Yale and a Howard Hughes Medical Institute investigator. However, "my feeling is that the only reason for knowing structure is to understand biological function. It has been established over the last three or four decades that one can only understand function when proteins are interacting with ligands."

	RODS Once crystallized, this prokaryotic protein can be studied using X-ray crystallography. COURTESY OF SGX
	WHOLE APPROACH Determining the structure of the 50S ribosomal subunit from the bacterium, Haloarcula marismortui shown here provides more assembly information than would be obtained by determining the structures of the isolated components. The nearly 3,000 nucleotides of its 23S and 5S rRNAs are in a ribbon representation in silver, and the proteins are in gold. COURTESY OF THOMAS A. STEITZ

Steitz, along with Moore, was among the coinvestigators who determined the structure of the large subunit of the ribosome--a landmark achievement in science. The group not only achieved unprecedented resolution of the structure, but its work has also provided new insights about how this subunit of the ribosome functions [Science, 289, 905 and 920 (2000)].

"It's taking away from something if you don't fund growth," Steitz says of his concerns about PSI and continued support for the kind of structural biology research that produced the ribosome structure. "I would rather see the money go to young structural biologists rather than PSI centers."

BALANCING THE DESIRE for quick structure determination against the need to study individual proteins "is one of the issues that needs to be dealt with," says Brian W. Matthews, a professor of physics at the University of Oregon, Eugene. "It's not clear that hypothesis-driven research is the only way."

But "we think we can have it both ways," says Thomas Terwilliger, principal investigator of the TB Structural Genomics Consortium at Los Alamos National Laboratory in New Mexico, one of the nine PSI centers. For example, he says, "we intend to solve structures for [the tuberculosis mycobacterium] that will be useful for drug design.

"The TB bacillus contains about 4,000 proteins, and we know the sequence of amino acids in each of them from the DNA sequence of the TB genome that was determined several years ago. We don't know what all these proteins do, but for many of them we can get an idea by comparing their sequences to those of proteins that have been studied in the laboratory."

"I have been astonished at how much you can learn about function from learning structure," says David S. Eisenberg, a Howard Hughes Medical Institute investigator and a professor of chemistry and biochemistry at the University of California, Los Angeles, and director of the UCLA-DOE Laboratory of Structural Biology & Molecular Medicine. He says, for example, that a student working on the structure of an Archaean protein has been able to learn about gene-splicing machinery.

In fact, "in every one of the structures we have determined so far, we have learned something about function," Eisenberg says, "often something surprising." Eisenberg is part of the TB Structural Genomics Consortium.

"So what is behind the criticisms of PSI?" Eisenberg asks. "Some scientists view the important problems to be large, difficult complexes like the ribosome or molecular machines. They say, 'Don't ignore these important complexes,' and I agree," Eisenberg explains.

"It's not true that it is data without meaning," says Andrzej Joachimiak, director of Argonne National Laboratory's Structural Biology Center in Illinois and principal investigator of the Midwest Center for Structural Genomics. "We have several examples of structures for which we were able to figure out function such as SurE, a stationary phase survival protein, and an acid phosphatase or YrdC that binds RNA. It's a real challenge to figure out what so-called hypothetical proteins do--many of these proteins are essential for living cells."

Argonne structural genomics researchers use the Advanced Photon Source, which produces the most brilliant X-ray beams available in the U.S., Joachimiak points out. That means the Midwest Center for Structural Genomics consortium is able to produce some of the highest resolution protein structures available.

What's more, Joachimiak says, Argonne's Robotic Molecular Biology Facility can produce 200 to 400 protein clones per week. Manual methods produce only 20 to 40. In time, he hopes to see automation increased to the point where "researchers can send us their crystals to be placed in the beamline using robots and have the data remotely collected and processed in real time."

The elements necessary for high-throughput structure determination began to materialize in the 1980s, says Helen M. Berman, professor of chemistry at Rutgers University, Piscataway, N.J., and director of the Protein Data Bank (PDB), an international repository for the processing and distribution of three-dimensional macromolecular structure data.

"There was an explosion of data in the 1980s," Berman says. Recombinant techniques, methodologies like crystallization kits for preparing protein crystals, the advent of facilities like the various DOE synchrotron sources for the purpose of gathering X-ray diffraction data from crystals, and improved computing capabilities, she continues, meant that methods for structure determination were dramatically improved.

In structural genomics today, Berman says, "everybody is trying new things." Of PSI, she adds, "it's a different kind of spin on the way we do science."

THE PDB will be an integral part of PSI. As GenBank has been the repository for sequence data from HGP, PDB will be the required repository for PSI structure data. More information about PDB is available at http://www.pdb.org/.

"Our requirement is that when a structure is complete it must be deposited in PDB in a timely fashion--for now, the time period is set at four to six weeks," PSI Director Norvell says.

Raw, fundamental data on the shape of natural protein molecules, including 3-D positional coordinates, "should be made freely available to researchers everywhere," according to an international agreement signed by NIH officials earlier this year that covers PSI and the structural genomics initiatives of other nations. But patent protection for inventions based on PSI structures can be obtained, Norvell says, and can play a role in stimulating health care product development.

	TIME-SAVER Liquid-handling robots are used in high-throughput structure determination. COURTESY OF SGX

Norvell says release of structure data may be accompanied by a short peer-reviewed paper. The key requirement is that publication be completed rapidly. Any publication prior to the end of the maximum six-week delay period will trigger data release. "People understand [the structures] are a public resource," he says.

But as PSI structure determination robotics and other techniques start piling up the work for PDB staff, there are other issues raised by PSI critics--training and education. In a 1998 memo to Cassman, Steitz wrote: "Projects of this sort would provide appalling research training for graduate students and have absolutely no place in the university setting. Students would not be trained to ask questions and seek experimental and theoretical answers, but rather would be trained as gene expression, crystallization, and structure determination technicians."

Graduate students involved with PSI "won't learn how to become scientists," Moore tells C&EN. "It's basically stamp collecting for somebody else's benefit."

"I wouldn't do it on a bet," Steitz adds. "Boring."

"It wouldn't be a good training experience," Norvell says. "PSI expects most of the work to be done by technicians and requires special justification for employment of graduate students and postdocs."

Wayne A. Hendrickson, a professor of biochemistry and molecular biophysics at Columbia University, says he appreciates the concerns of critics. His own laboratory studies focus on macromolecular structure to gain an in-depth understanding of biological activity. Among the work of his research group are 3-D atomic structures for binding fragments of the human T-cell co-receptors CD4 and CD8, and for murine class II major histocompatibility molecules complexed with peptide antigens.

But "it's a mistake to take a rigid view that you don't work a problem until there's lots of biological information," Hendrickson says. "You can speed biological discovery by putting structure in at an early stage.

"It doesn't makes sense to think [of the centers] as training programs," he adds. "If we're going to do this massive effort, we have to hire people to do the job. The so-called pipeline issue is a nonissue. If they get bored--tough.

"The real issue," Hendrickson continues, "is what's the most efficient way to get information that will enable future discoveries. PSI is incredibly enabling of other kinds of research."

Examples Hendrickson cites include the "tubby molecule" in mice that leads to obesity. There has been a lot of work to understand the molecule, he says, but finding the structure of the molecule "has driven the biological investigations."

Likewise with the human transcription factor molecule, Hendrickson continues. "The structure gave inspiration to a variety of experiments."

"The argument often leveled against the work by critics is that 'it's not really science. It's just doing structures for the heck of doing structures,' " Syrrx' McRee says.

The fact is, McRee continues, "the current method is solving the same sorts of [protein] folds over and over again. The number of unique folds is falling, so the NIGMS mission is to go after new folds.

"There is new information coming out of [PSI]," McRee continues. "And when you strike out into the unknown, you will find something you didn't expect."

[Previous Story] [Next Story]

Chemical & Engineering News
Copyright © 2001 American Chemical Society