SCIENCE IN PICTURES

Protean proteins

Through computer modeling, these minute molecules reveal form, function and a kind of natural beauty

By Scott LaFee

STAFF WRITER

April 13, 2005

DAVID S. GOODSELL /

The Scripps Research Institute

A ribbon diagram depicts the amino acid chain structure of a bacterial hemolysin protein. The cylindrical portion punctures red blood cells, releasing their contents.

The pictures on this page are the future of biomedicine taking shape.

They're proteins, or more precisely, computer-generated models of protein structures. If genes make up the "blueprint" for building a person, proteins pretty much do the rest. They are the construction materials, the tools, the workers. They provide the architecture of cells. They make up the working machinery. Hormones, enzymes and antibodies are all proteins.

Given their ubiquity and myriad responsibilities, scientists and pharmaceutical companies increasingly think proteins represent the best target and venue for designing new drugs and treatments. But to do so, researchers must first understand how proteins work, and that means teasing out their precise anatomical structure.

"Proteins represent real physical structures with real behaviors in the physical world," said Arthur Olson, a professor of molecular biology and director of The Scripps Research Institute's (TSRI) Molecular Graphics Laboratory in La Jolla.

"Their shape – and the distribution of the atoms within them – is fundamentally important. Asking how much is like asking how important is the structure of an auto engine. If the machining of an engine's parts isn't quite right, if the weight is not balanced, then nothing's going to happen."

And like a car's motor, proteins are the product of many parts. A gene consists of just four types of nucleotides: adenine, thymine, guanine and cytosine. In different sequences and lengths, base pairs of these chemical letters – A-T and G-C – create each of the estimated 35,000 genes in a human being.

Genes contain the instructions for building particular proteins, not unlike individual pages in a how-to manual. Proteins are assembled from 20 types of amino acids, combined in three-dimensional chains that range from a few dozen in length to more than 1,000. The number of possible variations – and thus, protein structures – is almost endless.

No one knows how many different proteins there are, but a huge effort akin to the Human Genome Project is ongoing to find out. The Protein Data Bank, a worldwide repository operated in part by the San Diego Supercomputer Center at UCSD, currently contains 30,179 three-dimensional protein structures. The ultimate tally will be much, much larger. Some estimates have put it at a million or more proteins.

Small wonders

How a protein functions – or doesn't – is dictated almost entirely by its shape, like a key and a lock. If a new protein folds its amino acid chains incorrectly, the portions of it meant to fit inside or receive other proteins won't work correctly or at all. Misfolded proteins, for example, are thought to be the cause of Alzheimer's disease.

As a result, cells can become dysfunctional or die, leading to any number of medical disorders. But if the structure of a protein is known, scientists can then build molecules specifically shaped to interact with protein targets to ameliorate deficiencies or as beneficial drugs. Similarly, understanding the structure of foreign proteins, like those used by disease-causing microbes, allows researchers to block or alter the sites where such proteins are active.

You can't actually see individual proteins. They're too small, smaller than a single wavelength of light. That means ordinary microscopes are useless. Even electron microscopes produce only "blobby shapes," said David Goodsell, an associate professor of molecular biology at TSRI.

Proteins are, in fact, simply clouds of electrons binding atoms together, said Olson. They have no real surface. To "see" them, researchers typically rely upon X-ray crystallography, a process in which a highly purified, crystalline protein sample is irradiated, producing an X-ray diffraction pattern that can be converted into a three-dimensional map of the sample's atomic structure.

The first protein structures – myoglobin and hemoglobin – were revealed in the late 1950s. It was time-consuming, arduous work, and earned Max Perutz and John C. Kendrew the 1962 Nobel Prize in chemistry.

The first structural images were relatively crude, the atomic structures converted into hand-drawn maps or crafted into physical models. In the years since, technological improvements – radiation equipment like synchrotrons, which can focus a very powerful X-ray beam, and more powerful computers – have both diversified and improved protein renderings.

These days, when scientists want to look at the structure of a protein, they have multiple options depending upon what kind of information they seek.

Perhaps the simplest protein structure image consists of balls and sticks. The balls depict individual atoms; the sticks are the chemical bonds that link them. The technique is used most often by organic chemists, but molecular biologists sometimes borrow it when they want to focus in on specific parts of a protein, such as sites where a protein is most likely to be chemically active.

"They're rarely used to display the whole structure because they're so complicated," said Goodsell.

"Space-filling" models, composed of spheres stuck together, were developed by Nobel laureate Linus Pauling and colleagues in the 1960s to reveal the fulsome shape of a protein. Each sphere represents all of the space occupied by an atom – its nucleus and orbiting electrons. Space-filling models define bulk, and help researchers locate active sites on a protein's surface.

"The active site on a protein, generally speaking, is going to be in a hole, groove or tunnel, and space-filling pictures are the best way to look at that shape," said Goodsell, who studies drug resistance in HIV.

Augmented reality

Currently, the most commonly used protein structure image is a "ribbon diagram," first popularized in the 1980s by Jane Richardson, a professor of biochemistry at Duke University in North Carolina.

Ribbon diagrams reveal the basic architecture of a protein. Chains of amino acids are represented as colored ribbons coiling and curling, sometimes in geometric weaves, sometimes in seemingly crumply chaos.

"I think of them as similar to stick pictures of people, which show the arms, legs, head, joints, but throw away all of the other detail," said Goodsell. "Ribbon diagrams show how amino acid chains fold into the protein's compact structure. All of the atomic details are left out."

Researchers frequently combine one or more of these techniques to illustrate internal relationships or present more information.

Increasingly, they are also returning to an old technique renewed.

Pauling's ball models helped predict the basic folding units of proteins. The double helix that John Watson and Francis Crick erected from brass wire helped them decipher the structure of DNA.

"When you're looking at complex objects, such as a protein, perceiving and understanding their three-dimensional structure on a computer screen is hard, even with current technologies," said Olson. "There are other modes of understanding, sensory input that kicks in when you're able to handle and touch something.

Olson, with colleagues Goodsell, Alexandre Gillet, Michel Sanner and Daniel Stoffler, recently published a paper advocating greater use of "auto-fabrication," an evolving technology in which a device similar to an ink-jet printer squirts out precise amounts of molding material, layer upon layer, to create three-dimensional, many-colored, biologically accurate models that can be handled, manipulated, made of moveable parts or fitted together to form larger molecules.

A new technology promises to take auto-fabrication even further.

"Augmented reality" combines real phenomena or scenes with virtual worlds generated by the computer. The goal is to seamlessly mesh the two, with the computer component adding information in the form of graphics, text, sound, touch or other sensory inputs.

For example, wearing special goggles linked to a computer, a scientist might pick up an autofabricated model of a protein and see, not just the actual model, but displayed data about its internal chemistry or how it physically interacts with other proteins or fits inside a cell.

Whatevertechnology or technique used, modeling proteins has proven to be increasingly and indisputably useful to science. But researchers say there's also an element of art. Their models reveal not just the complexity of life, but the beauty of it, too.