Since 1953, when James Watson and Francis Crick presented their double helix model of DNA, the field of structural biology has exploded. New technologies have led to rapid breakthroughs in understanding the form and function of a wide range of macromolecules.Indeed, most studies of life at the molecular level are now considered incomplete without three-dimensional structures to guide further research.

Structural biologists today use technologies such as nuclear magnetic resonance (NMR) spectroscopy, X-ray diffraction crystallography, electron microscopy, and computer modeling to discover and define the form of macromolecules. Used together, these technologies often provide a more complete picture of biological phenomena than any single technique.

Tom Darden
Steve McCaw, Image Associates
Researchers at the NIEHS have emphasized computer modeling and NMR. The institute was one of the first in the nation to purchase silicon graphics workstations that are now considered essential to modeling and visually displaying molecules. Today, using far more advanced systems, the NIEHS maintains a computer modeling group led by mathematical statistician Tom Darden that provides institute scientists with computational approaches to complex research questions. Several other groups at the NIEHS also use computer modeling in their research.

Computer modeling offers distinctive capabilities for deciphering results of other structural modeling techniques. Crystallography, for example, provides high-resolution snapshots of molecular structures in the solid state. However, molecules in solution are in constant motion, and a more complete picture requires a knowledge of their molecular dynamics. Quantum chemical calculations and various experimental techniques enable researchers to describe the force fields that determine how atoms interact with each other. With this information, scientists can use Newton's equations of motion to study the dynamics of the molecules. Because of the enormous number of equations that must be solved to simulate the motion of a macromolecule for even a billionth of a second, this can only be done by high-speed computers.

Not only is the computational speed offered by computers essential to understanding molecular form and function, so too is graphical representation. Computer graphics provide highly detailed information about the arrangement of molecules in space the only way humans can readily comprehend it--visually.

In the typical representation, called a wireframe, lines depict bonds between atoms with each endpoint representing an atom. These lines can be colored based on atom type for easy identification. Because a macromolecule may consist of more than 3,000 atoms, the ability to isolate and enlarge details is important. With a single click of the computer mouse, researchers can magnify a small portion of a molecule or view just the backbone structure. Another command allows the researcher to instantly measure the distance between atoms. The structure can even be rotated to observe how the atoms are arranged in space. "The important thing about a computer representation is that you can ask a lot of questions that are not answerable experimentally," says Darden, such as "What's the relationship between two atoms over time? Do they ever come together?"

Understanding molecular dynamics is especially crucial in drug design. In their search for the discovery of substrates or inhibitors that will bind to a key enzyme, researchers must know the precise shape of an enzyme's binding site and the dynamics of the enzyme and drug. Computer modeling offers a technique for characterizing substrate-enzyme binding that can assist researchers in identifying and synthesizing clinically effective compounds. Likewise, researchers seeking to understand the effects of environmental toxicants must identify the cellular receptors with which the compounds interact.

Thomas Kunkel and Kasia Bebenek
Steve McCaw, Image Associates
Research geneticist Thomas Kunkel heads a project that seeks to understand the fidelity of DNA replication, particularly as it relates to the HIV-1 virus that causes AIDS. HIV-1 escapes an individual's immune system and resists drugs such as AZT by rapidly mutating. Designing a drug to suppress the virus depends on understanding exactly how it mutates.

In 1988, Kunkel demonstrated that reverse transcriptase (RT), the enzyme that HIV-1 uses to replicate the genome, exhibits a high rate of mutations that result from slippage of the template with respect to the primer strand of DNA. Researchers then set out to understand what parts of the enzyme were responsible for this phenomenon. Kunkel's team chose a region of the RT protein and began making a series of mutant RT proteins, replacing various amino acids and analyzing their effect on DNA replication. The team identified several residues that lead to slippage of the DNA strands, but because they lacked high-resolution structural data, they were not able to rationalize how this occurred.

In 1993, researchers at Rutgers University successfully produced an X-ray crystallography structure of the HIV RT bound to DNA. Although this was a major breakthrough, it did not answer all the questions. Kunkel has since turned to the NIEHS's computer modelers to take this low resolution information and build a predictive model of the RT to suggest further areas of experimentation. "The crystallographic information basically provided me with a bunch of isolated points in space," says Darden. "Using a computer, I hunted through the database of known proteins to find fragments to fit these points, and ended up with a model of what the structure might look like."

According to Kunkel, the model suggested changes that could be made in the region of interest that will help researchers understand precisely how the enzyme binds its substrates. "With such understanding," Kunkel says, "there is a better chance of being able to design more effective drugs."

Masahiko Negishi
Steve McCaw, Image Associates
Computer modeling is also aiding Masahiko Negishi, head of the pharmacogenetics section of the Laboratory of Reproductive and Developmental Toxicology, in his study of P450s, a family of proteins that plays a key role in metabolism. P450s exhibit remarkable substrate and product diversity. Two P450s may differ by less than 10 amino acids out of 500, but this subtle difference may significantly affect an individual's ability to metabolize various drugs and chemicals. Specificity of P450 activity can be altered by a single amino acid substitution at a key position. Based on such differences, one person may be able to take 10 times more of a drug than another individual, or the same dose of a drug may cure one person's symptoms but cause adverse side effects in another. "If, for a given individual, we can pinpoint amino acid mutations at particular points, we might be able to predict their reaction to certain drugs or chemicals and warn them away from those that may cause them trouble," says Negishi.

To identify which amino acids are critical and at what times, researchers must understand the structure of the P450. Bacterial P450 is easy to crystallize, so its structure can be easily determined. However, mammalian P450 has so far been resistant to crystallization. Its structure can only be inferred through computer modeling. This is where Darden's team comes in.

Proteins with similar sequences of amino acids usually adopt similar orientations. Therefore, Darden gathered the structural information known about the bacterial P450 and information produced by Negishi's team through site-directed mutagenesis and enzyme analysis of mammalian P450. Using sequence homology, Darden was able to compute which amino acids in the bacterial enzyme correspond to the mammalian amino acids, resulting in a proposed structure for the mammalian protein. Negishi is currently experimenting with binding various substrates to the modeled mammalian P450 to predict how a particular mutation affects the activity of the enzyme.

"There are two benefits I get from the modeling," Negishi says. "I can get some hypothesis as to why the mutations we see in the mammalian enzyme affect the substrate specificity. And knowing why this might be so, I can do further research. We can go back to test the hypothesis that, for example, the size of a particular amino acid is crucial."

Rachelle Bienstock and Roger Wiseman
Steve McCaw, Image Associates
Computer modelers are using a similar process with Roger Wiseman, director of the Laboratory of Molecular Carcinogenesis, in his search to understand the workings of the BRCA1 gene. BRCA1 plays a key role in hereditary breast and ovarian cancers. The gene appears to function as a tumor suppressor, but a mutation affecting a relatively small number of amino acids at the amino terminus of the protein appears to decrease the gene's ability to regulate cell growth and to increase an individual's chance of developing cancer. Wiseman has called upon NIEHS computer modelers to create a three-dimensional model of this section of the BRCA1 protein to predict how it folds and to determine which residues are on the outside, interacting with other proteins and DNA.

Using the sequence of amino acids for BRCA1 and the sequence of a related herpes virus protein for which the structure is known, computer modeler Rachelle Bienstock ran a series of homology studies to come up with a model. "From this model, I suggested some residues in the protein that might bind to the DNA, and I suggested some experiments that could be done to mutate the protein to see if it changes the binding affinity," says Bienstock. "The more we learn about the pathway the gene takes in mutating, the more we have a chance to intervene."

Computer modeling is also helping to unravel the secrets of opioid peptides used in the treatment of pain. The medicinal use of these peptides, particularly those derived from frog skin secretions, dates back to ancient Assyria and Babylonia. The indigenous peoples of the Amazon Basin still use poisonous frog skin secretions to coat the tips of their blow-gun darts and in shamanistic rites. Scientists believe these unique peptides might have potential applications as therapeutic and clinical compounds in the management of chronic and acute pain and in the treatment of narcotic addiction, alcohol dependency, and suppression of the immune response during transplant surgery.

Sharon Bryant
Steve McCaw, Image Associates
Over the past six years, Lawrence Lazarus, head of the peptide neurochemistry section, has directed a project at the NIEHS that involves synthesizing and analyzing deltorphin and dermorphin analogs that are hundreds of times more potent that the endogenous mammalian opioid peptides (see EHP vol. 102, no.8, pp. 648-654). The ultimate goal of the project is to design new and more potent opioid agonists and antagonists that are clinically applicable. Sharon Bryant, a chemist in the group, uses molecular modeling to determine structural motifs of opioid peptides to predict structure-activity relationships. "Computer modeling helps us to visualize possible orientations of amino acids in the molecule," Bryant says. "It also helps us to understand the intermolecular dynamics. Knowing the orientation of the structure in three-dimensional space would enhance our efforts to design agonists and antagonists." Lazarus and his team have developed a number of dual-affinity agonists and ultraselective opioid antagonists which are currently being tested for their application to human health.

John Manuel