Computing Charged Relationships by Sara Latta Electrostatic interactions -- the attraction between two opposite charges, or the repulsion of two like charges -- play a critical role in many biological processes. A number of cell proteins (including some involved in cell signaling events) bind to negatively charged regions of cell membranes through positive regions on the protein surface. Proteins fold in the precise manner they do in part because of electrostatic interactions within the molecule. Moreover, these events are profoundly influenced by the electrostatic effects of surrounding water and the ions in it. [Image] Barry Honig, professor of biochemistry and molecular biophysics at Columbia University, is using NCSA's CM-5 and POWER CHALLENGE computing systems to construct accurate, high- resolution models of protein-membrane and protein-drug electrostatic interactions. These kinds of models, Honig says, will lead to new tools for understanding the physical interactions between biological molecules. New methods to solve an old equation Scientists use the Poisson-Boltzmann (PB) equation to describe electrostatic interactions of molecules in solution. Elegant and seemingly simple, the PB equation is so difficult to solve analytically that complex molecules must be reduced to simple geometrical shapes: i.e., proteins might be treated as spheres, DNA molecules as cylinders, and membranes as planes. Although the PB equation is nonlinear, scientists were using linear approximations when they dealt with simple proteins alone until recently. [Image] Surface representations of acetylcholine esterase. Space filling model. Atoms colored red are the oxygens of negatively charged carboxylate groups. Those colored blue, nitrogens of positively charged basic groups. The substrate acetylcholine, carrying a net positive charge, is shown in yellow. Courtesy Barry Honig and Anthony Nicholls. "Classical electrostatics had reached a dead end as a method in structural biology, because you needed to describe molecules as spheres, basically," says Honig. "The availability of fast computers and new numerical methods made it possible to describe their shape in detail. And shape is very important to the function of biological molecules." For example, a protein's crucial binding site may be nestled in a pocket having intense negative electrostatic potential, even though the proteinÍs overall electrostatic potential is positive. With the advent of fast workstations and numerical methods, scientists were able to construct detailed and accurate electrostatic models of individual proteins or to calculate the conformational free energies of small molecules. Really complex problems -- like constructing accurate, high-resolution models of electrostatic interactions between two complex molecules or adding molecular dynamics to the model -- require the power of a supercomputer. When protein meets membrane A number of cell proteins that bind directly to the cell membrane mediate intracellular communication -- they may relay signals to the nucleus that cause the cell to divide or to make a new protein, for example. Stuart McLaughlin and colleagues at SUNY-Stony Brook showed experimentally that there were strong electrostatic interactions between proteins and membranes. In collaboration with McLaughlin, Honig and his colleagues at Columbia have transferred McLaughlin's laboratory experiments to the computer in an attempt to better understand those interactions. Honig and his colleagues use a version of a program developed in his lab called DelPhi to solve the PB equation and to describe molecular surfaces. The DelPhi program uses finite difference methods. It maps molecules whose coordinates are known onto a 3D lattice. Every lattice point is associated with three variables: charge, dielectric constant, and ionic strength. The precision of the results depends on the resolution of the lattice. Although the water in the system is treated as a continuum, in terms of its dielectric constant, the membranes and proteins are another story. "Describing membranes and proteins in atomic detail require the inclusion of many thousands of atoms in the system," says Honig. Because they were unable to solve the problem using standard workstations, the team turned to NCSA's CM-5 (developed at Thinking Machines Corp., Cambridge, MA). Using a parallelized version of DelPhi developed by a postdoctoral fellow in Honig's laboratory, Nir Ben-Tal, Honig and his colleagues calculated energy as a function of distance between several peptides (protein fragments) and negatively charged membranes. "We're achieving excellent agreement between theory and experiment," says Honig. "We have learned about the nature of electrostatic binding of proteins to membranes, a fundamentally new result." Honig and his colleagues are in the process of moving their model to NCSA's POWER CHALLENGE (developed at Silicon Graphics, Inc., Mountain View, CA) [see access, Fall 1994]. "It's easier to use, and we have not gotten the speed-up from the parallel machine that we would have liked," says Honig. They look forward to using the POWER CHALLENGE ARRAY (also from Silicon Graphics, Inc.) in the future [see access, Summer 1995]. Honig adds, "We are very excited about it." Honig brings his electrostatic models to life using GRASP, a standard program for the display of the 3D structure of proteins and nucleic acids developed by Anthony Nicholls in his lab. "Seeing the electrostatic potential on 3D protein structures can reveal some surprising patterns. In addition to the specific location of charged and polar groups on the protein, the geometric shape of the molecular surface plays a crucial role in determining the electrostatic potential -- which, in turn, often has an important function. [Image] Surface potential displayed with GRASP software. The molecular surface is color coded by electrostatic potential, as calculated with DelPhi for the charges. Potentials less than 10kT are red; those greater than 10kT blue; and neutral potentials (0 kT) white. The active site is clearly distinguishable as a region of intense negative potential. A step toward drug design Honig is using the PB equation to understand the factors that determine the binding free energy of molecules -- a crucial step toward structure-based drug design. A drug must first have a lower free energy when bound to the protein than it has alone. "One has to consider many drugs, many binding modes," says Honig, "[and] conformational changes that occur to proteins when they bind to drugs. Even though one Poisson- Boltzmann calculation does not take that long, one has to do, in principle, many thousands of them to be sure that one's found the lowest energy conformation." Honig and his colleagues are using NCSA's POWER CHALLENGE to calculate the binding energies for a number of drug mimics that might bind to an HIV protein. Although, Honig hastens to add, "There are no direct applications to AIDS. We are just using it as a model system." Honig is taking advantage of other MetaCenter resources to study a number of problems, including molecular dynamics and protein folding. Postdoctoral Fellow Andreas Windemuth (formerly a graduate research associate in the Theoretical Biophysics Group of Klaus Schulten, UIUC biophysicist and long- time NCSA user [see access, Spring 1994) developed Parallel Molecular Dynamics (PMD), a scalable, parallel program for the simulation of the dynamics of biological macromolecules. In the future, according to Honig, PMD may be combined with programs like DelPhi to merge molecular dynamics and continuum electrostatics. Work in this direction is currently underway in a number of groups around the country. Classical electrostatics as a method of studying molecular interactions has backed out of its dead end. Honig says, "We believe this is going to lead to an exciting and fundamentally new research direction." NOTE: A report of this research, "Classical Electrostatics in Biology and Chemistry" by Barry Honig and Anthony Nicholls, appeared in a recent issue of Science (May 26, 1995). .