The CRAY Biology Lab

The two helical strands that comprise DNA's "backbone" are linked together by bonded pairs of "bases" -- four chemicals, guanine, cytosine, adenine and thymine (abbreviated G, C, A and T). The sequence of these bases define the basic genetic properties built into DNA. Like most of the many proteins involved in DNA processes, Eco RI recognizes a specific sequence -- GAATTC in this case, and it attaches to DNA to do its surgical duty only at these locations.

Rosenberg and his graduate students, now Ph.Ds, John Grable and Yongchang Kim used X-PLOR, an efficient program for refining the structure of biological molecules (developed by Axel Brünger of Yale), to help identify the mechanisms by which Eco RI recognizes the GAATTC sequence. These computations resulted in clarifications to the structural model of Eco RI (published in Science). "That's a very involved calculation," says Rosenberg. "Just about all our crystallographic refinement was done on the CRAY. We couldn't have done it without supercomputing power."

This study elaborated on Rosenberg's earlier work showing that Eco RI's structure of paired identical subunits conforms closely to the structural pattern of a number of other proteins that bind to "nucleotide" chains. A series of folds allows the protein to wedge into the large groove, the so-called "major groove" between the two DNA strands and spread them apart. This gives better access to the base-pairs, located at the floor of the groove. The protein wraps almost completely around the DNA, as if embracing it with extended arms, and at the same time kinks the DNA at the center of the GAATTC sequence.

Rosenberg believes that this folded structure, called the nucleotide-binding fold, may be one of the keys to understanding protein-DNA recognition. "This architecture is connected very deeply to how proteins recognize nucleic acids. In an evolutionary sense, it's one of the ancient patterns."

Large-Scale Molecular Dynamics

Rosenberg has also used the Y-MP to do molecular dynamics simulations of DNA. He and recent physics Ph.D. Shankar Kumar collaborated with Peter Kollman of the University of California at San Francisco, creator of the AMBER molecular dynamics program. The researchers designed their simulations of the kinked DNA to examine whether the kink was intrinsic to the DNA or was caused by binding with Eco RI.

"It's a very pronounced kink," says Rosenberg, "and the question is what does it mean? We set up two hypothetical models, the first one being that the kink represents a structural form of DNA not previously identified. The other model, what we call 'molecular strain,' is that the protein pushed the DNA into a distorted state, like a set of compressed coiled springs. We were definitely leaning toward the first model, a new structural form of DNA." The computation simulated the kinked DNA by itself (not bound to Eco RI) to see if the kinked structure was stable. The answer: not stable at all. "The computations gave us a very definite answer, that the kink results from binding with the protein, and this leads to a whole series of ramifications about how the system works."

In other simulations with AMBER, Rosenberg and Kumar have stretched the limits of computational biology. Their simulations of DNA in solution (with 1970 water molecules for 81 picoseconds) represent one of the most extensive molecular dynamics computations on DNA molecules to date.

go back to the main screen