by Michael Schneider
Pittsburgh, PA. — Although AIDS awareness programs and research leading to powerful drugs such as AZT have reduced AIDS-related mortality, AIDS remains one of the most pressing U.S. public-health problems, with incidence rising in some areas and population groups. In less developed parts of the world, furthermore, AIDS is virtually out of control. In South Africa, AIDS infects a third of the current populace of 43 million, compared to under 1 percent here, and is the leading cause of death, with no abatement in sight.
Because of their high cost, the drug “cocktails” that control HIV, the virus that causes AIDS, have been of little use in Africa. Even at their most successful, furthermore, these drugs aren’t a cure-all. By retarding HIV’s ability to reproduce in humans, they save many lives, but the quest for AIDS researchers remains, as it has been, to find a cure – not just therapeutic agents that manage the disease, but a knockout punch.
Physicist Marcela Madrid of the Pittsburgh Supercomputing Center and Carnegie Mellon biologist Jonathan Lukin are contributing to this effort. Their computer simulations have revealed new understanding of an enzyme, HIV-1 reverse transcriptase (RT), that enters the body in the package of proteins and ribonucleic acid that comprise HIV. RT plays an essential role in reproducing the virus and because of this is an important target for drugs.
Madrid specializes in simulating biological molecules, and in 1998-99 she began collaborating on RT with a team of structural biologists. Because RT is a very large molecule, about 1,000 amino-acids, it hadn’t been simulated before, and Madrid’s effort, using PSC’s CRAY T3E, broke new ground. By providing a moving picture of RT, her work filled in details unavailable from the molecule’s static structure. Beyond this, her results went a long way toward showing that this kind of computer simulation, called molecular dynamics, can be a valuable partner with laboratory studies in furthering AIDS research.
With these simulations as groundwork, Madrid and Lukin this year took the next step: including water molecules that surround RT in the living cellular environment. This greatly expanded the computational demands of the project, challenging memory limitations of PSC’s CRAY T3E, and Madrid turned to the SGI Origin 2000 at the National Computational Science Alliance in Illinois. Their results from this recent work, which also relied partly on PSC’s Intel Cluster, offer fresh insight into the details of how this enzyme interacts with other molecules to reproduce the virus.
Like many viruses, HIV carries its genetic information as RNA, which is single-stranded and must be converted to double-stranded DNA before the virus can reproduce. This is where RT comes into play. The enzyme takes its name, reverse transcriptase, from what it does: It transcribes the RNA bases and one-at-a-time creates a complementary base to form the DNA version of HIV’s genome, from which it then creates a mirror-image strand and pastes them together to form the DNA double-helix. It’s “reverse” transcription because in most cells transcription goes the other direction – DNA to RNA. To some extent, this RT copy-and-paste process is understood, and to some extent it’s a mystery – especially the fine-grained details of the molecular manipulations.
What’s clear is if you find a foolproof way to stop RT from doing its job you will have cured AIDS. Several existing AIDS drugs work by binding to RT and blocking transcription. These therapies are less than fully effective, however, because HIV transcription is highly prone to error, giving the virus a protean ability to mutate and, thereby, to defeat drugs. It’s estimated that HIV can undergo as much genetic change in 10 years as the human species does over millions of years.
For this reason especially, researchers want to delineate the atom-by-atom picture of how RT transcribes RNA to DNA. A major step in this direction, and in AIDS research in general since the late 80s, has been to deduce the static structure of RT using x-ray crystallography. This work shows that the active domain of RT, where transcription takes place, is analogous to a hand, with subdomains that roughly represent a thumb, fingers and palm. Interestingly, two versions of this structure – one bound with DNA, one not – show a big difference: With DNA, the thumb is extended and open, making space for the DNA to fit into the palm; without DNA, the thumb is folded over to almost touch the fingers.
This difference was Madrid’s entry to HIV research. Her CRAY T3E simulations at PSC last year investigated whether removing the DNA from the open-thumb structure would cause the thumb to close. It did. Her molecular dynamics movie showed the open thumb closing to a position that agrees well with the closed-thumb crystal structure, indicating that closed thumb is the low-energy, “native-state” of RT. These simulations, furthermore, support thinking that RT’s movable thumb may be a key to the transcription process, allowing the enzyme to slide along the RNA strand as it adds bases one-by-one to form DNA.
Along with possibly being involved in transcription, the joint-like flexibility of the RT thumb may be a factor in how one class of AIDS drugs inhibits HIV reproduction, since it’s believed that these drugs lodge in the palm of RT and interfere with transcription by locking the thumb in an open position. “We want to understand RT’s flexibility,” says Madrid, “when it’s by itself, when it’s bound with DNA and when it’s bound with drugs.”
With this objective, Madrid and Lukin used the SGI Origin 2000 to simulate RT structures with and without DNA. The plan was to compare the fluctuations of each atom as shown in molecular dynamics movies to the crystal structures, which are derived by methods that deduce atomic position according to electron density, and therefore include a measure – called the crystallographic B-factor – that represents the uncertainty in the position of an atom, evidence for motion.
By adding 37,000 water molecules to immerse the RT molecule in a water bath, these simulations had a degree of realism missing from the prior study. The simulation with DNA included more than 130,000 atoms, a huge number for this kind of computation. For each of the two structures, the researchers simulated a nanosecond (billionth of a second) of dynamics, recording the atomic coordinates every half picosecond (trillionth of a second). To do this, they used 32 SGI Origin processors, with each computation requiring about 600 hours of computing.
For the structure without DNA, the results showed good correlation between the simulations and B-factors, giving confidence in the accuracy of the simulation. For the structure with DNA, the simulation showed greater flexibility in the tips of the thumb and fingers than suggested by the crystal structure, and also more flexibility than the structure without DNA.
This flexibility of the fingertips, says Madrid, is consistent with the difference in fingertip position between the RT crystal structure with DNA and another recently reported RT structure that includes a molecule, deoxynucleoside triphosphate, that elongates the DNA. “The movement in the fingers,” says Madrid, “tends to validate the postulate that the fingers open to let this molecule in and then close to trap it in place.”
This simulation showed, furthermore, that movements in different regions correlate with each other. The thumb and fingers and a binding pocket in the palm are moving at the same time – information that can’t be obtained experimentally. “This is the first time the motion of the whole molecule has been simulated in water,” says Madrid, “and we can see that the whole molecule is moving in a concerted manner.”
Though well short of a complete atomic-level solution to the HIV transcription process, this dynamic, detailed picture complements the crystal structures and pulls together much of the available information into a consistent whole, offering a blueprint for future work. In their next project, Madrid and Lukin plan to simulate RT structures that include drug molecules that inhibit transcription. “We’ll be looking for clues,” says Madrid, “to see how the drugs work.” More information is available at http://www.psc.edu/science/madrid2000.html .