A team of computational biologists at the University of Illinois at Urbana-Champaign has performed an all-atom molecular dynamics simulation of the satellite tobacco mosaic virus — marking the first computational simulation of an entire life form and a potential advance for computational drug design.
The simulation, described in the March issue of the journal Structure, tracked the dynamics of the complete virus — comprising a 60-protein icosahedral capsid and a 1.1 kb RNA core — for 13 nanoseconds. The researchers also performed several 10-ns simulations of the capsid and the RNA alone, reaching the conclusion that the capsid is unstable and "implodes" when the RNA core is not present.
Previous experimental evidence had hinted at this behavior for STMV, which contradicts the "prevalent view" that all viruses assemble by first forming a rigid shell and then pull in the genetic material, said Klaus Schulten, director of the Theoretical and Computational Biophysics Group at the University of Illinois and a co-author on the paper.
"In this particular case, it's the opposite," Schulten said. Based on the simulation, it appears that STMV "first forms the RNA, it folds into some fuzzy ball, and then that recruits one by one the 60 proteins to build the shell around itself."
Schulten said that the simulation helps reinforce the experimental data, and serves as an example of how computing is becoming "a partner of experimental biology" and a useful tool for hypothesis generation in its own right.
The study serves as an example of how computing is becoming "a partner of experimental biology" and a useful tool for hypothesis generation in its own right.
In addition, he said, the simulation serves as proof of concept for the use of molecular dynamics in the study of larger viruses of interest to human healthcare.
Most molecular dynamics experiments in drug discovery are limited to single proteins or protein-ligand complexes, but Schulten said that the simulation of complete viruses will be the key to a systems-level understanding of living systems. In the case of STMV, for example, simulating a single protein would be "meaningless," he said. "It's only meaningful if you are studying the 59 other proteins that form the capsid."
Alex Perryman, a postdoc in Stephen Mayo's lab at the California Institute of Technology who recently published a molecular dynamics simulation of the HIV-1 protease, called the simulation of the full virus particle "extremely impressive," and added in an e-mail correspondence with BioInform that "their work will eventually inspire similar Herculean efforts on other viral systems."
Schulten said that his team chose STMV because of its small size. The entire simulation involved around 1 million atoms, which is near the upper limit of current molecular dynamics studies in biology. By comparison, IBM claimed it had reached a milestone for its BlueGene/L supercomputer last year when it simulated a rhodopsin complex comprising 43,000 atoms [BioInform 04-04-05].
The largest biological simulation to date was published last fall by researchers at Los Alamos National Laboratory, who simulated the behavior of 2.64 million atoms in the ribosome [BioInform 11-07-05].
Schulten said that the steady increase of computing power is the primary driver for more complex biological simulations. The STMV simulations took "several months" on 128 SGI Altix nodes at the National Center for Supercomputing Applications. "We are fortunate that we have that power now, but we would not have had that power available, say, two years ago," Schulten said.
On the software side, the University of Illinois team used its own NAMD package, which does not present any limitations in terms of scale-up, Schulten said. He noted that Kevin Sanbonmatsu, who led the LANL ribosome simulation, also used NAMD in that effort. "He just took the program off the shelf and put it on the computers at Los Alamos, so the program is really scalable," Schulten said.
In addition to improvements in computer power, Schulten credited the milestone to a bit of fortunate timing. The University of Illinois team was preparing to simulate the STMV capsid alone, but Alexander McPherson, a crystallographer from the University of California, Irvine, was able to resolve the structure of the RNA core just as the project got underway.
"It just happened that when we were looking at this small virus to simulate the entire thing for the first time, and we were ready to just simulate the capsid, that at that moment, to our great delight, Alexander McPherson resolved the RNA inside, and so then we could do the entire thing," he said.
The crystal structure alone was not enough to determine some properties of STMV, however — namely, how the virus disassembles once it has been taken into a host cell, because of the very high stability of the capsid's structure. The new hypothesis that the capsid is in fact unstable without the RNA core is a step toward understanding this mechanism, Schulten said.
While STMV doesn't pose much of a threat to human health — the virus is so small that it can only proliferate in cells infected by another virus — these initial simulations indicate that similar methods could be used to study larger viruses, Schulten said, which could have implications for the design of antiviral compounds.
"The main point here is that a virus has to do two rather incompatible things — it has to be rigid and sturdy to protect the genetic material and carry it around, but on the other hand, it also has to be very flexible and actually completely break apart at the moment of infection. So you must be able to trigger the breaking apart of a rigid shell, and that is very, very crucial for the infection process," he said.
By simulating a virus, researchers could study how antiviral drugs might "plug" the capsid so that it can't break apart and infect the cell, he said. "So we really need to understand [the] bistable system in order to interfere with its function and thereby protect people."
Caltech's Perryman agreed. "Understanding the assembly process and the critical interactions that stabilize the infective structure will aid in the development of new anti-viral drugs, because such data indicate the key steps and the most important interactions that should be targeted and disrupted for therapeutic purposes."
These advances may be a way off, however. Schulten stressed that "this is just the beginning" of the STMV project, which will now enter another cycle of experimental and computational studies.
One drawback of the recently published STMV simulation is that it is still a little on the rough side. The researchers performed an "icosahedral averaging" of the data based on the assumption that the capsid was perfectly symmetrical, but the simulation indicated that the virus "is moving in a way that doesn't reflect its icosahedral symmetry," Schulten said.
In addition, despite the crystal structure of the RNA core, it is still not understood how the sequence of the virus maps onto the RNA backbone, "so now we have to take the real sequence of the RNA and put it into the experimental electron density map — into the structure that has been identified — and see how the real sequence can be reconciled with the structure shown by experiment." Only then, he said, will the team be able to simulate "the real virus — meaning the virus that is not symmetric, with realistic genetic material inside."
After that, "once we establish that we can actually simulate the virus as a whole, we would like to go to larger viruses that are more interesting," Schulten said. "We took the smallest kid on the block because it was the easiest to handle, and now we want to go to the bigger kids that are more health-relevant, like the poliovirus and other viruses that are rising in human pathologies that need to be understood better."
Perryman said that these types of simulations would be "extremely helpful" if they could be extended to a complete HIV virus particle, but there are a number of significant obstacles that stand in the way of in achieving that goal. "Unfortunately, the structures of some of those viral components are not yet known to a sufficient level of detail, and significant computational advances would also need to occur before such a simulation of HIV could be feasible," he said.
Nevertheless, the work of Schulten's group offers hope that such an achievement might be possible. "Trying to simulate a full HIV particle and, especially, to simulate the way in which a full virus particle interacts with the host cell that it targets is a goal on which society should focus for the next few decades," he said. "The importance of such a goal approaches the magnitude of such things as landing a man on the moon or cloning a sheep."
— Bernadette Toner ([email protected])