|New Understanding of Life and Its Processes: Structure of Proteins and DNA|
In early 2001, researchers used the prototype Terascale Computing System for a pioneering simulation of a mechnanosensitive membrane protein.
Run your fingertips across a tabletop. Is it hard? Can you feel texture? Our sense of touch gives us this information. But how? For that matter, how do we sense the gorgeous music of a symphony or the throb of an electric bass or cocktail conversation in the next room?
It may come as a surprise that for all of the above the answer is proteins. While it's well known that hundreds of thousands of different proteins are ceaselessly busy in the body, there's an important group we're only recently learning about in detail. They're called mechanosensitive channels, and they're found in all living organisms, where they reside in the membranes that form cell walls. Like other membrane-channel proteins, MS channels open and close to provide a pathway for molecules to exit or enter the cell. The flow of ions calcium, sodium and others through membrane channels creates electrical signals that regulate neural and muscular activity. The special trait of MS channels, however, is their ability to open in response to mechanical stress such as the pressure of a fingertip on a tabletop or vibrations in the air and thereby trigger neural processes like touch and hearing.
An estimated 30 percent of the proteins in cells are membrane proteins, yet we've been slow to learn about them because it's difficult to determine their structure. "While we know about 10,000 proteins," explains biophysicist Klaus Schulten, who directs the Theoretical Biophysics Group at the University of Illinois Beckman Institute for Advanced Science and Technology, "only about 20 of these are membrane proteins a very small fraction."
Until quite recently, none of these were MS channels. In 1998, however, structural biologists at Caltech determined the structure of a bacterial MS channel, called the bacterial large conductance mechanosensitive channel, or MscL. This ground-breaking work became the raw material for Schulten and physics graduate student Justin Gullingsrud to carry out a series of simulations, for which they initially used PSC's CRAY T3E and then turned to the 256-processor prototype Terascale Computing System.
"When you have the structure, there's still a great deal that you don't know," says Schulten, "because the structure is a static snapshot, in this case with the channel closed. For this protein especially, you want to know the dynamical function, how it opens the details of what happens when you apply strain. The only way to learn this presently in detail is to take the structure and simulate its response to strain."
Opening the Gate
Although bacteria lack higher-animal perceptions like touch and hearing, MscL appears, nevertheless, to play an important role in their cells, acting as a safety valve during "osmotic shock." When a change in osmotic conditions induces water to flood into a cell, it can swell like a balloon to potentially burst and die. MscL protects against this, scientists believe, by opening like a gate, in response to pressure, allowing cytoplasmic material to escape. "It's a sort of uncontrolled safety valve," says Schulten, "where the cell loses some important interior molecules, but prefers this to bursting."
With an elegant laboratory technique called "patch-clamp" experiments, researchers have been able to gather data about this MscL gating mechanism. These studies measure electrical current through the channel in relation to strain on the membrane, and they show no current, an absolutely nonleaky channel, until significant strain is applied. They also show that when the gate fully opens, it provides a "conductance pore" 30 Angstroms across. The patch-clamp studies also indicate a staged process of opening, with several "subconductance states" of reduced current flow that precede the fully opened gate.
These important experiments, however, offer no insight into the molecular details, no picture of what structural changes MscL undergoes as strain on the membrane opens the channel. As a step in that direction, Schulten and his colleagues first constructed a computational model of a section of cellular membrane containing the MscL channel. They created a patch of membrane consisting of 195 lipids, long-chain fatty molecules that form cell walls. To realistically simulate the cellular environment, they "hydrated" the membrane placing it within a bilayer of 7,387 water molecules yielding a molecular system of 55,666 atoms.
Using the CRAY T3E, they simulated this membrane-bilayer system with realistic conditions of temperature and pressure to allow it to "equilibrate," to fluctuate and find its natural state. Because of the water environment, the electrical attractions and repulsions exerted by the molecules on each other play a significant role in the dynamics, which means these "electrostatic" properties need to be calculated in as exacting a manner as possible adding considerable complexity to an already complex system.
The results revealed that MscL is structurally stable in its closed state, in agreement with the patch-clamp experiments that show no leaks. The data also show that the stiffest part of the structure is the helices that form the narrowest part of the channel, a result that agrees with electron-spin resonance measuring the flexibility of different parts of the protein.
For this first simulation of an MS channel residing in a hydrated membrane bilayer, the equilibration computation alone was so computationally demanding that the researchers chose not to put strain on the channel-membrane system. Instead, they simulated the protein alone, applying surface tension directly to MscL. "If the membrane can pull on the protein, we reasoned," says Schulten, "why not pull on the protein directly, with mechanical forces in the simulation." Results were promising, showing an opening of 30 Angstroms as the channel helices flattened in response to pressure, in good agreement with patch-clamp data.
Uncorking the Bottle
"Of course, this isn't totally satisfactory," explains Schulten. "We wanted to do the same simulation but with stretching of the membrane, and the only way to do this was with a more powerful system, one of the most powerful available to civilian scientists, which is why we came to PSC."
In early 2001, Schulten and his team were among the first researchers to productively use the early model, 256-processor version of the Terascale Computing System. They rebuilt the membrane environment, to accommodate the dynamics of adding strain, with 242 lipids and 16,148 water molecules, a system of 88,097 atoms. Their pioneering, very large-scale simulation confirmed the earlier results and looked more closely at the staged process of channel opening. Results indicate correlations between the subconductance states shown in the patch-clamp data and details of the MscL structure. The narrowest part of the channel functions like a cork in a bottle, blocking the opening even as strain widens other parts of the channel, until the last instant.
"First you get the widening of the protein," explains Schulten, "through the flattening of the helices, like an iris shutter on a camera, but the 'cork' is still sitting there. Then this opens up radially also, and the whole channel becomes accessible to the molecules that pass through it. This is a feature in other membrane channels, an interesting feature that we can relate to other membrane properties. The analysis of the simulations was inspired by our experimental collaborators."
The simulations give a detailed picture of which parts of the protein, what amino-acid groups, move and how, as the channel goes from open to closed, details that elaborate upon and enrich the information developed from experiment. The next step? "If we can," says Schulten," we'd like to get information on the mammalian system. We're very interested in understanding how these cells that detect mechanical forces work."