Inside the human body are millions of miniature machines, the gatekeepers of the electrical impulses that keep our hearts beating and our minds thinking. They’re called ion channels; portals that allow small ions such as sodium, potassium, calcium, and chloride, to pass in or out of cells. A simple responsibility, with a complex and crucial outcome, as the flow of ions allows muscles to contract and action potentials to fire along the length of neurons.
While the job of an ion channel may seem straightforward, their design is anything but. Because the channels are awfully tiny, scientists have been forced to determine their workings through indirect experiments. A handful of pictures of the channels have also been taken, via the method of X-ray crystallography, but the photos can only capture an ion channel at rest – imagine trying to figure out how a car engine works from a single photograph.
But what if you could take the volumes of indirect information about an ion channel and instruct a computer to fill in the blanks? That was the approach taken recently by a team of scientists from the University of Illinois and the University of Chicago to tackle a target at the top of the list for ion channel researchers: the potassium channel voltage sensor.
“A potassium channel is a switch that opens and lets ions flow,” said Benoît Roux, professor of biochemistry and molecular biophysics at the University of Chicago and an author of the paper. “And that voltage dependence switch is necessary to understand how the nerve impulse works.”
Think of the ion channel as a garage door, and the voltage sensor as the control box. The channel only opens at a particular voltage, so it needs a way of both detecting voltage changes and powering the transition from closed to open states. It does so with the voltage sensor, a group of positively-charged amino-acids that can be pushed a small distance inward or outward by changes in voltage.
Scientists have a fuzzy idea of how that voltage sensor works from electrophysiological experiments, but the fine points of the mechanism are still unclear. As with any unknown territory in science, competing theories with colorful names like the paddle model, the helical-screw model and the transporter model attempt to fill the void. But Roux and his colleagues decided to test the movement of the voltage sensor with a complex computer simulation of one particular potassium channel, called Kv1.2, found in the heart and brain.
“This is not something you can do on a desktop computer,” Roux said. “Other people have had access to big computers, but it’s the strategy to compute that quantity that has never been done. This shows that it’s possible to address this kind of question at the most quantitative level with an atomistic model, and that has never been shown before.”
Essentially, the model – called all-atom molecular dynamics – took the Kv1.2 potassium channel, the surrounding membrane, and the molecules of water you would expect to find in the channel’s natural setting, and represented all their individual atoms: 350,000 of them, to be precise. The computer model then calculated the forces on each individual atom at a particular time, and predicted where each atom would be a femtosecond (one millionth of one billionth of a second) later. Then: repeat.
25 million computation hours later, the team had new information about how the Kv1.2 channel morphs in response to a change in voltage. Some of the results were structural – a component of the channel forms a tight helix shape so that all of the negatively-charged amino acids are lined up. Others were physical – the electrical field was found to drop unevenly across the thickness of the membrane, such that a small movement of the voltage sensor could generate more energy. The way these tricks add efficiency to the channel’s opening fit the experimental evidence and made good theoretical sense, Roux said.
“This is a pretty complex conformational change and you want to control that as much as possible,” Roux said. “If it was much weaker, this thing could slow down to 400 milliseconds, and then you wouldn’t be able to get the spark that you need for the nerve impulse. You would lose control. So it’s necessary that it’s fast.”
The model was repeatedly checked against experimental data to make sure the computer version was in line with reality. But Roux said that some of the discrepancies may actually be interesting leads, rather than computer errors.
“We’re at the stage where when you run these things and compare with experiments, if they disagree too much, it could be a sign that the model was really wrong or it could be a sign that the interpretation of the experiment was actually not correct,” Roux said. “The results are in good agreement enough with experiments that for the little parts where they poke into the unknown, it’s at least worth asking questions.”
And while just figuring out this one piece of the ion channel machinery took millions of hours on the best computers Argonne and Oak Ridge National Laboratories have to offer, Roux thinks that it’s a harbinger of how future research will unfold.
“To really understand all the intricacies of how a drug affects every ion channel, maybe we will have to really do that all in the computer ultimately,” Roux said. “You cannot expect somebody to go patch-clamp 150 different channels with the same drug to just see what the effect is on the tissue; that would be a little insane. If you have computer models that are reliable, then you would just do that in the computer. This is just the beginning. It’s promising to be able to do that.”
Khalili-Araghi F, Jogini V, Yarov-Yarovoy V, Tajkhorshid E, Roux B, & Schulten K (2010). Calculation of the gating charge for the Kv1.2 voltage-activated potassium channel. Biophysical journal, 98 (10), 2189-98 PMID: 20483327