janus1Janus Channel

Anton Simulations Reveal How Pain, Epilepsy Drugs Work through Same Target Protein

If you tried to imagine two neurological conditions as different from each other as possible, epilepsy and non-headache pain would be a good choice. Epilepsy develops when nerve cells in the brain start to fire in a too-regular pattern, overwhelming normal brain activity. Pain is a matter of sensation, when damage to tissues or other causes activate pain-sensitive nerves in the skin or other peripheral parts of the body, far from the brain.


Astonishingly, researchers recently discovered that drugs for these two very different maladies work by affecting the same nerve cell protein. 

The mystery of how such vastly different conditions could share a similar therapeutic lynchpin has recently been solved, thanks to molecular simulations using the Anton supercomputer at PSC. Cèline Boiteux of RMIT University in Melbourne, Australia, and Toby Allen’s research group at the University of California, Davis, have reported in the Proceedings of the National Academy of Sciences, USA, how they used the Anton supercomputer at PSC to identify two very different mechanisms by which the painkiller benzocaine and the antiepileptic drug phenytoin affect the voltage-gated sodium channel protein.

In the process, the researchers have uncovered unexpected new molecular targets that pharmaceutical researchers may use to develop new pain-and-seizure control medications.

janus1Topical painkiller benzocaine (colored blobs) sticks to the sodium channel protein (gray, looking down through the channel through the cell membrane) in a series of spots (Gb through Ab), finally settling into a number of positions in the pore of the channel, blocking any sodium passage (Ab).


Each nerve cell is like a little battery, storing electro-chemical energy by maintaining an imbalance of charged sodium and potassium ions outside and inside the cell. A nerve cell impulse happens when a trigger event—in the case of a pain-sensing neuron, a physical injury for example—causes “gated channel” proteins in the cell membrane to open up. This lets the ions flow across the membrane, collapsing the ion imbalance.

Gated channels both cause and react to this collapse of the ion imbalance. Unaffected channels nearby those that have activated open in response. This causes their neighbors to open as well, the cascade spreading the signal.

Pain and epilepsy are very different in many ways but share a common feature. Both require neural over-activity to happen—and so we could potentially stop both by targeting the sodium channel that underlies that activity.

“The idea is that if you stop the gated sodium channels from working, you stop the propagation of the pain message,” says Boiteux, first author of the study. “For an anti-epileptic agent it’s a bit more subtle…We want to slow down the affected neurons. And one way to do that is to stop some of the ion channels from working.”


The researchers believed that Anton, a supercomputer developed by D. E. Shaw Research and hosted at PSC, could provide a window into that difference. Anton is hardwired for long-timescale molecular dynamics simulations and so is ideal for this task.

janus2Figure (left) - Anti-epilepsy drug phenytoin (colored blobs) sticks to the sodium channel in a more random series of spots (Ep through Ap), eventually settling into a place outside the channel (Ap) that closes the central pore indirectly.

“We realized that one of the big issues with other computing systems was that on the typical time scale that we could access—in the range of 100 nanoseconds—we had to make a lot of assumptions about where the drug was going to bind to the channel,” says Allen, the principal investigator of the study. “We knew by experiment that some binding sites play a key role in drug binding,” but starting with the drugs in or near the known critical sites might prevent the researchers from discovering unknown critical sites. With an Anton allocation, by comparison, the researchers could simulate large biomolecules over time frames 10 to 100 times longer.

“We decided to use the power of Anton to not make any assumption at all of where the drug was going to bind,” Boiteux adds. “We just put the drug into the solution and waited to see what would happen without any other input on our part.”

The group concentrated on the voltage-gated sodium channel from bacteria, because the full three-dimensional structure of this protein was already known, unlike the human version. They could test human drugs against this protein because previous research had revealed it to react to the drugs in the same way human sodium channels do.


Simulating benzocaine and phenytoin with the membrane-bound gated channel provided both reassuringly predictable results and some big surprises. Benzocaine traveled to its known binding spot in the channel, blocking the ability of sodium ions to get through. But it used a highway of intermediate attachment points to get there that researchers had been entirely unaware of.

“We could see benzocaine going straight to the site we knew was really important, which was really reassuring,” suggesting that the simulation was recreating the real channel accurately, Boiteux says. “What we didn’t expect was that all these other sites kind of aligned along the protein, leading to this major site and allowing the drug to enter the protein and block it.” These intermediate sites could be potential drug-development targets.

Phenytoin had a very different way of approaching the channel. Unlike benzocaine, simulated phenytoin was unable to enter the channel quickly. Instead, it took a more random route, eventually sticking to a point on the exterior of the protein, shutting the channel indirectly and less completely than benzocaine—helping to explain the difference in the drugs’ effects.

“What was interesting about phenytoin was that it was able to close the channel even without getting inside,” Allen says, “and the amino acids it uses to close the gate are conserved between the bacterial and mammalian channels. This suggests the bacterial channel will be a good initial model for human anti-seizure drug discovery.”

Proteins & Life Sciences

User Information

Connect to PSC systems:
Technical questions:

Send mail to remarks@psc.edu or call the PSC hotline: 412-268-6350.