Projects in Scientific Computing

Download the PDF version of this article as it appeared in Projects in Scientific Computing, 2002.


If you could travel through the body, a swamp boat might be a good vehicle. Each of us is a walking, talking Everglades, about 75 percent water, and we can't survive for more than a few days without fresh intake of the life-giving juice. With water as the main ingredient, our internal machinery brews a variety of fluids — including blood, sweat, saliva and tears, and our ability to do this, to absorb water and move it place to place as needed, depends critically on a protein called aquaporin.

Completely unknown until 10 years ago, when it was discovered accidentally by researchers at Johns Hopkins, aquaporin is a family of related proteins that reside in cell walls. As their name suggests, aquaporins are pore-like channels for water to flow into and out of cells, and their discovery solved a physiological mystery.

Why does water pass through some cells — such as in the kidneys and the glands that produce saliva and tears — much faster than others? The answer is aquaporins. Lab studies show they have the remarkable ability to conduct water at rates of up to a billion molecules per second. In the human kidneys, this means as much as 400 pints a day.

Photo of Klaus Schulten and Emad Tajkhorshid. [Charles Hancock/University of Illinois Photographic Services]

Klaus Schulten (left) with Emad Tajkhorshid.

"They're very common proteins," says Klaus Schulten, director of Theoretical Biophysics at the University of Illinois Beckman Institute, "in plants, bacteria and mammals. More than 10 different aquaporins have been found in the human body, and several diseases have been traced to their malfunction." Defects in aquaporin underlie diabetes insipidus, a kidney malfunction that leads to thirst and frequent urination. Aquaporin is also a culprit in dry-mouth syndrome and in cataracts in the eyes.

When aquaporin was identified, several teams of molecular biologists began work to find its structure, an effort that culminated two years ago when Robert M. Stroud and colleagues at the University of California San Francisco succeeded. That success, however, still left important questions unanswered.

How is it that aquaporin lets water molecules flow unimpeded while, at the same time, remarkably, it acts like a molecular doorman, refusing passage to the uninvited? The uninvited in this case is protons — positively charged hydrogen ions, what you get when a hydrogen atom separates from water and leaves its single electron behind. Healthy metabolism depends on not allowing protons to travel at will through the cell wall.

Top View
Click Images to Enlarge

The simulation included more than 100,000 atoms, with the cell membrane (light green) immersed in water (aquamarine) and aquaporin embedded in the cell wall. Three of aquaporin's four channels are represented (in perspective) as colored coils (blue, red, golden). The simulation tracked water molecules moving in single file (aquamarine bubbles) through one channel.

"Every living cell is its own battery," explains Schulten. "If you let protons escape with the water, as they're naturally inclined to do, the battery would run down. Too many of these run-down batteries, and you're in big trouble."

Several other channel proteins allow protons to pass, and it's been perplexing that aquaporins don't. Experiments show what happens but can't say how. Working in collaboration with Stroud's group and using Pittsburgh Supercomputing Center's LeMieux, Schulten and University of Illinois colleagues Emad Tajkhorshid and Morten Jensen launched a series of aquaporin simulations. Their results, reported in SCIENCE (April 19, 2002), appear to answer this long-standing question.

Single File Water

When the structural data of aquaporin — spatial coordinates of each atom in relation to the others — became available, Schulten's team used it to build a computer model of the protein within its cellular environment. They inserted aquaporin into a patch of cell membrane, represented by long-chain fatty molecules, called lipids, that form a sandwich-like seal against water. They then, in effect, put this membrane with the embedded aquaporin into a pool of water, surrounding the membrane and protein with water molecules to realistically emulate the cell's watery environs.

Side view
Top View
Click Images to Enlarge

Building a Model
A single aquaporin has four separate channels, each represented as a rod-like coil (red, yellow, blue, gray). The protein (15,000 atoms) sits within a patch of cellular membrane, long-chain molecules (40,000 atoms) that align with their "head-groups" (red) facing outward as a seal against water. For the simulations, this model of protein and membrane swims in a bath of water, not shown (51,000 atoms). The simulation tracks a single file of water molecules (oxygen-red, hydrogen-white) in one channel (red).


With this kind of model, a computational approach called molecular dynamics can track atom-by-atom movements, in effect recording a movie, one frame at a time. Schulten and his team used molecular dynamics software called NAMD, developed in their laboratory to exploit large-scale parallel systems, such as PSC's Terascale, comprising thousands of processors. NAMD includes several features designed to capture the atomic-level details of protein movement with the highest possible realism. "With this machine, the Terascale, and this program that can use it effectively," says Schulten, "we have quite an achievement in technology development for our science."

The amount of computing depends, in large part, on the model's size, and since the aquaporin model included over 106,000 atoms, a huge number, availability of the Terascale system was crucial. "These are formidable simulations," says Schulten, "particularly since they have to be done at a high level of exactness, with the best simulation conditions that can be achieved today. Only the Terascale system at Pittsburgh permitted us to do this in a timely manner."

Earlier work, before the detailed aquaporin structure was available, indicated that water molecules line up single file in the aquaporin channel. "It's a narrow channel," says Schulten, "not much more than one molecule wide." A first round of simulations with the new model confirmed a single-file alignment of seven to nine water molecules, and it went further, identifying the position and orientation of each molecule in line as it moves through the channel.

This gave Stroud's group in California something to work with, and they found good agreement with the simulation for the position of the water molecules. The experimentalists, however, had no way to verify how each water molecule was oriented. "We saw something they couldn't recognize," says Schulten, "because the resolution of their observations isn't fine enough."

Slamming the Door on Protons

With another round of work, the researchers assessed the accuracy of their simulations. Using a version of aquaporin altered by changing two amino acids, results with the computer model closely matched the laboratory structure, giving a high degree of confidence. "We found very close agreement," says Schulten, "and we realized we could trust our findings about the orientation."

Side view
Click Images to Enlarge

Mid-Stream Flip-Flop
This closeup snapshot shows the single-file line of water molecules as they progress through the aquaporin channel. Because the water flips at the mid-point of the channel, protons can't pass through.

Those findings, it turns out, reveal the secret behind aquaporin's ability to let water flood through the cell membrane while, at the same time, locking out protons. As Schulten watches a movie from the simulation, he sees a molecular ballet. One after another, each water molecule enters the channel with its oxygen atom leading the way. At mid-stream, the molecule does a half-pirouette and exits facing the direction it came from.

"They orient with the oxygen atom into the channel," says Schulten, "and then they reverse, so that the hydrogens lead the way out. It's a very relevant finding because for water to conduct protons, orientation matters."

The ability of protons to pass along a single file of water molecules, as they do in other channel proteins, depends on a particular arrangement of water molecules and surrounding protein atoms. Called a "proton wire," this arrangement is like a stepping-stone pathway for protons. This pathway is disrupted and blocked by the mid-channel flip-flop that occurs in aquaporin.

The simulations show that aquaporin's mid-channel architecture doesn't vary under differing conditions and is a stable feature of the protein's structure. "We tested with simulations to see how strongly the channel preferred this orientation," says Schulten, "and we found it's very strongly preferred. As a result, it filters against the conduction of protons."

These results culminate more than ten years of effort, first, to find the three-dimensionsal structure of aquaporin and then to use structural data to understand how the protein works. By providing a picture of the atom-by-atom details of structure that experiment can't provide, computational simulations appear to have solved the second part of this problem. "We could see this orientation of the water molecules in simulation," says Schulten, "but you can't recognize this experimentally."

Knowing these atomic details could eventually help with treatment of aquaporin-related disease. Perhaps also, says Schulten, it can lead to improved approaches for water treatment. "The kidneys are a filtering system that relies on aquaporin, and it suggests what kind of technologies might give us better water purifying filters." Since aquaporin filters ions it also suggests an approach to removing salt, an ion, from sea water. "It's possible that we can learn from these proteins and make filters designed according to the principles mother nature uses."

Klaus Schulten, University of Illinois, Urbana-Champaign
Emad Tajkhorshid, University of Illinois, Urbana-Champaign
Morten Ø. Jensen, Technical University of Denmark

Terascale Computing System


Related Material on the Web:
Researchers Explain How Proteins Filter Water into Cells
Touchy Proteins, Projects in Scientific Computing, 2001
Univeristy of Illinois Theoretical Biophysics Group

E. Tajkhorshid, P. Nollert, M.Ø. Jensen, L.J.W. Miercke, J. O'Connell, R. M. Stroud & K. Schulten, "Control of the Selectivity of the Aquaporin Water Channel Family by Global Orientational Tuning," Science 296, 525-30 (April 19, 2002).
M. Ø. Jensen, S. Park, E. Tajkhorshid & K. Schulten, "Energetics of glycerol conduction through aquaglyceroporin GlpF," Proceedings of the National Academy of Sciences 99, 6731-36 (2002).
F. Zhu, E. Tajkhorshid & K. Schulten, "Pressure-induced water transport in membrane channels studied by molecular dynamics," Biophysical Journal, 83, 154-60 (2002).

Research Funding:
National Institutes of Health
National Science Foundation

Michael Schneider, Pittsburgh Supercomputing Center

Web Design:
Sean Fulton, Pittsburgh Supercomputing Center

Photo Credits:, Charles Hancock/University of Illinois Photographic Services

Revised: September 30, 2002