World’s Tiniest Test Tubes in Alzheimer’s Protein Simulation

University of Pennsylvania Team Uses PSC Systems to Explain Puzzling Lab Results

The progressive memory loss of Alzheimer’s disease is devastating. Doctors have known for more than 100 years that plaques of clumped-together protein form in the brains of people with Alzheimer’s. But they aren’t completely sure whether the clumps cause Alzheimer’s or are a consequence of the disease. Using a series of PSC systems, Paul Axelsen and fellow scientists at the University of Pennsylvania have undertaken computer simulations of plaque-forming beta amyloid molecules inside reverse micelles, the tiniest test tubes in the world. Their results offer insights into beta amyloid and why it clumps that can offer targets both for further investigation and possible drug therapies.

Why It’s Important: 

The progressive memory loss of Alzheimer’s disease is devastating to the people it affects and to their loved ones. And it’s a problem that isn’t going to go away. According to the Alzheimer’s Association, 5.5 million people in the U.S. had the condition in 2017. As the baby boomers age, that number is likely to grow: One person in the country develops Alzheimer’s every 66 seconds.

Scientists have learned a lot about the disease. Much research has concentrated on the beta amyloid protein. Doctors have known since 1906 that plaques of clumped-together protein—later identified as beta amyloid—form in the brains of people with Alzheimer’s. It seems likely that these plaques cause the disease and its symptoms. But scientists haven’t completely nailed down how that happens, and some believe that plaques may be a consequence rather than a cause of the disease. A better understanding of beta amyloid and how it forms clumps—aggregates—is crucial to understanding Alzheimer’s development.

Paul Axelsen and fellow scientists at the University of Pennsylvania Perelman School of Medicine are studying beta amyloid with an amazing new tool: the tiniest test tubes in the world. Called reverse micelles (RMs), these microscopic droplets of detergent and water are so small they can hold a single copy of the beta amyloid protein. The team’s previous lab results with RMs hinted at some intriguing possibilities for why and how beta amyloid aggregates. But before they could make sense of their findings, they had to understand their tiny test tubes better. Lab studies of RMs couldn’t even agree on their exact size, with different methods producing different answers.

“The cool thing about the simulations was that … the new, exact size [provided by] Anton could reproduce the experiments exactly … The machine was large enough and fast enough to give us some very deep insight into what our experiments had been showing.”—Paul Axelsen, University of Pennsylvania



Anton 2 simulation of the process of reverse micelle (RM) contact, fusion and fusion termination by the amyloid beta peptide. Times shown are at 2.3, 2.7 and 3 microseconds (millionths of a second).The shaded portions are detergent molecules that form the interface between the RM interior and the oil outside. The small dots are water and sodium ions. The coiled ribbon is the amyloid beta peptide.


How PSC Helped:

To put their lab results into perspective, Axelsen and his team turned to computer simulation using a series of high performance computing (HPC) systems at PSC. Early work on the now-retired Blacklight system, funded by the National Science Foundation, simulated a single RM alone. With the help of PSC’s Phil Blood, they created a virtual RM that showed how its structure forms. Earlier lab work had indicated that RMs form with incredibly uniform size. But different methods came up with different answers as to exactly what that size was. The Blacklight simulation cleared up the confusion. The different measurements, the UPenn scientists found, disagreed with each other because each was sensitive to a different part of the RM: the outer surface, the detergent molecules making up the RM and the water droplet contained within. Adjusting for the simulation’s results, the different lab methods now agreed with each other.

Armed with a better understanding of their tiny test tubes, and with Blood’s help, the team moved to Anton, another now-retired HPC system hosted at PSC thanks to the computer’s manufacturer, D.E. Shaw Research (DESRES) and funding from the National Institutes of Health (NIH). Anton was a super-specialized computer designed to simulate large biomolecules. Its focus allowed it to simulate much larger molecules for longer periods of time than general-purpose HPC systems. Now Axelsen and his colleagues simulated an RM containing a single beta amyloid protein molecule inside. They discovered that the protein created a huge deformity in the little droplets. Instead of taking the shape of little spheres, the protein-containing RMs became oblong. The beta amyloid even stuck itself into the RM’s outer surface. The protein also adopted the shape it normally takes when it’s aggregated with other copies of beta amyloid. This was important, because the group’s previous lab work with real amyloid in RMs had indicated this might be happening. But nobody had seen a single beta amyloid take that shape before, and so they weren’t sure what to make of the result. The simulation reproduced the lab result exactly, making the scientists more confident they’d interpreted the lab work correctly. The researchers believe their results hint at how amyloid beta aggregation may begin in living nerve cells. The protein may stick into the cell’s outer membrane in a similar way, pulling other beta amyloid molecules to itself to start the aggregation. If the simulation is confirmed with further lab experiments, it could offer a promising new target for designing drugs to prevent aggregation.

To better understand their reverse-micelle tool, the researchers returned to investigating the droplets’ behavior alone. They used the new, more powerful Anton 2 system, now hosted at PSC thanks to DESRES and the NIH. The increased power of Anton 2 allowed them to double the size of their simulation, creating two virtual RMs that could interact with each other. The scientists had wondered whether RMs maintain their size by water molecules escaping one RM and re-entering another. But they found something very different happening in the computer simulation. Instead, the RMs were in effect kissing each other, joining momentarily and exchanging water, with little or no water escaping on its own. That process of exchanging water can explain how the RMs are so uniform in size. When one contains a little more water, it joins briefly with another, and they exchange contents to even out. Importantly, when the scientists added beta amyloid to the new simulations, they found that the protein stopped the process, forcing the RMs to separate. The UPenn team doesn’t know yet what that last result means, but it’s an interesting finding they plan to study further.

The scientists reported their simulation results in three papers: the Journal of Physical Chemistry in 2016, the Journal of the American Chemical Society in 2017 and Langmuir in January 2018.

“If you have all these little reverse micelles and you start adding water, they’re going to get bigger. The more water you add, the bigger they’re going to get … What determines equilibrium size is the ratio of water to detergent. For any given ratio there is a very narrow size distribution … But the crazy thing was, even though we had techniques that told us the size distribution was narrow, we didn’t have a good technique to tell us what that size was.”—Paul Axelsen, University of Pennsylvania

Deeper Dive: What Are Reverse Micelles?

As you might expect, to explain what a reverse micelle (RM) is, you first have to understand what a micelle is. We all have daily experience with micelles. They form when a detergent is put into water. A detergent is a molecule that has a water-loving polar part and an oily nonpolar part. Oil and water really don’t mix: in water, the nonpolar parts of a detergent group together to avoid touching the water, forming a microscopic droplet in which the polar parts are outside touching the water, and the nonpolar parts inside touching each other. Detergents clean things because oily dirt and stains also avoid the water by moving inside the micelles. Dirt and micelles then get washed away in the rinse.

An RM forms when you put a detergent into oil, or another oily liquid. Now it’s the polar parts that are avoiding the surrounding liquid. They clump together to form the exact opposite of a micelle, with the oily nonpolar part sticking out and the polar part hidden inside. If you add water to the mix, the water molecules act like dirt does in micelles. They avoid the oily surroundings by moving inside the polar interior of the RMs. The size of the tiny droplet of water inside the RM depends on the ratio of water to detergent added, with a higher water content leading to bigger RMs containing bigger water droplets. The RM becomes a tiny test tube, containing a miniscule droplet of water.

Though lab measurements agreed that RMs maintain a surprisingly constant size given a particular water-to-detergent ratio, they disagreed on what that size was. This made it hard to simulate RMs realistically. The Axelsen team’s findings with Blacklight cleared up the confusion. They found that nuclear magnetic resonance measurements, which are sensitive to the water, were measuring only the size of the water droplet inside the RM. X-ray diffraction, which is sensitive to the polar parts of the detergent, were measuring the size of the polar lining of the RM. And light scattering, which is sensitive to the outer surface of the RM, was measuring the outer diameter. The team’s simulations explained the earlier lab experiments exactly and gave them a precise model for the RM they could use in their next simulations.