“Sticky and Loose Ends” Shed Light on Heart Health

Anton 2 Shows how APOA1 Protein Ends Link to Hold Together “the Good Cholesterol”

Until recently, scientists couldn’t agree how the protein APOA1 holds together the “good cholesterol” that protects us from heart disease. Scientists from the National Heart, Lung, and Blood Institute led a collaboration that used the D.E. Shaw Research Anton 2 supercomputer hosted at PSC to discover how the two APOA1 protein chains wrap around a disk of HDL fat, locking themselves in place via their “sticky ends” and nudging aside their “loose ends.” The finding will help doctors understand how the body regulates cholesterol, and may point to better heart-disease treatments.

Why It’s Important:

It’s a bit of scientific jargon that’s made its way into all of our lives. HDL, or high-density lipoprotein, is “the good cholesterol.” LDL, low-density lipoprotein, is “the bad cholesterol.” High HDL and low LDL in your bloodstream means your chance of heart disease is low. The reverse, and you’ve got a problem. It may be surprising to hear, then, that until recently scientists couldn’t agree on how the body assembles HDL particles. They knew that two copies of a protein called APOA1 form a kind of fence around a microscopic disk of mixed fats, trapping cholesterol and keeping it from making mischief in our systems. But laboratory studies and computer simulations hadn’t given a clear indication of how APOA1 pulls together an HDL particle. Understanding this process better could help doctors find better treatments that encourage HDL formation and so help prevent heart disease in patients.

“HDL clears the arteries and impedes the development of heart-disease plaques. The arrangement of the proteins around the HDL particle is very important for that biological effect, but has been a mystery for decades. People had come up with different models, but the models were contradictory to each other.”—Mohsen Pourmousa, NHLBI

A number of “models” had been suggested for how the chain-like proteins wind around the fat disk. But none of these proposals fit all the data. In particular, experiments designed to give a clue as to which parts of the APOA1s were near other parts by chemically “crosslinking” them in the test-tube gave confusing results that not all the models could explain. Mohsen Pourmousa and Richard Pastor of the National Heart, Lung, and Blood Institute (NHLBI) and Jere Segrest of the Vanderbilt University Medical Center decided to use the D.E. Shaw Research Anton 2 supercomputer hosted at PSC to see if they could shed some light on the question. Anton 2 is made available without cost by D.E. Shaw Research and operationally funded by the National Institutes of Health.

How PSC Helped: 

Anton 2 is specialized to accelerate simulation of molecular dynamics over long physical time scales. It enables researchers to create virtual molecules and watch them as they move, shake, and interact. The system’s specialized power offered Pourmousa and his colleagues a great performance advantage over previous computer simulations. Previous work had simulated the interaction of APOA1 with fat disks for at most 100 nanoseconds of virtual time. That’s one-ten-millionth of a second. Enough to capture some molecules’ motions. But the NHLBI-led scientists wondered whether the size and complexity of the HDL particle required a longer look. Thanks to Anton 2’s specialized hardware, the team could simulate HDL for 20 microseconds—200 times longer.

“Previous simulations shed some light on the HDL structure and essential features, but they were very short because the supercomputers were not fast enough. If we wanted to run 20 microseconds on a conventional supercomputer it would take five years! On Anton, it took five days.”—Mohsen Pourmousa, NHLBI

What the team found was that APOA1 is restless, not settling in for some time. When the two copies of the protein did settle down, they wound around the fat disk in an unexpected way. The amino acids at the ends of the protein chains took up unexpected positions. One end, called the N-terminal, of one copy of APOA1 forms a strong “salt bridge” with the N-terminal of the other copy. The other end, the C-terminal, of each copy is nudged aside by the other copy’s folded N-terminal chain so that it touches the surface of the fat disk. More interestingly, the team discovered two possible ways for the APOA1 proteins to wind around the disk. This explained the confusing crosslinking experiment results. Scientists had been trying to come up with one model for the proteins when in fact there were at least two different versions, each giving mutually exclusive results. Pourmousa’s and Pastor’s colleagues, Hyun Song and Jere Segrest at Vanderbilt University Medical Center and Yi He and Jay Heinecke at the University of Washington in Seattle, followed up on the Anton 2 results with further crosslinking experiments, a different type of computer simulation called “Rosetta modeling” and X-ray probes of APOA1 protein-only crystals. All of these agreed with the Anton 2 results, giving the scientists more confidence that the Anton 2 simulation got it right. They reported their results in the Proceedings of the National Academy of Sciences 115.20 (2018): 5163–5168.

The APOA1 protein chains (one red, one blue) wrapped around an HDL fat disk (molecules in other colors). The N-terminals of the two proteins (NH3+) link to each other via salt bridges, while each C-terminal (COO-) is nudged by the other protein’s folded chain to touch the surface of the fat disk.