Projects in Scientific Computing

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


Biochemically speaking, enzymes make things happen. Action is their motto. If metabolism were a party, it would be inert, as good as dead, until the enzymes arrive. When they show up, suddenly the whole place starts jumping.

Enzymes - nearly all of which are proteins - are such amazing catalysts that it's hard to appreciate how much difference they make. What they do seems simple enough. A molecule undergoing reaction - called the substrate - swims into contact with the enzyme for that reaction and finds a place where it fits, usually a pocket or crease in the enzyme's 3D shape - called the active site. Seemingly, the next thing you know the substrate has changed to the product of the reaction.

The enzyme itself is unchanged and from outward appearance has done nothing beyond provide a port in a storm for a wandering molecule. Yet incredibly, the reaction happens a million times faster than it would without the enzyme. And with some reactions, the speedup is much more than that, a billion-trillion times faster with enzymes.

John Hempel, University of Pittsburgh; Hugh Nicholas,
		David Deerfield and Troy Wymore, Pittsburgh Supercomputing Center.

(l to r): John Hempel, University of Pittsburgh; Hugh Nicholas, David Deerfield and Troy Wymore, Pittsburgh Supercomputing Center.

What biomolecular magic is happening here? "Over the past 50 years or so," says biochemist David Deerfield, "we've made enormous progress in filling in the big picture of what enzymes do and how they do it at an ever increasing level of detail. But we're a long way from precisely understanding what happens in the transition from substrate to product, the atom-by-atom details of these changes and exactly how enzymes play their role."

For the past few years, in research supported through the NIH National Center for Research Resources, PSC scientist Troy Wymore has worked with Deerfield, who directs the Pittsburgh Supercomputing Center biomedical program, and with others - PSC scientist Hugh Nicholas, University of Pittsburgh biologist John Hempel and computational chemist Martin Field - to elucidate the mechanism of an enzyme called aldehyde dehydrogenase.

"ALDH is a big family of enzymes that show up in one form or another in just about every living thing," says Nicholas. "They help to metabolize and detoxify chemicals called aldehydes." Among the aldehydes is ethanal, a byproduct of alcohol, and ALDH's ability to break down ethanal is what pulls us through a hangover. About half of all Asians lack a liver ALDH needed to metabolize ethanal and as a result become sick and potentially risk their lives if they drink alcoholic beverages.

A closeup of the ALDH3 active site with cofactor.
Click Image to Enlarge

A closeup of the ALDH3 active site with cofactor. The enzyme is represented as a surface with color corresponding to positive (blue) and negative (red) charge.

ALDHs also affect cancer treatment, where one of the most used chemotherapy drugs breaks down to an aldehyde. "With cancer therapy," says Hempel, "ALDH just does what it's supposed to do, detoxify an aldehyde. The aldehyde in this case, though, is a therapeutic agent. By doing its job, ALDH reduces effectiveness of the drug, which means you have to use a larger dose than would otherwise be necessary, which increases side effects. If we understood this enzyme better, we could presumably design drugs to block the ALDH that eats up the cancer drug and use smaller doses to attack the cancer."

With LeMieux, PSC's terascale system, Wymore has at hand a powerful tool to attack his problem. In 2001, with a fellowship from the Human Frontier Science Program, he spent a month in Grenoble, France working with Field, a leader in the development of computational methods to understand how enzymes work. Over the past year, with a creative approach that combines less costly, non-quantum methods with detailed quantum techniques, Wymore employed LeMieux in simulations that pin down previously unknown details about how ALDH does its job. In the process, he and his colleagues appear to have discovered an interesting chemical event no one has seen before - a proton transfer from the enzyme backbone.

Breaking Up is Hard to Do

Close-up of the active site. ALDH Full View
Click Images to Enlarge

Closing in on the Active Site
This ribbon model (inset) depicts the symmetrical two-part (dimer) structure of ALDH3, with the helices and sheet structure of one subunit (tan) differentiated from the other (deep green) and loops (rust & gray) also differentiated.

The closeup (above) identifies ALDH active-site amino acids according to Hempel and Nicholas' sequence-analysis studies. Some amino acids (pink, bonded representation) distinguish ALDH3 from other classes of ALDH. A few (dark blue) occur in all classes of ALDH. Others (green) show some variability but are nevertheless essential to structure and function of the enzyme.

Wymore's ALDH simulations arise from work started by Hempel, who in 1997 collaborated with others to determine the 3D structure of an ALDH enzyme for the first time. Not long after that, Hempel and Nicholas, a specialist in DNA and protein sequence analysis, teamed up on a sequencing study of the ALDHs. Their work grouped this "superfamily" of nearly 200 enzymes into 13 distinct sub-families or classes, structurally related by their amino-acid sequences and differing in the aldehydes they catalyze.

As an entry point into this huge enzyme family, Wymore focused on ALDH class 3, one of the smallest ALDHs and therefore a good choice, by the economics of computing time, to work out computational strategies. ALDH3 is the ALDH most involved in metabolizing cancer-therapy drugs, and it's also the subject of Hempel's lab work. "It's a big plus," says Deerfield, "that we have this strong connection with a wet biology group. Experiment and computation can feed each other. When the calculations identify certain amino-acid groups as being important, the lab team can do the experiments that test these findings."

What's going on at the ALDH3 active site? In other words, what happens chemically, atom-by-atom, between the substrate, an aldehyde called benzaldehyde, and the amino-acid groups that surround it like a big easy chair? Experiments help to determine which amino acids are involved in changing the substrate to the reaction product, but chemical reactions happen with many fleetingly transient intermediate states. Understanding these transient states can be the key to intervening with drugs to regulate an enzyme reaction. Wymore's challenge was to precisely identify each step in the ALDH3 mechanism.

It's a challenge that calls for quantum methods. The most widely used approach to simulating proteins - called molecular dynamics - shows structure and movement, but doesn't show reactions, the making and breaking of bonds. To get at these pesky, crucial details, Field pioneered hybrid methods that zoom in on the active site with quantum theory while, at the same time, maintaining the less costly approach for the rest of the enzyme. It's a complex approach - called QM/MM (quantum mechanics/molecular mechanics) - that Wymore added to his toolbox before going to work on ALDH3.

Keeping Track of Protons

A big question had to do with the chemistry involved in the first intermediate state of the reaction - in particular, what's going on with the hydrogen atoms? Experiments show the first intermediate forms when an active-site cysteine, an amino acid with a sulfur atom bonded to hydrogen, binds with benzaldehyde. Cysteine's sulfur binds with a benzaldehyde carbon, and the untested thinking was that the cysteine was "deprotonated" - gave up its hydrogen - before forming the bond with benzaldehyde.

The researchers had their doubts, and Wymore undertook a series of calculations, using various methods, to address this question, which was a prerequisite to using QM/MM to simulate the reaction. One strong result came from molecular dynamics computations that included benzaldehyde in the active site. With the cysteine deprotonated, benzaldehyde was unwilling to orient itself as needed to form the intermediate state.

Guided by this clue and others, Wymore did several different computations with the cysteine protonated. With a series of simulations of the enzyme with cofactors - molecules also involved in the enzyme mechanism - surrounded by water, a total of 56,000 atoms, he systematically studied the active-site interactions. The results show a two-stage process. The cysteine is protonated when it first interacts with the substrate and only then becomes deprotonated - through interaction with a nearby amino-acid group.

Molecular diagram.
Click Image to Enlarge

What Happens at the Active Site
Among the amino acids that surround benzaldehyde (six-sided figure) as it sits in the active site, are (from the top, clockwise) asparagine, threonine, lyseine, cysteine and glutamate. (Squiggly lines indicate where the enzyme structure continues before looping back to the active site.)

The first intermediate forms when the sulfur (S) from cysteine binds (RED ARROW OR SOME OTHER INDICATOR) with the carbon of the aldehyde (-CHO, C not shown). Contrary to thinking prior to Wymore's simulations, this sulfur keeps its hydrogen (H) when it binds to form the intermediate and only then donates it to glutamate.

Prior to the simulations, researchers believed that hydrogen bonds (dotted lines) from asparagine and the backbone nitrogen stabilize the intermediate in the active site. The simulations show, however, that the backbone proton actually transfers (ARROW) to the oxygen, a stronger chemical bond. Preliminary results indicate that this transfer occurs through a complex "proton relay mechanism" (arrows show the direction electrons move) that involves a hydrogen in the active-site lysine. With further calculations, the researchers are analyzing this mechanism.

It was also the prevalent view that another hydrogen atom at the active site formed a hydrogen bond with benzaldehyde to help stabilize this first intermediate. Wymore's simulations showed, however, that rather than forming a hydrogen bond, the enzyme actually donates a proton to the intermediate state. This proton transfer happens as a concerted reaction at the same time as cysteine's sulfur binds with benzaldehyde.

Perhaps the biggest surprise is that this donated proton comes from a nitrogen atom on ALDH3's backbone, next door to the cysteine. "It's quite interesting," says Wymore. "We seem to have discovered a novel enzyme mechanism. We wouldn't have looked for this because the backbone is structural. You expect the side chains to do all the chemistry."

The key, says Deerfield, is the QM/MM approach. "We did a series of calculations of just the active site and substrate, without all the protein interactions around it. Then we backed up again, using the quantum framework but now tied to the protein with molecular mechanics. This gives you a more realistic model."

Wymore has more work ahead with ALDH3 and then expects to move on to simulations of ALDH2, the ALDH involved in metabolizing ethanal. He has support from the National Institute on Alcohol Abuse and Alcoholism and from this standpoint ALDH3 is a first step to understanding the bigger enzyme's mechanism. "This has given us the clue to look for a proton donor," says Wymore, "where we wouldn't have thought to look."

Troy Wymore, Pittsburgh Supercomputing Center
David Deerfield, Pittsburgh Supercomputing Center
Hugh Nicholas, Pittsburgh Supercomputing Center
John Hempel, University of Pittsburgh
Martin Field, Institut de Biologie Structurale, Grenoble, France

Terascale Computing System

User-developed code.

Related Material on the Web:
All in the Family, Projects in Scientific Computing, 1999

Troy Wymore, Hugh B. Nicholas, John Hempel, "Molecular dynamics simulation of class 3 aldehyde dehydrogenase," Chemico-Biological Interactions 130-132, 201-07 (2001).

Martin J. Field, "Simulating Enzyme Reactions: Challenges and Perspectives," Journal of Computational Chemistry 23, 48-58 (2002).

John Perozich, Hugh Nicholas, Bi-Cheng Wang, Ronald Lindahl & John Hempel, "Relationships within the aldehyde dehydrogenase expanded family," Protein Science 8, 137-46 (1999).

Michael Schneider, Pittsburgh Supercomputing Center

Web Design:
Sean Fulton, Pittsburgh Supercomputing Center

Photo Credits:
Photography and Graphics Services, Carnegie Mellon University

Revised: November 3, 2002